Messages in this thread | | | Date | Wed, 26 Oct 2011 16:38:46 -0400 | From | Vivek Goyal <> | Subject | Re: [PATCH 10/13] block, cfq: unlink cfq_io_context's immediately |
| |
On Wed, Oct 26, 2011 at 12:55:06PM -0700, Tejun Heo wrote: > Hello, > > On Wed, Oct 26, 2011 at 03:48:57PM -0400, Vivek Goyal wrote: > > > + ioc_release_depth_inc(cic->q); > > > + cic->exit(cic); > > > + cic->release(cic); > > > + ioc_release_depth_dec(cic->q); > > > > cic->release(cic) can free the cic? Are we accessing cic after freeing it > > up in ioc_release_depth_dec(cic->q); > > As cic is RCU freed and we're running w/ irq disabled, this shouldn't > blow up, but yeah, I'll update it to use local variable instead.
Tejun,
So the only reason that cic needs to be rcu protected because in cfq_cic_lookup() we can access ioc->ioc_data which can point to a cic which is not ours and we are not holding queue lock for that and we can't access cic->q ?
Rest of the places we are either holding ioc_lock or accessing our own cic under associated queue lock so it should be safe?
I am wondering if there should be a small text file explaining all the locking in this area.
Good to see that you are cleaning it all up.
Thanks Vivek
| |