On Wed, Sep 26, 2012 at 8:33 AM, Robert P. J. Day <rpjday@crashcourse.ca> wrote:
in my online travels yesterday, i ran across this gem, "Linux Kernel Crash Book":
http://www.dedoimedo.com/computers/crash-book.html
and am now wondering about what would constitute a reasonable (and minimal?) list of canonical kernel debugging tools.
first, while the book above covers Linux Kernel Crash Dump (LKCD), the author freely admits that it's been pretty much obsoleted by the more recent and flexible kdump, so there seems to be little value in digging into LKCD (or, in my case, adding any coverage of it to a kernel debugging course, which i am currently designing).
next, someone else's course i'm teaching next week has a kernel debugging chapter which opens with netdump and diskdump before moving onto kdump and kexec, but those earlier utilities are *also* deprecated these days,
http://serverfault.com/questions/181554/how-should-i-capture-linux-kernel-pa...
so i would be tempted to skip any coverage of netdump and diskdump in favour of additional and more advanced coverage of kdump and kexec.
along those lines, i'm just digging into ftrace and was wondering if it in any way obsoleted systemtap, but i've heard from more than one source that while ftrace is allegedly more powerful, systemtap still has its place and is worth talking about.
While learning ftrace I found there's little good documentation for ftrace. Ftrace is huge and very flexible, you can do lot more than just trace your functions. I firmly believe ftrace + perf should be consider like a serious debugging technique. Regards, Ezequiel.