[Spice-devel] crashes / Assertion `uxa_pixmap_is_offscreen(src_pixmap)' failed

[Christophe Fergeau]

On Tue, May 02, 2017 at 02:16:00PM +0200, Daniel Pocock wrote:
> 
> 
> On 02/05/17 13:54, Christophe Fergeau wrote:
> > On Tue, May 02, 2017 at 01:46:20PM +0200, Daniel Pocock wrote:
> >> 
> >> 
> >> On 02/05/17 12:43, Christophe Fergeau wrote:
> >>> On Tue, May 02, 2017 at 11:39:08AM +0200, Daniel Pocock wrote:
> >>>> 
> >>>> 
> >>>> Hi,
> >>>> 
> >>>> I've observed that the Xorg process in my virtual server 
> >>>> sometimes crashes when I launch Firefox.  Xorg.0.log had an
> >>>> error about "Out of video memory: Could not allocate 7274816
> >>>> bytes", so I increased video RAM from 64MB to 256MB.  On the
> >>>> last crash the error about memory was not there, but I found
> >>>> this at the end of Xorg.0.log:
> >>>> 
> >>>> [170197.066] qxl_surface_create: Bad bpp: 1 (1) [170197.076] 
> >>>> qxl_surface_create: Bad bpp: 1 (1) [170228.641] 
> >>>> qxl_surface_create: Bad bpp: 1 (1) [170281.094] 
> >>>> qxl_surface_create: Bad bpp: 1 (1) [170281.641] 
> >>>> qxl_surface_create: Bad bpp: 1 (1)
> >>> 
> >>> These ones keep happening I think, and are harmless
> >>> 
> >>>> 
> >>>> and in journalctl I found:
> >>>> 
> >>>> gdm-Xorg-:0[1538]: qxl_surface_create: Bad bpp: 1 (1) 
> >>>> iceweasel.desktop[4967]: Gtk-Message: Failed to load module 
> >>>> "canberra-gtk-module" gdm-Xorg-:0[1538]: qxl_surface_create:
> >>>> Bad bpp: 1 (1) gdm-Xorg-:0[1538]: qxl_surface_create: Bad
> >>>> bpp: 1 (1) gdm-Xorg-:0[1538]: Xorg:
> >>>> ../../../src/uxa/uxa-glyphs.c:991: uxa_glyphs_via_mask:
> >>>> Assertion `uxa_pixmap_is_offscreen(src_pixmap)' failed.
> >>>> 
> >>>> at the moment of the crash.  I opened a Debian bug[1] with
> >>>> that error.
> >>> 
> >>> Would you have a core file when that happens?
> >>> 
> >> 
> >> Not yet, but I'd be happy to try and get one.  Would you be able
> >> to suggest how I tweak my system to generate core files for the
> >> Xorg process?  Should I just replace the Xorg binary with a
> >> wrapper script or is there a more official way to do it by
> >> modifying /lib/systemd/system/gdm3.service or something else?
> > 
> > On my f26 system, coredumpctl is just magic to grab coredumps, but
> > I don't know how much integration work is needed. An alternative is
> > to run a screen/tmux, and attach gdb to the running Xorg process in
> > there, and switch to a VT once it crashed (hopefully this should be
> > possible). Or I guess you can just enable core generation
> > system-wide, and grab it in the location where it landed (this is
> > defined in /proc/xxx)
> > 
> 
> Thanks for that tip, I'm always glad to learn something new.  I did this
> :
> 
>   apt-get install -t jessie-backports systemd-coredump
> 
> and then I checked it is there:
> 
> $ sysctl -a | grep core_patt
> kernel.core_pattern = |/lib/systemd/systemd-coredump %P %u %g %s %t
> 9223372036854775808 %e
> 
> I tried killing my shell with SEGV and it looks like I captured somethin
> g:
> 
> $ sudo coredumpctl list
> TIME                            PID   UID   GID SIG PRESENT EXE
> Tue 2017-05-02 14:13:14 CEST   6152  1000  1000  11 * /bin/bash
> 
> 
> Am I all set now for the next Xorg crash or is there anything else
> that might be worth tweaking?

I'd make sure that "coredumpctl gdb" is giving you a backtrace,
otherwise you are all set.

Christophe
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 801 bytes
Desc: not available
URL: <https://lists.freedesktop.org/archives/spice-devel/attachments/20170502/95200906/attachment.sig>