Re: [PATCH] fdset's leakage

[Andrew Morton]

On Tue, 11 Jul 2006 13:05:03 +0400
Kirill Korotaev <dev@xxxxxxxxxx> wrote:

> Andrew,
> 
> > But the code in there is really sick.   In all cases we do:
> > 
> > 	free_fdset(foo->open_fds, foo->max_fdset);
> > 	free_fdset(foo->close_on_exec, foo->max_fdset);
> > 
> > How much neater and more reliable would it be to do:
> > 
> > 	free_fdsets(foo);
> > 
> > ?
> agree. should I prepare a patch?

Is OK, I'll take care of it later.  We want to let your current patch bake
as-is in mainline for a while so that we can backport it into 2.6.17.x with
more confidence.  That's a bit excessive in this case, but the principle is
good.

> > Also,
> > 
> > 	nfds = NR_OPEN_DEFAULT;
> > 	/*
> > 	 * Expand to the max in easy steps, and keep expanding it until
> > 	 * we have enough for the requested fd array size.
> > 	 */
> > 	do {
> > #if NR_OPEN_DEFAULT < 256
> > 		if (nfds < 256)
> > 			nfds = 256;
> > 		else
> > #endif
> > 		if (nfds < (PAGE_SIZE / sizeof(struct file *)))
> > 			nfds = PAGE_SIZE / sizeof(struct file *);
> > 		else {
> > 			nfds = nfds * 2;
> > 			if (nfds > NR_OPEN)
> > 				nfds = NR_OPEN;
> >   		}
> > 	} while (nfds <= nr);
> > 
> > 
> > That's going to take a long time to compute if nr > NR_OPEN.  I just fixed
> > a similar infinite loop in this function.  Methinks this
> > 
> > 	nfds = max(NR_OPEN_DEFAULT, 256);
> > 	nfds = max(nfds, PAGE_SIZE/sizeof(struct file *));
> > 	nfds = max(nfds, round_up_pow_of_two(nr + 1));
> > 	nfds = min(nfds, NR_OPEN);
> > 
> > is clearer and less buggy.  I _think_ it's also equivalent (as long as
> > NR_OPEN>256).  But please check my logic.
> Yeah, I also noticed these nasty loops but was too lazy to bother :)
> Too much crap for my nerves :)
> 
> Your logic looks fine for me.

I usually get that stuff wrong.

> Do we have already round_up_pow_of_two() function

yep, in kernel.h.

-
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/