From: Johannes Berg johannes.berg@intel.com
This reverts commit fa7d81bb3c269a2ee38b6e4d569d9eb8be1a78ad.
As Peter explained: [...] lockless_dereference() is _stronger_ than READ_ONCE(), not weaker.
[...]
Also, clue is in the name: 'dereference', you don't actually dereference the pointer here, only load it.
My next patch breaks compile on this, because it assumes you want to deference and thus also need the struct type visible (which it isn't here), so revert it.
Cc: Peter Zijlstra peterz@infradead.org Signed-off-by: Johannes Berg johannes.berg@intel.com --- drivers/gpu/drm/drm_fb_helper.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/drm_fb_helper.c b/drivers/gpu/drm/drm_fb_helper.c index ce54e985d91b..0a06f9120b5a 100644 --- a/drivers/gpu/drm/drm_fb_helper.c +++ b/drivers/gpu/drm/drm_fb_helper.c @@ -464,7 +464,7 @@ static bool drm_fb_helper_is_bound(struct drm_fb_helper *fb_helper)
/* Sometimes user space wants everything disabled, so don't steal the * display if there's a master. */ - if (lockless_dereference(dev->master)) + if (READ_ONCE(dev->master)) return false;
drm_for_each_crtc(crtc, dev) {
From: Johannes Berg johannes.berg@intel.com
After Peter's commit (see below) we get a lot of sparse warnings (one for every rcu_dereference, and more) since the expression here is assigning to the wrong address space.
Instead of validating that 'p' is a pointer this way, instead make it fail compilation when it's not by using sizeof(*(p)). This will not cause any sparse warnings (tested, likely since the address space is irrelevant for sizeof), and will fail compilation when 'p' isn't a pointer type.
Fixes: 331b6d8c7afc ("locking/barriers: Validate lockless_dereference() is used on a pointer type") Signed-off-by: Johannes Berg johannes.berg@intel.com --- include/linux/compiler.h | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)
diff --git a/include/linux/compiler.h b/include/linux/compiler.h index 1bb954842725..436aa4e42221 100644 --- a/include/linux/compiler.h +++ b/include/linux/compiler.h @@ -527,13 +527,13 @@ static __always_inline void __write_once_size(volatile void *p, void *res, int s * object's lifetime is managed by something other than RCU. That * "something other" might be reference counting or simple immortality. * - * The seemingly unused void * variable is to validate @p is indeed a pointer - * type. All pointer types silently cast to void *. + * The seemingly unused size_t variable is to validate @p is indeed a pointer + * type by making sure it can be dereferenced. */ #define lockless_dereference(p) \ ({ \ typeof(p) _________p1 = READ_ONCE(p); \ - __maybe_unused const void * const _________p2 = _________p1; \ + size_t __maybe_unused __size_of_ptr = sizeof(*(p)); \ smp_read_barrier_depends(); /* Dependency order vs. p above. */ \ (_________p1); \ })
On Thu, Aug 11, 2016 at 11:50:21AM +0200, Johannes Berg wrote:
From: Johannes Berg johannes.berg@intel.com
This reverts commit fa7d81bb3c269a2ee38b6e4d569d9eb8be1a78ad.
As Peter explained: [...] lockless_dereference() is _stronger_ than READ_ONCE(), not weaker.
[...]
Also, clue is in the name: 'dereference', you don't actually dereference the pointer here, only load it.
My next patch breaks compile on this, because it assumes you want to deference and thus also need the struct type visible (which it isn't here), so revert it.
Cc: Peter Zijlstra peterz@infradead.org Signed-off-by: Johannes Berg johannes.berg@intel.com
Reviewed-by: Daniel Vetter daniel.vetter@ffwll.ch
And ack-by: me for merging through whatever tree this makes sense for. -Daniel
drivers/gpu/drm/drm_fb_helper.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/drm_fb_helper.c b/drivers/gpu/drm/drm_fb_helper.c index ce54e985d91b..0a06f9120b5a 100644 --- a/drivers/gpu/drm/drm_fb_helper.c +++ b/drivers/gpu/drm/drm_fb_helper.c @@ -464,7 +464,7 @@ static bool drm_fb_helper_is_bound(struct drm_fb_helper *fb_helper)
/* Sometimes user space wants everything disabled, so don't steal the * display if there's a master. */
- if (lockless_dereference(dev->master))
if (READ_ONCE(dev->master)) return false;
drm_for_each_crtc(crtc, dev) {
-- 2.8.1
dri-devel mailing list dri-devel@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/dri-devel
On Thu, Aug 11, 2016 at 12:38:59PM +0200, Daniel Vetter wrote:
On Thu, Aug 11, 2016 at 11:50:21AM +0200, Johannes Berg wrote:
From: Johannes Berg johannes.berg@intel.com
This reverts commit fa7d81bb3c269a2ee38b6e4d569d9eb8be1a78ad.
As Peter explained: [...] lockless_dereference() is _stronger_ than READ_ONCE(), not weaker.
[...]
Also, clue is in the name: 'dereference', you don't actually dereference the pointer here, only load it.
My next patch breaks compile on this, because it assumes you want to deference and thus also need the struct type visible (which it isn't here), so revert it.
Cc: Peter Zijlstra peterz@infradead.org Signed-off-by: Johannes Berg johannes.berg@intel.com
Reviewed-by: Daniel Vetter daniel.vetter@ffwll.ch
And ack-by: me for merging through whatever tree this makes sense for. -Daniel
Initial testing says that the change below must precede the change to the definition of lockless_dereference(), so the two should go together.
If my upcoming testing of the two changes together pans out, I will give you a Tested-by -- I am guessing that you don't want to wait until the next merge window for these changes.
Thanx, Paul
drivers/gpu/drm/drm_fb_helper.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/drm_fb_helper.c b/drivers/gpu/drm/drm_fb_helper.c index ce54e985d91b..0a06f9120b5a 100644 --- a/drivers/gpu/drm/drm_fb_helper.c +++ b/drivers/gpu/drm/drm_fb_helper.c @@ -464,7 +464,7 @@ static bool drm_fb_helper_is_bound(struct drm_fb_helper *fb_helper)
/* Sometimes user space wants everything disabled, so don't steal the * display if there's a master. */
- if (lockless_dereference(dev->master))
if (READ_ONCE(dev->master)) return false;
drm_for_each_crtc(crtc, dev) {
-- 2.8.1
dri-devel mailing list dri-devel@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/dri-devel
-- Daniel Vetter Software Engineer, Intel Corporation http://blog.ffwll.ch
Initial testing says that the change below must precede the change to the definition of lockless_dereference(), so the two should go together.
Indeed.
If my upcoming testing of the two changes together pans out, I will give you a Tested-by -- I am guessing that you don't want to wait until the next merge window for these changes.
I don't mind hugely since I have a fix now, but this one actually causes >1K warnings (including some "too many warnings!") for my build (net/wireless and net/mac80211 only!) and drowns out the real ones... I'm sure other parts of the tree are similarly affected.
johannes
On Thu, Aug 11, 2016 at 11:26:47AM -0700, Paul E. McKenney wrote:
If my upcoming testing of the two changes together pans out, I will give you a Tested-by -- I am guessing that you don't want to wait until the next merge window for these changes.
I was planning to stuff them in tip/locking/urgent, so they'd end up in this release.
On Fri, Aug 12, 2016 at 08:25:43PM +0200, Peter Zijlstra wrote:
On Thu, Aug 11, 2016 at 11:26:47AM -0700, Paul E. McKenney wrote:
If my upcoming testing of the two changes together pans out, I will give you a Tested-by -- I am guessing that you don't want to wait until the next merge window for these changes.
I was planning to stuff them in tip/locking/urgent, so they'd end up in this release.
They seem to work fine for me, so for both:
Tested-by: Paul E. McKenney paulmck@linux.vnet.ibm.com
Thanx, Paul
dri-devel@lists.freedesktop.org
-
Daniel Vetter -
Johannes Berg -
Paul E. McKenney -
Peter Zijlstra