Re: [PATCH] fix command retries in spi_transport class

[James Bottomley <James.Bottomley@SteelEye.com>]

On Mon, 2005-05-02 at 21:04 -0500, K.R. Foley wrote: 
>   Vendor: SEAGATE   Model: SX118273LC        Rev: 6679
>   Type:   Direct-Access                      ANSI SCSI revision: 02
> scsi1:A:0:0: Tagged Queuing enabled.  Depth 32
>  target1:0:0: Beginning Domain Validation
> (scsi1:A:0): 6.600MB/s transfers (16bit)
> (scsi1:A:0:0): parity error detected in Data-in phase. SEQADDR(0x6a) SCSIRATE(0x80)
> (scsi1:A:0:0): parity error detected in Data-in phase. SEQADDR(0x83) SCSIRATE(0x80)
> (scsi1:A:0:0): parity error detected in Data-in phase. SEQADDR(0x83) SCSIRATE(0x80)
> (scsi1:A:0:0): parity error detected in Data-in phase. SEQADDR(0x1a6) SCSIRATE(0x80)
> (scsi1:A:0:0): parity error detected in Command phase. SEQADDR(0x1a6) SCSIRATE(0x80)
>  target1:0:0: Wide Transfers Fail

This means that the driver is now correctly returning an error.

> (scsi1:A:0): 3.300MB/s transfers

Here we set the correct transfer parameters (narrow), which means all
the parity errors should go away.

I'm at a loss to either duplicate this or figure out from the code what
might be going on, so I really need a trace to help out.

If you can't set the kernel boot line options, you can just edit scsi.c
to set the variable scsi_logging_level to 0xffff (and remove the
module_param lines for it, just in case).  Remember that
CONFIG_SCSI_LOGGING must be set as well.

Thanks,

James