[Intel-gfx] [CI 1/2] drm/i915: Enable userspace to opt-out of implicit fencing

Chris Wilson chris at chris-wilson.co.uk
Fri Jan 27 09:40:07 UTC 2017

Userspace is faced with a dilemma. The kernel requires implicit fencing
to manage resource usage (we always must wait for the GPU to finish
before releasing its PTE) and for third parties. However, userspace may
wish to avoid this serialisation if it is either using explicit fencing
between parties and wants more fine-grained access to buffers (e.g. it
may partition the buffer between uses and track fences on ranges rather
than the implicit fences tracking the whole object). It follows that
userspace needs a mechanism to avoid the kernel's serialisation on its
implicit fences before execbuf execution.

The next question is whether this is an object, execbuf or context flag.
Hybrid users (such as using explicit EGL_ANDROID_native_sync fencing on
shared winsys buffers, but implicit fencing on internal surfaces)
require a per-object level flag. Given that this flag need to be only
set once for the lifetime of the object, this reduces the convenience of
having an execbuf or context level flag (and avoids having multiple
pieces of uABI controlling the same feature).

Incorrect use of this flag will result in rendering corruption and GPU
hangs - but will not result in use-after-free or similar resource
tracking issues.

Serious caveat: write ordering is not strictly correct after setting
this flag on a render target on multiple engines. This affects all
subsequent GEM operations (execbuf, set-domain, pread) and shared
dma-buf operations. A fix is possible - but costly (both in terms of
further ABI changes and runtime overhead).

Testcase: igt/gem_exec_async
Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen at linux.intel.com>
Acked-by: Chad Versace <chadversary at chromium.org>
 drivers/gpu/drm/i915/i915_drv.c            |  1 +
 drivers/gpu/drm/i915/i915_gem_execbuffer.c |  3 +++
 include/uapi/drm/i915_drm.h                | 30 +++++++++++++++++++++++++++++-
 3 files changed, 33 insertions(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c
index 0b87ae0ef942..8efa0d78ad98 100644
--- a/drivers/gpu/drm/i915/i915_drv.c
+++ b/drivers/gpu/drm/i915/i915_drv.c
@@ -349,6 +349,7 @@ static int i915_getparam(struct drm_device *dev, void *data,
 		/* For the time being all of these are always true;
 		 * if some supported hardware does not have one of these
 		 * features this value needs to be provided from
diff --git a/drivers/gpu/drm/i915/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/i915_gem_execbuffer.c
index c66e90571031..6fd60682bf93 100644
--- a/drivers/gpu/drm/i915/i915_gem_execbuffer.c
+++ b/drivers/gpu/drm/i915/i915_gem_execbuffer.c
@@ -1111,6 +1111,9 @@ i915_gem_execbuffer_move_to_gpu(struct drm_i915_gem_request *req,
 	list_for_each_entry(vma, vmas, exec_list) {
 		struct drm_i915_gem_object *obj = vma->obj;
+		if (vma->exec_entry->flags & EXEC_OBJECT_ASYNC)
+			continue;
 		ret = i915_gem_request_await_object
 			(req, obj, obj->base.pending_write_domain);
 		if (ret)
diff --git a/include/uapi/drm/i915_drm.h b/include/uapi/drm/i915_drm.h
index 57093b455db6..5842a652f596 100644
--- a/include/uapi/drm/i915_drm.h
+++ b/include/uapi/drm/i915_drm.h
@@ -397,6 +397,13 @@ typedef struct drm_i915_irq_wait {
 #define I915_PARAM_HAS_SCHEDULER	 41
 #define I915_PARAM_HUC_STATUS		 42
+/* Query whether DRM_I915_GEM_EXECBUFFER2 supports the ability to opt-out of
+ * synchronisation with implicit fencing on individual objects.
+ */
+#define I915_PARAM_HAS_EXEC_ASYNC	 43
 typedef struct drm_i915_getparam {
 	__s32 param;
@@ -737,8 +744,29 @@ struct drm_i915_gem_exec_object2 {
 #define EXEC_OBJECT_PINNED		 (1<<4)
 #define EXEC_OBJECT_PAD_TO_SIZE		 (1<<5)
+/* The kernel implicitly tracks GPU activity on all GEM objects, and
+ * synchronises operations with outstanding rendering. This includes
+ * rendering on other devices if exported via dma-buf. However, sometimes
+ * this tracking is too coarse and the user knows better. For example,
+ * if the object is split into non-overlapping ranges shared between different
+ * clients or engines (i.e. suballocating objects), the implicit tracking
+ * by kernel assumes that each operation affects the whole object rather
+ * than an individual range, causing needless synchronisation between clients.
+ * The kernel will also forgo any CPU cache flushes prior to rendering from
+ * the object as the client is expected to be also handling such domain
+ * tracking.
+ *
+ * The kernel maintains the implicit tracking in order to manage resources
+ * used by the GPU - this flag only disables the synchronisation prior to
+ * rendering with this object in this execbuf.
+ *
+ * Opting out of implicit synhronisation requires the user to do its own
+ * explicit tracking to avoid rendering corruption. See, for example,
+ * I915_PARAM_HAS_EXEC_FENCE to order execbufs and execute them asynchronously.
+ */
+#define EXEC_OBJECT_ASYNC		(1<<6)
 /* All remaining bits are MBZ and RESERVED FOR FUTURE USE */
 	__u64 flags;
 	union {

More information about the Intel-gfx mailing list