Re: [PATCH v2] libata: remove scsi host lock and serial numberusage

[Nicholas A. Bellinger]

On Wed, 2010-11-17 at 12:21 -0500, Jeff Garzik wrote:
> Much better...  Just updated #upstream-fixes with the below patch.
> 
> James just posted to linux-scsi
> 	[PATCH] Eliminate error handler overload of the SCSI serial number
> but as Tejun pointed out, we _completely and entirely_ replace the SCSI
> error handler apparatus with our own, thereby eliminating any
> possibility that we would ever read or use cmd->serial_number.
> 
> Our lock profile is quite a bit improved.
> 
> Further simplifications and cleanups related to 'done' are possible,
> but those are kept separate from the change below.
> 
> 
> commit 23e701e6208191ad103517ae7a700f2dc59ab2ec
> Author: Jeff Garzik <jeff@xxxxxxxxxx>
> Date:   Wed Nov 17 12:03:58 2010 -0500
> 
>     [libata] remove SCSI host lock and serial number usage from ata_scsi_queuecmd
>     
>     cmd->serial_number is never tested in any path we reach; therefore we may
>     remove the call to scsi_cmd_get_serial() inside DEF_SCSI_QCMD, the SCSI
>     host_lock acquisition surrounding it, and our own SCSI host_lock
>     unlock+relock cycle.
>     
>     Signed-off-by: Jeff Garzik <jgarzik@xxxxxxxxxx>

Hi Jeff,

I have merged this patch into lock_less-LLDs-for-38-v2, and dropped the
original:

[PATCH 02/12] libata: Convert to host_lock less w/ interrupts disabled internally

Thanks!

--nab

> 
>  drivers/ata/libata-scsi.c |   21 +++++++++------------
>  1 file changed, 9 insertions(+), 12 deletions(-)
> 
> diff --git a/drivers/ata/libata-scsi.c b/drivers/ata/libata-scsi.c
> index 19835d3..66aa4be 100644
> --- a/drivers/ata/libata-scsi.c
> +++ b/drivers/ata/libata-scsi.c
> @@ -3166,8 +3166,8 @@ static inline int __ata_scsi_queuecmd(struct scsi_cmnd *scmd,
>  
>  /**
>   *	ata_scsi_queuecmd - Issue SCSI cdb to libata-managed device
> + *	@shost: SCSI host of command to be sent
>   *	@cmd: SCSI command to be sent
> - *	@done: Completion function, called when command is complete
>   *
>   *	In some cases, this function translates SCSI commands into
>   *	ATA taskfiles, and queues the taskfiles to be sent to
> @@ -3177,42 +3177,39 @@ static inline int __ata_scsi_queuecmd(struct scsi_cmnd *scmd,
>   *	ATA and ATAPI devices appearing as SCSI devices.
>   *
>   *	LOCKING:
> - *	Releases scsi-layer-held lock, and obtains host lock.
> + *	ATA host lock
>   *
>   *	RETURNS:
>   *	Return value from __ata_scsi_queuecmd() if @cmd can be queued,
>   *	0 otherwise.
>   */
> -static int ata_scsi_queuecmd_lck(struct scsi_cmnd *cmd, void (*done)(struct scsi_cmnd *))
> +int ata_scsi_queuecmd(struct Scsi_Host *shost, struct scsi_cmnd *cmd)
>  {
>  	struct ata_port *ap;
>  	struct ata_device *dev;
>  	struct scsi_device *scsidev = cmd->device;
> -	struct Scsi_Host *shost = scsidev->host;
>  	int rc = 0;
> +	unsigned long irq_flags;
>  
>  	ap = ata_shost_to_port(shost);
>  
> -	spin_unlock(shost->host_lock);
> -	spin_lock(ap->lock);
> +	spin_lock_irqsave(ap->lock, irq_flags);
>  
>  	ata_scsi_dump_cdb(ap, cmd);
>  
>  	dev = ata_scsi_find_dev(ap, scsidev);
>  	if (likely(dev))
> -		rc = __ata_scsi_queuecmd(cmd, done, dev);
> +		rc = __ata_scsi_queuecmd(cmd, cmd->scsi_done, dev);
>  	else {
>  		cmd->result = (DID_BAD_TARGET << 16);
> -		done(cmd);
> +		cmd->scsi_done(cmd);
>  	}
>  
> -	spin_unlock(ap->lock);
> -	spin_lock(shost->host_lock);
> +	spin_unlock_irqrestore(ap->lock, irq_flags);
> +
>  	return rc;
>  }
>  
> -DEF_SCSI_QCMD(ata_scsi_queuecmd)
> -
>  /**
>   *	ata_scsi_simulate - simulate SCSI command on ATA device
>   *	@dev: the target device
> --
> To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
> the body of a message to majordomo@xxxxxxxxxxxxxxx
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/