Date: Thu, 26 Apr 2012 07:53:48 -0400 (EDT) From: Rick Macklem <rmacklem@uoguelph.ca> To: Oliver Brandmueller <ob@e-Gitt.NET> Cc: freebsd-stable@freebsd.org Subject: Re: 9-STABLE, ZFS, NFS, ggatec - suspected memory leak Message-ID: <1695893954.3415457.1335441228131.JavaMail.root@erie.cs.uoguelph.ca> In-Reply-To: <20120426095454.GX65313@e-Gitt.NET>
next in thread | previous in thread | raw e-mail | index | archive | help
Oliver Brandmueller wrote: > Hi, > > On Wed, Apr 25, 2012 at 05:34:05PM -0400, Rick Macklem wrote: > > Good work isolating this! > > Thank you! > > > I now see the problem. The new NFS server code assumed that > > VOP_LOOKUP() > > calls would not set SAVENAME, so it expected the path buffer to be > > free'd > > by the nfsvno_namei() when it hadn't set SAVENAME. > > > > It turns out ZFS sets SAVENAME in zfs_lookup() for the DELETE case. > > > > The attached patch, which is also here, should fix the problem for > > now: > > http://people.freebsd.org/~namei-leak.patch > > > > Please test this patch and let me know if it fixes the leak. > > Thanx for the explanation - anf coming up with a patch that fast! > > I applied the patch and in my testing environment I don't see the leak > anymore. I will not be able to apply it to our prod environment before > about mid of May, since I don't want to leave my fellow co-workers > with > any problems while being on holidays :) > > > jwd@ is working on a patch that will avoid using uma_zalloc() to get > > a path buffer for most cases for performance reasons. Once that > > patch > > goes it, the code should be patched so that it checks for SAVENAME > > being > > set for all cases where uma_zalloc() has allocated a path buffer, so > > that > > no more leaks like this will happen when underlying file systems set > > SAVENAME. > > So is itlikely, that this final version will at some time be included > into 9-STABLE or will the current patch (given more positive results > come up) make it into -STABLE until the other one is ready? > Well, I think I can commit it to head with an MFC of 1 month. That way, hopefully you will have been able to test it in your production environment before it gets MFC'd to 9-STABLE. I suspect John's patch will be committed sometime later, but I'll leave that up to him. (He runs a server with ZFS, so he should be able to check for the leak.) > > Greeting and many thanks. > And thanks for tracking it down. It's surprising only one other person noticed this. I guess others don't have enough removes going on for the to get serious. rick
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?1695893954.3415457.1335441228131.JavaMail.root>