Re: [PATCH v6 12/12] mmap locking API: convert mmap_sem comments

[Vlastimil Babka]

On 5/21/20 7:25 PM, Andrew Morton wrote:
> On Thu, 21 May 2020 00:50:56 -0700 Michel Lespinasse <walken@xxxxxxxxxx> wrote:
> 
>> > >   * Must be called holding task's alloc_lock to protect task's mems_allowed
>> > > - * and mempolicy.  May also be called holding the mmap_semaphore for write.
>> > > + * and mempolicy.  May also be called holding the mmap_lockaphore for write.
>> > >   */
>> > >  static int mpol_set_nodemask(struct mempolicy *pol,
>> > >                    const nodemask_t *nodes, struct nodemask_scratch *nsc)
>> >
>> > :)
>> 
>> Haha, good catch !
> 
> aww, you're all so cruel.  The world would be a better place if Linux
> had lockaphores!

Heh, in fact it does have them, but they are called "btrfs extent buffer locks",
see (or perhaps rather not) fs/btrfs/locking.c :)