Re: port-xen/57199: Pure PVH i386 guests hang on disk activity

[Brad Spencer <brad%anduin.eldar.org@localhost>]

Manuel Bouyer <bouyer%antioche.eu.org@localhost> writes:

> On Mon, Jul 15, 2024 at 03:00:03PM -0400, Brad Spencer wrote:
>> Taylor R Campbell <riastradh%NetBSD.org@localhost> writes:
>> 
>> > The following reply was made to PR kern/57199; it has been noted by GNATS.
>> >
>> > From: Taylor R Campbell <riastradh%NetBSD.org@localhost>
>> > To: Brad Spencer <brad%anduin.eldar.org@localhost>
>> > Cc: bouyer%antioche.eu.org@localhost, gnats-bugs%NetBSD.org@localhost,
>> > 	netbsd-bugs%NetBSD.org@localhost, gdt%lexort.com@localhost
>> > Subject: Re: port-xen/57199: Pure PVH i386 guests hang on disk activity
>> > Date: Mon, 15 Jul 2024 18:52:14 +0000
>> >
>> >  > Date: Mon, 15 Jul 2024 14:31:41 -0400
>> >  > From: Brad Spencer <brad%anduin.eldar.org@localhost>
>> >  > 
>> >  > b) I will test Taylor's patch on the DOMU.  Doing that on the DOM0 is
>> >  > probably not something I can manage right now.
>> >  
>> >  I should clarify: I suspect this may be needed on the dom0, but only
>> >  if the dom0 _also_ runs uniprocessor, i.e., single-vCPU.
>> >  
>> >  If you can run crash(8) or enter ddb on the dom0, you can check like
>> >  so:
>> >  
>> >  # crash
>> >  Crash version 10.0_STABLE, image version 10.0.
>> >  WARNING: versions differ, you may not be able to examine this image.
>> >  Kernel compiled without options LOCKDEBUG.
>> >  Output from a running system is unreliable.
>> >  crash> x/i membar_sync,3
>> >  _membar_enter:  lock addq       $0,fffffffffffffff8 (%rsp)
>> >  _membar_enter+0x7:      ret
>> >  _membar_enter+0x8:      nopl
>> >  
>> >  If it says `lock addq' or `lock addl', you're good.  If it's just
>> >  `addq' or `addl' with no `lock', the patch is needed.
>> >  
>> 
>> The DOM0 is NetBSD_9.x, but it appears to be ok according to what you
>> wrote above:
>> 
>> DOM0# crash
>> Crash version 9.3_STABLE, image version 9.3_STABLE.
>> Output from a running system is unreliable.
>> crash> x/i membar_sync,3
>> _membar_sync:   lock addq       $0,fffffffffffffff8 (%rsp)
>> _membar_sync+0x7:       ret
>> _membar_sync+0x8:       nopl
>> crash> 
>
> Yes, I don't think hotpatch is run when running PV

That may explain something that was strange to me in that the 32-bit
test system when it was a PV+PVSHIM didn't hang.



-- 
Brad Spencer - brad%anduin.eldar.org@localhost - KC8VKS - http://anduin.eldar.org