/*
 * Read-Copy Update mechanism for mutual exclusion (classic version)
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2 of the License, or
 * (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write to the Free Software
 * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
 *
 * Copyright IBM Corporation, 2001
 *
 * Author: Dipankar Sarma <dipankar@in.ibm.com>
 *
 * Based on the original work by Paul McKenney <paulmck@us.ibm.com>
 * and inputs from Rusty Russell, Andrea Arcangeli and Andi Kleen.
 * Papers:
 * http://www.rdrop.com/users/paulmck/paper/rclockpdcsproof.pdf
 * http://lse.sourceforge.net/locking/rclock_OLS.2001.05.01c.sc.pdf (OLS2001)
 *
 * For detailed explanation of Read-Copy Update mechanism see -
 * 		Documentation/RCU
 *
 * Rewrite based on a global state machine
 * (C) Manfred Spraul <manfred@colorfullife.com>, 2008
 */

#ifndef __LINUX_RCUCLASSIC_H
#define __LINUX_RCUCLASSIC_H

#include <linux/cache.h>
#include <linux/spinlock.h>
#include <linux/threads.h>
#include <linux/percpu.h>
#include <linux/cpumask.h>
#include <linux/seqlock.h>
#include <linux/cpumask.h>

/*
 * cpu bitmask:
 * default implementation, flat without hierarchy, not optimized for UP.
 */

struct rcu_cpumask {
	spinlock_t lock;
	cpumask_t cpus;
} ____cacheline_internodealigned_in_smp;

#define __RCU_CPUMASK_INIT(ptr) { .lock = __SPIN_LOCK_UNLOCKED(&(ptr)->lock) }

/*
 * global state machine:
 * - each cpu regularly check the global state and compares it with it's own local state.
 * - if both state do not match, then the cpus do the required work and afterwards
 *   - update their local state
 *   - clear their bit in the cpu bitmask.
 * The state machine is sequence lock protected. It's only read with disabled local interupts.
 * Since all cpus must do something to complete a state change, the current state cannot
 * jump forward by more than one state.
 */

/* RCU_STATE_DESTROY:
 * call callbacks that were registered by call_rcu for the objects in rcu_cpu_state.old
 */
#define RCU_STATE_DESTROY		1
/* RCU_STATE_DESTROY_AND_COLLECT:
 * - call callbacks that were registered by call_rcu for the objects in rcu_cpu_state.old
 * - move the objects from rcu_cpu_state.new to rcu_cpu_state.new
 */
#define RCU_STATE_DESTROY_AND_COLLECT	2
/* RCU_STATE_GRACE
 * - wait for a quiescent state
 */
#define RCU_STATE_GRACE			3

struct rcu_global_state {
	seqlock_t		lock;
	int			state;
	int			start_immediately;
	long			completed;
	struct rcu_cpumask	cpus;
} ____cacheline_internodealigned_in_smp;

struct rcu_cpu_state {

	int state;

	/* new objects, directly from call_rcu().
	 * objects are added LIFO, better for cache hits.
	 * the list are length-based, not NULL-terminated.
	 */
	struct rcu_head *new;	/* new objects */
	struct rcu_head **newtail;
	long            newqlen; 	 /* # of queued callbacks */

	/* objects that are in rcu grace processing. The actual
	 * state depends on rgs->state.
	 */
	struct rcu_head *old;		
	struct rcu_head **oldtail;
	long            oldqlen;
};

struct rcu_cpu_dead {
	/* objects that are scheduled for immediate call of
	 * ->func().
	 * objects are added FIFO, necessary for forward progress.
	 * only one structure for _bh and _normal.
	 */
	struct rcu_head *dead;
	long		deadqlen;

	long		batchcount;
};

DECLARE_PER_CPU(struct rcu_cpu_state, rcu_cpudata_normal);
DECLARE_PER_CPU(struct rcu_cpu_state, rcu_cpudata_bh);
DECLARE_PER_CPU(struct rcu_cpu_dead, rcu_cpudata_dead);

extern long rcu_batches_completed(void);
extern long rcu_batches_completed_bh(void);

extern int rcu_pending(int cpu);
extern int rcu_needs_cpu(int cpu);

#ifdef CONFIG_DEBUG_LOCK_ALLOC
extern struct lockdep_map rcu_lock_map;
# define rcu_read_acquire()	\
			lock_acquire(&rcu_lock_map, 0, 0, 2, 1, _THIS_IP_)
# define rcu_read_release()	lock_release(&rcu_lock_map, 1, _THIS_IP_)
#else
# define rcu_read_acquire()	do { } while (0)
# define rcu_read_release()	do { } while (0)
#endif

#define __rcu_read_lock() \
	do { \
		preempt_disable(); \
		__acquire(RCU); \
		rcu_read_acquire(); \
	} while (0)
#define __rcu_read_unlock() \
	do { \
		rcu_read_release(); \
		__release(RCU); \
		preempt_enable(); \
	} while (0)
#define __rcu_read_lock_bh() \
	do { \
		local_bh_disable(); \
		__acquire(RCU_BH); \
		rcu_read_acquire(); \
	} while (0)
#define __rcu_read_unlock_bh() \
	do { \
		rcu_read_release(); \
		__release(RCU_BH); \
		local_bh_enable(); \
	} while (0)

#define __synchronize_sched() synchronize_rcu()

#define call_rcu_sched(head, func) call_rcu(head, func)

extern void __rcu_init(void);
#define rcu_init_sched()	do { } while (0)
extern void rcu_check_callbacks(int cpu, int user);
extern void rcu_restart_cpu(int cpu);


#define rcu_enter_nohz()	do { } while (0)
#define rcu_exit_nohz()		do { } while (0)

#define rcu_qsctr_inc(cpu)	do { } while (0)
#define rcu_bh_qsctr_inc(cpu)	do { } while (0)

#endif /* __LINUX_RCUCLASSIC_H */