Re: [RFC][PATCH] ChunkFS: fs fission for faster fsck

[Andreas Dilger <adilger@clusterfs.com>]

On Apr 25, 2007  20:54 +1000, David Chinner wrote:
> On Tue, Apr 24, 2007 at 04:53:11PM -0500, Amit Gud wrote:
> > Right now, there is no distinction between an inode and continuation 
> > inode (also referred to as 'cnode' below), except for the 
> > EXT2_IS_CONT_FL flag. Every inode holds a list of static number of 
> > inodes, currently limited to 4.
> > 
> > The structure looks like this:
> > 
> >  ----------		----------
> > | cnode 0  |---------->| cnode 0  |----------> to another cnode or NULL
> >  ----------		----------
> > | cnode 1  |-----      | cnode 1  |-----
> >  ----------	|	----------	|
> > | cnode 2  |-- |      | cnode 2  |--   |
> >  ----------  | |	----------  |   |
> > | cnode 3  | | |      | cnode 3  | |   |
> >  ----------  | |	----------  |   |
> > 	  |  |  |		 |  |   |
> > 
> > 	   inodes		inodes or NULL
> 
> How do you recover if fsfuzzer takes out a cnode in the chain? The
> chunk is marked clean, but clearly corrupted and needs fixing and
> you don't know what it was pointing at.  Hence you have a pointer to
> a trashed cnode *somewhere* that you need to find and fix, and a
> bunch of orphaned cnodes that nobody points to *somewhere else* in
> the filesystem that you have to find. That's a full scan fsck case,
> isn't?

Presumably, the cnodes in the other chunks contain forward and back
references.  Those need to contain at minimum inode + generation + chunk
to avoid problem of pointing to a _different_ inode after such corruption
caused the old inode to be deleted and a new one allocated in its place.

If the cnode in each chunk is more than just a singly-linked list, the
file as a whole could survive multiple chunk corruptions, though there
would now be holes in the file.

> It seems that any sort of damage to the underlying storage (e.g.
> media error, I/O error or user brain explosion) results in the need
> to do a full fsck and hence chunkfs gives you no benefit in this
> case.

There are several cases where such corruption could be found:
- file access from the "parent" cnode will be missing corrupted cnode,
  probably causing a fsck of both the source and target chunks
- a fsck of the source chunk would find the dangling cnode reference
  and cause a fsck of the corrupt chunk
- a fsck of the later cnode chunks would find the dangling cnode reference
  and cause a fsck of the corrupt chunk
- a fsck of the corrupt chunk would find the original corruption

The case where only a fsck of the corrupt chunk is done would not find the
cnode references.  Maybe there needs to be per-chunk info which contains
a list/bitmap of other chunks that have cnodes shared with each chunk?

Cheers, Andreas
--
Andreas Dilger
Principal Software Engineer
Cluster File Systems, Inc.