On Tue, Apr 3, 2018 at 9:06 AM, Michel Dänzer <michel at daenzer.net> wrote: > On 2018-04-03 11:07 AM, Emily Deng wrote: >> For vega10 sriov, still need to use one package for invalidate tlb >> flush >> >> Emily Deng (8): >> Revert "Revert "drm/amdgpu: add emit_reg_write_reg_wait ring >> callback"" >> Revert "Revert "drm/amdgpu/sdma4: add emit_reg_write_reg_wait ring >> callback"" >> Revert "Revert "drm/amdgpu/uvd7: add emit_reg_write_reg_wait ring >> callback"" >> Revert "Revert "drm/amdgpu/vce4: add emit_reg_write_reg_wait ring >> callback"" >> Revert "Revert "drm/amdgpu/vcn1: add emit_reg_write_reg_wait ring >> callback"" >> Revert "Revert "drm/amdgpu/gmc9: use >> amdgpu_ring_emit_reg_write_reg_wait in gpu tlb flush"" >> Revert "Revert "drm/amdgpu/sdma4:fix sdma engine hang"" >> drm/amdgpu/gfx9: add emit_reg_write_reg_wait ring callback > > 'Revert "Revert "...""' is pretty ugly, better to use the original > commit logs again (possibly with amendments explaining why they're being > applied gain). I agree. If the fix from Ray fixes the hangs on gfx, please just generate new patches by cherry-picking the originals and squashing in any fixes and updating the commit messages if necessary. E.g., you can squash your sdma4 fix into the original sdma4 patch. Thanks, Alex
