[Ocfs2-devel] [PATCH linux-next v2] ocfs2: remove kfree() redundant null checks

[Andrew Morton]

On Wed, 13 Feb 2013 13:29:45 -0700
Tim Gardner <tim.gardner at canonical.com> wrote:

> smatch analysis indicates a number of redundant NULL checks before
> calling kfree(), e.g.,
> 
> fs/ocfs2/alloc.c:6138 ocfs2_begin_truncate_log_recovery() info:
>  redundant null check on *tl_copy calling kfree()
> 
> fs/ocfs2/alloc.c:6755 ocfs2_zero_range_for_truncate() info:
>  redundant null check on pages calling kfree()
> 
> ...
>
> diff --git a/fs/ocfs2/alloc.c b/fs/ocfs2/alloc.c
> index 31b9463..83a1a1d 100644
> --- a/fs/ocfs2/alloc.c
> +++ b/fs/ocfs2/alloc.c
> @@ -6134,7 +6134,7 @@ bail:
>  		iput(tl_inode);
>  	brelse(tl_bh);
>  
> -	if (status < 0 && (*tl_copy)) {
> +	if (status < 0) {
>  		kfree(*tl_copy);
>  		*tl_copy = NULL;
>  		mlog_errno(status);

This change does other things.  For example, if
ocfs2_begin_truncate_log_recovery()'s first "goto bail" is taken, we
will now call mlog_errno(status) twice.

That function is pretty confused about its error recovery and logging.
it needs some maintenance.  I'll omit this hunk of your patch.


> @@ -534,7 +533,7 @@ int ocfs2_begin_local_alloc_recovery(struct ocfs2_super *osb,
>  		mlog_errno(status);
>  
>  bail:
> -	if ((status < 0) && (*alloc_copy)) {
> +	if (status < 0) {
>  		kfree(*alloc_copy);
>  		*alloc_copy = NULL;
>  	}

Similar, but this change lokos OK.