[PATCH 0/2] drm/v3d: Expose global and per-context GPU reset counters

[PATCH 0/2] drm/v3d: Expose global and per-context GPU reset counters

[Maíra Canal]

The GL extension KHR_robustness requires a mechanism for a GL application
to learn about graphics resets that affect a GL context. With the goal
to provide support for such extension in Mesa, this series implements
global and per-context GPU reset counters that user-space can retrieve
using the DRM_IOCTL_V3D_GET_PARAM ioctl.

The corresponding user-space implementation for this series is available at [1].

[1] https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35965

Best Regards,
- Maíra

---
Maíra Canal (2):
      drm/v3d: Add parameter to retrieve the global number of GPU resets
      drm/v3d: Add parameter to retrieve the number of GPU resets per-fd

 drivers/gpu/drm/v3d/v3d_drv.c   | 11 +++++++++++
 drivers/gpu/drm/v3d/v3d_drv.h   | 11 +++++++++++
 drivers/gpu/drm/v3d/v3d_sched.c |  5 +++++
 include/uapi/drm/v3d_drm.h      |  2 ++
 4 files changed, 29 insertions(+)
---
base-commit: 917b10d90990fd2138b5dbc2d22cfa428c070ade
change-id: 20250708-v3d-reset-counter-ab231c87bd35

[PATCH 1/2] drm/v3d: Add parameter to retrieve the global number of GPU resets

[Maíra Canal]

The GL extension KHR_robustness uses the number of global and per-context
GPU resets to learn about graphics resets that affect a GL context. This
commit introduces a new V3D parameter to retrieve the global number of
GPU resets that have happened since the driver was probed.

To retrieve this information, user-space must use DRM_V3D_PARAM_GLOBAL_RESET_COUNTER.

Signed-off-by: Maíra Canal <mcanal@igalia.com>
---
 drivers/gpu/drm/v3d/v3d_drv.c   | 5 +++++
 drivers/gpu/drm/v3d/v3d_drv.h   | 5 +++++
 drivers/gpu/drm/v3d/v3d_sched.c | 2 ++
 include/uapi/drm/v3d_drm.h      | 1 +
 4 files changed, 13 insertions(+)

diff --git a/drivers/gpu/drm/v3d/v3d_drv.c b/drivers/gpu/drm/v3d/v3d_drv.c
index 5e997ae8bc9c7ac110d79df8b5cf81a59dc0a291..0f53ca7460044e2a74c31d6282914f49844066b2 100644
--- a/drivers/gpu/drm/v3d/v3d_drv.c
+++ b/drivers/gpu/drm/v3d/v3d_drv.c
@@ -107,6 +107,11 @@ static int v3d_get_param_ioctl(struct drm_device *dev, void *data,
 	case DRM_V3D_PARAM_SUPPORTS_SUPER_PAGES:
 		args->value = !!v3d->gemfs;
 		return 0;
+	case DRM_V3D_PARAM_GLOBAL_RESET_COUNTER:
+		mutex_lock(&v3d->reset_lock);
+		args->value = v3d->reset_counter;
+		mutex_unlock(&v3d->reset_lock);
+		return 0;
 	default:
 		DRM_DEBUG("Unknown parameter %d\n", args->param);
 		return -EINVAL;
diff --git a/drivers/gpu/drm/v3d/v3d_drv.h b/drivers/gpu/drm/v3d/v3d_drv.h
index b51f0b648a08011f737317ec1841d5ab316355b2..074cbccf88ddfe5961c779f0df93552264f54608 100644
--- a/drivers/gpu/drm/v3d/v3d_drv.h
+++ b/drivers/gpu/drm/v3d/v3d_drv.h
@@ -196,6 +196,11 @@ struct v3d_dev {
 	 * all jobs.
 	 */
 	struct v3d_perfmon *global_perfmon;
+
+	/* Global reset counter. The counter must be incremented when
+	 * a GPU reset happens. It must be protected by @reset_lock.
+	 */
+	unsigned int reset_counter;
 };
 
 static inline struct v3d_dev *
diff --git a/drivers/gpu/drm/v3d/v3d_sched.c b/drivers/gpu/drm/v3d/v3d_sched.c
index 35f131a46d0701cc8040d3b9654595a2bc260eab..29845917d7df5f9ff996fa52a728ff5d19dea90f 100644
--- a/drivers/gpu/drm/v3d/v3d_sched.c
+++ b/drivers/gpu/drm/v3d/v3d_sched.c
@@ -731,6 +731,8 @@ v3d_gpu_reset_for_timeout(struct v3d_dev *v3d, struct drm_sched_job *sched_job)
 	/* get the GPU back into the init state */
 	v3d_reset(v3d);
 
+	v3d->reset_counter++;
+
 	for (q = 0; q < V3D_MAX_QUEUES; q++)
 		drm_sched_resubmit_jobs(&v3d->queue[q].sched);
 
diff --git a/include/uapi/drm/v3d_drm.h b/include/uapi/drm/v3d_drm.h
index dbbc404d2b3dd6c8b8e3969b0b76b7d8b371dddb..0a7ce2f6be19d2cbccf090f0db301119eb4b28ba 100644
--- a/include/uapi/drm/v3d_drm.h
+++ b/include/uapi/drm/v3d_drm.h
@@ -294,6 +294,7 @@ enum drm_v3d_param {
 	DRM_V3D_PARAM_SUPPORTS_CPU_QUEUE,
 	DRM_V3D_PARAM_MAX_PERF_COUNTERS,
 	DRM_V3D_PARAM_SUPPORTS_SUPER_PAGES,
+	DRM_V3D_PARAM_GLOBAL_RESET_COUNTER,
 };
 
 struct drm_v3d_get_param {

-- 
2.50.0

[PATCH 2/2] drm/v3d: Add parameter to retrieve the number of GPU resets per-fd

[Maíra Canal]

The GL extension KHR_robustness uses the number of global and per-context
GPU resets to learn about graphics resets that affect a GL context. This
commit introduces a new V3D parameter to retrieve the number of GPU resets
triggered by jobs submitted through a file descriptor.

To retrieve this information, user-space must use DRM_V3D_PARAM_CONTEXT_RESET_COUNTER.

Signed-off-by: Maíra Canal <mcanal@igalia.com>
---
 drivers/gpu/drm/v3d/v3d_drv.c   | 6 ++++++
 drivers/gpu/drm/v3d/v3d_drv.h   | 6 ++++++
 drivers/gpu/drm/v3d/v3d_sched.c | 3 +++
 include/uapi/drm/v3d_drm.h      | 1 +
 4 files changed, 16 insertions(+)

diff --git a/drivers/gpu/drm/v3d/v3d_drv.c b/drivers/gpu/drm/v3d/v3d_drv.c
index 0f53ca7460044e2a74c31d6282914f49844066b2..2def155ce496ec5f159f6bda9929aeaae141d1a6 100644
--- a/drivers/gpu/drm/v3d/v3d_drv.c
+++ b/drivers/gpu/drm/v3d/v3d_drv.c
@@ -46,6 +46,7 @@ MODULE_PARM_DESC(super_pages, "Enable/Disable Super Pages support.");
 static int v3d_get_param_ioctl(struct drm_device *dev, void *data,
 			       struct drm_file *file_priv)
 {
+	struct v3d_file_priv *v3d_priv = file_priv->driver_priv;
 	struct v3d_dev *v3d = to_v3d_dev(dev);
 	struct drm_v3d_get_param *args = data;
 	static const u32 reg_map[] = {
@@ -112,6 +113,11 @@ static int v3d_get_param_ioctl(struct drm_device *dev, void *data,
 		args->value = v3d->reset_counter;
 		mutex_unlock(&v3d->reset_lock);
 		return 0;
+	case DRM_V3D_PARAM_CONTEXT_RESET_COUNTER:
+		mutex_lock(&v3d->reset_lock);
+		args->value = v3d_priv->reset_counter;
+		mutex_unlock(&v3d->reset_lock);
+		return 0;
 	default:
 		DRM_DEBUG("Unknown parameter %d\n", args->param);
 		return -EINVAL;
diff --git a/drivers/gpu/drm/v3d/v3d_drv.h b/drivers/gpu/drm/v3d/v3d_drv.h
index 074cbccf88ddfe5961c779f0df93552264f54608..dabda7aaf00074d8de42dcdb345d5f3331ac13b2 100644
--- a/drivers/gpu/drm/v3d/v3d_drv.h
+++ b/drivers/gpu/drm/v3d/v3d_drv.h
@@ -230,6 +230,12 @@ struct v3d_file_priv {
 
 	/* Stores the GPU stats for a specific queue for this fd. */
 	struct v3d_stats stats[V3D_MAX_QUEUES];
+
+	/* Per-fd reset counter, must be incremented when a job submitted
+	 * by this fd causes a GPU reset. It must be protected by
+	 * &struct v3d_dev->reset_lock.
+	 */
+	unsigned int reset_counter;
 };
 
 struct v3d_bo {
diff --git a/drivers/gpu/drm/v3d/v3d_sched.c b/drivers/gpu/drm/v3d/v3d_sched.c
index 29845917d7df5f9ff996fa52a728ff5d19dea90f..46003457a06e3c01637b2c0ed392660e4e569b4a 100644
--- a/drivers/gpu/drm/v3d/v3d_sched.c
+++ b/drivers/gpu/drm/v3d/v3d_sched.c
@@ -717,6 +717,8 @@ v3d_cache_clean_job_run(struct drm_sched_job *sched_job)
 static enum drm_gpu_sched_stat
 v3d_gpu_reset_for_timeout(struct v3d_dev *v3d, struct drm_sched_job *sched_job)
 {
+	struct v3d_job *job = to_v3d_job(sched_job);
+	struct v3d_file_priv *v3d_priv = job->file->driver_priv;
 	enum v3d_queue q;
 
 	mutex_lock(&v3d->reset_lock);
@@ -732,6 +734,7 @@ v3d_gpu_reset_for_timeout(struct v3d_dev *v3d, struct drm_sched_job *sched_job)
 	v3d_reset(v3d);
 
 	v3d->reset_counter++;
+	v3d_priv->reset_counter++;
 
 	for (q = 0; q < V3D_MAX_QUEUES; q++)
 		drm_sched_resubmit_jobs(&v3d->queue[q].sched);
diff --git a/include/uapi/drm/v3d_drm.h b/include/uapi/drm/v3d_drm.h
index 0a7ce2f6be19d2cbccf090f0db301119eb4b28ba..d9b01f4c3a047eac90899901cc603838ab5c9090 100644
--- a/include/uapi/drm/v3d_drm.h
+++ b/include/uapi/drm/v3d_drm.h
@@ -295,6 +295,7 @@ enum drm_v3d_param {
 	DRM_V3D_PARAM_MAX_PERF_COUNTERS,
 	DRM_V3D_PARAM_SUPPORTS_SUPER_PAGES,
 	DRM_V3D_PARAM_GLOBAL_RESET_COUNTER,
+	DRM_V3D_PARAM_CONTEXT_RESET_COUNTER,
 };
 
 struct drm_v3d_get_param {

-- 
2.50.0

Re: [PATCH 0/2] drm/v3d: Expose global and per-context GPU reset counters

[Iago Toral]

Looks good to me:
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>

Iago

El vie, 11-07-2025 a las 12:18 -0300, Maíra Canal escribió:
> The GL extension KHR_robustness requires a mechanism for a GL
> application
> to learn about graphics resets that affect a GL context. With the
> goal
> to provide support for such extension in Mesa, this series implements
> global and per-context GPU reset counters that user-space can
> retrieve
> using the DRM_IOCTL_V3D_GET_PARAM ioctl.
> 
> The corresponding user-space implementation for this series is
> available at [1].
> 
> [1] https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35965
> 
> Best Regards,
> - Maíra
> 
> ---
> Maíra Canal (2):
>       drm/v3d: Add parameter to retrieve the global number of GPU
> resets
>       drm/v3d: Add parameter to retrieve the number of GPU resets
> per-fd
> 
>  drivers/gpu/drm/v3d/v3d_drv.c   | 11 +++++++++++
>  drivers/gpu/drm/v3d/v3d_drv.h   | 11 +++++++++++
>  drivers/gpu/drm/v3d/v3d_sched.c |  5 +++++
>  include/uapi/drm/v3d_drm.h      |  2 ++
>  4 files changed, 29 insertions(+)
> ---
> base-commit: 917b10d90990fd2138b5dbc2d22cfa428c070ade
> change-id: 20250708-v3d-reset-counter-ab231c87bd35
> 
> 

Re: [PATCH 0/2] drm/v3d: Expose global and per-context GPU reset counters

[Maíra Canal]

On Fri, 11 Jul 2025 12:18:30 -0300, Maíra Canal wrote:
> The GL extension KHR_robustness requires a mechanism for a GL application
> to learn about graphics resets that affect a GL context. With the goal
> to provide support for such extension in Mesa, this series implements
> global and per-context GPU reset counters that user-space can retrieve
> using the DRM_IOCTL_V3D_GET_PARAM ioctl.
> 
> The corresponding user-space implementation for this series is available at [1].
> 
> [...]

Applied to drm-misc/drm-misc-next, thanks!

[1/2] drm/v3d: Add parameter to retrieve the global number of GPU resets
      commit: 5774b3cfdedb3624ef0d2c82cccbfd61bcb60fd5
[2/2] drm/v3d: Add parameter to retrieve the number of GPU resets per-fd
      commit: 769c153cfc3c6669c7b318f66c2b21ec3951fb4a

Best regards,
- Maíra