2.6.30-rc: reiserfs_delete_xattrs log spam (bug 13122)

2.6.30-rc: reiserfs_delete_xattrs log spam (bug 13122)

[Stefan Richter]

I just switched from 2.6.29.1 to 2.6.30-rc5, and I'm getting the same as 
the reporter of http://bugzilla.kernel.org/show_bug.cgi?id=13122 alias
http://marc.info/?l=linux-kernel&m=123990989515105, i.e. repetitions of
 >>>
May  9 14:27:22 stein REISERFS warning (device sda1): jdm-20004 
reiserfs_delete_xattrs: Couldn't delete all xattrs (-13)
May  9 14:27:22 stein
<<<

This happens in irregular intervals, but a lot of times --- presumably 
when certain files are accessed.  I'm on a recently set-up Gentoo 
machine.  I created its root filesystem using the Gentoo x86-64 minimal 
boot CD; my guess is that the kernel of the boot CD had 
CONFIG_REISERFS_FS_XATTR enabled.  All kernels which I used since then 
have it disabled:  2.6.27-gentoo-r8, 2.6.29.1, now 2.6.30-rc5.
-- 
Stefan Richter
-=====-=-=== -=-= -==-=
http://arcgraph.de/sr/

Re: 2.6.30-rc: reiserfs_delete_xattrs log spam (bug 13122)

[Stefan Richter]

On  9 May, Stefan Richter wrote:
> I just switched from 2.6.29.1 to 2.6.30-rc5, and I'm getting the same as 
> the reporter of http://bugzilla.kernel.org/show_bug.cgi?id=13122 alias
> http://marc.info/?l=linux-kernel&m=123990989515105, i.e. repetitions of
>  >>>
> May  9 14:27:22 stein REISERFS warning (device sda1): jdm-20004 
> reiserfs_delete_xattrs: Couldn't delete all xattrs (-13)
> May  9 14:27:22 stein
> <<<
> 
> This happens in irregular intervals, but a lot of times --- presumably 
> when certain files are accessed.  I'm on a recently set-up Gentoo 
> machine.  I created its root filesystem using the Gentoo x86-64 minimal 
> boot CD; my guess is that the kernel of the boot CD had 
> CONFIG_REISERFS_FS_XATTR enabled.  All kernels which I used since then 
> have it disabled:  2.6.27-gentoo-r8, 2.6.29.1, now 2.6.30-rc5.

PS:  Also, the following came up at this (= first) boot of 2.6.30-rc, as
a warm reboot from 2.6.29.1. I have no idea whether it's connected to
the issue.

May  9 13:56:53 stein REISERFS (device sda1): found reiserfs format "3.6" with standard journal
May  9 13:56:53 stein REISERFS (device sda1): using ordered data mode
May  9 13:56:53 stein REISERFS (device sda1): journal params: device sda1, size 8192, journal first block 18, max trans len 1024, max batch 900, max commit age 30, max trans age 30
May  9 13:56:53 stein REISERFS (device sda1): checking transaction log (sda1)
May  9 13:56:53 stein REISERFS (device sda1): Using r5 hash to sort names
May  9 13:56:53 stein ------------[ cut here ]------------
May  9 13:56:53 stein WARNING: at fs/namei.c:1251 lookup_one_len+0x39/0xc0()
May  9 13:56:53 stein Hardware name: System Product Name
May  9 13:56:53 stein Modules linked in:
May  9 13:56:53 stein Pid: 1, comm: swapper Not tainted 2.6.30-rc5 #1
May  9 13:56:53 stein Call Trace:
May  9 13:56:53 stein [<ffffffff80231fa6>] ? warn_slowpath_fmt+0xd9/0x10b
May  9 13:56:53 stein [<ffffffff8044b3f7>] ? printk+0x4e/0x57
May  9 13:56:53 stein [<ffffffff802e229e>] ? reiserfs_info+0x71/0x91
May  9 13:56:53 stein [<ffffffff80291617>] ? lookup_one_len+0x39/0xc0
May  9 13:56:53 stein [<ffffffff802f08c5>] ? reiserfs_xattr_init+0x2c/0x8b
May  9 13:56:53 stein [<ffffffff802e11bb>] ? reiserfs_fill_super+0x85f/0x950
May  9 13:56:53 stein [<ffffffff8037ce9b>] ? get_device+0x14/0x1a
May  9 13:56:53 stein [<ffffffff8044d818>] ? __down_write_nested+0x34/0xa9
May  9 13:56:53 stein [<ffffffff8028ad68>] ? sget+0x3d8/0x3ea
May  9 13:56:53 stein [<ffffffff8028b479>] ? get_sb_bdev+0x11a/0x16b
May  9 13:56:53 stein [<ffffffff802e095c>] ? reiserfs_fill_super+0x0/0x950
May  9 13:56:53 stein [<ffffffff8026f97c>] ? kstrdup+0x2b/0x43
May  9 13:56:53 stein [<ffffffff8028a2be>] ? vfs_kern_mount+0x4f/0x95
May  9 13:56:53 stein [<ffffffff8028a357>] ? do_kern_mount+0x43/0xdc
May  9 13:56:53 stein [<ffffffff8029ebb1>] ? do_mount+0x6fd/0x747
May  9 13:56:53 stein [<ffffffff8029ec85>] ? sys_mount+0x8a/0xd9
May  9 13:56:53 stein [<ffffffff8056ed3a>] ? mount_block_root+0xe3/0x27c
May  9 13:56:53 stein [<ffffffff8056f04e>] ? prepare_namespace+0x12c/0x156
May  9 13:56:53 stein [<ffffffff8056e61e>] ? kernel_init+0x173/0x183
May  9 13:56:53 stein [<ffffffff8020be2a>] ? child_rip+0xa/0x20
May  9 13:56:53 stein [<ffffffff8056e4ab>] ? kernel_init+0x0/0x183
May  9 13:56:53 stein [<ffffffff8020be20>] ? child_rip+0x0/0x20
May  9 13:56:53 stein ---[ end trace 36cf86746b5051e0 ]---
May  9 13:56:53 stein VFS: Mounted root (reiserfs filesystem) readonly on device 8:1.

-- 
Stefan Richter
-=====-==--= -=-= -=--=
http://arcgraph.de/sr/

Re: 2.6.30-rc: reiserfs_delete_xattrs log spam (bug 13122)

[Al Viro]

On Sat, May 09, 2009 at 02:44:25PM +0200, Stefan Richter wrote:
> On  9 May, Stefan Richter wrote:
> > I just switched from 2.6.29.1 to 2.6.30-rc5, and I'm getting the same as 
> > the reporter of http://bugzilla.kernel.org/show_bug.cgi?id=13122 alias
> > http://marc.info/?l=linux-kernel&m=123990989515105, i.e. repetitions of
> >  >>>
> > May  9 14:27:22 stein REISERFS warning (device sda1): jdm-20004 
> > reiserfs_delete_xattrs: Couldn't delete all xattrs (-13)
> > May  9 14:27:22 stein
> > <<<
> > 
> > This happens in irregular intervals, but a lot of times --- presumably 
> > when certain files are accessed.  I'm on a recently set-up Gentoo 
> > machine.  I created its root filesystem using the Gentoo x86-64 minimal 
> > boot CD; my guess is that the kernel of the boot CD had 
> > CONFIG_REISERFS_FS_XATTR enabled.  All kernels which I used since then 
> > have it disabled:  2.6.27-gentoo-r8, 2.6.29.1, now 2.6.30-rc5.
> 
> PS:  Also, the following came up at this (= first) boot of 2.6.30-rc, as
> a warm reboot from 2.6.29.1. I have no idea whether it's connected to
> the issue.

That WARN_ON() is fixed in the patches in VFS queue; should be gone in
-next and I'm going to push a bunch of fixes to Linus today, this one
among them.

The first one you've reported is a different story.  Could you try to
bisect it?

Re: 2.6.30-rc: reiserfs_delete_xattrs log spam (bug 13122)

[Jeff Mahoney]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Al Viro wrote:
> On Sat, May 09, 2009 at 02:44:25PM +0200, Stefan Richter wrote:
>> On  9 May, Stefan Richter wrote:
>>> I just switched from 2.6.29.1 to 2.6.30-rc5, and I'm getting the same as 
>>> the reporter of http://bugzilla.kernel.org/show_bug.cgi?id=13122 alias
>>> http://marc.info/?l=linux-kernel&m=123990989515105, i.e. repetitions of
>>>  >>>
>>> May  9 14:27:22 stein REISERFS warning (device sda1): jdm-20004 
>>> reiserfs_delete_xattrs: Couldn't delete all xattrs (-13)
>>> May  9 14:27:22 stein
>>> <<<
>>>
>>> This happens in irregular intervals, but a lot of times --- presumably 
>>> when certain files are accessed.  I'm on a recently set-up Gentoo 
>>> machine.  I created its root filesystem using the Gentoo x86-64 minimal 
>>> boot CD; my guess is that the kernel of the boot CD had 
>>> CONFIG_REISERFS_FS_XATTR enabled.  All kernels which I used since then 
>>> have it disabled:  2.6.27-gentoo-r8, 2.6.29.1, now 2.6.30-rc5.
>> PS:  Also, the following came up at this (= first) boot of 2.6.30-rc, as
>> a warm reboot from 2.6.29.1. I have no idea whether it's connected to
>> the issue.
> 
> That WARN_ON() is fixed in the patches in VFS queue; should be gone in
> -next and I'm going to push a bunch of fixes to Linus today, this one
> among them.
> 
> The first one you've reported is a different story.  Could you try to
> bisect it?

I'm looking into this right now. I have an idea, but I haven't been able
to reproduce it yet to test.

- -Jeff

- --
Jeff Mahoney
SUSE Labs
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.9 (GNU/Linux)
Comment: Using GnuPG with SUSE - http://enigmail.mozdev.org

iEYEARECAAYFAkoFjr0ACgkQLPWxlyuTD7LiowCgpZPFraYS+c7Y38LtNcVJaHiH
kMwAoIN8X7g37TyLs/7VskKScS+cap7G
=A7Xx
-----END PGP SIGNATURE-----

Re: 2.6.30-rc: reiserfs_delete_xattrs log spam (bug 13122)

[Stefan Richter]

On  9 May, Jeff Mahoney wrote:
> Al Viro wrote:
>>> On  9 May, Stefan Richter wrote:
>>>> May  9 14:27:22 stein REISERFS warning (device sda1): jdm-20004 
>>>> reiserfs_delete_xattrs: Couldn't delete all xattrs (-13)
>>>> May  9 14:27:22 stein
...
>> Could you try to bisect it?

I bisected it with a series of git-format-patch output on top of
2.6.29.3.  The last good one is "reiserfs: simplify xattr internal file
lookups/opens", the first bad one is "reiserfs: eliminate per-super xattr lock".
(Doh, what a surprise; grep'ing for the message in the diffs would have
pointed to this one too... Well, the bisection homed in quickly though.)
Of course that's only the patch which caused the symptoms to appear, not
necessarily the one which introduced the bug.

My config contains:
CONFIG_REISERFS_FS=y
# CONFIG_REISERFS_CHECK is not set
# CONFIG_REISERFS_PROC_INFO is not set
# CONFIG_REISERFS_FS_XATTR is not set

(As mentioned, the filesystem was created and initially populated with a
kernel which was differently configured; I don't have an immediate idea
where to obtain the config.)

The tested series:

0001-reiserfs-add-support-for-mount-count-incrementing.patch
0002-reiserfs-audit-transaction-ids-to-always-be-unsigne.patch
0003-reiserfs-use-buffer_info-for-leaf_paste_entries.patch
0004-reiserfs-use-more-consistent-printk-formatting.patch
0005-reiserfs-make-some-warnings-informational.patch
0006-reiserfs-rework-reiserfs_warning.patch
0007-reiserfs-prepare_error_buf-wrongly-consumes-va_arg.patch
0008-reiserfs-eliminate-reiserfs_warning-from-uniqueness.patch
0009-reiserfs-add-locking-around-error-buffer.patch
0010-reiserfs-rework-reiserfs_panic.patch
0011-reiserfs-rearrange-journal-abort.patch
0012-reiserfs-introduce-reiserfs_error.patch
0013-reiserfs-use-reiserfs_error.patch
0014-reiserfs-small-variable-cleanup.patch
0015-reiserfs-xattr-reiserfs_get_page-takes-offset-inste.patch
0016-reiserfs-remove-link-detection-code.patch
0017-reiserfs-remove-IS_PRIVATE-helpers.patch
0018-reiserfs-Clean-up-xattrs-when-REISERFS_FS_XATTR-is.patch
0019-reiserfs-simplify-xattr-internal-file-lookups-opens.patch
0020-reiserfs-eliminate-per-super-xattr-lock.patch
0021-reiserfs-make-per-inode-xattr-locking-more-fine-gra.patch
# 0021 doesn't build, I skipped it.
0022-reiserfs-remove-i_has_xattr_dir.patch
0023-reiserfs-use-generic-xattr-handlers.patch
0024-reiserfs-journaled-xattrs.patch
# 0024 and later build only with A. Bergalov's fix folded in.
0025-reiserfs-use-generic-readdir-for-operations-across.patch
0026-reiserfs-add-atomic-addition-of-selinux-attributes.patch
0027-reiserfs-factor-out-buffer_info-initialization.patch
0028-reiserfs-cleanup-path-functions.patch
0029-reiserfs-strip-trailing-whitespace.patch
0030-reiserfs-rename-p_s_sb-to-sb.patch
0031-reiserfs-rename-p_s_bh-to-bh.patch
0032-reiserfs-rename-p_s_inode-to-inode.patch
0033-reiserfs-rename-p_s_tb-to-tb.patch
0034-reiserfs-rename-p_._-variables.patch
0035-reiserfs-rename-cn-_-variables.patch

> I'm looking into this right now. I have an idea, but I haven't been able
> to reproduce it yet to test.

Thanks, I can test something for you.
-- 
Stefan Richter
-=====-==--= -=-= -=--=
http://arcgraph.de/sr/

Re: 2.6.30-rc: reiserfs_delete_xattrs log spam (bug 13122)

[Jeff Mahoney]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Al Viro wrote:
> On Sat, May 09, 2009 at 02:44:25PM +0200, Stefan Richter wrote:
>> On  9 May, Stefan Richter wrote:
>>> I just switched from 2.6.29.1 to 2.6.30-rc5, and I'm getting the same as 
>>> the reporter of http://bugzilla.kernel.org/show_bug.cgi?id=13122 alias
>>> http://marc.info/?l=linux-kernel&m=123990989515105, i.e. repetitions of
>>>  >>>
>>> May  9 14:27:22 stein REISERFS warning (device sda1): jdm-20004 
>>> reiserfs_delete_xattrs: Couldn't delete all xattrs (-13)
>>> May  9 14:27:22 stein
>>> <<<
>>>
>>> This happens in irregular intervals, but a lot of times --- presumably 
>>> when certain files are accessed.  I'm on a recently set-up Gentoo 
>>> machine.  I created its root filesystem using the Gentoo x86-64 minimal 
>>> boot CD; my guess is that the kernel of the boot CD had 
>>> CONFIG_REISERFS_FS_XATTR enabled.  All kernels which I used since then 
>>> have it disabled:  2.6.27-gentoo-r8, 2.6.29.1, now 2.6.30-rc5.
>> PS:  Also, the following came up at this (= first) boot of 2.6.30-rc, as
>> a warm reboot from 2.6.29.1. I have no idea whether it's connected to
>> the issue.
> 
> That WARN_ON() is fixed in the patches in VFS queue; should be gone in
> -next and I'm going to push a bunch of fixes to Linus today, this one
> among them.
> 
> The first one you've reported is a different story.  Could you try to
> bisect it?

Ok, I've looked into it some more.

The issue is that with xattrs disabled, reiserfs_permission isn't built.
The ACL part of it doesn't matter, but the IS_PRIVATE check does. What's
happening is that with generic_permission, a regular user doesn't have
permission to traverse the privroot. Thus, the -EACCES.

I'll post a patch this afternoon.

- -Jeff

- --
Jeff Mahoney
SUSE Labs
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.9 (GNU/Linux)
Comment: Using GnuPG with SUSE - http://enigmail.mozdev.org

iEYEARECAAYFAkoFo6QACgkQLPWxlyuTD7I1gwCfbPa4ipt7FB/gq2Xxsrxkq/K6
YggAmwf5eTyGGVwNeUmqp8yPNGc2mFbb
=Gobs
-----END PGP SIGNATURE-----