Thanks Vladimir,

Fantastic, I can now get all my data back! Your help is more appreciated than you 
know. This has saved me a lot of lost photos, work, emails etc which I foolishly never 
regularly backed up..

Thanks & Best Regards

Phil Lewis


Vladimir V. Saveliev wrote:

>Hi
>
>On Friday, October 12, 2001 03:42:32 PM +0400 "Vladimir V. Saveliev"
><[EMAIL PROTECTED]> wrote:
>
>>Hi, Chris
>>
>>I was looking over Philip J Lewis <[EMAIL PROTECTED]>'s
>>problem.
>>He was not able to mount clean filesystem.
>>
>>Oct 11 03:28:56 dual kernel: reiserfs: checking transaction log
>>
>(device
>
>>22:02) ...
>>Oct 11 03:28:57 dual kernel: journal-1204: REPLAY FAILURE fsck
>>
>required!
>
>>Trying to replay onto a log block
>>Oct 11 03:28:57 dual kernel: Replay Failure, unable to mount
>>Oct 11 03:28:57 dual kernel: reiserfs_read_super: unable to initialize
>>
>
>>journal space
>>
>>
>>The problem is that is_valid_transaction found a transaction which
>>actually is not a transaction.
>>
>
>>Journal header says that:
>>j_last_flush_trans_id 221337
>>j_first_unflushed_offset 1791
>>j_mount_id 165
>>
>>journal_read looked at offset 1791 in journal (block 1809 of a device)
>>
>
>>and found there a descriptor block:
>>Desc block 1809 (j_trans_id 10551298, j_mount_id 165, j_len 200)
>>then it looked at block 2010 in hope to see there a commit.
>>Desc block 2010 (j_trans_id 10551298, j_mount_id 165, j_len 200)
>>As commit block does not have its own signature and j_trans_id and
>>
>j_len
>
>>match - it thought that it found a valid transaction.
>>
>
>    Ouch.  I think that all we can do in journal replay is check to see
>if the
>    digest part of the commit block is filled with zeros.  I'll make a
>patch
>    that puts a magic string there as well.  Do we have a copy of block
>    number 2010 (the description block) I could look at?
>
>Yes. Here are blocks 16 and 2010 of the problem device.
>Note, that 2010 is internal node which has descriptor magic. So, you
>might want to edit print_block to get desired look at it.
>
>Phil, while the proper fix is not ready - you may want to work around
>the problem with
>dd if=/dev/zero of=/dev/problem-device bs=4096 count=1 conv=notrunc
>seek=1809
>That will wipe out the "transaction" which causes problem for journal
>replay code. So, after that you should be able to mount that filesystem.
>
>Thanks,
>vs
>
>




Reply via email to