Re: loopback over reiserfs broken in 2.6.1-mm1

[Andy Lutomirski <luto@stanford.edu>]

Ben Slusky wrote:
> On Tue, 13 Jan 2004 22:43:27 -0800, Andy Lutomirski wrote:
> 
>>Didn't help.
>>
>>I put in a few printk's -- it looks like the failure is in 
>>set_blocksize, probably due to a bogus blocksize of 131072, coming from 
>>'lo_blocksize = inode->i_blksize;'.
> 
> 
> Right... and this patch would have fixed it if ReiserFS' i_blksize were
> merely huge (on the order of 8192) instead of frickin' ludicrous (on the
> order of 131072).
> 
> Under the circumstances I think this is the best fix then, it ought
> to revert loop_set_fd's behavior to what it's always been:
> 
> --- drivers/block/loop.c.old	2004-01-11 13:05:05.000000000 -0500
> +++ drivers/block/loop.c	2004-01-14 07:49:24.795161024 -0500
> @@ -853,9 +853,7 @@
>  		blk_queue_merge_bvec(lo->lo_queue, q->merge_bvec_fn);
>  	}
>  
> -	error = set_blocksize(bdev, lo_blocksize);
> -	if (error)
> -		goto out_putf;
> +	set_blocksize(bdev, lo_blocksize);
>  
>  	kernel_thread(loop_thread, lo, CLONE_KERNEL);
>  	down(&lo->lo_sem);
> 

Fixed.  Thanks.

--Andy