can kprobes be modular?

[Jim Keniston]

On Mon, 2007-02-26 at 18:10, Roland McGrath wrote:
> I was reading over some things, and it occurred to me that kprobes ought to
> be a loadable kernel module.  I don't have any special motivation for this.
> It just seems like an unclean situation that it can't be a module now.
> Perhaps many kernels will want to build it in anyway, but I can't see why
> it isn't a module.  It's not very big, but neither are many other things
> that are used much more often and are built as modules.
> 
> The #ifdef CONFIG_KPROBES sections in e.g. arch/i386/kernel/traps.c look to
> me like things that ought to be enabled unconditionally, so kprobes or any
> other module could use them.  Things like register_page_fault_notifier
> ought to just be enabled and exported by default.
> 
> Thoughts?

Well, there's at least one call to kprobes from deep in the kernel:
finish_task_switch() calls kprobe_flush_task().

There may be other gotchas.  As far as I know, nobody has investigated
the idea for quite a while.

Jim

> 
> 
> Thanks,
> Roland