kernel/bpf/hashtab.c

   1 /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com
   2  * Copyright (c) 2016 Facebook
   3  *
   4  * This program is free software; you can redistribute it and/or
   5  * modify it under the terms of version 2 of the GNU General Public
   6  * License as published by the Free Software Foundation.
   7  *
   8  * This program is distributed in the hope that it will be useful, but
   9  * WITHOUT ANY WARRANTY; without even the implied warranty of
  10  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
  11  * General Public License for more details.
  12  */
  13 #include <linux/bpf.h>
  14 #include <linux/jhash.h>
  15 #include <linux/filter.h>
  16 #include <linux/vmalloc.h>
  17 #include "percpu_freelist.h"
  18 #include "bpf_lru_list.h"
  19
  20 struct bucket {
  21         struct hlist_head head;
  22         raw_spinlock_t lock;
  23 };
  24
  25 struct bpf_htab {
  26         struct bpf_map map;
  27         struct bucket *buckets;
  28         void *elems;
  29         union {
  30                 struct pcpu_freelist freelist;
  31                 struct bpf_lru lru;
  32         };
  33         void __percpu *extra_elems;
  34         atomic_t count; /* number of elements in this hashtable */
  35         u32 n_buckets;  /* number of hash buckets */
  36         u32 elem_size;  /* size of each element in bytes */
  37 };
  38
  39 enum extra_elem_state {
  40         HTAB_NOT_AN_EXTRA_ELEM = 0,
  41         HTAB_EXTRA_ELEM_FREE,
  42         HTAB_EXTRA_ELEM_USED
  43 };
  44
  45 /* each htab element is struct htab_elem + key + value */
  46 struct htab_elem {
  47         union {
  48                 struct hlist_node hash_node;
  49                 struct bpf_htab *htab;
  50                 struct pcpu_freelist_node fnode;
  51         };
  52         union {
  53                 struct rcu_head rcu;
  54                 enum extra_elem_state state;
  55                 struct bpf_lru_node lru_node;
  56         };
  57         u32 hash;
  58         char key[0] __aligned(8);
  59 };
  60
  61 static bool htab_lru_map_delete_node(void *arg, struct bpf_lru_node *node);
  62
  63 static bool htab_is_lru(const struct bpf_htab *htab)
  64 {
  65         return htab->map.map_type == BPF_MAP_TYPE_LRU_HASH ||
  66                 htab->map.map_type == BPF_MAP_TYPE_LRU_PERCPU_HASH;
  67 }
  68
  69 static bool htab_is_percpu(const struct bpf_htab *htab)
  70 {
  71         return htab->map.map_type == BPF_MAP_TYPE_PERCPU_HASH ||
  72                 htab->map.map_type == BPF_MAP_TYPE_LRU_PERCPU_HASH;
  73 }
  74
  75 static inline void htab_elem_set_ptr(struct htab_elem *l, u32 key_size,
  76                                      void __percpu *pptr)
  77 {
  78         *(void __percpu **)(l->key + key_size) = pptr;
  79 }
  80
  81 static inline void __percpu *htab_elem_get_ptr(struct htab_elem *l, u32 key_size)
  82 {
  83         return *(void __percpu **)(l->key + key_size);
  84 }
  85
  86 static struct htab_elem *get_htab_elem(struct bpf_htab *htab, int i)
  87 {
  88         return (struct htab_elem *) (htab->elems + i * htab->elem_size);
  89 }
  90
  91 static void htab_free_elems(struct bpf_htab *htab)
  92 {
  93         int i;
  94
  95         if (!htab_is_percpu(htab))
  96                 goto free_elems;
  97
  98         for (i = 0; i < htab->map.max_entries; i++) {
  99                 void __percpu *pptr;
 100
 101                 pptr = htab_elem_get_ptr(get_htab_elem(htab, i),
 102                                          htab->map.key_size);
 103                 free_percpu(pptr);
 104         }
 105 free_elems:
 106         vfree(htab->elems);
 107 }
 108
 109 static struct htab_elem *prealloc_lru_pop(struct bpf_htab *htab, void *key,
 110                                           u32 hash)
 111 {
 112         struct bpf_lru_node *node = bpf_lru_pop_free(&htab->lru, hash);
 113         struct htab_elem *l;
 114
 115         if (node) {
 116                 l = container_of(node, struct htab_elem, lru_node);
 117                 memcpy(l->key, key, htab->map.key_size);
 118                 return l;
 119         }
 120
 121         return NULL;
 122 }
 123
 124 static int prealloc_init(struct bpf_htab *htab)
 125 {
 126         int err = -ENOMEM, i;
 127
 128         htab->elems = vzalloc(htab->elem_size * htab->map.max_entries);
 129         if (!htab->elems)
 130                 return -ENOMEM;
 131
 132         if (!htab_is_percpu(htab))
 133                 goto skip_percpu_elems;
 134
 135         for (i = 0; i < htab->map.max_entries; i++) {
 136                 u32 size = round_up(htab->map.value_size, 8);
 137                 void __percpu *pptr;
 138
 139                 pptr = __alloc_percpu_gfp(size, 8, GFP_USER | __GFP_NOWARN);
 140                 if (!pptr)
 141                         goto free_elems;
 142                 htab_elem_set_ptr(get_htab_elem(htab, i), htab->map.key_size,
 143                                   pptr);
 144         }
 145
 146 skip_percpu_elems:
 147         if (htab_is_lru(htab))
 148                 err = bpf_lru_init(&htab->lru,
 149                                    htab->map.map_flags & BPF_F_NO_COMMON_LRU,
 150                                    offsetof(struct htab_elem, hash) -
 151                                    offsetof(struct htab_elem, lru_node),
 152                                    htab_lru_map_delete_node,
 153                                    htab);
 154         else
 155                 err = pcpu_freelist_init(&htab->freelist);
 156
 157         if (err)
 158                 goto free_elems;
 159
 160         if (htab_is_lru(htab))
 161                 bpf_lru_populate(&htab->lru, htab->elems,
 162                                  offsetof(struct htab_elem, lru_node),
 163                                  htab->elem_size, htab->map.max_entries);
 164         else
 165                 pcpu_freelist_populate(&htab->freelist, htab->elems,
 166                                        htab->elem_size, htab->map.max_entries);
 167
 168         return 0;
 169
 170 free_elems:
 171         htab_free_elems(htab);
 172         return err;
 173 }
 174
 175 static void prealloc_destroy(struct bpf_htab *htab)
 176 {
 177         htab_free_elems(htab);
 178
 179         if (htab_is_lru(htab))
 180                 bpf_lru_destroy(&htab->lru);
 181         else
 182                 pcpu_freelist_destroy(&htab->freelist);
 183 }
 184
 185 static int alloc_extra_elems(struct bpf_htab *htab)
 186 {
 187         void __percpu *pptr;
 188         int cpu;
 189
 190         pptr = __alloc_percpu_gfp(htab->elem_size, 8, GFP_USER | __GFP_NOWARN);
 191         if (!pptr)
 192                 return -ENOMEM;
 193
 194         for_each_possible_cpu(cpu) {
 195                 ((struct htab_elem *)per_cpu_ptr(pptr, cpu))->state =
 196                         HTAB_EXTRA_ELEM_FREE;
 197         }
 198         htab->extra_elems = pptr;
 199         return 0;
 200 }
 201
 202 /* Called from syscall */
 203 static struct bpf_map *htab_map_alloc(union bpf_attr *attr)
 204 {
 205         bool percpu = (attr->map_type == BPF_MAP_TYPE_PERCPU_HASH ||
 206                        attr->map_type == BPF_MAP_TYPE_LRU_PERCPU_HASH);
 207         bool lru = (attr->map_type == BPF_MAP_TYPE_LRU_HASH ||
 208                     attr->map_type == BPF_MAP_TYPE_LRU_PERCPU_HASH);
 209         /* percpu_lru means each cpu has its own LRU list.
 210          * it is different from BPF_MAP_TYPE_PERCPU_HASH where
 211          * the map's value itself is percpu.  percpu_lru has
 212          * nothing to do with the map's value.
 213          */
 214         bool percpu_lru = (attr->map_flags & BPF_F_NO_COMMON_LRU);
 215         bool prealloc = !(attr->map_flags & BPF_F_NO_PREALLOC);
 216         struct bpf_htab *htab;
 217         int err, i;
 218         u64 cost;
 219
 220         if (lru && !capable(CAP_SYS_ADMIN))
 221                 /* LRU implementation is much complicated than other
 222                  * maps.  Hence, limit to CAP_SYS_ADMIN for now.
 223                  */
 224                 return ERR_PTR(-EPERM);
 225
 226         if (attr->map_flags & ~(BPF_F_NO_PREALLOC | BPF_F_NO_COMMON_LRU))
 227                 /* reserved bits should not be used */
 228                 return ERR_PTR(-EINVAL);
 229
 230         if (!lru && percpu_lru)
 231                 return ERR_PTR(-EINVAL);
 232
 233         if (lru && !prealloc)
 234                 return ERR_PTR(-ENOTSUPP);
 235
 236         htab = kzalloc(sizeof(*htab), GFP_USER);
 237         if (!htab)
 238                 return ERR_PTR(-ENOMEM);
 239
 240         /* mandatory map attributes */
 241         htab->map.map_type = attr->map_type;
 242         htab->map.key_size = attr->key_size;
 243         htab->map.value_size = attr->value_size;
 244         htab->map.max_entries = attr->max_entries;
 245         htab->map.map_flags = attr->map_flags;
 246
 247         /* check sanity of attributes.
 248          * value_size == 0 may be allowed in the future to use map as a set
 249          */
 250         err = -EINVAL;
 251         if (htab->map.max_entries == 0 || htab->map.key_size == 0 ||
 252             htab->map.value_size == 0)
 253                 goto free_htab;
 254
 255         if (percpu_lru) {
 256                 /* ensure each CPU's lru list has >=1 elements.
 257                  * since we are at it, make each lru list has the same
 258                  * number of elements.
 259                  */
 260                 htab->map.max_entries = roundup(attr->max_entries,
 261                                                 num_possible_cpus());
 262                 if (htab->map.max_entries < attr->max_entries)
 263                         htab->map.max_entries = rounddown(attr->max_entries,
 264                                                           num_possible_cpus());
 265         }
 266
 267         /* hash table size must be power of 2 */
 268         htab->n_buckets = roundup_pow_of_two(htab->map.max_entries);
 269
 270         err = -E2BIG;
 271         if (htab->map.key_size > MAX_BPF_STACK)
 272                 /* eBPF programs initialize keys on stack, so they cannot be
 273                  * larger than max stack size
 274                  */
 275                 goto free_htab;
 276
 277         if (htab->map.value_size >= (1 << (KMALLOC_SHIFT_MAX - 1)) -
 278             MAX_BPF_STACK - sizeof(struct htab_elem))
 279                 /* if value_size is bigger, the user space won't be able to
 280                  * access the elements via bpf syscall. This check also makes
 281                  * sure that the elem_size doesn't overflow and it's
 282                  * kmalloc-able later in htab_map_update_elem()
 283                  */
 284                 goto free_htab;
 285
 286         if (percpu && round_up(htab->map.value_size, 8) > PCPU_MIN_UNIT_SIZE)
 287                 /* make sure the size for pcpu_alloc() is reasonable */
 288                 goto free_htab;
 289
 290         htab->elem_size = sizeof(struct htab_elem) +
 291                           round_up(htab->map.key_size, 8);
 292         if (percpu)
 293                 htab->elem_size += sizeof(void *);
 294         else
 295                 htab->elem_size += round_up(htab->map.value_size, 8);
 296
 297         /* prevent zero size kmalloc and check for u32 overflow */
 298         if (htab->n_buckets == 0 ||
 299             htab->n_buckets > U32_MAX / sizeof(struct bucket))
 300                 goto free_htab;
 301
 302         cost = (u64) htab->n_buckets * sizeof(struct bucket) +
 303                (u64) htab->elem_size * htab->map.max_entries;
 304
 305         if (percpu)
 306                 cost += (u64) round_up(htab->map.value_size, 8) *
 307                         num_possible_cpus() * htab->map.max_entries;
 308         else
 309                cost += (u64) htab->elem_size * num_possible_cpus();
 310
 311         if (cost >= U32_MAX - PAGE_SIZE)
 312                 /* make sure page count doesn't overflow */
 313                 goto free_htab;
 314
 315         htab->map.pages = round_up(cost, PAGE_SIZE) >> PAGE_SHIFT;
 316
 317         /* if map size is larger than memlock limit, reject it early */
 318         err = bpf_map_precharge_memlock(htab->map.pages);
 319         if (err)
 320                 goto free_htab;
 321
 322         err = -ENOMEM;
 323         htab->buckets = kmalloc_array(htab->n_buckets, sizeof(struct bucket),
 324                                       GFP_USER | __GFP_NOWARN);
 325
 326         if (!htab->buckets) {
 327                 htab->buckets = vmalloc(htab->n_buckets * sizeof(struct bucket));
 328                 if (!htab->buckets)
 329                         goto free_htab;
 330         }
 331
 332         for (i = 0; i < htab->n_buckets; i++) {
 333                 INIT_HLIST_HEAD(&htab->buckets[i].head);
 334                 raw_spin_lock_init(&htab->buckets[i].lock);
 335         }
 336
 337         if (!percpu && !lru) {
 338                 /* lru itself can remove the least used element, so
 339                  * there is no need for an extra elem during map_update.
 340                  */
 341                 err = alloc_extra_elems(htab);
 342                 if (err)
 343                         goto free_buckets;
 344         }
 345
 346         if (prealloc) {
 347                 err = prealloc_init(htab);
 348                 if (err)
 349                         goto free_extra_elems;
 350         }
 351
 352         return &htab->map;
 353
 354 free_extra_elems:
 355         free_percpu(htab->extra_elems);
 356 free_buckets:
 357         kvfree(htab->buckets);
 358 free_htab:
 359         kfree(htab);
 360         return ERR_PTR(err);
 361 }
 362
 363 static inline u32 htab_map_hash(const void *key, u32 key_len)
 364 {
 365         return jhash(key, key_len, 0);
 366 }
 367
 368 static inline struct bucket *__select_bucket(struct bpf_htab *htab, u32 hash)
 369 {
 370         return &htab->buckets[hash & (htab->n_buckets - 1)];
 371 }
 372
 373 static inline struct hlist_head *select_bucket(struct bpf_htab *htab, u32 hash)
 374 {
 375         return &__select_bucket(htab, hash)->head;
 376 }
 377
 378 static struct htab_elem *lookup_elem_raw(struct hlist_head *head, u32 hash,
 379                                          void *key, u32 key_size)
 380 {
 381         struct htab_elem *l;
 382
 383         hlist_for_each_entry_rcu(l, head, hash_node)
 384                 if (l->hash == hash && !memcmp(&l->key, key, key_size))
 385                         return l;
 386
 387         return NULL;
 388 }
 389
 390 /* Called from syscall or from eBPF program */
 391 static void *__htab_map_lookup_elem(struct bpf_map *map, void *key)
 392 {
 393         struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
 394         struct hlist_head *head;
 395         struct htab_elem *l;
 396         u32 hash, key_size;
 397
 398         /* Must be called with rcu_read_lock. */
 399         WARN_ON_ONCE(!rcu_read_lock_held());
 400
 401         key_size = map->key_size;
 402
 403         hash = htab_map_hash(key, key_size);
 404
 405         head = select_bucket(htab, hash);
 406
 407         l = lookup_elem_raw(head, hash, key, key_size);
 408
 409         return l;
 410 }
 411
 412 static void *htab_map_lookup_elem(struct bpf_map *map, void *key)
 413 {
 414         struct htab_elem *l = __htab_map_lookup_elem(map, key);
 415
 416         if (l)
 417                 return l->key + round_up(map->key_size, 8);
 418
 419         return NULL;
 420 }
 421
 422 static void *htab_lru_map_lookup_elem(struct bpf_map *map, void *key)
 423 {
 424         struct htab_elem *l = __htab_map_lookup_elem(map, key);
 425
 426         if (l) {
 427                 bpf_lru_node_set_ref(&l->lru_node);
 428                 return l->key + round_up(map->key_size, 8);
 429         }
 430
 431         return NULL;
 432 }
 433
 434 /* It is called from the bpf_lru_list when the LRU needs to delete
 435  * older elements from the htab.
 436  */
 437 static bool htab_lru_map_delete_node(void *arg, struct bpf_lru_node *node)
 438 {
 439         struct bpf_htab *htab = (struct bpf_htab *)arg;
 440         struct htab_elem *l, *tgt_l;
 441         struct hlist_head *head;
 442         unsigned long flags;
 443         struct bucket *b;
 444
 445         tgt_l = container_of(node, struct htab_elem, lru_node);
 446         b = __select_bucket(htab, tgt_l->hash);
 447         head = &b->head;
 448
 449         raw_spin_lock_irqsave(&b->lock, flags);
 450
 451         hlist_for_each_entry_rcu(l, head, hash_node)
 452                 if (l == tgt_l) {
 453                         hlist_del_rcu(&l->hash_node);
 454                         break;
 455                 }
 456
 457         raw_spin_unlock_irqrestore(&b->lock, flags);
 458
 459         return l == tgt_l;
 460 }
 461
 462 /* Called from syscall */
 463 static int htab_map_get_next_key(struct bpf_map *map, void *key, void *next_key)
 464 {
 465         struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
 466         struct hlist_head *head;
 467         struct htab_elem *l, *next_l;
 468         u32 hash, key_size;
 469         int i;
 470
 471         WARN_ON_ONCE(!rcu_read_lock_held());
 472
 473         key_size = map->key_size;
 474
 475         hash = htab_map_hash(key, key_size);
 476
 477         head = select_bucket(htab, hash);
 478
 479         /* lookup the key */
 480         l = lookup_elem_raw(head, hash, key, key_size);
 481
 482         if (!l) {
 483                 i = 0;
 484                 goto find_first_elem;
 485         }
 486
 487         /* key was found, get next key in the same bucket */
 488         next_l = hlist_entry_safe(rcu_dereference_raw(hlist_next_rcu(&l->hash_node)),
 489                                   struct htab_elem, hash_node);
 490
 491         if (next_l) {
 492                 /* if next elem in this hash list is non-zero, just return it */
 493                 memcpy(next_key, next_l->key, key_size);
 494                 return 0;
 495         }
 496
 497         /* no more elements in this hash list, go to the next bucket */
 498         i = hash & (htab->n_buckets - 1);
 499         i++;
 500
 501 find_first_elem:
 502         /* iterate over buckets */
 503         for (; i < htab->n_buckets; i++) {
 504                 head = select_bucket(htab, i);
 505
 506                 /* pick first element in the bucket */
 507                 next_l = hlist_entry_safe(rcu_dereference_raw(hlist_first_rcu(head)),
 508                                           struct htab_elem, hash_node);
 509                 if (next_l) {
 510                         /* if it's not empty, just return it */
 511                         memcpy(next_key, next_l->key, key_size);
 512                         return 0;
 513                 }
 514         }
 515
 516         /* iterated over all buckets and all elements */
 517         return -ENOENT;
 518 }
 519
 520 static void htab_elem_free(struct bpf_htab *htab, struct htab_elem *l)
 521 {
 522         if (htab->map.map_type == BPF_MAP_TYPE_PERCPU_HASH)
 523                 free_percpu(htab_elem_get_ptr(l, htab->map.key_size));
 524         kfree(l);
 525 }
 526
 527 static void htab_elem_free_rcu(struct rcu_head *head)
 528 {
 529         struct htab_elem *l = container_of(head, struct htab_elem, rcu);
 530         struct bpf_htab *htab = l->htab;
 531
 532         /* must increment bpf_prog_active to avoid kprobe+bpf triggering while
 533          * we're calling kfree, otherwise deadlock is possible if kprobes
 534          * are placed somewhere inside of slub
 535          */
 536         preempt_disable();
 537         __this_cpu_inc(bpf_prog_active);
 538         htab_elem_free(htab, l);
 539         __this_cpu_dec(bpf_prog_active);
 540         preempt_enable();
 541 }
 542
 543 static void free_htab_elem(struct bpf_htab *htab, struct htab_elem *l)
 544 {
 545         if (l->state == HTAB_EXTRA_ELEM_USED) {
 546                 l->state = HTAB_EXTRA_ELEM_FREE;
 547                 return;
 548         }
 549
 550         if (!(htab->map.map_flags & BPF_F_NO_PREALLOC)) {
 551                 pcpu_freelist_push(&htab->freelist, &l->fnode);
 552         } else {
 553                 atomic_dec(&htab->count);
 554                 l->htab = htab;
 555                 call_rcu(&l->rcu, htab_elem_free_rcu);
 556         }
 557 }
 558
 559 static void pcpu_copy_value(struct bpf_htab *htab, void __percpu *pptr,
 560                             void *value, bool onallcpus)
 561 {
 562         if (!onallcpus) {
 563                 /* copy true value_size bytes */
 564                 memcpy(this_cpu_ptr(pptr), value, htab->map.value_size);
 565         } else {
 566                 u32 size = round_up(htab->map.value_size, 8);
 567                 int off = 0, cpu;
 568
 569                 for_each_possible_cpu(cpu) {
 570                         bpf_long_memcpy(per_cpu_ptr(pptr, cpu),
 571                                         value + off, size);
 572                         off += size;
 573                 }
 574         }
 575 }
 576
 577 static struct htab_elem *alloc_htab_elem(struct bpf_htab *htab, void *key,
 578                                          void *value, u32 key_size, u32 hash,
 579                                          bool percpu, bool onallcpus,
 580                                          bool old_elem_exists)
 581 {
 582         u32 size = htab->map.value_size;
 583         bool prealloc = !(htab->map.map_flags & BPF_F_NO_PREALLOC);
 584         struct htab_elem *l_new;
 585         void __percpu *pptr;
 586         int err = 0;
 587
 588         if (prealloc) {
 589                 l_new = (struct htab_elem *)pcpu_freelist_pop(&htab->freelist);
 590                 if (!l_new)
 591                         err = -E2BIG;
 592         } else {
 593                 if (atomic_inc_return(&htab->count) > htab->map.max_entries) {
 594                         atomic_dec(&htab->count);
 595                         err = -E2BIG;
 596                 } else {
 597                         l_new = kmalloc(htab->elem_size,
 598                                         GFP_ATOMIC | __GFP_NOWARN);
 599                         if (!l_new)
 600                                 return ERR_PTR(-ENOMEM);
 601                 }
 602         }
 603
 604         if (err) {
 605                 if (!old_elem_exists)
 606                         return ERR_PTR(err);
 607
 608                 /* if we're updating the existing element and the hash table
 609                  * is full, use per-cpu extra elems
 610                  */
 611                 l_new = this_cpu_ptr(htab->extra_elems);
 612                 if (l_new->state != HTAB_EXTRA_ELEM_FREE)
 613                         return ERR_PTR(-E2BIG);
 614                 l_new->state = HTAB_EXTRA_ELEM_USED;
 615         } else {
 616                 l_new->state = HTAB_NOT_AN_EXTRA_ELEM;
 617         }
 618
 619         memcpy(l_new->key, key, key_size);
 620         if (percpu) {
 621                 /* round up value_size to 8 bytes */
 622                 size = round_up(size, 8);
 623
 624                 if (prealloc) {
 625                         pptr = htab_elem_get_ptr(l_new, key_size);
 626                 } else {
 627                         /* alloc_percpu zero-fills */
 628                         pptr = __alloc_percpu_gfp(size, 8,
 629                                                   GFP_ATOMIC | __GFP_NOWARN);
 630                         if (!pptr) {
 631                                 kfree(l_new);
 632                                 return ERR_PTR(-ENOMEM);
 633                         }
 634                 }
 635
 636                 pcpu_copy_value(htab, pptr, value, onallcpus);
 637
 638                 if (!prealloc)
 639                         htab_elem_set_ptr(l_new, key_size, pptr);
 640         } else {
 641                 memcpy(l_new->key + round_up(key_size, 8), value, size);
 642         }
 643
 644         l_new->hash = hash;
 645         return l_new;
 646 }
 647
 648 static int check_flags(struct bpf_htab *htab, struct htab_elem *l_old,
 649                        u64 map_flags)
 650 {
 651         if (l_old && map_flags == BPF_NOEXIST)
 652                 /* elem already exists */
 653                 return -EEXIST;
 654
 655         if (!l_old && map_flags == BPF_EXIST)
 656                 /* elem doesn't exist, cannot update it */
 657                 return -ENOENT;
 658
 659         return 0;
 660 }
 661
 662 /* Called from syscall or from eBPF program */
 663 static int htab_map_update_elem(struct bpf_map *map, void *key, void *value,
 664                                 u64 map_flags)
 665 {
 666         struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
 667         struct htab_elem *l_new = NULL, *l_old;
 668         struct hlist_head *head;
 669         unsigned long flags;
 670         struct bucket *b;
 671         u32 key_size, hash;
 672         int ret;
 673
 674         if (unlikely(map_flags > BPF_EXIST))
 675                 /* unknown flags */
 676                 return -EINVAL;
 677
 678         WARN_ON_ONCE(!rcu_read_lock_held());
 679
 680         key_size = map->key_size;
 681
 682         hash = htab_map_hash(key, key_size);
 683
 684         b = __select_bucket(htab, hash);
 685         head = &b->head;
 686
 687         /* bpf_map_update_elem() can be called in_irq() */
 688         raw_spin_lock_irqsave(&b->lock, flags);
 689
 690         l_old = lookup_elem_raw(head, hash, key, key_size);
 691
 692         ret = check_flags(htab, l_old, map_flags);
 693         if (ret)
 694                 goto err;
 695
 696         l_new = alloc_htab_elem(htab, key, value, key_size, hash, false, false,
 697                                 !!l_old);
 698         if (IS_ERR(l_new)) {
 699                 /* all pre-allocated elements are in use or memory exhausted */
 700                 ret = PTR_ERR(l_new);
 701                 goto err;
 702         }
 703
 704         /* add new element to the head of the list, so that
 705          * concurrent search will find it before old elem
 706          */
 707         hlist_add_head_rcu(&l_new->hash_node, head);
 708         if (l_old) {
 709                 hlist_del_rcu(&l_old->hash_node);
 710                 free_htab_elem(htab, l_old);
 711         }
 712         ret = 0;
 713 err:
 714         raw_spin_unlock_irqrestore(&b->lock, flags);
 715         return ret;
 716 }
 717
 718 static int htab_lru_map_update_elem(struct bpf_map *map, void *key, void *value,
 719                                     u64 map_flags)
 720 {
 721         struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
 722         struct htab_elem *l_new, *l_old = NULL;
 723         struct hlist_head *head;
 724         unsigned long flags;
 725         struct bucket *b;
 726         u32 key_size, hash;
 727         int ret;
 728
 729         if (unlikely(map_flags > BPF_EXIST))
 730                 /* unknown flags */
 731                 return -EINVAL;
 732
 733         WARN_ON_ONCE(!rcu_read_lock_held());
 734
 735         key_size = map->key_size;
 736
 737         hash = htab_map_hash(key, key_size);
 738
 739         b = __select_bucket(htab, hash);
 740         head = &b->head;
 741
 742         /* For LRU, we need to alloc before taking bucket's
 743          * spinlock because getting free nodes from LRU may need
 744          * to remove older elements from htab and this removal
 745          * operation will need a bucket lock.
 746          */
 747         l_new = prealloc_lru_pop(htab, key, hash);
 748         if (!l_new)
 749                 return -ENOMEM;
 750         memcpy(l_new->key + round_up(map->key_size, 8), value, map->value_size);
 751
 752         /* bpf_map_update_elem() can be called in_irq() */
 753         raw_spin_lock_irqsave(&b->lock, flags);
 754
 755         l_old = lookup_elem_raw(head, hash, key, key_size);
 756
 757         ret = check_flags(htab, l_old, map_flags);
 758         if (ret)
 759                 goto err;
 760
 761         /* add new element to the head of the list, so that
 762          * concurrent search will find it before old elem
 763          */
 764         hlist_add_head_rcu(&l_new->hash_node, head);
 765         if (l_old) {
 766                 bpf_lru_node_set_ref(&l_new->lru_node);
 767                 hlist_del_rcu(&l_old->hash_node);
 768         }
 769         ret = 0;
 770
 771 err:
 772         raw_spin_unlock_irqrestore(&b->lock, flags);
 773
 774         if (ret)
 775                 bpf_lru_push_free(&htab->lru, &l_new->lru_node);
 776         else if (l_old)
 777                 bpf_lru_push_free(&htab->lru, &l_old->lru_node);
 778
 779         return ret;
 780 }
 781
 782 static int __htab_percpu_map_update_elem(struct bpf_map *map, void *key,
 783                                          void *value, u64 map_flags,
 784                                          bool onallcpus)
 785 {
 786         struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
 787         struct htab_elem *l_new = NULL, *l_old;
 788         struct hlist_head *head;
 789         unsigned long flags;
 790         struct bucket *b;
 791         u32 key_size, hash;
 792         int ret;
 793
 794         if (unlikely(map_flags > BPF_EXIST))
 795                 /* unknown flags */
 796                 return -EINVAL;
 797
 798         WARN_ON_ONCE(!rcu_read_lock_held());
 799
 800         key_size = map->key_size;
 801
 802         hash = htab_map_hash(key, key_size);
 803
 804         b = __select_bucket(htab, hash);
 805         head = &b->head;
 806
 807         /* bpf_map_update_elem() can be called in_irq() */
 808         raw_spin_lock_irqsave(&b->lock, flags);
 809
 810         l_old = lookup_elem_raw(head, hash, key, key_size);
 811
 812         ret = check_flags(htab, l_old, map_flags);
 813         if (ret)
 814                 goto err;
 815
 816         if (l_old) {
 817                 /* per-cpu hash map can update value in-place */
 818                 pcpu_copy_value(htab, htab_elem_get_ptr(l_old, key_size),
 819                                 value, onallcpus);
 820         } else {
 821                 l_new = alloc_htab_elem(htab, key, value, key_size,
 822                                         hash, true, onallcpus, false);
 823                 if (IS_ERR(l_new)) {
 824                         ret = PTR_ERR(l_new);
 825                         goto err;
 826                 }
 827                 hlist_add_head_rcu(&l_new->hash_node, head);
 828         }
 829         ret = 0;
 830 err:
 831         raw_spin_unlock_irqrestore(&b->lock, flags);
 832         return ret;
 833 }
 834
 835 static int __htab_lru_percpu_map_update_elem(struct bpf_map *map, void *key,
 836                                              void *value, u64 map_flags,
 837                                              bool onallcpus)
 838 {
 839         struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
 840         struct htab_elem *l_new = NULL, *l_old;
 841         struct hlist_head *head;
 842         unsigned long flags;
 843         struct bucket *b;
 844         u32 key_size, hash;
 845         int ret;
 846
 847         if (unlikely(map_flags > BPF_EXIST))
 848                 /* unknown flags */
 849                 return -EINVAL;
 850
 851         WARN_ON_ONCE(!rcu_read_lock_held());
 852
 853         key_size = map->key_size;
 854
 855         hash = htab_map_hash(key, key_size);
 856
 857         b = __select_bucket(htab, hash);
 858         head = &b->head;
 859
 860         /* For LRU, we need to alloc before taking bucket's
 861          * spinlock because LRU's elem alloc may need
 862          * to remove older elem from htab and this removal
 863          * operation will need a bucket lock.
 864          */
 865         if (map_flags != BPF_EXIST) {
 866                 l_new = prealloc_lru_pop(htab, key, hash);
 867                 if (!l_new)
 868                         return -ENOMEM;
 869         }
 870
 871         /* bpf_map_update_elem() can be called in_irq() */
 872         raw_spin_lock_irqsave(&b->lock, flags);
 873
 874         l_old = lookup_elem_raw(head, hash, key, key_size);
 875
 876         ret = check_flags(htab, l_old, map_flags);
 877         if (ret)
 878                 goto err;
 879
 880         if (l_old) {
 881                 bpf_lru_node_set_ref(&l_old->lru_node);
 882
 883                 /* per-cpu hash map can update value in-place */
 884                 pcpu_copy_value(htab, htab_elem_get_ptr(l_old, key_size),
 885                                 value, onallcpus);
 886         } else {
 887                 pcpu_copy_value(htab, htab_elem_get_ptr(l_new, key_size),
 888                                 value, onallcpus);
 889                 hlist_add_head_rcu(&l_new->hash_node, head);
 890                 l_new = NULL;
 891         }
 892         ret = 0;
 893 err:
 894         raw_spin_unlock_irqrestore(&b->lock, flags);
 895         if (l_new)
 896                 bpf_lru_push_free(&htab->lru, &l_new->lru_node);
 897         return ret;
 898 }
 899
 900 static int htab_percpu_map_update_elem(struct bpf_map *map, void *key,
 901                                        void *value, u64 map_flags)
 902 {
 903         return __htab_percpu_map_update_elem(map, key, value, map_flags, false);
 904 }
 905
 906 static int htab_lru_percpu_map_update_elem(struct bpf_map *map, void *key,
 907                                            void *value, u64 map_flags)
 908 {
 909         return __htab_lru_percpu_map_update_elem(map, key, value, map_flags,
 910                                                  false);
 911 }
 912
 913 /* Called from syscall or from eBPF program */
 914 static int htab_map_delete_elem(struct bpf_map *map, void *key)
 915 {
 916         struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
 917         struct hlist_head *head;
 918         struct bucket *b;
 919         struct htab_elem *l;
 920         unsigned long flags;
 921         u32 hash, key_size;
 922         int ret = -ENOENT;
 923
 924         WARN_ON_ONCE(!rcu_read_lock_held());
 925
 926         key_size = map->key_size;
 927
 928         hash = htab_map_hash(key, key_size);
 929         b = __select_bucket(htab, hash);
 930         head = &b->head;
 931
 932         raw_spin_lock_irqsave(&b->lock, flags);
 933
 934         l = lookup_elem_raw(head, hash, key, key_size);
 935
 936         if (l) {
 937                 hlist_del_rcu(&l->hash_node);
 938                 free_htab_elem(htab, l);
 939                 ret = 0;
 940         }
 941
 942         raw_spin_unlock_irqrestore(&b->lock, flags);
 943         return ret;
 944 }
 945
 946 static int htab_lru_map_delete_elem(struct bpf_map *map, void *key)
 947 {
 948         struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
 949         struct hlist_head *head;
 950         struct bucket *b;
 951         struct htab_elem *l;
 952         unsigned long flags;
 953         u32 hash, key_size;
 954         int ret = -ENOENT;
 955
 956         WARN_ON_ONCE(!rcu_read_lock_held());
 957
 958         key_size = map->key_size;
 959
 960         hash = htab_map_hash(key, key_size);
 961         b = __select_bucket(htab, hash);
 962         head = &b->head;
 963
 964         raw_spin_lock_irqsave(&b->lock, flags);
 965
 966         l = lookup_elem_raw(head, hash, key, key_size);
 967
 968         if (l) {
 969                 hlist_del_rcu(&l->hash_node);
 970                 ret = 0;
 971         }
 972
 973         raw_spin_unlock_irqrestore(&b->lock, flags);
 974         if (l)
 975                 bpf_lru_push_free(&htab->lru, &l->lru_node);
 976         return ret;
 977 }
 978
 979 static void delete_all_elements(struct bpf_htab *htab)
 980 {
 981         int i;
 982
 983         for (i = 0; i < htab->n_buckets; i++) {
 984                 struct hlist_head *head = select_bucket(htab, i);
 985                 struct hlist_node *n;
 986                 struct htab_elem *l;
 987
 988                 hlist_for_each_entry_safe(l, n, head, hash_node) {
 989                         hlist_del_rcu(&l->hash_node);
 990                         if (l->state != HTAB_EXTRA_ELEM_USED)
 991                                 htab_elem_free(htab, l);
 992                 }
 993         }
 994 }
 995 /* Called when map->refcnt goes to zero, either from workqueue or from syscall */
 996 static void htab_map_free(struct bpf_map *map)
 997 {
 998         struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
 999
1000         /* at this point bpf_prog->aux->refcnt == 0 and this map->refcnt == 0,
1001          * so the programs (can be more than one that used this map) were
1002          * disconnected from events. Wait for outstanding critical sections in
1003          * these programs to complete
1004          */
1005         synchronize_rcu();
1006
1007         /* some of free_htab_elem() callbacks for elements of this map may
1008          * not have executed. Wait for them.
1009          */
1010         rcu_barrier();
1011         if (htab->map.map_flags & BPF_F_NO_PREALLOC)
1012                 delete_all_elements(htab);
1013         else
1014                 prealloc_destroy(htab);
1015
1016         free_percpu(htab->extra_elems);
1017         kvfree(htab->buckets);
1018         kfree(htab);
1019 }
1020
1021 static const struct bpf_map_ops htab_ops = {
1022         .map_alloc = htab_map_alloc,
1023         .map_free = htab_map_free,
1024         .map_get_next_key = htab_map_get_next_key,
1025         .map_lookup_elem = htab_map_lookup_elem,
1026         .map_update_elem = htab_map_update_elem,
1027         .map_delete_elem = htab_map_delete_elem,
1028 };
1029
1030 static struct bpf_map_type_list htab_type __read_mostly = {
1031         .ops = &htab_ops,
1032         .type = BPF_MAP_TYPE_HASH,
1033 };
1034
1035 static const struct bpf_map_ops htab_lru_ops = {
1036         .map_alloc = htab_map_alloc,
1037         .map_free = htab_map_free,
1038         .map_get_next_key = htab_map_get_next_key,
1039         .map_lookup_elem = htab_lru_map_lookup_elem,
1040         .map_update_elem = htab_lru_map_update_elem,
1041         .map_delete_elem = htab_lru_map_delete_elem,
1042 };
1043
1044 static struct bpf_map_type_list htab_lru_type __read_mostly = {
1045         .ops = &htab_lru_ops,
1046         .type = BPF_MAP_TYPE_LRU_HASH,
1047 };
1048
1049 /* Called from eBPF program */
1050 static void *htab_percpu_map_lookup_elem(struct bpf_map *map, void *key)
1051 {
1052         struct htab_elem *l = __htab_map_lookup_elem(map, key);
1053
1054         if (l)
1055                 return this_cpu_ptr(htab_elem_get_ptr(l, map->key_size));
1056         else
1057                 return NULL;
1058 }
1059
1060 static void *htab_lru_percpu_map_lookup_elem(struct bpf_map *map, void *key)
1061 {
1062         struct htab_elem *l = __htab_map_lookup_elem(map, key);
1063
1064         if (l) {
1065                 bpf_lru_node_set_ref(&l->lru_node);
1066                 return this_cpu_ptr(htab_elem_get_ptr(l, map->key_size));
1067         }
1068
1069         return NULL;
1070 }
1071
1072 int bpf_percpu_hash_copy(struct bpf_map *map, void *key, void *value)
1073 {
1074         struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
1075         struct htab_elem *l;
1076         void __percpu *pptr;
1077         int ret = -ENOENT;
1078         int cpu, off = 0;
1079         u32 size;
1080
1081         /* per_cpu areas are zero-filled and bpf programs can only
1082          * access 'value_size' of them, so copying rounded areas
1083          * will not leak any kernel data
1084          */
1085         size = round_up(map->value_size, 8);
1086         rcu_read_lock();
1087         l = __htab_map_lookup_elem(map, key);
1088         if (!l)
1089                 goto out;
1090         if (htab_is_lru(htab))
1091                 bpf_lru_node_set_ref(&l->lru_node);
1092         pptr = htab_elem_get_ptr(l, map->key_size);
1093         for_each_possible_cpu(cpu) {
1094                 bpf_long_memcpy(value + off,
1095                                 per_cpu_ptr(pptr, cpu), size);
1096                 off += size;
1097         }
1098         ret = 0;
1099 out:
1100         rcu_read_unlock();
1101         return ret;
1102 }
1103
1104 int bpf_percpu_hash_update(struct bpf_map *map, void *key, void *value,
1105                            u64 map_flags)
1106 {
1107         struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
1108         int ret;
1109
1110         rcu_read_lock();
1111         if (htab_is_lru(htab))
1112                 ret = __htab_lru_percpu_map_update_elem(map, key, value,
1113                                                         map_flags, true);
1114         else
1115                 ret = __htab_percpu_map_update_elem(map, key, value, map_flags,
1116                                                     true);
1117         rcu_read_unlock();
1118
1119         return ret;
1120 }
1121
1122 static const struct bpf_map_ops htab_percpu_ops = {
1123         .map_alloc = htab_map_alloc,
1124         .map_free = htab_map_free,
1125         .map_get_next_key = htab_map_get_next_key,
1126         .map_lookup_elem = htab_percpu_map_lookup_elem,
1127         .map_update_elem = htab_percpu_map_update_elem,
1128         .map_delete_elem = htab_map_delete_elem,
1129 };
1130
1131 static struct bpf_map_type_list htab_percpu_type __read_mostly = {
1132         .ops = &htab_percpu_ops,
1133         .type = BPF_MAP_TYPE_PERCPU_HASH,
1134 };
1135
1136 static const struct bpf_map_ops htab_lru_percpu_ops = {
1137         .map_alloc = htab_map_alloc,
1138         .map_free = htab_map_free,
1139         .map_get_next_key = htab_map_get_next_key,
1140         .map_lookup_elem = htab_lru_percpu_map_lookup_elem,
1141         .map_update_elem = htab_lru_percpu_map_update_elem,
1142         .map_delete_elem = htab_lru_map_delete_elem,
1143 };
1144
1145 static struct bpf_map_type_list htab_lru_percpu_type __read_mostly = {
1146         .ops = &htab_lru_percpu_ops,
1147         .type = BPF_MAP_TYPE_LRU_PERCPU_HASH,
1148 };
1149
1150 static int __init register_htab_map(void)
1151 {
1152         bpf_register_map_type(&htab_type);
1153         bpf_register_map_type(&htab_percpu_type);
1154         bpf_register_map_type(&htab_lru_type);
1155         bpf_register_map_type(&htab_lru_percpu_type);
1156         return 0;
1157 }
1158 late_initcall(register_htab_map);