Re: [PATCH 1/3] freevxfs: fix buffer_head leak

[Pekka Enberg <penberg@cs.helsinki.fi>]

On Tue, 2005-06-28 at 16:28 -0700, Andrew Morton wrote:
> But your change means that we'll always perform that kmalloc, even if the
> buffer came back !buffer_mapped().

I think I originally dropped that buffer_mapped() check (which doesn't
make sense to me) which is why I shuffled the code around.

On Tue, 2005-06-28 at 16:28 -0700, Andrew Morton wrote:
> Something like this?

Works for me. Thanks.

> diff -puN fs/freevxfs/vxfs_fshead.c~freevxfs-fix-buffer_head-leak fs/freevxfs/vxfs_fshead.c
> --- 25/fs/freevxfs/vxfs_fshead.c~freevxfs-fix-buffer_head-leak	Tue Jun 28 16:19:53 2005
> +++ 25-akpm/fs/freevxfs/vxfs_fshead.c	Tue Jun 28 16:24:53 2005
> @@ -78,14 +78,16 @@ vxfs_getfsh(struct inode *ip, int which)
>  	struct buffer_head		*bp;
>  
>  	bp = vxfs_bread(ip, which);
> -	if (buffer_mapped(bp)) {
> +	if (bp && buffer_mapped(bp)) {
>  		struct vxfs_fsh		*fhp;
>  
> -		if (!(fhp = kmalloc(sizeof(*fhp), SLAB_KERNEL)))
> +		if (!(fhp = kmalloc(sizeof(*fhp), GFP_KERNEL))) {
> +			put_bh(bp);
>  			return NULL;
> +		}
>  		memcpy(fhp, bp->b_data, sizeof(*fhp));
>  
> -		brelse(bp);
> +		put_bh(bp);
>  		return (fhp);
>  	}
>  
> _
>