On Wed, Dec 09, 2020 at 11:07:55PM +0000, Matthew Wilcox wrote: > On Wed, Dec 09, 2020 at 03:01:36PM -0800, Linus Torvalds wrote: > > On Wed, Dec 9, 2020 at 2:58 PM Al Viro <viro@xxxxxxxxxxxxxxxxxx> wrote: > > > > > > On Wed, Dec 09, 2020 at 07:49:38PM +0000, Matthew Wilcox wrote: > > > > > > > > Assuming this is safe, you can use RCU_INIT_POINTER() here because you're > > > > storing NULL, so you don't need the wmb() before storing the pointer. > > > > > > fs/file.c:pick_file() would make more interesting target for the same treatment... > > > > Actually, don't. > > > > rcu_assign_pointer() itself already does the optimization for the case > > of a constant NULL pointer assignment. > > > > So there's no need to manually change things to RCU_INIT_POINTER(). > > I missed that, and the documentation wasn't updated by > 3a37f7275cda5ad25c1fe9be8f20c76c60d175fa. Can't trust the author of that patch! ;-) > Paul, how about this? > > +++ b/Documentation/RCU/Design/Requirements/Requirements.rst > @@ -1668,8 +1668,10 @@ against mishaps and misuse: > this purpose. > #. It is not necessary to use rcu_assign_pointer() when creating > linked structures that are to be published via a single external > - pointer. The RCU_INIT_POINTER() macro is provided for this task > - and also for assigning ``NULL`` pointers at runtime. > + pointer. The RCU_INIT_POINTER() macro is provided for this task. > + It used to be more efficient to use RCU_INIT_POINTER() to store a > + ``NULL`` pointer, but rcu_assign_pointer() now optimises for a constant > + ``NULL`` pointer itself. > > This not a hard-and-fast list: RCU's diagnostic capabilities will > continue to be guided by the number and type of usage bugs found in Looks good to me! If you send a complete patch, I will be happy to pull it in. Thanx, Paul
