xfs
[Top] [All Lists]

[REVIEW 2/3] - xfs_repair enhancements (lost+found handling)

To: "xfs@xxxxxxxxxxx" <xfs@xxxxxxxxxxx>, xfs-dev <xfs-dev@xxxxxxx>
Subject: [REVIEW 2/3] - xfs_repair enhancements (lost+found handling)
From: "Barry Naujok" <bnaujok@xxxxxxx>
Date: Tue, 05 Jun 2007 11:54:00 +1000
Organization: SGI
Sender: xfs-bounce@xxxxxxxxxxx
User-agent: Opera Mail/9.10 (Win32)
This particular patch is in the sequence for part 3 which will rely on I/O being uniform for each piece of metadata.

It changes the behaviour of lost+found quite substantially in xfs_repair.

Currently, xfs_repair deletes any lost+found directories in phase 3, which will orphan inodes that still reside in lost+found. During phase 6, it's recreated and repopulated with orphaned inodes. When one leaves inodes in lost+found between successive runs, it can cause some confusion as xfs_repair keeps on discovering orphaned inodes.

This change leaves any lost+found directory or inode alone during phases 3/4 (other than checking for normal consistancy). During phase 6, if it finds lost+found in the root inode, it checks that it's a directory and is consistant and remembers it. If it's corrupted, it's junked or repaired like any other directory. If it's not a directory, it's junked. During the last part of phase 6, if any orphaned inodes are found, it will place them in the previously located lost+found directory, or create it and place them.

Attachment: lost+found
Description: Binary data

<Prev in Thread] Current Thread [Next in Thread>
  • [REVIEW 2/3] - xfs_repair enhancements (lost+found handling), Barry Naujok <=