On Thu, 3 Dec 1998, Tom Gelhause wrote:
> Gerard Roudier schrieb:
> >
> > Fast-20 Single Ended SCSI has very low electrical characteristics margin.
> > Please, would you want to describe accurately to me your SCSI BUS in terms
> > of:
> > - Cable length and quality.
> > - Connectors.
> > - Terminator types (active/passive).
> > - Position of each device and each termination.
> >
> > Regards,
> > Gerard.
> >
>
> I also have problems with the 53c875 using Linux. These started when I
> switched to a new Board (FIC VA-503+, Via Apollo MPV3), but they only
> occur using Linux, no problems on Win NT/95. I also tryed a 53c810 and
These O/Ses are known to castrate device features in order to run on
broken hardwares and donnot write relevant error messages to log files
when such exist. The Linux ncr driver tries to give you performances
you paid for and to inform you on any detected error.
> also had errors. I detected no errors with an AMD controller. I only
> have SCSI II devices connected to the SCSI bus, wich is passively
> terminated. (Only one cable, good quality, also tryed an other one.) The
> errors occure very irregularely, it isn't possible to tell when.
> Sometimes the box runs for over 5 h, sometimes just 10 min.
> Unfortunately, I don't understand the type of error. Sometimes it's a
> "phase change"-error, sometimes parity or something else. Has anybody
> any further proposals where I could look for the error?
This looks like the chipset being broken or badly configured.
(Or perhaps some device on the PCI bus behaves incorrectly)
You may try the following (1, 1+2 then 2 suggested):
1 - Configure the driver for NORMAL IOs (instead of MMIO which is the
default and the fastest)
2 - Boot your system using ncr53c8xx=specf:n
specf means 'special features', n means 'no'.
Regards,
Gerard.
PS: Could you let me know the PCI device ids and revision ids of your
53C8XX chips.
-
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to [EMAIL PROTECTED]
