[Intel-gfx] [PATCH 35/49] drm/i915: Remove request retirement before each batch
Chris Wilson
chris at chris-wilson.co.uk
Fri Mar 27 04:02:07 PDT 2015
This reimplements the denial-of-service protection against igt from
commit 227f782e4667fc622810bce8be8ccdeee45f89c2
Author: Chris Wilson <chris at chris-wilson.co.uk>
Date: Thu May 15 10:41:42 2014 +0100
drm/i915: Retire requests before creating a new one
and transfers the stall from before each batch into a new close handler.
The issue is that the stall is increasing latency between batches which
is detrimental in some cases (especially coupled with execlists) to
keeping the GPU well fed. Also we make the observation that retiring
requests can of itself free objects (and requests) and therefore makes
a good first step when shrinking. However, we do wish to do a retire
before forcing an allocation of a new batch pool object (prior to an
execbuffer), but we make the optimisation that we only need to do so if
the oldest available batch pool object is active.
Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
---
drivers/gpu/drm/i915/i915_drv.c | 1 +
drivers/gpu/drm/i915/i915_drv.h | 2 ++
drivers/gpu/drm/i915/i915_gem.c | 15 +++++++++++++++
drivers/gpu/drm/i915/i915_gem_batch_pool.c | 9 +++++++--
drivers/gpu/drm/i915/i915_gem_batch_pool.h | 6 +++++-
drivers/gpu/drm/i915/i915_gem_execbuffer.c | 2 --
drivers/gpu/drm/i915/intel_lrc.c | 2 +-
drivers/gpu/drm/i915/intel_ringbuffer.c | 2 +-
8 files changed, 32 insertions(+), 7 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c
index 82f8be4b6745..df85bfeabc40 100644
--- a/drivers/gpu/drm/i915/i915_drv.c
+++ b/drivers/gpu/drm/i915/i915_drv.c
@@ -1590,6 +1590,7 @@ static struct drm_driver driver = {
.debugfs_init = i915_debugfs_init,
.debugfs_cleanup = i915_debugfs_cleanup,
#endif
+ .gem_close_object = i915_gem_close_object,
.gem_free_object = i915_gem_free_object,
.gem_vm_ops = &i915_gem_vm_ops,
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index 4b51169c37ea..8fcf923aae3b 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -2638,6 +2638,8 @@ struct drm_i915_gem_object *i915_gem_alloc_object(struct drm_device *dev,
size_t size);
void i915_init_vm(struct drm_i915_private *dev_priv,
struct i915_address_space *vm);
+void i915_gem_close_object(struct drm_gem_object *obj,
+ struct drm_file *file);
void i915_gem_free_object(struct drm_gem_object *obj);
void i915_gem_vma_destroy(struct i915_vma *vma);
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 91b0c1db05ca..f87e7b90939c 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -2181,6 +2181,10 @@ i915_gem_object_get_pages_gtt(struct drm_i915_gem_object *obj)
for (i = 0; i < page_count; i++) {
page = shmem_read_mapping_page_gfp(mapping, i, gfp);
if (IS_ERR(page)) {
+ i915_gem_retire_requests(dev_priv->dev);
+ page = shmem_read_mapping_page_gfp(mapping, i, gfp);
+ }
+ if (IS_ERR(page)) {
i915_gem_shrink(dev_priv,
page_count,
I915_SHRINK_BOUND |
@@ -2873,6 +2877,17 @@ i915_gem_object_flush_active(struct drm_i915_gem_object *obj)
return 0;
}
+void i915_gem_close_object(struct drm_gem_object *gem,
+ struct drm_file *file)
+{
+ struct drm_i915_gem_object *obj = to_intel_bo(gem);
+
+ if (obj->active && mutex_trylock(&obj->base.dev->struct_mutex)) {
+ (void)i915_gem_object_flush_active(obj);
+ mutex_unlock(&obj->base.dev->struct_mutex);
+ }
+}
+
/**
* i915_gem_wait_ioctl - implements DRM_IOCTL_I915_GEM_WAIT
* @DRM_IOCTL_ARGS: standard ioctl arguments
diff --git a/drivers/gpu/drm/i915/i915_gem_batch_pool.c b/drivers/gpu/drm/i915/i915_gem_batch_pool.c
index 7bf2f3f2968e..03c67f4ad773 100644
--- a/drivers/gpu/drm/i915/i915_gem_batch_pool.c
+++ b/drivers/gpu/drm/i915/i915_gem_batch_pool.c
@@ -44,12 +44,13 @@
* @dev: the drm device
* @pool: the batch buffer pool
*/
-void i915_gem_batch_pool_init(struct drm_device *dev,
+void i915_gem_batch_pool_init(struct intel_engine_cs *ring,
struct i915_gem_batch_pool *pool)
{
int n;
- pool->dev = dev;
+ pool->dev = ring->dev;
+ pool->ring = ring;
for (n = 0; n < ARRAY_SIZE(pool->cache_list); n++)
INIT_LIST_HEAD(&pool->cache_list[n]);
@@ -113,6 +114,10 @@ i915_gem_batch_pool_get(struct i915_gem_batch_pool *pool,
n = ARRAY_SIZE(pool->cache_list) - 1;
list = &pool->cache_list[n];
+ tmp = list_first_entry_or_null(list, typeof(*tmp), batch_pool_link);
+ if (tmp && tmp->active)
+ i915_gem_retire_requests_ring(pool->ring);
+
list_for_each_entry_safe(tmp, next, list, batch_pool_link) {
/* The batches are strictly LRU ordered */
if (tmp->active)
diff --git a/drivers/gpu/drm/i915/i915_gem_batch_pool.h b/drivers/gpu/drm/i915/i915_gem_batch_pool.h
index 848e90703eed..467578c621bc 100644
--- a/drivers/gpu/drm/i915/i915_gem_batch_pool.h
+++ b/drivers/gpu/drm/i915/i915_gem_batch_pool.h
@@ -27,13 +27,17 @@
#include "i915_drv.h"
+struct drm_device;
+struct intel_engine_cs;
+
struct i915_gem_batch_pool {
struct drm_device *dev;
+ struct intel_engine_cs *ring;
struct list_head cache_list[4];
};
/* i915_gem_batch_pool.c */
-void i915_gem_batch_pool_init(struct drm_device *dev,
+void i915_gem_batch_pool_init(struct intel_engine_cs *ring,
struct i915_gem_batch_pool *pool);
void i915_gem_batch_pool_fini(struct i915_gem_batch_pool *pool);
struct drm_i915_gem_object*
diff --git a/drivers/gpu/drm/i915/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/i915_gem_execbuffer.c
index 9345436e4d95..403450f4e4ee 100644
--- a/drivers/gpu/drm/i915/i915_gem_execbuffer.c
+++ b/drivers/gpu/drm/i915/i915_gem_execbuffer.c
@@ -697,8 +697,6 @@ i915_gem_execbuffer_reserve(struct intel_engine_cs *ring,
bool has_fenced_gpu_access = INTEL_INFO(ring->dev)->gen < 4;
int retry;
- i915_gem_retire_requests_ring(ring);
-
vm = list_first_entry(vmas, struct i915_vma, exec_list)->vm;
INIT_LIST_HEAD(&ordered_vmas);
diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c
index 1c768c05e52e..80f00feb6bf4 100644
--- a/drivers/gpu/drm/i915/intel_lrc.c
+++ b/drivers/gpu/drm/i915/intel_lrc.c
@@ -1250,7 +1250,7 @@ static int logical_ring_init(struct drm_device *dev, struct intel_engine_cs *rin
ring->dev = dev;
INIT_LIST_HEAD(&ring->active_list);
INIT_LIST_HEAD(&ring->request_list);
- i915_gem_batch_pool_init(dev, &ring->batch_pool);
+ i915_gem_batch_pool_init(ring, &ring->batch_pool);
init_waitqueue_head(&ring->irq_queue);
INIT_LIST_HEAD(&ring->execlist_queue);
diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c
index f7097a80dea3..01eacaf4dac1 100644
--- a/drivers/gpu/drm/i915/intel_ringbuffer.c
+++ b/drivers/gpu/drm/i915/intel_ringbuffer.c
@@ -2010,7 +2010,7 @@ static int intel_init_ring_buffer(struct drm_device *dev,
INIT_LIST_HEAD(&ring->active_list);
INIT_LIST_HEAD(&ring->request_list);
INIT_LIST_HEAD(&ring->execlist_queue);
- i915_gem_batch_pool_init(dev, &ring->batch_pool);
+ i915_gem_batch_pool_init(ring, &ring->batch_pool);
ringbuf->size = 32 * PAGE_SIZE;
ringbuf->ring = ring;
memset(ring->semaphore.sync_seqno, 0, sizeof(ring->semaphore.sync_seqno));
--
2.1.4
More information about the Intel-gfx
mailing list