Did we consider the generic apporach to implement suspend inside vfs and VM?

All Makefile changed in this change includes src.opts.mk in the beginning of files.

rebase to -Current

updated a new revision to address comments.

Update

Address comments on man page

this change didn't touch efi boot code. need more working here.

This need more testing and some change in my private tree is dependency for this change.

update copyright line

Don't remove hyperv from GENERIC. Add boot loader logic only.

In D5958#127009, @sepherosa_gmail.com wrote:

Can we leave the device hyperv in GENERIC? Since atomic ops will have significant cost if hyperv drivers are loaded as modules. We can remove the device hyperv from GENERIC after atomic ops only have inlined version on amd64 (like other archs)

In D5922#126253, @hselasky wrote:

Hi,

In the USB core M_NOWAIT should be used, and NULL checks should be preserved. The NULL checks are active for embedded products (non-FreeBSD) where the malloc implementation does not support M_WAITOK.

--HPS

please use a different fix to the problem as we discussed

I have to split this into small changeset and manually cleanup more stlye(9)

mismerge one line, add it back

The patch is still refining. I submit this to see if there is strong objection not to do this. (remove null check)

the name of IPI sounds strange to me as you are enabling this in UP.

Even failed to boot in one of VM with 24 vCPU

In D5715#121996, @sepherosa_gmail.com wrote:

Yeah, the more NICs you have the higher chance you got this panic.

My worry is, now that the code sleeps in the globally shared taskqueue_thread. I'd suggest to create two vmbus taskqueues, instead of sharing the global taskqueue_fast/taskqueue_thread. Though that should be done in another patch.

a different version with same idea is already committed.

split patch

Use SINT4 for eventtimer instead of SINT2 that shares with others.

Can u review?

address review feedback

In D5651#120865, @decui_microsoft.com wrote:

Add const and KASSERT too?

I didn't see your comments before I made the change. I added const.

add const

Update guid print function

Add jhb@ to seek for input on IDT thing

fix comment

This is not for -current branch.

Let me split this change into several smaller ones for non-guid related changes.

Rebase to master. Rename {IDT|IRQ}_EVTCHN to _HYPERVISOR

Rebase to master and use taskqueue_enqueue for fast queue

Shall we commit this without changing the IDT name?

there is more wmb and mb() in the tree. please cleanup them together to switch to atomic(9)

update local_apic code to mark IDT_EVTCHN is used when HYPERV is enabled.

I suggest we don't create other node under sysctl. each device already have the node in sysctl tree. dev.hn.0 for example. I suggest to put the stats under it.

Implement is_perf_channel parameter in open_channel call

Update the change for utility drivers

Remove unrelated changes

No. Since read/peek are happened in single thread, no mb is needed. Peek is just a normal read except it copies less data and doesn't advance the read position.