Re: [RFC PATCH 1/2] Marker probes in futex.c

[Nicholas Miell <nmiell@comcast.net>]

On Sat, 2008-04-19 at 14:48 -0400, Frank Ch. Eigler wrote:
> Hi -
> 
> On Sat, Apr 19, 2008 at 08:29:27PM +0200, Peter Zijlstra wrote:
> > [...]
> > > OTOH I think this is a false dichotomy.  Debugging is not only done by
> > > a subsystem maintainer during the merge/rc period.  When something
> > > goes wrong on a deployed machine, problem diagnosis requires data,
> > > which ideally should be gathered as non-intrusively as possible - that
> > > means no recompiling / rebooting, and ideally very little slowdown.
> > > [...]
> 
> > This again tries into the argument about not making markers depend on
> > the code structure or implementation details.
> 
> It should not *unnecessarily* depend on those.
> 
> > I'm really wanting to avoid ever having to be obstructed by a marker. So
> > any marker that does not represent a solid high level event (to take
> > Mathieu's example: a context switch is a context switch, and we'll
> > always have one) I'm not comfortable with merging that upstream.
> 
> It is your prerogative as a subsystem maintainer to make a guess about
> this.  Others may make their own decisions differently, considering
> the small costs and potential benefits.
> 
> > So even though these ad-hoc markers might have some diagnostic value
> > - I'll never support merging them. If a customer might have some
> > issue I can hand him a custom kernel with these markers added in - I
> > see absolutely no reason to burden upstream with these.
> 
> Perhaps the kinds of bugs in your code, coupled with the kinds of
> customers who experience those bugs, make tolerable this means of
> diagnosis (requiring a reboot of their machines into a custom
> debugging kernel).  The customers I have dealt with (and frankly, I
> too) need to diagnose problems on a live running system as much as
> possible.  They don't run every kernel du jour, so they don't need the
> purely hypothetical tools that are dependent on a permanent set of
> markers.
>  

Sun solved this problem for DTrace by applying their stability
attributes to all probe points -- i.e. the probes that represent an ABI
are marked as such and the temporary debugging probes are marked as
unstable.

-- 
Nicholas Miell <nmiell@comcast.net>