phorgan@cup.portal.com (Patrick John Horgan) (03/03/90)
A friend of mine who's hacking a tape backup disassembled the driver for the 2090a and found out why the dern thing won't do direct SCSI writes. It seems that the high order bit of one of the DMA registers specifies read or write, and the code never pulls up the put for write. He also found a workaround. He set bit 24 (the 25th bit counting 0), of the data buffer address. Only 24 bits are used (0-23), but the higher bits aren't checked or cleared. This bit position luckily gets slid in to the high bit of the DMA register causing write to work:) He told me the day before yesterday that he's been able to write to a SCSI tape drive with the 2090a, and hopes to have a driver written soon. Patrick Horgan phorgan@cup.portal.com
greg@noel.CTS.COM (J. Gregory Noel) (03/04/90)
Patrick John Horgan (phorgan@cup.portal.com) writes: >A friend of mine who's hacking a tape backup disassembled the driver >for the 2090a ..... Can you put me in touch with this friend? Ever since I upgraded to 1.3 I have been unable to talk to my big SCSI disk -- it hangs in BindDrivers. I hoped that 1.3.2 would fix the problem, but it didn't, so the only thing that seems feasible is to reverse-engineer the driver for the 2090 and debug the problem myself. If your friend could make the disassembled driver available to me, it would save a lot of time. I sent the above via mail to Patrick, but I also have a question for the assembled net.wisdom: What changed from 1.2 to 1.3 that might cause my problem? Getting the drive to bind under 1.2 was somewhat flakey -- it would sometimes take two or three reboots before the drive was recognized as present, although it never just hung -- but once recognized, it ran very reliably. After spending all that money on a 400Mb drive, I'd like to be able to use it! Thanks in advance, -- -- J. Gregory Noel, UNIX Guru greg@noel.cts.com or greg@noel.uucp