Re: the only kernel debugging tools worth learning?

[Suresh Jayaraman <sjayaraman@xxxxxxxxxx>]

On Fri, 2007-09-07 at 12:15 -0400, Robert P. J. Day wrote:

> i'm putting together a short tutorial on this topic, so i wanted to
> be fairly complete but not repetitive.  at the moment, if i had to
> keep things really short, here's what i'd explain:
> 
> 1) using plain "gdb" with vmlinux and /proc/kcore to at least
> *examine* values in a running kernel,
> 
> 2) kprobes
> 
> 3) systemtap

I think it won't be complete without discussing atleast one of the 
crashdump mechanisms e.g. kdump + kexec and analysing with "crash". 
Also, you can touch upon NMI Watchdog which really helps in case 
of Hard lockups.

A reference to debugging options in the "Kernel Hacking" section while
configuring the kernel would help too. e.g. CONFIG_DEBUG_SPINLOCK etc.

> p.s.  it's possible to perhaps include at least a short reference to
> virtualization here, perhaps UML, but *only* as a reference.  actually
> covering how to use it would be beyond the scope of the tutorial.

Yeah, providing reference to UML or QEMU would be nice.

-- 
Suresh Jayaraman



--
To unsubscribe from this list: send an email with
"unsubscribe kernelnewbies" to ecartis@xxxxxxxxxxxx
Please read the FAQ at http://kernelnewbies.org/FAQ