Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]


Groups > linux.debian.kernel > #67414

Bug#961676: 5.6 kernel crashes host when using VFIO on KVM guests

From Simon John <simon@the-jedi.co.uk>
Newsgroups linux.debian.bugs.dist, linux.debian.kernel
Subject Bug#961676: 5.6 kernel crashes host when using VFIO on KVM guests
Date 2020-06-28 12:10 +0200
Message-ID <AmCq6-6Aj-11@gated-at.bofh.it> (permalink)
References (9 earlier) <Amt3r-RF-3@gated-at.bofh.it> <AmBaF-5FV-3@gated-at.bofh.it> <AmBaF-5FV-3@gated-at.bofh.it> <Aba4a-2Il-7@gated-at.bofh.it> <AmBaF-5FV-3@gated-at.bofh.it>
Organization linux.* mail to news gateway

Cross-posted to 2 groups.

Show all headers | View raw


On Sun, 28 Jun 2020 10:41:41 +0200 Salvatore Bonaccorso 
<carnil@debian.org> wrote:
> Hi Simon,

Hi Salvatore, thanks for looking into this.

> On Sun, Jun 28, 2020 at 01:01:01AM +0100, Simon John wrote:
> > This looks a likely culprit:
> > 
> > https://bugzilla.kernel.org/show_bug.cgi?id=207489
> 
> The issue you are seeing now seems different, though afaics.
> 
> The fixing commit from the above reference, 8be8f932e3db ("kvm:
> ioapic: Restrict lazy EOI update to edge-triggered interrupts") was
> applied in v5.7-rc3, which was as well backported to v5.6.13.

Yes, I was going to patch the kernel but noticed its already got that 
patch. I also tried passing through only one pci device and not the usb 
device but it didn't help, so probably isn't that issue.

> Now you kernel is tained, can you check if you see the issue as well
> when not loading the modules which taint the kernel?

I removed virtualbox which was tainting the kernel, confirmed cat 
/proc/sys/kernel/tainted returned 0 after a reboot.

I still managed to trigger the issue though, which then tainted the 
kernel itself, output from kernel-chktaint:

Kernel is "tainted" for the following reasons:
  * kernel died recently, i.e. there was an OOPS or BUG (#7)
  * kernel issued warning (#9)
  * soft lockup occurred (#14)
For a more detailed explanation of the various taint flags see
  Documentation/admin-guide/tainted-kernels.rst in the the Linux kernel 
sources
  or https://kernel.org/doc/html/latest/admin-guide/tainted-kernels.html
Raw taint value as int/string: 17024/'G      D W    L   '

The kvm-pit process was using 100% of a cpu core, i've never even 
noticed that process before, looks like the culprit now?:

watchdog: BUG: soft lockup - CPU#7 stuck for 22s! [kvm-pit/2888:2913]

CPU: 7 PID: 2913 Comm: kvm-pit/2888 Tainted: G      D W 
5.7.0-1-amd64 #1 Debian 5.7.6-1

I've added a strace and dmesg to the gist here:

https://gist.github.com/sej7278/766043a69c76308f84cfa14b3f3a924f

Any other diagnostics I can run?

> Next, if it still does show up, does it show up as well in current
> mainline?

Not sure what you mean there, do you mean upstream kernel.org kernel - 
i'm not sure how i'd run that.

Regards.

-- 
Simon John

Back to linux.debian.kernel | Previous | NextPrevious in thread | Next in thread | Find similar | Unroll thread


Thread

5.6 kernel crashes host when using VFIO on KVM guests Simon John <debian@the-jedi.co.uk> - 2020-05-25 22:40 +0200
  Bug#961676: 5.6 kernel crashes host when using VFIO on KVM guests Salvatore Bonaccorso <carnil@debian.org> - 2020-05-27 21:40 +0200
    Processed: Re: 5.6 kernel crashes host when using VFIO on KVM guests "Debian Bug Tracking System" <owner@bugs.debian.org> - 2020-05-27 21:40 +0200
    Bug#961676: 5.6 kernel crashes host when using VFIO on KVM guests Simon John <debian@the-jedi.co.uk> - 2020-05-27 22:50 +0200
      Bug#961676: 5.6 kernel crashes host when using VFIO on KVM guests Simon John <debian@the-jedi.co.uk> - 2020-06-12 00:30 +0200
        Bug#961676: 5.6 kernel crashes host when using VFIO on KVM guests Simon John <debian@the-jedi.co.uk> - 2020-06-14 00:50 +0200
          Bug#961676: 5.6 kernel crashes host when using VFIO on KVM guests Simon John <debian@the-jedi.co.uk> - 2020-06-28 02:00 +0200
            Bug#961676: 5.6 kernel crashes host when using VFIO on KVM guests Simon John <debian@the-jedi.co.uk> - 2020-06-28 02:10 +0200
              Bug#961676: 5.6 kernel crashes host when using VFIO on KVM guests Salvatore Bonaccorso <carnil@debian.org> - 2020-06-28 10:50 +0200
                Bug#961676: 5.6 kernel crashes host when using VFIO on KVM guests Simon John <simon@the-jedi.co.uk> - 2020-06-28 12:10 +0200
                Bug#961676: 5.6 kernel crashes host when using VFIO on KVM guests Simon John <debian@the-jedi.co.uk> - 2020-07-03 23:50 +0200
                Bug#961676: 5.6 kernel crashes host when using VFIO on KVM guests Simon John <debian@the-jedi.co.uk> - 2020-07-19 00:20 +0200
                Bug#961676: 5.6 kernel crashes host when using VFIO on KVM guests Simon John <debian@the-jedi.co.uk> - 2020-07-19 01:40 +0200
                Bug#961676: 5.6 kernel crashes host when using VFIO on KVM guests Simon John <debian@the-jedi.co.uk> - 2020-07-19 01:50 +0200
                Bug#961676: 5.6 kernel crashes host when using VFIO on KVM guests Simon John <debian@the-jedi.co.uk> - 2020-07-19 18:00 +0200
    Bug#961676: marked as done (5.6 kernel crashes host when using  VFIO on KVM guests) "Debian Bug Tracking System" <owner@bugs.debian.org> - 2020-07-19 18:20 +0200

csiph-web