Hi, The current reality is that after gpu hang or after first suspend/resume cycle, we have lost all our workarounds. It is somewhat troubling to think that what ghosts we have been hunting and what effort we have wasted on the testruns that have had unrelated single hang at some point of the 10k+ item testset. Also with gen8 we have been thinking that we have workarounds, but effectively we have run without some of them, even after boot until Arun's ring based wa setup got merged. Some of the workarounds touched a part of context that gen8 just threw away when we switched to our 'default' one. And when writes started to really stick, we see all kinds of interesting phenomena. For example: https://bugs.freedesktop.org/show_bug.cgi?id=83482 This series is my first attempt to get into deterministic, identical state after driver load, reset and resume. There is more work to be done but I posted this as RFC to get feedback on the direction before I go forward with other gens. So here are the patches, for BDW and IVB, to be able to verify the workarounds with igt, after reset/resume. Patch 3 was needed to get deterministic results on bdw. Any feedback will be greatly appreciated, -Mika Mika Kuoppala (6): drm/i915: Reinitialize default context after reset drm/i915: Reinitialize default context after resume drm/i915: Disable render idle on user forcewake drm/i915: Build workaround list in ring initialization drm/i915: Add ivb workarounds into the wa list. drm/i915: Check workaround status on dfs read time drivers/gpu/drm/i915/i915_debugfs.c | 30 ++-- drivers/gpu/drm/i915/i915_drv.c | 1 + drivers/gpu/drm/i915/i915_drv.h | 32 ++-- drivers/gpu/drm/i915/i915_gem_context.c | 42 +++-- drivers/gpu/drm/i915/i915_reg.h | 6 +- drivers/gpu/drm/i915/intel_lrc.c | 4 +- drivers/gpu/drm/i915/intel_pm.c | 61 +------ drivers/gpu/drm/i915/intel_ringbuffer.c | 282 +++++++++++++++++++++++--------- drivers/gpu/drm/i915/intel_uncore.c | 37 ++++- 9 files changed, 308 insertions(+), 187 deletions(-) -- 1.9.1 _______________________________________________ Intel-gfx mailing list Intel-gfx@xxxxxxxxxxxxxxxxxxxxx http://lists.freedesktop.org/mailman/listinfo/intel-gfx
