From mboxrd@z Thu Jan 1 00:00:00 1970 From: David Howells Subject: Re: [PATCH 1/4] CacheFiles: Use the ->write() file op rather than a special kernel aop Date: Fri, 03 Apr 2009 14:57:19 +0100 Message-ID: <11023.1238767039@redhat.com> References: <20090403134927.GA31907@infradead.org> <20090403094138.9510.80681.stgit@warthog.procyon.org.uk> Cc: dhowells@redhat.com, nickpiggin@yahoo.com.au, nfsv4@linux-nfs.org, linux-kernel@vger.kernel.org, linux-fsdevel@vger.kernel.org To: Christoph Hellwig Return-path: Received: from mx2.redhat.com ([66.187.237.31]:41584 "EHLO mx2.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1765326AbZDCN5b (ORCPT ); Fri, 3 Apr 2009 09:57:31 -0400 In-Reply-To: <20090403134927.GA31907@infradead.org> Sender: linux-fsdevel-owner@vger.kernel.org List-ID: Christoph Hellwig wrote: > > + if (!aops->bmap) > > goto check_error; > > What are you doing using bmap? You really shouldn't call into it from > anywhere but FIBMAP. Yes, swap currently does but it's a major pain > in the neck and Peter has been working on a proper interface for swap > for a while. Checking to see whether there's a hole in the file. If there's a hole, that represents data I need to fetch; if there isn't that represents data I have in the cache. I don't care _where_ the data is, only whether it exists or not. David