On Thursday 11 July 2002 01:36, Jesse Barnes wrote:
> On Thu, Jul 11, 2002 at 12:24:06AM +0200, Daniel Phillips wrote:
> > Acme, which is to replace all those above-the-function lock coverage
> > comments with assert-like thingies:
> >
> > spin_assert(&pagemap_lru_lock);
> >
> > And everbody knows what that does: when compiled with no spinlock
> > debugging it does nothing, but with spinlock debugging enabled, it oopses
> > unless pagemap_lru_lock is held at that point in the code. The practical
> > effect of this is that lots of 3 line comments get replaced with a
> > one line assert that actually does something useful. That is, besides
> > documenting the lock coverage, this thing will actually check to see if
> > you're telling the truth, if you ask it to.
> >
> > Oh, and they will stay up to date much better than the comments do,
> > because nobody needs to to be an ueber-hacker to turn on the option and
> > post any resulting oopses to lkml.
>
> Sounds like a great idea to me. Were you thinking of something along
> the lines of what I have below or perhaps something more
> sophisticated? I suppose it would be helpful to have the name of the
> lock in addition to the file and line number...
I was thinking of something as simple as:
#define spin_assert_locked(LOCK) BUG_ON(!spin_is_locked(LOCK))
but in truth I'd be happy regardless of the internal implementation. A note
on names: Linus likes to shout the names of his BUG macros. I've never been
one for shouting, but it's not my kernel, and anyway, I'm happy he now likes
asserts. I bet he'd like it more spelled like this though:
MUST_HOLD(&lock);
And, dare I say it, what I'd *really* like to happen when the thing triggers
is to get dropped into kdb. Ah well, perhaps in a parallel universe...
When one of these things triggers I do think you want everything to come to
a screeching halt, since, to misquote Matrix, "you're already dead", and you
don't want any one-per-year warnings to slip off into the gloomy depths of
some forgotten log file.
-- Daniel - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Mon Jul 15 2002 - 22:00:19 EST