Messages in this thread | | | Date | Sat, 14 Mar 2009 07:32:19 -0700 (PDT) | From | Davide Libenzi <> | Subject | Re: [PATCH] fs: fput() can be called from interrupt context |
| |
On Sat, 14 Mar 2009, Trond Myklebust wrote:
> > Did I miss something? This wouldn't be (eventually) on keventd shoulders, > > but on aio work queue (aio_wq). > > I guess we could do the same optimization we're already doing for ki_filp, > > for ki_eventfd ... > > Last I checked, a call to schedule_work(&fddef->wq) would still run > fd_defer_queue() under keventd.
Oh, you were commenting about Eric's patch-set. I thought you were talking about the current status (that my reply to Andrew was hinting to), and was wondering for a while from where fd_defer_queue() came from.
> That said, even if you were to run it under aio_wq, the argument remains > the same: do you really want to add potentially long lasting sleeps onto > a work queue that was designed to service fast i/o requests?
No, and if you roll back, you couldn't miss to notice that mine was a question about the f_count direct manipulation/optimization. We can do the same optimization for eventfd, although both will fall in the aio kthread in any case, if we're holding the last insteance of the file* (that I guess we can consider an unlikely case). Dunno how worth it could be adding a new NCPU*thread(s) just to handle delayed fputs.
- Davide
| |