Subject: Re: ksyms patches.
To: Jason Thorpe <>
From: None <>
List: tech-kern
Date: 04/21/2003 19:11:04
> On Monday, April 21, 2003, at 07:01  AM, <> wrote:
> > Comments?
> And now for comments on the code :-)
> In addsymtab(), I think, in fact, the shstrtab is loaded ... see 
> db_elf.c for how the .symtab and .strtab sections are linked up.  I 
> think you'll want to duplicate that algorithm.
Yeah, good idea. There are some funny fiddling with the headers
that must be obeyed, I see.

> In general, it would be nice if locking were in place in this code.  
> Once less thing to fix later :-)  I don't think that's a barrier to 
> this code being checked in, though ... mostly a "would be nice" :-)
I will do that, but not right now. I'll deal with that when starting
to fix some other internals of it, like the fast-lookup stuff.

> What's the deal with NKSYMS?  I'd say if kern_ksyms.c is built at all 
> that /dev/ksyms should be mandatory.
May be true. It was for the case if say DDB is defined and 
no pseudo-device ksyms. This is probably only of academic interest.

> Other than the (pretty minor :-) comments above, this looks great!  
> Thanks a ton for doing this (this is also going to make embedded 
> systems a LOT happier :-)
Thanks, it has been in the closet for too many years, fortunately
it hadn't bit-rotten too much :-)

-- Ragge