linux-ext4 - fstest generic/696 failure on ext4 fs with quotas+idmap

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]

Message-ID: <87jzmxisqm.fsf@suse.de>
Date: Wed, 21 Feb 2024 18:20:49 +0000
From: Luis Henriques <lhenriques@...e.de>
To: linux-ext4@...r.kernel.org
Cc: Christian Brauner <brauner@...nel.org>
Subject: fstest generic/696 failure on ext4 fs with quotas+idmap

Hi!

The fstest generic/696 (and 697) fail on ext4 when the filesystem is
created with quota support (-O quota).  It's really easy to reproduce, and
it fails when doing the idmapped tests (setgid_create_umask_idmapped() and
setgid_create_umask_idmapped_in_userns()).

The failure happens when the test does an openat() with O_TMPFILE:

  ext4_tmpfile()
    __ext4_new_inode()
      dquot_initialize()
        dqget()

and at this point the error occurs:

	if (!qid_has_mapping(sb->s_user_ns, qid))
		return ERR_PTR(-EINVAL);

qid is '-1', which is invalid, but I'm failing to understand if it should
really be invalid or if dqget() should handle this invalid qid some other
way.  Earlier, __ext4_new_inode() called inode_init_owner(), which indeed
sets inode->i_uid with '-1'.

I've been trying to figure it out, but it's very tricky to follow all the
details, so I decided to ask here and see if anyone has any idea.  Is this
a known issue?  Maybe the issue is with the test itself, and not with
ext4, quota or idmapped code.

Cheers,
-- 
Luís