ZFS part looks good to me.

LGTM. Thank you!

• add a barrier function to ensure that the interrupt handling code sees that a handler is suspended after intr_event_suspend_handler() returns

Terry, thank you!

I've just tested this and I see a couple of issues.

rebase

I am considering all previous requests for changes as outdated and I am going to commit this.
Please state if you still have any objections.

In D16459#350746, @eugen_grosbein.net wrote:

I'm going to commit latest revision soon unless an objection is raised.

In D15905#348331, @cem wrote:

It was this: https://reviews.freebsd.org/D15905#338982

Which has been completed. I don't know of a way in phabricator to un-request changes other than approving, which isn't quite the same thing.

Well, the standard says (0 ..64) and that means from zero to 64.
Also, my impression is that, according to the RFC, ifAlias is something that should be settable over SNMP and it should be persistent.

Some comments inline, but in general looks good to me.
Thanks!

remove TESTING / KFAIL code

Per suggestions from Samy Al Bahra:

@cem, you are shown as having requested a change but looking into the comment history I am not sure what that change might be.
Could you please clarify?

rebase

Use CK_SLIST instead of SLIST to avoid a need for hand-rolled list
operations with fences.
This change adds CK_SLIST_{INSERT,REMOVE}_PREVPTR.

LGTM, thanks!
Maybe also add MODULE_PNP_INFO now that the IDs are in a table?

fix a mismerge

In D15985#339084, @jeff wrote:

The majority of the patch is actually unrelated to the specific issue. Only the timeshare delta and tdq_runq_elevate are for the problem. The others were just issues I noticed when reviewing the code while looking into this issue. I can break them off and commit them one at a time if it becomes more clear. I'm not even certain I want to commit the timeshare_delta as it is because I am hoping someone may suggest a more elegant fix.

In D15630#339073, @dch wrote:

also to confirm it works for me - just used it to get a crash dump from a system freeze. many thanks!

rebase on top of D16016

In D15905#338982, @cem wrote:

Please update with TAILQ->SLIST patch removed so it's easier for reviewers to see the meat of the change. (And then I'd ask for 24h.)

slide up to rS335655

In D15985#338992, @jeff wrote:

I had just forgotten about IPI_AST but I like the way I have implemented it here better. It gives the remote scheduler a chance to look at what's going on and make a new decision.

Stop me or I will commit this.
I will [pretend to] ignore any post-commit comments from the reviewers too.

Also, what's a practical difference between TDF_NEEDRESCHED + IPI_AST and IPI_PREEMPT ?

Hmm, now I see a difference between 4BSD and ULE.
If kick_other_cpu does not preempt the remote CPU, then it does this:

pcpu->pc_curthread->td_flags |= TDF_NEEDRESCHED;
ipi_cpu(cpuid, IPI_AST);

On the other hand, tdq_notify either preempts or does nothing at all.

@jeff I do not completely understand why in this scenario cksum runs behind the loop. Does cksum get a priority worse than the loop?
Otherwise, I would expect that even without preemption TDF_NEEDRESCHED would produce a similar effect.