Re: [PATCH 0/2] utrace

[Ananth N Mavinakayanahalli <ananth@in.ibm.com>]

On Wed, Aug 27, 2008 at 09:54:53AM -0400, Christoph Hellwig wrote:
> On Tue, Aug 26, 2008 at 08:17:12PM -0400, Frank Ch. Eigler wrote:
> > Among others, utrace is an enablement layer for systemtap user-space
> > probing, through another subsequent part that implements a
> > kprobes-like API for user-space tasks.  All this code now exists in at
> > least prototype form, so if you need to see the bigger picture, look
> > that way.  Other users are anticipated, but first we need to get past
> > the chicken-and-egg.
> 
> As usual nothing of that stuff has any real in-kernel users so the same
> argument applies here.  If it did have real uses it could be merged at
> the same time.  But the current uprobes mess is not a reason to merge
> utrace.

Uprobes is just one user of utrace. It is intended for use for simple
tracing where we need a kernel+userspace look at the problem at hand.
The intention is for use in simple cases (when condition x is met, what
is the value of variable y, etc).

For more advanced tracing though, there are other ideas being proposed
(see ntrace discussions on utrace-devel, but I guess now lkml is the
right place for that discussion too).

However, there are components of uprobes such as breakpoint
insertion/removal and single-stepping infrastructure that are
potentially useful to other userspace debuggers. We are working on
factoring those out to live independent of uprobes. You should be seeing
those patches soon.

Ananth