[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Pinpointing faulty kernel driver?



On Apr 5, 2012 16:12 "Camaleón" <noelamac@gmail.com> wrote:



> You can send the output to another machine using a serial cable and
instructing the kernel to dump the message there. I had to do this years
> ago to debug a kernel crash from a VM but to be sincere, I doubt I can
nowadays repeat that milestone, I don't remember the steps :-)
>
> Look, Ubuntu has some good doc about this:
>
> Kernel Debugging Tricks
> <https://wiki.ubuntu.com/Kernel/KernelDebuggingTricks>
>
>

Thank you for this reference! It is indeed an awesome reference. I tried
getting netconsole to work, because I don't have a serial-to-USB cable. I
will be looking into this problem over the next few days. Perhaps I will
have to buy such a cable.

There is also, according to the InitramfsDebug Debian wiki page, a way to
get log data in /dev/.initramfs/initramfs.debug. I can not make this work
either, for some reason.

I tried the boot_delay option, but the delay seems to revert to full speed
at a certain point in the boot sequence, so it is no use.
There is odd behavior when the crash happens: The visible screen area
seems to "scroll back" so that when the hang occurs, the timestamps show
approximately 64.xxxxxx seconds (but actually, the crash occurs at
68.xxxxxx seconds).

On a hunch, I tried setting acpi=off - and now the new kernel boots! Of
course, this is a sub-optimal solution, so I got dmesg listings from both
the old and the new kernel.

The diff is here: http://pastebin.com/L4YXTmJh

I would be very grateful if you'd take a look.

Thanks,

Lars
On Apr 5, 2012 16:12 "Camaleón" <noelamac@gmail.com> wrote:


You can send the output to another machine using a serial cable and
instructing the kernel to dump the message there. I had to do this years
ago to debug a kernel crash from a VM but to be sincere, I doubt I can
nowadays repeat that milestone, I don't remember the steps :-)

Look, Ubuntu has some good doc about this:

Kernel Debugging Tricks
https://wiki.ubuntu.com/Kernel/KernelDebuggingTricks

Thank you for this reference! It is indeed an awesome reference. I tried getting netconsole to work, because I don't have a serial-to-USB cable. I will be looking into this problem over the next few days.

There is also, according to the InitramfsDebug Debian wiki page, a way to get log data in /dev/.initramfs/initramfs.debug. I can not make this work either, for some reason.

The easiest way (which will allow you to read the screen) would be
slowing the kernel output messages, I would start from there.
I am running the latest packages in both kernel series, and the 3.1
series does not have the problem.

Are you using any special kernel module that may require to be re-
compiled (as for example, the VGA driver) or any special module for the
hard disk controller?

My list of loaded modules follows (taken from 3.1.x). I have had some
complaints about missing wifi firmware for a while (this is a desktop
box, so I don't care about it) even though I have both
linux-firmware-(non)free packages installed.

A message warning for a missing firmware could render that device
inoperable but nothing more.
Your help is very much appreciated.

(...)

Thanks for the module listing :-)

It seems that you're using a usual SATA configuration at the BIOS (ahci),
I mean, nothing "fancy" that may require for specific hard disk
controller module to be present and thus preventing the system from
booting normally.

Anyway, you can check if by appending "rootdelay=9" to the kernel line at
GRUB's boot menu you can get rid of the "waiting... to be fully
populated" message.

Greetings,

Reply to: