diff --git a/documentation/content/en/articles/committers-guide/_index.adoc b/documentation/content/en/articles/committers-guide/_index.adoc index 63f2db466e..8cc8a8beeb 100644 --- a/documentation/content/en/articles/committers-guide/_index.adoc +++ b/documentation/content/en/articles/committers-guide/_index.adoc @@ -1,3947 +1,3947 @@ --- title: Committer's Guide authors: - author: The FreeBSD Documentation Project copyright: 1999-2021 The FreeBSD Documentation Project description: Introductory information for FreeBSD committers trademarks: ["freebsd", "coverity", "ibm", "intel", "general"] weight: 25 tags: ["FreeBSD Committer's Guide", "Guide", "Community"] --- = Committer's Guide :doctype: article :toc: macro :toclevels: 1 :icons: font :sectnums: :sectnumlevels: 6 :source-highlighter: rouge :experimental: ifeval::["{backend}" == "html5"] include::shared/authors.adoc[] include::shared/en/teams.adoc[lines=16..-1] include::shared/en/mailing-lists.adoc[] include::shared/en/urls.adoc[] endif::[] ifeval::["{backend}" == "pdf"] include::../../../../shared/authors.adoc[] include::../../../../shared/en/teams.adoc[lines=16..-1] include::../../../../shared/en/mailing-lists.adoc[] include::../../../../shared/en/urls.adoc[] endif::[] ifeval::["{backend}" == "epub3"] include::../../../../shared/authors.adoc[] include::../../../../shared/en/teams.adoc[lines=16..-1] include::../../../../shared/en/mailing-lists.adoc[] include::../../../../shared/en/urls.adoc[] endif::[] [.abstract-title] Abstract This document provides information for the FreeBSD committer community. All new committers should read this document before they start, and existing committers are strongly encouraged to review it from time to time. Almost all FreeBSD developers have commit rights to one or more repositories. However, a few developers do not, and some of the information here applies to them as well. (For instance, some people only have rights to work with the Problem Report database). Please see <> for more information. This document may also be of interest to members of the FreeBSD community who want to learn more about how the project works. ''' toc::[] [[admin]] == Administrative Details [.informaltable] [cols="1,1", frame="none"] |=== |_Login Methods_ |man:ssh[1], protocol 2 only |_Main Shell Host_ |`freefall.FreeBSD.org` |_Reference Machines_ |`ref*.FreeBSD.org`, `universe*.freeBSD.org` (see also link:https://www.FreeBSD.org/internal/machines/[FreeBSD Project Hosts]) |_SMTP Host_ |`smtp.FreeBSD.org:587` (see also <>). |`_src/_` Git Repository |`ssh://git@gitrepo.FreeBSD.org/src.git` (see also <>). |`_doc/_` Git Repository |`ssh://git@gitrepo.FreeBSD.org/doc.git` (see also <>). |`_ports/_` Git Repository |`ssh://git@gitrepo.FreeBSD.org/ports.git` (see also <>). |_Internal Mailing Lists_ |developers (technically called all-developers), doc-developers, doc-committers, ports-developers, ports-committers, src-developers, src-committers. (Each project repository has its own -developers and -committers mailing lists. Archives for these lists can be found in the files [.filename]#/local/mail/repository-name-developers-archive# and [.filename]#/local/mail/repository-name-committers-archive# on the `FreeBSD.org` cluster.) |_Core Team monthly reports_ |[.filename]#/home/core/public/monthly-reports# on the `FreeBSD.org` cluster. |_Ports Management Team monthly reports_ |[.filename]#/home/portmgr/public/monthly-reports# on the `FreeBSD.org` cluster. |_Noteworthy `src/` Git Branches:_ |`stable/n` (`n`-STABLE), `main` (-CURRENT) |=== man:ssh[1] is required to connect to the project hosts. For more information, see <>. Useful links: * link:https://www.FreeBSD.org/internal/[FreeBSD Project Internal Pages] * link:https://www.FreeBSD.org/internal/machines/[FreeBSD Project Hosts] * link:https://www.FreeBSD.org/administration/[FreeBSD Project Administrative Groups] [[pgpkeys]] == OpenPGP Keys for FreeBSD Cryptographic keys conforming to the OpenPGP (__Pretty Good Privacy__) standard are used by the FreeBSD project to authenticate committers. Messages carrying important information like public SSH keys can be signed with the OpenPGP key to prove that they are really from the committer. See http://www.nostarch.com/pgp_ml.htm[PGP & GPG: Email for the Practical Paranoid by Michael Lucas] and http://en.wikipedia.org/wiki/Pretty_Good_Privacy[] for more information. [[pgpkeys-creating]] === Creating a Key Existing keys can be used, but should be checked with [.filename]#documentation/tools/checkkey.sh# first. In this case, make sure the key has a FreeBSD user ID. For those who do not yet have an OpenPGP key, or need a new key to meet FreeBSD security requirements, here we show how to generate one. [[pgpkeys-create-steps]] [.procedure] ==== . Install [.filename]#security/gnupg#. Enter these lines in [.filename]#~/.gnupg/gpg.conf# to set minimum acceptable defaults: + [.programlisting] .... fixed-list-mode keyid-format 0xlong personal-digest-preferences SHA512 SHA384 SHA256 SHA224 default-preference-list SHA512 SHA384 SHA256 SHA224 AES256 AES192 AES CAST5 BZIP2 ZLIB ZIP Uncompressed use-agent verify-options show-uid-validity list-options show-uid-validity sig-notation issuer-fpr@notations.openpgp.fifthhorseman.net=%g cert-digest-algo SHA512 .... . Generate a key: + [source,shell] .... % gpg --full-gen-key gpg (GnuPG) 2.1.8; Copyright (C) 2015 Free Software Foundation, Inc. This is free software: you are free to change and redistribute it. There is NO WARRANTY, to the extent permitted by law. Warning: using insecure memory! Please select what kind of key you want: (1) RSA and RSA (default) (2) DSA and Elgamal (3) DSA (sign only) (4) RSA (sign only) Your selection? 1 RSA keys may be between 1024 and 4096 bits long. What keysize do you want? (2048) 2048 <.> Requested keysize is 2048 bits Please specify how long the key should be valid. 0 = key does not expire = key expires in n days w = key expires in n weeks m = key expires in n months y = key expires in n years Key is valid for? (0) 3y <.> Key expires at Wed Nov 4 17:20:20 2015 MST Is this correct? (y/N) y GnuPG needs to construct a user ID to identify your key. Real name: Chucky Daemon <.> Email address: notreal@example.com Comment: You selected this USER-ID: "Chucky Daemon " Change (N)ame, (C)omment, (E)mail or (O)kay/(Q)uit? o You need a Passphrase to protect your secret key. .... <.> 2048-bit keys with a three-year expiration provide adequate protection at present (2013-12). http://danielpocock.com/rsa-key-sizes-2048-or-4096-bits[] describes the situation in more detail. <.> A three year key lifespan is short enough to obsolete keys weakened by advancing computer power, but long enough to reduce key management problems. <.> Use your real name here, preferably matching that shown on government-issued ID to make it easier for others to verify your identity. Text that may help others identify you can be entered in the `Comment` section. + After the email address is entered, a passphrase is requested. Methods of creating a secure passphrase are contentious. Rather than suggest a single way, here are some links to sites that describe various methods: http://world.std.com/~reinhold/diceware.html[], http://www.iusmentis.com/security/passphrasefaq/[], http://xkcd.com/936/[], http://en.wikipedia.org/wiki/Passphrase[]. ==== Protect the private key and passphrase. If either the private key or passphrase may have been compromised or disclosed, immediately notify mailto:accounts@FreeBSD.org[accounts@FreeBSD.org] and revoke the key. Committing the new key is shown in <>. [[kerberos-ldap]] == Kerberos and LDAP web Password for FreeBSD Cluster The FreeBSD cluster requires a Kerberos password to access certain services. The Kerberos password also serves as the LDAP web password, since LDAP is proxying to Kerberos in the cluster. Some of the services which require this include: * https://bugs.freebsd.org/bugzilla[Bugzilla] * https://ci.freebsd.org[Jenkins] To create a new Kerberos account in the FreeBSD cluster, or to reset a Kerberos password for an existing account using a random password generator: [source,shell] .... % ssh kpasswd.freebsd.org .... [NOTE] ==== This must be done from a machine outside of the FreeBSD.org cluster. ==== A Kerberos password can also be set manually by logging into `freefall.FreeBSD.org` and running: [source,shell] .... % kpasswd .... [NOTE] ==== Unless the Kerberos-authenticated services of the FreeBSD.org cluster have been used previously, `Client unknown` will be shown. This error means that the `ssh kpasswd.freebsd.org` method shown above must be used first to initialize the Kerberos account. ==== [[committer.types]] == Commit Bit Types The FreeBSD repository has a number of components which, when combined, support the basic operating system source, documentation, third party application ports infrastructure, and various maintained utilities. When FreeBSD commit bits are allocated, the areas of the tree where the bit may be used are specified. Generally, the areas associated with a bit reflect who authorized the allocation of the commit bit. Additional areas of authority may be added at a later date: when this occurs, the committer should follow normal commit bit allocation procedures for that area of the tree, seeking approval from the appropriate entity and possibly getting a mentor for that area for some period of time. [.informaltable] [cols="1,1,1", frame="none"] |=== |__Committer Type__ |__Responsible__ |__Tree Components__ |src |core@ |src/ |doc |doceng@ |doc/, ports/, src/ documentation |ports |portmgr@ |ports/ |=== Commit bits allocated prior to the development of the notion of areas of authority may be appropriate for use in many parts of the tree. However, common sense dictates that a committer who has not previously worked in an area of the tree seek review prior to committing, seek approval from the appropriate responsible party, and/or work with a mentor. Since the rules regarding code maintenance differ by area of the tree, this is as much for the benefit of the committer working in an area of less familiarity as it is for others working on the tree. Committers are encouraged to seek review for their work as part of the normal development process, regardless of the area of the tree where the work is occurring. === Policy for Committer Activity in Other Trees * All committers may modify [.filename]#src/share/misc/committers-*.dot#, [.filename]#src/usr.bin/calendar/calendars/calendar.freebsd#, and [.filename]#ports/astro/xearth/files#. * doc committers may commit documentation changes to [.filename]#src# files, such as man pages, READMEs, fortune databases, calendar files, and comment fixes without approval from a src committer, subject to the normal care and tending of commits. * Any committer may make changes to any other tree with an "Approved by" from a non-mentored committer with the appropriate bit. Mentored committers can provide a "Reviewed by" but not an "Approved by". * Committers can acquire an additional bit by the usual process of finding a mentor who will propose them to core, doceng, or portmgr, as appropriate. When approved, they will be added to 'access' and the normal mentoring period will ensue, which will involve a continuing of "Approved by" for some period. [[git-primer]] == Git Primer [[git-basics]] === Git basics There are many primers on how to use Git on the web. There's a lot of them (search for "Git primer"). https://danielmiessler.com/study/git/ and https://gist.github.com/williewillus/068e9a8543de3a7ef80adb2938657b6b are good overviews. The Git book is also complete, but much longer https://git-scm.com/book/en/v2. There is also this website https://ohshitgit.com/ for common traps and pitfalls of Git, in case you need guidance to fix things up. This document will assume that you've read through it and will try not to belabor the basics (though it will cover them briefly). [[git-mini-primer]] === Git Mini Primer This primer is less ambitiously scoped than the old Subversion Primer, but should cover the basics. ==== Scope If you want to download FreeBSD, compile it from sources, and generally keep up to date that way, this primer is for you. It covers getting the sources, updating the sources, bisecting and touches briefly on how to cope with a few local changes. It covers the basics, and tries to give good pointers to more in-depth treatment for when the reader finds the basics insufficient. Other sections of this guide cover more advanced topics related to contributing to the project. The goal of this section is to highlight those bits of Git needed to track sources. They assume a basic understanding of Git. There are many primers for Git on the web, but the https://git-scm.com/book/en/v2[Git Book] provides one of the better treatments. [[git-mini-primer-getting-started]] ==== Getting Started For Developers This section describes the read-write access for committers to push the commits from developers or contributors. ===== Daily use * Clone the repository: + [source,shell] .... % git clone -o freebsd --config remote.freebsd.fetch='+refs/notes/*:refs/notes/*' https://git.freebsd.org/${repo}.git .... + Then you should have the official mirrors as your remote: + [source,shell] .... % git remote -v freebsd https://git.freebsd.org/${repo}.git (fetch) freebsd https://git.freebsd.org/${repo}.git (push) .... * Configure the FreeBSD committer data: + The commit hook in repo.freebsd.org checks the "Commit" field matches the committer's information in FreeBSD.org. The easiest way to get the suggested config is by executing `/usr/local/bin/gen-gitconfig.sh` script on freefall: + [source,shell] .... % gen-gitconfig.sh [...] % git config user.name (your name in gecos) % git config user.email (your login)@FreeBSD.org .... * Set the push URL: + [source,shell] .... % git remote set-url --push freebsd git@gitrepo.freebsd.org:${repo}.git .... + Then you should have separated fetch and push URLs as the most efficient setup: + [source,shell] .... % git remote -v freebsd https://git.freebsd.org/${repo}.git (fetch) freebsd git@gitrepo.freebsd.org:${repo}.git (push) .... + Again, note that `gitrepo.freebsd.org` will be canonicalized to `repo.freebsd.org` in the future. * Install commit message template hook: + [source,shell] .... % fetch https://cgit.freebsd.org/src/plain/tools/tools/git/hooks/prepare-commit-msg -o .git/hooks % chmod 755 .git/hooks/prepare-commit-msg .... [[admin-branch]] ===== "admin" branch The `access` and `mentors` files are stored in an orphan branch, `internal/admin`, in each repository. Following example is how to check out the `internal/admin` branch to a local branch named `admin`: [source,shell] .... % git config --add remote.freebsd.fetch '+refs/internal/*:refs/internal/*' % git fetch % git checkout -b admin internal/admin .... Alternatively, you can add a worktree for the `admin` branch: [source,shell] .... git worktree add -b admin ../${repo}-admin internal/admin .... For browsing `internal/admin` branch on web: https://cgit.freebsd.org/${repo}/log/?h=internal/admin For pushing, either specify the full refspec: [source,shell] .... git push freebsd HEAD:refs/internal/admin .... Or set `push.default` to `freebsd` which will make `git push` to push the current branch back to its upstream by default, which is more suitable for our workflow: [source,shell] .... git config push.default freebsd .... ==== Keeping Current With The FreeBSD src Tree [[keeping_current]] First step: cloning a tree. This downloads the entire tree. There are two ways to download. Most people will want to do a deep clone of the repository. However, there are times when you may wish to do a shallow clone. ===== Branch names The branch names in the new Git repository are similar to the old names. For the stable branches, they are stable/X where X is the major release (like 11 or 12). The main branch in the new repository is 'main'. The main branch in the old GitHub mirror was 'master', but is now 'main'. Both reflect the defaults of Git at the time they were created. The 'main' branch is the default branch if you omit the '-b branch' or '--branch branch' options below. ===== Repositories Please see the <> for the latest information on where to get FreeBSD sources. $URL below can be obtained from that page. Note: The project doesn't use submodules as they are a poor fit for our workflows and development model. How we track changes in third-party applications is discussed elsewhere and generally of little concern to the casual user. ===== Deep Clone A deep clone pulls in the entire tree, as well as all the history and branches. It is the easiest to do. It also allows you to use Git's worktree feature to have all your active branches checked out into separate directories but with only one copy of the repository. [source,shell] .... % git clone -o freebsd $URL -b branch [dir] .... is how you make a deep clone. 'branch' should be one of the branches listed in the previous section. It is optional if it is the main branch. 'dir' is an optional directory to place it in (the default will be the name of the repo you are cloning (src, doc, etc)). You will want a deep clone if you are interested in the history, plan on making local changes, or plan on working on more than one branch. It is the easiest to keep up to date as well. If you are interested in the history, but are working with only one branch and are short on space, you can also use --single-branch to only download the one branch (though some merge commits will not reference the merged-from branch which may be important for some users who are interested in detailed versions of history). ===== Shallow Clone A shallow clone copies just the most current code, but none or little of the history. This can be useful when you need to build a specific revision of FreeBSD, or when you are just starting out and plan to track the tree more fully. You can also use it to limit history to only so many revisions. However, see below for a significant limitation of this approach. [source,shell] .... % git clone -o freebsd -b branch --depth 1 $URL [dir] .... This clones the repository, but only has the most recent version in the repository. The rest of the history is not downloaded. Should you change your mind later, you can do 'git fetch --unshallow' to get the old history. [WARNING] ==== When you make a shallow clone, you will lose the commit count in your uname output. This can make it more difficult to determine if your system needs to be updated when a security advisory is issued. ==== ===== Building Once you've downloaded, building is done as described in the handbook, eg: [source,shell] .... % cd src % make buildworld % make buildkernel % make installkernel % make installworld .... so that won't be covered in depth here. If you want to build a custom kernel, link:{handbook}#kernelconfig[the kernel config section] of the FreeBSD Handbook recommends creating a file MYKERNEL under sys/${ARCH}/conf with your changes against GENERIC. To have MYKERNEL disregarded by Git, it can be added to .git/info/exclude. ===== Updating To update both types of trees uses the same commands. This pulls in all the revisions since your last update. [source,shell] .... % git pull --ff-only .... will update the tree. In Git, a 'fast forward' merge is one that only needs to set a new branch pointer and doesn't need to re-create the commits. By always doing a 'fast forward' merge/pull, you'll ensure that you have an exact copy of the FreeBSD tree. This will be important if you want to maintain local patches. See below for how to manage local changes. The simplest is to use --autostash on the 'git pull' command, but more sophisticated options are available. ==== Selecting a Specific Version In Git, the 'git checkout' checks out both branches and specific versions. Git's versions are the long hashes rather than a sequential number. When you checkout a specific version, just specify the hash you want on the command line (the git log command can help you decide which hash you might want): [source,shell] .... % git checkout 08b8197a74 .... and you have that checked out. You will be greeted with a message similar to the following: [source,shell] .... Note: checking out '08b8197a742a96964d2924391bf9fdfeb788865d'. You are in a 'detached HEAD' state. You can look around, make experimental changes and commit them, and you can discard any commits you make in this state without impacting any branches by performing another checkout. If you want to create a new branch to retain commits you create, you may do so (now or later) by using -b with the checkout command again. Example: git checkout -b HEAD is now at 08b8197a742a hook gpiokeys.4 to the build .... where the last line is generated from the hash you are checking out and the first line of the commit message from that revision. The hash can be abbreviated to the shortest unique length. Git itself is inconsistent about how many digits it displays. ==== Bisecting Sometimes, things go wrong. The last version worked, but the one you just updated to does not. A developer may ask you to bisect the problem to track down which commit caused the regression. Git makes bisecting changes easy with a powerful 'git bisect' command. Here's a brief outline of how to use it. For more information, you can view https://www.metaltoad.com/blog/beginners-guide-git-bisect-process-elimination or https://git-scm.com/docs/git-bisect for more details. The man git-bisect page is good at describing what can go wrong, what to do when versions won't build, when you want to use terms other than 'good' and 'bad', etc, none of which will be covered here. `git bisect start` will start the bisection process. Next, you need to tell a range to go through. 'git bisect good XXXXXX' will tell it the working version and 'git bisect bad XXXXX' will tell it the bad version. The bad version will almost always be HEAD (a special tag for what you have checked out). The good version will be the last one you checked out. [TIP] ==== If you want to know the last version you checked out, you should use 'git reflog': [source,shell] .... 5ef0bd68b515 (HEAD -> main, freebsd/main, freebsd/HEAD) HEAD@{0}: pull --ff-only: Fast-forward a8163e165c5b (upstream/main) HEAD@{1}: checkout: moving from b6fb97efb682994f59b21fe4efb3fcfc0e5b9eeb to main ... .... shows me moving the working tree to the main branch (a816...) and then updating from upstream (to 5ef0...). In this case, bad would be HEAD (or 5rf0bd68) and good would be a8163e165. As you can see from the output, HEAD@{1} also often works, but isn't foolproof if you have done other things to your Git tree after updating, but before you discover the need to bisect. ==== Set the 'good' version first, then set the bad (though the order doesn't matter). When you set the bad version, it will give you some statistics on the process: [source,shell] .... % git bisect start % git bisect good a8163e165c5b % git bisect bad HEAD Bisecting: 1722 revisions left to test after this (roughly 11 steps) [c427b3158fd8225f6afc09e7e6f62326f9e4de7e] Fixup r361997 by balancing parens. Duh. .... You would then build/install that version. If it's good you'd type 'git bisect good' otherwise 'git bisect bad'. If the version doesn't compile, type 'git bisect skip'. You will get a similar message to the above after each step. When you are done, report the bad version to the developer (or fix the bug yourself and send a patch). 'git bisect reset' will end the process and return you back to where you started (usually tip of main). Again, the git-bisect manual (linked above) is a good resource for when things go wrong or for unusual cases. [[git-gpg-signing]] ==== Signing the commits, tags, and pushes, with GnuPG Git knows how to sign commits, tags, and pushes. When you sign a Git commit or a tag, you can prove that the code you submitted came from you and wasn't altered while you were transferring it. You also can prove that you submitted the code and not someone else. A more in-depth documentation on signing commits and tags can be found in the https://git-scm.com/book/en/v2/Git-Tools-Signing-Your-Work[Git Tools - Signing Your Work] chapter of the Git's book. The rationale behind signing pushes can be found in the https://github.com/git/git/commit/a85b377d0419a9dfaca8af2320cc33b051cbed04[commit that introduced the feature]. The best way is to simply tell Git you always want to sign commits, tags, and pushes. You can do this by setting a few configuration variables: [source,shell] .... % git config --add user.signingKey=LONG-KEY-ID % git config --add commit.gpgSign=true % git config --add tag.gpgSign=true % git config --add push.gpgSign=if-asked .... // push.gpgSign should probably be set to `yes` once we enable it, or be set with --global, so that it is enabled for all repositories. [NOTE] ====== To avoid possible collisions, make sure you give a long key id to Git. You can get the long id with: `gpg --list-secret-keys --keyid-format LONG`. ====== [TIP] ====== To use specific subkeys, and not have GnuPG to resolve the subkey to a primary key, attach `!` to the key. For example, to encrypt for the subkey `DEADBEEF`, use `DEADBEEF!`. ====== ===== Verifying signatures Commit signatures can be verified by running either `git verify-commit `, or `git log --show-signature`. Tag signatures can be verifed with `git verity-tag `, or `git tag -v `. //// Commented out for now until we decide what to do. Git pushes are a bit different, they live in a special ref in the repository. TODO: write how to verify them //// ==== Ports Considerations The ports tree operates the same way. The branch names are different and the repositories are in different locations. The cgit repository web interface for use with web browsers is at https://cgit.FreeBSD.org/ports/ . The production Git repository is at https://git.FreeBSD.org/ports.git and at ssh://anongit@git.FreeBSD.org/ports.git (or anongit@git.FreeBSD.org:ports.git). There is also a mirror on GitHub, see link:{handbook}mirrors/#mirrors[External mirrors] for an overview. The 'current' branch is 'main' . The quarterly branches are named 'yyyyQn' for year 'yyyy' and quarter 'n'. ===== Commit message formats A hook is available in the ports repository to help you write up your commit messages in https://cgit.freebsd.org/ports/tree/.hooks/prepare-commit-msg[.hooks/prepare-commit-message]. It can be enabled by running ``git config --add core.hooksPath .hooks``. The main point being that a commit message should be formatted in the following way: .... category/port: Summary. Description of why the changes where made. PR: 12345 .... [IMPORTANT] ==== The first line is the subject of the commit, it contains what port was changed, and a summary of the commit. It should contain 50 characters or less. A blank line should separate it from the rest of the commit message. The rest of the commit message should be wrapped at the 72 characters boundary. Another blank line should be added if there are any metadata fields, so that they are easily distinguishable from the commit message. ==== ==== Managing Local Changes This section addresses tracking local changes. If you have no local changes, you can stop reading now (it is the last section and OK to skip). One item that is important for all of them: all changes are local until pushed. Unlike Subversion, Git uses a distributed model. For users, for most things, there is very little difference. However, if you have local changes, you can use the same tool to manage them as you use to pull in changes from FreeBSD. All changes that you have not pushed are local and can easily be modified (git rebase, discussed below does this). ===== Keeping local changes The simplest way to keep local changes (especially trivial ones) is to use 'git stash'. In its simplest form, you use 'git stash' to record the changes (which pushes them onto the stash stack). Most people use this to save changes before updating the tree as described above. They then use 'git stash apply' to re-apply them to the tree. The stash is a stack of changes that can be examined with 'git stash list'. The git-stash man page (https://git-scm.com/docs/git-stash) has all the details. This method is suitable when you have tiny tweaks to the tree. When you have anything non trivial, you'll likely be better off keeping a local branch and rebasing. Stashing is also integrated with the 'git pull' command: just add '--autostash' to the command line. ===== Keeping a local branch [[keeping_a_local_branch]] It is much easier to keep a local branch with Git than Subversion. In Subversion you need to merge the commit, and resolve the conflicts. This is manageable, but can lead to a convoluted history that's hard to upstream should that ever be necessary, or hard to replicate if you need to do so. Git also allows one to merge, along with the same problems. That's one way to manage the branch, but it's the least flexible. In addition to merging, Git supports the concept of 'rebasing' which avoids these issues. The 'git rebase' command replays all the commits of a branch at a newer location on the parent branch. We will cover the most common scenarios that arise using it. ====== Create a branch Let's say you want to make a change to FreeBSD's ls command to never, ever do color. There are many reasons to do this, but this example will use that as a baseline. The FreeBSD ls command changes from time to time, and you'll need to cope with those changes. Fortunately, with Git rebase it usually is automatic. [source,shell] .... % cd src % git checkout main % git checkout -b no-color-ls % cd bin/ls % vi ls.c # hack the changes in % git diff # check the changes diff --git a/bin/ls/ls.c b/bin/ls/ls.c index 7378268867ef..cfc3f4342531 100644 --- a/bin/ls/ls.c +++ b/bin/ls/ls.c @@ -66,6 +66,7 @@ __FBSDID("$FreeBSD$"); #include #include #include +#undef COLORLS #ifdef COLORLS #include #include % # these look good, make the commit... % git commit ls.c .... The commit will pop you into an editor to describe what you've done. Once you enter that, you have your own **local** branch in the Git repo. Build and install it like you normally would, following the directions in the handbook. Git differs from other version control systems in that you have to tell it explicitly which files to commit. I have opted to do it on the commit command line, but you can also do it with 'git add' which many of the more in depth tutorials cover. ====== Time to update When it is time to bring in a new version, it is almost the same as w/o the branches. You would update like you would above, but there is one extra command before you update, and one after. The following assumes you are starting with an unmodified tree. It is important to start rebasing operations with a clean tree (Git usually requires this). [source,shell] .... % git checkout main % git pull --no-ff % git rebase -i main no-color-ls .... This will bring up an editor that lists all the commits in it. For this example, do not change it at all. This is typically what you are doing while updating the baseline (though you also use the Git rebase command to curate the commits you have in the branch). Once you are done with the above, you have to move the commits to ls.c forward from the old version of FreeBSD to the newer one. Sometimes there are merge conflicts. That is OK. Do not panic. Instead, handle them the same as any other merge conflicts. To keep it simple, I will just describe a common issue that may arise. A pointer to a more complete treatment can be found at the end of this section. Let's say the includes changes upstream in a radical shift to terminfo as well as a name change for the option. When you updated, you might see something like this: [source,shell] .... Auto-merging bin/ls/ls.c CONFLICT (content): Merge conflict in bin/ls/ls.c error: could not apply 646e0f9cda11... no color ls Resolve all conflicts manually, mark them as resolved with "git add/rm ", then run "git rebase --continue". You can instead skip this commit: run "git rebase --skip". To abort and get back to the state before "git rebase", run "git rebase --abort". Could not apply 646e0f9cda11... no color ls .... which looks scary. If you bring up an editor, you will see it is a typical 3-way merge conflict resolution that you may be familiar with from other source code systems (the rest of ls.c has been omitted): [source,shell] .... <<<<<<< HEAD #ifdef COLORLS_NEW #include ======= #undef COLORLS #ifdef COLORLS #include >>>>>>> 646e0f9cda11... no color ls .... The new code is first, and your code is second. The right fix here is to just add a #undef COLORLS_NEW before #ifdef and then delete the old changes: [source,shell] .... #undef COLORLS_NEW #ifdef COLORLS_NEW #include .... save the file. The rebase was interrupted, so you have to complete it: [source,shell] .... % git add ls.c % git rebase --continue .... which tells Git that ls.c has been fixed and to continue the rebase operation. Since there was a conflict, you will get kicked into the editor to update the commit message if necessary. If the commit message is still accurate, just exit the editor. If you get stuck during the rebase, do not panic. git rebase --abort will take you back to a clean slate. It is important, though, to start with an unmodified tree. An aside: The above mentioned 'git reflog' comes in handy here, as it will have a list of all the (intermediate) commits that you can view or inspect or cherry-pick. For more on this topic, https://www.freecodecamp.org/news/the-ultimate-guide-to-git-merge-and-git-rebase/ provides a rather extensive treatment. It is a good resource for issues that arise occasionally but are too obscure for this guide. ===== Switching to a Different FreeBSD Branch If you wish to shift from stable/12 to the current branch. If you have a deep clone, the following will suffice: [source,shell] .... % git checkout main % # build and install here... .... If you have a local branch, though, there are one or two caveats. First, rebase will rewrite history, so you will likely want to do something to save it. Second, jumping branches tends to cause more conflicts. If we pretend the example above was relative to stable/12, then to move to main, I'd suggest the following: [source,shell] .... % git checkout no-color-ls % git checkout -b no-color-ls-stable-12 # create another name for this branch % git rebase -i stable/12 no-color-ls --onto main .... What the above does is checkout no-color-ls. Then create a new name for it (no-color-ls-stable-12) in case you need to get back to it. Then you rebase onto the main branch. This will find all the commits to the current no-color-ls branch (back to where it meets up with the stable/12 branch) and then it will replay them onto the main branch creating a new no-color-ls branch there (which is why I had you create a place holder name). ===== Migrating from an existing Git clone If you have work based on a previous Git conversion or a locally running git-svn conversion, migrating to new repository can encounter problems because Git has no knowledge about the connection between the two. When you have only a few local changes, the easiest way would be to cherry-pick those changes to the new base: [source,shell] .... % git checkout main % git cherry-pick old_branch..your_branch .... Or alternatively, do the same thing with rebase: [source,shell] .... % git rebase --onto main master your_branch .... If you do have a lot of changes, you would probably want to perform a merge instead. The idea is to create a merge point that consolidates the history of the old_branch, and the new FreeBSD repository (main). You can find out by looking up the same commit that are found on both parents: [source,shell] .... % git show old_branch .... You will see a commit message, now search for that in the new branch: [source,shell] .... % git log --grep="commit message on old_branch" freebsd/main .... You would help locate the commit hash on the new main branch, create a helper branch (in the example we call it 'stage') from that hash: [source,shell] .... % git checkout -b stage _hash_found_from_git_log_ .... Then perform a merge of the old branch: [source,shell] .... % git merge -s ours -m "Mark old branch as merged" old_branch .... With that, it's possible to merge your work branch or the main branch in any order without problem. Eventually, when you are ready to commit your work back to main, you can perform a rebase to main, or do a squash commit by combining everything into one commit. [[mfc-with-git]] === MFC (Merge From Current) Procedures ==== Summary MFC workflow can be summarized as `git cherry-pick -x` plus `git commit --amend` to adjust the commit message. For multiple commits, use `git rebase -i` to squash them together and edit the commit message. ==== Single commit MFC [source,shell] .... % git checkout stable/X % git cherry-pick -x $HASH --edit .... For MFC commits, for example a vendor import, you would need to specify one parent for cherry-pick purposes. Normally, that would be the "first parent" of the branch you are cherry-picking from, so: [source,shell] .... % git checkout stable/X % git cherry-pick -x $HASH -m 1 --edit .... If things go wrong, you'll either need to abort the cherry-pick with `git cherry-pick --abort` or fix it up and do a `git cherry-pick --continue`. Once the cherry-pick is finished, push with `git push`. If you get an error due to losing the commit race, use `git pull --rebase` and try to push again. ==== MFC to RELENG branch MFCs to branches that require approval require a bit more care. The process is the same for either a typical merge or an exceptional direct commit. * Merge or direct commit to the appropriate `stable/X` branch first before merging to the `releng/X.Y` branch. * Use the hash that's in the `stable/X` branch for the MFC to `releng/X.Y` branch. * Leave both "cherry picked from" lines in the commit message. * Be sure to add the `Approved by:` line when you are in the editor. [source,shell] .... % git checkout releng/13.0 % git cherry-pick -x $HASH --edit .... If you forget to to add the `Approved by:` line, you can do a `git commit --amend` to edit the commit message before you push the change. ==== Multiple commit MFC [source,shell] .... % git checkout -b tmp-branch stable/X % for h in $HASH_LIST; do git cherry-pick -x $h; done % git rebase -i stable/X # mark each of the commits after the first as 'squash' # Update the commit message to reflect all elements of commit, if necessary. # Be sure to retain the "cherry picked from" lines. % git push freebsd HEAD:stable/X .... If the push fails due to losing the commit race, rebase and try again: [source,shell] .... % git checkout stable/X % git pull % git checkout tmp-branch % git rebase stable/X % git push freebsd HEAD:stable/X .... Once the MFC is complete, you can delete the temporary branch: [source,shell] .... % git checkout stable/X % git branch -d tmp-branch .... ==== MFC a vendor import Vendor imports are the only thing in the tree that creates a merge commit in the main line. Cherry picking merge commits into stable/XX presents an additional difficulty because there are two parents for a merge commit. Generally, you'll want the first parent's diff since that's the diff to mainline (though there may be some exceptions). [source,shell] .... % git cherry-pick -x -m 1 $HASH .... is typically what you want. This will tell cherry-pick to apply the correct diff. There are some, hopefully, rare cases where it's possible that the mainline was merged backwards by the conversion script. Should that be the case (and we've not found any yet), you'd change the above to '-m 2' to pickup the proper parent. Just do [source,shell] .... % git cherry-pick --abort % git cherry-pick -x -m 2 $HASH .... to do that. The `--aboort` will cleanup the failed first attempt. ==== Redoing a MFC If you do a MFC, and it goes horribly wrong and you want to start over, then the easiest way is to use `git reset --hard` like so: [source,shell] .... % git reset --hard freebsd/stable/12 .... though if you have some revs you want to keep, and others you don't, using 'git rebase -i' is better. ==== Considerations when MFCing When committing source commits to stable and releng branches, we have the following goals: * Clearly mark direct commits distinct from commits that land a change from another branch. * Avoid introducing known breakage into stable and releng branches. * Allow developers to determine which changes have or have not been landed from one branch to another. With Subversion, we used the following practices to achieve these goals: * Using 'MFC' and 'MFS' tags to mark commits that merged changes from another branch. * Squashing fixup commits into the main commit when merging a change. * Recording mergeinfo so that `svn mergeinfo --show-revs` worked. With Git, we will need to use different strategies to achieve the same goals. This document aims to define best practices when merging source commits using Git that achieve these goals. In general, we aim to use Git's native support to achieve these goals rather than enforcing practices built on Subversion's model. One general note: due to technical differences with Git, we will not be using Git "merge commits" (created via `git merge`) in stable or releng branches. Instead, when this document refers to "merge commits", it means a commit originally made to `main` that is replicated or "landed" to a stable branch, or a commit from a stable branch that is replicated to a releng branch with some variation of `git cherry-pick`. ==== Finding Eligible Hashes to MFC Git provides some built-in support for this via the `git cherry` and `git log --cherry` commands. These commands compare the raw diffs of commits (but not other metadata such as log messages) to determine if two commits are identical. This works well when each commit from head is landed as a single commit to a stable branch, but it falls over if multiple commits from main are squashed together as a single commit to a stable branch. There are a few options for resolving this: 1. We could ban squashing of commits and instead require that committers stage all of the fixup / follow-up commits to stable into a single push. This would still achieve the goal of stability in stable and releng branches since pushes are atomic and users doing a simple pull will never end up with a tree that has the main commit without the fixup(s). `git bisect` is also able to cope with this model via `git bisect skip`. 2. We could adopt a consistent style for describing MFCs and write our own tooling to wrap around `git cherry` to determine the list of eligible commits. A simple approach here might be to use the syntax from `git cherry-pick -x`, but require that a squashed commit list all of the hashes (one line per hash) at the end of the commit message. Developers could do this by using `git cherry-pick -x` of each individual commit into a branch and then use `git rebase` to squash the commits down into a single commit, but collecting the `-x` annotations at the end of the landed commit log. ==== Commit message standards ===== Marking MFCs The project has adopted the following practice for marking MFCs: * Use the `-x` flag with `git cherry-pick`. This adds a line to the commit message that includes the hash of the original commit when merging. Since it is added by Git directly, committers do not have to manually edit the commit log when merging. When merging multiple commits, keep all the "cherry picked from" lines. ===== Trim Metadata? One area that was not clearly documented with Subversion (or even CVS) is how to format metadata in log messages for MFC commits. Should it include the metadata from the original commit unchanged, or should it be altered to reflect information about the MFC commit itself? Historical practice has varied, though some of the variance is by field. For example, MFCs that are relevant to a PR generally include the PR field in the MFC so that MFC commits are included in the bug tracker's audit trail. Other fields are less clear. For example, Phabricator shows the diff of the last commit tagged to a review, so including Phabricator URLs replaces the `main` commit with the landed commits. The list of reviewers is also not clear. If a reviewer has approved a change to `main`, does that mean they have approved the MFC commit? Is that true if it's identical code only, or with merely trivial reworkes? It's clearly not true for more extensive reworks. Even for identical code what if the commit doesn't conflict but introduces an ABI change? A reviewer may have ok'd a commit for `main` due to the ABI breakage but may not approve of merging the same commit as-is. One will have to use one's best judgement until clear guidelines can be agreed upon. For MFCs regulated by re@, new metadata fields are added, such as the Approved by tag for approved commits. This new metadata will have to be added via `git commit --amend` or similar after the original commit has been reviewed and approved. We may also want to reserve some metadata fields in MFC commits such as Phabricator URLs for use by re@ in the future. Preserving existing metadata provides a very simple workflow. Developers can just use `git cherry-pick -x` without having to edit the log message. If instead we choose to adjust metadata in MFCs, developers will have to edit log messages explicitly via the use of `git cherry-pick --edit` or `git commit --amend`. However, as compared to svn, at least the existing commit message can be pre-populated and metadata fields can be added or removed without having to re-enter the entire commit message. The bottom line is that developers will likely need to curate their commit message for MFCs that are non-trivial. ==== Examples ===== Merging a Single Subversion Commit This walks through the process of merging a commit to stable/12 that was originally committed to head in Subversion. In this case, the original commit is r368685. The first step is to map the Subversion commit to a Git hash. Once you have fetched refs/notes/commits, you can pass the revision number to `git log --grep`: [source,shell] .... % git log main --grep 368685 commit ce8395ecfda2c8e332a2adf9a9432c2e7f35ea81 Author: John Baldwin Date: Wed Dec 16 00:11:30 2020 +0000 Use the 't' modifier to print a ptrdiff_t. Reviewed by: imp Obtained from: CheriBSD Sponsored by: DARPA Differential Revision: https://reviews.freebsd.org/D27576 Notes: svn path=/head/; revision=368685 .... Next, MFC the commit to a `stable/12` checkout: [source,shell] .... git checkout stable/12 git cherry-pick -x ce8395ecfda2c8e332a2adf9a9432c2e7f35ea81 --edit .... Git will invoke the editor. Use this to remove the metadata that only applied to the original commit (Phabricator URL and Reviewed by). After the editor saves the updated log message, Git completes the commit: [source,shell] .... [stable/12 3e3a548c4874] Use the 't' modifier to print a ptrdiff_t. Date: Wed Dec 16 00:11:30 2020 +0000 1 file changed, 1 insertion(+), 1 deletion(-) .... The contents of the MFCd commit can be examined via `git show`: [source,shell] .... % git show commit 3e3a548c487450825679e4bd63d8d1a67fd8bd2d (HEAD -> stable/12) Author: John Baldwin Date: Wed Dec 16 00:11:30 2020 +0000 Use the 't' modifier to print a ptrdiff_t. Obtained from: CheriBSD Sponsored by: DARPA (cherry picked from commit ce8395ecfda2c8e332a2adf9a9432c2e7f35ea81) diff --git a/sys/compat/linuxkpi/common/include/linux/printk.h b/sys/compat/linuxkpi/common/include/linux/printk.h index 31802bdd2c99..e6510e9e9834 100644 --- a/sys/compat/linuxkpi/common/include/linux/printk.h +++ b/sys/compat/linuxkpi/common/include/linux/printk.h @@ -68,7 +68,7 @@ print_hex_dump(const char *level, const char *prefix_str, printf("[%p] ", buf); break; case DUMP_PREFIX_OFFSET: - printf("[%p] ", (const char *)((const char *)buf - + printf("[%#tx] ", ((const char *)buf - (const char *)buf_old)); break; default: .... The MFC commit can now be published via `git push` [source,shell] .... % git push freebsd Enumerating objects: 17, done. Counting objects: 100% (17/17), done. Delta compression using up to 4 threads Compressing objects: 100% (7/7), done. Writing objects: 100% (9/9), 817 bytes | 204.00 KiB/s, done. Total 9 (delta 5), reused 1 (delta 1), pack-reused 0 To gitrepo-dev.FreeBSD.org:src.git 525bd9c9dda7..3e3a548c4874 stable/12 -> stable/12 .... ===== Merging a Single Subversion Commit with a Conflict This example is similar to the previous example except that the commit in question encounters a merge conflict. In this case, the original commit is r368314. As above, the first step is to map the Subversion commit to a Git hash: [source,shell] .... % git log main --grep 368314 commit 99963f5343a017e934e4d8ea2371a86789a46ff9 Author: John Baldwin Date: Thu Dec 3 22:01:13 2020 +0000 Don't transmit mbufs that aren't yet ready on TOE sockets. This includes mbufs waiting for data from sendfile() I/O requests, or mbufs awaiting encryption for KTLS. Reviewed by: np MFC after: 2 weeks Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D27469 Notes: svn path=/head/; revision=368314 .... Next, MFC the commit to a `stable/12` checkout: [source,shell] .... % git checkout stable/12 % git cherry-pick -x 99963f5343a017e934e4d8ea2371a86789a46ff9 --edit Auto-merging sys/dev/cxgbe/tom/t4_cpl_io.c CONFLICT (content): Merge conflict in sys/dev/cxgbe/tom/t4_cpl_io.c warning: inexact rename detection was skipped due to too many files. warning: you may want to set your merge.renamelimit variable to at least 7123 and retry the command. error: could not apply 99963f5343a0... Don't transmit mbufs that aren't yet ready on TOE sockets. hint: after resolving the conflicts, mark the corrected paths hint: with 'git add ' or 'git rm ' hint: and commit the result with 'git commit' .... In this case, the commit encountered a merge conflict in sys/dev/cxge/tom/t4_cpl_io.c as kernel TLS is not present in stable/12. Note that Git does not invoke an editor to adjust the commit message due to the conflict. `git status` confirms that this file has merge conflicts: [source,shell] .... % git status On branch stable/12 Your branch is up to date with 'upstream/stable/12'. You are currently cherry-picking commit 99963f5343a0. (fix conflicts and run "git cherry-pick --continue") (use "git cherry-pick --skip" to skip this patch) (use "git cherry-pick --abort" to cancel the cherry-pick operation) Unmerged paths: (use "git add ..." to mark resolution) both modified: sys/dev/cxgbe/tom/t4_cpl_io.c no changes added to commit (use "git add" and/or "git commit -a") .... After editing the file to resolve the conflict, `git status` shows the conflict as resolved: [source,shell] .... % git status On branch stable/12 Your branch is up to date with 'upstream/stable/12'. You are currently cherry-picking commit 99963f5343a0. (all conflicts fixed: run "git cherry-pick --continue") (use "git cherry-pick --skip" to skip this patch) (use "git cherry-pick --abort" to cancel the cherry-pick operation) Changes to be committed: modified: sys/dev/cxgbe/tom/t4_cpl_io.c .... The cherry-pick can now be completed: [source,shell] .... % git cherry-pick --continue .... Since there was a merge conflict, Git invokes the editor to adjust the commit message. Trim the metadata fields from the commit log from the original commit to head and save the updated log message. The contents of the MFC commit can be examined via `git show`: [source,shell] .... % git show commit 525bd9c9dda7e7c7efad2d4570c7fd8e1a8ffabc (HEAD -> stable/12) Author: John Baldwin Date: Thu Dec 3 22:01:13 2020 +0000 Don't transmit mbufs that aren't yet ready on TOE sockets. This includes mbufs waiting for data from sendfile() I/O requests, or mbufs awaiting encryption for KTLS. Sponsored by: Chelsio Communications (cherry picked from commit 99963f5343a017e934e4d8ea2371a86789a46ff9) diff --git a/sys/dev/cxgbe/tom/t4_cpl_io.c b/sys/dev/cxgbe/tom/t4_cpl_io.c index 8e8c2b8639e6..43861f10b689 100644 --- a/sys/dev/cxgbe/tom/t4_cpl_io.c +++ b/sys/dev/cxgbe/tom/t4_cpl_io.c @@ -746,6 +746,8 @@ t4_push_frames(struct adapter *sc, struct toepcb *toep, int drop) for (m = sndptr; m != NULL; m = m->m_next) { int n; + if ((m->m_flags & M_NOTAVAIL) != 0) + break; if (IS_AIOTX_MBUF(m)) n = sglist_count_vmpages(aiotx_mbuf_pages(m), aiotx_mbuf_pgoff(m), m->m_len); @@ -821,8 +823,9 @@ t4_push_frames(struct adapter *sc, struct toepcb *toep, int drop) /* nothing to send */ if (plen == 0) { - KASSERT(m == NULL, - ("%s: nothing to send, but m != NULL", __func__)); + KASSERT(m == NULL || (m->m_flags & M_NOTAVAIL) != 0, + ("%s: nothing to send, but m != NULL is ready", + __func__)); break; } @@ -910,7 +913,7 @@ t4_push_frames(struct adapter *sc, struct toepcb *toep, int drop) toep->txsd_avail--; t4_l2t_send(sc, wr, toep->l2te); - } while (m != NULL); + } while (m != NULL && (m->m_flags & M_NOTAVAIL) == 0); /* Send a FIN if requested, but only if there's no more data to send */ if (m == NULL && toep->flags & TPF_SEND_FIN) .... The MFC commit can now be published via `git push` [source,shell] .... git push freebsd Enumerating objects: 13, done. Counting objects: 100% (13/13), done. Delta compression using up to 4 threads Compressing objects: 100% (7/7), done. Writing objects: 100% (7/7), 819 bytes | 117.00 KiB/s, done. Total 7 (delta 6), reused 0 (delta 0), pack-reused 0 To gitrepo.FreeBSD.org:src.git f4d0bc6aa6b9..525bd9c9dda7 stable/12 -> stable/12 .... [[vendor-import-git]] === Vendor Imports with Git This section describes the vendor import procedure with Git in detail. ==== Branch naming convention All vendor branches and tags start with `vendor/`. These branches and tags are visible by default. [NOTE] ==== This chapter follows the convention that the `freebsd` origin is the origin name for the official FreeBSD Git repository. If you use a different convention, replace `freebsd` with the name you use instead in the examples below. ==== We will explore an example for updating NetBSD's mtree that is in our tree. The vendor branch for this is `vendor/NetBSD/mtree`. ==== Updating an old vendor import The vendor trees usually have only the subset of the third-party software that is appropriate to FreeBSD. These trees are usually tiny in comparison to the FreeBSD tree. Git worktrees are thus quite small and fast and the preferred method to use. Make sure that whatever directory you choose below (the `../mtree`) does not currently exist. [source,shell] .... % git worktree add ../mtree vendor/NetBSD/mtree .... ==== Update the Sources in the Vendor Branch Prepare a full, clean tree of the vendor sources. Import everything but merge only what is needed. This example assumes the NetBSD checked out from their GitHub mirror in `~/git/NetBSD`. Note that "upstream" might have added or removed files, so we want to make sure deletions are propagated as well. rsync(1) is commonly installed, so I'll use that. [source,shell] .... % cd ../mtree % rsync -va --del --exclude=".git" ~/git/NetBSD/usr.sbin/mtree/ . % git add -A % git status ... % git diff --staged ... % git commit -m"Vendor import of NetBSD's mtree at 2020-12-11" [vendor/NetBSD/mtree 8e7aa25fcf1] Vendor import of NetBSD's mtree at 2020-12-11 7 files changed, 114 insertions(+), 82 deletions(-) % git tag -a vendor/NetBSD/mtree/20201211 .... Note: I run the `git diff` and `git status` commands to make sure nothing weird was present. Also I used `-m` to illustrate, but you should compose a proper message in an editor (using a commit message template). It is also important to create an annotated tag, otherwise the push will be rejected. Only annotated tags are allowed to be pushed. The annotated tag gives you a chance to enter a commit message. Enter the version you are importing, along with any salient new features or fixes in that version. ==== Updating the FreeBSD Copy At this point you can push the import to vendor into our repo. [source,shell] .... % git push --follow-tags freebsd vendor/NetBSD/mtree .... `--follow-tags` tells `git push` to also push tags associated with the locally committed revision. ==== Updating the FreeBSD source tree Now you need to update the mtree in FreeBSD. The sources live in `contrib/mtree` since it is upstream software. [source,shell] .... % cd ../src % git subtree merge -P contrib/mtree vendor/NetBSD/mtree .... This would generate a subtree merge commit of `contrib/mtree` against the local `vendor/NetBSD/mtree` branch. If there were conflicts, you would need to fix them before committing. Include details about the changes being merged in the merge commit message. ==== Rebasing your change against latest FreeBSD source tree Because the current policy recommends against using merges, if the upstream FreeBSD `main` moved forward before you get a chance to push, you would have to redo the merge. Regular `git rebase` or `git pull --rebase` doesn't know how to rebase a merge commit **as a merge commit**, so instead of that you would have to recreate the commit. The easiest way to do this would be to create a side branch with the **contents** of the merged tree: [source,shell] .... % cd ../src % git fetch freebsd % git checkout -b merge_result % git merge freebsd/main .... Typically, there would be no merge conflicts here (because developers tend to work on different components). In the worst case scenario, you would still have to resolve merge conflicts, if there was any, but this should be really rare. Now, checkout `freebsd/main` again as `new_merge`, and redo the merge: [source,shell] .... % git checkout -b new_merge freebsd/main % git subtree merge -P contrib/mtree vendor/NetBSD/mtree .... Instead of resolving the conflicts, perform this instead: [source,shell] .... % git checkout merge_result . .... Which will overwrite the files with conflicts with the version found in `merge_result`. Examine the tree against `merge_result` to make sure that you haven't missed deleted files: [source,shell] .... % git diff merge_result .... ==== Pushing the changes Once you are sure that you have a set of deltas you think is good, you can push it to a fork off GitHub or GitLab for others to review. One nice thing about Git is that it allows you to publish rough drafts of your work for others to review. While phabricator is good for content review, publishing the updated vendor branch and merge commits lets others check the details as they will eventually appear in the repository. After review, when you are sure it is a good change, you can push it to the FreeBSD repo: [source,shell] .... % git push freebsd main .... === Creating a new vendor branch There are a number of ways to create a new vendor branch. The recommended way is to create a new repository and then merge that with FreeBSD. If one is importing `glorbnitz` into the FreeBSD tree, release 3.1415. For the sake of simplicity, we will not trim this release. It is a simple user command that puts the nitz device into different magical glorb states and is small enough trimming will not save much. ==== Create the repo [source,shell] .... % cd /some/where % mkdir glorbnitz % cd glorbnitz % git init % git checkout -b vendor/glorbnitz .... At this point, you have a new repo, where all new commits will go on the `vendor/glorbnitz` branch. Git experts can also do this right in their FreeBSD clone, using `git checkout --orphan vendor/glorbnitz` if they are more comfortable with that. ==== Copy the sources in Since this is a new import, you can just cp the sources in, or use tar or even rsync as shown above. And we will add everything, assuming no dot files. [source,shell] .... % cp -r ~/glorbnitz/* . % git add * .... At this point, you should have a pristine copy of glorbnitz ready to commit. [source,shell] .... % git commit -m"Import GlorbNitz frobnosticator revision 3.1415" .... As above, I used `-m` for simplicity, but you should likely create a commit message that explains what a Glorb is and why you'd use a Nitz to get it. Not everybody will know. But for your actual commit, you should follow the <> section instead of emulating the brief style used here. ==== Now import it into our repository Now you need to import the branch into our repository. [source,shell] .... % cd /path/to/freebsd/repo/src % git remote add glorbnitz /some/where/glorbnitz % git fetch glorbnitz vendor/glorbnitz .... Note the vendor/glorbnitz branch is in the repo. At this point the `/some/where/glorbnitz` can be deleted, if you like. It was only a means to an end. ==== Tag and push Steps from here on out are much the same as they are in the case of updating a vendor branch, though without the updating the vendor branch step. [source,shell] .... % git worktree add ../glorbnitz vendor/glorbnitz % cd ../glorbnitz % git tag --annotate vendor/glorbnitz/3.1415 # Make sure the commit is good with "git show" % git push --follow-tags freebsd vendor/glorbnitz .... By 'good' we mean: . All the right files are present . None of the wrong files are present . The vendor branch points at something sensible . The tag looks good, and is annotated . The commit message for the tag has a quick summary of what's new since the last tag ==== Time to finally merge it into the base tree [source,shell] .... % cd ../src % git subtree add -P contrib/glorbnitz vendor/glorbnitz # Make sure the commit is good with "git show" % git commit --amend # one last sanity check on commit message % git push freebsd .... Here 'good' means: . All the right files, and none of the wrong ones, were merged into contrib/glorbnitz. . No other changes are in the tree. . The commit messages look <>. It should contain a summary of what's changed since the last merge to the FreeBSD main line and any caveats. . UPDATING should be updated if there is anything of note, such as user visible changes, important upgrade concerns, etc. [NOTE] ==== This hasn't connected `glorbnitz` to the build yet. How so do that is specific to the software being imported and is beyond the scope of this tutorial. ==== === FreeBSD Src Committer Transition Guide This section is designed to walk people through the conversion process from Subversion to Git, written from the source committer's point of view. ==== Migrating from a Subversion tree This section will cover a couple of common scenarios for migrating from using the FreeBSD Subversion repo to the FreeBSD source Git repo. The FreeBSD Git conversion is still in beta status, so some minor things may change between this and going into production. The first thing to do is install Git. Any version of Git will do, though the latest one in ports / packages generally will be good. Either build it from ports, or install it using pkg (though some folks might use `su` or `doas` instead of `sudo`): [source,shell] .... % sudo pkg install git .... ===== No staged changes migration If you have no changes pending, the migration is straightforward. In this, you abandon the Subversion tree and clone the Git repository. It's likely best to retain your Subversion tree, in case there's something you've forgotten about there. First, let's clone the repository: [source,shell] .... % git clone -o freebsd --config remote.freebsd.fetch='+refs/notes/*:refs/notes/*' https://git.freebsd.org/src.git freebsd-src .... will create a clone of the FreeBSD src repository into a subdirectory called `freebsd-src` and include the 'notes' about the revisions. We are currently mirroring the source repository to https://github.com/freebsd/freebsd-src.git as well. https://github.com/freebsd/freebsd-legacy.git has the old GitHub mirror with the old hashes should you need that for your migration. The GitHub `master` branch has been frozen. As the default in Git has changed, we've shifted from `master` to `main`; the new repository uses `main`. We also mirror the repository to GitLab at https://gitlab.com/FreeBSD/src.git . It's useful to have the old Subversion revisions available. This data is stored using Git notes, but Git doesn't fetch those by default. The --config and the argument above changed the default to fetch the notes. If you've cloned the repository without this, or wish to add notes to a previously cloned repository, use the following commands: [source,shell] .... % git config --add remote.freebsd.fetch "+refs/notes/*:refs/notes/*" % git fetch .... At this point you have the src checked out into a Git tree, ready to do other things. ===== But I have changes that I've not committed If you are migrating from a tree that has changes you've not yet committed to FreeBSD, you'll need to follow the steps from the previous section first, and then follow these. [source,shell] .... % cd path-to-svn-checkout-tree % svn diff > /tmp/src.diff % cd _mumble_/freebsd-src % git checkout -b working .... This will create a diff of your current changes. The last command creates a branch called `working` though you can call it whatever you want. [source,shell] .... % git apply /tmp/src.diff .... this will apply all your pending changes to the working tree. This doesn't commit the change, so you'll need to make this permanent: [source,shell] .... % git add _files_ % git commit .... The last command will commit these changes to the branch. The editor will prompt you for a commit message. Enter one as if you were committing to FreeBSD. At this point, your work is preserved, and in the Git repository. ===== Keeping current So, time passes. It's time now to update the tree for the latest changes upstream. When you checkout `main` make sure that you have no diffs. It's a lot easier to commit those to a branch (or use `git stash`) before doing the following. If you are used to `git pull`, we strongly recommend using the `--ff-only` option, and further setting it as the default option. Alternatively, `git pull --rebase` is useful if you have changes staged in the main branch. [source,shell] .... % git config --global pull.ff only .... You may need to omit the --global if you want this setting to apply to only this repository. [source,shell] .... % cd freebsd-src % git checkout main % git pull (--ff-only|--rebase) .... There is a common trap, that the combination command `git pull` will try to perform a merge, which would sometimes creates a merge commit that didn't exist before. This can be harder to recover from. The longer form is also recommended. [source,shell] .... % cd freebsd-src % git checkout main % git fetch freebsd % git merge --ff-only freebsd/main .... These commands reset your tree to the main branch, and then update it from where you pulled the tree from originally. It's important to switch to `main` before doing this so it moves forward. Now, it's time to move the changes forward: [source,shell] .... % git rebase -i main working .... This will bring up an interactive screen to change the defaults. For now, just exit the editor. Everything should just apply. If not, then you'll need to resolve the diffs. https://docs.github.com/en/free-pro-team@latest/github/using-git/resolving-merge-conflicts-after-a-git-rebase[This github document] can help you navigate this process. ===== Time to push changes upstream First, ensure that the push URL is properly configured for the upstream repository. [source,shell] .... % git remote set-url --push freebsd ssh://git@gitrepo.freebsd.org/src.git .... Then, verify that user name and email are configured right. We require that they exactly match the passwd entry in FreeBSD cluster. Use [source,shell] .... freefall% gen-gitconfig.sh .... on freefall.freebsd.org to get a recipe that you can use directly, assuming /usr/local/bin is in the PATH. The below command merges the 'working' branch into the upstream main line. It's important that you curate your changes to be just like you want them in the FreeBSD source repo before doing this. [source,shell] .... % git push freebsd working:main .... If your push is rejected due to losing a commit race, rebase your branch before trying again: [source,shell] .... % git checkout working % git fetch freebsd % git rebase freebsd/main % git push freebsd working:main .... ===== Finding the Subversion Revision You'll need to make sure that you've fetched the notes (see the `No staged changes migration` section above for details. Once you have these, notes will show up in the git log command like so: [source,shell] .... % git log .... If you have a specific version in mind, you can use this construct: [source,shell] .... % git log --grep revision=XXXX .... to find the specific revision. The hex number after 'commit' is the hash you can use to refer to this commit. ==== Migrating from GitHub fork Note: as of this writing, https://github.com/freebsd/freebsd-src is mirroring all official branches, along with a `master` branch which is the legacy svn2git result. The `master` branch will not be updated anymore, and the link:https://github.com/freebsd/freebsd-src/commit/de1aa3dab23c06fec962a14da3e7b4755c5880cf[last commit] contains the instructions for migrating to the new `main` branch. We'll retain the `master` branch for a certain time, but in the future it will only be kept in the link:https://github.com/freebsd/freebsd-legacy[freebsd-legacy] repository. When migrating branches from a GitHub fork from the old GitHub mirror to the official repo, the process is straight forward. This assumes that you have a `freebsd` upstream pointing to GitHub, adjust if necessary. This also assumes a clean tree before starting... ===== Add the new `freebsd` upstream repository: [source,shell] .... % git remote add freebsd https://git.freebsd.org/src.git % git fetch freebsd % git checkout --track freebsd/main .... ===== Rebase all your WIP branches. For each branch FOO, do the following after fetching the `freebsd` sources and creating a local `main` branch with the above checkout: [source,shell] .... % git rebase -i freebsd/master FOO --onto main .... And you'll now be tracking the official repository. You can then follow the `Keeping Current` section above to stay up to date. If you need to then commit work to FreeBSD, you can do so following the `Time to push changes upstream` instructions. You'll need to do the following once to update the push URL if you are a FreeBSD committer: [source,shell] .... % git remote set-url --push freebsd ssh://git@gitrepo.freebsd.org/src.git .... (note that gitrepo.freebsd.org will be change to repo.freebsd.org in the future.) You will also need to add `freebsd` as the location to push to. The author recommends that your upstream GitHub repository remain the default push location so that you only push things into FreeBSD you intend to by making it explicit. [[git-faq]] === Git FAQ This section provides a number of targeted answers to questions that are likely to come up often for users and developers. [NOTE] ==== We use the common convention of having the origin for the FreeBSD repository being 'freebsd' rather than the default 'origin' to allow people to use that for their own development and to minimize "whoopse" pushes to the wrong repository. ==== ==== Users ===== How do I track -current and -stable with only one copy of the repository? **Q:** Although disk space is not a huge issue, it's more efficient to use only one copy of the repository. With SVN mirroring, I could checkout multiple trees from the same repository. How do I do this with Git? **A:** You can use Git worktrees. There's a number of ways to do this, but the simplest way is to use a clone to track -current, and a worktree to track stable releases. While using a 'bare repository' has been put forward as a way to cope, it's more complicated and will not be documented here. First, you need to clone the FreeBSD repository, shown here cloning into `freebsd-current` to reduce confusion. $URL is whatever mirror works best for you: [source,shell] .... % git clone -o freebsd --config remote.freebsd.fetch='+refs/notes/*:refs/notes/*' $URL freebsd-current .... then once that's cloned, you can simply create a worktree from it: [source,shell] .... % cd freebsd-current % git worktree add ../freebsd-stable-12 stable/12 .... this will checkout `stable/12` into a directory named `freebsd-stable-12` that's a peer to the `freebsd-current` directory. Once created, it's updated very similarly to how you might expect: [source,shell] .... % cd freebsd-current % git checkout main % git pull --ff-only # changes from upstream now local and current tree updated % cd ../freebsd-stable-12 % git merge --ff-only freebsd/stable/12 # now your stable/12 is up to date too .... I recommend using `--ff-only` because it's safer and you avoid accidentally getting into a 'merge nightmare' where you have an extra change in your tree, forcing a complicated merge rather than a simple one. Here's https://adventurist.me/posts/00296[a good writeup] that goes into more detail. ==== Developers ===== Ooops! I committed to `main` instead of a branch. **Q:** From time to time, I goof up and commit to main instead of to a branch. What do I do? **A:** First, don't panic. Second, don't push. In fact, you can fix almost anything if you haven't pushed. All the answers in this section assume no push has happened. The following answer assumes you committed to `main` and want to create a branch called `issue`: [source,shell] .... % git branch issue # Create the 'issue' branch % git reset --hard freebsd/main # Reset 'main' back to the official tip % git checkout issue # Back to where you were .... ===== Ooops! I committed something to the wrong branch! **Q:** I was working on feature on the `wilma` branch, but accidentally committed a change relevant to the `fred` branch in 'wilma'. What do I do? **A:** The answer is similar to the previous one, but with cherry picking. This assumes there's only one commit on wilma, but will generalize to more complicated situations. It also assumes that it's the last commit on wilma (hence using wilma in the `git cherry-pick` command), but that too can be generalized. [source,shell] .... # We're on branch wilma % git checkout fred # move to fred branch % git cherry-pick wilma # copy the misplaced commit % git checkout wilma # go back to wilma branch % git reset --hard HEAD^ # move what wilma refers to back 1 commit .... Git experts would first rewind the wilma branch by 1 commit, switch over to fred and then use `git reflog` to see what that 1 deleted commit was and cherry-pick it over. **Q:** But what if I want to commit a few changes to `main`, but keep the rest in `wilma` for some reason? **A:** The same technique above also works if you are wanting to 'land' parts of the branch you are working on into `main` before the rest of the branch is ready (say you noticed an unrelated typo, or fixed an incidental bug). You can cherry pick those changes into main, then push to the parent repository. Once you've done that, cleanup couldn't be simpler: just `git rebase -i`. Git will notice you've done this and skip the common changes automatically (even if you had to change the commit message or tweak the commit slightly). There's no need to switch back to wilma to adjust it: just rebase! **Q:** I want to split off some changes from branch `wilma` into branch `fred` **A:** The more general answer would be the same as the previous. You'd checkout/create the `fred` branch, cherry pick the changes you want from `wilma` one at a time, then rebase `wilma` to remove those changes you cherry picked. `git rebase -i main wilma` will toss you into an editor, and remove the `pick` lines that correspond to the commits you copied to `fred`. If all goes well, and there are no conflicts, you're done. If not, you'll need to resolve the conflicts as you go. The other way to do this would be to checkout `wilma` and then create the branch `fred` to point to the same point in the tree. You can then `git rebase -i` both these branches, selecting the changes you want in `fred` or `wilma` by retaining the pick likes, and deleting the rest from the editor. Some people would create a tag/branch called `pre-split` before starting in case something goes wrong in the split. You can undo it with the following sequence: [source,shell] .... % git checkout pre-split # Go back % git branch -D fred # delete the fred branch % git checkout -B wilma # reset the wilma branch % git branch -d pre-split # Pretend it didn't happen .... The last step is optional. If you are going to try again to split, you'd omit it. **Q:** But I did things as I read along and didn't see your advice at the end to create a branch, and now `fred` and `wilma` are all screwed up. How do I find what `wilma` was before I started. I don't know how many times I moved things around. **A:** All is not lost. You can figure out it, so long as it hasn't been too long, or too many commits (hundreds). So I created a wilma branch and committed a couple of things to it, then decided I wanted to split it into fred and wilma. Nothing weird happened when I did that, but let's say it did. The way to look at what you've done is with the `git reflog`: [source,shell] .... % git reflog 6ff9c25 (HEAD -> wilma) HEAD@{0}: rebase -i (finish): returning to refs/heads/wilma 6ff9c25 (HEAD -> wilma) HEAD@{1}: rebase -i (start): checkout main 869cbd3 HEAD@{2}: rebase -i (start): checkout wilma a6a5094 (fred) HEAD@{3}: rebase -i (finish): returning to refs/heads/fred a6a5094 (fred) HEAD@{4}: rebase -i (pick): Encourage contributions 1ccd109 (freebsd/main, main) HEAD@{5}: rebase -i (start): checkout main 869cbd3 HEAD@{6}: rebase -i (start): checkout fred 869cbd3 HEAD@{7}: checkout: moving from wilma to fred 869cbd3 HEAD@{8}: commit: Encourage contributions ... % .... Here we see the changes I've made. You can use it to figure out where things went wrong. I'll just point out a few things here. The first one is that HEAD@{X} is a 'commitish' thing, so you can use that as an argument to a command. Although if that command commits anything to the repository, the X numbers change. You can also use the hash (first column). Next, 'Encourage contributions' was the last commit I made to `wilma` before I decided to split things up. You can also see the same hash is there when I created the `fred` branch to do that. I started by rebasing `fred` and you see the 'start', each step, and the 'finish' for that process. While we don't need it here, you can figure out exactly what happened. Fortunately, to fix this, you can follow the prior answer's steps, but with the hash `869cbd3` instead of `pre-split`. While that seems a bit verbose, it's easy to remember since you're doing one thing at a time. You can also stack: [source,shell] .... % git checkout -B wilma 869cbd3 % git branch -D fred .... and you are ready to try again. The 'checkout -B' with the hash combines checking out and creating a branch for it. The -B instead of -b forces the movement of a pre-existing branch. Either way works, which is what's great (and awful) about Git. One reason I tend to use `git checkout -B xxxx hash` instead of checking out the hash, and then creating / moving the branch is purely to avoid the slightly distressing message about detached heads: [source,shell] .... % git checkout 869cbd3 M faq.md Note: checking out '869cbd3'. You are in 'detached HEAD' state. You can look around, make experimental changes and commit them, and you can discard any commits you make in this state without impacting any branches by performing another checkout. If you want to create a new branch to retain commits you create, you may do so (now or later) by using -b with the checkout command again. Example: git checkout -b HEAD is now at 869cbd3 Encourage contributions % git checkout -B wilma .... this produces the same effect, but I have to read a lot more and severed heads aren't an image I like to contemplate. ===== Ooops! I did a `git pull` and it created a merge commit, what do I do? **Q:** I was on autopilot and did a `git pull` for my development tree and that created a merge commit on the mainline. How do I recover? **A:** This can happen when you invoke the pull with your development branch checked out. Right after the pull, you will have the new merge commit checked out. Git supports a `HEAD^#` syntax to examine the parents of a merge commit: [source,shell] .... git log --oneline HEAD^1 # Look at the first parent's commits git log --oneline HEAD^2 # Look at the second parent's commits .... From those logs, you can easily identify which commit is your development work. Then you simply reset your branch to the corresponding `HEAD^#`: [source,shell] .... git reset --hard HEAD^2 .... **Q:** But I also need to fix my `main` branch. How do I do that? **A:** Git keeps track of the remote repository branches in a `freebsd/` namespace. To fix your `main` branch, just make it point to the remote's `main`: [source,shell] .... git branch -f main freebsd/main .... There's nothing magical about branches in Git: they are just labels on a graph that are automatically moved forward by making commits. So the above works because you're just moving a label. There's no metadata about the branch that needs to be preserved due to this. ===== Mixing and matching branches **Q:** So I have two branches `worker` and `async` that I'd like to combine into one branch called `feature` while maintaining the commits in both. **A:** This is a job for cherry pick. [source,shell] .... % git checkout worker % git checkout -b feature # create a new branch % git cherry-pick main..async # bring in the changes .... You now have a new branch called `feature`. This branch combines commits from both branches. You can further curate it with `git rebase`. **Q:** I have a branch called `driver` and I'd like to break it up into `kernel` and `userland` so I can evolve them separately and commit each branch as it becomes ready. **A:** This takes a little bit of prep work, but `git rebase` will do the heavy lifting here. [source,shell] .... % git checkout driver # Checkout the driver % git checkout -b kernel # Create kernel branch % git checkout -b userland # Create userland branch .... Now you have two identical branches. So, it's time to separate out the commits. We'll assume first that all the commits in `driver` go into either the `kernel` or the `userland` branch, but not both. [source,shell] .... % git rebase -i main kernel .... and just include the changes you want (with a 'p' or 'pick' line) and just delete the commits you don't (this sounds scary, but if worse comes to worse, you can throw this all away and start over with the `driver` branch since you've not yet moved it). [source,shell] .... % git rebase -i main userland .... and do the same thing you did with the `kernel` branch. **Q:** Oh great! I followed the above and forgot a commit in the `kernel` branch. How do I recover? **A:** You can use the `driver` branch to find the hash of the commit is missing and cherry pick it. [source,shell] .... % git checkout kernel % git log driver % git cherry-pick $HASH .... **Q:** OK. I have the same situation as the above, but my commits are all mixed up. I need parts of one commit to go to one branch and the rest to go to the other. In fact, I have several. Your rebase method to select sounds tricky. **A:** In this situation, you'd be better off to curate the original branch to separate out the commits, and then use the above method to split the branch. So let's assume that there's just one commit with a clean tree. You can either use `git rebase` with an `edit` line, or you can use this with the commit on the tip. The steps are the same either way. The first thing we need to do is to back up one commit while leaving the changes uncommitted in the tree: [source,shell] .... % git reset HEAD^ .... Note: Do not, repeat do not, add `--hard` here since that also removes the changes from your tree. Now, if you are lucky, the change needing to be split up falls entirely along file lines. In that case you can just do the usual `git add` for the files in each group than do a `git commit`. Note: when you do this, you'll lose the commit message when you do the reset, so if you need it for some reason, you should save a copy (though `git log $HASH` can recover it). If you are not lucky, you'll need to split apart files. There's another tool to do that which you can apply one file at a time. [source,shell] .... git add -i foo/bar.c .... will step through the diffs, prompting you, one at time, whether to include or exclude the hunk. Once you're done, `git commit` and you'll have the remainder in your tree. You can run it multiple times as well, and even over multiple files (though I find it easier to do one file at a time and use the `git rebase -i` to fold the related commits together). ==== Cloning and Mirroring **Q:** I'd like to mirror the entire Git repository, how do I do that? **A:** If all you want to do is mirror, then [source,shell] .... % git clone --mirror $URL .... will do the trick. However, there are two disadvantages to this if you want to use it for anything other than a mirror you'll reclone. First, this is a 'bare repository' which has the repository database, but no checked out worktree. This is great for mirroring, but terrible for day to day work. There's a number of ways around this with 'git worktree': [source,shell] .... % git clone --mirror https://git.freebsd.org/ports.git ports.git % cd ports.git % git worktree add ../ports main % git worktree add ../quarterly branches/2020Q4 % cd ../ports .... But if you aren't using your mirror for further local clones, then it's a poor match. The second disadvantage is that Git normally rewrites the refs (branch name, tags, etc) from upstream so that your local refs can evolve independently of upstream. This means that you'll lose changes if you are committing to this repository on anything other than private project branches. **Q:** So what can I do instead? **A:** Well, you can stuff all of the upstream repository's refs into a private namespace in your local repository. Git clones everything via a 'refspec' and the default refspec is: [source,shell] .... fetch = +refs/heads/*:refs/remotes/freebsd/* .... which says just fetch the branch refs. However, the FreeBSD repository has a number of other things in it. To see those, you can add explicit refspecs for each ref namespace, or you can fetch everything. To setup your repository to do that: [source,shell] .... git config --add remote.freebsd.fetch '+refs/*:refs/freebsd/*' .... which will put everything in the upstream repository into your local repository's 'refs/freebsd/' namespace. Please note, that this also grabs all the unconverted vendor branches and the number of refs associated with them is quite large. You'll need to refer to these 'refs' with their full name because they aren't in and of Git's regular namespaces. [source,shell] .... git log refs/freebsd/vendor/zlib/1.2.10 .... would look at the log for the vendor branch for zlib starting at 1.2.10. === Collaborating with others One of the keys to good software development on a project as large as FreeBSD is the ability to collaborate with others before you push your changes to the tree. The FreeBSD project's Git repositories do not, yet, allow user-created branches to be pushed to the repository, and therefore if you wish to share your changes with others you must use another mechanism, such as a hosted GitLab or GitHub, in order to share changes in a user-generated branch. The following instructions show how to set up a user-generated branch, based on the FreeBSD main branch, and push it to GitHub. Before you begin, make sure that your local Git repo is up to date and has the correct origins set <> [source,shell] ```` % git remote -v freebsd https://git.freebsd.org/src.git (fetch) freebsd ssh://git@gitrepo.freebsd.org/src.git (push) ```` The first step is to create a fork of https://github.com/freebsd/freebsd-src[FreeBSD] on GitHub following these https://docs.github.com/en/github/getting-started-with-github/fork-a-repo[guidelines]. The destination of the fork should be your own, personal, GitHub account (gvnn3 in my case). Now add a remote on your local system that points to your fork: [source,shell] .... % git remote add github git@github.com:gvnn3/freebsd-src.git % git remote -v github git@github.com:gvnn3/freebsd-src.git (fetch) github git@github.com:gvnn3/freebsd-src.git (push) freebsd https://git.freebsd.org/src.git (fetch) freebsd ssh://git@gitrepo.freebsd.org/src.git (push) .... With this in place you can create a branch <> [source,shell] .... % git checkout -b gnn-pr2001-fix .... Make whatever modifications you wish in your branch. Build, test, and once you're ready to collaborate with others it's time to push your changes into your hosted branch. Before you can push you'll have to set the appropriate upstream, as Git will tell you the first time you try to push to your +github+ remote: [source,shell] .... % git push github fatal: The current branch gnn-pr2001-fix has no upstream branch. To push the current branch and set the remote as upstream, use git push --set-upstream github gnn-pr2001-fix .... Setting the push as +git+ advises allows it to succeed: [source,shell] .... % git push --set-upstream github gnn-feature Enumerating objects: 20486, done. Counting objects: 100% (20486/20486), done. Delta compression using up to 8 threads Compressing objects: 100% (12202/12202), done. Writing objects: 100% (20180/20180), 56.25 MiB | 13.15 MiB/s, done. Total 20180 (delta 11316), reused 12972 (delta 7770), pack-reused 0 remote: Resolving deltas: 100% (11316/11316), completed with 247 local objects. remote: remote: Create a pull request for 'gnn-feature' on GitHub by visiting: remote: https://github.com/gvnn3/freebsd-src/pull/new/gnn-feature remote: To github.com:gvnn3/freebsd-src.git * [new branch] gnn-feature -> gnn-feature Branch 'gnn-feature' set up to track remote branch 'gnn-feature' from 'github'. .... Subsequent changes to the same branch will push correctly by default: [source,shell] .... % git push Enumerating objects: 4, done. Counting objects: 100% (4/4), done. Delta compression using up to 8 threads Compressing objects: 100% (2/2), done. Writing objects: 100% (3/3), 314 bytes | 1024 bytes/s, done. Total 3 (delta 1), reused 1 (delta 0), pack-reused 0 remote: Resolving deltas: 100% (1/1), completed with 1 local object. To github.com:gvnn3/freebsd-src.git 9e5243d7b659..cf6aeb8d7dda gnn-feature -> gnn-feature .... At this point your work is now in your branch on +GitHub+ and you can share the link with other collaborators. === Landing a github pull request This section documents how to land a GitHub pull request that's submitted against the FreeBSD Git mirrors at GitHub. While this is not an official way to submit patches at this time, sometimes good fixes come in this way and it is easiest just to bring them into a committer's tree and have them pushed into the FreeBSD's tree from there. Similar steps can be used to pull branches from other repositories and land those. When committing pull requests from others, one should take extra care to examine all the changes to ensure they are exactly as represented. Before beginning, make sure that the local Git repo is up to date and has the correct origins set <> In addition, make sure to have the following origins: [source,shell] .... % git remote -v freebsd https://git.freebsd.org/src.git (fetch) freebsd ssh://git@gitrepo.freebsd.org/src.git (push) github https://github.com/freebsd/freebsd-src (fetch) github https://github.com/freebsd/freebsd-src (fetch) .... Often pull requests are simple: requests that contain only a single commit. In this case, a streamlined approach may be used, though the approach in the prior section will also work. Here, a branch is created, the change is cherry picked, the commit message adjusted, and sanity-checked before being pushed. The branch `staging` is used in this example but it can be any name. This technique works for any number of commits in the pull request, especailly when the changes apply cleanly to the FreeBSD tree. However, when there's multiple commits, especially when minor adjustments are needed, `git rebase -i` works better than `git cherry-pick`. Briefly, these commands create a branch; cherry-picks the changes from the pull request; tests it; adjusts the commit messages; and fast forward merges it back to `main`. The PR number is `$PR` below. When adjusting the message, add `Pull Request: https://github.com/freebsd-src/pull/$PR`. [source,shell] .... % git fetch github pull/$PR/head:staging % git rebase -i main staging # to move the staging branch forward, adjust commit message here % git checkout main % git pull --ff-only # to get the latest if time has passed % git checkout main % git merge --ff-only staging % git push freebsd --push-option=confirm-author .... [.procedure] ==== For complicated pull requests that have multiple commits with conflicts, follow the following outline. . checkout the pull request `git checkout github/pull/XXX` . create a branch to rebase `git checkout -b staging` . rebase the `staging` branch to the latest `main` with `git rebase -i main staging` . resolve conflicts and do whatever testing is needed . fast forward the `staging` branch into `main` as above . final sanity check of changes to make sure all is well . push to FreeBSD's Git repository. This will also work when bringing branches developed elsewhere into the local tree for committing. ==== Once finished with the pull request, close it using GitHub's web interface. It is worth noting that if your `github` origin uses `https://`, the only step you'll need a GitHub account for is closing the pull request. [[vcs-history]] == Version Control History The project has moved to <>. The FreeBSD source repository switched from CVS to Subversion on May 31st, 2008. The first real SVN commit is __r179447__. The source repository switched from Subversion to Git on December 23rd, 2020. The last real svn commit is __r368820__. The first real git commit hash is __5ef5f51d2bef80b0ede9b10ad5b0e9440b60518c__ The FreeBSD `doc/www` repository switched from CVS to Subversion on May 19th, 2012. The first real SVN commit is __r38821__. The documentation repository switched from Subversion to Git on December 8th, 2020. The last SVN commit is __r54737__. The first real git commit hash is __3be01a475855e7511ad755b2defd2e0da5d58bbe__. The FreeBSD `ports` repository switched from CVS to Subversion on July 14th, 2012. The first real SVN commit is __r300894__. The ports repository switched from Subversion to Git on April 6, 2021. The last SVN commit is __r569609__ The first real git commit hash is __ed8d3eda309dd863fb66e04bccaa513eee255cbf__. [[conventions]] == Setup, Conventions, and Traditions There are a number of things to do as a new developer. The first set of steps is specific to committers only. These steps must be done by a mentor for those who are not committers. [[conventions-committers]] === For New Committers Those who have been given commit rights to the FreeBSD repositories must follow these steps. * Get mentor approval before committing each of these changes! * All [.filename]#src# commits go to FreeBSD-CURRENT first before being merged to FreeBSD-STABLE. The FreeBSD-STABLE branch must maintain ABI and API compatibility with earlier versions of that branch. Do not merge changes that break this compatibility. [[commit-steps]] [.procedure] ==== *Procedure 1. Steps for New Committers* . Add an Author Entity + [.filename]#doc/shared/authors.adoc# - Add an author entity. Later steps depend on this entity, and missing this step will cause the [.filename]#doc/# build to fail. This is a relatively easy task, but remains a good first test of version control skills. . Update the List of Developers and Contributors + [.filename]#doc/documentation/content/en/articles/contributors/contrib-committers.adoc# - Add an entry, which will then appear in the "Developers" section of the link:{contributors}#staff-committers[Contributors List]. Entries are sorted by last name. + [.filename]#doc/documentation/content/en/articles/contributors/contrib-additional.adoc# - _Remove_ the entry. Entries are sorted by first name. . Add a News Item + [.filename]#doc/website/data/en/news/news.toml# - Add an entry. Look for the other entries that announce new committers and follow the format. Use the date from the commit bit approval email from mailto:core@FreeBSD.org[core@FreeBSD.org]. . Add a PGP Key + `{des}` has written a shell script ([.filename]#doc/documentation/tools/addkey.sh#) to make this easier. See the https://cgit.freebsd.org/doc/plain/documentation/static/pgpkeys/README[README] file for more information. + Use [.filename]#doc/documentation/tools/checkkey.sh# to verify that keys meet minimal best-practices standards. + After adding and checking a key, add both updated files to source control and then commit them. Entries in this file are sorted by last name. + [NOTE] ====== It is very important to have a current PGP/GnuPG key in the repository. The key may be required for positive identification of a committer. For example, the `{admins}` might need it for account recovery. A complete keyring of `FreeBSD.org` users is available for download from link:https://www.FreeBSD.org/doc/pgpkeyring.txt[https://www.FreeBSD.org/doc/pgpkeyring.txt]. ====== . Update Mentor and Mentee Information + [.filename]#src/share/misc/committers-.dot# - Add an entry to the current committers section, where _repository_ is `doc`, `ports`, or `src`, depending on the commit privileges granted. + Add an entry for each additional mentor/mentee relationship in the bottom section. . Generate a Kerberos Password + See <> to generate or set a Kerberos for use with other FreeBSD services like the bug tracking database. . Optional: Enable Wiki Account + https://wiki.freebsd.org[FreeBSD Wiki] Account - A wiki account allows sharing projects and ideas. Those who do not yet have an account can follow instructions on the https://wiki.freebsd.org/AboutWiki[AboutWiki Page] to obtain one. Contact mailto:wiki-admin@FreeBSD.org[wiki-admin@FreeBSD.org] if you need help with your Wiki account. . Optional: Update Wiki Information + Wiki Information - After gaining access to the wiki, some people add entries to the https://wiki.freebsd.org/HowWeGotHere[How We Got Here], https://wiki.freebsd.org/IRC/Nicknames[IRC Nicks], and https://wiki.freebsd.org/Community/Dogs[Dogs of FreeBSD] pages. . Optional: Update Ports with Personal Information + [.filename]#ports/astro/xearth/files/freebsd.committers.markers# and [.filename]#src/usr.bin/calendar/calendars/calendar.freebsd# - Some people add entries for themselves to these files to show where they are located or the date of their birthday. . Optional: Prevent Duplicate Mailings + Subscribers to {dev-commits-doc-all}, {dev-commits-ports-all} or {dev-commits-src-all} might wish to unsubscribe to avoid receiving duplicate copies of commit messages and followups. ==== [[conventions-everyone]] === For Everyone [[conventions-everyone-steps]] [.procedure] ==== . Introduce yourself to the other developers, otherwise no one will have any idea who you are or what you are working on. The introduction need not be a comprehensive biography, just write a paragraph or two about who you are, what you plan to be working on as a developer in FreeBSD, and who will be your mentor. Email this to the {developers-name} and you will be on your way! . Log into `freefall.FreeBSD.org` and create a [.filename]#/var/forward/user# (where _user_ is your username) file containing the e-mail address where you want mail addressed to _yourusername_@FreeBSD.org to be forwarded. This includes all of the commit messages as well as any other mail addressed to the {committers-name} and the {developers-name}. Really large mailboxes which have taken up permanent residence on `freefall` may get truncated without warning if space needs to be freed, so forward it or save it elsewhere. + [NOTE] ====== -If your e-mail system uses SPF with strict rules, you should whitelist `mx2.FreeBSD.org` from SPF checks. +If your e-mail system uses SPF with strict rules, you should exclude `mx2.FreeBSD.org` from SPF checks. ====== + Due to the severe load dealing with SPAM places on the central mail servers that do the mailing list processing, the front-end server does do some basic checks and will drop some messages based on these checks. At the moment proper DNS information for the connecting host is the only check in place but that may change. Some people blame these checks for bouncing valid email. To have these checks turned off for your email, create a file named [.filename]#~/.spam_lover# on `freefall.FreeBSD.org`. + [NOTE] ====== Those who are developers but not committers will not be subscribed to the committers or developers mailing lists. The subscriptions are derived from the access rights. ====== ==== [[smtp-setup]] ==== SMTP Access Setup For those willing to send e-mail messages through the FreeBSD.org infrastructure, follow the instructions below: [.procedure] ==== . Point your mail client at `smtp.FreeBSD.org:587`. . Enable STARTTLS. . Ensure your `From:` address is set to `_yourusername_@FreeBSD.org`. . For authentication, you can use your FreeBSD Kerberos username and password (see <>). The `_yourusername_/mail` principal is preferred, as it is only valid for authenticating to mail resources. + [NOTE] ====== Do not include `@FreeBSD.org` when entering in your username. ====== + .Additional Notes [NOTE] ====== * Will only accept mail from `_yourusername_@FreeBSD.org`. If you are authenticated as one user, you are not permitted to send mail from another. * A header will be appended with the SASL username: (`Authenticated sender: _username_`). * Host has various rate limits in place to cut down on brute force attempts. ====== ==== [[smtp-setup-local-mta]] ===== Using a Local MTA to Forward Emails to the FreeBSD.org SMTP Service It is also possible to use a local MTA to forward locally sent emails to the FreeBSD.org SMTP servers. [[smtp-setup-local-postfix]] .Using Postfix [example] ==== To tell a local Postfix instance that anything from `_yourusername_@FreeBSD.org` should be forwarded to the FreeBSD.org servers, add this to your [.filename]#main.cf#: [.programlisting] .... sender_dependent_relayhost_maps = hash:/usr/local/etc/postfix/relayhost_maps smtp_sasl_auth_enable = yes smtp_sasl_security_options = noanonymous smtp_sasl_password_maps = hash:/usr/local/etc/postfix/sasl_passwd smtp_use_tls = yes .... Create [.filename]#/usr/local/etc/postfix/relayhost_maps# with the following content: [.programlisting] .... yourusername@FreeBSD.org [smtp.freebsd.org]:587 .... Create [.filename]#/usr/local/etc/postfix/sasl_passwd# with the following content: [.programlisting] .... [smtp.freebsd.org]:587 yourusername:yourpassword .... If the email server is used by other people, you may want to prevent them from sending e-mails from your address. To achieve this, add this to your [.filename]#main.cf#: [.programlisting] .... smtpd_sender_login_maps = hash:/usr/local/etc/postfix/sender_login_maps smtpd_sender_restrictions = reject_known_sender_login_mismatch .... Create [.filename]#/usr/local/etc/postfix/sender_login_maps# with the following content: [.programlisting] .... yourusername@FreeBSD.org yourlocalusername .... Where _yourlocalusername_ is the SASL username used to connect to the local instance of Postfix. ==== [[smtp-setup-local-opensmtpd]] .Using OpenSMTPD [example] ==== To tell a local OpenSMTPD instance that anything from `_yourusername_@FreeBSD.org` should be forwarded to the FreeBSD.org servers, add this to your [.filename]#smtpd.conf#: [.programlisting] .... action "freebsd" relay host smtp+tls://freebsd@smtp.freebsd.org:587 auth match from any auth yourlocalusername mail-from "_yourusername_@freebsd.org" for any action "freebsd" .... Where _yourlocalusername_ is the SASL username used to connect to the local instance of OpenSMTPD. Create [.filename]#/usr/local/etc/mail/secrets# with the following content: [.programlisting] .... freebsd yourusername:yourpassword .... ==== [[mentors]] === Mentors All new developers have a mentor assigned to them for the first few months. A mentor is responsible for teaching the mentee the rules and conventions of the project and guiding their first steps in the developer community. The mentor is also personally responsible for the mentee's actions during this initial period. For committers: do not commit anything without first getting mentor approval. Document that approval with an `Approved by:` line in the commit message. When the mentor decides that a mentee has learned the ropes and is ready to commit on their own, the mentor announces it with a commit to [.filename]#mentors#. This file is in the [.filename]#admin# orphan branch of each repository. Detailed information on how to access these branches can be found in <>. [[pre-commit-review]] == Pre-Commit Review Code review is one way to increase the quality of software. The following guidelines apply to commits to the `head` (-CURRENT) branch of the `src` repository. Other branches and the `ports` and `docs` trees have their own review policies, but these guidelines generally apply to commits requiring review: * All non-trivial changes should be reviewed before they are committed to the repository. * Reviews may be conducted by email, in Bugzilla, in Phabricator, or by another mechanism. Where possible, reviews should be public. * The developer responsible for a code change is also responsible for making all necessary review-related changes. * Code review can be an iterative process, which continues until the patch is ready to be committed. Specifically, once a patch is sent out for review, it should receive an explicit "looks good" before it is committed. So long as it is explicit, this can take whatever form makes sense for the review method. * Timeouts are not a substitute for review. Sometimes code reviews will take longer than you would hope for, especially for larger features. Accepted ways to speed up review times for your patches are: * Review other people's patches. If you help out, everybody will be more willing to do the same for you; goodwill is our currency. * Ping the patch. If it is urgent, provide reasons why it is important to you to get this patch landed and ping it every couple of days. If it is not urgent, the common courtesy ping rate is one week. Remember that you are asking for valuable time from other professional developers. * Ask for help on mailing lists, IRC, etc. Others may be able to either help you directly, or suggest a reviewer. * Split your patch into multiple smaller patches that build on each other. The smaller your patch, the higher the probability that somebody will take a quick look at it. + When making large changes, it is helpful to keep this in mind from the beginning of the effort as breaking large changes into smaller ones is often difficult after the fact. Developers should participate in code reviews as both reviewers and reviewees. If someone is kind enough to review your code, you should return the favor for someone else. Note that while anyone is welcome to review and give feedback on a patch, only an appropriate subject-matter expert can approve a change. This will usually be a committer who works with the code in question on a regular basis. In some cases, no subject-matter expert may be available. In those cases, a review by an experienced developer is sufficient when coupled with appropriate testing. [[commit-log-message]] == Commit Log Messages This section contains some suggestions and traditions for how commit logs are formatted. === Why are commit messages important? When you commit a change in Git, Subversion, or another version control system (VCS), you're prompted to write some text describing the commit -- a commit message. How important is this commit message? Should you spend some significant effort writing it? Does it really matter if you write simply fixed a bug? Most projects have more than one developer and last for some length of time. Commit messages are a very important method of communicating with other developers, in the present and for the future. FreeBSD has hundreds of active developers and hundreds of thousands of commits spanning decades of history. Over that time the developer community has learned how valuable good commit messages are; sometimes these are hard-learned lessons. Commit messages serve at least three purposes: * Communicating with other developers + FreeBSD commits generate email to various mailing lists. These include the commit message along with a copy of the patch itself. Commit messages are also viewed through commands like git log. These serve to make other developers aware of changes that are ongoing; that other developer may want to test the change, may have an interest in the topic and will want to review in more detail, or may have their own projects underway that would benefit from interaction. * Making Changes Discoverable + In a large project with a long history it may be difficult to find changes of interest when investigating an issue or change in behaviour. Verbose, detailed commit messages allow searches for changes that might be relevant. For example, `git log --since 1year --grep 'USB timeout'`. * Providing historical documentation + Commit messages serve to document changes for future developers, perhaps years or decades later. This future developer may even be you, the original author. A change that seems obvious today may be decidedly not so much later on. The `git blame` command annotates each line of a source file with the change (hash and subject line) that brought it in. Having established the importance, here are elements of a good FreeBSD commit message: === Start with a subject line Commit messages should start with a single-line subject that briefly summarizes the change. The subject should, by itself, allow the reader to quickly determine if the change is of interest or not. === Keep subject lines short The subject line should be as short as possible while still retaining the required information. This is to make browsing Git log more efficient, and so that git log --oneline can display the short hash and subject on a single 80-column line. A good rule of thumb is to stay below 63 characters, and aim for about 50 or fewer if possible. === Prefix the subject line with a component, if applicable If the change relates to a specific component the subject line may be prefixed with that component name and a colon (:). ✓ `foo: Add -k option to keep temporary data` Include the prefix in the 63-character limit suggested above, so that `git log --oneline` avoids wrapping. === Capitalize the first letter of the subject Capitalize the first letter of the subject itself. The prefix, if any, is not capitalized unless necessary (e.g., `USB:` is capitalized). === Do not end the subject line with punctuation Do not end with a period or other punctuation. In this regard the subject line is like a newspaper headline. === Separate the subject and body with a blank line Separate the body from the subject with a blank line. Some trivial commits do not require a body, and will have only a subject. ✓ `ls: Fix typo in usage text` === Limit messages to 72 columns `git log` and `git format-patch` indent the commit message by four spaces. Wrapping at 72 columns provides a matching margin on the right edge. Limiting messages to 72 characters also keeps the commit message in formatted patches below RFC 2822's suggested email line length limit of 78 characters. This limit works well with a variety of tools that may render commit messages; line wrapping might be inconsistent with longer line length. === Use the present tense, imperative mood This facilitates short subject lines and provides consistency, including with automatically generated commit messages (e.g., as generated by git revert). This is important when reading a list of commit subjects. Think of the subject as finishing the sentence "when applied, this change will ...". ✓ `foo: Implement the -k (keep) option` + ✗ `foo: Implemented the -k option` + ✗ `This change implements the -k option in foo` + ✗ `-k option added` === Focus on what and why, not how Explain what the change accomplishes and why it is being done, rather than how. Do not assume that the reader is familiar with the issue. Explain the background and motivation for the change. Include benchmark data if you have it. If there are limitations or incomplete aspects of the change, describe them in the commit message. === Consider whether parts of the commit message could be code comments instead Sometimes while writing a commit message you may find yourself writing a sentence or two explaining some tricky or confusing aspect of the change. When this happens consider whether it would be valuable to have that explanation as a comment in the code itself. === Write commit messages for your future self While writing the commit message for a change you have all of the context in mind - what prompted the change, alternate approaches that were considered and rejected, limitations of the change, and so on. Imagine yourself revisiting the change a year or two in the future, and write the commit message in a way that would provide that necessary context. === Commit messages should stand alone You may include references to mailing list postings, benchmark result web sites, or code review links. However, the commit message should contain all of the relevant information in case these references are no longer available in the future. Similarly, a commit may refer to a previous commit, for example in the case of a bug fix or revert. In addition to the commit identifier (revision or hash), include the subject line from the referenced commit (or another suitable brief reference). With each VCS migration (from CVS to Subversion to Git) revision identifiers from previous systems may become difficult to follow. === Include appropriate metadata in a footer As well as including an informative message with each commit, some additional information may be needed. This information consists of one or more lines containing the key word or phrase, a colon, tabs for formatting, and then the additional information. The key words or phrases are: [.informaltable] [cols="20%,80%", frame="none"] |=== |`PR:` |The problem report (if any) which is affected (typically, by being closed) by this commit. Multiple PRs may be specified on one line, separated by commas or spaces. |`Reported by:` |The name and e-mail address of the person that reported the issue; for developers, just the username on the FreeBSD cluster. Typically used when there is no PR, for example if the issue was reported on a mailing list. |`Submitted by:` |The name and e-mail address of the person that submitted the fix; for developers, just the username on the FreeBSD cluster. Typically not used with Git; submitted patches should have the author set by using `git commit --author`. If the submitter is the maintainer of the port being committed, include "(maintainer)" after the email address. Avoid obfuscating the email address of the submitter as this adds additional work when searching logs. |`Reviewed by:` |The name and e-mail address of the person or people that reviewed the change; for developers, just the username on the FreeBSD cluster. If a patch was submitted to a mailing list for review, and the review was favorable, then just include the list name. |`Tested by:` |The name and e-mail address of the person or people that tested the change; for developers, just the username on the FreeBSD cluster. |`Approved by:` a| The name and e-mail address of the person or people that approved the change; for developers, just the username on the FreeBSD cluster. There are several cases where approval is customary: * while a new committer is under mentorship * commits to an area of the tree to which you do not usually commit * during a release cycle * committing to a repo where you do not hold a commit bit (e.g. src committer committing to docs) While under mentorship, get mentor approval before the commit. Enter the mentor's username in this field, and note that they are a mentor: [source,shell] .... Approved by: username-of-mentor (mentor) .... If a team approved these commits then include the team name followed by the username of the approver in parentheses. For example: [source,shell] .... Approved by: re (username) .... |`Obtained from:` |The name of the project (if any) from which the code was obtained. Do not use this line for the name of an individual person. |`Fixes:` |The Git short hash and the title line of a commit that is fixed by this change as returned by `git log -n 1 --oneline GIT-COMMIT-HASH`. |`MFC after:` |To receive an e-mail reminder to MFC at a later date, specify the number of days, weeks, or months after which an MFC is planned. |`MFC to:` |If the commit should be merged to a subset of stable branches, specify the branch names. |`MFC with:` |If the commit should be merged together with a previous one in a single MFC commit (for example, where this commit corrects a bug in the previous change), specify the corresponding Git hash. |`MFH:` |If the commit is to be merged into a ports quarterly branch name, specify the quarterly branch. For example `2021Q2`. |`Relnotes:` |If the change is a candidate for inclusion in the release notes for the next release from the branch, set to `yes`. |`Security:` |If the change is related to a security vulnerability or security exposure, include one or more references or a description of the issue. If possible, include a VuXML URL or a CVE ID. |`Event:` |The description for the event where this commit was made. If this is a recurring event, add the year or even the month to it. For example, this could be `FooBSDcon 2019`. The idea behind this line is to put recognition to conferences, gatherings, and other types of meetups and to show that these are useful to have. Please do not use the `Sponsored by:` line for this as that is meant for organizations sponsoring certain features or developers working on them. |`Sponsored by:` |Sponsoring organizations for this change, if any. Separate multiple organizations with commas. If only a portion of the work was sponsored, or different amounts of sponsorship were provided to different authors, please give appropriate credit in parentheses after each sponsor name. For example, `Example.com (alice, code refactoring), Wormulon (bob), Momcorp (cindy)` shows that Alice was sponsored by Example.com to do code refactoring, while Wormulon sponsored Bob's work and Momcorp sponsored Cindy's work. Other authors were either not sponsored or chose not to list sponsorship. |`Differential Revision:` |The full URL of the Phabricator review. This line __must be the last line__. For example: `https://reviews.freebsd.org/D1708`. |`Signed-off-by:` |ID certifies compliance with https://developercertificate.org/ |=== .Commit Log for a Commit Based on a PR [example] ==== The commit is based on a patch from a PR submitted by John Smith. The commit message "PR" field is filled. [.programlisting] .... ... PR: 12345 .... The committer sets the author of the patch with `git commit --author "John Smith "`. ==== .Commit Log for a Commit Needing Review [example] ==== The virtual memory system is being changed. After posting patches to the appropriate mailing list (in this case, `freebsd-arch`) and the changes have been approved. [.programlisting] .... ... Reviewed by: -arch .... ==== .Commit Log for a Commit Needing Approval [example] ==== Commit a port, after working with the listed MAINTAINER, who said to go ahead and commit. [.programlisting] .... ... Approved by: abc (maintainer) .... Where _abc_ is the account name of the person who approved. ==== .Commit Log for a Commit Bringing in Code from OpenBSD [example] ==== Committing some code based on work done in the OpenBSD project. [.programlisting] .... ... Obtained from: OpenBSD .... ==== .Commit Log for a Change to FreeBSD-CURRENT with a Planned Commit to FreeBSD-STABLE to Follow at a Later Date. [example] ==== Committing some code which will be merged from FreeBSD-CURRENT into the FreeBSD-STABLE branch after two weeks. [.programlisting] .... ... MFC after: 2 weeks .... Where _2_ is the number of days, weeks, or months after which an MFC is planned. The _weeks_ option may be `day`, `days`, `week`, `weeks`, `month`, `months`. ==== It is often necessary to combine these. Consider the situation where a user has submitted a PR containing code from the NetBSD project. Looking at the PR, the developer sees it is not an area of the tree they normally work in, so they have the change reviewed by the `arch` mailing list. Since the change is complex, the developer opts to MFC after one month to allow adequate testing. The extra information to include in the commit would look something like .Example Combined Commit Log [example] ==== [.programlisting] .... PR: 54321 Reviewed by: -arch Obtained from: NetBSD MFC after: 1 month Relnotes: yes .... ==== [[pref-license]] == Preferred License for New Files The FreeBSD Project's full license policy can be found at link:https://www.FreeBSD.org/internal/software-license/[https://www.FreeBSD.org/internal/software-license]. The rest of this section is intended to help you get started. As a rule, when in doubt, ask. It is much easier to give advice than to fix the source tree. The FreeBSD Project suggests and uses this text as the preferred license scheme: [.programlisting] .... /*- * SPDX-License-Identifier: BSD-2-Clause-FreeBSD * * Copyright (c) [year] [your name] * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * [id for your version control system, if any] */ .... The FreeBSD project strongly discourages the so-called "advertising clause" in new code. Due to the large number of contributors to the FreeBSD project, complying with this clause for many commercial vendors has become difficult. If you have code in the tree with the advertising clause, please consider removing it. In fact, please consider using the above license for your code. The FreeBSD project discourages completely new licenses and variations on the standard licenses. New licenses require the approval of the {core-email} to reside in the main repository. The more different licenses that are used in the tree, the more problems that this causes to those wishing to utilize this code, typically from unintended consequences from a poorly worded license. Project policy dictates that code under some non-BSD licenses must be placed only in specific sections of the repository, and in some cases, compilation must be conditional or even disabled by default. For example, the GENERIC kernel must be compiled under only licenses identical to or substantially similar to the BSD license. GPL, APSL, CDDL, etc, licensed software must not be compiled into GENERIC. Developers are reminded that in open source, getting "open" right is just as important as getting "source" right, as improper handling of intellectual property has serious consequences. Any questions or concerns should immediately be brought to the attention of the core team. [[tracking.license.grants]] == Keeping Track of Licenses Granted to the FreeBSD Project Various software or data exist in the repositories where the FreeBSD project has been granted a special licence to be able to use them. A case in point are the Terminus fonts for use with man:vt[4]. Here the author Dimitar Zhekov has allowed us to use the "Terminus BSD Console" font under a 2-clause BSD license rather than the regular Open Font License he normally uses. It is clearly sensible to keep a record of any such license grants. To that end, the {core-email} has decided to keep an archive of them. Whenever the FreeBSD project is granted a special license we require the {core-email} to be notified. Any developers involved in arranging such a license grant, please send details to the {core-email} including: * Contact details for people or organizations granting the special license. * What files, directories etc. in the repositories are covered by the license grant including the revision numbers where any specially licensed material was committed. * The date the license comes into effect from. Unless otherwise agreed, this will be the date the license was issued by the authors of the software in question. * The license text. * A note of any restrictions, limitations or exceptions that apply specifically to FreeBSD's usage of the licensed material. * Any other relevant information. Once the {core-email} is satisfied that all the necessary details have been gathered and are correct, the secretary will send a PGP-signed acknowledgement of receipt including the license details. This receipt will be persistently archived and serve as our permanent record of the license grant. The license archive should contain only details of license grants; this is not the place for any discussions around licensing or other subjects. Access to data within the license archive will be available on request to the {core-email}. [[spdx.tags]] == SPDX Tags in the tree The project uses https://spdx.dev[SPDX] tags in our source base. At present, these tags are indented to help automated tools reconstruct license requirements mechanically. All _SPDX-License-Identifier_ tags in the tree should be considered to be informative. All files in the FreeBSD source tree with these tags also have a copy of the license which governs use of that file. In the event of a discrepency, the verbatim license is controlling. The project tries to follow the https://spdx.github.io/spdx-spec/[SPDX Specification, Version 2.2]. How to mark source files and valid algebraic expressions are found in https://spdx.github.io/spdx-spec/appendix-IV-SPDX-license-expressions/[Appendix IV] and https://spdx.github.io/spdx-spec/appendix-V-using-SPDX-short-identifiers-in-source-files/[Appendix V]. The project draws identifiers from SPDX's list of valid https://spdx.org/licenses/[short license identifiers]. The project uses only the _SPDX-License-Identifier_ tag. As of March 2021, approximately 25,000 out of 90,000 files in the tree have been marked. [[developer.relations]] == Developer Relations When working directly on your own code or on code which is already well established as your responsibility, then there is probably little need to check with other committers before jumping in with a commit. Working on a bug in an area of the system which is clearly orphaned (and there are a few such areas, to our shame), the same applies. When modifying parts of the system which are maintained, formally, or informally, consider asking for review just as a developer would have before becoming a committer. For ports, contact the listed `MAINTAINER` in the [.filename]#Makefile#. To determine if an area of the tree is maintained, check the MAINTAINERS file at the root of the tree. If nobody is listed, scan the revision history to see who has committed changes in the past. An example script that lists each person who has committed to a given file along with the number of commits each person has made can be found at on `freefall` at [.filename]#~eadler/bin/whodid#. If queries go unanswered or the committer otherwise indicates a lack of interest in the area affected, go ahead and commit it. [IMPORTANT] ==== Avoid sending private emails to maintainers. Other people might be interested in the conversation, not just the final output. ==== If there is any doubt about a commit for any reason at all, have it reviewed before committing. Better to have it flamed then and there rather than when it is part of the repository. If a commit does results in controversy erupting, it may be advisable to consider backing the change out again until the matter is settled. Remember, with a version control system we can always change it back. Do not impugn the intentions of others. If they see a different solution to a problem, or even a different problem, it is probably not because they are stupid, because they have questionable parentage, or because they are trying to destroy hard work, personal image, or FreeBSD, but basically because they have a different outlook on the world. Different is good. Disagree honestly. Argue your position from its merits, be honest about any shortcomings it may have, and be open to seeing their solution, or even their vision of the problem, with an open mind. Accept correction. We are all fallible. When you have made a mistake, apologize and get on with life. Do not beat up yourself, and certainly do not beat up others for your mistake. Do not waste time on embarrassment or recrimination, just fix the problem and move on. Ask for help. Seek out (and give) peer reviews. One of the ways open source software is supposed to excel is in the number of eyeballs applied to it; this does not apply if nobody will review code. [[if-in-doubt]] == If in Doubt... When unsure about something, whether it be a technical issue or a project convention be sure to ask. If you stay silent you will never make progress. If it relates to a technical issue ask on the public mailing lists. Avoid the temptation to email the individual person that knows the answer. This way everyone will be able to learn from the question and the answer. For project specific or administrative questions ask, in order: * Your mentor or former mentor. * An experienced committer on IRC, email, etc. * Any team with a "hat", as they can give you a definitive answer. * If still not sure, ask on {developers-name}. Once your question is answered, if no one pointed you to documentation that spelled out the answer to your question, document it, as others will have the same question. [[bugzilla]] == Bugzilla The FreeBSD Project utilizes Bugzilla for tracking bugs and change requests. Be sure that if you commit a fix or suggestion found in the PR database to close it. It is also considered nice if you take time to close any PRs associated with your commits, if appropriate. Committers with non-``FreeBSD.org`` Bugzilla accounts can have the old account merged with the `FreeBSD.org` account by following these steps: [.procedure] ==== . Log in using your old account. . Open new bug. Choose `Services` as the Product, and `Bug Tracker` as the Component. In bug description list accounts you wish to be merged. . Log in using `FreeBSD.org` account and post comment to newly opened bug to confirm ownership. See <> for more details on how to generate or set a password for your `FreeBSD.org` account. . If there are more than two accounts to merge, post comments from each of them. ==== You can find out more about Bugzilla at: * link:{pr-guidelines}[FreeBSD Problem Report Handling Guidelines] * link:https://www.FreeBSD.org/support/[https://www.FreeBSD.org/support] [[phabricator]] == Phabricator The FreeBSD Project utilizes https://reviews.freebsd.org[Phabricator] for code review requests. See the https://wiki.freebsd.org/CodeReview[CodeReview] wiki page for details. Committers with non-``FreeBSD.org`` Phabricator accounts can have the old account renamed to the ``FreeBSD.org`` account by following these steps: [.procedure] ==== . Change your Phabricator account email to your `FreeBSD.org` email. . Open new bug on our bug tracker using your `FreeBSD.org` account, see <> for more information. Choose `Services` as the Product, and `Code Review` as the Component. In bug description request that your Phabricator account be renamed, and provide a link to your Phabricator user. For example, `https://reviews.freebsd.org/p/bob_example.com/` ==== [IMPORTANT] ==== Phabricator accounts cannot be merged, please do not open a new account. ==== [[people]] == Who's Who Besides the repository meisters, there are other FreeBSD project members and teams whom you will probably get to know in your role as a committer. Briefly, and by no means all-inclusively, these are: `{doceng}`:: doceng is the group responsible for the documentation build infrastructure, approving new documentation committers, and ensuring that the FreeBSD website and documentation on the FTP site is up to date with respect to the Subversion tree. It is not a conflict resolution body. The vast majority of documentation related discussion takes place on the {freebsd-doc}. More details regarding the doceng team can be found in its https://www.FreeBSD.org/internal/doceng/[charter]. Committers interested in contributing to the documentation should familiarize themselves with the link:{fdp-primer}[Documentation Project Primer]. `{re-members}`:: These are the members of the `{re}`. This team is responsible for setting release deadlines and controlling the release process. During code freezes, the release engineers have final authority on all changes to the system for whichever branch is pending release status. If there is something you want merged from FreeBSD-CURRENT to FreeBSD-STABLE (whatever values those may have at any given time), these are the people to talk to about it. `{so}`:: `{so-name}` is the link:https://www.FreeBSD.org/security/[FreeBSD Security Officer] and oversees the `{security-officer}`. `{wollman}`:: If you need advice on obscure network internals or are not sure of some potential change to the networking subsystem you have in mind, Garrett is someone to talk to. Garrett is also very knowledgeable on the various standards applicable to FreeBSD. {committers-name}:: {svn-src-all}, {svn-ports-all} and {svn-doc-all} are the mailing lists that the version control system uses to send commit messages to. _Never_ send email directly to these lists. Only send replies to this list when they are short and are directly related to a commit. {developers-name}:: All committers are subscribed to -developers. This list was created to be a forum for the committers "community" issues. Examples are Core voting, announcements, etc. + The {developers-name} is for the exclusive use of FreeBSD committers. To develop FreeBSD, committers must have the ability to openly discuss matters that will be resolved before they are publicly announced. Frank discussions of work in progress are not suitable for open publication and may harm FreeBSD. + All FreeBSD committers are expected not to not publish or forward messages from the {developers-name} outside the list membership without permission of all of the authors. Violators will be removed from the {developers-name}, resulting in a suspension of commit privileges. Repeated or flagrant violations may result in permanent revocation of commit privileges. + This list is _not_ intended as a place for code reviews or for any technical discussion. In fact using it as such hurts the FreeBSD Project as it gives a sense of a closed list where general decisions affecting all of the FreeBSD using community are made without being "open". Last, but not least __never, never ever, email the {developers-name} and CC:/BCC: another FreeBSD list__. Never, ever email another FreeBSD email list and CC:/BCC: the {developers-name}. Doing so can greatly diminish the benefits of this list. [[ssh.guide]] == SSH Quick-Start Guide [.procedure] ==== . If you do not wish to type your password in every time you use man:ssh[1], and you use keys to authenticate, man:ssh-agent[1] is there for your convenience. If you want to use man:ssh-agent[1], make sure that you run it before running other applications. X users, for example, usually do this from their [.filename]#.xsession# or [.filename]#.xinitrc#. See man:ssh-agent[1] for details. . Generate a key pair using man:ssh-keygen[1]. The key pair will wind up in your [.filename]#$HOME/.ssh/# directory. + [IMPORTANT] ====== Only ECDSA, Ed25519 or RSA keys are supported. ====== . Send your public key ([.filename]#$HOME/.ssh/id_ecdsa.pub#, [.filename]#$HOME/.ssh/id_ed25519.pub#, or [.filename]#$HOME/.ssh/id_rsa.pub#) to the person setting you up as a committer so it can be put into [.filename]#yourlogin# in [.filename]#/etc/ssh-keys/# on `freefall`. ==== Now man:ssh-add[1] can be used for authentication once per session. It prompts for the private key's pass phrase, and then stores it in the authentication agent (man:ssh-agent[1]). Use `ssh-add -d` to remove keys stored in the agent. Test with a simple remote command: `ssh freefall.FreeBSD.org ls /usr`. For more information, see package:security/openssh-portable[], man:ssh[1], man:ssh-add[1], man:ssh-agent[1], man:ssh-keygen[1], and man:scp[1]. For information on adding, changing, or removing man:ssh[1] keys, see https://wiki.freebsd.org/clusteradm/ssh-keys[this article]. [[coverity]] == Coverity(R) Availability for FreeBSD Committers All FreeBSD developers can obtain access to Coverity analysis results of all FreeBSD Project software. All who are interested in obtaining access to the analysis results of the automated Coverity runs, can sign up at http://scan.coverity.com/[Coverity Scan]. The FreeBSD wiki includes a mini-guide for developers who are interested in working with the Coverity(R) analysis reports: https://wiki.freebsd.org/CoverityPrevent[https://wiki.freebsd.org/CoverityPrevent]. Please note that this mini-guide is only readable by FreeBSD developers, so if you cannot access this page, you will have to ask someone to add you to the appropriate Wiki access list. Finally, all FreeBSD developers who are going to use Coverity(R) are always encouraged to ask for more details and usage information, by posting any questions to the mailing list of the FreeBSD developers. [[rules]] == The FreeBSD Committers' Big List of Rules Everyone involved with the FreeBSD project is expected to abide by the _Code of Conduct_ available from link:https://www.FreeBSD.org/internal/code-of-conduct/[https://www.FreeBSD.org/internal/code-of-conduct]. As committers, you form the public face of the project, and how you behave has a vital impact on the public perception of it. This guide expands on the parts of the _Code of Conduct_ specific to committers. . Respect other committers. . Respect other contributors. . Discuss any significant change _before_ committing. . Respect existing maintainers (if listed in the `MAINTAINER` field in [.filename]#Makefile# or in [.filename]#MAINTAINER# in the top-level directory). . Any disputed change must be backed out pending resolution of the dispute if requested by a maintainer. Security related changes may override a maintainer's wishes at the Security Officer's discretion. . Changes go to FreeBSD-CURRENT before FreeBSD-STABLE unless specifically permitted by the release engineer or unless they are not applicable to FreeBSD-CURRENT. Any non-trivial or non-urgent change which is applicable should also be allowed to sit in FreeBSD-CURRENT for at least 3 days before merging so that it can be given sufficient testing. The release engineer has the same authority over the FreeBSD-STABLE branch as outlined for the maintainer in rule #5. . Do not fight in public with other committers; it looks bad. . Respect all code freezes and read the `committers` and `developers` mailing lists in a timely manner so you know when a code freeze is in effect. . When in doubt on any procedure, ask first! . Test your changes before committing them. . Do not commit to contributed software without _explicit_ approval from the respective maintainers. As noted, breaking some of these rules can be grounds for suspension or, upon repeated offense, permanent removal of commit privileges. Individual members of core have the power to temporarily suspend commit privileges until core as a whole has the chance to review the issue. In case of an "emergency" (a committer doing damage to the repository), a temporary suspension may also be done by the repository meisters. Only a 2/3 majority of core has the authority to suspend commit privileges for longer than a week or to remove them permanently. This rule does not exist to set core up as a bunch of cruel dictators who can dispose of committers as casually as empty soda cans, but to give the project a kind of safety fuse. If someone is out of control, it is important to be able to deal with this immediately rather than be paralyzed by debate. In all cases, a committer whose privileges are suspended or revoked is entitled to a "hearing" by core, the total duration of the suspension being determined at that time. A committer whose privileges are suspended may also request a review of the decision after 30 days and every 30 days thereafter (unless the total suspension period is less than 30 days). A committer whose privileges have been revoked entirely may request a review after a period of 6 months has elapsed. This review policy is _strictly informal_ and, in all cases, core reserves the right to either act on or disregard requests for review if they feel their original decision to be the right one. In all other aspects of project operation, core is a subset of committers and is bound by the __same rules__. Just because someone is in core this does not mean that they have special dispensation to step outside any of the lines painted here; core's "special powers" only kick in when it acts as a group, not on an individual basis. As individuals, the core team members are all committers first and core second. === Details [[respect]] . Respect other committers. + This means that you need to treat other committers as the peer-group developers that they are. Despite our occasional attempts to prove the contrary, one does not get to be a committer by being stupid and nothing rankles more than being treated that way by one of your peers. Whether we always feel respect for one another or not (and everyone has off days), we still have to _treat_ other committers with respect at all times, on public forums and in private email. + Being able to work together long term is this project's greatest asset, one far more important than any set of changes to the code, and turning arguments about code into issues that affect our long-term ability to work harmoniously together is just not worth the trade-off by any conceivable stretch of the imagination. + To comply with this rule, do not send email when you are angry or otherwise behave in a manner which is likely to strike others as needlessly confrontational. First calm down, then think about how to communicate in the most effective fashion for convincing the other persons that your side of the argument is correct, do not just blow off some steam so you can feel better in the short term at the cost of a long-term flame war. Not only is this very bad "energy economics", but repeated displays of public aggression which impair our ability to work well together will be dealt with severely by the project leadership and may result in suspension or termination of your commit privileges. The project leadership will take into account both public and private communications brought before it. It will not seek the disclosure of private communications, but it will take it into account if it is volunteered by the committers involved in the complaint. + All of this is never an option which the project's leadership enjoys in the slightest, but unity comes first. No amount of code or good advice is worth trading that away. . Respect other contributors. + You were not always a committer. At one time you were a contributor. Remember that at all times. Remember what it was like trying to get help and attention. Do not forget that your work as a contributor was very important to you. Remember what it was like. Do not discourage, belittle, or demean contributors. Treat them with respect. They are our committers in waiting. They are every bit as important to the project as committers. Their contributions are as valid and as important as your own. After all, you made many contributions before you became a committer. Always remember that. + Consider the points raised under <> and apply them also to contributors. . Discuss any significant change _before_ committing. + The repository is not where changes are initially submitted for correctness or argued over, that happens first in the mailing lists or by use of the Phabricator service. The commit will only happen once something resembling consensus has been reached. This does not mean that permission is required before correcting every obvious syntax error or manual page misspelling, just that it is good to develop a feel for when a proposed change is not quite such a no-brainer and requires some feedback first. People really do not mind sweeping changes if the result is something clearly better than what they had before, they just do not like being _surprised_ by those changes. The very best way of making sure that things are on the right track is to have code reviewed by one or more other committers. + When in doubt, ask for review! . Respect existing maintainers if listed. + Many parts of FreeBSD are not "owned" in the sense that any specific individual will jump up and yell if you commit a change to "their" area, but it still pays to check first. One convention we use is to put a maintainer line in the [.filename]#Makefile# for any package or subtree which is being actively maintained by one or more people; see link:{developers-handbook}#policies[Source Tree Guidelines and Policies] for documentation on this. Where sections of code have several maintainers, commits to affected areas by one maintainer need to be reviewed by at least one other maintainer. In cases where the "maintainer-ship" of something is not clear, look at the repository logs for the files in question and see if someone has been working recently or predominantly in that area. . Any disputed change must be backed out pending resolution of the dispute if requested by a maintainer. Security related changes may override a maintainer's wishes at the Security Officer's discretion. + This may be hard to swallow in times of conflict (when each side is convinced that they are in the right, of course) but a version control system makes it unnecessary to have an ongoing dispute raging when it is far easier to simply reverse the disputed change, get everyone calmed down again and then try to figure out what is the best way to proceed. If the change turns out to be the best thing after all, it can be easily brought back. If it turns out not to be, then the users did not have to live with the bogus change in the tree while everyone was busily debating its merits. People _very_ rarely call for back-outs in the repository since discussion generally exposes bad or controversial changes before the commit even happens, but on such rare occasions the back-out should be done without argument so that we can get immediately on to the topic of figuring out whether it was bogus or not. . Changes go to FreeBSD-CURRENT before FreeBSD-STABLE unless specifically permitted by the release engineer or unless they are not applicable to FreeBSD-CURRENT. Any non-trivial or non-urgent change which is applicable should also be allowed to sit in FreeBSD-CURRENT for at least 3 days before merging so that it can be given sufficient testing. The release engineer has the same authority over the FreeBSD-STABLE branch as outlined in rule #5. + This is another "do not argue about it" issue since it is the release engineer who is ultimately responsible (and gets beaten up) if a change turns out to be bad. Please respect this and give the release engineer your full cooperation when it comes to the FreeBSD-STABLE branch. The management of FreeBSD-STABLE may frequently seem to be overly conservative to the casual observer, but also bear in mind the fact that conservatism is supposed to be the hallmark of FreeBSD-STABLE and different rules apply there than in FreeBSD-CURRENT. There is also really no point in having FreeBSD-CURRENT be a testing ground if changes are merged over to FreeBSD-STABLE immediately. Changes need a chance to be tested by the FreeBSD-CURRENT developers, so allow some time to elapse before merging unless the FreeBSD-STABLE fix is critical, time sensitive or so obvious as to make further testing unnecessary (spelling fixes to manual pages, obvious bug/typo fixes, etc.) In other words, apply common sense. + Changes to the security branches (for example, `releng/9.3`) must be approved by a member of the `{security-officer}`, or in some cases, by a member of the `{re}`. . Do not fight in public with other committers; it looks bad. + This project has a public image to uphold and that image is very important to all of us, especially if we are to continue to attract new members. There will be occasions when, despite everyone's very best attempts at self-control, tempers are lost and angry words are exchanged. The best thing that can be done in such cases is to minimize the effects of this until everyone has cooled back down. Do not air angry words in public and do not forward private correspondence or other private communications to public mailing lists, mail aliases, instant messaging channels or social media sites. What people say one-to-one is often much less sugar-coated than what they would say in public, and such communications therefore have no place there - they only serve to inflame an already bad situation. If the person sending a flame-o-gram at least had the grace to send it privately, then have the grace to keep it private yourself. If you feel you are being unfairly treated by another developer, and it is causing you anguish, bring the matter up with core rather than taking it public. Core will do its best to play peace makers and get things back to sanity. In cases where the dispute involves a change to the codebase and the participants do not appear to be reaching an amicable agreement, core may appoint a mutually-agreeable third party to resolve the dispute. All parties involved must then agree to be bound by the decision reached by this third party. . Respect all code freezes and read the `committers` and `developers` mailing list on a timely basis so you know when a code freeze is in effect. + Committing unapproved changes during a code freeze is a really big mistake and committers are expected to keep up-to-date on what is going on before jumping in after a long absence and committing 10 megabytes worth of accumulated stuff. People who abuse this on a regular basis will have their commit privileges suspended until they get back from the FreeBSD Happy Reeducation Camp we run in Greenland. . When in doubt on any procedure, ask first! + Many mistakes are made because someone is in a hurry and just assumes they know the right way of doing something. If you have not done it before, chances are good that you do not actually know the way we do things and really need to ask first or you are going to completely embarrass yourself in public. There is no shame in asking "how in the heck do I do this?" We already know you are an intelligent person; otherwise, you would not be a committer. . Test your changes before committing them. + This may sound obvious, but if it really were so obvious then we probably would not see so many cases of people clearly not doing this. If your changes are to the kernel, make sure you can still compile both GENERIC and LINT. If your changes are anywhere else, make sure you can still make world. If your changes are to a branch, make sure your testing occurs with a machine which is running that code. If you have a change which also may break another architecture, be sure and test on all supported architectures. Please refer to the https://www.FreeBSD.org/internal/[FreeBSD Internal Page] for a list of available resources. As other architectures are added to the FreeBSD supported platforms list, the appropriate shared testing resources will be made available. . Do not commit to contributed software without _explicit_ approval from the respective maintainers. + Contributed software is anything under the [.filename]#src/contrib#, [.filename]#src/crypto#, or [.filename]#src/sys/contrib# trees. + The trees mentioned above are for contributed software usually imported onto a vendor branch. Committing something there may cause unnecessary headaches when importing newer versions of the software. As a general consider sending patches upstream to the vendor. Patches may be committed to FreeBSD first with permission of the maintainer. + Reasons for modifying upstream software range from wanting strict control over a tightly coupled dependency to lack of portability in the canonical repository's distribution of their code. Regardless of the reason, effort to minimize the maintenance burden of fork is helpful to fellow maintainers. Avoid committing trivial or cosmetic changes to files since it makes every merge thereafter more difficult: such patches need to be manually re-verified every import. + If a particular piece of software lacks a maintainer, you are encouraged to take up ownership. If you are unsure of the current maintainership email {freebsd-arch} and ask. === Policy on Multiple Architectures FreeBSD has added several new architecture ports during recent release cycles and is truly no longer an i386(TM) centric operating system. In an effort to make it easier to keep FreeBSD portable across the platforms we support, core has developed this mandate: [.blockquote] Our 32-bit reference platform is i386, and our 64-bit reference platform is amd64. Major design work (including major API and ABI changes) must prove itself on at least one 32-bit and at least one 64-bit platform, preferably the primary reference platforms, before it may be committed to the source tree. The i386 and amd64 platforms were chosen due to being more readily available to developers and as representatives of more diverse processor and system designs - big versus little endian, register file versus register stack, different DMA and cache implementations, hardware page tables versus software TLB management etc. We will continue to re-evaluate this policy as cost and availability of the 64-bit platforms change. Developers should also be aware of our Tier Policy for the long term support of hardware architectures. The rules here are intended to provide guidance during the development process, and are distinct from the requirements for features and architectures listed in that section. The Tier rules for feature support on architectures at release-time are more strict than the rules for changes during the development process. === Other Suggestions When committing documentation changes, use a spell checker before committing. For all XML docs, verify that the formatting directives are correct by running `make lint` and package:textproc/igor[]. For manual pages, run package:sysutils/manck[] and package:textproc/igor[] over the manual page to verify all of the cross references and file references are correct and that the man page has all of the appropriate `MLINKS` installed. Do not mix style fixes with new functionality. A style fix is any change which does not modify the functionality of the code. Mixing the changes obfuscates the functionality change when asking for differences between revisions, which can hide any new bugs. Do not include whitespace changes with content changes in commits to [.filename]#doc/#. The extra clutter in the diffs makes the translators' job much more difficult. Instead, make any style or whitespace changes in separate commits that are clearly labeled as such in the commit message. === Deprecating Features When it is necessary to remove functionality from software in the base system, follow these guidelines whenever possible: . Mention is made in the manual page and possibly the release notes that the option, utility, or interface is deprecated. Use of the deprecated feature generates a warning. . The option, utility, or interface is preserved until the next major (point zero) release. . The option, utility, or interface is removed and no longer documented. It is now obsolete. It is also generally a good idea to note its removal in the release notes. === Privacy and Confidentiality . Most FreeBSD business is done in public. + FreeBSD is an _open_ project. Which means that not only can anyone use the source code, but that most of the development process is open to public scrutiny. . Certain sensitive matters must remain private or held under embargo. + There unfortunately cannot be complete transparency. As a FreeBSD developer you will have a certain degree of privileged access to information. Consequently you are expected to respect certain requirements for confidentiality. Sometimes the need for confidentiality comes from external collaborators or has a specific time limit. Mostly though, it is a matter of not releasing private communications. . The Security Officer has sole control over the release of security advisories. + Where there are security problems that affect many different operating systems, FreeBSD frequently depends on early access to be able to prepare advisories for coordinated release. Unless FreeBSD developers can be trusted to maintain security, such early access will not be made available. The Security Officer is responsible for controlling pre-release access to information about vulnerabilities, and for timing the release of all advisories. He may request help under condition of confidentiality from any developer with relevant knowledge to prepare security fixes. . Communications with Core are kept confidential for as long as necessary. + Communications to core will initially be treated as confidential. Eventually however, most of Core's business will be summarized into the monthly or quarterly core reports. Care will be taken to avoid publicising any sensitive details. Records of some particularly sensitive subjects may not be reported on at all and will be retained only in Core's private archives. . Non-disclosure Agreements may be required for access to certain commercially sensitive data. + Access to certain commercially sensitive data may only be available under a Non-Disclosure Agreement. The FreeBSD Foundation legal staff must be consulted before any binding agreements are entered into. . Private communications must not be made public without permission. + Beyond the specific requirements above there is a general expectation not to publish private communications between developers without the consent of all parties involved. Ask permission before forwarding a message onto a public mailing list, or posting it to a forum or website that can be accessed by other than the original correspondents. . Communications on project-only or restricted access channels must be kept private. + Similarly to personal communications, certain internal communications channels, including FreeBSD Committer only mailing lists and restricted access IRC channels are considered private communications. Permission is required to publish material from these sources. . Core may approve publication. + Where it is impractical to obtain permission due to the number of correspondents or where permission to publish is unreasonably withheld, Core may approve release of such private matters that merit more general publication. [[archs]] == Support for Multiple Architectures FreeBSD is a highly portable operating system intended to function on many different types of hardware architectures. Maintaining clean separation of Machine Dependent (MD) and Machine Independent (MI) code, as well as minimizing MD code, is an important part of our strategy to remain agile with regards to current hardware trends. Each new hardware architecture supported by FreeBSD adds substantially to the cost of code maintenance, toolchain support, and release engineering. It also dramatically increases the cost of effective testing of kernel changes. As such, there is strong motivation to differentiate between classes of support for various architectures while remaining strong in a few key architectures that are seen as the FreeBSD "target audience". === Statement of General Intent The FreeBSD Project targets "production quality commercial off-the-shelf (COTS) workstation, server, and high-end embedded systems". By retaining a focus on a narrow set of architectures of interest in these environments, the FreeBSD Project is able to maintain high levels of quality, stability, and performance, as well as minimize the load on various support teams on the project, such as the ports team, documentation team, security officer, and release engineering teams. Diversity in hardware support broadens the options for FreeBSD consumers by offering new features and usage opportunities, but these benefits must always be carefully considered in terms of the real-world maintenance cost associated with additional platform support. The FreeBSD Project differentiates platform targets into four tiers. Each tier includes a list of guarantees consumers may rely on as well as obligations by the Project and developers to fulfill those guarantees. These lists define the minimum guarantees for each tier. The Project and developers may provide additional levels of support beyond the minimum guarantees for a given tier, but such additional support is not guaranteed. Each platform target is assigned to a specific tier for each stable branch. As a result, a platform target might be assigned to different tiers on concurrent stable branches. === Platform Targets Support for a hardware platform consists of two components: kernel support and userland Application Binary Interfaces (ABIs). Kernel platform support includes things needed to run a FreeBSD kernel on a hardware platform such as machine-dependent virtual memory management and device drivers. A userland ABI specifies an interface for user processes to interact with a FreeBSD kernel and base system libraries. A userland ABI includes system call interfaces, the layout and semantics of public data structures, and the layout and semantics of arguments passed to subroutines. Some components of an ABI may be defined by specifications such as the layout of C++ exception objects or calling conventions for C functions. A FreeBSD kernel also uses an ABI (sometimes referred to as the Kernel Binary Interface (KBI)) which includes the semantics and layouts of public data structures and the layout and semantics of arguments to public functions within the kernel itself. A FreeBSD kernel may support multiple userland ABIs. For example, FreeBSD's amd64 kernel supports FreeBSD amd64 and i386 userland ABIs as well as Linux x86_64 and i386 userland ABIs. A FreeBSD kernel should support a "native" ABI as the default ABI. The native "ABI" generally shares certain properties with the kernel ABI such as the C calling convention, sizes of basic types, etc. Tiers are defined for both kernels and userland ABIs. In the common case, a platform's kernel and FreeBSD ABIs are assigned to the same tier. === Tier 1: Fully-Supported Architectures Tier 1 platforms are the most mature FreeBSD platforms. They are supported by the security officer, release engineering, and port management teams. Tier 1 architectures are expected to be Production Quality with respect to all aspects of the FreeBSD operating system, including installation and development environments. The FreeBSD Project provides the following guarantees to consumers of Tier 1 platforms: * Official FreeBSD release images will be provided by the release engineering team. * Binary updates and source patches for Security Advisories and Errata Notices will be provided for supported releases. * Source patches for Security Advisories will be provided for supported branches. * Binary updates and source patches for cross-platform Security Advisories will typically be provided at the time of the announcement. * Changes to userland ABIs will generally include compatibility shims to ensure correct operation of binaries compiled against any stable branch where the platform is Tier 1. These shims might not be enabled in the default install. If compatibility shims are not provided for an ABI change, the lack of shims will be clearly documented in the release notes. * Changes to certain portions of the kernel ABI will include compatibility shims to ensure correct operation of kernel modules compiled against the oldest supported release on the branch. Note that not all parts of the kernel ABI are protected. * Official binary packages for third party software will be provided by the ports team. For embedded architectures, these packages may be cross-built from a different architecture. * Most relevant ports should either build or have the appropriate filters to prevent inappropriate ones from building. * New features which are not inherently platform-specific will be fully functional on all Tier 1 architectures. * Features and compatibility shims used by binaries compiled against older stable branches may be removed in newer major versions. Such removals will be clearly documented in the release notes. * Tier 1 platforms should be fully documented. Basic operations will be documented in the FreeBSD Handbook. * Tier 1 platforms will be included in the source tree. * Tier 1 platforms should be self-hosting either via the in-tree toolchain or an external toolchain. If an external toolchain is required, official binary packages for an external toolchain will be provided. To maintain maturity of Tier 1 platforms, the FreeBSD Project will maintain the following resources to support development: * Build and test automation support either in the FreeBSD.org cluster or some other location easily available for all developers. Embedded platforms may substitute an emulator available in the FreeBSD.org cluster for actual hardware. * Inclusion in the `make universe` and `make tinderbox` targets. * Dedicated hardware in one of the FreeBSD clusters for package building (either natively or via qemu-user). Collectively, developers are required to provide the following to maintain the Tier 1 status of a platform: * Changes to the source tree should not knowingly break the build of a Tier 1 platform. * Tier 1 architectures must have a mature, healthy ecosystem of users and active developers. * Developers should be able to build packages on commonly available, non-embedded Tier 1 systems. This can mean either native builds if non-embedded systems are commonly available for the platform in question, or it can mean cross-builds hosted on some other Tier 1 architecture. * Changes cannot break the userland ABI. If an ABI change is required, ABI compatibility for existing binaries should be provided via use of symbol versioning or shared library version bumps. * Changes merged to stable branches cannot break the protected portions of the kernel ABI. If a kernel ABI change is required, the change should be modified to preserve functionality of existing kernel modules. === Tier 2: Developmental and Niche Architectures Tier 2 platforms are functional, but less mature FreeBSD platforms. They are not supported by the security officer, release engineering, and port management teams. Tier 2 platforms may be Tier 1 platform candidates that are still under active development. Architectures reaching end of life may also be moved from Tier 1 status to Tier 2 status as the availability of resources to continue to maintain the system in a Production Quality state diminishes. Well-supported niche architectures may also be Tier 2. The FreeBSD Project provides the following guarantees to consumers of Tier 2 platforms: * The ports infrastructure should include basic support for Tier 2 architectures sufficient to support building ports and packages. This includes support for basic packages such as ports-mgmt/pkg, but there is no guarantee that arbitrary ports will be buildable or functional. * New features which are not inherently platform-specific should be feasible on all Tier 2 architectures if not implemented. * Tier 2 platforms will be included in the source tree. * Tier 2 platforms should be self-hosting either via the in-tree toolchain or an external toolchain. If an external toolchain is required, official binary packages for an external toolchain will be provided. * Tier 2 platforms should provide functional kernels and userlands even if an official release distribution is not provided. To maintain maturity of Tier 2 platforms, the FreeBSD Project will maintain the following resources to support development: * Inclusion in the `make universe` and `make tinderbox` targets. Collectively, developers are required to provide the following to maintain the Tier 2 status of a platform: * Changes to the source tree should not knowingly break the build of a Tier 2 platform. * Tier 2 architectures must have an active ecosystem of users and developers. * While changes are permitted to break the userland ABI, the ABI should not be broken gratuitously. Significant userland ABI changes should be restricted to major versions. * New features that are not yet implemented on Tier 2 architectures should provide a means of disabling them on those architectures. === Tier 3: Experimental Architectures Tier 3 platforms have at least partial FreeBSD support. They are _not_ supported by the security officer, release engineering, and port management teams. Tier 3 platforms are architectures in the early stages of development, for non-mainstream hardware platforms, or which are considered legacy systems unlikely to see broad future use. Initial support for Tier 3 platforms may exist in a separate repository rather than the main source repository. The FreeBSD Project provides no guarantees to consumers of Tier 3 platforms and is not committed to maintaining resources to support development. Tier 3 platforms may not always be buildable, nor are any kernel or userland ABIs considered stable. === Tier 4: Unsupported Architectures Tier 4 platforms are not supported in any form by the project. All systems not otherwise classified are Tier 4 systems. When a platform transitions to Tier 4, all support for the platform is removed from the source and ports trees. Note that ports support should remain as long as the platform is supported in a branch supported by ports. === Policy on Changing the Tier of an Architecture Systems may only be moved from one tier to another by approval of the FreeBSD Core Team, which shall make that decision in collaboration with the Security Officer, Release Engineering, and ports management teams. For a platform to be promoted to a higher tier, any missing support guarantees must be satisfied before the promotion is completed. [[ports]] == Ports Specific FAQ [[ports-qa-adding]] === Adding a New Port [[ports-qa-add-new]] ==== How do I add a new port? First, please read the section about repository copies. The easiest way to add a new port is the `addport` script located in the [.filename]#ports/Tools/scripts# directory. It adds a port from the directory specified, determining the category automatically from the port [.filename]#Makefile#. It also adds an entry to the port's category [.filename]#Makefile#. It was written by `{mharo}`, `{will}`, and `{garga}`. When sending questions about this script to the {freebsd-ports}, please also CC `{crees}`, the current maintainer. [[ports-qa-add-new-extra]] ==== Any other things I need to know when I add a new port? Check the port, preferably to make sure it compiles and packages correctly. The link:{porters-handbook}testing[Porters Handbook's Testing Chapter] contains more detailed instructions. See the link:{porters-handbook}testing#testing-portclippy[Portclippy / Portfmt] and the link:{porters-handbook}testing#testing-poudriere[Poudriere] sections. You do not necessarily have to eliminate all warnings but make sure you have fixed the simple ones. If the port came from a submitter who has not contributed to the Project before, add that person's name to the link:{contributors}#contrib-additional[Additional Contributors] section of the FreeBSD Contributors List. Close the PR if the port came in as a PR. To close a PR, change the state to `Issue Resolved` and the resolution as `Fixed`. [NOTE] ==== If for some reason using link:{porters-handbook}testing#testing-poudriere[Poudriere] to test the new port is not possible, the bare minimum of testing includes this sequence: [source,shell] .... # make install # make package # make deinstall # pkg add package you built above # make deinstall # make reinstall # make package .... Note that poudriere is the reference for package building, it the port does not build in poudriere, it will be removed. ==== [[ports-qa-removing]] === Removing an Existing Port [[ports-qa-remove-one]] ==== How do I remove an existing port? First, please read the section about repository copies. Before you remove the port, you have to verify there are no other ports depending on it. * Make sure there is no dependency on the port in the ports collection: ** The port's PKGNAME appears in exactly one line in a recent INDEX file. ** No other ports contains any reference to the port's directory or PKGNAME in their Makefiles + [TIP] ==== When using Git, consider using man:git-grep[1], it is much faster than `grep -r`. ==== + * Then, remove the port: + [.procedure] ==== * Remove the port's files and directory with `git rm`. * Remove the `SUBDIR` listing of the port in the parent directory [.filename]#Makefile#. * Add an entry to [.filename]#ports/MOVED#. * Search for entries in [.filename]#ports/security/vuxml/vuln.xml# and adjust them accordingly. In particular, check for previous packages with the new name which version could include the new port. * Remove the port from [.filename]#ports/LEGAL# if it is there. ==== Alternatively, you can use the rmport script, from [.filename]#ports/Tools/scripts#. This script was written by {vd}. When sending questions about this script to the {freebsd-ports}, please also CC {crees}, the current maintainer. [[ports-qa-move-port]] === How do I move a port to a new location? [.procedure] ==== . Perform a thorough check of the ports collection for any dependencies on the old port location/name, and update them. Running `grep` on [.filename]#INDEX# is not enough because some ports have dependencies enabled by compile-time options. A full man:git-grep[1] of the ports collection is recommended. . Remove the `SUBDIR` entry from the old category Makefile and add a `SUBDIR` entry to the new category Makefile. . Add an entry to [.filename]#ports/MOVED#. . Move the port with `git mv`. . Commit the changes. ==== [[ports-qa-copy-port]] === How do I copy a port to a new location? [.procedure] ==== . Copy port with `cp -R old-cat/old-port new-cat/new-port`. . Add the new port to the [.filename]#new-cat/Makefile#. . Change stuff in [.filename]#new-cat/new-port#. . Commit the changes. ==== [[ports-qa-freeze]] === Ports Freeze [[ports-qa-freeze-what]] ==== What is a “ports freeze”? A “ports freeze” was a restricted state the ports tree was put in before a release. It was used to ensure a higher quality for the packages shipped with a release. It usually lasted a couple of weeks. During that time, build problems were fixed, and the release packages were built. This practice is no longer used, as the packages for the releases are built from the current stable, quarterly branch. For more information on how to merge commits to the quarterly branch, see <>. [[ports-qa-quarterly]] === Quarterly Branches [[ports-qa-misc-request-mfh]] ==== What is the procedure to request authorization for merging a commit to the quarterly branch? As of November 30, 2020, there is no need to seek explicit approval to commit to the quarterly branch. [[ports-qa-misc-commit-mfh]] ==== What is the procedure for merging commits to the quarterly branch? Merging commits to the quarterly branch is very similar to MFC'ing a commit in the src repository, so basically: [source,shell] .... % git checkout 2021Q2 % git cherry-pick -x $HASH (verify everything is OK, for example by doing a build test) % git push .... where '$HASH' is the hash of the commit you want to copy over to the quarterly branch. The -x parameter ensures the hash '$HASH' of the main branch is included in the new commit message of the quarterly branch. [[ports-qa-new-category]] === Creating a New Category [[ports-qa-new-category-how]] ==== What is the procedure for creating a new category? Please see link:{porters-handbook}#proposing-categories[Proposing a New Category] in the Porter's Handbook. Once that procedure has been followed and the PR has been assigned to the {portmgr}, it is their decision whether or not to approve it. If they do, it is their responsibility to: [.procedure] ==== . Perform any needed moves. (This only applies to physical categories.) . Update the `VALID_CATEGORIES` definition in [.filename]#ports/Mk/bsd.port.mk#. . Assign the PR back to you. ==== [[ports-qa-new-category-physical]] ==== What do I need to do to implement a new physical category? [.procedure] ==== . Upgrade each moved port's [.filename]#Makefile#. Do not connect the new category to the build yet. + To do this, you will need to: + [.procedure] ====== . Change the port's `CATEGORIES` (this was the point of the exercise, remember?) The new category is listed first. This will help to ensure that the PKGORIGIN is correct. . Run a `make describe`. Since the top-level `make index` that you will be running in a few steps is an iteration of `make describe` over the entire ports hierarchy, catching any errors here will save you having to re-run that step later on. . If you want to be really thorough, now might be a good time to run man:portlint[1]. ====== + . Check that the ``PKGORIGIN``s are correct. The ports system uses each port's `CATEGORIES` entry to create its `PKGORIGIN`, which is used to connect installed packages to the port directory they were built from. If this entry is wrong, common port tools like man:pkg_version[1] and man:portupgrade[1] fail. + To do this, use the [.filename]#chkorigin.sh# tool: `env PORTSDIR=/path/to/ports sh -e /path/to/ports/Tools/scripts/chkorigin.sh`. This will check every port in the ports tree, even those not connected to the build, so you can run it directly after the move operation. Hint: do not forget to look at the ``PKGORIGIN``s of any slave ports of the ports you just moved! . On your own local system, test the proposed changes: first, comment out the SUBDIR entries in the old ports' categories' [.filename]##Makefile##s; then enable building the new category in [.filename]#ports/Makefile#. Run make checksubdirs in the affected category directories to check the SUBDIR entries. Next, in the [.filename]#ports/# directory, run make index. This can take over 40 minutes on even modern systems; however, it is a necessary step to prevent problems for other people. . Once this is done, you can commit the updated [.filename]#ports/Makefile# to connect the new category to the build and also commit the [.filename]#Makefile# changes for the old category or categories. . Add appropriate entries to [.filename]#ports/MOVED#. . Update the documentation by modifying: ** the link:{porters-handbook}#PORTING-CATEGORIES[list of categories] in the Porter's Handbook + . Only once all the above have been done, and no one is any longer reporting problems with the new ports, should the old ports be deleted from their previous locations in the repository. ==== ==== What do I need to do to implement a new virtual category? This is much simpler than a physical category. Only a few modifications are needed: * the link:{porters-handbook}#PORTING-CATEGORIES[list of categories] in the Porter's Handbook [[ports-qa-misc-questions]] === Miscellaneous Questions [[ports-qa-misc-blanket-approval]] ==== Are there changes that can be committed without asking the maintainer for approval? Blanket approval for most ports applies to these types of fixes: * Most infrastructure changes to a port (that is, modernizing, but not changing the functionality). For example, the blanket covers converting to new `USES` macros, enabling verbose builds, and switching to new ports system syntaxes. * Trivial and _tested_ build and runtime fixes. * Documentations or metadata changes to ports, like [.filename]#pkg-descr# or `COMMENT`. [IMPORTANT] ==== Exceptions to this are anything maintained by the {portmgr}, or the {security-officer}. No unauthorized commits may ever be made to ports maintained by those groups. ==== [[ports-qa-misc-correctly-building]] ==== How do I know if my port is building correctly or not? The packages are built multiple times each week. If a port fails, the maintainer will receive an email from `pkg-fallout@FreeBSD.org`. Reports for all the package builds (official, experimental, and non-regression) are aggregated at link:pkg-status.FreeBSD.org[pkg-status.FreeBSD.org]. [[ports-qa-misc-INDEX]] ==== I added a new port. Do I need to add it to the [.filename]#INDEX#? No. The file can either be generated by running `make index`, or a pre-generated version can be downloaded with `make fetchindex`. [[ports-qa-misc-no-touch]] ==== Are there any other files I am not allowed to touch? Any file directly under [.filename]#ports/#, or any file under a subdirectory that starts with an uppercase letter ([.filename]#Mk/#, [.filename]#Tools/#, etc.). In particular, the {portmgr} is very protective of [.filename]#ports/Mk/bsd.port*.mk# so do not commit changes to those files unless you want to face their wrath. [[ports-qa-misc-updated-distfile]] ==== What is the proper procedure for updating the checksum for a port distfile when the file changes without a version change? When the checksum for a distribution file is updated due to the author updating the file without changing the port revision, the commit message includes a summary of the relevant diffs between the original and new distfile to ensure that the distfile has not been corrupted or maliciously altered. If the current version of the port has been in the ports tree for a while, a copy of the old distfile will usually be available on the ftp servers; otherwise the author or maintainer should be contacted to find out why the distfile has changed. [[ports-exp-run]] ==== How can an experimental test build of the ports tree (exp-run) be requested? An exp-run must be completed before patches with a significant ports impact are committed. The patch can be against the ports tree or the base system. Full package builds will be done with the patches provided by the submitter, and the submitter is required to fix detected problems _(fallout)_ before commit. [.procedure] ==== . Go to the link:https://bugs.freebsd.org/submit[Bugzilla new PR page]. . Select the product your patch is about. . Fill in the bug report as normal. Remember to attach the patch. . If at the top it says “Show Advanced Fields” click on it. It will now say “Hide Advanced Fields”. Many new fields will be available. If it already says “Hide Advanced Fields”, no need to do anything. . In the “Flags” section, set the “exp-run” one to `?`. As for all other fields, hovering the mouse over any field shows more details. . Submit. Wait for the build to run. . {portmgr} will reply with a possible fallout. . Depending on the fallout: ** If there is no fallout, the procedure stops here, and the change can be committed, pending any other approval required. ... If there is fallout, it _must_ be fixed, either by fixing the ports directly in the ports tree, or adding to the submitted patch. ... When this is done, go back to step 6 saying the fallout was fixed and wait for the exp-run to be run again. Repeat as long as there are broken ports. ==== [[non-committers]] == Issues Specific to Developers Who Are Not Committers A few people who have access to the FreeBSD machines do not have commit bits. Almost all of this document will apply to these developers as well (except things specific to commits and the mailing list memberships that go with them). In particular, we recommend that you read: * <> * <> + [NOTE] ==== Get your mentor to add you to the "Additional Contributors" ([.filename]#~/documentation/content/en/articles/contributors/contrib-additional.adoc#), if you are not already listed there. ==== * <> * <> * <> [[google-analytics]] == Information About Google Analytics As of December 12, 2012, Google Analytics was enabled on the FreeBSD Project website to collect anonymized usage statistics regarding usage of the site. The information collected is valuable to the FreeBSD Documentation Project, to identify various problems on the FreeBSD website. [[google-analytics-policy]] === Google Analytics General Policy The FreeBSD Project takes visitor privacy very seriously. As such, the FreeBSD Project website honors the "Do Not Track" header _before_ fetching the tracking code from Google. For more information, please see the https://www.FreeBSD.org/privacy/[FreeBSD Privacy Policy]. Google Analytics access is _not_ arbitrarily allowed - access must be requested, voted on by the `{doceng}`, and explicitly granted. Requests for Google Analytics data must include a specific purpose. For example, a valid reason for requesting access would be "to see the most frequently used web browsers when viewing FreeBSD web pages to ensure page rendering speeds are acceptable." Conversely, "to see what web browsers are most frequently used" (without stating __why__) would be rejected. All requests must include the timeframe for which the data would be required. For example, it must be explicitly stated if the requested data would be needed for a timeframe covering a span of 3 weeks, or if the request would be one-time only. Any request for Google Analytics data without a clear, reasonable reason beneficial to the FreeBSD Project will be rejected. [[google-analytics-data]] === Data Available Through Google Analytics A few examples of the types of Google Analytics data available include: * Commonly used web browsers * Page load times * Site access by language [[misc]] == Miscellaneous Questions === Are there changes that can be committed without asking the maintainer for approval? Blanket approval for most ports applies to these types of fixes: * Most infrastructure changes to a port (that is, modernizing, but not changing the functionality). For example, the blanket covers converting to new `USES` macros, enabling verbose builds, and switching to new ports system syntaxes. * Trivial and _tested_ build and runtime fixes. * Documentations or metadata changes to ports, like [.filename]#pkg-descr# or `COMMENT`. === How do I access people.FreeBSD.org to put up personal or project information? `people.FreeBSD.org` is the same as `freefall.FreeBSD.org`. Just create a [.filename]#public_html# directory. Anything you place in that directory will automatically be visible under https://people.FreeBSD.org/[https://people.FreeBSD.org/]. === Where are the mailing list archives stored? The mailing lists are archived under [.filename]#/local/mail# on `freefall.FreeBSD.org`. === I would like to mentor a new committer. What process do I need to follow? See the https://www.freebsd.org/internal/new-account/[New Account Creation Procedure] document on the internal pages. [[benefits]] == Benefits and Perks for FreeBSD Committers [[benefits-recognition]] === Recognition Recognition as a competent software engineer is the longest lasting value. In addition, getting a chance to work with some of the best people that every engineer would dream of meeting is a great perk! [[benefits-freebsdmall]] === FreeBSD Mall FreeBSD committers can get a free 4-CD or DVD set at conferences from http://www.freebsdmall.com[FreeBSD Mall, Inc.]. [[benefits-irc]] === IRC In addition, developers may request a cloaked hostmask for their account on the Freenode IRC network in the form of `freebsd/developer/` or `freebsd/developer/`. To request a cloak, send an email to `{irc-email}` with your requested hostmask and NickServ account name. See the https://wiki.freebsd.org/IRC/Cloaks[IRC Cloaks] wiki page for more details. [[benefits-gandi]] === `Gandi.net` https://gandi.net[Gandi] provides website hosting, cloud computing, domain registration, and X.509 certificate services. Gandi offers an E-rate discount to all FreeBSD developers. In order to streamline the process of getting the discount first set up a Gandi account, fill in the billing information and select the currency. Then send an mail to mailto:non-profit@gandi.net[non-profit@gandi.net] using your `@freebsd.org` mail address, and indicate your Gandi handle. [[benefits-rsync]] === `rsync.net` https://rsync.net[rsync.net] provides cloud storage for offsite backup that is optimized for UNIX users. Their service runs entirely on FreeBSD and ZFS. rsync.net offers a free-forever 500 GB account to FreeBSD developers. Simply sign up at https://www.rsync.net/freebsd.html[https://www.rsync.net/freebsd.html] using your `@freebsd.org` address to receive this free account. diff --git a/documentation/content/en/articles/filtering-bridges/_index.adoc b/documentation/content/en/articles/filtering-bridges/_index.adoc index c729c8f934..4f54b24ace 100644 --- a/documentation/content/en/articles/filtering-bridges/_index.adoc +++ b/documentation/content/en/articles/filtering-bridges/_index.adoc @@ -1,285 +1,285 @@ --- title: Filtering Bridges authors: - author: Alex Dupre email: ale@FreeBSD.org description: Configuring firewalls and filtering on FreeBSD hosts acting as bridges rather than routers trademarks: ["freebsd", "3com", "intel", "general"] tags: ["network", "filtering", "bridges", "FreeBSD"] --- = Filtering Bridges :doctype: article :toc: macro :toclevels: 1 :icons: font :sectnums: :source-highlighter: rouge :experimental: :sectnumlevels: 6 ifeval::["{backend}" == "html5"] include::shared/en/urls.adoc[] endif::[] ifeval::["{backend}" == "pdf"] include::../../../../shared/en/urls.adoc[] endif::[] ifeval::["{backend}" == "epub3"] include::../../../../shared/en/urls.adoc[] endif::[] [.abstract-title] Abstract Often it is useful to divide one physical network (like an Ethernet) into two separate segments without having to create subnets, and use a router to link them together. The device that connects the two networks in this way is called a bridge. A FreeBSD system with two network interfaces is enough in order to act as a bridge. A bridge works by scanning the addresses of MAC level (Ethernet addresses) of the devices connected to each of its network interfaces and then forwarding the traffic between the two networks only if the source and the destination are on different segments. Under many points of view a bridge is similar to an Ethernet switch with only two ports. ''' toc::[] [[filtering-bridges-why]] == Why use a filtering bridge? More and more frequently, thanks to the lowering costs of broad band Internet connections (xDSL) and also because of the reduction of available IPv4 addresses, many companies are connected to the Internet 24 hours on 24 and with few (sometimes not even a power of 2) IP addresses. In these situations it is often desirable to have a firewall that filters incoming and outgoing traffic from and towards Internet, but a packet filtering solution based on router may not be applicable, either due to subnetting issues, the router is owned by the connectivity supplier (ISP), or because it does not support such functionalities. In these scenarios the use of a filtering bridge is highly advised. A bridge-based firewall can be configured and inserted between the xDSL router and your Ethernet hub/switch without any IP numbering issues. [[filtering-bridges-how]] == How to Install Adding bridge functionalities to a FreeBSD system is not difficult. Since 4.5 release it is possible to load such functionalities as modules instead of having to rebuild the kernel, simplifying the procedure a great deal. In the following subsections I will explain both installation ways. [IMPORTANT] ==== _Do not_ follow both instructions: a procedure _excludes_ the other one. Select the best choice according to your needs and abilities. ==== Before going on, be sure to have at least two Ethernet cards that support the promiscuous mode for both reception and transmission, since they must be able to send Ethernet packets with any address, not just their own. Moreover, to have a good throughput, the cards should be PCI bus mastering cards. The best choices are still the Intel EtherExpress(TM) Pro, followed by the 3Com(R) 3c9xx series. To simplify the firewall configuration it may be useful to have two cards of different manufacturers (using different drivers) in order to distinguish clearly which interface is connected to the router and which to the inner network. [[filtering-bridges-kernel]] === Kernel Configuration So you have decided to use the older but well tested installation method. To begin, you have to add the following rows to your kernel configuration file: [.programlisting] .... options BRIDGE options IPFIREWALL options IPFIREWALL_VERBOSE .... The first line is to compile the bridge support, the second one is the firewall and the third one is the logging functions of the firewall. Now it is necessary to build and install the new kernel. You may find detailed instructions in the link:{handbook}#kernelconfig-building[Building and Installing a Custom Kernel] section of the FreeBSD Handbook. [[filtering-bridges-modules]] === Modules Loading If you have chosen to use the new and simpler installation method, the only thing to do now is add the following row to [.filename]#/boot/loader.conf#: [.programlisting] .... bridge_load="YES" .... In this way, during the system startup, the [.filename]#bridge.ko# module will be loaded together with the kernel. It is not required to add a similar row for the [.filename]#ipfw.ko# module, since it will be loaded automatically after the execution of the steps in the following section. [[filtering-bridges-finalprep]] == Final Preparation Before rebooting in order to load the new kernel or the required modules (according to the previously chosen installation method), you have to make some changes to the [.filename]#/etc/rc.conf# configuration file. The default rule of the firewall is to reject all IP packets. Initially we will set up an `open` firewall, in order to verify its operation without any issue related to packet filtering (in case you are going to execute this procedure remotely, such configuration will avoid you to remain isolated from the network). Put these lines in [.filename]#/etc/rc.conf#: [.programlisting] .... firewall_enable="YES" firewall_type="open" firewall_quiet="YES" firewall_logging="YES" .... The first row will enable the firewall (and will load the module [.filename]#ipfw.ko# if it is not compiled in the kernel), the second one to set up it in `open` mode (as explained in [.filename]#/etc/rc.firewall#), the third one to not show rules loading and the fourth one to enable logging support. About the configuration of the network interfaces, the most used way is to assign an IP to only one of the network cards, but the bridge will work equally even if both interfaces or none has a configured IP. In the last case (IP-less) the bridge machine will be still more hidden, as inaccessible from the network: to configure it, you have to login from console or through a third network interface separated from the bridge. Sometimes, during the system startup, some programs require network access, say for domain resolution: in this case it is necessary to assign an IP to the external interface (the one connected to Internet, where DNS server resides), since the bridge will be activated at the end of the startup procedure. It means that the [.filename]#fxp0# interface (in our case) must be mentioned in the ifconfig section of the [.filename]#/etc/rc.conf# file, while the [.filename]#xl0# is not. Assigning an IP to both the network cards does not make much sense, unless, during the start procedure, applications should access to services on both Ethernet segments. There is another important thing to know. When running IP over Ethernet, there are actually two Ethernet protocols in use: one is IP, the other is ARP. ARP does the conversion of the IP address of a host into its Ethernet address (MAC layer). In order to allow the communication between two hosts separated by the bridge, it is necessary that the bridge will forward ARP packets. Such protocol is not included in the IP layer, since it exists only with IP over Ethernet. The FreeBSD firewall filters exclusively on the IP layer and therefore all non-IP packets (ARP included) will be forwarded without being filtered, even if the firewall is configured to not permit anything. Now it is time to reboot the system and use it as before: there will be some new messages about the bridge and the firewall, but the bridge will not be activated and the firewall, being in `open` mode, will not avoid any operations. If there are any problems, you should sort them out now before proceeding. [[filtering-bridges-enabling]] == Enabling the Bridge At this point, to enable the bridge, you have to execute the following commands (having the shrewdness to replace the names of the two network interfaces [.filename]#fxp0# and [.filename]#xl0# with your own ones): [source,shell] .... # sysctl net.link.ether.bridge.config=fxp0:0,xl0:0 # sysctl net.link.ether.bridge.ipfw=1 # sysctl net.link.ether.bridge.enable=1 .... The first row specifies which interfaces should be activated by the bridge, the second one will enable the firewall on the bridge and finally the third one will enable the bridge. At this point you should be able to insert the machine between two sets of hosts without compromising any communication abilities between them. If so, the next step is to add the `net.link.ether.bridge._[blah]_=_[blah]_` portions of these rows to the [.filename]#/etc/sysctl.conf# file, in order to have them execute at startup. [[filtering-bridges-ipfirewall]] == Configuring The Firewall Now it is time to create your own file with custom firewall rules, in order to secure the inside network. There will be some complication in doing this because not all of the firewall functionalities are available on bridged packets. Furthermore, there is a difference between the packets that are in the process of being forwarded and packets that are being received by the local machine. In general, incoming packets are run through the firewall only once, not twice as is normally the case; in fact they are filtered only upon receipt, so rules that use `out` or `xmit` will never match. Personally, I use `in via` which is an older syntax, but one that has a sense when you read it. Another limitation is that you are restricted to use only `pass` or `drop` commands for packets filtered by a bridge. Sophisticated things like `divert`, `forward` or `reject` are not available. Such options can still be used, but only on traffic to or from the bridge machine itself (if it has an IP address). New in FreeBSD 4.0, is the concept of stateful filtering. This is a big improvement for UDP traffic, which typically is a request going out, followed shortly thereafter by a response with the exact same set of IP addresses and port numbers (but with source and destination reversed, of course). For firewalls that have no statekeeping, there is almost no way to deal with this sort of traffic as a single session. But with a firewall that can "remember" an outgoing UDP packet and, for the next few minutes, allow a response, handling UDP services is trivial. The following example shows how to do it. It is possible to do the same thing with TCP packets. This allows you to avoid some denial of service attacks and other nasty tricks, but it also typically makes your state table grow quickly in size. Let's look at an example setup. Note first that at the top of [.filename]#/etc/rc.firewall# there are already standard rules for the loopback interface [.filename]#lo0#, so we should not have to care for them anymore. Custom rules should be put in a separate file (say [.filename]#/etc/rc.firewall.local#) and loaded at system startup, by modifying the row of [.filename]#/etc/rc.conf# where we defined the `open` firewall: [.programlisting] .... firewall_type="/etc/rc.firewall.local" .... [IMPORTANT] ==== You have to specify the _full_ path, otherwise it will not be loaded with the risk to remain isolated from the network. ==== For our example imagine to have the [.filename]#fxp0# interface connected towards the outside (Internet) and the [.filename]#xl0# towards the inside (LAN). The bridge machine has the IP `1.2.3.4` (it is not possible that your ISP can give you an address quite like this, but for our example it is good). [.programlisting] .... # Things that we have kept state on before get to go through in a hurry add check-state # Throw away RFC 1918 networks add drop all from 10.0.0.0/8 to any in via fxp0 add drop all from 172.16.0.0/12 to any in via fxp0 add drop all from 192.168.0.0/16 to any in via fxp0 # Allow the bridge machine to say anything it wants # (if the machine is IP-less do not include these rows) add pass tcp from 1.2.3.4 to any setup keep-state add pass udp from 1.2.3.4 to any keep-state add pass ip from 1.2.3.4 to any # Allow the inside hosts to say anything they want add pass tcp from any to any in via xl0 setup keep-state add pass udp from any to any in via xl0 keep-state add pass ip from any to any in via xl0 # TCP section # Allow SSH add pass tcp from any to any 22 in via fxp0 setup keep-state # Allow SMTP only towards the mail server add pass tcp from any to relay 25 in via fxp0 setup keep-state -# Allow zone transfers only by the slave name server [dns2.nic.it] +# Allow zone transfers only by the secondary name server [dns2.nic.it] add pass tcp from 193.205.245.8 to ns 53 in via fxp0 setup keep-state # Pass ident probes. It is better than waiting for them to timeout add pass tcp from any to any 113 in via fxp0 setup keep-state # Pass the "quarantine" range add pass tcp from any to any 49152-65535 in via fxp0 setup keep-state # UDP section # Allow DNS only towards the name server add pass udp from any to ns 53 in via fxp0 keep-state # Pass the "quarantine" range add pass udp from any to any 49152-65535 in via fxp0 keep-state # ICMP section # Pass 'ping' add pass icmp from any to any icmptypes 8 keep-state # Pass error messages generated by 'traceroute' add pass icmp from any to any icmptypes 3 add pass icmp from any to any icmptypes 11 # Everything else is suspect add drop log all from any to any .... Those of you who have set up firewalls before may notice some things missing. In particular, there are no anti-spoofing rules, in fact we did _not_ add: [.programlisting] .... add deny all from 1.2.3.4/8 to any in via fxp0 .... That is, drop packets that are coming in from the outside claiming to be from our network. This is something that you would commonly do to be sure that someone does not try to evade the packet filter, by generating nefarious packets that look like they are from the inside. The problem with that is that there is _at least_ one host on the outside interface that you do not want to ignore: the router. But usually, the ISP anti-spoofs at their router, so we do not need to bother that much. The last rule seems to be an exact duplicate of the default rule, that is, do not let anything pass that is not specifically allowed. But there is a difference: all suspected traffic will be logged. There are two rules for passing SMTP and DNS traffic towards the mail server and the name server, if you have them. Obviously the whole rule set should be flavored to personal taste, this is only a specific example (rule format is described accurately in the man:ipfw[8] man page). Note that in order for "relay" and "ns" to work, name service lookups must work _before_ the bridge is enabled. This is an example of making sure that you set the IP on the correct network card. Alternatively it is possible to specify the IP address instead of the host name (required if the machine is IP-less). People that are used to setting up firewalls are probably also used to either having a `reset` or a `forward` rule for ident packets (TCP port 113). Unfortunately, this is not an applicable option with the bridge, so the best thing is to simply pass them to their destination. As long as that destination machine is not running an ident daemon, this is relatively harmless. The alternative is dropping connections on port 113, which creates some problems with services like IRC (the ident probe must timeout). The only other thing that is a little weird that you may have noticed is that there is a rule to let the bridge machine speak, and another for internal hosts. Remember that this is because the two sets of traffic will take different paths through the kernel and into the packet filter. The inside net will go through the bridge, while the local machine will use the normal IP stack to speak. Thus the two rules to handle the different cases. The `in via fxp0` rules work for both paths. In general, if you use `in via` rules throughout the filter, you will need to make an exception for locally generated packets, because they did not come in via any of our interfaces. [[filtering-bridges-contributors]] == Contributors Many parts of this article have been taken, updated and adapted from an old text about bridging, edited by Nick Sayer. A pair of inspirations are due to an introduction on bridging by Steve Peterson. A big thanks to Luigi Rizzo for the implementation of the bridge code in FreeBSD and for the time he has dedicated to me answering all of my related questions. A thanks goes out also to Tom Rhodes who looked over my job of translation from Italian (the original language of this article) into English. diff --git a/documentation/content/en/articles/linux-emulation/_index.adoc b/documentation/content/en/articles/linux-emulation/_index.adoc index 6512d55b80..e9d86ae201 100644 --- a/documentation/content/en/articles/linux-emulation/_index.adoc +++ b/documentation/content/en/articles/linux-emulation/_index.adoc @@ -1,1427 +1,1427 @@ --- title: Linux® emulation in FreeBSD authors: - author: Roman Divacky email: rdivacky@FreeBSD.org description: A technical description about the internals of the Linux emulation layer in FreeBSD trademarks: ["freebsd", "ibm", "adobe", "netbsd", "realnetworks", "oracle", "linux", "sun", "general"] tags: ["Emulation", "Linuxulator", "kernel", "FreeBSD"] --- = Linux(R) emulation in FreeBSD :doctype: article :toc: macro :toclevels: 1 :icons: font :sectnums: :sectnumlevels: 6 :source-highlighter: rouge :experimental: ifeval::["{backend}" == "html5"] include::shared/authors.adoc[] endif::[] ifeval::["{backend}" == "pdf"] include::../../../../shared/authors.adoc[] endif::[] ifeval::["{backend}" == "epub3"] include::../../../../shared/authors.adoc[] endif::[] [.abstract-title] Abstract This masters thesis deals with updating the Linux(R) emulation layer (the so called _Linuxulator_). The task was to update the layer to match the functionality of Linux(R) 2.6. As a reference implementation, the Linux(R) 2.6.16 kernel was chosen. The concept is loosely based on the NetBSD implementation. Most of the work was done in the summer of 2006 as a part of the Google Summer of Code students program. The focus was on bringing the _NPTL_ (new POSIX(R) thread library) support into the emulation layer, including _TLS_ (thread local storage), _futexes_ (fast user space mutexes), _PID mangling_, and some other minor things. Many small problems were identified and fixed in the process. My work was integrated into the main FreeBSD source repository and will be shipped in the upcoming 7.0R release. We, the emulation development team, are working on making the Linux(R) 2.6 emulation the default emulation layer in FreeBSD. ''' toc::[] [[intro]] == Introduction In the last few years the open source UNIX(R) based operating systems started to be widely deployed on server and client machines. Among these operating systems I would like to point out two: FreeBSD, for its BSD heritage, time proven code base and many interesting features and Linux(R) for its wide user base, enthusiastic open developer community and support from large companies. FreeBSD tends to be used on server class machines serving heavy duty networking tasks with less usage on desktop class machines for ordinary users. While Linux(R) has the same usage on servers, but it is used much more by home based users. This leads to a situation where there are many binary only programs available for Linux(R) that lack support for FreeBSD. Naturally, a need for the ability to run Linux(R) binaries on a FreeBSD system arises and this is what this thesis deals with: the emulation of the Linux(R) kernel in the FreeBSD operating system. During the Summer of 2006 Google Inc. sponsored a project which focused on extending the Linux(R) emulation layer (the so called Linuxulator) in FreeBSD to include Linux(R) 2.6 facilities. This thesis is written as a part of this project. [[inside]] == A look inside... In this section we are going to describe every operating system in question. How they deal with syscalls, trapframes etc., all the low-level stuff. We also describe the way they understand common UNIX(R) primitives like what a PID is, what a thread is, etc. In the third subsection we talk about how UNIX(R) on UNIX(R) emulation could be done in general. [[what-is-unix]] === What is UNIX(R) UNIX(R) is an operating system with a long history that has influenced almost every other operating system currently in use. Starting in the 1960s, its development continues to this day (although in different projects). UNIX(R) development soon forked into two main ways: the BSDs and System III/V families. They mutually influenced themselves by growing a common UNIX(R) standard. Among the contributions originated in BSD we can name virtual memory, TCP/IP networking, FFS, and many others. The System V branch contributed to SysV interprocess communication primitives, copy-on-write, etc. UNIX(R) itself does not exist any more but its ideas have been used by many other operating systems world wide thus forming the so called UNIX(R)-like operating systems. These days the most influential ones are Linux(R), Solaris, and possibly (to some extent) FreeBSD. There are in-company UNIX(R) derivatives (AIX, HP-UX etc.), but these have been more and more migrated to the aforementioned systems. Let us summarize typical UNIX(R) characteristics. [[tech-details]] === Technical details Every running program constitutes a process that represents a state of the computation. Running process is divided between kernel-space and user-space. Some operations can be done only from kernel space (dealing with hardware etc.), but the process should spend most of its lifetime in the user space. The kernel is where the management of the processes, hardware, and low-level details take place. The kernel provides a standard unified UNIX(R) API to the user space. The most important ones are covered below. [[kern-proc-comm]] ==== Communication between kernel and user space process Common UNIX(R) API defines a syscall as a way to issue commands from a user space process to the kernel. The most common implementation is either by using an interrupt or specialized instruction (think of `SYSENTER`/`SYSCALL` instructions for ia32). Syscalls are defined by a number. For example in FreeBSD, the syscall number 85 is the man:swapon[2] syscall and the syscall number 132 is man:mkfifo[2]. Some syscalls need parameters, which are passed from the user-space to the kernel-space in various ways (implementation dependant). Syscalls are synchronous. Another possible way to communicate is by using a _trap_. Traps occur asynchronously after some event occurs (division by zero, page fault etc.). A trap can be transparent for a process (page fault) or can result in a reaction like sending a _signal_ (division by zero). [[proc-proc-comm]] ==== Communication between processes There are other APIs (System V IPC, shared memory etc.) but the single most important API is signal. Signals are sent by processes or by the kernel and received by processes. Some signals can be ignored or handled by a user supplied routine, some result in a predefined action that cannot be altered or ignored. [[proc-mgmt]] ==== Process management Kernel instances are processed first in the system (so called init). Every running process can create its identical copy using the man:fork[2] syscall. Some slightly modified versions of this syscall were introduced but the basic semantic is the same. Every running process can morph into some other process using the man:exec[3] syscall. Some modifications of this syscall were introduced but all serve the same basic purpose. Processes end their lives by calling the man:exit[2] syscall. Every process is identified by a unique number called PID. Every process has a defined parent (identified by its PID). [[thread-mgmt]] ==== Thread management Traditional UNIX(R) does not define any API nor implementation for threading, while POSIX(R) defines its threading API but the implementation is undefined. Traditionally there were two ways of implementing threads. Handling them as separate processes (1:1 threading) or envelope the whole thread group in one process and managing the threading in userspace (1:N threading). Comparing main features of each approach: 1:1 threading * - heavyweight threads * - the scheduling cannot be altered by the user (slightly mitigated by the POSIX(R) API) * + no syscall wrapping necessary * + can utilize multiple CPUs 1:N threading * + lightweight threads * + scheduling can be easily altered by the user * - syscalls must be wrapped * - cannot utilize more than one CPU [[what-is-freebsd]] === What is FreeBSD? The FreeBSD project is one of the oldest open source operating systems currently available for daily use. It is a direct descendant of the genuine UNIX(R) so it could be claimed that it is a true UNIX(R) although licensing issues do not permit that. The start of the project dates back to the early 1990's when a crew of fellow BSD users patched the 386BSD operating system. Based on this patchkit a new operating system arose named FreeBSD for its liberal license. Another group created the NetBSD operating system with different goals in mind. We will focus on FreeBSD. FreeBSD is a modern UNIX(R)-based operating system with all the features of UNIX(R). Preemptive multitasking, multiuser facilities, TCP/IP networking, memory protection, symmetric multiprocessing support, virtual memory with merged VM and buffer cache, they are all there. One of the interesting and extremely useful features is the ability to emulate other UNIX(R)-like operating systems. As of December 2006 and 7-CURRENT development, the following emulation functionalities are supported: * FreeBSD/i386 emulation on FreeBSD/amd64 * FreeBSD/i386 emulation on FreeBSD/ia64 * Linux(R)-emulation of Linux(R) operating system on FreeBSD * NDIS-emulation of Windows networking drivers interface * NetBSD-emulation of NetBSD operating system * PECoff-support for PECoff FreeBSD executables * SVR4-emulation of System V revision 4 UNIX(R) Actively developed emulations are the Linux(R) layer and various FreeBSD-on-FreeBSD layers. Others are not supposed to work properly nor be usable these days. [[freebsd-tech-details]] ==== Technical details FreeBSD is traditional flavor of UNIX(R) in the sense of dividing the run of processes into two halves: kernel space and user space run. There are two types of process entry to the kernel: a syscall and a trap. There is only one way to return. In the subsequent sections we will describe the three gates to/from the kernel. The whole description applies to the i386 architecture as the Linuxulator only exists there but the concept is similar on other architectures. The information was taken from [1] and the source code. [[freebsd-sys-entries]] ===== System entries FreeBSD has an abstraction called an execution class loader, which is a wedge into the man:execve[2] syscall. This employs a structure `sysentvec`, which describes an executable ABI. It contains things like errno translation table, signal translation table, various functions to serve syscall needs (stack fixup, coredumping, etc.). Every ABI the FreeBSD kernel wants to support must define this structure, as it is used later in the syscall processing code and at some other places. System entries are handled by trap handlers, where we can access both the kernel-space and the user-space at once. [[freebsd-syscalls]] ===== Syscalls Syscalls on FreeBSD are issued by executing interrupt `0x80` with register `%eax` set to a desired syscall number with arguments passed on the stack. When a process issues an interrupt `0x80`, the `int0x80` syscall trap handler is issued (defined in [.filename]#sys/i386/i386/exception.s#), which prepares arguments (i.e. copies them on to the stack) for a call to a C function man:syscall[2] (defined in [.filename]#sys/i386/i386/trap.c#), which processes the passed in trapframe. The processing consists of preparing the syscall (depending on the `sysvec` entry), determining if the syscall is 32-bit or 64-bit one (changes size of the parameters), then the parameters are copied, including the syscall. Next, the actual syscall function is executed with processing of the return code (special cases for `ERESTART` and `EJUSTRETURN` errors). Finally an `userret()` is scheduled, switching the process back to the users-pace. The parameters to the actual syscall handler are passed in the form of `struct thread *td`, `struct syscall args *` arguments where the second parameter is a pointer to the copied in structure of parameters. [[freebsd-traps]] ===== Traps Handling of traps in FreeBSD is similar to the handling of syscalls. Whenever a trap occurs, an assembler handler is called. It is chosen between alltraps, alltraps with regs pushed or calltrap depending on the type of the trap. This handler prepares arguments for a call to a C function `trap()` (defined in [.filename]#sys/i386/i386/trap.c#), which then processes the occurred trap. After the processing it might send a signal to the process and/or exit to userland using `userret()`. [[freebsd-exits]] ===== Exits Exits from kernel to userspace happen using the assembler routine `doreti` regardless of whether the kernel was entered via a trap or via a syscall. This restores the program status from the stack and returns to the userspace. [[freebsd-unix-primitives]] ===== UNIX(R) primitives FreeBSD operating system adheres to the traditional UNIX(R) scheme, where every process has a unique identification number, the so called _PID_ (Process ID). PID numbers are allocated either linearly or randomly ranging from `0` to `PID_MAX`. The allocation of PID numbers is done using linear searching of PID space. Every thread in a process receives the same PID number as result of the man:getpid[2] call. There are currently two ways to implement threading in FreeBSD. The first way is M:N threading followed by the 1:1 threading model. The default library used is M:N threading (`libpthread`) and you can switch at runtime to 1:1 threading (`libthr`). The plan is to switch to 1:1 library by default soon. Although those two libraries use the same kernel primitives, they are accessed through different API(es). The M:N library uses the `kse_*` family of syscalls while the 1:1 library uses the `thr_*` family of syscalls. Due to this, there is no general concept of thread ID shared between kernel and userspace. Of course, both threading libraries implement the pthread thread ID API. Every kernel thread (as described by `struct thread`) has td tid identifier but this is not directly accessible from userland and solely serves the kernel's needs. It is also used for 1:1 threading library as pthread's thread ID but handling of this is internal to the library and cannot be relied on. As stated previously there are two implementations of threading in FreeBSD. The M:N library divides the work between kernel space and userspace. Thread is an entity that gets scheduled in the kernel but it can represent various number of userspace threads. M userspace threads get mapped to N kernel threads thus saving resources while keeping the ability to exploit multiprocessor parallelism. Further information about the implementation can be obtained from the man page or [1]. The 1:1 library directly maps a userland thread to a kernel thread thus greatly simplifying the scheme. None of these designs implement a fairness mechanism (such a mechanism was implemented but it was removed recently because it caused serious slowdown and made the code more difficult to deal with). [[what-is-linux]] === What is Linux(R) Linux(R) is a UNIX(R)-like kernel originally developed by Linus Torvalds, and now being contributed to by a massive crowd of programmers all around the world. From its mere beginnings to today, with wide support from companies such as IBM or Google, Linux(R) is being associated with its fast development pace, full hardware support and benevolent dictator model of organization. Linux(R) development started in 1991 as a hobbyist project at University of Helsinki in Finland. Since then it has obtained all the features of a modern UNIX(R)-like OS: multiprocessing, multiuser support, virtual memory, networking, basically everything is there. There are also highly advanced features like virtualization etc. As of 2006 Linux(R) seems to be the most widely used open source operating system with support from independent software vendors like Oracle, RealNetworks, Adobe, etc. Most of the commercial software distributed for Linux(R) can only be obtained in a binary form so recompilation for other operating systems is impossible. Most of the Linux(R) development happens in a Git version control system. Git is a distributed system so there is no central source of the Linux(R) code, but some branches are considered prominent and official. The version number scheme implemented by Linux(R) consists of four numbers A.B.C.D. Currently development happens in 2.6.C.D, where C represents major version, where new features are added or changed while D is a minor version for bugfixes only. More information can be obtained from [3]. [[linux-tech-details]] ==== Technical details Linux(R) follows the traditional UNIX(R) scheme of dividing the run of a process in two halves: the kernel and user space. The kernel can be entered in two ways: via a trap or via a syscall. The return is handled only in one way. The further description applies to Linux(R) 2.6 on the i386(TM) architecture. This information was taken from [2]. [[linux-syscalls]] ===== Syscalls Syscalls in Linux(R) are performed (in userspace) using `syscallX` macros where X substitutes a number representing the number of parameters of the given syscall. This macro translates to a code that loads `%eax` register with a number of the syscall and executes interrupt `0x80`. After this syscall return is called, which translates negative return values to positive `errno` values and sets `res` to `-1` in case of an error. Whenever the interrupt `0x80` is called the process enters the kernel in system call trap handler. This routine saves all registers on the stack and calls the selected syscall entry. Note that the Linux(R) calling convention expects parameters to the syscall to be passed via registers as shown here: . parameter -> `%ebx` . parameter -> `%ecx` . parameter -> `%edx` . parameter -> `%esi` . parameter -> `%edi` . parameter -> `%ebp` There are some exceptions to this, where Linux(R) uses different calling convention (most notably the `clone` syscall). [[linux-traps]] ===== Traps The trap handlers are introduced in [.filename]#arch/i386/kernel/traps.c# and most of these handlers live in [.filename]#arch/i386/kernel/entry.S#, where handling of the traps happens. [[linux-exits]] ===== Exits Return from the syscall is managed by syscall man:exit[3], which checks for the process having unfinished work, then checks whether we used user-supplied selectors. If this happens stack fixing is applied and finally the registers are restored from the stack and the process returns to the userspace. [[linux-unix-primitives]] ===== UNIX(R) primitives In the 2.6 version, the Linux(R) operating system redefined some of the traditional UNIX(R) primitives, notably PID, TID and thread. PID is defined not to be unique for every process, so for some processes (threads) man:getppid[2] returns the same value. Unique identification of process is provided by TID. This is because _NPTL_ (New POSIX(R) Thread Library) defines threads to be normal processes (so called 1:1 threading). Spawning a new process in Linux(R) 2.6 happens using the `clone` syscall (fork variants are reimplemented using it). This clone syscall defines a set of flags that affect behavior of the cloning process regarding thread implementation. The semantic is a bit fuzzy as there is no single flag telling the syscall to create a thread. Implemented clone flags are: * `CLONE_VM` - processes share their memory space * `CLONE_FS` - share umask, cwd and namespace * `CLONE_FILES` - share open files * `CLONE_SIGHAND` - share signal handlers and blocked signals * `CLONE_PARENT` - share parent * `CLONE_THREAD` - be thread (further explanation below) * `CLONE_NEWNS` - new namespace * `CLONE_SYSVSEM` - share SysV undo structures * `CLONE_SETTLS` - setup TLS at supplied address * `CLONE_PARENT_SETTID` - set TID in the parent * `CLONE_CHILD_CLEARTID` - clear TID in the child * `CLONE_CHILD_SETTID` - set TID in the child `CLONE_PARENT` sets the real parent to the parent of the caller. This is useful for threads because if thread A creates thread B we want thread B to be parented to the parent of the whole thread group. `CLONE_THREAD` does exactly the same thing as `CLONE_PARENT`, `CLONE_VM` and `CLONE_SIGHAND`, rewrites PID to be the same as PID of the caller, sets exit signal to be none and enters the thread group. `CLONE_SETTLS` sets up GDT entries for TLS handling. The `CLONE_*_*TID` set of flags sets/clears user supplied address to TID or 0. As you can see the `CLONE_THREAD` does most of the work and does not seem to fit the scheme very well. The original intention is unclear (even for authors, according to comments in the code) but I think originally there was one threading flag, which was then parcelled among many other flags but this separation was never fully finished. It is also unclear what this partition is good for as glibc does not use that so only hand-written use of the clone permits a programmer to access this features. For non-threaded programs the PID and TID are the same. For threaded programs the first thread PID and TID are the same and every created thread shares the same PID and gets assigned a unique TID (because `CLONE_THREAD` is passed in) also parent is shared for all processes forming this threaded program. The code that implements man:pthread_create[3] in NPTL defines the clone flags like this: [.programlisting] .... int clone_flags = (CLONE_VM | CLONE_FS | CLONE_FILES | CLONE_SIGNAL | CLONE_SETTLS | CLONE_PARENT_SETTID | CLONE_CHILD_CLEARTID | CLONE_SYSVSEM #if __ASSUME_NO_CLONE_DETACHED == 0 | CLONE_DETACHED #endif | 0); .... The `CLONE_SIGNAL` is defined like [.programlisting] .... #define CLONE_SIGNAL (CLONE_SIGHAND | CLONE_THREAD) .... the last 0 means no signal is sent when any of the threads exits. [[what-is-emu]] === What is emulation According to a dictionary definition, emulation is the ability of a program or device to imitate another program or device. This is achieved by providing the same reaction to a given stimulus as the emulated object. In practice, the software world mostly sees three types of emulation - a program used to emulate a machine (QEMU, various game console emulators etc.), software emulation of a hardware facility (OpenGL emulators, floating point units emulation etc.) and operating system emulation (either in kernel of the operating system or as a userspace program). Emulation is usually used in a place, where using the original component is not feasible nor possible at all. For example someone might want to use a program developed for a different operating system than they use. Then emulation comes in handy. Sometimes there is no other way but to use emulation - e.g. when the hardware device you try to use does not exist (yet/anymore) then there is no other way but emulation. This happens often when porting an operating system to a new (non-existent) platform. Sometimes it is just cheaper to emulate. Looking from an implementation point of view, there are two main approaches to the implementation of emulation. You can either emulate the whole thing - accepting possible inputs of the original object, maintaining inner state and emitting correct output based on the state and/or input. This kind of emulation does not require any special conditions and basically can be implemented anywhere for any device/program. The drawback is that implementing such emulation is quite difficult, time-consuming and error-prone. In some cases we can use a simpler approach. Imagine you want to emulate a printer that prints from left to right on a printer that prints from right to left. It is obvious that there is no need for a complex emulation layer but simply reversing of the printed text is sufficient. Sometimes the emulating environment is very similar to the emulated one so just a thin layer of some translation is necessary to provide fully working emulation! As you can see this is much less demanding to implement, so less time-consuming and error-prone than the previous approach. But the necessary condition is that the two environments must be similar enough. The third approach combines the two previous. Most of the time the objects do not provide the same capabilities so in a case of emulating the more powerful one on the less powerful we have to emulate the missing features with full emulation described above. This master thesis deals with emulation of UNIX(R) on UNIX(R), which is exactly the case, where only a thin layer of translation is sufficient to provide full emulation. The UNIX(R) API consists of a set of syscalls, which are usually self contained and do not affect some global kernel state. There are a few syscalls that affect inner state but this can be dealt with by providing some structures that maintain the extra state. No emulation is perfect and emulations tend to lack some parts but this usually does not cause any serious drawbacks. Imagine a game console emulator that emulates everything but music output. No doubt that the games are playable and one can use the emulator. It might not be that comfortable as the original game console but its an acceptable compromise between price and comfort. The same goes with the UNIX(R) API. Most programs can live with a very limited set of syscalls working. Those syscalls tend to be the oldest ones (man:read[2]/man:write[2], man:fork[2] family, man:signal[3] handling, man:exit[3], man:socket[2] API) hence it is easy to emulate because their semantics is shared among all UNIX(R)es, which exist todays. [[freebsd-emulation]] == Emulation === How emulation works in FreeBSD As stated earlier, FreeBSD supports running binaries from several other UNIX(R)es. This works because FreeBSD has an abstraction called the execution class loader. This wedges into the man:execve[2] syscall, so when man:execve[2] is about to execute a binary it examines its type. There are basically two types of binaries in FreeBSD. Shell-like text scripts which are identified by `#!` as their first two characters and normal (typically _ELF_) binaries, which are a representation of a compiled executable object. The vast majority (one could say all of them) of binaries in FreeBSD are from type ELF. ELF files contain a header, which specifies the OS ABI for this ELF file. By reading this information, the operating system can accurately determine what type of binary the given file is. Every OS ABI must be registered in the FreeBSD kernel. This applies to the FreeBSD native OS ABI, as well. So when man:execve[2] executes a binary it iterates through the list of registered APIs and when it finds the right one it starts to use the information contained in the OS ABI description (its syscall table, `errno` translation table, etc.). So every time the process calls a syscall, it uses its own set of syscalls instead of some global one. This effectively provides a very elegant and easy way of supporting execution of various binary formats. The nature of emulation of different OSes (and also some other subsystems) led developers to invite a handler event mechanism. There are various places in the kernel, where a list of event handlers are called. Every subsystem can register an event handler and they are called accordingly. For example, when a process exits there is a handler called that possibly cleans up whatever the subsystem needs to be cleaned. Those simple facilities provide basically everything that is needed for the emulation infrastructure and in fact these are basically the only things necessary to implement the Linux(R) emulation layer. [[freebsd-common-primitives]] === Common primitives in the FreeBSD kernel Emulation layers need some support from the operating system. I am going to describe some of the supported primitives in the FreeBSD operating system. [[freebsd-locking-primitives]] ==== Locking primitives Contributed by: `{attilio}` The FreeBSD synchronization primitive set is based on the idea to supply a rather huge number of different primitives in a way that the better one can be used for every particular, appropriate situation. To a high level point of view you can consider three kinds of synchronization primitives in the FreeBSD kernel: * atomic operations and memory barriers * locks * scheduling barriers Below there are descriptions for the 3 families. For every lock, you should really check the linked manpage (where possible) for more detailed explanations. [[freebsd-atomic-op]] ===== Atomic operations and memory barriers Atomic operations are implemented through a set of functions performing simple arithmetics on memory operands in an atomic way with respect to external events (interrupts, preemption, etc.). Atomic operations can guarantee atomicity just on small data types (in the magnitude order of the `.long.` architecture C data type), so should be rarely used directly in the end-level code, if not only for very simple operations (like flag setting in a bitmap, for example). In fact, it is rather simple and common to write down a wrong semantic based on just atomic operations (usually referred as lock-less). The FreeBSD kernel offers a way to perform atomic operations in conjunction with a memory barrier. The memory barriers will guarantee that an atomic operation will happen following some specified ordering with respect to other memory accesses. For example, if we need that an atomic operation happen just after all other pending writes (in terms of instructions reordering buffers activities) are completed, we need to explicitly use a memory barrier in conjunction to this atomic operation. So it is simple to understand why memory barriers play a key role for higher-level locks building (just as refcounts, mutexes, etc.). For a detailed explanatory on atomic operations, please refer to man:atomic[9]. It is far, however, noting that atomic operations (and memory barriers as well) should ideally only be used for building front-ending locks (as mutexes). [[freebsd-refcounts]] ===== Refcounts Refcounts are interfaces for handling reference counters. They are implemented through atomic operations and are intended to be used just for cases, where the reference counter is the only one thing to be protected, so even something like a spin-mutex is deprecated. Using the refcount interface for structures, where a mutex is already used is often wrong since we should probably close the reference counter in some already protected paths. A manpage discussing refcount does not exist currently, just check [.filename]#sys/refcount.h# for an overview of the existing API. [[freebsd-locks]] ===== Locks FreeBSD kernel has huge classes of locks. Every lock is defined by some peculiar properties, but probably the most important is the event linked to contesting holders (or in other terms, the behavior of threads unable to acquire the lock). FreeBSD's locking scheme presents three different behaviors for contenders: . spinning . blocking . sleeping [NOTE] ==== numbers are not casual ==== [[freebsd-spinlocks]] ===== Spinning locks Spin locks let waiters to spin until they cannot acquire the lock. An important matter do deal with is when a thread contests on a spin lock if it is not descheduled. Since the FreeBSD kernel is preemptive, this exposes spin lock at the risk of deadlocks that can be solved just disabling interrupts while they are acquired. For this and other reasons (like lack of priority propagation support, poorness in load balancing schemes between CPUs, etc.), spin locks are intended to protect very small paths of code, or ideally not to be used at all if not explicitly requested (explained later). [[freebsd-blocking]] ===== Blocking Block locks let waiters to be descheduled and blocked until the lock owner does not drop it and wakes up one or more contenders. In order to avoid starvation issues, blocking locks do priority propagation from the waiters to the owner. Block locks must be implemented through the turnstile interface and are intended to be the most used kind of locks in the kernel, if no particular conditions are met. [[freebsd-sleeping]] ===== Sleeping Sleep locks let waiters to be descheduled and fall asleep until the lock holder does not drop it and wakes up one or more waiters. Since sleep locks are intended to protect large paths of code and to cater asynchronous events, they do not do any form of priority propagation. They must be implemented through the man:sleepqueue[9] interface. The order used to acquire locks is very important, not only for the possibility to deadlock due at lock order reversals, but even because lock acquisition should follow specific rules linked to locks natures. If you give a look at the table above, the practical rule is that if a thread holds a lock of level n (where the level is the number listed close to the kind of lock) it is not allowed to acquire a lock of superior levels, since this would break the specified semantic for a path. For example, if a thread holds a block lock (level 2), it is allowed to acquire a spin lock (level 1) but not a sleep lock (level 3), since block locks are intended to protect smaller paths than sleep lock (these rules are not about atomic operations or scheduling barriers, however). This is a list of lock with their respective behaviors: * spin mutex - spinning - man:mutex[9] * sleep mutex - blocking - man:mutex[9] * pool mutex - blocking - man:mtx[pool] * sleep family - sleeping - man:sleep[9] pause tsleep msleep msleep spin msleep rw msleep sx * condvar - sleeping - man:condvar[9] * rwlock - blocking - man:rwlock[9] * sxlock - sleeping - man:sx[9] * lockmgr - sleeping - man:lockmgr[9] * semaphores - sleeping - man:sema[9] Among these locks only mutexes, sxlocks, rwlocks and lockmgrs are intended to handle recursion, but currently recursion is only supported by mutexes and lockmgrs. [[freebsd-scheduling]] ===== Scheduling barriers Scheduling barriers are intended to be used in order to drive scheduling of threading. They consist mainly of three different stubs: * critical sections (and preemption) * sched_bind * sched_pin Generally, these should be used only in a particular context and even if they can often replace locks, they should be avoided because they do not let the diagnose of simple eventual problems with locking debugging tools (as man:witness[4]). [[freebsd-critical]] ===== Critical sections The FreeBSD kernel has been made preemptive basically to deal with interrupt threads. In fact, in order to avoid high interrupt latency, time-sharing priority threads can be preempted by interrupt threads (in this way, they do not need to wait to be scheduled as the normal path previews). Preemption, however, introduces new racing points that need to be handled, as well. Often, in order to deal with preemption, the simplest thing to do is to completely disable it. A critical section defines a piece of code (borderlined by the pair of functions man:critical_enter[9] and man:critical_exit[9], where preemption is guaranteed to not happen (until the protected code is fully executed). This can often replace a lock effectively but should be used carefully in order to not lose the whole advantage that preemption brings. [[freebsd-schedpin]] ===== sched_pin/sched_unpin Another way to deal with preemption is the `sched_pin()` interface. If a piece of code is closed in the `sched_pin()` and `sched_unpin()` pair of functions it is guaranteed that the respective thread, even if it can be preempted, it will always be executed on the same CPU. Pinning is very effective in the particular case when we have to access at per-cpu datas and we assume other threads will not change those data. The latter condition will determine a critical section as a too strong condition for our code. [[freebsd-schedbind]] ===== sched_bind/sched_unbind `sched_bind` is an API used in order to bind a thread to a particular CPU for all the time it executes the code, until a `sched_unbind` function call does not unbind it. This feature has a key role in situations where you cannot trust the current state of CPUs (for example, at very early stages of boot), as you want to avoid your thread to migrate on inactive CPUs. Since `sched_bind` and `sched_unbind` manipulate internal scheduler structures, they need to be enclosed in `sched_lock` acquisition/releasing when used. [[freebsd-proc]] ==== Proc structure Various emulation layers sometimes require some additional per-process data. It can manage separate structures (a list, a tree etc.) containing these data for every process but this tends to be slow and memory consuming. To solve this problem the FreeBSD `proc` structure contains `p_emuldata`, which is a void pointer to some emulation layer specific data. This `proc` entry is protected by the proc mutex. The FreeBSD `proc` structure contains a `p_sysent` entry that identifies, which ABI this process is running. In fact, it is a pointer to the `sysentvec` described above. So by comparing this pointer to the address where the `sysentvec` structure for the given ABI is stored we can effectively determine whether the process belongs to our emulation layer. The code typically looks like: [.programlisting] .... if (__predict_true(p->p_sysent != &elf_Linux(R)_sysvec)) return; .... As you can see, we effectively use the `__predict_true` modifier to collapse the most common case (FreeBSD process) to a simple return operation thus preserving high performance. This code should be turned into a macro because currently it is not very flexible, i.e. we do not support Linux(R)64 emulation nor A.OUT Linux(R) processes on i386. [[freebsd-vfs]] ==== VFS The FreeBSD VFS subsystem is very complex but the Linux(R) emulation layer uses just a small subset via a well defined API. It can either operate on vnodes or file handlers. Vnode represents a virtual vnode, i.e. representation of a node in VFS. Another representation is a file handler, which represents an opened file from the perspective of a process. A file handler can represent a socket or an ordinary file. A file handler contains a pointer to its vnode. More then one file handler can point to the same vnode. [[freebsd-namei]] ===== namei The man:namei[9] routine is a central entry point to pathname lookup and translation. It traverses the path point by point from the starting point to the end point using lookup function, which is internal to VFS. The man:namei[9] syscall can cope with symlinks, absolute and relative paths. When a path is looked up using man:namei[9] it is inputed to the name cache. This behavior can be suppressed. This routine is used all over the kernel and its performance is very critical. [[freebsd-vn]] ===== vn_fullpath The man:vn_fullpath[9] function takes the best effort to traverse VFS name cache and returns a path for a given (locked) vnode. This process is unreliable but works just fine for the most common cases. The unreliability is because it relies on VFS cache (it does not traverse the on medium structures), it does not work with hardlinks, etc. This routine is used in several places in the Linuxulator. [[freebsd-vnode]] ===== Vnode operations * `fgetvp` - given a thread and a file descriptor number it returns the associated vnode * man:vn_lock[9] - locks a vnode * `vn_unlock` - unlocks a vnode * man:VOP_READDIR[9] - reads a directory referenced by a vnode * man:VOP_GETATTR[9] - gets attributes of a file or a directory referenced by a vnode * man:VOP_LOOKUP[9] - looks up a path to a given directory * man:VOP_OPEN[9] - opens a file referenced by a vnode * man:VOP_CLOSE[9] - closes a file referenced by a vnode * man:vput[9] - decrements the use count for a vnode and unlocks it * man:vrele[9] - decrements the use count for a vnode * man:vref[9] - increments the use count for a vnode [[freebsd-file-handler]] ===== File handler operations * `fget` - given a thread and a file descriptor number it returns associated file handler and references it * `fdrop` - drops a reference to a file handler * `fhold` - references a file handler [[md]] == Linux(R) emulation layer -MD part This section deals with implementation of Linux(R) emulation layer in FreeBSD operating system. It first describes the machine dependent part talking about how and where interaction between userland and kernel is implemented. It talks about syscalls, signals, ptrace, traps, stack fixup. This part discusses i386 but it is written generally so other architectures should not differ very much. The next part is the machine independent part of the Linuxulator. This section only covers i386 and ELF handling. A.OUT is obsolete and untested. [[syscall-handling]] === Syscall handling Syscall handling is mostly written in [.filename]#linux_sysvec.c#, which covers most of the routines pointed out in the `sysentvec` structure. When a Linux(R) process running on FreeBSD issues a syscall, the general syscall routine calls linux prepsyscall routine for the Linux(R) ABI. [[linux-prepsyscall]] ==== Linux(R) prepsyscall Linux(R) passes arguments to syscalls via registers (that is why it is limited to 6 parameters on i386) while FreeBSD uses the stack. The Linux(R) prepsyscall routine must copy parameters from registers to the stack. The order of the registers is: `%ebx`, `%ecx`, `%edx`, `%esi`, `%edi`, `%ebp`. The catch is that this is true for only _most_ of the syscalls. Some (most notably `clone`) uses a different order but it is luckily easy to fix by inserting a dummy parameter in the `linux_clone` prototype. [[syscall-writing]] ==== Syscall writing Every syscall implemented in the Linuxulator must have its prototype with various flags in [.filename]#syscalls.master#. The form of the file is: [.programlisting] .... ... AUE_FORK STD { int linux_fork(void); } ... AUE_CLOSE NOPROTO { int close(int fd); } ... .... The first column represents the syscall number. The second column is for auditing support. The third column represents the syscall type. It is either `STD`, `OBSOL`, `NOPROTO` and `UNIMPL`. `STD` is a standard syscall with full prototype and implementation. `OBSOL` is obsolete and defines just the prototype. `NOPROTO` means that the syscall is implemented elsewhere so do not prepend ABI prefix, etc. `UNIMPL` means that the syscall will be substituted with the `nosys` syscall (a syscall just printing out a message about the syscall not being implemented and returning `ENOSYS`). From [.filename]#syscalls.master# a script generates three files: [.filename]#linux_syscall.h#, [.filename]#linux_proto.h# and [.filename]#linux_sysent.c#. The [.filename]#linux_syscall.h# contains definitions of syscall names and their numerical value, e.g.: [.programlisting] .... ... #define LINUX_SYS_linux_fork 2 ... #define LINUX_SYS_close 6 ... .... The [.filename]#linux_proto.h# contains structure definitions of arguments to every syscall, e.g.: [.programlisting] .... struct linux_fork_args { register_t dummy; }; .... And finally, [.filename]#linux_sysent.c# contains structure describing the system entry table, used to actually dispatch a syscall, e.g.: [.programlisting] .... { 0, (sy_call_t *)linux_fork, AUE_FORK, NULL, 0, 0 }, /* 2 = linux_fork */ { AS(close_args), (sy_call_t *)close, AUE_CLOSE, NULL, 0, 0 }, /* 6 = close */ .... As you can see `linux_fork` is implemented in Linuxulator itself so the definition is of `STD` type and has no argument, which is exhibited by the dummy argument structure. On the other hand `close` is just an alias for real FreeBSD man:close[2] so it has no linux arguments structure associated and in the system entry table it is not prefixed with linux as it calls the real man:close[2] in the kernel. [[dummy-syscalls]] ==== Dummy syscalls The Linux(R) emulation layer is not complete, as some syscalls are not implemented properly and some are not implemented at all. The emulation layer employs a facility to mark unimplemented syscalls with the `DUMMY` macro. These dummy definitions reside in [.filename]#linux_dummy.c# in a form of `DUMMY(syscall);`, which is then translated to various syscall auxiliary files and the implementation consists of printing a message saying that this syscall is not implemented. The `UNIMPL` prototype is not used because we want to be able to identify the name of the syscall that was called in order to know what syscalls are more important to implement. [[signal-handling]] === Signal handling Signal handling is done generally in the FreeBSD kernel for all binary compatibilities with a call to a compat-dependent layer. Linux(R) compatibility layer defines `linux_sendsig` routine for this purpose. [[linux-sendsig]] ==== Linux(R) sendsig This routine first checks whether the signal has been installed with a `SA_SIGINFO` in which case it calls `linux_rt_sendsig` routine instead. Furthermore, it allocates (or reuses an already existing) signal handle context, then it builds a list of arguments for the signal handler. It translates the signal number based on the signal translation table, assigns a handler, translates sigset. Then it saves context for the `sigreturn` routine (various registers, translated trap number and signal mask). Finally, it copies out the signal context to the userspace and prepares context for the actual signal handler to run. [[linux-rt-sendsig]] ==== linux_rt_sendsig This routine is similar to `linux_sendsig` just the signal context preparation is different. It adds `siginfo`, `ucontext`, and some POSIX(R) parts. It might be worth considering whether those two functions could not be merged with a benefit of less code duplication and possibly even faster execution. [[linux-sigreturn]] ==== linux_sigreturn This syscall is used for return from the signal handler. It does some security checks and restores the original process context. It also unmasks the signal in process signal mask. [[ptrace]] === Ptrace Many UNIX(R) derivates implement the man:ptrace[2] syscall in order to allow various tracking and debugging features. This facility enables the tracing process to obtain various information about the traced process, like register dumps, any memory from the process address space, etc. and also to trace the process like in stepping an instruction or between system entries (syscalls and traps). man:ptrace[2] also lets you set various information in the traced process (registers etc.). man:ptrace[2] is a UNIX(R)-wide standard implemented in most UNIX(R)es around the world. Linux(R) emulation in FreeBSD implements the man:ptrace[2] facility in [.filename]#linux_ptrace.c#. The routines for converting registers between Linux(R) and FreeBSD and the actual man:ptrace[2] syscall emulation syscall. The syscall is a long switch block that implements its counterpart in FreeBSD for every man:ptrace[2] command. The man:ptrace[2] commands are mostly equal between Linux(R) and FreeBSD so usually just a small modification is needed. For example, `PT_GETREGS` in Linux(R) operates on direct data while FreeBSD uses a pointer to the data so after performing a (native) man:ptrace[2] syscall, a copyout must be done to preserve Linux(R) semantics. The man:ptrace[2] implementation in Linuxulator has some known weaknesses. There have been panics seen when using `strace` (which is a man:ptrace[2] consumer) in the Linuxulator environment. Also `PT_SYSCALL` is not implemented. [[traps]] === Traps Whenever a Linux(R) process running in the emulation layer traps the trap itself is handled transparently with the only exception of the trap translation. Linux(R) and FreeBSD differs in opinion on what a trap is so this is dealt with here. The code is actually very short: [.programlisting] .... static int translate_traps(int signal, int trap_code) { if (signal != SIGBUS) return signal; switch (trap_code) { case T_PROTFLT: case T_TSSFLT: case T_DOUBLEFLT: case T_PAGEFLT: return SIGSEGV; default: return signal; } } .... [[stack-fixup]] === Stack fixup The RTLD run-time link-editor expects so called AUX tags on stack during an `execve` so a fixup must be done to ensure this. Of course, every RTLD system is different so the emulation layer must provide its own stack fixup routine to do this. So does Linuxulator. The `elf_linux_fixup` simply copies out AUX tags to the stack and adjusts the stack of the user space process to point right after those tags. So RTLD works in a smart way. [[aout-support]] === A.OUT support The Linux(R) emulation layer on i386 also supports Linux(R) A.OUT binaries. Pretty much everything described in the previous sections must be implemented for A.OUT support (beside traps translation and signals sending). The support for A.OUT binaries is no longer maintained, especially the 2.6 emulation does not work with it but this does not cause any problem, as the linux-base in ports probably do not support A.OUT binaries at all. This support will probably be removed in future. Most of the stuff necessary for loading Linux(R) A.OUT binaries is in [.filename]#imgact_linux.c# file. [[mi]] == Linux(R) emulation layer -MI part This section talks about machine independent part of the Linuxulator. It covers the emulation infrastructure needed for Linux(R) 2.6 emulation, the thread local storage (TLS) implementation (on i386) and futexes. Then we talk briefly about some syscalls. [[nptl-desc]] === Description of NPTL One of the major areas of progress in development of Linux(R) 2.6 was threading. Prior to 2.6, the Linux(R) threading support was implemented in the linuxthreads library. The library was a partial implementation of POSIX(R) threading. The threading was implemented using separate processes for each thread using the `clone` syscall to let them share the address space (and other things). The main weaknesses of this approach was that every thread had a different PID, signal handling was broken (from the pthreads perspective), etc. Also the performance was not very good (use of `SIGUSR` signals for threads synchronization, kernel resource consumption, etc.) so to overcome these problems a new threading system was developed and named NPTL. The NPTL library focused on two things but a third thing came along so it is usually considered a part of NPTL. Those two things were embedding of threads into a process structure and futexes. The additional third thing was TLS, which is not directly required by NPTL but the whole NPTL userland library depends on it. Those improvements yielded in much improved performance and standards conformance. NPTL is a standard threading library in Linux(R) systems these days. The FreeBSD Linuxulator implementation approaches the NPTL in three main areas. The TLS, futexes and PID mangling, which is meant to simulate the Linux(R) threads. Further sections describe each of these areas. [[linux26-emu]] === Linux(R) 2.6 emulation infrastructure These sections deal with the way Linux(R) threads are managed and how we simulate that in FreeBSD. [[linux26-runtime]] ==== Runtime determining of 2.6 emulation The Linux(R) emulation layer in FreeBSD supports runtime setting of the emulated version. This is done via man:sysctl[8], namely `compat.linux.osrelease`. Setting this man:sysctl[8] affects runtime behavior of the emulation layer. When set to 2.6.x it sets the value of `linux_use_linux26` while setting to something else keeps it unset. This variable (plus per-prison variables of the very same kind) determines whether 2.6 infrastructure (mainly PID mangling) is used in the code or not. The version setting is done system-wide and this affects all Linux(R) processes. The man:sysctl[8] should not be changed when running any Linux(R) binary as it might harm things. [[linux-proc-thread]] ==== Linux(R) processes and thread identifiers The semantics of Linux(R) threading are a little confusing and uses entirely different nomenclature to FreeBSD. A process in Linux(R) consists of a `struct task` embedding two identifier fields - PID and TGID. PID is _not_ a process ID but it is a thread ID. The TGID identifies a thread group in other words a process. For single-threaded process the PID equals the TGID. The thread in NPTL is just an ordinary process that happens to have TGID not equal to PID and have a group leader not equal to itself (and shared VM etc. of course). Everything else happens in the same way as to an ordinary process. There is no separation of a shared status to some external structure like in FreeBSD. This creates some duplication of information and possible data inconsistency. The Linux(R) kernel seems to use task -> group information in some places and task information elsewhere and it is really not very consistent and looks error-prone. Every NPTL thread is created by a call to the `clone` syscall with a specific set of flags (more in the next subsection). The NPTL implements strict 1:1 threading. In FreeBSD we emulate NPTL threads with ordinary FreeBSD processes that share VM space, etc. and the PID gymnastic is just mimicked in the emulation specific structure attached to the process. The structure attached to the process looks like: [.programlisting] .... struct linux_emuldata { pid_t pid; int *child_set_tid; /* in clone(): Child.s TID to set on clone */ int *child_clear_tid;/* in clone(): Child.s TID to clear on exit */ struct linux_emuldata_shared *shared; int pdeath_signal; /* parent death signal */ LIST_ENTRY(linux_emuldata) threads; /* list of linux threads */ }; .... The PID is used to identify the FreeBSD process that attaches this structure. The `child_se_tid` and `child_clear_tid` are used for TID address copyout when a process exits and is created. The `shared` pointer points to a structure shared among threads. The `pdeath_signal` variable identifies the parent death signal and the `threads` pointer is used to link this structure to the list of threads. The `linux_emuldata_shared` structure looks like: [.programlisting] .... struct linux_emuldata_shared { int refs; pid_t group_pid; LIST_HEAD(, linux_emuldata) threads; /* head of list of linux threads */ }; .... The `refs` is a reference counter being used to determine when we can free the structure to avoid memory leaks. The `group_pid` is to identify PID ( = TGID) of the whole process ( = thread group). The `threads` pointer is the head of the list of threads in the process. The `linux_emuldata` structure can be obtained from the process using `em_find`. The prototype of the function is: [.programlisting] .... struct linux_emuldata *em_find(struct proc *, int locked); .... Here, `proc` is the process we want the emuldata structure from and the locked parameter determines whether we want to lock or not. The accepted values are `EMUL_DOLOCK` and `EMUL_DOUNLOCK`. More about locking later. [[pid-mangling]] ==== PID mangling As there is a difference in view as what to the idea of a process ID and thread ID is between FreeBSD and Linux(R) we have to translate the view somehow. We do it by PID mangling. This means that we fake what a PID (=TGID) and TID (=PID) is between kernel and userland. The rule of thumb is that in kernel (in Linuxulator) PID = PID and TGID = shared -> group pid and to userland we present `PID = shared -> group_pid` and `TID = proc -> p_pid`. The PID member of `linux_emuldata structure` is a FreeBSD PID. The above affects mainly getpid, getppid, gettid syscalls. Where we use PID/TGID respectively. In copyout of TIDs in `child_clear_tid` and `child_set_tid` we copy out FreeBSD PID. [[clone-syscall]] ==== Clone syscall The `clone` syscall is the way threads are created in Linux(R). The syscall prototype looks like this: [.programlisting] .... int linux_clone(l_int flags, void *stack, void *parent_tidptr, int dummy, void * child_tidptr); .... The `flags` parameter tells the syscall how exactly the processes should be cloned. As described above, Linux(R) can create processes sharing various things independently, for example two processes can share file descriptors but not VM, etc. Last byte of the `flags` parameter is the exit signal of the newly created process. The `stack` parameter if non-`NULL` tells, where the thread stack is and if it is `NULL` we are supposed to copy-on-write the calling process stack (i.e. do what normal man:fork[2] routine does). The `parent_tidptr` parameter is used as an address for copying out process PID (i.e. thread id) once the process is sufficiently instantiated but is not runnable yet. The `dummy` parameter is here because of the very strange calling convention of this syscall on i386. It uses the registers directly and does not let the compiler do it what results in the need of a dummy syscall. The `child_tidptr` parameter is used as an address for copying out PID once the process has finished forking and when the process exits. The syscall itself proceeds by setting corresponding flags depending on the flags passed in. For example, `CLONE_VM` maps to RFMEM (sharing of VM), etc. The only nit here is `CLONE_FS` and `CLONE_FILES` because FreeBSD does not allow setting this separately so we fake it by not setting RFFDG (copying of fd table and other fs information) if either of these is defined. This does not cause any problems, because those flags are always set together. After setting the flags the process is forked using the internal `fork1` routine, the process is instrumented not to be put on a run queue, i.e. not to be set runnable. After the forking is done we possibly reparent the newly created process to emulate `CLONE_PARENT` semantics. Next part is creating the emulation data. Threads in Linux(R) does not signal their parents so we set exit signal to be 0 to disable this. After that setting of `child_set_tid` and `child_clear_tid` is performed enabling the functionality later in the code. At this point we copy out the PID to the address specified by `parent_tidptr`. The setting of process stack is done by simply rewriting thread frame `%esp` register (`%rsp` on amd64). Next part is setting up TLS for the newly created process. After this man:vfork[2] semantics might be emulated and finally the newly created process is put on a run queue and copying out its PID to the parent process via `clone` return value is done. The `clone` syscall is able and in fact is used for emulating classic man:fork[2] and man:vfork[2] syscalls. Newer glibc in a case of 2.6 kernel uses `clone` to implement man:fork[2] and man:vfork[2] syscalls. [[locking]] ==== Locking The locking is implemented to be per-subsystem because we do not expect a lot of contention on these. There are two locks: `emul_lock` used to protect manipulating of `linux_emuldata` and `emul_shared_lock` used to manipulate `linux_emuldata_shared`. The `emul_lock` is a nonsleepable blocking mutex while `emul_shared_lock` is a sleepable blocking `sx_lock`. Due to of the per-subsystem locking we can coalesce some locks and that is why the em find offers the non-locking access. [[tls]] === TLS This section deals with TLS also known as thread local storage. [[trheading-intro]] ==== Introduction to threading Threads in computer science are entities within a process that can be scheduled independently from each other. The threads in the process share process wide data (file descriptors, etc.) but also have their own stack for their own data. Sometimes there is a need for process-wide data specific to a given thread. Imagine a name of the thread in execution or something like that. The traditional UNIX(R) threading API, pthreads provides a way to do it via man:pthread_key_create[3], man:pthread_setspecific[3] and man:pthread_getspecific[3] where a thread can create a key to the thread local data and using man:pthread_getspecific[3] or man:pthread_getspecific[3] to manipulate those data. You can easily see that this is not the most comfortable way this could be accomplished. So various producers of C/C++ compilers introduced a better way. They defined a new modifier keyword thread that specifies that a variable is thread specific. A new method of accessing such variables was developed as well (at least on i386). The pthreads method tends to be implemented in userspace as a trivial lookup table. The performance of such a solution is not very good. So the new method uses (on i386) segment registers to address a segment, where TLS area is stored so the actual accessing of a thread variable is just appending the segment register to the address thus addressing via it. The segment registers are usually `%gs` and `%fs` acting like segment selectors. Every thread has its own area where the thread local data are stored and the segment must be loaded on every context switch. This method is very fast and used almost exclusively in the whole i386 UNIX(R) world. Both FreeBSD and Linux(R) implement this approach and it yields very good results. The only drawback is the need to reload the segment on every context switch which can slowdown context switches. FreeBSD tries to avoid this overhead by using only 1 segment descriptor for this while Linux(R) uses 3. Interesting thing is that almost nothing uses more than 1 descriptor (only Wine seems to use 2) so Linux(R) pays this unnecessary price for context switches. [[i386-segs]] ==== Segments on i386 The i386 architecture implements the so called segments. A segment is a description of an area of memory. The base address (bottom) of the memory area, the end of it (ceiling), type, protection, etc. The memory described by a segment can be accessed using segment selector registers (`%cs`, `%ds`, `%ss`, `%es`, `%fs`, `%gs`). For example let us suppose we have a segment which base address is 0x1234 and length and this code: [.programlisting] .... mov %edx,%gs:0x10 .... This will load the content of the `%edx` register into memory location 0x1244. Some segment registers have a special use, for example `%cs` is used for code segment and `%ss` is used for stack segment but `%fs` and `%gs` are generally unused. Segments are either stored in a global GDT table or in a local LDT table. LDT is accessed via an entry in the GDT. The LDT can store more types of segments. LDT can be per process. Both tables define up to 8191 entries. [[linux-i386]] ==== Implementation on Linux(R) i386 There are two main ways of setting up TLS in Linux(R). It can be set when cloning a process using the `clone` syscall or it can call `set_thread_area`. When a process passes `CLONE_SETTLS` flag to `clone`, the kernel expects the memory pointed to by the `%esi` register a Linux(R) user space representation of a segment, which gets translated to the machine representation of a segment and loaded into a GDT slot. The GDT slot can be specified with a number or -1 can be used meaning that the system itself should choose the first free slot. In practice, the vast majority of programs use only one TLS entry and does not care about the number of the entry. We exploit this in the emulation and in fact depend on it. [[tls-emu]] ==== Emulation of Linux(R) TLS [[tls-i386]] ===== i386 Loading of TLS for the current thread happens by calling `set_thread_area` while loading TLS for a second process in `clone` is done in the separate block in `clone`. Those two functions are very similar. The only difference being the actual loading of the GDT segment, which happens on the next context switch for the newly created process while `set_thread_area` must load this directly. The code basically does this. It copies the Linux(R) form segment descriptor from the userland. The code checks for the number of the descriptor but because this differs between FreeBSD and Linux(R) we fake it a little. We only support indexes of 6, 3 and -1. The 6 is genuine Linux(R) number, 3 is genuine FreeBSD one and -1 means autoselection. Then we set the descriptor number to constant 3 and copy out this to the userspace. We rely on the userspace process using the number from the descriptor but this works most of the time (have never seen a case where this did not work) as the userspace process typically passes in 1. Then we convert the descriptor from the Linux(R) form to a machine dependant form (i.e. operating system independent form) and copy this to the FreeBSD defined segment descriptor. Finally we can load it. We assign the descriptor to threads PCB (process control block) and load the `%gs` segment using `load_gs`. This loading must be done in a critical section so that nothing can interrupt us. The `CLONE_SETTLS` case works exactly like this just the loading using `load_gs` is not performed. The segment used for this (segment number 3) is shared for this use between FreeBSD processes and Linux(R) processes so the Linux(R) emulation layer does not add any overhead over plain FreeBSD. [[tls-amd64]] ===== amd64 The amd64 implementation is similar to the i386 one but there was initially no 32bit segment descriptor used for this purpose (hence not even native 32bit TLS users worked) so we had to add such a segment and implement its loading on every context switch (when a flag signaling use of 32bit is set). Apart from this the TLS loading is exactly the same just the segment numbers are different and the descriptor format and the loading differs slightly. [[futexes]] === Futexes [[sync-intro]] ==== Introduction to synchronization Threads need some kind of synchronization and POSIX(R) provides some of them: mutexes for mutual exclusion, read-write locks for mutual exclusion with biased ratio of reads and writes and condition variables for signaling a status change. It is interesting to note that POSIX(R) threading API lacks support for semaphores. Those synchronization routines implementations are heavily dependant on the type threading support we have. In pure 1:M (userspace) model the implementation can be solely done in userspace and thus be very fast (the condition variables will probably end up being implemented using signals, i.e. not fast) and simple. In 1:1 model, the situation is also quite clear - the threads must be synchronized using kernel facilities (which is very slow because a syscall must be performed). The mixed M:N scenario just combines the first and second approach or rely solely on kernel. Threads synchronization is a vital part of thread-enabled programming and its performance can affect resulting program a lot. Recent benchmarks on FreeBSD operating system showed that an improved sx_lock implementation yielded 40% speedup in _ZFS_ (a heavy sx user), this is in-kernel stuff but it shows clearly how important the performance of synchronization primitives is. Threaded programs should be written with as little contention on locks as possible. Otherwise, instead of doing useful work the thread just waits on a lock. As a result of this, the most well written threaded programs show little locks contention. [[futex-intro]] ==== Futexes introduction Linux(R) implements 1:1 threading, i.e. it has to use in-kernel synchronization primitives. As stated earlier, well written threaded programs have little lock contention. So a typical sequence could be performed as two atomic increase/decrease mutex reference counter, which is very fast, as presented by the following example: [.programlisting] .... pthread_mutex_lock(&mutex); ... pthread_mutex_unlock(&mutex); .... 1:1 threading forces us to perform two syscalls for those mutex calls, which is very slow. The solution Linux(R) 2.6 implements is called futexes. Futexes implement the check for contention in userspace and call kernel primitives only in a case of contention. Thus the typical case takes place without any kernel intervention. This yields reasonably fast and flexible synchronization primitives implementation. [[futex-api]] ==== Futex API The futex syscall looks like this: [.programlisting] .... int futex(void *uaddr, int op, int val, struct timespec *timeout, void *uaddr2, int val3); .... In this example `uaddr` is an address of the mutex in userspace, `op` is an operation we are about to perform and the other parameters have per-operation meaning. Futexes implement the following operations: * `FUTEX_WAIT` * `FUTEX_WAKE` * `FUTEX_FD` * `FUTEX_REQUEUE` * `FUTEX_CMP_REQUEUE` * `FUTEX_WAKE_OP` [[futex-wait]] ===== FUTEX_WAIT This operation verifies that on address `uaddr` the value `val` is written. If not, `EWOULDBLOCK` is returned, otherwise the thread is queued on the futex and gets suspended. If the argument `timeout` is non-zero it specifies the maximum time for the sleeping, otherwise the sleeping is infinite. [[futex-wake]] ===== FUTEX_WAKE This operation takes a futex at `uaddr` and wakes up `val` first futexes queued on this futex. [[futex-fd]] ===== FUTEX_FD This operations associates a file descriptor with a given futex. [[futex-requeue]] ===== FUTEX_REQUEUE This operation takes `val` threads queued on futex at `uaddr`, wakes them up, and takes `val2` next threads and requeues them on futex at `uaddr2`. [[futex-cmp-requeue]] ===== FUTEX_CMP_REQUEUE This operation does the same as `FUTEX_REQUEUE` but it checks that `val3` equals to `val` first. [[futex-wake-op]] ===== FUTEX_WAKE_OP This operation performs an atomic operation on `val3` (which contains coded some other value) and `uaddr`. Then it wakes up `val` threads on futex at `uaddr` and if the atomic operation returned a positive number it wakes up `val2` threads on futex at `uaddr2`. The operations implemented in `FUTEX_WAKE_OP`: * `FUTEX_OP_SET` * `FUTEX_OP_ADD` * `FUTEX_OP_OR` * `FUTEX_OP_AND` * `FUTEX_OP_XOR` [NOTE] ==== There is no `val2` parameter in the futex prototype. The `val2` is taken from the `struct timespec *timeout` parameter for operations `FUTEX_REQUEUE`, `FUTEX_CMP_REQUEUE` and `FUTEX_WAKE_OP`. ==== [[futex-emu]] ==== Futex emulation in FreeBSD The futex emulation in FreeBSD is taken from NetBSD and further extended by us. It is placed in `linux_futex.c` and [.filename]#linux_futex.h# files. The `futex` structure looks like: [.programlisting] .... struct futex { void *f_uaddr; int f_refcount; LIST_ENTRY(futex) f_list; TAILQ_HEAD(lf_waiting_paroc, waiting_proc) f_waiting_proc; }; .... And the structure `waiting_proc` is: [.programlisting] .... struct waiting_proc { struct thread *wp_t; struct futex *wp_new_futex; TAILQ_ENTRY(waiting_proc) wp_list; }; .... [[futex-get]] ===== futex_get / futex_put A futex is obtained using the `futex_get` function, which searches a linear list of futexes and returns the found one or creates a new futex. When releasing a futex from the use we call the `futex_put` function, which decreases a reference counter of the futex and if the refcount reaches zero it is released. [[futex-sleep]] ===== futex_sleep When a futex queues a thread for sleeping it creates a `working_proc` structure and puts this structure to the list inside the futex structure then it just performs a man:tsleep[9] to suspend the thread. The sleep can be timed out. After man:tsleep[9] returns (the thread was woken up or it timed out) the `working_proc` structure is removed from the list and is destroyed. All this is done in the `futex_sleep` function. If we got woken up from `futex_wake` we have `wp_new_futex` set so we sleep on it. This way the actual requeueing is done in this function. [[futex-wake-2]] ===== futex_wake Waking up a thread sleeping on a futex is performed in the `futex_wake` function. First in this function we mimic the strange Linux(R) behavior, where it wakes up N threads for all operations, the only exception is that the REQUEUE operations are performed on N+1 threads. But this usually does not make any difference as we are waking up all threads. Next in the function in the loop we wake up n threads, after this we check if there is a new futex for requeueing. If so, we requeue up to n2 threads on the new futex. This cooperates with `futex_sleep`. [[futex-wake-op-2]] ===== futex_wake_op The `FUTEX_WAKE_OP` operation is quite complicated. First we obtain two futexes at addresses `uaddr` and `uaddr2` then we perform the atomic operation using `val3` and `uaddr2`. Then `val` waiters on the first futex is woken up and if the atomic operation condition holds we wake up `val2` (i.e. `timeout`) waiter on the second futex. [[futex-atomic-op]] ===== futex atomic operation The atomic operation takes two parameters `encoded_op` and `uaddr`. The encoded operation encodes the operation itself, comparing value, operation argument, and comparing argument. The pseudocode for the operation is like this one: [.programlisting] .... oldval = *uaddr2 *uaddr2 = oldval OP oparg .... And this is done atomically. First a copying in of the number at `uaddr` is performed and the operation is done. The code handles page faults and if no page fault occurs `oldval` is compared to `cmparg` argument with cmp comparator. [[futex-locking]] ===== Futex locking Futex implementation uses two lock lists protecting `sx_lock` and global locks (either Giant or another `sx_lock`). Every operation is performed locked from the start to the very end. [[syscall-impl]] === Various syscalls implementation In this section I am going to describe some smaller syscalls that are worth mentioning because their implementation is not obvious or those syscalls are interesting from other point of view. [[syscall-at]] ==== *at family of syscalls During development of Linux(R) 2.6.16 kernel, the *at syscalls were added. Those syscalls (`openat` for example) work exactly like their at-less counterparts with the slight exception of the `dirfd` parameter. This parameter changes where the given file, on which the syscall is to be performed, is. When the `filename` parameter is absolute `dirfd` is ignored but when the path to the file is relative, it comes to the play. The `dirfd` parameter is a directory relative to which the relative pathname is checked. The `dirfd` parameter is a file descriptor of some directory or `AT_FDCWD`. So for example the `openat` syscall can be like this: [.programlisting] .... file descriptor 123 = /tmp/foo/, current working directory = /tmp/ openat(123, /tmp/bah\, flags, mode) /* opens /tmp/bah */ openat(123, bah\, flags, mode) /* opens /tmp/foo/bah */ openat(AT_FDWCWD, bah\, flags, mode) /* opens /tmp/bah */ openat(stdio, bah\, flags, mode) /* returns error because stdio is not a directory */ .... This infrastructure is necessary to avoid races when opening files outside the working directory. Imagine that a process consists of two threads, thread A and thread B. Thread A issues `open(./tmp/foo/bah., flags, mode)` and before returning it gets preempted and thread B runs. Thread B does not care about the needs of thread A and renames or removes [.filename]#/tmp/foo/#. We got a race. To avoid this we can open [.filename]#/tmp/foo# and use it as `dirfd` for `openat` syscall. This also enables user to implement per-thread working directories. Linux(R) family of *at syscalls contains: `linux_openat`, `linux_mkdirat`, `linux_mknodat`, `linux_fchownat`, `linux_futimesat`, `linux_fstatat64`, `linux_unlinkat`, `linux_renameat`, `linux_linkat`, `linux_symlinkat`, `linux_readlinkat`, `linux_fchmodat` and `linux_faccessat`. All these are implemented using the modified man:namei[9] routine and simple wrapping layer. [[implementation]] ===== Implementation The implementation is done by altering the man:namei[9] routine (described above) to take additional parameter `dirfd` in its `nameidata` structure, which specifies the starting point of the pathname lookup instead of using the current working directory every time. The resolution of `dirfd` from file descriptor number to a vnode is done in native *at syscalls. When `dirfd` is `AT_FDCWD` the `dvp` entry in `nameidata` structure is `NULL` but when `dirfd` is a different number we obtain a file for this file descriptor, check whether this file is valid and if there is vnode attached to it then we get a vnode. Then we check this vnode for being a directory. In the actual man:namei[9] routine we simply substitute the `dvp` vnode for `dp` variable in the man:namei[9] function, which determines the starting point. The man:namei[9] is not used directly but via a trace of different functions on various levels. For example the `openat` goes like this: [.programlisting] .... openat() --> kern_openat() --> vn_open() -> namei() .... For this reason `kern_open` and `vn_open` must be altered to incorporate the additional `dirfd` parameter. No compat layer is created for those because there are not many users of this and the users can be easily converted. This general implementation enables FreeBSD to implement their own *at syscalls. This is being discussed right now. [[ioctl]] ==== Ioctl The ioctl interface is quite fragile due to its generality. We have to bear in mind that devices differ between Linux(R) and FreeBSD so some care must be applied to do ioctl emulation work right. The ioctl handling is implemented in [.filename]#linux_ioctl.c#, where `linux_ioctl` function is defined. This function simply iterates over sets of ioctl handlers to find a handler that implements a given command. The ioctl syscall has three parameters, the file descriptor, command and an argument. The command is a 16-bit number, which in theory is divided into high 8 bits determining class of the ioctl command and low 8 bits, which are the actual command within the given set. The emulation takes advantage of this division. We implement handlers for each set, like `sound_handler` or `disk_handler`. Each handler has a maximum command and a minimum command defined, which is used for determining what handler is used. There are slight problems with this approach because Linux(R) does not use the set division consistently so sometimes ioctls for a different set are inside a set they should not belong to (SCSI generic ioctls inside cdrom set, etc.). FreeBSD currently does not implement many Linux(R) ioctls (compared to NetBSD, for example) but the plan is to port those from NetBSD. The trend is to use Linux(R) ioctls even in the native FreeBSD drivers because of the easy porting of applications. [[debugging]] ==== Debugging Every syscall should be debuggable. For this purpose we introduce a small infrastructure. We have the ldebug facility, which tells whether a given syscall should be debugged (settable via a sysctl). For printing we have LMSG and ARGS macros. Those are used for altering a printable string for uniform debugging messages. [[conclusion]] == Conclusion [[results]] === Results As of April 2007 the Linux(R) emulation layer is capable of emulating the Linux(R) 2.6.16 kernel quite well. The remaining problems concern futexes, unfinished *at family of syscalls, problematic signals delivery, missing `epoll` and `inotify` and probably some bugs we have not discovered yet. Despite this we are capable of running basically all the Linux(R) programs included in FreeBSD Ports Collection with Fedora Core 4 at 2.6.16 and there are some rudimentary reports of success with Fedora Core 6 at 2.6.16. The Fedora Core 6 linux_base was recently committed enabling some further testing of the emulation layer and giving us some more hints where we should put our effort in implementing missing stuff. We are able to run the most used applications like package:www/linux-firefox[], package:net-im/skype[] and some games from the Ports Collection. Some of the programs exhibit bad behavior under 2.6 emulation but this is currently under investigation and hopefully will be fixed soon. The only big application that is known not to work is the Linux(R) Java(TM) Development Kit and this is because of the requirement of `epoll` facility which is not directly related to the Linux(R) kernel 2.6. We hope to enable 2.6.16 emulation by default some time after FreeBSD 7.0 is released at least to expose the 2.6 emulation parts for some wider testing. Once this is done we can switch to Fedora Core 6 linux_base, which is the ultimate plan. [[future-work]] === Future work Future work should focus on fixing the remaining issues with futexes, implement the rest of the *at family of syscalls, fix the signal delivery and possibly implement the `epoll` and `inotify` facilities. We hope to be able to run the most important programs flawlessly soon, so we will be able to switch to the 2.6 emulation by default and make the Fedora Core 6 the default linux_base because our currently used Fedora Core 4 is not supported any more. The other possible goal is to share our code with NetBSD and DragonflyBSD. NetBSD has some support for 2.6 emulation but its far from finished and not really tested. DragonflyBSD has expressed some interest in porting the 2.6 improvements. Generally, as Linux(R) develops we would like to keep up with their development, implementing newly added syscalls. Splice comes to mind first. -Some already implemented syscalls are also heavily crippled, for example `mremap` and others. +Some already implemented syscalls are also suboptimal, for example `mremap` and others. Some performance improvements can also be made, finer grained locking and others. [[team]] === Team I cooperated on this project with (in alphabetical order): * `{jhb}` * `{kib}` * Emmanuel Dreyfus * Scot Hetzel * `{jkim}` * `{netchild}` * `{ssouhlal}` * Li Xiao * `{davidxu}` I would like to thank all those people for their advice, code reviews and general support. [[literatures]] == Literatures . Marshall Kirk McKusick - George V. Nevile-Neil. Design and Implementation of the FreeBSD operating system. Addison-Wesley, 2005. . https://tldp.org[https://tldp.org] . https://www.kernel.org[https://www.kernel.org] diff --git a/documentation/content/en/books/handbook/disks/_index.adoc b/documentation/content/en/books/handbook/disks/_index.adoc index 238c075973..5390274807 100644 --- a/documentation/content/en/books/handbook/disks/_index.adoc +++ b/documentation/content/en/books/handbook/disks/_index.adoc @@ -1,2558 +1,2558 @@ --- title: Chapter 18. Storage part: Part III. System Administration prev: books/handbook/audit next: books/handbook/geom description: This chapter covers the use of disks and storage media in FreeBSD. This includes SCSI and IDE disks, CD and DVD media, memory-backed disks, and USB storage devices. tags: ["storage", "disks", "gpart", "mount", "quotas", "encrypt", "GPT", "cdrecord", "NTFS", "quotas", "swap", "HAST", "CD", "DVD", "resizing", "growing"] --- [[disks]] = Storage :doctype: book :toc: macro :toclevels: 1 :icons: font :sectnums: :sectnumlevels: 6 :source-highlighter: rouge :experimental: :skip-front-matter: :xrefstyle: basic :relfileprefix: ../ :outfilesuffix: :sectnumoffset: 18 ifeval::["{backend}" == "html5"] :imagesdir: ../../../../images/books/handbook/disks/ endif::[] ifeval::["{backend}" == "pdf"] :imagesdir: ../../../../static/images/books/handbook/disks/ endif::[] ifeval::["{backend}" == "epub3"] :imagesdir: ../../../../static/images/books/handbook/disks/ endif::[] include::shared/authors.adoc[] include::shared/releases.adoc[] include::shared/en/mailing-lists.adoc[] include::shared/en/teams.adoc[] include::shared/en/urls.adoc[] toc::[] [[disks-synopsis]] == Synopsis This chapter covers the use of disks and storage media in FreeBSD. This includes SCSI and IDE disks, CD and DVD media, memory-backed disks, and USB storage devices. After reading this chapter, you will know: * How to add additional hard disks to a FreeBSD system. * How to grow the size of a disk's partition on FreeBSD. * How to configure FreeBSD to use USB storage devices. * How to use CD and DVD media on a FreeBSD system. * How to use the backup programs available under FreeBSD. * How to set up memory disks. * What file system snapshots are and how to use them efficiently. * How to use quotas to limit disk space usage. * How to encrypt disks and swap to secure them against attackers. * How to configure a highly available storage network. Before reading this chapter, you should: * Know how to crossref:kernelconfig[kernelconfig,configure and install a new FreeBSD kernel]. [[disks-adding]] == Adding Disks This section describes how to add a new SATA disk to a machine that currently only has a single drive. First, turn off the computer and install the drive in the computer following the instructions of the computer, controller, and drive manufacturers. Reboot the system and become `root`. Inspect [.filename]#/var/run/dmesg.boot# to ensure the new disk was found. In this example, the newly added SATA drive will appear as [.filename]#ada1#. For this example, a single large partition will be created on the new disk. The http://en.wikipedia.org/wiki/GUID_Partition_Table[GPT] partitioning scheme will be used in preference to the older and less versatile MBR scheme. [NOTE] ==== If the disk to be added is not blank, old partition information can be removed with `gpart delete`. See man:gpart[8] for details. ==== The partition scheme is created, and then a single partition is added. To improve performance on newer disks with larger hardware block sizes, the partition is aligned to one megabyte boundaries: [source,shell] .... # gpart create -s GPT ada1 # gpart add -t freebsd-ufs -a 1M ada1 .... Depending on use, several smaller partitions may be desired. See man:gpart[8] for options to create partitions smaller than a whole disk. The disk partition information can be viewed with `gpart show`: [source,shell] .... % gpart show ada1 => 34 1465146988 ada1 GPT (699G) 34 2014 - free - (1.0M) 2048 1465143296 1 freebsd-ufs (699G) 1465145344 1678 - free - (839K) .... A file system is created in the new partition on the new disk: [source,shell] .... # newfs -U /dev/ada1p1 .... An empty directory is created as a _mountpoint_, a location for mounting the new disk in the original disk's file system: [source,shell] .... # mkdir /newdisk .... Finally, an entry is added to [.filename]#/etc/fstab# so the new disk will be mounted automatically at startup: [.programlisting] .... /dev/ada1p1 /newdisk ufs rw 2 2 .... The new disk can be mounted manually, without restarting the system: [source,shell] .... # mount /newdisk .... [[disks-growing]] == Resizing and Growing Disks A disk's capacity can increase without any changes to the data already present. This happens commonly with virtual machines, when the virtual disk turns out to be too small and is enlarged. Sometimes a disk image is written to a USB memory stick, but does not use the full capacity. Here we describe how to resize or _grow_ disk contents to take advantage of increased capacity. Determine the device name of the disk to be resized by inspecting [.filename]#/var/run/dmesg.boot#. In this example, there is only one SATA disk in the system, so the drive will appear as [.filename]#ada0#. List the partitions on the disk to see the current configuration: [source,shell] .... # gpart show ada0 => 34 83886013 ada0 GPT (48G) [CORRUPT] 34 128 1 freebsd-boot (64k) 162 79691648 2 freebsd-ufs (38G) 79691810 4194236 3 freebsd-swap (2G) 83886046 1 - free - (512B) .... [NOTE] ==== If the disk was formatted with the http://en.wikipedia.org/wiki/GUID_Partition_Table[GPT] partitioning scheme, it may show as "corrupted" because the GPT backup partition table is no longer at the end of the drive. Fix the backup partition table with `gpart`: [source,shell] .... # gpart recover ada0 ada0 recovered .... ==== Now the additional space on the disk is available for use by a new partition, or an existing partition can be expanded: [source,shell] .... # gpart show ada0 => 34 102399933 ada0 GPT (48G) 34 128 1 freebsd-boot (64k) 162 79691648 2 freebsd-ufs (38G) 79691810 4194236 3 freebsd-swap (2G) 83886046 18513921 - free - (8.8G) .... Partitions can only be resized into contiguous free space. Here, the last partition on the disk is the swap partition, but the second partition is the one that needs to be resized. Swap partitions only contain temporary data, so it can safely be unmounted, deleted, and then recreate the third partition after resizing the second partition. Disable the swap partition: [source,shell] .... # swapoff /dev/ada0p3 .... Delete the third partition, specified by the `-i` flag, from the disk _ada0_. [source,shell] .... # gpart delete -i 3 ada0 ada0p3 deleted # gpart show ada0 => 34 102399933 ada0 GPT (48G) 34 128 1 freebsd-boot (64k) 162 79691648 2 freebsd-ufs (38G) 79691810 22708157 - free - (10G) .... [WARNING] ==== There is risk of data loss when modifying the partition table of a mounted file system. It is best to perform the following steps on an unmounted file system while running from a live CD-ROM or USB device. However, if absolutely necessary, a mounted file system can be resized after disabling GEOM safety features: [source,shell] .... # sysctl kern.geom.debugflags=16 .... ==== Resize the partition, leaving room to recreate a swap partition of the desired size. The partition to resize is specified with `-i`, and the new desired size with `-s`. Optionally, alignment of the partition is controlled with `-a`. This only modifies the size of the partition. The file system in the partition will be expanded in a separate step. [source,shell] .... # gpart resize -i 2 -s 47G -a 4k ada0 ada0p2 resized # gpart show ada0 => 34 102399933 ada0 GPT (48G) 34 128 1 freebsd-boot (64k) 162 98566144 2 freebsd-ufs (47G) 98566306 3833661 - free - (1.8G) .... Recreate the swap partition and activate it. If no size is specified with `-s`, all remaining space is used: [source,shell] .... # gpart add -t freebsd-swap -a 4k ada0 ada0p3 added # gpart show ada0 => 34 102399933 ada0 GPT (48G) 34 128 1 freebsd-boot (64k) 162 98566144 2 freebsd-ufs (47G) 98566306 3833661 3 freebsd-swap (1.8G) # swapon /dev/ada0p3 .... Grow the UFS file system to use the new capacity of the resized partition: [source,shell] .... # growfs /dev/ada0p2 Device is mounted read-write; resizing will result in temporary write suspension for /. It's strongly recommended to make a backup before growing the file system. OK to grow file system on /dev/ada0p2, mounted on /, from 38GB to 47GB? [Yes/No] Yes super-block backups (for fsck -b #) at: 80781312, 82063552, 83345792, 84628032, 85910272, 87192512, 88474752, 89756992, 91039232, 92321472, 93603712, 94885952, 96168192, 97450432 .... If the file system is ZFS, the resize is triggered by running the `online` subcommand with `-e`: [source,shell] .... # zpool online -e zroot /dev/ada0p2 .... Both the partition and the file system on it have now been resized to use the newly-available disk space. [[usb-disks]] == USB Storage Devices Many external storage solutions, such as hard drives, USB thumbdrives, and CD and DVD burners, use the Universal Serial Bus (USB). FreeBSD provides support for USB 1.x, 2.0, and 3.0 devices. [NOTE] ==== USB 3.0 support is not compatible with some hardware, including Haswell (Lynx point) chipsets. If FreeBSD boots with a `failed with error 19` message, disable xHCI/USB3 in the system BIOS. ==== Support for USB storage devices is built into the [.filename]#GENERIC# kernel. For a custom kernel, be sure that the following lines are present in the kernel configuration file: [.programlisting] .... device scbus # SCSI bus (required for ATA/SCSI) device da # Direct Access (disks) device pass # Passthrough device (direct ATA/SCSI access) device uhci # provides USB 1.x support device ohci # provides USB 1.x support device ehci # provides USB 2.0 support device xhci # provides USB 3.0 support device usb # USB Bus (required) device umass # Disks/Mass storage - Requires scbus and da device cd # needed for CD and DVD burners .... FreeBSD uses the man:umass[4] driver which uses the SCSI subsystem to access USB storage devices. Since any USB device will be seen as a SCSI device by the system, if the USB device is a CD or DVD burner, do _not_ include `device atapicam` in a custom kernel configuration file. The rest of this section demonstrates how to verify that a USB storage device is recognized by FreeBSD and how to configure the device so that it can be used. === Device Configuration To test the USB configuration, plug in the USB device. Use `dmesg` to confirm that the drive appears in the system message buffer. It should look something like this: [source,shell] .... umass0: on usbus0 umass0: SCSI over Bulk-Only; quirks = 0x0100 umass0:4:0:-1: Attached to scbus4 da0 at umass-sim0 bus 0 scbus4 target 0 lun 0 da0: Fixed Direct Access SCSI-4 device da0: Serial Number WD-WXE508CAN263 da0: 40.000MB/s transfers da0: 152627MB (312581808 512 byte sectors: 255H 63S/T 19457C) da0: quirks=0x2 .... The brand, device node ([.filename]#da0#), speed, and size will differ according to the device. Since the USB device is seen as a SCSI one, `camcontrol` can be used to list the USB storage devices attached to the system: [source,shell] .... # camcontrol devlist at scbus4 target 0 lun 0 (pass3,da0) .... Alternately, `usbconfig` can be used to list the device. Refer to man:usbconfig[8] for more information about this command. [source,shell] .... # usbconfig ugen0.3: at usbus0, cfg=0 md=HOST spd=HIGH (480Mbps) pwr=ON (2mA) .... If the device has not been formatted, refer to <> for instructions on how to format and create partitions on the USB drive. If the drive comes with a file system, it can be mounted by `root` using the instructions in crossref:basics[mount-unmount,“Mounting and Unmounting File Systems”]. [WARNING] ==== Allowing untrusted users to mount arbitrary media, by enabling `vfs.usermount` as described below, should not be considered safe from a security point of view. Most file systems were not built to safeguard against malicious devices. ==== To make the device mountable as a normal user, one solution is to make all users of the device a member of the `operator` group using man:pw[8]. Next, ensure that `operator` is able to read and write the device by adding these lines to [.filename]#/etc/devfs.rules#: [.programlisting] .... [localrules=5] add path 'da*' mode 0660 group operator .... [NOTE] ==== If internal SCSI disks are also installed in the system, change the second line as follows: [.programlisting] .... add path 'da[3-9]*' mode 0660 group operator .... This will exclude the first three SCSI disks ([.filename]#da0# to [.filename]#da2#) from belonging to the `operator` group. Replace _3_ with the number of internal SCSI disks. Refer to man:devfs.rules[5] for more information about this file. ==== Next, enable the ruleset in [.filename]#/etc/rc.conf#: [.programlisting] .... devfs_system_ruleset="localrules" .... Then, instruct the system to allow regular users to mount file systems by adding the following line to [.filename]#/etc/sysctl.conf#: [.programlisting] .... vfs.usermount=1 .... Since this only takes effect after the next reboot, use `sysctl` to set this variable now: [source,shell] .... # sysctl vfs.usermount=1 vfs.usermount: 0 -> 1 .... The final step is to create a directory where the file system is to be mounted. This directory needs to be owned by the user that is to mount the file system. One way to do that is for `root` to create a subdirectory owned by that user as [.filename]#/mnt/username#. In the following example, replace _username_ with the login name of the user and _usergroup_ with the user's primary group: [source,shell] .... # mkdir /mnt/username # chown username:usergroup /mnt/username .... Suppose a USB thumbdrive is plugged in, and a device [.filename]#/dev/da0s1# appears. If the device is formatted with a FAT file system, the user can mount it using: [source,shell] .... % mount -t msdosfs -o -m=644,-M=755 /dev/da0s1 /mnt/username .... Before the device can be unplugged, it _must_ be unmounted first: [source,shell] .... % umount /mnt/username .... After device removal, the system message buffer will show messages similar to the following: [source,shell] .... umass0: at uhub3, port 2, addr 3 (disconnected) da0 at umass-sim0 bus 0 scbus4 target 0 lun 0 da0: s/n WD-WXE508CAN263 detached (da0:umass-sim0:0:0:0): Periph destroyed .... === Automounting Removable Media USB devices can be automatically mounted by uncommenting this line in [.filename]#/etc/auto_master#: [source,shell] .... /media -media -nosuid .... Then add these lines to [.filename]#/etc/devd.conf#: [source,shell] .... notify 100 { match "system" "GEOM"; match "subsystem" "DEV"; action "/usr/sbin/automount -c"; }; .... Reload the configuration if man:autofs[5] and man:devd[8] are already running: [source,shell] .... # service automount restart # service devd restart .... man:autofs[5] can be set to start at boot by adding this line to [.filename]#/etc/rc.conf#: [.programlisting] .... autofs_enable="YES" .... man:autofs[5] requires man:devd[8] to be enabled, as it is by default. Start the services immediately with: [source,shell] .... # service automount start # service automountd start # service autounmountd start # service devd start .... Each file system that can be automatically mounted appears as a directory in [.filename]#/media/#. The directory is named after the file system label. If the label is missing, the directory is named after the device node. The file system is transparently mounted on the first access, and unmounted after a period of inactivity. Automounted drives can also be unmounted manually: [source,shell] .... # automount -fu .... This mechanism is typically used for memory cards and USB memory sticks. It can be used with any block device, including optical drives or iSCSILUNs. [[creating-cds]] == Creating and Using CD Media Compact Disc (CD) media provide a number of features that differentiate them from conventional disks. They are designed so that they can be read continuously without delays to move the head between tracks. While CD media do have tracks, these refer to a section of data to be read continuously, and not a physical property of the disk. The ISO 9660 file system was designed to deal with these differences. The FreeBSD Ports Collection provides several utilities for burning and duplicating audio and data CDs. This chapter demonstrates the use of several command line utilities. For CD burning software with a graphical utility, consider installing the package:sysutils/xcdroast[] or package:sysutils/k3b[] packages or ports. [[atapicam]] === Supported Devices The [.filename]#GENERIC# kernel provides support for SCSI, USB, and ATAPICD readers and burners. If a custom kernel is used, the options that need to be present in the kernel configuration file vary by the type of device. For a SCSI burner, make sure these options are present: [.programlisting] .... device scbus # SCSI bus (required for ATA/SCSI) device da # Direct Access (disks) device pass # Passthrough device (direct ATA/SCSI access) device cd # needed for CD and DVD burners .... For a USB burner, make sure these options are present: [.programlisting] .... device scbus # SCSI bus (required for ATA/SCSI) device da # Direct Access (disks) device pass # Passthrough device (direct ATA/SCSI access) device cd # needed for CD and DVD burners device uhci # provides USB 1.x support device ohci # provides USB 1.x support device ehci # provides USB 2.0 support device xhci # provides USB 3.0 support device usb # USB Bus (required) device umass # Disks/Mass storage - Requires scbus and da .... For an ATAPI burner, make sure these options are present: [.programlisting] .... device ata # Legacy ATA/SATA controllers device scbus # SCSI bus (required for ATA/SCSI) device pass # Passthrough device (direct ATA/SCSI access) device cd # needed for CD and DVD burners .... [NOTE] ==== On FreeBSD versions prior to 10.x, this line is also needed in the kernel configuration file if the burner is an ATAPI device: [.programlisting] .... device atapicam .... Alternately, this driver can be loaded at boot time by adding the following line to [.filename]#/boot/loader.conf#: [.programlisting] .... atapicam_load="YES" .... This will require a reboot of the system as this driver can only be loaded at boot time. ==== To verify that FreeBSD recognizes the device, run `dmesg` and look for an entry for the device. On systems prior to 10.x, the device name in the first line of the output will be [.filename]#acd0# instead of [.filename]#cd0#. [source,shell] .... % dmesg | grep cd cd0 at ahcich1 bus 0 scbus1 target 0 lun 0 cd0: Removable CD-ROM SCSI-0 device cd0: Serial Number M3OD3S34152 cd0: 150.000MB/s transfers (SATA 1.x, UDMA6, ATAPI 12bytes, PIO 8192bytes) cd0: Attempt to query device size failed: NOT READY, Medium not present - tray closed .... [[cdrecord]] === Burning a CD In FreeBSD, `cdrecord` can be used to burn CDs. This command is installed with the package:sysutils/cdrtools[] package or port. While `cdrecord` has many options, basic usage is simple. Specify the name of the ISO file to burn and, if the system has multiple burner devices, specify the name of the device to use: [source,shell] .... # cdrecord dev=device imagefile.iso .... To determine the device name of the burner, use `-scanbus` which might produce results like this: [source,shell] .... # cdrecord -scanbus ProDVD-ProBD-Clone 3.00 (amd64-unknown-freebsd10.0) Copyright (C) 1995-2010 Jörg Schilling Using libscg version 'schily-0.9' scsibus0: 0,0,0 0) 'SEAGATE ' 'ST39236LW ' '0004' Disk 0,1,0 1) 'SEAGATE ' 'ST39173W ' '5958' Disk 0,2,0 2) * 0,3,0 3) 'iomega ' 'jaz 1GB ' 'J.86' Removable Disk 0,4,0 4) 'NEC ' 'CD-ROM DRIVE:466' '1.26' Removable CD-ROM 0,5,0 5) * 0,6,0 6) * 0,7,0 7) * scsibus1: 1,0,0 100) * 1,1,0 101) * 1,2,0 102) * 1,3,0 103) * 1,4,0 104) * 1,5,0 105) 'YAMAHA ' 'CRW4260 ' '1.0q' Removable CD-ROM 1,6,0 106) 'ARTEC ' 'AM12S ' '1.06' Scanner 1,7,0 107) * .... Locate the entry for the CD burner and use the three numbers separated by commas as the value for `dev`. In this case, the Yamaha burner device is `1,5,0`, so the appropriate input to specify that device is `dev=1,5,0`. Refer to the manual page for `cdrecord` for other ways to specify this value and for information on writing audio tracks and controlling the write speed. Alternately, run the following command to get the device address of the burner: [source,shell] .... # camcontrol devlist at scbus1 target 0 lun 0 (cd0,pass0) .... Use the numeric values for `scbus`, `target`, and `lun`. For this example, `1,0,0` is the device name to use. [[mkisofs]] === Writing Data to an ISO File System In order to produce a data CD, the data files that are going to make up the tracks on the CD must be prepared before they can be burned to the CD. In FreeBSD, package:sysutils/cdrtools[] installs `mkisofs`, which can be used to produce an ISO 9660 file system that is an image of a directory tree within a UNIX(R) file system. The simplest usage is to specify the name of the ISO file to create and the path to the files to place into the ISO 9660 file system: [source,shell] .... # mkisofs -o imagefile.iso /path/to/tree .... This command maps the file names in the specified path to names that fit the limitations of the standard ISO 9660 file system, and will exclude files that do not meet the standard for ISO file systems. A number of options are available to overcome the restrictions imposed by the standard. In particular, `-R` enables the Rock Ridge extensions common to UNIX(R) systems and `-J` enables Joliet extensions used by Microsoft(R) systems. For CDs that are going to be used only on FreeBSD systems, `-U` can be used to disable all filename restrictions. When used with `-R`, it produces a file system image that is identical to the specified FreeBSD tree, even if it violates the ISO 9660 standard. The last option of general use is `-b`. This is used to specify the location of a boot image for use in producing an "El Torito" bootable CD. This option takes an argument which is the path to a boot image from the top of the tree being written to the CD. By default, `mkisofs` creates an ISO image in "floppy disk emulation" mode, and thus expects the boot image to be exactly 1200, 1440 or 2880 KB in size. Some boot loaders, like the one used by the FreeBSD distribution media, do not use emulation mode. In this case, `-no-emul-boot` should be used. So, if [.filename]#/tmp/myboot# holds a bootable FreeBSD system with the boot image in [.filename]#/tmp/myboot/boot/cdboot#, this command would produce [.filename]#/tmp/bootable.iso#: [source,shell] .... # mkisofs -R -no-emul-boot -b boot/cdboot -o /tmp/bootable.iso /tmp/myboot .... The resulting ISO image can be mounted as a memory disk with: [source,shell] .... # mdconfig -a -t vnode -f /tmp/bootable.iso -u 0 # mount -t cd9660 /dev/md0 /mnt .... One can then verify that [.filename]#/mnt# and [.filename]#/tmp/myboot# are identical. There are many other options available for `mkisofs` to fine-tune its behavior. Refer to man:mkisofs[8] for details. [NOTE] ==== It is possible to copy a data CD to an image file that is functionally equivalent to the image file created with `mkisofs`. To do so, use [.filename]#dd# with the device name as the input file and the name of the ISO to create as the output file: [source,shell] .... # dd if=/dev/cd0 of=file.iso bs=2048 .... The resulting image file can be burned to CD as described in <>. ==== [[mounting-cd]] === Using Data CDs Once an ISO has been burned to a CD, it can be mounted by specifying the file system type, the name of the device containing the CD, and an existing mount point: [source,shell] .... # mount -t cd9660 /dev/cd0 /mnt .... Since `mount` assumes that a file system is of type `ufs`, an `Incorrect super block` error will occur if `-t cd9660` is not included when mounting a data CD. While any data CD can be mounted this way, disks with certain ISO 9660 extensions might behave oddly. For example, Joliet disks store all filenames in two-byte Unicode characters. If some non-English characters show up as question marks, specify the local charset with `-C`. For more information, refer to man:mount_cd9660[8]. [NOTE] ==== In order to do this character conversion with the help of `-C`, the kernel requires the [.filename]#cd9660_iconv.ko# module to be loaded. This can be done either by adding this line to [.filename]#loader.conf#: [.programlisting] .... cd9660_iconv_load="YES" .... and then rebooting the machine, or by directly loading the module with `kldload`. ==== Occasionally, `Device not configured` will be displayed when trying to mount a data CD. This usually means that the CD drive has not detected a disk in the tray, or that the drive is not visible on the bus. It can take a couple of seconds for a CD drive to detect media, so be patient. Sometimes, a SCSICD drive may be missed because it did not have enough time to answer the bus reset. To resolve this, a custom kernel can be created which increases the default SCSI delay. Add the following option to the custom kernel configuration file and rebuild the kernel using the instructions in crossref:kernelconfig[kernelconfig-building,“Building and Installing a Custom Kernel”]: [.programlisting] .... options SCSI_DELAY=15000 .... This tells the SCSI bus to pause 15 seconds during boot, to give the CD drive every possible chance to answer the bus reset. [NOTE] ==== It is possible to burn a file directly to CD, without creating an ISO 9660 file system. This is known as burning a raw data CD and some people do this for backup purposes. This type of disk can not be mounted as a normal data CD. In order to retrieve the data burned to such a CD, the data must be read from the raw device node. For example, this command will extract a compressed tar file located on the second CD device into the current working directory: [source,shell] .... # tar xzvf /dev/cd1 .... In order to mount a data CD, the data must be written using `mkisofs`. ==== [[duplicating-audiocds]] === Duplicating Audio CDs To duplicate an audio CD, extract the audio data from the CD to a series of files, then write these files to a blank CD. <> describes how to duplicate and burn an audio CD. If the FreeBSD version is less than 10.0 and the device is ATAPI, the `atapicam` module must be first loaded using the instructions in <>. [[using-cdrecord]] [.procedure] .Procedure: Duplicating an Audio CD . The package:sysutils/cdrtools[] package or port installs `cdda2wav`. This command can be used to extract all of the audio tracks, with each track written to a separate WAV file in the current working directory: + [source,shell] .... % cdda2wav -vall -B -Owav .... + A device name does not need to be specified if there is only one CD device on the system. Refer to the `cdda2wav` manual page for instructions on how to specify a device and to learn more about the other options available for this command. . Use `cdrecord` to write the [.filename]#.wav# files: + [source,shell] .... % cdrecord -v dev=2,0 -dao -useinfo *.wav .... + Make sure that _2,0_ is set appropriately, as described in <>. [[creating-dvds]] == Creating and Using DVD Media Compared to the CD, the DVD is the next generation of optical media storage technology. The DVD can hold more data than any CD and is the standard for video publishing. Five physical recordable formats can be defined for a recordable DVD: * DVD-R: This was the first DVD recordable format available. The DVD-R standard is defined by the http://www.dvdforum.org/forum.shtml[DVD Forum]. This format is write once. * DVD-RW: This is the rewritable version of the DVD-R standard. A DVD-RW can be rewritten about 1000 times. * DVD-RAM: This is a rewritable format which can be seen as a removable hard drive. However, this media is not compatible with most DVD-ROM drives and DVD-Video players as only a few DVD writers support the DVD-RAM format. Refer to <> for more information on DVD-RAM use. * DVD+RW: This is a rewritable format defined by the https://en.wikipedia.org/wiki/DVD%2BRW_Alliance[DVD+RW Alliance]. A DVD+RW can be rewritten about 1000 times. * DVD+R: This format is the write once variation of the DVD+RW format. A single layer recordable DVD can hold up to 4,700,000,000 bytes which is actually 4.38 GB or 4485 MB as 1 kilobyte is 1024 bytes. [NOTE] ==== A distinction must be made between the physical media and the application. For example, a DVD-Video is a specific file layout that can be written on any recordable DVD physical media such as DVD-R, DVD+R, or DVD-RW. Before choosing the type of media, ensure that both the burner and the DVD-Video player are compatible with the media under consideration. ==== === Configuration To perform DVD recording, use man:growisofs[1]. This command is part of the package:sysutils/dvd+rw-tools[] utilities which support all DVD media types. These tools use the SCSI subsystem to access the devices, therefore <> must be loaded or statically compiled into the kernel. This support is not needed if the burner uses the USB interface. Refer to <> for more details on USB device configuration. DMA access must also be enabled for ATAPI devices, by adding the following line to [.filename]#/boot/loader.conf#: [.programlisting] .... hw.ata.atapi_dma="1" .... Before attempting to use dvd+rw-tools, consult the http://fy.chalmers.se/~appro/linux/DVD+RW/hcn.html[Hardware Compatibility Notes]. [NOTE] ==== For a graphical user interface, consider using package:sysutils/k3b[] which provides a user friendly interface to man:growisofs[1] and many other burning tools. ==== === Burning Data DVDs Since man:growisofs[1] is a front-end to <>, it will invoke man:mkisofs[8] to create the file system layout and perform the write on the DVD. This means that an image of the data does not need to be created before the burning process. To burn to a DVD+R or a DVD-R the data in [.filename]#/path/to/data#, use the following command: [source,shell] .... # growisofs -dvd-compat -Z /dev/cd0 -J -R /path/to/data .... In this example, `-J -R` is passed to man:mkisofs[8] to create an ISO 9660 file system with Joliet and Rock Ridge extensions. Refer to man:mkisofs[8] for more details. For the initial session recording, `-Z` is used for both single and multiple sessions. Replace _/dev/cd0_, with the name of the DVD device. Using `-dvd-compat` indicates that the disk will be closed and that the recording will be unappendable. This should also provide better media compatibility with DVD-ROM drives. To burn a pre-mastered image, such as _imagefile.iso_, use: [source,shell] .... # growisofs -dvd-compat -Z /dev/cd0=imagefile.iso .... The write speed should be detected and automatically set according to the media and the drive being used. To force the write speed, use `-speed=`. Refer to man:growisofs[1] for example usage. [NOTE] ==== In order to support working files larger than 4.38GB, an UDF/ISO-9660 hybrid file system must be created by passing `-udf -iso-level 3` to man:mkisofs[8] and all related programs, such as man:growisofs[1]. This is required only when creating an ISO image file or when writing files directly to a disk. Since a disk created this way must be mounted as an UDF file system with man:mount_udf[8], it will be usable only on an UDF aware operating system. Otherwise it will look as if it contains corrupted files. To create this type of ISO file: [source,shell] .... % mkisofs -R -J -udf -iso-level 3 -o imagefile.iso /path/to/data .... To burn files directly to a disk: [source,shell] .... # growisofs -dvd-compat -udf -iso-level 3 -Z /dev/cd0 -J -R /path/to/data .... When an ISO image already contains large files, no additional options are required for man:growisofs[1] to burn that image on a disk. Be sure to use an up-to-date version of package:sysutils/cdrtools[], which contains man:mkisofs[8], as an older version may not contain large files support. If the latest version does not work, install package:sysutils/cdrtools-devel[] and read its man:mkisofs[8]. ==== === Burning a DVD-Video A DVD-Video is a specific file layout based on the ISO 9660 and micro-UDF (M-UDF) specifications. Since DVD-Video presents a specific data structure hierarchy, a particular program such as package:multimedia/dvdauthor[] is needed to author the DVD. If an image of the DVD-Video file system already exists, it can be burned in the same way as any other image. If `dvdauthor` was used to make the DVD and the result is in [.filename]#/path/to/video#, the following command should be used to burn the DVD-Video: [source,shell] .... # growisofs -Z /dev/cd0 -dvd-video /path/to/video .... `-dvd-video` is passed to man:mkisofs[8] to instruct it to create a DVD-Video file system layout. This option implies the `-dvd-compat` man:growisofs[1] option. === Using a DVD+RW Unlike CD-RW, a virgin DVD+RW needs to be formatted before first use. It is _recommended_ to let man:growisofs[1] take care of this automatically whenever appropriate. However, it is possible to use `dvd+rw-format` to format the DVD+RW: [source,shell] .... # dvd+rw-format /dev/cd0 .... Only perform this operation once and keep in mind that only virgin DVD+RW medias need to be formatted. Once formatted, the DVD+RW can be burned as usual. To burn a totally new file system and not just append some data onto a DVD+RW, the media does not need to be blanked first. Instead, write over the previous recording like this: [source,shell] .... # growisofs -Z /dev/cd0 -J -R /path/to/newdata .... The DVD+RW format supports appending data to a previous recording. This operation consists of merging a new session to the existing one as it is not considered to be multi-session writing. man:growisofs[1] will _grow_ the ISO 9660 file system present on the media. For example, to append data to a DVD+RW, use the following: [source,shell] .... # growisofs -M /dev/cd0 -J -R /path/to/nextdata .... The same man:mkisofs[8] options used to burn the initial session should be used during next writes. [NOTE] ==== Use `-dvd-compat` for better media compatibility with DVD-ROM drives. When using DVD+RW, this option will not prevent the addition of data. ==== To blank the media, use: [source,shell] .... # growisofs -Z /dev/cd0=/dev/zero .... === Using a DVD-RW A DVD-RW accepts two disc formats: incremental sequential and restricted overwrite. By default, DVD-RW discs are in sequential format. A virgin DVD-RW can be directly written without being formatted. However, a non-virgin DVD-RW in sequential format needs to be blanked before writing a new initial session. To blank a DVD-RW in sequential mode: [source,shell] .... # dvd+rw-format -blank=full /dev/cd0 .... [NOTE] ==== A full blanking using `-blank=full` will take about one hour on a 1x media. A fast blanking can be performed using `-blank`, if the DVD-RW will be recorded in Disk-At-Once (DAO) mode. To burn the DVD-RW in DAO mode, use the command: [source,shell] .... # growisofs -use-the-force-luke=dao -Z /dev/cd0=imagefile.iso .... Since man:growisofs[1] automatically attempts to detect fast blanked media and engage DAO write, `-use-the-force-luke=dao` should not be required. One should instead use restricted overwrite mode with any DVD-RW as this format is more flexible than the default of incremental sequential. ==== To write data on a sequential DVD-RW, use the same instructions as for the other DVD formats: [source,shell] .... # growisofs -Z /dev/cd0 -J -R /path/to/data .... To append some data to a previous recording, use `-M` with man:growisofs[1]. However, if data is appended on a DVD-RW in incremental sequential mode, a new session will be created on the disc and the result will be a multi-session disc. A DVD-RW in restricted overwrite format does not need to be blanked before a new initial session. Instead, overwrite the disc with `-Z`. It is also possible to grow an existing ISO 9660 file system written on the disc with `-M`. The result will be a one-session DVD. To put a DVD-RW in restricted overwrite format, the following command must be used: [source,shell] .... # dvd+rw-format /dev/cd0 .... To change back to sequential format, use: [source,shell] .... # dvd+rw-format -blank=full /dev/cd0 .... === Multi-Session Few DVD-ROM drives support multi-session DVDs and most of the time only read the first session. DVD+R, DVD-R and DVD-RW in sequential format can accept multiple sessions. The notion of multiple sessions does not exist for the DVD+RW and the DVD-RW restricted overwrite formats. Using the following command after an initial non-closed session on a DVD+R, DVD-R, or DVD-RW in sequential format, will add a new session to the disc: [source,shell] .... # growisofs -M /dev/cd0 -J -R /path/to/nextdata .... Using this command with a DVD+RW or a DVD-RW in restricted overwrite mode will append data while merging the new session to the existing one. The result will be a single-session disc. Use this method to add data after an initial write on these types of media. [NOTE] ==== Since some space on the media is used between each session to mark the end and start of sessions, one should add sessions with a large amount of data to optimize media space. The number of sessions is limited to 154 for a DVD+R, about 2000 for a DVD-R, and 127 for a DVD+R Double Layer. ==== === For More Information To obtain more information about a DVD, use `dvd+rw-mediainfo _/dev/cd0_` while the disc in the specified drive. More information about dvd+rw-tools can be found in man:growisofs[1], on the http://fy.chalmers.se/~appro/linux/DVD+RW/[dvd+rw-tools web site], and in the http://lists.debian.org/cdwrite/[cdwrite mailing list] archives. [NOTE] ==== When creating a problem report related to the use of dvd+rw-tools, always include the output of `dvd+rw-mediainfo`. ==== [[creating-dvd-ram]] === Using a DVD-RAM DVD-RAM writers can use either a SCSI or ATAPI interface. For ATAPI devices, DMA access has to be enabled by adding the following line to [.filename]#/boot/loader.conf#: [.programlisting] .... hw.ata.atapi_dma="1" .... A DVD-RAM can be seen as a removable hard drive. Like any other hard drive, the DVD-RAM must be formatted before it can be used. In this example, the whole disk space will be formatted with a standard UFS2 file system: [source,shell] .... # dd if=/dev/zero of=/dev/acd0 bs=2k count=1 # bsdlabel -Bw acd0 # newfs /dev/acd0 .... The DVD device, [.filename]#acd0#, must be changed according to the configuration. Once the DVD-RAM has been formatted, it can be mounted as a normal hard drive: [source,shell] .... # mount /dev/acd0 /mnt .... Once mounted, the DVD-RAM will be both readable and writeable. [[floppies]] == Creating and Using Floppy Disks This section explains how to format a 3.5 inch floppy disk in FreeBSD. [.procedure] ==== *Procedure: Steps to Format a Floppy* A floppy disk needs to be low-level formatted before it can be used. This is usually done by the vendor, but formatting is a good way to check media integrity. To low-level format the floppy disk on FreeBSD, use man:fdformat[1]. When using this utility, make note of any error messages, as these can help determine if the disk is good or bad. . To format the floppy, insert a new 3.5 inch floppy disk into the first floppy drive and issue: + [source,shell] .... # /usr/sbin/fdformat -f 1440 /dev/fd0 .... + . After low-level formatting the disk, create a disk label as it is needed by the system to determine the size of the disk and its geometry. The supported geometry values are listed in [.filename]#/etc/disktab#. + To write the disk label, use man:bsdlabel[8]: + [source,shell] .... # /sbin/bsdlabel -B -w /dev/fd0 fd1440 .... + . The floppy is now ready to be high-level formatted with a file system. The floppy's file system can be either UFS or FAT, where FAT is generally a better choice for floppies. + To format the floppy with FAT, issue: + [source,shell] .... # /sbin/newfs_msdos /dev/fd0 .... ==== The disk is now ready for use. To use the floppy, mount it with man:mount_msdosfs[8]. One can also install and use package:emulators/mtools[] from the Ports Collection. [[using-ntfs]] == Using NTFS Disks This section explains how to mount NTFS disks in FreeBSD. NTFS (New Technology File System) is a proprietary journaling file system developed by Microsoft(R). It has been the default file system in Microsoft Windows(R) for many years. FreeBSD can mount NTFS volumes using a FUSE file system. These file systems are implemented as user space programs which interact with the man:fusefs[5] kernel module via a well defined interface. [.procedure] ==== *Procedure: Steps to Mount a NTFS Disk* . Before using a FUSE file system we need to load the man:fusefs[5] kernel module: + [source,shell] .... # kldload fusefs .... + Use man:sysrc[8] to load the module at startup: + [source,shell] .... # sysrc kld_list+=fusefs .... . Install the actual NTFS file system from packages as in the example (see crossref:ports[pkgng-intro,Using pkg for Binary Package Management]) or from ports (see crossref:ports[ports-using,Using the Ports Collection]): + [source,shell] .... # pkg install fusefs-ntfs .... . Last we need to create a directory where the file system will be mounted: + [source,shell] .... # mkdir /mnt/usb .... . Suppose a USB disk is plugged in. The disk partition information can be viewed with man:gpart[8]: + [source,shell] .... # gpart show da0 => 63 1953525105 da0 MBR (932G) 63 1953525105 1 ntfs (932G) .... . We can mount the disk using the following command: + [source,shell] .... # ntfs-3g /dev/da0s1 /mnt/usb/ .... The disk is now ready to use. + . Additionally, an entry can be added to /etc/fstab: + [.programlisting] .... /dev/da0s1 /mnt/usb ntfs mountprog=/usr/local/bin/ntfs-3g,noauto,rw 0 0 .... + Now the disk can be now mounted with: + [source,shell] .... # mount /mnt/usb .... . The disk can be unmounted with: + [source,shell] .... # umount /mnt/usb/ .... ==== [[backup-basics]] == Backup Basics Implementing a backup plan is essential in order to have the ability to recover from disk failure, accidental file deletion, random file corruption, or complete machine destruction, including destruction of on-site backups. The backup type and schedule will vary, depending upon the importance of the data, the granularity needed for file restores, and the amount of acceptable downtime. Some possible backup techniques include: * Archives of the whole system, backed up onto permanent, off-site media. This provides protection against all of the problems listed above, but is slow and inconvenient to restore from, especially for non-privileged users. * File system snapshots, which are useful for restoring deleted files or previous versions of files. * Copies of whole file systems or disks which are synchronized with another system on the network using a scheduled package:net/rsync[]. * Hardware or software RAID, which minimizes or avoids downtime when a disk fails. Typically, a mix of backup techniques is used. For example, one could create a schedule to automate a weekly, full system backup that is stored off-site and to supplement this backup with hourly ZFS snapshots. In addition, one could make a manual backup of individual directories or files before making file edits or deletions. This section describes some of the utilities which can be used to create and manage backups on a FreeBSD system. === File System Backups The traditional UNIX(R) programs for backing up a file system are man:dump[8], which creates the backup, and man:restore[8], which restores the backup. These utilities work at the disk block level, below the abstractions of the files, links, and directories that are created by file systems. Unlike other backup software, `dump` backs up an entire file system and is unable to backup only part of a file system or a directory tree that spans multiple file systems. Instead of writing files and directories, `dump` writes the raw data blocks that comprise files and directories. [NOTE] ==== If `dump` is used on the root directory, it will not back up [.filename]#/home#, [.filename]#/usr# or many other directories since these are typically mount points for other file systems or symbolic links into those file systems. ==== When used to restore data, `restore` stores temporary files in [.filename]#/tmp/# by default. When using a recovery disk with a small [.filename]#/tmp#, set `TMPDIR` to a directory with more free space in order for the restore to succeed. When using `dump`, be aware that some quirks remain from its early days in Version 6 of AT&T UNIX(R),circa 1975. The default parameters assume a backup to a 9-track tape, rather than to another type of media or to the high-density tapes available today. These defaults must be overridden on the command line. It is possible to backup a file system across the network to a another system or to a tape drive attached to another computer. While the man:rdump[8] and man:rrestore[8] utilities can be used for this purpose, they are not considered to be secure. Instead, one can use `dump` and `restore` in a more secure fashion over an SSH connection. This example creates a full, compressed backup of [.filename]#/usr# and sends the backup file to the specified host over a SSH connection. .Using `dump` over ssh [example] ==== [source,shell] .... # /sbin/dump -0uan -f - /usr | gzip -2 | ssh -c blowfish \ targetuser@targetmachine.example.com dd of=/mybigfiles/dump-usr-l0.gz .... ==== This example sets `RSH` in order to write the backup to a tape drive on a remote system over a SSH connection: .Using `dump` over ssh with `RSH` Set [example] ==== [source,shell] .... # env RSH=/usr/bin/ssh /sbin/dump -0uan -f targetuser@targetmachine.example.com:/dev/sa0 /usr .... ==== === Directory Backups Several built-in utilities are available for backing up and restoring specified files and directories as needed. A good choice for making a backup of all of the files in a directory is man:tar[1]. This utility dates back to Version 6 of AT&T UNIX(R) and by default assumes a recursive backup to a local tape device. Switches can be used to instead specify the name of a backup file. This example creates a compressed backup of the current directory and saves it to [.filename]#/tmp/mybackup.tgz#. When creating a backup file, make sure that the backup is not saved to the same directory that is being backed up. .Backing Up the Current Directory with `tar` [example] ==== [source,shell] .... # tar czvf /tmp/mybackup.tgz . .... ==== To restore the entire backup, `cd` into the directory to restore into and specify the name of the backup. Note that this will overwrite any newer versions of files in the restore directory. When in doubt, restore to a temporary directory or specify the name of the file within the backup to restore. .Restoring Up the Current Directory with `tar` [example] ==== [source,shell] .... # tar xzvf /tmp/mybackup.tgz .... ==== There are dozens of available switches which are described in man:tar[1]. This utility also supports the use of exclude patterns to specify which files should not be included when backing up the specified directory or restoring files from a backup. To create a backup using a specified list of files and directories, man:cpio[1] is a good choice. Unlike `tar`, `cpio` does not know how to walk the directory tree and it must be provided the list of files to backup. For example, a list of files can be created using `ls` or `find`. This example creates a recursive listing of the current directory which is then piped to `cpio` in order to create an output backup file named [.filename]#/tmp/mybackup.cpio#. .Using `ls` and `cpio` to Make a Recursive Backup of the Current Directory [example] ==== [source,shell] .... # ls -R | cpio -ovF /tmp/mybackup.cpio .... ==== A backup utility which tries to bridge the features provided by `tar` and `cpio` is man:pax[1]. Over the years, the various versions of `tar` and `cpio` became slightly incompatible. POSIX(R) created `pax` which attempts to read and write many of the various `cpio` and `tar` formats, plus new formats of its own. The `pax` equivalent to the previous examples would be: .Backing Up the Current Directory with `pax` [example] ==== [source,shell] .... # pax -wf /tmp/mybackup.pax . .... ==== [[backups-tapebackups]] === Using Data Tapes for Backups While tape technology has continued to evolve, modern backup systems tend to combine off-site backups with local removable media. FreeBSD supports any tape drive that uses SCSI, such as LTO or DAT. There is limited support for SATA and USB tape drives. For SCSI tape devices, FreeBSD uses the man:sa[4] driver and the [.filename]#/dev/sa0#, [.filename]#/dev/nsa0#, and [.filename]#/dev/esa0# devices. The physical device name is [.filename]#/dev/sa0#. When [.filename]#/dev/nsa0# is used, the backup application will not rewind the tape after writing a file, which allows writing more than one file to a tape. Using [.filename]#/dev/esa0# ejects the tape after the device is closed. In FreeBSD, `mt` is used to control operations of the tape drive, such as seeking through files on a tape or writing tape control marks to the tape. For example, the first three files on a tape can be preserved by skipping past them before writing a new file: [source,shell] .... # mt -f /dev/nsa0 fsf 3 .... This utility supports many operations. Refer to man:mt[1] for details. To write a single file to tape using `tar`, specify the name of the tape device and the file to backup: [source,shell] .... # tar cvf /dev/sa0 file .... To recover files from a `tar` archive on tape into the current directory: [source,shell] .... # tar xvf /dev/sa0 .... To backup a UFS file system, use `dump`. This examples backs up [.filename]#/usr# without rewinding the tape when finished: [source,shell] .... # dump -0aL -b64 -f /dev/nsa0 /usr .... To interactively restore files from a `dump` file on tape into the current directory: [source,shell] .... # restore -i -f /dev/nsa0 .... [[backups-programs-amanda]] === Third-Party Backup Utilities The FreeBSD Ports Collection provides many third-party utilities which can be used to schedule the creation of backups, simplify tape backup, and make backups easier and more convenient. Many of these applications are client/server based and can be used to automate the backups of a single system or all of the computers in a network. Popular utilities include Amanda, Bacula, rsync, and duplicity. === Emergency Recovery In addition to regular backups, it is recommended to perform the following steps as part of an emergency preparedness plan. Create a print copy of the output of the following commands: * `gpart show` * `more /etc/fstab` * `dmesg` Store this printout and a copy of the installation media in a secure location. Should an emergency restore be needed, boot into the installation media and select `Live CD` to access a rescue shell. This rescue mode can be used to view the current state of the system, and if needed, to reformat disks and restore data from backups. [NOTE] ==== The installation media for FreeBSD/i386 {rel112-current}-RELEASE does not include a rescue shell. For this version, instead download and burn a Livefs CD image from link:ftp://ftp.FreeBSD.org/pub/FreeBSD/releases/i386/ISO-IMAGES/{rel112-current}/FreeBSD-{rel112-current}-RELEASE-i386-livefs.iso[ftp://ftp.FreeBSD.org/pub/FreeBSD/releases/i386/ISO-IMAGES/{rel112-current}/FreeBSD-{rel112-current}-RELEASE-i386-livefs.iso]. ==== Next, test the rescue shell and the backups. Make notes of the procedure. Store these notes with the media, the printouts, and the backups. These notes may prevent the inadvertent destruction of the backups while under the stress of performing an emergency recovery. For an added measure of security, store the latest backup at a remote location which is physically separated from the computers and disk drives by a significant distance. [[disks-virtual]] == Memory Disks In addition to physical disks, FreeBSD also supports the creation and use of memory disks. One possible use for a memory disk is to access the contents of an ISO file system without the overhead of first burning it to a CD or DVD, then mounting the CD/DVD media. In FreeBSD, the man:md[4] driver is used to provide support for memory disks. The [.filename]#GENERIC# kernel includes this driver. When using a custom kernel configuration file, ensure it includes this line: [.programlisting] .... device md .... [[disks-mdconfig]] === Attaching and Detaching Existing Images To mount an existing file system image, use `mdconfig` to specify the name of the ISO file and a free unit number. Then, refer to that unit number to mount it on an existing mount point. Once mounted, the files in the ISO will appear in the mount point. This example attaches _diskimage.iso_ to the memory device [.filename]#/dev/md0# then mounts that memory device on [.filename]#/mnt#: [source,shell] .... # mdconfig -f diskimage.iso -u 0 # mount -t cd9660 /dev/md0 /mnt .... Notice that `-t cd9660` was used to mount an ISO format. If a unit number is not specified with `-u`, `mdconfig` will automatically allocate an unused memory device and output the name of the allocated unit, such as [.filename]#md4#. Refer to man:mdconfig[8] for more details about this command and its options. When a memory disk is no longer in use, its resources should be released back to the system. First, unmount the file system, then use `mdconfig` to detach the disk from the system and release its resources. To continue this example: [source,shell] .... # umount /mnt # mdconfig -d -u 0 .... To determine if any memory disks are still attached to the system, type `mdconfig -l`. [[disks-md-freebsd5]] === Creating a File- or Memory-Backed Memory Disk FreeBSD also supports memory disks where the storage to use is allocated from either a hard disk or an area of memory. The first method is commonly referred to as a file-backed file system and the second method as a memory-backed file system. Both types can be created using `mdconfig`. To create a new memory-backed file system, specify a type of `swap` and the size of the memory disk to create. Then, format the memory disk with a file system and mount as usual. This example creates a 5M memory disk on unit `1`. That memory disk is then formatted with the UFS file system before it is mounted: [source,shell] .... # mdconfig -a -t swap -s 5m -u 1 # newfs -U md1 /dev/md1: 5.0MB (10240 sectors) block size 16384, fragment size 2048 using 4 cylinder groups of 1.27MB, 81 blks, 192 inodes. with soft updates super-block backups (for fsck -b #) at: 160, 2752, 5344, 7936 # mount /dev/md1 /mnt # df /mnt Filesystem 1K-blocks Used Avail Capacity Mounted on /dev/md1 4718 4 4338 0% /mnt .... To create a new file-backed memory disk, first allocate an area of disk to use. This example creates an empty 5MB file named [.filename]#newimage#: [source,shell] .... # dd if=/dev/zero of=newimage bs=1k count=5k 5120+0 records in 5120+0 records out .... Next, attach that file to a memory disk, label the memory disk and format it with the UFS file system, mount the memory disk, and verify the size of the file-backed disk: [source,shell] .... # mdconfig -f newimage -u 0 # bsdlabel -w md0 auto # newfs -U md0a /dev/md0a: 5.0MB (10224 sectors) block size 16384, fragment size 2048 using 4 cylinder groups of 1.25MB, 80 blks, 192 inodes. super-block backups (for fsck -b #) at: 160, 2720, 5280, 7840 # mount /dev/md0a /mnt # df /mnt Filesystem 1K-blocks Used Avail Capacity Mounted on /dev/md0a 4710 4 4330 0% /mnt .... It takes several commands to create a file- or memory-backed file system using `mdconfig`. FreeBSD also comes with `mdmfs` which automatically configures a memory disk, formats it with the UFS file system, and mounts it. For example, after creating _newimage_ with `dd`, this one command is equivalent to running the `bsdlabel`, `newfs`, and `mount` commands shown above: [source,shell] .... # mdmfs -F newimage -s 5m md0 /mnt .... To instead create a new memory-based memory disk with `mdmfs`, use this one command: [source,shell] .... # mdmfs -s 5m md1 /mnt .... If the unit number is not specified, `mdmfs` will automatically select an unused memory device. For more details about `mdmfs`, refer to man:mdmfs[8]. [[snapshots]] == File System Snapshots FreeBSD offers a feature in conjunction with crossref:config[soft-updates,Soft Updates]: file system snapshots. UFS snapshots allow a user to create images of specified file systems, and treat them as a file. Snapshot files must be created in the file system that the action is performed on, and a user may create no more than 20 snapshots per file system. Active snapshots are recorded in the superblock so they are persistent across unmount and remount operations along with system reboots. When a snapshot is no longer required, it can be removed using man:rm[1]. While snapshots may be removed in any order, all the used space may not be acquired because another snapshot will possibly claim some of the released blocks. The un-alterable `snapshot` file flag is set by man:mksnap_ffs[8] after initial creation of a snapshot file. man:unlink[1] makes an exception for snapshot files since it allows them to be removed. Snapshots are created using man:mount[8]. To place a snapshot of [.filename]#/var# in the file [.filename]#/var/snapshot/snap#, use the following command: [source,shell] .... # mount -u -o snapshot /var/snapshot/snap /var .... Alternatively, use man:mksnap_ffs[8] to create the snapshot: [source,shell] .... # mksnap_ffs /var /var/snapshot/snap .... One can find snapshot files on a file system, such as [.filename]#/var#, using man:find[1]: [source,shell] .... # find /var -flags snapshot .... Once a snapshot has been created, it has several uses: * Some administrators will use a snapshot file for backup purposes, because the snapshot can be transferred to CDs or tape. * The file system integrity checker, man:fsck[8], may be run on the snapshot. Assuming that the file system was clean when it was mounted, this should always provide a clean and unchanging result. * Running man:dump[8] on the snapshot will produce a dump file that is consistent with the file system and the timestamp of the snapshot. man:dump[8] can also take a snapshot, create a dump image, and then remove the snapshot in one command by using `-L`. * The snapshot can be mounted as a frozen image of the file system. To man:mount[8] the snapshot [.filename]#/var/snapshot/snap# run: + [source,shell] .... # mdconfig -a -t vnode -o readonly -f /var/snapshot/snap -u 4 # mount -r /dev/md4 /mnt .... The frozen [.filename]#/var# is now available through [.filename]#/mnt#. Everything will initially be in the same state it was during the snapshot creation time. The only exception is that any earlier snapshots will appear as zero length files. To unmount the snapshot, use: [source,shell] .... # umount /mnt # mdconfig -d -u 4 .... For more information about `softupdates` and file system snapshots, including technical papers, visit Marshall Kirk McKusick's website at http://www.mckusick.com/[http://www.mckusick.com/]. [[quotas]] == Disk Quotas Disk quotas can be used to limit the amount of disk space or the number of files a user or members of a group may allocate on a per-file system basis. This prevents one user or group of users from consuming all of the available disk space. This section describes how to configure disk quotas for the UFS file system. To configure quotas on the ZFS file system, refer to crossref:zfs[zfs-zfs-quota,"Dataset, User, and Group Quotas"] === Enabling Disk Quotas To determine if the FreeBSD kernel provides support for disk quotas: [source,shell] .... % sysctl kern.features.ufs_quota kern.features.ufs_quota: 1 .... In this example, the `1` indicates quota support. If the value is instead `0`, add the following line to a custom kernel configuration file and rebuild the kernel using the instructions in crossref:kernelconfig[kernelconfig,Configuring the FreeBSD Kernel]: [.programlisting] .... options QUOTA .... Next, enable disk quotas in [.filename]#/etc/rc.conf#: [.programlisting] .... quota_enable="YES" .... Normally on bootup, the quota integrity of each file system is checked by man:quotacheck[8]. This program insures that the data in the quota database properly reflects the data on the file system. This is a time consuming process that will significantly affect the time the system takes to boot. To skip this step, add this variable to [.filename]#/etc/rc.conf#: [.programlisting] .... check_quotas="NO" .... Finally, edit [.filename]#/etc/fstab# to enable disk quotas on a per-file system basis. To enable per-user quotas on a file system, add `userquota` to the options field in the [.filename]#/etc/fstab# entry for the file system to enable quotas on. For example: [.programlisting] .... /dev/da1s2g /home ufs rw,userquota 1 2 .... To enable group quotas, use `groupquota` instead. To enable both user and group quotas, separate the options with a comma: [.programlisting] .... /dev/da1s2g /home ufs rw,userquota,groupquota 1 2 .... By default, quota files are stored in the root directory of the file system as [.filename]#quota.user# and [.filename]#quota.group#. Refer to man:fstab[5] for more information. Specifying an alternate location for the quota files is not recommended. Once the configuration is complete, reboot the system and [.filename]#/etc/rc# will automatically run the appropriate commands to create the initial quota files for all of the quotas enabled in [.filename]#/etc/fstab#. In the normal course of operations, there should be no need to manually run man:quotacheck[8], man:quotaon[8], or man:quotaoff[8]. However, one should read these manual pages to be familiar with their operation. === Setting Quota Limits To verify that quotas are enabled, run: [source,shell] .... # quota -v .... There should be a one line summary of disk usage and current quota limits for each file system that quotas are enabled on. The system is now ready to be assigned quota limits with `edquota`. Several options are available to enforce limits on the amount of disk space a user or group may allocate, and how many files they may create. Allocations can be limited based on disk space (block quotas), number of files (inode quotas), or a combination of both. Each limit is further broken down into two categories: hard and soft limits. A hard limit may not be exceeded. Once a user reaches a hard limit, no further allocations can be made on that file system by that user. For example, if the user has a hard limit of 500 kbytes on a file system and is currently using 490 kbytes, the user can only allocate an additional 10 kbytes. Attempting to allocate an additional 11 kbytes will fail. Soft limits can be exceeded for a limited amount of time, known as the grace period, which is one week by default. If a user stays over their limit longer than the grace period, the soft limit turns into a hard limit and no further allocations are allowed. When the user drops back below the soft limit, the grace period is reset. In the following example, the quota for the `test` account is being edited. When `edquota` is invoked, the editor specified by `EDITOR` is opened in order to edit the quota limits. The default editor is set to vi. [source,shell] .... # edquota -u test Quotas for user test: /usr: kbytes in use: 65, limits (soft = 50, hard = 75) inodes in use: 7, limits (soft = 50, hard = 60) /usr/var: kbytes in use: 0, limits (soft = 50, hard = 75) inodes in use: 0, limits (soft = 50, hard = 60) .... There are normally two lines for each file system that has quotas enabled. One line represents the block limits and the other represents the inode limits. Change the value to modify the quota limit. For example, to raise the block limit on [.filename]#/usr# to a soft limit of `500` and a hard limit of `600`, change the values in that line as follows: [.programlisting] .... /usr: kbytes in use: 65, limits (soft = 500, hard = 600) .... The new quota limits take effect upon exiting the editor. Sometimes it is desirable to set quota limits on a range of users. This can be done by first assigning the desired quota limit to a user. Then, use `-p` to duplicate that quota to a specified range of user IDs (UIDs). The following command will duplicate those quota limits for UIDs `10,000` through `19,999`: [source,shell] .... # edquota -p test 10000-19999 .... For more information, refer to man:edquota[8]. === Checking Quota Limits and Disk Usage To check individual user or group quotas and disk usage, use man:quota[1]. A user may only examine their own quota and the quota of a group they are a member of. Only the superuser may view all user and group quotas. To get a summary of all quotas and disk usage for file systems with quotas enabled, use man:repquota[8]. Normally, file systems that the user is not using any disk space on will not show in the output of `quota`, even if the user has a quota limit assigned for that file system. Use `-v` to display those file systems. The following is sample output from `quota -v` for a user that has quota limits on two file systems. [.programlisting] .... Disk quotas for user test (uid 1002): Filesystem usage quota limit grace files quota limit grace /usr 65* 50 75 5days 7 50 60 /usr/var 0 50 75 0 50 60 .... In this example, the user is currently 15 kbytes over the soft limit of 50 kbytes on [.filename]#/usr# and has 5 days of grace period left. The asterisk `*` indicates that the user is currently over the quota limit. === Quotas over NFS Quotas are enforced by the quota subsystem on the NFS server. The man:rpc.rquotad[8] daemon makes quota information available to `quota` on NFS clients, allowing users on those machines to see their quota statistics. On the NFS server, enable `rpc.rquotad` by removing the `#` from this line in [.filename]*/etc/inetd.conf*: [.programlisting] .... rquotad/1 dgram rpc/udp wait root /usr/libexec/rpc.rquotad rpc.rquotad .... Then, restart `inetd`: [source,shell] .... # service inetd restart .... [[disks-encrypting]] == Encrypting Disk Partitions FreeBSD offers excellent online protections against unauthorized data access. File permissions and crossref:mac[mac,Mandatory Access Control] (MAC) help prevent unauthorized users from accessing data while the operating system is active and the computer is powered up. However, the permissions enforced by the operating system are irrelevant if an attacker has physical access to a computer and can move the computer's hard drive to another system to copy and analyze the data. Regardless of how an attacker may have come into possession of a hard drive or powered-down computer, the GEOM-based cryptographic subsystems built into FreeBSD are able to protect the data on the computer's file systems against even highly-motivated attackers with significant resources. Unlike encryption methods that encrypt individual files, the built-in `gbde` and `geli` utilities can be used to transparently encrypt entire file systems. No cleartext ever touches the hard drive's platter. This chapter demonstrates how to create an encrypted file system on FreeBSD. It first demonstrates the process using `gbde` and then demonstrates the same example using `geli`. === Disk Encryption with gbde The objective of the man:gbde[4] facility is to provide a formidable challenge for an attacker to gain access to the contents of a _cold_ storage device. However, if the computer is compromised while up and running and the storage device is actively attached, or the attacker has access to a valid passphrase, it offers no protection to the contents of the storage device. Thus, it is important to provide physical security while the system is running and to protect the passphrase used by the encryption mechanism. This facility provides several barriers to protect the data stored in each disk sector. It encrypts the contents of a disk sector using 128-bit AES in CBC mode. Each sector on the disk is encrypted with a different AES key. For more information on the cryptographic design, including how the sector keys are derived from the user-supplied passphrase, refer to man:gbde[4]. FreeBSD provides a kernel module for gbde which can be loaded with this command: [source,shell] .... # kldload geom_bde .... If using a custom kernel configuration file, ensure it contains this line: `options GEOM_BDE` The following example demonstrates adding a new hard drive to a system that will hold a single encrypted partition that will be mounted as [.filename]#/private#. [.procedure] .Procedure: Encrypting a Partition with gbde . Add the New Hard Drive + Install the new drive to the system as explained in <>. For the purposes of this example, a new hard drive partition has been added as [.filename]#/dev/ad4s1c# and [.filename]#/dev/ad0s1*# represents the existing standard FreeBSD partitions. + [source,shell] .... # ls /dev/ad* /dev/ad0 /dev/ad0s1b /dev/ad0s1e /dev/ad4s1 /dev/ad0s1 /dev/ad0s1c /dev/ad0s1f /dev/ad4s1c /dev/ad0s1a /dev/ad0s1d /dev/ad4 .... . Create a Directory to Hold `gbde` Lock Files + [source,shell] .... # mkdir /etc/gbde .... + The gbde lock file contains information that gbde requires to access encrypted partitions. Without access to the lock file, gbde will not be able to decrypt the data contained in the encrypted partition without significant manual intervention which is not supported by the software. Each encrypted partition uses a separate lock file. . Initialize the `gbde` Partition + A gbde partition must be initialized before it can be used. This initialization needs to be performed only once. This command will open the default editor, in order to set various configuration options in a template. For use with the UFS file system, set the sector_size to 2048: + [source,shell] .... # gbde init /dev/ad4s1c -i -L /etc/gbde/ad4s1c.lock # $FreeBSD: src/sbin/gbde/template.txt,v 1.1.36.1 2009/08/03 08:13:06 kensmith Exp $ # # Sector size is the smallest unit of data which can be read or written. # Making it too small decreases performance and decreases available space. # Making it too large may prevent filesystems from working. 512 is the # minimum and always safe. For UFS, use the fragment size # sector_size = 2048 [...] .... + Once the edit is saved, the user will be asked twice to type the passphrase used to secure the data. The passphrase must be the same both times. The ability of gbde to protect data depends entirely on the quality of the passphrase. For tips on how to select a secure passphrase that is easy to remember, see http://world.std.com/\~reinhold/diceware.html[http://world.std.com/~reinhold/diceware.htm]. + This initialization creates a lock file for the gbde partition. In this example, it is stored as [.filename]#/etc/gbde/ad4s1c.lock#. Lock files must end in ".lock" in order to be correctly detected by the [.filename]#/etc/rc.d/gbde# start up script. + [CAUTION] ==== Lock files _must_ be backed up together with the contents of any encrypted partitions. Without the lock file, the legitimate owner will be unable to access the data on the encrypted partition. ==== . Attach the Encrypted Partition to the Kernel + [source,shell] .... # gbde attach /dev/ad4s1c -l /etc/gbde/ad4s1c.lock .... + This command will prompt to input the passphrase that was selected during the initialization of the encrypted partition. The new encrypted device will appear in [.filename]#/dev# as [.filename]#/dev/device_name.bde#: + [source,shell] .... # ls /dev/ad* /dev/ad0 /dev/ad0s1b /dev/ad0s1e /dev/ad4s1 /dev/ad0s1 /dev/ad0s1c /dev/ad0s1f /dev/ad4s1c /dev/ad0s1a /dev/ad0s1d /dev/ad4 /dev/ad4s1c.bde .... . Create a File System on the Encrypted Device + Once the encrypted device has been attached to the kernel, a file system can be created on the device. This example creates a UFS file system with soft updates enabled. Be sure to specify the partition which has a [.filename]#*.bde# extension: + [source,shell] .... # newfs -U /dev/ad4s1c.bde .... . Mount the Encrypted Partition + Create a mount point and mount the encrypted file system: + [source,shell] .... # mkdir /private # mount /dev/ad4s1c.bde /private .... . Verify That the Encrypted File System is Available + The encrypted file system should now be visible and available for use: + [source,shell] .... % df -H Filesystem Size Used Avail Capacity Mounted on /dev/ad0s1a 1037M 72M 883M 8% / /devfs 1.0K 1.0K 0B 100% /dev /dev/ad0s1f 8.1G 55K 7.5G 0% /home /dev/ad0s1e 1037M 1.1M 953M 0% /tmp /dev/ad0s1d 6.1G 1.9G 3.7G 35% /usr /dev/ad4s1c.bde 150G 4.1K 138G 0% /private .... After each boot, any encrypted file systems must be manually re-attached to the kernel, checked for errors, and mounted, before the file systems can be used. To configure these steps, add the following lines to [.filename]#/etc/rc.conf#: [.programlisting] .... gbde_autoattach_all="YES" gbde_devices="ad4s1c" gbde_lockdir="/etc/gbde" .... This requires that the passphrase be entered at the console at boot time. After typing the correct passphrase, the encrypted partition will be mounted automatically. Additional gbde boot options are available and listed in man:rc.conf[5]. [NOTE] ==== sysinstall is incompatible with gbde-encrypted devices. All [.filename]#*.bde# devices must be detached from the kernel before starting sysinstall or it will crash during its initial probing for devices. To detach the encrypted device used in the example, use the following command: [source,shell] .... # gbde detach /dev/ad4s1c .... ==== [[disks-encrypting-geli]] === Disk Encryption with `geli` An alternative cryptographic GEOM class is available using `geli`. This control utility adds some features and uses a different scheme for doing cryptographic work. It provides the following features: * Utilizes the man:crypto[9] framework and automatically uses cryptographic hardware when it is available. * Supports multiple cryptographic algorithms such as AES, Blowfish, and 3DES. * Allows the root partition to be encrypted. The passphrase used to access the encrypted root partition will be requested during system boot. * Allows the use of two independent keys. * It is fast as it performs simple sector-to-sector encryption. * Allows backup and restore of master keys. If a user destroys their keys, it is still possible to get access to the data by restoring keys from the backup. * Allows a disk to attach with a random, one-time key which is useful for swap partitions and temporary file systems. More features and usage examples can be found in man:geli[8]. The following example describes how to generate a key file which will be used as part of the master key for the encrypted provider mounted under [.filename]#/private#. The key file will provide some random data used to encrypt the master key. The master key will also be protected by a passphrase. The provider's sector size will be 4kB. The example describes how to attach to the `geli` provider, create a file system on it, mount it, work with it, and finally, how to detach it. [.procedure] .Procedure: Encrypting a Partition with `geli` . Load `geli` Support + Support for `geli` is available as a loadable kernel module. To configure the system to automatically load the module at boot time, add the following line to [.filename]#/boot/loader.conf#: + [.programlisting] .... geom_eli_load="YES" .... + To load the kernel module now: + [source,shell] .... # kldload geom_eli .... + For a custom kernel, ensure the kernel configuration file contains these lines: + [.programlisting] .... options GEOM_ELI device crypto .... . Generate the Master Key + The following commands generate a master key that all data will be encrypted with. This key can never be changed. Rather than using it directly, it is encrypted with one or more user keys. The user keys are made up of an optional combination of random bytes from a file, [.filename]#/root/da2.key#, and/or a passphrase. In this case, the data source for the key file is [.filename]#/dev/random#. This command also configures the sector size of the provider ([.filename]#/dev/da2.eli#) as 4kB, for better performance: + [source,shell] .... # dd if=/dev/random of=/root/da2.key bs=64 count=1 # geli init -K /root/da2.key -s 4096 /dev/da2 Enter new passphrase: Reenter new passphrase: .... + It is not mandatory to use both a passphrase and a key file as either method of securing the master key can be used in isolation. + If the key file is given as "-", standard input will be used. For example, this command generates three key files: + [source,shell] .... # cat keyfile1 keyfile2 keyfile3 | geli init -K - /dev/da2 .... . Attach the Provider with the Generated Key + To attach the provider, specify the key file, the name of the disk, and the passphrase: + [source,shell] .... # geli attach -k /root/da2.key /dev/da2 Enter passphrase: .... + This creates a new device with an [.filename]#.eli# extension: + [source,shell] .... # ls /dev/da2* /dev/da2 /dev/da2.eli .... . Create the New File System + Next, format the device with the UFS file system and mount it on an existing mount point: + [source,shell] .... # dd if=/dev/random of=/dev/da2.eli bs=1m # newfs /dev/da2.eli # mount /dev/da2.eli /private .... + The encrypted file system should now be available for use: + [source,shell] .... # df -H Filesystem Size Used Avail Capacity Mounted on /dev/ad0s1a 248M 89M 139M 38% / /devfs 1.0K 1.0K 0B 100% /dev /dev/ad0s1f 7.7G 2.3G 4.9G 32% /usr /dev/ad0s1d 989M 1.5M 909M 0% /tmp /dev/ad0s1e 3.9G 1.3G 2.3G 35% /var /dev/da2.eli 150G 4.1K 138G 0% /private .... Once the work on the encrypted partition is done, and the [.filename]#/private# partition is no longer needed, it is prudent to put the device into cold storage by unmounting and detaching the `geli` encrypted partition from the kernel: [source,shell] .... # umount /private # geli detach da2.eli .... An [.filename]#rc.d# script is provided to simplify the mounting of `geli`-encrypted devices at boot time. For this example, add these lines to [.filename]#/etc/rc.conf#: [.programlisting] .... geli_devices="da2" geli_da2_flags="-k /root/da2.key" .... This configures [.filename]#/dev/da2# as a `geli` provider with a master key of [.filename]#/root/da2.key#. The system will automatically detach the provider from the kernel before the system shuts down. During the startup process, the script will prompt for the passphrase before attaching the provider. Other kernel messages might be shown before and after the password prompt. If the boot process seems to stall, look carefully for the password prompt among the other messages. Once the correct passphrase is entered, the provider is attached. The file system is then mounted, typically by an entry in [.filename]#/etc/fstab#. Refer to crossref:basics[mount-unmount,“Mounting and Unmounting File Systems”] for instructions on how to configure a file system to mount at boot time. [[swap-encrypting]] == Encrypting Swap Like the encryption of disk partitions, encryption of swap space is used to protect sensitive information. Consider an application that deals with passwords. As long as these passwords stay in physical memory, they are not written to disk and will be cleared after a reboot. However, if FreeBSD starts swapping out memory pages to free space, the passwords may be written to the disk unencrypted. Encrypting swap space can be a solution for this scenario. This section demonstrates how to configure an encrypted swap partition using man:gbde[8] or man:geli[8] encryption. It assumes that [.filename]#/dev/ada0s1b# is the swap partition. === Configuring Encrypted Swap Swap partitions are not encrypted by default and should be cleared of any sensitive data before continuing. To overwrite the current swap partition with random garbage, execute the following command: [source,shell] .... # dd if=/dev/random of=/dev/ada0s1b bs=1m .... To encrypt the swap partition using man:gbde[8], add the `.bde` suffix to the swap line in [.filename]#/etc/fstab#: [.programlisting] .... # Device Mountpoint FStype Options Dump Pass# /dev/ada0s1b.bde none swap sw 0 0 .... To instead encrypt the swap partition using man:geli[8], use the `.eli` suffix: [.programlisting] .... # Device Mountpoint FStype Options Dump Pass# /dev/ada0s1b.eli none swap sw 0 0 .... By default, man:geli[8] uses the AES algorithm with a key length of 128 bits. Normally the default settings will suffice. If desired, these defaults can be altered in the options field in [.filename]#/etc/fstab#. The possible flags are: aalgo:: Data integrity verification algorithm used to ensure that the encrypted data has not been tampered with. See man:geli[8] for a list of supported algorithms. ealgo:: Encryption algorithm used to protect the data. See man:geli[8] for a list of supported algorithms. keylen:: The length of the key used for the encryption algorithm. See man:geli[8] for the key lengths that are supported by each encryption algorithm. sectorsize:: The size of the blocks data is broken into before it is encrypted. Larger sector sizes increase performance at the cost of higher storage overhead. The recommended size is 4096 bytes. This example configures an encrypted swap partition using the Blowfish algorithm with a key length of 128 bits and a sectorsize of 4 kilobytes: [.programlisting] .... # Device Mountpoint FStype Options Dump Pass# /dev/ada0s1b.eli none swap sw,ealgo=blowfish,keylen=128,sectorsize=4096 0 0 .... === Encrypted Swap Verification Once the system has rebooted, proper operation of the encrypted swap can be verified using `swapinfo`. If man:gbde[8] is being used: [source,shell] .... % swapinfo Device 1K-blocks Used Avail Capacity /dev/ada0s1b.bde 542720 0 542720 0 .... If man:geli[8] is being used: [source,shell] .... % swapinfo Device 1K-blocks Used Avail Capacity /dev/ada0s1b.eli 542720 0 542720 0 .... [[disks-hast]] == Highly Available Storage (HAST) High availability is one of the main requirements in serious business applications and highly-available storage is a key component in such environments. In FreeBSD, the Highly Available STorage (HAST) framework allows transparent storage of the same data across several physically separated machines connected by a TCP/IP network. HAST can be understood as a network-based RAID1 (mirror), and is similar to the DRBD(R) storage system used in the GNU/Linux(R) platform. In combination with other high-availability features of FreeBSD like CARP, HAST makes it possible to build a highly-available storage cluster that is resistant to hardware failures. The following are the main features of HAST: * Can be used to mask I/O errors on local hard drives. * File system agnostic as it works with any file system supported by FreeBSD. * Efficient and quick resynchronization as only the blocks that were modified during the downtime of a node are synchronized. * Can be used in an already deployed environment to add additional redundancy. * Together with CARP, Heartbeat, or other tools, it can be used to build a robust and durable storage system. After reading this section, you will know: * What HAST is, how it works, and which features it provides. * How to set up and use HAST on FreeBSD. * How to integrate CARP and man:devd[8] to build a robust storage system. Before reading this section, you should: * Understand UNIX(R) and FreeBSD basics (crossref:basics[basics,FreeBSD Basics]). * Know how to configure network interfaces and other core FreeBSD subsystems (crossref:config[config-tuning,Configuration and Tuning]). * Have a good understanding of FreeBSD networking (crossref:partiv[network-communication,"Network Communication"]). The HAST project was sponsored by The FreeBSD Foundation with support from http://www.omc.net/[http://www.omc.net/] and http://www.transip.nl/[http://www.transip.nl/]. === HAST Operation -HAST provides synchronous block-level replication between two physical machines: the _primary_, also known as the _master_ node, and the _secondary_, or _slave_ node. +HAST provides synchronous block-level replication between two physical machines: the _primary_ nodeand the _secondary_ node. These two machines together are referred to as a cluster. Since HAST works in a primary-secondary configuration, it allows only one of the cluster nodes to be active at any given time. The primary node, also called _active_, is the one which will handle all the I/O requests to HAST-managed devices. The secondary node is automatically synchronized from the primary node. The physical components of the HAST system are the local disk on primary node, and the disk on the remote, secondary node. HAST operates synchronously on a block level, making it transparent to file systems and applications. HAST provides regular GEOM providers in [.filename]#/dev/hast/# for use by other tools or applications. There is no difference between using HAST-provided devices and raw disks or partitions. Each write, delete, or flush operation is sent to both the local disk and to the remote disk over TCP/IP. Each read operation is served from the local disk, unless the local disk is not up-to-date or an I/O error occurs. In such cases, the read operation is sent to the secondary node. HAST tries to provide fast failure recovery. For this reason, it is important to reduce synchronization time after a node's outage. To provide fast synchronization, HAST manages an on-disk bitmap of dirty extents and only synchronizes those during a regular synchronization, with an exception of the initial sync. There are many ways to handle synchronization. HAST implements several replication modes to handle different synchronization methods: * _memsync_: This mode reports a write operation as completed when the local write operation is finished and when the remote node acknowledges data arrival, but before actually storing the data. The data on the remote node will be stored directly after sending the acknowledgement. This mode is intended to reduce latency, but still provides good reliability. This mode is the default. * _fullsync_: This mode reports a write operation as completed when both the local write and the remote write complete. This is the safest and the slowest replication mode. * _async_: This mode reports a write operation as completed when the local write completes. This is the fastest and the most dangerous replication mode. It should only be used when replicating to a distant node where latency is too high for other modes. === HAST Configuration The HAST framework consists of several components: * The man:hastd[8] daemon which provides data synchronization. When this daemon is started, it will automatically load `geom_gate.ko`. * The userland management utility, man:hastctl[8]. * The man:hast.conf[5] configuration file. This file must exist before starting hastd. Users who prefer to statically build `GEOM_GATE` support into the kernel should add this line to the custom kernel configuration file, then rebuild the kernel using the instructions in crossref:kernelconfig[kernelconfig,Configuring the FreeBSD Kernel]: [.programlisting] .... options GEOM_GATE .... -The following example describes how to configure two nodes in master-slave/primary-secondary operation using HAST to replicate the data between the two. +The following example describes how to configure two nodes in primary-secondary operation using HAST to replicate the data between the two. The nodes will be called `hasta`, with an IP address of `172.16.0.1`, and `hastb`, with an IP address of `172.16.0.2`. Both nodes will have a dedicated hard drive [.filename]#/dev/ad6# of the same size for HAST operation. The HAST pool, sometimes referred to as a resource or the GEOM provider in [.filename]#/dev/hast/#, will be called `test`. Configuration of HAST is done using [.filename]#/etc/hast.conf#. This file should be identical on both nodes. The simplest configuration is: [.programlisting] .... resource test { on hasta { local /dev/ad6 remote 172.16.0.2 } on hastb { local /dev/ad6 remote 172.16.0.1 } } .... For more advanced configuration, refer to man:hast.conf[5]. [TIP] ==== It is also possible to use host names in the `remote` statements if the hosts are resolvable and defined either in [.filename]#/etc/hosts# or in the local DNS. ==== Once the configuration exists on both nodes, the HAST pool can be created. Run these commands on both nodes to place the initial metadata onto the local disk and to start man:hastd[8]: [source,shell] .... # hastctl create test # service hastd onestart .... [NOTE] ==== It is _not_ possible to use GEOM providers with an existing file system or to convert an existing storage to a HAST-managed pool. This procedure needs to store some metadata on the provider and there will not be enough required space available on an existing provider. ==== A HAST node's `primary` or `secondary` role is selected by an administrator, or software like Heartbeat, using man:hastctl[8]. On the primary node, `hasta`, issue this command: [source,shell] .... # hastctl role primary test .... Run this command on the secondary node, `hastb`: [source,shell] .... # hastctl role secondary test .... Verify the result by running `hastctl` on each node: [source,shell] .... # hastctl status test .... Check the `status` line in the output. If it says `degraded`, something is wrong with the configuration file. It should say `complete` on each node, meaning that the synchronization between the nodes has started. The synchronization completes when `hastctl status` reports 0 bytes of `dirty` extents. The next step is to create a file system on the GEOM provider and mount it. This must be done on the `primary` node. Creating the file system can take a few minutes, depending on the size of the hard drive. This example creates a UFS file system on [.filename]#/dev/hast/test#: [source,shell] .... # newfs -U /dev/hast/test # mkdir /hast/test # mount /dev/hast/test /hast/test .... Once the HAST framework is configured properly, the final step is to make sure that HAST is started automatically during system boot. Add this line to [.filename]#/etc/rc.conf#: [.programlisting] .... hastd_enable="YES" .... ==== Failover Configuration The goal of this example is to build a robust storage system which is resistant to the failure of any given node. If the primary node fails, the secondary node is there to take over seamlessly, check and mount the file system, and continue to work without missing a single bit of data. To accomplish this task, the Common Address Redundancy Protocol (CARP) is used to provide for automatic failover at the IP layer. CARP allows multiple hosts on the same network segment to share an IP address. Set up CARP on both nodes of the cluster according to the documentation available in crossref:advanced-networking[carp,“Common Address Redundancy Protocol (CARP)”]. In this example, each node will have its own management IP address and a shared IP address of _172.16.0.254_. -The primary HAST node of the cluster must be the master CARP node. +The primary HAST node of the cluster must be the primary CARP node. The HAST pool created in the previous section is now ready to be exported to the other hosts on the network. This can be accomplished by exporting it through NFS or Samba, using the shared IP address _172.16.0.254_. The only problem which remains unresolved is an automatic failover should the primary node fail. In the event of CARP interfaces going up or down, the FreeBSD operating system generates a man:devd[8] event, making it possible to watch for state changes on the CARP interfaces. A state change on the CARP interface is an indication that one of the nodes failed or came back online. These state change events make it possible to run a script which will automatically handle the HAST failover. To catch state changes on the CARP interfaces, add this configuration to [.filename]#/etc/devd.conf# on each node: [.programlisting] .... notify 30 { match "system" "IFNET"; match "subsystem" "carp0"; match "type" "LINK_UP"; - action "/usr/local/sbin/carp-hast-switch master"; + action "/usr/local/sbin/carp-hast-switch primary"; }; notify 30 { match "system" "IFNET"; match "subsystem" "carp0"; match "type" "LINK_DOWN"; - action "/usr/local/sbin/carp-hast-switch slave"; + action "/usr/local/sbin/carp-hast-switch secondary"; }; .... [NOTE] ==== If the systems are running FreeBSD 10 or higher, replace [.filename]#carp0# with the name of the CARP-configured interface. ==== Restart man:devd[8] on both nodes to put the new configuration into effect: [source,shell] .... # service devd restart .... When the specified interface state changes by going up or down , the system generates a notification, allowing the man:devd[8] subsystem to run the specified automatic failover script, [.filename]#/usr/local/sbin/carp-hast-switch#. For further clarification about this configuration, refer to man:devd.conf[5]. Here is an example of an automated failover script: [.programlisting] .... #!/bin/sh # Original script by Freddie Cash # Modified by Michael W. Lucas # and Viktor Petersson # The names of the HAST resources, as listed in /etc/hast.conf resources="test" -# delay in mounting HAST resource after becoming master +# delay in mounting HAST resource after becoming primary # make your best guess delay=3 # logging log="local0.debug" name="carp-hast" # end of user configurable stuff case "$1" in - master) + primary) logger -p $log -t $name "Switching to primary provider for ${resources}." sleep ${delay} # Wait for any "hastd secondary" processes to stop for disk in ${resources}; do while $( pgrep -lf "hastd: ${disk} \(secondary\)" > /dev/null 2>&1 ); do sleep 1 done # Switch role for each disk hastctl role primary ${disk} if [ $? -ne 0 ]; then logger -p $log -t $name "Unable to change role to primary for resource ${disk}." exit 1 fi done # Wait for the /dev/hast/* devices to appear for disk in ${resources}; do for I in $( jot 60 ); do [ -c "/dev/hast/${disk}" ] && break sleep 0.5 done if [ ! -c "/dev/hast/${disk}" ]; then logger -p $log -t $name "GEOM provider /dev/hast/${disk} did not appear." exit 1 fi done logger -p $log -t $name "Role for HAST resources ${resources} switched to primary." logger -p $log -t $name "Mounting disks." for disk in ${resources}; do mkdir -p /hast/${disk} fsck -p -y -t ufs /dev/hast/${disk} mount /dev/hast/${disk} /hast/${disk} done ;; - slave) + secondary) logger -p $log -t $name "Switching to secondary provider for ${resources}." # Switch roles for the HAST resources for disk in ${resources}; do if ! mount | grep -q "^/dev/hast/${disk} on " then else umount -f /hast/${disk} fi sleep $delay hastctl role secondary ${disk} 2>&1 if [ $? -ne 0 ]; then logger -p $log -t $name "Unable to switch role to secondary for resource ${disk}." exit 1 fi logger -p $log -t $name "Role switched to secondary for resource ${disk}." done ;; esac .... -In a nutshell, the script takes these actions when a node becomes master: +In a nutshell, the script takes these actions when a node becomes primary: * Promotes the HAST pool to primary on the other node. * Checks the file system under the HAST pool. * Mounts the pool. When a node becomes secondary: * Unmounts the HAST pool. * Degrades the HAST pool to secondary. [CAUTION] ==== This is just an example script which serves as a proof of concept. It does not handle all the possible scenarios and can be extended or altered in any way, for example, to start or stop required services. ==== [TIP] ==== For this example, a standard UFS file system was used. To reduce the time needed for recovery, a journal-enabled UFS or ZFS file system can be used instead. ==== More detailed information with additional examples can be found at http://wiki.FreeBSD.org/HAST[http://wiki.FreeBSD.org/HAST]. === Troubleshooting HAST should generally work without issues. However, as with any other software product, there may be times when it does not work as supposed. The sources of the problems may be different, but the rule of thumb is to ensure that the time is synchronized between the nodes of the cluster. When troubleshooting HAST, the debugging level of man:hastd[8] should be increased by starting `hastd` with `-d`. This argument may be specified multiple times to further increase the debugging level. Consider also using `-F`, which starts `hastd` in the foreground. [[disks-hast-sb]] ==== Recovering from the Split-brain Condition _Split-brain_ occurs when the nodes of the cluster are unable to communicate with each other, and both are configured as primary. This is a dangerous condition because it allows both nodes to make incompatible changes to the data. This problem must be corrected manually by the system administrator. The administrator must either decide which node has more important changes, or perform the merge manually. Then, let HAST perform full synchronization of the node which has the broken data. To do this, issue these commands on the node which needs to be resynchronized: [source,shell] .... # hastctl role init test # hastctl create test # hastctl role secondary test .... diff --git a/documentation/content/en/books/handbook/firewalls/_index.adoc b/documentation/content/en/books/handbook/firewalls/_index.adoc index 47429914bc..a70334c309 100644 --- a/documentation/content/en/books/handbook/firewalls/_index.adoc +++ b/documentation/content/en/books/handbook/firewalls/_index.adoc @@ -1,2681 +1,2680 @@ --- title: Chapter 31. Firewalls part: IV. Network Communication prev: books/handbook/network-servers next: books/handbook/advanced-networking description: "FreeBSD has three firewalls built into the base system: PF, IPFW, and IPFILTER. This chapter covers how to define packet filtering rules, the differences between the firewalls built into FreeBSD and how to use them" tags: ["firewall", "pf", "ipfw", "ipfilter", "blacklistd", "filtering"] --- [[firewalls]] = Firewalls :doctype: book :toc: macro :toclevels: 1 :icons: font :sectnums: :sectnumlevels: 6 :source-highlighter: rouge :experimental: :skip-front-matter: :xrefstyle: basic :relfileprefix: ../ :outfilesuffix: :sectnumoffset: 31 ifeval::["{backend}" == "html5"] :imagesdir: ../../../../images/books/handbook/firewalls/ endif::[] ifeval::["{backend}" == "pdf"] :imagesdir: ../../../../static/images/books/handbook/firewalls/ endif::[] ifeval::["{backend}" == "epub3"] :imagesdir: ../../../../static/images/books/handbook/firewalls/ endif::[] include::shared/authors.adoc[] include::shared/releases.adoc[] include::shared/en/mailing-lists.adoc[] include::shared/en/teams.adoc[] include::shared/en/urls.adoc[] toc::[] [[firewalls-intro]] == Synopsis Firewalls make it possible to filter the incoming and outgoing traffic that flows through a system. A firewall can use one or more sets of "rules" to inspect network packets as they come in or go out of network connections and either allows the traffic through or blocks it. The rules of a firewall can inspect one or more characteristics of the packets such as the protocol type, source or destination host address, and source or destination port. Firewalls can enhance the security of a host or a network. They can be used to do one or more of the following: * Protect and insulate the applications, services, and machines of an internal network from unwanted traffic from the public Internet. * Limit or disable access from hosts of the internal network to services of the public Internet. * Support network address translation (NAT), which allows an internal network to use private IP addresses and share a single connection to the public Internet using either a single IP address or a shared pool of automatically assigned public addresses. FreeBSD has three firewalls built into the base system: PF, IPFW, and IPFILTER, also known as IPF. FreeBSD also provides two traffic shapers for controlling bandwidth usage: man:altq[4] and man:dummynet[4]. ALTQ has traditionally been closely tied with PF and dummynet with IPFW. Each firewall uses rules to control the access of packets to and from a FreeBSD system, although they go about it in different ways and each has a different rule syntax. FreeBSD provides multiple firewalls in order to meet the different requirements and preferences for a wide variety of users. Each user should evaluate which firewall best meets their needs. After reading this chapter, you will know: * How to define packet filtering rules. * The differences between the firewalls built into FreeBSD. * How to use and configure the PF firewall. * How to use and configure the IPFW firewall. * How to use and configure the IPFILTER firewall. Before reading this chapter, you should: * Understand basic FreeBSD and Internet concepts. [NOTE] ==== Since all firewalls are based on inspecting the values of selected packet control fields, the creator of the firewall ruleset must have an understanding of how TCP/IP works, what the different values in the packet control fields are, and how these values are used in a normal session conversation. For a good introduction, refer to http://www.ipprimer.com[Daryl's TCP/IP Primer]. ==== [[firewalls-concepts]] == Firewall Concepts A ruleset contains a group of rules which pass or block packets based on the values contained in the packet. The bi-directional exchange of packets between hosts comprises a session conversation. The firewall ruleset processes both the packets arriving from the public Internet, as well as the packets produced by the system as a response to them. Each TCP/IP service is predefined by its protocol and listening port. Packets destined for a specific service originate from the source address using an unprivileged port and target the specific service port on the destination address. All the above parameters can be used as selection criteria to create rules which will pass or block services. To lookup unknown port numbers, refer to [.filename]#/etc/services#. Alternatively, visit http://en.wikipedia.org/wiki/List_of_TCP_and_UDP_port_numbers[http://en.wikipedia.org/wiki/List_of_TCP_and_UDP_port_numbers] and do a port number lookup to find the purpose of a particular port number. Check out this link for http://web.archive.org/web/20150803024617/http://www.sans.org/security-resources/idfaq/oddports.php[port numbers used by Trojans]. FTP has two modes: active mode and passive mode. The difference is in how the data channel is acquired. Passive mode is more secure as the data channel is acquired by the ordinal ftp session requester. For a good explanation of FTP and the different modes, see http://www.slacksite.com/other/ftp.html[http://www.slacksite.com/other/ftp.html]. A firewall ruleset can be either "exclusive" or "inclusive". An exclusive firewall allows all traffic through except for the traffic matching the ruleset. An inclusive firewall does the reverse as it only allows traffic matching the rules through and blocks everything else. An inclusive firewall offers better control of the outgoing traffic, making it a better choice for systems that offer services to the public Internet. It also controls the type of traffic originating from the public Internet that can gain access to a private network. All traffic that does not match the rules is blocked and logged. Inclusive firewalls are generally safer than exclusive firewalls because they significantly reduce the risk of allowing unwanted traffic. [NOTE] ==== Unless noted otherwise, all configuration and example rulesets in this chapter create inclusive firewall rulesets. ==== Security can be tightened further using a "stateful firewall". This type of firewall keeps track of open connections and only allows traffic which either matches an existing connection or opens a new, allowed connection. Stateful filtering treats traffic as a bi-directional exchange of packets comprising a session. When state is specified on a matching rule the firewall dynamically generates internal rules for each anticipated packet being exchanged during the session. It has sufficient matching capabilities to determine if a packet is valid for a session. Any packets that do not properly fit the session template are automatically rejected. When the session completes, it is removed from the dynamic state table. Stateful filtering allows one to focus on blocking/passing new sessions. If the new session is passed, all its subsequent packets are allowed automatically and any impostor packets are automatically rejected. If a new session is blocked, none of its subsequent packets are allowed. Stateful filtering provides advanced matching abilities capable of defending against the flood of different attack methods employed by attackers. NAT stands for _Network Address Translation_. NAT function enables the private LAN behind the firewall to share a single ISP-assigned IP address, even if that address is dynamically assigned. NAT allows each computer in the LAN to have Internet access, without having to pay the ISP for multiple Internet accounts or IP addresses. NAT will automatically translate the private LAN IP address for each system on the LAN to the single public IP address as packets exit the firewall bound for the public Internet. It also performs the reverse translation for returning packets. According to RFC 1918, the following IP address ranges are reserved for private networks which will never be routed directly to the public Internet, and therefore are available for use with NAT: * `10.0.0.0/8`. * `172.16.0.0/12`. * `192.168.0.0/16`. [WARNING] ==== When working with the firewall rules, be _very careful_. Some configurations _can lock the administrator out_ of the server. To be on the safe side, consider performing the initial firewall configuration from the local console rather than doing it remotely over ssh. ==== [[firewalls-pf]] == PF Since FreeBSD 5.3, a ported version of OpenBSD's PF firewall has been included as an integrated part of the base system. PF is a complete, full-featured firewall that has optional support for ALTQ (Alternate Queuing), which provides Quality of Service (QoS). The OpenBSD Project maintains the definitive reference for PF in the http://www.openbsd.org/faq/pf/[PF FAQ]. Peter Hansteen maintains a thorough PF tutorial at http://home.nuug.no/\~peter/pf/[http://home.nuug.no/~peter/pf/]. [WARNING] ==== When reading the http://www.openbsd.org/faq/pf/[PF FAQ], keep in mind that FreeBSD's version of PF has diverged substantially from the upstream OpenBSD version over the years. Not all features work the same way on FreeBSD as they do in OpenBSD and vice versa. ==== The {freebsd-pf} is a good place to ask questions about configuring and running the PF firewall. Check the mailing list archives before asking a question as it may have already been answered. This section of the Handbook focuses on PF as it pertains to FreeBSD. It demonstrates how to enable PF and ALTQ. It also provides several examples for creating rulesets on a FreeBSD system. === Enabling PF To use PF, its kernel module must be first loaded. This section describes the entries that can be added to [.filename]#/etc/rc.conf# to enable PF. Start by adding `pf_enable=yes` to [.filename]#/etc/rc.conf#: [source,shell] .... # sysrc pf_enable=yes .... Additional options, described in man:pfctl[8], can be passed to PF when it is started. Add or change this entry in [.filename]#/etc/rc.conf# and specify any required flags between the two quotes (`""`): [.programlisting] .... pf_flags="" # additional flags for pfctl startup .... PF will not start if it cannot find its ruleset configuration file. By default, FreeBSD does not ship with a ruleset and there is no [.filename]#/etc/pf.conf#. Example rulesets can be found in [.filename]#/usr/share/examples/pf/#. If a custom ruleset has been saved somewhere else, add a line to [.filename]#/etc/rc.conf# which specifies the full path to the file: [.programlisting] .... pf_rules="/path/to/pf.conf" .... Logging support for PF is provided by man:pflog[4]. To enable logging support, add `pflog_enable=yes` to [.filename]#/etc/rc.conf#: [source,shell] .... # sysrc pflog_enable=yes .... The following lines can also be added to change the default location of the log file or to specify any additional flags to pass to man:pflog[4] when it is started: [.programlisting] .... pflog_logfile="/var/log/pflog" # where pflogd should store the logfile pflog_flags="" # additional flags for pflogd startup .... Finally, if there is a LAN behind the firewall and packets need to be forwarded for the computers on the LAN, or NAT is required, enable the following option: [.programlisting] .... gateway_enable="YES" # Enable as LAN gateway .... After saving the needed edits, PF can be started with logging support by typing: [source,shell] .... # service pf start # service pflog start .... By default, PF reads its configuration rules from [.filename]#/etc/pf.conf# and modifies, drops, or passes packets according to the rules or definitions specified in this file. The FreeBSD installation includes several sample files located in [.filename]#/usr/share/examples/pf/#. Refer to the http://www.openbsd.org/faq/pf/[PF FAQ] for complete coverage of PF rulesets. To control PF, use `pfctl`. <> summarizes some useful options to this command. Refer to man:pfctl[8] for a description of all available options: [[pfctl]] .Useful `pfctl` Options [cols="1,1", frame="none", options="header"] |=== | Command | Purpose |`pfctl -e` |Enable PF. |`pfctl -d` |Disable PF. |`pfctl -F all -f /etc/pf.conf` |Flush all NAT, filter, state, and table rules and reload [.filename]#/etc/pf.conf#. |`pfctl -s [ rules \| nat \| states ]` |Report on the filter rules, NAT rules, or state table. |`pfctl -vnf /etc/pf.conf` |Check [.filename]#/etc/pf.conf# for errors, but do not load ruleset. |=== [TIP] ==== package:security/sudo[] is useful for running commands like `pfctl` that require elevated privileges. It can be installed from the Ports Collection. ==== To keep an eye on the traffic that passes through the PF firewall, consider installing the package:sysutils/pftop[] package or port. Once installed, pftop can be run to view a running snapshot of traffic in a format which is similar to man:top[1]. [[pf-tutorial]] === PF Rulesets This section demonstrates how to create a customized ruleset. It starts with the simplest of rulesets and builds upon its concepts using several examples to demonstrate real-world usage of PF's many features. The simplest possible ruleset is for a single machine that does not run any services and which needs access to one network, which may be the Internet. To create this minimal ruleset, edit [.filename]#/etc/pf.conf# so it looks like this: [.programlisting] .... block in all pass out all keep state .... The first rule denies all incoming traffic by default. The second rule allows connections created by this system to pass out, while retaining state information on those connections. This state information allows return traffic for those connections to pass back and should only be used on machines that can be trusted. The ruleset can be loaded with: [source,shell] .... # pfctl -e ; pfctl -f /etc/pf.conf .... In addition to keeping state, PF provides _lists_ and _macros_ which can be defined for use when creating rules. Macros can include lists and need to be defined before use. As an example, insert these lines at the very top of the ruleset: [.programlisting] .... tcp_services = "{ ssh, smtp, domain, www, pop3, auth, pop3s }" udp_services = "{ domain }" .... PF understands port names as well as port numbers, as long as the names are listed in [.filename]#/etc/services#. This example creates two macros. The first is a list of seven TCP port names and the second is one UDP port name. Once defined, macros can be used in rules. In this example, all traffic is blocked except for the connections initiated by this system for the seven specified TCP services and the one specified UDP service: [.programlisting] .... tcp_services = "{ ssh, smtp, domain, www, pop3, auth, pop3s }" udp_services = "{ domain }" block all pass out proto tcp to any port $tcp_services keep state pass proto udp to any port $udp_services keep state .... Even though UDP is considered to be a stateless protocol, PF is able to track some state information. For example, when a UDP request is passed which asks a name server about a domain name, PF will watch for the response to pass it back. Whenever an edit is made to a ruleset, the new rules must be loaded so they can be used: [source,shell] .... # pfctl -f /etc/pf.conf .... If there are no syntax errors, `pfctl` will not output any messages during the rule load. Rules can also be tested before attempting to load them: [source,shell] .... # pfctl -nf /etc/pf.conf .... Including `-n` causes the rules to be interpreted only, but not loaded. This provides an opportunity to correct any errors. At all times, the last valid ruleset loaded will be enforced until either PF is disabled or a new ruleset is loaded. [TIP] ==== Adding `-v` to a `pfctl` ruleset verify or load will display the fully parsed rules exactly the way they will be loaded. This is extremely useful when debugging rules. ==== [[pftut-gateway]] ==== A Simple Gateway with NAT This section demonstrates how to configure a FreeBSD system running PF to act as a gateway for at least one other machine. The gateway needs at least two network interfaces, each connected to a separate network. In this example, [.filename]#xl0# is connected to the Internet and [.filename]#xl1# is connected to the internal network. First, enable the gateway to let the machine forward the network traffic it receives on one interface to another interface. This sysctl setting will forward IPv4 packets: [source,shell] .... # sysctl net.inet.ip.forwarding=1 .... To forward IPv6 traffic, use: [source,shell] .... # sysctl net.inet6.ip6.forwarding=1 .... To enable these settings at system boot, use man:sysrc[8] to add them to [.filename]#/etc/rc.conf#: [source,shell] .... # sysrc gateway_enable=yes # sysrc ipv6_gateway_enable=yes .... Verify with `ifconfig` that both of the interfaces are up and running. Next, create the PF rules to allow the gateway to pass traffic. While the following rule allows stateful traffic from hosts of the internal network to pass to the gateway, the `to` keyword does not guarantee passage all the way from source to destination: [.programlisting] .... pass in on xl1 from xl1:network to xl0:network port $ports keep state .... That rule only lets the traffic pass in to the gateway on the internal interface. To let the packets go further, a matching rule is needed: [.programlisting] .... pass out on xl0 from xl1:network to xl0:network port $ports keep state .... While these two rules will work, rules this specific are rarely needed. For a busy network admin, a readable ruleset is a safer ruleset. The remainder of this section demonstrates how to keep the rules as simple as possible for readability. For example, those two rules could be replaced with one rule: [.programlisting] .... pass from xl1:network to any port $ports keep state .... The `interface:network` notation can be replaced with a macro to make the ruleset even more readable. For example, a `$localnet` macro could be defined as the network directly attached to the internal interface (`$xl1:network`). Alternatively, the definition of `$localnet` could be changed to an _IP address/netmask_ notation to denote a network, such as `192.168.100.1/24` for a subnet of private addresses. If required, `$localnet` could even be defined as a list of networks. Whatever the specific needs, a sensible `$localnet` definition could be used in a typical pass rule as follows: [.programlisting] .... pass from $localnet to any port $ports keep state .... The following sample ruleset allows all traffic initiated by machines on the internal network. It first defines two macros to represent the external and internal 3COM interfaces of the gateway. [NOTE] ==== For dialup users, the external interface will use [.filename]#tun0#. For an ADSL connection, specifically those using PPP over Ethernet (PPPoE), the correct external interface is [.filename]#tun0#, not the physical Ethernet interface. ==== [.programlisting] .... ext_if = "xl0" # macro for external interface - use tun0 for PPPoE int_if = "xl1" # macro for internal interface localnet = $int_if:network # ext_if IP address could be dynamic, hence ($ext_if) nat on $ext_if from $localnet to any -> ($ext_if) block all pass from { lo0, $localnet } to any keep state .... This ruleset introduces the `nat` rule which is used to handle the network address translation from the non-routable addresses inside the internal network to the IP address assigned to the external interface. The parentheses surrounding the last part of the nat rule `($ext_if)` is included when the IP address of the external interface is dynamically assigned. It ensures that network traffic runs without serious interruptions even if the external IP address changes. Note that this ruleset probably allows more traffic to pass out of the network than is needed. One reasonable setup could create this macro: [.programlisting] .... client_out = "{ ftp-data, ftp, ssh, domain, pop3, auth, nntp, http, \ https, cvspserver, 2628, 5999, 8000, 8080 }" .... to use in the main pass rule: [.programlisting] .... pass inet proto tcp from $localnet to any port $client_out \ flags S/SA keep state .... A few other pass rules may be needed. This one enables SSH on the external interface: [.programlisting] .... pass in inet proto tcp to $ext_if port ssh .... This macro definition and rule allows DNS and NTP for internal clients: [.programlisting] .... udp_services = "{ domain, ntp }" pass quick inet proto { tcp, udp } to any port $udp_services keep state .... Note the `quick` keyword in this rule. Since the ruleset consists of several rules, it is important to understand the relationships between the rules in a ruleset. Rules are evaluated from top to bottom, in the sequence they are written. For each packet or connection evaluated by PF, _the last matching rule_ in the ruleset is the one which is applied. However, when a packet matches a rule which contains the `quick` keyword, the rule processing stops and the packet is treated according to that rule. This is very useful when an exception to the general rules is needed. [[pftut-ftp]] ==== Creating an FTP Proxy Configuring working FTP rules can be problematic due to the nature of the FTP protocol. FTP pre-dates firewalls by several decades and is insecure in its design. The most common points against using FTP include: * Passwords are transferred in the clear. * The protocol demands the use of at least two TCP connections (control and data) on separate ports. * When a session is established, data is communicated using randomly selected ports. All of these points present security challenges, even before considering any potential security weaknesses in client or server software. More secure alternatives for file transfer exist, such as man:sftp[1] or man:scp[1], which both feature authentication and data transfer over encrypted connections.. For those situations when FTP is required, PF provides redirection of FTP traffic to a small proxy program called man:ftp-proxy[8], which is included in the base system of FreeBSD. The role of the proxy is to dynamically insert and delete rules in the ruleset, using a set of anchors, to correctly handle FTP traffic. To enable the FTP proxy, add this line to [.filename]#/etc/rc.conf#: [.programlisting] .... ftpproxy_enable="YES" .... Then start the proxy by running `service ftp-proxy start`. For a basic configuration, three elements need to be added to [.filename]#/etc/pf.conf#. First, the anchors which the proxy will use to insert the rules it generates for the FTP sessions: [.programlisting] .... nat-anchor "ftp-proxy/*" rdr-anchor "ftp-proxy/*" .... Second, a pass rule is needed to allow FTP traffic in to the proxy. Third, redirection and NAT rules need to be defined before the filtering rules. Insert this `rdr` rule immediately after the `nat` rule: [.programlisting] .... rdr pass on $int_if proto tcp from any to any port ftp -> 127.0.0.1 port 8021 .... Finally, allow the redirected traffic to pass: [.programlisting] .... pass out proto tcp from $proxy to any port ftp .... where `$proxy` expands to the address the proxy daemon is bound to. Save [.filename]#/etc/pf.conf#, load the new rules, and verify from a client that FTP connections are working: [source,shell] .... # pfctl -f /etc/pf.conf .... This example covers a basic setup where the clients in the local network need to contact FTP servers elsewhere. This basic configuration should work well with most combinations of FTP clients and servers. As shown in man:ftp-proxy[8], the proxy's behavior can be changed in various ways by adding options to the `ftpproxy_flags=` line. Some clients or servers may have specific quirks that must be compensated for in the configuration, or there may be a need to integrate the proxy in specific ways such as assigning FTP traffic to a specific queue. For ways to run an FTP server protected by PF and man:ftp-proxy[8], configure a separate `ftp-proxy` in reverse mode, using `-R`, on a separate port with its own redirecting pass rule. [[pftut-icmp]] ==== Managing ICMP Many of the tools used for debugging or troubleshooting a TCP/IP network rely on the Internet Control Message Protocol (ICMP), which was designed specifically with debugging in mind. The ICMP protocol sends and receives _control messages_ between hosts and gateways, mainly to provide feedback to a sender about any unusual or difficult conditions enroute to the target host. Routers use ICMP to negotiate packet sizes and other transmission parameters in a process often referred to as _path MTU discovery_. From a firewall perspective, some ICMP control messages are vulnerable to known attack vectors. Also, letting all diagnostic traffic pass unconditionally makes debugging easier, but it also makes it easier for others to extract information about the network. For these reasons, the following rule may not be optimal: [.programlisting] .... pass inet proto icmp from any to any .... One solution is to let all ICMP traffic from the local network through while stopping all probes from outside the network: [.programlisting] .... pass inet proto icmp from $localnet to any keep state pass inet proto icmp from any to $ext_if keep state .... Additional options are available which demonstrate some of PF's flexibility. For example, rather than allowing all ICMP messages, one can specify the messages used by man:ping[8] and man:traceroute[8]. Start by defining a macro for that type of message: [.programlisting] .... icmp_types = "echoreq" .... and a rule which uses the macro: [.programlisting] .... pass inet proto icmp all icmp-type $icmp_types keep state .... If other types of ICMP packets are needed, expand `icmp_types` to a list of those packet types. Type `more /usr/src/sbin/pfctl/pfctl_parser.c` to see the list of ICMP message types supported by PF. Refer to http://www.iana.org/assignments/icmp-parameters/icmp-parameters.xhtml[http://www.iana.org/assignments/icmp-parameters/icmp-parameters.xhtml] for an explanation of each message type. Since Unix `traceroute` uses UDP by default, another rule is needed to allow Unix `traceroute`: [.programlisting] .... # allow out the default range for traceroute(8): pass out on $ext_if inet proto udp from any to any port 33433 >< 33626 keep state .... Since `TRACERT.EXE` on Microsoft Windows systems uses ICMP echo request messages, only the first rule is needed to allow network traces from those systems. Unix `traceroute` can be instructed to use other protocols as well, and will use ICMP echo request messages if `-I` is used. Check the man:traceroute[8] man page for details. [[pftut-pathmtudisc]] ===== Path MTU Discovery Internet protocols are designed to be device independent, and one consequence of device independence is that the optimal packet size for a given connection cannot always be predicted reliably. The main constraint on packet size is the _Maximum Transmission Unit_ (MTU) which sets the upper limit on the packet size for an interface. Type `ifconfig` to view the MTUs for a system's network interfaces. TCP/IP uses a process known as path MTU discovery to determine the right packet size for a connection. This process sends packets of varying sizes with the "Do not fragment" flag set, expecting an ICMP return packet of "type 3, code 4" when the upper limit has been reached. Type 3 means "destination unreachable", and code 4 is short for "fragmentation needed, but the do-not-fragment flag is set". To allow path MTU discovery in order to support connections to other MTUs, add the `destination unreachable` type to the `icmp_types` macro: [.programlisting] .... icmp_types = "{ echoreq, unreach }" .... Since the pass rule already uses that macro, it does not need to be modified to support the new ICMP type: [.programlisting] .... pass inet proto icmp all icmp-type $icmp_types keep state .... PF allows filtering on all variations of ICMP types and codes. The list of possible types and codes are documented in man:icmp[4] and man:icmp6[4]. [[pftut-tables]] ==== Using Tables Some types of data are relevant to filtering and redirection at a given time, but their definition is too long to be included in the ruleset file. PF supports the use of tables, which are defined lists that can be manipulated without needing to reload the entire ruleset, and which can provide fast lookups. Table names are always enclosed within `< >`, like this: [.programlisting] .... table { 192.168.2.0/24, !192.168.2.5 } .... In this example, the `192.168.2.0/24` network is part of the table, except for the address `192.168.2.5`, which is excluded using the `!` operator. It is also possible to load tables from files where each item is on a separate line, as seen in this example [.filename]#/etc/clients#: [.programlisting] .... 192.168.2.0/24 !192.168.2.5 .... To refer to the file, define the table like this: [.programlisting] .... table persist file "/etc/clients" .... Once the table is defined, it can be referenced by a rule: [.programlisting] .... pass inet proto tcp from to any port $client_out flags S/SA keep state .... A table's contents can be manipulated live, using `pfctl`. This example adds another network to the table: [source,shell] .... # pfctl -t clients -T add 192.168.1.0/16 .... Note that any changes made this way will take affect now, making them ideal for testing, but will not survive a power failure or reboot. To make the changes permanent, modify the definition of the table in the ruleset or edit the file that the table refers to. One can maintain the on-disk copy of the table using a man:cron[8] job which dumps the table's contents to disk at regular intervals, using a command such as `pfctl -t clients -T show >/etc/clients`. Alternatively, [.filename]#/etc/clients# can be updated with the in-memory table contents: [source,shell] .... # pfctl -t clients -T replace -f /etc/clients .... [[pftut-overload]] ==== Using Overload Tables to Protect SSH Those who run SSH on an external interface have probably seen something like this in the authentication logs: [.programlisting] .... Sep 26 03:12:34 skapet sshd[25771]: Failed password for root from 200.72.41.31 port 40992 ssh2 Sep 26 03:12:34 skapet sshd[5279]: Failed password for root from 200.72.41.31 port 40992 ssh2 Sep 26 03:12:35 skapet sshd[5279]: Received disconnect from 200.72.41.31: 11: Bye Bye Sep 26 03:12:44 skapet sshd[29635]: Invalid user admin from 200.72.41.31 Sep 26 03:12:44 skapet sshd[24703]: input_userauth_request: invalid user admin Sep 26 03:12:44 skapet sshd[24703]: Failed password for invalid user admin from 200.72.41.31 port 41484 ssh2 .... This is indicative of a brute force attack where somebody or some program is trying to discover the user name and password which will let them into the system. If external SSH access is needed for legitimate users, changing the default port used by SSH can offer some protection. However, PF provides a more elegant solution. Pass rules can contain limits on what connecting hosts can do and violators can be banished to a table of addresses which are denied some or all access. It is even possible to drop all existing connections from machines which overreach the limits. To configure this, create this table in the tables section of the ruleset: [.programlisting] .... table persist .... Then, somewhere early in the ruleset, add rules to block brute access while allowing legitimate access: [.programlisting] .... block quick from pass inet proto tcp from any to $localnet port $tcp_services \ flags S/SA keep state \ (max-src-conn 100, max-src-conn-rate 15/5, \ overload flush global) .... The part in parentheses defines the limits and the numbers should be changed to meet local requirements. It can be read as follows: `max-src-conn` is the number of simultaneous connections allowed from one host. `max-src-conn-rate` is the rate of new connections allowed from any single host (_15_) per number of seconds (_5_). `overload ` means that any host which exceeds these limits gets its address added to the `bruteforce` table. The ruleset blocks all traffic from addresses in the `bruteforce` table. Finally, `flush global` says that when a host reaches the limit, that all (`global`) of that host's connections will be terminated (`flush`). [NOTE] ==== These rules will _not_ block slow bruteforcers, as described in http://home.nuug.no/\~peter/hailmary2013/[http://home.nuug.no/~peter/hailmary2013/]. ==== This example ruleset is intended mainly as an illustration. For example, if a generous number of connections in general are wanted, but the desire is to be more restrictive when it comes to ssh, supplement the rule above with something like the one below, early on in the rule set: [.programlisting] .... pass quick proto { tcp, udp } from any to any port ssh \ flags S/SA keep state \ (max-src-conn 15, max-src-conn-rate 5/3, \ overload flush global) .... [NOTE] ==== *It May Not be Necessary to Block All Overloaders:* + It is worth noting that the overload mechanism is a general technique which does not apply exclusively to SSH, and it is not always optimal to entirely block all traffic from offenders. For example, an overload rule could be used to protect a mail service or a web service, and the overload table could be used in a rule to assign offenders to a queue with a minimal bandwidth allocation or to redirect to a specific web page. ==== Over time, tables will be filled by overload rules and their size will grow incrementally, taking up more memory. Sometimes an IP address that is blocked is a dynamically assigned one, which has since been assigned to a host who has a legitimate reason to communicate with hosts in the local network. For situations like these, pfctl provides the ability to expire table entries. For example, this command will remove `` table entries which have not been referenced for `86400` seconds: [source,shell] .... # pfctl -t bruteforce -T expire 86400 .... Similar functionality is provided by package:security/expiretable[], which removes table entries which have not been accessed for a specified period of time. Once installed, expiretable can be run to remove `` table entries older than a specified age. This example removes all entries older than 24 hours: [.programlisting] .... /usr/local/sbin/expiretable -v -d -t 24h bruteforce .... [[pftut-spamd]] ==== Protecting Against SPAM Not to be confused with the spamd daemon which comes bundled with spamassassin, package:mail/spamd[] can be configured with PF to provide an outer defense against SPAM. This spamd hooks into the PF configuration using a set of redirections. -Spammers tend to send a large number of messages, and SPAM is mainly sent from a few spammer friendly networks and a large number of hijacked machines, both of which are reported to _blacklists_ fairly quickly. +Spammers tend to send a large number of messages, and SPAM is mainly sent from a few spammer friendly networks and a large number of hijacked machines, both of which are reported to _blocklists_ fairly quickly. -When an SMTP connection from an address in a blacklist is received, spamd presents its banner and immediately switches to a mode where it answers SMTP traffic one byte at a time. +When an SMTP connection from an address in a blocklist is received, spamd presents its banner and immediately switches to a mode where it answers SMTP traffic one byte at a time. This technique, which is intended to waste as much time as possible on the spammer's end, is called _tarpitting_. The specific implementation which uses one byte SMTP replies is often referred to as _stuttering_. -This example demonstrates the basic procedure for setting up spamd with automatically updated blacklists. +This example demonstrates the basic procedure for setting up spamd with automatically updated blocklists. Refer to the man pages which are installed with package:mail/spamd[] for more information. [.procedure] **** .Procedure: Configuring spamd . Install the package:mail/spamd[] package or port. To use spamd's greylisting features, man:fdescfs[5] must be mounted at [.filename]#/dev/fd#. Add the following line to [.filename]#/etc/fstab#: + [.programlisting] .... fdescfs /dev/fd fdescfs rw 0 0 .... + Then, mount the filesystem: + [.programlisting] .... # mount fdescfs .... . Next, edit the PF ruleset to include: + [.programlisting] .... table persist table persist rdr pass on $ext_if inet proto tcp from to \ { $ext_if, $localnet } port smtp -> 127.0.0.1 port 8025 rdr pass on $ext_if inet proto tcp from ! to \ { $ext_if, $localnet } port smtp -> 127.0.0.1 port 8025 .... + The two tables `` and `` are essential. SMTP traffic from an address listed in `` but not in `` is redirected to the spamd daemon listening at port 8025. . The next step is to configure spamd in [.filename]#/usr/local/etc/spamd.conf# and to add some [.filename]#rc.conf# parameters. + The installation of package:mail/spamd[] includes a sample configuration file ([.filename]#/usr/local/etc/spamd.conf.sample#) and a man page for [.filename]#spamd.conf#. Refer to these for additional configuration options beyond those shown in this example. + One of the first lines in the configuration file that does not begin with a `#` comment sign contains the block which defines the `all` list, which specifies the lists to use: + [.programlisting] .... all:\ - :traplist:whitelist: + :traplist:allowlist: .... + -This entry adds the desired blacklists, separated by colons (`:`). -To use a whitelist to subtract addresses from a blacklist, add the name of the whitelist _immediately_ after the name of that blacklist. For example: `:blacklist:whitelist:`. +This entry adds the desired blocklists, separated by colons (`:`). To use an allowlist to subtract addresses from a blocklist, add the name of the allowlist _immediately_ after the name of that blocklist. For example: `:blocklist:allowlist:`. + -This is followed by the specified blacklist's definition: +This is followed by the specified blocklist's definition: + [.programlisting] .... traplist:\ :black:\ :msg="SPAM. Your address %A has sent spam within the last 24 hours":\ :method=http:\ :file=www.openbsd.org/spamd/traplist.gz .... + -where the first line is the name of the blacklist and the second line specifies the list type. -The `msg` field contains the message to display to blacklisted senders during the SMTP dialogue. +where the first line is the name of the blocklist and the second line specifies the list type. +The `msg` field contains the message to display to blocklisted senders during the SMTP dialogue. The `method` field specifies how spamd-setup fetches the list data; supported methods are `http`, `ftp`, from a `file` in a mounted file system, and via `exec` of an external program. Finally, the `file` field specifies the name of the file spamd expects to receive. + -The definition of the specified whitelist is similar, but omits the `msg` field since a message is not needed: +The definition of the specified allowlist is similar, but omits the `msg` field since a message is not needed: + [.programlisting] .... -whitelist:\ +allowlist:\ :white:\ :method=file:\ - :file=/var/mail/whitelist.txt + :file=/var/mail/allowlist.txt .... + [TIP] ==== *Choose Data Sources with Care:* + -Using all the blacklists in the sample [.filename]#spamd.conf# will blacklist large blocks of the Internet. +Using all the blocklists in the sample [.filename]#spamd.conf# will block large blocks of the Internet. Administrators need to edit the file to create an optimal configuration which uses applicable data sources and, when necessary, uses custom lists. ==== + Next, add this entry to [.filename]#/etc/rc.conf#. Additional flags are described in the man page specified by the comment: + [.programlisting] .... spamd_flags="-v" # use "" and see spamd-setup(8) for flags .... + When finished, reload the ruleset, start spamd by typing `service obspamd start`, and complete the configuration using `spamd-setup`. Finally, create a man:cron[8] job which calls `spamd-setup` to update the tables at reasonable intervals. **** On a typical gateway in front of a mail server, hosts will soon start getting trapped within a few seconds to several minutes. PF also supports _greylisting_, which temporarily rejects messages from unknown hosts with _45n_ codes. Messages from greylisted hosts which try again within a reasonable time are let through. Traffic from senders which are set up to behave within the limits set by RFC 1123 and RFC 2821 are immediately let through. More information about greylisting as a technique can be found at the http://www.greylisting.org/[greylisting.org] web site. The most amazing thing about greylisting, apart from its simplicity, is that it still works. Spammers and malware writers have been very slow to adapt to bypass this technique. The basic procedure for configuring greylisting is as follows: [.procedure] .Procedure: Configuring Greylisting . Make sure that man:fdescfs[5] is mounted as described in Step 1 of the previous Procedure. . To run spamd in greylisting mode, add this line to [.filename]#/etc/rc.conf#: + [.programlisting] .... spamd_grey="YES" # use spamd greylisting if YES .... + Refer to the spamd man page for descriptions of additional related parameters. . To complete the greylisting setup: + [.programlisting] .... # service obspamd restart # service obspamlogd start .... Behind the scenes, the spamdb database tool and the spamlogd whitelist updater perform essential functions for the greylisting feature. -spamdb is the administrator's main interface to managing the black, grey, and white lists via the contents of the [.filename]#/var/db/spamdb# database. +spamdb is the administrator's main interface to managing the block, grey, and allow lists via the contents of the [.filename]#/var/db/spamdb# database. [[pftut-hygiene]] ==== Network Hygiene This section describes how `block-policy`, `scrub`, and `antispoof` can be used to make the ruleset behave sanely. The `block-policy` is an option which can be set in the `options` part of the ruleset, which precedes the redirection and filtering rules. This option determines which feedback, if any, PF sends to hosts that are blocked by a rule. The option has two possible values: `drop` drops blocked packets with no feedback, and `return` returns a status code such as `Connection refused`. If not set, the default policy is `drop`. To change the `block-policy`, specify the desired value: [.programlisting] .... set block-policy return .... In PF, `scrub` is a keyword which enables network packet normalization. This process reassembles fragmented packets and drops TCP packets that have invalid flag combinations. Enabling `scrub` provides a measure of protection against certain kinds of attacks based on incorrect handling of packet fragments. A number of options are available, but the simplest form is suitable for most configurations: [.programlisting] .... scrub in all .... Some services, such as NFS, require specific fragment handling options. Refer to https://home.nuug.no/\~peter/pf/en/scrub.html[https://home.nuug.no/~peter/pf/en/scrub.html] for more information. This example reassembles fragments, clears the "do not fragment" bit, and sets the maximum segment size to 1440 bytes: [.programlisting] .... scrub in all fragment reassemble no-df max-mss 1440 .... The `antispoof` mechanism protects against activity from spoofed or forged IP addresses, mainly by blocking packets appearing on interfaces and in directions which are logically not possible. These rules weed out spoofed traffic coming in from the rest of the world as well as any spoofed packets which originate in the local network: [.programlisting] .... antispoof for $ext_if antispoof for $int_if .... [[pftut-unrouteables]] ==== Handling Non-Routable Addresses Even with a properly configured gateway to handle network address translation, one may have to compensate for other people's misconfigurations. A common misconfiguration is to let traffic with non-routable addresses out to the Internet. Since traffic from non-routeable addresses can play a part in several DoS attack techniques, consider explicitly blocking traffic from non-routeable addresses from entering the network through the external interface. In this example, a macro containing non-routable addresses is defined, then used in blocking rules. Traffic to and from these addresses is quietly dropped on the gateway's external interface. [.programlisting] .... martians = "{ 127.0.0.0/8, 192.168.0.0/16, 172.16.0.0/12, \ 10.0.0.0/8, 169.254.0.0/16, 192.0.2.0/24, \ 0.0.0.0/8, 240.0.0.0/4 }" block drop in quick on $ext_if from $martians to any block drop out quick on $ext_if from any to $martians .... === Enabling ALTQ On FreeBSD, ALTQ can be used with PF to provide Quality of Service (QOS). Once ALTQ is enabled, queues can be defined in the ruleset which determine the processing priority of outbound packets. Before enabling ALTQ, refer to man:altq[4] to determine if the drivers for the network cards installed on the system support it. ALTQ is not available as a loadable kernel module. If the system's interfaces support ALTQ, create a custom kernel using the instructions in crossref:kernelconfig[kernelconfig,Configuring the FreeBSD Kernel]. The following kernel options are available. The first is needed to enable ALTQ. At least one of the other options is necessary to specify the queueing scheduler algorithm: [.programlisting] .... options ALTQ options ALTQ_CBQ # Class Based Queuing (CBQ) options ALTQ_RED # Random Early Detection (RED) options ALTQ_RIO # RED In/Out options ALTQ_HFSC # Hierarchical Packet Scheduler (HFSC) options ALTQ_PRIQ # Priority Queuing (PRIQ) .... The following scheduler algorithms are available: CBQ:: Class Based Queuing (CBQ) is used to divide a connection's bandwidth into different classes or queues to prioritize traffic based on filter rules. RED:: Random Early Detection (RED) is used to avoid network congestion by measuring the length of the queue and comparing it to the minimum and maximum thresholds for the queue. When the queue is over the maximum, all new packets are randomly dropped. RIO:: In Random Early Detection In and Out (RIO) mode, RED maintains multiple average queue lengths and multiple threshold values, one for each QOS level. HFSC:: Hierarchical Fair Service Curve Packet Scheduler (HFSC) is described in http://www-2.cs.cmu.edu/\~hzhang/HFSC/main.html[http://www-2.cs.cmu.edu/~hzhang/HFSC/main.html]. PRIQ:: Priority Queuing (PRIQ) always passes traffic that is in a higher queue first. More information about the scheduling algorithms and example rulesets are available at the https://web.archive.org/web/20151109213426/http://www.openbsd.org/faq/pf/queueing.html[OpenBSD's web archive]. [[firewalls-ipfw]] == IPFW IPFW is a stateful firewall written for FreeBSD which supports both IPv4 and IPv6. It is comprised of several components: the kernel firewall filter rule processor and its integrated packet accounting facility, the logging facility, NAT, the man:dummynet[4] traffic shaper, a forward facility, a bridge facility, and an ipstealth facility. FreeBSD provides a sample ruleset in [.filename]#/etc/rc.firewall# which defines several firewall types for common scenarios to assist novice users in generating an appropriate ruleset. IPFW provides a powerful syntax which advanced users can use to craft customized rulesets that meet the security requirements of a given environment. This section describes how to enable IPFW, provides an overview of its rule syntax, and demonstrates several rulesets for common configuration scenarios. [[firewalls-ipfw-enable]] === Enabling IPFW IPFW is included in the basic FreeBSD install as a kernel loadable module, meaning that a custom kernel is not needed in order to enable IPFW. For those users who wish to statically compile IPFW support into a custom kernel, see <>. To configure the system to enable IPFW at boot time, add `firewall_enable="YES"` to [.filename]#/etc/rc.conf#: [source,shell] .... # sysrc firewall_enable="YES" .... To use one of the default firewall types provided by FreeBSD, add another line which specifies the type: [source,shell] .... # sysrc firewall_type="open" .... The available types are: * `open`: passes all traffic. * `client`: protects only this machine. * `simple`: protects the whole network. * `closed`: entirely disables IP traffic except for the loopback interface. * `workstation`: protects only this machine using stateful rules. * `UNKNOWN`: disables the loading of firewall rules. * [.filename]#filename#: full path of the file containing the firewall ruleset. If `firewall_type` is set to either `client` or `simple`, modify the default rules found in [.filename]#/etc/rc.firewall# to fit the configuration of the system. Note that the `filename` type is used to load a custom ruleset. An alternate way to load a custom ruleset is to set the `firewall_script` variable to the absolute path of an _executable script_ that includes IPFW commands. The examples used in this section assume that the `firewall_script` is set to [.filename]#/etc/ipfw.rules#: [source,shell] .... # sysrc firewall_script="/etc/ipfw.rules" .... To enable logging through man:syslogd[8], include this line: [source,shell] .... # sysrc firewall_logging="YES" .... [WARNING] ==== Only firewall rules with the `log` option will be logged. The default rules do not include this option and it must be manually added. Therefore it is advisable that the default ruleset is edited for logging. In addition, log rotation may be desired if the logs are stored in a separate file. ==== There is no [.filename]#/etc/rc.conf# variable to set logging limits. To limit the number of times a rule is logged per connection attempt, specify the number using this line in [.filename]#/etc/sysctl.conf#: [source,shell] .... # echo "net.inet.ip.fw.verbose_limit=5" >> /etc/sysctl.conf .... To enable logging through a dedicated interface named `ipfw0`, add this line to [.filename]#/etc/rc.conf# instead: [source,shell] .... # sysrc firewall_logif="YES" .... Then use tcpdump to see what is being logged: [source,shell] .... # tcpdump -t -n -i ipfw0 .... [TIP] ==== There is no overhead due to logging unless tcpdump is attached. ==== After saving the needed edits, start the firewall. To enable logging limits now, also set the `sysctl` value specified above: [source,shell] .... # service ipfw start # sysctl net.inet.ip.fw.verbose_limit=5 .... [[firewalls-ipfw-rules]] === IPFW Rule Syntax When a packet enters the IPFW firewall, it is compared against the first rule in the ruleset and progresses one rule at a time, moving from top to bottom in sequence. When the packet matches the selection parameters of a rule, the rule's action is executed and the search of the ruleset terminates for that packet. This is referred to as "first match wins". If the packet does not match any of the rules, it gets caught by the mandatory IPFW default rule number 65535, which denies all packets and silently discards them. However, if the packet matches a rule that contains the `count`, `skipto`, or `tee` keywords, the search continues. Refer to man:ipfw[8] for details on how these keywords affect rule processing. When creating an IPFW rule, keywords must be written in the following order. Some keywords are mandatory while other keywords are optional. The words shown in uppercase represent a variable and the words shown in lowercase must precede the variable that follows it. The `#` symbol is used to mark the start of a comment and may appear at the end of a rule or on its own line. Blank lines are ignored. `_CMD RULE_NUMBER set SET_NUMBER ACTION log LOG_AMOUNT PROTO from SRC SRC_PORT to DST DST_PORT OPTIONS_` This section provides an overview of these keywords and their options. It is not an exhaustive list of every possible option. Refer to man:ipfw[8] for a complete description of the rule syntax that can be used when creating IPFW rules. CMD:: Every rule must start with `ipfw add`. RULE_NUMBER:: Each rule is associated with a number from `1` to `65534`. The number is used to indicate the order of rule processing. Multiple rules can have the same number, in which case they are applied according to the order in which they have been added. SET_NUMBER:: Each rule is associated with a set number from `0` to `31`. Sets can be individually disabled or enabled, making it possible to quickly add or delete a set of rules. If a SET_NUMBER is not specified, the rule will be added to set `0`. ACTION:: A rule can be associated with one of the following actions. The specified action will be executed when the packet matches the selection criterion of the rule. + `allow | accept | pass | permit`: these keywords are equivalent and allow packets that match the rule. + `check-state`: checks the packet against the dynamic state table. If a match is found, execute the action associated with the rule which generated this dynamic rule, otherwise move to the next rule. A `check-state` rule does not have selection criterion. If no `check-state` rule is present in the ruleset, the dynamic rules table is checked at the first `keep-state` or `limit` rule. + `count`: updates counters for all packets that match the rule. The search continues with the next rule. + `deny | drop`: either word silently discards packets that match this rule. + Additional actions are available. Refer to man:ipfw[8] for details. LOG_AMOUNT:: When a packet matches a rule with the `log` keyword, a message will be logged to man:syslogd[8] with a facility name of `SECURITY`. Logging only occurs if the number of packets logged for that particular rule does not exceed a specified LOG_AMOUNT. If no LOG_AMOUNT is specified, the limit is taken from the value of `net.inet.ip.fw.verbose_limit`. A value of zero removes the logging limit. Once the limit is reached, logging can be re-enabled by clearing the logging counter or the packet counter for that rule, using `ipfw resetlog`. + [NOTE] ==== Logging is done after all other packet matching conditions have been met, and before performing the final action on the packet. The administrator decides which rules to enable logging on. ==== PROTO:: This optional value can be used to specify any protocol name or number found in [.filename]#/etc/protocols#. SRC:: The `from` keyword must be followed by the source address or a keyword that represents the source address. An address can be represented by `any`, `me` (any address configured on an interface on this system), `me6`, (any IPv6 address configured on an interface on this system), or `table` followed by the number of a lookup table which contains a list of addresses. When specifying an IP address, it can be optionally followed by its CIDR mask or subnet mask. For example, `1.2.3.4/25` or `1.2.3.4:255.255.255.128`. SRC_PORT:: An optional source port can be specified using the port number or name from [.filename]#/etc/services#. DST:: The `to` keyword must be followed by the destination address or a keyword that represents the destination address. The same keywords and addresses described in the SRC section can be used to describe the destination. DST_PORT:: An optional destination port can be specified using the port number or name from [.filename]#/etc/services#. OPTIONS:: Several keywords can follow the source and destination. As the name suggests, OPTIONS are optional. Commonly used options include `in` or `out`, which specify the direction of packet flow, `icmptypes` followed by the type of ICMP message, and `keep-state`. + When a `keep-state` rule is matched, the firewall will create a dynamic rule which matches bidirectional traffic between the source and destination addresses and ports using the same protocol. + The dynamic rules facility is vulnerable to resource depletion from a SYN-flood attack which would open a huge number of dynamic rules. To counter this type of attack with IPFW, use `limit`. This option limits the number of simultaneous sessions by checking the open dynamic rules, counting the number of times this rule and IP address combination occurred. If this count is greater than the value specified by `limit`, the packet is discarded. + Dozens of OPTIONS are available. Refer to man:ipfw[8] for a description of each available option. === Example Ruleset This section demonstrates how to create an example stateful firewall ruleset script named [.filename]#/etc/ipfw.rules#. In this example, all connection rules use `in` or `out` to clarify the direction. They also use `via` _interface-name_ to specify the interface the packet is traveling over. [NOTE] ==== When first creating or testing a firewall ruleset, consider temporarily setting this tunable: [.programlisting] .... net.inet.ip.fw.default_to_accept="1" .... This sets the default policy of man:ipfw[8] to be more permissive than the default `deny ip from any to any`, making it slightly more difficult to get locked out of the system right after a reboot. ==== The firewall script begins by indicating that it is a Bourne shell script and flushes any existing rules. It then creates the `cmd` variable so that `ipfw add` does not have to be typed at the beginning of every rule. It also defines the `pif` variable which represents the name of the interface that is attached to the Internet. [.programlisting] .... #!/bin/sh # Flush out the list before we begin. ipfw -q -f flush # Set rules command prefix cmd="ipfw -q add" pif="dc0" # interface name of NIC attached to Internet .... The first two rules allow all traffic on the trusted internal interface and on the loopback interface: [.programlisting] .... # Change xl0 to LAN NIC interface name $cmd 00005 allow all from any to any via xl0 # No restrictions on Loopback Interface $cmd 00010 allow all from any to any via lo0 .... The next rule allows the packet through if it matches an existing entry in the dynamic rules table: [.programlisting] .... $cmd 00101 check-state .... The next set of rules defines which stateful connections internal systems can create to hosts on the Internet: [.programlisting] .... # Allow access to public DNS # Replace x.x.x.x with the IP address of a public DNS server # and repeat for each DNS server in /etc/resolv.conf $cmd 00110 allow tcp from any to x.x.x.x 53 out via $pif setup keep-state $cmd 00111 allow udp from any to x.x.x.x 53 out via $pif keep-state # Allow access to ISP's DHCP server for cable/DSL configurations. # Use the first rule and check log for IP address. # Then, uncomment the second rule, input the IP address, and delete the first rule $cmd 00120 allow log udp from any to any 67 out via $pif keep-state #$cmd 00120 allow udp from any to x.x.x.x 67 out via $pif keep-state # Allow outbound HTTP and HTTPS connections $cmd 00200 allow tcp from any to any 80 out via $pif setup keep-state $cmd 00220 allow tcp from any to any 443 out via $pif setup keep-state # Allow outbound email connections $cmd 00230 allow tcp from any to any 25 out via $pif setup keep-state $cmd 00231 allow tcp from any to any 110 out via $pif setup keep-state # Allow outbound ping $cmd 00250 allow icmp from any to any out via $pif keep-state # Allow outbound NTP $cmd 00260 allow udp from any to any 123 out via $pif keep-state # Allow outbound SSH $cmd 00280 allow tcp from any to any 22 out via $pif setup keep-state # deny and log all other outbound connections $cmd 00299 deny log all from any to any out via $pif .... The next set of rules controls connections from Internet hosts to the internal network. It starts by denying packets typically associated with attacks and then explicitly allows specific types of connections. All the authorized services that originate from the Internet use `limit` to prevent flooding. [.programlisting] .... # Deny all inbound traffic from non-routable reserved address spaces $cmd 00300 deny all from 192.168.0.0/16 to any in via $pif #RFC 1918 private IP $cmd 00301 deny all from 172.16.0.0/12 to any in via $pif #RFC 1918 private IP $cmd 00302 deny all from 10.0.0.0/8 to any in via $pif #RFC 1918 private IP $cmd 00303 deny all from 127.0.0.0/8 to any in via $pif #loopback $cmd 00304 deny all from 0.0.0.0/8 to any in via $pif #loopback $cmd 00305 deny all from 169.254.0.0/16 to any in via $pif #DHCP auto-config $cmd 00306 deny all from 192.0.2.0/24 to any in via $pif #reserved for docs $cmd 00307 deny all from 204.152.64.0/23 to any in via $pif #Sun cluster interconnect $cmd 00308 deny all from 224.0.0.0/3 to any in via $pif #Class D & E multicast # Deny public pings $cmd 00310 deny icmp from any to any in via $pif # Deny ident $cmd 00315 deny tcp from any to any 113 in via $pif # Deny all Netbios services. $cmd 00320 deny tcp from any to any 137 in via $pif $cmd 00321 deny tcp from any to any 138 in via $pif $cmd 00322 deny tcp from any to any 139 in via $pif $cmd 00323 deny tcp from any to any 81 in via $pif # Deny fragments $cmd 00330 deny all from any to any frag in via $pif # Deny ACK packets that did not match the dynamic rule table $cmd 00332 deny tcp from any to any established in via $pif # Allow traffic from ISP's DHCP server. # Replace x.x.x.x with the same IP address used in rule 00120. #$cmd 00360 allow udp from any to x.x.x.x 67 in via $pif keep-state # Allow HTTP connections to internal web server $cmd 00400 allow tcp from any to me 80 in via $pif setup limit src-addr 2 # Allow inbound SSH connections $cmd 00410 allow tcp from any to me 22 in via $pif setup limit src-addr 2 # Reject and log all other incoming connections $cmd 00499 deny log all from any to any in via $pif .... The last rule logs all packets that do not match any of the rules in the ruleset: [.programlisting] .... # Everything else is denied and logged $cmd 00999 deny log all from any to any .... [[in-kernel-nat]] === In-kernel NAT FreeBSD's IPFW firewall has two implementations of NAT: the userland implementation man:natd[8], and the more recent in-kernel NAT implementation. Both work in conjunction with IPFW to provide network address translation. This can be used to provide an Internet Connection Sharing solution so that several internal computers can connect to the Internet using a single public IP address. To do this, the FreeBSD machine connected to the Internet must act as a gateway. This system must have two NICs, where one is connected to the Internet and the other is connected to the internal LAN. Each machine connected to the LAN should be assigned an IP address in the private network space, as defined by https://www.ietf.org/rfc/rfc1918.txt[RFC 1918]. Some additional configuration is needed in order to enable the in-kernel NAT facility of IPFW. To enable in-kernel NAT support at boot time, the following must be set in [.filename]#/etc/rc.conf#: [.programlisting] .... gateway_enable="YES" firewall_enable="YES" firewall_nat_enable="YES" .... [NOTE] ==== When `firewall_nat_enable` is set but `firewall_enable` is not, it will have no effect and do nothing. This is because the in-kernel NAT implementation is only compatible with IPFW. ==== When the ruleset contains stateful rules, the positioning of the NAT rule is critical and the `skipto` action is used. The `skipto` action requires a rule number so that it knows which rule to jump to. The example below builds upon the firewall ruleset shown in the previous section. It adds some additional entries and modifies some existing rules in order to configure the firewall for in-kernel NAT. It starts by adding some additional variables which represent the rule number to skip to, the `keep-state` option, and a list of TCP ports which will be used to reduce the number of rules. [.programlisting] .... #!/bin/sh ipfw -q -f flush cmd="ipfw -q add" skip="skipto 1000" pif=dc0 ks="keep-state" good_tcpo="22,25,37,53,80,443,110" .... With in-kernel NAT it is necessary to disable TCP segmentation offloading (TSO) due to the architecture of man:libalias[3], a library implemented as a kernel module to provide the in-kernel NAT facility of IPFW. TSO can be disabled on a per network interface basis using man:ifconfig[8] or on a system wide basis using man:sysctl[8]. To disable TSO system wide, the following must be set it [.filename]#/etc/sysctl.conf#: [.programlisting] .... net.inet.tcp.tso="0" .... A NAT instance will also be configured. It is possible to have multiple NAT instances each with their own configuration. For this example only one NAT instance is needed, NAT instance number 1. The configuration can take a few options such as: `if` which indicates the public interface, `same_ports` which takes care that alliased ports and local port numbers are mapped the same, `unreg_only` will result in only unregistered (private) address spaces to be processed by the NAT instance, and `reset` which will help to keep a functioning NAT instance even when the public IP address of the IPFW machine changes. For all possible options that can be passed to a single NAT instance configuration consult man:ipfw[8]. When configuring a stateful NATing firewall, it is necessary to allow translated packets to be reinjected in the firewall for further processing. This can be achieved by disabling `one_pass` behavior at the start of the firewall script. [.programlisting] .... ipfw disable one_pass ipfw -q nat 1 config if $pif same_ports unreg_only reset .... The inbound NAT rule is inserted _after_ the two rules which allow all traffic on the trusted and loopback interfaces and after the reassemble rule but _before_ the `check-state` rule. It is important that the rule number selected for this NAT rule, in this example `100`, is higher than the first three rules and lower than the `check-state` rule. Furthermore, because of the behavior of in-kernel NAT it is advised to place a reassemble rule just before the first NAT rule and after the rules that allow traffic on trusted interface. Normally, IP fragmentation should not happen, but when dealing with IPSEC/ESP/GRE tunneling traffic it might and the reassembling of fragments is necessary before handing the complete packet over to the in-kernel NAT facility. [NOTE] ==== The reassemble rule was not needed with userland man:natd[8] because the internal workings of the IPFW `divert` action already takes care of reassembling packets before delivery to the socket as also stated in man:ipfw[8]. The NAT instance and rule number used in this example does not match with the default NAT instance and rule number created by [.filename]#rc.firewall#. [.filename]#rc.firewall# is a script that sets up the default firewall rules present in FreeBSD. ==== [.programlisting] .... $cmd 005 allow all from any to any via xl0 # exclude LAN traffic $cmd 010 allow all from any to any via lo0 # exclude loopback traffic $cmd 099 reass all from any to any in # reassemble inbound packets $cmd 100 nat 1 ip from any to any in via $pif # NAT any inbound packets # Allow the packet through if it has an existing entry in the dynamic rules table $cmd 101 check-state .... The outbound rules are modified to replace the `allow` action with the `$skip` variable, indicating that rule processing will continue at rule `1000`. The seven `tcp` rules have been replaced by rule `125` as the `$good_tcpo` variable contains the seven allowed outbound ports. [NOTE] ==== Remember that IPFW's performance is largely determined by the number of rules present in the ruleset. ==== [.programlisting] .... # Authorized outbound packets $cmd 120 $skip udp from any to x.x.x.x 53 out via $pif $ks $cmd 121 $skip udp from any to x.x.x.x 67 out via $pif $ks $cmd 125 $skip tcp from any to any $good_tcpo out via $pif setup $ks $cmd 130 $skip icmp from any to any out via $pif $ks .... The inbound rules remain the same, except for the very last rule which removes the `via $pif` in order to catch both inbound and outbound rules. The NAT rule must follow this last outbound rule, must have a higher number than that last rule, and the rule number must be referenced by the `skipto` action. In this ruleset, rule number `1000` handles passing all packets to our configured instance for NAT processing. The next rule allows any packet which has undergone NAT processing to pass. [.programlisting] .... $cmd 999 deny log all from any to any $cmd 1000 nat 1 ip from any to any out via $pif # skipto location for outbound stateful rules $cmd 1001 allow ip from any to any .... In this example, rules `100`, `101`, `125`, `1000`, and `1001` control the address translation of the outbound and inbound packets so that the entries in the dynamic state table always register the private LANIP address. Consider an internal web browser which initializes a new outbound HTTP session over port 80. When the first outbound packet enters the firewall, it does not match rule `100` because it is headed out rather than in. It passes rule `101` because this is the first packet and it has not been posted to the dynamic state table yet. The packet finally matches rule `125` as it is outbound on an allowed port and has a source IP address from the internal LAN. On matching this rule, two actions take place. First, the `keep-state` action adds an entry to the dynamic state table and the specified action, `skipto rule 1000`, is executed. Next, the packet undergoes NAT and is sent out to the Internet. This packet makes its way to the destination web server, where a response packet is generated and sent back. This new packet enters the top of the ruleset. It matches rule `100` and has its destination IP address mapped back to the original internal address. It then is processed by the `check-state` rule, is found in the table as an existing session, and is released to the LAN. On the inbound side, the ruleset has to deny bad packets and allow only authorized services. A packet which matches an inbound rule is posted to the dynamic state table and the packet is released to the LAN. The packet generated as a response is recognized by the `check-state` rule as belonging to an existing session. It is then sent to rule `1000` to undergo NAT before being released to the outbound interface. [NOTE] ==== Transitioning from userland man:natd[8] to in-kernel NAT might appear seamless at first but there is small catch. When using the GENERIC kernel, IPFW will load the [.filename]#libalias.ko# kernel module, when `firewall_nat_enable` is enabled in [.filename]#/etc/rc.conf#. The [.filename]#libalias.ko# kernel module only provides basic NAT functionality, whereas the userland implementation man:natd[8] has all NAT functionality available in its userland library without any extra configuration. All functionality refers to the following kernel modules that can additionally be loaded when needed besides the standard [.filename]#libalias.ko# kernel module: [.filename]#alias_ftp.ko#, [.filename]#alias_bbt.ko#, [.filename]#skinny.ko#, [.filename]#irc.ko#, [.filename]#alias_pptp.ko# and [.filename]#alias_smedia.ko# using the `kld_list` directive in [.filename]#/etc/rc.conf#. If a custom kernel is used, the full functionality of the userland library can be compiled in, in the kernel, using the `options LIBALIAS`. ==== ==== Port Redirection The drawback with NAT in general is that the LAN clients are not accessible from the Internet. Clients on the LAN can make outgoing connections to the world but cannot receive incoming ones. This presents a problem if trying to run Internet services on one of the LAN client machines. A simple way around this is to redirect selected Internet ports on the NAT providing machine to a LAN client. For example, an IRC server runs on client `A` and a web server runs on client `B`. For this to work properly, connections received on ports 6667 (IRC) and 80 (HTTP) must be redirected to the respective machines. With in-kernel NAT all configuration is done in the NAT instance configuration. For a full list of options that an in-kernel NAT instance can use, consult man:ipfw[8]. The IPFW syntax follows the syntax of natd. The syntax for `redirect_port` is as follows: [.programlisting] .... redirect_port proto targetIP:targetPORT[-targetPORT] [aliasIP:]aliasPORT[-aliasPORT] [remoteIP[:remotePORT[-remotePORT]]] .... To configure the above example setup, the arguments should be: [.programlisting] .... redirect_port tcp 192.168.0.2:6667 6667 redirect_port tcp 192.168.0.3:80 80 .... After adding these arguments to the configuration of NAT instance 1 in the above ruleset, the TCP ports will be port forwarded to the LAN client machines running the IRC and HTTP services. [.programlisting] .... ipfw -q nat 1 config if $pif same_ports unreg_only reset \ redirect_port tcp 192.168.0.2:6667 6667 \ redirect_port tcp 192.168.0.3:80 80 .... Port ranges over individual ports can be indicated with `redirect_port`. For example, _tcp 192.168.0.2:2000-3000 2000-3000_ would redirect all connections received on ports 2000 to 3000 to ports 2000 to 3000 on client `A`. ==== Address Redirection Address redirection is useful if more than one IP address is available. Each LAN client can be assigned its own external IP address by man:ipfw[8], which will then rewrite outgoing packets from the LAN clients with the proper external IP address and redirects all traffic incoming on that particular IP address back to the specific LAN client. This is also known as static NAT. For example, if IP addresses `128.1.1.1`, `128.1.1.2`, and `128.1.1.3` are available, `128.1.1.1` can be used as the man:ipfw[8] machine's external IP address, while `128.1.1.2` and `128.1.1.3` are forwarded back to LAN clients `A` and `B`. The `redirect_address` syntax is as below, where `localIP` is the internal IP address of the LAN client, and `publicIP` the external IP address corresponding to the LAN client. [.programlisting] .... redirect_address localIP publicIP .... In the example, the arguments would read: [.programlisting] .... redirect_address 192.168.0.2 128.1.1.2 redirect_address 192.168.0.3 128.1.1.3 .... Like `redirect_port`, these arguments are placed in a NAT instance configuration. With address redirection, there is no need for port redirection, as all data received on a particular IP address is redirected. The external IP addresses on the man:ipfw[8] machine must be active and aliased to the external interface. Refer to man:rc.conf[5] for details. ==== Userspace NAT Let us start with a statement: the userspace NAT implementation: man:natd[8], has more overhead than in-kernel NAT. For man:natd[8] to translate packets, the packets have to be copied from the kernel to userspace and back which brings in extra overhead that is not present with in-kernel NAT. To enable the userpace NAT daemon man:natd[8] at boot time, the following is a minimum configuration in [.filename]#/etc/rc.conf#. Where `natd_interface` is set to the name of the NIC attached to the Internet. The man:rc[8] script of man:natd[8] will automatically check if a dynamic IP address is used and configure itself to handle that. [.programlisting] .... gateway_enable="YES" natd_enable="YES" natd_interface="rl0" .... In general, the above ruleset as explained for in-kernel NAT can also be used together with man:natd[8]. The exceptions are the configuration of the in-kernel NAT instance `(ipfw -q nat 1 config ...)` which is not needed together with reassemble rule 99 because its functionality is included in the `divert` action. Rule number 100 and 1000 will have to change sligthly as shown below. [.programlisting] .... $cmd 100 divert natd ip from any to any in via $pif $cmd 1000 divert natd ip from any to any out via $pif .... To configure port or address redirection, a similar syntax as with in-kernel NAT is used. Although, now, instead of specifying the configuration in our ruleset script like with in-kernel NAT, configuration of man:natd[8] is best done in a configuration file. To do this, an extra flag must be passed via [.filename]#/etc/rc.conf# which specifies the path of the configuration file. [.programlisting] .... natd_flags="-f /etc/natd.conf" .... [NOTE] ==== The specified file must contain a list of configuration options, one per line. For more information about the configuration file and possible variables, consult man:natd[8]. Below are two example entries, one per line: [.programlisting] .... redirect_port tcp 192.168.0.2:6667 6667 redirect_address 192.168.0.3 128.1.1.3 .... ==== [[firewalls-ipfw-cmd]] === The IPFW Command `ipfw` can be used to make manual, single rule additions or deletions to the active firewall while it is running. The problem with using this method is that all the changes are lost when the system reboots. It is recommended to instead write all the rules in a file and to use that file to load the rules at boot time and to replace the currently running firewall rules whenever that file changes. `ipfw` is a useful way to display the running firewall rules to the console screen. The IPFW accounting facility dynamically creates a counter for each rule that counts each packet that matches the rule. During the process of testing a rule, listing the rule with its counter is one way to determine if the rule is functioning as expected. To list all the running rules in sequence: [source,shell] .... # ipfw list .... To list all the running rules with a time stamp of when the last time the rule was matched: [source,shell] .... # ipfw -t list .... The next example lists accounting information and the packet count for matched rules along with the rules themselves. The first column is the rule number, followed by the number of matched packets and bytes, followed by the rule itself. [source,shell] .... # ipfw -a list .... To list dynamic rules in addition to static rules: [source,shell] .... # ipfw -d list .... To also show the expired dynamic rules: [source,shell] .... # ipfw -d -e list .... To zero the counters: [source,shell] .... # ipfw zero .... To zero the counters for just the rule with number _NUM_: [source,shell] .... # ipfw zero NUM .... ==== Logging Firewall Messages Even with the logging facility enabled, IPFW will not generate any rule logging on its own. The firewall administrator decides which rules in the ruleset will be logged, and adds the `log` keyword to those rules. Normally only deny rules are logged. It is customary to duplicate the "ipfw default deny everything" rule with the `log` keyword included as the last rule in the ruleset. This way, it is possible to see all the packets that did not match any of the rules in the ruleset. Logging is a two edged sword. If one is not careful, an over abundance of log data or a DoS attack can fill the disk with log files. Log messages are not only written to syslogd, but also are displayed on the root console screen and soon become annoying. The `IPFIREWALL_VERBOSE_LIMIT=5` kernel option limits the number of consecutive messages sent to man:syslogd[8], concerning the packet matching of a given rule. When this option is enabled in the kernel, the number of consecutive messages concerning a particular rule is capped at the number specified. There is nothing to be gained from 200 identical log messages. With this option set to five, five consecutive messages concerning a particular rule would be logged to syslogd and the remainder identical consecutive messages would be counted and posted to syslogd with a phrase like the following: [.programlisting] .... last message repeated 45 times .... All logged packets messages are written by default to [.filename]#/var/log/security#, which is defined in [.filename]#/etc/syslog.conf#. [[firewalls-ipfw-rules-script]] ==== Building a Rule Script Most experienced IPFW users create a file containing the rules and code them in a manner compatible with running them as a script. The major benefit of doing this is the firewall rules can be refreshed in mass without the need of rebooting the system to activate them. This method is convenient in testing new rules as the procedure can be executed as many times as needed. Being a script, symbolic substitution can be used for frequently used values to be substituted into multiple rules. This example script is compatible with the syntax used by the man:sh[1], man:csh[1], and man:tcsh[1] shells. Symbolic substitution fields are prefixed with a dollar sign ($). Symbolic fields do not have the $ prefix. The value to populate the symbolic field must be enclosed in double quotes (""). Start the rules file like this: [.programlisting] .... ############### start of example ipfw rules script ############# # ipfw -q -f flush # Delete all rules # Set defaults oif="tun0" # out interface odns="192.0.2.11" # ISP's DNS server IP address cmd="ipfw -q add " # build rule prefix ks="keep-state" # just too lazy to key this each time $cmd 00500 check-state $cmd 00502 deny all from any to any frag $cmd 00501 deny tcp from any to any established $cmd 00600 allow tcp from any to any 80 out via $oif setup $ks $cmd 00610 allow tcp from any to $odns 53 out via $oif setup $ks $cmd 00611 allow udp from any to $odns 53 out via $oif $ks ################### End of example ipfw rules script ############ .... The rules are not important as the focus of this example is how the symbolic substitution fields are populated. If the above example was in [.filename]#/etc/ipfw.rules#, the rules could be reloaded by the following command: [source,shell] .... # sh /etc/ipfw.rules .... [.filename]#/etc/ipfw.rules# can be located anywhere and the file can have any name. The same thing could be accomplished by running these commands by hand: [source,shell] .... # ipfw -q -f flush # ipfw -q add check-state # ipfw -q add deny all from any to any frag # ipfw -q add deny tcp from any to any established # ipfw -q add allow tcp from any to any 80 out via tun0 setup keep-state # ipfw -q add allow tcp from any to 192.0.2.11 53 out via tun0 setup keep-state # ipfw -q add 00611 allow udp from any to 192.0.2.11 53 out via tun0 keep-state .... [[firewalls-ipfw-kernelconfig]] === IPFW Kernel Options In order to statically compile IPFW support into a custom kernel, refer to the instructions in crossref:kernelconfig[kernelconfig,Configuring the FreeBSD Kernel]. The following options are available for the custom kernel configuration file: [.programlisting] .... options IPFIREWALL # enables IPFW options IPFIREWALL_VERBOSE # enables logging for rules with log keyword to syslogd(8) options IPFIREWALL_VERBOSE_LIMIT=5 # limits number of logged packets per-entry options IPFIREWALL_DEFAULT_TO_ACCEPT # sets default policy to pass what is not explicitly denied options IPFIREWALL_NAT # enables basic in-kernel NAT support options LIBALIAS # enables full in-kernel NAT support options IPFIREWALL_NAT64 # enables in-kernel NAT64 support options IPFIREWALL_NPTV6 # enables in-kernel IPv6 NPT support options IPFIREWALL_PMOD # enables protocols modification module support options IPDIVERT # enables NAT through natd(8) .... [NOTE] ==== IPFW can be loaded as a kernel module: options above are built by default as modules or can be set at runtime using tunables. ==== [[firewalls-ipf]] == IPFILTER (IPF) IPFILTER, also known as IPF, is a cross-platform, open source firewall which has been ported to several operating systems, including FreeBSD, NetBSD, OpenBSD, and Solaris(TM). IPFILTER is a kernel-side firewall and NAT mechanism that can be controlled and monitored by userland programs. Firewall rules can be set or deleted using ipf, NAT rules can be set or deleted using ipnat, run-time statistics for the kernel parts of IPFILTER can be printed using ipfstat, and ipmon can be used to log IPFILTER actions to the system log files. IPF was originally written using a rule processing logic of "the last matching rule wins" and only used stateless rules. Since then, IPF has been enhanced to include the `quick` and `keep state` options. The IPF FAQ is at http://www.phildev.net/ipf/index.html[http://www.phildev.net/ipf/index.html]. A searchable archive of the IPFilter mailing list is available at http://marc.info/?l=ipfilter[http://marc.info/?l=ipfilter]. This section of the Handbook focuses on IPF as it pertains to FreeBSD. It provides examples of rules that contain the `quick` and `keep state` options. === Enabling IPF IPF is included in the basic FreeBSD install as a kernel loadable module, meaning that a custom kernel is not needed in order to enable IPF. For users who prefer to statically compile IPF support into a custom kernel, refer to the instructions in crossref:kernelconfig[kernelconfig,Configuring the FreeBSD Kernel]. The following kernel options are available: [.programlisting] .... options IPFILTER options IPFILTER_LOG options IPFILTER_LOOKUP options IPFILTER_DEFAULT_BLOCK .... where `options IPFILTER` enables support for IPFILTER, `options IPFILTER_LOG` enables IPF logging using the [.filename]#ipl# packet logging pseudo-device for every rule that has the `log` keyword, `IPFILTER_LOOKUP` enables IP pools in order to speed up IP lookups, and `options IPFILTER_DEFAULT_BLOCK` changes the default behavior so that any packet not matching a firewall `pass` rule gets blocked. To configure the system to enable IPF at boot time, add the following entries to [.filename]#/etc/rc.conf#. These entries will also enable logging and `default pass all`. To change the default policy to `block all` without compiling a custom kernel, remember to add a `block all` rule at the end of the ruleset. [.programlisting] .... ipfilter_enable="YES" # Start ipf firewall ipfilter_rules="/etc/ipf.rules" # loads rules definition text file ipv6_ipfilter_rules="/etc/ipf6.rules" # loads rules definition text file for IPv6 ipmon_enable="YES" # Start IP monitor log ipmon_flags="-Ds" # D = start as daemon # s = log to syslog # v = log tcp window, ack, seq # n = map IP & port to names .... If NAT functionality is needed, also add these lines: [.programlisting] .... gateway_enable="YES" # Enable as LAN gateway ipnat_enable="YES" # Start ipnat function ipnat_rules="/etc/ipnat.rules" # rules definition file for ipnat .... Then, to start IPF now: [.programlisting] .... # service ipfilter start .... To load the firewall rules, specify the name of the ruleset file using `ipf`. The following command can be used to replace the currently running firewall rules: [source,shell] .... # ipf -Fa -f /etc/ipf.rules .... where `-Fa` flushes all the internal rules tables and `-f` specifies the file containing the rules to load. This provides the ability to make changes to a custom ruleset and update the running firewall with a fresh copy of the rules without having to reboot the system. This method is convenient for testing new rules as the procedure can be executed as many times as needed. Refer to man:ipf[8] for details on the other flags available with this command. === IPF Rule Syntax This section describes the IPF rule syntax used to create stateful rules. When creating rules, keep in mind that unless the `quick` keyword appears in a rule, every rule is read in order, with the _last matching rule_ being the one that is applied. This means that even if the first rule to match a packet is a `pass`, if there is a later matching rule that is a `block`, the packet will be dropped. Sample rulesets can be found in [.filename]#/usr/share/examples/ipfilter#. When creating rules, a `#` character is used to mark the start of a comment and may appear at the end of a rule, to explain that rule's function, or on its own line. Any blank lines are ignored. The keywords which are used in rules must be written in a specific order, from left to right. Some keywords are mandatory while others are optional. Some keywords have sub-options which may be keywords themselves and also include more sub-options. The keyword order is as follows, where the words shown in uppercase represent a variable and the words shown in lowercase must precede the variable that follows it: `_ACTION DIRECTION OPTIONS proto PROTO_TYPE from SRC_ADDR SRC_PORT to DST_ADDR DST_PORT TCP_FLAG|ICMP_TYPE keep state STATE_` This section describes each of these keywords and their options. It is not an exhaustive list of every possible option. Refer to man:ipf[5] for a complete description of the rule syntax that can be used when creating IPF rules and examples for using each keyword. ACTION:: The action keyword indicates what to do with the packet if it matches that rule. Every rule _must_ have an action. The following actions are recognized: + `block`: drops the packet. + `pass`: allows the packet. + `log`: generates a log record. + `count`: counts the number of packets and bytes which can provide an indication of how often a rule is used. + `auth`: queues the packet for further processing by another program. + `call`: provides access to functions built into IPF that allow more complex actions. + `decapsulate`: removes any headers in order to process the contents of the packet. DIRECTION:: Next, each rule must explicitly state the direction of traffic using one of these keywords: + `in`: the rule is applied against an inbound packet. + `out`: the rule is applied against an outbound packet. + `all`: the rule applies to either direction. + If the system has multiple interfaces, the interface can be specified along with the direction. An example would be `in on fxp0`. OPTIONS:: Options are optional. However, if multiple options are specified, they must be used in the order shown here. + `log`: when performing the specified ACTION, the contents of the packet's headers will be written to the man:ipl[4] packet log pseudo-device. + `quick`: if a packet matches this rule, the ACTION specified by the rule occurs and no further processing of any following rules will occur for this packet. + `on`: must be followed by the interface name as displayed by man:ifconfig[8]. The rule will only match if the packet is going through the specified interface in the specified direction. + When using the `log` keyword, the following qualifiers may be used in this order: + `body`: indicates that the first 128 bytes of the packet contents will be logged after the headers. + `first`: if the `log` keyword is being used in conjunction with a `keep state` option, this option is recommended so that only the triggering packet is logged and not every packet which matches the stateful connection. + Additional options are available to specify error return messages. Refer to man:ipf[5] for more details. PROTO_TYPE:: The protocol type is optional. However, it is mandatory if the rule needs to specify a SRC_PORT or a DST_PORT as it defines the type of protocol. When specifying the type of protocol, use the `proto` keyword followed by either a protocol number or name from [.filename]#/etc/protocols#. Example protocol names include `tcp`, `udp`, or `icmp`. If PROTO_TYPE is specified but no SRC_PORT or DST_PORT is specified, all port numbers for that protocol will match that rule. SRC_ADDR:: The `from` keyword is mandatory and is followed by a keyword which represents the source of the packet. The source can be a hostname, an IP address followed by the CIDR mask, an address pool, or the keyword `all`. Refer to man:ipf[5] for examples. + There is no way to match ranges of IP addresses which do not express themselves easily using the dotted numeric form / mask-length notation. The package:net-mgmt/ipcalc[] package or port may be used to ease the calculation of the CIDR mask. Additional information is available at the utility's web page: http://jodies.de/ipcalc[http://jodies.de/ipcalc]. SRC_PORT:: The port number of the source is optional. However, if it is used, it requires PROTO_TYPE to be first defined in the rule. The port number must also be preceded by the `proto` keyword. + A number of different comparison operators are supported: `=` (equal to), `!=` (not equal to), `<` (less than), `>` (greater than), `<=` (less than or equal to), and `>=` (greater than or equal to). + To specify port ranges, place the two port numbers between `<>` (less than and greater than ), `><` (greater than and less than ), or `:` (greater than or equal to and less than or equal to). DST_ADDR:: The `to` keyword is mandatory and is followed by a keyword which represents the destination of the packet. Similar to SRC_ADDR, it can be a hostname, an IP address followed by the CIDR mask, an address pool, or the keyword `all`. DST_PORT:: Similar to SRC_PORT, the port number of the destination is optional. However, if it is used, it requires PROTO_TYPE to be first defined in the rule. The port number must also be preceded by the `proto` keyword. TCP_FLAG|ICMP_TYPE:: If `tcp` is specified as the PROTO_TYPE, flags can be specified as letters, where each letter represents one of the possible TCP flags used to determine the state of a connection. Possible values are: `S` (SYN), `A` (ACK), `P` (PSH), `F` (FIN), `U` (URG), `R` (RST), `C` (CWN), and `E` (ECN). + If `icmp` is specified as the PROTO_TYPE, the ICMP type to match can be specified. Refer to man:ipf[5] for the allowable types. STATE:: If a `pass` rule contains `keep state`, IPF will add an entry to its dynamic state table and allow subsequent packets that match the connection. IPF can track state for TCP, UDP, and ICMP sessions. Any packet that IPF can be certain is part of an active session, even if it is a different protocol, will be allowed. + In IPF, packets destined to go out through the interface connected to the public Internet are first checked against the dynamic state table. If the packet matches the next expected packet comprising an active session conversation, it exits the firewall and the state of the session conversation flow is updated in the dynamic state table. Packets that do not belong to an already active session are checked against the outbound ruleset. Packets coming in from the interface connected to the public Internet are first checked against the dynamic state table. If the packet matches the next expected packet comprising an active session, it exits the firewall and the state of the session conversation flow is updated in the dynamic state table. Packets that do not belong to an already active session are checked against the inbound ruleset. + Several keywords can be added after `keep state`. If used, these keywords set various options that control stateful filtering, such as setting connection limits or connection age. Refer to man:ipf[5] for the list of available options and their descriptions. === Example Ruleset This section demonstrates how to create an example ruleset which only allows services matching `pass` rules and blocks all others. FreeBSD uses the loopback interface ([.filename]#lo0#) and the IP address `127.0.0.1` for internal communication. The firewall ruleset must contain rules to allow free movement of these internally used packets: [.programlisting] .... # no restrictions on loopback interface pass in quick on lo0 all pass out quick on lo0 all .... The public interface connected to the Internet is used to authorize and control access of all outbound and inbound connections. If one or more interfaces are cabled to private networks, those internal interfaces may require rules to allow packets originating from the LAN to flow between the internal networks or to the interface attached to the Internet. The ruleset should be organized into three major sections: any trusted internal interfaces, outbound connections through the public interface, and inbound connections through the public interface. These two rules allow all traffic to pass through a trusted LAN interface named [.filename]#xl0#: [.programlisting] .... # no restrictions on inside LAN interface for private network pass out quick on xl0 all pass in quick on xl0 all .... The rules for the public interface's outbound and inbound sections should have the most frequently matched rules placed before less commonly matched rules, with the last rule in the section blocking and logging all packets for that interface and direction. This set of rules defines the outbound section of the public interface named [.filename]#dc0#. These rules keep state and identify the specific services that internal systems are authorized for public Internet access. All the rules use `quick` and specify the appropriate port numbers and, where applicable, destination addresses. [.programlisting] .... # interface facing Internet (outbound) # Matches session start requests originating from or behind the # firewall, destined for the Internet. # Allow outbound access to public DNS servers. # Replace x.x.x. with address listed in /etc/resolv.conf. # Repeat for each DNS server. pass out quick on dc0 proto tcp from any to x.x.x. port = 53 flags S keep state pass out quick on dc0 proto udp from any to xxx port = 53 keep state # Allow access to ISP's specified DHCP server for cable or DSL networks. # Use the first rule, then check log for the IP address of DHCP server. # Then, uncomment the second rule, replace z.z.z.z with the IP address, # and comment out the first rule pass out log quick on dc0 proto udp from any to any port = 67 keep state #pass out quick on dc0 proto udp from any to z.z.z.z port = 67 keep state # Allow HTTP and HTTPS pass out quick on dc0 proto tcp from any to any port = 80 flags S keep state pass out quick on dc0 proto tcp from any to any port = 443 flags S keep state # Allow email pass out quick on dc0 proto tcp from any to any port = 110 flags S keep state pass out quick on dc0 proto tcp from any to any port = 25 flags S keep state # Allow NTP pass out quick on dc0 proto tcp from any to any port = 37 flags S keep state # Allow FTP pass out quick on dc0 proto tcp from any to any port = 21 flags S keep state # Allow SSH pass out quick on dc0 proto tcp from any to any port = 22 flags S keep state # Allow ping pass out quick on dc0 proto icmp from any to any icmp-type 8 keep state # Block and log everything else block out log first quick on dc0 all .... This example of the rules in the inbound section of the public interface blocks all undesirable packets first. This reduces the number of packets that are logged by the last rule. [.programlisting] .... # interface facing Internet (inbound) # Block all inbound traffic from non-routable or reserved address spaces block in quick on dc0 from 192.168.0.0/16 to any #RFC 1918 private IP block in quick on dc0 from 172.16.0.0/12 to any #RFC 1918 private IP block in quick on dc0 from 10.0.0.0/8 to any #RFC 1918 private IP block in quick on dc0 from 127.0.0.0/8 to any #loopback block in quick on dc0 from 0.0.0.0/8 to any #loopback block in quick on dc0 from 169.254.0.0/16 to any #DHCP auto-config block in quick on dc0 from 192.0.2.0/24 to any #reserved for docs block in quick on dc0 from 204.152.64.0/23 to any #Sun cluster interconnect block in quick on dc0 from 224.0.0.0/3 to any #Class D & E multicast # Block fragments and too short tcp packets block in quick on dc0 all with frags block in quick on dc0 proto tcp all with short # block source routed packets block in quick on dc0 all with opt lsrr block in quick on dc0 all with opt ssrr # Block OS fingerprint attempts and log first occurrence block in log first quick on dc0 proto tcp from any to any flags FUP # Block anything with special options block in quick on dc0 all with ipopts # Block public pings and ident block in quick on dc0 proto icmp all icmp-type 8 block in quick on dc0 proto tcp from any to any port = 113 # Block incoming Netbios services block in log first quick on dc0 proto tcp/udp from any to any port = 137 block in log first quick on dc0 proto tcp/udp from any to any port = 138 block in log first quick on dc0 proto tcp/udp from any to any port = 139 block in log first quick on dc0 proto tcp/udp from any to any port = 81 .... Any time there are logged messages on a rule with the `log first` option, run `ipfstat -hio` to evaluate how many times the rule has been matched. A large number of matches may indicate that the system is under attack. The rest of the rules in the inbound section define which connections are allowed to be initiated from the Internet. The last rule denies all connections which were not explicitly allowed by previous rules in this section. [.programlisting] .... # Allow traffic in from ISP's DHCP server. Replace z.z.z.z with # the same IP address used in the outbound section. pass in quick on dc0 proto udp from z.z.z.z to any port = 68 keep state # Allow public connections to specified internal web server pass in quick on dc0 proto tcp from any to x.x.x.x port = 80 flags S keep state # Block and log only first occurrence of all remaining traffic. block in log first quick on dc0 all .... === Configuring NAT To enable NAT, add these statements to [.filename]#/etc/rc.conf# and specify the name of the file containing the NAT rules: [.programlisting] .... gateway_enable="YES" ipnat_enable="YES" ipnat_rules="/etc/ipnat.rules" .... NAT rules are flexible and can accomplish many different things to fit the needs of both commercial and home users. The rule syntax presented here has been simplified to demonstrate common usage. For a complete rule syntax description, refer to man:ipnat[5]. The basic syntax for a NAT rule is as follows, where `map` starts the rule and _IF_ should be replaced with the name of the external interface: [.programlisting] .... map IF LAN_IP_RANGE -> PUBLIC_ADDRESS .... The _LAN_IP_RANGE_ is the range of IP addresses used by internal clients. Usually, it is a private address range such as `192.168.1.0/24`. The _PUBLIC_ADDRESS_ can either be the static external IP address or the keyword `0/32` which represents the IP address assigned to _IF_. In IPF, when a packet arrives at the firewall from the LAN with a public destination, it first passes through the outbound rules of the firewall ruleset. Then, the packet is passed to the NAT ruleset which is read from the top down, where the first matching rule wins. IPF tests each NAT rule against the packet's interface name and source IP address. When a packet's interface name matches a NAT rule, the packet's source IP address in the private LAN is checked to see if it falls within the IP address range specified in _LAN_IP_RANGE_. On a match, the packet has its source IP address rewritten with the public IP address specified by _PUBLIC_ADDRESS_. IPF posts an entry in its internal NAT table so that when the packet returns from the Internet, it can be mapped back to its original private IP address before being passed to the firewall rules for further processing. For networks that have large numbers of internal systems or multiple subnets, the process of funneling every private IP address into a single public IP address becomes a resource problem. Two methods are available to relieve this issue. The first method is to assign a range of ports to use as source ports. By adding the `portmap` keyword, NAT can be directed to only use source ports in the specified range: [.programlisting] .... map dc0 192.168.1.0/24 -> 0/32 portmap tcp/udp 20000:60000 .... Alternately, use the `auto` keyword which tells NAT to determine the ports that are available for use: [.programlisting] .... map dc0 192.168.1.0/24 -> 0/32 portmap tcp/udp auto .... The second method is to use a pool of public addresses. This is useful when there are too many LAN addresses to fit into a single public address and a block of public IP addresses is available. These public addresses can be used as a pool from which NAT selects an IP address as a packet's address is mapped on its way out. The range of public IP addresses can be specified using a netmask or CIDR notation. These two rules are equivalent: [.programlisting] .... map dc0 192.168.1.0/24 -> 204.134.75.0/255.255.255.0 map dc0 192.168.1.0/24 -> 204.134.75.0/24 .... A common practice is to have a publically accessible web server or mail server segregated to an internal network segment. The traffic from these servers still has to undergo NAT, but port redirection is needed to direct inbound traffic to the correct server. For example, to map a web server using the internal address `10.0.10.25` to its public IP address of `20.20.20.5`, use this rule: [.programlisting] .... rdr dc0 20.20.20.5/32 port 80 -> 10.0.10.25 port 80 .... If it is the only web server, this rule would also work as it redirects all external HTTP requests to `10.0.10.25`: [.programlisting] .... rdr dc0 0.0.0.0/0 port 80 -> 10.0.10.25 port 80 .... IPF has a built in FTP proxy which can be used with NAT. It monitors all outbound traffic for active or passive FTP connection requests and dynamically creates temporary filter rules containing the port number used by the FTP data channel. This eliminates the need to open large ranges of high order ports for FTP connections. In this example, the first rule calls the proxy for outbound FTP traffic from the internal LAN. The second rule passes the FTP traffic from the firewall to the Internet, and the third rule handles all non-FTP traffic from the internal LAN: [.programlisting] .... map dc0 10.0.10.0/29 -> 0/32 proxy port 21 ftp/tcp map dc0 0.0.0.0/0 -> 0/32 proxy port 21 ftp/tcp map dc0 10.0.10.0/29 -> 0/32 .... The FTP `map` rules go before the NAT rule so that when a packet matches an FTP rule, the FTP proxy creates temporary filter rules to let the FTP session packets pass and undergo NAT. All LAN packets that are not FTP will not match the FTP rules but will undergo NAT if they match the third rule. Without the FTP proxy, the following firewall rules would instead be needed. Note that without the proxy, all ports above `1024` need to be allowed: [.programlisting] .... # Allow out LAN PC client FTP to public Internet # Active and passive modes pass out quick on rl0 proto tcp from any to any port = 21 flags S keep state # Allow out passive mode data channel high order port numbers pass out quick on rl0 proto tcp from any to any port > 1024 flags S keep state # Active mode let data channel in from FTP server pass in quick on rl0 proto tcp from any to any port = 20 flags S keep state .... Whenever the file containing the NAT rules is edited, run `ipnat` with `-CF` to delete the current NAT rules and flush the contents of the dynamic translation table. Include `-f` and specify the name of the NAT ruleset to load: [source,shell] .... # ipnat -CF -f /etc/ipnat.rules .... To display the NAT statistics: [source,shell] .... # ipnat -s .... To list the NAT table's current mappings: [source,shell] .... # ipnat -l .... To turn verbose mode on and display information relating to rule processing and active rules and table entries: [source,shell] .... # ipnat -v .... === Viewing IPF Statistics IPF includes man:ipfstat[8] which can be used to retrieve and display statistics which are gathered as packets match rules as they go through the firewall. Statistics are accumulated since the firewall was last started or since the last time they were reset to zero using `ipf -Z`. The default `ipfstat` output looks like this: [source,shell] .... input packets: blocked 99286 passed 1255609 nomatch 14686 counted 0 output packets: blocked 4200 passed 1284345 nomatch 14687 counted 0 input packets logged: blocked 99286 passed 0 output packets logged: blocked 0 passed 0 packets logged: input 0 output 0 log failures: input 3898 output 0 fragment state(in): kept 0 lost 0 fragment state(out): kept 0 lost 0 packet state(in): kept 169364 lost 0 packet state(out): kept 431395 lost 0 ICMP replies: 0 TCP RSTs sent: 0 Result cache hits(in): 1215208 (out): 1098963 IN Pullups succeeded: 2 failed: 0 OUT Pullups succeeded: 0 failed: 0 Fastroute successes: 0 failures: 0 TCP cksum fails(in): 0 (out): 0 Packet log flags set: (0) .... Several options are available. When supplied with either `-i` for inbound or `-o` for outbound, the command will retrieve and display the appropriate list of filter rules currently installed and in use by the kernel. To also see the rule numbers, include `-n`. For example, `ipfstat -on` displays the outbound rules table with rule numbers: [source,shell] .... @1 pass out on xl0 from any to any @2 block out on dc0 from any to any @3 pass out quick on dc0 proto tcp/udp from any to any keep state .... Include `-h` to prefix each rule with a count of how many times the rule was matched. For example, `ipfstat -oh` displays the outbound internal rules table, prefixing each rule with its usage count: [source,shell] .... 2451423 pass out on xl0 from any to any 354727 block out on dc0 from any to any 430918 pass out quick on dc0 proto tcp/udp from any to any keep state .... To display the state table in a format similar to man:top[1], use `ipfstat -t`. When the firewall is under attack, this option provides the ability to identify and see the attacking packets. The optional sub-flags give the ability to select the destination or source IP, port, or protocol to be monitored in real time. Refer to man:ipfstat[8] for details. === IPF Logging IPF provides `ipmon`, which can be used to write the firewall's logging information in a human readable format. It requires that `options IPFILTER_LOG` be first added to a custom kernel using the instructions in crossref:kernelconfig[kernelconfig,Configuring the FreeBSD Kernel]. This command is typically run in daemon mode in order to provide a continuous system log file so that logging of past events may be reviewed. Since FreeBSD has a built in man:syslogd[8] facility to automatically rotate system logs, the default [.filename]#rc.conf# `ipmon_flags` statement uses `-Ds`: [.programlisting] .... ipmon_flags="-Ds" # D = start as daemon # s = log to syslog # v = log tcp window, ack, seq # n = map IP & port to names .... Logging provides the ability to review, after the fact, information such as which packets were dropped, what addresses they came from, and where they were going. This information is useful in tracking down attackers. Once the logging facility is enabled in [.filename]#rc.conf# and started with `service ipmon start`, IPF will only log the rules which contain the `log` keyword. The firewall administrator decides which rules in the ruleset should be logged and normally only deny rules are logged. It is customary to include the `log` keyword in the last rule in the ruleset. This makes it possible to see all the packets that did not match any of the rules in the ruleset. By default, `ipmon -Ds` mode uses `local0` as the logging facility. The following logging levels can be used to further segregate the logged data: [source,shell] .... LOG_INFO - packets logged using the "log" keyword as the action rather than pass or block. LOG_NOTICE - packets logged which are also passed LOG_WARNING - packets logged which are also blocked LOG_ERR - packets which have been logged and which can be considered short due to an incomplete header .... In order to setup IPF to log all data to [.filename]#/var/log/ipfilter.log#, first create the empty file: [source,shell] .... # touch /var/log/ipfilter.log .... Then, to write all logged messages to the specified file, add the following statement to [.filename]#/etc/syslog.conf#: [.programlisting] .... local0.* /var/log/ipfilter.log .... To activate the changes and instruct man:syslogd[8] to read the modified [.filename]#/etc/syslog.conf#, run `service syslogd reload`. Do not forget to edit [.filename]#/etc/newsyslog.conf# to rotate the new log file. Messages generated by `ipmon` consist of data fields separated by white space. Fields common to all messages are: . The date of packet receipt. . The time of packet receipt. This is in the form HH:MM:SS.F, for hours, minutes, seconds, and fractions of a second. . The name of the interface that processed the packet. . The group and rule number of the rule in the format `@0:17`. . The action: `p` for passed, `b` for blocked, `S` for a short packet, `n` did not match any rules, and `L` for a log rule. . The addresses written as three fields: the source address and port separated by a comma, the -> symbol, and the destination address and port. For example: `209.53.17.22,80 -> 198.73.220.17,1722`. . `PR` followed by the protocol name or number: for example, `PR tcp`. . `len` followed by the header length and total length of the packet: for example, `len 20 40`. If the packet is a TCP packet, there will be an additional field starting with a hyphen followed by letters corresponding to any flags that were set. Refer to man:ipf[5] for a list of letters and their flags. If the packet is an ICMP packet, there will be two fields at the end: the first always being "icmp" and the next being the ICMP message and sub-message type, separated by a slash. For example: `icmp 3/3` for a port unreachable message. [[firewalls-blacklistd]] == Blacklistd Blacklistd is a daemon listening to sockets to receive notifications from other daemons about connection attempts that failed or were successful. It is most widely used in blocking too many connection attempts on open ports. A prime example is SSH running on the internet getting a lot of requests from bots or scripts trying to guess passwords and gain access. Using blacklistd, the daemon can notify the firewall to create a filter rule to block excessive connection attempts from a single source after a number of tries. Blacklistd was first developed on NetBSD and appeared there in version 7. FreeBSD 11 imported blacklistd from NetBSD. This chapter describes how to set up blacklistd, configure it, and provides examples on how to use it. Readers should be familiar with basic firewall concepts like rules. For details, refer to the firewall chapter. PF is used in the examples, but other firewalls available on FreeBSD should be able to work with blacklistd, too. === Enabling Blacklistd The main configuration for blacklistd is stored in man:blacklistd.conf[5]. Various command line options are also available to change blacklistd's run-time behavior. Persistent configuration across reboots should be stored in [.filename]#/etc/blacklistd.conf#. To enable the daemon during system boot, add a `blacklistd_enable` line to [.filename]#/etc/rc.conf# like this: [source,shell] .... # sysrc blacklistd_enable=yes .... To start the service manually, run this command: [source,shell] .... # service blacklistd start .... === Creating a Blacklistd Ruleset Rules for blacklistd are configured in man:blacklistd.conf[5] with one entry per line. Each rule contains a tuple separated by spaces or tabs. Rules either belong to a `local` or a `remote`, which applies to the machine where blacklistd is running or an outside source, respectively. ==== Local Rules An example blacklistd.conf entry for a local rule looks like this: [.programlisting] .... [local] ssh stream * * * 3 24h .... All rules that follow the `[local]` section are treated as local rules (which is the default), applying to the local machine. When a `[remote]` section is encountered, all rules that follow it are handled as remote machine rules. Seven fields define a rule separated by either tabs or spaces. -The first four fields identify the traffic that should be blacklisted. +The first four fields identify the traffic that should be blocklisted. The three fields that follow define backlistd's behavior. Wildcards are denoted as asterisks (`*`), matching anything in this field. The first field defines the location. In local rules, these are the network ports. The syntax for the location field is as follows: [.programlisting] .... [address|interface][/mask][:port] .... Adressses can be specified as IPv4 in numeric format or IPv6 in square brackets. An interface name like `_em0_` can also be used. The socket type is defined by the second field. TCP sockets are of type `stream`, whereas UDP is denoted as `dgram`. The example above uses TCP, since SSH is using that protocol. A protocol can be used in the third field of a blacklistd rule. The following protocols can be used: `tcp`, `udp`, `tcp6`, `udp6`, or numeric. A wildcard, like in the example, is typically used to match all protocols unless there is a reason to distinguish traffic by a certain protocol. In the fourth field, the effective user or owner of the daemon process that is reporting the event is defined. The username or UID can be used here, as well as a wildcard (see example rule above). The packet filter rule name is declared by the fifth field, which starts the behavior part of the rule. By default, blacklistd puts all blocks under a pf anchor called `blacklistd` in [.filename]#pf.conf# like this: [.programlisting] .... anchor "blacklistd/*" in on $ext_if block in pass out .... -For separate blacklists, an anchor name can be used in this field. +For separate blocklists, an anchor name can be used in this field. In other cases, the wildcard will suffice. When a name starts with a hyphen (`-`) it means that an anchor with the default rule name prepended should be used. A modified example from the above using the hyphen would look like this: [.programlisting] .... ssh stream * * -ssh 3 24h .... -With such a rule, any new blacklist rules are added to an anchor called `blacklistd-ssh`. +With such a rule, any new blocklist rules are added to an anchor called `blacklistd-ssh`. To block whole subnets for a single rule violation, a `/` in the rule name can be used. This causes the remaining portion of the name to be interpreted as the mask to be applied to the address specified in the rule. For example, this rule would block every address adjoining `/24`. [.programlisting] .... 22 stream tcp * */24 3 24h .... [NOTE] ==== It is important to specify the proper protocol here. IPv4 and IPv6 treat /24 differently, that is the reason why `*` cannot be used in the third field for this rule. ==== This rule defines that if any one host in that network is misbehaving, everything else on that network will be blocked, too. -The sixth field, called `nfail`, sets the number of login failures required to blacklist the remote IP in question. +The sixth field, called `nfail`, sets the number of login failures required to blocklist the remote IP in question. When a wildcard is used at this position, it means that blocks will never happen. In the example rule above, a limit of three is defined meaning that after three attempts to log into SSH on one connection, the IP is blocked. -The last field in a blacklistd rule definition specifies how long a host is blacklisted. +The last field in a blacklistd rule definition specifies how long a host is blocklisted. The default unit is seconds, but suffixes like `m`, `h`, and `d` can also be specified for minutes, hours, and days, respectively. The example rule in its entirety means that after three times authenticating to SSH will result in a new PF block rule for that host. Rule matches are performed by first checking local rules one after another, from most specific to least specific. When a match occurs, the `remote` rules are applied and the name, `nfail`, and disable fields are changed by the `remote` rule that matched. ==== Remote Rules Remote rules are used to specify how blacklistd changes its behavior depending on the remote host currently being evaluated. Each field in a remote rule is the same as in a local rule. The only difference is in the way blacklistd is using them. To explain it, this example rule is used: [.programlisting] .... [remote] 203.0.113.128/25 * * * =/25 = 48h .... The address field can be an IP address (either v4 or v6), a port or both. This allows setting special rules for a specific remote address range like in this example. The fields for type, protocol and owner are identically interpreted as in the local rule. The name fields is different though: the equal sign (`=`) in a remote rule tells blacklistd to use the value from the matching local rule. It means that the firewall rule entry is taken and the `/25` prefix (a netmask of `255.255.255.128`) is added. -When a connection from that address range is blacklisted, the entire subnet is affected. +When a connection from that address range is blocklisted, the entire subnet is affected. A PF anchor name can also be used here, in which case blacklistd will add rules for this address block to the anchor of that name. The default table is used when a wildcard is specified. A custom number of failures in the `nfail` column can be defined for an address. This is useful for exceptions to a specific rule, to maybe allow someone a less strict application of rules or a bit more leniency in login tries. Blocking is disabled when an asterisk is used in this sixth field. Remote rules allow a stricter enforcement of limits on attempts to log in compared to attempts coming from a local network like an office. === Blacklistd Client Configuration There are a few software packages in FreeBSD that can utilize blacklistd's functionality. The two most prominent ones are man:ftpd[8] and man:sshd[8] to block excessive connection attempts. To activate blacklistd in the SSH daemon, add the following line to [.filename]#/etc/ssh/sshd_config#: [.programlisting] .... UseBlacklist yes .... Restart sshd afterwards to make these changes take effect. Blacklisting for man:ftpd[8] is enabled using `-B`, either in [.filename]#/etc/inetd.conf# or as a flag in [.filename]#/etc/rc.conf# like this: [.programlisting] .... ftpd_flags="-B" .... That is all that is needed to make these programs talk to blacklistd. === Blacklistd Management Blacklistd provides the user with a management utility called man:blacklistctl[8]. -It displays blocked addresses and networks that are blacklisted by the rules defined in man:blacklistd.conf[5]. +It displays blocked addresses and networks that are blocklisted by the rules defined in man:blacklistd.conf[5]. To see the list of currently blocked hosts, use `dump` combined with `-b` like this. [source,shell] .... # blacklistctl dump -b address/ma:port id nfail last access 213.0.123.128/25:22 OK 6/3 2019/06/08 14:30:19 .... This example shows that there were 6 out of three permitted attempts on port 22 coming from the address range `213.0.123.128/25`. There are more attempts listed than are allowed because SSH allows a client to try multiple logins on a single TCP connection. A connection that is currently going on is not stopped by blacklistd. The last connection attempt is listed in the `last access` column of the output. -To see the remaining time that this host will be on the blacklist, add `-r` to the previous command. +To see the remaining time that this host will be on the blocklist, add `-r` to the previous command. [source,shell] .... # blacklistctl dump -br address/ma:port id nfail remaining time 213.0.123.128/25:22 OK 6/3 36s .... In this example, there are 36s seconds left until this host will not be blocked any more. === Removing Hosts from the Block List Sometimes it is necessary to remove a host from the block list before the remaining time expires. Unfortunately, there is no functionality in blacklistd to do that. However, it is possible to remove the address from the PF table using pfctl. For each blocked port, there is a child anchor inside the blacklistd anchor defined in [.filename]#/etc/pf.conf#. For example, if there is a child anchor for blocking port 22 it is called `blacklistd/22`. There is a table inside that child anchor that contains the blocked addresses. This table is called port followed by the port number. In this example, it would be called `port22`. With that information at hand, it is now possible to use man:pfctl[8] to display all addresses listed like this: [source,shell] .... # pfctl -a blacklistd/22 -t port22 -T show ... 213.0.123.128/25 ... .... After identifying the address to be unblocked from the list, the following command removes it from the list: [source,shell] .... # pfctl -a blacklistd/22 -t port22 -T delete 213.0.123.128/25 .... The address is now removed from PF, but will still show up in the blacklistctl list, since it does not know about any changes made in PF. The entry in blacklistd's database will eventually expire and be removed from its output eventually. The entry will be added again if the host is matching one of the block rules in blacklistd again.