@@ -437,6 +437,7 @@ struct i915_hw_ppgtt {
/* This must match up with the value previously used for execbuf2.rsvd1. */
#define DEFAULT_CONTEXT_ID 0
struct i915_hw_context {
+ struct kref ref;
int id;
bool is_initialized;
struct drm_i915_file_private *file_priv;
@@ -1239,6 +1240,9 @@ struct drm_i915_gem_request {
/** Postion in the ringbuffer of the end of the request */
u32 tail;
+ /** Context related to this request */
+ struct i915_hw_context *ctx;
+
/** Time at which this request was emitted, in jiffies. */
unsigned long emitted_jiffies;
@@ -1628,9 +1632,10 @@ int __must_check i915_gpu_idle(struct drm_device *dev);
int __must_check i915_gem_idle(struct drm_device *dev);
int i915_do_add_request(struct intel_ring_buffer *ring,
u32 *seqno,
- struct drm_file *file);
+ struct drm_file *file,
+ struct i915_hw_context *ctx);
#define i915_add_request(ring, seqno) \
- i915_do_add_request(ring, seqno, NULL)
+ i915_do_add_request(ring, seqno, NULL, NULL)
int __must_check i915_wait_seqno(struct intel_ring_buffer *ring,
uint32_t seqno);
int i915_gem_fault(struct vm_area_struct *vma, struct vm_fault *vmf);
@@ -1674,6 +1679,7 @@ void i915_gem_context_close(struct drm_device *dev, struct drm_file *file);
struct i915_hw_context * __must_check
i915_switch_context(struct intel_ring_buffer *ring,
struct drm_file *file, int to_id);
+void i915_gem_context_free(struct kref *ctx_ref);
int i915_gem_context_create_ioctl(struct drm_device *dev, void *data,
struct drm_file *file);
int i915_gem_context_destroy_ioctl(struct drm_device *dev, void *data,
@@ -1997,7 +1997,8 @@ i915_gem_get_seqno(struct drm_device *dev, u32 *seqno)
int
i915_do_add_request(struct intel_ring_buffer *ring,
u32 *out_seqno,
- struct drm_file *file)
+ struct drm_file *file,
+ struct i915_hw_context *ctx)
{
drm_i915_private_t *dev_priv = ring->dev->dev_private;
struct drm_i915_gem_request *request;
@@ -2037,6 +2038,11 @@ i915_do_add_request(struct intel_ring_buffer *ring,
request->seqno = intel_ring_get_seqno(ring);
request->ring = ring;
request->tail = request_ring_position;
+ request->ctx = ctx;
+
+ if (request->ctx)
+ kref_get(&request->ctx->ref);
+
request->emitted_jiffies = jiffies;
was_empty = list_empty(&ring->request_list);
list_add_tail(&request->list, &ring->request_list);
@@ -2101,6 +2107,10 @@ static void i915_gem_reset_ring_lists(struct drm_i915_private *dev_priv,
list_del(&request->list);
i915_gem_request_remove_from_client(request);
+
+ if (request->ctx)
+ kref_put(&request->ctx->ref, i915_gem_context_free);
+
kfree(request);
}
@@ -2195,6 +2205,10 @@ i915_gem_retire_requests_ring(struct intel_ring_buffer *ring)
list_del(&request->list);
i915_gem_request_remove_from_client(request);
+
+ if (request->ctx)
+ kref_put(&request->ctx->ref, i915_gem_context_free);
+
kfree(request);
}
@@ -124,12 +124,24 @@ static int get_context_size(struct drm_device *dev)
return ret;
}
-static void do_destroy(struct i915_hw_context *ctx)
+void i915_gem_context_free(struct kref *ctx_ref)
+{
+ struct i915_hw_context *ctx = container_of(ctx_ref,
+ typeof(*ctx), ref);
+ kfree(ctx);
+}
+
+static void do_release(struct i915_hw_context *ctx)
{
if (ctx->file_priv)
idr_remove(&ctx->file_priv->context_idr, ctx->id);
drm_gem_object_unreference(&ctx->obj->base);
+}
+
+static void do_destroy(struct i915_hw_context *ctx)
+{
+ do_release(ctx);
kfree(ctx);
}
@@ -145,6 +157,7 @@ create_hw_context(struct drm_device *dev,
if (ctx == NULL)
return ERR_PTR(-ENOMEM);
+ kref_init(&ctx->ref);
ctx->obj = i915_gem_alloc_object(dev, dev_priv->hw_context_size);
if (ctx->obj == NULL) {
kfree(ctx);
@@ -286,8 +299,8 @@ static int context_idr_cleanup(int id, void *p, void *data)
BUG_ON(id == DEFAULT_CONTEXT_ID);
- do_destroy(ctx);
-
+ do_release(ctx);
+ kref_put(&ctx->ref, i915_gem_context_free);
return 0;
}
@@ -522,7 +535,8 @@ int i915_gem_context_destroy_ioctl(struct drm_device *dev, void *data,
return -ENOENT;
}
- do_destroy(ctx);
+ do_release(ctx);
+ kref_put(&ctx->ref, i915_gem_context_free);
mutex_unlock(&dev->struct_mutex);
@@ -793,13 +793,14 @@ i915_gem_execbuffer_move_to_active(struct list_head *objects,
static void
i915_gem_execbuffer_retire_commands(struct drm_device *dev,
struct drm_file *file,
- struct intel_ring_buffer *ring)
+ struct intel_ring_buffer *ring,
+ struct i915_hw_context *ctx)
{
/* Unconditionally force add_request to emit a full flush. */
ring->gpu_caches_dirty = true;
/* Add a breadcrumb for the completion of the batch buffer */
- (void)i915_do_add_request(ring, NULL, file);
+ (void)i915_do_add_request(ring, NULL, file, ctx);
}
static int
@@ -1074,7 +1075,7 @@ i915_gem_do_execbuffer(struct drm_device *dev, void *data,
trace_i915_gem_ring_dispatch(ring, intel_ring_get_seqno(ring), flags);
i915_gem_execbuffer_move_to_active(&eb->objects, ring);
- i915_gem_execbuffer_retire_commands(dev, file, ring);
+ i915_gem_execbuffer_retire_commands(dev, file, ring, ctx);
err:
eb_destroy(eb);
In preparation to do analysis of which context was guilty of gpu hung, store kreffed context pointer into request struct. This allows us to inspect contexts when gpu is reset even if those contexts would already be released by userspace. v2: track i915_hw_context pointers instead of using ctx_ids (from Chris Wilson) Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com> --- drivers/gpu/drm/i915/i915_drv.h | 10 ++++++++-- drivers/gpu/drm/i915/i915_gem.c | 16 +++++++++++++++- drivers/gpu/drm/i915/i915_gem_context.c | 22 ++++++++++++++++++---- drivers/gpu/drm/i915/i915_gem_execbuffer.c | 7 ++++--- 4 files changed, 45 insertions(+), 10 deletions(-)