On Mon, Mar 27, 2006 at 01:26:02PM -0500, Steven M. Bellovin wrote:
>On Mon, 27 Mar 2006 12:04:42 -0500, Thor Lancelot Simon
><tls@rek.tjls.com> wrote:
>
>> On Mon, Mar 27, 2006 at 12:00:51PM -0500, Steven M. Bellovin wrote:
>> > On Mon, 27 Mar 2006 18:58:03 +0200, Manuel Bouyer
>> > <bouyer@antioche.eu.org> wrote:
>> > 
>> > > On Mon, Mar 27, 2006 at 11:53:25AM -0500, Steven M. Bellovin wrote:
>> > > > 
>> > > > The badsect will grab the sector, but is the problem exactly that
>> > > > sector or all sectors past that point?
>> > > 
>> > > Exactly this sector.
>> > > 
>> > 
>> > Then badsect(8) should do exactly what's desired.
>> 
>> But why would one want to use it? 
>
>Someone, I forget who, thought that there might be a serious
>performance hit from LBA48.  If there's no such problem, or you're
>going to encounter it anyway on much of the disk, I agree that badsect
>is not the right way to go.

This all sounds a lot like Silicon Image SATA / seagate issues 
http://www.linux-m32r.org/lxr/http/source/drivers/scsi/sata_sil.c

356         /* limit requests to 15 sectors */
357         if (slow_down ||
358             ((ap->flags & SIL_FLAG_MOD15WRITE) &&
359              (quirks & SIL_QUIRK_MOD15WRITE))) {
360                 printk(KERN_INFO "ata%u(%u): applying Seagate errata fix (mod15write workaround)\n",
361                        ap->id, dev->devno);
362                 dev->max_sectors = 15;
363                 return;
364         }

*that* gave me a 60% performance hit on my linux sil 160Gb seagate
box. even though I reported it wasn't necessary in my case. it
was pushed in the kernel for my hw. I stuck with the old linux
kernel and has been stable a year (or three?) now.  Hope I'm not
confusing the heck out of the present bad block issue.

// George


-- 
George Georgalis, systems architect, administrator <IXOYE><
http://galis.org/ cell:646-331-2027 mailto:george@galis.org