Re: [PATCH 0/4] futex: get_user_pages_fast() for shared futexes

From: Peter Zijlstra
Date: Tue Sep 30 2008 - 07:16:53 EST


On Tue, 2008-09-30 at 12:55 +0200, Eric Dumazet wrote:
> Peter Zijlstra a Ãcrit :

> > On a regular modern Linux system, not much. But I've been told there are
> > applications out there that do indeed make heavy use of them - as
> > they're part of POSIX etc.. blah blah :-)
>
> inter-process futexes are still used for pthread creation/join
> (aka clear_child_tid / CLONE_CHILD_CLEARTID)
>
> kernel/fork.c, functions mm_release() & sys_set_tid_address()
>
> I am not sure how it could be converted to private futexes, since
> old binaries (static glibc) will use FUTEX_WAKE like calls.

Ah, thanks, didn't know that.

> > ---
> > Subject: futex: fixup get_futex_key() for private futexes
> > From: Peter Zijlstra <a.p.zijlstra@xxxxxxxxx>
> >
> > With the get_user_pages_fast() patches we made get_futex_key() obtain a
> > reference on the returned key, but failed to do so for private futexes.
> >
>
> Sorry I am lost...
> private futexes dont need to get references at all...

Ah, right - its a NOP, that's why it didn't show up in testing.

The thing is, I changed the semantics of get_futex_key() to return a key
with reference taken. And then noticed I didn't take one in the private
futex path, and failed to notice the ref ops are nops for private
futexes.

So yeah, the below patch is basically a NOP, but we might consider
retaining it to maintain the symmetry... dunno

> > Signed-off-by: Peter Zijlstra <a.p.zijlstra@xxxxxxxxx>
> > ---
> > diff --git a/kernel/futex.c b/kernel/futex.c
> > index 197fdab..beee9af 100644
> > --- a/kernel/futex.c
> > +++ b/kernel/futex.c
> > @@ -227,6 +227,7 @@ static int get_futex_key(u32 __user *uaddr, int
> > fshared, union futex_key *key)
> > return -EFAULT;
> > key->private.mm = mm;
> > key->private.address = address;
> > + get_futex_key_refs(key);
> > return 0;
> > }


--
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/