Documentation/kref.txt

   1 ===================================================
   2 Adding reference counters (krefs) to kernel objects
   3 ===================================================
   4
   5 :Author: Corey Minyard <minyard@acm.org>
   6 :Author: Thomas Hellstrom <thellstrom@vmware.com>
   7
   8 A lot of this was lifted from Greg Kroah-Hartman's 2004 OLS paper and
   9 presentation on krefs, which can be found at:
  10
  11   - http://www.kroah.com/linux/talks/ols_2004_kref_paper/Reprint-Kroah-Hartman-OLS2004.pdf
  12   - http://www.kroah.com/linux/talks/ols_2004_kref_talk/
  13
  14 Introduction
  15 ============
  16
  17 krefs allow you to add reference counters to your objects.  If you
  18 have objects that are used in multiple places and passed around, and
  19 you don't have refcounts, your code is almost certainly broken.  If
  20 you want refcounts, krefs are the way to go.
  21
  22 To use a kref, add one to your data structures like::
  23
  24     struct my_data
  25     {
  26         .
  27         .
  28         struct kref refcount;
  29         .
  30         .
  31     };
  32
  33 The kref can occur anywhere within the data structure.
  34
  35 Initialization
  36 ==============
  37
  38 You must initialize the kref after you allocate it.  To do this, call
  39 kref_init as so::
  40
  41      struct my_data *data;
  42
  43      data = kmalloc(sizeof(*data), GFP_KERNEL);
  44      if (!data)
  45             return -ENOMEM;
  46      kref_init(&data->refcount);
  47
  48 This sets the refcount in the kref to 1.
  49
  50 Kref rules
  51 ==========
  52
  53 Once you have an initialized kref, you must follow the following
  54 rules:
  55
  56 1) If you make a non-temporary copy of a pointer, especially if
  57    it can be passed to another thread of execution, you must
  58    increment the refcount with kref_get() before passing it off::
  59
  60        kref_get(&data->refcount);
  61
  62    If you already have a valid pointer to a kref-ed structure (the
  63    refcount cannot go to zero) you may do this without a lock.
  64
  65 2) When you are done with a pointer, you must call kref_put()::
  66
  67        kref_put(&data->refcount, data_release);
  68
  69    If this is the last reference to the pointer, the release
  70    routine will be called.  If the code never tries to get
  71    a valid pointer to a kref-ed structure without already
  72    holding a valid pointer, it is safe to do this without
  73    a lock.
  74
  75 3) If the code attempts to gain a reference to a kref-ed structure
  76    without already holding a valid pointer, it must serialize access
  77    where a kref_put() cannot occur during the kref_get(), and the
  78    structure must remain valid during the kref_get().
  79
  80 For example, if you allocate some data and then pass it to another
  81 thread to process::
  82
  83     void data_release(struct kref *ref)
  84     {
  85         struct my_data *data = container_of(ref, struct my_data, refcount);
  86         kfree(data);
  87     }
  88
  89     void more_data_handling(void *cb_data)
  90     {
  91         struct my_data *data = cb_data;
  92         .
  93         . do stuff with data here
  94         .
  95         kref_put(&data->refcount, data_release);
  96     }
  97
  98     int my_data_handler(void)
  99     {
 100         int rv = 0;
 101         struct my_data *data;
 102         struct task_struct *task;
 103         data = kmalloc(sizeof(*data), GFP_KERNEL);
 104         if (!data)
 105                 return -ENOMEM;
 106         kref_init(&data->refcount);
 107
 108         kref_get(&data->refcount);
 109         task = kthread_run(more_data_handling, data, "more_data_handling");
 110         if (task == ERR_PTR(-ENOMEM)) {
 111                 rv = -ENOMEM;
 112                 kref_put(&data->refcount, data_release);
 113                 goto out;
 114         }
 115
 116         .
 117         . do stuff with data here
 118         .
 119     out:
 120         kref_put(&data->refcount, data_release);
 121         return rv;
 122     }
 123
 124 This way, it doesn't matter what order the two threads handle the
 125 data, the kref_put() handles knowing when the data is not referenced
 126 any more and releasing it.  The kref_get() does not require a lock,
 127 since we already have a valid pointer that we own a refcount for.  The
 128 put needs no lock because nothing tries to get the data without
 129 already holding a pointer.
 130
 131 Note that the "before" in rule 1 is very important.  You should never
 132 do something like::
 133
 134         task = kthread_run(more_data_handling, data, "more_data_handling");
 135         if (task == ERR_PTR(-ENOMEM)) {
 136                 rv = -ENOMEM;
 137                 goto out;
 138         } else
 139                 /* BAD BAD BAD - get is after the handoff */
 140                 kref_get(&data->refcount);
 141
 142 Don't assume you know what you are doing and use the above construct.
 143 First of all, you may not know what you are doing.  Second, you may
 144 know what you are doing (there are some situations where locking is
 145 involved where the above may be legal) but someone else who doesn't
 146 know what they are doing may change the code or copy the code.  It's
 147 bad style.  Don't do it.
 148
 149 There are some situations where you can optimize the gets and puts.
 150 For instance, if you are done with an object and enqueuing it for
 151 something else or passing it off to something else, there is no reason
 152 to do a get then a put::
 153
 154         /* Silly extra get and put */
 155         kref_get(&obj->ref);
 156         enqueue(obj);
 157         kref_put(&obj->ref, obj_cleanup);
 158
 159 Just do the enqueue.  A comment about this is always welcome::
 160
 161         enqueue(obj);
 162         /* We are done with obj, so we pass our refcount off
 163            to the queue.  DON'T TOUCH obj AFTER HERE! */
 164
 165 The last rule (rule 3) is the nastiest one to handle.  Say, for
 166 instance, you have a list of items that are each kref-ed, and you wish
 167 to get the first one.  You can't just pull the first item off the list
 168 and kref_get() it.  That violates rule 3 because you are not already
 169 holding a valid pointer.  You must add a mutex (or some other lock).
 170 For instance::
 171
 172         static DEFINE_MUTEX(mutex);
 173         static LIST_HEAD(q);
 174         struct my_data
 175         {
 176                 struct kref      refcount;
 177                 struct list_head link;
 178         };
 179
 180         static struct my_data *get_entry()
 181         {
 182                 struct my_data *entry = NULL;
 183                 mutex_lock(&mutex);
 184                 if (!list_empty(&q)) {
 185                         entry = container_of(q.next, struct my_data, link);
 186                         kref_get(&entry->refcount);
 187                 }
 188                 mutex_unlock(&mutex);
 189                 return entry;
 190         }
 191
 192         static void release_entry(struct kref *ref)
 193         {
 194                 struct my_data *entry = container_of(ref, struct my_data, refcount);
 195
 196                 list_del(&entry->link);
 197                 kfree(entry);
 198         }
 199
 200         static void put_entry(struct my_data *entry)
 201         {
 202                 mutex_lock(&mutex);
 203                 kref_put(&entry->refcount, release_entry);
 204                 mutex_unlock(&mutex);
 205         }
 206
 207 The kref_put() return value is useful if you do not want to hold the
 208 lock during the whole release operation.  Say you didn't want to call
 209 kfree() with the lock held in the example above (since it is kind of
 210 pointless to do so).  You could use kref_put() as follows::
 211
 212         static void release_entry(struct kref *ref)
 213         {
 214                 /* All work is done after the return from kref_put(). */
 215         }
 216
 217         static void put_entry(struct my_data *entry)
 218         {
 219                 mutex_lock(&mutex);
 220                 if (kref_put(&entry->refcount, release_entry)) {
 221                         list_del(&entry->link);
 222                         mutex_unlock(&mutex);
 223                         kfree(entry);
 224                 } else
 225                         mutex_unlock(&mutex);
 226         }
 227
 228 This is really more useful if you have to call other routines as part
 229 of the free operations that could take a long time or might claim the
 230 same lock.  Note that doing everything in the release routine is still
 231 preferred as it is a little neater.
 232
 233 The above example could also be optimized using kref_get_unless_zero() in
 234 the following way::
 235
 236         static struct my_data *get_entry()
 237         {
 238                 struct my_data *entry = NULL;
 239                 mutex_lock(&mutex);
 240                 if (!list_empty(&q)) {
 241                         entry = container_of(q.next, struct my_data, link);
 242                         if (!kref_get_unless_zero(&entry->refcount))
 243                                 entry = NULL;
 244                 }
 245                 mutex_unlock(&mutex);
 246                 return entry;
 247         }
 248
 249         static void release_entry(struct kref *ref)
 250         {
 251                 struct my_data *entry = container_of(ref, struct my_data, refcount);
 252
 253                 mutex_lock(&mutex);
 254                 list_del(&entry->link);
 255                 mutex_unlock(&mutex);
 256                 kfree(entry);
 257         }
 258
 259         static void put_entry(struct my_data *entry)
 260         {
 261                 kref_put(&entry->refcount, release_entry);
 262         }
 263
 264 Which is useful to remove the mutex lock around kref_put() in put_entry(), but
 265 it's important that kref_get_unless_zero is enclosed in the same critical
 266 section that finds the entry in the lookup table,
 267 otherwise kref_get_unless_zero may reference already freed memory.
 268 Note that it is illegal to use kref_get_unless_zero without checking its
 269 return value. If you are sure (by already having a valid pointer) that
 270 kref_get_unless_zero() will return true, then use kref_get() instead.
 271
 272 Krefs and RCU
 273 =============
 274
 275 The function kref_get_unless_zero also makes it possible to use rcu
 276 locking for lookups in the above example::
 277
 278         struct my_data
 279         {
 280                 struct rcu_head rhead;
 281                 .
 282                 struct kref refcount;
 283                 .
 284                 .
 285         };
 286
 287         static struct my_data *get_entry_rcu()
 288         {
 289                 struct my_data *entry = NULL;
 290                 rcu_read_lock();
 291                 if (!list_empty(&q)) {
 292                         entry = container_of(q.next, struct my_data, link);
 293                         if (!kref_get_unless_zero(&entry->refcount))
 294                                 entry = NULL;
 295                 }
 296                 rcu_read_unlock();
 297                 return entry;
 298         }
 299
 300         static void release_entry_rcu(struct kref *ref)
 301         {
 302                 struct my_data *entry = container_of(ref, struct my_data, refcount);
 303
 304                 mutex_lock(&mutex);
 305                 list_del_rcu(&entry->link);
 306                 mutex_unlock(&mutex);
 307                 kfree_rcu(entry, rhead);
 308         }
 309
 310         static void put_entry(struct my_data *entry)
 311         {
 312                 kref_put(&entry->refcount, release_entry_rcu);
 313         }
 314
 315 But note that the struct kref member needs to remain in valid memory for a
 316 rcu grace period after release_entry_rcu was called. That can be accomplished
 317 by using kfree_rcu(entry, rhead) as done above, or by calling synchronize_rcu()
 318 before using kfree, but note that synchronize_rcu() may sleep for a
 319 substantial amount of time.