> On Thu, 27 Aug 1998, Steven N. Hirsch wrote:
>
> > I am having terrible problems with the 3.0 series of drivers on an Alpha
> > UDB. They work perfectly under Intel Linux (I have C810a-based adapters
> > in two different boxes - one an SMP), so I'm wondering if this is a fluke
> > in the UDB hardware.
> >
> > It boots in 2.1.117, but blows up fairly quickly with an endless cascade
> > of time-outs on /dev/sdb. Eventually, it tries a reset, pauses, and just
> Your system probably does not boot from this drive (does it?).
> Does Linux print either the capacity or the partition table of this drive
> before starting to endless loop with timeouts?
No, the initial bootstrap (Milo) comes up from sdba, the internal SCSI
drive. However, Linux comes up all the way; correctly identifying the
drive capacity and partitions. It actually gets to a login prompt and
will (usually) function long enough to start doing something. On
occasion, it's been up long enough to begin a kernel compile, though it's
never made it to the end.
> > There is nothing whatever wrong with the drive. It is a Maxtor MXT540S
> > which works perfectly under version 2.5f (and earlier) of your driver.
> This may be a bug in the new driver version.
I'm sure some "feature" of the Maxtor drive is triggering it, as others
are having good success on the UDB with other drives + v3.0 driver.
> For now, you may apply the following patch that disables the done queue
> handling and let me know if it makes differences.
>
> --- linux/drivers/scsi/ncr53c8xx.c.SAVE Fri Aug 28 14:47:31 1998
> +++ linux/drivers/scsi/ncr53c8xx.c Fri Aug 28 14:59:50 1998
> @@ -282,7 +282,7 @@
> **==========================================================
> */
>
> -#define SCSI_NCR_CCB_DONE_SUPPORT
> +/* #define SCSI_NCR_CCB_DONE_SUPPORT */
> #ifdef SCSI_NCR_CCB_DONE_SUPPORT
>
> #define MAX_DONE 24
>
> ---------------------- Cut here --------------------------
>
>
> You also could boot your system with conservative SCSI features:
> (No tags, 5MHz sync data transfers (50ns x 4 = 200ns))
>
> ncr53c8xx=tags:0,sync:50
^^^^^^^^^^^^^^^
Hmm. The docs imply that "sync:5" is 5Mhz. I'll try your suggestions and
let you know if the (mis)behavior changes.
Steve
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.altern.org/andrebalsa/doc/lkml-faq.html