linux-ext4 - [Bug 196405] mkdir mishandles st_nlink in ext4 directory with 64997 subdirectories

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

Date:   Wed, 19 Jul 2017 19:44:49 +0000
From:   bugzilla-daemon@...zilla.kernel.org
To:     linux-ext4@...nel.org
Subject: [Bug 196405] mkdir mishandles st_nlink in ext4 directory with 64997
 subdirectories

https://bugzilla.kernel.org/show_bug.cgi?id=196405

--- Comment #12 from Paul Eggert (eggert@...ucla.edu) ---
(In reply to Theodore Tso from comment #1)
> If you don't want this behavior because strict POSIX compliance is
> preferable to failing with ENOSPC when you hit that overflow
> condition, you can create the file system with the dir_nlink feature
> turned off.

POSIX says that mkdir should fail with EMLINK if the link count in the parent
directory would exceed LINK_MAX. If the POSIX-compliant mode is setting
errno==ENOSPC then that sounds like a bug; it should fail with errno==EMLINK.

Although the documentation for tune2fs says that dir_nlink can be cleared in an
existing ext4 filesystem, this did not work for me on Fedora 26 x86-64. I
unmounted the filesystem, ran "tune2fs -O ^dir_nlink /dev/sda2", remounted the
filesystem, and verified that dir_nlink was turned off by running "dumpe2fs -h
/dev/sda2". The fts-test.c program attached to this bug report still failed in
the same way. Either this is a filesystem bug that should be fixed, or the
documentation should be changed so that it does not incorrectly say that
tune2fs can usefully change the dir_nlink option.

I then created a new ext4 file system with dir_nlink disabled from birth. Here,
the test program failed in mkdir ("d/32757", 0777) with ENOSPC. Yet there was
plenty of space in the file system, and the link count was only 32758, which is
way lower than LINK_MAX (which is 65000). Presumably the ext4 code arbitrarily
limits the link count to 32758 (i.e., to 2**15 - 10) when dir_nlink is clear.
This sounds like a bug; the limit should be 65000.

One more thing. In the typical case where dir_nlink is set, there is an
off-by-one error, as LINK_MAX is 65000 and in fts-test.c the link count should
be 65000 so this should work. That is, the link count should wrap around to 1
when it would otherwise reach 65001, not when it would otherwise reach 65000.

-- 
You are receiving this mail because:
You are watching the assignee of the bug.