Re: qemu emulated machine crashes due to disk timeouts

[Jaromír Doleček <jaromir.dolecek%gmail.com@localhost>]

Le ven. 15 mai 2020 à 15:53, Jonathan A. Kollasch
<jakllsch%kollasch.net@localhost> a écrit :
>
> On Sat, May 02, 2020 at 12:02:45PM +1000, Paul Ripke wrote:
> > Since I have my qemu disk images on slow spinning rust host disks, when the
> > host disk is busy (esp. daily+security runs), I find my qemu vm's see disk
> > timeouts, and end up crashing. This isn't great behaviour.
>
> Timeout issue aside, crashing because of it is a bug in the error
> handling code somewhere.

Yes, this is root cause of the panic:

panic: LOCKDEBUG: Mutex error: mutex_vector_enter,509: assertion
failed: !cpu_intr_p()
[ 13493.1166960] cpu0: Begin traceback...
[ 13493.1166960] vpanic() at netbsd:vpanic+0x178
[ 13493.1166960] snprintf() at netbsd:snprintf
[ 13493.1166960] lockdebug_more() at netbsd:lockdebug_more
[ 13493.1166960] mutex_enter() at netbsd:mutex_enter+0x656
[ 13493.1166960] suspendsched() at netbsd:suspendsched+0x19
[ 13493.1166960] cpu_reboot() at netbsd:cpu_reboot+0x46
[ 13493.1166960] sys_reboot() at netbsd:sys_reboot
[ 13493.1166960] vpanic() at netbsd:vpanic+0x181
[ 13493.1166960] snprintf() at netbsd:snprintf
[ 13493.1166960] startlwp() at netbsd:startlwp
[ 13493.1166960] alltraps() at netbsd:alltraps+0xc3
[ 13493.1166960] wdc_ata_bio_start() at netbsd:wdc_ata_bio_start+0xcbd
[ 13493.1166960] ata_xfer_start() at netbsd:ata_xfer_start+0x4f
[ 13493.1166960] wdc_ata_bio_intr() at netbsd:wdc_ata_bio_intr+0x3b9
[ 13493.1166960] wdcintr() at netbsd:wdcintr+0x10a
[ 13493.1166960] intr_biglock_wrapper() at netbsd:intr_biglock_wrapper+0x36
[ 13493.1166960] Xhandle_ioapic_edge3() at netbsd:Xhandle_ioapic_edge3+0x6d

I're-recheck the path, something seems to still fool the code to run
handling in interrupt context when it things it's in process context.

Jaromir