linux-ext4 - Re: test for EA validity checking

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <20071115184514.GA18130@webber.adilger.int>
Date:	Thu, 15 Nov 2007 11:45:14 -0700
From:	Andreas Dilger <adilger@....com>
To:	Theodore Ts'o <tytso@....edu>
Cc:	linux-ext4@...r.kernel.org
Subject: Re: test for EA validity checking

On Nov 10, 2007  05:10 -0700, Andreas Dilger wrote:
> Attached is a test image for extended attribute block checking.
> This small image contains a number of different kinds of corruptions
> (bad magic, bad checksum, empty block, bad EA block number), as well
> as an old-format (v1) EA, and a valid v2 EA.

Attached is an updated patch with a new test image, that includes a
corrupt EA set on a block device.  That had been causing us some
e2fsck heartburn because of random inode table corruption.

This goes along with an updated "expand-extra-isize" patch.  The root of
the problem was that I had added an extra check to verify the EA magic
in ext2fs_read_ext_attr(), but the return value of this function was used
in check_blocks() to set pctx->errcode, and later in check_ext_attr() the
non-zero value of pctx->errcode caused an abort when
ext2fs_inode_has_valid_blocks() was false and it caused PR_1_BLOCK_ITERATE
to be hit (a fatal error).

I ended up checking the magic in all of the ext2fs_read_ext_attr() callsites,
but IMHO the library should at least be doing basic validity checking like
this.

One question that remains unclear is whether pctx->errcode being set from
the early call of check_ext_attr() should cause check_blocks() to abort?
In the common regular-file/directory/slow-symlink case pctx->errcode is
reset by the call to ext2fs_block_iterate2(), so it would seem reasonable
to go back to checking EA magic in ext2fs_read_ext_attr(), and then clearing
pctx->errcode in check_ext_attr() if the problem is fixed.

Cheers, Andreas
--
Andreas Dilger
Sr. Software Engineer, Lustre Group
Sun Microsystems of Canada, Inc.

Download attachment "e2fsprogs-tests-f_ea_checks.patch" of type "application/octet-stream" (4938 bytes)

View attachment "e2fsprogs-expand-extra-isize.patch" of type "text/plain" (50457 bytes)