@@ -1193,7 +1193,7 @@ static void a6xx_recover(struct msm_gpu *gpu)
{
struct adreno_gpu *adreno_gpu = to_adreno_gpu(gpu);
struct a6xx_gpu *a6xx_gpu = to_a6xx_gpu(adreno_gpu);
- int i;
+ int i, active_submits;
adreno_dump_info(gpu);
@@ -1210,8 +1210,34 @@ static void a6xx_recover(struct msm_gpu *gpu)
*/
gmu_write(&a6xx_gpu->gmu, REG_A6XX_GMU_GMU_PWR_COL_KEEPALIVE, 0);
- gpu->funcs->pm_suspend(gpu);
- gpu->funcs->pm_resume(gpu);
+ pm_runtime_dont_use_autosuspend(&gpu->pdev->dev);
+
+ /* active_submit won't change until we make a submission */
+ mutex_lock(&gpu->active_lock);
+ active_submits = gpu->active_submits;
+
+ /*
+ * Temporarily clear active_submits count to silence a WARN() in the
+ * runtime suspend cb
+ */
+ gpu->active_submits = 0;
+
+ /* Drop the rpm refcount from active submits */
+ if (active_submits)
+ pm_runtime_put(&gpu->pdev->dev);
+
+ /* And the final one from recover worker */
+ pm_runtime_put_sync(&gpu->pdev->dev);
+
+ pm_runtime_use_autosuspend(&gpu->pdev->dev);
+
+ if (active_submits)
+ pm_runtime_get(&gpu->pdev->dev);
+
+ pm_runtime_get_sync(&gpu->pdev->dev);
+
+ gpu->active_submits = active_submits;
+ mutex_unlock(&gpu->active_lock);
msm_gpu_hw_init(gpu);
}
@@ -426,9 +426,7 @@ static void recover_worker(struct kthread_work *work)
/* retire completed submits, plus the one that hung: */
retire_submits(gpu);
- pm_runtime_get_sync(&gpu->pdev->dev);
gpu->funcs->recover(gpu);
- pm_runtime_put_sync(&gpu->pdev->dev);
/*
* Replay all remaining submits starting with highest priority
@@ -445,7 +443,7 @@ static void recover_worker(struct kthread_work *work)
}
}
- pm_runtime_put_sync(&gpu->pdev->dev);
+ pm_runtime_put(&gpu->pdev->dev);
mutex_unlock(&gpu->lock);
There are some hardware logic under CX domain. For a successful recovery, we should ensure cx headswitch collapses to ensure all the stale states are cleard out. This is especially true to for a6xx family where we can GMU co-processor. Currently, cx doesn't collapse due to a devlink between gpu and its smmu. So the *struct gpu device* needs to be runtime suspended to ensure that the iommu driver removes its vote on cx gdsc. Signed-off-by: Akhil P Oommen <quic_akhilpo@quicinc.com> --- (no changes since v4) Changes in v4: - Keep active_submit lock across the suspend & resume (Rob) - Clear gpu->active_submits to silence a WARN() during runpm suspend (Rob) Changes in v3: - Simplied the pm refcount drop since we have just a single refcount now for all active submits drivers/gpu/drm/msm/adreno/a6xx_gpu.c | 32 +++++++++++++++++++++++++++++--- drivers/gpu/drm/msm/msm_gpu.c | 4 +--- 2 files changed, 30 insertions(+), 6 deletions(-)