From mboxrd@z Thu Jan 1 00:00:00 1970 From: Evgeniy Polyakov Subject: Re: NFS/credentials leak in 2.6.29-rc1 Date: Tue, 20 Jan 2009 18:23:04 +0300 Message-ID: <20090120152304.GA28592@ioremap.net> References: <20090120114649.GA15832@ioremap.net> <20090120151125.GB24266@fieldses.org> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: Trond Myklebust , linux-fsdevel@vger.kernel.org, David Howells To: "J. Bruce Fields" Return-path: Received: from genesysrack.ru ([195.178.208.66]:47823 "EHLO tservice.net.ru" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1755769AbZATPXH (ORCPT ); Tue, 20 Jan 2009 10:23:07 -0500 Content-Disposition: inline In-Reply-To: <20090120151125.GB24266@fieldses.org> Sender: linux-fsdevel-owner@vger.kernel.org List-ID: On Tue, Jan 20, 2009 at 10:11:25AM -0500, J. Bruce Fields (bfields@fieldses.org) wrote: > This doesn't look familiar, no; thanks for the report. I guess we > should take a careful look at the recent changes to fs/nfsd/auth.c? If creds are allocated in nfsd_setuser() and never freed? groups_alloc() can also explain size-256 slab grew, so this may be the place where things are allocated, but why they are not freed? This may also explain why I did not see this for the large sequential IO, since number of requests to the server was noticebly smaller, than in random IO test. -- Evgeniy Polyakov