From: Ian Kent Subject: Re: [PATCH 1/4] Make hash table scale to thousands of entries Date: Fri, 16 Jan 2009 15:39:12 +0900 Message-ID: <1232087952.3166.98.camel@zeus.themaw.net> References: <1231526823-18595-1-git-send-email-vaurora@redhat.com> <1232032573.3100.53.camel@zeus.themaw.net> <408431290901151233q33699529r6119788c998866ed@mail.gmail.com> <408431290901151257i70400351yc3ab9360136c8e9c@mail.gmail.com> <1232065653.3166.19.camel@zeus.themaw.net> <1232066406.3166.28.camel@zeus.themaw.net> <1232070195.3166.36.camel@zeus.themaw.net> <408431290901151959o7c504c79k88681dddb468359f@mail.gmail.com> <1232079714.3166.56.camel@zeus.themaw.net> <408431290901152204i8d2da71vbf8ee5f2e2f84fa@mail.gmail.com> <1232087057.3166.88.camel@zeus.themaw.net> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <1232087057.3166.88.camel@zeus.themaw.net> List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Sender: autofs-bounces@linux.kernel.org Errors-To: autofs-bounces@linux.kernel.org To: Paul Wankadia Cc: autofs@linux.kernel.org On Fri, 2009-01-16 at 15:24 +0900, Ian Kent wrote: > On Fri, 2009-01-16 at 17:04 +1100, Paul Wankadia wrote: > > On Fri, Jan 16, 2009 at 3:21 PM, Ian Kent wrote: > > > > > + if (me && ctxt->mtime <= me->age) > > > > > > You have to call stat(2) or fstat(2) at some point or else > > ctxt->mtime > > > will never change. > > > > > > We set the mtime when we read the entire map, see > > modules/lookup_file.c:lookup_read_map(). This is totally > > taking > > advantage of the existing code to read a map that is marked as > > stale. It > > doesn't however attempt to simplify the code which works out > > if we need > > to mark the map stale but we probably don't want to anyway as > > we need to > > ensure we use up to date entries for lookups done while > > waiting for the > > map update to be done. > > > > Sorry, I don't understand how that condition will evaluate to false. > > You won't read the map until ctxt->mtime changes, but it won't change > > until you read the map. > > Mmmm .. before that last change I was assuming the code which sets the > stale flag would cause the read but then I removed that bit and ... > oops! Because, after a mount lookup we check the stale status and queue a map read if the map is stale. But then I went and removed the bit that sets the map stale because it's lazy in identifying if the map has changed so we could get quite a few mount lookups before a map read gets queued. I'll add back the stat(2) I originally had their before I posted the patch. I'm also thinking of checking if a map read task is in progress when a mount lookup comes in and waiting for it to complete before continuing with the lookup (to reduce the number of file scans a bit). But I'll need to have a look around to see if that will be a problem. Ian