linux-kernel - 2.6.37-next problems with fs dcache patches.

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: <19902.1293047238@localhost>
Date:	Wed, 22 Dec 2010 14:47:18 -0500
From:	Valdis.Kletnieks@...edu
To:	Nick Piggin <npiggin@...nel.dk>
Cc:	linux-kernel@...r.kernel.org, linux-fs@...r.kernel.org
Subject: 2.6.37-next problems with fs dcache patches.

Wish I was able to bisect this down better, but a number of steps in the bisect
had to be skipped, this series is *not* bisect-clean.  Sorry I didn't do this
earlier, but I got detoured chasing the mlockall() issue which originally looked
like the culprit.

Symptoms:  My initramfs fails to mount the root filesystem, and gets
dropped down to a debug shell running off the initramfs.

Git bisect log so far:

git bisect start
# bad: [e935fc8e7828d083dde7b696ee5121cbbaadda07] Add linux-next specific files for 20101217
git bisect bad e935fc8e7828d083dde7b696ee5121cbbaadda07
# good: [e8a7e48bb248a1196484d3f8afa53bded2b24e71] Linux 2.6.37-rc4
git bisect good e8a7e48bb248a1196484d3f8afa53bded2b24e71
# bad: [e18448d78a4893a4714be343e0c8688a12a54e78] Merge remote branch 'net/master'
git bisect bad e18448d78a4893a4714be343e0c8688a12a54e78
# bad: [8698cfbb100845603f167fadb0432535122c453a] Merge branch 'quilt/kernel-doc'
git bisect bad 8698cfbb100845603f167fadb0432535122c453a
# good: [b4b604c1cd3e34a0bf03dd19e030b698a3438b05] Merge remote branch 'davinci/davinci-next'
git bisect good b4b604c1cd3e34a0bf03dd19e030b698a3438b05
# good: [752a83e8514ddc27c9767d235cb0858dec2a0266] Merge remote branch '52xx-and-virtex/next'
git bisect good 752a83e8514ddc27c9767d235cb0858dec2a0266
# good: [87fab72d587fe60a9d023e424a77a527e92781b3] Merge remote branch 'nilfs2/for-next'
git bisect good 87fab72d587fe60a9d023e424a77a527e92781b3
# bad: [493d590cb2ea1cef5a5b58bcce0f311f4c3f9efa] Merge remote branch 'pci/linux-next'
git bisect bad 493d590cb2ea1cef5a5b58bcce0f311f4c3f9efa
# bad: [68eb4446938b2ec9cb0b3e4c8e76f71940723f81] fs: dcache per-bucket dcache hash locking
git bisect bad 68eb4446938b2ec9cb0b3e4c8e76f71940723f81
# skip: [17e24aaaf2dd0fc305f19cc8c3a70dba5be3aeb7] fs: dcache avoid starvation in dcache multi-step operations
git bisect skip 17e24aaaf2dd0fc305f19cc8c3a70dba5be3aeb7
# good: [eb7fe777998ef1b46f650fe3b5a35eabb2854a82] fs: dcache reduce dput locking
git bisect good eb7fe777998ef1b46f650fe3b5a35eabb2854a82
# good: [d36c2a8e3deef473a74874a87a587807020879d8] kernel: optimise seqlock
git bisect good d36c2a8e3deef473a74874a87a587807020879d8
# skip: [27056795db2de7bfefc5ea5cbc4486ba8f3fe90d] fs: cache optimise dentry and inode for rcu-walk
git bisect skip 27056795db2de7bfefc5ea5cbc4486ba8f3fe90d
# skip: [c06020c90ab49382419e71f87ed341d28f87a166] fs: d_revalidate_rcu for rcu-walk
git bisect skip c06020c90ab49382419e71f87ed341d28f87a166
# skip: [40f9b00323a73859ac53d7d4f2f820c4c8d33165] fs: dcache remove d_mounted
git bisect skip 40f9b00323a73859ac53d7d4f2f820c4c8d33165
# bad: [0c259f9c9ee263a5d907030e29c52fd8d206c631] fs: provide rcu-walk aware permission i_ops
git bisect bad 0c259f9c9ee263a5d907030e29c52fd8d206c631
# skip: [1f42f4566770e4d7bda4be9bfc0f39a237685f54] fs: dcache reduce branches in lookup path
git bisect skip 1f42f4566770e4d7bda4be9bfc0f39a237685f54
# skip: [ff7f5c4c2c1542cd8a8c538b11441aac47d87709] fs: fs_struct use seqlock
git bisect skip ff7f5c4c2c1542cd8a8c538b11441aac47d87709
# skip: [3c884da2830985495ddbc8817871070cb6e9d86d] fs: rcu-walk for path lookup
git bisect skip 3c884da2830985495ddbc8817871070cb6e9d86d

The first 'skip' was due to a kernel that completely and totally failed immediately,
grub loaded it and *thud* it was dead, needed to hold down the power button for
5 seconds.

The later skips were all due to build failures:

  CC      security/security.o
security/security.c: In function 'security_inode_exec_permission':
security/security.c:520:6: error: 'flags' undeclared (first use in this function)
security/security.c:520:6: note: each undeclared identifier is reported only once for each function it appears in
make[1]: *** [security/security.o] Error 1

And we end with:

git bisect skip
There are only 'skip'ped commits left to test.
The first bad commit could be any of:
1f42f4566770e4d7bda4be9bfc0f39a237685f54
40f9b00323a73859ac53d7d4f2f820c4c8d33165
27056795db2de7bfefc5ea5cbc4486ba8f3fe90d
ff7f5c4c2c1542cd8a8c538b11441aac47d87709
3c884da2830985495ddbc8817871070cb6e9d86d
c06020c90ab49382419e71f87ed341d28f87a166
0c259f9c9ee263a5d907030e29c52fd8d206c631
We cannot bisect more!

Any ideas?  Patches that weren't in linux-next as of 12-17?

Content of type "application/pgp-signature" skipped