rcu: Remove CONFIG_PREEMPT_RCU
Paul E. McKenney [Sat, 22 Aug 2009 20:56:53 +0000 (13:56 -0700)]
Now that CONFIG_TREE_PREEMPT_RCU is in place, there is no
further need for CONFIG_PREEMPT_RCU.  Remove it, along with
whatever subtle bugs it may (or may not) contain.

Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: laijs@cn.fujitsu.com
Cc: dipankar@in.ibm.com
Cc: akpm@linux-foundation.org
Cc: mathieu.desnoyers@polymtl.ca
Cc: josht@linux.vnet.ibm.com
Cc: dvhltc@us.ibm.com
Cc: niv@us.ibm.com
Cc: peterz@infradead.org
Cc: rostedt@goodmis.org
LKML-Reference: <125097461396-git-send-email->
Signed-off-by: Ingo Molnar <mingo@elte.hu>

12 files changed:
Documentation/RCU/rcu.txt
Documentation/RCU/whatisRCU.txt
include/linux/init_task.h
include/linux/rcupdate.h
include/linux/rcupreempt.h [deleted file]
include/linux/rcupreempt_trace.h [deleted file]
include/linux/sched.h
init/Kconfig
kernel/Makefile
kernel/rcupreempt.c [deleted file]
kernel/rcupreempt_trace.c [deleted file]
lib/Kconfig.debug

index 7aa2002..2a23523 100644 (file)
@@ -36,7 +36,7 @@ o     How can the updater tell when a grace period has completed
        executed in user mode, or executed in the idle loop, we can
        safely free up that item.
 
-       Preemptible variants of RCU (CONFIG_PREEMPT_RCU) get the
+       Preemptible variants of RCU (CONFIG_TREE_PREEMPT_RCU) get the
        same effect, but require that the readers manipulate CPU-local
        counters.  These counters allow limited types of blocking
        within RCU read-side critical sections.  SRCU also uses
@@ -79,10 +79,10 @@ o   I hear that RCU is patented?  What is with that?
 o      I hear that RCU needs work in order to support realtime kernels?
 
        This work is largely completed.  Realtime-friendly RCU can be
-       enabled via the CONFIG_PREEMPT_RCU kernel configuration parameter.
-       However, work is in progress for enabling priority boosting of
-       preempted RCU read-side critical sections.  This is needed if you
-       have CPU-bound realtime threads.
+       enabled via the CONFIG_TREE_PREEMPT_RCU kernel configuration
+       parameter.  However, work is in progress for enabling priority
+       boosting of preempted RCU read-side critical sections.  This is
+       needed if you have CPU-bound realtime threads.
 
 o      Where can I find more information on RCU?
 
index 97ded24..e41a7fe 100644 (file)
@@ -136,10 +136,10 @@ rcu_read_lock()
        Used by a reader to inform the reclaimer that the reader is
        entering an RCU read-side critical section.  It is illegal
        to block while in an RCU read-side critical section, though
-       kernels built with CONFIG_PREEMPT_RCU can preempt RCU read-side
-       critical sections.  Any RCU-protected data structure accessed
-       during an RCU read-side critical section is guaranteed to remain
-       unreclaimed for the full duration of that critical section.
+       kernels built with CONFIG_TREE_PREEMPT_RCU can preempt RCU
+       read-side critical sections.  Any RCU-protected data structure
+       accessed during an RCU read-side critical section is guaranteed to
+       remain unreclaimed for the full duration of that critical section.
        Reference counts may be used in conjunction with RCU to maintain
        longer-term references to data structures.
 
index 971a968..79d4bae 100644 (file)
@@ -94,11 +94,7 @@ extern struct group_info init_groups;
 # define CAP_INIT_BSET  CAP_INIT_EFF_SET
 #endif
 
-#ifdef CONFIG_PREEMPT_RCU
-#define INIT_TASK_RCU_PREEMPT(tsk)                                     \
-       .rcu_read_lock_nesting = 0,                                     \
-       .rcu_flipctr_idx = 0,
-#elif defined(CONFIG_TREE_PREEMPT_RCU)
+#ifdef CONFIG_TREE_PREEMPT_RCU
 #define INIT_TASK_RCU_PREEMPT(tsk)                                     \
        .rcu_read_lock_nesting = 0,                                     \
        .rcu_read_unlock_special = 0,                                   \
index 26892f5..ec90fc3 100644 (file)
@@ -68,11 +68,9 @@ extern int rcu_scheduler_active;
 
 #if defined(CONFIG_TREE_RCU) || defined(CONFIG_TREE_PREEMPT_RCU)
 #include <linux/rcutree.h>
-#elif defined(CONFIG_PREEMPT_RCU)
-#include <linux/rcupreempt.h>
 #else
 #error "Unknown RCU implementation specified to kernel configuration"
-#endif /* #else #if defined(CONFIG_CLASSIC_RCU) */
+#endif
 
 #define RCU_HEAD_INIT  { .next = NULL, .func = NULL }
 #define RCU_HEAD(head) struct rcu_head head = RCU_HEAD_INIT
diff --git a/include/linux/rcupreempt.h b/include/linux/rcupreempt.h
deleted file mode 100644 (file)
index a42ab88..0000000
+++ /dev/null
@@ -1,140 +0,0 @@
-/*
- * Read-Copy Update mechanism for mutual exclusion (RT implementation)
- *
- * 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 (C) IBM Corporation, 2006
- *
- * Author:  Paul McKenney <paulmck@us.ibm.com>
- *
- * Based on the original work by Paul McKenney <paul.mckenney@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
- *
- */
-
-#ifndef __LINUX_RCUPREEMPT_H
-#define __LINUX_RCUPREEMPT_H
-
-#include <linux/cache.h>
-#include <linux/spinlock.h>
-#include <linux/threads.h>
-#include <linux/smp.h>
-#include <linux/cpumask.h>
-#include <linux/seqlock.h>
-
-extern void rcu_sched_qs(int cpu);
-static inline void rcu_bh_qs(int cpu) { }
-
-/*
- * Someone might want to pass call_rcu_bh as a function pointer.
- * So this needs to just be a rename and not a macro function.
- *  (no parentheses)
- */
-#define call_rcu_bh            call_rcu
-
-/**
- * call_rcu_sched - Queue RCU callback for invocation after sched grace period.
- * @head: structure to be used for queueing the RCU updates.
- * @func: actual update function to be invoked after the grace period
- *
- * The update function will be invoked some time after a full
- * synchronize_sched()-style grace period elapses, in other words after
- * all currently executing preempt-disabled sections of code (including
- * hardirq handlers, NMI handlers, and local_irq_save() blocks) have
- * completed.
- */
-extern void call_rcu_sched(struct rcu_head *head,
-                          void (*func)(struct rcu_head *head));
-
-extern void __rcu_read_lock(void);
-extern void __rcu_read_unlock(void);
-extern int rcu_needs_cpu(int cpu);
-
-#define __rcu_read_lock_bh()   { rcu_read_lock(); local_bh_disable(); }
-#define __rcu_read_unlock_bh() { local_bh_enable(); rcu_read_unlock(); }
-
-extern void __synchronize_sched(void);
-
-static inline void synchronize_rcu_expedited(void)
-{
-       synchronize_rcu();  /* Placeholder for new rcupreempt implementation. */
-}
-
-static inline void synchronize_rcu_bh_expedited(void)
-{
-       synchronize_rcu_bh();  /* Placeholder for new rcupreempt impl. */
-}
-
-extern void __rcu_init(void);
-extern void rcu_init_sched(void);
-extern void rcu_check_callbacks(int cpu, int user);
-extern void rcu_restart_cpu(int cpu);
-extern long rcu_batches_completed(void);
-
-/*
- * Return the number of RCU batches processed thus far. Useful for debug
- * and statistic. The _bh variant is identifcal to straight RCU
- */
-static inline long rcu_batches_completed_bh(void)
-{
-       return rcu_batches_completed();
-}
-
-static inline void exit_rcu(void)
-{
-}
-
-#ifdef CONFIG_RCU_TRACE
-struct rcupreempt_trace;
-extern long *rcupreempt_flipctr(int cpu);
-extern long rcupreempt_data_completed(void);
-extern int rcupreempt_flip_flag(int cpu);
-extern int rcupreempt_mb_flag(int cpu);
-extern char *rcupreempt_try_flip_state_name(void);
-extern struct rcupreempt_trace *rcupreempt_trace_cpu(int cpu);
-#endif
-
-struct softirq_action;
-
-#ifdef CONFIG_NO_HZ
-extern void rcu_enter_nohz(void);
-extern void rcu_exit_nohz(void);
-#else
-# define rcu_enter_nohz()      do { } while (0)
-# define rcu_exit_nohz()       do { } while (0)
-#endif
-
-/*
- * A context switch is a grace period for rcupreempt synchronize_rcu()
- * only during early boot, before the scheduler has been initialized.
- * So, how the heck do we get a context switch?  Well, if the caller
- * invokes synchronize_rcu(), they are willing to accept a context
- * switch, so we simply pretend that one happened.
- *
- * After boot, there might be a blocked or preempted task in an RCU
- * read-side critical section, so we cannot then take the fastpath.
- */
-static inline int rcu_blocking_is_gp(void)
-{
-       return num_online_cpus() == 1 && !rcu_scheduler_active;
-}
-
-#endif /* __LINUX_RCUPREEMPT_H */
diff --git a/include/linux/rcupreempt_trace.h b/include/linux/rcupreempt_trace.h
deleted file mode 100644 (file)
index b99ae07..0000000
+++ /dev/null
@@ -1,97 +0,0 @@
-/*
- * Read-Copy Update mechanism for mutual exclusion (RT implementation)
- *
- * 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 (C) IBM Corporation, 2006
- *
- * Author:  Paul McKenney <paulmck@us.ibm.com>
- *
- * Based on the original work by Paul McKenney <paul.mckenney@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 the Preemptible Read-Copy Update mechanism see -
- *              http://lwn.net/Articles/253651/
- */
-
-#ifndef __LINUX_RCUPREEMPT_TRACE_H
-#define __LINUX_RCUPREEMPT_TRACE_H
-
-#include <linux/types.h>
-#include <linux/kernel.h>
-
-#include <asm/atomic.h>
-
-/*
- * PREEMPT_RCU data structures.
- */
-
-struct rcupreempt_trace {
-       long            next_length;
-       long            next_add;
-       long            wait_length;
-       long            wait_add;
-       long            done_length;
-       long            done_add;
-       long            done_remove;
-       atomic_t        done_invoked;
-       long            rcu_check_callbacks;
-       atomic_t        rcu_try_flip_1;
-       atomic_t        rcu_try_flip_e1;
-       long            rcu_try_flip_i1;
-       long            rcu_try_flip_ie1;
-       long            rcu_try_flip_g1;
-       long            rcu_try_flip_a1;
-       long            rcu_try_flip_ae1;
-       long            rcu_try_flip_a2;
-       long            rcu_try_flip_z1;
-       long            rcu_try_flip_ze1;
-       long            rcu_try_flip_z2;
-       long            rcu_try_flip_m1;
-       long            rcu_try_flip_me1;
-       long            rcu_try_flip_m2;
-};
-
-#ifdef CONFIG_RCU_TRACE
-#define RCU_TRACE(fn, arg)     fn(arg);
-#else
-#define RCU_TRACE(fn, arg)
-#endif
-
-extern void rcupreempt_trace_move2done(struct rcupreempt_trace *trace);
-extern void rcupreempt_trace_move2wait(struct rcupreempt_trace *trace);
-extern void rcupreempt_trace_try_flip_1(struct rcupreempt_trace *trace);
-extern void rcupreempt_trace_try_flip_e1(struct rcupreempt_trace *trace);
-extern void rcupreempt_trace_try_flip_i1(struct rcupreempt_trace *trace);
-extern void rcupreempt_trace_try_flip_ie1(struct rcupreempt_trace *trace);
-extern void rcupreempt_trace_try_flip_g1(struct rcupreempt_trace *trace);
-extern void rcupreempt_trace_try_flip_a1(struct rcupreempt_trace *trace);
-extern void rcupreempt_trace_try_flip_ae1(struct rcupreempt_trace *trace);
-extern void rcupreempt_trace_try_flip_a2(struct rcupreempt_trace *trace);
-extern void rcupreempt_trace_try_flip_z1(struct rcupreempt_trace *trace);
-extern void rcupreempt_trace_try_flip_ze1(struct rcupreempt_trace *trace);
-extern void rcupreempt_trace_try_flip_z2(struct rcupreempt_trace *trace);
-extern void rcupreempt_trace_try_flip_m1(struct rcupreempt_trace *trace);
-extern void rcupreempt_trace_try_flip_me1(struct rcupreempt_trace *trace);
-extern void rcupreempt_trace_try_flip_m2(struct rcupreempt_trace *trace);
-extern void rcupreempt_trace_check_callbacks(struct rcupreempt_trace *trace);
-extern void rcupreempt_trace_done_remove(struct rcupreempt_trace *trace);
-extern void rcupreempt_trace_invoke(struct rcupreempt_trace *trace);
-extern void rcupreempt_trace_next_add(struct rcupreempt_trace *trace);
-
-#endif /* __LINUX_RCUPREEMPT_TRACE_H */
index d7f98f6..bfca26d 100644 (file)
@@ -1205,11 +1205,6 @@ struct task_struct {
        unsigned int policy;
        cpumask_t cpus_allowed;
 
-#ifdef CONFIG_PREEMPT_RCU
-       int rcu_read_lock_nesting;
-       int rcu_flipctr_idx;
-#endif /* #ifdef CONFIG_PREEMPT_RCU */
-
 #ifdef CONFIG_TREE_PREEMPT_RCU
        int rcu_read_lock_nesting;
        char rcu_read_unlock_special;
@@ -1744,14 +1739,6 @@ static inline void rcu_copy_process(struct task_struct *p)
        INIT_LIST_HEAD(&p->rcu_node_entry);
 }
 
-#elif defined(CONFIG_PREEMPT_RCU)
-
-static inline void rcu_copy_process(struct task_struct *p)
-{
-       p->rcu_read_lock_nesting = 0;
-       p->rcu_flipctr_idx = 0;
-}
-
 #else
 
 static inline void rcu_copy_process(struct task_struct *p)
index f88da2d..8e8b76d 100644 (file)
@@ -324,17 +324,6 @@ config TREE_RCU
          thousands of CPUs.  It also scales down nicely to
          smaller systems.
 
-config PREEMPT_RCU
-       bool "Preemptible RCU"
-       depends on PREEMPT
-       help
-         This option reduces the latency of the kernel by making certain
-         RCU sections preemptible. Normally RCU code is non-preemptible, if
-         this option is selected then read-only RCU sections become
-         preemptible. This helps latency, but may expose bugs due to
-         now-naive assumptions about each RCU read-side critical section
-         remaining on a given CPU through its execution.
-
 config TREE_PREEMPT_RCU
        bool "Preemptable tree-based hierarchical RCU"
        depends on PREEMPT
@@ -348,7 +337,7 @@ endchoice
 
 config RCU_TRACE
        bool "Enable tracing for RCU"
-       depends on TREE_RCU || PREEMPT_RCU || TREE_PREEMPT_RCU
+       depends on TREE_RCU || TREE_PREEMPT_RCU
        help
          This option provides tracing in RCU which presents stats
          in debugfs for debugging RCU implementation.
@@ -395,13 +384,6 @@ config TREE_RCU_TRACE
          TREE_PREEMPT_RCU implementations, permitting Makefile to
          trivially select kernel/rcutree_trace.c.
 
-config PREEMPT_RCU_TRACE
-       def_bool RCU_TRACE && PREEMPT_RCU
-       select DEBUG_FS
-       help
-         This option provides tracing for the PREEMPT_RCU implementation,
-         permitting Makefile to trivially select kernel/rcupreempt_trace.c.
-
 endmenu # "RCU Subsystem"
 
 config IKCONFIG
index 1a38b47..b833bd5 100644 (file)
@@ -82,9 +82,7 @@ obj-$(CONFIG_SECCOMP) += seccomp.o
 obj-$(CONFIG_RCU_TORTURE_TEST) += rcutorture.o
 obj-$(CONFIG_TREE_RCU) += rcutree.o
 obj-$(CONFIG_TREE_PREEMPT_RCU) += rcutree.o
-obj-$(CONFIG_PREEMPT_RCU) += rcupreempt.o
 obj-$(CONFIG_TREE_RCU_TRACE) += rcutree_trace.o
-obj-$(CONFIG_PREEMPT_RCU_TRACE) += rcupreempt_trace.o
 obj-$(CONFIG_RELAY) += relay.o
 obj-$(CONFIG_SYSCTL) += utsname_sysctl.o
 obj-$(CONFIG_TASK_DELAY_ACCT) += delayacct.o
diff --git a/kernel/rcupreempt.c b/kernel/rcupreempt.c
deleted file mode 100644 (file)
index 0053ce5..0000000
+++ /dev/null
@@ -1,1518 +0,0 @@
-/*
- * Read-Copy Update mechanism for mutual exclusion, realtime implementation
- *
- * 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, 2006
- *
- * Authors: Paul E. McKenney <paulmck@us.ibm.com>
- *             With thanks to Esben Nielsen, Bill Huey, and Ingo Molnar
- *             for pushing me away from locks and towards counters, and
- *             to Suparna Bhattacharya for pushing me completely away
- *             from atomic instructions on the read side.
- *
- *  - Added handling of Dynamic Ticks
- *      Copyright 2007 - Paul E. Mckenney <paulmck@us.ibm.com>
- *                     - Steven Rostedt <srostedt@redhat.com>
- *
- * Papers:  http://www.rdrop.com/users/paulmck/RCU
- *
- * Design Document: http://lwn.net/Articles/253651/
- *
- * For detailed explanation of Read-Copy Update mechanism see -
- *             Documentation/RCU/ *.txt
- *
- */
-#include <linux/types.h>
-#include <linux/kernel.h>
-#include <linux/init.h>
-#include <linux/spinlock.h>
-#include <linux/smp.h>
-#include <linux/rcupdate.h>
-#include <linux/interrupt.h>
-#include <linux/sched.h>
-#include <asm/atomic.h>
-#include <linux/bitops.h>
-#include <linux/module.h>
-#include <linux/kthread.h>
-#include <linux/completion.h>
-#include <linux/moduleparam.h>
-#include <linux/percpu.h>
-#include <linux/notifier.h>
-#include <linux/cpu.h>
-#include <linux/random.h>
-#include <linux/delay.h>
-#include <linux/cpumask.h>
-#include <linux/rcupreempt_trace.h>
-#include <asm/byteorder.h>
-
-/*
- * PREEMPT_RCU data structures.
- */
-
-/*
- * GP_STAGES specifies the number of times the state machine has
- * to go through the all the rcu_try_flip_states (see below)
- * in a single Grace Period.
- *
- * GP in GP_STAGES stands for Grace Period ;)
- */
-#define GP_STAGES    2
-struct rcu_data {
-       spinlock_t      lock;           /* Protect rcu_data fields. */
-       long            completed;      /* Number of last completed batch. */
-       int             waitlistcount;
-       struct rcu_head *nextlist;
-       struct rcu_head **nexttail;
-       struct rcu_head *waitlist[GP_STAGES];
-       struct rcu_head **waittail[GP_STAGES];
-       struct rcu_head *donelist;      /* from waitlist & waitschedlist */
-       struct rcu_head **donetail;
-       long rcu_flipctr[2];
-       struct rcu_head *nextschedlist;
-       struct rcu_head **nextschedtail;
-       struct rcu_head *waitschedlist;
-       struct rcu_head **waitschedtail;
-       int rcu_sched_sleeping;
-#ifdef CONFIG_RCU_TRACE
-       struct rcupreempt_trace trace;
-#endif /* #ifdef CONFIG_RCU_TRACE */
-};
-
-/*
- * States for rcu_try_flip() and friends.
- */
-
-enum rcu_try_flip_states {
-
-       /*
-        * Stay here if nothing is happening. Flip the counter if somthing
-        * starts happening. Denoted by "I"
-        */
-       rcu_try_flip_idle_state,
-
-       /*
-        * Wait here for all CPUs to notice that the counter has flipped. This
-        * prevents the old set of counters from ever being incremented once
-        * we leave this state, which in turn is necessary because we cannot
-        * test any individual counter for zero -- we can only check the sum.
-        * Denoted by "A".
-        */
-       rcu_try_flip_waitack_state,
-
-       /*
-        * Wait here for the sum of the old per-CPU counters to reach zero.
-        * Denoted by "Z".
-        */
-       rcu_try_flip_waitzero_state,
-
-       /*
-        * Wait here for each of the other CPUs to execute a memory barrier.
-        * This is necessary to ensure that these other CPUs really have
-        * completed executing their RCU read-side critical sections, despite
-        * their CPUs wildly reordering memory. Denoted by "M".
-        */
-       rcu_try_flip_waitmb_state,
-};
-
-/*
- * States for rcu_ctrlblk.rcu_sched_sleep.
- */
-
-enum rcu_sched_sleep_states {
-       rcu_sched_not_sleeping, /* Not sleeping, callbacks need GP.  */
-       rcu_sched_sleep_prep,   /* Thinking of sleeping, rechecking. */
-       rcu_sched_sleeping,     /* Sleeping, awaken if GP needed. */
-};
-
-struct rcu_ctrlblk {
-       spinlock_t      fliplock;       /* Protect state-machine transitions. */
-       long            completed;      /* Number of last completed batch. */
-       enum rcu_try_flip_states rcu_try_flip_state; /* The current state of
-                                                       the rcu state machine */
-       spinlock_t      schedlock;      /* Protect rcu_sched sleep state. */
-       enum rcu_sched_sleep_states sched_sleep; /* rcu_sched state. */
-       wait_queue_head_t sched_wq;     /* Place for rcu_sched to sleep. */
-};
-
-struct rcu_dyntick_sched {
-       int dynticks;
-       int dynticks_snap;
-       int sched_qs;
-       int sched_qs_snap;
-       int sched_dynticks_snap;
-};
-
-static DEFINE_PER_CPU_SHARED_ALIGNED(struct rcu_dyntick_sched, rcu_dyntick_sched) = {
-       .dynticks = 1,
-};
-
-static int rcu_pending(int cpu);
-
-void rcu_sched_qs(int cpu)
-{
-       struct rcu_dyntick_sched *rdssp = &per_cpu(rcu_dyntick_sched, cpu);
-
-       rdssp->sched_qs++;
-}
-
-#ifdef CONFIG_NO_HZ
-
-void rcu_enter_nohz(void)
-{
-       static DEFINE_RATELIMIT_STATE(rs, 10 * HZ, 1);
-
-       smp_mb(); /* CPUs seeing ++ must see prior RCU read-side crit sects */
-       __get_cpu_var(rcu_dyntick_sched).dynticks++;
-       WARN_ON_RATELIMIT(__get_cpu_var(rcu_dyntick_sched).dynticks & 0x1, &rs);
-}
-
-void rcu_exit_nohz(void)
-{
-       static DEFINE_RATELIMIT_STATE(rs, 10 * HZ, 1);
-
-       __get_cpu_var(rcu_dyntick_sched).dynticks++;
-       smp_mb(); /* CPUs seeing ++ must see later RCU read-side crit sects */
-       WARN_ON_RATELIMIT(!(__get_cpu_var(rcu_dyntick_sched).dynticks & 0x1),
-                               &rs);
-}
-
-#endif /* CONFIG_NO_HZ */
-
-
-static DEFINE_PER_CPU(struct rcu_data, rcu_data);
-
-static struct rcu_ctrlblk rcu_ctrlblk = {
-       .fliplock = __SPIN_LOCK_UNLOCKED(rcu_ctrlblk.fliplock),
-       .completed = 0,
-       .rcu_try_flip_state = rcu_try_flip_idle_state,
-       .schedlock = __SPIN_LOCK_UNLOCKED(rcu_ctrlblk.schedlock),
-       .sched_sleep = rcu_sched_not_sleeping,
-       .sched_wq = __WAIT_QUEUE_HEAD_INITIALIZER(rcu_ctrlblk.sched_wq),
-};
-
-static struct task_struct *rcu_sched_grace_period_task;
-
-#ifdef CONFIG_RCU_TRACE
-static char *rcu_try_flip_state_names[] =
-       { "idle", "waitack", "waitzero", "waitmb" };
-#endif /* #ifdef CONFIG_RCU_TRACE */
-
-static DECLARE_BITMAP(rcu_cpu_online_map, NR_CPUS) __read_mostly
-       = CPU_BITS_NONE;
-
-/*
- * Enum and per-CPU flag to determine when each CPU has seen
- * the most recent counter flip.
- */
-
-enum rcu_flip_flag_values {
-       rcu_flip_seen,          /* Steady/initial state, last flip seen. */
-                               /* Only GP detector can update. */
-       rcu_flipped             /* Flip just completed, need confirmation. */
-                               /* Only corresponding CPU can update. */
-};
-static DEFINE_PER_CPU_SHARED_ALIGNED(enum rcu_flip_flag_values, rcu_flip_flag)
-                                                               = rcu_flip_seen;
-
-/*
- * Enum and per-CPU flag to determine when each CPU has executed the
- * needed memory barrier to fence in memory references from its last RCU
- * read-side critical section in the just-completed grace period.
- */
-
-enum rcu_mb_flag_values {
-       rcu_mb_done,            /* Steady/initial state, no mb()s required. */
-                               /* Only GP detector can update. */
-       rcu_mb_needed           /* Flip just completed, need an mb(). */
-                               /* Only corresponding CPU can update. */
-};
-static DEFINE_PER_CPU_SHARED_ALIGNED(enum rcu_mb_flag_values, rcu_mb_flag)
-                                                               = rcu_mb_done;
-
-/*
- * RCU_DATA_ME: find the current CPU's rcu_data structure.
- * RCU_DATA_CPU: find the specified CPU's rcu_data structure.
- */
-#define RCU_DATA_ME()          (&__get_cpu_var(rcu_data))
-#define RCU_DATA_CPU(cpu)      (&per_cpu(rcu_data, cpu))
-
-/*
- * Helper macro for tracing when the appropriate rcu_data is not
- * cached in a local variable, but where the CPU number is so cached.
- */
-#define RCU_TRACE_CPU(f, cpu) RCU_TRACE(f, &(RCU_DATA_CPU(cpu)->trace));
-
-/*
- * Helper macro for tracing when the appropriate rcu_data is not
- * cached in a local variable.
- */
-#define RCU_TRACE_ME(f) RCU_TRACE(f, &(RCU_DATA_ME()->trace));
-
-/*
- * Helper macro for tracing when the appropriate rcu_data is pointed
- * to by a local variable.
- */
-#define RCU_TRACE_RDP(f, rdp) RCU_TRACE(f, &((rdp)->trace));
-
-#define RCU_SCHED_BATCH_TIME (HZ / 50)
-
-/*
- * Return the number of RCU batches processed thus far.  Useful
- * for debug and statistics.
- */
-long rcu_batches_completed(void)
-{
-       return rcu_ctrlblk.completed;
-}
-EXPORT_SYMBOL_GPL(rcu_batches_completed);
-
-void __rcu_read_lock(void)
-{
-       int idx;
-       struct task_struct *t = current;
-       int nesting;
-
-       nesting = ACCESS_ONCE(t->rcu_read_lock_nesting);
-       if (nesting != 0) {
-
-               /* An earlier rcu_read_lock() covers us, just count it. */
-
-               t->rcu_read_lock_nesting = nesting + 1;
-
-       } else {
-               unsigned long flags;
-
-               /*
-                * We disable interrupts for the following reasons:
-                * - If we get scheduling clock interrupt here, and we
-                *   end up acking the counter flip, it's like a promise
-                *   that we will never increment the old counter again.
-                *   Thus we will break that promise if that
-                *   scheduling clock interrupt happens between the time
-                *   we pick the .completed field and the time that we
-                *   increment our counter.
-                *
-                * - We don't want to be preempted out here.
-                *
-                * NMIs can still occur, of course, and might themselves
-                * contain rcu_read_lock().
-                */
-
-               local_irq_save(flags);
-
-               /*
-                * Outermost nesting of rcu_read_lock(), so increment
-                * the current counter for the current CPU.  Use volatile
-                * casts to prevent the compiler from reordering.
-                */
-
-               idx = ACCESS_ONCE(rcu_ctrlblk.completed) & 0x1;
-               ACCESS_ONCE(RCU_DATA_ME()->rcu_flipctr[idx])++;
-
-               /*
-                * Now that the per-CPU counter has been incremented, we
-                * are protected from races with rcu_read_lock() invoked
-                * from NMI handlers on this CPU.  We can therefore safely
-                * increment the nesting counter, relieving further NMIs
-                * of the need to increment the per-CPU counter.
-                */
-
-               ACCESS_ONCE(t->rcu_read_lock_nesting) = nesting + 1;
-
-               /*
-                * Now that we have preventing any NMIs from storing
-                * to the ->rcu_flipctr_idx, we can safely use it to
-                * remember which counter to decrement in the matching
-                * rcu_read_unlock().
-                */
-
-               ACCESS_ONCE(t->rcu_flipctr_idx) = idx;
-               local_irq_restore(flags);
-       }
-}
-EXPORT_SYMBOL_GPL(__rcu_read_lock);
-
-void __rcu_read_unlock(void)
-{
-       int idx;
-       struct task_struct *t = current;
-       int nesting;
-
-       nesting = ACCESS_ONCE(t->rcu_read_lock_nesting);
-       if (nesting > 1) {
-
-               /*
-                * We are still protected by the enclosing rcu_read_lock(),
-                * so simply decrement the counter.
-                */
-
-               t->rcu_read_lock_nesting = nesting - 1;
-
-       } else {
-               unsigned long flags;
-
-               /*
-                * Disable local interrupts to prevent the grace-period
-                * detection state machine from seeing us half-done.
-                * NMIs can still occur, of course, and might themselves
-                * contain rcu_read_lock() and rcu_read_unlock().
-                */
-
-               local_irq_save(flags);
-
-               /*
-                * Outermost nesting of rcu_read_unlock(), so we must
-                * decrement the current counter for the current CPU.
-                * This must be done carefully, because NMIs can
-                * occur at any point in this code, and any rcu_read_lock()
-                * and rcu_read_unlock() pairs in the NMI handlers
-                * must interact non-destructively with this code.
-                * Lots of volatile casts, and -very- careful ordering.
-                *
-                * Changes to this code, including this one, must be
-                * inspected, validated, and tested extremely carefully!!!
-                */
-
-               /*
-                * First, pick up the index.
-                */
-
-               idx = ACCESS_ONCE(t->rcu_flipctr_idx);
-
-               /*
-                * Now that we have fetched the counter index, it is
-                * safe to decrement the per-task RCU nesting counter.
-                * After this, any interrupts or NMIs will increment and
-                * decrement the per-CPU counters.
-                */
-               ACCESS_ONCE(t->rcu_read_lock_nesting) = nesting - 1;
-
-               /*
-                * It is now safe to decrement this task's nesting count.
-                * NMIs that occur after this statement will route their
-                * rcu_read_lock() calls through this "else" clause, and
-                * will thus start incrementing the per-CPU counter on
-                * their own.  They will also clobber ->rcu_flipctr_idx,
-                * but that is OK, since we have already fetched it.
-                */
-
-               ACCESS_ONCE(RCU_DATA_ME()->rcu_flipctr[idx])--;
-               local_irq_restore(flags);
-       }
-}
-EXPORT_SYMBOL_GPL(__rcu_read_unlock);
-
-/*
- * If a global counter flip has occurred since the last time that we
- * advanced callbacks, advance them.  Hardware interrupts must be
- * disabled when calling this function.
- */
-static void __rcu_advance_callbacks(struct rcu_data *rdp)
-{
-       int cpu;
-       int i;
-       int wlc = 0;
-
-       if (rdp->completed != rcu_ctrlblk.completed) {
-               if (rdp->waitlist[GP_STAGES - 1] != NULL) {
-                       *rdp->donetail = rdp->waitlist[GP_STAGES - 1];
-                       rdp->donetail = rdp->waittail[GP_STAGES - 1];
-                       RCU_TRACE_RDP(rcupreempt_trace_move2done, rdp);
-               }
-               for (i = GP_STAGES - 2; i >= 0; i--) {
-                       if (rdp->waitlist[i] != NULL) {
-                               rdp->waitlist[i + 1] = rdp->waitlist[i];
-                               rdp->waittail[i + 1] = rdp->waittail[i];
-                               wlc++;
-                       } else {
-                               rdp->waitlist[i + 1] = NULL;
-                               rdp->waittail[i + 1] =
-                                       &rdp->waitlist[i + 1];
-                       }
-               }
-               if (rdp->nextlist != NULL) {
-                       rdp->waitlist[0] = rdp->nextlist;
-                       rdp->waittail[0] = rdp->nexttail;
-                       wlc++;
-                       rdp->nextlist = NULL;
-                       rdp->nexttail = &rdp->nextlist;
-                       RCU_TRACE_RDP(rcupreempt_trace_move2wait, rdp);
-               } else {
-                       rdp->waitlist[0] = NULL;
-                       rdp->waittail[0] = &rdp->waitlist[0];
-               }
-               rdp->waitlistcount = wlc;
-               rdp->completed = rcu_ctrlblk.completed;
-       }
-
-       /*
-        * Check to see if this CPU needs to report that it has seen
-        * the most recent counter flip, thereby declaring that all
-        * subsequent rcu_read_lock() invocations will respect this flip.
-        */
-
-       cpu = raw_smp_processor_id();
-       if (per_cpu(rcu_flip_flag, cpu) == rcu_flipped) {
-               smp_mb();  /* Subsequent counter accesses must see new value */
-               per_cpu(rcu_flip_flag, cpu) = rcu_flip_seen;
-               smp_mb();  /* Subsequent RCU read-side critical sections */
-                          /*  seen -after- acknowledgement. */
-       }
-}
-
-#ifdef CONFIG_NO_HZ
-static DEFINE_PER_CPU(int, rcu_update_flag);
-
-/**
- * rcu_irq_enter - Called from Hard irq handlers and NMI/SMI.
- *
- * If the CPU was idle with dynamic ticks active, this updates the
- * rcu_dyntick_sched.dynticks to let the RCU handling know that the
- * CPU is active.
- */
-void rcu_irq_enter(void)
-{
-       int cpu = smp_processor_id();
-       struct rcu_dyntick_sched *rdssp = &per_cpu(rcu_dyntick_sched, cpu);
-
-       if (per_cpu(rcu_update_flag, cpu))
-               per_cpu(rcu_update_flag, cpu)++;
-
-       /*
-        * Only update if we are coming from a stopped ticks mode
-        * (rcu_dyntick_sched.dynticks is even).
-        */
-       if (!in_interrupt() &&
-           (rdssp->dynticks & 0x1) == 0) {
-               /*
-                * The following might seem like we could have a race
-                * with NMI/SMIs. But this really isn't a problem.
-                * Here we do a read/modify/write, and the race happens
-                * when an NMI/SMI comes in after the read and before
-                * the write. But NMI/SMIs will increment this counter
-                * twice before returning, so the zero bit will not
-                * be corrupted by the NMI/SMI which is the most important
-                * part.
-                *
-                * The only thing is that we would bring back the counter
-                * to a postion that it was in during the NMI/SMI.
-                * But the zero bit would be set, so the rest of the
-                * counter would again be ignored.
-                *
-                * On return from the IRQ, the counter may have the zero
-                * bit be 0 and the counter the same as the return from
-                * the NMI/SMI. If the state machine was so unlucky to
-                * see that, it still doesn't matter, since all
-                * RCU read-side critical sections on this CPU would
-                * have already completed.
-                */
-               rdssp->dynticks++;
-               /*
-                * The following memory barrier ensures that any
-                * rcu_read_lock() primitives in the irq handler
-                * are seen by other CPUs to follow the above
-                * increment to rcu_dyntick_sched.dynticks. This is
-                * required in order for other CPUs to correctly
-                * determine when it is safe to advance the RCU
-                * grace-period state machine.
-                */
-               smp_mb(); /* see above block comment. */
-               /*
-                * Since we can't determine the dynamic tick mode from
-                * the rcu_dyntick_sched.dynticks after this routine,
-                * we use a second flag to acknowledge that we came
-                * from an idle state with ticks stopped.
-                */
-               per_cpu(rcu_update_flag, cpu)++;
-               /*
-                * If we take an NMI/SMI now, they will also increment
-                * the rcu_update_flag, and will not update the
-                * rcu_dyntick_sched.dynticks on exit. That is for
-                * this IRQ to do.
-                */
-       }
-}
-
-/**
- * rcu_irq_exit - Called from exiting Hard irq context.
- *
- * If the CPU was idle with dynamic ticks active, update the
- * rcu_dyntick_sched.dynticks to let the RCU handling be
- * aware that the CPU is going back to idle with no ticks.
- */
-void rcu_irq_exit(void)
-{
-       int cpu = smp_processor_id();
-       struct rcu_dyntick_sched *rdssp = &per_cpu(rcu_dyntick_sched, cpu);
-
-       /*
-        * rcu_update_flag is set if we interrupted the CPU
-        * when it was idle with ticks stopped.
-        * Once this occurs, we keep track of interrupt nesting
-        * because a NMI/SMI could also come in, and we still
-        * only want the IRQ that started the increment of the
-        * rcu_dyntick_sched.dynticks to be the one that modifies
-        * it on exit.
-        */
-       if (per_cpu(rcu_update_flag, cpu)) {
-               if (--per_cpu(rcu_update_flag, cpu))
-                       return;
-
-               /* This must match the interrupt nesting */
-               WARN_ON(in_interrupt());
-
-               /*
-                * If an NMI/SMI happens now we are still
-                * protected by the rcu_dyntick_sched.dynticks being odd.
-                */
-
-               /*
-                * The following memory barrier ensures that any
-                * rcu_read_unlock() primitives in the irq handler
-                * are seen by other CPUs to preceed the following
-                * increment to rcu_dyntick_sched.dynticks. This
-                * is required in order for other CPUs to determine
-                * when it is safe to advance the RCU grace-period
-                * state machine.
-                */
-               smp_mb(); /* see above block comment. */
-               rdssp->dynticks++;
-               WARN_ON(rdssp->dynticks & 0x1);
-       }
-}
-
-void rcu_nmi_enter(void)
-{
-       rcu_irq_enter();
-}
-
-void rcu_nmi_exit(void)
-{
-       rcu_irq_exit();
-}
-
-static void dyntick_save_progress_counter(int cpu)
-{
-       struct rcu_dyntick_sched *rdssp = &per_cpu(rcu_dyntick_sched, cpu);
-
-       rdssp->dynticks_snap = rdssp->dynticks;
-}
-
-static inline int
-rcu_try_flip_waitack_needed(int cpu)
-{
-       long curr;
-       long snap;
-       struct rcu_dyntick_sched *rdssp = &per_cpu(rcu_dyntick_sched, cpu);
-
-       curr = rdssp->dynticks;
-       snap = rdssp->dynticks_snap;
-       smp_mb(); /* force ordering with cpu entering/leaving dynticks. */
-
-       /*
-        * If the CPU remained in dynticks mode for the entire time
-        * and didn't take any interrupts, NMIs, SMIs, or whatever,
-        * then it cannot be in the middle of an rcu_read_lock(), so
-        * the next rcu_read_lock() it executes must use the new value
-        * of the counter.  So we can safely pretend that this CPU
-        * already acknowledged the counter.
-        */
-
-       if ((curr == snap) && ((curr & 0x1) == 0))
-               return 0;
-
-       /*
-        * If the CPU passed through or entered a dynticks idle phase with
-        * no active irq handlers, then, as above, we can safely pretend
-        * that this CPU already acknowledged the counter.
-        */
-
-       if ((curr - snap) > 2 || (curr & 0x1) == 0)
-               return 0;
-
-       /* We need this CPU to explicitly acknowledge the counter flip. */
-
-       return 1;
-}
-
-static inline int
-rcu_try_flip_waitmb_needed(int cpu)
-{
-       long curr;
-       long snap;
-       struct rcu_dyntick_sched *rdssp = &per_cpu(rcu_dyntick_sched, cpu);
-
-       curr = rdssp->dynticks;
-       snap = rdssp->dynticks_snap;
-       smp_mb(); /* force ordering with cpu entering/leaving dynticks. */
-
-       /*
-        * If the CPU remained in dynticks mode for the entire time
-        * and didn't take any interrupts, NMIs, SMIs, or whatever,
-        * then it cannot have executed an RCU read-side critical section
-        * during that time, so there is no need for it to execute a
-        * memory barrier.
-        */
-
-       if ((curr == snap) && ((curr & 0x1) == 0))
-               return 0;
-
-       /*
-        * If the CPU either entered or exited an outermost interrupt,
-        * SMI, NMI, or whatever handler, then we know that it executed
-        * a memory barrier when doing so.  So we don't need another one.
-        */
-       if (curr != snap)
-               return 0;
-
-       /* We need the CPU to execute a memory barrier. */
-
-       return 1;
-}
-
-static void dyntick_save_progress_counter_sched(int cpu)
-{
-       struct rcu_dyntick_sched *rdssp = &per_cpu(rcu_dyntick_sched, cpu);
-
-       rdssp->sched_dynticks_snap = rdssp->dynticks;
-}
-
-static int rcu_qsctr_inc_needed_dyntick(int cpu)
-{
-       long curr;
-       long snap;
-       struct rcu_dyntick_sched *rdssp = &per_cpu(rcu_dyntick_sched, cpu);
-
-       curr = rdssp->dynticks;
-       snap = rdssp->sched_dynticks_snap;
-       smp_mb(); /* force ordering with cpu entering/leaving dynticks. */
-
-       /*
-        * If the CPU remained in dynticks mode for the entire time
-        * and didn't take any interrupts, NMIs, SMIs, or whatever,
-        * then it cannot be in the middle of an rcu_read_lock(), so
-        * the next rcu_read_lock() it executes must use the new value
-        * of the counter.  Therefore, this CPU has been in a quiescent
-        * state the entire time, and we don't need to wait for it.
-        */
-
-       if ((curr == snap) && ((curr & 0x1) == 0))
-               return 0;
-
-       /*
-        * If the CPU passed through or entered a dynticks idle phase with
-        * no active irq handlers, then, as above, this CPU has already
-        * passed through a quiescent state.
-        */
-
-       if ((curr - snap) > 2 || (snap & 0x1) == 0)
-               return 0;
-
-       /* We need this CPU to go through a quiescent state. */
-
-       return 1;
-}
-
-#else /* !CONFIG_NO_HZ */
-
-# define dyntick_save_progress_counter(cpu)            do { } while (0)
-# define rcu_try_flip_waitack_needed(cpu)              (1)
-# define rcu_try_flip_waitmb_needed(cpu)               (1)
-
-# define dyntick_save_progress_counter_sched(cpu)      do { } while (0)
-# define rcu_qsctr_inc_needed_dyntick(cpu)             (1)
-
-#endif /* CONFIG_NO_HZ */
-
-static void save_qsctr_sched(int cpu)
-{
-       struct rcu_dyntick_sched *rdssp = &per_cpu(rcu_dyntick_sched, cpu);
-
-       rdssp->sched_qs_snap = rdssp->sched_qs;
-}
-
-static inline int rcu_qsctr_inc_needed(int cpu)
-{
-       struct rcu_dyntick_sched *rdssp = &per_cpu(rcu_dyntick_sched, cpu);
-
-       /*
-        * If there has been a quiescent state, no more need to wait
-        * on this CPU.
-        */
-
-       if (rdssp->sched_qs != rdssp->sched_qs_snap) {
-               smp_mb(); /* force ordering with cpu entering schedule(). */
-               return 0;
-       }
-
-       /* We need this CPU to go through a quiescent state. */
-
-       return 1;
-}
-
-/*
- * Get here when RCU is idle.  Decide whether we need to
- * move out of idle state, and return non-zero if so.
- * "Straightforward" approach for the moment, might later
- * use callback-list lengths, grace-period duration, or
- * some such to determine when to exit idle state.
- * Might also need a pre-idle test that does not acquire
- * the lock, but let's get the simple case working first...
- */
-
-static int
-rcu_try_flip_idle(void)
-{
-       int cpu;
-
-       RCU_TRACE_ME(rcupreempt_trace_try_flip_i1);
-       if (!rcu_pending(smp_processor_id())) {
-               RCU_TRACE_ME(rcupreempt_trace_try_flip_ie1);
-               return 0;
-       }
-
-       /*
-        * Do the flip.
-        */
-
-       RCU_TRACE_ME(rcupreempt_trace_try_flip_g1);
-       rcu_ctrlblk.completed++;  /* stands in for rcu_try_flip_g2 */
-
-       /*
-        * Need a memory barrier so that other CPUs see the new
-        * counter value before they see the subsequent change of all
-        * the rcu_flip_flag instances to rcu_flipped.
-        */
-
-       smp_mb();       /* see above block comment. */
-
-       /* Now ask each CPU for acknowledgement of the flip. */
-
-       for_each_cpu(cpu, to_cpumask(rcu_cpu_online_map)) {
-               per_cpu(rcu_flip_flag, cpu) = rcu_flipped;
-               dyntick_save_progress_counter(cpu);
-       }
-
-       return 1;
-}
-
-/*
- * Wait for CPUs to acknowledge the flip.
- */
-
-static int
-rcu_try_flip_waitack(void)
-{
-       int cpu;
-
-       RCU_TRACE_ME(rcupreempt_trace_try_flip_a1);
-       for_each_cpu(cpu, to_cpumask(rcu_cpu_online_map))
-               if (rcu_try_flip_waitack_needed(cpu) &&
-                   per_cpu(rcu_flip_flag, cpu) != rcu_flip_seen) {
-                       RCU_TRACE_ME(rcupreempt_trace_try_flip_ae1);
-                       return 0;
-               }
-
-       /*
-        * Make sure our checks above don't bleed into subsequent
-        * waiting for the sum of the counters to reach zero.
-        */
-
-       smp_mb();       /* see above block comment. */
-       RCU_TRACE_ME(rcupreempt_trace_try_flip_a2);
-       return 1;
-}
-
-/*
- * Wait for collective ``last'' counter to reach zero,
- * then tell all CPUs to do an end-of-grace-period memory barrier.
- */
-
-static int
-rcu_try_flip_waitzero(void)
-{
-       int cpu;
-       int lastidx = !(rcu_ctrlblk.completed & 0x1);
-       int sum = 0;
-
-       /* Check to see if the sum of the "last" counters is zero. */
-
-       RCU_TRACE_ME(rcupreempt_trace_try_flip_z1);
-       for_each_possible_cpu(cpu)
-               sum += RCU_DATA_CPU(cpu)->rcu_flipctr[lastidx];
-       if (sum != 0) {
-               RCU_TRACE_ME(rcupreempt_trace_try_flip_ze1);
-               return 0;
-       }
-
-       /*
-        * This ensures that the other CPUs see the call for
-        * memory barriers -after- the sum to zero has been
-        * detected here
-        */
-       smp_mb();  /*  ^^^^^^^^^^^^ */
-
-       /* Call for a memory barrier from each CPU. */
-       for_each_cpu(cpu, to_cpumask(rcu_cpu_online_map)) {
-               per_cpu(rcu_mb_flag, cpu) = rcu_mb_needed;
-               dyntick_save_progress_counter(cpu);
-       }
-
-       RCU_TRACE_ME(rcupreempt_trace_try_flip_z2);
-       return 1;
-}
-
-/*
- * Wait for all CPUs to do their end-of-grace-period memory barrier.
- * Return 0 once all CPUs have done so.
- */
-
-static int
-rcu_try_flip_waitmb(void)
-{
-       int cpu;
-
-       RCU_TRACE_ME(rcupreempt_trace_try_flip_m1);
-       for_each_cpu(cpu, to_cpumask(rcu_cpu_online_map))
-               if (rcu_try_flip_waitmb_needed(cpu) &&
-                   per_cpu(rcu_mb_flag, cpu) != rcu_mb_done) {
-                       RCU_TRACE_ME(rcupreempt_trace_try_flip_me1);
-                       return 0;
-               }
-
-       smp_mb(); /* Ensure that the above checks precede any following flip. */
-       RCU_TRACE_ME(rcupreempt_trace_try_flip_m2);
-       return 1;
-}
-
-/*
- * Attempt a single flip of the counters.  Remember, a single flip does
- * -not- constitute a grace period.  Instead, the interval between
- * at least GP_STAGES consecutive flips is a grace period.
- *
- * If anyone is nuts enough to run this CONFIG_PREEMPT_RCU implementation
- * on a large SMP, they might want to use a hierarchical organization of
- * the per-CPU-counter pairs.
- */
-static void rcu_try_flip(void)
-{
-       unsigned long flags;
-
-       RCU_TRACE_ME(rcupreempt_trace_try_flip_1);
-       if (unlikely(!spin_trylock_irqsave(&rcu_ctrlblk.fliplock, flags))) {
-               RCU_TRACE_ME(rcupreempt_trace_try_flip_e1);
-               return;
-       }
-
-       /*
-        * Take the next transition(s) through the RCU grace-period
-        * flip-counter state machine.
-        */
-
-       switch (rcu_ctrlblk.rcu_try_flip_state) {
-       case rcu_try_flip_idle_state:
-               if (rcu_try_flip_idle())
-                       rcu_ctrlblk.rcu_try_flip_state =
-                               rcu_try_flip_waitack_state;
-               break;
-       case rcu_try_flip_waitack_state:
-               if (rcu_try_flip_waitack())
-                       rcu_ctrlblk.rcu_try_flip_state =
-                               rcu_try_flip_waitzero_state;
-               break;
-       case rcu_try_flip_waitzero_state:
-               if (rcu_try_flip_waitzero())
-                       rcu_ctrlblk.rcu_try_flip_state =
-                               rcu_try_flip_waitmb_state;
-               break;
-       case rcu_try_flip_waitmb_state:
-               if (rcu_try_flip_waitmb())
-                       rcu_ctrlblk.rcu_try_flip_state =
-                               rcu_try_flip_idle_state;
-       }
-       spin_unlock_irqrestore(&rcu_ctrlblk.fliplock, flags);
-}
-
-/*
- * Check to see if this CPU needs to do a memory barrier in order to
- * ensure that any prior RCU read-side critical sections have committed
- * their counter manipulations and critical-section memory references
- * before declaring the grace period to be completed.
- */
-static void rcu_check_mb(int cpu)
-{
-       if (per_cpu(rcu_mb_flag, cpu) == rcu_mb_needed) {
-               smp_mb();  /* Ensure RCU read-side accesses are visible. */
-               per_cpu(rcu_mb_flag, cpu) = rcu_mb_done;
-       }
-}
-
-void rcu_check_callbacks(int cpu, int user)
-{
-       unsigned long flags;
-       struct rcu_data *rdp;
-
-       if (!rcu_pending(cpu))
-               return; /* if nothing for RCU to do. */
-
-       /*
-        * If this CPU took its interrupt from user mode or from the
-        * idle loop, and this is not a nested interrupt, then
-        * this CPU has to have exited all prior preept-disable
-        * sections of code.  So invoke rcu_sched_qs() to note this.
-        *
-        * The memory barrier is needed to handle the case where
-        * writes from a preempt-disable section of code get reordered
-        * into schedule() by this CPU's write buffer.  So the memory
-        * barrier makes sure that the rcu_sched_qs() is seen by other
-        * CPUs to happen after any such write.
-        */
-
-       rdp = RCU_DATA_CPU(cpu);
-       if (user ||
-           (idle_cpu(cpu) && !in_softirq() &&
-            hardirq_count() <= (1 << HARDIRQ_SHIFT))) {
-               smp_mb();       /* Guard against aggressive schedule(). */
-               rcu_sched_qs(cpu);
-       }
-
-       rcu_check_mb(cpu);
-       if (rcu_ctrlblk.completed == rdp->completed)
-               rcu_try_flip();
-       spin_lock_irqsave(&rdp->lock, flags);
-       RCU_TRACE_RDP(rcupreempt_trace_check_callbacks, rdp);
-       __rcu_advance_callbacks(rdp);
-       if (rdp->donelist == NULL) {
-               spin_unlock_irqrestore(&rdp->lock, flags);
-       } else {
-               spin_unlock_irqrestore(&rdp->lock, flags);
-               raise_softirq(RCU_SOFTIRQ);
-       }
-}
-
-/*
- * Needed by dynticks, to make sure all RCU processing has finished
- * when we go idle:
- */
-void rcu_advance_callbacks(int cpu, int user)
-{
-       unsigned long flags;
-       struct rcu_data *rdp = RCU_DATA_CPU(cpu);
-
-       if (rcu_ctrlblk.completed == rdp->completed) {
-               rcu_try_flip();
-               if (rcu_ctrlblk.completed == rdp->completed)
-                       return;
-       }
-       spin_lock_irqsave(&rdp->lock, flags);
-       RCU_TRACE_RDP(rcupreempt_trace_check_callbacks, rdp);
-       __rcu_advance_callbacks(rdp);
-       spin_unlock_irqrestore(&rdp->lock, flags);
-}
-
-#ifdef CONFIG_HOTPLUG_CPU
-#define rcu_offline_cpu_enqueue(srclist, srctail, dstlist, dsttail) do { \
-               *dsttail = srclist; \
-               if (srclist != NULL) { \
-                       dsttail = srctail; \
-                       srclist = NULL; \
-                       srctail = &srclist;\
-               } \
-       } while (0)
-
-void rcu_offline_cpu(int cpu)
-{
-       int i;
-       struct rcu_head *list = NULL;
-       unsigned long flags;
-       struct rcu_data *rdp = RCU_DATA_CPU(cpu);
-       struct rcu_head *schedlist = NULL;
-       struct rcu_head **schedtail = &schedlist;
-       struct rcu_head **tail = &list;
-
-       /*
-        * Remove all callbacks from the newly dead CPU, retaining order.
-        * Otherwise rcu_barrier() will fail
-        */
-
-       spin_lock_irqsave(&rdp->lock, flags);
-       rcu_offline_cpu_enqueue(rdp->donelist, rdp->donetail, list, tail);
-       for (i = GP_STAGES - 1; i >= 0; i--)
-               rcu_offline_cpu_enqueue(rdp->waitlist[i], rdp->waittail[i],
-                                               list, tail);
-       rcu_offline_cpu_enqueue(rdp->nextlist, rdp->nexttail, list, tail);
-       rcu_offline_cpu_enqueue(rdp->waitschedlist, rdp->waitschedtail,
-                               schedlist, schedtail);
-       rcu_offline_cpu_enqueue(rdp->nextschedlist, rdp->nextschedtail,
-                               schedlist, schedtail);
-       rdp->rcu_sched_sleeping = 0;
-       spin_unlock_irqrestore(&rdp->lock, flags);
-       rdp->waitlistcount = 0;
-
-       /* Disengage the newly dead CPU from the grace-period computation. */
-
-       spin_lock_irqsave(&rcu_ctrlblk.fliplock, flags);
-       rcu_check_mb(cpu);
-       if (per_cpu(rcu_flip_flag, cpu) == rcu_flipped) {
-               smp_mb();  /* Subsequent counter accesses must see new value */
-               per_cpu(rcu_flip_flag, cpu) = rcu_flip_seen;
-               smp_mb();  /* Subsequent RCU read-side critical sections */
-                          /*  seen -after- acknowledgement. */
-       }
-
-       cpumask_clear_cpu(cpu, to_cpumask(rcu_cpu_online_map));
-
-       spin_unlock_irqrestore(&rcu_ctrlblk.fliplock, flags);
-
-       /*
-        * Place the removed callbacks on the current CPU's queue.
-        * Make them all start a new grace period: simple approach,
-        * in theory could starve a given set of callbacks, but
-        * you would need to be doing some serious CPU hotplugging
-        * to make this happen.  If this becomes a problem, adding
-        * a synchronize_rcu() to the hotplug path would be a simple
-        * fix.
-        */
-
-       local_irq_save(flags);  /* disable preempt till we know what lock. */
-       rdp = RCU_DATA_ME();
-       spin_lock(&rdp->lock);
-       *rdp->nexttail = list;
-       if (list)
-               rdp->nexttail = tail;
-       *rdp->nextschedtail = schedlist;
-       if (schedlist)
-               rdp->nextschedtail = schedtail;
-       spin_unlock_irqrestore(&rdp->lock, flags);
-}
-
-#else /* #ifdef CONFIG_HOTPLUG_CPU */
-
-void rcu_offline_cpu(int cpu)
-{
-}
-
-#endif /* #else #ifdef CONFIG_HOTPLUG_CPU */
-
-void __cpuinit rcu_online_cpu(int cpu)
-{
-       unsigned long flags;
-       struct rcu_data *rdp;
-
-       spin_lock_irqsave(&rcu_ctrlblk.fliplock, flags);
-       cpumask_set_cpu(cpu, to_cpumask(rcu_cpu_online_map));
-       spin_unlock_irqrestore(&rcu_ctrlblk.fliplock, flags);
-
-       /*
-        * The rcu_sched grace-period processing might have bypassed
-        * this CPU, given that it was not in the rcu_cpu_online_map
-        * when the grace-period scan started.  This means that the
-        * grace-period task might sleep.  So make sure that if this
-        * should happen, the first callback posted to this CPU will
-        * wake up the grace-period task if need be.
-        */
-
-       rdp = RCU_DATA_CPU(cpu);
-       spin_lock_irqsave(&rdp->lock, flags);
-       rdp->rcu_sched_sleeping = 1;
-       spin_unlock_irqrestore(&rdp->lock, flags);
-}
-
-static void rcu_process_callbacks(struct softirq_action *unused)
-{
-       unsigned long flags;
-       struct rcu_head *next, *list;
-       struct rcu_data *rdp;
-
-       local_irq_save(flags);
-       rdp = RCU_DATA_ME();
-       spin_lock(&rdp->lock);
-       list = rdp->donelist;
-       if (list == NULL) {
-               spin_unlock_irqrestore(&rdp->lock, flags);
-               return;
-       }
-       rdp->donelist = NULL;
-       rdp->donetail = &rdp->donelist;
-       RCU_TRACE_RDP(rcupreempt_trace_done_remove, rdp);
-       spin_unlock_irqrestore(&rdp->lock, flags);
-       while (list) {
-               next = list->next;
-               list->func(list);
-               list = next;
-               RCU_TRACE_ME(rcupreempt_trace_invoke);
-       }
-}
-
-void call_rcu(struct rcu_head *head, void (*func)(struct rcu_head *rcu))
-{
-       unsigned long flags;
-       struct rcu_data *rdp;
-
-       head->func = func;
-       head->next = NULL;
-       local_irq_save(flags);
-       rdp = RCU_DATA_ME();
-       spin_lock(&rdp->lock);
-       __rcu_advance_callbacks(rdp);
-       *rdp->nexttail = head;
-       rdp->nexttail = &head->next;
-       RCU_TRACE_RDP(rcupreempt_trace_next_add, rdp);
-       spin_unlock_irqrestore(&rdp->lock, flags);
-}
-EXPORT_SYMBOL_GPL(call_rcu);
-
-void call_rcu_sched(struct rcu_head *head, void (*func)(struct rcu_head *rcu))
-{
-       unsigned long flags;
-       struct rcu_data *rdp;
-       int wake_gp = 0;
-
-       head->func = func;
-       head->next = NULL;
-       local_irq_save(flags);
-       rdp = RCU_DATA_ME();
-       spin_lock(&rdp->lock);
-       *rdp->nextschedtail = head;
-       rdp->nextschedtail = &head->next;
-       if (rdp->rcu_sched_sleeping) {
-
-               /* Grace-period processing might be sleeping... */
-
-               rdp->rcu_sched_sleeping = 0;
-               wake_gp = 1;
-       }
-       spin_unlock_irqrestore(&rdp->lock, flags);
-       if (wake_gp) {
-
-               /* Wake up grace-period processing, unless someone beat us. */
-
-               spin_lock_irqsave(&rcu_ctrlblk.schedlock, flags);
-               if (rcu_ctrlblk.sched_sleep != rcu_sched_sleeping)
-                       wake_gp = 0;
-               rcu_ctrlblk.sched_sleep = rcu_sched_not_sleeping;
-               spin_unlock_irqrestore(&rcu_ctrlblk.schedlock, flags);
-               if (wake_gp)
-                       wake_up_interruptible(&rcu_ctrlblk.sched_wq);
-       }
-}
-EXPORT_SYMBOL_GPL(call_rcu_sched);
-
-/*
- * Wait until all currently running preempt_disable() code segments
- * (including hardware-irq-disable segments) complete.  Note that
- * in -rt this does -not- necessarily result in all currently executing
- * interrupt -handlers- having completed.
- */
-void __synchronize_sched(void)
-{
-       struct rcu_synchronize rcu;
-
-       if (num_online_cpus() == 1)
-               return;  /* blocking is gp if only one CPU! */
-
-       init_completion(&rcu.completion);
-       /* Will wake me after RCU finished. */
-       call_rcu_sched(&rcu.head, wakeme_after_rcu);
-       /* Wait for it. */
-       wait_for_completion(&rcu.completion);
-}
-EXPORT_SYMBOL_GPL(__synchronize_sched);
-
-/*
- * kthread function that manages call_rcu_sched grace periods.
- */
-static int rcu_sched_grace_period(void *arg)
-{
-       int couldsleep;         /* might sleep after current pass. */
-       int couldsleepnext = 0; /* might sleep after next pass. */
-       int cpu;
-       unsigned long flags;
-       struct rcu_data *rdp;
-       int ret;
-
-       /*
-        * Each pass through the following loop handles one
-        * rcu_sched grace period cycle.
-        */
-       do {
-               /* Save each CPU's current state. */
-
-               for_each_online_cpu(cpu) {
-                       dyntick_save_progress_counter_sched(cpu);
-                       save_qsctr_sched(cpu);
-               }
-
-               /*
-                * Sleep for about an RCU grace-period's worth to
-                * allow better batching and to consume less CPU.
-                */
-               schedule_timeout_interruptible(RCU_SCHED_BATCH_TIME);
-
-               /*
-                * If there was nothing to do last time, prepare to
-                * sleep at the end of the current grace period cycle.
-                */
-               couldsleep = couldsleepnext;
-               couldsleepnext = 1;
-               if (couldsleep) {
-                       spin_lock_irqsave(&rcu_ctrlblk.schedlock, flags);
-                       rcu_ctrlblk.sched_sleep = rcu_sched_sleep_prep;
-                       spin_unlock_irqrestore(&rcu_ctrlblk.schedlock, flags);
-               }
-
-               /*
-                * Wait on each CPU in turn to have either visited
-                * a quiescent state or been in dynticks-idle mode.
-                */
-               for_each_online_cpu(cpu) {
-                       while (rcu_qsctr_inc_needed(cpu) &&
-                              rcu_qsctr_inc_needed_dyntick(cpu)) {
-                               /* resched_cpu(cpu); @@@ */
-                               schedule_timeout_interruptible(1);
-                       }
-               }
-
-               /* Advance callbacks for each CPU.  */
-
-               for_each_online_cpu(cpu) {
-
-                       rdp = RCU_DATA_CPU(cpu);
-                       spin_lock_irqsave(&rdp->lock, flags);
-
-                       /*
-                        * We are running on this CPU irq-disabled, so no
-                        * CPU can go offline until we re-enable irqs.
-                        * The current CPU might have already gone
-                        * offline (between the for_each_offline_cpu and
-                        * the spin_lock_irqsave), but in that case all its
-                        * callback lists will be empty, so no harm done.
-                        *
-                        * Advance the callbacks!  We share normal RCU's
-                        * donelist, since callbacks are invoked the
-                        * same way in either case.
-                        */
-                       if (rdp->waitschedlist != NULL) {
-                               *rdp->donetail = rdp->waitschedlist;
-                               rdp->donetail = rdp->waitschedtail;
-
-                               /*
-                                * Next rcu_check_callbacks() will
-                                * do the required raise_softirq().
-                                */
-                       }
-                       if (rdp->nextschedlist != NULL) {
-                               rdp->waitschedlist = rdp->nextschedlist;
-                               rdp->waitschedtail = rdp->nextschedtail;
-                               couldsleep = 0;
-                               couldsleepnext = 0;
-                       } else {
-                               rdp->waitschedlist = NULL;
-                               rdp->waitschedtail = &rdp->waitschedlist;
-                       }
-                       rdp->nextschedlist = NULL;
-                       rdp->nextschedtail = &rdp->nextschedlist;
-
-                       /* Mark sleep intention. */
-
-                       rdp->rcu_sched_sleeping = couldsleep;
-
-                       spin_unlock_irqrestore(&rdp->lock, flags);
-               }
-
-               /* If we saw callbacks on the last scan, go deal with them. */
-
-               if (!couldsleep)
-                       continue;
-
-               /* Attempt to block... */
-
-               spin_lock_irqsave(&rcu_ctrlblk.schedlock, flags);
-               if (rcu_ctrlblk.sched_sleep != rcu_sched_sleep_prep) {
-
-                       /*
-                        * Someone posted a callback after we scanned.
-                        * Go take care of it.
-                        */
-                       spin_unlock_irqrestore(&rcu_ctrlblk.schedlock, flags);
-                       couldsleepnext = 0;
-                       continue;
-               }
-
-               /* Block until the next person posts a callback. */
-
-               rcu_ctrlblk.sched_sleep = rcu_sched_sleeping;
-               spin_unlock_irqrestore(&rcu_ctrlblk.schedlock, flags);
-               ret = 0; /* unused */
-               __wait_event_interruptible(rcu_ctrlblk.sched_wq,
-                       rcu_ctrlblk.sched_sleep != rcu_sched_sleeping,
-                       ret);
-
-               couldsleepnext = 0;
-
-       } while (!kthread_should_stop());
-
-       return (0);
-}
-
-/*
- * Check to see if any future RCU-related work will need to be done
- * by the current CPU, even if none need be done immediately, returning
- * 1 if so.  Assumes that notifiers would take care of handling any
- * outstanding requests from the RCU core.
- *
- * This function is part of the RCU implementation; it is -not-
- * an exported member of the RCU API.
- */
-int rcu_needs_cpu(int cpu)
-{
-       struct rcu_data *rdp = RCU_DATA_CPU(cpu);
-
-       return (rdp->donelist != NULL ||
-               !!rdp->waitlistcount ||
-               rdp->nextlist != NULL ||
-               rdp->nextschedlist != NULL ||
-               rdp->waitschedlist != NULL);
-}
-
-static int rcu_pending(int cpu)
-{
-       struct rcu_data *rdp = RCU_DATA_CPU(cpu);
-
-       /* The CPU has at least one callback queued somewhere. */
-
-       if (rdp->donelist != NULL ||
-           !!rdp->waitlistcount ||
-           rdp->nextlist != NULL ||
-           rdp->nextschedlist != NULL ||
-           rdp->waitschedlist != NULL)
-               return 1;
-
-       /* The RCU core needs an acknowledgement from this CPU. */
-
-       if ((per_cpu(rcu_flip_flag, cpu) == rcu_flipped) ||
-           (per_cpu(rcu_mb_flag, cpu) == rcu_mb_needed))
-               return 1;
-
-       /* This CPU has fallen behind the global grace-period number. */
-
-       if (rdp->completed != rcu_ctrlblk.completed)
-               return 1;
-
-       /* Nothing needed from this CPU. */
-
-       return 0;
-}
-
-int __cpuinit rcu_cpu_notify(struct notifier_block *self,
-                            unsigned long action, void *hcpu)
-{
-       long cpu = (long)hcpu;
-
-       switch (action) {
-       case CPU_UP_PREPARE:
-       case CPU_UP_PREPARE_FROZEN:
-               rcu_online_cpu(cpu);
-               break;
-       case CPU_UP_CANCELED:
-       case CPU_UP_CANCELED_FROZEN:
-       case CPU_DEAD:
-       case CPU_DEAD_FROZEN:
-               rcu_offline_cpu(cpu);
-               break;
-       default:
-               break;
-       }
-       return NOTIFY_OK;
-}
-
-void __init __rcu_init(void)
-{
-       int cpu;
-       int i;
-       struct rcu_data *rdp;
-
-       printk(KERN_NOTICE "Preemptible RCU implementation.\n");
-       for_each_possible_cpu(cpu) {
-               rdp = RCU_DATA_CPU(cpu);
-               spin_lock_init(&rdp->lock);
-               rdp->completed = 0;
-               rdp->waitlistcount = 0;
-               rdp->nextlist = NULL;
-               rdp->nexttail = &rdp->nextlist;
-               for (i = 0; i < GP_STAGES; i++) {
-                       rdp->waitlist[i] = NULL;
-                       rdp->waittail[i] = &rdp->waitlist[i];
-               }
-               rdp->donelist = NULL;
-               rdp->donetail = &rdp->donelist;
-               rdp->rcu_flipctr[0] = 0;
-               rdp->rcu_flipctr[1] = 0;
-               rdp->nextschedlist = NULL;
-               rdp->nextschedtail = &rdp->nextschedlist;
-               rdp->waitschedlist = NULL;
-               rdp->waitschedtail = &rdp->waitschedlist;
-               rdp->rcu_sched_sleeping = 0;
-       }
-       open_softirq(RCU_SOFTIRQ, rcu_process_callbacks);
-}
-
-/*
- * Late-boot-time RCU initialization that must wait until after scheduler
- * has been initialized.
- */
-void __init rcu_init_sched(void)
-{
-       rcu_sched_grace_period_task = kthread_run(rcu_sched_grace_period,
-                                                 NULL,
-                                                 "rcu_sched_grace_period");
-       WARN_ON(IS_ERR(rcu_sched_grace_period_task));
-}
-
-#ifdef CONFIG_RCU_TRACE
-long *rcupreempt_flipctr(int cpu)
-{
-       return &RCU_DATA_CPU(cpu)->rcu_flipctr[0];
-}
-EXPORT_SYMBOL_GPL(rcupreempt_flipctr);
-
-int rcupreempt_flip_flag(int cpu)
-{
-       return per_cpu(rcu_flip_flag, cpu);
-}
-EXPORT_SYMBOL_GPL(rcupreempt_flip_flag);
-
-int rcupreempt_mb_flag(int cpu)
-{
-       return per_cpu(rcu_mb_flag, cpu);
-}
-EXPORT_SYMBOL_GPL(rcupreempt_mb_flag);
-
-char *rcupreempt_try_flip_state_name(void)
-{
-       return rcu_try_flip_state_names[rcu_ctrlblk.rcu_try_flip_state];
-}
-EXPORT_SYMBOL_GPL(rcupreempt_try_flip_state_name);
-
-struct rcupreempt_trace *rcupreempt_trace_cpu(int cpu)
-{
-       struct rcu_data *rdp = RCU_DATA_CPU(cpu);
-
-       return &rdp->trace;
-}
-EXPORT_SYMBOL_GPL(rcupreempt_trace_cpu);
-
-#endif /* #ifdef RCU_TRACE */
diff --git a/kernel/rcupreempt_trace.c b/kernel/rcupreempt_trace.c
deleted file mode 100644 (file)
index 1164034..0000000
+++ /dev/null
@@ -1,335 +0,0 @@
-/*
- * Read-Copy Update tracing for realtime implementation
- *
- * 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, 2006
- *
- * Papers:  http://www.rdrop.com/users/paulmck/RCU
- *
- * For detailed explanation of Read-Copy Update mechanism see -
- *             Documentation/RCU/ *.txt
- *
- */
-#include <linux/types.h>
-#include <linux/kernel.h>
-#include <linux/init.h>
-#include <linux/spinlock.h>
-#include <linux/smp.h>
-#include <linux/rcupdate.h>
-#include <linux/interrupt.h>
-#include <linux/sched.h>
-#include <asm/atomic.h>
-#include <linux/bitops.h>
-#include <linux/module.h>
-#include <linux/completion.h>
-#include <linux/moduleparam.h>
-#include <linux/percpu.h>
-#include <linux/notifier.h>
-#include <linux/cpu.h>
-#include <linux/mutex.h>
-#include <linux/rcupreempt_trace.h>
-#include <linux/debugfs.h>
-
-static struct mutex rcupreempt_trace_mutex;
-static char *rcupreempt_trace_buf;
-#define RCUPREEMPT_TRACE_BUF_SIZE 4096
-
-void rcupreempt_trace_move2done(struct rcupreempt_trace *trace)
-{
-       trace->done_length += trace->wait_length;
-       trace->done_add += trace->wait_length;
-       trace->wait_length = 0;
-}
-void rcupreempt_trace_move2wait(struct rcupreempt_trace *trace)
-{
-       trace->wait_length += trace->next_length;
-       trace->wait_add += trace->next_length;
-       trace->next_length = 0;
-}
-void rcupreempt_trace_try_flip_1(struct rcupreempt_trace *trace)
-{
-       atomic_inc(&trace->rcu_try_flip_1);
-}
-void rcupreempt_trace_try_flip_e1(struct rcupreempt_trace *trace)
-{
-       atomic_inc(&trace->rcu_try_flip_e1);
-}
-void rcupreempt_trace_try_flip_i1(struct rcupreempt_trace *trace)
-{
-       trace->rcu_try_flip_i1++;
-}
-void rcupreempt_trace_try_flip_ie1(struct rcupreempt_trace *trace)
-{
-       trace->rcu_try_flip_ie1++;
-}
-void rcupreempt_trace_try_flip_g1(struct rcupreempt_trace *trace)
-{
-       trace->rcu_try_flip_g1++;
-}
-void rcupreempt_trace_try_flip_a1(struct rcupreempt_trace *trace)
-{
-       trace->rcu_try_flip_a1++;
-}
-void rcupreempt_trace_try_flip_ae1(struct rcupreempt_trace *trace)
-{
-       trace->rcu_try_flip_ae1++;
-}
-void rcupreempt_trace_try_flip_a2(struct rcupreempt_trace *trace)
-{
-       trace->rcu_try_flip_a2++;
-}
-void rcupreempt_trace_try_flip_z1(struct rcupreempt_trace *trace)
-{
-       trace->rcu_try_flip_z1++;
-}
-void rcupreempt_trace_try_flip_ze1(struct rcupreempt_trace *trace)
-{
-       trace->rcu_try_flip_ze1++;
-}
-void rcupreempt_trace_try_flip_z2(struct rcupreempt_trace *trace)
-{
-       trace->rcu_try_flip_z2++;
-}
-void rcupreempt_trace_try_flip_m1(struct rcupreempt_trace *trace)
-{
-       trace->rcu_try_flip_m1++;
-}
-void rcupreempt_trace_try_flip_me1(struct rcupreempt_trace *trace)
-{
-       trace->rcu_try_flip_me1++;
-}
-void rcupreempt_trace_try_flip_m2(struct rcupreempt_trace *trace)
-{
-       trace->rcu_try_flip_m2++;
-}
-void rcupreempt_trace_check_callbacks(struct rcupreempt_trace *trace)
-{
-       trace->rcu_check_callbacks++;
-}
-void rcupreempt_trace_done_remove(struct rcupreempt_trace *trace)
-{
-       trace->done_remove += trace->done_length;
-       trace->done_length = 0;
-}
-void rcupreempt_trace_invoke(struct rcupreempt_trace *trace)
-{
-       atomic_inc(&trace->done_invoked);
-}
-void rcupreempt_trace_next_add(struct rcupreempt_trace *trace)
-{
-       trace->next_add++;
-       trace->next_length++;
-}
-
-static void rcupreempt_trace_sum(struct rcupreempt_trace *sp)
-{
-       struct rcupreempt_trace *cp;
-       int cpu;
-
-       memset(sp, 0, sizeof(*sp));
-       for_each_possible_cpu(cpu) {
-               cp = rcupreempt_trace_cpu(cpu);
-               sp->next_length += cp->next_length;
-               sp->next_add += cp->next_add;
-               sp->wait_length += cp->wait_length;
-               sp->wait_add += cp->wait_add;
-               sp->done_length += cp->done_length;
-               sp->done_add += cp->done_add;
-               sp->done_remove += cp->done_remove;
-               atomic_add(atomic_read(&cp->done_invoked), &sp->done_invoked);
-               sp->rcu_check_callbacks += cp->rcu_check_callbacks;
-               atomic_add(atomic_read(&cp->rcu_try_flip_1),
-                          &sp->rcu_try_flip_1);
-               atomic_add(atomic_read(&cp->rcu_try_flip_e1),
-                          &sp->rcu_try_flip_e1);
-               sp->rcu_try_flip_i1 += cp->rcu_try_flip_i1;
-               sp->rcu_try_flip_ie1 += cp->rcu_try_flip_ie1;
-               sp->rcu_try_flip_g1 += cp->rcu_try_flip_g1;
-               sp->rcu_try_flip_a1 += cp->rcu_try_flip_a1;
-               sp->rcu_try_flip_ae1 += cp->rcu_try_flip_ae1;
-               sp->rcu_try_flip_a2 += cp->rcu_try_flip_a2;
-               sp->rcu_try_flip_z1 += cp->rcu_try_flip_z1;
-               sp->rcu_try_flip_ze1 += cp->rcu_try_flip_ze1;
-               sp->rcu_try_flip_z2 += cp->rcu_try_flip_z2;
-               sp->rcu_try_flip_m1 += cp->rcu_try_flip_m1;
-               sp->rcu_try_flip_me1 += cp->rcu_try_flip_me1;
-               sp->rcu_try_flip_m2 += cp->rcu_try_flip_m2;
-       }
-}
-
-static ssize_t rcustats_read(struct file *filp, char __user *buffer,
-                               size_t count, loff_t *ppos)
-{
-       struct rcupreempt_trace trace;
-       ssize_t bcount;
-       int cnt = 0;
-
-       rcupreempt_trace_sum(&trace);
-       mutex_lock(&rcupreempt_trace_mutex);
-       snprintf(&rcupreempt_trace_buf[cnt], RCUPREEMPT_TRACE_BUF_SIZE - cnt,
-                "ggp=%ld rcc=%ld\n",
-                rcu_batches_completed(),
-                trace.rcu_check_callbacks);
-       snprintf(&rcupreempt_trace_buf[cnt], RCUPREEMPT_TRACE_BUF_SIZE - cnt,
-                "na=%ld nl=%ld wa=%ld wl=%ld da=%ld dl=%ld dr=%ld di=%d\n"
-                "1=%d e1=%d i1=%ld ie1=%ld g1=%ld a1=%ld ae1=%ld a2=%ld\n"
-                "z1=%ld ze1=%ld z2=%ld m1=%ld me1=%ld m2=%ld\n",
-
-                trace.next_add, trace.next_length,
-                trace.wait_add, trace.wait_length,
-                trace.done_add, trace.done_length,
-                trace.done_remove, atomic_read(&trace.done_invoked),
-                atomic_read(&trace.rcu_try_flip_1),
-                atomic_read(&trace.rcu_try_flip_e1),
-                trace.rcu_try_flip_i1, trace.rcu_try_flip_ie1,
-                trace.rcu_try_flip_g1,
-                trace.rcu_try_flip_a1, trace.rcu_try_flip_ae1,
-                        trace.rcu_try_flip_a2,
-                trace.rcu_try_flip_z1, trace.rcu_try_flip_ze1,
-                        trace.rcu_try_flip_z2,
-                trace.rcu_try_flip_m1, trace.rcu_try_flip_me1,
-                       trace.rcu_try_flip_m2);
-       bcount = simple_read_from_buffer(buffer, count, ppos,
-                       rcupreempt_trace_buf, strlen(rcupreempt_trace_buf));
-       mutex_unlock(&rcupreempt_trace_mutex);
-       return bcount;
-}
-
-static ssize_t rcugp_read(struct file *filp, char __user *buffer,
-                               size_t count, loff_t *ppos)
-{
-       long oldgp = rcu_batches_completed();
-       ssize_t bcount;
-
-       mutex_lock(&rcupreempt_trace_mutex);
-       synchronize_rcu();
-       snprintf(rcupreempt_trace_buf, RCUPREEMPT_TRACE_BUF_SIZE,
-               "oldggp=%ld  newggp=%ld\n", oldgp, rcu_batches_completed());
-       bcount = simple_read_from_buffer(buffer, count, ppos,
-                       rcupreempt_trace_buf, strlen(rcupreempt_trace_buf));
-       mutex_unlock(&rcupreempt_trace_mutex);
-       return bcount;
-}
-
-static ssize_t rcuctrs_read(struct file *filp, char __user *buffer,
-                               size_t count, loff_t *ppos)
-{
-       int cnt = 0;
-       int cpu;
-       int f = rcu_batches_completed() & 0x1;
-       ssize_t bcount;
-
-       mutex_lock(&rcupreempt_trace_mutex);
-
-       cnt += snprintf(&rcupreempt_trace_buf[cnt], RCUPREEMPT_TRACE_BUF_SIZE,
-                               "CPU last cur F M\n");
-       for_each_possible_cpu(cpu) {
-               long *flipctr = rcupreempt_flipctr(cpu);
-               cnt += snprintf(&rcupreempt_trace_buf[cnt],
-                               RCUPREEMPT_TRACE_BUF_SIZE - cnt,
-                                       "%3d%c %4ld %3ld %d %d\n",
-                              cpu,
-                              cpu_is_offline(cpu) ? '!' : ' ',
-                              flipctr[!f],
-                              flipctr[f],
-                              rcupreempt_flip_flag(cpu),
-                              rcupreempt_mb_flag(cpu));
-       }
-       cnt += snprintf(&rcupreempt_trace_buf[cnt],
-                       RCUPREEMPT_TRACE_BUF_SIZE - cnt,
-                       "ggp = %ld, state = %s\n",
-                       rcu_batches_completed(),
-                       rcupreempt_try_flip_state_name());
-       cnt += snprintf(&rcupreempt_trace_buf[cnt],
-                       RCUPREEMPT_TRACE_BUF_SIZE - cnt,
-                       "\n");
-       bcount = simple_read_from_buffer(buffer, count, ppos,
-                       rcupreempt_trace_buf, strlen(rcupreempt_trace_buf));
-       mutex_unlock(&rcupreempt_trace_mutex);
-       return bcount;
-}
-
-static struct file_operations rcustats_fops = {
-       .owner = THIS_MODULE,
-       .read = rcustats_read,
-};
-
-static struct file_operations rcugp_fops = {
-       .owner = THIS_MODULE,
-       .read = rcugp_read,
-};
-
-static struct file_operations rcuctrs_fops = {
-       .owner = THIS_MODULE,
-       .read = rcuctrs_read,
-};
-
-static struct dentry *rcudir, *statdir, *ctrsdir, *gpdir;
-static int rcupreempt_debugfs_init(void)
-{
-       rcudir = debugfs_create_dir("rcu", NULL);
-       if (!rcudir)
-               goto out;
-       statdir = debugfs_create_file("rcustats", 0444, rcudir,
-                                               NULL, &rcustats_fops);
-       if (!statdir)
-               goto free_out;
-
-       gpdir = debugfs_create_file("rcugp", 0444, rcudir, NULL, &rcugp_fops);
-       if (!gpdir)
-               goto free_out;
-
-       ctrsdir = debugfs_create_file("rcuctrs", 0444, rcudir,
-                                               NULL, &rcuctrs_fops);
-       if (!ctrsdir)
-               goto free_out;
-       return 0;
-free_out:
-       if (statdir)
-               debugfs_remove(statdir);
-       if (gpdir)
-               debugfs_remove(gpdir);
-       debugfs_remove(rcudir);
-out:
-       return 1;
-}
-
-static int __init rcupreempt_trace_init(void)
-{
-       int ret;
-
-       mutex_init(&rcupreempt_trace_mutex);
-       rcupreempt_trace_buf = kmalloc(RCUPREEMPT_TRACE_BUF_SIZE, GFP_KERNEL);
-       if (!rcupreempt_trace_buf)
-               return 1;
-       ret = rcupreempt_debugfs_init();
-       if (ret)
-               kfree(rcupreempt_trace_buf);
-       return ret;
-}
-
-static void __exit rcupreempt_trace_cleanup(void)
-{
-       debugfs_remove(statdir);
-       debugfs_remove(gpdir);
-       debugfs_remove(ctrsdir);
-       debugfs_remove(rcudir);
-       kfree(rcupreempt_trace_buf);
-}
-
-
-module_init(rcupreempt_trace_init);
-module_exit(rcupreempt_trace_cleanup);
index f87fb0c..82fbc49 100644 (file)
@@ -725,7 +725,7 @@ config RCU_TORTURE_TEST_RUNNABLE
 
 config RCU_CPU_STALL_DETECTOR
        bool "Check for stalled CPUs delaying RCU grace periods"
-       depends on CLASSIC_RCU || TREE_RCU || TREE_PREEMPT_RCU
+       depends on TREE_RCU || TREE_PREEMPT_RCU
        default n
        help
          This option causes RCU to printk information on which