linux-kernel - Re: Random guest crashes since 5c34d002dcc7 ("virtio_pci: use shared interrupts for virtqueues")

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Fri, 31 Mar 2017 19:47:02 +0300
From:   "Michael S. Tsirkin" <mst@...hat.com>
To:     Christoph Hellwig <hch@....de>
Cc:     Mike Galbraith <efault@....de>,
        Thorsten Leemhuis <linux@...mhuis.info>,
        virtio-dev@...ts.oasis-open.org,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
        rjones@...hat.com
Subject: Re: Random guest crashes since 5c34d002dcc7 ("virtio_pci: use shared
 interrupts for virtqueues")

On Fri, Mar 31, 2017 at 10:20:49AM +0200, Christoph Hellwig wrote:
> On Fri, Mar 31, 2017 at 06:22:31AM +0300, Michael S. Tsirkin wrote:
> > I'm not sure why does it fail after 32 on 64 bit, but as
> > virtio devices aren't limited to 32 vqs it looks like we
> > should go back to requesting the irq only once for all vqs.
> 
> Meh.
> 
> > 
> > Christoph, should I just revert for now, or do you
> > want to look into a smaller patch for this?
> 
> I think we'll need to do a different patch than just a simple revert,
> mostly because so much infrastructure depends on the patch.
> 
> I'll take a look over the weekend.
> 
> > Another question is looking into intx support - that
> > should work but it seems to be broken at the moment.
> 
> Does it?  I'm pretty sure I tested it back when I came up with the
> series by artifically disabling MSI-X in the kernel.  I can try this
> again, though.

I'm not 100% sure - what I see is that we do not handle failure to
request irqs correctly, we seem to fall back on intx but
the following freeze then blows up trying to free non-existing
vectors.

Does not seem to trigger with just msix off so maybe that is
simply failure to recover from an error correctly.


-- 
MST