To: ncr53c810@colorado.edu, linux-scsi@vger.rutgers.edu
cc:
Subject: NCR53c810 Linux driver syncronous bug
Fcc: +outbox
--------

If you look in drivers/scsi/53c7,8xx.h,

#define SXFER_TP_SHIFT		4

should be 

#define SXFER_TP_SHIFT		5

The net effects of this bug are that the NCR chip talks to < 5M/sec drives
faster than they've said we can.  In most cases, SCSI-I devices will claim
5M/sec support, in which case you get no error (the value is biased by 4, so
0 is used for the TP field in this case); if they're a bit slower, you may
see a 20% error.

This fix will be in the rel9 patch, which will probably also include 
disconnect/reconnect support (My Maxtor, Syquests, and tape drive 
disconnect/reconnect fine now, but something bogus is happening with 
my RZ55).