[PATCH AUTOSEL 5.12 14/49] drm/tegra: sor: Do not leak runtime PM reference

[PATCH AUTOSEL 5.12 14/49] drm/tegra: sor: Do not leak runtime PM reference

[Sasha Levin]

From: "Pavel Machek (CIP)" <pavel@denx.de>

[ Upstream commit 73a395c46704304b96bc5e2ee19be31124025c0c ]

It's theoretically possible for the runtime PM reference to leak if the
code fails anywhere between the pm_runtime_resume_and_get() and
pm_runtime_put() calls, so make sure to release the runtime PM reference
in that case.

Practically this will never happen because none of the functions will
fail on Tegra, but it's better for the code to be pedantic in case these
assumptions will ever become wrong.

Signed-off-by: Pavel Machek (CIP) <pavel@denx.de>
[treding@nvidia.com: add commit message]
Signed-off-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 drivers/gpu/drm/tegra/sor.c | 14 ++++++++++----
 1 file changed, 10 insertions(+), 4 deletions(-)

diff --git a/drivers/gpu/drm/tegra/sor.c b/drivers/gpu/drm/tegra/sor.c
index 7b88261f57bb..67a80dae1c00 100644
--- a/drivers/gpu/drm/tegra/sor.c
+++ b/drivers/gpu/drm/tegra/sor.c
@@ -3125,21 +3125,21 @@ static int tegra_sor_init(struct host1x_client *client)
 		if (err < 0) {
 			dev_err(sor->dev, "failed to acquire SOR reset: %d\n",
 				err);
-			return err;
+			goto rpm_put;
 		}
 
 		err = reset_control_assert(sor->rst);
 		if (err < 0) {
 			dev_err(sor->dev, "failed to assert SOR reset: %d\n",
 				err);
-			return err;
+			goto rpm_put;
 		}
 	}
 
 	err = clk_prepare_enable(sor->clk);
 	if (err < 0) {
 		dev_err(sor->dev, "failed to enable clock: %d\n", err);
-		return err;
+		goto rpm_put;
 	}
 
 	usleep_range(1000, 3000);
@@ -3150,7 +3150,7 @@ static int tegra_sor_init(struct host1x_client *client)
 			dev_err(sor->dev, "failed to deassert SOR reset: %d\n",
 				err);
 			clk_disable_unprepare(sor->clk);
-			return err;
+			goto rpm_put;
 		}
 
 		reset_control_release(sor->rst);
@@ -3171,6 +3171,12 @@ static int tegra_sor_init(struct host1x_client *client)
 	}
 
 	return 0;
+
+rpm_put:
+	if (sor->rst)
+		pm_runtime_put(sor->dev);
+
+	return err;
 }
 
 static int tegra_sor_exit(struct host1x_client *client)
-- 
2.30.2

[PATCH AUTOSEL 5.12 15/49] gpu: host1x: Split up client initalization and registration

[Sasha Levin]

From: Thierry Reding <treding@nvidia.com>

[ Upstream commit 0cfe5a6e758fb20be8ad3e8f10cb087cc8033eeb ]

In some cases we may need to initialize the host1x client first before
registering it. This commit adds a new helper that will do nothing but
the initialization of the data structure.

At the same time, the initialization is removed from the registration
function. Note, however, that for simplicity we explicitly initialize
the client when the host1x_client_register() function is called, as
opposed to the low-level __host1x_client_register() function. This
allows existing callers to remain unchanged.

Signed-off-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 drivers/gpu/host1x/bus.c | 30 ++++++++++++++++++++++++------
 include/linux/host1x.h   | 30 ++++++++++++++++++++++++------
 2 files changed, 48 insertions(+), 12 deletions(-)

diff --git a/drivers/gpu/host1x/bus.c b/drivers/gpu/host1x/bus.c
index 68a766ff0e9d..b98d746141a8 100644
--- a/drivers/gpu/host1x/bus.c
+++ b/drivers/gpu/host1x/bus.c
@@ -704,6 +704,29 @@ void host1x_driver_unregister(struct host1x_driver *driver)
 }
 EXPORT_SYMBOL(host1x_driver_unregister);
 
+/**
+ * __host1x_client_init() - initialize a host1x client
+ * @client: host1x client
+ * @key: lock class key for the client-specific mutex
+ */
+void __host1x_client_init(struct host1x_client *client, struct lock_class_key *key)
+{
+	INIT_LIST_HEAD(&client->list);
+	__mutex_init(&client->lock, "host1x client lock", key);
+	client->usecount = 0;
+}
+EXPORT_SYMBOL(__host1x_client_init);
+
+/**
+ * host1x_client_exit() - uninitialize a host1x client
+ * @client: host1x client
+ */
+void host1x_client_exit(struct host1x_client *client)
+{
+	mutex_destroy(&client->lock);
+}
+EXPORT_SYMBOL(host1x_client_exit);
+
 /**
  * __host1x_client_register() - register a host1x client
  * @client: host1x client
@@ -716,16 +739,11 @@ EXPORT_SYMBOL(host1x_driver_unregister);
  * device and call host1x_device_init(), which will in turn call each client's
  * &host1x_client_ops.init implementation.
  */
-int __host1x_client_register(struct host1x_client *client,
-			     struct lock_class_key *key)
+int __host1x_client_register(struct host1x_client *client)
 {
 	struct host1x *host1x;
 	int err;
 
-	INIT_LIST_HEAD(&client->list);
-	__mutex_init(&client->lock, "host1x client lock", key);
-	client->usecount = 0;
-
 	mutex_lock(&devices_lock);
 
 	list_for_each_entry(host1x, &devices, list) {
diff --git a/include/linux/host1x.h b/include/linux/host1x.h
index 9eb77c87a83b..ed0005ce4285 100644
--- a/include/linux/host1x.h
+++ b/include/linux/host1x.h
@@ -320,12 +320,30 @@ static inline struct host1x_device *to_host1x_device(struct device *dev)
 int host1x_device_init(struct host1x_device *device);
 int host1x_device_exit(struct host1x_device *device);
 
-int __host1x_client_register(struct host1x_client *client,
-			     struct lock_class_key *key);
-#define host1x_client_register(class) \
-	({ \
-		static struct lock_class_key __key; \
-		__host1x_client_register(class, &__key); \
+void __host1x_client_init(struct host1x_client *client, struct lock_class_key *key);
+void host1x_client_exit(struct host1x_client *client);
+
+#define host1x_client_init(client)			\
+	({						\
+		static struct lock_class_key __key;	\
+		__host1x_client_init(client, &__key);	\
+	})
+
+int __host1x_client_register(struct host1x_client *client);
+
+/*
+ * Note that this wrapper calls __host1x_client_init() for compatibility
+ * with existing callers. Callers that want to separately initialize and
+ * register a host1x client must first initialize using either of the
+ * __host1x_client_init() or host1x_client_init() functions and then use
+ * the low-level __host1x_client_register() function to avoid the client
+ * getting reinitialized.
+ */
+#define host1x_client_register(client)			\
+	({						\
+		static struct lock_class_key __key;	\
+		__host1x_client_init(client, &__key);	\
+		__host1x_client_register(client);	\
 	})
 
 int host1x_client_unregister(struct host1x_client *client);
-- 
2.30.2

[PATCH AUTOSEL 5.12 16/49] drm/tegra: sor: Fully initialize SOR before registration

[Sasha Levin]

From: Thierry Reding <treding@nvidia.com>

[ Upstream commit 5dea42759bcef74b0802ea64b904409bc37f9045 ]

Before registering the SOR host1x client, make sure that it is fully
initialized. This avoids a potential race condition between the SOR's
probe and the host1x device initialization in cases where the SOR is
the final sub-device to register to a host1x instance.

Reported-by: Jonathan Hunter <jonathanh@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
Tested-by: Jon Hunter <jonathanh@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 drivers/gpu/drm/tegra/sor.c | 27 +++++++++++++--------------
 1 file changed, 13 insertions(+), 14 deletions(-)

diff --git a/drivers/gpu/drm/tegra/sor.c b/drivers/gpu/drm/tegra/sor.c
index 67a80dae1c00..32c83f2e386c 100644
--- a/drivers/gpu/drm/tegra/sor.c
+++ b/drivers/gpu/drm/tegra/sor.c
@@ -3922,17 +3922,10 @@ static int tegra_sor_probe(struct platform_device *pdev)
 	platform_set_drvdata(pdev, sor);
 	pm_runtime_enable(&pdev->dev);
 
-	INIT_LIST_HEAD(&sor->client.list);
+	host1x_client_init(&sor->client);
 	sor->client.ops = &sor_client_ops;
 	sor->client.dev = &pdev->dev;
 
-	err = host1x_client_register(&sor->client);
-	if (err < 0) {
-		dev_err(&pdev->dev, "failed to register host1x client: %d\n",
-			err);
-		goto rpm_disable;
-	}
-
 	/*
 	 * On Tegra210 and earlier, provide our own implementation for the
 	 * pad output clock.
@@ -3944,13 +3937,13 @@ static int tegra_sor_probe(struct platform_device *pdev)
 				      sor->index);
 		if (!name) {
 			err = -ENOMEM;
-			goto unregister;
+			goto uninit;
 		}
 
 		err = host1x_client_resume(&sor->client);
 		if (err < 0) {
 			dev_err(sor->dev, "failed to resume: %d\n", err);
-			goto unregister;
+			goto uninit;
 		}
 
 		sor->clk_pad = tegra_clk_sor_pad_register(sor, name);
@@ -3961,14 +3954,20 @@ static int tegra_sor_probe(struct platform_device *pdev)
 		err = PTR_ERR(sor->clk_pad);
 		dev_err(sor->dev, "failed to register SOR pad clock: %d\n",
 			err);
-		goto unregister;
+		goto uninit;
+	}
+
+	err = __host1x_client_register(&sor->client);
+	if (err < 0) {
+		dev_err(&pdev->dev, "failed to register host1x client: %d\n",
+			err);
+		goto uninit;
 	}
 
 	return 0;
 
-unregister:
-	host1x_client_unregister(&sor->client);
-rpm_disable:
+uninit:
+	host1x_client_exit(&sor->client);
 	pm_runtime_disable(&pdev->dev);
 remove:
 	tegra_output_remove(&sor->output);
-- 
2.30.2

[PATCH AUTOSEL 5.12 41/49] drm/amd/display: Allow bandwidth validation for 0 streams.

[Sasha Levin]

From: Bindu Ramamurthy <bindu.r@amd.com>

[ Upstream commit ba8e59773ae59818695d1e20b8939282da80ec8c ]

[Why]
Bandwidth calculations are triggered for non zero streams, and
in case of 0 streams, these calculations were skipped with
pstate status not being updated.

[How]
As the pstate status is applicable for non zero streams, check
added for allowing 0 streams inline with dcn internal bandwidth
validations.

Signed-off-by: Bindu Ramamurthy <bindu.r@amd.com>
Reviewed-by: Roman Li <Roman.Li@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_resource.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/amd/display/dc/dcn20/dcn20_resource.c b/drivers/gpu/drm/amd/display/dc/dcn20/dcn20_resource.c
index bfbc23b76cd5..3e3c898848bd 100644
--- a/drivers/gpu/drm/amd/display/dc/dcn20/dcn20_resource.c
+++ b/drivers/gpu/drm/amd/display/dc/dcn20/dcn20_resource.c
@@ -3231,7 +3231,7 @@ static noinline bool dcn20_validate_bandwidth_fp(struct dc *dc,
 	voltage_supported = dcn20_validate_bandwidth_internal(dc, context, false);
 	dummy_pstate_supported = context->bw_ctx.bw.dcn.clk.p_state_change_support;
 
-	if (voltage_supported && dummy_pstate_supported) {
+	if (voltage_supported && (dummy_pstate_supported || !(context->stream_count))) {
 		context->bw_ctx.bw.dcn.clk.p_state_change_support = false;
 		goto restore_dml_state;
 	}
-- 
2.30.2

[PATCH AUTOSEL 5.12 42/49] drm/amdgpu: refine amdgpu_fru_get_product_info

[Sasha Levin]

From: Jiansong Chen <Jiansong.Chen@amd.com>

[ Upstream commit 5cfc912582e13b05d71fb7acc4ec69ddfa9af320 ]

1. eliminate potential array index out of bounds.
2. return meaningful value for failure.

Signed-off-by: Jiansong Chen <Jiansong.Chen@amd.com>
Reviewed-by: Jack Gui <Jack.Gui@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 .../gpu/drm/amd/amdgpu/amdgpu_fru_eeprom.c    | 42 ++++++++++---------
 1 file changed, 23 insertions(+), 19 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_fru_eeprom.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_fru_eeprom.c
index 8f4a8f8d8146..39b6c6bfab45 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_fru_eeprom.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_fru_eeprom.c
@@ -101,7 +101,8 @@ static int amdgpu_fru_read_eeprom(struct amdgpu_device *adev, uint32_t addrptr,
 int amdgpu_fru_get_product_info(struct amdgpu_device *adev)
 {
 	unsigned char buff[34];
-	int addrptr = 0, size = 0;
+	int addrptr, size;
+	int len;
 
 	if (!is_fru_eeprom_supported(adev))
 		return 0;
@@ -109,7 +110,7 @@ int amdgpu_fru_get_product_info(struct amdgpu_device *adev)
 	/* If algo exists, it means that the i2c_adapter's initialized */
 	if (!adev->pm.smu_i2c.algo) {
 		DRM_WARN("Cannot access FRU, EEPROM accessor not initialized");
-		return 0;
+		return -ENODEV;
 	}
 
 	/* There's a lot of repetition here. This is due to the FRU having
@@ -128,7 +129,7 @@ int amdgpu_fru_get_product_info(struct amdgpu_device *adev)
 	size = amdgpu_fru_read_eeprom(adev, addrptr, buff);
 	if (size < 1) {
 		DRM_ERROR("Failed to read FRU Manufacturer, ret:%d", size);
-		return size;
+		return -EINVAL;
 	}
 
 	/* Increment the addrptr by the size of the field, and 1 due to the
@@ -138,43 +139,45 @@ int amdgpu_fru_get_product_info(struct amdgpu_device *adev)
 	size = amdgpu_fru_read_eeprom(adev, addrptr, buff);
 	if (size < 1) {
 		DRM_ERROR("Failed to read FRU product name, ret:%d", size);
-		return size;
+		return -EINVAL;
 	}
 
+	len = size;
 	/* Product name should only be 32 characters. Any more,
 	 * and something could be wrong. Cap it at 32 to be safe
 	 */
-	if (size > 32) {
+	if (len >= sizeof(adev->product_name)) {
 		DRM_WARN("FRU Product Number is larger than 32 characters. This is likely a mistake");
-		size = 32;
+		len = sizeof(adev->product_name) - 1;
 	}
 	/* Start at 2 due to buff using fields 0 and 1 for the address */
-	memcpy(adev->product_name, &buff[2], size);
-	adev->product_name[size] = '\0';
+	memcpy(adev->product_name, &buff[2], len);
+	adev->product_name[len] = '\0';
 
 	addrptr += size + 1;
 	size = amdgpu_fru_read_eeprom(adev, addrptr, buff);
 	if (size < 1) {
 		DRM_ERROR("Failed to read FRU product number, ret:%d", size);
-		return size;
+		return -EINVAL;
 	}
 
+	len = size;
 	/* Product number should only be 16 characters. Any more,
 	 * and something could be wrong. Cap it at 16 to be safe
 	 */
-	if (size > 16) {
+	if (len >= sizeof(adev->product_number)) {
 		DRM_WARN("FRU Product Number is larger than 16 characters. This is likely a mistake");
-		size = 16;
+		len = sizeof(adev->product_number) - 1;
 	}
-	memcpy(adev->product_number, &buff[2], size);
-	adev->product_number[size] = '\0';
+	memcpy(adev->product_number, &buff[2], len);
+	adev->product_number[len] = '\0';
 
 	addrptr += size + 1;
 	size = amdgpu_fru_read_eeprom(adev, addrptr, buff);
 
 	if (size < 1) {
 		DRM_ERROR("Failed to read FRU product version, ret:%d", size);
-		return size;
+		return -EINVAL;
 	}
 
 	addrptr += size + 1;
@@ -182,18 +185,19 @@ int amdgpu_fru_get_product_info(struct amdgpu_device *adev)
 
 	if (size < 1) {
 		DRM_ERROR("Failed to read FRU serial number, ret:%d", size);
-		return size;
+		return -EINVAL;
 	}
 
+	len = size;
 	/* Serial number should only be 16 characters. Any more,
 	 * and something could be wrong. Cap it at 16 to be safe
 	 */
-	if (size > 16) {
+	if (len >= sizeof(adev->serial)) {
 		DRM_WARN("FRU Serial Number is larger than 16 characters. This is likely a mistake");
-		size = 16;
+		len = sizeof(adev->serial) - 1;
 	}
-	memcpy(adev->serial, &buff[2], size);
-	adev->serial[size] = '\0';
+	memcpy(adev->serial, &buff[2], len);
+	adev->serial[len] = '\0';
 
 	return 0;
 }
-- 
2.30.2

[PATCH AUTOSEL 5.12 43/49] drm/amd/display: Fix overlay validation by considering cursors

[Sasha Levin]

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

[ Upstream commit 33f409e60eb0c59a4d0d06a62ab4642a988e17f7 ]

A few weeks ago, we saw a two cursor issue in a ChromeOS system. We
fixed it in the commit:

 drm/amd/display: Fix two cursor duplication when using overlay
 (read the commit message for more details)

After this change, we noticed that some IGT subtests related to
kms_plane and kms_plane_scaling started to fail. After investigating
this issue, we noticed that all subtests that fail have a primary plane
covering the overlay plane, which is currently rejected by amdgpu dm.
Fail those IGT tests highlight that our verification was too broad and
compromises the overlay usage in our drive. This patch fixes this issue
by ensuring that we only reject commits where the primary plane is not
fully covered by the overlay when the cursor hardware is enabled. With
this fix, all IGT tests start to pass again, which means our overlay
support works as expected.

Cc: Tianci.Yin <tianci.yin@amd.com>
Cc: Harry Wentland <harry.wentland@amd.com>
Cc: Nicholas Choi <nicholas.choi@amd.com>
Cc: Bhawanpreet Lakha <bhawanpreet.lakha@amd.com>
Cc: Nicholas Kazlauskas <Nicholas.Kazlauskas@amd.com>
Cc: Mark Yacoub <markyacoub@google.com>
Cc: Daniel Wheeler <daniel.wheeler@amd.com>

Tested-by: Daniel Wheeler <daniel.wheeler@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>
---
 drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c | 10 +++++++++-
 1 file changed, 9 insertions(+), 1 deletion(-)

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 b63f55ea8758..69023b4b0a8b 100644
--- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c
+++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c
@@ -9349,7 +9349,7 @@ static int validate_overlay(struct drm_atomic_state *state)
 	int i;
 	struct drm_plane *plane;
 	struct drm_plane_state *old_plane_state, *new_plane_state;
-	struct drm_plane_state *primary_state, *overlay_state = NULL;
+	struct drm_plane_state *primary_state, *cursor_state, *overlay_state = NULL;
 
 	/* Check if primary plane is contained inside overlay */
 	for_each_oldnew_plane_in_state_reverse(state, plane, old_plane_state, new_plane_state, i) {
@@ -9379,6 +9379,14 @@ static int validate_overlay(struct drm_atomic_state *state)
 	if (!primary_state->crtc)
 		return 0;
 
+	/* check if cursor plane is enabled */
+	cursor_state = drm_atomic_get_plane_state(state, overlay_state->crtc->cursor);
+	if (IS_ERR(cursor_state))
+		return PTR_ERR(cursor_state);
+
+	if (drm_atomic_plane_disabling(plane->state, cursor_state))
+		return 0;
+
 	/* Perform the bounds check to ensure the overlay plane covers the primary */
 	if (primary_state->crtc_x < overlay_state->crtc_x ||
 	    primary_state->crtc_y < overlay_state->crtc_y ||
-- 
2.30.2

[PATCH AUTOSEL 5.12 44/49] drm/amd/display: Fix potential memory leak in DMUB hw_init

[Sasha Levin]

From: Roman Li <roman.li@amd.com>

[ Upstream commit c5699e2d863f58221044efdc3fa712dd32d55cde ]

[Why]
On resume we perform DMUB hw_init which allocates memory:
dm_resume->dm_dmub_hw_init->dc_dmub_srv_create->kzalloc
That results in memory leak in suspend/resume scenarios.

[How]
Allocate memory for the DC wrapper to DMUB only if it was not
allocated before.
No need to reallocate it on suspend/resume.

Signed-off-by: Lang Yu <Lang.Yu@amd.com>
Signed-off-by: Roman Li <roman.li@amd.com>
Reviewed-by: Nicholas Kazlauskas <Nicholas.Kazlauskas@amd.com>
Acked-by: Qingqing Zhuo <qingqing.zhuo@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, 2 insertions(+), 2 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 69023b4b0a8b..95d5bc2da178 100644
--- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c
+++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c
@@ -871,7 +871,8 @@ static int dm_dmub_hw_init(struct amdgpu_device *adev)
 		abm->dmcu_is_running = dmcu->funcs->is_dmcu_initialized(dmcu);
 	}
 
-	adev->dm.dc->ctx->dmub_srv = dc_dmub_srv_create(adev->dm.dc, dmub_srv);
+	if (!adev->dm.dc->ctx->dmub_srv)
+		adev->dm.dc->ctx->dmub_srv = dc_dmub_srv_create(adev->dm.dc, dmub_srv);
 	if (!adev->dm.dc->ctx->dmub_srv) {
 		DRM_ERROR("Couldn't allocate DC DMUB server!\n");
 		return -ENOMEM;
@@ -1863,7 +1864,6 @@ static int dm_suspend(void *handle)
 
 	amdgpu_dm_irq_suspend(adev);
 
-
 	dc_set_power_state(dm->dc, DC_ACPI_CM_POWER_STATE_D3);
 
 	return 0;
-- 
2.30.2

[PATCH AUTOSEL 5.12 45/49] drm/amd/amdgpu:save psp ring wptr to avoid attack

[Sasha Levin]

From: Victor Zhao <Victor.Zhao@amd.com>

[ Upstream commit 2370eba9f552eaae3d8aa1f70b8e9eec5c560f9e ]

[Why]
When some tools performing psp mailbox attack, the readback value
of register can be a random value which may break psp.

[How]
Use a psp wptr cache machanism to aovid the change made by attack.

v2: unify change and add detailed reason

Signed-off-by: Victor Zhao <Victor.Zhao@amd.com>
Signed-off-by: Jingwen Chen <Jingwen.Chen2@amd.com>
Reviewed-by: Monk Liu <monk.liu@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_psp.h | 1 +
 drivers/gpu/drm/amd/amdgpu/psp_v11_0.c  | 3 ++-
 drivers/gpu/drm/amd/amdgpu/psp_v3_1.c   | 3 ++-
 3 files changed, 5 insertions(+), 2 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h
index cb50ba445f8c..0fd62a8e68c2 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h
@@ -76,6 +76,7 @@ struct psp_ring
 	uint64_t			ring_mem_mc_addr;
 	void				*ring_mem_handle;
 	uint32_t			ring_size;
+	uint32_t			ring_wptr;
 };
 
 /* More registers may will be supported */
diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v11_0.c b/drivers/gpu/drm/amd/amdgpu/psp_v11_0.c
index c325d6f53a71..d39735a89a25 100644
--- a/drivers/gpu/drm/amd/amdgpu/psp_v11_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/psp_v11_0.c
@@ -720,7 +720,7 @@ static uint32_t psp_v11_0_ring_get_wptr(struct psp_context *psp)
 	struct amdgpu_device *adev = psp->adev;
 
 	if (amdgpu_sriov_vf(adev))
-		data = RREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_102);
+		data = psp->km_ring.ring_wptr;
 	else
 		data = RREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_67);
 
@@ -734,6 +734,7 @@ static void psp_v11_0_ring_set_wptr(struct psp_context *psp, uint32_t value)
 	if (amdgpu_sriov_vf(adev)) {
 		WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_102, value);
 		WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_101, GFX_CTRL_CMD_ID_CONSUME_CMD);
+		psp->km_ring.ring_wptr = value;
 	} else
 		WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_67, value);
 }
diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v3_1.c b/drivers/gpu/drm/amd/amdgpu/psp_v3_1.c
index f2e725f72d2f..908664a5774b 100644
--- a/drivers/gpu/drm/amd/amdgpu/psp_v3_1.c
+++ b/drivers/gpu/drm/amd/amdgpu/psp_v3_1.c
@@ -379,7 +379,7 @@ static uint32_t psp_v3_1_ring_get_wptr(struct psp_context *psp)
 	struct amdgpu_device *adev = psp->adev;
 
 	if (amdgpu_sriov_vf(adev))
-		data = RREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_102);
+		data = psp->km_ring.ring_wptr;
 	else
 		data = RREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_67);
 	return data;
@@ -394,6 +394,7 @@ static void psp_v3_1_ring_set_wptr(struct psp_context *psp, uint32_t value)
 		/* send interrupt to PSP for SRIOV ring write pointer update */
 		WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_101,
 			GFX_CTRL_CMD_ID_CONSUME_CMD);
+		psp->km_ring.ring_wptr = value;
 	} else
 		WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_67, value);
 }
-- 
2.30.2