[RFC,2/8] drm/i915: Introduce mode for capture of multi ctx OA reports synchronized with RCS

Commit Message

[e-mail] Aug. 5, 2015, 5:52 a.m. UTC

From: Sourab Gupta <[e-mail]>

This patch introduces a mode of capturing OA counter reports belonging to
multiple contexts, which can be mapped back to individual contexts. The OA
reports captured in this way are synchronized with Render command stream.

There may be usecases wherein we need more than periodic OA capture mode
which is supported by perf_event currently. We may need to insert RCS
synchronized commands to capture the OA counter snapshots.
This mode is primarily used for two usecases:
    - Ability to capture system wide metrics, alongwith the ability to map
      the reports back to individual contexts.
    - Ability to inject tags for work, into the reports. This provides
      visibility into the multiple stages of work within single context.

The OA reports generated in this way will be forwarded to userspace after
appending a footer, which will have this metadata information. This will
enable the usecases mentioned above.

This patch introduces an additional field in the oa attr structure for
supporting this capture mode. The data thus captured needs to be stored in
a separate buffer, which will be different from the buffer used otherwise
for periodic OA capture mode. Again this buffer address will not need to be
mapped to OA unit register addresses such as OASTATUS1, OASTATUS2 and
OABUFFER.

The subsequent patches introduce the  mechanism for forwarding reports to
userspace, handling the command synchronization and mechanism for
inserting corresponding commands into the ringbuffer.

v2: Following changes:
    - Save the gtt offset while pinning the buffer, to be retrieved when
      required.
    - Use spin_lock instead of spin_lock_irqsave

Signed-off-by: Sourab Gupta <[e-mail]>
---
 drivers/gpu/drm/i915/i915_drv.h     |  10 +++
 drivers/gpu/drm/i915/i915_oa_perf.c | 170 +++++++++++++++++++++++++++---------
 include/uapi/drm/i915_drm.h         |   3 +-
 3 files changed, 141 insertions(+), 42 deletions(-)

Patch

diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index 3436f3b..050bdda 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -1931,6 +1931,7 @@  struct drm_i915_private {
 		bool event_active;
 
 		bool periodic;
+		bool multiple_ctx_mode;
 		u32 period_exponent;
 
 		u32 metrics_set;
@@ -1945,6 +1946,15 @@  struct drm_i915_private {
 			int format_size;
 			spinlock_t flush_lock;
 		} oa_buffer;
+
+		/* Fields for multiple context capture mode */
+		struct {
+			struct drm_i915_gem_object *obj;
+			u32 gtt_offset;
+			u8 *addr;
+			int format;
+			int format_size;
+		} oa_rcs_buffer;
 	} oa_pmu;
 #endif
 
diff --git a/drivers/gpu/drm/i915/i915_oa_perf.c b/drivers/gpu/drm/i915/i915_oa_perf.c
index e7e0b2b..fd0c3a0 100644
--- a/drivers/gpu/drm/i915/i915_oa_perf.c
+++ b/drivers/gpu/drm/i915/i915_oa_perf.c
@@ -166,19 +166,36 @@  static void flush_oa_snapshots(struct drm_i915_private *dev_priv,
 }
 
 static void
-oa_buffer_destroy(struct drm_i915_private *i915)
+oa_rcs_buffer_destroy(struct drm_i915_private *i915)
 {
 	mutex_lock(&i915->dev->struct_mutex);
+	vunmap(i915->oa_pmu.oa_rcs_buffer.addr);
+	i915_gem_object_ggtt_unpin(i915->oa_pmu.oa_rcs_buffer.obj);
+	drm_gem_object_unreference(&i915->oa_pmu.oa_rcs_buffer.obj->base);
+	mutex_unlock(&i915->dev->struct_mutex);
+
+	spin_lock(&i915->oa_pmu.lock);
+	i915->oa_pmu.oa_rcs_buffer.obj = NULL;
+	i915->oa_pmu.oa_rcs_buffer.gtt_offset = 0;
+	i915->oa_pmu.oa_rcs_buffer.addr = NULL;
+	spin_unlock(&i915->oa_pmu.lock);
+}
 
+static void
+oa_buffer_destroy(struct drm_i915_private *i915)
+{
+	mutex_lock(&i915->dev->struct_mutex);
 	vunmap(i915->oa_pmu.oa_buffer.addr);
 	i915_gem_object_ggtt_unpin(i915->oa_pmu.oa_buffer.obj);
 	drm_gem_object_unreference(&i915->oa_pmu.oa_buffer.obj->base);
+	mutex_unlock(&i915->dev->struct_mutex);
 
+	spin_lock(&i915->oa_pmu.lock);
 	i915->oa_pmu.oa_buffer.obj = NULL;
 	i915->oa_pmu.oa_buffer.gtt_offset = 0;
 	i915->oa_pmu.oa_buffer.addr = NULL;
+	spin_unlock(&i915->oa_pmu.lock);
 
-	mutex_unlock(&i915->dev->struct_mutex);
 }
 
 static void i915_oa_event_destroy(struct perf_event *event)
@@ -207,6 +224,9 @@  static void i915_oa_event_destroy(struct perf_event *event)
 	I915_WRITE(GDT_CHICKEN_BITS, (I915_READ(GDT_CHICKEN_BITS) &
 				      ~GT_NOA_ENABLE));
 
+	if (dev_priv->oa_pmu.multiple_ctx_mode)
+		oa_rcs_buffer_destroy(dev_priv);
+
 	oa_buffer_destroy(dev_priv);
 
 	BUG_ON(dev_priv->oa_pmu.exclusive_event != event);
@@ -216,6 +236,59 @@  static void i915_oa_event_destroy(struct perf_event *event)
 	intel_runtime_pm_put(dev_priv);
 }
 
+static int alloc_obj(struct drm_i915_private *dev_priv,
+				struct drm_i915_gem_object **obj)
+{
+	struct drm_i915_gem_object *bo;
+	int ret;
+
+	/* NB: We over allocate the OA buffer due to the way raw sample data
+	 * gets copied from the gpu mapped circular buffer into the perf
+	 * circular buffer so that only one copy is required.
+	 *
+	 * For each perf sample (raw->size + 4) needs to be 8 byte aligned,
+	 * where the 4 corresponds to the 32bit raw->size member that's
+	 * added to the sample header that userspace sees.
+	 *
+	 * Due to the + 4 for the size member: when we copy a report to the
+	 * userspace facing perf buffer we always copy an additional 4 bytes
+	 * from the subsequent report to make up for the miss alignment, but
+	 * when a report is at the end of the gpu mapped buffer we need to
+	 * read 4 bytes past the end of the buffer.
+	 */
+	intel_runtime_pm_get(dev_priv);
+
+	ret = i915_mutex_lock_interruptible(dev_priv->dev);
+	if (ret)
+		goto out;
+
+	bo = i915_gem_alloc_object(dev_priv->dev, OA_BUFFER_SIZE + PAGE_SIZE);
+	if (bo == NULL) {
+		DRM_ERROR("Failed to allocate OA buffer\n");
+		ret = -ENOMEM;
+		goto unlock;
+	}
+	ret = i915_gem_object_set_cache_level(bo, I915_CACHE_LLC);
+	if (ret)
+		goto err_unref;
+
+	/* PreHSW required 512K alignment, HSW requires 16M */
+	ret = i915_gem_obj_ggtt_pin(bo, SZ_16M, 0);
+	if (ret)
+		goto err_unref;
+
+	*obj = bo;
+	goto unlock;
+
+err_unref:
+	drm_gem_object_unreference(&bo->base);
+unlock:
+	mutex_unlock(&dev_priv->dev->struct_mutex);
+out:
+	intel_runtime_pm_put(dev_priv);
+	return ret;
+}
+
 static void *vmap_oa_buffer(struct drm_i915_gem_object *obj)
 {
 	int i;
@@ -257,42 +330,13 @@  static int init_oa_buffer(struct perf_event *event)
 	BUG_ON(!IS_HASWELL(dev_priv->dev));
 	BUG_ON(dev_priv->oa_pmu.oa_buffer.obj);
 
-	ret = i915_mutex_lock_interruptible(dev_priv->dev);
-	if (ret)
-		return ret;
-
 	spin_lock_init(&dev_priv->oa_pmu.oa_buffer.flush_lock);
 
-	/* NB: We over allocate the OA buffer due to the way raw sample data
-	 * gets copied from the gpu mapped circular buffer into the perf
-	 * circular buffer so that only one copy is required.
-	 *
-	 * For each perf sample (raw->size + 4) needs to be 8 byte aligned,
-	 * where the 4 corresponds to the 32bit raw->size member that's
-	 * added to the sample header that userspace sees.
-	 *
-	 * Due to the + 4 for the size member: when we copy a report to the
-	 * userspace facing perf buffer we always copy an additional 4 bytes
-	 * from the subsequent report to make up for the miss alignment, but
-	 * when a report is at the end of the gpu mapped buffer we need to
-	 * read 4 bytes past the end of the buffer.
-	 */
-	bo = i915_gem_alloc_object(dev_priv->dev, OA_BUFFER_SIZE + PAGE_SIZE);
-	if (bo == NULL) {
-		DRM_ERROR("Failed to allocate OA buffer\n");
-		ret = -ENOMEM;
-		goto unlock;
-	}
-	dev_priv->oa_pmu.oa_buffer.obj = bo;
-
-	ret = i915_gem_object_set_cache_level(bo, I915_CACHE_LLC);
+	ret = alloc_obj(dev_priv, &bo);
 	if (ret)
-		goto err_unref;
+		return ret;
 
-	/* PreHSW required 512K alignment, HSW requires 16M */
-	ret = i915_gem_obj_ggtt_pin(bo, SZ_16M, 0);
-	if (ret)
-		goto err_unref;
+	dev_priv->oa_pmu.oa_buffer.obj = bo;
 
 	dev_priv->oa_pmu.oa_buffer.gtt_offset = i915_gem_obj_ggtt_offset(bo);
 	dev_priv->oa_pmu.oa_buffer.addr = vmap_oa_buffer(bo);
@@ -309,14 +353,31 @@  static int init_oa_buffer(struct perf_event *event)
 			 dev_priv->oa_pmu.oa_buffer.gtt_offset,
 			 dev_priv->oa_pmu.oa_buffer.addr);
 
-	goto unlock;
+	return 0;
+}
 
-err_unref:
-	drm_gem_object_unreference(&bo->base);
+static int init_oa_rcs_buffer(struct perf_event *event)
+{
+	struct drm_i915_private *dev_priv =
+		container_of(event->pmu, typeof(*dev_priv), oa_pmu.pmu);
+	struct drm_i915_gem_object *bo;
+	int ret;
 
-unlock:
-	mutex_unlock(&dev_priv->dev->struct_mutex);
-	return ret;
+	BUG_ON(dev_priv->oa_pmu.oa_rcs_buffer.obj);
+
+	ret = alloc_obj(dev_priv, &bo);
+	if (ret)
+		return ret;
+
+	dev_priv->oa_pmu.oa_rcs_buffer.obj = bo;
+	dev_priv->oa_pmu.oa_rcs_buffer.gtt_offset =
+				i915_gem_obj_ggtt_offset(bo);
+	dev_priv->oa_pmu.oa_rcs_buffer.addr = vmap_oa_buffer(bo);
+
+	DRM_DEBUG_DRIVER("OA RCS Buffer initialized, vaddr = %p",
+			 dev_priv->oa_pmu.oa_rcs_buffer.addr);
+
+	return 0;
 }
 
 static enum hrtimer_restart hrtimer_sample(struct hrtimer *hrtimer)
@@ -439,11 +500,28 @@  static int i915_oa_event_init(struct perf_event *event)
 	/* To avoid the complexity of having to accurately filter
 	 * counter snapshots and marshal to the appropriate client
 	 * we currently only allow exclusive access */
-	if (dev_priv->oa_pmu.oa_buffer.obj)
+	spin_lock(&dev_priv->oa_pmu.lock);
+	if (dev_priv->oa_pmu.oa_buffer.obj) {
+		spin_unlock(&dev_priv->oa_pmu.lock);
 		return -EBUSY;
+	}
+	spin_unlock(&dev_priv->oa_pmu.lock);
+
+	/*
+	 * In case of multiple context mode, we need to check for
+	 * CAP_SYS_ADMIN capability as we need to profile all the running
+	 * contexts
+	 */
+	if (oa_attr.multiple_context_mode) {
+		if (!capable(CAP_SYS_ADMIN))
+			return -EACCES;
+		dev_priv->oa_pmu.multiple_ctx_mode = true;
+	}
 
 	report_format = oa_attr.format;
 	dev_priv->oa_pmu.oa_buffer.format = report_format;
+	if (oa_attr.multiple_context_mode)
+		dev_priv->oa_pmu.oa_rcs_buffer.format = report_format;
 	dev_priv->oa_pmu.metrics_set = oa_attr.metrics_set;
 
 	if (IS_HASWELL(dev_priv->dev)) {
@@ -457,6 +535,9 @@  static int i915_oa_event_init(struct perf_event *event)
 			return -EINVAL;
 
 		dev_priv->oa_pmu.oa_buffer.format_size = snapshot_size;
+		if (oa_attr.multiple_context_mode)
+			dev_priv->oa_pmu.oa_rcs_buffer.format_size =
+					snapshot_size;
 
 		if (oa_attr.metrics_set > I915_OA_METRICS_SET_MAX)
 			return -EINVAL;
@@ -465,6 +546,7 @@  static int i915_oa_event_init(struct perf_event *event)
 		return -ENODEV;
 	}
 
+
 	/* Since we are limited to an exponential scale for
 	 * programming the OA sampling period we don't allow userspace
 	 * to pass a precise attr.sample_period. */
@@ -528,6 +610,12 @@  static int i915_oa_event_init(struct perf_event *event)
 	if (ret)
 		return ret;
 
+	if (oa_attr.multiple_context_mode) {
+		ret = init_oa_rcs_buffer(event);
+		if (ret)
+			return ret;
+	}
+
 	BUG_ON(dev_priv->oa_pmu.exclusive_event);
 	dev_priv->oa_pmu.exclusive_event = event;
 
diff --git a/include/uapi/drm/i915_drm.h b/include/uapi/drm/i915_drm.h
index 9d866e8..dcf7c87 100644
--- a/include/uapi/drm/i915_drm.h
+++ b/include/uapi/drm/i915_drm.h
@@ -92,7 +92,8 @@  typedef struct _drm_i915_oa_attr {
 	__u32 ctx_id;
 
 	__u64 single_context : 1,
-		__reserved_1 : 63;
+		multiple_context_mode:1,
+		__reserved_1:62;
 } drm_i915_oa_attr_t;
 
 /* Header for PERF_RECORD_DEVICE type events */