Re: [PATCH, take4] FUTEX : new PRIVATE futexes

[Nick Piggin <nickpiggin@yahoo.com.au>]

Eric Dumazet wrote:
> On Wed, 11 Apr 2007 17:22:57 +1000
> Nick Piggin <nickpiggin@yahoo.com.au> wrote:
>
>
> > Eric Dumazet wrote:
> >
> > > On Sat, 07 Apr 2007 19:30:14 +1000
> > > Nick Piggin <nickpiggin@yahoo.com.au> wrote:
> > >
> > >
> > >
> > > > Eric Dumazet wrote:
> > >
> > >
> > >
> > >
> > > > > - Current mm code have a problem with 64bit futexes, as spoted by Nick :
> > > > >
> > > > > get_futex_key() does a check against sizeof(u32) regardless of futex being 64bits or not.
> > > > > So it is possible a 64bit futex spans two pages of memory...
> > > > > I had to change get_futex_key() prototype to be able to do a correct test.
> > > >
> > > > I wonder if it should be encfocing alignment to keep in on 1 page?
> > >
> > >
> > > I believe I just did that :)
> >
> > Yes :P What I was trying to say before jumping on a plane is that
> > sys_futex/sys_futex64 calls should each check their own address alignment, so
> > the deeper parts of the call stack always know alignment is correct.
> >
> > This will remove all the fsize you pass around, and also sanitise the userspace
> > argument much higher in the call stack, which is very preferable and more
> > conventional.
> >
> > Maybe this isn't possible (it's very obvious, so there may be a good reason it
> > hasn't been done).
>
>
> I had this idea as well, but considering get_futex_key() is exported in include/linux/futex.h, I believe some out-of tree thing is using it.

You're changing the API anyway, so just get rid of the declaration from futex.h
(that doesn't actually make for an export anyway, unless it is inline).

But... that isn't there in mainline. Why is it in -mm? At any rate, that makes
it a no brainer to change.

> As this external thing certainly is not doing the check itself, to be on the safe side we should enforce it in get_futex_key(). I agree with you : If we want to maximize performance, we could say : The check *must* be done by the caller.

Well we _control_ the API, so let's make it as clean and performant as possible
from the start.

--
SUSE Labs, Novell Inc.
-
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/