Re: kallsyms 2.6.8 address ordering

[Robin Holt <holt@sgi.com>]

On Thu, Aug 19, 2004 at 06:10:25PM +0000, jmerkey@comcast.net wrote:
> I've noticed that LKML of late is unresponsive to a lot bug posts and
> that email is being blocked for a lot of folks.  It smells like partisan
> politics based on economic motivations and its not really "open" any
> more when people stoop to this level of behavior.  That aside:

I attribute this to people being over busy right now.

> kallsyms in 2.6.8 is presenting module symbol tables with out of order
> addresses in 2.6.X.  This makes maintaining a commercial kernel debugger
> for Linux 2.6 kernels nighmareish.  Also, the need to kmalloc name strings
> (like kdb does) from kallsyms in kdbsupport.c while IN THE DEBUGGER makes
> it impossible to debug large portions of the kernel code with kdb, so I
> have rewritten large sections of kallsyms.c to handle all these broken,
> brain-dead cases in mdb and I am not relying much on kdb hooks anymore.
> Why on earth does Linux need to have shifting tables of test strings
> for module names requiring all this complexity in the symbol tables
> and kallsyms.

It must be useful for people using small memory footprint machines.
Check with the folks doing embedded stuff.

I remember a discussion about kallsyms and scaling problems with
top reading some /proc/<pid> file.

Look at this:
http://marc.theaimsgroup.com/?l=linux-kernel&m=108758995727517&w=2

> I don't expect a response so I'll keep coding around the broken Linux
> 2.6 code but I wanted to post a record of this so perhaps someone will
> think about over-engineering systems which should be left alone.
> 
> Jeff

Good Luck,
Robin Holt