Re: [PATCH 2/3] of: Introduce safe accessors for node->data

[Anton Vorontsov <avorontsov@ru.mvista.com>]

On Tue, Feb 09, 2010 at 10:25:22AM -0700, Grant Likely wrote:
> On Fri, Feb 5, 2010 at 1:50 PM, Anton Vorontsov
> <avorontsov@ru.mvista.com> wrote:
> > Platform code use node->data to store some private information
> > associated with a node.
> >
> > Previously there was no need for any locks and accessors since we were
> > initializing the data mostly at boot time and never modified it later.
> >
> > Though, nowadays OF GPIO infrastructure supports GPIO chips detaching,
> > so to handle this correctly we have to introduce locking for the
> > node->data field.
> 
> I'm not convinced this is needed.  What's wrong with using the
> whole-tree devtree_lock?

Why are you concerned? It doesn't add much of any footprint.

$ grep -c { -r arch/powerpc/boot/dts/ | cut -d: -f2 | sort -n | tail -n1
84

So far we have max 84 nodes, so it's a few hundreds of bytes for all
the dev tree.

Anyway, yes, we can use the devtree lock. Though, this will require a
bit more modifications, and I'm not sure if it's a great idea in
general (i.e. using the global lock in contrast to fine grained
locking).

The thing is that you can't use most of the of_ functions when you
hold the devtree lock (IIRC, rwlock has the same restrictions as a
spinlock, so you can't nest these locks).

I can try to rework OF GPIO calls so that they won't require of_
calls when they hold the lock, and let's see how it'll look like.

Thanks!

-- 
Anton Vorontsov
email: cbouatmailru@gmail.com
irc://irc.freenode.net/bd2