Regression 2.6.34+ -> 2.6.34-rc5: radeon KMS rs780 problems
in [Kernel]
|
From: Ed Tomlinson on 17 Jul 2010 19:00 Hi, My 790gx chipset supplies an rs780. With .34+ all is ok. With .35-rc5 basic function seems ok (kde works) but games like extreme tux racer have problems. The cursor does not display when on the tux racer window and various objects flicker on and off as I attempt to play the game. Here is what .34+ detects (think this kernel has a drm-radeon-testing from may or early june) [ 1.370735] [drm] Initialized drm 1.1.0 20060810 [ 1.376014] [drm] radeon defaulting to kernel modesetting. [ 1.382041] [drm] radeon kernel modesetting enabled. [ 1.387560] radeon 0000:01:05.0: PCI INT A -> GSI 18 (level, low) -> IRQ 18 [ 1.395205] radeon 0000:01:05.0: setting latency timer to 64 [ 1.397347] [drm] initializing kernel modesetting (RS780 0x1002:0x9614). [ 1.404961] [drm] register mmio base: 0xFE9F0000 [ 1.410036] [drm] register mmio size: 65536 [ 1.415328] ATOM BIOS: B27722 [ 1.418612] [drm] Clocks initialized ! [ 1.422740] [drm] 3 Power State(s) [ 1.426486] [drm] State 0 Default (default) [ 1.431081] [drm] 1 Clock Mode(s) [ 1.434831] [drm] 0 engine: 300000 [ 1.438773] [drm] State 1 Performance [ 1.442897] [drm] 1 Clock Mode(s) [ 1.446648] [drm] 0 engine: 500000 [ 1.450589] [drm] State 2 Default [ 1.454344] [drm] 1 Clock Mode(s) [ 1.458093] [drm] 0 engine: 500000 [ 1.462038] [drm] radeon: power management initialized [ 1.467705] radeon 0000:01:05.0: VRAM: 256M 0xC0000000 - 0xCFFFFFFF (256M used) [ 1.475724] radeon 0000:01:05.0: GTT: 512M 0xA0000000 - 0xBFFFFFFF [ 1.485117] [drm] Detected VRAM RAM=256M, BAR=256M [ 1.490391] [drm] RAM width 32bits DDR [ 1.494785] [TTM] Zone kernel: Available graphics memory: 3956838 kiB. [ 1.502042] [TTM] Zone dma32: Available graphics memory: 2097152 kiB. [ 1.509465] [drm] radeon: 256M of VRAM memory ready [ 1.514825] [drm] radeon: 512M of GTT memory ready. [ 1.520236] [drm] radeon: irq initialized. [ 1.524739] [drm] GART: num cpu pages 131072, num gpu pages 131072 [ 1.532253] [drm] Loading RS780 Microcode with .35-rc5 it finds a different ammount of VRAM. Hardware wise there is 128M of sideport ram. [ 1.361622] [drm] Initialized drm 1.1.0 20060810 [ 1.366843] [drm] radeon defaulting to kernel modesetting. [ 1.372863] [drm] radeon kernel modesetting enabled. [ 1.378371] radeon 0000:01:05.0: PCI INT A -> GSI 18 (level, low) -> IRQ 18 [ 1.386003] radeon 0000:01:05.0: setting latency timer to 64 [ 1.388242] [drm] initializing kernel modesetting (RS780 0x1002:0x9614). [ 1.395870] [drm] register mmio base: 0xFE9F0000 [ 1.400940] [drm] register mmio size: 65536 [ 1.406238] ATOM BIOS: B27722 [ 1.409531] [drm] Clocks initialized ! [ 1.413688] radeon 0000:01:05.0: VRAM: 384M 0xC0000000 - 0xD7FFFFFF (384M used) [ 1.421706] radeon 0000:01:05.0: GTT: 512M 0xA0000000 - 0xBFFFFFFF [ 1.429481] [drm] Detected VRAM RAM=384M, BAR=256M [ 1.436429] [drm] RAM width 32bits DDR [ 1.440829] [TTM] Zone kernel: Available graphics memory: 3956866 kiB. [ 1.448096] [TTM] Zone dma32: Available graphics memory: 2097152 kiB. [ 1.455348] [TTM] Initializing pool allocator. [ 1.460471] [drm] radeon: 384M of VRAM memory ready [ 1.465824] [drm] radeon: 512M of GTT memory ready. [ 1.471255] [drm] radeon: irq initialized. [ 1.475760] [drm] GART: num cpu pages 131072, num gpu pages 131072 [ 1.483280] [drm] Loading RS780 Microcode User space/firmware/bios is the same in both cases. Idea? Ed Tomlinson -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo(a)vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
From: Alex Deucher on 19 Jul 2010 10:30 On Sat, Jul 17, 2010 at 6:56 PM, Ed Tomlinson <edt(a)aei.ca> wrote: > Hi, > > My 790gx chipset supplies an rs780. �With .34+ all is ok. �With .35-rc5 basic function seems ok (kde works) but > games like extreme tux racer have problems. �The cursor does not display when on the tux racer window and > various objects flicker on and off as I attempt to play the game. Did you update anything on the userspace side (ddx, mesa, xserver)? Can you bisect what commit is causing the problems? > > Here is what .34+ detects (think this kernel has a drm-radeon-testing from may or early june) > > [ � �1.370735] [drm] Initialized drm 1.1.0 20060810 > [ � �1.376014] [drm] radeon defaulting to kernel modesetting. > [ � �1.382041] [drm] radeon kernel modesetting enabled. > [ � �1.387560] radeon 0000:01:05.0: PCI INT A -> GSI 18 (level, low) -> IRQ 18 > [ � �1.395205] radeon 0000:01:05.0: setting latency timer to 64 > [ � �1.397347] [drm] initializing kernel modesetting (RS780 0x1002:0x9614). > [ � �1.404961] [drm] register mmio base: 0xFE9F0000 > [ � �1.410036] [drm] register mmio size: 65536 > [ � �1.415328] ATOM BIOS: B27722 > [ � �1.418612] [drm] Clocks initialized ! > [ � �1.422740] [drm] 3 Power State(s) > [ � �1.426486] [drm] State 0 Default (default) > [ � �1.431081] [drm] � �1 Clock Mode(s) > [ � �1.434831] [drm] � � � � � �0 engine: 300000 > [ � �1.438773] [drm] State 1 Performance > [ � �1.442897] [drm] � �1 Clock Mode(s) > [ � �1.446648] [drm] � � � � � �0 engine: 500000 > [ � �1.450589] [drm] State 2 Default > [ � �1.454344] [drm] � �1 Clock Mode(s) > [ � �1.458093] [drm] � � � � � �0 engine: 500000 > [ � �1.462038] [drm] radeon: power management initialized > [ � �1.467705] radeon 0000:01:05.0: VRAM: 256M 0xC0000000 - 0xCFFFFFFF (256M used) > [ � �1.475724] radeon 0000:01:05.0: GTT: 512M 0xA0000000 - 0xBFFFFFFF > [ � �1.485117] [drm] Detected VRAM RAM=256M, BAR=256M > [ � �1.490391] [drm] RAM width 32bits DDR > [ � �1.494785] [TTM] Zone �kernel: Available graphics memory: 3956838 kiB. > [ � �1.502042] [TTM] Zone � dma32: Available graphics memory: 2097152 kiB. > [ � �1.509465] [drm] radeon: 256M of VRAM memory ready > [ � �1.514825] [drm] radeon: 512M of GTT memory ready. > [ � �1.520236] [drm] radeon: irq initialized. > [ � �1.524739] [drm] GART: num cpu pages 131072, num gpu pages 131072 > [ � �1.532253] [drm] Loading RS780 Microcode > > with .35-rc5 it finds a different ammount of VRAM. �Hardware wise there is 128M of sideport ram. 2.6.35 added support for vram greater then the size of the PCI framebuffer BAR (256 MB in your case). The amount of vram accessible to the CPU is limited by the size of the PCI framebuffer BAR, however the GPU can access the entire amount. 2.6.35 works around this restriction so that the CPU can access the entire amount of vram. Alex > > �[ � �1.361622] [drm] Initialized drm 1.1.0 20060810 > �[ � �1.366843] [drm] radeon defaulting to kernel modesetting. > �[ � �1.372863] [drm] radeon kernel modesetting enabled. > �[ � �1.378371] radeon 0000:01:05.0: PCI INT A -> GSI 18 (level, low) -> IRQ 18 > �[ � �1.386003] radeon 0000:01:05.0: setting latency timer to 64 > �[ � �1.388242] [drm] initializing kernel modesetting (RS780 0x1002:0x9614). > �[ � �1.395870] [drm] register mmio base: 0xFE9F0000 > �[ � �1.400940] [drm] register mmio size: 65536 > �[ � �1.406238] ATOM BIOS: B27722 > �[ � �1.409531] [drm] Clocks initialized ! > �[ � �1.413688] radeon 0000:01:05.0: VRAM: 384M 0xC0000000 - 0xD7FFFFFF (384M used) > �[ � �1.421706] radeon 0000:01:05.0: GTT: 512M 0xA0000000 - 0xBFFFFFFF > �[ � �1.429481] [drm] Detected VRAM RAM=384M, BAR=256M > �[ � �1.436429] [drm] RAM width 32bits DDR > �[ � �1.440829] [TTM] Zone �kernel: Available graphics memory: 3956866 kiB. > �[ � �1.448096] [TTM] Zone � dma32: Available graphics memory: 2097152 kiB. > �[ � �1.455348] [TTM] Initializing pool allocator. > �[ � �1.460471] [drm] radeon: 384M of VRAM memory ready > �[ � �1.465824] [drm] radeon: 512M of GTT memory ready. > �[ � �1.471255] [drm] radeon: irq initialized. > �[ � �1.475760] [drm] GART: num cpu pages 131072, num gpu pages 131072 > �[ � �1.483280] [drm] Loading RS780 Microcode > > User space/firmware/bios is the same in both cases. > > Idea? > Ed Tomlinson > _______________________________________________ > dri-devel mailing list > dri-devel(a)lists.freedesktop.org > http://lists.freedesktop.org/mailman/listinfo/dri-devel > -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo(a)vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
From: Michel Dänzer on 19 Jul 2010 10:50 On Sam, 2010-07-17 at 18:56 -0400, Ed Tomlinson wrote: > > My 790gx chipset supplies an rs780. With .34+ all is ok. > With .35-rc5 basic function seems ok (kde works) but games like > extreme tux racer have problems. The cursor does not display when on > the tux racer window and various objects flicker on and off as I > attempt to play the game. Sounds like it might be https://bugs.freedesktop.org/show_bug.cgi?id=28341 . Does the X driver patch attached there fix the problem? -- Earthling Michel Dänzer | http://www.vmware.com Libre software enthusiast | Debian, X and DRI developer -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo(a)vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
From: Ed Tomlinson on 19 Jul 2010 23:10 On Monday 19 July 2010 10:34:20 Michel Dänzer wrote: > On Sam, 2010-07-17 at 18:56 -0400, Ed Tomlinson wrote: > > > > My 790gx chipset supplies an rs780. With .34+ all is ok. > > With .35-rc5 basic function seems ok (kde works) but games like > > extreme tux racer have problems. The cursor does not display when on > > the tux racer window and various objects flicker on and off as I > > attempt to play the game. > > Sounds like it might be > https://bugs.freedesktop.org/show_bug.cgi?id=28341 . Does the X driver > patch attached there fix the problem? Michel, xf86-video-ati git from today fails. Applying the revert patch in the above message fixes the problem. Its an interesting bug in that the kernel change triggers the driver bug. Might be a good idea to get this fixed, one way or another, before -35 gets released? Let me know if you want me to test anything. Thanks, Ed Tomlinson -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo(a)vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
From: Michel Dänzer on 20 Jul 2010 02:10 On Mon, 2010-07-19 at 23:00 -0400, Ed Tomlinson wrote: > On Monday 19 July 2010 10:34:20 Michel Dänzer wrote: > > On Sam, 2010-07-17 at 18:56 -0400, Ed Tomlinson wrote: > > > > > > My 790gx chipset supplies an rs780. With .34+ all is ok. > > > With .35-rc5 basic function seems ok (kde works) but games like > > > extreme tux racer have problems. The cursor does not display when on > > > the tux racer window and various objects flicker on and off as I > > > attempt to play the game. > > > > Sounds like it might be > > https://bugs.freedesktop.org/show_bug.cgi?id=28341 . Does the X driver > > patch attached there fix the problem? > > Michel, > > xf86-video-ati git from today fails. Applying the revert patch in the above message > fixes the problem. Its an interesting bug in that the kernel change triggers the > driver bug. Might be a good idea to get this fixed, one way or another, before > -35 gets released? It's basically a userspace bug, the buggy userspace code is just only active with new functionality in the new kernel. -- Earthling Michel Dänzer | http://www.vmware.com Libre software enthusiast | Debian, X and DRI developer -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo(a)vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
|
Pages: 1 Prev: zcache: page cache compression support Next: charity projects, |