Messages in this thread | | | Date | Thu, 27 Oct 2011 09:17:01 -0700 | From | Tejun Heo <> | Subject | Re: [PATCH UPDATED 10/13] block, cfq: unlink cfq_io_context's immediately |
| |
Hello,
On Thu, Oct 27, 2011 at 10:31:19AM -0400, Vivek Goyal wrote: > On Wed, Oct 26, 2011 at 02:31:07PM -0700, Tejun Heo wrote: > > BUG_ON(atomic_long_read(&ioc->refcount) <= 0); > > + if (locked_q) > > + lockdep_assert_held(locked_q->queue_lock); > > > > Looks like lockdep_assert_held() can't be used if !CONFIG_LOCKDEP. So > Jens had put following patch to fix one compilation issue. You might > want to provide a null definition of lockdep_assert_held() in case > of !CONFIG_LOCKDEP. > > commit 334c2b0b8b2ab186fa198413386cba41fffcb4f2 > Author: Jens Axboe <axboe@kernel.dk> > Date: Tue Oct 25 15:51:48 2011 +0200 > > blk-throttle: use queue_is_locked() instead of lockdep_is_held() > > We can't use the latter if !CONFIG_LOCKDEP. > > Reported-by: Sedat Dilek <sedat.dilek@googlemail.com> > Signed-off-by: Jens Axboe <axboe@kernel.dk>
Eh? lockdep.h has dummy definition if !CONFIG_LOCKDEP. If lockdep_is_held() is causing compilation failure on !CONFIG_LOCKDEP, it's more likely to be have been caused by missing include than anything else. Jens, what was the failure? lockdep_is_held() provides way better protection against locking mistakes than spin_is_locked().
Thanks.
-- tejun
| |