kernel/srcu.c

   1 /*
   2  * Sleepable Read-Copy Update mechanism for mutual exclusion.
   3  *
   4  * This program is free software; you can redistribute it and/or modify
   5  * it under the terms of the GNU General Public License as published by
   6  * the Free Software Foundation; either version 2 of the License, or
   7  * (at your option) any later version.
   8  *
   9  * This program is distributed in the hope that it will be useful,
  10  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  11  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  12  * GNU General Public License for more details.
  13  *
  14  * You should have received a copy of the GNU General Public License
  15  * along with this program; if not, write to the Free Software
  16  * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
  17  *
  18  * Copyright (C) IBM Corporation, 2006
  19  *
  20  * Author: Paul McKenney <paulmck@us.ibm.com>
  21  *
  22  * For detailed explanation of Read-Copy Update mechanism see -
  23  *              Documentation/RCU/ *.txt
  24  *
  25  */
  26
  27 #include <linux/module.h>
  28 #include <linux/mutex.h>
  29 #include <linux/percpu.h>
  30 #include <linux/preempt.h>
  31 #include <linux/rcupdate.h>
  32 #include <linux/sched.h>
  33 #include <linux/smp.h>
  34 #include <linux/srcu.h>
  35
  36 static int init_srcu_struct_fields(struct srcu_struct *sp)
  37 {
  38         sp->completed = 0;
  39         mutex_init(&sp->mutex);
  40         sp->per_cpu_ref = alloc_percpu(struct srcu_struct_array);
  41         return sp->per_cpu_ref ? 0 : -ENOMEM;
  42 }
  43
  44 #ifdef CONFIG_DEBUG_LOCK_ALLOC
  45
  46 int __init_srcu_struct(struct srcu_struct *sp, const char *name,
  47                        struct lock_class_key *key)
  48 {
  49 #ifdef CONFIG_DEBUG_LOCK_ALLOC
  50         /* Don't re-initialize a lock while it is held. */
  51         debug_check_no_locks_freed((void *)sp, sizeof(*sp));
  52         lockdep_init_map(&sp->dep_map, name, key, 0);
  53 #endif /* #ifdef CONFIG_DEBUG_LOCK_ALLOC */
  54         return init_srcu_struct_fields(sp);
  55 }
  56 EXPORT_SYMBOL_GPL(__init_srcu_struct);
  57
  58 #else /* #ifdef CONFIG_DEBUG_LOCK_ALLOC */
  59
  60 /**
  61  * init_srcu_struct - initialize a sleep-RCU structure
  62  * @sp: structure to initialize.
  63  *
  64  * Must invoke this on a given srcu_struct before passing that srcu_struct
  65  * to any other function.  Each srcu_struct represents a separate domain
  66  * of SRCU protection.
  67  */
  68 int init_srcu_struct(struct srcu_struct *sp)
  69 {
  70         return init_srcu_struct_fields(sp);
  71 }
  72 EXPORT_SYMBOL_GPL(init_srcu_struct);
  73
  74 #endif /* #else #ifdef CONFIG_DEBUG_LOCK_ALLOC */
  75
  76 /*
  77  * srcu_readers_active_idx -- returns approximate number of readers
  78  *      active on the specified rank of per-CPU counters.
  79  */
  80
  81 static int srcu_readers_active_idx(struct srcu_struct *sp, int idx)
  82 {
  83         int cpu;
  84         int sum;
  85
  86         sum = 0;
  87         for_each_possible_cpu(cpu)
  88                 sum += per_cpu_ptr(sp->per_cpu_ref, cpu)->c[idx];
  89         return sum;
  90 }
  91
  92 /**
  93  * srcu_readers_active - returns approximate number of readers.
  94  * @sp: which srcu_struct to count active readers (holding srcu_read_lock).
  95  *
  96  * Note that this is not an atomic primitive, and can therefore suffer
  97  * severe errors when invoked on an active srcu_struct.  That said, it
  98  * can be useful as an error check at cleanup time.
  99  */
 100 static int srcu_readers_active(struct srcu_struct *sp)
 101 {
 102         return srcu_readers_active_idx(sp, 0) + srcu_readers_active_idx(sp, 1);
 103 }
 104
 105 /**
 106  * cleanup_srcu_struct - deconstruct a sleep-RCU structure
 107  * @sp: structure to clean up.
 108  *
 109  * Must invoke this after you are finished using a given srcu_struct that
 110  * was initialized via init_srcu_struct(), else you leak memory.
 111  */
 112 void cleanup_srcu_struct(struct srcu_struct *sp)
 113 {
 114         int sum;
 115
 116         sum = srcu_readers_active(sp);
 117         WARN_ON(sum);  /* Leakage unless caller handles error. */
 118         if (sum != 0)
 119                 return;
 120         free_percpu(sp->per_cpu_ref);
 121         sp->per_cpu_ref = NULL;
 122 }
 123 EXPORT_SYMBOL_GPL(cleanup_srcu_struct);
 124
 125 /*
 126  * Counts the new reader in the appropriate per-CPU element of the
 127  * srcu_struct.  Must be called from process context.
 128  * Returns an index that must be passed to the matching srcu_read_unlock().
 129  */
 130 int __srcu_read_lock(struct srcu_struct *sp)
 131 {
 132         int idx;
 133
 134         preempt_disable();
 135         idx = sp->completed & 0x1;
 136         barrier();  /* ensure compiler looks -once- at sp->completed. */
 137         per_cpu_ptr(sp->per_cpu_ref, smp_processor_id())->c[idx]++;
 138         srcu_barrier();  /* ensure compiler won't misorder critical section. */
 139         preempt_enable();
 140         return idx;
 141 }
 142 EXPORT_SYMBOL_GPL(__srcu_read_lock);
 143
 144 /*
 145  * Removes the count for the old reader from the appropriate per-CPU
 146  * element of the srcu_struct.  Note that this may well be a different
 147  * CPU than that which was incremented by the corresponding srcu_read_lock().
 148  * Must be called from process context.
 149  */
 150 void __srcu_read_unlock(struct srcu_struct *sp, int idx)
 151 {
 152         preempt_disable();
 153         srcu_barrier();  /* ensure compiler won't misorder critical section. */
 154         per_cpu_ptr(sp->per_cpu_ref, smp_processor_id())->c[idx]--;
 155         preempt_enable();
 156 }
 157 EXPORT_SYMBOL_GPL(__srcu_read_unlock);
 158
 159 /*
 160  * Helper function for synchronize_srcu() and synchronize_srcu_expedited().
 161  */
 162 static void __synchronize_srcu(struct srcu_struct *sp, void (*sync_func)(void))
 163 {
 164         int idx;
 165
 166         idx = sp->completed;
 167         mutex_lock(&sp->mutex);
 168
 169         /*
 170          * Check to see if someone else did the work for us while we were
 171          * waiting to acquire the lock.  We need -two- advances of
 172          * the counter, not just one.  If there was but one, we might have
 173          * shown up -after- our helper's first synchronize_sched(), thus
 174          * having failed to prevent CPU-reordering races with concurrent
 175          * srcu_read_unlock()s on other CPUs (see comment below).  So we
 176          * either (1) wait for two or (2) supply the second ourselves.
 177          */
 178
 179         if ((sp->completed - idx) >= 2) {
 180                 mutex_unlock(&sp->mutex);
 181                 return;
 182         }
 183
 184         sync_func();  /* Force memory barrier on all CPUs. */
 185
 186         /*
 187          * The preceding synchronize_sched() ensures that any CPU that
 188          * sees the new value of sp->completed will also see any preceding
 189          * changes to data structures made by this CPU.  This prevents
 190          * some other CPU from reordering the accesses in its SRCU
 191          * read-side critical section to precede the corresponding
 192          * srcu_read_lock() -- ensuring that such references will in
 193          * fact be protected.
 194          *
 195          * So it is now safe to do the flip.
 196          */
 197
 198         idx = sp->completed & 0x1;
 199         sp->completed++;
 200
 201         sync_func();  /* Force memory barrier on all CPUs. */
 202
 203         /*
 204          * At this point, because of the preceding synchronize_sched(),
 205          * all srcu_read_lock() calls using the old counters have completed.
 206          * Their corresponding critical sections might well be still
 207          * executing, but the srcu_read_lock() primitives themselves
 208          * will have finished executing.
 209          */
 210
 211         while (srcu_readers_active_idx(sp, idx))
 212                 schedule_timeout_interruptible(1);
 213
 214         sync_func();  /* Force memory barrier on all CPUs. */
 215
 216         /*
 217          * The preceding synchronize_sched() forces all srcu_read_unlock()
 218          * primitives that were executing concurrently with the preceding
 219          * for_each_possible_cpu() loop to have completed by this point.
 220          * More importantly, it also forces the corresponding SRCU read-side
 221          * critical sections to have also completed, and the corresponding
 222          * references to SRCU-protected data items to be dropped.
 223          *
 224          * Note:
 225          *
 226          *      Despite what you might think at first glance, the
 227          *      preceding synchronize_sched() -must- be within the
 228          *      critical section ended by the following mutex_unlock().
 229          *      Otherwise, a task taking the early exit can race
 230          *      with a srcu_read_unlock(), which might have executed
 231          *      just before the preceding srcu_readers_active() check,
 232          *      and whose CPU might have reordered the srcu_read_unlock()
 233          *      with the preceding critical section.  In this case, there
 234          *      is nothing preventing the synchronize_sched() task that is
 235          *      taking the early exit from freeing a data structure that
 236          *      is still being referenced (out of order) by the task
 237          *      doing the srcu_read_unlock().
 238          *
 239          *      Alternatively, the comparison with "2" on the early exit
 240          *      could be changed to "3", but this increases synchronize_srcu()
 241          *      latency for bulk loads.  So the current code is preferred.
 242          */
 243
 244         mutex_unlock(&sp->mutex);
 245 }
 246
 247 /**
 248  * synchronize_srcu - wait for prior SRCU read-side critical-section completion
 249  * @sp: srcu_struct with which to synchronize.
 250  *
 251  * Flip the completed counter, and wait for the old count to drain to zero.
 252  * As with classic RCU, the updater must use some separate means of
 253  * synchronizing concurrent updates.  Can block; must be called from
 254  * process context.
 255  *
 256  * Note that it is illegal to call synchronize_srcu() from the corresponding
 257  * SRCU read-side critical section; doing so will result in deadlock.
 258  * However, it is perfectly legal to call synchronize_srcu() on one
 259  * srcu_struct from some other srcu_struct's read-side critical section.
 260  */
 261 void synchronize_srcu(struct srcu_struct *sp)
 262 {
 263         __synchronize_srcu(sp, synchronize_sched);
 264 }
 265 EXPORT_SYMBOL_GPL(synchronize_srcu);
 266
 267 /**
 268  * synchronize_srcu_expedited - like synchronize_srcu, but less patient
 269  * @sp: srcu_struct with which to synchronize.
 270  *
 271  * Flip the completed counter, and wait for the old count to drain to zero.
 272  * As with classic RCU, the updater must use some separate means of
 273  * synchronizing concurrent updates.  Can block; must be called from
 274  * process context.
 275  *
 276  * Note that it is illegal to call synchronize_srcu_expedited()
 277  * from the corresponding SRCU read-side critical section; doing so
 278  * will result in deadlock.  However, it is perfectly legal to call
 279  * synchronize_srcu_expedited() on one srcu_struct from some other
 280  * srcu_struct's read-side critical section.
 281  */
 282 void synchronize_srcu_expedited(struct srcu_struct *sp)
 283 {
 284         __synchronize_srcu(sp, synchronize_sched_expedited);
 285 }
 286 EXPORT_SYMBOL_GPL(synchronize_srcu_expedited);
 287
 288 /**
 289  * srcu_batches_completed - return batches completed.
 290  * @sp: srcu_struct on which to report batch completion.
 291  *
 292  * Report the number of batches, correlated with, but not necessarily
 293  * precisely the same as, the number of grace periods that have elapsed.
 294  */
 295
 296 long srcu_batches_completed(struct srcu_struct *sp)
 297 {
 298         return sp->completed;
 299 }
 300 EXPORT_SYMBOL_GPL(srcu_batches_completed);