Re: [10.99.12] Panic (softints stuck)

[Taylor R Campbell <riastradh%NetBSD.org@localhost>]

> Date: Wed, 22 Jan 2025 10:31:07 +0100
> From: BERTRAND Joël <joel.bertrand%systella.fr@localhost>
> 
> 	As there is a regression in recent -10.x kernel with iscsi initiator,

Is there a PR for this?  If not, can you please file one with details
of the regression?

>                                                                             I
> have built a 10.99.12 kernel. iscsi initiator runs now as expected. I
> haven't seen hard lock for a while. Nevertheless, system randomly panics
> with :
> 
> [ 347098.708000] ccb_timeout: num=1 total=0 disp=0 invalid
> ccb=0xffff860021cec3e0

Can you please file a PR with this symptom (and the rest of the panic
and stack trace)?  Once we can get a crash dump out, we can probably
find what's going on with this.  It looks like something is hogging
the mbuf pool mutex but it's not clear why and this `XXX Should never
happen' printf is probably related.

> 	I have checked, always the same panic message. I would have more
> information, but as you see, "dump device 'is' bad". For a while, my
> system is not able to write kernel dump. With -9, it worked. Never with
> -10 (but without message). -10.99 complains against bad device. My swap
> is on raid0b. Maybe kernel cannot write a dump on a raid slice.

I have filed PR kern/59024: dump fails on raid0b
<https://gnats.NetBSD.org/59024> to track the dump issue.  Can you
please update your kernel to HEAD (with sys/arch/amd64/amd64/machdep.c
rev. 1.370) and try again?

Can you confirm that you had the same disk configuration in 9, 10, and
HEAD -- that is, dump device configured to be raid0b, and the same
disklabel type of swap for raid0b?