Re: [patch 02/17] Kernel Tracepoints

[Steven Rostedt]

On Thu, 24 Jul 2008, Mathieu Desnoyers wrote:
> > > +
> > > +	if (nr_probes - nr_del == 0) {
> > > +		/* N -> 0, (N > 1) */
> > > +		entry->funcs = NULL;
> > > +		entry->refcount = 0;
> > > +		debug_print_probes(entry);
> > > +		return old;
> > > +	} else {
> > > +		int j = 0;
> > > +		/* N -> M, (N > 1, M > 0) */
> > > +		/* + 1 for NULL */
> > > +		new = kzalloc((nr_probes - nr_del + 1)
> > > +			* sizeof(void *), GFP_KERNEL);
> > > +		if (new == NULL)
> > > +			return ERR_PTR(-ENOMEM);
> >
> > Hmm, on failure of allocating a new array, we could simply use the
> > old array, and remove the one probe from it instead of just failing.
> >
>
> Nay, because of RCU constraints. So we have the readers in the current
> RCU window who need to see the old version, and readers of the following
> window who need to see the next version. Both can live at the same time
> on the system. We cannot reuse the same memory to perform the array
> shrink without corrupting the data seen by the previous readers. We
> really have to perform a copy here.

Ah, good point. I forgot the whole RCU factor here.

-- Steve

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/