From: Thierry Reding treding@nvidia.com
Hi Ben,
these are fixes for a couple of issues that I've been running into when testing on various Tegra boards. The first two patches fix up issues in the fix that I had sent out earlier to fix the regression introduced in drm-misc-next. The first one is critical because it avoids a BUG_ON as reported by Ilia, while the second is less critical, but restores the locking correctness (at least to the best of my knowledge).
Patch 3 is something that I think was also caused by the reservation object rework and is kind of a continuation of my earlier attempt to fix the VMA node sharing breakage. The current ordering between TTM and GEM teardown is causing a DEBUG_LOCKS_WARN_ON() because GEM cleanup already freed a mutex that TTM teardown will still want to use.
Lastly, patch 4 is quite uncritical, but it's a one-line change that is causing an ugly (but harmless) external memory address decode error on Tegra210 and later. It seems that for some reason clearing this register will cause a DMA operation to be started by the GPU. I've verified that it's tied to exactly that register write by modifying the value written to the register, and stalling for a couple of seconds after the register write. The address decode error reflects the value written into this register exactly and it always happens a couple of milliseconds after this write.
Thierry
Thierry Reding (4): drm/nouveau: Fix fallout from reservation object rework drm/nouveau: prime: Extend DMA reservation object lock drm/nouveau: Fix ordering between TTM and GEM release drm/nouveau: gm20b: Avoid BAR1 teardown during init
drivers/gpu/drm/nouveau/nouveau_bo.c | 26 +++++++++++------- drivers/gpu/drm/nouveau/nouveau_bo.h | 4 +-- drivers/gpu/drm/nouveau/nouveau_gem.c | 7 ++--- drivers/gpu/drm/nouveau/nouveau_prime.c | 27 ++++++++++++------- .../gpu/drm/nouveau/nvkm/subdev/bar/gm20b.c | 1 - 5 files changed, 39 insertions(+), 26 deletions(-)