> Thanks to those who responded to my earlier message. I decided to buy the
> ASUS PCI SCSI card (since I got a good deal on it). However, it has a
> Symbios Logic 53c810a chip which caused my 2.0.30 kernel to hang with a
> "scsi bus busy" error during the SCSI host probe. Fortunately, after
> patching to 2.0.33, the kernel successfully recognised the card, but
> I got a (hopefully, minor) warning that the chipset revision was greater
> than 2 (I've included the text of this warning at the end of my message).
Yeah - looks harmless. The driver just tells you, that it doesn't know that
revision, and there might be things that have changed ...
> Is this anything to be concerned about (maybe I should use one of the
> 2.1.x kernels which I understand have a 53c810a driver) ?
Don't think so.
> On another note, does the scsi card become associated with any particular
> /dev file
No. Eventually with something in /proc/scsi/* to allow for setting parameters,
but not with a /dev.
> or are the disks, scanners, CD-Rs, etc. on the SCSI chain the ones
> that receive a device file (I know the 1st disk would be /dev/sda) ?
Yes.
> I understand that I have to do a 'ln -s /dev/sga /dev/scanner',
Wait - this depends on the SCSI-ID of the scanner and the other devices.
Linux assigns SCSI device names in the order in which devices are found.
/dev/sga is the device with the lowest ID on the Bus. sgb the next one
etc.
The same goes for disks and cdrom, where only devices of that type count.
> but what is sga associated with (I'm assuming it's the card itself) ?
No. It is something like a "raw data channel" to the devicxe with the
lowest ID number.
> I'm also a bit fuzzy regarding SCSI termination: The manual states that
> termination is required on the end device,
This is correct. Let me explain what termination is about:
Electrical waves that travel through a cable will suffer a reflection,
when they hit the open end of a cable. Thus a reflected pulse will go
back the line and interfere with the data on the cable.
If you put a "terminator", which is a resistor with a special value on the
end, the electrical wave will be kind of absorbed in that resistor, thus
avoiding the reflection problem.
This should illustrate where terminators need to be put :
At the ends of the device chains.
Putting them earlier will cut the signal from the later devices, leaving
them out will cause reflection.
Thus you should turn on termination on the last devices on the chain
on either side. If the device has no internal termination, put an
external terminator on the outgoing SCSI plug of the device or on
the next connector of a flat cable.
> but I was told that I don't have to worry about it (auto-termination).
This only applies to the SCSI host (the adaptor), if it happens to be on
one end of the chain. It will sense the termination resistors, see if it
is in the middle or at the end of the cable and thus terminate itself
correctly.
However this only works on the assumption, that alll other devices are
set up correctly.
> Does the scanner have to be switched on before I boot-up for association
> to /dev/sga to take place
This would be better for the beginning. Linux can add devices later, but
this can cause confusion and should be avoided.
> Finally (for the sake of argument) if I put 2 scanners on the SCSI chain
> how could I associate both with sga ?
You don't. Both scanners have different SCSI-IDs (or it won't work). Thus they
get different /dev/sg? entries.
CU,ANdy
-- = Andreas Beck | Email : <andreas.beck@ggi-project.org> = ======== GGI - The Right Thing To Do : http://www.ggi-project.org/ ========
-- Source code, list archive, and docs: http://www.mostang.com/sane/ To unsubscribe: echo unsubscribe sane-devel | mail majordomo@mostang.com