[Mesa-dev] piglit: Segmentation fault running fbo-depthstencil -auto readpixels default

[Brian Paul]

On 11/16/2011 01:38 AM, Theiss, Ingo wrote:
> Dear devs,
>
> I am getting segmentation faults when running piglit r600.tests with latest mesa build from git:
>
> Nov 16 09:21:33 spoc kernel: [74538.198983] radeon 0000:05:00.0: object_init failed for (1431699456, 0x00000006)
> Nov 16 09:21:33 spoc kernel: [74538.198987] [drm:radeon_gem_object_create] *ERROR* Failed to allocate GEM object (1431699456, 4, 4096, -12)
> Nov 16 09:21:37 spoc kernel: [74542.266472] fbo-depthstenci[6867]: segfault at 0 ip 00007fb271b5414a sp 00007fffbfdf8fd0 error 4 in r600_dri.so[7fb2716c8000+dba000]
>
> ---
> Core was generated by `/home/itheiss/build/mesa.git/piglit/bin/fbo-depthstencil -auto readpixels defau'.
> Program terminated with signal 11, Segmentation fault.
> #0  u_transfer_unmap_vtbl (pipe=0x2441df0, transfer=0x0) at util/u_resource.c:63
> 63	   ur->vtbl->transfer_unmap(pipe, transfer);
> (gdb) bt
> #0  u_transfer_unmap_vtbl (pipe=0x2441df0, transfer=0x0) at util/u_resource.c:63
> #1  0x00007f4cb36cc0eb in pipe_transfer_unmap (transfer=<optimized out>, context=0x2441df0)
>      at ../../src/gallium/auxiliary/util/u_inlines.h:398
> #2  st_UnmapRenderbuffer (ctx=<optimized out>, rb=0x25b7fc0) at state_tracker/st_cb_fbo.c:704
> #3  0x00007f4cb365f925 in slow_read_depth_stencil_pixels_separate (ctx=0x254b720, x=<optimized out>, y=<optimized out>, width=123,
>      height=123, type=36269, packing=0x7fff653d0750, dst=0x7fff653ee138 "", dstStride=984) at main/readpix.c:427
> #4  0x00007f4cb36601d6 in read_depth_stencil_pixels (packing=0x7fff653d0750, pixels=<optimized out>, type=36269, height=<optimized out>,
>      width=123, y=0, x=0, ctx=0x254b720) at main/readpix.c:469
> #5  _mesa_readpixels (ctx=0x254b720, x=0, y=0, width=123, height=123, format=<optimized out>, type=36269, packing=0x255afb8,
>      pixels=<optimized out>) at main/readpix.c:508
> #6  0x00007f4cb3660bbd in _mesa_ReadnPixelsARB (pixels=0x7fff653d0870, type=36269, format=34041, height=123, width=123, y=0, x=0,
>      bufSize=<optimized out>) at main/readpix.c:753
> #7  _mesa_ReadPixels (x=0, y=0, width=123, height=123, format=34041, type=36269, pixels=0x7fff653d0870) at main/readpix.c:761
> #8  0x000000000042eb4a in read_32f_8 ()
> #9  0x000000000042ec0b in compare ()
> #10 0x000000000042ef73 in test_readpixels ()
> #11 0x000000000042f7aa in piglit_display ()
> #12 0x0000000000430241 in display ()
> #13 0x00007f4cb6f11808 in ?? () from /usr/lib/libglut.so.3
> #14 0x00007f4cb6f15339 in fgEnumWindows () from /usr/lib/libglut.so.3
> #15 0x00007f4cb6f11ca1 in glutMainLoopEvent () from /usr/lib/libglut.so.3
> #16 0x00007f4cb6f12540 in glutMainLoop () from /usr/lib/libglut.so.3
> #17 0x0000000000430967 in main ()
> ---
>
> My glxinfo:
>
> OpenGL vendor string: X.Org
> OpenGL renderer string: Gallium 0.4 on AMD BARTS
> OpenGL version string: 2.1 Mesa 7.12-devel (git-4f677ca)
> OpenGL shading language version string: 1.20
>
>
> A general question: I am interested in running regular piglit tests on my hardware and would like to help by reporting problems/bugs. What´s the correct way of doing so? Should I post any problems to mesa-dev or should I open a bug.
>
> Thanks and keep up the good work!

I think that this will be fixed by the patch I just posted (mesa: 
don't map depth+stencil buffer twice in glReadPixels())  Can you try it?

-Brian