Skip to contentSkip to conversation

drm: Add comm/cmdline fdinfo fields

RFC In review Rob Clark 5 participants 18 messages Apr 17, 2023 – Apr 27, 2023

Subsystems (6)
lore
Share & tools
lore

Rob Clark

author
#0 lore raw

cover drm: Add comm/cmdline fdinfo fields

From: Rob Clark <[email protected]>

When many of the things using the GPU are processes in a VM guest, the actual client process is just a proxy. The msm driver has a way to let the proxy tell the kernel the actual VM client process's executable name and command-line, which has until now been used simply for GPU crash devcore dumps. Lets also expose this via fdinfo so that tools can expose who the actual user of the GPU is.

Rob Clark (3):
drm/doc: Relax fdinfo string constraints
drm/msm: Rework get_comm_cmdline() helper
drm/msm: Add comm/cmdline fields

 Documentation/gpu/drm-usage-stats.rst   | 37 +++++++++++++++----------
 drivers/gpu/drm/msm/adreno/adreno_gpu.c |  4 +--
 drivers/gpu/drm/msm/msm_drv.c           |  2 ++
 drivers/gpu/drm/msm/msm_gpu.c           | 27 +++++++++++++-----
 drivers/gpu/drm/msm/msm_gpu.h           | 12 ++++++--
 drivers/gpu/drm/msm/msm_submitqueue.c   |  1 +
 6 files changed, 58 insertions(+), 25 deletions(-)
Signature · 1 line
2.39.2
author
#1 lore raw

RFC 1/3 drm/doc: Relax fdinfo string constraints

From: Rob Clark <[email protected]>

The restriction about no whitespace, etc, really only applies to the usage of strings in keys. Values can contain anything (other than newline).

---
 Documentation/gpu/drm-usage-stats.rst | 29 ++++++++++++++-------------
 1 file changed, 15 insertions(+), 14 deletions(-)
Documentation/gpu/drm-usage-stats.rst +15 -14
index 258bdcc8fb86..8e00d53231e0 100644
@@ -24,7 +24,7 @@ File format specification
 - All keys shall be prefixed with `drm-`.
 - Whitespace between the delimiter and first non-whitespace character shall be
   ignored when parsing.
-- Neither keys or values are allowed to contain whitespace characters.
+- Keys are not allowed to contain whitespace characters.
 - Numerical key value pairs can end with optional unit string.
 - Data type of the value is fixed as defined in the specification.
 
@@ -39,12 +39,13 @@ Data types
 ----------
 
 - <uint> - Unsigned integer without defining the maximum value.
-- <str> - String excluding any above defined reserved characters or whitespace.
+- <keystr> - String excluding any above defined reserved characters or whitespace.
+- <valstr> - String.
 
 Mandatory fully standardised keys
 ---------------------------------
 
-- drm-driver: <str>
+- drm-driver: <valstr>
 
 String shall contain the name this driver registered as via the respective
 `struct drm_driver` data structure.
@@ -69,10 +70,10 @@ scope of each device, in which case `drm-pdev` shall be present as well.
 Userspace should make sure to not double account any usage statistics by using
 the above described criteria in order to associate data to individual clients.
 
-- drm-engine-<str>: <uint> ns
+- drm-engine-<keystr>: <uint> ns
 
 GPUs usually contain multiple execution engines. Each shall be given a stable
-and unique name (str), with possible values documented in the driver specific
+and unique name (keystr), with possible values documented in the driver specific
 documentation.
 
 Value shall be in specified time units which the respective GPU engine spent
@@ -84,16 +85,16 @@ larger value within a reasonable period. Upon observing a value lower than what
 was previously read, userspace is expected to stay with that larger previous
 value until a monotonic update is seen.
 
-- drm-engine-capacity-<str>: <uint>
+- drm-engine-capacity-<keystr>: <uint>
 
 Engine identifier string must be the same as the one specified in the
-drm-engine-<str> tag and shall contain a greater than zero number in case the
+drm-engine-<keystr> tag and shall contain a greater than zero number in case the
 exported engine corresponds to a group of identical hardware engines.
 
 In the absence of this tag parser shall assume capacity of one. Zero capacity
 is not allowed.
 
-- drm-memory-<str>: <uint> [KiB|MiB]
+- drm-memory-<keystr>: <uint> [KiB|MiB]
 
 Each possible memory type which can be used to store buffer objects by the
 GPU in question shall be given a stable and unique name to be returned as the
@@ -126,10 +127,10 @@ The total size of buffers that are purgeable.
 
 The total size of buffers that are active on one or more rings.
 
-- drm-cycles-<str>: <uint>
+- drm-cycles-<keystr>: <uint>
 
 Engine identifier string must be the same as the one specified in the
-drm-engine-<str> tag and shall contain the number of busy cycles for the given
+drm-engine-<keystr> tag and shall contain the number of busy cycles for the given
 engine.
 
 Values are not required to be constantly monotonic if it makes the driver
@@ -138,12 +139,12 @@ larger value within a reasonable period. Upon observing a value lower than what
 was previously read, userspace is expected to stay with that larger previous
 value until a monotonic update is seen.
 
-- drm-maxfreq-<str>: <uint> [Hz|MHz|KHz]
+- drm-maxfreq-<keystr>: <uint> [Hz|MHz|KHz]
 
 Engine identifier string must be the same as the one specified in the
-drm-engine-<str> tag and shall contain the maximum frequency for the given
-engine.  Taken together with drm-cycles-<str>, this can be used to calculate
-percentage utilization of the engine, whereas drm-engine-<str> only reflects
+drm-engine-<keystr> tag and shall contain the maximum frequency for the given
+engine.  Taken together with drm-cycles-<keystr>, this can be used to calculate
+percentage utilization of the engine, whereas drm-engine-<keystr> only reflects
 time active without considering what frequency the engine is operating as a
 percentage of it's maximum frequency.
 
Signature · 1 line
2.39.2
Acked-by on v1 1/3
#2 lore raw

Re: [RFC 1/3] drm/doc: Relax fdinfo string constraints

Quoted 25 lines · Rob Clark

From: Rob Clark <[email protected]>

The restriction about no whitespace, etc, really only applies to the usage of strings in keys. Values can contain anything (other than newline).

---
  Documentation/gpu/drm-usage-stats.rst | 29 ++++++++++++++-------------
  1 file changed, 15 insertions(+), 14 deletions(-)
Documentation/gpu/drm-usage-stats.rst +1 -1
index 258bdcc8fb86..8e00d53231e0 100644
@@ -24,7 +24,7 @@ File format specification
  - All keys shall be prefixed with `drm-`.
  - Whitespace between the delimiter and first non-whitespace character shall be
    ignored when parsing.
-- Neither keys or values are allowed to contain whitespace characters.
+- Keys are not allowed to contain whitespace characters.
  - Numerical key value pairs can end with optional unit string.
  - Data type of the value is fixed as defined in the specification.
 

Documentation/gpu/drm-usage-stats.rst Data types · line 43

Quoted 6 lines
Documentation/gpu/drm-usage-stats.rst +2 -1
@@ -39,12 +39,13 @@ Data types
  ----------
 
  - <uint> - Unsigned integer without defining the maximum value.
-- <str> - String excluding any above defined reserved characters or whitespace.
+- <keystr> - String excluding any above defined reserved characters or whitespace.
+- <valstr> - String.

Makes sense I think. At least I can't remember that I had special reason to word it as strict as it was. Lets give it some time to marinade so for later:

Regards,

Tvrtko

Quoted 72 lines
Documentation/gpu/drm-usage-stats.rst +12 -12
  Mandatory fully standardised keys
  ---------------------------------
 
-- drm-driver: <str>
+- drm-driver: <valstr>
 
  String shall contain the name this driver registered as via the respective
  `struct drm_driver` data structure.
@@ -69,10 +70,10 @@ scope of each device, in which case `drm-pdev` shall be present as well.
  Userspace should make sure to not double account any usage statistics by using
  the above described criteria in order to associate data to individual clients.
 
-- drm-engine-<str>: <uint> ns
+- drm-engine-<keystr>: <uint> ns
 
  GPUs usually contain multiple execution engines. Each shall be given a stable
-and unique name (str), with possible values documented in the driver specific
+and unique name (keystr), with possible values documented in the driver specific
  documentation.
 
  Value shall be in specified time units which the respective GPU engine spent
@@ -84,16 +85,16 @@ larger value within a reasonable period. Upon observing a value lower than what
  was previously read, userspace is expected to stay with that larger previous
  value until a monotonic update is seen.
 
-- drm-engine-capacity-<str>: <uint>
+- drm-engine-capacity-<keystr>: <uint>
 
  Engine identifier string must be the same as the one specified in the
-drm-engine-<str> tag and shall contain a greater than zero number in case the
+drm-engine-<keystr> tag and shall contain a greater than zero number in case the
  exported engine corresponds to a group of identical hardware engines.
 
  In the absence of this tag parser shall assume capacity of one. Zero capacity
  is not allowed.
 
-- drm-memory-<str>: <uint> [KiB|MiB]
+- drm-memory-<keystr>: <uint> [KiB|MiB]
 
  Each possible memory type which can be used to store buffer objects by the
  GPU in question shall be given a stable and unique name to be returned as the
@@ -126,10 +127,10 @@ The total size of buffers that are purgeable.
 
  The total size of buffers that are active on one or more rings.
 
-- drm-cycles-<str>: <uint>
+- drm-cycles-<keystr>: <uint>
 
  Engine identifier string must be the same as the one specified in the
-drm-engine-<str> tag and shall contain the number of busy cycles for the given
+drm-engine-<keystr> tag and shall contain the number of busy cycles for the given
  engine.
 
  Values are not required to be constantly monotonic if it makes the driver
@@ -138,12 +139,12 @@ larger value within a reasonable period. Upon observing a value lower than what
  was previously read, userspace is expected to stay with that larger previous
  value until a monotonic update is seen.
 
-- drm-maxfreq-<str>: <uint> [Hz|MHz|KHz]
+- drm-maxfreq-<keystr>: <uint> [Hz|MHz|KHz]
 
  Engine identifier string must be the same as the one specified in the
-drm-engine-<str> tag and shall contain the maximum frequency for the given
-engine.  Taken together with drm-cycles-<str>, this can be used to calculate
-percentage utilization of the engine, whereas drm-engine-<str> only reflects
+drm-engine-<keystr> tag and shall contain the maximum frequency for the given
+engine.  Taken together with drm-cycles-<keystr>, this can be used to calculate
+percentage utilization of the engine, whereas drm-engine-<keystr> only reflects
  time active without considering what frequency the engine is operating as a
  percentage of it's maximum frequency.
 
author
#3 lore raw

RFC 2/3 drm/msm: Rework get_comm_cmdline() helper

In reply to Rob Clark

From: Rob Clark <[email protected]>

Make it work in terms of ctx so that it can be re-used for fdinfo.

---
 drivers/gpu/drm/msm/adreno/adreno_gpu.c |  4 ++--
 drivers/gpu/drm/msm/msm_drv.c           |  2 ++
 drivers/gpu/drm/msm/msm_gpu.c           | 13 ++++++-------
 drivers/gpu/drm/msm/msm_gpu.h           | 12 ++++++++++--
 drivers/gpu/drm/msm/msm_submitqueue.c   |  1 +
 5 files changed, 21 insertions(+), 11 deletions(-)
drivers/gpu/drm/msm/adreno/adreno_gpu.c +2 -2
index bb38e728864d..43c4e1fea83f 100644
@@ -412,7 +412,7 @@ int adreno_set_param(struct msm_gpu *gpu, struct msm_file_private *ctx,
 		/* Ensure string is null terminated: */
 		str[len] = '\0';
 
-		mutex_lock(&gpu->lock);
+		mutex_lock(&ctx->lock);
 
 		if (param == MSM_PARAM_COMM) {
 			paramp = &ctx->comm;
@@ -423,7 +423,7 @@ int adreno_set_param(struct msm_gpu *gpu, struct msm_file_private *ctx,
 		kfree(*paramp);
 		*paramp = str;
 
-		mutex_unlock(&gpu->lock);
+		mutex_unlock(&ctx->lock);
 
 		return 0;
 	}
drivers/gpu/drm/msm/msm_drv.c +2 -0
index 3d73b98d6a9c..ca0e89e46e13 100644
@@ -581,6 +581,8 @@ static int context_init(struct drm_device *dev, struct drm_file *file)
 	rwlock_init(&ctx->queuelock);
 
 	kref_init(&ctx->ref);
+	ctx->pid = get_pid(task_pid(current));
+	mutex_init(&ctx->lock);
 	msm_submitqueue_init(dev, ctx);
 
 	ctx->aspace = msm_gpu_create_private_address_space(priv->gpu, current);
drivers/gpu/drm/msm/msm_gpu.c +6 -7
index c403912d13ab..f0f4f845c32d 100644
@@ -327,18 +327,17 @@ find_submit(struct msm_ringbuffer *ring, uint32_t fence)
 
 static void retire_submits(struct msm_gpu *gpu);
 
-static void get_comm_cmdline(struct msm_gem_submit *submit, char **comm, char **cmd)
+static void get_comm_cmdline(struct msm_file_private *ctx, char **comm, char **cmd)
 {
-	struct msm_file_private *ctx = submit->queue->ctx;
 	struct task_struct *task;
 
-	WARN_ON(!mutex_is_locked(&submit->gpu->lock));
-
 	/* Note that kstrdup will return NULL if argument is NULL: */
+	mutex_lock(&ctx->lock);
 	*comm = kstrdup(ctx->comm, GFP_KERNEL);
 	*cmd  = kstrdup(ctx->cmdline, GFP_KERNEL);
+	mutex_unlock(&ctx->lock);
 
-	task = get_pid_task(submit->pid, PIDTYPE_PID);
+	task = get_pid_task(ctx->pid, PIDTYPE_PID);
 	if (!task)
 		return;
 
@@ -372,7 +371,7 @@ static void recover_worker(struct kthread_work *work)
 		if (submit->aspace)
 			submit->aspace->faults++;
 
-		get_comm_cmdline(submit, &comm, &cmd);
+		get_comm_cmdline(submit->queue->ctx, &comm, &cmd);
 
 		if (comm && cmd) {
 			DRM_DEV_ERROR(dev->dev, "%s: offending task: %s (%s)\n",
@@ -460,7 +459,7 @@ static void fault_worker(struct kthread_work *work)
 		goto resume_smmu;
 
 	if (submit) {
-		get_comm_cmdline(submit, &comm, &cmd);
+		get_comm_cmdline(submit->queue->ctx, &comm, &cmd);
 
 		/*
 		 * When we get GPU iova faults, we can get 1000s of them,
drivers/gpu/drm/msm/msm_gpu.h +10 -2
index 7a4fa1b8655b..b2023a42116b 100644
@@ -377,17 +377,25 @@ struct msm_file_private {
 	 */
 	int sysprof;
 
+	/** @pid: Process that opened this file. */
+	struct pid *pid;
+
+	/**
+	 * lock: Protects comm and cmdline
+	 */
+	struct mutex lock;
+
 	/**
 	 * comm: Overridden task comm, see MSM_PARAM_COMM
 	 *
-	 * Accessed under msm_gpu::lock
+	 * Accessed under msm_file_private::lock
 	 */
 	char *comm;
 
 	/**
 	 * cmdline: Overridden task cmdline, see MSM_PARAM_CMDLINE
 	 *
-	 * Accessed under msm_gpu::lock
+	 * Accessed under msm_file_private::lock
 	 */
 	char *cmdline;
 
drivers/gpu/drm/msm/msm_submitqueue.c +1 -0
index 0e803125a325..0444ba04fa06 100644
@@ -61,6 +61,7 @@ void __msm_file_private_destroy(struct kref *kref)
 	}
 
 	msm_gem_address_space_put(ctx->aspace);
+	put_pid(ctx->pid);
 	kfree(ctx->comm);
 	kfree(ctx->cmdline);
 	kfree(ctx);
Signature · 1 line
2.39.2
#4 lore raw

Re: [RFC 2/3] drm/msm: Rework get_comm_cmdline() helper

Quoted 40 lines · Rob Clark

From: Rob Clark <[email protected]>

Make it work in terms of ctx so that it can be re-used for fdinfo.

---
  drivers/gpu/drm/msm/adreno/adreno_gpu.c |  4 ++--
  drivers/gpu/drm/msm/msm_drv.c           |  2 ++
  drivers/gpu/drm/msm/msm_gpu.c           | 13 ++++++-------
  drivers/gpu/drm/msm/msm_gpu.h           | 12 ++++++++++--
  drivers/gpu/drm/msm/msm_submitqueue.c   |  1 +
  5 files changed, 21 insertions(+), 11 deletions(-)
drivers/gpu/drm/msm/adreno/adreno_gpu.c +2 -2
index bb38e728864d..43c4e1fea83f 100644
@@ -412,7 +412,7 @@ int adreno_set_param(struct msm_gpu *gpu, struct msm_file_private *ctx,
  		/* Ensure string is null terminated: */
  		str[len] = '\0';
 
-		mutex_lock(&gpu->lock);
+		mutex_lock(&ctx->lock);
 
  		if (param == MSM_PARAM_COMM) {
  			paramp = &ctx->comm;
@@ -423,7 +423,7 @@ int adreno_set_param(struct msm_gpu *gpu, struct msm_file_private *ctx,
  		kfree(*paramp);
  		*paramp = str;
 
-		mutex_unlock(&gpu->lock);
+		mutex_unlock(&ctx->lock);
 
  		return 0;
  	}
drivers/gpu/drm/msm/msm_drv.c +0 -0
index 3d73b98d6a9c..ca0e89e46e13 100644

drivers/gpu/drm/msm/msm_drv.c static int context_init(struct drm_device *dev, struct drm_file *file) · line 584

Quoted 4 lines
drivers/gpu/drm/msm/msm_drv.c +1 -0
@@ -581,6 +581,8 @@ static int context_init(struct drm_device *dev, struct drm_file *file)
  	rwlock_init(&ctx->queuelock);
 
  	kref_init(&ctx->ref);
+	ctx->pid = get_pid(task_pid(current));

Would it simplify things for msm if DRM core had an up to date file->pid as proposed in https://patchwork.freedesktop.org/patch/526752/?series=109902&rev=4 ? It gets updated if ioctl issuer is different than fd opener and this being context_init here reminded me of it. Maybe you wouldn't have to track the pid in msm?

Regards,

Tvrtko

Quoted 93 lines
drivers/gpu/drm/msm/msm_drv.c +1 -0
+	mutex_init(&ctx->lock);
  	msm_submitqueue_init(dev, ctx);
 
  	ctx->aspace = msm_gpu_create_private_address_space(priv->gpu, current);
drivers/gpu/drm/msm/msm_gpu.c +6 -7
index c403912d13ab..f0f4f845c32d 100644
@@ -327,18 +327,17 @@ find_submit(struct msm_ringbuffer *ring, uint32_t fence)
 
  static void retire_submits(struct msm_gpu *gpu);
 
-static void get_comm_cmdline(struct msm_gem_submit *submit, char **comm, char **cmd)
+static void get_comm_cmdline(struct msm_file_private *ctx, char **comm, char **cmd)
  {
-	struct msm_file_private *ctx = submit->queue->ctx;
  	struct task_struct *task;
 
-	WARN_ON(!mutex_is_locked(&submit->gpu->lock));
-
  	/* Note that kstrdup will return NULL if argument is NULL: */
+	mutex_lock(&ctx->lock);
  	*comm = kstrdup(ctx->comm, GFP_KERNEL);
  	*cmd  = kstrdup(ctx->cmdline, GFP_KERNEL);
+	mutex_unlock(&ctx->lock);
 
-	task = get_pid_task(submit->pid, PIDTYPE_PID);
+	task = get_pid_task(ctx->pid, PIDTYPE_PID);
  	if (!task)
  		return;
 
@@ -372,7 +371,7 @@ static void recover_worker(struct kthread_work *work)
  		if (submit->aspace)
  			submit->aspace->faults++;
 
-		get_comm_cmdline(submit, &comm, &cmd);
+		get_comm_cmdline(submit->queue->ctx, &comm, &cmd);
 
  		if (comm && cmd) {
  			DRM_DEV_ERROR(dev->dev, "%s: offending task: %s (%s)\n",
@@ -460,7 +459,7 @@ static void fault_worker(struct kthread_work *work)
  		goto resume_smmu;
 
  	if (submit) {
-		get_comm_cmdline(submit, &comm, &cmd);
+		get_comm_cmdline(submit->queue->ctx, &comm, &cmd);
 
  		/*
  		 * When we get GPU iova faults, we can get 1000s of them,
drivers/gpu/drm/msm/msm_gpu.h +10 -2
index 7a4fa1b8655b..b2023a42116b 100644
@@ -377,17 +377,25 @@ struct msm_file_private {
  	 */
  	int sysprof;
 
+	/** @pid: Process that opened this file. */
+	struct pid *pid;
+
+	/**
+	 * lock: Protects comm and cmdline
+	 */
+	struct mutex lock;
+
  	/**
  	 * comm: Overridden task comm, see MSM_PARAM_COMM
  	 *
-	 * Accessed under msm_gpu::lock
+	 * Accessed under msm_file_private::lock
  	 */
  	char *comm;
 
  	/**
  	 * cmdline: Overridden task cmdline, see MSM_PARAM_CMDLINE
  	 *
-	 * Accessed under msm_gpu::lock
+	 * Accessed under msm_file_private::lock
  	 */
  	char *cmdline;
 
drivers/gpu/drm/msm/msm_submitqueue.c +1 -0
index 0e803125a325..0444ba04fa06 100644
@@ -61,6 +61,7 @@ void __msm_file_private_destroy(struct kref *kref)
  	}
 
  	msm_gem_address_space_put(ctx->aspace);
+	put_pid(ctx->pid);
  	kfree(ctx->comm);
  	kfree(ctx->cmdline);
  	kfree(ctx);
#5 lore raw

Re: [RFC 2/3] drm/msm: Rework get_comm_cmdline() helper

Quoted 40 lines · Tvrtko Ursulin
Quoted 38 lines · Rob Clark

From: Rob Clark <[email protected]>

Make it work in terms of ctx so that it can be re-used for fdinfo.

---
  drivers/gpu/drm/msm/adreno/adreno_gpu.c |  4 ++--
  drivers/gpu/drm/msm/msm_drv.c           |  2 ++
  drivers/gpu/drm/msm/msm_gpu.c           | 13 ++++++-------
  drivers/gpu/drm/msm/msm_gpu.h           | 12 ++++++++++--
  drivers/gpu/drm/msm/msm_submitqueue.c   |  1 +
  5 files changed, 21 insertions(+), 11 deletions(-)
drivers/gpu/drm/msm/adreno/adreno_gpu.c +2 -2
index bb38e728864d..43c4e1fea83f 100644
@@ -412,7 +412,7 @@ int adreno_set_param(struct msm_gpu *gpu, struct msm_file_private *ctx,
  		/* Ensure string is null terminated: */
  		str[len] = '\0';
-		mutex_lock(&gpu->lock);
+		mutex_lock(&ctx->lock);
  		if (param == MSM_PARAM_COMM) {
  			paramp = &ctx->comm;
@@ -423,7 +423,7 @@ int adreno_set_param(struct msm_gpu *gpu, struct msm_file_private *ctx,
  		kfree(*paramp);
  		*paramp = str;
-		mutex_unlock(&gpu->lock);
+		mutex_unlock(&ctx->lock);
  		return 0;
  	}
drivers/gpu/drm/msm/msm_drv.c +0 -0
index 3d73b98d6a9c..ca0e89e46e13 100644
@@ -581,6 +581,8 @@ static int context_init(struct drm_device *dev, struct drm_file *file)
  	rwlock_init(&ctx->queuelock);
  	kref_init(&ctx->ref);
Quoted 8 lines

drivers/gpu/drm/msm/msm_drv.c static int context_init(struct drm_device *dev, struct drm_file *file) · line 584

Quoted 1 line
drivers/gpu/drm/msm/msm_drv.c +1 -0
+	ctx->pid = get_pid(task_pid(current));

Would it simplify things for msm if DRM core had an up to date file->pid as proposed in
https://patchwork.freedesktop.org/patch/526752/?series=109902&rev=4 ? It gets updated if ioctl issuer is different than fd opener and this being context_init here reminded me of it. Maybe you wouldn't have to track the pid in msm?

Can we go one step further and let the drm fdinfo stuff print these new additions? Consistency across drivers and all that.

Also for a generic trigger I think any driver ioctl is good enough (we only really need to avoid the auth dance when you're not on a render node).
-Daniel

Quoted 84 lines

Regards,

Tvrtko

Quoted 79 lines
drivers/gpu/drm/msm/msm_drv.c +1 -0
+	mutex_init(&ctx->lock);
  	msm_submitqueue_init(dev, ctx);
  	ctx->aspace = msm_gpu_create_private_address_space(priv->gpu, current);
drivers/gpu/drm/msm/msm_gpu.c +6 -7
index c403912d13ab..f0f4f845c32d 100644
@@ -327,18 +327,17 @@ find_submit(struct msm_ringbuffer *ring, uint32_t fence)
  static void retire_submits(struct msm_gpu *gpu);
-static void get_comm_cmdline(struct msm_gem_submit *submit, char **comm, char **cmd)
+static void get_comm_cmdline(struct msm_file_private *ctx, char **comm, char **cmd)
  {
-	struct msm_file_private *ctx = submit->queue->ctx;
  	struct task_struct *task;
-	WARN_ON(!mutex_is_locked(&submit->gpu->lock));
-
  	/* Note that kstrdup will return NULL if argument is NULL: */
+	mutex_lock(&ctx->lock);
  	*comm = kstrdup(ctx->comm, GFP_KERNEL);
  	*cmd  = kstrdup(ctx->cmdline, GFP_KERNEL);
+	mutex_unlock(&ctx->lock);
-	task = get_pid_task(submit->pid, PIDTYPE_PID);
+	task = get_pid_task(ctx->pid, PIDTYPE_PID);
  	if (!task)
  		return;
@@ -372,7 +371,7 @@ static void recover_worker(struct kthread_work *work)
  		if (submit->aspace)
  			submit->aspace->faults++;
-		get_comm_cmdline(submit, &comm, &cmd);
+		get_comm_cmdline(submit->queue->ctx, &comm, &cmd);
  		if (comm && cmd) {
  			DRM_DEV_ERROR(dev->dev, "%s: offending task: %s (%s)\n",
@@ -460,7 +459,7 @@ static void fault_worker(struct kthread_work *work)
  		goto resume_smmu;
  	if (submit) {
-		get_comm_cmdline(submit, &comm, &cmd);
+		get_comm_cmdline(submit->queue->ctx, &comm, &cmd);
  		/*
  		 * When we get GPU iova faults, we can get 1000s of them,
drivers/gpu/drm/msm/msm_gpu.h +10 -2
index 7a4fa1b8655b..b2023a42116b 100644
@@ -377,17 +377,25 @@ struct msm_file_private {
  	 */
  	int sysprof;
+	/** @pid: Process that opened this file. */
+	struct pid *pid;
+
+	/**
+	 * lock: Protects comm and cmdline
+	 */
+	struct mutex lock;
+
  	/**
  	 * comm: Overridden task comm, see MSM_PARAM_COMM
  	 *
-	 * Accessed under msm_gpu::lock
+	 * Accessed under msm_file_private::lock
  	 */
  	char *comm;
  	/**
  	 * cmdline: Overridden task cmdline, see MSM_PARAM_CMDLINE
  	 *
-	 * Accessed under msm_gpu::lock
+	 * Accessed under msm_file_private::lock
  	 */
  	char *cmdline;
drivers/gpu/drm/msm/msm_submitqueue.c +1 -0
index 0e803125a325..0444ba04fa06 100644
@@ -61,6 +61,7 @@ void __msm_file_private_destroy(struct kref *kref)
  	}
  	msm_gem_address_space_put(ctx->aspace);
+	put_pid(ctx->pid);
  	kfree(ctx->comm);
  	kfree(ctx->cmdline);
  	kfree(ctx);
Signature · 3 lines
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
author
#6 lore raw

Re: [RFC 2/3] drm/msm: Rework get_comm_cmdline() helper

Quoted 42 lines · Daniel Vetter
Quoted 40 lines · Tvrtko Ursulin
Quoted 38 lines · Rob Clark

From: Rob Clark <[email protected]>

Make it work in terms of ctx so that it can be re-used for fdinfo.

---
  drivers/gpu/drm/msm/adreno/adreno_gpu.c |  4 ++--
  drivers/gpu/drm/msm/msm_drv.c           |  2 ++
  drivers/gpu/drm/msm/msm_gpu.c           | 13 ++++++-------
  drivers/gpu/drm/msm/msm_gpu.h           | 12 ++++++++++--
  drivers/gpu/drm/msm/msm_submitqueue.c   |  1 +
  5 files changed, 21 insertions(+), 11 deletions(-)
drivers/gpu/drm/msm/adreno/adreno_gpu.c +2 -2
index bb38e728864d..43c4e1fea83f 100644
@@ -412,7 +412,7 @@ int adreno_set_param(struct msm_gpu *gpu, struct msm_file_private *ctx,
            /* Ensure string is null terminated: */
            str[len] = '\0';
-           mutex_lock(&gpu->lock);
+           mutex_lock(&ctx->lock);
            if (param == MSM_PARAM_COMM) {
                    paramp = &ctx->comm;
@@ -423,7 +423,7 @@ int adreno_set_param(struct msm_gpu *gpu, struct msm_file_private *ctx,
            kfree(*paramp);
            *paramp = str;
-           mutex_unlock(&gpu->lock);
+           mutex_unlock(&ctx->lock);
            return 0;
    }
drivers/gpu/drm/msm/msm_drv.c +0 -0
index 3d73b98d6a9c..ca0e89e46e13 100644
@@ -581,6 +581,8 @@ static int context_init(struct drm_device *dev, struct drm_file *file)
    rwlock_init(&ctx->queuelock);
    kref_init(&ctx->ref);
Quoted 8 lines
Quoted 8 lines

drivers/gpu/drm/msm/msm_drv.c static int context_init(struct drm_device *dev, struct drm_file *file) · line 584

Quoted 1 line
drivers/gpu/drm/msm/msm_drv.c +1 -0
+   ctx->pid = get_pid(task_pid(current));

Would it simplify things for msm if DRM core had an up to date file->pid as proposed in
https://patchwork.freedesktop.org/patch/526752/?series=109902&rev=4 ? It gets updated if ioctl issuer is different than fd opener and this being context_init here reminded me of it. Maybe you wouldn't have to track the pid in msm?

The problem is that we also need this for gpu devcore dumps, which could happen after the drm_file is closed. The ctx can outlive the file.

But the ctx->pid has the same problem as the existing file->pid when it comes to Xorg.. hopefully over time that problem just goes away. I guess I could do a similar dance to your patch to update the pid whenever (for ex) a submitqueue is created.

Quoted 2 lines

Can we go one step further and let the drm fdinfo stuff print these new additions? Consistency across drivers and all that.

Hmm, I guess I could _also_ store the overridden comm/cmdline in drm_file. I still need to track it in ctx (msm_file_private) because I could need it after the file is closed.

Maybe it could be useful to have a gl extension to let the app set a name on the context so that this is useful beyond native-ctx (ie. maybe it would be nice to see that "chrome: lwn.net" is using less gpu memory than "chrome: phoronix.com", etc)

BR,
-R

Quoted 94 lines

Also for a generic trigger I think any driver ioctl is good enough (we only really need to avoid the auth dance when you're not on a render node).
-Daniel

Quoted 76 lines

Regards,

Tvrtko

Quoted 71 lines
drivers/gpu/drm/msm/msm_drv.c +1 -0
+   mutex_init(&ctx->lock);
    msm_submitqueue_init(dev, ctx);
    ctx->aspace = msm_gpu_create_private_address_space(priv->gpu, current);
drivers/gpu/drm/msm/msm_gpu.c +6 -7
index c403912d13ab..f0f4f845c32d 100644
@@ -327,18 +327,17 @@ find_submit(struct msm_ringbuffer *ring, uint32_t fence)
  static void retire_submits(struct msm_gpu *gpu);
-static void get_comm_cmdline(struct msm_gem_submit *submit, char **comm, char **cmd)
+static void get_comm_cmdline(struct msm_file_private *ctx, char **comm, char **cmd)
  {
-   struct msm_file_private *ctx = submit->queue->ctx;
    struct task_struct *task;
-   WARN_ON(!mutex_is_locked(&submit->gpu->lock));
-
    /* Note that kstrdup will return NULL if argument is NULL: */
+   mutex_lock(&ctx->lock);
    *comm = kstrdup(ctx->comm, GFP_KERNEL);
    *cmd  = kstrdup(ctx->cmdline, GFP_KERNEL);
+   mutex_unlock(&ctx->lock);
-   task = get_pid_task(submit->pid, PIDTYPE_PID);
+   task = get_pid_task(ctx->pid, PIDTYPE_PID);
    if (!task)
            return;
@@ -372,7 +371,7 @@ static void recover_worker(struct kthread_work *work)
            if (submit->aspace)
                    submit->aspace->faults++;
-           get_comm_cmdline(submit, &comm, &cmd);
+           get_comm_cmdline(submit->queue->ctx, &comm, &cmd);
            if (comm && cmd) {
                    DRM_DEV_ERROR(dev->dev, "%s: offending task: %s (%s)\n",
@@ -460,7 +459,7 @@ static void fault_worker(struct kthread_work *work)
            goto resume_smmu;
    if (submit) {
-           get_comm_cmdline(submit, &comm, &cmd);
+           get_comm_cmdline(submit->queue->ctx, &comm, &cmd);
            /*
             * When we get GPU iova faults, we can get 1000s of them,
drivers/gpu/drm/msm/msm_gpu.h +10 -2
index 7a4fa1b8655b..b2023a42116b 100644
@@ -377,17 +377,25 @@ struct msm_file_private {
     */
    int sysprof;
+   /** @pid: Process that opened this file. */
+   struct pid *pid;
+
+   /**
+    * lock: Protects comm and cmdline
+    */
+   struct mutex lock;
+
    /**
     * comm: Overridden task comm, see MSM_PARAM_COMM
     *
-    * Accessed under msm_gpu::lock
+    * Accessed under msm_file_private::lock
     */
    char *comm;
    /**
     * cmdline: Overridden task cmdline, see MSM_PARAM_CMDLINE
     *
-    * Accessed under msm_gpu::lock
+    * Accessed under msm_file_private::lock
     */
    char *cmdline;
drivers/gpu/drm/msm/msm_submitqueue.c +0 -1
index 0e803125a325..0444ba04fa06 100644
--- a/drivers/gpu/drm/msm/msm_submitqueue.c
Quoted 8 lines

drivers/gpu/drm/msm/msm_submitqueue.c void __msm_file_private_destroy(struct kref *kref) · line 66

Quoted 8 lines

+++ b/drivers/gpu/drm/msm/msm_submitqueue.c

drivers/gpu/drm/msm/msm_submitqueue.c +1 -0
@@ -61,6 +61,7 @@ void __msm_file_private_destroy(struct kref *kref)
    }
    msm_gem_address_space_put(ctx->aspace);
+   put_pid(ctx->pid);
    kfree(ctx->comm);
    kfree(ctx->cmdline);
    kfree(ctx);

--
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch

Reviewed-by
#7 lore raw

Re: [RFC 2/3] drm/msm: Rework get_comm_cmdline() helper

Greeting all,

Sorry for the delay - Easter Holidays, food coma and all that :-)

Quoted 53 lines · Rob Clark
Quoted 42 lines · Daniel Vetter
Quoted 40 lines · Tvrtko Ursulin
Quoted 38 lines · Rob Clark

From: Rob Clark <[email protected]>

Make it work in terms of ctx so that it can be re-used for fdinfo.

---
  drivers/gpu/drm/msm/adreno/adreno_gpu.c |  4 ++--
  drivers/gpu/drm/msm/msm_drv.c           |  2 ++
  drivers/gpu/drm/msm/msm_gpu.c           | 13 ++++++-------
  drivers/gpu/drm/msm/msm_gpu.h           | 12 ++++++++++--
  drivers/gpu/drm/msm/msm_submitqueue.c   |  1 +
  5 files changed, 21 insertions(+), 11 deletions(-)
drivers/gpu/drm/msm/adreno/adreno_gpu.c +2 -2
index bb38e728864d..43c4e1fea83f 100644
@@ -412,7 +412,7 @@ int adreno_set_param(struct msm_gpu *gpu, struct msm_file_private *ctx,
            /* Ensure string is null terminated: */
            str[len] = '\0';
-           mutex_lock(&gpu->lock);
+           mutex_lock(&ctx->lock);
            if (param == MSM_PARAM_COMM) {
                    paramp = &ctx->comm;
@@ -423,7 +423,7 @@ int adreno_set_param(struct msm_gpu *gpu, struct msm_file_private *ctx,
            kfree(*paramp);
            *paramp = str;
-           mutex_unlock(&gpu->lock);
+           mutex_unlock(&ctx->lock);
            return 0;
    }
drivers/gpu/drm/msm/msm_drv.c +0 -0
index 3d73b98d6a9c..ca0e89e46e13 100644
@@ -581,6 +581,8 @@ static int context_init(struct drm_device *dev, struct drm_file *file)
    rwlock_init(&ctx->queuelock);
    kref_init(&ctx->ref);
Quoted 8 lines
Quoted 8 lines

drivers/gpu/drm/msm/msm_drv.c static int context_init(struct drm_device *dev, struct drm_file *file) · line 584

Quoted 1 line
drivers/gpu/drm/msm/msm_drv.c +1 -0
+   ctx->pid = get_pid(task_pid(current));

Would it simplify things for msm if DRM core had an up to date file->pid as proposed in
https://patchwork.freedesktop.org/patch/526752/?series=109902&rev=4 ? It gets updated if ioctl issuer is different than fd opener and this being context_init here reminded me of it. Maybe you wouldn't have to track the pid in msm?

Quoted 4 lines

The problem is that we also need this for gpu devcore dumps, which could happen after the drm_file is closed. The ctx can outlive the file.

I think we all kept forgetting about that. MSM had support for ages, while AMDGPU is the second driver to land support - just a release ago.

Quoted 2 lines

But the ctx->pid has the same problem as the existing file->pid when it comes to Xorg.. hopefully over time that problem just goes away.

Out of curiosity: what do you mean with "when it comes to Xorg" - the "was_master" handling or something else?

Quoted 6 lines

guess I could do a similar dance to your patch to update the pid whenever (for ex) a submitqueue is created.

Quoted 2 lines

Can we go one step further and let the drm fdinfo stuff print these new additions? Consistency across drivers and all that.

Quoted 9 lines

Hmm, I guess I could _also_ store the overridden comm/cmdline in drm_file. I still need to track it in ctx (msm_file_private) because I could need it after the file is closed.

Maybe it could be useful to have a gl extension to let the app set a name on the context so that this is useful beyond native-ctx (ie. maybe it would be nice to see that "chrome: lwn.net" is using less gpu memory than "chrome: phoronix.com", etc)

/me awaits for the series to hit the respective websites ;-)

But seriously - the series from Tvrtko (thanks for the link, will check in a moment) makes sense. Although given the livespan issue mentioned above, I don't think it's applicable here.

So if it were me, I would consider the two orthogonal for the short/mid term. Fwiw this and patch 1/3 are:

HTH
-Emil

author
#8 lore raw

Re: [RFC 2/3] drm/msm: Rework get_comm_cmdline() helper

Quoted 67 lines · Emil Velikov

Greeting all,

Sorry for the delay - Easter Holidays, food coma and all that :-)

Quoted 53 lines · Rob Clark
Quoted 42 lines · Daniel Vetter
Quoted 40 lines · Tvrtko Ursulin
Quoted 38 lines · Rob Clark

From: Rob Clark <[email protected]>

Make it work in terms of ctx so that it can be re-used for fdinfo.

---
  drivers/gpu/drm/msm/adreno/adreno_gpu.c |  4 ++--
  drivers/gpu/drm/msm/msm_drv.c           |  2 ++
  drivers/gpu/drm/msm/msm_gpu.c           | 13 ++++++-------
  drivers/gpu/drm/msm/msm_gpu.h           | 12 ++++++++++--
  drivers/gpu/drm/msm/msm_submitqueue.c   |  1 +
  5 files changed, 21 insertions(+), 11 deletions(-)
drivers/gpu/drm/msm/adreno/adreno_gpu.c +2 -2
index bb38e728864d..43c4e1fea83f 100644
@@ -412,7 +412,7 @@ int adreno_set_param(struct msm_gpu *gpu, struct msm_file_private *ctx,
            /* Ensure string is null terminated: */
            str[len] = '\0';
-           mutex_lock(&gpu->lock);
+           mutex_lock(&ctx->lock);
            if (param == MSM_PARAM_COMM) {
                    paramp = &ctx->comm;
@@ -423,7 +423,7 @@ int adreno_set_param(struct msm_gpu *gpu, struct msm_file_private *ctx,
            kfree(*paramp);
            *paramp = str;
-           mutex_unlock(&gpu->lock);
+           mutex_unlock(&ctx->lock);
            return 0;
    }
drivers/gpu/drm/msm/msm_drv.c +0 -0
index 3d73b98d6a9c..ca0e89e46e13 100644
@@ -581,6 +581,8 @@ static int context_init(struct drm_device *dev, struct drm_file *file)
    rwlock_init(&ctx->queuelock);
    kref_init(&ctx->ref);
Quoted 8 lines
Quoted 8 lines

drivers/gpu/drm/msm/msm_drv.c static int context_init(struct drm_device *dev, struct drm_file *file) · line 584

Quoted 1 line
drivers/gpu/drm/msm/msm_drv.c +1 -0
+   ctx->pid = get_pid(task_pid(current));

Would it simplify things for msm if DRM core had an up to date file->pid as proposed in
https://patchwork.freedesktop.org/patch/526752/?series=109902&rev=4 ? It gets updated if ioctl issuer is different than fd opener and this being context_init here reminded me of it. Maybe you wouldn't have to track the pid in msm?

Quoted 4 lines

The problem is that we also need this for gpu devcore dumps, which could happen after the drm_file is closed. The ctx can outlive the file.

I think we all kept forgetting about that. MSM had support for ages, while AMDGPU is the second driver to land support - just a release ago.

Quoted 5 lines
Quoted 2 lines

But the ctx->pid has the same problem as the existing file->pid when it comes to Xorg.. hopefully over time that problem just goes away.

Out of curiosity: what do you mean with "when it comes to Xorg" - the "was_master" handling or something else?

The problem is that Xorg is the one to open the drm fd, and then passes the fd to the client.. so the pid of drm_file is the Xorg pid, not the client. Making it not terribly informative.

Tvrtko's patch he linked above would address that for drm_file, but not for other driver internal usages. Maybe it could be wired up as a helper so that drivers don't have to re-invent that dance. Idk, I have to think about it.

Btw, with my WIP drm sched fence signalling patch lockdep is unhappy when gpu devcore dumps are triggered. I'm still pondering how to decouple the locking so that anything coming from fs (ie. show_fdinfo()) is decoupled from anything that happens in the fence signaling path. But will repost this series once I get that sorted out.

BR,
-R

Quoted 29 lines
Quoted 7 lines

guess I could do a similar dance to your patch to update the pid whenever (for ex) a submitqueue is created.

Quoted 2 lines

Can we go one step further and let the drm fdinfo stuff print these new additions? Consistency across drivers and all that.

Quoted 9 lines

Hmm, I guess I could _also_ store the overridden comm/cmdline in drm_file. I still need to track it in ctx (msm_file_private) because I could need it after the file is closed.

Maybe it could be useful to have a gl extension to let the app set a name on the context so that this is useful beyond native-ctx (ie. maybe it would be nice to see that "chrome: lwn.net" is using less gpu memory than "chrome: phoronix.com", etc)

/me awaits for the series to hit the respective websites ;-)

But seriously - the series from Tvrtko (thanks for the link, will check in a moment) makes sense. Although given the livespan issue mentioned above, I don't think it's applicable here.

So if it were me, I would consider the two orthogonal for the short/mid term. Fwiw this and patch 1/3 are:

HTH
-Emil

#9 lore raw

Re: [RFC 2/3] drm/msm: Rework get_comm_cmdline() helper

Quoted 83 lines · Rob Clark
Quoted 67 lines · Emil Velikov

Greeting all,

Sorry for the delay - Easter Holidays, food coma and all that :-)

Quoted 53 lines · Rob Clark
Quoted 42 lines · Daniel Vetter
Quoted 40 lines · Tvrtko Ursulin
Quoted 38 lines · Rob Clark

From: Rob Clark <[email protected]>

Make it work in terms of ctx so that it can be re-used for fdinfo.

---
  drivers/gpu/drm/msm/adreno/adreno_gpu.c |  4 ++--
  drivers/gpu/drm/msm/msm_drv.c           |  2 ++
  drivers/gpu/drm/msm/msm_gpu.c           | 13 ++++++-------
  drivers/gpu/drm/msm/msm_gpu.h           | 12 ++++++++++--
  drivers/gpu/drm/msm/msm_submitqueue.c   |  1 +
  5 files changed, 21 insertions(+), 11 deletions(-)
drivers/gpu/drm/msm/adreno/adreno_gpu.c +2 -2
index bb38e728864d..43c4e1fea83f 100644
@@ -412,7 +412,7 @@ int adreno_set_param(struct msm_gpu *gpu, struct msm_file_private *ctx,
            /* Ensure string is null terminated: */
            str[len] = '\0';
-           mutex_lock(&gpu->lock);
+           mutex_lock(&ctx->lock);
            if (param == MSM_PARAM_COMM) {
                    paramp = &ctx->comm;
@@ -423,7 +423,7 @@ int adreno_set_param(struct msm_gpu *gpu, struct msm_file_private *ctx,
            kfree(*paramp);
            *paramp = str;
-           mutex_unlock(&gpu->lock);
+           mutex_unlock(&ctx->lock);
            return 0;
    }
drivers/gpu/drm/msm/msm_drv.c +0 -0
index 3d73b98d6a9c..ca0e89e46e13 100644
@@ -581,6 +581,8 @@ static int context_init(struct drm_device *dev, struct drm_file *file)
    rwlock_init(&ctx->queuelock);
    kref_init(&ctx->ref);
Quoted 8 lines
Quoted 8 lines

drivers/gpu/drm/msm/msm_drv.c static int context_init(struct drm_device *dev, struct drm_file *file) · line 584

Quoted 1 line
drivers/gpu/drm/msm/msm_drv.c +1 -0
+   ctx->pid = get_pid(task_pid(current));

Would it simplify things for msm if DRM core had an up to date file->pid as proposed in
https://patchwork.freedesktop.org/patch/526752/?series=109902&rev=4 ? It gets updated if ioctl issuer is different than fd opener and this being context_init here reminded me of it. Maybe you wouldn't have to track the pid in msm?

Quoted 4 lines

The problem is that we also need this for gpu devcore dumps, which could happen after the drm_file is closed. The ctx can outlive the file.

I think we all kept forgetting about that. MSM had support for ages, while AMDGPU is the second driver to land support - just a release ago.

Quoted 5 lines
Quoted 2 lines

But the ctx->pid has the same problem as the existing file->pid when it comes to Xorg.. hopefully over time that problem just goes away.

Out of curiosity: what do you mean with "when it comes to Xorg" - the "was_master" handling or something else?

The problem is that Xorg is the one to open the drm fd, and then passes the fd to the client.. so the pid of drm_file is the Xorg pid, not the client. Making it not terribly informative.

Tvrtko's patch he linked above would address that for drm_file, but not for other driver internal usages. Maybe it could be wired up as a helper so that drivers don't have to re-invent that dance. Idk, I have to think about it.

Quoted 6 lines

Btw, with my WIP drm sched fence signalling patch lockdep is unhappy when gpu devcore dumps are triggered. I'm still pondering how to decouple the locking so that anything coming from fs (ie. show_fdinfo()) is decoupled from anything that happens in the fence signaling path. But will repost this series once I get that sorted out.

So the cleanest imo is that you push most of the capturing into a worker that's entirely decoupled. If you have terminal context (i.e. on first hang they stop all further cmd submission, which is anyway what vk/arb_robustness want), then you don't have to capture at tdr time, because there's no subsequent batch that will wreck the state.

But it only works if your gpu ctx don't have recoverable semantics.

If you can't do that it's a _lot_ of GFP_ATOMIC and trylock and bailing out if any fails :-/
-Daniel

Quoted 33 lines

BR,
-R

Quoted 29 lines
Quoted 7 lines

guess I could do a similar dance to your patch to update the pid whenever (for ex) a submitqueue is created.

Quoted 2 lines

Can we go one step further and let the drm fdinfo stuff print these new additions? Consistency across drivers and all that.

Quoted 9 lines

Hmm, I guess I could _also_ store the overridden comm/cmdline in drm_file. I still need to track it in ctx (msm_file_private) because I could need it after the file is closed.

Maybe it could be useful to have a gl extension to let the app set a name on the context so that this is useful beyond native-ctx (ie. maybe it would be nice to see that "chrome: lwn.net" is using less gpu memory than "chrome: phoronix.com", etc)

/me awaits for the series to hit the respective websites ;-)

But seriously - the series from Tvrtko (thanks for the link, will check in a moment) makes sense. Although given the livespan issue mentioned above, I don't think it's applicable here.

So if it were me, I would consider the two orthogonal for the short/mid term. Fwiw this and patch 1/3 are:

HTH
-Emil

Signature · 3 lines
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
author
#10 lore raw

Re: [RFC 2/3] drm/msm: Rework get_comm_cmdline() helper

Quoted 92 lines · Daniel Vetter
Quoted 83 lines · Rob Clark
Quoted 67 lines · Emil Velikov

Greeting all,

Sorry for the delay - Easter Holidays, food coma and all that :-)

Quoted 53 lines · Rob Clark
Quoted 42 lines · Daniel Vetter
Quoted 40 lines · Tvrtko Ursulin
Quoted 38 lines · Rob Clark

From: Rob Clark <[email protected]>

Make it work in terms of ctx so that it can be re-used for fdinfo.

---
  drivers/gpu/drm/msm/adreno/adreno_gpu.c |  4 ++--
  drivers/gpu/drm/msm/msm_drv.c           |  2 ++
  drivers/gpu/drm/msm/msm_gpu.c           | 13 ++++++-------
  drivers/gpu/drm/msm/msm_gpu.h           | 12 ++++++++++--
  drivers/gpu/drm/msm/msm_submitqueue.c   |  1 +
  5 files changed, 21 insertions(+), 11 deletions(-)
drivers/gpu/drm/msm/adreno/adreno_gpu.c +2 -2
index bb38e728864d..43c4e1fea83f 100644
@@ -412,7 +412,7 @@ int adreno_set_param(struct msm_gpu *gpu, struct msm_file_private *ctx,
            /* Ensure string is null terminated: */
            str[len] = '\0';
-           mutex_lock(&gpu->lock);
+           mutex_lock(&ctx->lock);
            if (param == MSM_PARAM_COMM) {
                    paramp = &ctx->comm;
@@ -423,7 +423,7 @@ int adreno_set_param(struct msm_gpu *gpu, struct msm_file_private *ctx,
            kfree(*paramp);
            *paramp = str;
-           mutex_unlock(&gpu->lock);
+           mutex_unlock(&ctx->lock);
            return 0;
    }
drivers/gpu/drm/msm/msm_drv.c +0 -0
index 3d73b98d6a9c..ca0e89e46e13 100644
@@ -581,6 +581,8 @@ static int context_init(struct drm_device *dev, struct drm_file *file)
    rwlock_init(&ctx->queuelock);
    kref_init(&ctx->ref);
Quoted 8 lines
Quoted 8 lines

drivers/gpu/drm/msm/msm_drv.c static int context_init(struct drm_device *dev, struct drm_file *file) · line 583

Quoted 1 line
drivers/gpu/drm/msm/msm_drv.c +1 -0
+   ctx->pid = get_pid(task_pid(current));

Would it simplify things for msm if DRM core had an up to date file->pid as proposed in
https://patchwork.freedesktop.org/patch/526752/?series=109902&rev=4 ? It gets updated if ioctl issuer is different than fd opener and this being context_init here reminded me of it. Maybe you wouldn't have to track the pid in msm?

Quoted 4 lines

The problem is that we also need this for gpu devcore dumps, which could happen after the drm_file is closed. The ctx can outlive the file.

I think we all kept forgetting about that. MSM had support for ages, while AMDGPU is the second driver to land support - just a release ago.

Quoted 5 lines
Quoted 2 lines

But the ctx->pid has the same problem as the existing file->pid when it comes to Xorg.. hopefully over time that problem just goes away.

Out of curiosity: what do you mean with "when it comes to Xorg" - the "was_master" handling or something else?

The problem is that Xorg is the one to open the drm fd, and then passes the fd to the client.. so the pid of drm_file is the Xorg pid, not the client. Making it not terribly informative.

Tvrtko's patch he linked above would address that for drm_file, but not for other driver internal usages. Maybe it could be wired up as a helper so that drivers don't have to re-invent that dance. Idk, I have to think about it.

Quoted 6 lines

Btw, with my WIP drm sched fence signalling patch lockdep is unhappy when gpu devcore dumps are triggered. I'm still pondering how to decouple the locking so that anything coming from fs (ie. show_fdinfo()) is decoupled from anything that happens in the fence signaling path. But will repost this series once I get that sorted out.

Quoted 5 lines

So the cleanest imo is that you push most of the capturing into a worker that's entirely decoupled. If you have terminal context (i.e. on first hang they stop all further cmd submission, which is anyway what vk/arb_robustness want), then you don't have to capture at tdr time, because there's no subsequent batch that will wreck the state.

It is already in a worker, but we (a) need to block other contexts from submitting while at the same time (b) using the GPU itself to capture its state.. (yes, the way the hw works is overly complicated in this regard)

Quoted 1 line

But it only works if your gpu ctx don't have recoverable semantics.

We do have recoverable semantics.. but that is pretty orthogonal. We just need a different lock.. I have a plan to move (a copy) of the override strings to drm_file with it's own locking decoupled from what we need in the recovery path.. and hopefully will finally have time to type it up today and post it (just before disappearing off into the woods to go backpacking ;-))

BR,
-R

Quoted 42 lines

If you can't do that it's a _lot_ of GFP_ATOMIC and trylock and bailing out if any fails :-/
-Daniel

Quoted 33 lines

BR,
-R

Quoted 29 lines
Quoted 7 lines

guess I could do a similar dance to your patch to update the pid whenever (for ex) a submitqueue is created.

Quoted 2 lines

Can we go one step further and let the drm fdinfo stuff print these new additions? Consistency across drivers and all that.

Quoted 9 lines

Hmm, I guess I could _also_ store the overridden comm/cmdline in drm_file. I still need to track it in ctx (msm_file_private) because I could need it after the file is closed.

Maybe it could be useful to have a gl extension to let the app set a name on the context so that this is useful beyond native-ctx (ie. maybe it would be nice to see that "chrome: lwn.net" is using less gpu memory than "chrome: phoronix.com", etc)

/me awaits for the series to hit the respective websites ;-)

But seriously - the series from Tvrtko (thanks for the link, will check in a moment) makes sense. Although given the livespan issue mentioned above, I don't think it's applicable here.

So if it were me, I would consider the two orthogonal for the short/mid term. Fwiw this and patch 1/3 are:

HTH
-Emil

--
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch

author
#11 lore raw

RFC 3/3 drm/msm: Add comm/cmdline fields

In reply to Rob Clark

From: Rob Clark <[email protected]>

Normally this would be the same information that can be obtained in other ways. But in some cases the process opening the drm fd is merely a sort of proxy for the actual process using the GPU. This is the case for guest VM processes using the GPU via virglrenderer, in which case the msm native-context renderer in virglrenderer overrides the comm/ cmdline to be the guest process's values.

Exposing this via fdinfo allows tools like gputop to show something more meaningful than just a bunch of "pcivirtio-gpu" users.

---
 Documentation/gpu/drm-usage-stats.rst |  8 ++++++++
 drivers/gpu/drm/msm/msm_gpu.c         | 14 ++++++++++++++
 2 files changed, 22 insertions(+)
Documentation/gpu/drm-usage-stats.rst +8 -0
index 8e00d53231e0..bc90bed455e3 100644
@@ -148,6 +148,14 @@ percentage utilization of the engine, whereas drm-engine-<keystr> only reflects
 time active without considering what frequency the engine is operating as a
 percentage of it's maximum frequency.
 
+- drm-comm: <valstr>
+
+Returns the clients executable path.
+
+- drm-cmdline: <valstr>
+
+Returns the clients cmdline.
+
 Implementation Details
 ======================
 
drivers/gpu/drm/msm/msm_gpu.c +14 -0
index f0f4f845c32d..1150dcbf28aa 100644
@@ -148,12 +148,26 @@ int msm_gpu_pm_suspend(struct msm_gpu *gpu)
 	return 0;
 }
 
+static void get_comm_cmdline(struct msm_file_private *ctx, char **comm, char **cmd);
+
 void msm_gpu_show_fdinfo(struct msm_gpu *gpu, struct msm_file_private *ctx,
 			 struct drm_printer *p)
 {
+	char *comm, *cmdline;
+
+	get_comm_cmdline(ctx, &comm, &cmdline);
+
 	drm_printf(p, "drm-engine-gpu:\t%llu ns\n", ctx->elapsed_ns);
 	drm_printf(p, "drm-cycles-gpu:\t%llu\n", ctx->cycles);
 	drm_printf(p, "drm-maxfreq-gpu:\t%u Hz\n", gpu->fast_rate);
+
+	if (comm)
+		drm_printf(p, "drm-comm:\t%s\n", comm);
+	if (cmdline)
+		drm_printf(p, "drm-cmdline:\t%s\n", cmdline);
+
+	kfree(comm);
+	kfree(cmdline);
 }
 
 int msm_gpu_hw_init(struct msm_gpu *gpu)
Signature · 1 line
2.39.2
#12 lore raw

Re: [RFC 3/3] drm/msm: Add comm/cmdline fields

Quoted 11 lines · Rob Clark

From: Rob Clark <[email protected]>

Normally this would be the same information that can be obtained in other ways. But in some cases the process opening the drm fd is merely a sort of proxy for the actual process using the GPU. This is the case for guest VM processes using the GPU via virglrenderer, in which case the msm native-context renderer in virglrenderer overrides the comm/ cmdline to be the guest process's values.

Exposing this via fdinfo allows tools like gputop to show something more meaningful than just a bunch of "pcivirtio-gpu" users.

You also later expanded with:

"""
I should have also mentioned, in the VM/proxy scenario we have a single process with separate drm_file's for each guest VM process. So it isn't an option to just change the proxy process's name to match the client.
"""

So how does that work - this single process temporarily changes it's name for each drm fd it opens and creates a context or it is actually in the native context protocol?

Quoted 12 lines
---
  Documentation/gpu/drm-usage-stats.rst |  8 ++++++++
  drivers/gpu/drm/msm/msm_gpu.c         | 14 ++++++++++++++
  2 files changed, 22 insertions(+)
Documentation/gpu/drm-usage-stats.rst +0 -0
index 8e00d53231e0..bc90bed455e3 100644

Documentation/gpu/drm-usage-stats.rst percentage utilization of the engine, whereas drm-engine-<keystr> only reflects · line 153

Quoted 6 lines
Documentation/gpu/drm-usage-stats.rst +3 -0
@@ -148,6 +148,14 @@ percentage utilization of the engine, whereas drm-engine-<keystr> only reflects
  time active without considering what frequency the engine is operating as a
  percentage of it's maximum frequency.
 
+- drm-comm: <valstr>
+
+Returns the clients executable path.

Full path and not just current->comm? In this case probably give it a more descriptive name here.

drm-client-executable
drm-client-command-line

So we stay in the drm-client- namespace?

Or if the former is absolute path could one key be enough for both?

drm-client-command-line: /path/to/executable --arguments

Documentation/gpu/drm-usage-stats.rst percentage utilization of the engine, whereas drm-engine-<keystr> only reflects · line 157

Quoted 4 lines
Documentation/gpu/drm-usage-stats.rst +4 -0
+
+- drm-cmdline: <valstr>
+
+Returns the clients cmdline.

I think drm-usage-stats.rst text should provide some more text with these two. To precisely define their content and outline the use case under which driver authors may want to add them, and fdinfo consumer therefore expect to see them. Just so everything is completely clear and people do not start adding them for drivers which do not support native context (or like).

But on the overall it sounds reasonable to me - it would be really cool to not just see pcivirtio-gpu as you say. Even if the standard virtiogpu use case (not native context) could show real users.

Regards,

Tvrtko

Quoted 35 lines
Documentation/gpu/drm-usage-stats.rst +1 -0
+
  Implementation Details
  ======================
 
drivers/gpu/drm/msm/msm_gpu.c +14 -0
index f0f4f845c32d..1150dcbf28aa 100644
@@ -148,12 +148,26 @@ int msm_gpu_pm_suspend(struct msm_gpu *gpu)
  	return 0;
  }
 
+static void get_comm_cmdline(struct msm_file_private *ctx, char **comm, char **cmd);
+
  void msm_gpu_show_fdinfo(struct msm_gpu *gpu, struct msm_file_private *ctx,
  			 struct drm_printer *p)
  {
+	char *comm, *cmdline;
+
+	get_comm_cmdline(ctx, &comm, &cmdline);
+
  	drm_printf(p, "drm-engine-gpu:\t%llu ns\n", ctx->elapsed_ns);
  	drm_printf(p, "drm-cycles-gpu:\t%llu\n", ctx->cycles);
  	drm_printf(p, "drm-maxfreq-gpu:\t%u Hz\n", gpu->fast_rate);
+
+	if (comm)
+		drm_printf(p, "drm-comm:\t%s\n", comm);
+	if (cmdline)
+		drm_printf(p, "drm-cmdline:\t%s\n", cmdline);
+
+	kfree(comm);
+	kfree(cmdline);
  }
 
  int msm_gpu_hw_init(struct msm_gpu *gpu)
author
#13 lore raw

Re: [RFC 3/3] drm/msm: Add comm/cmdline fields

Quoted 24 lines · Tvrtko Ursulin
Quoted 11 lines · Rob Clark

From: Rob Clark <[email protected]>

Normally this would be the same information that can be obtained in other ways. But in some cases the process opening the drm fd is merely a sort of proxy for the actual process using the GPU. This is the case for guest VM processes using the GPU via virglrenderer, in which case the msm native-context renderer in virglrenderer overrides the comm/ cmdline to be the guest process's values.

Exposing this via fdinfo allows tools like gputop to show something more meaningful than just a bunch of "pcivirtio-gpu" users.

You also later expanded with:

"""
I should have also mentioned, in the VM/proxy scenario we have a single process with separate drm_file's for each guest VM process. So it isn't an option to just change the proxy process's name to match the client.
"""

Quoted 3 lines

So how does that work - this single process temporarily changes it's name for each drm fd it opens and creates a context or it is actually in the native context protocol?

It is part of the protocol, the mesa driver in the VM sends[1] this info to the native-context "shim" in host userspace which uses the SET_PARAM ioctl to pass this to the kernel. In the host userspace there is just a single process (you see the host PID below) but it does a separate open() of the drm dev for each guest process (so that they each have their own GPU address space for isolation):

DRM minor 128
PID MEM ACTIV NAME gpu
5297 200M 82M com.mojang.minecr |██████████████▏ |
1859 199M 0B chrome |█▉ |
5297 64M 9M surfaceflinger | |
5297 12M 0B org.chromium.arc. | |
5297 12M 0B com.android.syste | |
5297 12M 0B org.chromium.arc. | |
5297 26M 0B com.google.androi | |
5297 65M 0B system_server | |

[1] https://gitlab.freedesktop.org/virgl/virglrenderer/-/blob/master/src/drm/msm/msm_proto.h#L326
[2] https://gitlab.freedesktop.org/virgl/virglrenderer/-/blob/master/src/drm/msm/msm_renderer.c#L1050

Quoted 22 lines
Quoted 12 lines
---
  Documentation/gpu/drm-usage-stats.rst |  8 ++++++++
  drivers/gpu/drm/msm/msm_gpu.c         | 14 ++++++++++++++
  2 files changed, 22 insertions(+)
Documentation/gpu/drm-usage-stats.rst +0 -0
index 8e00d53231e0..bc90bed455e3 100644

Documentation/gpu/drm-usage-stats.rst percentage utilization of the engine, whereas drm-engine-<keystr> only reflects · line 153

Quoted 6 lines
Documentation/gpu/drm-usage-stats.rst +3 -0
@@ -148,6 +148,14 @@ percentage utilization of the engine, whereas drm-engine-<keystr> only reflects
  time active without considering what frequency the engine is operating as a
  percentage of it's maximum frequency.
 
+- drm-comm: <valstr>
+
+Returns the clients executable path.

Full path and not just current->comm? In this case probably give it a more descriptive name here.

Quoted 8 lines

drm-client-executable
drm-client-command-line

So we stay in the drm-client- namespace?

Or if the former is absolute path could one key be enough for both?

drm-client-command-line: /path/to/executable --arguments

comm and cmdline can be different. Android seems to change the comm to the apk name, for example (and w/ the zygote stuff cmdline isn't really a thing)

I guess it could be drm-client-comm and drm-client-cmdline? Although comm/cmdline aren't the best names, they are just following what the kernel calls them elsewhere.

Quoted 11 lines

Documentation/gpu/drm-usage-stats.rst percentage utilization of the engine, whereas drm-engine-<keystr> only reflects · line 157

Quoted 4 lines
Documentation/gpu/drm-usage-stats.rst +4 -0
+
+- drm-cmdline: <valstr>
+
+Returns the clients cmdline.

I think drm-usage-stats.rst text should provide some more text with these two. To precisely define their content and outline the use case under which driver authors may want to add them, and fdinfo consumer therefore expect to see them. Just so everything is completely clear and people do not start adding them for drivers which do not support native context (or like).

I really was just piggy-backing on existing comm/cmdline.. but I'll try to write up something better.

I think it maybe should not be limited just to native context.. for ex. if the browser did somehow manage to create different displays associated with different drm_file instances (I guess it would have to use gbm to do this?) it would be nice to see browser tab names.

Quoted 3 lines

But on the overall it sounds reasonable to me - it would be really cool to not just see pcivirtio-gpu as you say. Even if the standard virtiogpu use case (not native context) could show real users.

For vrend/virgl, we'd first need to solve the issue that there is just a single drm_file for all guest processes. But really, just don't use virgl. (I mean, like seriously, would you put a gl driver in the kernel? Vrend has access to all guest memory, so this is essentially what you have with virgl. This is just not a sane thing to do.) The only "valid" reason for not doing native-context is if you don't have the src code for your UMD to be able to modify it to talk native-context to virtgpu in the guest. ;-)

BR,
-R

Quoted 39 lines

Regards,

Tvrtko

Quoted 35 lines
Documentation/gpu/drm-usage-stats.rst +1 -0
+
  Implementation Details
  ======================
 
drivers/gpu/drm/msm/msm_gpu.c +14 -0
index f0f4f845c32d..1150dcbf28aa 100644
@@ -148,12 +148,26 @@ int msm_gpu_pm_suspend(struct msm_gpu *gpu)
      return 0;
  }
 
+static void get_comm_cmdline(struct msm_file_private *ctx, char **comm, char **cmd);
+
  void msm_gpu_show_fdinfo(struct msm_gpu *gpu, struct msm_file_private *ctx,
                       struct drm_printer *p)
  {
+     char *comm, *cmdline;
+
+     get_comm_cmdline(ctx, &comm, &cmdline);
+
      drm_printf(p, "drm-engine-gpu:\t%llu ns\n", ctx->elapsed_ns);
      drm_printf(p, "drm-cycles-gpu:\t%llu\n", ctx->cycles);
      drm_printf(p, "drm-maxfreq-gpu:\t%u Hz\n", gpu->fast_rate);
+
+     if (comm)
+             drm_printf(p, "drm-comm:\t%s\n", comm);
+     if (cmdline)
+             drm_printf(p, "drm-cmdline:\t%s\n", cmdline);
+
+     kfree(comm);
+     kfree(cmdline);
  }
 
  int msm_gpu_hw_init(struct msm_gpu *gpu)
#14 lore raw

Re: [RFC 3/3] drm/msm: Add comm/cmdline fields

Quoted 83 lines · Rob Clark
Quoted 24 lines · Tvrtko Ursulin
Quoted 11 lines · Rob Clark

From: Rob Clark <[email protected]>

Normally this would be the same information that can be obtained in other ways. But in some cases the process opening the drm fd is merely a sort of proxy for the actual process using the GPU. This is the case for guest VM processes using the GPU via virglrenderer, in which case the msm native-context renderer in virglrenderer overrides the comm/ cmdline to be the guest process's values.

Exposing this via fdinfo allows tools like gputop to show something more meaningful than just a bunch of "pcivirtio-gpu" users.

You also later expanded with:

"""
I should have also mentioned, in the VM/proxy scenario we have a single process with separate drm_file's for each guest VM process. So it isn't an option to just change the proxy process's name to match the client.
"""

Quoted 3 lines

So how does that work - this single process temporarily changes it's name for each drm fd it opens and creates a context or it is actually in the native context protocol?

It is part of the protocol, the mesa driver in the VM sends[1] this info to the native-context "shim" in host userspace which uses the SET_PARAM ioctl to pass this to the kernel. In the host userspace there is just a single process (you see the host PID below) but it does a separate open() of the drm dev for each guest process (so that they each have their own GPU address space for isolation):

DRM minor 128
PID MEM ACTIV NAME gpu
5297 200M 82M com.mojang.minecr |██████████████▏ |
1859 199M 0B chrome |█▉ |
5297 64M 9M surfaceflinger | |
5297 12M 0B org.chromium.arc. | |
5297 12M 0B com.android.syste | |
5297 12M 0B org.chromium.arc. | |
5297 26M 0B com.google.androi | |
5297 65M 0B system_server | |

[1] https://gitlab.freedesktop.org/virgl/virglrenderer/-/blob/master/src/drm/msm/msm_proto.h#L326
[2] https://gitlab.freedesktop.org/virgl/virglrenderer/-/blob/master/src/drm/msm/msm_renderer.c#L1050

Quoted 22 lines
Quoted 12 lines
---
   Documentation/gpu/drm-usage-stats.rst |  8 ++++++++
   drivers/gpu/drm/msm/msm_gpu.c         | 14 ++++++++++++++
   2 files changed, 22 insertions(+)
Documentation/gpu/drm-usage-stats.rst +0 -0
index 8e00d53231e0..bc90bed455e3 100644

Documentation/gpu/drm-usage-stats.rst percentage utilization of the engine, whereas drm-engine-<keystr> only reflects · line 153

Quoted 6 lines
Documentation/gpu/drm-usage-stats.rst +3 -0
@@ -148,6 +148,14 @@ percentage utilization of the engine, whereas drm-engine-<keystr> only reflects
   time active without considering what frequency the engine is operating as a
   percentage of it's maximum frequency.
 
+- drm-comm: <valstr>
+
+Returns the clients executable path.

Full path and not just current->comm? In this case probably give it a more descriptive name here.

Quoted 8 lines

drm-client-executable
drm-client-command-line

So we stay in the drm-client- namespace?

Or if the former is absolute path could one key be enough for both?

drm-client-command-line: /path/to/executable --arguments

Quoted 7 lines

comm and cmdline can be different. Android seems to change the comm to the apk name, for example (and w/ the zygote stuff cmdline isn't really a thing)

I guess it could be drm-client-comm and drm-client-cmdline? Although comm/cmdline aren't the best names, they are just following what the kernel calls them elsewhere.

I wasn't sure what do you plan to do given mention of a path under the drm-comm description. If it is a path then comm would be misleading, since comm as defined in procfs is not a path, I don't think so at least. Which is why I was suggesting executable. But if you remove the mention of a path from rst and rather refer to processes' comm value I think that is then okay.

Quoted 12 lines
Quoted 11 lines

Documentation/gpu/drm-usage-stats.rst percentage utilization of the engine, whereas drm-engine-<keystr> only reflects · line 157

Quoted 4 lines
Documentation/gpu/drm-usage-stats.rst +4 -0
+
+- drm-cmdline: <valstr>
+
+Returns the clients cmdline.

I think drm-usage-stats.rst text should provide some more text with these two. To precisely define their content and outline the use case under which driver authors may want to add them, and fdinfo consumer therefore expect to see them. Just so everything is completely clear and people do not start adding them for drivers which do not support native context (or like).

Quoted 7 lines

I really was just piggy-backing on existing comm/cmdline.. but I'll try to write up something better.

I think it maybe should not be limited just to native context.. for ex. if the browser did somehow manage to create different displays associated with different drm_file instances (I guess it would have to use gbm to do this?) it would be nice to see browser tab names.

Would be cool yes.

My thinking behind why we maybe do not want to blanket add them is because for common case is it the same information which can be obtained from procfs. Like in igt_drm_clients.c I get the pid and comm from /proc/$pid/stat. So I was thinking it is only interesting to add to fdinfo for drivers where it could differ by the explicit override like you have with native context.

It can be added once there is a GL/whatever extension which would allow it? (I am not familiar with how browsers manage rendering contexts so maybe I am missing something.)

Quoted 4 lines
Quoted 3 lines

But on the overall it sounds reasonable to me - it would be really cool to not just see pcivirtio-gpu as you say. Even if the standard virtiogpu use case (not native context) could show real users.

Quoted 8 lines

For vrend/virgl, we'd first need to solve the issue that there is just a single drm_file for all guest processes. But really, just don't use virgl. (I mean, like seriously, would you put a gl driver in the kernel? Vrend has access to all guest memory, so this is essentially what you have with virgl. This is just not a sane thing to do.) The only "valid" reason for not doing native-context is if you don't have the src code for your UMD to be able to modify it to talk native-context to virtgpu in the guest. ;-)

I am just observing the current state of things on an Intel based Chromebook. :) Presumably the custom name for a context would be passable via the virtio-gpu protocol or something?

Regards,

Tvrtko

Quoted 43 lines

BR,
-R

Quoted 39 lines

Regards,

Tvrtko

Quoted 35 lines
Documentation/gpu/drm-usage-stats.rst +1 -0
+
   Implementation Details
   ======================
 
drivers/gpu/drm/msm/msm_gpu.c +14 -0
index f0f4f845c32d..1150dcbf28aa 100644
@@ -148,12 +148,26 @@ int msm_gpu_pm_suspend(struct msm_gpu *gpu)
       return 0;
   }
 
+static void get_comm_cmdline(struct msm_file_private *ctx, char **comm, char **cmd);
+
   void msm_gpu_show_fdinfo(struct msm_gpu *gpu, struct msm_file_private *ctx,
                        struct drm_printer *p)
   {
+     char *comm, *cmdline;
+
+     get_comm_cmdline(ctx, &comm, &cmdline);
+
       drm_printf(p, "drm-engine-gpu:\t%llu ns\n", ctx->elapsed_ns);
       drm_printf(p, "drm-cycles-gpu:\t%llu\n", ctx->cycles);
       drm_printf(p, "drm-maxfreq-gpu:\t%u Hz\n", gpu->fast_rate);
+
+     if (comm)
+             drm_printf(p, "drm-comm:\t%s\n", comm);
+     if (cmdline)
+             drm_printf(p, "drm-cmdline:\t%s\n", cmdline);
+
+     kfree(comm);
+     kfree(cmdline);
   }
 
   int msm_gpu_hw_init(struct msm_gpu *gpu)
author
#15 lore raw

Re: [RFC 3/3] drm/msm: Add comm/cmdline fields

Quoted 94 lines · Tvrtko Ursulin
Quoted 83 lines · Rob Clark
Quoted 24 lines · Tvrtko Ursulin
Quoted 11 lines · Rob Clark

From: Rob Clark <[email protected]>

Normally this would be the same information that can be obtained in other ways. But in some cases the process opening the drm fd is merely a sort of proxy for the actual process using the GPU. This is the case for guest VM processes using the GPU via virglrenderer, in which case the msm native-context renderer in virglrenderer overrides the comm/ cmdline to be the guest process's values.

Exposing this via fdinfo allows tools like gputop to show something more meaningful than just a bunch of "pcivirtio-gpu" users.

You also later expanded with:

"""
I should have also mentioned, in the VM/proxy scenario we have a single process with separate drm_file's for each guest VM process. So it isn't an option to just change the proxy process's name to match the client.
"""

Quoted 3 lines

So how does that work - this single process temporarily changes it's name for each drm fd it opens and creates a context or it is actually in the native context protocol?

It is part of the protocol, the mesa driver in the VM sends[1] this info to the native-context "shim" in host userspace which uses the SET_PARAM ioctl to pass this to the kernel. In the host userspace there is just a single process (you see the host PID below) but it does a separate open() of the drm dev for each guest process (so that they each have their own GPU address space for isolation):

DRM minor 128
PID MEM ACTIV NAME gpu
5297 200M 82M com.mojang.minecr |██████████████▏ |
1859 199M 0B chrome |█▉ |
5297 64M 9M surfaceflinger | |
5297 12M 0B org.chromium.arc. | |
5297 12M 0B com.android.syste | |
5297 12M 0B org.chromium.arc. | |
5297 26M 0B com.google.androi | |
5297 65M 0B system_server | |

[1] https://gitlab.freedesktop.org/virgl/virglrenderer/-/blob/master/src/drm/msm/msm_proto.h#L326
[2] https://gitlab.freedesktop.org/virgl/virglrenderer/-/blob/master/src/drm/msm/msm_renderer.c#L1050

Quoted 22 lines
Quoted 12 lines
---
   Documentation/gpu/drm-usage-stats.rst |  8 ++++++++
   drivers/gpu/drm/msm/msm_gpu.c         | 14 ++++++++++++++
   2 files changed, 22 insertions(+)
Documentation/gpu/drm-usage-stats.rst +0 -0
index 8e00d53231e0..bc90bed455e3 100644

Documentation/gpu/drm-usage-stats.rst percentage utilization of the engine, whereas drm-engine-<keystr> only reflects · line 153

Quoted 6 lines
Documentation/gpu/drm-usage-stats.rst +3 -0
@@ -148,6 +148,14 @@ percentage utilization of the engine, whereas drm-engine-<keystr> only reflects
   time active without considering what frequency the engine is operating as a
   percentage of it's maximum frequency.
 
+- drm-comm: <valstr>
+
+Returns the clients executable path.

Full path and not just current->comm? In this case probably give it a more descriptive name here.

Quoted 8 lines

drm-client-executable
drm-client-command-line

So we stay in the drm-client- namespace?

Or if the former is absolute path could one key be enough for both?

drm-client-command-line: /path/to/executable --arguments

Quoted 7 lines

comm and cmdline can be different. Android seems to change the comm to the apk name, for example (and w/ the zygote stuff cmdline isn't really a thing)

I guess it could be drm-client-comm and drm-client-cmdline? Although comm/cmdline aren't the best names, they are just following what the kernel calls them elsewhere.

Quoted 6 lines

I wasn't sure what do you plan to do given mention of a path under the drm-comm description. If it is a path then comm would be misleading, since comm as defined in procfs is not a path, I don't think so at least. Which is why I was suggesting executable. But if you remove the mention of a path from rst and rather refer to processes' comm value I think that is then okay.

Oh, whoops the mention of "path" for comm was a mistake. task->comm is described as executable name without path, and that is what the fdinfo field was intending to follow.

Quoted 20 lines
Quoted 12 lines
Quoted 11 lines

Documentation/gpu/drm-usage-stats.rst percentage utilization of the engine, whereas drm-engine-<keystr> only reflects · line 157

Quoted 4 lines
Documentation/gpu/drm-usage-stats.rst +4 -0
+
+- drm-cmdline: <valstr>
+
+Returns the clients cmdline.

I think drm-usage-stats.rst text should provide some more text with these two. To precisely define their content and outline the use case under which driver authors may want to add them, and fdinfo consumer therefore expect to see them. Just so everything is completely clear and people do not start adding them for drivers which do not support native context (or like).

Quoted 7 lines

I really was just piggy-backing on existing comm/cmdline.. but I'll try to write up something better.

I think it maybe should not be limited just to native context.. for ex. if the browser did somehow manage to create different displays associated with different drm_file instances (I guess it would have to use gbm to do this?) it would be nice to see browser tab names.

Quoted 8 lines

Would be cool yes.

My thinking behind why we maybe do not want to blanket add them is because for common case is it the same information which can be obtained from procfs. Like in igt_drm_clients.c I get the pid and comm from /proc/$pid/stat. So I was thinking it is only interesting to add to fdinfo for drivers where it could differ by the explicit override like you have with native context.

Yeah, I suppose I could define them as drm-client-comm-override and drm-client-cmdline-override

Quoted 17 lines

It can be added once there is a GL/whatever extension which would allow it? (I am not familiar with how browsers manage rendering contexts so maybe I am missing something.)

Quoted 4 lines
Quoted 3 lines

But on the overall it sounds reasonable to me - it would be really cool to not just see pcivirtio-gpu as you say. Even if the standard virtiogpu use case (not native context) could show real users.

Quoted 8 lines

For vrend/virgl, we'd first need to solve the issue that there is just a single drm_file for all guest processes. But really, just don't use virgl. (I mean, like seriously, would you put a gl driver in the kernel? Vrend has access to all guest memory, so this is essentially what you have with virgl. This is just not a sane thing to do.) The only "valid" reason for not doing native-context is if you don't have the src code for your UMD to be able to modify it to talk native-context to virtgpu in the guest. ;-)

Quoted 3 lines

I am just observing the current state of things on an Intel based Chromebook. :) Presumably the custom name for a context would be passable via the virtio-gpu protocol or something?

It is part of the context-type specific protocol. Ie. some parts of the protocol are "core" and dealt with in virtgpu guest kernel driver. But on top of that there are various context-types with their own protocol (ie. virgl, venus, cross-domain, msm native ctx, and some WIP native ctx types floating around)

BR,
-R

Quoted 47 lines

Regards,

Tvrtko

Quoted 43 lines

BR,
-R

Quoted 39 lines

Regards,

Tvrtko

Quoted 35 lines
Documentation/gpu/drm-usage-stats.rst +1 -0
+
   Implementation Details
   ======================
 
drivers/gpu/drm/msm/msm_gpu.c +14 -0
index f0f4f845c32d..1150dcbf28aa 100644
@@ -148,12 +148,26 @@ int msm_gpu_pm_suspend(struct msm_gpu *gpu)
       return 0;
   }
 
+static void get_comm_cmdline(struct msm_file_private *ctx, char **comm, char **cmd);
+
   void msm_gpu_show_fdinfo(struct msm_gpu *gpu, struct msm_file_private *ctx,
                        struct drm_printer *p)
   {
+     char *comm, *cmdline;
+
+     get_comm_cmdline(ctx, &comm, &cmdline);
+
       drm_printf(p, "drm-engine-gpu:\t%llu ns\n", ctx->elapsed_ns);
       drm_printf(p, "drm-cycles-gpu:\t%llu\n", ctx->cycles);
       drm_printf(p, "drm-maxfreq-gpu:\t%u Hz\n", gpu->fast_rate);
+
+     if (comm)
+             drm_printf(p, "drm-comm:\t%s\n", comm);
+     if (cmdline)
+             drm_printf(p, "drm-cmdline:\t%s\n", cmdline);
+
+     kfree(comm);
+     kfree(cmdline);
   }
 
   int msm_gpu_hw_init(struct msm_gpu *gpu)
author
#16 lore raw
In reply to Rob Clark
Quoted 9 lines · Rob Clark

From: Rob Clark <[email protected]>

When many of the things using the GPU are processes in a VM guest, the actual client process is just a proxy. The msm driver has a way to let the proxy tell the kernel the actual VM client process's executable name and command-line, which has until now been used simply for GPU crash devcore dumps. Lets also expose this via fdinfo so that tools can expose who the actual user of the GPU is.

I should have also mentioned, in the VM/proxy scenario we have a single process with separate drm_file's for each guest VM process. So it isn't an option to just change the proxy process's name to match the client.

Quoted 16 lines

Rob Clark (3):
drm/doc: Relax fdinfo string constraints
drm/msm: Rework get_comm_cmdline() helper
drm/msm: Add comm/cmdline fields

 Documentation/gpu/drm-usage-stats.rst   | 37 +++++++++++++++----------
 drivers/gpu/drm/msm/adreno/adreno_gpu.c |  4 +--
 drivers/gpu/drm/msm/msm_drv.c           |  2 ++
 drivers/gpu/drm/msm/msm_gpu.c           | 27 +++++++++++++-----
 drivers/gpu/drm/msm/msm_gpu.h           | 12 ++++++--
 drivers/gpu/drm/msm/msm_submitqueue.c   |  1 +
 6 files changed, 58 insertions(+), 25 deletions(-)

--
2.39.2

#17 lore raw
In reply to Rob Clark

Looks like the 'PATCH' part of your subject was cut off!

Konrad

Quoted 22 lines · Rob Clark

From: Rob Clark <[email protected]>

When many of the things using the GPU are processes in a VM guest, the actual client process is just a proxy. The msm driver has a way to let the proxy tell the kernel the actual VM client process's executable name and command-line, which has until now been used simply for GPU crash devcore dumps. Lets also expose this via fdinfo so that tools can expose who the actual user of the GPU is.

Rob Clark (3):
drm/doc: Relax fdinfo string constraints
drm/msm: Rework get_comm_cmdline() helper
drm/msm: Add comm/cmdline fields

 Documentation/gpu/drm-usage-stats.rst   | 37 +++++++++++++++----------
 drivers/gpu/drm/msm/adreno/adreno_gpu.c |  4 +--
 drivers/gpu/drm/msm/msm_drv.c           |  2 ++
 drivers/gpu/drm/msm/msm_gpu.c           | 27 +++++++++++++-----
 drivers/gpu/drm/msm/msm_gpu.h           | 12 ++++++--
 drivers/gpu/drm/msm/msm_submitqueue.c   |  1 +
 6 files changed, 58 insertions(+), 25 deletions(-)