Re: [patch 1/1] schedule removal of FUTEX_FD

[Rusty Russell]

On Tue, 2006-10-31 at 15:09 -0800, akpm@xxxxxxxx wrote:
> From: Andrew Morton <akpm@xxxxxxxx>
> 
> Apparently FUTEX_FD is unfixably racy and nothing uses it (or if it does, it
> shouldn't).
> 
> Add a warning printk, give any remaining users six months to migrate off it.

This makes sense.  FUTEX_FD was for the NGPT project which did userspace
threading, and hence couldn't block.  It was always kind of a hack
(although unfixably racy isn't quite right, it depends on usage).

However, the existence of FUTEX_FD is what made Ingo complain that we
couldn't simply pin the futex page in memory, because now a process
could pin one page per fd.  Removing it would seem to indicate that we
can return to a much simpler scheme of (1) pinning a page when someone
does futex_wait, and (2) simply comparing futexes by physical address.

Now, I realize with some dismay that simplicity is no longer a futex
feature, but it might be worth considering?

Cheers,
Rusty.
PS.  I used to have a patch for "ratelim_printk()" which hashed on the
format string to reduce the chance that one message limit would clobber
other messages.  I'll dig it out...

-
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/