Re: Task watchers v2

[Matt Helsley <matthltc@us.ibm.com>]

On Fri, 2006-12-15 at 09:34 +0100, Christoph Hellwig wrote:
> On Thu, Dec 14, 2006 at 04:07:55PM -0800, Matt Helsley wrote:
> > Associate function calls with significant events in a task's lifetime much like
> > we handle kernel and module init/exit functions. This creates a table for each
> > of the following events in the task_watchers_table ELF section:
> >
> > WATCH_TASK_INIT at the beginning of a fork/clone system call when the
> > new task struct first becomes available.
> >
> > WATCH_TASK_CLONE just before returning successfully from a fork/clone.
> >
> > WATCH_TASK_EXEC just before successfully returning from the exec
> > system call.
> >
> > WATCH_TASK_UID every time a task's real or effective user id changes.
> >
> > WATCH_TASK_GID every time a task's real or effective group id changes.
> >
> > WATCH_TASK_EXIT at the beginning of do_exit when a task is exiting
> > for any reason.
> >
> > WATCH_TASK_FREE is called before critical task structures like
> > the mm_struct become inaccessible and the task is subsequently freed.
> >
> > The next patch will add a debugfs interface for measuring fork and exit rates
> > which can be used to calculate the overhead of the task watcher infrastructure.
> 
> What's the point of the ELF hackery? This code would be a lot simpler
> and more understandable if you simply had task_watcher_ops and a
> register / unregister function for it.

A bit more verbose response:

	I posted a notifier chain implementation back in June that bears some
resemblance to your suggestion -- a structure needed to be registered at
runtime. There was a single global list of them to iterate over for each
event.

	This patch and the following patches are significantly shorter than
their counterparts in that series. They avoid iterating over elements
with empty ops. The way function pointers and function bodies are
grouped together by this series should improve locality. The fact that
there's no locking required also makes it simpler to analyze and use.

	The patches to allow modules to register task watchers does make things
more complex though -- that does require a list and a lock. However, the
lock does not need to be taken in the fork/exec/etc paths if we pin the
module. In contrast your suggested approach is simpler because it
doesn't treat modules any differently. However overall I think the
balance still favors these patches.

Cheers,
	-Matt Helsley