When running Team Fortress 2 in Wine with the DRI driver from the radeon-rewrite branch(with KMS et al set up), the game crashes in glCopyTexSubImage2D and in turn brings the X server down. This is the backtrace I got from the crash: =>0 0x7e1ce123 _generic_read_RGBA_span_BGRA8888_REV_SSE2+0xb3() in r300_dri.so (0x0032dca4) 1 0x7e18d974 _swrast_read_rgba_span+0xe4(ctx=0x7d1e6950, rb=<register EDI not in topmost frame>, n=512, x=0, y=0, dstType=5121, rgba=) [/home/stefan/src/mesa/src/mesa/swrast/s_span.c:1585] in r300_dri.so (0x0033dce4) 2 0x7e1a3ed0 read_color_image+0xc0(ctx=0x7d1e6950, x=0, y=0, type=5121, width=512, height=512) [/home/stefan/src/mesa/src/mesa/swrast/s_texstore.c:85] in r300_dri.so (0x0033dd34) 3 0x7e1a426d _swrast_copy_texsubimage2d+0xad(ctx=<register EDI not in topmost frame>, target=3553, level=0, xoffset=0, yoffset=0, x=0, y=0, width=512, height=512) [/home/stefan/src/mesa/src/mesa/swrast/s_texstore.c:512] in r300_dri.so (0x0033dd94) 4 0x7e120f79 _mesa_CopyTexSubImage2D+0x249(target=3553, level=0, xoffset=0, yoffset=0, x=0, y=0, width=512, height=512) [/home/stefan/src/mesa/src/mesa/main/teximage.c:3241] in r300_dri.so (0x0033dde4) 5 0x7db4363e read_from_framebuffer_texture+0x19e(This=<register EDI not in topmost frame>, srgb=<is not available>) [/home/stefan/src/wine/linux/dlls/wined3d/../../../dlls/wined3d/surface.c:1048] in wined3d<elf> (0x0033de44) <more wine code following> The X server writes the following: RADEON DRM CS failure - corruptions/glitches may occur -22 bufmgr: last submission : r:192 vs g:534769664 w:3356160 vs v:110455596 And then the screen seems dead. ssh still works It may be possible that Wine calls glTexSubImage2D with a deleted texture name or the 0 texture bound. Team Fortress 2 releases the current rendertarget, and Wine does not handle this properly, resulting in complaints from the GL driver with other drivers. The logs indicate that I get the same errors back from Mesa, so this might be a red herring, and the crash occurs during a valid operation. In any case, the X server should not crash, no matter how badly the app behaves.
Any better with a more recent kernel & mesa checkout ?
The kernel is David Airlie's radeon modeset kernel. Is there any newer kernel? Mesa was the git version from yesterday, I'll retry it with your patches later today.
More fix were pushed in mesa radeon-rewrite branch would be nice if you could test with lastest commit.
I'll retest tomorrow. I only had time for a r200 test session today.
I've switched to your drm-next-radeon kernel branch(b1e6187f5ad3c8b836a68009006584337d914e99) and DDX driver(fba534017e581fcd9b9e49ba0b281fb500f576a7). The bug is different, but TF2 still does not start. I don't know yet if it is the same bug not. Mesa is updated to d7cc0eb47930d6c8ebfd18fefbe48fe8eec696a0. The symptoms now are: *) TF2 starts up and the driver complains(correctly) about invalid texture operations. TF2 continues startup, but the static loading background is not visible. *) Without any obvious error message, the whole system slows down. TF2 doesn't crash, but I can move the mouse pointer only every other second and overall redraw is slow. I can kill the X server via Ctrl+C(started it from an SSH connection). The framebuffer redraws properly. After an X server restart the server is still slow. There are some suspicious messages in the dmesg. I'll try to get a log.
Do you still see this crash with current Mesa master after the merge?
I'll have to retest. I am currently busy with other things unfortunately, so it will take a while before I have the time to update and retest.
It should be fixed on current mesa master branch (at least for r300-r500 gpus). Can you confirm?
Same state as in my last comment unfortunately. My git xorg setup is fairly outdated, and I'll be away for the Christmas holidays. In the meantime I have however updated my xorg git installation on my r250 card. That one has made big progress, but there are still a few more bugs to report. That's off-topic for this bug report though since TF2 won't run in Wine on that card. (Lack of a GL_ATI_fragment_shader based pixel shader backend)
(In reply to comment #9) > Same state as in my last comment unfortunately. My git xorg setup is fairly > outdated, and I'll be away for the Christmas holidays. > > In the meantime I have however updated my xorg git installation on my r250 > card. That one has made big progress, but there are still a few more bugs to > report. That's off-topic for this bug report though since TF2 won't run in Wine > on that card. (Lack of a GL_ATI_fragment_shader based pixel shader backend) > Can you try mesa/7.8 branch? There are a few r200 fixes, also it contains hw accelerated support for glCopyTex(Sub)Image, so the game should run much faster.
The classic r300 driver has been abandoned long ago. It was replaced by the Gallium driver r300g. If you have issues with r300g please file a new bug report with component Drivers/Gallium/r300 Thanks.
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.