msm: eva: Enhance session error handling

Ensure smmu fault dump before calling BUG_ON. Rely on WD timeout
for FW hang debugging.

Change-Id: I0d7fba31dbc602b244af74da6774885c56af0fcc
Signed-off-by: George Shen <quic_sqiao@quicinc.com>

msm/eva/msm_cvp_buf.c

@@ -633,9 +633,9 @@ int msm_cvp_unmap_buf_wncc(struct msm_cvp_inst *inst,
 		inst->unused_wncc_bufs.smem[idx] = *(cbuf->smem);
 		inst->unused_wncc_bufs.nr++;
 		inst->unused_wncc_bufs.nr =
-			(inst->unused_wncc_bufs.nr > MAX_FRAME_BUFFER_NUMS)?
+			(inst->unused_wncc_bufs.nr > NUM_WNCC_BUFS)?
-			MAX_FRAME_BUFFER_NUMS : inst->unused_wncc_bufs.nr;
+			NUM_WNCC_BUFS : inst->unused_wncc_bufs.nr;
-		inst->unused_wncc_bufs.ktid = ++idx % MAX_FRAME_BUFFER_NUMS;
+		inst->unused_wncc_bufs.ktid = ++idx % NUM_WNCC_BUFS;
 	}
 	mutex_unlock(&inst->cvpwnccbufs.lock);
 
@@ -1880,6 +1880,7 @@ int msm_cvp_session_deinit_buffers(struct msm_cvp_inst *inst)
 	return rc;
 }
 
+#define MAX_NUM_FRAMES_DUMP 4
 void msm_cvp_print_inst_bufs(struct msm_cvp_inst *inst, bool log)
 {
 	struct cvp_internal_buf *buf;
@@ -1916,10 +1917,18 @@ void msm_cvp_print_inst_bufs(struct msm_cvp_inst *inst, bool log)
 	dprintk(CVP_ERR, "frame buffer list\n");
 	mutex_lock(&inst->frames.lock);
 	list_for_each_entry(frame, &inst->frames.list, list) {
-		dprintk(CVP_ERR, "frame no %d tid %llx bufs\n", i++, frame->ktid);
+		i++;
-		for (c = 0; c < frame->nr; c++)
+		if (i <= MAX_NUM_FRAMES_DUMP) {
-			_log_smem(snap, inst, frame->bufs[c].smem, log);
+			dprintk(CVP_ERR, "frame no %d tid %llx bufs\n",
+					i, frame->ktid);
+			for (c = 0; c < frame->nr; c++)
+				_log_smem(snap, inst, frame->bufs[c].smem,
+						log);
+		}
 	}
+	if (i > MAX_NUM_FRAMES_DUMP)
+		dprintk(CVP_ERR, "Skipped %d frames' buffers\n",
+				(i - MAX_NUM_FRAMES_DUMP));
 	mutex_unlock(&inst->frames.lock);
 
 	mutex_lock(&inst->cvpdspbufs.lock);

msm/eva/msm_cvp_common.c

@@ -22,11 +22,6 @@
 
 static void handle_session_error(enum hal_command_response cmd, void *data);
 
-static void msm_cvp_comm_generate_session_error(struct msm_cvp_inst *inst)
-{
-	dprintk(CVP_WARN, "%s function is deprecated\n", __func__);
-}
-
 static void dump_hfi_queue(struct iris_hfi_device *device)
 {
 	struct cvp_hfi_queue_header *queue;
@@ -567,28 +562,6 @@ static void handle_session_error(enum hal_command_response cmd, void *data)
 	cvp_put_inst(inst);
 }
 
-static void msm_comm_clean_notify_client(struct msm_cvp_core *core)
-{
-	struct msm_cvp_inst *inst = NULL;
-
-	if (!core) {
-		dprintk(CVP_ERR, "%s: Invalid params\n", __func__);
-		return;
-	}
-
-	dprintk(CVP_WARN, "%s: Core %pK\n", __func__, core);
-	mutex_lock(&core->lock);
-
-	list_for_each_entry(inst, &core->instances, list) {
-		mutex_lock(&inst->lock);
-		inst->state = MSM_CVP_CORE_INVALID;
-		mutex_unlock(&inst->lock);
-		dprintk(CVP_WARN,
-			"%s Send sys error for inst %pK\n", __func__, inst);
-	}
-	mutex_unlock(&core->lock);
-}
-
 void handle_sys_error(enum hal_command_response cmd, void *data)
 {
 	struct msm_cvp_cb_cmd_done *response = data;
@@ -848,11 +821,18 @@ static int msm_comm_session_abort(struct msm_cvp_inst *inst)
 	int rc = 0, abort_completion = 0;
 	struct cvp_hfi_device *hdev;
 
+
 	if (!inst || !inst->core || !inst->core->device) {
 		dprintk(CVP_ERR, "%s invalid params\n", __func__);
 		return -EINVAL;
 	}
+
 	hdev = inst->core->device;
+	print_hfi_queue_info(hdev);
+	if (1)
+		return 0;
+
+	/* Activate code below for Watchdog timeout testing */
 	abort_completion = SESSION_MSG_INDEX(HAL_SESSION_ABORT_DONE);
 
 	dprintk(CVP_WARN, "%s: inst %pK session %x\n", __func__,
@@ -882,44 +862,7 @@ exit:
 
 static void handle_thermal_event(struct msm_cvp_core *core)
 {
-	int rc = 0;
+	dprintk(CVP_WARN, "Deprecated thermal_event handler\n");
-	struct msm_cvp_inst *inst;
-
-	if (!core || !core->device) {
-		dprintk(CVP_ERR, "%s Invalid params\n", __func__);
-		return;
-	}
-	mutex_lock(&core->lock);
-	list_for_each_entry(inst, &core->instances, list) {
-		if (!inst->session)
-			continue;
-
-		mutex_unlock(&core->lock);
-		if (inst->state >= MSM_CVP_OPEN_DONE &&
-			inst->state < MSM_CVP_CLOSE_DONE) {
-			dprintk(CVP_WARN, "%s: abort inst %pK\n",
-				__func__, inst);
-			rc = msm_comm_session_abort(inst);
-			if (rc) {
-				dprintk(CVP_ERR,
-					"%s session_abort failed rc: %d\n",
-					__func__, rc);
-				goto err_sess_abort;
-			}
-			change_cvp_inst_state(inst, MSM_CVP_CORE_INVALID);
-			dprintk(CVP_WARN,
-				"%s Send sys error for inst %pK\n",
-				__func__, inst);
-		} else {
-			msm_cvp_comm_generate_session_error(inst);
-		}
-		mutex_lock(&core->lock);
-	}
-	mutex_unlock(&core->lock);
-	return;
-
-err_sess_abort:
-	msm_comm_clean_notify_client(core);
 }
 
 void msm_cvp_comm_handle_thermal_event(void)

msm/eva/msm_cvp_dsp.h

@@ -15,7 +15,7 @@
 #include <linux/sched.h>
 
 #ifdef CVP_FASTRPC_ENABLED
-#include <linux/fastrpc.h>
+#include <fastrpc.h>
 #else
 struct fastrpc_device {
 	int handle;

msm/eva/msm_cvp_res_parse.c

@@ -1038,13 +1038,13 @@ int msm_cvp_smmu_fault_handler(struct iommu_domain *domain,
 		return -EINVAL;
 	}
 
+	mutex_lock(&core->lock);
 	core->smmu_fault_count++;
 	if (!core->last_fault_addr)
 		core->last_fault_addr = iova;
 	dprintk(CVP_ERR, "%s - faulting address: %lx, %d\n",
 		__func__, iova, core->smmu_fault_count);
 
-	mutex_lock(&core->lock);
 	log = (core->log.snapshot_index > 0)? false : true;
 	list_for_each_entry(inst, &core->instances, list) {
 		cvp_print_inst(CVP_ERR, inst);