mm/list_lru.c

   1 // SPDX-License-Identifier: GPL-2.0-only
   2 /*
   3  * Copyright (c) 2013 Red Hat, Inc. and Parallels Inc. All rights reserved.
   4  * Authors: David Chinner and Glauber Costa
   5  *
   6  * Generic LRU infrastructure
   7  */
   8 #include <linux/kernel.h>
   9 #include <linux/module.h>
  10 #include <linux/mm.h>
  11 #include <linux/list_lru.h>
  12 #include <linux/slab.h>
  13 #include <linux/mutex.h>
  14 #include <linux/memcontrol.h>
  15 #include "slab.h"
  16 #include "internal.h"
  17
  18 #ifdef CONFIG_MEMCG_KMEM
  19 static LIST_HEAD(memcg_list_lrus);
  20 static DEFINE_MUTEX(list_lrus_mutex);
  21
  22 static inline bool list_lru_memcg_aware(struct list_lru *lru)
  23 {
  24         return lru->memcg_aware;
  25 }
  26
  27 static void list_lru_register(struct list_lru *lru)
  28 {
  29         if (!list_lru_memcg_aware(lru))
  30                 return;
  31
  32         mutex_lock(&list_lrus_mutex);
  33         list_add(&lru->list, &memcg_list_lrus);
  34         mutex_unlock(&list_lrus_mutex);
  35 }
  36
  37 static void list_lru_unregister(struct list_lru *lru)
  38 {
  39         if (!list_lru_memcg_aware(lru))
  40                 return;
  41
  42         mutex_lock(&list_lrus_mutex);
  43         list_del(&lru->list);
  44         mutex_unlock(&list_lrus_mutex);
  45 }
  46
  47 static int lru_shrinker_id(struct list_lru *lru)
  48 {
  49         return lru->shrinker_id;
  50 }
  51
  52 static inline struct list_lru_one *
  53 list_lru_from_memcg_idx(struct list_lru *lru, int nid, int idx)
  54 {
  55         if (list_lru_memcg_aware(lru) && idx >= 0) {
  56                 struct list_lru_memcg *mlru = xa_load(&lru->xa, idx);
  57
  58                 return mlru ? &mlru->node[nid] : NULL;
  59         }
  60         return &lru->node[nid].lru;
  61 }
  62 #else
  63 static void list_lru_register(struct list_lru *lru)
  64 {
  65 }
  66
  67 static void list_lru_unregister(struct list_lru *lru)
  68 {
  69 }
  70
  71 static int lru_shrinker_id(struct list_lru *lru)
  72 {
  73         return -1;
  74 }
  75
  76 static inline bool list_lru_memcg_aware(struct list_lru *lru)
  77 {
  78         return false;
  79 }
  80
  81 static inline struct list_lru_one *
  82 list_lru_from_memcg_idx(struct list_lru *lru, int nid, int idx)
  83 {
  84         return &lru->node[nid].lru;
  85 }
  86 #endif /* CONFIG_MEMCG_KMEM */
  87
  88 bool list_lru_add(struct list_lru *lru, struct list_head *item, int nid,
  89                     struct mem_cgroup *memcg)
  90 {
  91         struct list_lru_node *nlru = &lru->node[nid];
  92         struct list_lru_one *l;
  93
  94         spin_lock(&nlru->lock);
  95         if (list_empty(item)) {
  96                 l = list_lru_from_memcg_idx(lru, nid, memcg_kmem_id(memcg));
  97                 list_add_tail(item, &l->list);
  98                 /* Set shrinker bit if the first element was added */
  99                 if (!l->nr_items++)
 100                         set_shrinker_bit(memcg, nid, lru_shrinker_id(lru));
 101                 nlru->nr_items++;
 102                 spin_unlock(&nlru->lock);
 103                 return true;
 104         }
 105         spin_unlock(&nlru->lock);
 106         return false;
 107 }
 108 EXPORT_SYMBOL_GPL(list_lru_add);
 109
 110 bool list_lru_add_obj(struct list_lru *lru, struct list_head *item)
 111 {
 112         int nid = page_to_nid(virt_to_page(item));
 113         struct mem_cgroup *memcg = list_lru_memcg_aware(lru) ?
 114                 mem_cgroup_from_slab_obj(item) : NULL;
 115
 116         return list_lru_add(lru, item, nid, memcg);
 117 }
 118 EXPORT_SYMBOL_GPL(list_lru_add_obj);
 119
 120 bool list_lru_del(struct list_lru *lru, struct list_head *item, int nid,
 121                     struct mem_cgroup *memcg)
 122 {
 123         struct list_lru_node *nlru = &lru->node[nid];
 124         struct list_lru_one *l;
 125
 126         spin_lock(&nlru->lock);
 127         if (!list_empty(item)) {
 128                 l = list_lru_from_memcg_idx(lru, nid, memcg_kmem_id(memcg));
 129                 list_del_init(item);
 130                 l->nr_items--;
 131                 nlru->nr_items--;
 132                 spin_unlock(&nlru->lock);
 133                 return true;
 134         }
 135         spin_unlock(&nlru->lock);
 136         return false;
 137 }
 138 EXPORT_SYMBOL_GPL(list_lru_del);
 139
 140 bool list_lru_del_obj(struct list_lru *lru, struct list_head *item)
 141 {
 142         int nid = page_to_nid(virt_to_page(item));
 143         struct mem_cgroup *memcg = list_lru_memcg_aware(lru) ?
 144                 mem_cgroup_from_slab_obj(item) : NULL;
 145
 146         return list_lru_del(lru, item, nid, memcg);
 147 }
 148 EXPORT_SYMBOL_GPL(list_lru_del_obj);
 149
 150 void list_lru_isolate(struct list_lru_one *list, struct list_head *item)
 151 {
 152         list_del_init(item);
 153         list->nr_items--;
 154 }
 155 EXPORT_SYMBOL_GPL(list_lru_isolate);
 156
 157 void list_lru_isolate_move(struct list_lru_one *list, struct list_head *item,
 158                            struct list_head *head)
 159 {
 160         list_move(item, head);
 161         list->nr_items--;
 162 }
 163 EXPORT_SYMBOL_GPL(list_lru_isolate_move);
 164
 165 void list_lru_putback(struct list_lru *lru, struct list_head *item, int nid,
 166                       struct mem_cgroup *memcg)
 167 {
 168         struct list_lru_one *list =
 169                 list_lru_from_memcg_idx(lru, nid, memcg_kmem_id(memcg));
 170
 171         if (list_empty(item)) {
 172                 list_add_tail(item, &list->list);
 173                 if (!list->nr_items++)
 174                         set_shrinker_bit(memcg, nid, lru_shrinker_id(lru));
 175         }
 176 }
 177 EXPORT_SYMBOL_GPL(list_lru_putback);
 178
 179 unsigned long list_lru_count_one(struct list_lru *lru,
 180                                  int nid, struct mem_cgroup *memcg)
 181 {
 182         struct list_lru_one *l;
 183         long count;
 184
 185         rcu_read_lock();
 186         l = list_lru_from_memcg_idx(lru, nid, memcg_kmem_id(memcg));
 187         count = l ? READ_ONCE(l->nr_items) : 0;
 188         rcu_read_unlock();
 189
 190         if (unlikely(count < 0))
 191                 count = 0;
 192
 193         return count;
 194 }
 195 EXPORT_SYMBOL_GPL(list_lru_count_one);
 196
 197 unsigned long list_lru_count_node(struct list_lru *lru, int nid)
 198 {
 199         struct list_lru_node *nlru;
 200
 201         nlru = &lru->node[nid];
 202         return nlru->nr_items;
 203 }
 204 EXPORT_SYMBOL_GPL(list_lru_count_node);
 205
 206 static unsigned long
 207 __list_lru_walk_one(struct list_lru *lru, int nid, int memcg_idx,
 208                     list_lru_walk_cb isolate, void *cb_arg,
 209                     unsigned long *nr_to_walk)
 210 {
 211         struct list_lru_node *nlru = &lru->node[nid];
 212         struct list_lru_one *l;
 213         struct list_head *item, *n;
 214         unsigned long isolated = 0;
 215
 216 restart:
 217         l = list_lru_from_memcg_idx(lru, nid, memcg_idx);
 218         if (!l)
 219                 goto out;
 220
 221         list_for_each_safe(item, n, &l->list) {
 222                 enum lru_status ret;
 223
 224                 /*
 225                  * decrement nr_to_walk first so that we don't livelock if we
 226                  * get stuck on large numbers of LRU_RETRY items
 227                  */
 228                 if (!*nr_to_walk)
 229                         break;
 230                 --*nr_to_walk;
 231
 232                 ret = isolate(item, l, &nlru->lock, cb_arg);
 233                 switch (ret) {
 234                 case LRU_REMOVED_RETRY:
 235                         assert_spin_locked(&nlru->lock);
 236                         fallthrough;
 237                 case LRU_REMOVED:
 238                         isolated++;
 239                         nlru->nr_items--;
 240                         /*
 241                          * If the lru lock has been dropped, our list
 242                          * traversal is now invalid and so we have to
 243                          * restart from scratch.
 244                          */
 245                         if (ret == LRU_REMOVED_RETRY)
 246                                 goto restart;
 247                         break;
 248                 case LRU_ROTATE:
 249                         list_move_tail(item, &l->list);
 250                         break;
 251                 case LRU_SKIP:
 252                         break;
 253                 case LRU_RETRY:
 254                         /*
 255                          * The lru lock has been dropped, our list traversal is
 256                          * now invalid and so we have to restart from scratch.
 257                          */
 258                         assert_spin_locked(&nlru->lock);
 259                         goto restart;
 260                 default:
 261                         BUG();
 262                 }
 263         }
 264 out:
 265         return isolated;
 266 }
 267
 268 unsigned long
 269 list_lru_walk_one(struct list_lru *lru, int nid, struct mem_cgroup *memcg,
 270                   list_lru_walk_cb isolate, void *cb_arg,
 271                   unsigned long *nr_to_walk)
 272 {
 273         struct list_lru_node *nlru = &lru->node[nid];
 274         unsigned long ret;
 275
 276         spin_lock(&nlru->lock);
 277         ret = __list_lru_walk_one(lru, nid, memcg_kmem_id(memcg), isolate,
 278                                   cb_arg, nr_to_walk);
 279         spin_unlock(&nlru->lock);
 280         return ret;
 281 }
 282 EXPORT_SYMBOL_GPL(list_lru_walk_one);
 283
 284 unsigned long
 285 list_lru_walk_one_irq(struct list_lru *lru, int nid, struct mem_cgroup *memcg,
 286                       list_lru_walk_cb isolate, void *cb_arg,
 287                       unsigned long *nr_to_walk)
 288 {
 289         struct list_lru_node *nlru = &lru->node[nid];
 290         unsigned long ret;
 291
 292         spin_lock_irq(&nlru->lock);
 293         ret = __list_lru_walk_one(lru, nid, memcg_kmem_id(memcg), isolate,
 294                                   cb_arg, nr_to_walk);
 295         spin_unlock_irq(&nlru->lock);
 296         return ret;
 297 }
 298
 299 unsigned long list_lru_walk_node(struct list_lru *lru, int nid,
 300                                  list_lru_walk_cb isolate, void *cb_arg,
 301                                  unsigned long *nr_to_walk)
 302 {
 303         long isolated = 0;
 304
 305         isolated += list_lru_walk_one(lru, nid, NULL, isolate, cb_arg,
 306                                       nr_to_walk);
 307
 308 #ifdef CONFIG_MEMCG_KMEM
 309         if (*nr_to_walk > 0 && list_lru_memcg_aware(lru)) {
 310                 struct list_lru_memcg *mlru;
 311                 unsigned long index;
 312
 313                 xa_for_each(&lru->xa, index, mlru) {
 314                         struct list_lru_node *nlru = &lru->node[nid];
 315
 316                         spin_lock(&nlru->lock);
 317                         isolated += __list_lru_walk_one(lru, nid, index,
 318                                                         isolate, cb_arg,
 319                                                         nr_to_walk);
 320                         spin_unlock(&nlru->lock);
 321
 322                         if (*nr_to_walk <= 0)
 323                                 break;
 324                 }
 325         }
 326 #endif
 327
 328         return isolated;
 329 }
 330 EXPORT_SYMBOL_GPL(list_lru_walk_node);
 331
 332 static void init_one_lru(struct list_lru_one *l)
 333 {
 334         INIT_LIST_HEAD(&l->list);
 335         l->nr_items = 0;
 336 }
 337
 338 #ifdef CONFIG_MEMCG_KMEM
 339 static struct list_lru_memcg *memcg_init_list_lru_one(gfp_t gfp)
 340 {
 341         int nid;
 342         struct list_lru_memcg *mlru;
 343
 344         mlru = kmalloc(struct_size(mlru, node, nr_node_ids), gfp);
 345         if (!mlru)
 346                 return NULL;
 347
 348         for_each_node(nid)
 349                 init_one_lru(&mlru->node[nid]);
 350
 351         return mlru;
 352 }
 353
 354 static void memcg_list_lru_free(struct list_lru *lru, int src_idx)
 355 {
 356         struct list_lru_memcg *mlru = xa_erase_irq(&lru->xa, src_idx);
 357
 358         /*
 359          * The __list_lru_walk_one() can walk the list of this node.
 360          * We need kvfree_rcu() here. And the walking of the list
 361          * is under lru->node[nid]->lock, which can serve as a RCU
 362          * read-side critical section.
 363          */
 364         if (mlru)
 365                 kvfree_rcu(mlru, rcu);
 366 }
 367
 368 static inline void memcg_init_list_lru(struct list_lru *lru, bool memcg_aware)
 369 {
 370         if (memcg_aware)
 371                 xa_init_flags(&lru->xa, XA_FLAGS_LOCK_IRQ);
 372         lru->memcg_aware = memcg_aware;
 373 }
 374
 375 static void memcg_destroy_list_lru(struct list_lru *lru)
 376 {
 377         XA_STATE(xas, &lru->xa, 0);
 378         struct list_lru_memcg *mlru;
 379
 380         if (!list_lru_memcg_aware(lru))
 381                 return;
 382
 383         xas_lock_irq(&xas);
 384         xas_for_each(&xas, mlru, ULONG_MAX) {
 385                 kfree(mlru);
 386                 xas_store(&xas, NULL);
 387         }
 388         xas_unlock_irq(&xas);
 389 }
 390
 391 static void memcg_reparent_list_lru_node(struct list_lru *lru, int nid,
 392                                          int src_idx, struct mem_cgroup *dst_memcg)
 393 {
 394         struct list_lru_node *nlru = &lru->node[nid];
 395         int dst_idx = dst_memcg->kmemcg_id;
 396         struct list_lru_one *src, *dst;
 397
 398         /*
 399          * Since list_lru_{add,del} may be called under an IRQ-safe lock,
 400          * we have to use IRQ-safe primitives here to avoid deadlock.
 401          */
 402         spin_lock_irq(&nlru->lock);
 403
 404         src = list_lru_from_memcg_idx(lru, nid, src_idx);
 405         if (!src)
 406                 goto out;
 407         dst = list_lru_from_memcg_idx(lru, nid, dst_idx);
 408
 409         list_splice_init(&src->list, &dst->list);
 410
 411         if (src->nr_items) {
 412                 dst->nr_items += src->nr_items;
 413                 set_shrinker_bit(dst_memcg, nid, lru_shrinker_id(lru));
 414                 src->nr_items = 0;
 415         }
 416 out:
 417         spin_unlock_irq(&nlru->lock);
 418 }
 419
 420 static void memcg_reparent_list_lru(struct list_lru *lru,
 421                                     int src_idx, struct mem_cgroup *dst_memcg)
 422 {
 423         int i;
 424
 425         for_each_node(i)
 426                 memcg_reparent_list_lru_node(lru, i, src_idx, dst_memcg);
 427
 428         memcg_list_lru_free(lru, src_idx);
 429 }
 430
 431 void memcg_reparent_list_lrus(struct mem_cgroup *memcg, struct mem_cgroup *parent)
 432 {
 433         struct cgroup_subsys_state *css;
 434         struct list_lru *lru;
 435         int src_idx = memcg->kmemcg_id;
 436
 437         /*
 438          * Change kmemcg_id of this cgroup and all its descendants to the
 439          * parent's id, and then move all entries from this cgroup's list_lrus
 440          * to ones of the parent.
 441          *
 442          * After we have finished, all list_lrus corresponding to this cgroup
 443          * are guaranteed to remain empty. So we can safely free this cgroup's
 444          * list lrus in memcg_list_lru_free().
 445          *
 446          * Changing ->kmemcg_id to the parent can prevent memcg_list_lru_alloc()
 447          * from allocating list lrus for this cgroup after memcg_list_lru_free()
 448          * call.
 449          */
 450         rcu_read_lock();
 451         css_for_each_descendant_pre(css, &memcg->css) {
 452                 struct mem_cgroup *child;
 453
 454                 child = mem_cgroup_from_css(css);
 455                 WRITE_ONCE(child->kmemcg_id, parent->kmemcg_id);
 456         }
 457         rcu_read_unlock();
 458
 459         mutex_lock(&list_lrus_mutex);
 460         list_for_each_entry(lru, &memcg_list_lrus, list)
 461                 memcg_reparent_list_lru(lru, src_idx, parent);
 462         mutex_unlock(&list_lrus_mutex);
 463 }
 464
 465 static inline bool memcg_list_lru_allocated(struct mem_cgroup *memcg,
 466                                             struct list_lru *lru)
 467 {
 468         int idx = memcg->kmemcg_id;
 469
 470         return idx < 0 || xa_load(&lru->xa, idx);
 471 }
 472
 473 int memcg_list_lru_alloc(struct mem_cgroup *memcg, struct list_lru *lru,
 474                          gfp_t gfp)
 475 {
 476         int i;
 477         unsigned long flags;
 478         struct list_lru_memcg_table {
 479                 struct list_lru_memcg *mlru;
 480                 struct mem_cgroup *memcg;
 481         } *table;
 482         XA_STATE(xas, &lru->xa, 0);
 483
 484         if (!list_lru_memcg_aware(lru) || memcg_list_lru_allocated(memcg, lru))
 485                 return 0;
 486
 487         gfp &= GFP_RECLAIM_MASK;
 488         table = kmalloc_array(memcg->css.cgroup->level, sizeof(*table), gfp);
 489         if (!table)
 490                 return -ENOMEM;
 491
 492         /*
 493          * Because the list_lru can be reparented to the parent cgroup's
 494          * list_lru, we should make sure that this cgroup and all its
 495          * ancestors have allocated list_lru_memcg.
 496          */
 497         for (i = 0; memcg; memcg = parent_mem_cgroup(memcg), i++) {
 498                 if (memcg_list_lru_allocated(memcg, lru))
 499                         break;
 500
 501                 table[i].memcg = memcg;
 502                 table[i].mlru = memcg_init_list_lru_one(gfp);
 503                 if (!table[i].mlru) {
 504                         while (i--)
 505                                 kfree(table[i].mlru);
 506                         kfree(table);
 507                         return -ENOMEM;
 508                 }
 509         }
 510
 511         xas_lock_irqsave(&xas, flags);
 512         while (i--) {
 513                 int index = READ_ONCE(table[i].memcg->kmemcg_id);
 514                 struct list_lru_memcg *mlru = table[i].mlru;
 515
 516                 xas_set(&xas, index);
 517 retry:
 518                 if (unlikely(index < 0 || xas_error(&xas) || xas_load(&xas))) {
 519                         kfree(mlru);
 520                 } else {
 521                         xas_store(&xas, mlru);
 522                         if (xas_error(&xas) == -ENOMEM) {
 523                                 xas_unlock_irqrestore(&xas, flags);
 524                                 if (xas_nomem(&xas, gfp))
 525                                         xas_set_err(&xas, 0);
 526                                 xas_lock_irqsave(&xas, flags);
 527                                 /*
 528                                  * The xas lock has been released, this memcg
 529                                  * can be reparented before us. So reload
 530                                  * memcg id. More details see the comments
 531                                  * in memcg_reparent_list_lrus().
 532                                  */
 533                                 index = READ_ONCE(table[i].memcg->kmemcg_id);
 534                                 if (index < 0)
 535                                         xas_set_err(&xas, 0);
 536                                 else if (!xas_error(&xas) && index != xas.xa_index)
 537                                         xas_set(&xas, index);
 538                                 goto retry;
 539                         }
 540                 }
 541         }
 542         /* xas_nomem() is used to free memory instead of memory allocation. */
 543         if (xas.xa_alloc)
 544                 xas_nomem(&xas, gfp);
 545         xas_unlock_irqrestore(&xas, flags);
 546         kfree(table);
 547
 548         return xas_error(&xas);
 549 }
 550 #else
 551 static inline void memcg_init_list_lru(struct list_lru *lru, bool memcg_aware)
 552 {
 553 }
 554
 555 static void memcg_destroy_list_lru(struct list_lru *lru)
 556 {
 557 }
 558 #endif /* CONFIG_MEMCG_KMEM */
 559
 560 int __list_lru_init(struct list_lru *lru, bool memcg_aware,
 561                     struct lock_class_key *key, struct shrinker *shrinker)
 562 {
 563         int i;
 564
 565 #ifdef CONFIG_MEMCG_KMEM
 566         if (shrinker)
 567                 lru->shrinker_id = shrinker->id;
 568         else
 569                 lru->shrinker_id = -1;
 570 #endif
 571
 572         lru->node = kcalloc(nr_node_ids, sizeof(*lru->node), GFP_KERNEL);
 573         if (!lru->node)
 574                 return -ENOMEM;
 575
 576         for_each_node(i) {
 577                 spin_lock_init(&lru->node[i].lock);
 578                 if (key)
 579                         lockdep_set_class(&lru->node[i].lock, key);
 580                 init_one_lru(&lru->node[i].lru);
 581         }
 582
 583         memcg_init_list_lru(lru, memcg_aware);
 584         list_lru_register(lru);
 585
 586         return 0;
 587 }
 588 EXPORT_SYMBOL_GPL(__list_lru_init);
 589
 590 void list_lru_destroy(struct list_lru *lru)
 591 {
 592         /* Already destroyed or not yet initialized? */
 593         if (!lru->node)
 594                 return;
 595
 596         list_lru_unregister(lru);
 597
 598         memcg_destroy_list_lru(lru);
 599         kfree(lru->node);
 600         lru->node = NULL;
 601
 602 #ifdef CONFIG_MEMCG_KMEM
 603         lru->shrinker_id = -1;
 604 #endif
 605 }
 606 EXPORT_SYMBOL_GPL(list_lru_destroy);