[PATCH 1/6] mm: change locked_vm's type from unsigned long to atomic64_t

Daniel Jordan daniel.m.jordan at oracle.com
Thu Apr 4 03:09:03 AEDT 2019


On Wed, Apr 03, 2019 at 06:46:07AM +0200, Christophe Leroy wrote:
> 
> 
> Le 02/04/2019 à 22:41, Daniel Jordan a écrit :
> > Taking and dropping mmap_sem to modify a single counter, locked_vm, is
> > overkill when the counter could be synchronized separately.
> > 
> > Make mmap_sem a little less coarse by changing locked_vm to an atomic,
> > the 64-bit variety to avoid issues with overflow on 32-bit systems.
> 
> Can you elaborate on the above ? Previously it was 'unsigned long', what
> were the issues ?

Sure, I responded to this in another thread from this series.

> If there was such issues, shouldn't there be a first patch
> moving it from unsigned long to u64 before this atomic64_t change ? Or at
> least it should be clearly explain here what the issues are and how
> switching to a 64 bit counter fixes them.

Yes, I can explain the motivation in the next version.


More information about the Linuxppc-dev mailing list