[Unionfs] RHEL 2.6.9 hangs with unionfs-1.1.5

Benoit Guillon guillon at thalescomputers.fr
Tue Nov 28 04:40:40 EST 2006


Josef Sipek wrote:

>On Mon, Nov 27, 2006 at 03:09:26PM +0100, Benoit Guillon wrote:
>...
>  
>
>>I had no answer to the previous mail; does it mean that 1.1.5 is not
>>maintained?
>>    
>>
>It is still maintained.
>  
>
Thansks much!

>>Nov 27 14:36:35 node1 kernel: Modules linked in: unionfs(U) i915 nfsd 
>>exportfs lockd nfs_acl sunrpc i2c_dev i2c_core poweroff(U) allocator(U) 
>>dm_mirror dm_mod button battery ac md5 ipv6 joydev uhci_hcd ehci_hcd 
>>hw_random snd_intel8x0 snd_ac97_codec snd_pcm_oss snd_mixer_oss snd_pcm 
>>snd_timer snd_page_alloc snd_mpu401_uart snd_rawmidi snd_seq_device snd 
>>soundcore tg3 8139too mii floppy ext3 jbd ata_piix libata sd_mod scsi_mod
>>Nov 27 14:36:35 node1 kernel: CPU:    0
>>Nov 27 14:36:35 node1 kernel: EIP:    0060:[<e07bf9f6>]    Tainted: 
>>P      VLI
>>    
>>
>
>Can you reproduce the problem without the proprietary module?
>  
>
Ok, i'll try without poweroff and allocator. The unionfs is patched as 
you have seen.

>>Nov 27 14:36:35 node1 kernel: EFLAGS: 00010246   (2.6.9-34.ELsmp)
>>Nov 27 14:36:35 node1 kernel: EIP is at do_get_dentry+0xb56/0xd70 [unionfs]
>>Nov 27 14:36:35 node1 kernel: eax: 00000000   ebx: d71e0ba4   ecx: 
>>00000000   edx: c038ff80
>>Nov 27 14:36:35 node1 kernel: esi: 00000001   edi: 00000000   ebp: 
>>ffffff8c   esp: d5158e30
>>Nov 27 14:36:35 node1 kernel: ds: 007b   es: 007b   ss: 0068
>>Nov 27 14:36:35 node1 kernel: Process nfsd (pid: 2992, 
>>threadinfo=d5158000 task=de844b30)
>>Nov 27 14:36:35 node1 kernel: Stack: e07e86c0 e07e767c 00000233 00000004 
>>e07e89d0 e07f0ae0 e07e767c e07e86c0
>>Nov 27 14:36:35 node1 kernel:        00000233 df0ba474 0008ce37 0008ce1a 
>>00000002 e07be00f 00000002 00000003
>>Nov 27 14:36:35 node1 kernel:        dee58000 ffffff8c 00000000 000002dd 
>>df23a400 00000002 00000012 dee58ff9
>>Nov 27 14:36:35 node1 kernel: Call Trace:
>>Nov 27 14:36:35 node1 kernel:  [<e07be00f>] unionfs_destroy_inode+0xbf/0xd0 [unionfs]
>>Nov 27 14:36:35 node1 kernel:  [<e07c1052>] unionfs_decode_fh+0x132/0x164 [unionfs]
>>    
>>
>
>Ok, this is really odd...Are you using any patches on top of 1.1.5? I don't
>see unionfs_decode_fh anywhere in the 1.1.5 source.
>  
>
Yes, the patches from this thread are applied :

http://www.fsl.cs.sunysb.edu/pipermail/unionfs/2006-June/004557.html

They are applied in this order:

# Compilation fixes for RHEL4 2.6.9 kernel
cat rhel4-build-fix.diff | patch
cat imap-build-fix-2.6.9.diff | patch

# Fix for NFS export (use persistent imap!)
cat nfs-imap-fix.diff | patch

# Fix Makefile to install imap manpage
cat imap-manpage.diff | patch


I attach the used patches. The only differences between the tests done 
by Vincent Vaneckere and the configuration I use is that the remote 
diskless nodes (that NFS mount the unionfs stack) run on a ppc64 Fedora 
Core 4 distribution. I don't know if the architecture can have an impact 
on NFS. I thought that it was architecture independent.

Regards,

-- 
Benoît Guillon                guillon at thalescomputers.fr
TRT/SML                       tel. : 33 (0)4 98 16 33 90
 
THALES RESEARCH & TECHNOLOGY

-------------- next part --------------
A non-text attachment was scrubbed...
Name: unionfs-1.1.5-patches.tar.bz2
Type: application/x-bzip
Size: 4335 bytes
Desc: not available
Url : http://www.fsl.cs.sunysb.edu/pipermail/unionfs/attachments/20061128/646c4b6d/unionfs-1.1.5-patches.tar-0001.bin


More information about the unionfs mailing list