From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1756209AbcCCKAE (ORCPT ); Thu, 3 Mar 2016 05:00:04 -0500 Received: from mx2.suse.de ([195.135.220.15]:48787 "EHLO mx2.suse.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751243AbcCCKAC (ORCPT ); Thu, 3 Mar 2016 05:00:02 -0500 Date: Thu, 3 Mar 2016 11:00:21 +0100 From: Jan Kara To: Nikolay Borisov Cc: jack@suse.com, linux-kernel@vger.kernel.org Subject: Re: [RFC PATCH] quota: Fix possible GFP due to uninitialised pointers Message-ID: <20160303100021.GD2307@quack.suse.cz> References: <1456935569-20053-1-git-send-email-kernel@kyup.com> MIME-Version: 1.0 Content-Type: multipart/mixed; boundary="x+6KMIRAuhnl3hBn" Content-Disposition: inline In-Reply-To: <1456935569-20053-1-git-send-email-kernel@kyup.com> User-Agent: Mutt/1.5.24 (2015-08-30) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org --x+6KMIRAuhnl3hBn Content-Type: text/plain; charset=us-ascii Content-Disposition: inline On Wed 02-03-16 18:19:29, Nikolay Borisov wrote: > While debugging some issues with quota I realized that > it's possible to pass array with bogus dquot pointers from > __dquot_initialize to dqput. This can happen if the initialisation > of the dquot objects for an inode fail and the control flow is > transferred to the out_put label. In case only the USR or GRP quota > are initialised then the PRJ pointer in the "got" array would remain > uninitialised. This will cause the NULL ptr check in dqput to pass > but actually the pointer is going to be invalid. Eventually this would > cause a GFP. > > To fix this just zero out the got array > > Signed-off-by: Nikolay Borisov Thanks for spotting this and for the fix. There are couple of issues with your patch: a) You should use MAXQUOTAS instead of hardcoded 3 in the memset(). Even better just leave, specify proper initializer directly. b) You could remove the array initialization from the for loop. I have cleaned up the patch as attached and merged it into my tree. Honza > --- > fs/quota/dquot.c | 2 ++ > 1 file changed, 2 insertions(+) > > diff --git a/fs/quota/dquot.c b/fs/quota/dquot.c > index ef0d64b2a6d9..a0ab58fd85ae 100644 > --- a/fs/quota/dquot.c > +++ b/fs/quota/dquot.c > @@ -1408,6 +1408,8 @@ static int __dquot_initialize(struct inode *inode, int type) > > dquots = i_dquot(inode); > > + memset(got, 0, 3 * sizeof(struct dquot *)); > + > /* First get references to structures we might need. */ > for (cnt = 0; cnt < MAXQUOTAS; cnt++) { > struct kqid qid; > -- > 2.5.0 > > -- Jan Kara SUSE Labs, CR --x+6KMIRAuhnl3hBn Content-Type: text/x-patch; charset=us-ascii Content-Disposition: attachment; filename="0001-quota-Fix-possible-GPF-due-to-uninitialised-pointers.patch" >>From 130248216a62ab08b628d7c35b33842fb213fc7b Mon Sep 17 00:00:00 2001 From: Nikolay Borisov Date: Thu, 3 Mar 2016 10:54:57 +0100 Subject: [PATCH] quota: Fix possible GPF due to uninitialised pointers When dqget() in __dquot_initialize() fails e.g. due to IO error, __dquot_initialize() will pass an array of uninitialized pointers to dqput_all() and thus can lead to deference of random data. Fix the problem by properly initializing the array. Signed-off-by: Nikolay Borisov Signed-off-by: Jan Kara --- fs/quota/dquot.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) diff --git a/fs/quota/dquot.c b/fs/quota/dquot.c index e8467dc90c7e..dcec1edf579f 100644 --- a/fs/quota/dquot.c +++ b/fs/quota/dquot.c @@ -1410,7 +1410,7 @@ static int dquot_active(const struct inode *inode) static int __dquot_initialize(struct inode *inode, int type) { int cnt, init_needed = 0; - struct dquot **dquots, *got[MAXQUOTAS]; + struct dquot **dquots, *got[MAXQUOTAS] = {}; struct super_block *sb = inode->i_sb; qsize_t rsv; int ret = 0; @@ -1427,7 +1427,6 @@ static int __dquot_initialize(struct inode *inode, int type) int rc; struct dquot *dquot; - got[cnt] = NULL; if (type != -1 && cnt != type) continue; /* -- 2.6.2 --x+6KMIRAuhnl3hBn--