Merge "msm: ais: Remove the active_req_cnt < 2 limitation" (b3de3ad6) · Commits · e / devices / android_kernel_xiaomi_nabu

drivers/media/platform/msm/ais/cam_isp/cam_isp_context.c

+3 −35

Original line number	Diff line number	Diff line
		@@ -1698,9 +1698,7 @@ static int __cam_isp_ctx_apply_req_in_activated_state(
		{
		int rc = 0;
		struct cam_ctx_request *req;
		struct cam_ctx_request *active_req = NULL;
		struct cam_isp_ctx_req *req_isp;
		struct cam_isp_ctx_req *active_req_isp;
		struct cam_isp_context *ctx_isp = NULL;
		struct cam_hw_config_args cfg;

		@@ -1755,32 +1753,6 @@ static int __cam_isp_ctx_apply_req_in_activated_state(
		req->request_id, ctx_isp->substate_activated, ctx->ctx_id);
		req_isp = (struct cam_isp_ctx_req *) req->req_priv;

		if (ctx_isp->active_req_cnt >= 2) {
		CAM_ERR_RATE_LIMIT(CAM_ISP,
		"Reject apply request (id %lld) due to congestion(cnt = %d) ctx %u",
		req->request_id,
		ctx_isp->active_req_cnt,
		ctx->ctx_id);

		spin_lock_bh(&ctx->lock);
		if (!list_empty(&ctx->active_req_list))
		active_req = list_first_entry(&ctx->active_req_list,
		struct cam_ctx_request, list);
		else
		CAM_ERR_RATE_LIMIT(CAM_ISP,
		"WARNING: should not happen (cnt = %d) but active_list empty",
		ctx_isp->active_req_cnt);
		spin_unlock_bh(&ctx->lock);

		if (active_req) {
		active_req_isp =
		(struct cam_isp_ctx_req *) active_req->req_priv;
		__cam_isp_ctx_handle_buf_done_fail_log(active_req_isp);
		}

		rc = -EFAULT;
		goto end;
		}
		req_isp->bubble_report = apply->report_if_bubble;

		cfg.ctxt_to_hw_map = ctx_isp->hw_ctx;
		@@ -2113,14 +2085,10 @@ static int __cam_isp_ctx_rdi_only_sof_in_top_state(
		ctx_isp->frame_id, ctx_isp->sof_timestamp_val);

		/*
		* notify reqmgr with sof signal. Note, due to scheduling delay
		* we can run into situation that two active requests has already
		* be in the active queue while we try to do the notification.
		* In this case, we need to skip the current notification. This
		* helps the state machine to catch up the delay.
		* Note, always notify SOF to CRM, since there is
		* another mechnism to handle irq delay
		*/
		if (ctx->ctx_crm_intf && ctx->ctx_crm_intf->notify_trigger &&
		ctx_isp->active_req_cnt <= 2) {
		if (ctx->ctx_crm_intf && ctx->ctx_crm_intf->notify_trigger) {
		notify.link_hdl = ctx->link_hdl;
		notify.dev_hdl = ctx->dev_hdl;
		notify.frame_id = ctx_isp->frame_id;