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:   Wed, 5 Apr 2017 08:36:06 +0200
From:   Christoph Hellwig <hch@....de>
To:     "Michael S. Tsirkin" <mst@...hat.com>
Cc:     Christoph Hellwig <hch@....de>, 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 Mon, Apr 03, 2017 at 07:14:22PM +0300, Michael S. Tsirkin wrote:
> On Mon, Apr 03, 2017 at 04:18:23PM +0200, Christoph Hellwig wrote:
> > Mike,
> > 
> > can you try the patch below?
> > 
> > ---
> > >From fe41a30b54878cc631623b7511267125e0da4b15 Mon Sep 17 00:00:00 2001
> > From: Christoph Hellwig <hch@....de>
> > Date: Mon, 3 Apr 2017 14:51:35 +0200
> > Subject: virtio_pci: don't use shared irq for virtqueues
> > 
> > Reimplement the shared irq feature manually, as we might have a larger
> > number of virtqueues than the core shared interrupt code can handle
> > in threaded interrupt mode.
> > 
> > Signed-off-by: Christoph Hellwig <hch@....de>
> > ---
> >  drivers/virtio/virtio_pci_common.c | 142 +++++++++++++++++++++----------------
> >  drivers/virtio/virtio_pci_common.h |   1 +
> >  2 files changed, 83 insertions(+), 60 deletions(-)
> 
> Well the original patch this is trying to fix is
> 07ec51480b5eb1233f8c1b0f5d7a7c8d1247c507 which dropped just 40 lines
> with documentation. It did this by re-using error handling to switch
> from per-vq to non-per-vq mode. Now this has separate flows for errors
> and per-vq non-per-vq switch and (I think, as a result) is adding 140
> lines which doesn't make me very happy.

The above adds 23 lines.  We could entangle both loops again, but I'm
not sure it's going to buy us much.