On Wed, 2003-03-12 at 16:50, Mike Anderson wrote:
> Alan Cox [alan@lxorguk.ukuu.org.uk] wrote:
> > On Wed, 2003-03-12 at 23:55, Douglas Gilbert wrote:
> > > /*
> > > * Limit max queue depth on a single lun to 256 for now. Remember,
> > > * we allocate a struct scsi_command for each of these and keep it
> > > * around forever. Too deep of a depth just wastes memory.
> > > */
> > > if(tags > 256)
> > > return;
> > > ....
> >
> > I can see the memory consideration. However the thing can really handle big
> > queues well. Possibly we should be setting the queue to 512 / somefunction(volumes)
> > though to avoid the worst case overcommit here
>
> I agree with Doug that the previous comment is out of date and we could
> raise the value, but we also should never leave the function with the
> possibility that the queue_depth is 0.
>
> The patch below is something Patrick and I where discussing though I
> believe he indicated that I should print out the value we where setting
> the queue_depth to. It was only compiled and not tested on any devices.
>
> -andmike
> --
> Michael Anderson
> andmike@us.ibm.com
>
> scsi.c | 32 +++++++++++++++++++++++---------
> 1 files changed, 23 insertions(+), 9 deletions(-)
>
> ------
>
> --- 1.96/drivers/scsi/scsi.c Fri Feb 21 13:46:58 2003
> +++ edited/drivers/scsi/scsi.c Wed Mar 12 16:05:42 2003
> @@ -926,15 +926,28 @@
> /*
> * refuse to set tagged depth to an unworkable size
> */
> - if(tags <= 0)
> - return;
> + if(tags <= 0) {
> + printk(KERN_WARNING "(scsi%d:%d:%d:%d) "
> + "%s, tag value to small\n"
> + "disabled\n", SDpnt->host->host_no,
> + SDpnt->channel, SDpnt->id, SDpnt->lun,
> + __FUNCTION__);
> +
> + SDpnt->queue_depth = 1;
> + }
> /*
> - * Limit max queue depth on a single lun to 256 for now. Remember,
> - * we allocate a struct scsi_command for each of these and keep it
> - * around forever. Too deep of a depth just wastes memory.
> + * Limit max queue depth on a single lun to 256 for now.
> + * Too deep of a depth just wastes memory.
> */
> - if(tags > 256)
> - return;
> + if(tags > 256) {
> + printk(KERN_WARNING "(scsi%d:%d:%d:%d) "
> + "%s, tag value to big\n"
> + "disabled\n", SDpnt->host->host_no,
> + SDpnt->channel, SDpnt->id, SDpnt->lun,
> + __FUNCTION__);
> +
> + SDpnt->queue_depth = 256;
> + }
>
> spin_lock_irqsave(&device_request_lock, flags);
> SDpnt->queue_depth = tags;
> @@ -949,9 +962,10 @@
> break;
> default:
> printk(KERN_WARNING "(scsi%d:%d:%d:%d) "
> - "scsi_adjust_queue_depth, bad queue type, "
> + "%s, bad queue type, "
> "disabled\n", SDpnt->host->host_no,
> - SDpnt->channel, SDpnt->id, SDpnt->lun);
> + SDpnt->channel, SDpnt->id, SDpnt->lun,
> + __FUNCTION__);
> case 0:
> SDpnt->ordered_tags = SDpnt->simple_tags = 0;
> SDpnt->queue_depth = tags;
>
>
This looks like a good idea. This is better than a hung system and no
indication why.
Mark.
-- Mark Haverkamp <markh@osdl.org>- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Sat Mar 15 2003 - 22:00:37 EST