Backport armv7 architecture support to stable/11
AbandonedPublic
Actions

Authored by gonzo on Apr 15 2018, 2:39 AM.

Details

Reviewers

imp
linimon
manu

Summary

This backport includes revisions 322453, 324339, 324340, 324341, 324363
plus some manual fixes for armv7eb cehcks in Makefiles in gnu.
I booted it on RPi2, boots fine:

root@rpi2-11:~ # uname -a
FreeBSD rpi2-11 11.1-STABLE FreeBSD 11.1-STABLE #0 r332507M: Sat Apr 14 16:14:06 PDT 2018 gonzo@station.bluezbox.com:/src/FreeBSD/obj/arm.armv7/src/FreeBSD/11/sys/RPI2 arm

Diff Detail

Repository

rS FreeBSD src repository - subversion

Lint

Lint Passed

Unit

No Test Coverage

Build Status

Buildable 16140
Build 16093: arc lint + arc unit

Event Timeline

gonzo created this revision.Apr 15 2018, 2:39 AM

Herald added a reviewer: manu. · View Herald TranscriptApr 15 2018, 2:39 AM

Herald added subscribers: emaste, andrew, bdrewery. · View Herald Transcript

Harbormaster completed remote builds in B16140: Diff 41476.Apr 15 2018, 2:39 AM

This looks sane to me. But I've not done end-to-end tests to prove it doesn't cause more harm.
Make sure releases still build.
And ports aren't screwed.

There are a few things missing.

https://svnweb.freebsd.org/base/head/UPDATING?r1=324259&r2=324342
https://svnweb.freebsd.org/base/head/UPDATING?r1=324342&r2=324365

The following were missing, but now should _not_ be included:

https://svnweb.freebsd.org/base/head/sys/arm/conf/GENERIC?r1=323967&r2=324341
https://svnweb.freebsd.org/base/head/sys/arm/conf/SOCFPGA?r1=320777&r2=324341

(Please ignore the earlier note about Makefile.mirrors; the change is OBE.)

After email discussion, I think we now agree that it is incorrect to switch the kernels for 11 to armv7 by default. So, most of the changes in head/sys/arm/ are incorrect. (The addition of head/sys/arm/std.armv7 must stay, of course.)

I am working on a patch that does all the above.

In D15071#318760, @linimon wrote:

After email discussion, I think we now agree that it is incorrect to switch the kernels for 11 to armv7 by default. So, most of the changes in head/sys/arm/ are incorrect. (The addition of head/sys/arm/std.armv7 must stay, of course.)

I am working on a patch that does all the above.

It's not clear to me that it's the case that switching the kernels is incorrect.

People are confusing too many issues on this, and the email discussions aren't complete.

It's not clear to me that it's the case that switching the kernels is incorrect.

Unless we explicitly tell all users to use the pkg.conf remap, or the lonesome.com armv7 packages, stable/11 users will regress.

In D15071#318762, @linimon wrote:

It's not clear to me that it's the case that switching the kernels is incorrect.

Unless we explicitly tell all users to use the pkg.conf remap, or the lonesome.com armv7 packages, stable/11 users will regress.

Since arm is tier 2, I'm not sure what the problem here is. We make no guarantees for tier 2 platforms. Users will still be ABLE to boot a kernel where they won't see a regression, which is another path forward as well. It's a release notes item.

Since arm is tier 2, I'm not sure what the problem here is. We make no guarantees for tier 2 platforms.

Going from "there have been supported FreeBSD packages for a year or so" to "there are no longer any packages" is POLA out the ass.

In D15071#318764, @linimon wrote:

Since arm is tier 2, I'm not sure what the problem here is. We make no guarantees for tier 2 platforms.

Going from "there have been supported FreeBSD packages for a year or so" to "there are no longer any packages" is POLA out the ass.

And there's at least three remedies that can be documented in the release notes. Oh, wait, 4 ways: we can put armv6 alias in the pkg.conf file by default and there'd be no issue.