kern/166568: [sched_ule] intr stuck in WAIT state

[Oliver Pinter]

On 4/2/12, Andriy Gapon <avg at freebsd.org> wrote:
> The following reply was made to PR kern/166568; it has been noted by GNATS.
>
> From: Andriy Gapon <avg at FreeBSD.org>
> To: Oliver Pinter <oliver.pntr at gmail.com>, bug-followup at FreeBSD.org
> Cc:
> Subject: Re: kern/166568: [sched_ule] intr stuck in WAIT state
> Date: Mon, 02 Apr 2012 12:10:58 +0300
>
>  on 02/04/2012 11:21 Oliver Pinter said the following:
>  > Hi!
>  >
>  > Today night have I more time to play with this issue (DDB, kgdb..) .
>  > At this time the machine running on non-debug kernel, and the symptom
>  > are only to see, that the [intr] kernel process stucked in WAIT state
>  > and the load are constantly near 1.00 no others are visible.
>  > The machine is a HP 5310m laptop with core2duo CPU - P9400.
>  >
>  > The fast solution are reverting the commit, that I previously reported.
>  >
>  > When you have some suggestion what must check, than feel free to say.
>
>  Basically I don't understand what you mean by "stuck in WAIT state".
>  What's so bad about the WAIT state in your opinion?  Why being in that
> state
>  implies being "stuck"?
>  As far as I understand it's a normal state for the interrupt threads when
> they
>  are not servicing any interrupts.  Do you have a different understanding?
>
>
>  > On 4/2/12, Andriy Gapon <avg at freebsd.org> wrote:
>  >> The following reply was made to PR kern/166568; it has been noted by
> GNATS.
>  >>
>  >> From: Andriy Gapon <avg at FreeBSD.org>
>  >> To: bug-followup at FreeBSD.org, pinter at tresorium.hu
>  >> Cc:
>  >> Subject: Re: kern/166568: [sched_ule] intr stuck in WAIT state
>  >> Date: Mon, 02 Apr 2012 10:17:35 +0300
>  >>
>  >>  Apologies, but could you please be a little bit less terse in
> describing
>  >> what
>  >>  problem you are reporting?
>  >>
>  >>  --
>  >>  Andriy Gapon
>  >> _______________________________________________
>  >> freebsd-bugs at freebsd.org mailing list
>  >> http://lists.freebsd.org/mailman/listinfo/freebsd-bugs
>  >> To unsubscribe, send any mail to "freebsd-bugs-unsubscribe at freebsd.org"
>  >>
>

You are right, this is independent from intr, while it is "always" in
WAIT, but the load average dependend on r233599. See the attached
graph.

The datasets are generated in single user boot ( top -s 1 -S -d 200 -m
cpu ) , with/without r233599. Without the ule patch, the load
decreased normally, but with r233599 it's hold to 0.5.
,

>
>  --
>  Andriy Gapon
> _______________________________________________
> freebsd-bugs at freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-bugs
> To unsubscribe, send any mail to "freebsd-bugs-unsubscribe at freebsd.org"
>