[PATCH] intel: Add support for GPU reset status query ioctl

[PATCH] intel: Add support for GPU reset status query ioctl

[Ian Romanick]

From: Ian Romanick <ian.d.romanick@intel.com>

I would have just used the drmIoctl interface directly in Mesa, but the
ioctl needs some data from the drm_intel_context that is not exposed
outside libdrm.

v2: Update based on Mika's kernel work.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
---
 include/drm/i915_drm.h   | 17 +++++++++++++++++
 intel/intel_bufmgr.h     |  5 +++++
 intel/intel_bufmgr_gem.c | 34 ++++++++++++++++++++++++++++++++++
 3 files changed, 56 insertions(+)

diff --git a/include/drm/i915_drm.h b/include/drm/i915_drm.h
index aa983f3..008e4ed 100644
--- a/include/drm/i915_drm.h
+++ b/include/drm/i915_drm.h
@@ -198,6 +198,7 @@ typedef struct _drm_i915_sarea {
 #define DRM_I915_GEM_SET_CACHEING	0x2f
 #define DRM_I915_GEM_GET_CACHEING	0x30
 #define DRM_I915_REG_READ		0x31
+#define DRM_I915_GET_RESET_STATUS	0x32
 
 #define DRM_IOCTL_I915_INIT		DRM_IOW( DRM_COMMAND_BASE + DRM_I915_INIT, drm_i915_init_t)
 #define DRM_IOCTL_I915_FLUSH		DRM_IO ( DRM_COMMAND_BASE + DRM_I915_FLUSH)
@@ -247,6 +248,7 @@ typedef struct _drm_i915_sarea {
 #define DRM_IOCTL_I915_GEM_CONTEXT_CREATE	DRM_IOWR (DRM_COMMAND_BASE + DRM_I915_GEM_CONTEXT_CREATE, struct drm_i915_gem_context_create)
 #define DRM_IOCTL_I915_GEM_CONTEXT_DESTROY	DRM_IOW (DRM_COMMAND_BASE + DRM_I915_GEM_CONTEXT_DESTROY, struct drm_i915_gem_context_destroy)
 #define DRM_IOCTL_I915_REG_READ			DRM_IOWR (DRM_COMMAND_BASE + DRM_I915_REG_READ, struct drm_i915_reg_read)
+#define DRM_IOCTL_I915_GET_RESET_STATUS		DRM_IOWR (DRM_COMMAND_BASE + DRM_I915_GET_RESET_STATUS, struct drm_i915_reset_stats)
 
 /* Allow drivers to submit batchbuffers directly to hardware, relying
  * on the security mechanisms provided by hardware.
@@ -943,4 +945,19 @@ struct drm_i915_reg_read {
 	__u64 offset;
 	__u64 val; /* Return value */
 };
+
+struct drm_i915_reset_stats {
+	__u32 ctx_id;
+	__u32 flags;
+
+	/* For all contexts */
+	__u32 reset_count;
+
+	/* For this context */
+	__u32 batch_active;
+	__u32 batch_pending;
+
+	__u32 pad;
+};
+
 #endif				/* _I915_DRM_H_ */
diff --git a/intel/intel_bufmgr.h b/intel/intel_bufmgr.h
index 15f818e..2eb9742 100644
--- a/intel/intel_bufmgr.h
+++ b/intel/intel_bufmgr.h
@@ -248,6 +248,11 @@ int drm_intel_reg_read(drm_intel_bufmgr *bufmgr,
 		       uint32_t offset,
 		       uint64_t *result);
 
+int drm_intel_get_reset_stats(drm_intel_context *ctx,
+			      uint32_t *reset_count,
+			      uint32_t *active,
+			      uint32_t *pending);
+
 /** @{ Compatibility defines to keep old code building despite the symbol rename
  * from dri_* to drm_intel_*
  */
diff --git a/intel/intel_bufmgr_gem.c b/intel/intel_bufmgr_gem.c
index 278f5c8..f2c8f0d 100644
--- a/intel/intel_bufmgr_gem.c
+++ b/intel/intel_bufmgr_gem.c
@@ -3009,6 +3009,40 @@ drm_intel_gem_context_destroy(drm_intel_context *ctx)
 }
 
 int
+drm_intel_get_reset_stats(drm_intel_context *ctx,
+			  uint32_t *reset_count,
+			  uint32_t *active,
+			  uint32_t *pending)
+{
+	drm_intel_bufmgr_gem *bufmgr_gem;
+	struct drm_i915_reset_stats stats;
+	int ret;
+
+	if (ctx == NULL)
+		return -EINVAL;
+
+	VG_CLEAR(stats);
+
+	bufmgr_gem = (drm_intel_bufmgr_gem *)ctx->bufmgr;
+	reset_status.ctx_id = ctx->ctx_id;
+	ret = drmIoctl(bufmgr_gem->fd,
+		       DRM_IOCTL_I915_GET_RESET_STATS,
+		       &stats);
+	if (ret == 0) {
+		if (reset_count != NULL)
+			*reset_count = stats.reset_count;
+
+		if (active != NULL)
+			*active = stats.batch_active;
+
+		if (pending != NULL)
+			*pending = stats.batch_pending;
+	}
+
+	return ret;
+}
+
+int
 drm_intel_reg_read(drm_intel_bufmgr *bufmgr,
 		   uint32_t offset,
 		   uint64_t *result)
-- 
1.8.1.4

[PATCH v3] intel: Add support for GPU reset status query ioctl

[Ian Romanick]

From: Ian Romanick <ian.d.romanick@intel.com>

I would have just used the drmIoctl interface directly in Mesa, but the
ioctl needs some data from the drm_intel_context that is not exposed
outside libdrm.

v2: Update based on Mika's kernel work.

v3: Fix compile failures from last-minute typos.  Sigh.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
---
 include/drm/i915_drm.h   | 17 +++++++++++++++++
 intel/intel_bufmgr.h     |  5 +++++
 intel/intel_bufmgr_gem.c | 34 ++++++++++++++++++++++++++++++++++
 3 files changed, 56 insertions(+)

diff --git a/include/drm/i915_drm.h b/include/drm/i915_drm.h
index aa983f3..2703ded 100644
--- a/include/drm/i915_drm.h
+++ b/include/drm/i915_drm.h
@@ -198,6 +198,7 @@ typedef struct _drm_i915_sarea {
 #define DRM_I915_GEM_SET_CACHEING	0x2f
 #define DRM_I915_GEM_GET_CACHEING	0x30
 #define DRM_I915_REG_READ		0x31
+#define DRM_I915_GET_RESET_STATS	0x32
 
 #define DRM_IOCTL_I915_INIT		DRM_IOW( DRM_COMMAND_BASE + DRM_I915_INIT, drm_i915_init_t)
 #define DRM_IOCTL_I915_FLUSH		DRM_IO ( DRM_COMMAND_BASE + DRM_I915_FLUSH)
@@ -247,6 +248,7 @@ typedef struct _drm_i915_sarea {
 #define DRM_IOCTL_I915_GEM_CONTEXT_CREATE	DRM_IOWR (DRM_COMMAND_BASE + DRM_I915_GEM_CONTEXT_CREATE, struct drm_i915_gem_context_create)
 #define DRM_IOCTL_I915_GEM_CONTEXT_DESTROY	DRM_IOW (DRM_COMMAND_BASE + DRM_I915_GEM_CONTEXT_DESTROY, struct drm_i915_gem_context_destroy)
 #define DRM_IOCTL_I915_REG_READ			DRM_IOWR (DRM_COMMAND_BASE + DRM_I915_REG_READ, struct drm_i915_reg_read)
+#define DRM_IOCTL_I915_GET_RESET_STATS		DRM_IOWR (DRM_COMMAND_BASE + DRM_I915_GET_RESET_STATS, struct drm_i915_reset_stats)
 
 /* Allow drivers to submit batchbuffers directly to hardware, relying
  * on the security mechanisms provided by hardware.
@@ -943,4 +945,19 @@ struct drm_i915_reg_read {
 	__u64 offset;
 	__u64 val; /* Return value */
 };
+
+struct drm_i915_reset_stats {
+	__u32 ctx_id;
+	__u32 flags;
+
+	/* For all contexts */
+	__u32 reset_count;
+
+	/* For this context */
+	__u32 batch_active;
+	__u32 batch_pending;
+
+	__u32 pad;
+};
+
 #endif				/* _I915_DRM_H_ */
diff --git a/intel/intel_bufmgr.h b/intel/intel_bufmgr.h
index 15f818e..2eb9742 100644
--- a/intel/intel_bufmgr.h
+++ b/intel/intel_bufmgr.h
@@ -248,6 +248,11 @@ int drm_intel_reg_read(drm_intel_bufmgr *bufmgr,
 		       uint32_t offset,
 		       uint64_t *result);
 
+int drm_intel_get_reset_stats(drm_intel_context *ctx,
+			      uint32_t *reset_count,
+			      uint32_t *active,
+			      uint32_t *pending);
+
 /** @{ Compatibility defines to keep old code building despite the symbol rename
  * from dri_* to drm_intel_*
  */
diff --git a/intel/intel_bufmgr_gem.c b/intel/intel_bufmgr_gem.c
index 278f5c8..437d72e 100644
--- a/intel/intel_bufmgr_gem.c
+++ b/intel/intel_bufmgr_gem.c
@@ -3009,6 +3009,40 @@ drm_intel_gem_context_destroy(drm_intel_context *ctx)
 }
 
 int
+drm_intel_get_reset_stats(drm_intel_context *ctx,
+			  uint32_t *reset_count,
+			  uint32_t *active,
+			  uint32_t *pending)
+{
+	drm_intel_bufmgr_gem *bufmgr_gem;
+	struct drm_i915_reset_stats stats;
+	int ret;
+
+	if (ctx == NULL)
+		return -EINVAL;
+
+	VG_CLEAR(stats);
+
+	bufmgr_gem = (drm_intel_bufmgr_gem *)ctx->bufmgr;
+	stats.ctx_id = ctx->ctx_id;
+	ret = drmIoctl(bufmgr_gem->fd,
+		       DRM_IOCTL_I915_GET_RESET_STATS,
+		       &stats);
+	if (ret == 0) {
+		if (reset_count != NULL)
+			*reset_count = stats.reset_count;
+
+		if (active != NULL)
+			*active = stats.batch_active;
+
+		if (pending != NULL)
+			*pending = stats.batch_pending;
+	}
+
+	return ret;
+}
+
+int
 drm_intel_reg_read(drm_intel_bufmgr *bufmgr,
 		   uint32_t offset,
 		   uint64_t *result)
-- 
1.8.1.4

Re: [PATCH v3] intel: Add support for GPU reset status query ioctl

[Kenneth Graunke]

On 10/29/2013 06:02 PM, Ian Romanick wrote:
> From: Ian Romanick <ian.d.romanick@intel.com>
> 
> I would have just used the drmIoctl interface directly in Mesa, but the
> ioctl needs some data from the drm_intel_context that is not exposed
> outside libdrm.
> 
> v2: Update based on Mika's kernel work.
> 
> v3: Fix compile failures from last-minute typos.  Sigh.
> 
> Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
> Cc: Mika Kuoppala <mika.kuoppala@intel.com>
> Cc: Daniel Vetter <daniel.vetter@ffwll.ch>

Assuming Mika's kernel patch actually lands, and the ioctl numbers don't
get out of sync, this is:

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

Once it lands, we'll need to do a new release of libdrm.

[PATCH] intel: Add support for GPU reset status query ioctl

[Ian Romanick]

From: Ian Romanick <ian.d.romanick@intel.com>

I would have just used the drmIoctl interface directly in Mesa, but the
ioctl needs some data from the drm_intel_context that is not exposed
outside libdrm.

This ioctl is in the drm-intel-next tree as b635991.

v2: Update based on Mika's kernel work.

v3: Fix compile failures from last-minute typos.  Sigh.

v4: Import the actual changes from the kernel i915_drm.h.  Only comments
on some fields of drm_i915_reset_stats differed.  There are still some
deltas between the kernel i915_drm.h and the one in libdrm, but those
can be resolved in other patches.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org> [v3]
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
---
 include/drm/i915_drm.h   | 19 +++++++++++++++++++
 intel/intel_bufmgr.h     |  5 +++++
 intel/intel_bufmgr_gem.c | 34 ++++++++++++++++++++++++++++++++++
 3 files changed, 58 insertions(+)

diff --git a/include/drm/i915_drm.h b/include/drm/i915_drm.h
index aa983f3..c1914d6 100644
--- a/include/drm/i915_drm.h
+++ b/include/drm/i915_drm.h
@@ -198,6 +198,7 @@ typedef struct _drm_i915_sarea {
 #define DRM_I915_GEM_SET_CACHEING	0x2f
 #define DRM_I915_GEM_GET_CACHEING	0x30
 #define DRM_I915_REG_READ		0x31
+#define DRM_I915_GET_RESET_STATS	0x32
 
 #define DRM_IOCTL_I915_INIT		DRM_IOW( DRM_COMMAND_BASE + DRM_I915_INIT, drm_i915_init_t)
 #define DRM_IOCTL_I915_FLUSH		DRM_IO ( DRM_COMMAND_BASE + DRM_I915_FLUSH)
@@ -247,6 +248,7 @@ typedef struct _drm_i915_sarea {
 #define DRM_IOCTL_I915_GEM_CONTEXT_CREATE	DRM_IOWR (DRM_COMMAND_BASE + DRM_I915_GEM_CONTEXT_CREATE, struct drm_i915_gem_context_create)
 #define DRM_IOCTL_I915_GEM_CONTEXT_DESTROY	DRM_IOW (DRM_COMMAND_BASE + DRM_I915_GEM_CONTEXT_DESTROY, struct drm_i915_gem_context_destroy)
 #define DRM_IOCTL_I915_REG_READ			DRM_IOWR (DRM_COMMAND_BASE + DRM_I915_REG_READ, struct drm_i915_reg_read)
+#define DRM_IOCTL_I915_GET_RESET_STATS		DRM_IOWR (DRM_COMMAND_BASE + DRM_I915_GET_RESET_STATS, struct drm_i915_reset_stats)
 
 /* Allow drivers to submit batchbuffers directly to hardware, relying
  * on the security mechanisms provided by hardware.
@@ -943,4 +945,21 @@ struct drm_i915_reg_read {
 	__u64 offset;
 	__u64 val; /* Return value */
 };
+
+struct drm_i915_reset_stats {
+	__u32 ctx_id;
+	__u32 flags;
+
+	/* All resets since boot/module reload, for all contexts */
+	__u32 reset_count;
+
+	/* Number of batches lost when active in GPU, for this context */
+	__u32 batch_active;
+
+	/* Number of batches lost pending for execution, for this context */
+	__u32 batch_pending;
+
+	__u32 pad;
+};
+
 #endif				/* _I915_DRM_H_ */
diff --git a/intel/intel_bufmgr.h b/intel/intel_bufmgr.h
index 7b28a70..34a5750 100644
--- a/intel/intel_bufmgr.h
+++ b/intel/intel_bufmgr.h
@@ -249,6 +249,11 @@ int drm_intel_reg_read(drm_intel_bufmgr *bufmgr,
 		       uint32_t offset,
 		       uint64_t *result);
 
+int drm_intel_get_reset_stats(drm_intel_context *ctx,
+			      uint32_t *reset_count,
+			      uint32_t *active,
+			      uint32_t *pending);
+
 /** @{ Compatibility defines to keep old code building despite the symbol rename
  * from dri_* to drm_intel_*
  */
diff --git a/intel/intel_bufmgr_gem.c b/intel/intel_bufmgr_gem.c
index dbadc52..0b9252e 100644
--- a/intel/intel_bufmgr_gem.c
+++ b/intel/intel_bufmgr_gem.c
@@ -3029,6 +3029,40 @@ drm_intel_gem_context_get_hw_context_id(const drm_intel_context *ctx)
 }
 
 int
+drm_intel_get_reset_stats(drm_intel_context *ctx,
+			  uint32_t *reset_count,
+			  uint32_t *active,
+			  uint32_t *pending)
+{
+	drm_intel_bufmgr_gem *bufmgr_gem;
+	struct drm_i915_reset_stats stats;
+	int ret;
+
+	if (ctx == NULL)
+		return -EINVAL;
+
+	VG_CLEAR(stats);
+
+	bufmgr_gem = (drm_intel_bufmgr_gem *)ctx->bufmgr;
+	stats.ctx_id = ctx->ctx_id;
+	ret = drmIoctl(bufmgr_gem->fd,
+		       DRM_IOCTL_I915_GET_RESET_STATS,
+		       &stats);
+	if (ret == 0) {
+		if (reset_count != NULL)
+			*reset_count = stats.reset_count;
+
+		if (active != NULL)
+			*active = stats.batch_active;
+
+		if (pending != NULL)
+			*pending = stats.batch_pending;
+	}
+
+	return ret;
+}
+
+int
 drm_intel_reg_read(drm_intel_bufmgr *bufmgr,
 		   uint32_t offset,
 		   uint64_t *result)
-- 
1.8.1.4

Re: [PATCH] intel: Add support for GPU reset status query ioctl

[Damien Lespiau]

On Fri, Nov 15, 2013 at 10:41:57AM -0800, Ian Romanick wrote:
> From: Ian Romanick <ian.d.romanick@intel.com>
> 
> I would have just used the drmIoctl interface directly in Mesa, but the
> ioctl needs some data from the drm_intel_context that is not exposed
> outside libdrm.
> 
> This ioctl is in the drm-intel-next tree as b635991.
> 
> v2: Update based on Mika's kernel work.
> 
> v3: Fix compile failures from last-minute typos.  Sigh.
> 
> v4: Import the actual changes from the kernel i915_drm.h.  Only comments
> on some fields of drm_i915_reset_stats differed.  There are still some
> deltas between the kernel i915_drm.h and the one in libdrm, but those
> can be resolved in other patches.
> 
> Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
> Reviewed-by: Kenneth Graunke <kenneth@whitecape.org> [v3]
> Cc: Mika Kuoppala <mika.kuoppala@intel.com>
> Cc: Daniel Vetter <daniel.vetter@ffwll.ch>

Looks good to me.

Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>

> ---
>  include/drm/i915_drm.h   | 19 +++++++++++++++++++
>  intel/intel_bufmgr.h     |  5 +++++
>  intel/intel_bufmgr_gem.c | 34 ++++++++++++++++++++++++++++++++++
>  3 files changed, 58 insertions(+)
> 
> diff --git a/include/drm/i915_drm.h b/include/drm/i915_drm.h
> index aa983f3..c1914d6 100644
> --- a/include/drm/i915_drm.h
> +++ b/include/drm/i915_drm.h
> @@ -198,6 +198,7 @@ typedef struct _drm_i915_sarea {
>  #define DRM_I915_GEM_SET_CACHEING	0x2f
>  #define DRM_I915_GEM_GET_CACHEING	0x30
>  #define DRM_I915_REG_READ		0x31
> +#define DRM_I915_GET_RESET_STATS	0x32
>  
>  #define DRM_IOCTL_I915_INIT		DRM_IOW( DRM_COMMAND_BASE + DRM_I915_INIT, drm_i915_init_t)
>  #define DRM_IOCTL_I915_FLUSH		DRM_IO ( DRM_COMMAND_BASE + DRM_I915_FLUSH)
> @@ -247,6 +248,7 @@ typedef struct _drm_i915_sarea {
>  #define DRM_IOCTL_I915_GEM_CONTEXT_CREATE	DRM_IOWR (DRM_COMMAND_BASE + DRM_I915_GEM_CONTEXT_CREATE, struct drm_i915_gem_context_create)
>  #define DRM_IOCTL_I915_GEM_CONTEXT_DESTROY	DRM_IOW (DRM_COMMAND_BASE + DRM_I915_GEM_CONTEXT_DESTROY, struct drm_i915_gem_context_destroy)
>  #define DRM_IOCTL_I915_REG_READ			DRM_IOWR (DRM_COMMAND_BASE + DRM_I915_REG_READ, struct drm_i915_reg_read)
> +#define DRM_IOCTL_I915_GET_RESET_STATS		DRM_IOWR (DRM_COMMAND_BASE + DRM_I915_GET_RESET_STATS, struct drm_i915_reset_stats)
>  
>  /* Allow drivers to submit batchbuffers directly to hardware, relying
>   * on the security mechanisms provided by hardware.
> @@ -943,4 +945,21 @@ struct drm_i915_reg_read {
>  	__u64 offset;
>  	__u64 val; /* Return value */
>  };
> +
> +struct drm_i915_reset_stats {
> +	__u32 ctx_id;
> +	__u32 flags;
> +
> +	/* All resets since boot/module reload, for all contexts */
> +	__u32 reset_count;
> +
> +	/* Number of batches lost when active in GPU, for this context */
> +	__u32 batch_active;
> +
> +	/* Number of batches lost pending for execution, for this context */
> +	__u32 batch_pending;
> +
> +	__u32 pad;
> +};
> +
>  #endif				/* _I915_DRM_H_ */
> diff --git a/intel/intel_bufmgr.h b/intel/intel_bufmgr.h
> index 7b28a70..34a5750 100644
> --- a/intel/intel_bufmgr.h
> +++ b/intel/intel_bufmgr.h
> @@ -249,6 +249,11 @@ int drm_intel_reg_read(drm_intel_bufmgr *bufmgr,
>  		       uint32_t offset,
>  		       uint64_t *result);
>  
> +int drm_intel_get_reset_stats(drm_intel_context *ctx,
> +			      uint32_t *reset_count,
> +			      uint32_t *active,
> +			      uint32_t *pending);
> +
>  /** @{ Compatibility defines to keep old code building despite the symbol rename
>   * from dri_* to drm_intel_*
>   */
> diff --git a/intel/intel_bufmgr_gem.c b/intel/intel_bufmgr_gem.c
> index dbadc52..0b9252e 100644
> --- a/intel/intel_bufmgr_gem.c
> +++ b/intel/intel_bufmgr_gem.c
> @@ -3029,6 +3029,40 @@ drm_intel_gem_context_get_hw_context_id(const drm_intel_context *ctx)
>  }
>  
>  int
> +drm_intel_get_reset_stats(drm_intel_context *ctx,
> +			  uint32_t *reset_count,
> +			  uint32_t *active,
> +			  uint32_t *pending)
> +{
> +	drm_intel_bufmgr_gem *bufmgr_gem;
> +	struct drm_i915_reset_stats stats;
> +	int ret;
> +
> +	if (ctx == NULL)
> +		return -EINVAL;
> +
> +	VG_CLEAR(stats);
> +
> +	bufmgr_gem = (drm_intel_bufmgr_gem *)ctx->bufmgr;
> +	stats.ctx_id = ctx->ctx_id;
> +	ret = drmIoctl(bufmgr_gem->fd,
> +		       DRM_IOCTL_I915_GET_RESET_STATS,
> +		       &stats);
> +	if (ret == 0) {
> +		if (reset_count != NULL)
> +			*reset_count = stats.reset_count;
> +
> +		if (active != NULL)
> +			*active = stats.batch_active;
> +
> +		if (pending != NULL)
> +			*pending = stats.batch_pending;
> +	}
> +
> +	return ret;
> +}
> +
> +int
>  drm_intel_reg_read(drm_intel_bufmgr *bufmgr,
>  		   uint32_t offset,
>  		   uint64_t *result)
> -- 
> 1.8.1.4
> 
> _______________________________________________
> dri-devel mailing list
> dri-devel@lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/dri-devel