Re: [PATCH V2 0/6][RFC] futex: FUTEX_LOCK with optional adaptive spinning

Previous message: [thread] [date] [author]
Next message: [thread] [date] [author]
From: Alan Cox
Date: Saturday, April 10, 2010 - 4:35 pm

> The non-swappable part will be tricky.  One doesn't know how many

You only need one page per 4096 threads or so if you make it create the
page on the first request, tied to say the signals and the like in
threaded tasks, and you then allocate 'slots' in the page for future
calls until you've got 4096 live.

ie you'd see something like

	addr = set_tid_notifier_addr();

	[1st call
	 map page at x to x + 4095, probably R/O]
	[returns x]

next thread
	addr = set_tid_notifier_addr()

	[returns x + 1]

Alan
--
Previous message: [thread] [date] [author]
Next message: [thread] [date] [author]

Messages in current thread:
[PATCH 2/6] futex: add futex_q static initializer, Darren Hart, (Mon Apr 5, 1:23 pm)
[PATCH 3/6] futex: refactor futex_lock_pi_atomic, Darren Hart, (Mon Apr 5, 1:23 pm)
[No subject], Ulrich Drepper, (Tue Apr 6, 7:47 am)
[No subject], Peter Zijlstra, (Tue Apr 6, 7:51 am)
[No subject], Darren Hart, (Tue Apr 6, 8:28 am)
[No subject], Peter Zijlstra, (Tue Apr 6, 8:29 am)
[No subject], Darren Hart, (Tue Apr 6, 8:33 am)
[No subject], Peter Zijlstra, (Tue Apr 6, 8:37 am)
[No subject], Avi Kivity, (Tue Apr 6, 9:06 am)
[No subject], Avi Kivity, (Tue Apr 6, 9:10 am)
[No subject], Thomas Gleixner, (Tue Apr 6, 9:14 am)
[No subject], Avi Kivity, (Tue Apr 6, 9:20 am)
[No subject], Alan Cox, (Tue Apr 6, 9:44 am)
[No subject], Alan Cox, (Tue Apr 6, 9:53 am)
[No subject], Alan Cox, (Tue Apr 6, 9:54 am)
[No subject], Thomas Gleixner, (Tue Apr 6, 9:55 am)
[No subject], Ulrich Drepper, (Tue Apr 6, 10:34 am)
Re: [PATCH V2 0/6][RFC] futex: FUTEX_LOCK with optional ad ..., Alan Cox, (Sat Apr 10, 4:35 pm)