| Message ID | 20210608092846.64198-2-thomas.hellstrom@linux.intel.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | Prereqs for TTM accelerated migration | expand |
On Tue, 8 Jun 2021 at 10:29, Thomas Hellström <thomas.hellstrom@linux.intel.com> wrote: > > Since the ww transaction endpoint easily end up far out-of-scope of > the objects on the ww object list, particularly for contending lock > objects, make sure we reference objects on the list so they don't > disappear under us. > > This comes with a performance penalty so it's been debated whether this > is really needed. But I think this is motivated by the fact that locking > is typically difficult to get right, and whatever we can do to make it > simpler for developers moving forward should be done, unless the > performance impact is far too high. > > Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com> Reviewed-by: Matthew Auld <matthew.auld@intel.com>
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object.h b/drivers/gpu/drm/i915/gem/i915_gem_object.h index 7c0eb425cb3b..1fafcc89ecee 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_object.h +++ b/drivers/gpu/drm/i915/gem/i915_gem_object.h @@ -169,13 +169,17 @@ static inline int __i915_gem_object_lock(struct drm_i915_gem_object *obj, else ret = dma_resv_lock(obj->base.resv, ww ? &ww->ctx : NULL); - if (!ret && ww) + if (!ret && ww) { + i915_gem_object_get(obj); list_add_tail(&obj->obj_link, &ww->obj_list); + } if (ret == -EALREADY) ret = 0; - if (ret == -EDEADLK) + if (ret == -EDEADLK) { + i915_gem_object_get(obj); ww->contended = obj; + } return ret; } diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 589388dec48a..3f060ab58c5d 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -1219,6 +1219,7 @@ static void i915_gem_ww_ctx_unlock_all(struct i915_gem_ww_ctx *ww) while ((obj = list_first_entry_or_null(&ww->obj_list, struct drm_i915_gem_object, obj_link))) { list_del(&obj->obj_link); i915_gem_object_unlock(obj); + i915_gem_object_put(obj); } } @@ -1226,6 +1227,7 @@ void i915_gem_ww_unlock_single(struct drm_i915_gem_object *obj) { list_del(&obj->obj_link); i915_gem_object_unlock(obj); + i915_gem_object_put(obj); } void i915_gem_ww_ctx_fini(struct i915_gem_ww_ctx *ww) @@ -1250,6 +1252,8 @@ int __must_check i915_gem_ww_ctx_backoff(struct i915_gem_ww_ctx *ww) if (!ret) list_add_tail(&ww->contended->obj_link, &ww->obj_list); + else + i915_gem_object_put(ww->contended); ww->contended = NULL;
Since the ww transaction endpoint easily end up far out-of-scope of the objects on the ww object list, particularly for contending lock objects, make sure we reference objects on the list so they don't disappear under us. This comes with a performance penalty so it's been debated whether this is really needed. But I think this is motivated by the fact that locking is typically difficult to get right, and whatever we can do to make it simpler for developers moving forward should be done, unless the performance impact is far too high. Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com> --- drivers/gpu/drm/i915/gem/i915_gem_object.h | 8 ++++++-- drivers/gpu/drm/i915/i915_gem.c | 4 ++++ 2 files changed, 10 insertions(+), 2 deletions(-)