Hi Christian,
The patch overall LGTM, I just have one small nit.
On 17/11/24 18:41, Christian Gmeiner wrote:
From: Christian Gmeiner <[email protected]>
Add a new ioctl, DRM_IOCTL_V3D_PERFMON_SET_GLOBAL, to allow
configuration of a global performance monitor (perfmon).
Use the global perfmon for all jobs to ensure consistent
performance tracking across submissions. This feature is
needed to implement a Perfetto datasources in user-space.
Signed-off-by: Christian Gmeiner <[email protected]>
---
Changes in v3:
- Reworked commit message.
- Refined some code comments.
- Added missing v3d_perfmon_stop(..) call to v3d_perfmon_destroy_ioctl(..).
Changes in v2:
- Reworked commit message.
- Removed num_perfmon counter for tracking perfmon allocations.
- Allowing allocation of perfmons when the global perfmon is active.
- Return -EAGAIN for submissions with a per job perfmon if the global perfmon
is active.
---
---
drivers/gpu/drm/v3d/v3d_drv.c | 1 +
drivers/gpu/drm/v3d/v3d_drv.h | 8 ++++++
drivers/gpu/drm/v3d/v3d_perfmon.c | 42 +++++++++++++++++++++++++++++++
drivers/gpu/drm/v3d/v3d_sched.c | 14 ++++++++---
drivers/gpu/drm/v3d/v3d_submit.c | 10 ++++++++
include/uapi/drm/v3d_drm.h | 15 +++++++++++
6 files changed, 87 insertions(+), 3 deletions(-)
diff --git a/drivers/gpu/drm/v3d/v3d_drv.c b/drivers/gpu/drm/v3d/v3d_drv.c
index d7ff1f5fa481..3c89f0daa5b8 100644
--- a/drivers/gpu/drm/v3d/v3d_drv.c
+++ b/drivers/gpu/drm/v3d/v3d_drv.c
@@ -214,6 +214,7 @@ static const struct drm_ioctl_desc v3d_drm_ioctls[] = {
DRM_IOCTL_DEF_DRV(V3D_PERFMON_GET_VALUES, v3d_perfmon_get_values_ioctl,
DRM_RENDER_ALLOW),
DRM_IOCTL_DEF_DRV(V3D_SUBMIT_CPU, v3d_submit_cpu_ioctl,
DRM_RENDER_ALLOW | DRM_AUTH),
DRM_IOCTL_DEF_DRV(V3D_PERFMON_GET_COUNTER,
v3d_perfmon_get_counter_ioctl, DRM_RENDER_ALLOW),
+ DRM_IOCTL_DEF_DRV(V3D_PERFMON_SET_GLOBAL, v3d_perfmon_set_global_ioctl,
DRM_RENDER_ALLOW),
};
static const struct drm_driver v3d_drm_driver = {
diff --git a/drivers/gpu/drm/v3d/v3d_drv.h b/drivers/gpu/drm/v3d/v3d_drv.h
index cf4b23369dc4..a0d920ec2b1d 100644
--- a/drivers/gpu/drm/v3d/v3d_drv.h
+++ b/drivers/gpu/drm/v3d/v3d_drv.h
@@ -179,6 +179,12 @@ struct v3d_dev {
u32 num_allocated;
u32 pages_allocated;
} bo_stats;
+
+ /* To support a performance analysis tool in user space, we require
+ * a single, globally configured performance monitor (perfmon) for
+ * all jobs.
+ */
+ struct v3d_perfmon *global_perfmon;
};
static inline struct v3d_dev *
@@ -584,6 +590,8 @@ int v3d_perfmon_get_values_ioctl(struct drm_device *dev,
void *data,
struct drm_file *file_priv);
int v3d_perfmon_get_counter_ioctl(struct drm_device *dev, void *data,
struct drm_file *file_priv);
+int v3d_perfmon_set_global_ioctl(struct drm_device *dev, void *data,
+ struct drm_file *file_priv);
/* v3d_sysfs.c */
int v3d_sysfs_init(struct device *dev);
diff --git a/drivers/gpu/drm/v3d/v3d_perfmon.c
b/drivers/gpu/drm/v3d/v3d_perfmon.c
index 156be13ab2ef..df72171d8c80 100644
--- a/drivers/gpu/drm/v3d/v3d_perfmon.c
+++ b/drivers/gpu/drm/v3d/v3d_perfmon.c
@@ -312,6 +312,9 @@ static int v3d_perfmon_idr_del(int id, void *elem, void
*data)
if (perfmon == v3d->active_perfmon)
v3d_perfmon_stop(v3d, perfmon, false);
+ /* If the global perfmon is being destroyed, set it to NULL */
+ cmpxchg(&v3d->global_perfmon, perfmon, NULL);
+
v3d_perfmon_put(perfmon);
return 0;
@@ -383,6 +386,7 @@ int v3d_perfmon_destroy_ioctl(struct drm_device *dev, void
*data,
{
struct v3d_file_priv *v3d_priv = file_priv->driver_priv;
struct drm_v3d_perfmon_destroy *req = data;
+ struct v3d_dev *v3d = v3d_priv->v3d;
struct v3d_perfmon *perfmon;
mutex_lock(&v3d_priv->perfmon.lock);
@@ -392,6 +396,13 @@ int v3d_perfmon_destroy_ioctl(struct drm_device *dev, void
*data,
if (!perfmon)
return -EINVAL;
+ /* If the active perfmon is being destroyed, stop it first */
+ if (perfmon == v3d->active_perfmon)
+ v3d_perfmon_stop(v3d, perfmon, false);
Could you please move this snippet to different patch? This snippet is a
global fix as it can prevent NULL pointer dereferences triggered by the
user-space. I would like to push this snippet to drm-misc-fixes.
+
+ /* If the global perfmon is being destroyed, set it to NULL */
+ cmpxchg(&v3d->global_perfmon, perfmon, NULL);
+
v3d_perfmon_put(perfmon);
return 0;
@@ -451,3 +462,34 @@ int v3d_perfmon_get_counter_ioctl(struct drm_device *dev,
void *data,
return 0;
}
+
+int v3d_perfmon_set_global_ioctl(struct drm_device *dev, void *data,
+ struct drm_file *file_priv)
+{
+ struct v3d_file_priv *v3d_priv = file_priv->driver_priv;
+ struct drm_v3d_perfmon_set_global *req = data;
+ struct v3d_dev *v3d = to_v3d_dev(dev);
+ struct v3d_perfmon *perfmon;
+
+ if (req->flags & ~DRM_V3D_PERFMON_CLEAR_GLOBAL)
+ return -EINVAL;
+
+ perfmon = v3d_perfmon_find(v3d_priv, req->id);
+ if (!perfmon)
+ return -EINVAL;
+
+ /* If the request is to clear the global performance monitor */
+ if (req->flags & DRM_V3D_PERFMON_CLEAR_GLOBAL) {
+ if (!v3d->global_perfmon)
+ return -EINVAL;
+
+ xchg(&v3d->global_perfmon, NULL);
What if v3d->global_perfmon is running?
+
+ return 0;
+ }
+
+ if (cmpxchg(&v3d->global_perfmon, NULL, perfmon))
+ return -EBUSY;
+
+ return 0;
+}
[...]
diff --git a/include/uapi/drm/v3d_drm.h b/include/uapi/drm/v3d_drm.h
index 87fc5bb0a61e..7c2ad9d0270d 100644
--- a/include/uapi/drm/v3d_drm.h
+++ b/include/uapi/drm/v3d_drm.h
@@ -43,6 +43,7 @@ extern "C" {
#define DRM_V3D_PERFMON_GET_VALUES 0x0a
#define DRM_V3D_SUBMIT_CPU 0x0b
#define DRM_V3D_PERFMON_GET_COUNTER 0x0c
+#define DRM_V3D_PERFMON_SET_GLOBAL 0x0d
#define DRM_IOCTL_V3D_SUBMIT_CL DRM_IOWR(DRM_COMMAND_BASE + DRM_V3D_SUBMIT_CL, struct drm_v3d_submit_cl)
#define DRM_IOCTL_V3D_WAIT_BO DRM_IOWR(DRM_COMMAND_BASE +
DRM_V3D_WAIT_BO, struct drm_v3d_wait_bo)
@@ -61,6 +62,8 @@ extern "C" {
#define DRM_IOCTL_V3D_SUBMIT_CPU DRM_IOW(DRM_COMMAND_BASE +
DRM_V3D_SUBMIT_CPU, struct drm_v3d_submit_cpu)
#define DRM_IOCTL_V3D_PERFMON_GET_COUNTER DRM_IOWR(DRM_COMMAND_BASE +
DRM_V3D_PERFMON_GET_COUNTER, \
struct
drm_v3d_perfmon_get_counter)
+#define DRM_IOCTL_V3D_PERFMON_SET_GLOBAL DRM_IOW(DRM_COMMAND_BASE +
DRM_V3D_PERFMON_SET_GLOBAL, \
+ struct
drm_v3d_perfmon_set_global)
#define DRM_V3D_SUBMIT_CL_FLUSH_CACHE 0x01
#define DRM_V3D_SUBMIT_EXTENSION 0x02
@@ -765,6 +768,18 @@ struct drm_v3d_perfmon_get_counter {
__u8 reserved[7];
};
+#define DRM_V3D_PERFMON_CLEAR_GLOBAL 0x0001
+
+/**
+ * struct drm_v3d_perfmon_set_global - ioctl to define a global performance
+ * monitor that is used for all job. If a global performance monitor is
s/job/jobs
Best Regards,
- Maíra
+ * defined, jobs with a self-defined performance monitor are not allowed.
+ */
+struct drm_v3d_perfmon_set_global {
+ __u32 flags;
+ __u32 id;
+};
+
#if defined(__cplusplus)
}
#endif
