[SOLVED] Kernel Panic in Red Hat Fedora 2

[rtroxel]

Reposted from linuxquestions.org (I'm a newbie and trying to meet people.)

While trying to install Red Hat Fedora 2, I twice received the message:

DMA Disabled
Drive Ready Error
Drive Status Error
I/O Error
Kernel Panic
Fatal exception In interrupt
In interrupt handler - not syncing

I tried two different hard drives and also upgraded the motherboard BIOS

The kernel is 2.6.

It's a homemade PC with an ASUS P3B-F motherboard, with version 2.21 BIOS.

It has 256MB of RAM, with a Pentium III processor.

The hard drive is an 80MB Maxtor, 7200 RPM. I also tried to install it on a WD 400BB, but that drive was defective. (It wasn't recognized by the motherboard.)

Also, I booted up with a RH 6.1 CD, and there was no kernel panic.

Help!:([

[bwkaz]

Originally posted by rtroxel
Drive Ready Error
Drive Status Error Those are not the full errors that you got. :) The hugely important part of them is which actual error bits are set in the IDE status word.

DriveReady is one bit in the status word. Error is another bit (it means check the error word). There's almost always another bit set in the status word, though, named SeekComplete. The first message is almost always { DriveReady SeekComplete Error }, indicating that those three bits are set.

Then, since the Error bit is set, the IDE code dumps the error word in the same way. This is the really important part. I've seen { DriveStatusError BadCRC } messages when I was using a round IDE cable, which went away as soon as I went to a better (i.e. a flat) cable. The round one wasn't cancelling radio noise well enough.

(CRC errors can be caused anywhere between the disk circuitry and the processor, inclusive, but AFAIK, not the drive platter itself. I believe the disk's onboard controller generates the CRCs on the data that it reads off the drive, and sends both the CRCs and the data over the IDE cable to the IDE controller. The IDE controller reports them to the OS, and I believe it's also the piece of circuitry that checks them. So either the disk controller is flipping bits, the IDE cable is flipping bits, or the IDE controller is miscalculating its CRC.)

I've also seen a case where the error word was just { UncorrectableError } -- that one went away when I wrote zeros to the entire drive and tried again. (That doesn't make any sense to me, because the sector that was giving uncorrectable errors already contained zeros, but if it worked, I won't argue too much. ;) Plus it was a new drive, so I didn't destroy anything important.)

So -- what were the exact messages generated by the kernel's IDE code?

[rtroxel]

Those are not the full errors that you got. The hugely important part of them is which actual error bits are set in the IDE status word.
My PC returned about fifty lines of error code, mostly pertaining to I/O. I did change the IDE cable, but that didn't solve anything. I'll have to reboot, and then write down the actual error messages. I'll let you know what I find.

Many thanks.
:)

[rtroxel]

OK. Since I first posted, I learned that the hard drive I was using was defective, so I shipped it back to Newegg. The motherboard BIOS wasn't recognizing the drive.

So, I no longer get these errors:

DMA Disabled
Drive Ready Error
Drive Status Error
I/O Error

However, when I boot up with the install CD, I get these two lines:

<0> Kernel panic: Fatal exception in interrupt
In interrupt handling - not syncing

However, when I boot with the RH 6.1 CD, I get no errors, and go right into the install GUI.

Does this mean the problem is with Fedora 2?

[bwkaz]

Hmm... I figured the kernel oops would have been related to the IDE errors. Obviously not.

Does it tell you which oops number is happening? Can you get at a kernel log from the session with the oops in it? The panic is because it oopsed inside an interrupt handler, and didn't know what to do other than shut itself down.

Are you sure your memory is OK? Try running a memtest86 on it overnight (hey, you never know). Alternately, if you have more than one stick, try removing sticks until there's only one in there (if you use dual-channel memory, then remove sticks until only two are in there), and see if the problem goes away. If not, replace the one (or two, if dual-channel) sticks with one (or two) others, and try again.

If you only have one (or two, for DDR) stick, try the memtest86.

[rtroxel]

Does it tell you which oops number is happening? Can you get at a kernel log from the session with the oops in it? The panic is because it oopsed inside an interrupt handler, and didn't know what to do other than shut itself down.
Hmmmm. I'm a newbie here, so I don't know what an oops is. I don't know what memtest86 is, either. Is it a Windows utility?

I did think about printing out a log, but since the PC never boots with the Fedora CD, I don't have access to the printer!

Actually, the PC already has two drives installed on it, with Windows XP running. My initial plan was to add a third drive, install Linux on that, and then make a dual-boot setup. When I realized the new drive was defective, I decided to see if I could boot up with Fedora, delete one of the XP partitions, and install Linux there, but then I got this second kernel panic.:confused:

[bwkaz]

www.memtest86.com

Oopses are when the kernel goes "oops, something impossible just happened". It logs an oops number plus the description of that oops, along with the kernel code that was executing at the time (in case it's a kernel bug), plus all modules that are loaded, and some other information.

Not all oopses are fatal -- some just kill whatever process they were executing on behalf of. But many are fatal to the entire system.

[rtroxel]

Thanks. I'll give memtest a try.

[rtroxel]

The problem is solved.

I took out the CD (an 8x Mitsumi) and subsituted a 16x Generic, and everything booted up OK.

Maybe this is just a temporary solution, but I can now install Fedora.

Thanks, everybody, for all the advice!

[bwkaz]

Ahhh, so it was bad bits being read off the CD. (Does Fedora have a CD test option when it boots? Some distros do -- they basically run an md5sum of the actual CD and make sure it matches whatever it should be, to try to ensure that no bits are getting flipped when the CD was burned or read.)

But whatever. Glad it's working!

[rtroxel]

Hey, hey!

I'm now posting this through Fedora 2!

Does Fedora have a CD test option when it boots?

I saw only an option called linux mediacheck. It didn't detect any flaws with the CD.