Recently installed MD 9.1 on K6 233, 200MB, 2 nics and ATI Rage Video.

Have tried 2 different drives (seagate ST36530A, 6.5GB and Fujitsu MPB3021AT, 2.1 GB) and here's what happens....

can boot into Icewm no prob, but KDE and Gnome both cause the drives to start spinning uncontollably and system freezes.

also when in Icewm, launching Nautilus or Konqueror has the same effect.....

have tried both drives by themselves on primary IDE channel.


any ideas???

whew, pretty weird...

This only happens when opeing up one of the monster GUIs?

The are some old bugs that occasionally crop up where KDE and Gnome make constant accesses until the machine hangs...

Got anything from either the.... XFree86 log, /var/log/messages or /var/log/syslog?

Also, what's the chipset off of this thing, just in case it is hardware. You can find out from /sbin/lspci

Cheers,

Finegan

Processor is AMD K6 233, TX chipset - (P82C439TX & PC82C371TX), only 96MB RAM (not 200)

/sbin/lspci doesn't exist however var/log/messages turns up some interesting things...

ALI15X3: chipset Rev 32
ALI15X3: not 100% native mode: will probe irqs later

hda0: task_no_data_intr: status 0x51 {DriveReady SeekComplete Error}
hda0: task_no_data_intr: error0x04 {DriveStatusError}

xFree86 log and syslog turn up no errors...

this post is from the machine, so there are a few things that do work on it. pretty strange.

the system doesn't always hang when the drive goes nuts. ive tried just letting it go, after a half hour i decided to power down b4 it blew up....

is there something in hdparm that i could try???

thanx

Check the info off of the drive:

hdparm -i /dev/hda

And see what DMA mode the drive is in, it'll be the one with * next to it.

Looking through the kernel config I bumped into CONFIG_BLK_DEV_ALI15X3
which allows DMA to work with that chipset... something tells me though that if the drives are too old it may flake out, I had similar problems with some old 4Gb drives on an SiS controller until I booted nodma. (actually my filesystems self destructed, but that's neither here nor there...)

Evidently there's more in: /usr/src/linux/drivers/ide/alim15x3.c at the top of the C file for the driver itself.

Cheers,

Finegan

I had similar errors when I ran with a buggy IDE chipset. I still use that
chipset but in the meantime the kernel IDE driver has been enhanced to work around my hardware bugs. The way to fix this problem, if this happens to be your problem, is to find out the model number of your IDE chipset. Then find out if there is a special kernel IDE module specifically for your chipset. If so, then do a kernel compile and include that module in your new kernel.

The kernel is detecting your chipset as:
ALI15X3: chipset Rev 32

finigan says:
Looking through the kernel config I bumped into CONFIG_BLK_DEV_ALI15X3 which allows DMA to work with that chipset... something tells me though that if the drives are too old it may flake out, I had similar problems with some old 4Gb drives on an SiS controller until I booted nodma. (actually my filesystems self destructed, but that's neither here nor there...)

That means that for a ALI15X3 to work with Linux a special IDE driver must be compiled into the kernel.

So check your Mandrake documentation and see if ALI15X3 support was included in your Mandrake kernel. If not then do a kernel compile and include ALI15X3 support.

the word you want is thrashes, not spins.
i have a couple of ali ddr motherboards and just gave
up trying to get the drives to work in dma mode on them.
and linux would start and endless loop of kernel error
messages when i tried ot shutdown on those
motherboards too.
Good luck.

thanx guys, still looking into it. This is my 1st foray into Linux (besides the coyote router distro that got me started), and I love it so far. I'm no stranger to the command line (1st pc was a trs-80), and can appreciate the power that it provides.

hdparm says its using mdma2 even though the board mfg and the drive mfg says they will support udma33

kernal does detect the chipset as ALI15x3 rev32, i'm trying to dig a little deeper to see if this is the correct revision...

both these drives worked fine on this machine when it was windoze, so i'm hoping against hopes that i can work this out. would like all my machines to be MS free.....

if and after you get the ali chipset support in the kernel
ok, you can play with different settings to get the drives
running as fast as reliable. in general the commands
are
hdparm -X<mode> -d1<for dma> -k1 < for keeping settings /dev/<drivedevice>
over a drive reset, there's several other little tweaks too.

hdparm -X67 -d1 would be udma33 i think, it might be 66.
the mdma modes are in the 30's or 40's.
usually the motherboard will set the drive at the highest
supported mode, and if dma doesn't get turned on
then hdparm -d1 /dev/hd<drive letter> will turn it on.
you can test the drives speed with hdparm -Tt /dev/hd?
after changing settings to see the results.

more clues
hdparm -g returns LBA 791/255/63, manufaturer lit says s/b 841/240/63

chs numbers are correct, and i tried changing the lba values in the bios, didn't make a difference.

tried to change proc/ide/ide0/hda/geometry to manufacture's values, wouldn't let me (--r--r--r read only)

is there another way i can force the system to try the manufacturer's lba numbers?? i read the hdparm man and saw some things that scared me a little to just go and punch them in..

still digging on the ali15x3 driver, found i2c-ali15x3.0.gz, looks like the driver....seems to me its not reading the system bios correctly for the hd params....

grrrrr

more clues
hdparm -g returns LBA 791/255/63, manufaturer lit says s/b 841/240/63

chs numbers are correct, and i tried changing the lba values in the bios, didn't make a difference.

tried to change proc/ide/ide0/hda/geometry to manufacture's values, wouldn't let me (--r--r--r read only)




that stuff doesn't matter. you don't want to change it
after you partition your disk anyway. that stuff only
has to do with the hard drive capacity. They aren't
true values anyway. multiply them out, and they'll
be close to the same.

Although I can't imagine they wouldn't, you might want to see if they compiled in proper support for the chipset into the Mandrake default kernel.

Cheers,

Finegan

bit the bullet and considered that MB to be a lost cause, it has now been relegated to being a coyote router once again.

picked up an asus p2b w/ pII 350 for next to nothing and am going to give that a try....just seemed like alot of grief for what it was worth...

at least this mb shows up under supported hardware on the mandrake site.....