kernel/bpf/hashtab.c

   1 /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com
   2  *
   3  * This program is free software; you can redistribute it and/or
   4  * modify it under the terms of version 2 of the GNU General Public
   5  * License as published by the Free Software Foundation.
   6  *
   7  * This program is distributed in the hope that it will be useful, but
   8  * WITHOUT ANY WARRANTY; without even the implied warranty of
   9  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
  10  * General Public License for more details.
  11  */
  12 #include <linux/bpf.h>
  13 #include <linux/jhash.h>
  14 #include <linux/filter.h>
  15 #include <linux/vmalloc.h>
  16
  17 struct bpf_htab {
  18         struct bpf_map map;
  19         struct hlist_head *buckets;
  20         raw_spinlock_t lock;
  21         atomic_t count; /* number of elements in this hashtable */
  22         u32 n_buckets;  /* number of hash buckets */
  23         u32 elem_size;  /* size of each element in bytes */
  24 };
  25
  26 /* each htab element is struct htab_elem + key + value */
  27 struct htab_elem {
  28         struct hlist_node hash_node;
  29         struct rcu_head rcu;
  30         u32 hash;
  31         char key[0] __aligned(8);
  32 };
  33
  34 /* Called from syscall */
  35 static struct bpf_map *htab_map_alloc(union bpf_attr *attr)
  36 {
  37         struct bpf_htab *htab;
  38         int err, i;
  39
  40         htab = kzalloc(sizeof(*htab), GFP_USER);
  41         if (!htab)
  42                 return ERR_PTR(-ENOMEM);
  43
  44         /* mandatory map attributes */
  45         htab->map.key_size = attr->key_size;
  46         htab->map.value_size = attr->value_size;
  47         htab->map.max_entries = attr->max_entries;
  48
  49         /* check sanity of attributes.
  50          * value_size == 0 may be allowed in the future to use map as a set
  51          */
  52         err = -EINVAL;
  53         if (htab->map.max_entries == 0 || htab->map.key_size == 0 ||
  54             htab->map.value_size == 0)
  55                 goto free_htab;
  56
  57         /* hash table size must be power of 2 */
  58         htab->n_buckets = roundup_pow_of_two(htab->map.max_entries);
  59
  60         err = -E2BIG;
  61         if (htab->map.key_size > MAX_BPF_STACK)
  62                 /* eBPF programs initialize keys on stack, so they cannot be
  63                  * larger than max stack size
  64                  */
  65                 goto free_htab;
  66
  67         if (htab->map.value_size >= (1 << (KMALLOC_SHIFT_MAX - 1)) -
  68             MAX_BPF_STACK - sizeof(struct htab_elem))
  69                 /* if value_size is bigger, the user space won't be able to
  70                  * access the elements via bpf syscall. This check also makes
  71                  * sure that the elem_size doesn't overflow and it's
  72                  * kmalloc-able later in htab_map_update_elem()
  73                  */
  74                 goto free_htab;
  75
  76         htab->elem_size = sizeof(struct htab_elem) +
  77                           round_up(htab->map.key_size, 8) +
  78                           htab->map.value_size;
  79
  80         /* prevent zero size kmalloc and check for u32 overflow */
  81         if (htab->n_buckets == 0 ||
  82             htab->n_buckets > U32_MAX / sizeof(struct hlist_head))
  83                 goto free_htab;
  84
  85         if ((u64) htab->n_buckets * sizeof(struct hlist_head) +
  86             (u64) htab->elem_size * htab->map.max_entries >=
  87             U32_MAX - PAGE_SIZE)
  88                 /* make sure page count doesn't overflow */
  89                 goto free_htab;
  90
  91         htab->map.pages = round_up(htab->n_buckets * sizeof(struct hlist_head) +
  92                                    htab->elem_size * htab->map.max_entries,
  93                                    PAGE_SIZE) >> PAGE_SHIFT;
  94
  95         err = -ENOMEM;
  96         htab->buckets = kmalloc_array(htab->n_buckets, sizeof(struct hlist_head),
  97                                       GFP_USER | __GFP_NOWARN);
  98
  99         if (!htab->buckets) {
 100                 htab->buckets = vmalloc(htab->n_buckets * sizeof(struct hlist_head));
 101                 if (!htab->buckets)
 102                         goto free_htab;
 103         }
 104
 105         for (i = 0; i < htab->n_buckets; i++)
 106                 INIT_HLIST_HEAD(&htab->buckets[i]);
 107
 108         raw_spin_lock_init(&htab->lock);
 109         atomic_set(&htab->count, 0);
 110
 111         return &htab->map;
 112
 113 free_htab:
 114         kfree(htab);
 115         return ERR_PTR(err);
 116 }
 117
 118 static inline u32 htab_map_hash(const void *key, u32 key_len)
 119 {
 120         return jhash(key, key_len, 0);
 121 }
 122
 123 static inline struct hlist_head *select_bucket(struct bpf_htab *htab, u32 hash)
 124 {
 125         return &htab->buckets[hash & (htab->n_buckets - 1)];
 126 }
 127
 128 static struct htab_elem *lookup_elem_raw(struct hlist_head *head, u32 hash,
 129                                          void *key, u32 key_size)
 130 {
 131         struct htab_elem *l;
 132
 133         hlist_for_each_entry_rcu(l, head, hash_node)
 134                 if (l->hash == hash && !memcmp(&l->key, key, key_size))
 135                         return l;
 136
 137         return NULL;
 138 }
 139
 140 /* Called from syscall or from eBPF program */
 141 static void *htab_map_lookup_elem(struct bpf_map *map, void *key)
 142 {
 143         struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
 144         struct hlist_head *head;
 145         struct htab_elem *l;
 146         u32 hash, key_size;
 147
 148         /* Must be called with rcu_read_lock. */
 149         WARN_ON_ONCE(!rcu_read_lock_held());
 150
 151         key_size = map->key_size;
 152
 153         hash = htab_map_hash(key, key_size);
 154
 155         head = select_bucket(htab, hash);
 156
 157         l = lookup_elem_raw(head, hash, key, key_size);
 158
 159         if (l)
 160                 return l->key + round_up(map->key_size, 8);
 161
 162         return NULL;
 163 }
 164
 165 /* Called from syscall */
 166 static int htab_map_get_next_key(struct bpf_map *map, void *key, void *next_key)
 167 {
 168         struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
 169         struct hlist_head *head;
 170         struct htab_elem *l, *next_l;
 171         u32 hash, key_size;
 172         int i;
 173
 174         WARN_ON_ONCE(!rcu_read_lock_held());
 175
 176         key_size = map->key_size;
 177
 178         hash = htab_map_hash(key, key_size);
 179
 180         head = select_bucket(htab, hash);
 181
 182         /* lookup the key */
 183         l = lookup_elem_raw(head, hash, key, key_size);
 184
 185         if (!l) {
 186                 i = 0;
 187                 goto find_first_elem;
 188         }
 189
 190         /* key was found, get next key in the same bucket */
 191         next_l = hlist_entry_safe(rcu_dereference_raw(hlist_next_rcu(&l->hash_node)),
 192                                   struct htab_elem, hash_node);
 193
 194         if (next_l) {
 195                 /* if next elem in this hash list is non-zero, just return it */
 196                 memcpy(next_key, next_l->key, key_size);
 197                 return 0;
 198         }
 199
 200         /* no more elements in this hash list, go to the next bucket */
 201         i = hash & (htab->n_buckets - 1);
 202         i++;
 203
 204 find_first_elem:
 205         /* iterate over buckets */
 206         for (; i < htab->n_buckets; i++) {
 207                 head = select_bucket(htab, i);
 208
 209                 /* pick first element in the bucket */
 210                 next_l = hlist_entry_safe(rcu_dereference_raw(hlist_first_rcu(head)),
 211                                           struct htab_elem, hash_node);
 212                 if (next_l) {
 213                         /* if it's not empty, just return it */
 214                         memcpy(next_key, next_l->key, key_size);
 215                         return 0;
 216                 }
 217         }
 218
 219         /* itereated over all buckets and all elements */
 220         return -ENOENT;
 221 }
 222
 223 /* Called from syscall or from eBPF program */
 224 static int htab_map_update_elem(struct bpf_map *map, void *key, void *value,
 225                                 u64 map_flags)
 226 {
 227         struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
 228         struct htab_elem *l_new, *l_old;
 229         struct hlist_head *head;
 230         unsigned long flags;
 231         u32 key_size;
 232         int ret;
 233
 234         if (map_flags > BPF_EXIST)
 235                 /* unknown flags */
 236                 return -EINVAL;
 237
 238         WARN_ON_ONCE(!rcu_read_lock_held());
 239
 240         /* allocate new element outside of lock */
 241         l_new = kmalloc(htab->elem_size, GFP_ATOMIC | __GFP_NOWARN);
 242         if (!l_new)
 243                 return -ENOMEM;
 244
 245         key_size = map->key_size;
 246
 247         memcpy(l_new->key, key, key_size);
 248         memcpy(l_new->key + round_up(key_size, 8), value, map->value_size);
 249
 250         l_new->hash = htab_map_hash(l_new->key, key_size);
 251
 252         /* bpf_map_update_elem() can be called in_irq() */
 253         raw_spin_lock_irqsave(&htab->lock, flags);
 254
 255         head = select_bucket(htab, l_new->hash);
 256
 257         l_old = lookup_elem_raw(head, l_new->hash, key, key_size);
 258
 259         if (!l_old && unlikely(atomic_read(&htab->count) >= map->max_entries)) {
 260                 /* if elem with this 'key' doesn't exist and we've reached
 261                  * max_entries limit, fail insertion of new elem
 262                  */
 263                 ret = -E2BIG;
 264                 goto err;
 265         }
 266
 267         if (l_old && map_flags == BPF_NOEXIST) {
 268                 /* elem already exists */
 269                 ret = -EEXIST;
 270                 goto err;
 271         }
 272
 273         if (!l_old && map_flags == BPF_EXIST) {
 274                 /* elem doesn't exist, cannot update it */
 275                 ret = -ENOENT;
 276                 goto err;
 277         }
 278
 279         /* add new element to the head of the list, so that concurrent
 280          * search will find it before old elem
 281          */
 282         hlist_add_head_rcu(&l_new->hash_node, head);
 283         if (l_old) {
 284                 hlist_del_rcu(&l_old->hash_node);
 285                 kfree_rcu(l_old, rcu);
 286         } else {
 287                 atomic_inc(&htab->count);
 288         }
 289         raw_spin_unlock_irqrestore(&htab->lock, flags);
 290
 291         return 0;
 292 err:
 293         raw_spin_unlock_irqrestore(&htab->lock, flags);
 294         kfree(l_new);
 295         return ret;
 296 }
 297
 298 /* Called from syscall or from eBPF program */
 299 static int htab_map_delete_elem(struct bpf_map *map, void *key)
 300 {
 301         struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
 302         struct hlist_head *head;
 303         struct htab_elem *l;
 304         unsigned long flags;
 305         u32 hash, key_size;
 306         int ret = -ENOENT;
 307
 308         WARN_ON_ONCE(!rcu_read_lock_held());
 309
 310         key_size = map->key_size;
 311
 312         hash = htab_map_hash(key, key_size);
 313
 314         raw_spin_lock_irqsave(&htab->lock, flags);
 315
 316         head = select_bucket(htab, hash);
 317
 318         l = lookup_elem_raw(head, hash, key, key_size);
 319
 320         if (l) {
 321                 hlist_del_rcu(&l->hash_node);
 322                 atomic_dec(&htab->count);
 323                 kfree_rcu(l, rcu);
 324                 ret = 0;
 325         }
 326
 327         raw_spin_unlock_irqrestore(&htab->lock, flags);
 328         return ret;
 329 }
 330
 331 static void delete_all_elements(struct bpf_htab *htab)
 332 {
 333         int i;
 334
 335         for (i = 0; i < htab->n_buckets; i++) {
 336                 struct hlist_head *head = select_bucket(htab, i);
 337                 struct hlist_node *n;
 338                 struct htab_elem *l;
 339
 340                 hlist_for_each_entry_safe(l, n, head, hash_node) {
 341                         hlist_del_rcu(&l->hash_node);
 342                         atomic_dec(&htab->count);
 343                         kfree(l);
 344                 }
 345         }
 346 }
 347
 348 /* Called when map->refcnt goes to zero, either from workqueue or from syscall */
 349 static void htab_map_free(struct bpf_map *map)
 350 {
 351         struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
 352
 353         /* at this point bpf_prog->aux->refcnt == 0 and this map->refcnt == 0,
 354          * so the programs (can be more than one that used this map) were
 355          * disconnected from events. Wait for outstanding critical sections in
 356          * these programs to complete
 357          */
 358         synchronize_rcu();
 359
 360         /* some of kfree_rcu() callbacks for elements of this map may not have
 361          * executed. It's ok. Proceed to free residual elements and map itself
 362          */
 363         delete_all_elements(htab);
 364         kvfree(htab->buckets);
 365         kfree(htab);
 366 }
 367
 368 static const struct bpf_map_ops htab_ops = {
 369         .map_alloc = htab_map_alloc,
 370         .map_free = htab_map_free,
 371         .map_get_next_key = htab_map_get_next_key,
 372         .map_lookup_elem = htab_map_lookup_elem,
 373         .map_update_elem = htab_map_update_elem,
 374         .map_delete_elem = htab_map_delete_elem,
 375 };
 376
 377 static struct bpf_map_type_list htab_type __read_mostly = {
 378         .ops = &htab_ops,
 379         .type = BPF_MAP_TYPE_HASH,
 380 };
 381
 382 static int __init register_htab_map(void)
 383 {
 384         bpf_register_map_type(&htab_type);
 385         return 0;
 386 }
 387 late_initcall(register_htab_map);