Messages in this thread | | | Date | Tue, 10 Nov 2009 15:20:41 -0500 (EST) | From | Christoph Lameter <> | Subject | Re: Subject: [RFC MM] mmap_sem scaling: Use mutex and percpu counter instead |
| |
On Sat, 7 Nov 2009, Minchan Kim wrote:
> On Sat, Nov 7, 2009 at 2:10 AM, Christoph Lameter > <cl@linux-foundation.org> wrote: > > On Sat, 7 Nov 2009, Minchan Kim wrote: > > > >> How about change from 'mm_readers' to 'is_readers' to improve your > >> goal 'scalibility'? > > > > Good idea. Thanks. Next rev will use your suggestion. > > > > Any creative thoughts on what to do about the 1 millisecond wait period? > > > > Hmm, > it would be importatn to prevent livelock for reader to hold lock > continuously before > hodling writer than 1 msec write ovhead.
Livelock because there are too frequent readers?
We could just keep the mutex locked to ensure that no new readers arrive.
> First of all, After we solve it, second step is that optimize write > overhead, I think.
| |