Subject: Re: kern/21893: division by zero in scsipi_sync_factor_to_freq(): one more quirk in scsiconf.c ?
To: Anthony Mallet <>
From: Manuel Bouyer <>
List: netbsd-bugs
Date: 06/15/2003 21:49:17
On Sun, Jun 15, 2003 at 06:35:03PM +0200, Anthony Mallet wrote:
> >Description:
> Something changed in the kernel scsi code between March 30th and today:
> my cdrom which was normally detect as 
> cd0 at scsibus0 target 3 lun 0: <YAMAHA, CRW8824S, 1.00> cdrom removable
> cd0: async, 8-bit transfers
> started to insist on printing:
> cd0: sync (0.00ns offset 0), 8-bit
> followed by a division by zero in scsipi_sync_factor_to_period(0) (the
> function was called with factor==0).
> Adding a new quirk in scsiconf.c (PQUIRK_NOSYNC) seems to restore the
> correct behavior (see the patch below).
> Since I actually don't understand anything in the scsi code, I'm not sure
> wether this is the right fix or not...

No, it's not. The problem seems to be that async is not handled properly
any more.

As far as I can see, nothing changed in the scsipi code that could explain
this. The problem may be in the HBA code.
Can you post the boot message for your scsi controller ?

Manuel Bouyer <>
     NetBSD: 24 ans d'experience feront toujours la difference