[patch 7/12] s390: fba dasd i/o errors.

[patch 7/12] s390: fba dasd i/o errors.

[Martin Schwidefsky]

[patch 7/12] s390: fba dasd i/o errors.

From: Horst Hummel <horst.hummel@de.ibm.com>

The FBA discipline does not use retries for failed requests. A request
fails after the first unsuccessful start attempt. There are some rare
conditions (e.g. CIO path recovery) in which the start of an i/o on
a fba device can fail. A tiny amount of retries is therefore
reasonable.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>

diffstat:
 drivers/s390/block/dasd_fba.c |    4 +++-
 1 files changed, 3 insertions(+), 1 deletion(-)

diff -urpN linux-2.6/drivers/s390/block/dasd_fba.c linux-2.6-patched/drivers/s390/block/dasd_fba.c
--- linux-2.6/drivers/s390/block/dasd_fba.c	2005-06-17 21:48:29.000000000 +0200
+++ linux-2.6-patched/drivers/s390/block/dasd_fba.c	2005-07-11 17:37:46.000000000 +0200
@@ -4,7 +4,7 @@
  * Bugreports.to..: <Linux390@de.ibm.com>
  * (C) IBM Corporation, IBM Deutschland Entwicklung GmbH, 1999,2000
  *
- * $Revision: 1.39 $
+ * $Revision: 1.40 $
  */
 
 #include <linux/config.h>
@@ -354,6 +354,8 @@ dasd_fba_build_cp(struct dasd_device * d
 	}
 	cqr->device = device;
 	cqr->expires = 5 * 60 * HZ;	/* 5 minutes */
+	cqr->retries = 32;
+	cqr->buildclk = get_clock();
 	cqr->status = DASD_CQR_FILLED;
 	return cqr;
 }

Re: [patch 7/12] s390: fba dasd i/o errors.

[Denis Vlasenko]

On Monday 11 July 2005 19:36, Martin Schwidefsky wrote:
> [patch 7/12] s390: fba dasd i/o errors.
> 
> From: Horst Hummel <horst.hummel@de.ibm.com>
> 
> The FBA discipline does not use retries for failed requests. A request
> fails after the first unsuccessful start attempt. There are some rare
> conditions (e.g. CIO path recovery) in which the start of an i/o on
> a fba device can fail. A tiny amount of retries is therefore
> reasonable.
> 
> Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
> 
> diffstat:
>  drivers/s390/block/dasd_fba.c |    4 +++-
>  1 files changed, 3 insertions(+), 1 deletion(-)
> 
> diff -urpN linux-2.6/drivers/s390/block/dasd_fba.c linux-2.6-patched/drivers/s390/block/dasd_fba.c
> --- linux-2.6/drivers/s390/block/dasd_fba.c	2005-06-17 21:48:29.000000000 +0200
> +++ linux-2.6-patched/drivers/s390/block/dasd_fba.c	2005-07-11 17:37:46.000000000 +0200
> @@ -4,7 +4,7 @@
>   * Bugreports.to..: <Linux390@de.ibm.com>
>   * (C) IBM Corporation, IBM Deutschland Entwicklung GmbH, 1999,2000
>   *
> - * $Revision: 1.39 $
> + * $Revision: 1.40 $
>   */
>  
>  #include <linux/config.h>
> @@ -354,6 +354,8 @@ dasd_fba_build_cp(struct dasd_device * d
>  	}
>  	cqr->device = device;
>  	cqr->expires = 5 * 60 * HZ;	/* 5 minutes */
> +	cqr->retries = 32;

2..4 maybe, but 32? This isn't tiny by any account.

> +	cqr->buildclk = get_clock();
>  	cqr->status = DASD_CQR_FILLED;
>  	return cqr;
>  }

Re: [patch 7/12] s390: fba dasd i/o errors.

[Martin Schwidefsky]

> > @@ -354,6 +354,8 @@ dasd_fba_build_cp(struct dasd_device * d
> >          }
> >          cqr->device = device;
> >          cqr->expires = 5 * 60 * HZ;         /* 5 minutes */
> > +        cqr->retries = 32;
>
> 2..4 maybe, but 32? This isn't tiny by any account.

Are you arguing the use of the adjective "tiny" or the technical
aspects of using 32 as the number of retries for dasd fba?
In the dasd driver we use a retry count of 255 as "standard", so
32 is indeed much smaller than that. If you can call it tiny,
well who cares??

blue skies,
   Martin

Martin Schwidefsky
Linux for zSeries Development & Services
IBM Deutschland Entwicklung GmbH

Re: [patch 7/12] s390: fba dasd i/o errors.

[Denis Vlasenko]

On Tuesday 12 July 2005 13:36, Martin Schwidefsky wrote:
> > > @@ -354,6 +354,8 @@ dasd_fba_build_cp(struct dasd_device * d
> > >          }
> > >          cqr->device = device;
> > >          cqr->expires = 5 * 60 * HZ;         /* 5 minutes */
> > > +        cqr->retries = 32;
> >
> > 2..4 maybe, but 32? This isn't tiny by any account.
> 
> Are you arguing the use of the adjective "tiny" or the technical
> aspects of using 32 as the number of retries for dasd fba?
> In the dasd driver we use a retry count of 255 as "standard", so
> 32 is indeed much smaller than that. If you can call it tiny,
> well who cares??

I meant that 32 retries is too many. Retries tend to multiply.
If OS does N retries per failed attempt and disk drive does M
attempts per attempt, you end up with N*M retries.
Add a few 'retrying' layers, and you have a 'I cannot umount
this fscking scratched CDROM, maybe tomorrow' type disaster.

It's better to err to the smaller number of attempts.
--
vda