Bug 43802

Created attachment 54401 [details]
dmesg

It would be useful if you could grab that stack again after rebuilding the ddx with debugging symbols.

(In reply to comment #2)
> It would be useful if you could grab that stack again after rebuilding the ddx
> with debugging symbols.

you mean with the --enable-debug configure switch?

(In reply to comment #3)
> (In reply to comment #2)
> > It would be useful if you could grab that stack again after rebuilding the ddx
> > with debugging symbols.
> 
> you mean with the --enable-debug configure switch?

Not necessarily, but if it dies extremely quickly then it would also help. No, it appears that the installed intel_drv.so is stripped of its debug symbols and so gdb can't resolve the line number or local variables, which are extremely useful in working out what the problem is.

(In reply to comment #4)
> (In reply to comment #3)
> > (In reply to comment #2)
> > > It would be useful if you could grab that stack again after rebuilding the ddx
> > > with debugging symbols.
> > 
> > you mean with the --enable-debug configure switch?
> 
> Not necessarily, but if it dies extremely quickly then it would also help. No,
> it appears that the installed intel_drv.so is stripped of its debug symbols and
> so gdb can't resolve the line number or local variables, which are extremely
> useful in working out what the problem is.

Hi Chris,

you're right. using --enable-debug didn't expose any debug symbols. however you should be able to reproduce the issue very quick when you fire up eclipse for example. X dies as soon as the workbench comes up. ot try to start firefox or claws-mail. any app kills my X instantly after fired up.

Regards
Nic

Now you would think that if nearly any app killed X, I'd have a problem typing this into firefox...

Can you try installing the intel_drv.so by hand to avoid any stripping? Just
./configure --enable-sna && make && sudo make src/.libs/intel_drv.so /usr/lib/xorg/modules/drivers/intel_drv.so should do the trick. (Give or take /usr/lib64.)

(In reply to comment #6)
> Now you would think that if nearly any app killed X, I'd have a problem typing
> this into firefox...
> 
> Can you try installing the intel_drv.so by hand to avoid any stripping? Just
> ./configure --enable-sna && make && sudo make src/.libs/intel_drv.so
> /usr/lib/xorg/modules/drivers/intel_drv.so should do the trick. (Give or take
> /usr/lib64.)


Hi Chris,

ok I've compiled xf86-video-intel latest git from scratch using --enable-sna --enable-dri --enable-xvmc and replaced intel_drv.so.

Here is what GDB spits out shortly after firing up firefox and doing a little webbrowsing. Hope this helps.

 Segmentation fault.
0x00007f18a193b869 in sna_poly_fill_rect_stippled_1_blt (clipped=0, extents=0x7fff0d7a5490, r=0x18eb5a8, 
    n=1, gc=0x18edb30, damage=0x0, drawable=0x18d9390, bo=<optimized out>) at sna_accel.c:6567
6567		br13 = priv->gpu_bo->pitch;
(gdb) backtrace full
#0  0x00007f18a193b869 in sna_poly_fill_rect_stippled_1_blt (clipped=0, extents=0x7fff0d7a5490, 
    r=0x18eb5a8, n=1, gc=0x18edb30, damage=0x0, drawable=0x18d9390, bo=<optimized out>)
    at sna_accel.c:6567
        stipple = 0x181e860
        origin = <optimized out>
        dx = <optimized out>
        br00 = 3145728
        br13 = <optimized out>
        sna = 0x1335720
        priv = 0x18b77f0
        dy = <optimized out>
#1  sna_poly_fill_rect_stippled_blt (drawable=0x18d9390, bo=0x18d8fd0, damage=0x0, gc=0x18edb30, n=1, 
    rect=0x18eb5a8, extents=0x7fff0d7a5490, clipped=0) at sna_accel.c:7026
        stipple = <optimized out>
#2  0x00007f18a1943795 in sna_poly_fill_rect (draw=0x18d9390, gc=0x18edb30, n=1, rect=0x18eb5a8)
    at sna_accel.c:7154
        priv = 0x18b77f0
        damage = <optimized out>
        pixmap = 0x0
        region = {extents = {x1 = 5, y1 = 4, x2 = 11, y2 = 13}, data = 0x18edb30}
        flags = 0
#3  0x00000000004f99a9 in damagePolyFillRect ()
No symbol table info available.
#4  0x000000000043201e in ProcPolyFillRectangle ()
No symbol table info available.
#5  0x00000000004357e1 in Dispatch ()
No symbol table info available.
#6  0x000000000042485a in main ()
No symbol table info available.

Gotcha! Fallen afoul of the recent availability of the CPU bo, instead of using priv->gpu_bo, we need to use the bo passed into the function. Thanks.

(In reply to comment #8)
> Gotcha! Fallen afoul of the recent availability of the CPU bo, instead of using
> priv->gpu_bo, we need to use the bo passed into the function. Thanks.

Great! :) Will test the fix as soon as it hits git and close the ticket if successful.

Thx
Nic

commit 43a22743124a83310379122d509c35840b583b2e
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Wed Dec 14 13:26:05 2011 +0000

    sna: Use the provided bo for blitting rather than assume priv->gpu_bo
    
    Reported-by: nkalkhof@web.de
    Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=43802
    Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

Patch is good! Closing Bug. Thanks!