Massive IDE problems. Who leaves data here?

Massive IDE problems. Who leaves data here?

[Manuel Reimer]

Hello,

anything started with a try to burn Slackware 12.0 from the original DVD 
to an new medium with different boot settings. I always got corrupted 
results and didn't know why.

So I started with an "md5sum -c CHECKSUMS.md5" directly on the original 
media. This resulted in "anything OK".

Now I copied the whole DVD to my hard drive and created an ISO from it. 
I mounted the ISO locally and my md5sum now results in 5 corrupted files.

--> A Bug in mkisofs?

No, unfortunately not, as a md5sum on the copy, I have created from the 
original DVD by using "cp -vr" is corrupted, too!

So md5sum on the original DVD is OK, but after copying to my hard drive, 
several files are corrupted.

I'm using kernel 2.6.21.5. Distribution is Slackware 12.0
All my "partitions" are LVs in LVM2

I also updated the kernel to 2.6.23.12 to test with this one, but I 
still get corrupted files.

Is this a LVM bug? Do I already have a corrupted LVM filesystem? How to 
check/fix it? Is this a known kernel bug? Which may be the reason for 
corrupted files?

I've created a backup of my important data to a second disc to a "real 
ext2 partition" (without LVM), but this is connected to the same IDE 
controller and I don't even know if I may still trust my mainboard...

I also get those kernel messages via dmesg:

http://pastebin.org/16537

Thank you very much in advance for any help!

Yours

Manuel

Re: Massive IDE problems. Who leaves data here?

[Jeff Garzik]

Manuel Reimer wrote:
> Hello,
> 
> anything started with a try to burn Slackware 12.0 from the original DVD 
> to an new medium with different boot settings. I always got corrupted 
> results and didn't know why.
> 
> So I started with an "md5sum -c CHECKSUMS.md5" directly on the original 
> media. This resulted in "anything OK".
> 
> Now I copied the whole DVD to my hard drive and created an ISO from it. 
> I mounted the ISO locally and my md5sum now results in 5 corrupted files.
> 
> --> A Bug in mkisofs?
> 
> No, unfortunately not, as a md5sum on the copy, I have created from the 
> original DVD by using "cp -vr" is corrupted, too!
> 
> So md5sum on the original DVD is OK, but after copying to my hard drive, 
> several files are corrupted.
> 
> I'm using kernel 2.6.21.5. Distribution is Slackware 12.0
> All my "partitions" are LVs in LVM2
> 
> I also updated the kernel to 2.6.23.12 to test with this one, but I 
> still get corrupted files.
> 
> Is this a LVM bug? Do I already have a corrupted LVM filesystem? How to 
> check/fix it? Is this a known kernel bug? Which may be the reason for 
> corrupted files?
> 
> I've created a backup of my important data to a second disc to a "real 
> ext2 partition" (without LVM), but this is connected to the same IDE 
> controller and I don't even know if I may still trust my mainboard...
> 
> I also get those kernel messages via dmesg:
> 
> http://pastebin.org/16537

If your IDE interface is complaining about BadCRC errors, then it's 
complaining about hardware problems (bad cable, etc.)

	Jeff

Re: Massive IDE problems. Who leaves data here?

[Manuel Reimer]

Jeff Garzik wrote:
> If your IDE interface is complaining about BadCRC errors, then it's 
> complaining about hardware problems (bad cable, etc.)

The cable already has been replaced three times. I even got sure to 
*not* bend the cable. This all doesn't help.

May a bad cable really cause corrupted data?

I also have such messages for my cdrom drive (the bad sector messages at 
the end of the log for drive hdc). I replaced the 40 pin cable with a 
new 80 pin one and still have the errors...

Yours

Manuel

Re: Massive IDE problems. Who leaves data here?

[Ray Lee]

(Please always do a reply-to-all for this email list.)

On Jan 22, 2008 12:40 PM, Manuel Reimer <Manuel.Spam@nurfuerspam.de> wrote:
> Jeff Garzik wrote:
> > If your IDE interface is complaining about BadCRC errors, then it's
> > complaining about hardware problems (bad cable, etc.)
>
> The cable already has been replaced three times. I even got sure to
> *not* bend the cable. This all doesn't help.
>
> May a bad cable really cause corrupted data?
>
> I also have such messages for my cdrom drive (the bad sector messages at
> the end of the log for drive hdc). I replaced the 40 pin cable with a
> new 80 pin one and still have the errors...

Bad or almost bad power supplies can cause lots of unhappy problems
such as these. If you have another power supply laying around, it can
be worth swapping them out. Double-checking that your CPU fan is still
spinning, and that memtest86+ doesn't show any errors doesn't hurt
either.

Re: Massive IDE problems. Who leaves data here?

[Manuel Reimer]

Ray Lee wrote:
> (Please always do a reply-to-all for this email list.)

Currently I don't have a SMTP server configured. As soon as my system is 
trustworthy, again, I'll do that.

> Bad or almost bad power supplies can cause lots of unhappy problems
> such as these. If you have another power supply laying around, it can
> be worth swapping them out. Double-checking that your CPU fan is still
> spinning, and that memtest86+ doesn't show any errors doesn't hurt
> either.

... that may be possible...

I have all PCI slots filled with cards, two big IDE hard drives and one 
DVD RAM writer. I already disconnected the DVD drive, as this was nearly 
unusable with all those IDE errors. The power supply is (AFAIR) a 200 
watt one.

But may the power supply cause corrupted data? Shouldn't that crash down 
the PC completely?

CU

Manuel

Re: Massive IDE problems. Who leaves data here?

[Ray Lee]

On Jan 22, 2008 1:04 PM, Manuel Reimer <Manuel.Spam@nurfuerspam.de> wrote:
> Ray Lee wrote:
> > (Please always do a reply-to-all for this email list.)
>
> Currently I don't have a SMTP server configured. As soon as my system is
> trustworthy, again, I'll do that.

Oy. Just know that without CC:ing people, I'm having to add Jeff back
in by hand, and we may not notice your messages. Please, with whatever
you're using to send email, CC: people that you want to read your
messages, okay? LKML gets a lot of messages, and it's easy to miss
some.

> > Bad or almost bad power supplies can cause lots of unhappy problems
> > such as these. If you have another power supply laying around, it can
> > be worth swapping them out. Double-checking that your CPU fan is still
> > spinning, and that memtest86+ doesn't show any errors doesn't hurt
> > either.
>
> ... that may be possible...
>
> I have all PCI slots filled with cards, two big IDE hard drives and one
> DVD RAM writer. I already disconnected the DVD drive, as this was nearly
> unusable with all those IDE errors. The power supply is (AFAIR) a 200
> watt one.
>
> But may the power supply cause corrupted data? Shouldn't that crash down
> the PC completely?

No, not in the least. When power supplies are on the edge and
overloaded, the quality of power drops. 12V lines may only supply 11
or less, etc. Some components in your computer will deal with it fine,
others won't.

In general, if you have weird hardware problems and you're ruled out
memory and overheating, the power supply is almost always the next
thing to check.

Re: Massive IDE problems. Who leaves data here?

[Alan Cox]

On Tue, 22 Jan 2008 21:40:21 +0100
Manuel Reimer <Manuel.Spam@nurfuerspam.de> wrote:

> Jeff Garzik wrote:
> > If your IDE interface is complaining about BadCRC errors, then it's 
> > complaining about hardware problems (bad cable, etc.)
> 
> The cable already has been replaced three times. I even got sure to 
> *not* bend the cable. This all doesn't help.
> 
> May a bad cable really cause corrupted data?

Yes, or dodgy connectors or other problems. The CRC itself is computed by
the hardware each end so a BadCRC error really means the two ends
disagree about the data. 

That is *usually* a cable/electrical noise problem. It can be power and
it can occasionally show up if the software misprograms the timing so the
two ends are a bit out of sync.

Alan

Re: Massive IDE problems. Who leaves data here?

[Jan Engelhardt]

On Jan 22 2008 22:04, Manuel Reimer wrote:
>
> ... that may be possible...
>
> I have all PCI slots filled with cards, two big IDE hard drives and one DVD RAM
> writer. I already disconnected the DVD drive, as this was nearly unusable with
> all those IDE errors. The power supply is (AFAIR) a 200 watt one.

Ugh, 250 W was a safe^[0] minimum even back in '98 when harddisks were
like 4.3GB and a dual-speed CD-RW drive was like $400 and up. You
seriously need to replace that.

[0] e.g. suited for 2 CD drives and 2 harddisks, 2 PCI and 1 AGP

Re: Massive IDE problems. Who leaves data here?

[Bill Davidsen]

Manuel Reimer wrote:
> Hello,
> 
> anything started with a try to burn Slackware 12.0 from the original DVD 
> to an new medium with different boot settings. I always got corrupted 
> results and didn't know why.
> 
> So I started with an "md5sum -c CHECKSUMS.md5" directly on the original 
> media. This resulted in "anything OK".
> 
> Now I copied the whole DVD to my hard drive and created an ISO from it. 
> I mounted the ISO locally and my md5sum now results in 5 corrupted files.
> 
> --> A Bug in mkisofs?
> 
> No, unfortunately not, as a md5sum on the copy, I have created from the 
> original DVD by using "cp -vr" is corrupted, too!
> 
Possibly a known kernel problem, you may have read past the end of data 
into the pad sectors of the DVD and gotten garbage at the end of the ISO 
image. Use isoinfo to determine the correct size of the ISO filesystem, 
and compare. You can try setting readahead on the DVD reader to zero 
with blockdev.

If the file is smaller, other bug, if readahead hit EOF it returns no 
data instead of a short read, the blockdev fix should handle that as 
well. This was supposed to be fixed in recent kernels, that may be true.

I suggest the cdwrite@other.debian.org mailing list is a better forum 
for CD/DVD/BR problems, good technical people, unfortunately with 
personal agendas in some cases.

> So md5sum on the original DVD is OK, but after copying to my hard drive, 
> several files are corrupted.
> 
That's odd, I would expect the data on the disk to just be the wrong 
size, and get a CRC on that. You might also use readcd to pull the data, 
that almost always does what it should.

> I'm using kernel 2.6.21.5. Distribution is Slackware 12.0
> All my "partitions" are LVs in LVM2
> 
> I also updated the kernel to 2.6.23.12 to test with this one, but I 
> still get corrupted files.
> 
> Is this a LVM bug? Do I already have a corrupted LVM filesystem? How to 
> check/fix it? Is this a known kernel bug? Which may be the reason for 
> corrupted files?
> 
> I've created a backup of my important data to a second disc to a "real 
> ext2 partition" (without LVM), but this is connected to the same IDE 
> controller and I don't even know if I may still trust my mainboard...
> 
> I also get those kernel messages via dmesg:
> 
> http://pastebin.org/16537
> 
Could be anything, in no order dirty lens, bad drive, bad DVD, firmware 
error, cable, power supply, acpi confused... could even be a poorly 
handled end of data on the DVD. Not enough info for me to tell, for 
sure. Trying readcd is cheap, turning off readahead on the DVD drive is 
easy, if the problem persists you probably want to take it to the 
mailing list.

> Thank you very much in advance for any help!
> 
I'm not sure I helped, but you now have more and better things about 
which to be confused.  ;-)

> Yours
> 
> Manuel
> 


-- 
Bill Davidsen <davidsen@tmr.com>
   "We have more to fear from the bungling of the incompetent than from
the machinations of the wicked."  - from Slashdot

Re: Massive IDE problems. Who leaves data here?

[Manuel Reimer]

Ray Lee wrote:
> Oy. Just know that without CC:ing people, I'm having to add Jeff back
> in by hand, and we may not notice your messages. Please, with whatever
> you're using to send email, CC: people that you want to read your
> messages, okay? LKML gets a lot of messages, and it's easy to miss
> some.

Trying to do this with this message...

> No, not in the least. When power supplies are on the edge and
> overloaded, the quality of power drops. 12V lines may only supply 11
> or less, etc. Some components in your computer will deal with it fine,
> others won't.
> 
> In general, if you have weird hardware problems and you're ruled out
> memory and overheating, the power supply is almost always the next
> thing to check.

I didn't rule out "memory" for sure. I'll try "memcheck" this evening.

Yours

Manuel

Re: Massive IDE problems. Who leaves data here?

[Manuel Reimer]

Ray Lee wrote:
> Oy. Just know that without CC:ing people, I'm having to add Jeff back
> in by hand, and we may not notice your messages. Please, with whatever
> you're using to send email, CC: people that you want to read your
> messages, okay? LKML gets a lot of messages, and it's easy to miss
> some.

Trying to do this with this message...

> No, not in the least. When power supplies are on the edge and
> overloaded, the quality of power drops. 12V lines may only supply 11
> or less, etc. Some components in your computer will deal with it fine,
> others won't.
> 
> In general, if you have weird hardware problems and you're ruled out
> memory and overheating, the power supply is almost always the next
> thing to check.

I didn't rule out "memory" for sure. I'll try "memcheck" this evening.

Yours

Manuel

Re: Massive IDE problems. Who leaves data here?

[Manuel Reimer]

Jan Engelhardt wrote:
> Ugh, 250 W was a safe^[0] minimum even back in '98 when harddisks were
> like 4.3GB and a dual-speed CD-RW drive was like $400 and up. You
> seriously need to replace that.
> 
> [0] e.g. suited for 2 CD drives and 2 harddisks, 2 PCI and 1 AGP

I replaced with an 400 Watt one for my tests, but this didn't change 
anything.

Yours

Manuel

Re: Massive IDE problems. Who leaves data here?

[Manuel Reimer]

Alan Cox wrote:
> Yes, or dodgy connectors or other problems. The CRC itself is computed by
> the hardware each end so a BadCRC error really means the two ends
> disagree about the data. 

I'll have a closer look at the connectors, but as I replaced the whole 
cable for the fourth time, today, I don't think, that the connectors are 
the reason. Plugging in and out should have cleaned the connectors.

> That is *usually* a cable/electrical noise problem. It can be power and
> it can occasionally show up if the software misprograms the timing so the
> two ends are a bit out of sync.

Even if it *would* be a timing problem, I would be unable to debug or 
fix this.

Yours

Manuel