Re: What breaks aic7xxx in post 2.6.12-rc2 ?

[Frank van Maarseveen <frankvm@frankvm.com>]

On Mon, Jun 13, 2005 at 05:59:53PM -0500, James Bottomley wrote:
> On Mon, 2005-06-13 at 23:42 +0200, Frank van Maarseveen wrote:
> > kernel:    Vendor: WANGTEK   Model: 5525ES SCSI       Rev: 73F 
> > kernel:    Type:   Sequential-Access                  ANSI SCSI
> > revision: 02
> > kernel:   target0:0:4: Beginning Domain Validation
> > kernel:   target0:0:4: Domain Validation skipping write tests
> > kernel:  scsi0:0:4:0: Attempting to queue an ABORT message
> 
> The aic7xxx error handling is still very unreconstructed and DV takes
> the driver through this if there's a mismatch.  My best guess is that
> like Gr\'egoir's CD-ROM, your wangtek is claiming to support a speed it
> cannot.  We find this out in DV, but not until we've gone through all
> the error paths.  The quick fix is simply to set the bios to whatever
> the tape eventually configures with (although you'll need the current
> patch set to make that work properly).

Well, I can't seem to find the right collection of patches applying
cleanly on 2.6.12-rc6 and the Adaptec BIOS says "ASYNC". It cannot
go below 10MB/s and the tape unit can't go above 3.3 according to
/proc/scsi/aic7xxx/0 I think:

Target 4 Negotiation Settings
        User: 6.600MB/s transfers (16bit)
        Goal: 3.300MB/s transfers
        Curr: 3.300MB/s transfers
        Channel A Target 4 Lun 0 Settings
                Commands Queued 61
                Commands Active 0
                Command Openings 1
                Max Tagged Openings 0
                Device Queue Frozen Count 0

So, I'm kind of stuck with a very slow driver initialization. I think it
takes much longer than the 30 seconds visible in the logfile, at least
that's the impression I have (I'll use a stopwatch next time).

-- 
Frank