From: Arnd Bergmann Subject: Re: [PATCH] Remove BKL from fs/locks.c Date: Thu, 16 Sep 2010 10:28:56 +0200 Message-ID: <201009161028.56837.arnd@arndb.de> References: <201009142206.54130.arnd@arndb.de> <1284501356.10782.139.camel@heimdal.trondhjem.org> <20100915204203.GA27837@fieldses.org> Mime-Version: 1.0 Content-Type: Text/Plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Cc: Trond Myklebust , Linus Torvalds , linux-kernel@vger.kernel.org, linux-fsdevel@vger.kernel.org, Matthew Wilcox , Christoph Hellwig , Miklos Szeredi , Frederic Weisbecker , Ingo Molnar , John Kacur , Stephen Rothwell , Andrew Morton , Thomas Gleixner To: "J. Bruce Fields" Return-path: In-Reply-To: <20100915204203.GA27837@fieldses.org> Sender: linux-kernel-owner@vger.kernel.org List-Id: linux-fsdevel.vger.kernel.org On Wednesday 15 September 2010 22:42:03 J. Bruce Fields wrote: > Looking over the server code.... The only code I see under the BKL is: > > - a few lease callbacks in fs/nfsd/nfs4state.c, none of which > sleep at this point, so all should be fine under a spinlock or > whatever we want. > - fs/nfsd/nfs4state.c:check_for_locks(), which explicitly takes > the BKL itself. All it does, though, is walk the lock list > for a given file and check whether any of them have a given > owner. It would be trivial to put it under some other lock > and/or move it to locks.c. Ok. In the version of the patch I sent out yesterday, I came to the same conclusion and put both of these under lock_flocks(), which is still the BKL but can be converted to a spinlock after we have sorted out ceph and lockd. If you and others are fine with this patch, I'll add it to my bkl/config series. Note that walking the i_flock list needs to use lock_flocks() (or the BKL), not a private lock. I guess that's what you are saying anyway, just making sure. Arnd