Problems with reiserfs when start OpenAFS 1.2.11-fc1 Segmentation fault with kernel 2.4.22-1.2174.nptl

Problems with reiserfs when start OpenAFS 1.2.11-fc1 Segmentation fault with kernel 2.4.22-1.2174.nptl

[jds]

Hi: Iam compile OpenAFS 1.2.11-fc1 with kernel 2.4.22-1.2174.nptl is OK

problems with reiserfs 3.6 and Fedora Core 1
the problems is when start the client AFS recive the messages:

[root@angelinux modload]# service afs start
Found libafs-2.4.22-1.2174.nptl-i686.o from SymTable... Loading...
Starting AFS services.....
/etc/init.d/afs: line 305:  3300 Segmentation fault      /usr/vice/etc/afsd
${AFSD_OPTIONS}
[root@angelinux modload]# lsmod
Module                  Size  Used by    Tainted: PF
libafs-2.4.22-1.2174.nptl-i686  483360   0  (unused)
parport_pc             18756   1  (autoclean)
etc,etc

In log messages:
Starting AFS cache scan...Can't open inode 135536
Unable to handle kernel paging request at virtual address ffffffff
printing eip:
f093e1b0
*pde = 00002067
*pte = 00000000
Oops: 0002
libafs-2.4.22-1.2174.nptl-i686 parport_pc lp parport autofs tg3 floppy sg
keybdev hid ehci-hcd usb-uhci usbcore mousedev input i830 agpgart reiserfs
ata_piix
CPU:    0
EIP:    0060:[<f093e1b0>]    Tainted: PF
EFLAGS: 00010282

EIP is at osi_Panic [libafs-2.4.22-1.2174.nptl-i686] 0x20
(2.4.22-1.2174.nptl)
eax: 00000018   ebx: e7bd2580   ecx: 00000001   edx: e80d6000
esi: ea666b80   edi: ea666c08   ebp: 00000400   esp: e80d7c14
ds: 0068   es: 0068   ss: 0068
Process afsd (pid: 3300, stackpage=e80d7000)
Stack: f095c03c 00021170 c0341640 c0341640 ea666b80 ea666c08 c19b4800
f0948130
     f095c03c 00021170 c0341640 c0341640 ef724c10 00000000 00000000
00000000
     f0b5b000 00000000 00000001 f090e52d 00021170 00000000 00000ce4
00000ce4
Call Trace:   [<f095c03c>] .rodata.str1.1 [libafs-2.4.22-1.2174.nptl-i686]
0xe0c (0xe80d7c14)
[<f0948130>] osi_UFSOpen [libafs-2.4.22-1.2174.nptl-i686] 0x120 (0xe80d7c30)
[<f095c03c>] .rodata.str1.1 [libafs-2.4.22-1.2174.nptl-i686] 0xe0c
(0xe80d7c34)
[<f090e52d>] afs_InitCacheFile [libafs-2.4.22-1.2174.nptl-i686] 0x10d
(0xe80d7c60)
[<f006d260>] reiserfs_file_operations [reiserfs] 0x0 (0xe80d7c7c)
[<f006d260>] reiserfs_file_operations [reiserfs] 0x0 (0xe80d7c80)
[<f0910efb>] afs_InitVolumeInfo [libafs-2.4.22-1.2174.nptl-i686] 0x5b
(0xe80d7c90)
[<f094c7de>] afs_syscall_call [libafs-2.4.22-1.2174.nptl-i686] 0x5ce
(0xe80d7cb0)
[<c013a94b>] activate_page [kernel] 0xab (0xe80d7ce4)
[<c0146189>] getblk [kernel] 0x59 (0xe80d7d00)
[<c0146460>] bread [kernel] 0x20 (0xe80d7d24)
[<f0054b74>] is_tree_node [reiserfs] 0x74 (0xe80d7d40)
[<f0055131>] search_by_key [reiserfs] 0x5b1 (0xe80d7d54)
[<f005e2b7>] journal_end [reiserfs] 0x27 (0xe80d7d94)
[<f004ca3e>] reiserfs_dirty_inode [reiserfs] 0x7e (0xe80d7da
[<f00405a8>] search_by_entry_key [reiserfs] 0xc8 (0xe80d7dd4)
[<f0054861>] pathrelse [reiserfs] 0x21 (0xe80d7df4)
[<f0048077>] reiserfs_readdir [reiserfs] 0x3c7 (0xe80d7e04)
[<c013c2f4>] __alloc_pages [kernel] 0x64 (0xe80d7e44)
[<c013c2f4>] __alloc_pages [kernel] 0x64 (0xe80d7e68)
[<c0131013>] vm_enough_memory [kernel] 0x33 (0xe80d7e84)
[<c026762f>] rb_insert_color [kernel] 0x8f (0xe80d7ea0)
[<c013043d>] do_wp_page [kernel] 0x4d (0xe80d7ebc)
[<c0130dd9>] handle_mm_fault [kernel] 0xf9 (0xe80d7ee0)
[<c0117bf8>] do_page_fault [kernel] 0x138 (0xe80d7f0c)
[<f094cec0>] afs_syscall [libafs-2.4.22-1.2174.nptl-i686] 0x190 (0xe80d7f40)
[<c0117ac0>] do_page_fault [kernel] 0x0 (0xe80d7fb0)
[<c0109747>] system_call [kernel] 0x33 (0xe80d7fc0)

Code: c6 05 ff ff ff ff 2a 83 c4 1c c3 90    74 26 00 b8 1e bd 95

with kernel 2.4.22-2166 and is the same version OpenAFS working good and
startup client AFS good.

Helpme please.
------- End of Forwarded Message -------

Re: Problems with reiserfs when start OpenAFS 1.2.11-fc1 Segmentation fault with kernel 2.4.22-1.2174.nptl

[Ryan Underwood]

[-- Attachment #1: Type: text/plain, Size: 1025 bytes --]


On Wed, Mar 03, 2004 at 07:00:06PM -0600, jds wrote:
> Hi: Iam compile OpenAFS 1.2.11-fc1 with kernel 2.4.22-1.2174.nptl is OK
> 
> problems with reiserfs 3.6 and Fedora Core 1
> the problems is when start the client AFS recive the messages:
> 
> [root@angelinux modload]# service afs start
> Found libafs-2.4.22-1.2174.nptl-i686.o from SymTable... Loading...
> Starting AFS services.....
> /etc/init.d/afs: line 305:  3300 Segmentation fault      /usr/vice/etc/afsd
> ${AFSD_OPTIONS}

You can't use the OpenAFS client on a ReiserFS filesystem.  It currently
only supports ext2/ext3 filesystems.  If you only have reiser
filesystems, just make a big empty file:
# dd if=/dev/zero of=/var/afscache
and make an ext2 filesystem on it:
# mke2fs -F /var/afscache
and loop-mount it wherever your afs cache needs to be.
# mount -o loop,rw /var/afscache /usr/vice/cache

You can automate this in your fstab:
/var/afscache /usr/vice/cache ext2 defaults,loop 0 0

-- 
Ryan Underwood, <nemesis@icequake.net>

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: Problems with reiserfs when start OpenAFS 1.2.11-fc1 Segmentation fault with kernel 2.4.22-1.2174.nptl

[jds]

Hi Ryan:

   Thanks for your help, the client openafs working again.

loop: loaded (max 8 devices)
Journalled Block Device driver loaded
kjournald starting.  Commit interval 5 seconds
EXT3 FS 2.4-0.9.19, 19 August 2002 on loop(7,0), internal journal
EXT3-fs: mounted filesystem with ordered data mode.
Found sys_call_table at c033d9f0
Starting AFS cache scan...found 0 non-empty cache files (0%).

[root@angelinux root]# df -T
Filesystem    Type   1K-blocks      Used Available Use% Mounted on
/dev/hda2 reiserfs    12947988   2838624  10109364  22% /
/dev/hda3 reiserfs    10241116     33320  10207796   1% /home
none         tmpfs      382776         0    382776   0% /dev/shm
/var/afscache ext3       92659      4722     83153   6% /usr/vice/cache
AFS            afs     9000000         0   9000000   0% /afs

Regards.

Thanks at all people for help me ( Derek, Pete, Derrick )


---------- Original Message -----------
From: Ryan Underwood <nemesis-lists@icequake.net>
To: reiserfs-list@namesys.com
Sent: Wed, 3 Mar 2004 18:28:04 -0600
Subject: Re: Problems with reiserfs when start OpenAFS 1.2.11-fc1 Segmentation
fault with kernel 2.4.22-1.2174.nptl

> On Wed, Mar 03, 2004 at 07:00:06PM -0600, jds wrote:
> > Hi: Iam compile OpenAFS 1.2.11-fc1 with kernel 2.4.22-1.2174.nptl is OK
> > 
> > problems with reiserfs 3.6 and Fedora Core 1
> > the problems is when start the client AFS recive the messages:
> > 
> > [root@angelinux modload]# service afs start
> > Found libafs-2.4.22-1.2174.nptl-i686.o from SymTable... Loading...
> > Starting AFS services.....
> > /etc/init.d/afs: line 305:  3300 Segmentation fault      /usr/vice/etc/afsd
> > ${AFSD_OPTIONS}
> 
> You can't use the OpenAFS client on a ReiserFS filesystem.  It currently
> only supports ext2/ext3 filesystems.  If you only have reiser
> filesystems, just make a big empty file:
> # dd if=/dev/zero of=/var/afscache
> and make an ext2 filesystem on it:
> # mke2fs -F /var/afscache
> and loop-mount it wherever your afs cache needs to be.
> # mount -o loop,rw /var/afscache /usr/vice/cache
> 
> You can automate this in your fstab:
> /var/afscache /usr/vice/cache ext2 defaults,loop 0 0
> 
> -- 
> Ryan Underwood, <nemesis@icequake.net>
------- End of Original Message -------

Re: Problems with reiserfs when start OpenAFS 1.2.11-fc1 Segmentation fault with kernel 2.4.22-1.2174.nptl

[Mike Fedyk]

jds wrote:
> Hi Ryan:
> 
>    Thanks for your help, the client openafs working again.
> 
> loop: loaded (max 8 devices)
> Journalled Block Device driver loaded
> kjournald starting.  Commit interval 5 seconds
> EXT3 FS 2.4-0.9.19, 19 August 2002 on loop(7,0), internal journal
> EXT3-fs: mounted filesystem with ordered data mode.

No, if you're going to be stacking filesystems, use ext2, not ext3.  You 
will get better performance, and there's no use in having two levels of 
journaling in this case.

Mike

Re: Problems with reiserfs when start OpenAFS 1.2.11-fc1 Segmentation fault with kernel 2.4.22-1.2174.nptl

[Ryan Underwood]

[-- Attachment #1: Type: text/plain, Size: 826 bytes --]

On Wed, Mar 03, 2004 at 07:46:48PM -0800, Mike Fedyk wrote:
> jds wrote:
> >Hi Ryan:
> >
> >   Thanks for your help, the client openafs working again.
> >
> >loop: loaded (max 8 devices)
> >Journalled Block Device driver loaded
> >kjournald starting.  Commit interval 5 seconds
> >EXT3 FS 2.4-0.9.19, 19 August 2002 on loop(7,0), internal journal
> >EXT3-fs: mounted filesystem with ordered data mode.
> 
> No, if you're going to be stacking filesystems, use ext2, not ext3.  You 
> will get better performance, and there's no use in having two levels of 
> journaling in this case.

You might also think about bigger than 100MB cache file.  A big cache
helps out AFS client performance tremendously.  But definitely use ext2
and not ext3 as others have said.

-- 
Ryan Underwood, <nemesis@icequake.net>

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: [OpenAFS] Re: Problems with reiserfs when start OpenAFS 1.2.11-fc1 Segmentation fault with kernel 2.4.22-1.2174.nptl

[Dan Pritts]

On Thu, Mar 04, 2004 at 01:18:46AM -0600, Ryan Underwood wrote:
> > No, if you're going to be stacking filesystems, use ext2, not ext3.  You 
> > will get better performance, and there's no use in having two levels of 
> > journaling in this case.
> 
> You might also think about bigger than 100MB cache file.  A big cache
> helps out AFS client performance tremendously.  But definitely use ext2
> and not ext3 as others have said.

I realize it won't work out of the box, but has anyone given any
thought to putting the cache in tmpfs?

Seems like this might be the best of both worlds - when the system is low
on RAM it can use swap, but when you've got plenty of RAM you avoid the
overhead of writing the cache files to disk.

more flexible than memcache, faster than disk cache.  My understanding
(limited) of the tmpfs on linux suggests that it might be an easy change
to make in the code for someone who understood the data structions better
than i do ;)

mostly, just curious...can anyone say?

danno
--
dan pritts                                       danno@internet2.edu
systems administrator                            734/352-4953 office
internet2                                        734/834-7224 mobile

Re: [OpenAFS] Re: Problems with reiserfs when start OpenAFS 1.2.11-fc1 Segmentation fault with kernel 2.4.22-1.2174.nptl

[Ryan Underwood]

[-- Attachment #1: Type: text/plain, Size: 1514 bytes --]

On Thu, Mar 04, 2004 at 06:07:25PM -0500, Dan Pritts wrote:
> > 
> > You might also think about bigger than 100MB cache file.  A big cache
> > helps out AFS client performance tremendously.  But definitely use ext2
> > and not ext3 as others have said.
> 
> I realize it won't work out of the box, but has anyone given any
> thought to putting the cache in tmpfs?
> 
> Seems like this might be the best of both worlds - when the system is low
> on RAM it can use swap, but when you've got plenty of RAM you avoid the
> overhead of writing the cache files to disk.

I thought about it, and indeed it would be beneficial when running a
server with no writeable medium.  But there was some discussion on the
OpenAFS list a while back, and it would take some work to make the cache
manager work with tmpfs.  Basically it's the same problem with tmpfs as
with reiser, xfs, etc as a cache; the cache manager was developer rather
closely coupled to ext2, and other filesystems just don't support the
operations that the cache manager currently needs.

I'm not sure if there are any plans to fix that at the moment, perhaps
bigger fish to fry, such as getting some real AFS functionality going
for 2.6 users.  For 2.6, we need a per-process credentials API as well
as a way for the OpenAFS module to install its hooks.  I haven't seen 
anything happening recently except for the very feature-limited kAFS
client (no authentication, no cache, etc).

-- 
Ryan Underwood, <nemesis@icequake.net>

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 189 bytes --]