Very Computer

The reason I need it as kernel code is so that I can put /usr on a
seperate 8.2g drive.

When aha1542 is built as a module, it pops up immediately before the
login screen, and everything works as planned.

Remaking the kernel with it set to be kernel code, it starts the bus
scan during the boot, not too long after the ide devices are found, but
then goes into a double loop of resetting the device 2-4 times, then the
whole bus, then repeats the device, all the while saying it can't abort
the call, but evertually it will find the drive at address 0 on the bus
and go on, at which point it repeats this procedure for address one,
each pass thru the loop taking about 5 minutes, finding the quatum 8.2g
at each addreess in turn until it gets to address 4, at which point the
discovery phase eventually finds a Colorado T4000s at address 4, which
is its correct address, then also at 5 & 6.

Then it apparently repeats the loop, asking each device for its capacity
details, and going through this same error recovery time killer, except
that this time it only gets partial answers only, and only at the
correct address for the device.

I let it go through all that, and finally got a login screen and did so,
but was unable to open either device, not even with fdisk /dev/sda.

I've used several of these cards in years past, and have always had to
edit the source to lower the address of its port by 0x100, to the 0x230
area, and to reset the irq from 11 to 10.  Without those changes the
module doesn't init itself, ever, and it has required those changes to
the code in order to function on 3 different motherboards I've run it
on.

But here, I need the direct in the kernel version so I can put /usr on
it.  Put it in the kernel, and it fails miserably, both for 2.2.18,
2.2.19, and 2.4.2.  The 2.4.2 failure surprised me as that code has a
diff that looks to be in excess of 10k from the older code.

So why can't I put aha1542.o into the (apparently any) kernel?
Inquiring minds want to know. (-;

Cheers, Gene
--

        email gene underscore heskett at iolinc dot net
#Amiga based X10 home automation program EZHome, see at:#
 <http://www.iolinc.net/gene_heskett>
This messages reply content, but not any previously quoted material,
is ? 2001 by Gene Heskett, all rights reserved.  Due to recent
changes in M$ lusers TOS, mail from msn.com, msn.net, microsoft.com,
microsoft.net, hotmail.com, and hotmail.net is auto-deleted, unread.
--

I don't know the answer I'm afraid, but I have a small machine with a
1542 with /usr on the SCSI disk (/ is on an IDE as the 1542 isn't
bootable), using an initial Ramdisk containing the aha1542 module. I
must admit that I thought that the reason I had to do this was because I
also have my swap partition on the SCSI (the IDE is only 120M). I didn't
think /usr needed it. I havn't tested this mind. My kernel is 2.2.16.

My 1542 is using IRQ 11 and I/O 330. I'm not exactly sure what the MB
is, this machine was cobbled together out of spare bits. This is a
simple machine used as an ISDN router/DNS/mailserver, it only has the
SCSI card, an RTL8139 ethernet and an AVM Fritz! PCI ISDN card. No sound
cards or such to*things up.

You might try the Ramdisk approach to see if it works for you.

I have come to the conclusion (with a little help from the more
experienced) that Ramdisks are best avoided so had been thinking to
change this setup to the one you are attempting (I have already done
this on the large machine with a 2940). So I will exercise a little
caution in this, and follow this thread.

Regards, Ian

Gene Heskett sends Greetings to Ian Northeast;

Quote:

>> The reason I need it as kernel code is so that I can put /usr on a
>> seperate 8.2g drive.

>> When aha1542 is built as a module, it pops up immediately before
>> the login screen, and everything works as planned.

>> I've used several of these cards in years past, and have always had
>> to edit the source to lower the address of its port by 0x100, to
>> the 0x230 area, and to reset the irq from 11 to 10.  Without those
>> changes the module doesn't init itself, ever, and it has required
>> those changes to the code in order to function on 3 different
>> motherboards I've run it on.

 IN> I don't know the answer I'm afraid, but I have a small machine
 IN> with a
 IN> 1542 with /usr on the SCSI disk (/ is on an IDE as the 1542 isn't
 IN> bootable), using an initial Ramdisk containing the aha1542
 IN> module. I must admit that I thought that the reason I had to do
 IN> this was because I also have my swap partition on the SCSI (the
 IN> IDE is only 120M). I didn't think /usr needed it. I havn't tested
 IN> this mind. My kernel is 2.2.16.

 IN> My 1542 is using IRQ 11 and I/O 330. I'm not exactly sure what
 IN> the MB is, this machine was cobbled together out of spare bits.
 IN> This is a simple machine used as an ISDN router/DNS/mailserver,
 IN> it only has the SCSI card, an RTL8139 ethernet and an AVM Fritz!
 IN> PCI ISDN card. No sound cards or such to*things up.

I finally went to the adaptec site and got the docs on the card, reset
it from 0x230 to 0x334, edited the source to restrict it to a MAXBOARDS
of 1 and gave it the 0x334 address as the choice.

Its working now, albeit rather slowly since its an ISA card.  I had this
card in my home machine at one point, and with an older sheet metal
quantum 820 megger on it, could not get it above 900k/second, and
accessing it resulted in a severely lagged machine even though its dma
channel was registered into /proc/dma.  A later advansys pci card gets a
couple of megs/second out of that drive, without the lags.

It looks as if its going to do that on this 8.2g drive too, as I have
about 3g in /usr and it only had 2 of it copied when I left for the
night.  It had been working for over an hour on the copy then.

So I'll use it (and * about it) till I stumble across a good pci
card.

Cheers, Gene
--

        email gene underscore heskett at iolinc dot net
#Amiga based X10 home automation program EZHome, see at:#
 <http://www.veryComputer.com/;
This messages reply content, but not any previously quoted material,
is ? 2001 by Gene Heskett, all rights reserved.  Due to recent
changes in M$ lusers TOS, mail from msn.com, msn.net, microsoft.com,
microsoft.net, hotmail.com, and hotmail.net is auto-deleted, unread.
--