The retire worker is kicked for each fence, either the normal way by signaling the fence from the event completion interrupt or by the recover worker if the GPU got stuck. Moving the RPM put into the retire worker allows us to have it in a single place for both cases.
This also shaves off quite a bit of the CPU time spent in hardirq context, as arming the autosuspend timer when the RPM refcount drops to 0 is a relatively costly operation.
Signed-off-by: Lucas Stach l.stach@pengutronix.de --- drivers/gpu/drm/etnaviv/etnaviv_gpu.c | 22 ++++++++-------------- 1 file changed, 8 insertions(+), 14 deletions(-)
diff --git a/drivers/gpu/drm/etnaviv/etnaviv_gpu.c b/drivers/gpu/drm/etnaviv/etnaviv_gpu.c index 056a72e6ed26..ca949ae19933 100644 --- a/drivers/gpu/drm/etnaviv/etnaviv_gpu.c +++ b/drivers/gpu/drm/etnaviv/etnaviv_gpu.c @@ -806,11 +806,6 @@ static void recover_worker(struct work_struct *work) gpu->event[i].fence = NULL; gpu->event[i].used = false; complete(&gpu->event_free); - /* - * Decrement the PM count for each stuck event. This is safe - * even in atomic context as we use ASYNC RPM here. - */ - pm_runtime_put_autosuspend(gpu->dev); } spin_unlock_irqrestore(&gpu->event_spinlock, flags); gpu->completed_fence = gpu->active_fence; @@ -1093,6 +1088,14 @@ static void retire_worker(struct work_struct *work) }
etnaviv_gpu_cmdbuf_free(cmdbuf); + /* + * We need to balance the runtime PM count caused by + * each submission. Upon submission, we increment + * the runtime PM counter, and allocate one event. + * So here, we put the runtime PM count for each + * completed event. + */ + pm_runtime_put_autosuspend(gpu->dev); }
gpu->retired_fence = fence; @@ -1313,15 +1316,6 @@ static irqreturn_t irq_handler(int irq, void *data) gpu->completed_fence = fence->seqno;
event_free(gpu, event); - - /* - * We need to balance the runtime PM count caused by - * each submission. Upon submission, we increment - * the runtime PM counter, and allocate one event. - * So here, we put the runtime PM count for each - * completed event. - */ - pm_runtime_put_autosuspend(gpu->dev); }
/* Retire the buffer objects in a work */