Messages in this thread | | | Date | Mon, 30 Sep 2013 12:03:16 -0400 | From | Waiman Long <> | Subject | Re: [PATCH] rwsem: reduce spinlock contention in wakeup code path |
| |
On 09/30/2013 03:05 AM, Ingo Molnar wrote: > * Michel Lespinasse<walken@google.com> wrote: > >> That said, I am very scared of using rwlock_t here, and I would much >> prefer we choose a fair lock (either spinlock or a new rwlock >> implementation which guarantees not to starve any locker thread) > Given how few users rwlock_t has today we could attempt to make it > reader-writer fair, as long as the usecase of a hardirq or softirq > context always getting nested read access on the same CPU is preserved. > > (but that can be done - if nothing else then with an explicit context > check.) > > Thanks, > > Ingo
My queue rwlock patch is able to handle the use case of a hardirq and softirq context and can be made almost as fair as the ticket spinlock.
-Longman
| |