Re: [bug] ata subsystem related crash with latest -git

[Benny Halevy <bhalevy@panasas.com>]

On Oct. 18, 2007, 15:32 +0200, Jens Axboe <jens.axboe@oracle.com> wrote:
>  static inline struct scatterlist *sg_next(struct scatterlist *sg)
>  {
> -	sg++;
> -
> -	if (unlikely(sg_is_chain(sg)))
> +#ifdef CONFIG_DEBUG_SG
> +	BUG_ON(sg->sg_magic != SG_MAGIC);
> +#endif
> +	if (sg_is_last(sg))
> +		sg = NULL;
> +	else if (sg_is_chain(sg))
>  		sg = sg_chain_ptr(sg);
> +	else
> +		sg++;
>  

Jens, again, please correct me if I'm wrong, but when sg points at the
entry right before a chain entry this implementation of sg_next will
return a pointer to the chain entry here, which I believe it must not.

>  	return sg;
>  }
> 

here's how I think sg_next should be implemented:

  */
 static inline struct scatterlist *sg_next(struct scatterlist *sg)
 {
+#ifdef CONFIG_DEBUG_SG
+	BUG_ON(sg->sg_magic != SG_MAGIC);
+#endif
+	if (sg_is_last(sg))
+		return NULL;
+
 	sg++;
 
 	if (unlikely(sg_is_chain(sg)))
 		sg = sg_chain_ptr(sg);
 
 	return sg;
 }