Hi. This is my story:
My desktop system:
CPU AMD A4-3400 (2.7GHz)
MB: ASrock A75M-HVS
(output lspci attached)
HDD: Hitachi HDS721050CLA362 500GB (SATA)
Slakcware-current, tried versions of 14.1-rc1(3.10.16-smp) and a bit older one(3.8.8-smp kernel).

LILO loads kernel/initrd image successfully, but, when kernel wants to mount root partition it reports absence of root disk. The same problem on Slax-7.0.8 on USB (HDD not detected). Other devices work perfectly.

Found temporary workaround of this problem: a bit earlier I have overclocked CPU from 2.7GHz to 3.2GHz. When I reset overclocking kernel can detect HDD, and boots successfully.

So question: is there any boot-time kernel parameters or build-time configurations which can help kernel to detect my HDD? This is not hardware problem in any case, because WindowsXP does detects HDD in overclocked mode anyway. And I not going to reset overclock every time I boot linux, because I want system work overclocked.

Regards

Attached Files

lspci-t-vv-M-nn-Q-k.txt (24.5 KB, 9 views)

My computer runs at 3.3GHz (not overclocked), and the kernel has no issues with that (so it's not like the kernel can't handle 3.2GHz if the hardware is stable). I'd say that more than likely the situation with your machine is, in fact, a hardware problem due to overclocking, even if XP isn't showing any immediate symptoms.

1 members found this post helpful.

What partitioning method did you use?

The kernel used to have a parameter rootdelay=N where N was the number of seconds for it to sleep before trying to mount the root filesystem.

It is/was for dealing with root devices that were on external storage and might require some spin up time. You can add it to the append= line in /etc/lilo. It may or may not help, worth a try I supposed.

I am inclined toward volkerdi's opinion however, sounds like stability problems related to the over clock. Even if you can make it work, I would not be real excited about running a machine that exhibits quirkiness related to disk I/O.

> What partitioning method did you use?
Old good MBR created by cfdisk long time ago(since 13.0 I think). I don't think it influences problem.

> The kernel used to have a parameter rootdelay=N where N was the number of seconds for it to sleep before trying to mount the root filesystem.

As I said earlier, I tried boot slax(doesn't requires hdd to boot) on this machine, and long after full boot no HDD-related /dev/sd* devices was detected.
But I tried to boot using this additional parameter(rootdelay=10) on both 3.8.8 and 3.10.16 ... not changes detected except notification of waiting in boot log. During boot of windows I do not hear any special spin-up sounds(all spin-ups done during POST of 1st cold boot).
Can anybody describe what mean 4 lines at time 2.228-2.244? I don't like them. (I shall try reset overclock and see if they disappeared).


> a hardware problem due to overclocking,
Looks like for me too, but logic votes for opposite side: same conditions, only software differs, logically problem in software. So I hope there is some soft- solution.

Attached Thumbnails

 

So, I have reset overclocking back to native 2.7GHz and examined log. Now I can see, there is only 3 lines(SATA link down) left(time 2.324), one disappeared(of ata1, my hdd connected to 1st of 6 sata sockets).

And I noticed weird:
"SATA link down" notifications in overclocked mode printed relatively late then in not-overclocked mode, for instance relatively to message "Using IPI No-Shortcut mode" and to messages of net subsystem initialization. I shall search for sources of this message, probably there are some specific parameter that influence behavior.

Attached Thumbnails

 

To follow your own logic: Same condition, only hardware differs, so a hardware problem. How have you actually determined that your system is stable at 3.2GHz?

Well, at least 2-3 month(or even more, cant remember precisely) I work on this PC in overclocked mode on Windows XP, and no malfunction was detected. Well, I consider this stable.

In other words, no stress tests were used to confirm stability and you have malfunctions in Linux when running the system overclocked. I wouldn't consider that stable, but to each his own. Reduce your overclock and see when you get a functioning Linux system, use stress tests to confirm stability.

It's quite possible that it's specific machine instructions which fail consistently because of the overclock, and if XP never executes those instructions (different embedded assembler code, or different compiler with different optimisations, or different strategy for accessing PCI memory, or 32/64 bit) then XP never fails.

As I ran into a big problems to start firewire controller yesterday I can say that the biggest problem is here: