x11perf -all leads to Xorg crash on 8-STABLE

[Carlos A. M. dos Santos]

On Sun, Dec 27, 2009 at 2:33 AM, Warren Block <wblock at wonkity.com> wrote:
> On Sat, 26 Dec 2009, Carlos A. M. dos Santos wrote:
>
>> Environment: 8-STABLE, as of Dec 23 and fresh built with
>> xorg-server-1.6.1,1 and xf86-video-intel-2.7.1. Running x11perf with
>> the "-all" argument I got this:
>>
>> avatar:/local[152]# x11perf -all > x11perf.default
>> Setting master
>
> Don't see that line on intel or radeon.  Maybe a -verbose setting?

There is not verbose setting.

>> Fatal server error:
>> Caught signal 11.  Server aborting
>
> After a couple of hours on 8-stable Dec 25 with radeon, I do see that.
>
>> Dropping master
>> XIO:  fatal IO error 35 (Resource temporarily unavailable) on X server
>> ":0.0"
>>     after 522479794 requests (522479789 known processed) with 0
>> events remaining.
>
> But not those lines.  The last thing in the output was "ShmPutImage 500x500
> square".

I figured out that "x11perf -shmputxy10" triggers the crash. Mys first
guess was that something related to shared memory was causing the
problem. However, running the X server with

     truss /usr/local/bin/Xorg >& Xorg.truss.log

shows that all shm* calls succeed. Anyway, I attempted to increase the
available shared memory, using the values recommended by the Xine
port:

     sysctl kern.ipc.shmmax=67108864
     sysclt kern.ipc.shmall=32768

But it keeps crashing.