[PATCH] libata: don't initialize sg in ata_exec_internal() if DMA_NONE

[Jeff Garzik]

Tejun Heo wrote:
> Calling sg_init_one() with NULL buf causes oops on certain
> configurations.  Don't initialize sg in ata_exec_internal() if
> DMA_NONE and make the function complain if @buf is NULL when dma_dir
> isn't DMA_NONE.  While at it, fix comment.
> 
> The problem is discovered and initial patch was submitted by Arnd
> Bergmann.
> 
> Signed-off-by: Tejun Heo <htejun at gmail.com>
> Cc: Arnd Bergmann <arnd.bergmann at de.ibm.com>
> ---
> 
> Hello, Arnd Bergmann.
> 
> Thanks for spotting and fixing this but ata_exec_internal_nodma() is
> almost identical to ata_do_simple_cmd() and ata_exec_internal() itself
> needs fixing anyway.  This patch just fixes ata_exec_internal().  I'll
> follow up with conversion to ata_do_simple_cmd().
> 
> Thanks.
> 
> diff --git a/drivers/ata/libata-core.c b/drivers/ata/libata-core.c
> index 011c0a8..70e02e9 100644
> --- a/drivers/ata/libata-core.c
> +++ b/drivers/ata/libata-core.c
> @@ -1332,7 +1332,7 @@ unsigned ata_exec_internal_sg(struct ata_device *dev,
>  }
>  
>  /**
> - *	ata_exec_internal_sg - execute libata internal command
> + *	ata_exec_internal - execute libata internal command
>   *	@dev: Device to which the command is sent
>   *	@tf: Taskfile registers for the command and the result
>   *	@cdb: CDB for packet command
> @@ -1354,10 +1354,15 @@ unsigned ata_exec_internal(struct ata_device *dev,
>  			   int dma_dir, void *buf, unsigned int buflen)
>  {
>  	struct scatterlist sg;
> +	unsigned int n_elem = 0;
>  
> -	sg_init_one(&sg, buf, buflen);
> +	if (dma_dir != DMA_NONE) {
> +		WARN_ON(!buf);
> +		sg_init_one(&sg, buf, buflen);
> +		n_elem++;
> +	}
>  
> -	return ata_exec_internal_sg(dev, tf, cdb, dma_dir, &sg, 1);
> +	return ata_exec_internal_sg(dev, tf, cdb, dma_dir, &sg, n_elem);

ACK, if you conditionally replace "&sg" with NULL.  That's the safer 
choice, as it guarantees (via an oops) that the user will not be 
touching sg, if dma_dir==DMA_NONE.

	Jeff