[PATCH AUTOSEL 5.18 11/93] drm/amd/display: fix system hang when PSR exits

[PATCH AUTOSEL 5.18 11/93] drm/amd/display: fix system hang when PSR exits

[Sasha Levin]

From: David Zhang <dingchen.zhang@amd.com>

[ Upstream commit 6cc5c77300afbb285c4f41e04f3435ae3c484c40 ]

[why]
When DC driver send PSR exit dmub command to DMUB FW, it might not
wait until PSR exit. Then it may hit the following deadlock situation.
1. DC driver send HW LOCK command to DMUB FW due to frame update
2. DMUB FW Set the HW lock
3. DMUB execute PSR exit sequence and stuck at polling DPG Pending
register due to the HW Lock is set
4. DC driver ask DMUB FW to unlock HW lock, but DMUB FW is polling
DPG pending register

[how]
The reason why DC driver doesn't wait until PSR exit is because some of
the PSR state machine state is not update the dc driver. So when DC
driver read back the PSR state, it take the state for PSR inactive.

Signed-off-by: David Zhang <dingchen.zhang@amd.com>
Acked-by: Leo Li <sunpeng.li@amd.com>
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 drivers/gpu/drm/amd/display/dc/dc_types.h     |  7 +++++++
 drivers/gpu/drm/amd/display/dc/dce/dmub_psr.c | 16 ++++++++++++++++
 2 files changed, 23 insertions(+)

diff --git a/drivers/gpu/drm/amd/display/dc/dc_types.h b/drivers/gpu/drm/amd/display/dc/dc_types.h
index 2ba9f528c0fe..f1f11b3c205f 100644
--- a/drivers/gpu/drm/amd/display/dc/dc_types.h
+++ b/drivers/gpu/drm/amd/display/dc/dc_types.h
@@ -657,10 +657,17 @@ enum dc_psr_state {
 	PSR_STATE4b,
 	PSR_STATE4c,
 	PSR_STATE4d,
+	PSR_STATE4_FULL_FRAME,
+	PSR_STATE4a_FULL_FRAME,
+	PSR_STATE4b_FULL_FRAME,
+	PSR_STATE4c_FULL_FRAME,
+	PSR_STATE4_FULL_FRAME_POWERUP,
 	PSR_STATE5,
 	PSR_STATE5a,
 	PSR_STATE5b,
 	PSR_STATE5c,
+	PSR_STATE_HWLOCK_MGR,
+	PSR_STATE_POLLVUPDATE,
 	PSR_STATE_INVALID = 0xFF
 };
 
diff --git a/drivers/gpu/drm/amd/display/dc/dce/dmub_psr.c b/drivers/gpu/drm/amd/display/dc/dce/dmub_psr.c
index 312c68172689..ce2f70134669 100644
--- a/drivers/gpu/drm/amd/display/dc/dce/dmub_psr.c
+++ b/drivers/gpu/drm/amd/display/dc/dce/dmub_psr.c
@@ -74,6 +74,22 @@ static enum dc_psr_state convert_psr_state(uint32_t raw_state)
 		state = PSR_STATE5b;
 	else if (raw_state == 0x53)
 		state = PSR_STATE5c;
+	else if (raw_state == 0x4A)
+		state = PSR_STATE4_FULL_FRAME;
+	else if (raw_state == 0x4B)
+		state = PSR_STATE4a_FULL_FRAME;
+	else if (raw_state == 0x4C)
+		state = PSR_STATE4b_FULL_FRAME;
+	else if (raw_state == 0x4D)
+		state = PSR_STATE4c_FULL_FRAME;
+	else if (raw_state == 0x4E)
+		state = PSR_STATE4_FULL_FRAME_POWERUP;
+	else if (raw_state == 0x60)
+		state = PSR_STATE_HWLOCK_MGR;
+	else if (raw_state == 0x61)
+		state = PSR_STATE_POLLVUPDATE;
+	else
+		state = PSR_STATE_INVALID;
 
 	return state;
 }
-- 
2.35.1

[PATCH AUTOSEL 5.18 12/93] drm/amd/display: Detect dpcd_rev when hotplug mst monitor

[Sasha Levin]

From: Wayne Lin <Wayne.Lin@amd.com>

[ Upstream commit 453b0016a054df0f442fda8a145b97a33816cab9 ]

[Why]
Once mst topology is constructed, later on new connected monitors
are reported to source by CSN message. Within CSN, there is no
carried info of DPCD_REV comparing to LINK_ADDRESS reply. As the
result, we might leave some ports connected to DP but without DPCD
revision number which will affect us determining the capability of
the DP Rx.

[How]
Send out remote DPCD read when the port's dpcd_rev is 0x0 in
detect_ctx(). Firstly, read out the value from DPCD 0x2200. If the
return value is 0x0, it's likely the DP1.2 DP Rx then we reques
revision from DPCD 0x0 again.

Reviewed-by: Hersen Wu <hersenwu@amd.com>
Acked-by: Hamza Mahfooz <hamza.mahfooz@amd.com>
Signed-off-by: Wayne Lin <Wayne.Lin@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 .../display/amdgpu_dm/amdgpu_dm_mst_types.c   | 38 ++++++++++++++++++-
 1 file changed, 37 insertions(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c
index d864cae1af67..bcbb6f6f8c82 100644
--- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c
+++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c
@@ -363,12 +363,48 @@ dm_dp_mst_detect(struct drm_connector *connector,
 {
 	struct amdgpu_dm_connector *aconnector = to_amdgpu_dm_connector(connector);
 	struct amdgpu_dm_connector *master = aconnector->mst_port;
+	struct drm_dp_mst_port *port = aconnector->port;
+	int connection_status;
 
 	if (drm_connector_is_unregistered(connector))
 		return connector_status_disconnected;
 
-	return drm_dp_mst_detect_port(connector, ctx, &master->mst_mgr,
+	connection_status = drm_dp_mst_detect_port(connector, ctx, &master->mst_mgr,
 				      aconnector->port);
+
+	if (port->pdt != DP_PEER_DEVICE_NONE && !port->dpcd_rev) {
+		uint8_t dpcd_rev;
+		int ret;
+
+		ret = drm_dp_dpcd_readb(&port->aux, DP_DP13_DPCD_REV, &dpcd_rev);
+
+		if (ret == 1) {
+			port->dpcd_rev = dpcd_rev;
+
+			/* Could be DP1.2 DP Rx case*/
+			if (!dpcd_rev) {
+				ret = drm_dp_dpcd_readb(&port->aux, DP_DPCD_REV, &dpcd_rev);
+
+				if (ret == 1)
+					port->dpcd_rev = dpcd_rev;
+			}
+
+			if (!dpcd_rev)
+				DRM_DEBUG_KMS("Can't decide DPCD revision number!");
+		}
+
+		/*
+		 * Could be legacy sink, logical port etc on DP1.2.
+		 * Will get Nack under these cases when issue remote
+		 * DPCD read.
+		 */
+		if (ret != 1)
+			DRM_DEBUG_KMS("Can't access DPCD");
+	} else if (port->pdt == DP_PEER_DEVICE_NONE) {
+		port->dpcd_rev = 0;
+	}
+
+	return connection_status;
 }
 
 static int dm_dp_mst_atomic_check(struct drm_connector *connector,
-- 
2.35.1

[PATCH AUTOSEL 5.18 16/93] drm/amd/display: Fix dpp dto for disabled pipes

[Sasha Levin]

From: Duncan Ma <duncan.ma@amd.com>

[ Upstream commit d4965c53b95d7533dfc2309d2fc25838bd33220e ]

[Why]
When switching from 1 pipe to 4to1 mpc combine,
DppDtoClk aren't enabled for the disabled pipes
pior to programming the pipes. Upon optimizing
bandwidth, DppDto are enabled causing intermittent
underflow.

[How]
Update dppclk dto whenever pipe are flagged to
enable.

Reviewed-by: Dmytro Laktyushkin <Dmytro.Laktyushkin@amd.com>
Reviewed-by: Nicholas Kazlauskas <Nicholas.Kazlauskas@amd.com>
Reviewed-by: Hansen Dsouza <Hansen.Dsouza@amd.com>
Acked-by: Hamza Mahfooz <hamza.mahfooz@amd.com>
Signed-off-by: Duncan Ma <duncan.ma@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 drivers/gpu/drm/amd/display/dc/dcn20/dcn20_hwseq.c | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/drivers/gpu/drm/amd/display/dc/dcn20/dcn20_hwseq.c b/drivers/gpu/drm/amd/display/dc/dcn20/dcn20_hwseq.c
index b627c41713cc..76decaed5acd 100644
--- a/drivers/gpu/drm/amd/display/dc/dcn20/dcn20_hwseq.c
+++ b/drivers/gpu/drm/amd/display/dc/dcn20/dcn20_hwseq.c
@@ -1412,11 +1412,15 @@ static void dcn20_update_dchubp_dpp(
 	struct hubp *hubp = pipe_ctx->plane_res.hubp;
 	struct dpp *dpp = pipe_ctx->plane_res.dpp;
 	struct dc_plane_state *plane_state = pipe_ctx->plane_state;
+	struct dccg *dccg = dc->res_pool->dccg;
 	bool viewport_changed = false;
 
 	if (pipe_ctx->update_flags.bits.dppclk)
 		dpp->funcs->dpp_dppclk_control(dpp, false, true);
 
+	if (pipe_ctx->update_flags.bits.enable)
+		dccg->funcs->update_dpp_dto(dccg, dpp->inst, pipe_ctx->plane_res.bw.dppclk_khz);
+
 	/* TODO: Need input parameter to tell current DCHUB pipe tie to which OTG
 	 * VTG is within DCHUBBUB which is commond block share by each pipe HUBP.
 	 * VTG is 1:1 mapping with OTG. Each pipe HUBP will select which VTG
-- 
2.35.1

[PATCH AUTOSEL 5.18 17/93] drm/amd/display: Reduce frame size in the bouding box for DCN20

[Sasha Levin]

From: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>

[ Upstream commit c55300ad4a1814bac9191a4d2c7b0d74273aec7c ]

GCC throw warnings for the function dcn20_update_bounding_box due to its
frame size that looks like this:

 error: the frame size of 1936 bytes is larger than 1024 bytes [-Werror=frame-larger-than=]

This commit fixes this issue by eliminating an intermediary variable
that creates a large array.

Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Cc: Hamza Mahfooz <hamza.mahfooz@amd.com>
Cc: Aurabindo Pillai <aurabindo.pillai@amd.com>
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 .../drm/amd/display/dc/dml/dcn20/dcn20_fpu.c  | 38 +++++++++----------
 1 file changed, 18 insertions(+), 20 deletions(-)

diff --git a/drivers/gpu/drm/amd/display/dc/dml/dcn20/dcn20_fpu.c b/drivers/gpu/drm/amd/display/dc/dml/dcn20/dcn20_fpu.c
index f93af45aeab4..c758f84baaf5 100644
--- a/drivers/gpu/drm/amd/display/dc/dml/dcn20/dcn20_fpu.c
+++ b/drivers/gpu/drm/amd/display/dc/dml/dcn20/dcn20_fpu.c
@@ -1430,21 +1430,20 @@ void dcn20_calculate_wm(
 void dcn20_update_bounding_box(struct dc *dc, struct _vcs_dpi_soc_bounding_box_st *bb,
 		struct pp_smu_nv_clock_table *max_clocks, unsigned int *uclk_states, unsigned int num_states)
 {
-	struct _vcs_dpi_voltage_scaling_st calculated_states[DC__VOLTAGE_STATES];
-	int i;
 	int num_calculated_states = 0;
 	int min_dcfclk = 0;
+	int i;
 
 	dc_assert_fp_enabled();
 
 	if (num_states == 0)
 		return;
 
-	memset(calculated_states, 0, sizeof(calculated_states));
+	memset(bb->clock_limits, 0, sizeof(bb->clock_limits));
 
-	if (dc->bb_overrides.min_dcfclk_mhz > 0)
+	if (dc->bb_overrides.min_dcfclk_mhz > 0) {
 		min_dcfclk = dc->bb_overrides.min_dcfclk_mhz;
-	else {
+	} else {
 		if (ASICREV_IS_NAVI12_P(dc->ctx->asic_id.hw_internal_rev))
 			min_dcfclk = 310;
 		else
@@ -1455,36 +1454,35 @@ void dcn20_update_bounding_box(struct dc *dc, struct _vcs_dpi_soc_bounding_box_s
 
 	for (i = 0; i < num_states; i++) {
 		int min_fclk_required_by_uclk;
-		calculated_states[i].state = i;
-		calculated_states[i].dram_speed_mts = uclk_states[i] * 16 / 1000;
+		bb->clock_limits[i].state = i;
+		bb->clock_limits[i].dram_speed_mts = uclk_states[i] * 16 / 1000;
 
 		// FCLK:UCLK ratio is 1.08
 		min_fclk_required_by_uclk = div_u64(((unsigned long long)uclk_states[i]) * 1080,
 			1000000);
 
-		calculated_states[i].fabricclk_mhz = (min_fclk_required_by_uclk < min_dcfclk) ?
+		bb->clock_limits[i].fabricclk_mhz = (min_fclk_required_by_uclk < min_dcfclk) ?
 				min_dcfclk : min_fclk_required_by_uclk;
 
-		calculated_states[i].socclk_mhz = (calculated_states[i].fabricclk_mhz > max_clocks->socClockInKhz / 1000) ?
-				max_clocks->socClockInKhz / 1000 : calculated_states[i].fabricclk_mhz;
+		bb->clock_limits[i].socclk_mhz = (bb->clock_limits[i].fabricclk_mhz > max_clocks->socClockInKhz / 1000) ?
+				max_clocks->socClockInKhz / 1000 : bb->clock_limits[i].fabricclk_mhz;
 
-		calculated_states[i].dcfclk_mhz = (calculated_states[i].fabricclk_mhz > max_clocks->dcfClockInKhz / 1000) ?
-				max_clocks->dcfClockInKhz / 1000 : calculated_states[i].fabricclk_mhz;
+		bb->clock_limits[i].dcfclk_mhz = (bb->clock_limits[i].fabricclk_mhz > max_clocks->dcfClockInKhz / 1000) ?
+				max_clocks->dcfClockInKhz / 1000 : bb->clock_limits[i].fabricclk_mhz;
 
-		calculated_states[i].dispclk_mhz = max_clocks->displayClockInKhz / 1000;
-		calculated_states[i].dppclk_mhz = max_clocks->displayClockInKhz / 1000;
-		calculated_states[i].dscclk_mhz = max_clocks->displayClockInKhz / (1000 * 3);
+		bb->clock_limits[i].dispclk_mhz = max_clocks->displayClockInKhz / 1000;
+		bb->clock_limits[i].dppclk_mhz = max_clocks->displayClockInKhz / 1000;
+		bb->clock_limits[i].dscclk_mhz = max_clocks->displayClockInKhz / (1000 * 3);
 
-		calculated_states[i].phyclk_mhz = max_clocks->phyClockInKhz / 1000;
+		bb->clock_limits[i].phyclk_mhz = max_clocks->phyClockInKhz / 1000;
 
 		num_calculated_states++;
 	}
 
-	calculated_states[num_calculated_states - 1].socclk_mhz = max_clocks->socClockInKhz / 1000;
-	calculated_states[num_calculated_states - 1].fabricclk_mhz = max_clocks->socClockInKhz / 1000;
-	calculated_states[num_calculated_states - 1].dcfclk_mhz = max_clocks->dcfClockInKhz / 1000;
+	bb->clock_limits[num_calculated_states - 1].socclk_mhz = max_clocks->socClockInKhz / 1000;
+	bb->clock_limits[num_calculated_states - 1].fabricclk_mhz = max_clocks->socClockInKhz / 1000;
+	bb->clock_limits[num_calculated_states - 1].dcfclk_mhz = max_clocks->dcfClockInKhz / 1000;
 
-	memcpy(bb->clock_limits, calculated_states, sizeof(bb->clock_limits));
 	bb->num_states = num_calculated_states;
 
 	// Duplicate the last state, DML always an extra state identical to max state to work
-- 
2.35.1

[PATCH AUTOSEL 5.18 18/93] drm/radeon: integer overflow in radeon_mode_dumb_create()

[Sasha Levin]

From: Xiaohui Zhang <xiaohuizhang@ruc.edu.cn>

[ Upstream commit feb54650bae25f2a2adfc493e3e254e7c27a3fba ]

Similar to the handling of amdgpu_mode_dumb_create in commit 54ef0b5461c0
("drm/amdgpu: integer overflow in amdgpu_mode_dumb_create()"),
we thought a patch might be needed here as well.

args->size is a u64.  arg->pitch and args->height are u32.  The
multiplication will overflow instead of using the high 32 bits as
intended.

Signed-off-by: Xiaohui Zhang <xiaohuizhang@ruc.edu.cn>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 drivers/gpu/drm/radeon/radeon_gem.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/radeon/radeon_gem.c b/drivers/gpu/drm/radeon/radeon_gem.c
index f563284a7fac..6c9cc0061a8d 100644
--- a/drivers/gpu/drm/radeon/radeon_gem.c
+++ b/drivers/gpu/drm/radeon/radeon_gem.c
@@ -830,7 +830,7 @@ int radeon_mode_dumb_create(struct drm_file *file_priv,
 
 	args->pitch = radeon_align_pitch(rdev, args->width,
 					 DIV_ROUND_UP(args->bpp, 8), 0);
-	args->size = args->pitch * args->height;
+	args->size = (u64)args->pitch * args->height;
 	args->size = ALIGN(args->size, PAGE_SIZE);
 
 	r = radeon_gem_object_create(rdev, args->size, 0,
-- 
2.35.1

[PATCH AUTOSEL 5.18 19/93] drm/radeon: Initialize fences array entries in radeon_sa_bo_next_hole

[Sasha Levin]

From: Xiaohui Zhang <xiaohuizhang@ruc.edu.cn>

[ Upstream commit 0381ac3ca2e727d4dfb7264d9416a8ba6bb6c18b ]

Similar to the handling of amdgpu_sa_bo_next_hole in commit 6a15f3ff19a8
("drm/amdgpu: Initialize fences array entries in amdgpu_sa_bo_next_hole"),
we thought a patch might be needed here as well.

The entries were only initialized once in radeon_sa_bo_new. If a fence
wasn't signalled yet in the first radeon_sa_bo_next_hole call, but then
got signalled before a later radeon_sa_bo_next_hole call, it could
destroy the fence but leave its pointer in the array, resulting in
use-after-free in radeon_sa_bo_new.

Signed-off-by: Xiaohui Zhang <xiaohuizhang@ruc.edu.cn>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 drivers/gpu/drm/radeon/radeon_sa.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/drivers/gpu/drm/radeon/radeon_sa.c b/drivers/gpu/drm/radeon/radeon_sa.c
index 310c322c7112..0981948bd9ed 100644
--- a/drivers/gpu/drm/radeon/radeon_sa.c
+++ b/drivers/gpu/drm/radeon/radeon_sa.c
@@ -267,6 +267,8 @@ static bool radeon_sa_bo_next_hole(struct radeon_sa_manager *sa_manager,
 	for (i = 0; i < RADEON_NUM_RINGS; ++i) {
 		struct radeon_sa_bo *sa_bo;
 
+		fences[i] = NULL;
+
 		if (list_empty(&sa_manager->flist[i])) {
 			continue;
 		}
@@ -332,10 +334,8 @@ int radeon_sa_bo_new(struct radeon_device *rdev,
 
 	spin_lock(&sa_manager->wq.lock);
 	do {
-		for (i = 0; i < RADEON_NUM_RINGS; ++i) {
-			fences[i] = NULL;
+		for (i = 0; i < RADEON_NUM_RINGS; ++i)
 			tries[i] = 0;
-		}
 
 		do {
 			radeon_sa_bo_try_free(sa_manager);
-- 
2.35.1

[PATCH AUTOSEL 5.18 24/93] drm/amd/display: Fix monitor flash issue

[Sasha Levin]

From: Shah Dharati <dharshah@amd.com>

[ Upstream commit b840b64bc8ed3fc46f6d6aa7f97c43862a33bea5 ]

[Why & How]
For a some specific monitors, when connected on boot or hot plug,
monitor flash for 1/2 seconds can happen during first HDCP query
operation. Ading some delay in the init sequence for these monitors
fixes the issue, so it is implemented as monitor specific patch.

Co-authored-by: Shah Dharati <dharshah@amd.com>
Reviewed-by: Hansen Dsouza <Hansen.Dsouza@amd.com>
Reviewed-by: Nicholas Kazlauskas <Nicholas.Kazlauskas@amd.com>
Acked-by: Hamza Mahfooz <hamza.mahfooz@amd.com>
Signed-off-by: Shah Dharati <dharshah@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 drivers/gpu/drm/amd/display/modules/hdcp/hdcp2_transition.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/amd/display/modules/hdcp/hdcp2_transition.c b/drivers/gpu/drm/amd/display/modules/hdcp/hdcp2_transition.c
index 1f4095b26409..c5f6c11de7e5 100644
--- a/drivers/gpu/drm/amd/display/modules/hdcp/hdcp2_transition.c
+++ b/drivers/gpu/drm/amd/display/modules/hdcp/hdcp2_transition.c
@@ -524,7 +524,7 @@ enum mod_hdcp_status mod_hdcp_hdcp2_dp_transition(struct mod_hdcp *hdcp,
 			set_watchdog_in_ms(hdcp, 3000, output);
 			set_state_id(hdcp, output, D2_A6_WAIT_FOR_RX_ID_LIST);
 		} else {
-			callback_in_ms(0, output);
+			callback_in_ms(1, output);
 			set_state_id(hdcp, output, D2_SEND_CONTENT_STREAM_TYPE);
 		}
 		break;
-- 
2.35.1

[PATCH AUTOSEL 5.18 41/93] amdgpu/pm: Fix possible array out-of-bounds if SCLK levels != 2

[Sasha Levin]

From: Darren Powell <darren.powell@amd.com>

[ Upstream commit ceb180361e3851007547c55035cd1de03f108f75 ]

 [v2]
simplified fix after Lijo's feedback
 removed clocks.num_levels from calculation of loop count
   removed unsafe accesses to shim table freq_values
 retained corner case output only min,now if
   clocks.num_levels == 1 && now > min

 [v1]
added a check to populate and use SCLK shim table freq_values only
   if using dpm_level == AMD_DPM_FORCED_LEVEL_MANUAL or
                         AMD_DPM_FORCED_LEVEL_PERF_DETERMINISM
removed clocks.num_levels from calculation of shim table size
removed unsafe accesses to shim table freq_values
   output gfx_table values if using other dpm levels
added check for freq_match when using freq_values for when now == min_clk

== Test ==
LOGFILE=aldebaran-sclk.test.log
AMDGPU_PCI_ADDR=`lspci -nn | grep "VGA\|Display" | cut -d " " -f 1`
AMDGPU_HWMON=`ls -la /sys/class/hwmon | grep $AMDGPU_PCI_ADDR | awk '{print $9}'`
HWMON_DIR=/sys/class/hwmon/${AMDGPU_HWMON}

lspci -nn | grep "VGA\|Display"  > $LOGFILE
FILES="pp_od_clk_voltage
pp_dpm_sclk"

for f in $FILES
do
  echo === $f === >> $LOGFILE
  cat $HWMON_DIR/device/$f >> $LOGFILE
done
cat $LOGFILE

Signed-off-by: Darren Powell <darren.powell@amd.com>
Reviewed-by: Kenneth Feng <kenneth.feng@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 .../drm/amd/pm/swsmu/smu13/aldebaran_ppt.c    | 34 +++++++------------
 1 file changed, 12 insertions(+), 22 deletions(-)

diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu13/aldebaran_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu13/aldebaran_ppt.c
index 7f998f24af81..b57750cb27c4 100644
--- a/drivers/gpu/drm/amd/pm/swsmu/smu13/aldebaran_ppt.c
+++ b/drivers/gpu/drm/amd/pm/swsmu/smu13/aldebaran_ppt.c
@@ -733,7 +733,7 @@ static int aldebaran_print_clk_levels(struct smu_context *smu,
 	struct smu_13_0_dpm_table *single_dpm_table;
 	struct smu_dpm_context *smu_dpm = &smu->smu_dpm;
 	struct smu_13_0_dpm_context *dpm_context = NULL;
-	uint32_t display_levels;
+	int display_levels;
 	uint32_t freq_values[3] = {0};
 	uint32_t min_clk, max_clk;
 
@@ -765,7 +765,7 @@ static int aldebaran_print_clk_levels(struct smu_context *smu,
 			return ret;
 		}
 
-		display_levels = clocks.num_levels;
+		display_levels = (clocks.num_levels == 1) ? 1 : 2;
 
 		min_clk = pstate_table->gfxclk_pstate.curr.min;
 		max_clk = pstate_table->gfxclk_pstate.curr.max;
@@ -775,30 +775,20 @@ static int aldebaran_print_clk_levels(struct smu_context *smu,
 
 		/* fine-grained dpm has only 2 levels */
 		if (now > min_clk && now < max_clk) {
-			display_levels = clocks.num_levels + 1;
+			display_levels++;
 			freq_values[2] = max_clk;
 			freq_values[1] = now;
 		}
 
-		/*
-		 * For DPM disabled case, there will be only one clock level.
-		 * And it's safe to assume that is always the current clock.
-		 */
-		if (display_levels == clocks.num_levels) {
-			for (i = 0; i < clocks.num_levels; i++)
-				size += sysfs_emit_at(buf, size, "%d: %uMhz %s\n", i,
-					freq_values[i],
-					(clocks.num_levels == 1) ?
-						"*" :
-						(aldebaran_freqs_in_same_level(
-							 freq_values[i], now) ?
-							 "*" :
-							 ""));
-		} else {
-			for (i = 0; i < display_levels; i++)
-				size += sysfs_emit_at(buf, size, "%d: %uMhz %s\n", i,
-						freq_values[i], i == 1 ? "*" : "");
-		}
+		for (i = 0; i < display_levels; i++)
+			size += sysfs_emit_at(buf, size, "%d: %uMhz %s\n", i,
+				freq_values[i],
+				(display_levels == 1) ?
+					"*" :
+					(aldebaran_freqs_in_same_level(
+						 freq_values[i], now) ?
+						 "*" :
+						 ""));
 
 		break;
 
-- 
2.35.1

[PATCH AUTOSEL 5.18 47/93] drm/amdgpu/display/dc: Fix null pointer exception

[Sasha Levin]

From: Rahul Kumar <rahul.kumar1@amd.com>

[ Upstream commit 1c4dae3e4639540fb567e570cc56a3c292afb6fe ]

We observed hard hang due to NULL derefrence This issue is seen after
running system all the time after two or three days

struct dc *dc = plane_state->ctx->dc; Randomly in long run we found
plane_state or plane_state->ctx is found NULL which causes exception.

BUG: kernel NULL pointer dereference, address: 0000000000000000
PF: supervisor read access in kernel mode
PF: error_code(0x0000) - not-present page
PGD 1dc7f2067 P4D 1dc7f2067 PUD 222c75067 PMD 0
Oops: 0000 [#1] SMP NOPTI
CPU: 5 PID: 29855 Comm: kworker/u16:4 ...
...
Workqueue: events_unbound commit_work [drm_kms_helper]
RIP: 0010:dcn10_update_pending_status+0x1f/0xee [amdgpu]
Code: 41 5f c3 0f 1f 44 00 00 b0 01 c3 0f 1f 44 00 00 41 55 41 54 55 53 48 8b 1f 4c 8b af f8 00 00 00 48 8b 83 88 03 00 00 48 85 db <4c> 8b 20 0f 84 bf 00 00 00 48 89 fd 48 8b bf b8 00 00 00 48 8b 07
RSP: 0018:ffff942941997ab8 EFLAGS: 00010286
RAX: 0000000000000000 RBX: ffff8d7fd98d2000 RCX: 0000000000000000
RDX: 0000000000000000 RSI: ffff8d7e3e87c708 RDI: ffff8d7f2d8c0690
RBP: ffff8d7f2d8c0000 R08: ffff942941997a34 R09: 00000000ffffffff
R10: 0000000000005000 R11: 00000000000000f0 R12: ffff8d7f2d8c0690
R13: ffff8d8035a41680 R14: 00000000000186a0 R15: ffff8d7f2d8c1dd8
FS:  0000000000000000(0000) GS:ffff8d8037340000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 0000000000000000 CR3: 0000000148030000 CR4: 00000000003406e0
Call Trace:
 dc_commit_state+0x6a2/0x7f0 [amdgpu]
 amdgpu_dm_atomic_commit_tail+0x460/0x19bb [amdgpu]

Tested-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Rahul Kumar <rahul.kumar1@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hw_sequencer.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hw_sequencer.c b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hw_sequencer.c
index 83fbea2df410..b12a46c4e872 100644
--- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hw_sequencer.c
+++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hw_sequencer.c
@@ -3205,7 +3205,7 @@ void dcn10_update_pending_status(struct pipe_ctx *pipe_ctx)
 	struct dc_plane_state *plane_state = pipe_ctx->plane_state;
 	struct timing_generator *tg = pipe_ctx->stream_res.tg;
 	bool flip_pending;
-	struct dc *dc = plane_state->ctx->dc;
+	struct dc *dc = pipe_ctx->stream->ctx->dc;
 
 	if (plane_state == NULL)
 		return;
-- 
2.35.1

[PATCH AUTOSEL 5.18 51/93] drm/amdgpu: skip whole ras bad page framework on sriov

[Sasha Levin]

From: "Stanley.Yang" <Stanley.Yang@amd.com>

[ Upstream commit e0e146d5567317d6ba7d0169bed55d1d9ea05a61 ]

It should not init whole ras bad page framework on sriov guest side
due to it is handled on host side.

Signed-off-by: Stanley.Yang <Stanley.Yang@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Tao Zhou <tao.zhou1@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
index 3f96dadf2698..7506b0b9c2cb 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
@@ -2062,7 +2062,7 @@ int amdgpu_ras_recovery_init(struct amdgpu_device *adev)
 	bool exc_err_limit = false;
 	int ret;
 
-	if (!con)
+	if (!con || amdgpu_sriov_vf(adev))
 		return 0;
 
 	/* Allow access to RAS EEPROM via debugfs, when the ASIC
-- 
2.35.1

[PATCH AUTOSEL 5.18 52/93] drm/amd/display: Guard against ddc_pin being NULL for AUX

[Sasha Levin]

From: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>

[ Upstream commit bc19909f19fdc8253d720d11c948935786fbfa08 ]

[Why]
In the case where we don't support DMUB aux but we have DPIA links
in the configuration we might try to message AUX using the legacy
path - where DDC pin is NULL. This causes a NULL pointer dereference.

[How]
Guard against NULL DDC pin, return a failure for aux engine acquire.

Reviewed-by: Michael Strauss <Michael.Strauss@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 drivers/gpu/drm/amd/display/dc/dce/dce_aux.c | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/drivers/gpu/drm/amd/display/dc/dce/dce_aux.c b/drivers/gpu/drm/amd/display/dc/dce/dce_aux.c
index 8e814000db62..5620a3345e4d 100644
--- a/drivers/gpu/drm/amd/display/dc/dce/dce_aux.c
+++ b/drivers/gpu/drm/amd/display/dc/dce/dce_aux.c
@@ -573,6 +573,11 @@ int dce_aux_transfer_raw(struct ddc_service *ddc,
 	memset(&aux_req, 0, sizeof(aux_req));
 	memset(&aux_rep, 0, sizeof(aux_rep));
 
+	if (ddc_pin == NULL) {
+		*operation_result = AUX_RET_ERROR_ENGINE_ACQUIRE;
+		return -1;
+	}
+
 	aux_engine = ddc->ctx->dc->res_pool->engines[ddc_pin->pin_data->en];
 	if (!acquire(aux_engine, ddc_pin)) {
 		*operation_result = AUX_RET_ERROR_ENGINE_ACQUIRE;
-- 
2.35.1

[PATCH AUTOSEL 5.18 53/93] drm/amd/display: Fix dmub soft hang for PSR 1

[Sasha Levin]

From: Fangzhi Zuo <Jerry.Zuo@amd.com>

[ Upstream commit 22676bc500c27d987a0b42cbe162aebf783f1c38 ]

[Why]
Unexpected change of aux hw mapping causes dmub soft hang when
initiate aux transation at wrong aux channel.

ddc_channel stands for hw dp aux index which is from vbios,
but link_index is pure software concept for link count depending on which link
is probed first. They are not interchangeable.

dmub aux transaction could pass if happens eDP link_index gets
the same value as vbios ddc_channel, e.g., ddc_channel = 1, link_index = 1
if they gets different, e.g., ddc_channel = 2, link_index = 0, overwrite
ddc_channel with link_index will have wrong ddc channel being used for aux
transaction in dmub PSR, cause aux transaction soft hang.

[How]
ddc_channel mapping to each link is determined by vbios and further
parsed in dc. Such info. should not be touched in any kind, otherwise
the mapping is screwed up leading to aux transaction timeout.

Reviewed-by: Aurabindo Jayamohanan Pillai <Aurabindo.Pillai@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Fangzhi Zuo <Jerry.Zuo@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c | 4 +---
 1 file changed, 1 insertion(+), 3 deletions(-)

diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c
index a2575195c4e0..1991a8a4826e 100644
--- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c
+++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c
@@ -8563,7 +8563,7 @@ static int amdgpu_dm_i2c_xfer(struct i2c_adapter *i2c_adap,
 
 	if (dc_submit_i2c(
 			ddc_service->ctx->dc,
-			ddc_service->ddc_pin->hw_info.ddc_channel,
+			ddc_service->link->link_index,
 			&cmd))
 		result = num;
 
@@ -8599,8 +8599,6 @@ create_i2c(struct ddc_service *ddc_service,
 	snprintf(i2c->base.name, sizeof(i2c->base.name), "AMDGPU DM i2c hw bus %d", link_index);
 	i2c_set_adapdata(&i2c->base, i2c);
 	i2c->ddc_service = ddc_service;
-	if (i2c->ddc_service->ddc_pin)
-		i2c->ddc_service->ddc_pin->hw_info.ddc_channel = link_index;
 
 	return i2c;
 }
-- 
2.35.1

[PATCH AUTOSEL 5.18 54/93] drm/amd/display: disable otg toggle w/a on boot

[Sasha Levin]

From: Dmytro Laktyushkin <Dmytro.Laktyushkin@amd.com>

[ Upstream commit 8a077d9caa3a274de36ee2fe7b608041f5690343 ]

This w/a has a bad interaction with seamless boot toggling an
active stream. Most panels recover, however some fail leading
to display corruption.

Reviewed-by: Charlene Liu <Charlene.Liu@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Dmytro Laktyushkin <Dmytro.Laktyushkin@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 .../gpu/drm/amd/display/dc/clk_mgr/dcn315/dcn315_clk_mgr.c | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn315/dcn315_clk_mgr.c b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn315/dcn315_clk_mgr.c
index 2918ad07d489..8f8680e89a8f 100644
--- a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn315/dcn315_clk_mgr.c
+++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn315/dcn315_clk_mgr.c
@@ -173,11 +173,14 @@ static void dcn315_update_clocks(struct clk_mgr *clk_mgr_base,
 	}
 
 	if (should_set_clock(safe_to_lower, new_clocks->dispclk_khz, clk_mgr_base->clks.dispclk_khz)) {
-		dcn315_disable_otg_wa(clk_mgr_base, true);
+		/* No need to apply the w/a if we haven't taken over from bios yet */
+		if (clk_mgr_base->clks.dispclk_khz)
+			dcn315_disable_otg_wa(clk_mgr_base, true);
 
 		clk_mgr_base->clks.dispclk_khz = new_clocks->dispclk_khz;
 		dcn315_smu_set_dispclk(clk_mgr, clk_mgr_base->clks.dispclk_khz);
-		dcn315_disable_otg_wa(clk_mgr_base, false);
+		if (clk_mgr_base->clks.dispclk_khz)
+			dcn315_disable_otg_wa(clk_mgr_base, false);
 
 		update_dispclk = true;
 	}
-- 
2.35.1

[PATCH AUTOSEL 5.18 63/93] drm/amdgpu: fix file permissions on some files

[Sasha Levin]

From: Alex Deucher <alexander.deucher@amd.com>

[ Upstream commit 0a94608f0f7de9b1135ffea3546afe68eafef57f ]

Drop execute.

Bug: https://gitlab.freedesktop.org/drm/amd/-/issues/2085
Reviewed-by: Guchun Chen <guchun.chen@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 drivers/gpu/drm/amd/include/asic_reg/clk/clk_11_0_1_offset.h   | 0
 drivers/gpu/drm/amd/include/asic_reg/clk/clk_11_0_1_sh_mask.h  | 0
 drivers/gpu/drm/amd/include/asic_reg/dcn/dcn_2_0_3_offset.h    | 0
 drivers/gpu/drm/amd/include/asic_reg/dcn/dcn_2_0_3_sh_mask.h   | 0
 drivers/gpu/drm/amd/include/asic_reg/dpcs/dpcs_2_0_3_offset.h  | 0
 drivers/gpu/drm/amd/include/asic_reg/dpcs/dpcs_2_0_3_sh_mask.h | 0
 6 files changed, 0 insertions(+), 0 deletions(-)
 mode change 100755 => 100644 drivers/gpu/drm/amd/include/asic_reg/clk/clk_11_0_1_offset.h
 mode change 100755 => 100644 drivers/gpu/drm/amd/include/asic_reg/clk/clk_11_0_1_sh_mask.h
 mode change 100755 => 100644 drivers/gpu/drm/amd/include/asic_reg/dcn/dcn_2_0_3_offset.h
 mode change 100755 => 100644 drivers/gpu/drm/amd/include/asic_reg/dcn/dcn_2_0_3_sh_mask.h
 mode change 100755 => 100644 drivers/gpu/drm/amd/include/asic_reg/dpcs/dpcs_2_0_3_offset.h
 mode change 100755 => 100644 drivers/gpu/drm/amd/include/asic_reg/dpcs/dpcs_2_0_3_sh_mask.h

diff --git a/drivers/gpu/drm/amd/include/asic_reg/clk/clk_11_0_1_offset.h b/drivers/gpu/drm/amd/include/asic_reg/clk/clk_11_0_1_offset.h
old mode 100755
new mode 100644
diff --git a/drivers/gpu/drm/amd/include/asic_reg/clk/clk_11_0_1_sh_mask.h b/drivers/gpu/drm/amd/include/asic_reg/clk/clk_11_0_1_sh_mask.h
old mode 100755
new mode 100644
diff --git a/drivers/gpu/drm/amd/include/asic_reg/dcn/dcn_2_0_3_offset.h b/drivers/gpu/drm/amd/include/asic_reg/dcn/dcn_2_0_3_offset.h
old mode 100755
new mode 100644
diff --git a/drivers/gpu/drm/amd/include/asic_reg/dcn/dcn_2_0_3_sh_mask.h b/drivers/gpu/drm/amd/include/asic_reg/dcn/dcn_2_0_3_sh_mask.h
old mode 100755
new mode 100644
diff --git a/drivers/gpu/drm/amd/include/asic_reg/dpcs/dpcs_2_0_3_offset.h b/drivers/gpu/drm/amd/include/asic_reg/dpcs/dpcs_2_0_3_offset.h
old mode 100755
new mode 100644
diff --git a/drivers/gpu/drm/amd/include/asic_reg/dpcs/dpcs_2_0_3_sh_mask.h b/drivers/gpu/drm/amd/include/asic_reg/dpcs/dpcs_2_0_3_sh_mask.h
old mode 100755
new mode 100644
-- 
2.35.1

[PATCH AUTOSEL 5.18 67/93] drm/amdgpu: Call trace info was found in dmesg when loading amdgpu

[Sasha Levin]

From: lin cao <lin.cao@amd.com>

[ Upstream commit 748262eb400e809aa13e3485f4983c3db3d0ebb3 ]

In the case of SRIOV, the register smnMp1_PMI_3_FIFO will get an invalid
value which will cause the "shift out of bound". In Ubuntu22.04, this
issue will be checked an related call trace will be reported in dmesg.

Signed-off-by: lin cao <lin.cao@amd.com>
Reviewed-by: Jingwen Chen <Jingwen.Chen2@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 drivers/gpu/drm/amd/pm/swsmu/smu11/sienna_cichlid_ppt.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu11/sienna_cichlid_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu11/sienna_cichlid_ppt.c
index 7a1e225fb823..971a9712fa63 100644
--- a/drivers/gpu/drm/amd/pm/swsmu/smu11/sienna_cichlid_ppt.c
+++ b/drivers/gpu/drm/amd/pm/swsmu/smu11/sienna_cichlid_ppt.c
@@ -758,6 +758,7 @@ static void sienna_cichlid_stb_init(struct smu_context *smu);
 
 static int sienna_cichlid_init_smc_tables(struct smu_context *smu)
 {
+	struct amdgpu_device *adev = smu->adev;
 	int ret = 0;
 
 	ret = sienna_cichlid_tables_init(smu);
@@ -768,7 +769,8 @@ static int sienna_cichlid_init_smc_tables(struct smu_context *smu)
 	if (ret)
 		return ret;
 
-	sienna_cichlid_stb_init(smu);
+	if (!amdgpu_sriov_vf(adev))
+		sienna_cichlid_stb_init(smu);
 
 	return smu_v11_0_init_smc_tables(smu);
 }
-- 
2.35.1

[PATCH AUTOSEL 5.18 82/93] drm/amdkfd: Correct mmu_notifier_get failure handling

[Sasha Levin]

From: Philip Yang <Philip.Yang@amd.com>

[ Upstream commit 0593ad215359d51514c1e6c81ce28ea598efed6b ]

If process has signal pending, mmu_notifier_get_locked fails and calls
ops->free_notifier, kfd_process_free_notifier will schedule
kfd_process_wq_release as process refcount is 1, but process structure
is already freed. This use after free bug causes system crash with
different backtrace.

The fix is to increase process refcount and then decrease the refcount
after mmu_notifier_get success.

Signed-off-by: Philip Yang <Philip.Yang@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 drivers/gpu/drm/amd/amdkfd/kfd_process.c | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
index 59c04b2d383b..19d4089a0b1c 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
@@ -1397,6 +1397,11 @@ static struct kfd_process *create_process(const struct task_struct *thread)
 	hash_add_rcu(kfd_processes_table, &process->kfd_processes,
 			(uintptr_t)process->mm);
 
+	/* Avoid free_notifier to start kfd_process_wq_release if
+	 * mmu_notifier_get failed because of pending signal.
+	 */
+	kref_get(&process->ref);
+
 	/* MMU notifier registration must be the last call that can fail
 	 * because after this point we cannot unwind the process creation.
 	 * After this point, mmu_notifier_put will trigger the cleanup by
@@ -1409,6 +1414,7 @@ static struct kfd_process *create_process(const struct task_struct *thread)
 	}
 	BUG_ON(mn != &process->mmu_notifier);
 
+	kfd_unref_process(process);
 	get_task_struct(process->lead_thread);
 
 	return process;
-- 
2.35.1

[PATCH AUTOSEL 5.18 83/93] drm/amdkfd: Process notifier release callback don't take mutex

[Sasha Levin]

From: Philip Yang <Philip.Yang@amd.com>

[ Upstream commit 74097f9fd2f5ebdae04fcba59da345386415cbf3 ]

Move process queues cleanup to deferred work kfd_process_wq_release, to
avoid potential deadlock circular locking warning:

 WARNING: possible circular locking dependency detected
               the existing dependency chain (in reverse order) is:
      -> #2
        ((work_completion)(&svms->deferred_list_work)){+.+.}-{0:0}:
        __flush_work+0x343/0x4a0
        svm_range_list_lock_and_flush_work+0x39/0xc0
        svm_range_set_attr+0xe8/0x1080 [amdgpu]
        kfd_ioctl+0x19b/0x600 [amdgpu]
        __x64_sys_ioctl+0x81/0xb0
        do_syscall_64+0x34/0x80
        entry_SYSCALL_64_after_hwframe+0x44/0xae

      -> #1 (&info->lock#2){+.+.}-{3:3}:
        __mutex_lock+0xa4/0x940
        amdgpu_amdkfd_gpuvm_acquire_process_vm+0x2e3/0x590
        kfd_process_device_init_vm+0x61/0x200 [amdgpu]
        kfd_ioctl_acquire_vm+0x83/0xb0 [amdgpu]
        kfd_ioctl+0x19b/0x600 [amdgpu]
        __x64_sys_ioctl+0x81/0xb0
        do_syscall_64+0x34/0x80
       entry_SYSCALL_64_after_hwframe+0x44/0xae

      -> #0 (&process->mutex){+.+.}-{3:3}:
        __lock_acquire+0x1365/0x23d0
        lock_acquire+0xc9/0x2e0
        __mutex_lock+0xa4/0x940
        kfd_process_notifier_release+0x96/0xe0 [amdgpu]
        __mmu_notifier_release+0x94/0x210
        exit_mmap+0x35/0x1f0
        mmput+0x63/0x120
        svm_range_deferred_list_work+0x177/0x2c0 [amdgpu]
        process_one_work+0x2a4/0x600
        worker_thread+0x39/0x3e0
        kthread+0x16d/0x1a0

  Possible unsafe locking scenario:

      CPU0                    CPU1
        ----                    ----
   lock((work_completion)(&svms->deferred_list_work));
                                lock(&info->lock#2);
             lock((work_completion)(&svms->deferred_list_work));
   lock(&process->mutex);

Signed-off-by: Philip Yang <Philip.Yang@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 drivers/gpu/drm/amd/amdkfd/kfd_process.c | 21 +++++++++------------
 1 file changed, 9 insertions(+), 12 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
index 19d4089a0b1c..3f2383db1b73 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
@@ -1111,6 +1111,15 @@ static void kfd_process_wq_release(struct work_struct *work)
 	struct kfd_process *p = container_of(work, struct kfd_process,
 					     release_work);
 
+	kfd_process_dequeue_from_all_devices(p);
+	pqm_uninit(&p->pqm);
+
+	/* Signal the eviction fence after user mode queues are
+	 * destroyed. This allows any BOs to be freed without
+	 * triggering pointless evictions or waiting for fences.
+	 */
+	dma_fence_signal(p->ef);
+
 	kfd_process_remove_sysfs(p);
 	kfd_iommu_unbind_process(p);
 
@@ -1175,20 +1184,8 @@ static void kfd_process_notifier_release(struct mmu_notifier *mn,
 	cancel_delayed_work_sync(&p->eviction_work);
 	cancel_delayed_work_sync(&p->restore_work);
 
-	mutex_lock(&p->mutex);
-
-	kfd_process_dequeue_from_all_devices(p);
-	pqm_uninit(&p->pqm);
-
 	/* Indicate to other users that MM is no longer valid */
 	p->mm = NULL;
-	/* Signal the eviction fence after user mode queues are
-	 * destroyed. This allows any BOs to be freed without
-	 * triggering pointless evictions or waiting for fences.
-	 */
-	dma_fence_signal(p->ef);
-
-	mutex_unlock(&p->mutex);
 
 	mmu_notifier_put(&p->mmu_notifier);
 }
-- 
2.35.1