[ale] System.map

[Ben Coleman]

On 18 Jan 2001 09:44:03 CST, scott thomason wrote:

>It's my understanding that with modern kernels (including recent 2.2.x
>releases), System.map is unnecessary unless you're trying to translate kernel
>oopses into symbolic names to make debugging easier. So if you aren't
>planning on debugging kernel problems, System.map is unnecessary. I
>intentionally read up on this in the 2.4 kernel doc, as I have often wondered
>about it. However, I'm only 90% certain. Comments, anyone?

My own reason for copying System.map with a new kernel is that the
first time I compiled a kernel on one machine for use on a different
machine, I checked Slackware's kernel packages to see what they
included.  They included System.map, so I did, too (as a matter of
fact, my preferred procedure nowadays is to create a Slackware-style
package to make installation on the target machine easier.  I also
mimicked the habit of copying the .config file for a kernel to
/boot/Config, which makes it easier to tweak or upgrade a kernel for a
particular machine, because I can simply copy /boot/Config to .config,
and then run make xconfig to make whatever tweaks are necessary.).  If
it seemed like a good idea for Patrick, who was I to argue?

The original poster's problem may very well be that he's got older
versions of ps, top, etc. that don't use /proc, and therefore probably
need System.map to find information in the kernel.  In his case, it's
probably easier in the short term(assuming the files on the machine
where he compiled his kernel are still there) to copy the appropriate
System.map to the target machines than to try to find and upgrade all
of the utilities that look at System.map.

Ben
-- 
Ben Coleman oloryn at mindspring.com    | The attempt to legislatively
http://oloryn.home.mindspring.com/   | micromanage equality results, at
Amateur Radio NJ8J                   | best, in equal misery for all.


--
To unsubscribe: mail majordomo at ale.org with "unsubscribe ale" in message body.