From 143034daab1d90fe9a996e914cfd5325424ed43d Mon Sep 17 00:00:00 2001
From: Debarshi Dutta <ddutta@nvidia.com>
Date: Wed, 24 Aug 2022 16:58:05 +0530
Subject: [PATCH] gpu: nvgpu: modify wait_pending

The wait_pending HAL is now modified to simply
check the pending status of a given runlist.
The while loop is removed from this HAL.

A new function nvgpu_runlist_wait_pending_legacy() is
added that emulates the older wait_pending() HAL.

nvgpu_runlist_tick() is modified to accept a 64 bit
"preempt_grace_ns" value.

These changes prepare for upcoming control-fifo parser
changes.

Jira NVGPU-8619

Signed-off-by: Debarshi Dutta <ddutta@nvidia.com>
Change-Id: If3f288eb6f2181743c53b657219b3b30d56d26bc
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2766100
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
---
 drivers/gpu/nvgpu/common/fifo/runlist.c       | 98 +++++++++++--------
 drivers/gpu/nvgpu/common/nvs/nvs_sched.c      |  2 +-
 .../gpu/nvgpu/hal/fifo/runlist_fifo_ga10b.h   |  2 +-
 .../nvgpu/hal/fifo/runlist_fifo_ga10b_fusa.c  | 21 +---
 .../gpu/nvgpu/hal/fifo/runlist_fifo_gk20a.h   |  4 +-
 .../nvgpu/hal/fifo/runlist_fifo_gk20a_fusa.c  | 26 +----
 .../gpu/nvgpu/hal/fifo/runlist_fifo_tu104.c   | 25 ++---
 .../gpu/nvgpu/hal/fifo/runlist_fifo_tu104.h   |  4 +-
 .../gpu/nvgpu/hal/fifo/runlist_ga10b_fusa.c   |  2 +-
 drivers/gpu/nvgpu/hal/init/hal_ga100.c        |  2 +-
 drivers/gpu/nvgpu/hal/init/hal_ga10b.c        |  2 +-
 drivers/gpu/nvgpu/hal/init/hal_gm20b.c        |  2 +-
 drivers/gpu/nvgpu/hal/init/hal_gv11b.c        |  2 +-
 drivers/gpu/nvgpu/hal/init/hal_tu104.c        |  2 +-
 .../gpu/nvgpu/hal/vgpu/init/vgpu_hal_ga10b.c  |  2 +-
 .../gpu/nvgpu/hal/vgpu/init/vgpu_hal_gv11b.c  |  2 +-
 .../gpu/nvgpu/include/nvgpu/gops/runlist.h    |  2 +-
 drivers/gpu/nvgpu/include/nvgpu/runlist.h     | 17 +++-
 libs/dgpu/libnvgpu-drv-dgpu_safe.export       |  3 +-
 libs/igpu/libnvgpu-drv-igpu_safe.export       |  3 +-
 userspace/required_tests.ini                  |  2 +-
 .../fifo/runlist/gk20a/nvgpu-runlist-gk20a.c  | 10 +-
 .../fifo/runlist/gk20a/nvgpu-runlist-gk20a.h  | 14 +--
 23 files changed, 122 insertions(+), 127 deletions(-)

diff --git a/drivers/gpu/nvgpu/common/fifo/runlist.c b/drivers/gpu/nvgpu/common/fifo/runlist.c
index 07922ae83..bd8bd52e0 100644
--- a/drivers/gpu/nvgpu/common/fifo/runlist.c
+++ b/drivers/gpu/nvgpu/common/fifo/runlist.c
@@ -453,11 +453,31 @@ void nvgpu_runlist_swap_mem(struct gk20a *g, struct nvgpu_runlist_domain *domain
 	nvgpu_spinlock_release(&domain->lock);
 }
 
-static int nvgpu_runlist_domain_actual_submit(struct gk20a *g, struct nvgpu_runlist *rl,
-		bool wait_for_finish)
+int nvgpu_runlist_wait_pending_legacy(struct gk20a *g, struct nvgpu_runlist *rl)
 {
-	int ret = 0;
+	struct nvgpu_timeout timeout;
+	u32 delay = POLL_DELAY_MIN_US;
+	int ret;
 
+	nvgpu_timeout_init_cpu_timer(g, &timeout, nvgpu_get_poll_timeout(g));
+
+	ret = -ETIMEDOUT;
+	do {
+		if (g->ops.runlist.check_pending(g, rl) == 0) {
+			ret = 0;
+			break;
+		}
+
+		/* Avoid sleeping */
+		nvgpu_usleep_range(delay, delay * 2U);
+		delay = min_t(u32, delay << 1, POLL_DELAY_MAX_US);
+	} while (nvgpu_timeout_expired(&timeout) == 0);
+
+	return ret;
+}
+
+static void nvgpu_runlist_domain_actual_submit(struct gk20a *g, struct nvgpu_runlist *rl)
+{
 	rl_dbg(g, "Runlist[%u]: submitting domain[%llu]",
 		rl->id, rl->domain->domain_id);
 
@@ -467,18 +487,6 @@ static int nvgpu_runlist_domain_actual_submit(struct gk20a *g, struct nvgpu_runl
 	nvgpu_spinlock_acquire(&rl->domain->lock);
 	g->ops.runlist.hw_submit(g, rl);
 	nvgpu_spinlock_release(&rl->domain->lock);
-
-	if (wait_for_finish) {
-		ret = g->ops.runlist.wait_pending(g, rl);
-		if (ret == -ETIMEDOUT) {
-			nvgpu_err(g, "runlist %d update timeout", rl->id);
-			/* trigger runlist update timeout recovery */
-			return ret;
-
-		}
-	}
-
-	return ret;
 }
 
 static int nvgpu_runlist_update_mem_locked(struct gk20a *g, struct nvgpu_runlist *rl,
@@ -594,7 +602,7 @@ int nvgpu_runlist_reschedule(struct nvgpu_channel *ch, bool preempt_next,
 		}
 	}
 
-	if (g->ops.runlist.wait_pending(g, runlist) != 0) {
+	if (nvgpu_runlist_wait_pending_legacy(g, runlist) != 0) {
 		nvgpu_err(g, "wait pending failed for runlist %u",
 				runlist->id);
 	}
@@ -668,19 +676,15 @@ static int nvgpu_runlist_do_update(struct gk20a *g, struct nvgpu_runlist *rl,
 /*
  * This is expected to be called only when device is powered on.
  */
-static int runlist_submit_powered(struct gk20a *g, struct nvgpu_runlist *runlist,
-		struct nvgpu_runlist_domain *next_domain, bool wait_for_finish)
+static void runlist_submit_powered(struct gk20a *g, struct nvgpu_runlist *runlist,
+		struct nvgpu_runlist_domain *next_domain)
 {
-	int err;
-
 	runlist->domain = next_domain;
 
 	rl_dbg(g, "Runlist[%u]: switching to domain %llu",
 		runlist->id, next_domain->domain_id);
 
-	err = nvgpu_runlist_domain_actual_submit(g, runlist, wait_for_finish);
-
-	return err;
+	nvgpu_runlist_domain_actual_submit(g, runlist);
 }
 
 int nvgpu_rl_domain_sync_submit(struct gk20a *g, struct nvgpu_runlist *runlist,
@@ -692,26 +696,28 @@ int nvgpu_rl_domain_sync_submit(struct gk20a *g, struct nvgpu_runlist *runlist,
 		next_domain = runlist->shadow_rl_domain;
 	}
 
-	err = runlist_submit_powered(g, runlist, next_domain, wait_for_finish);
+	runlist_submit_powered(g, runlist, next_domain);
+	if (wait_for_finish) {
+		err = nvgpu_runlist_wait_pending_legacy(g, runlist);
+		if (err != 0) {
+			nvgpu_err(g, "runlist %d update timeout", runlist->id);
+			/* trigger runlist update timeout recovery */
+			return err;
+		}
+	}
 
 	return err;
 }
 
-static int runlist_switch_domain_and_submit(struct gk20a *g,
-		struct nvgpu_runlist *runlist, struct nvgpu_runlist_domain *rl_domain)
-{
-	int ret = 0;
-
-	ret = runlist_submit_powered(g, runlist, rl_domain, false);
-
-	return ret;
-}
-
-void nvgpu_runlist_tick(struct gk20a *g, struct nvgpu_runlist_domain **rl_domain)
+int nvgpu_runlist_tick(struct gk20a *g, struct nvgpu_runlist_domain **rl_domain,
+	u64 preempt_grace_ns)
 {
 	struct nvgpu_fifo *f = &g->fifo;
 	u32 i;
-	int err = 0;
+	int ret = 0;
+	int err = -ETIMEDOUT;
+	u64 start_time = nvgpu_safe_cast_s64_to_u64(nvgpu_current_time_ns());
+	u64 current_time;
 
 	rl_dbg(g, "domain tick");
 
@@ -719,11 +725,23 @@ void nvgpu_runlist_tick(struct gk20a *g, struct nvgpu_runlist_domain **rl_domain
 		struct nvgpu_runlist *runlist;
 
 		runlist = &f->active_runlists[i];
-		err = runlist_switch_domain_and_submit(g, runlist, rl_domain[i]);
-		if (err != 0) {
-			nvgpu_err(g, "Failed to schedule domain [%llu]", rl_domain[i]->domain_id);
-		}
+		runlist_submit_powered(g, runlist, rl_domain[i]);
+
+		do {
+			ret = g->ops.runlist.check_pending(g, runlist);
+			if (ret == 0) {
+				break;
+			}
+			current_time = nvgpu_safe_cast_s64_to_u64(nvgpu_current_time_ns());
+		} while ((preempt_grace_ns == 0ULL)
+			|| (nvgpu_safe_sub_u64(current_time, start_time) <= preempt_grace_ns));
 	}
+
+	if (i == f->num_runlists) {
+		err = 0;
+	}
+
+	return err;
 }
 
 int nvgpu_runlist_update(struct gk20a *g, struct nvgpu_runlist *rl,
diff --git a/drivers/gpu/nvgpu/common/nvs/nvs_sched.c b/drivers/gpu/nvgpu/common/nvs/nvs_sched.c
index cee5b3afe..e35dba184 100644
--- a/drivers/gpu/nvgpu/common/nvs/nvs_sched.c
+++ b/drivers/gpu/nvgpu/common/nvs/nvs_sched.c
@@ -144,7 +144,7 @@ static u64 nvgpu_nvs_tick(struct gk20a *g)
 	timeslice = nvs_next->timeslice_ns;
 	nvgpu_domain_next = nvs_next->priv;
 
-	nvgpu_runlist_tick(g, nvgpu_domain_next->rl_domains);
+	(void)nvgpu_runlist_tick(g, nvgpu_domain_next->rl_domains, 0ULL);
 	sched->active_domain = nvs_next->priv;
 
 	nvgpu_mutex_release(&g->sched_mutex);
diff --git a/drivers/gpu/nvgpu/hal/fifo/runlist_fifo_ga10b.h b/drivers/gpu/nvgpu/hal/fifo/runlist_fifo_ga10b.h
index 7020facfa..1dab63b8e 100644
--- a/drivers/gpu/nvgpu/hal/fifo/runlist_fifo_ga10b.h
+++ b/drivers/gpu/nvgpu/hal/fifo/runlist_fifo_ga10b.h
@@ -33,7 +33,7 @@ struct nvgpu_runlist;
 u32 ga10b_runlist_count_max(struct gk20a *g);
 u32 ga10b_runlist_length_max(struct gk20a *g);
 void ga10b_runlist_hw_submit(struct gk20a *g, struct nvgpu_runlist *runlist);
-int ga10b_runlist_wait_pending(struct gk20a *g, struct nvgpu_runlist *runlist);
+int ga10b_runlist_check_pending(struct gk20a *g, struct nvgpu_runlist *runlist);
 void ga10b_runlist_write_state(struct gk20a *g, u32 runlists_mask,
 				u32 runlist_state);
 u32 ga10b_get_runlist_aperture(struct gk20a *g, struct nvgpu_runlist *runlist);
diff --git a/drivers/gpu/nvgpu/hal/fifo/runlist_fifo_ga10b_fusa.c b/drivers/gpu/nvgpu/hal/fifo/runlist_fifo_ga10b_fusa.c
index c2a8be3d5..0fed0ea5b 100644
--- a/drivers/gpu/nvgpu/hal/fifo/runlist_fifo_ga10b_fusa.c
+++ b/drivers/gpu/nvgpu/hal/fifo/runlist_fifo_ga10b_fusa.c
@@ -81,25 +81,14 @@ void ga10b_runlist_hw_submit(struct gk20a *g, struct nvgpu_runlist *runlist)
 			runlist_submit_length_f(runlist->domain->mem_hw->count));
 }
 
-int ga10b_runlist_wait_pending(struct gk20a *g, struct nvgpu_runlist *runlist)
+int ga10b_runlist_check_pending(struct gk20a *g, struct nvgpu_runlist *runlist)
 {
-	struct nvgpu_timeout timeout;
-	u32 delay = POLL_DELAY_MIN_US;
-	int ret;
+	int ret = 1;
 
-	nvgpu_timeout_init_cpu_timer(g, &timeout, nvgpu_get_poll_timeout(g));
-
-	ret = -ETIMEDOUT;
-	do {
-		if ((nvgpu_runlist_readl(g, runlist, runlist_submit_info_r()) &
+	if ((nvgpu_runlist_readl(g, runlist, runlist_submit_info_r()) &
 			runlist_submit_info_pending_true_f()) == 0U) {
-			ret = 0;
-			break;
-		}
-
-		nvgpu_usleep_range(delay, delay * 2U);
-		delay = min_t(u32, delay << 1, POLL_DELAY_MAX_US);
-	} while (nvgpu_timeout_expired(&timeout) == 0);
+		ret = 0;
+	}
 
 	return ret;
 }
diff --git a/drivers/gpu/nvgpu/hal/fifo/runlist_fifo_gk20a.h b/drivers/gpu/nvgpu/hal/fifo/runlist_fifo_gk20a.h
index 5fe05f922..17237cea7 100644
--- a/drivers/gpu/nvgpu/hal/fifo/runlist_fifo_gk20a.h
+++ b/drivers/gpu/nvgpu/hal/fifo/runlist_fifo_gk20a.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2011-2021, NVIDIA CORPORATION.  All rights reserved.
+ * Copyright (c) 2011-2022, 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 @@ u32 gk20a_runlist_count_max(struct gk20a *g);
 
 u32 gk20a_runlist_length_max(struct gk20a *g);
 void gk20a_runlist_hw_submit(struct gk20a *g, struct nvgpu_runlist *runlist);
-int gk20a_runlist_wait_pending(struct gk20a *g, struct nvgpu_runlist *runlist);
+int gk20a_runlist_check_pending(struct gk20a *g, struct nvgpu_runlist *runlist);
 void gk20a_runlist_write_state(struct gk20a *g, u32 runlists_mask,
 		u32 runlist_state);
 
diff --git a/drivers/gpu/nvgpu/hal/fifo/runlist_fifo_gk20a_fusa.c b/drivers/gpu/nvgpu/hal/fifo/runlist_fifo_gk20a_fusa.c
index a92ce6fbb..ff4d242eb 100644
--- a/drivers/gpu/nvgpu/hal/fifo/runlist_fifo_gk20a_fusa.c
+++ b/drivers/gpu/nvgpu/hal/fifo/runlist_fifo_gk20a_fusa.c
@@ -61,29 +61,13 @@ void gk20a_runlist_hw_submit(struct gk20a *g, struct nvgpu_runlist *runlist)
        nvgpu_spinlock_release(&g->fifo.runlist_submit_lock);
 }
 
-int gk20a_runlist_wait_pending(struct gk20a *g, struct nvgpu_runlist *runlist)
+int gk20a_runlist_check_pending(struct gk20a *g, struct nvgpu_runlist *runlist)
 {
-	struct nvgpu_timeout timeout;
-	u32 delay = POLL_DELAY_MIN_US;
-	int ret = 0;
+	int ret = 1;
 
-	nvgpu_timeout_init_cpu_timer(g, &timeout, nvgpu_get_poll_timeout(g));
-
-	ret = -ETIMEDOUT;
-	do {
-		if ((nvgpu_readl(g, fifo_eng_runlist_r(runlist->id)) &
-				fifo_eng_runlist_pending_true_f()) == 0U) {
-			ret = 0;
-			break;
-		}
-
-		nvgpu_usleep_range(delay, delay * 2U);
-		delay = min_t(u32, delay << 1U, POLL_DELAY_MAX_US);
-	} while (nvgpu_timeout_expired(&timeout) == 0);
-
-	if (ret != 0) {
-		nvgpu_err(g, "runlist wait timeout: runlist id: %u",
-			runlist->id);
+	if ((nvgpu_readl(g, fifo_eng_runlist_r(runlist->id)) &
+			fifo_eng_runlist_pending_true_f()) == 0U) {
+		ret = 0;
 	}
 
 	return ret;
diff --git a/drivers/gpu/nvgpu/hal/fifo/runlist_fifo_tu104.c b/drivers/gpu/nvgpu/hal/fifo/runlist_fifo_tu104.c
index 827bd6045..f9fb5a376 100644
--- a/drivers/gpu/nvgpu/hal/fifo/runlist_fifo_tu104.c
+++ b/drivers/gpu/nvgpu/hal/fifo/runlist_fifo_tu104.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2018-2021, 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"),
@@ -61,25 +61,14 @@ void tu104_runlist_hw_submit(struct gk20a *g, struct nvgpu_runlist *runlist)
 		fifo_runlist_submit_length_f(runlist->domain->mem_hw->count));
 }
 
-int tu104_runlist_wait_pending(struct gk20a *g, struct nvgpu_runlist *runlist)
+int tu104_runlist_check_pending(struct gk20a *g, struct nvgpu_runlist *runlist)
 {
-	struct nvgpu_timeout timeout;
-	u32 delay = POLL_DELAY_MIN_US;
-	int ret;
+	int ret = 1;
 
-	nvgpu_timeout_init_cpu_timer(g, &timeout, nvgpu_get_poll_timeout(g));
-
-	ret = -ETIMEDOUT;
-	do {
-		if ((nvgpu_readl(g, fifo_runlist_submit_info_r(runlist->id)) &
-			fifo_runlist_submit_info_pending_true_f()) == 0U) {
-			ret = 0;
-			break;
-		}
-
-		nvgpu_usleep_range(delay, delay * 2U);
-		delay = min_t(u32, delay << 1, POLL_DELAY_MAX_US);
-	} while (nvgpu_timeout_expired(&timeout) == 0);
+	if ((nvgpu_readl(g, fifo_runlist_submit_info_r(runlist->id)) &
+		fifo_runlist_submit_info_pending_true_f()) == 0U) {
+		ret = 0;
+	}
 
 	return ret;
 }
diff --git a/drivers/gpu/nvgpu/hal/fifo/runlist_fifo_tu104.h b/drivers/gpu/nvgpu/hal/fifo/runlist_fifo_tu104.h
index 0e127bba7..ef2c2ecc8 100644
--- a/drivers/gpu/nvgpu/hal/fifo/runlist_fifo_tu104.h
+++ b/drivers/gpu/nvgpu/hal/fifo/runlist_fifo_tu104.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2018-2021, 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"),
@@ -30,6 +30,6 @@ struct nvgpu_runlist;
 
 u32 tu104_runlist_count_max(struct gk20a *g);
 void tu104_runlist_hw_submit(struct gk20a *g, struct nvgpu_runlist *runlist);
-int tu104_runlist_wait_pending(struct gk20a *g, struct nvgpu_runlist *runlist);
+int tu104_runlist_check_pending(struct gk20a *g, struct nvgpu_runlist *runlist);
 
 #endif /* NVGPU_RUNLIST_FIFO_TU104_H */
diff --git a/drivers/gpu/nvgpu/hal/fifo/runlist_ga10b_fusa.c b/drivers/gpu/nvgpu/hal/fifo/runlist_ga10b_fusa.c
index 0048d504d..f28ebad51 100644
--- a/drivers/gpu/nvgpu/hal/fifo/runlist_ga10b_fusa.c
+++ b/drivers/gpu/nvgpu/hal/fifo/runlist_ga10b_fusa.c
@@ -1,7 +1,7 @@
 /*
  * GA10B runlist
  *
- * 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"),
diff --git a/drivers/gpu/nvgpu/hal/init/hal_ga100.c b/drivers/gpu/nvgpu/hal/init/hal_ga100.c
index d97b4d615..34bdb2071 100644
--- a/drivers/gpu/nvgpu/hal/init/hal_ga100.c
+++ b/drivers/gpu/nvgpu/hal/init/hal_ga100.c
@@ -1151,7 +1151,7 @@ static const struct gops_runlist ga100_ops_runlist = {
 	.get_tsg_entry = gv11b_runlist_get_tsg_entry,
 	.get_ch_entry = gv11b_runlist_get_ch_entry,
 	.hw_submit = ga100_runlist_hw_submit,
-	.wait_pending = ga10b_runlist_wait_pending,
+	.check_pending = ga10b_runlist_check_pending,
 	.write_state = ga10b_runlist_write_state,
 	.get_runlist_id = ga10b_runlist_get_runlist_id,
 	.get_engine_id_from_rleng_id = ga10b_runlist_get_engine_id_from_rleng_id,
diff --git a/drivers/gpu/nvgpu/hal/init/hal_ga10b.c b/drivers/gpu/nvgpu/hal/init/hal_ga10b.c
index a0053c703..2342f5edb 100644
--- a/drivers/gpu/nvgpu/hal/init/hal_ga10b.c
+++ b/drivers/gpu/nvgpu/hal/init/hal_ga10b.c
@@ -1165,7 +1165,7 @@ static const struct gops_runlist ga10b_ops_runlist = {
 	.get_tsg_entry = gv11b_runlist_get_tsg_entry,
 	.get_ch_entry = gv11b_runlist_get_ch_entry,
 	.hw_submit = ga10b_runlist_hw_submit,
-	.wait_pending = ga10b_runlist_wait_pending,
+	.check_pending = ga10b_runlist_check_pending,
 	.write_state = ga10b_runlist_write_state,
 	.get_runlist_id = ga10b_runlist_get_runlist_id,
 	.get_runlist_aperture = ga10b_get_runlist_aperture,
diff --git a/drivers/gpu/nvgpu/hal/init/hal_gm20b.c b/drivers/gpu/nvgpu/hal/init/hal_gm20b.c
index f68ea7577..35604bc97 100644
--- a/drivers/gpu/nvgpu/hal/init/hal_gm20b.c
+++ b/drivers/gpu/nvgpu/hal/init/hal_gm20b.c
@@ -710,7 +710,7 @@ static const struct gops_runlist gm20b_ops_runlist = {
 	.get_tsg_entry = gk20a_runlist_get_tsg_entry,
 	.get_ch_entry = gk20a_runlist_get_ch_entry,
 	.hw_submit = gk20a_runlist_hw_submit,
-	.wait_pending = gk20a_runlist_wait_pending,
+	.check_pending = gk20a_runlist_check_pending,
 	.write_state = gk20a_runlist_write_state,
 	.init_enginfo = nvgpu_runlist_init_enginfo,
 	.get_tsg_max_timeslice = gk20a_runlist_max_timeslice,
diff --git a/drivers/gpu/nvgpu/hal/init/hal_gv11b.c b/drivers/gpu/nvgpu/hal/init/hal_gv11b.c
index cf6c1cc1b..e691fad13 100644
--- a/drivers/gpu/nvgpu/hal/init/hal_gv11b.c
+++ b/drivers/gpu/nvgpu/hal/init/hal_gv11b.c
@@ -989,7 +989,7 @@ static const struct gops_runlist gv11b_ops_runlist = {
 	.get_tsg_entry = gv11b_runlist_get_tsg_entry,
 	.get_ch_entry = gv11b_runlist_get_ch_entry,
 	.hw_submit = gk20a_runlist_hw_submit,
-	.wait_pending = gk20a_runlist_wait_pending,
+	.check_pending = gk20a_runlist_check_pending,
 	.write_state = gk20a_runlist_write_state,
 	.init_enginfo = nvgpu_runlist_init_enginfo,
 	.get_tsg_max_timeslice = gv11b_runlist_max_timeslice,
diff --git a/drivers/gpu/nvgpu/hal/init/hal_tu104.c b/drivers/gpu/nvgpu/hal/init/hal_tu104.c
index 3d1454bd6..74790477d 100644
--- a/drivers/gpu/nvgpu/hal/init/hal_tu104.c
+++ b/drivers/gpu/nvgpu/hal/init/hal_tu104.c
@@ -1044,7 +1044,7 @@ static const struct gops_runlist tu104_ops_runlist = {
 	.get_tsg_entry = gv11b_runlist_get_tsg_entry,
 	.get_ch_entry = gv11b_runlist_get_ch_entry,
 	.hw_submit = tu104_runlist_hw_submit,
-	.wait_pending = tu104_runlist_wait_pending,
+	.check_pending = tu104_runlist_check_pending,
 	.write_state = gk20a_runlist_write_state,
 	.init_enginfo = nvgpu_runlist_init_enginfo,
 	.get_tsg_max_timeslice = gv11b_runlist_max_timeslice,
diff --git a/drivers/gpu/nvgpu/hal/vgpu/init/vgpu_hal_ga10b.c b/drivers/gpu/nvgpu/hal/vgpu/init/vgpu_hal_ga10b.c
index 7ecb38a68..dc3244ca5 100644
--- a/drivers/gpu/nvgpu/hal/vgpu/init/vgpu_hal_ga10b.c
+++ b/drivers/gpu/nvgpu/hal/vgpu/init/vgpu_hal_ga10b.c
@@ -710,7 +710,7 @@ static const struct gops_runlist vgpu_ga10b_ops_runlist = {
 	.get_tsg_entry = gv11b_runlist_get_tsg_entry,
 	.get_ch_entry = gv11b_runlist_get_ch_entry,
 	.hw_submit = NULL,
-	.wait_pending = NULL,
+	.check_pending = NULL,
 	.init_enginfo = nvgpu_runlist_init_enginfo,
 	.get_max_channels_per_tsg = gv11b_runlist_get_max_channels_per_tsg,
 };
diff --git a/drivers/gpu/nvgpu/hal/vgpu/init/vgpu_hal_gv11b.c b/drivers/gpu/nvgpu/hal/vgpu/init/vgpu_hal_gv11b.c
index 0db32ef25..770f5e7e6 100644
--- a/drivers/gpu/nvgpu/hal/vgpu/init/vgpu_hal_gv11b.c
+++ b/drivers/gpu/nvgpu/hal/vgpu/init/vgpu_hal_gv11b.c
@@ -677,7 +677,7 @@ static const struct gops_runlist vgpu_gv11b_ops_runlist = {
 	.get_tsg_entry = gv11b_runlist_get_tsg_entry,
 	.get_ch_entry = gv11b_runlist_get_ch_entry,
 	.hw_submit = NULL,
-	.wait_pending = NULL,
+	.check_pending = NULL,
 	.init_enginfo = nvgpu_runlist_init_enginfo,
 	.get_tsg_max_timeslice = gv11b_runlist_max_timeslice,
 	.get_max_channels_per_tsg = gv11b_runlist_get_max_channels_per_tsg,
diff --git a/drivers/gpu/nvgpu/include/nvgpu/gops/runlist.h b/drivers/gpu/nvgpu/include/nvgpu/gops/runlist.h
index 6ab1d580b..ba821186e 100644
--- a/drivers/gpu/nvgpu/include/nvgpu/gops/runlist.h
+++ b/drivers/gpu/nvgpu/include/nvgpu/gops/runlist.h
@@ -93,7 +93,7 @@ struct gops_runlist {
 			u32 *runlist, u32 timeslice);
 	void (*get_ch_entry)(struct nvgpu_channel *ch, u32 *runlist);
 	void (*hw_submit)(struct gk20a *g, struct nvgpu_runlist *runlist);
-	int (*wait_pending)(struct gk20a *g, struct nvgpu_runlist *runlist);
+	int (*check_pending)(struct gk20a *g, struct nvgpu_runlist *runlist);
 	void (*write_state)(struct gk20a *g, u32 runlists_mask,
 			u32 runlist_state);
 	int (*reschedule)(struct nvgpu_channel *ch, bool preempt_next);
diff --git a/drivers/gpu/nvgpu/include/nvgpu/runlist.h b/drivers/gpu/nvgpu/include/nvgpu/runlist.h
index 7a6f79bd3..042e02a56 100644
--- a/drivers/gpu/nvgpu/include/nvgpu/runlist.h
+++ b/drivers/gpu/nvgpu/include/nvgpu/runlist.h
@@ -198,7 +198,7 @@ struct nvgpu_runlist {
 	const struct nvgpu_device *rl_dev_list[RLENG_PER_RUNLIST_SIZE];
 	/** @endcond DOXYGEN_SHOULD_SKIP_THIS */
 };
-
+int nvgpu_runlist_wait_pending_legacy(struct gk20a *g, struct nvgpu_runlist *rl);
 bool nvgpu_rl_domain_exists(struct gk20a *g, const char *name);
 struct nvgpu_runlist_domain *nvgpu_runlist_domain_alloc(struct gk20a *g,
 		u64 domain_id);
@@ -233,7 +233,20 @@ nvgpu_runlist_domain_from_domains_list(struct nvgpu_list_node *node)
 	((uintptr_t)node - offsetof(struct nvgpu_runlist_domain, domains_list));
 }
 
-void nvgpu_runlist_tick(struct gk20a *g, struct nvgpu_runlist_domain **rl_domain);
+/**
+ * @brief Submit the nvgpu_runlist_domain instance corresponding to a given domain.
+ *
+ * Attempts to switch all the rl_domains and wait appropriately. If preempt_grace_ns
+ * equals 0, the attempt continues infinitely, otherwise it would timeout if the
+ * total time exceeds preempt_grace_ns.
+ *
+ * @param g			Global gk20a struct
+ * @param rl_domain	> Actual array of rl domain thats meant to be scheduled.
+ * @param preempt_grace_ns	-> Max Grace Period for Domain switches.
+ * @return int 0 if domain switch was successful, else -ETIMEDOUT
+ */
+int nvgpu_runlist_tick(struct gk20a *g, struct nvgpu_runlist_domain **rl_domain,
+	u64 preempt_grace_ns);
 
 /**
  * @brief Rebuild runlist
diff --git a/libs/dgpu/libnvgpu-drv-dgpu_safe.export b/libs/dgpu/libnvgpu-drv-dgpu_safe.export
index fd477b3e7..3b38c32e7 100644
--- a/libs/dgpu/libnvgpu-drv-dgpu_safe.export
+++ b/libs/dgpu/libnvgpu-drv-dgpu_safe.export
@@ -28,7 +28,8 @@ gk20a_ramin_alloc_size
 gk20a_ramin_base_shift
 gk20a_runlist_length_max
 gk20a_runlist_hw_submit
-gk20a_runlist_wait_pending
+nvgpu_runlist_wait_pending_legacy
+gk20a_runlist_check_pending
 gk20a_runlist_write_state
 gk20a_userd_entry_size
 gk20a_vm_release_share
diff --git a/libs/igpu/libnvgpu-drv-igpu_safe.export b/libs/igpu/libnvgpu-drv-igpu_safe.export
index 2b1ba8408..48e6653ae 100644
--- a/libs/igpu/libnvgpu-drv-igpu_safe.export
+++ b/libs/igpu/libnvgpu-drv-igpu_safe.export
@@ -27,7 +27,8 @@ gk20a_ramin_alloc_size
 gk20a_ramin_base_shift
 gk20a_runlist_length_max
 gk20a_runlist_hw_submit
-gk20a_runlist_wait_pending
+nvgpu_runlist_wait_pending_legacy
+gk20a_runlist_check_pending
 gk20a_runlist_write_state
 gk20a_userd_entry_size
 gk20a_vm_release_share
diff --git a/userspace/required_tests.ini b/userspace/required_tests.ini
index 8c959bd9b..e8041e9d0 100644
--- a/userspace/required_tests.ini
+++ b/userspace/required_tests.ini
@@ -744,7 +744,7 @@ test_fifo_init_support.init_support=0
 test_fifo_remove_support.remove_support=0
 test_gk20a_runlist_hw_submit.hw_submit=0
 test_gk20a_runlist_length_max.length_max=0
-test_gk20a_runlist_wait_pending.wait_pending=0
+test_gk20a_runlist_wait_pending.check_pending=0
 test_gk20a_runlist_write_state.write_state=0
 
 [nvgpu_runlist_gv11b]
diff --git a/userspace/units/fifo/runlist/gk20a/nvgpu-runlist-gk20a.c b/userspace/units/fifo/runlist/gk20a/nvgpu-runlist-gk20a.c
index 7435cee70..44b17e5a5 100644
--- a/userspace/units/fifo/runlist/gk20a/nvgpu-runlist-gk20a.c
+++ b/userspace/units/fifo/runlist/gk20a/nvgpu-runlist-gk20a.c
@@ -177,22 +177,22 @@ int test_gk20a_runlist_wait_pending(struct unit_module *m,
 
 	/* no wait */
 	ctx->count = 0;
-	err = gk20a_runlist_wait_pending(g, runlist);
+	err = nvgpu_runlist_wait_pending_legacy(g, runlist);
 	unit_assert(err == 0, goto done);
 
 	/* 1 loop */
 	ctx->count = 1;
-	err = gk20a_runlist_wait_pending(g, runlist);
+	err = nvgpu_runlist_wait_pending_legacy(g, runlist);
 	unit_assert(err == 0, goto done);
 
 	/* 2 loops */
 	ctx->count = 2;
-	err = gk20a_runlist_wait_pending(g, runlist);
+	err = nvgpu_runlist_wait_pending_legacy(g, runlist);
 	unit_assert(err == 0, goto done);
 
 	/* timeout  */
 	ctx->count = U32_MAX;
-	err = gk20a_runlist_wait_pending(g, runlist);
+	err = nvgpu_runlist_wait_pending_legacy(g, runlist);
 	unit_assert(err == -ETIMEDOUT, goto done);
 
 	ret = UNIT_SUCCESS;
@@ -235,7 +235,7 @@ struct unit_module_test nvgpu_runlist_gk20a_tests[] = {
 	UNIT_TEST(init_support, test_fifo_init_support, NULL, 0),
 	UNIT_TEST(length_max, test_gk20a_runlist_length_max, NULL, 0),
 	UNIT_TEST(hw_submit, test_gk20a_runlist_hw_submit, NULL, 0),
-	UNIT_TEST(wait_pending, test_gk20a_runlist_wait_pending, NULL, 0),
+	UNIT_TEST(check_pending, test_gk20a_runlist_wait_pending, NULL, 0),
 	UNIT_TEST(write_state, test_gk20a_runlist_write_state, NULL, 0),
 	UNIT_TEST(remove_support, test_fifo_remove_support, NULL, 0),
 };
diff --git a/userspace/units/fifo/runlist/gk20a/nvgpu-runlist-gk20a.h b/userspace/units/fifo/runlist/gk20a/nvgpu-runlist-gk20a.h
index cf8c5b617..9ec7e6332 100644
--- a/userspace/units/fifo/runlist/gk20a/nvgpu-runlist-gk20a.h
+++ b/userspace/units/fifo/runlist/gk20a/nvgpu-runlist-gk20a.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"),
@@ -76,28 +76,28 @@ int test_gk20a_runlist_hw_submit(struct unit_module *m,
 /**
  * Test specification for: test_gk20a_runlist_wait_pending
  *
- * Description: Branch coverage for gk20a_runlist_wait_pending
+ * Description: Branch coverage for nvgpu_runlist_wait_pending_legacy
  *
  * Test Type: Feature
  *
- * Targets: gops_runlist.wait_pending, gk20a_runlist_wait_pending
+ * Targets: gops_runlist.check_pending, nvgpu_runlist_wait_pending_legacy
  *
  * Input: test_fifo_init_support() run for this GPU
  *
  * Steps:
  * - Check case where runlist is not pending (not wait).
  *   - Set register to indicate that runlist is NOT pending.
- *   - Call gk20a_runlist_wait_pending.
+ *   - Call nvgpu_runlist_wait_pending_legacy.
  * - Check case where some polling is needed until runlist is not pending:
  *   - Install register IO callbacks in order to control
  *     value read from fifo_eng_runlist_r register.
  *   - Configure callback to clear pending bit after one nvgpu_readl.
- *   - Call gk20a_runlist_wait_pending.
+ *   - Call nvgpu_runlist_wait_pending_legacy.
  *   - Configure callback to clear pending bit after two nvgpu_readl.
- *   - Call gk20a_runlist_wait_pending.
+ *   - Call nvgpu_runlist_wait_pending_legacy.
  * - Check case where polling times out:
  *   - Set register to indicate that runlist is pending.
- *   - Call gk20a_runlist_wait_pending.
+ *   - Call nvgpu_runlist_wait_pending_legacy.
  *
  * Output: Returns PASS if all branches gave expected results. FAIL otherwise.
  */