Re: [PATCH] e2fsprogs : Add stricter checks for blocksize in ext2fs_open

[Theodore Tso <tytso@mit.edu>]

On Fri, Jul 11, 2008 at 02:19:06PM +0530, Manish Katiyar wrote:
> Below patch adds stricter checks in ext2fs_open() so that we catch bad
> block sizes earlier than later.

That concept seems fine; I'm curious why you found this necessary?
Did you have a corrupted filesystem where this caused major problems?
If so, can I have more details?

>  	fs->blocksize = EXT2_BLOCK_SIZE(fs->super);
> -	if (fs->blocksize == 0) {
> +	if ((fs->blocksize < EXT2_MIN_BLOCK_SIZE) ||
> +	    (fs->blocksize > EXT2_MAX_BLOCK_SIZE) ||
> +	    (fs->blocksize % EXT2_MIN_BLOCK_SIZE != 0)) {

The first and last check is not necessary, given that EXT2_bLOCK_SIZE
is defined as:

#define EXT2_BLOCK_SIZE(s)	 (EXT2_MIN_BLOCK_SIZE << (s)->s_log_block_size)

So by definition, the blocksize will *always* be greater than or equal
to MIN_BLOCK_SIZE, and it always will be a multiple of EXT2_MIN_BLOCK_SIZE.

The more direct check which we could do would be something like this:

	if ((fs->super->s_log_block_size < EXT2_MIN_BLOCK_LOG_SIZE) ||
	    (fs->super->s_log_block_size > EXT2_MAX_BLOCK_LOG_SIZE))
  		retval = EXT2_ET_CORRUPT_SUPERBLOCK;
  		goto cleanup;
  	}

... before setting fs->blocksize.

I'm curious what problem you were worried about that might happen if
fs->blocksize were greater than 64k, though.

							- Ted