Major enhancement in this series is the support for a minimal gmu coredump which can be captured inline instead of through our usual recover worker. It is helpful in the case of gmu errors during gpu wake-up/suspend path and helps to capture a snapshot of gmu before we do a suspend. I had to introduce a lock to synchronize the crashstate because the runtime-suspend can happen from an asynchronous RPM thread.
Apart from this, there are some improvements to gracefully handle the gmu errors by propagating the error back to parent or by retrying. Also, a few patches to fix some trivial bugs in the related code.
Akhil P Oommen (10): drm/msm/a6xx: Add helper to check smmu is stalled drm/msm/a6xx: Send NMI to gmu when it is hung drm/msm/a6xx: Avoid gmu lock in pm ops drm/msm/a6xx: Enhance debugging of gmu faults drm/msm: Do recovery on hw_init failure drm/msm/a6xx: Propagate OOB set error drm/msm/adreno: Retry on gpu resume failure drm/msm/a6xx: Remove clk votes on failure drm/msm: Remove pm_runtime_get() from msm_job_run() drm/msm/a6xx: Free gmu_debug crashstate bo
drivers/gpu/drm/msm/adreno/a6xx_gmu.c | 89 +++++++++++++++++++++++------ drivers/gpu/drm/msm/adreno/a6xx_gmu.h | 1 + drivers/gpu/drm/msm/adreno/a6xx_gpu.c | 31 +++++++--- drivers/gpu/drm/msm/adreno/a6xx_gpu.h | 4 +- drivers/gpu/drm/msm/adreno/a6xx_gpu_state.c | 79 +++++++++++++++++++++---- drivers/gpu/drm/msm/adreno/adreno_device.c | 10 +++- drivers/gpu/drm/msm/adreno/adreno_gpu.c | 10 +++- drivers/gpu/drm/msm/adreno/adreno_gpu.h | 2 + drivers/gpu/drm/msm/msm_gpu.c | 28 ++++++++- drivers/gpu/drm/msm/msm_gpu.h | 11 ++-- drivers/gpu/drm/msm/msm_ringbuffer.c | 4 -- 11 files changed, 218 insertions(+), 51 deletions(-)