Re: [Xen-devel] PV dom0 crash: kernel NULL pointer dereference in evtchn_from_irq

[Jürgen Groß <jgross@xxxxxxxx>]

On 19.03.20 03:26, Marek Marczykowski-Górecki wrote:
> Hi,
>
>  From time to time, during intensive tests I get the dom0 crash like
> below. This is PV dom0, running on Xen nested inside KVM.
> I don't really know when it started happening, I've got it on at least
> those versions:
>   - Xen 4.8.5 + Linux dom0 4.19.94
>   - Xen 4.13.0 + Linux dom0 5.4.25
>   - at least once also on physical hardware (Xen 4.13.0 + Linux dom0
>     5.4.x)
>
> Contrary to the other issue, here suspend is not involved, it is just
> intensive system usage - multiple VM startups, involving some I/O,
> network traffic etc. This happens rather rarely (I'd say about 1-3% of
> tests).
> To be honest, I'm not really sure if the bug is in Xen-related code at
> all, or if Xen functions are in the call trace only because it is PV
> dom0.
>
> Full crash message:
>
> [14474.613706] BUG: kernel NULL pointer dereference, address: 000000000000001c
> [14474.615832] #PF: supervisor read access in kernel mode
> [14474.617321] #PF: error_code(0x0000) - not-present page
> [14474.618702] PGD 0 P4D 0
> [14474.619452] Oops: 0000 [#1] SMP NOPTI
> [14474.620452] CPU: 0 PID: 431254 Comm: rm Not tainted 5.4.25-1.qubes.x86_64 #1
> [14474.622900] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 
> rel-1.12.0-59-gc9ba527-rebuilt.opensuse.org 04/01/2014
> [14474.626322] RIP: e030:evtchn_from_irq+0x1f/0x40
I have seen this while testing some event channel related patches and
thought I was introducing this case. Seems as if it can happen even
without my patches.

I'll send the fixup I've added to my series soon.


Juergen

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/xen-devel