diff --git a/drivers/gpu/nvgpu/common/fifo/pbdma_status.c b/drivers/gpu/nvgpu/common/fifo/pbdma_status.c
index 0c214d273..4b434ac21 100644
--- a/drivers/gpu/nvgpu/common/fifo/pbdma_status.c
+++ b/drivers/gpu/nvgpu/common/fifo/pbdma_status.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2019, NVIDIA CORPORATION.  All rights reserved.
+ * Copyright (c) 2019-2022, NVIDIA CORPORATION.  All rights reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -42,6 +42,11 @@ bool nvgpu_pbdma_status_is_chsw_valid(struct nvgpu_pbdma_status_info
 {
 	return pbdma_status->chsw_status == NVGPU_PBDMA_CHSW_STATUS_VALID;
 }
+bool nvgpu_pbdma_status_ch_not_loaded(struct nvgpu_pbdma_status_info
+		*pbdma_status)
+{
+	return pbdma_status->chsw_status == NVGPU_PBDMA_CHSW_STATUS_INVALID;
+}
 bool nvgpu_pbdma_status_is_id_type_tsg(struct nvgpu_pbdma_status_info
 		*pbdma_status)
 {
diff --git a/drivers/gpu/nvgpu/common/fifo/tsg.c b/drivers/gpu/nvgpu/common/fifo/tsg.c
index 8e22be557..df68830f9 100644
--- a/drivers/gpu/nvgpu/common/fifo/tsg.c
+++ b/drivers/gpu/nvgpu/common/fifo/tsg.c
@@ -621,9 +621,8 @@ void nvgpu_tsg_set_error_notifier(struct gk20a *g, struct nvgpu_tsg *tsg,
 		u32 error_notifier)
 {
 	struct nvgpu_channel *ch = NULL;
-	u32 max_error_notifier_id = NVGPU_ERR_NOTIFIER_PBDMA_PUSHBUFFER_CRC_MISMATCH;
 
-	if (error_notifier > max_error_notifier_id) {
+	if (error_notifier >= NVGPU_ERR_NOTIFIER_INVAL) {
 		return;
 	}
 
diff --git a/drivers/gpu/nvgpu/common/rc/rc.c b/drivers/gpu/nvgpu/common/rc/rc.c
index ddab7014e..23023f9b3 100644
--- a/drivers/gpu/nvgpu/common/rc/rc.c
+++ b/drivers/gpu/nvgpu/common/rc/rc.c
@@ -35,6 +35,7 @@
 #include <nvgpu/error_notifier.h>
 #include <nvgpu/pbdma_status.h>
 #include <nvgpu/rc.h>
+#include <nvgpu/nvgpu_init.h>
 
 void nvgpu_rc_fifo_recover(struct gk20a *g, u32 eng_bitmask,
 			u32 hw_id, bool id_is_tsg,
@@ -94,11 +95,18 @@ void nvgpu_rc_ctxsw_timeout(struct gk20a *g, u32 eng_bitmask,
 #endif
 }
 
-void nvgpu_rc_pbdma_fault(struct gk20a *g, u32 pbdma_id, u32 error_notifier,
+int nvgpu_rc_pbdma_fault(struct gk20a *g, u32 pbdma_id, enum nvgpu_err_notif error_notifier,
 			struct nvgpu_pbdma_status_info *pbdma_status)
 {
-	u32 id;
 	u32 id_type = PBDMA_STATUS_ID_TYPE_INVALID;
+	int err = 0;
+	u32 id;
+
+	if (error_notifier >= NVGPU_ERR_NOTIFIER_INVAL) {
+		nvgpu_err(g, "Invalid error notifier %u", error_notifier);
+		err = -EINVAL;
+		goto out;
+	}
 
 	nvgpu_log(g, gpu_dbg_info, "pbdma id %d error notifier %d",
 			pbdma_id, error_notifier);
@@ -111,10 +119,14 @@ void nvgpu_rc_pbdma_fault(struct gk20a *g, u32 pbdma_id, u32 error_notifier,
 			nvgpu_pbdma_status_is_chsw_switch(pbdma_status)) {
 		id = pbdma_status->next_id;
 		id_type = pbdma_status->next_id_type;
-	} else {
+	} else if (nvgpu_pbdma_status_ch_not_loaded(pbdma_status)) {
 		/* Nothing to do here */
-		nvgpu_err(g, "Invalid pbdma_status.id");
-		return;
+		nvgpu_log_info(g, "no channel loaded on pbdma.");
+		goto out;
+	} else {
+		nvgpu_err(g, "pbdma status not valid");
+		err = -EINVAL;
+		goto out;
 	}
 
 	if (id_type == PBDMA_STATUS_ID_TYPE_TSGID) {
@@ -128,7 +140,8 @@ void nvgpu_rc_pbdma_fault(struct gk20a *g, u32 pbdma_id, u32 error_notifier,
 		struct nvgpu_tsg *tsg;
 		if (ch == NULL) {
 			nvgpu_err(g, "channel is not referenceable");
-			return;
+			err = -EINVAL;
+			goto out;
 		}
 
 		tsg = nvgpu_tsg_from_ch(ch);
@@ -138,12 +151,21 @@ void nvgpu_rc_pbdma_fault(struct gk20a *g, u32 pbdma_id, u32 error_notifier,
 				RC_TYPE_PBDMA_FAULT);
 		} else {
 			nvgpu_err(g, "chid: %d is not bound to tsg", ch->chid);
+			err = -EINVAL;
 		}
 
 		nvgpu_channel_put(ch);
 	} else {
-		nvgpu_err(g, "Invalid pbdma_status.id_type");
+		nvgpu_err(g, "Invalid pbdma_status id_type or next_id_type");
+		err = -EINVAL;
 	}
+
+out:
+	if (err != 0) {
+		nvgpu_sw_quiesce(g);
+	}
+
+	return err;
 }
 
 void nvgpu_rc_runlist_update(struct gk20a *g, u32 runlist_id)
diff --git a/drivers/gpu/nvgpu/hal/fifo/fifo_intr_ga10b.h b/drivers/gpu/nvgpu/hal/fifo/fifo_intr_ga10b.h
index fad9272d2..cda6624f3 100644
--- a/drivers/gpu/nvgpu/hal/fifo/fifo_intr_ga10b.h
+++ b/drivers/gpu/nvgpu/hal/fifo/fifo_intr_ga10b.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2020-2021, NVIDIA CORPORATION.  All rights reserved.
+ * Copyright (c) 2020-2022, NVIDIA CORPORATION.  All rights reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -36,7 +36,6 @@ void ga10b_fifo_intr_0_isr(struct gk20a *g);
 void ga10b_fifo_intr_set_recover_mask(struct gk20a *g);
 void ga10b_fifo_intr_unset_recover_mask(struct gk20a *g);
 
-void ga10b_fifo_pbdma_isr(struct gk20a *g, struct nvgpu_runlist *runlist, u32 pbdma_idx);
 void ga10b_fifo_runlist_intr_retrigger(struct gk20a *g, u32 intr_tree);
 
 #endif /* NVGPU_FIFO_INTR_GA10B_H */
diff --git a/drivers/gpu/nvgpu/hal/fifo/fifo_intr_ga10b_fusa.c b/drivers/gpu/nvgpu/hal/fifo/fifo_intr_ga10b_fusa.c
index cfa0deb3d..df67ce015 100644
--- a/drivers/gpu/nvgpu/hal/fifo/fifo_intr_ga10b_fusa.c
+++ b/drivers/gpu/nvgpu/hal/fifo/fifo_intr_ga10b_fusa.c
@@ -317,6 +317,36 @@ static void ga10b_fifo_runlist_intr_clear(struct gk20a *g)
 	}
 }
 
+static int ga10b_fifo_pbdma_isr(struct gk20a *g, struct nvgpu_runlist *runlist,
+				u32 pbdma_idx)
+{
+	u32 pbdma_id;
+	const struct nvgpu_pbdma_info *pbdma_info;
+	int err;
+
+	if (pbdma_idx >= PBDMA_PER_RUNLIST_SIZE) {
+		nvgpu_err(g, "pbdma_idx(%d) >= max_pbdmas_per_runlist(%d)",
+				pbdma_idx, PBDMA_PER_RUNLIST_SIZE);
+		return -EINVAL;
+	}
+	pbdma_info = runlist->pbdma_info;
+	pbdma_id = pbdma_info->pbdma_id[pbdma_idx];
+	if (pbdma_id == PBDMA_ID_INVALID) {
+		nvgpu_err(g, "runlist_id(%d), pbdma_idx(%d): invalid PBDMA",
+				runlist->id, pbdma_idx);
+		return -EINVAL;
+	}
+
+	err = g->ops.pbdma.handle_intr(g, pbdma_id, true);
+	if (err != 0) {
+		nvgpu_err(g, "pbdma intr failed id: %u", pbdma_idx);
+		return err;
+	}
+
+	return err;
+}
+
+
 void ga10b_fifo_intr_0_isr(struct gk20a *g)
 {
 	u32 i, intr_0, handled_intr_0 = 0U;
@@ -324,6 +354,7 @@ void ga10b_fifo_intr_0_isr(struct gk20a *g)
 	u32 pbdma_idx = 0U;
 	u32 intr_tree_0 = 0U, intr_tree_1 = 1U;
 	struct nvgpu_runlist *runlist;
+	int err = 0;
 
 	/* TODO: sw_ready is needed only for recovery part */
 	if (!g->fifo.sw_ready) {
@@ -349,7 +380,17 @@ void ga10b_fifo_intr_0_isr(struct gk20a *g)
 			pbdma_idx++) {
 			if (intr_0 &
 				runlist_intr_0_pbdmai_intr_tree_j_pending_f(pbdma_idx, intr_tree_0)) {
-				ga10b_fifo_pbdma_isr(g, runlist, pbdma_idx);
+				/**
+				 * Quiesce is triggered as part of nvgpu_rc_pbdma_fault
+				 * failure case, so -
+				 * 1. Avoid looping through the rest of the PBDMAs by
+				 *    adding a return statement here.
+				 * 2. Avoid re-triggering the PBDMA ISR by returning
+				 *    pbdma_intr field value here in handled_intr_0.
+				 */
+				if (err == 0) {
+					err = ga10b_fifo_pbdma_isr(g, runlist, pbdma_idx);
+				}
 				handled_intr_0 |= runlist_intr_0_pbdmai_intr_tree_j_pending_f(pbdma_idx, intr_tree_0);
 			}
 		}
@@ -456,27 +497,6 @@ void ga10b_fifo_intr_unset_recover_mask(struct gk20a *g)
 
 }
 
-
-void ga10b_fifo_pbdma_isr(struct gk20a *g, struct nvgpu_runlist *runlist, u32 pbdma_idx)
-{
-	u32 pbdma_id;
-	const struct nvgpu_pbdma_info *pbdma_info;
-
-	if (pbdma_idx >= PBDMA_PER_RUNLIST_SIZE) {
-		nvgpu_err(g, "pbdma_idx(%d) >= max_pbdmas_per_runlist(%d)",
-				pbdma_idx, PBDMA_PER_RUNLIST_SIZE);
-		return;
-	}
-	pbdma_info = runlist->pbdma_info;
-	pbdma_id = pbdma_info->pbdma_id[pbdma_idx];
-	if (pbdma_id == PBDMA_ID_INVALID) {
-		nvgpu_err(g, "runlist_id(%d), pbdma_idx(%d): invalid PBDMA",
-				runlist->id, pbdma_idx);
-		return;
-	}
-	g->ops.pbdma.handle_intr(g, pbdma_id, true);
-}
-
 void ga10b_fifo_runlist_intr_retrigger(struct gk20a *g, u32 intr_tree)
 {
 	u32 i = 0U;
diff --git a/drivers/gpu/nvgpu/hal/fifo/fifo_intr_gk20a_fusa.c b/drivers/gpu/nvgpu/hal/fifo/fifo_intr_gk20a_fusa.c
index 41a30a174..2e25cecac 100644
--- a/drivers/gpu/nvgpu/hal/fifo/fifo_intr_gk20a_fusa.c
+++ b/drivers/gpu/nvgpu/hal/fifo/fifo_intr_gk20a_fusa.c
@@ -32,6 +32,7 @@
 #include <nvgpu/nvgpu_err.h>
 #include <nvgpu/cic_mon.h>
 #include <nvgpu/engines.h>
+#include <nvgpu/nvgpu_init.h>
 
 #include <hal/fifo/fifo_intr_gk20a.h>
 #include <hal/fifo/mmu_fault_gk20a.h>
@@ -92,12 +93,25 @@ u32 gk20a_fifo_pbdma_isr(struct gk20a *g)
 	u32 pbdma_id;
 	u32 num_pbdma = nvgpu_get_litter_value(g, GPU_LIT_HOST_NUM_PBDMA);
 	u32 pbdma_pending_bitmask = nvgpu_readl(g, fifo_intr_pbdma_id_r());
+	int err;
 
 	for (pbdma_id = 0; pbdma_id < num_pbdma; pbdma_id++) {
 		if (fifo_intr_pbdma_id_status_v(pbdma_pending_bitmask, pbdma_id) != 0U) {
 			nvgpu_log(g, gpu_dbg_intr, "pbdma id %d intr pending",
 				pbdma_id);
-			g->ops.pbdma.handle_intr(g, pbdma_id, true);
+			/**
+			 * Quiesce is triggered as part of nvgpu_rc_pbdma_fault
+			 * failure case, so -
+			 * 1. Avoid looping through the rest of the PBDMAs by
+			 *    adding a return statement here.
+			 * 2. Avoid re-triggering the PBDMA ISR by returning
+			 *    pbdma_intr field value here.
+			 */
+			err = g->ops.pbdma.handle_intr(g, pbdma_id, true);
+			if (err != 0) {
+				nvgpu_err(g, "pbdma intr failed id: %u", pbdma_id);
+				break;
+			}
 		}
 	}
 	return fifo_intr_0_pbdma_intr_pending_f();
diff --git a/drivers/gpu/nvgpu/hal/fifo/pbdma_ga10b.h b/drivers/gpu/nvgpu/hal/fifo/pbdma_ga10b.h
index 3656fe7a9..fa240c2c1 100644
--- a/drivers/gpu/nvgpu/hal/fifo/pbdma_ga10b.h
+++ b/drivers/gpu/nvgpu/hal/fifo/pbdma_ga10b.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2020-2021, NVIDIA CORPORATION.  All rights reserved.
+ * Copyright (c) 2020-2022, NVIDIA CORPORATION.  All rights reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -41,7 +41,7 @@ struct nvgpu_pbdma_status_info;
 struct nvgpu_device;
 
 void ga10b_pbdma_intr_enable(struct gk20a *g, bool enable);
-void ga10b_pbdma_handle_intr(struct gk20a *g, u32 pbdma_id, bool recover);
+int ga10b_pbdma_handle_intr(struct gk20a *g, u32 pbdma_id, bool recover);
 bool ga10b_pbdma_handle_intr_0(struct gk20a *g, u32 pbdma_id, u32 pbdma_intr_0,
 			u32 *error_notifier);
 bool ga10b_pbdma_handle_intr_1(struct gk20a *g, u32 pbdma_id, u32 pbdma_intr_1,
diff --git a/drivers/gpu/nvgpu/hal/fifo/pbdma_ga10b_fusa.c b/drivers/gpu/nvgpu/hal/fifo/pbdma_ga10b_fusa.c
index 1ae99bc3b..e7cf37c2f 100644
--- a/drivers/gpu/nvgpu/hal/fifo/pbdma_ga10b_fusa.c
+++ b/drivers/gpu/nvgpu/hal/fifo/pbdma_ga10b_fusa.c
@@ -358,11 +358,11 @@ void ga10b_pbdma_intr_enable(struct gk20a *g, bool enable)
 	}
 }
 
-
-void ga10b_pbdma_handle_intr(struct gk20a *g, u32 pbdma_id, bool recover)
+int ga10b_pbdma_handle_intr(struct gk20a *g, u32 pbdma_id, bool recover)
 {
 	struct nvgpu_pbdma_status_info pbdma_status;
 	u32 intr_error_notifier = NVGPU_ERR_NOTIFIER_PBDMA_ERROR;
+	int err = 0;
 
 	u32 pbdma_intr_0 = nvgpu_readl(g, pbdma_intr_0_r(pbdma_id));
 	u32 pbdma_intr_1 = nvgpu_readl(g, pbdma_intr_1_r(pbdma_id));
@@ -377,9 +377,12 @@ void ga10b_pbdma_handle_intr(struct gk20a *g, u32 pbdma_id, bool recover)
 			g->ops.pbdma_status.read_pbdma_status_info(g,
 				pbdma_id, &pbdma_status);
 			if (recover) {
-				nvgpu_rc_pbdma_fault(g, pbdma_id,
+				err = nvgpu_rc_pbdma_fault(g, pbdma_id,
 						intr_error_notifier,
 						&pbdma_status);
+				if (err != 0) {
+					nvgpu_err(g, "recovery failed");
+				}
 			}
 		}
 		nvgpu_writel(g, pbdma_intr_0_r(pbdma_id), pbdma_intr_0);
@@ -391,17 +394,22 @@ void ga10b_pbdma_handle_intr(struct gk20a *g, u32 pbdma_id, bool recover)
 			pbdma_id, pbdma_intr_1);
 
 		if (g->ops.pbdma.handle_intr_1(g, pbdma_id, pbdma_intr_1,
-			&intr_error_notifier)) {
+			&intr_error_notifier) && (err == 0)) {
 			g->ops.pbdma_status.read_pbdma_status_info(g,
 				pbdma_id, &pbdma_status);
 			if (recover) {
-				nvgpu_rc_pbdma_fault(g, pbdma_id,
+				err = nvgpu_rc_pbdma_fault(g, pbdma_id,
 						intr_error_notifier,
 						&pbdma_status);
+				if (err != 0) {
+					nvgpu_err(g, "recovery failed");
+				}
 			}
 		}
 		nvgpu_writel(g, pbdma_intr_1_r(pbdma_id), pbdma_intr_1);
 	}
+
+	return err;
 }
 
 static bool ga10b_pbdma_handle_intr_0_legacy(struct gk20a *g, u32 pbdma_id,
diff --git a/drivers/gpu/nvgpu/hal/fifo/pbdma_gm20b.h b/drivers/gpu/nvgpu/hal/fifo/pbdma_gm20b.h
index 0a99a7c4a..191bc6434 100644
--- a/drivers/gpu/nvgpu/hal/fifo/pbdma_gm20b.h
+++ b/drivers/gpu/nvgpu/hal/fifo/pbdma_gm20b.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2014-2020, NVIDIA CORPORATION.  All rights reserved.
+ * Copyright (c) 2014-2022, NVIDIA CORPORATION.  All rights reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -34,7 +34,7 @@ struct nvgpu_device;
 
 bool gm20b_pbdma_handle_intr_0(struct gk20a *g, u32 pbdma_id,
 			u32 pbdma_intr_0, u32 *error_notifier);
-void gm20b_pbdma_handle_intr(struct gk20a *g, u32 pbdma_id, bool recover);
+int gm20b_pbdma_handle_intr(struct gk20a *g, u32 pbdma_id, bool recover);
 
 u32 gm20b_pbdma_read_data(struct gk20a *g, u32 pbdma_id);
 void gm20b_pbdma_reset_header(struct gk20a *g, u32 pbdma_id);
diff --git a/drivers/gpu/nvgpu/hal/fifo/pbdma_gm20b_fusa.c b/drivers/gpu/nvgpu/hal/fifo/pbdma_gm20b_fusa.c
index 695c03a73..dd0b056b5 100644
--- a/drivers/gpu/nvgpu/hal/fifo/pbdma_gm20b_fusa.c
+++ b/drivers/gpu/nvgpu/hal/fifo/pbdma_gm20b_fusa.c
@@ -323,10 +323,11 @@ u32 gm20b_pbdma_restartable_0_intr_descs(void)
 	return restartable_0_intr_descs;
 }
 
-void gm20b_pbdma_handle_intr(struct gk20a *g, u32 pbdma_id, bool recover)
+int gm20b_pbdma_handle_intr(struct gk20a *g, u32 pbdma_id, bool recover)
 {
 	struct nvgpu_pbdma_status_info pbdma_status;
 	u32 intr_error_notifier = NVGPU_ERR_NOTIFIER_PBDMA_ERROR;
+	int err = 0;
 
 	u32 pbdma_intr_0 = nvgpu_readl(g, pbdma_intr_0_r(pbdma_id));
 	u32 pbdma_intr_1 = nvgpu_readl(g, pbdma_intr_1_r(pbdma_id));
@@ -341,9 +342,12 @@ void gm20b_pbdma_handle_intr(struct gk20a *g, u32 pbdma_id, bool recover)
 			g->ops.pbdma_status.read_pbdma_status_info(g,
 				pbdma_id, &pbdma_status);
 			if (recover) {
-				nvgpu_rc_pbdma_fault(g, pbdma_id,
+				err = nvgpu_rc_pbdma_fault(g, pbdma_id,
 						intr_error_notifier,
 						&pbdma_status);
+				if (err != 0) {
+					nvgpu_err(g, "recovery failed");
+				}
 			}
 		}
 		nvgpu_writel(g, pbdma_intr_0_r(pbdma_id), pbdma_intr_0);
@@ -355,18 +359,23 @@ void gm20b_pbdma_handle_intr(struct gk20a *g, u32 pbdma_id, bool recover)
 			pbdma_id, pbdma_intr_1);
 
 		if (g->ops.pbdma.handle_intr_1(g, pbdma_id, pbdma_intr_1,
-			&intr_error_notifier)) {
+			&intr_error_notifier) && (err == 0)) {
 			g->ops.pbdma_status.read_pbdma_status_info(g,
 				pbdma_id, &pbdma_status);
 			if (recover) {
-				nvgpu_rc_pbdma_fault(g, pbdma_id,
+				err = nvgpu_rc_pbdma_fault(g, pbdma_id,
 						intr_error_notifier,
 						&pbdma_status);
+				if (err != 0) {
+					nvgpu_err(g, "recovery failed");
+				}
 			}
 		}
 
 		nvgpu_writel(g, pbdma_intr_1_r(pbdma_id), pbdma_intr_1);
 	}
+
+	return err;
 }
 
 u32 gm20b_pbdma_get_gp_base(u64 gpfifo_base)
diff --git a/drivers/gpu/nvgpu/hal/fifo/preempt_gv11b_fusa.c b/drivers/gpu/nvgpu/hal/fifo/preempt_gv11b_fusa.c
index ff10839e4..38d202a5f 100644
--- a/drivers/gpu/nvgpu/hal/fifo/preempt_gv11b_fusa.c
+++ b/drivers/gpu/nvgpu/hal/fifo/preempt_gv11b_fusa.c
@@ -131,7 +131,11 @@ int gv11b_fifo_preempt_poll_pbdma(struct gk20a *g, u32 tsgid,
 		 * reported to SW.
 		 */
 
-		g->ops.pbdma.handle_intr(g, pbdma_id, false);
+		ret = g->ops.pbdma.handle_intr(g, pbdma_id, false);
+		if (ret != 0) {
+			nvgpu_err(g, "pbdma intr failed id: %u %d", pbdma_id, ret);
+			break;
+		}
 
 		g->ops.pbdma_status.read_pbdma_status_info(g,
 			pbdma_id, &pbdma_status);
diff --git a/drivers/gpu/nvgpu/include/nvgpu/error_notifier.h b/drivers/gpu/nvgpu/include/nvgpu/error_notifier.h
index d97c67002..d01cdacf2 100644
--- a/drivers/gpu/nvgpu/include/nvgpu/error_notifier.h
+++ b/drivers/gpu/nvgpu/include/nvgpu/error_notifier.h
@@ -27,7 +27,7 @@
 
 struct nvgpu_channel;
 
-enum {
+enum nvgpu_err_notif {
 	NVGPU_ERR_NOTIFIER_FIFO_ERROR_IDLE_TIMEOUT = 0,
 	NVGPU_ERR_NOTIFIER_GR_ERROR_SW_METHOD,
 	NVGPU_ERR_NOTIFIER_GR_ERROR_SW_NOTIFY,
@@ -40,6 +40,7 @@ enum {
 	NVGPU_ERR_NOTIFIER_RESETCHANNEL_VERIF_ERROR,
 	NVGPU_ERR_NOTIFIER_PBDMA_PUSHBUFFER_CRC_MISMATCH,
 	NVGPU_ERR_NOTIFIER_CE_ERROR,
+	NVGPU_ERR_NOTIFIER_INVAL,
 };
 
 void nvgpu_set_err_notifier_locked(struct nvgpu_channel *ch, u32 error);
diff --git a/drivers/gpu/nvgpu/include/nvgpu/gops/pbdma.h b/drivers/gpu/nvgpu/include/nvgpu/gops/pbdma.h
index d903afc5c..722d4b508 100644
--- a/drivers/gpu/nvgpu/include/nvgpu/gops/pbdma.h
+++ b/drivers/gpu/nvgpu/include/nvgpu/gops/pbdma.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2019-2021, NVIDIA CORPORATION.  All rights reserved.
+ * Copyright (c) 2019-2022, NVIDIA CORPORATION.  All rights reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -51,7 +51,7 @@ struct gops_pbdma {
 	bool (*handle_intr_1)(struct gk20a *g,
 			u32 pbdma_id, u32 pbdma_intr_1,
 			u32 *error_notifier);
-	void (*handle_intr)(struct gk20a *g, u32 pbdma_id, bool recover);
+	int (*handle_intr)(struct gk20a *g, u32 pbdma_id, bool recover);
 	u32 (*set_clear_intr_offsets) (struct gk20a *g,
 			u32 set_clear_size);
 	u32 (*get_signature)(struct gk20a *g);
diff --git a/drivers/gpu/nvgpu/include/nvgpu/pbdma_status.h b/drivers/gpu/nvgpu/include/nvgpu/pbdma_status.h
index 95c9e78ae..c0ca97756 100644
--- a/drivers/gpu/nvgpu/include/nvgpu/pbdma_status.h
+++ b/drivers/gpu/nvgpu/include/nvgpu/pbdma_status.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2019-2021, NVIDIA CORPORATION.  All rights reserved.
+ * Copyright (c) 2019-2022, NVIDIA CORPORATION.  All rights reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -150,6 +150,17 @@ bool nvgpu_pbdma_status_is_chsw_save(struct nvgpu_pbdma_status_info
  */
 bool nvgpu_pbdma_status_is_chsw_valid(struct nvgpu_pbdma_status_info
 		*pbdma_status);
+/**
+ * @brief Check if chsw_status is set to invalid.
+ *
+ * @param pbdma_status [in]	Pointer to struct containing pbdma_status h/w
+ *				reg/field value.
+ *
+ * @return Interprets #pbdma_status and returns true if channel
+ *         status is set to #NVGPU_PBDMA_CHSW_STATUS_INVALID else returns false.
+ */
+bool nvgpu_pbdma_status_ch_not_loaded(struct nvgpu_pbdma_status_info
+		*pbdma_status);
 /**
  * @brief Check if id_type is tsg.
  *
diff --git a/drivers/gpu/nvgpu/include/nvgpu/rc.h b/drivers/gpu/nvgpu/include/nvgpu/rc.h
index 08cbb7e14..a75ed9f13 100644
--- a/drivers/gpu/nvgpu/include/nvgpu/rc.h
+++ b/drivers/gpu/nvgpu/include/nvgpu/rc.h
@@ -103,6 +103,7 @@ struct nvgpu_tsg;
 struct nvgpu_channel;
 struct nvgpu_pbdma_status_info;
 struct mmu_fault_info;
+enum nvgpu_err_notif;
 
 static inline const char *nvgpu_rc_type_to_str(unsigned int rc_type)
 {
@@ -172,8 +173,16 @@ void nvgpu_rc_ctxsw_timeout(struct gk20a *g, u32 eng_bitmask,
  *
  * Do PBDMA fault recovery. Set error notifier as per \a error_notifier and call
  * \a nvgpu_rc_tsg_and_related_engines to do the recovery.
+ *
+ * @return 0 in case of success, < 0 in case of failure.
+ * @retval -EINVAL in case of following cases:
+ * 1. the error_notifier is invalid.
+ * 2. the pbdma status is invalid.
+ * 3. the channel is not referenceable.
+ * 4. the channel is not bound to tsg.
+ * 5. the id type or next_id type are not indicating channel id type or tsg id type.
  */
-void nvgpu_rc_pbdma_fault(struct gk20a *g, u32 pbdma_id, u32 error_notifier,
+int nvgpu_rc_pbdma_fault(struct gk20a *g, u32 pbdma_id, enum nvgpu_err_notif error_notifier,
 			struct nvgpu_pbdma_status_info *pbdma_status);
 
 /**
diff --git a/libs/dgpu/libnvgpu-drv-dgpu_safe.export b/libs/dgpu/libnvgpu-drv-dgpu_safe.export
index 0be8774aa..2a7e0b6d5 100644
--- a/libs/dgpu/libnvgpu-drv-dgpu_safe.export
+++ b/libs/dgpu/libnvgpu-drv-dgpu_safe.export
@@ -594,6 +594,7 @@ nvgpu_pbdma_status_is_chsw_load
 nvgpu_pbdma_status_is_chsw_save
 nvgpu_pbdma_status_is_chsw_switch
 nvgpu_pbdma_status_is_chsw_valid
+nvgpu_pbdma_status_ch_not_loaded
 nvgpu_pbdma_status_is_id_type_tsg
 nvgpu_pbdma_status_is_next_id_type_tsg
 nvgpu_pbdma_setup_sw
diff --git a/libs/igpu/libnvgpu-drv-igpu_safe.export b/libs/igpu/libnvgpu-drv-igpu_safe.export
index 041be4cde..be292eb22 100644
--- a/libs/igpu/libnvgpu-drv-igpu_safe.export
+++ b/libs/igpu/libnvgpu-drv-igpu_safe.export
@@ -614,6 +614,7 @@ nvgpu_pbdma_status_is_chsw_switch
 nvgpu_pbdma_status_is_chsw_valid
 nvgpu_pbdma_status_is_id_type_tsg
 nvgpu_pbdma_status_is_next_id_type_tsg
+nvgpu_pbdma_status_ch_not_loaded
 nvgpu_pbdma_setup_sw
 nvgpu_pd_alloc
 nvgpu_pd_cache_fini
diff --git a/userspace/units/fifo/fifo/gk20a/nvgpu-fifo-intr-gk20a.c b/userspace/units/fifo/fifo/gk20a/nvgpu-fifo-intr-gk20a.c
index 0d2a1d1ac..024c81da9 100644
--- a/userspace/units/fifo/fifo/gk20a/nvgpu-fifo-intr-gk20a.c
+++ b/userspace/units/fifo/fifo/gk20a/nvgpu-fifo-intr-gk20a.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2019-2021, NVIDIA CORPORATION.  All rights reserved.
+ * Copyright (c) 2019-2022, NVIDIA CORPORATION.  All rights reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -172,13 +172,15 @@ done:
 	return ret;
 }
 
-static void stub_pbdma_handle_intr(struct gk20a *g, u32 pbdma_id, bool recover)
+static int stub_pbdma_handle_intr(struct gk20a *g, u32 pbdma_id, bool recover)
 {
 	if (nvgpu_readl(g, fifo_intr_pbdma_id_r()) != BIT(pbdma_id)) {
 		u.fail = true;
 	}
 
 	u.count++;
+
+	return 0;
 }
 
 int test_gk20a_fifo_pbdma_isr(struct unit_module *m,
diff --git a/userspace/units/fifo/pbdma/nvgpu-pbdma.c b/userspace/units/fifo/pbdma/nvgpu-pbdma.c
index c10d3f09f..14ab9a62e 100644
--- a/userspace/units/fifo/pbdma/nvgpu-pbdma.c
+++ b/userspace/units/fifo/pbdma/nvgpu-pbdma.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2018-2020, NVIDIA CORPORATION.  All rights reserved.
+ * Copyright (c) 2018-2022, NVIDIA CORPORATION.  All rights reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -180,6 +180,9 @@ int test_pbdma_status(struct unit_module *m,
 		unit_assert(nvgpu_pbdma_status_is_chsw_valid(&pbdma_status) ==
 			(pbdma_status.chsw_status ==
 				NVGPU_PBDMA_CHSW_STATUS_VALID), goto done);
+		unit_assert(nvgpu_pbdma_status_ch_not_loaded(&pbdma_status) ==
+			(pbdma_status.chsw_status ==
+				NVGPU_PBDMA_CHSW_STATUS_INVALID), goto done);
 	}
 
 	pbdma_status.id_type = PBDMA_STATUS_ID_TYPE_CHID;
diff --git a/userspace/units/fifo/preempt/gv11b/nvgpu-preempt-gv11b.c b/userspace/units/fifo/preempt/gv11b/nvgpu-preempt-gv11b.c
index e86b4e56d..ecea5b68f 100644
--- a/userspace/units/fifo/preempt/gv11b/nvgpu-preempt-gv11b.c
+++ b/userspace/units/fifo/preempt/gv11b/nvgpu-preempt-gv11b.c
@@ -144,8 +144,9 @@ done:
 	return ret;
 }
 
-static void stub_pbdma_handle_intr(struct gk20a *g, u32 pbdma_id, bool recover)
+static int stub_pbdma_handle_intr(struct gk20a *g, u32 pbdma_id, bool recover)
 {
+	return 0;
 }
 
 static int stub_fifo_preempt_tsg(struct gk20a *g, struct nvgpu_tsg *tsg)
diff --git a/userspace/units/fifo/tsg/nvgpu-tsg.c b/userspace/units/fifo/tsg/nvgpu-tsg.c
index 3ad96acd5..a7d8e4adf 100644
--- a/userspace/units/fifo/tsg/nvgpu-tsg.c
+++ b/userspace/units/fifo/tsg/nvgpu-tsg.c
@@ -1579,7 +1579,7 @@ int test_nvgpu_tsg_set_error_notifier_bvec(struct unit_module *m,
 	int ret = 0;
 
 	u32 valid_error_notifier_ids[][2] = {{NVGPU_ERR_NOTIFIER_FIFO_ERROR_IDLE_TIMEOUT, NVGPU_ERR_NOTIFIER_PBDMA_PUSHBUFFER_CRC_MISMATCH}};
-	u32 invalid_error_notifier_ids[][2] = {{NVGPU_ERR_NOTIFIER_PBDMA_PUSHBUFFER_CRC_MISMATCH + 1, U32_MAX}};
+	u32 invalid_error_notifier_ids[][2] = {{NVGPU_ERR_NOTIFIER_INVAL, U32_MAX}};
 	u32 (*working_list)[2];
 	u32 error_code, error_notifier_range_len;
 	/*
diff --git a/userspace/units/fifo/tsg/nvgpu-tsg.h b/userspace/units/fifo/tsg/nvgpu-tsg.h
index 161cfbfb8..4cbebfc09 100644
--- a/userspace/units/fifo/tsg/nvgpu-tsg.h
+++ b/userspace/units/fifo/tsg/nvgpu-tsg.h
@@ -448,7 +448,7 @@ int test_tsg_mark_error(struct unit_module *m,
  * Input: None
  * Equivalence classes:
  * error_notifier
- * - Invalid : { NVGPU_ERR_NOTIFIER_PBDMA_PUSHBUFFER_CRC_MISMATCH + 1, U32_MAX }
+ * - Invalid : { NVGPU_ERR_NOTIFIER_INVAL, U32_MAX }
  * - Valid :   { NVGPU_ERR_NOTIFIER_FIFO_ERROR_IDLE_TIMEOUT, NVGPU_ERR_NOTIFIER_PBDMA_PUSHBUFFER_CRC_MISMATCH }
  *
  * Steps:
diff --git a/userspace/units/rc/nvgpu-rc.c b/userspace/units/rc/nvgpu-rc.c
index 457006142..cb59283ca 100644
--- a/userspace/units/rc/nvgpu-rc.c
+++ b/userspace/units/rc/nvgpu-rc.c
@@ -23,6 +23,7 @@
 #include <unistd.h>
 #include <unit/unit.h>
 #include <unit/io.h>
+#include <unit/utils.h>
 #include <nvgpu/types.h>
 #include <nvgpu/gk20a.h>
 #include <nvgpu/hal_init.h>
@@ -146,6 +147,9 @@ int test_rc_init(struct unit_module *m, struct gk20a *g, void *args)
 		goto clear_posix_channel;
 	}
 
+	/* initialize the seed for random number generation needed in bvec tests */
+	srand(time(0));
+
 	return UNIT_SUCCESS;
 
 clear_posix_channel:
@@ -294,13 +298,18 @@ int test_rc_mmu_fault(struct unit_module *m, struct gk20a *g, void *args)
 	return UNIT_SUCCESS;
 }
 
-#define F_RC_IS_CHSW_VALID_OR_SAVE  0U
-#define F_RC_IS_CHSW_LOAD_OR_SWITCH 1U
-#define F_RC_IS_CHSW_INVALID        2U
+#define F_RC_IS_CHSW_VALID_OR_SAVE                0U
+#define F_RC_IS_CHSW_LOAD_OR_SWITCH               1U
+#define F_RC_IS_CHSW_INVALID                      2U
+#define F_RC_IS_CHSW_INVALID_STATE_MIN            3U
+#define F_RC_IS_CHSW_INVALID_STATE_RANDOM         4U
+#define F_RC_IS_CHSW_INVALID_STATE_MAX            5U
 
 #define F_RC_ID_TYPE_TSG     	    0U
 #define F_RC_ID_TYPE_CH             1U
-#define F_RC_ID_TYPE_INVALID        2U
+#define F_RC_ID_TYPE_INVALID_MIN    2U
+#define F_RC_ID_TYPE_INVALID_RANDOM 3U
+#define F_RC_ID_TYPE_INVALID_MAX    4U
 
 #define F_RC_ID_TYPE_CH_NULL_CHANNEL  0U
 #define F_RC_ID_TYPE_CH_NULL_TSG      1U
@@ -309,13 +318,18 @@ int test_rc_mmu_fault(struct unit_module *m, struct gk20a *g, void *args)
 static const char *f_rc_chsw[] = {
 	"is_chsw_valid_or_save",
 	"is_chsw_load_or_switch",
-	"is_chsw_invalid",
+	"is_chsw_invalid channel not loaded on engine",
+	"is_chsw_inval_min",
+	"is_chsw_inval_random",
+	"is_chsw_inval_max",
 };
 
 static const char *f_rc_id_type[] = {
 	"id_type_tsg",
 	"id_type_ch",
-	"id_type_invalid",
+	"id_type_invalid_min",
+	"id_type_invalid_random",
+	"id_type_invalid_max",
 };
 
 static const char *f_rc_id_ch_subbranch[] = {
@@ -366,7 +380,13 @@ static void set_pbdma_info_id_type(u32 chsw_branches,
 			info->next_id_type = (chsw_branches == F_RC_IS_CHSW_LOAD_OR_SWITCH) ?
 				PBDMA_STATUS_NEXT_ID_TYPE_CHID : PBDMA_STATUS_NEXT_ID_TYPE_INVALID;
 		}
-	} else {
+	} else if (id_type_branches == F_RC_ID_TYPE_INVALID_MIN) {
+		info->id_type = PBDMA_STATUS_ID_TYPE_TSGID + 1;
+		info->next_id_type = PBDMA_STATUS_ID_TYPE_TSGID + 1;
+	} else if (id_type_branches == F_RC_ID_TYPE_INVALID_RANDOM) {
+		info->id_type = PBDMA_STATUS_ID_TYPE_TSGID + 2 + get_random_u32(PBDMA_STATUS_ID_TYPE_TSGID + 1, U32_MAX);
+		info->next_id_type = PBDMA_STATUS_ID_TYPE_TSGID + 2 + get_random_u32(PBDMA_STATUS_ID_TYPE_TSGID + 1, U32_MAX);
+	} else if (id_type_branches == F_RC_ID_TYPE_INVALID_MAX) {
 		info->id_type = PBDMA_STATUS_ID_INVALID;
 		info->next_id_type = PBDMA_STATUS_ID_INVALID;
 	}
@@ -374,7 +394,13 @@ static void set_pbdma_info_id_type(u32 chsw_branches,
 
 int test_rc_pbdma_fault(struct unit_module *m, struct gk20a *g, void *args)
 {
+	int notifiers[] = {NVGPU_ERR_NOTIFIER_FIFO_ERROR_IDLE_TIMEOUT, NVGPU_ERR_NOTIFIER_PBDMA_PUSHBUFFER_CRC_MISMATCH,
+			   NVGPU_ERR_NOTIFIER_INVAL,
+			   NVGPU_ERR_NOTIFIER_INVAL + 1 + get_random_u32(NVGPU_ERR_NOTIFIER_INVAL, INT_MAX), INT_MAX};
+	struct nvgpu_pbdma_status_info info = {0};
 	u32 chsw_branches, id_type_branches;
+	int err = UNIT_SUCCESS;
+	u32 i;
 	u32 chsw_subbranch;
 
 	struct nvgpu_channel *ch_without_tsg = NULL;
@@ -388,18 +414,19 @@ int test_rc_pbdma_fault(struct unit_module *m, struct gk20a *g, void *args)
 
 	g->sw_quiesce_pending = true;
 
-	for (chsw_branches = F_RC_IS_CHSW_VALID_OR_SAVE;
-		chsw_branches <= F_RC_IS_CHSW_INVALID; chsw_branches++) {
-		struct nvgpu_pbdma_status_info info = {0};
-
-		if (chsw_branches == F_RC_IS_CHSW_INVALID) {
-			info.chsw_status = NVGPU_PBDMA_CHSW_STATUS_INVALID;
-			unit_info(m, "%s branch: %s\n", __func__, f_rc_chsw[chsw_branches]);
-			nvgpu_rc_pbdma_fault(g, 0U, NVGPU_ERR_NOTIFIER_PBDMA_ERROR, &info);
-			continue;
+	for (i = 0; i < ARRAY_SIZE(notifiers); i++) {
+		err = nvgpu_rc_pbdma_fault(g, 0U, notifiers[i], &info);
+		if (err != (i < 2 ? 0 : -EINVAL)) {
+			unit_err(m, "fault processing error with notifier %d", notifiers[i]);
+			err = UNIT_FAIL;
+			goto out;
 		}
+	}
 
-		for (chsw_subbranch = 0U; chsw_subbranch < 2U; chsw_subbranch++) {
+	for (chsw_branches = F_RC_IS_CHSW_VALID_OR_SAVE;
+		chsw_branches <= F_RC_IS_CHSW_LOAD_OR_SWITCH; chsw_branches++) {
+
+		for (chsw_subbranch = 0U; chsw_subbranch <= chsw_branches; chsw_subbranch++) {
 			if (chsw_branches == F_RC_IS_CHSW_VALID_OR_SAVE) {
 				info.chsw_status =
 					(chsw_subbranch * NVGPU_PBDMA_CHSW_STATUS_VALID) +
@@ -411,7 +438,7 @@ int test_rc_pbdma_fault(struct unit_module *m, struct gk20a *g, void *args)
 			}
 		}
 
-		for (id_type_branches = F_RC_ID_TYPE_TSG; id_type_branches <= F_RC_ID_TYPE_INVALID;
+		for (id_type_branches = F_RC_ID_TYPE_TSG; id_type_branches <= F_RC_ID_TYPE_INVALID_MAX;
 				id_type_branches++) {
 			u32 id_type_ch_sub_branches = 0U;
 			if (id_type_branches == F_RC_ID_TYPE_CH) {
@@ -425,27 +452,81 @@ int test_rc_pbdma_fault(struct unit_module *m, struct gk20a *g, void *args)
 						f_rc_id_type[id_type_branches],
 						f_rc_id_ch_subbranch[id_type_ch_sub_branches]);
 
-					nvgpu_rc_pbdma_fault(g, 0U, NVGPU_ERR_NOTIFIER_PBDMA_ERROR, &info);
+					err = nvgpu_rc_pbdma_fault(g, 0U, NVGPU_ERR_NOTIFIER_PBDMA_ERROR, &info);
+					if ((id_type_branches >= F_RC_ID_TYPE_INVALID_MIN) ||
+					    (id_type_ch_sub_branches < F_RC_ID_TYPE_CH_FULL)) {
+						if (err != -EINVAL) {
+							unit_err(m, "invalid id type or null ch/tsg passed");
+							err = UNIT_FAIL;
+							goto out;
+						}
+					} else if (err != 0) {
+						unit_err(m, "valid id type with full ch failed");
+						err = UNIT_FAIL;
+						goto out;
+					}
 				}
 			} else {
 				set_pbdma_info_id_type(chsw_branches, &info, ch_without_tsg,
 					id_type_branches, id_type_ch_sub_branches);
 
-
 				unit_info(m, "%s branch: %s - %s\n", __func__,
 					f_rc_chsw[chsw_branches],
 					f_rc_id_type[id_type_branches]);
 
-				nvgpu_rc_pbdma_fault(g, 0U, NVGPU_ERR_NOTIFIER_PBDMA_ERROR, &info);
+				err = nvgpu_rc_pbdma_fault(g, 0U, NVGPU_ERR_NOTIFIER_PBDMA_ERROR, &info);
+				if (id_type_branches >= F_RC_ID_TYPE_INVALID_MIN) {
+					if (err != -EINVAL) {
+						unit_err(m, "invalid id type passed");
+						err = UNIT_FAIL;
+						goto out;
+					}
+				} else if (err != 0) {
+					unit_err(m, "valid id type with tsg failed");
+					err = UNIT_FAIL;
+					goto out;
+				}
 			}
 		}
 	}
 
+	for (chsw_branches = F_RC_IS_CHSW_INVALID;
+		chsw_branches <= F_RC_IS_CHSW_INVALID_STATE_MAX; chsw_branches++) {
+
+		if (chsw_branches == F_RC_IS_CHSW_INVALID) {
+			info.chsw_status = NVGPU_PBDMA_CHSW_STATUS_INVALID;
+		}
+
+		if (chsw_branches == F_RC_IS_CHSW_INVALID_STATE_MIN) {
+			info.chsw_status = NVGPU_PBDMA_CHSW_STATUS_SWITCH + 1;
+		}
+
+		if (chsw_branches == F_RC_IS_CHSW_INVALID_STATE_RANDOM) {
+			info.chsw_status = NVGPU_PBDMA_CHSW_STATUS_SWITCH + 2 +
+						get_random_u32(NVGPU_PBDMA_CHSW_STATUS_SWITCH + 1, INT_MAX);
+		}
+
+		if (chsw_branches == F_RC_IS_CHSW_INVALID_STATE_MAX) {
+			info.chsw_status = INT_MAX;
+		}
+
+		unit_info(m, "%s branch: %s\n", __func__, f_rc_chsw[chsw_branches]);
+		err = nvgpu_rc_pbdma_fault(g, 0U, NVGPU_ERR_NOTIFIER_PBDMA_ERROR, &info);
+		if (err != ((chsw_branches == F_RC_IS_CHSW_INVALID) ? 0 : -EINVAL)) {
+			unit_err(m, "pbdma status check failed");
+			err = UNIT_FAIL;
+			goto out;
+		}
+	}
+
+	err = UNIT_SUCCESS;
+
+out:
 	g->sw_quiesce_pending = false;
 
 	nvgpu_channel_close(ch_without_tsg);
 
-	return UNIT_SUCCESS;
+	return err;
 }
 
 struct unit_module_test nvgpu_rc_tests[] = {
diff --git a/userspace/units/rc/nvgpu-rc.h b/userspace/units/rc/nvgpu-rc.h
index f53326c36..3b5b10803 100644
--- a/userspace/units/rc/nvgpu-rc.h
+++ b/userspace/units/rc/nvgpu-rc.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2019, NVIDIA CORPORATION.  All rights reserved.
+ * Copyright (c) 2019-2022, NVIDIA CORPORATION.  All rights reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -249,33 +249,49 @@ int test_rc_mmu_fault(struct unit_module *m, struct gk20a *g, void *args);
  *
  * Description: Coverage test for nvgpu_rc_pbdma_fault
  *
- * Test Type: Feature
+ * Test Type: Feature, Boundary Value
  *
  * Targets: nvgpu_rc_pbdma_fault
  *
  * Input: test_rc_init run for this GPU
  *
+ * Equivalence classes:
+ * Variable: error_notifier
+ * - Valid: [NVGPU_ERR_NOTIFIER_FIFO_ERROR_IDLE_TIMEOUT, NVGPU_ERR_NOTIFIER_PBDMA_PUSHBUFFER_CRC_MISMATCH]
+ * - Invalid: [NVGPU_ERR_NOTIFIER_INVAL, INT_MAX]
+ * Variable: chsw state
+ * - Valid: [NVGPU_PBDMA_CHSW_STATUS_INVALID, NVGPU_PBDMA_CHSW_STATUS_SWITCH]
+ * - Invalid: [NVGPU_PBDMA_CHSW_STATUS_SWITCH + 1, INT_MAX]
+ * Variable: id_type
+ * - Valid: [PBDMA_STATUS_ID_TYPE_CHID, PBDMA_STATUS_ID_TYPE_TSGID]
+ * - Invalid: [PBDMA_STATUS_ID_TYPE_TSGID + 1, PBDMA_STATUS_ID_TYPE_INVALID]
+ *
  * Steps:
  * - initialize Channel error_notifier
+ * - test with valid and invalid error notifier values types
  * - set g->sw_quiesce_pending = true
  * - For each branch check with the following pbdma_status values
  * - set chsw_status to chsw_valid_or_save
  *   - set id_type to TSG
  *   - set id_type to Channel
  *     - set Channel Id to Invalid
- *	   - set Channel Id to a channel without TSG
- *	   - set Channel Id to a channel with a valid TSG
- *	 - set id_type to Invalid
+ *     - set Channel Id to a channel without TSG
+ *     - set Channel Id to a channel with a valid TSG
+ *     - set id_type to chid, tsgid, tsgid + 1, tsgid + 1 + random, invalid_id
+ *   - verify that nvgpu_rc_pbdma_fault fails for invalid id_types and invalid channel ids and succeeds otherwise.
  * - set chsw_status to is_chsw_load_or_switch
  *   - set id_type to TSG
  *   - set id_type to Channel
  *     - set Channel Id to Invalid
- *	   - set Channel Id to a channel without TSG
- *	   - set Channel Id to a channel with a valid TSG
- *	 - set id_type to Invalid
- * - set chsw_status to chsw_invalid
+ *     - set Channel Id to a channel without TSG
+ *     - set Channel Id to a channel with a valid TSG
+ *     - set id_type to chid, tsgid, tsgid + 1, tsgid + 1 + random, invalid_id
+ *   - verify that nvgpu_rc_pbdma_fault fails for invalid id_types and invalid channel ids and succeeds otherwise.
+ * - set chsw_status to chsw_invalid and verify that nvgpu_rc_pbdma_fault succeeds.
+ * - set chsw_status to invalid states and verify that nvgpu_rc_pbdma_fault fails.
  *
- * Output: Cover all branch in safety build.
+ * Output: Returns PASS if nvgpu_rc_pbdma_fault succeeds for valid inputs
+ *         and fails for invalid inputs. Returns FAIL otherwise.
  */
 int test_rc_pbdma_fault(struct unit_module *m, struct gk20a *g, void *args);