Re: [Qemu-devel] [PATCH] virtio-serial: Fix segfault on guest boot

[Amit Shah <amit.shah@redhat.com>]

On (Fri) 17 Jun 2011 [09:47:31], Markus Armbruster wrote:
> Amit Shah <amit.shah@redhat.com> writes:
> 
> > On (Thu) 16 Jun 2011 [13:38:49], Luiz Capitulino wrote:
> >> If I start qemu with:
> >> 
> >>  # qemu -hda disks/test.img -enable-kvm -m 1G -snapshot \
> >>         -device virtio-serial \
> >>         -chardev socket,host=localhost,port=1234,server,nowait,id=foo \
> >>         -device virtserialport,chardev=foo,name=org.qemu.guest_agent
> >> 
> >> I get a segfault when booting a Fedora 14 guest. The backtrace says:
> >>
> >>   Program terminated with signal 11, Segmentation fault.
> >>   #0  0x0000000000420850 in handle_control_message (vser=0x3732bd0, buf=0x2c173e0, len=8) at /home/lcapitulino/src/qmp-unstable/hw/virtio-serial-bus.c:335
> >>   335     info = DO_UPCAST(VirtIOSerialPortInfo, qdev, port->dev.info);
> >
> > Strange, I've not seen it so far in my testing (neither in the daily
> > test runs of the virtio-serial testsuite).
> >
> >> I've also bisected this and git points out to commit:
> >> 
> >>   commit a15bb0d6a981de749452a5180fc8084d625671da
> >>   Author: Markus Armbruster <armbru@redhat.com>
> >>   Date:   Wed May 25 14:21:13 2011 +0200
> >> 
> >>       virtio-serial: Drop redundant VirtIOSerialPort member info
> >> 
> >> I think what's happening is that the device is not initialized on a
> >> VIRTIO_CONSOLE_DEVICE_READY event.
> 
> Really?
> 
> I believe the device is initialized just fine, but the message refers to
> a nonexistant port.  The check after find_port_by_id() suggests that's
> expected for VIRTIO_CONSOLE_DEVICE_READY.

What's happening is VIRTIO_CONSOLE_DEVICE_READY is a message for the
whole device, not for an individual port.  So port is NULL.  All other
messages (currently) are port-specific, so they have 'port' set to
something.  The commit message needs to be tweaked for this.

> 
> >>                                    Moving the DO_UPCAST() call to
> >> the other events fixes the problem to me.
> >> 
> >> Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
> >> ---
> >>  hw/virtio-serial-bus.c |    4 ++--
> >>  1 files changed, 2 insertions(+), 2 deletions(-)
> >> 
> >> diff --git a/hw/virtio-serial-bus.c b/hw/virtio-serial-bus.c
> >> index 9a12104..579f676 100644
> >> --- a/hw/virtio-serial-bus.c
> >> +++ b/hw/virtio-serial-bus.c
> >> @@ -332,8 +332,6 @@ static void handle_control_message(VirtIOSerial *vser, void *buf, size_t len)
> >>      if (!port && cpkt.event != VIRTIO_CONSOLE_DEVICE_READY)
> >>          return;
> >>  
> >> -    info = DO_UPCAST(VirtIOSerialPortInfo, qdev, port->dev.info);
> >> -
> >
> > Ah - this missed the !port check.  It should be possible to do this in
> > a 'if (port)' block instead of replicating in the individual case
> > statements.
> 
> You might have to do something like info = port ? DO_UPCAST(...) : NULL
> to avoid warnings about info used uninitialized.

That won't happen -- in the case where port is NULL, we shouldn't try
to use info.  If we do get such a warning, it means there's a bug :-)


		Amit