linux-kernel - utimensat() non-conformances and fixes [v4] (test results)

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>] [day] [month] [year] [list]

Date:	Tue, 03 Jun 2008 22:15:07 +0200
From:	Michael Kerrisk <mtk.manpages@...glemail.com>
To:	Andrew Morton <akpm@...ux-foundation.org>
CC:	lkml <linux-kernel@...r.kernel.org>,
	Christoph Hellwig <hch@....de>,
	Miklos Szeredi <miklos@...redi.hu>,
	Al Viro <viro@...iv.linux.org.uk>, jamie@...reable.org,
	Ulrich Drepper <drepper@...hat.com>,
	linux-fsdevel@...r.kernel.org,
	Subrata Modak <subrata@...ux.vnet.ibm.com>
Subject: utimensat() non-conformances and fixes [v4] (test results)

Andrew,

The following is why I think my patch should
preferably be applied for 2.6.26, rather than
waiting for .27.

The first reason is the obvious.  The test suite
runs 99 tests.  With my patch applied to 2.6.24-rc4,
all tests pass (see the attached test results).

There is another reason why I think we should apply the
patch for .26.

On 2.6.25, we have:

Total tests: 99; passed: 73; failed: 26

On 2.6.26-rc4, we have:
Total tests: 99; passed: 74; failed: 25

As far as I can see, this is due to the changes from
Miklos's

commit 02c6be615f1fcd37ac5ed93a3ad6692ad8991cd9
Author: Miklos Szeredi <mszeredi@...e.cz>
Date:   Thu May 1 04:34:45 2008 -0700

But the small improvement in the test results isn't
quite what it seems.  There is for 2.6.25:

Failed tests:  21 22 23 24 28 29 30 31 32 33 38 46 47 48 49 50 51 58 59 60 75 76 87 88 98 99

And for 2.6.24-rc4:

Failed tests:  21 24 28 29 30 31 32 33 38 41 46 47 48 49 50 51 58 59 60 75 76 87 88 98 99

Looking at the list shows that 2.6.24-rc4 fixes tests
22 and 23, but breaks test 41[*].  I don't think we
should do that.

Cheers,

Michael


[*] Test 41 is:

Readable file descriptor (futimens(3)) test
Owner=root; perms=-rw-rw-rw-; EFAs=---
./test_utimensat -q -d /tmp/utimensat_tests/utimensat.test_file NULL 0 n 0 n
EXPECTED: SUCCESS y  y
RESULT:   EACCES
FAILED test 41

In other words, make the call

utimensat(fd, NULL, {{0, UTIME_NOW}, {0. UTIME_NOW}}, 0)

where fd is a read-only (O_RDONLY) descriptor that refers to
a file that is not owned by the caller, but is writable by
the caller.  This call should succeed.

-- 
Michael Kerrisk
Linux man-pages maintainer; http://www.kernel.org/doc/man-pages/
Found a bug? http://www.kernel.org/doc/man-pages/reporting_bugs.html



View attachment "utimensat_tests-2.6.25.log" of type "text/x-log" (24016 bytes)

View attachment "utimensat_tests-2.6.26-rc4.log" of type "text/x-log" (23883 bytes)

View attachment "utimensat_tests-2.6.26-rc4-utimensat-fix-v4.log" of type "text/x-log" (23819 bytes)