Re: [PATCH scsi] Add NULL checking of return value from scsi_cmd_to_driver()

[James Bottomley <James.Bottomley@HansenPartnership.com>]

On Mon, 2012-09-24 at 17:25 +0800, Li Zhong wrote:
> On Mon, 2012-09-24 at 11:35 +0400, James Bottomley wrote:
> > On Mon, 2012-09-24 at 15:03 +0800, Li Zhong wrote:
> > > On Mon, 2012-09-24 at 09:44 +0400, James Bottomley wrote:
> > > > On Mon, 2012-09-24 at 13:30 +0800, Li Zhong wrote:
> > > > > Just noticed that after commit 919f797, it is possible that 
> > > > > scsi_cmd_to_driver() returns NULL. This patch adds the NULL checking for drv 
> > > > > returned from the above function. 
> > > > > 
> > > > > Maybe it is not possible at run time, but from the code itself, we'd better
> > > > > have this check?  
> > > > 
> > > > There's not much point having a check that never trips, unless it's an
> > > > assert, in which case a NULL deref does that.  All it does is add
> > > > pointless instructions to the critical path.  only REQ_TYPE_BLOCK_PC
> > > > commands can be submitted without a driver, so the check above would
> > > > seem to preclude that.
> > > 
> > > Hi James, 
> > > 
> > > Thank you, it sounds reasonable to me. Let's drop it. 
> > 
> > Well, there is another thing you might do: The path length of
> > scsi_cmd_to_driver() increased a lot thanks to 18a4d0a22ed6 it might be
> > worth getting it back to what it was (this looks to be doable with the
> > same != REQ_TYPE_BLOCK_PC test in the error handler.  Plus, I think it
> > fixes a bug where you get different behaviours from REQ_TYPE_BLOCK_PC
> > commands when a driver is and isn't attached (I've cc'd Martin to see
> > what he thinks).
> 
> Hi James, 
> 
> Do you mean something like this: 

Pretty much, yes.

James

> =======
> 
> diff --git a/drivers/scsi/scsi_error.c b/drivers/scsi/scsi_error.c
> index de2337f..c1b05a8 100644
> --- a/drivers/scsi/scsi_error.c
> +++ b/drivers/scsi/scsi_error.c
> @@ -789,7 +789,6 @@ static int scsi_send_eh_cmnd(struct scsi_cmnd *scmd, unsigned char *cmnd,
>  			     int cmnd_size, int timeout, unsigned sense_bytes)
>  {
>  	struct scsi_device *sdev = scmd->device;
> -	struct scsi_driver *sdrv = scsi_cmd_to_driver(scmd);
>  	struct Scsi_Host *shost = sdev->host;
>  	DECLARE_COMPLETION_ONSTACK(done);
>  	unsigned long timeleft;
> @@ -845,8 +844,11 @@ static int scsi_send_eh_cmnd(struct scsi_cmnd *scmd, unsigned char *cmnd,
>  
>  	scsi_eh_restore_cmnd(scmd, &ses);
>  
> -	if (sdrv && sdrv->eh_action)
> -		rtn = sdrv->eh_action(scmd, cmnd, cmnd_size, rtn);
> +	if (scmd->request->cmd_type != REQ_TYPE_BLOCK_PC) {
> +		struct scsi_driver *sdrv = scsi_cmd_to_driver(scmd);
> +		if (sdrv->eh_action)
> +			rtn = sdrv->eh_action(scmd, cmnd, cmnd_size, rtn);
> +	}
>  
>  	return rtn;
>  }
> diff --git a/include/scsi/scsi_cmnd.h b/include/scsi/scsi_cmnd.h
> index ac06cc5..377df4a 100644
> --- a/include/scsi/scsi_cmnd.h
> +++ b/include/scsi/scsi_cmnd.h
> @@ -134,16 +134,7 @@ struct scsi_cmnd {
>  
>  static inline struct scsi_driver *scsi_cmd_to_driver(struct scsi_cmnd *cmd)
>  {
> -	struct scsi_driver **sdp;
> -
> -	if (!cmd->request->rq_disk)
> -		return NULL;
> -
> -	sdp = (struct scsi_driver **)cmd->request->rq_disk->private_data;
> -	if (!sdp)
> -		return NULL;
> -
> -	return *sdp;
> +	return *(struct scsi_driver **)cmd->request->rq_disk->private_data;
>  }
>  
>  extern struct scsi_cmnd *scsi_get_command(struct scsi_device *, gfp_t);
> 
> 
> --
> To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html