From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1754058AbcAGXjz (ORCPT ); Thu, 7 Jan 2016 18:39:55 -0500 Received: from mga02.intel.com ([134.134.136.20]:1443 "EHLO mga02.intel.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752738AbcAGXjs (ORCPT ); Thu, 7 Jan 2016 18:39:48 -0500 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.20,535,1444719600"; d="scan'208";a="886104056" Date: Thu, 7 Jan 2016 16:39:46 -0700 From: Ross Zwisler To: Dave Chinner Cc: Jan Kara , Dan Williams , Ross Zwisler , "linux-kernel@vger.kernel.org" , "H. Peter Anvin" , "J. Bruce Fields" , "Theodore Ts'o" , Alexander Viro , Andreas Dilger , Andrew Morton , Dave Hansen , Ingo Molnar , Jan Kara , Jeff Layton , Matthew Wilcox , Matthew Wilcox , Thomas Gleixner , linux-ext4 , linux-fsdevel , Linux MM , "linux-nvdimm@lists.01.org" , X86 ML , XFS Developers Subject: Re: [PATCH v7 1/9] dax: fix NULL pointer dereference in __dax_dbg() Message-ID: <20160107233946.GD20802@linux.intel.com> Mail-Followup-To: Ross Zwisler , Dave Chinner , Jan Kara , Dan Williams , "linux-kernel@vger.kernel.org" , "H. Peter Anvin" , "J. Bruce Fields" , Theodore Ts'o , Alexander Viro , Andreas Dilger , Andrew Morton , Dave Hansen , Ingo Molnar , Jan Kara , Jeff Layton , Matthew Wilcox , Matthew Wilcox , Thomas Gleixner , linux-ext4 , linux-fsdevel , Linux MM , "linux-nvdimm@lists.01.org" , X86 ML , XFS Developers References: <1452103263-1592-1-git-send-email-ross.zwisler@linux.intel.com> <1452103263-1592-2-git-send-email-ross.zwisler@linux.intel.com> <20160107093402.GA8380@quack.suse.cz> <20160107231000.GO21461@dastard> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20160107231000.GO21461@dastard> 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 On Fri, Jan 08, 2016 at 10:10:00AM +1100, Dave Chinner wrote: > On Thu, Jan 07, 2016 at 10:34:02AM +0100, Jan Kara wrote: > > On Wed 06-01-16 11:14:09, Dan Williams wrote: > > > On Wed, Jan 6, 2016 at 10:00 AM, Ross Zwisler > > > wrote: > > > > __dax_dbg() currently assumes that bh->b_bdev is non-NULL, passing it into > > > > bdevname() where is is dereferenced. This assumption isn't always true - > > > > when called for reads of holes, ext4_dax_mmap_get_block() returns a buffer > > > > head where bh->b_bdev is never set. I hit this BUG while testing the DAX > > > > PMD fault path. > > > > > > > > Instead, verify that we have a valid bh->b_bdev, else just say "unknown" > > > > for the block device. > > > > > > > > Signed-off-by: Ross Zwisler > > > > Cc: Dan Williams > > > > --- > > > > fs/dax.c | 7 ++++++- > > > > 1 file changed, 6 insertions(+), 1 deletion(-) > > > > > > > > diff --git a/fs/dax.c b/fs/dax.c > > > > index 7af8797..03cc4a3 100644 > > > > --- a/fs/dax.c > > > > +++ b/fs/dax.c > > > > @@ -563,7 +563,12 @@ static void __dax_dbg(struct buffer_head *bh, unsigned long address, > > > > { > > > > if (bh) { > > > > char bname[BDEVNAME_SIZE]; > > > > - bdevname(bh->b_bdev, bname); > > > > + > > > > + if (bh->b_bdev) > > > > + bdevname(bh->b_bdev, bname); > > > > + else > > > > + snprintf(bname, BDEVNAME_SIZE, "unknown"); > > > > + > > > > pr_debug("%s: %s addr: %lx dev %s state %lx start %lld " > > > > "length %zd fallback: %s\n", fn, current->comm, > > > > address, bname, bh->b_state, (u64)bh->b_blocknr, > > > > > > I'm assuming there's no danger of a such a buffer_head ever being used > > > for the bdev parameter to dax_map_atomic()? Shouldn't we also/instead > > > go fix ext4 to not send partially filled buffer_heads? > > > > No. The real problem is a long-standing abuse of struct buffer_head to be > > used for passing block mapping information (it's on my todo list to remove > > that at least from DAX code and use cleaner block mapping interface but > > first I want basic DAX functionality to settle down to avoid unnecessary > > conflicts). Filesystem is not supposed to touch bh->b_bdev. > > That has not been true for a long, long time. e.g. XFS always > rewrites bh->b_bdev in get_blocks because the file may not reside on > the primary block device of the filesystem. i.e.: > > /* > * If this is a realtime file, data may be on a different device. > * to that pointed to from the buffer_head b_bdev currently. > */ > bh_result->b_bdev = xfs_find_bdev_for_inode(inode); > > > If you need > > that filled in, set it yourself in before passing bh to the block mapping > > function. > > That may be true, but we cannot assume that the bdev coming back > out of get_block is the same one that was passed in. For our use case I think this is fine - we just need the bdev to be filled in so that we can print reasonable error messages. If the filesystem updates bh->b_bdev during get_blocks(), we are fine with that.