hook to run on attach/exec?

hook to run on attach/exec?

[Roland McGrath]

I read someone mention somewhere recently that a new hook might be in the
works to be the spot where PIE support would figure out where the
executable got loaded.  Can someone give me the skinny?

For vsyscall DSO detection, the hook I need goes at the same place, I
think.  That is, a spot that gets run after an inferior execs (from "run"
or otherwise, skipping first "run" exec for the shell of course) or when
gdb attaches to an inferior or to a core dump image; i.e., whenever the
inferior's address space is freshly made or presumed to be.  Those are the
times you need to figure out where a PIE actually resides in memory, and
also the times you need to figure out where a system-supplied DSO is.

Last time I was looking into the vsyscall DSO support, it seemed like there
was no precisely appropriate hook already.  But I didn't know exactly what
to do about adding one.  Is my life easier now?


Thanks,
Roland

Re: hook to run on attach/exec?

[Daniel Jacobowitz]

On Wed, Jan 28, 2004 at 08:08:31PM -0800, Roland McGrath wrote:
> I read someone mention somewhere recently that a new hook might be in the
> works to be the spot where PIE support would figure out where the
> executable got loaded.  Can someone give me the skinny?
> 
> For vsyscall DSO detection, the hook I need goes at the same place, I
> think.  That is, a spot that gets run after an inferior execs (from "run"
> or otherwise, skipping first "run" exec for the shell of course) or when
> gdb attaches to an inferior or to a core dump image; i.e., whenever the
> inferior's address space is freshly made or presumed to be.  Those are the
> times you need to figure out where a PIE actually resides in memory, and
> also the times you need to figure out where a system-supplied DSO is.
> 
> Last time I was looking into the vsyscall DSO support, it seemed like there
> was no precisely appropriate hook already.  But I didn't know exactly what
> to do about adding one.  Is my life easier now?

Well, some PIE work is checked in on a branch; you can take a look at
it and see.  I don't remember whether Elena added such a hook.  Right
now, post_attach and post_startup_inferior for various targets together
cover all the cases; that's what I'm using in my port of your patches
to 6.0.

Should this be a target hook, or an observer?

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer