Created attachment 87932 [details] /sys/class/drm/card0/error Hi, Linux x30 3.12.0-994-generic #201310160505 SMP Wed Oct 16 09:15:44 UTC 2013 i686 GNU/Linux from Ubuntu drm-intel-nightly. On 3.11 and 3.12rc kernels the system freezes after few minutes, with drm-intel-nightly it continues witg acceleration disabled. "AccelMethod" "uxa" works, but it's rather slow. lcpci: 00:00.0 Host bridge: Intel Corporation 82830M/MG/MP Host Bridge (rev 04) 00:02.0 VGA compatible controller: Intel Corporation 82830M/MG Integrated Graphics Controller (rev 04) 00:02.1 Display controller: Intel Corporation 82830M/MG Integrated Graphics Controller ...
The error-state is bogus...
Please apply diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 76e54352111e..d10a0c3156f1 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -1740,7 +1740,7 @@ struct drm_i915_file_private { #define OVERLAY_NEEDS_PHYSICAL(dev) (INTEL_INFO(dev)->overlay_needs_physical) /* Early gen2 have a totally busted CS tlb and require pinned batches. */ -#define HAS_BROKEN_CS_TLB(dev) (IS_I830(dev) || IS_845G(dev)) +#define HAS_BROKEN_CS_TLB(dev) (0) /* With the 945 and later, Y tiling got adjusted so that it was 32 128-byte * rows, which changed the alignment requirements and fence programming. so that we can capture the right GPU dump.
The the below patch help by any chance? http://cgit.freedesktop.org/~danvet/xf86-video-intel/commit/
It's not the usual incoherence...
Created attachment 87968 [details] /sys/class/drm/card0/error (#define HAS_BROKEN_CS_TLB(dev) (0))
Can you please try this hack to xf86-video-intel: diff --git a/src/sna/kgem.c b/src/sna/kgem.c index f959875..927a432 100644 --- a/src/sna/kgem.c +++ b/src/sna/kgem.c @@ -943,6 +943,9 @@ static bool test_has_caching(struct kgem *kgem) if (kgem->gen == 040) return false; + if (kgem->gen == 020) + return false; + handle = gem_create(kgem->fd, 1); if (handle == 0) return false;
With #define HAS_BROKEN_CS_TLB(dev) (0)), i have a crash on SNA and UXA, but with patched from Chris Wilson SNA works. I have no errors for 2 hours, but with option "DRI" "false". Without it, glxgears makes GPU crash: 4551.004051] [drm] stuck on render ring [ 4551.004077] [drm] GPU crash dump saved to /sys/class/drm/card0/error [ 4551.004083] [drm] GPU hangs can indicate a bug anywhere in the entire gfx stack, including userspace. [ 4551.004088] [drm] Please file a _new_ bug report on bugs.freedesktop.org against DRI -> DRM/Intel [ 4551.004094] [drm] drm/i915 developers can then reassign to the right component if it's not a kernel issue. [ 4551.004099] [drm] The gpu crash dump is required to analyze gpu hangs, so please always attach it. [ 4551.014974] [drm:i915_reset] *ERROR* Failed to reset chip.
Daniel, can you set your 830gm running through i-g-t, perhaps we need to beat the snoop tests a bit harder?
The w/a may only be possible in the DDX, but at the moment this looks to be a hardware issue so we may be missing some protections in the kernel for example.
I was waiting on Daniel doing some testing on 830..
Is it still happening on latest drm-intel-nightly and with latest xf86-video-intel?
More than a year since we've heard from reporter, and apparently Daniel isn't testing with the 830 either... I'll close this next time I come across here.
Oh well, timed out. Hopefully this is fixed.
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.