I currently store all of my kernels and System.map files
in /boot. So I have zImage-2.0.24 and a corresponding
System.map-2.0.24. And I use (in rc.local):
VERSION=`/bin/uname -a | /usr/bin/cut -d\ -f3`
in my rc.local to access the 'right' map (and create /etc/issue).
I like the idea of making a link to /System.map such that
it would always be the right one, and immediately ran into
the problem where if I apply the reverse patch to sock.c to
profile the result, I have two 2.0.24 kernels. I'm thinking
about something like using the following:
VEXT=`/bin/uname -a | /usr/bin/md5sum | cut -b1-4`
and having System.map-${VEXT} files in /boot.
Then a:
ln -sf /boot/System.map-${VEXT} /System.map
should get the right file (if 4 digits is enough to always
be unique).
Two questions:
1. Is there a standard way to get the exact uname info
from an unbooted kernel image so I know what VEXT is
before rebooting ( to install System.map-xxxx)? I can
find the string in my /usr/src/linux/vmlinux file.
2. Is there a way to know what boot image was booted
in a running system? I could just run the kernel image
through md5sum to get an extension.
Maybe there's a better way to get a unique identifier. Something
that can be determined before and after booting a new kernel.
Thanks,
Robert
-- mailto:rwuest@ix.netcom.com Empowered by Linux.