Quoting Patchwork (2019-08-26 08:57:32) > * igt@i915_selftest@live_gem_contexts: > - fi-ivb-3770: [PASS][3] -> [DMESG-FAIL][4] > [3]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_6783/fi-ivb-3770/igt@i915_selftest@live_gem_contexts.html > [4]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_14187/fi-ivb-3770/igt@i915_selftest@live_gem_contexts.html ivb/hsw looked to be much more of a timing issue wrt to TLB invalidation. Adding msleep() and whacking memory barriers around the place appeared to improve the situation, but that may have just been wishful thinking. Short term, I think we need to switch gen7 back to aliasing-ppgtt until the TLB mess can be resolved (all we've changing in this series is timing, so the same incorrect page access should be possible in current kernels just harder to hit). -Chris _______________________________________________ Intel-gfx mailing list Intel-gfx@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/intel-gfx
