| Date | Wed, 28 Nov 2012 19:46:31 -0500 | From | Benjamin LaHaise <> | Subject | Re: [PATCH 12/25] aio: Refcounting cleanup |
| |
Hi Kent,
On Wed, Nov 28, 2012 at 08:43:36AM -0800, Kent Overstreet wrote: > + * now it's safe to cancel any that need to be. > + */ > +static void free_ioctx(struct kioctx *ctx) ... > + aio_nr -= ctx->max_reqs; > + spin_unlock(&aio_nr_lock); > + > + synchronize_rcu(); > + > + pr_debug("freeing %p\n", ctx); > + kmem_cache_free(kioctx_cachep, ctx); > +}
As mentioned on irc, we probably want to avoid the synchronize_rcu() overhead, since delays here will impact the time it takes for a task to exit. Cheers,
-ben -- "Thought is the essence of where you are now."
|