linux-mm.kvack.org archive mirror
 help / color / mirror / Atom feed
From: Andrew Morton <akpm@digeo.com>
To: Andrea Arcangeli <andrea@suse.de>
Cc: davem@redhat.com, shemminger@osdl.org, rmk@arm.linux.org.uk,
	ak@muc.de, davidm@napali.hpl.hp.com, anton@samba.org,
	linux-mm@kvack.org, rth@twiddle.net
Subject: Re: Linus rollup
Date: Wed, 29 Jan 2003 18:00:54 -0800	[thread overview]
Message-ID: <20030129180054.03ac0d48.akpm@digeo.com> (raw)
In-Reply-To: <20030130013522.GP1237@dualathlon.random>

Andrea Arcangeli <andrea@suse.de> wrote:
>
> On Wed, Jan 29, 2003 at 05:27:43PM -0800, Andrew Morton wrote:
> > @@ -82,11 +85,12 @@ static inline int fr_write_trylock(frloc
> >  
> >  	if (ret) {
> >  		++rw->pre_sequence;
> > -		wmb();
> > +		mb();
> >  	}
> 
> this isn't needed
> 
> 
> if we hold the spinlock, the serialized memory can't be change under us,
> so there's no need to put a read barrier, we only care that pre_sequence
> is visible before the chagnes are visible and before post_sequence is
> visible, hence only wmb() (after spin_lock and pre_sequence++) is
> needed there and only rmb() is needed in the read-side.
> 

OK, thanks muchly.

Lots more updates.  Here's the version which I currently have.  Looks like
fr_write_lock() and fr_write_unlock() need to be switched back to rmb()?



#ifndef __LINUX_FRLOCK_H
#define __LINUX_FRLOCK_H

/*
 * Fast read-write spinlocks.
 *
 * Fast reader/writer locks without starving writers. This type of
 * lock for data where the reader wants a consitent set of information
 * and is willing to retry if the information changes.  Readers never
 * block but they may have to retry if a writer is in
 * progress. Writers do not wait for readers. 
 *
 * Generalization on sequence variables used for gettimeofday on x86-64 
 * by Andrea Arcangeli
 *
 * This is not as cache friendly as brlock. Also, this will not work
 * for data that contains pointers, because any writer could
 * invalidate a pointer that a reader was following.
 *
 * Expected reader usage:
 * 	do {
 *	    seq = fr_read_begin();
 * 	...
 *      } while (seq != fr_read_end());
 *
 * On non-SMP the spin locks disappear but the writer still needs
 * to increment the sequence variables because an interrupt routine could
 * change the state of the data.
 */

#include <linux/config.h>
#include <linux/spinlock.h>
#include <linux/preempt.h>

typedef struct {
	unsigned pre_sequence;
	unsigned post_sequence;
	spinlock_t lock;
} frlock_t;

/*
 * These macros triggered gcc-3.x compile-time problems.  We think these are
 * OK now.  Be cautious.
 */
#define FR_LOCK_UNLOCKED { 0, 0, SPIN_LOCK_UNLOCKED }
#define frlock_init(x)	do { *(x) = (frlock_t) FR_LOCK_UNLOCKED; } while (0)

/* Update sequence count only
 * Assumes caller is doing own mutual exclusion with other lock
 * or semaphore.
 */
static inline void fr_write_begin(frlock_t *rw)
{
	preempt_disable();
	rw->pre_sequence++;
	mb();
}

static inline void fr_write_end(frlock_t *rw)
{
	mb();
	rw->post_sequence++;
	BUG_ON(rw->post_sequence != rw->pre_sequence);
	preempt_enable();
}

/* Lock out other writers and update the count.
 * Acts like a normal spin_lock/unlock.
 */
static inline void fr_write_lock(frlock_t *rw)
{
	spin_lock(&rw->lock);
	rw->pre_sequence++;
	mb();
}	

static inline void fr_write_unlock(frlock_t *rw) 
{
	mb();
	rw->post_sequence++;
	spin_unlock(&rw->lock);
}

static inline int fr_write_trylock(frlock_t *rw)
{
	int ret = spin_trylock(&rw->lock);

	if (ret) {
		++rw->pre_sequence;
		wmb();
	}
	return ret;
}

static inline unsigned fr_read_begin(const frlock_t *rw) 
{
	unsigned ret = rw->post_sequence;
	rmb();
	return ret;
	
}

/* End of reader calculation -- fetch last writer start token */
static inline unsigned fr_read_end(const frlock_t *rw)
{
	rmb();
	return rw->pre_sequence;
}

/*
 * Possible sw/hw IRQ protected versions of the interfaces.
 */
#define fr_write_lock_irqsave(lock, flags)				\
	do { local_irq_save(flags);	fr_write_lock(lock); } while (0)
#define fr_write_lock_irq(lock)						\
	do { local_irq_disable();	fr_write_lock(lock); } while (0)
#define fr_write_lock_bh(lock)						\
        do { local_bh_disable();	fr_write_lock(lock); } while (0)

#define fr_write_unlock_irqrestore(lock, flags)				\
	do { fr_write_unlock(lock); local_irq_restore(flags); } while(0)
#define fr_write_unlock_irq(lock)					\
	do { fr_write_unlock(lock); local_irq_enable(); } while(0)
#define fr_write_unlock_bh(lock)					\
	do { fr_write_unlock(lock); local_bh_enable(); } while(0)

#define fr_read_begin_irqsave(lock, flags)				\
	({ local_irq_save(flags);	fr_read_begin(lock); })

#define fr_read_end_irqrestore(lock, flags)				\
	({	unsigned ret = fr_read_end(lock);			\
		local_irq_save(flags);					\
		ret;							\
	})

#endif /* __LINUX_FRLOCK_H */

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/

  reply	other threads:[~2003-01-30  1:43 UTC|newest]

Thread overview: 36+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2003-01-29  6:07 Andrew Morton
2003-01-29  6:53 ` David Mosberger
2003-01-29  7:25 ` David S. Miller
2003-01-29  9:33 ` Andrew Morton
2003-01-29  9:35   ` David S. Miller
2003-01-29  9:54 ` Anton Blanchard
2003-01-29  9:59 ` Russell King
2003-01-29  9:51   ` David S. Miller
2003-01-29 10:26     ` Andrew Morton
2003-01-29 22:35       ` Stephen Hemminger
2003-01-29 23:12         ` Andrew Morton
2003-01-30  0:30           ` David S. Miller
2003-01-30  1:27             ` Andrew Morton
2003-01-30  1:24               ` Andi Kleen
2003-01-30  2:01                 ` Andrew Morton
2003-01-30  1:35               ` Andrea Arcangeli
2003-01-30  2:00                 ` Andrew Morton [this message]
2003-01-30  1:50                   ` Jeff Garzik
2003-01-30  2:03                     ` Andrea Arcangeli
2003-01-30 17:09                       ` Stephen Hemminger
2003-01-30 17:15                         ` Randy.Dunlap
2003-01-30 17:25                           ` Andi Kleen
2003-01-30 17:23                             ` Randy.Dunlap
2003-01-30  1:52                   ` Richard Henderson
2003-01-30  2:06                     ` Andrea Arcangeli
2003-01-30  1:54                   ` Andrea Arcangeli
2003-01-30  2:19                     ` Andrew Morton
2003-01-30 17:37                     ` Stephen Hemminger
2003-01-30 17:50                       ` Andrea Arcangeli
2003-01-30  2:00                   ` Richard Henderson
2003-01-30  0:46         ` Andrea Arcangeli
2003-01-30  0:43       ` Andrea Arcangeli
2003-01-29 10:16   ` Andrew Morton
2003-01-29 17:04   ` David Mosberger
2003-01-29 17:25     ` Russell King
2003-01-29 19:05       ` David Mosberger

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20030129180054.03ac0d48.akpm@digeo.com \
    --to=akpm@digeo.com \
    --cc=ak@muc.de \
    --cc=andrea@suse.de \
    --cc=anton@samba.org \
    --cc=davem@redhat.com \
    --cc=davidm@napali.hpl.hp.com \
    --cc=linux-mm@kvack.org \
    --cc=rmk@arm.linux.org.uk \
    --cc=rth@twiddle.net \
    --cc=shemminger@osdl.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox