Re: firefox52 quits on netbsd-9 w/i915drmkmsfb

2020-01-03 Thread Michael van Elst
jdba...@consolidated.net ("John D. Baker") writes:

>  $ ssh -X user@localhost
>  $ firefox52 > /dev/null 2>&1 &
>  $

>and the problem sites don't tickle the i915drmkms/intel driver's sensitive
>spot that makes firefox52 suddenly quit.

>(X11-forwarding otherwise represents too much of a performance hit for
>using video sites, so need an instance running directly to avoid this.


The X11-forwarding basically prevents direct rendering. Then everything
is done by the X server which usually supports only some 2D hardware
acceleration. There is also some minor overhead from the network code.

You can test this by setting the LIBGL_ALWAYS_INDIRECT=1 environment
variable instead. That also prevents direct rendering, but also
avoids the network overhead of ssh.

Another option is to set LIBGL_ALWAYS_SOFTWARE=1, that still allows
direct rendering but doesn't use hardware acceleration.

Playing video without hardware acceleration is possible even on some
older PC hardware, maybe not at full screen resolution. 3D and compositing
on the other hand requires hardware support to be useful.

-- 
-- 
Michael van Elst
Internet: mlel...@serpens.de
"A potential Snark may lurk in every tree."


Re: firefox52 quits on netbsd-9 w/i915drmkmsfb

2019-11-05 Thread Jaromír Doleček
Le mar. 5 nov. 2019 à 20:14, John D. Baker  a
écrit :

> [...]
> i965: Failed to submit batchbuffer: Input/output error
> assertion "pthread__tsd_destructors[key] != NULL" failed: file
> "/x/netbsd-9/src/lib/libpthread/pthread_tsd.c", line 177, function
> "pthread__add_specific"
>
>
It seems like something called pthread_key_delete() while/before the
pthread_setspecific() call.

I believe the code there should be changed to simply return without doing
anything if specified key doesn't exist, and the assert() removed.

Jaromir


Re: firefox52 quits on netbsd-9 w/i915drmkmsfb

2019-11-05 Thread John D. Baker
On Tue, 5 Nov 2019, John D. Baker wrote:

> I have been noticing a lot of messages on the console (xconsole) about:

Also, this message:

  [ 2717.7371974] kern error: 
[drm:(/x/netbsd-9/src/sys/external/bsd/drm2/dist/drm/i915/intel_sprite.c:132)intel_pipe_update_start]
 *ERROR* Potential atomic update failure on pipe A: -35

-- 
|/"\ John D. Baker, KN5UKS   NetBSD Darwin/MacOS X
|\ / jdbaker[snail]consolidated[flyspeck]net  OpenBSDFreeBSD
| X  No HTML/proprietary data in email.   BSD just sits there and works!
|/ \ GPGkeyID:  D703 4A7E 479F 63F8 D3F4  BD99 9572 8F23 E4AD 1645


Re: firefox52 quits on netbsd-9 w/i915drmkmsfb

2019-11-05 Thread John D. Baker
On Tue, 5 Nov 2019, John D. Baker wrote:

> Restarting firefox52, it recovers the open tabs and tries to repaint the
> one that was last open, but then it just quits--no core, just the same
> message at the end of the ".xsession-errors" file.

It seems to be a problem when repainting tabs that were on particularly
complex web sites.  I my case, trying to repaint the one from the US
National Weather Service (www.weather.gov).  Curiously, reloading a
YouTube user page works fine.

I seem to recall seeing this before and the only way to "fix" it is to
at least terminate the Xsession (resetting the server) or reboot the
workstation.

I tried refreshing the NWS page again and got a little more information
when firefox52 quit:

[...]
i965: Failed to submit batchbuffer: Input/output error
assertion "pthread__tsd_destructors[key] != NULL" failed: file 
"/x/netbsd-9/src/lib/libpthread/pthread_tsd.c", line 177, function 
"pthread__add_specific"

-- 
|/"\ John D. Baker, KN5UKS   NetBSD Darwin/MacOS X
|\ / jdbaker[snail]consolidated[flyspeck]net  OpenBSDFreeBSD
| X  No HTML/proprietary data in email.   BSD just sits there and works!
|/ \ GPGkeyID:  D703 4A7E 479F 63F8 D3F4  BD99 9572 8F23 E4AD 1645