linux-ext4 - e2fsprogs: f_eofblocks and f_pgsize_gt

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-next>] [day] [month] [year] [list]

Date:   Tue, 26 Jun 2018 18:44:24 +0200
From:   Lukas Czerner <lczerner@...hat.com>
To:     linux-ext4@...r.kernel.org
Subject: e2fsprogs: f_eofblocks and f_pgsize_gt_blksize fails on 1.44.3-rc1

Hi all,

I've noticed that f_eofblocks fails on systems with page size > 4k
(like ppc46) due to this commit.

commit 8d541f641ef3861276f7138b2f3e601036f03110
Author: Andreas Dilger <andreas.dilger@...el.com>
Date:   Sat Jun 23 14:30:52 2018 -0400

    e2fsck: handle preallocation for large PAGE_SIZE
    
    Fix handling of block preallocation support in cases where the kernel
    PAGE_SIZE is larger than the filesystem blocksize.

Test f_pgsize_gt_blksize was added right after this commit and it also fails
on ppc64

I belive that the reason is this

		if ((pb.last_init_lblock >= 0) &&
		    /* if size is smaller than expected by the block count,
		     * allow allocated blocks to end of PAGE_SIZE.
		     * last_init_lblock is the last in-use block, so it is
		     * the minimum expected file size, but +1 because it is
		     * the base-zero block number and not the block count. */
		    (size < (__u64)pb.last_init_lblock * fs->blocksize) &&
>>>>		    ((pb.last_init_lblock + 1) / blkpg * blkpg !=
>>>>		     (pb.last_init_lblock + 1) ||
		     size < (__u64)(pb.last_init_lblock & ~(blkpg - 1)) *
		     fs->blocksize))
			bad_size = 3;

while the fix seems correct in context I do not understand the point
of the check.  Seems like the reason for it is to make sure that if we do
have any initialized blocks after i_size, it's always up the the page size
boundary.

This was added with commit

commit 9f0288d3bbbf47bb05e5abf1e570df368476a8cd
Author: Theodore Ts'o <tytso@....edu>
Date:   Fri Aug 3 20:43:37 2007 -0400

    e2fsck: Allow i_size to be rounded up to the size of a VM page

    Allow files to be preallocated on-disk up to the next multiple of the
    system's page size without complaining about extra blocks.


It's not explained why we need this. Do we ever want to allow initialized
extents (blocks) past i_size ? I though we do not, what am I missing ?
Shouldn't we only be doing this ?

	if ((pb.last_block >= 0) &&
	    (size < (__u64) pb.last_block * fs->blocksize))
		bad_size = 3;


Thanks!
-Lukas