Re: [PATCH 1/2 v4] e2fsprogs: Fix how we treat user-spcified filesystem size

[Ted Ts'o <tytso@mit.edu>]

On Sun, Jun 05, 2011 at 12:30:44PM +0200, Lukas Czerner wrote:
> We can not specify filesystem size in blocks count without specifying
> blocksize as well. It is because we need blocks count to determine
> filesystem type, and we need filesystem type to determine blocksize. So
> it should not be allowed, however due to compatibility reason it should
> be still possible, so at least print warning message for now, so we can
> easily restrict that later.

What we currently have is actually very well defined.  The fs-size
parameter is either an absolute size, if a suffixed value (i.e., 12k,
64M, 128G, 2T) is specified.  If the value does not have a suffix,
then it is either interpreted as a block count if a blocksize is
specified, or as kilobytes if no blocksize is explicitly specified.

And this is really only a problem with mke2fs; it's not a problem with
resize2fs, since we know what the blocksize is.

> diff --git a/lib/e2p/parse_num.c b/lib/e2p/parse_num.c
> index 83a329a..8e2751b 100644
> --- a/lib/e2p/parse_num.c
> +++ b/lib/e2p/parse_num.c
> @@ -37,7 +37,6 @@ unsigned long long parse_num_blocks2(const char *arg, int log_block_size)
>  		num >>= (1+log_block_size);
>  		break;
>  	case '\0':
> -		break;
>  	default:
>  		return 0;
>  	}

and this doesn't do what the commit description claims.  This change
actually breaks the use of a number that does not have a suffix.

I'd rather not break resize2fs at all.  If script wants to specify a
block count, that's fine.  If we want to add a warning which gets
printed by mke2fs in the case where there is no explicit blockcount
specified by the user and no explicit size suffix to the fs-size
parameter stating that parameter was interpreted as kilobytes, that's
fine.

Basically, how mke2fs parses that field has been the way it has been
for over a decade, and it's really not caused any problems for users.

    	   	       	    	       - Ted