Hi

This has been occuring ever since i started my first installation of mandrake onto a newly formatted computer. It is a Mandrake 10.0. Whenever it crashes, ill hit the reset button to restart. This is what they call the Unclean shutdown right? So before it loads into mandrake, theres a "Force Integrity Check" Option.

This is the part when it gets hung again and i have to reset my system another time for it to boot back into mandrake.


This hanging has been happening ever since installation.



Is this a mandrake issue or my harddisk bad sector issue (im not sure which one is the case)?

If you have a bad sector it can cause this. I often get crashes if I try to overclock too much, usually shows up in memory problems first, check with >5 passes of a memory check program run from a rescue disk.

I would suggest you check your drive from a boot cd, then you can do the whole thing w/o it having to be mounted at boot time.

Eh, a bad sector should produce error messages but not always a freeze.

As suggested it would be a GREAT idea to first run memtest on your system.

Linux taxes the computers more so than say Winblows.

Windows can run fine all day long and systems which Linux will crash immediately.

The problem is usually NOT with Linux, but rather an inherent instability in the machine especially if you get ERR-11's when running applications.

That said it also may be an issue with AGP and your video card.

Try bringing up Linux in TEXT mode only, not permitting X windows to start.

(Safe mode).

If Linux brings you to the text login and you can log in and execute text based interface commands, then you are probably suffering from the latter, and not memory or hard disk problems.

Your're right about the immediate crashing by linux.

I'll probably do a Memory Test with memtest86. Are there any (BEFOREBOOT) test for HDD to scan for bad sectors?

Seriously i have no idea how do i begin this two with linux.

Re: HDD

Yes you can use a LiveLinux CD and boot it up.

Then you can use the badblocks command to non-destructively locate any bad blocks.

Now for the kicker...

If you suspect your drive has a few bad blocks and don't care that it may fail on you at some point... (e.g. you may be able to continue to use the drive for a good while this way...) you can have Linux mark the bad blocks and avoid them using bad blocks and fsck.

If the bad blocks occur in non-directory areas in a Reiser file system (dunno about ext3), fsck, once instructed will attempt to even move things around the bad blocks preserving what it can.

I was able to salvage a couple of known non-dying 120 gig drives that had hard bad block errors caused by a peculiar "event", and use them under Linux this way.

They have been in use on non-critical systems for about a year now w/o a single subsequent problem.

However if I ever reformat the partition I MUST run the same proceedure again.

Linux also seems to like RAM. The more Ram the better. Disk caching into the swap partition not only slows things down but can also lead to corrupt data I think.

I had an old PII 400 box with 128MB of RAM and it would freeze every now and then while I was on the internet. I added 128MB more for 256 MB total and I haven't had a freeze since. I was using 2 old 3 gig HD's with it. UDMA2 33, 40 conductor IDE cables.

hdparm -Tt /dev/hda
cache 152 MB/sec
disk 12MB/sec

Saying that the old slow hard drives caching data caused problems may be a stretch, but after adding more RAM the problem disappeared.

It's is indeed a stretch.

While your observation about Linux liking RAM is very true, stating that swapping leads to disk corruption, isn't.

I have a bunch of meager 128 meg systems "out in the field" as routers running X windows and KDE upon startup for the local admins.

These are lowly Celeron or PII machines with 40 gig drives and dual interfaces.

I've NEVER had a single crash on these machines.

No, swapper problems are indicative of more severe low level problems on the machine.

While it could be due to settings (e.g. pushing the drive DMA and 32 bit modes, while using a "braindead" older WD drive) causing HDD problems and thereby crashing, this is symptomatic of a vendor level problem.

Case in point is WD's infamous problem with higher drive data rates over IDE.

Seems that they disable Error correction on the IDE bus... very dumb.

You can read about this in the kernel documentation.

Also there are clock timer instability problems (again see the kernel problems) with certain VIA and AliMagik chipsets. These lead to the problems with SoundBlasters and TV Tuners (anything which does DMA bus mastering) under Windows as well.

Clearly this isn't a linux problem but rather design flaws which crept in and were to expensive for the manufacturers to ever correct.

VIA and AliMagik both denied any problem and instead issued WinBlows chipset driver hacks to slow down their chipset access speeds to try to diminish the errors.

I'm using a Western Digital HDD JB Series. and I realise it is due to this HDD that my system is crashing.

Are there any solutions on this section?

YES!

And it's very easy.

DO NOT run hdparm at startup on the drive.

10.0 has a entry in /etc/sysconfig, this is a file called harddisks or harddrives.

Edit it and uncomment the DMA line.

This causes the system to boot up using DMA mode but it will use it in SYNC mode.

When you manually enable DMA on the drive you do so in ASYNC mode which suffers from these problems.

Bringing it up in SYNC mode normally does away with HD corruption.

BTW: The reason for this problem is lack of ERROR correction on the WD drives when transmitting data over the IDE bus.

If you are having problems not only should you do the above but you should immediately go out and buy a NEW IDE ribbon cable.

DO NOT under ANY circumstances use rounded IDE cables!

These put the wires too closely together and they induce errors on the lines. Since the WD doesn't correct the error, rounded IDE cables are anathma for the WD drives.

Take your NEW flat 80 pin cable and carefully fold it so that it is ALWAYS up against a metal portion of your case. Keep the areas where it freestands (away from the case metal) as short as possible.

Keep the cable away from the motherboard as well. It should be laided so that it goes away from the motherboard as quickly as possible to minimize noise.



I had exactly the same problems with the WD JB drives and Linux.

After I did the above all the errors stopped.

Thanks for all of the good info.

Thanks for this!
I couldn't find this information for WD hdd anywhere around.

3 Cheers!