2019-11-04 17:39:03

by Daniel Vetter

Subject: [PATCH 3/3] drm/i915: use might_lock_nested in get_pages annotation

So strictly speaking the existing annotation is also ok, because we
have a chain of

obj->mm.lock#I915_MM_GET_PAGES -> fs_reclaim -> obj->mm.lock

(the shrinker cannot get at an object while we're in get_pages, hence
this is safe). But it's confusing, so try to take the right subclass
of the lock.

This does a bit reduce our lockdep based checking, but then it's also
less fragile, in case we ever change the nesting around.

Signed-off-by: Daniel Vetter <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Will Deacon <[email protected]>
Cc: [email protected]
---
drivers/gpu/drm/i915/gem/i915_gem_object.h | 36 +++++++++++-----------
1 file changed, 18 insertions(+), 18 deletions(-)

diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object.h b/drivers/gpu/drm/i915/gem/i915_gem_object.h
index edaf7126a84d..e5750d506cc9 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_object.h
+++ b/drivers/gpu/drm/i915/gem/i915_gem_object.h
@@ -271,10 +271,27 @@ void __i915_gem_object_set_pages(struct drm_i915_gem_object *obj,
int ____i915_gem_object_get_pages(struct drm_i915_gem_object *obj);
int __i915_gem_object_get_pages(struct drm_i915_gem_object *obj);

+enum i915_mm_subclass { /* lockdep subclass for obj->mm.lock/struct_mutex */
+ I915_MM_NORMAL = 0,
+ /*
+ * Only used by struct_mutex, when called "recursively" from
+ * direct-reclaim-esque. Safe because there is only every one
+ * struct_mutex in the entire system.
+ */
+ I915_MM_SHRINKER = 1,
+ /*
+ * Used for obj->mm.lock when allocating pages. Safe because the object
+ * isn't yet on any LRU, and therefore the shrinker can't deadlock on
+ * it. As soon as the object has pages, obj->mm.lock nests within
+ * fs_reclaim.
+ */
+ I915_MM_GET_PAGES = 1,
+};
+
static inline int __must_check
i915_gem_object_pin_pages(struct drm_i915_gem_object *obj)
{
- might_lock(&obj->mm.lock);
+ might_lock_nested(&obj->mm.lock, I915_MM_GET_PAGES);

if (atomic_inc_not_zero(&obj->mm.pages_pin_count))
return 0;
@@ -317,23 +334,6 @@ i915_gem_object_unpin_pages(struct drm_i915_gem_object *obj)
__i915_gem_object_unpin_pages(obj);
}

-enum i915_mm_subclass { /* lockdep subclass for obj->mm.lock/struct_mutex */
- I915_MM_NORMAL = 0,
- /*
- * Only used by struct_mutex, when called "recursively" from
- * direct-reclaim-esque. Safe because there is only every one
- * struct_mutex in the entire system.
- */
- I915_MM_SHRINKER = 1,
- /*
- * Used for obj->mm.lock when allocating pages. Safe because the object
- * isn't yet on any LRU, and therefore the shrinker can't deadlock on
- * it. As soon as the object has pages, obj->mm.lock nests within
- * fs_reclaim.
- */
- I915_MM_GET_PAGES = 1,
-};
-
int __i915_gem_object_put_pages(struct drm_i915_gem_object *obj);
void i915_gem_object_truncate(struct drm_i915_gem_object *obj);
void i915_gem_object_writeback(struct drm_i915_gem_object *obj);
--
2.24.0.rc2

2019-11-05 09:03:39

by Joonas Lahtinen

Subject: Re: [Intel-gfx] [PATCH 3/3] drm/i915: use might_lock_nested in get_pages annotation

Quoting Daniel Vetter (2019-11-04 19:37:20)
> So strictly speaking the existing annotation is also ok, because we
> have a chain of
>
> obj->mm.lock#I915_MM_GET_PAGES -> fs_reclaim -> obj->mm.lock
>
> (the shrinker cannot get at an object while we're in get_pages, hence
> this is safe). But it's confusing, so try to take the right subclass
> of the lock.
>
> This does a bit reduce our lockdep based checking, but then it's also
> less fragile, in case we ever change the nesting around.
>
> Signed-off-by: Daniel Vetter <[email protected]>
> Cc: Peter Zijlstra <[email protected]>
> Cc: Ingo Molnar <[email protected]>
> Cc: Will Deacon <[email protected]>
> Cc: [email protected]

Reviewed-by: Joonas Lahtinen <[email protected]>

Regards, Joonas