Hello all:

I have installed a second hard disk in my Linux box. All goes ok,
and there is not problem except that whenn the machine start.
Here is a extract of dmesg:

If I cat /proc/pci (I have erased ethernet and video):

The machine is a old Pentium 100MHz that goes very well in Linux, but
this problem ...
I am looking for a solution in google but I have found nothing.
I have set mode 3 for my new drive in my BIOS. But nothing.
I must to say also that I have an old Award BIOS (95).
I know that I can put in my start script hdparm -d0, but, there is another solution?

Can you help me?

Thanks all.

BIOS mode 3 will 'dumb-down' the drive to a lower PIO, but you may need this as it shares the IDE1 chain with the 40x CD-ROM drive (which prob only supports DMA(33)).

You could try to remove the ide-dma reference in your start-up scripts (check /etc/intit.d/rc2.d + rc3.d).

As a last resort you could get a bootdisk util from Fujitsu to set the _drive's_ bios down to DMA33 default.

You could also throw the drive on the IDE0 controller, but that may be a performance issue for you as the ATA standard does not supprt the diconnect call - meaning that either the master OR the slave can process a reques. depends on how you throw your partitions around.

I also have been having that trouble, with no luck in finding a solution. Last night I recompiled my kernel using make xconfig, and somewhere in there, I think in the ATA/IDE/MFM/RLL support section I had to turn a feature on. Reading the help for this particular 'switch' mentioned the error message you are getting, and by saying Y to it (default is N), it could fix the problem.

Sorry I can't remember exactly where it is in the xconfig, but if you do some snooping you will come across it.

OK, I found it.

Now, keep in mind that this is in xconfig (cuz I'm not a 733t hax0r, merely a newbie), but here it is:

make xconfig

then, in ATA/IDE/MFM/RLL support, click the IDE, ATA and ATAPI Block devices button. the thing you need to say Y to (that is set to N by default) is the 'Use multi-mode by default' selection.

Understand that this is the frist step to recompiling your kernel, and you must be willing to follow through to the end for this 'fix' to work.

PLus you'll want to include your drive controller as well be it PIIX/4 or Promise or Highpoint...

I'll test recompiling my kernel with this option enabled. I will say you how it gets.

Thanks all

Hello, I have downloaded the latest kernel and I have selected 'Use multi-mode by default' and PIIX/4.
I have done it in this machine (I have done it in the Fujitsu). I have must to restart a pair times the 'make dep' and 'make bzImage': kernel failures and memory dumps.
I have tryed to restart with the new kernel and hard disk errors: the superblock was kaput. The old kernel doesn't boot too.
I have reinstalled today the machine without the Fujitsu: I have download a test utility from their site and says that the disk is corrupt. (There isn't a utility that down to DMA33 for my model)

I must to test a disk before go into the server.

Thanks all.

Ouch.

What is wrong with the disk? Here's a story about a Fuju 10GB ATA/66. (these are also the same drives populating my desktop RAID array) Needless to say I am awaiting my first failure.

I had a client with one that 'mysteriously' died after they moved it to another room. It was a Win98 box and it started making a loud click/bonk noise every now and again. Then the rotation speed went _very_ high and then very low. This went on for a while until they finally called me. By then it was too late. I was able to back up the critical stuff but a full recovery was impossible. I ran the drive outside of the case to see what was happening. The electronics were all very hot and cooling them incresed the time before a complete I/O failure. I was able to get some more data from the dirve with a can of compressed air and an old heatsink.