drm/i915/gem: Try to flush pending unbind events (aa5e4453) · Commits · 戴 / test

If we cannot handle a vma within the unbind loop, try to flush the pending events (i915_vma_parked, i915_vm_release) and try again. This avoids a round trip to userspace that is not guaranteed to make forward progress, as the events we wait upon require being idle. References: ("drm/i915/gem: Avoid parking the vma as we unbind") Signed-off-by:

Chris Wilson <chris@chris-wilson.co.uk> Cc: Matthew Auld <matthew.auld@intel.com> Reviewed-by:

Matthew Auld <matthew.auld@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20191204123556.3740002-1-chris@chris-wilson.co.uk

drivers/gpu/drm/i915/i915_gem.c

+9 −1

Original line number	Diff line number	Diff line
		@@ -124,7 +124,7 @@ int i915_gem_object_unbind(struct drm_i915_gem_object *obj,
		LIST_HEAD(still_in_list);
		intel_wakeref_t wakeref;
		struct i915_vma *vma;
		int ret = 0;
		int ret;

		if (!atomic_read(&obj->bind_count))
		return 0;
		@@ -137,6 +137,8 @@ int i915_gem_object_unbind(struct drm_i915_gem_object *obj,
		*/
		wakeref = intel_runtime_pm_get(rpm);

		try_again:
		ret = 0;
		spin_lock(&obj->vma.lock);
		while (!ret && (vma = list_first_entry_or_null(&obj->vma.list,
		struct i915_vma,
		@@ -154,6 +156,7 @@ int i915_gem_object_unbind(struct drm_i915_gem_object *obj,
		} else {
		if (i915_vma_is_closed(vma)) {
		spin_unlock(&obj->vma.lock);
		i915_vma_parked(vm->gt);
		goto err_vm;
		}
		}
		@@ -175,6 +178,11 @@ err_vm:
		list_splice(&still_in_list, &obj->vma.list);
		spin_unlock(&obj->vma.lock);

		if (ret == -EAGAIN && flags & I915_GEM_OBJECT_UNBIND_ACTIVE) {
		rcu_barrier(); /* flush the i915_vm_release() */
		goto try_again;
		}

		intel_runtime_pm_put(rpm, wakeref);

		return ret;

Admin message