include/linux/slab.h

   1 /* SPDX-License-Identifier: GPL-2.0 */
   2 /*
   3  * Written by Mark Hemment, 1996 (markhe@nextd.demon.co.uk).
   4  *
   5  * (C) SGI 2006, Christoph Lameter
   6  *      Cleaned up and restructured to ease the addition of alternative
   7  *      implementations of SLAB allocators.
   8  * (C) Linux Foundation 2008-2013
   9  *      Unified interface for all slab allocators
  10  */
  11
  12 #ifndef _LINUX_SLAB_H
  13 #define _LINUX_SLAB_H
  14
  15 #include <linux/gfp.h>
  16 #include <linux/types.h>
  17 #include <linux/workqueue.h>
  18
  19
  20 /*
  21  * Flags to pass to kmem_cache_create().
  22  * The ones marked DEBUG are only valid if CONFIG_DEBUG_SLAB is set.
  23  */
  24 /* DEBUG: Perform (expensive) checks on alloc/free */
  25 #define SLAB_CONSISTENCY_CHECKS ((slab_flags_t __force)0x00000100U)
  26 /* DEBUG: Red zone objs in a cache */
  27 #define SLAB_RED_ZONE           ((slab_flags_t __force)0x00000400U)
  28 /* DEBUG: Poison objects */
  29 #define SLAB_POISON             ((slab_flags_t __force)0x00000800U)
  30 /* Align objs on cache lines */
  31 #define SLAB_HWCACHE_ALIGN      ((slab_flags_t __force)0x00002000U)
  32 /* Use GFP_DMA memory */
  33 #define SLAB_CACHE_DMA          ((slab_flags_t __force)0x00004000U)
  34 /* DEBUG: Store the last owner for bug hunting */
  35 #define SLAB_STORE_USER         ((slab_flags_t __force)0x00010000U)
  36 /* Panic if kmem_cache_create() fails */
  37 #define SLAB_PANIC              ((slab_flags_t __force)0x00040000U)
  38 /*
  39  * SLAB_TYPESAFE_BY_RCU - **WARNING** READ THIS!
  40  *
  41  * This delays freeing the SLAB page by a grace period, it does _NOT_
  42  * delay object freeing. This means that if you do kmem_cache_free()
  43  * that memory location is free to be reused at any time. Thus it may
  44  * be possible to see another object there in the same RCU grace period.
  45  *
  46  * This feature only ensures the memory location backing the object
  47  * stays valid, the trick to using this is relying on an independent
  48  * object validation pass. Something like:
  49  *
  50  *  rcu_read_lock()
  51  * again:
  52  *  obj = lockless_lookup(key);
  53  *  if (obj) {
  54  *    if (!try_get_ref(obj)) // might fail for free objects
  55  *      goto again;
  56  *
  57  *    if (obj->key != key) { // not the object we expected
  58  *      put_ref(obj);
  59  *      goto again;
  60  *    }
  61  *  }
  62  *  rcu_read_unlock();
  63  *
  64  * This is useful if we need to approach a kernel structure obliquely,
  65  * from its address obtained without the usual locking. We can lock
  66  * the structure to stabilize it and check it's still at the given address,
  67  * only if we can be sure that the memory has not been meanwhile reused
  68  * for some other kind of object (which our subsystem's lock might corrupt).
  69  *
  70  * rcu_read_lock before reading the address, then rcu_read_unlock after
  71  * taking the spinlock within the structure expected at that address.
  72  *
  73  * Note that SLAB_TYPESAFE_BY_RCU was originally named SLAB_DESTROY_BY_RCU.
  74  */
  75 /* Defer freeing slabs to RCU */
  76 #define SLAB_TYPESAFE_BY_RCU    ((slab_flags_t __force)0x00080000U)
  77 /* Spread some memory over cpuset */
  78 #define SLAB_MEM_SPREAD         ((slab_flags_t __force)0x00100000U)
  79 /* Trace allocations and frees */
  80 #define SLAB_TRACE              ((slab_flags_t __force)0x00200000U)
  81
  82 /* Flag to prevent checks on free */
  83 #ifdef CONFIG_DEBUG_OBJECTS
  84 # define SLAB_DEBUG_OBJECTS     ((slab_flags_t __force)0x00400000U)
  85 #else
  86 # define SLAB_DEBUG_OBJECTS     0
  87 #endif
  88
  89 /* Avoid kmemleak tracing */
  90 #define SLAB_NOLEAKTRACE        ((slab_flags_t __force)0x00800000U)
  91
  92 /* Fault injection mark */
  93 #ifdef CONFIG_FAILSLAB
  94 # define SLAB_FAILSLAB          ((slab_flags_t __force)0x02000000U)
  95 #else
  96 # define SLAB_FAILSLAB          0
  97 #endif
  98 /* Account to memcg */
  99 #if defined(CONFIG_MEMCG) && !defined(CONFIG_SLOB)
 100 # define SLAB_ACCOUNT           ((slab_flags_t __force)0x04000000U)
 101 #else
 102 # define SLAB_ACCOUNT           0
 103 #endif
 104
 105 #ifdef CONFIG_KASAN
 106 #define SLAB_KASAN              ((slab_flags_t __force)0x08000000U)
 107 #else
 108 #define SLAB_KASAN              0
 109 #endif
 110
 111 /* The following flags affect the page allocator grouping pages by mobility */
 112 /* Objects are reclaimable */
 113 #define SLAB_RECLAIM_ACCOUNT    ((slab_flags_t __force)0x00020000U)
 114 #define SLAB_TEMPORARY          SLAB_RECLAIM_ACCOUNT    /* Objects are short-lived */
 115 /*
 116  * ZERO_SIZE_PTR will be returned for zero sized kmalloc requests.
 117  *
 118  * Dereferencing ZERO_SIZE_PTR will lead to a distinct access fault.
 119  *
 120  * ZERO_SIZE_PTR can be passed to kfree though in the same way that NULL can.
 121  * Both make kfree a no-op.
 122  */
 123 #define ZERO_SIZE_PTR ((void *)16)
 124
 125 #define ZERO_OR_NULL_PTR(x) ((unsigned long)(x) <= \
 126                                 (unsigned long)ZERO_SIZE_PTR)
 127
 128 #include <linux/kmemleak.h>
 129 #include <linux/kasan.h>
 130
 131 struct mem_cgroup;
 132 /*
 133  * struct kmem_cache related prototypes
 134  */
 135 void __init kmem_cache_init(void);
 136 bool slab_is_available(void);
 137
 138 extern bool usercopy_fallback;
 139
 140 struct kmem_cache *kmem_cache_create(const char *name, size_t size,
 141                         size_t align, slab_flags_t flags,
 142                         void (*ctor)(void *));
 143 struct kmem_cache *kmem_cache_create_usercopy(const char *name,
 144                         size_t size, size_t align, slab_flags_t flags,
 145                         size_t useroffset, size_t usersize,
 146                         void (*ctor)(void *));
 147 void kmem_cache_destroy(struct kmem_cache *);
 148 int kmem_cache_shrink(struct kmem_cache *);
 149
 150 void memcg_create_kmem_cache(struct mem_cgroup *, struct kmem_cache *);
 151 void memcg_deactivate_kmem_caches(struct mem_cgroup *);
 152 void memcg_destroy_kmem_caches(struct mem_cgroup *);
 153
 154 /*
 155  * Please use this macro to create slab caches. Simply specify the
 156  * name of the structure and maybe some flags that are listed above.
 157  *
 158  * The alignment of the struct determines object alignment. If you
 159  * f.e. add ____cacheline_aligned_in_smp to the struct declaration
 160  * then the objects will be properly aligned in SMP configurations.
 161  */
 162 #define KMEM_CACHE(__struct, __flags)                                   \
 163                 kmem_cache_create(#__struct, sizeof(struct __struct),   \
 164                         __alignof__(struct __struct), (__flags), NULL)
 165
 166 /*
 167  * To whitelist a single field for copying to/from usercopy, use this
 168  * macro instead for KMEM_CACHE() above.
 169  */
 170 #define KMEM_CACHE_USERCOPY(__struct, __flags, __field)                 \
 171                 kmem_cache_create_usercopy(#__struct,                   \
 172                         sizeof(struct __struct),                        \
 173                         __alignof__(struct __struct), (__flags),        \
 174                         offsetof(struct __struct, __field),             \
 175                         sizeof_field(struct __struct, __field), NULL)
 176
 177 /*
 178  * Common kmalloc functions provided by all allocators
 179  */
 180 void * __must_check __krealloc(const void *, size_t, gfp_t);
 181 void * __must_check krealloc(const void *, size_t, gfp_t);
 182 void kfree(const void *);
 183 void kzfree(const void *);
 184 size_t ksize(const void *);
 185
 186 #ifdef CONFIG_HAVE_HARDENED_USERCOPY_ALLOCATOR
 187 void __check_heap_object(const void *ptr, unsigned long n, struct page *page,
 188                         bool to_user);
 189 #else
 190 static inline void __check_heap_object(const void *ptr, unsigned long n,
 191                                        struct page *page, bool to_user) { }
 192 #endif
 193
 194 /*
 195  * Some archs want to perform DMA into kmalloc caches and need a guaranteed
 196  * alignment larger than the alignment of a 64-bit integer.
 197  * Setting ARCH_KMALLOC_MINALIGN in arch headers allows that.
 198  */
 199 #if defined(ARCH_DMA_MINALIGN) && ARCH_DMA_MINALIGN > 8
 200 #define ARCH_KMALLOC_MINALIGN ARCH_DMA_MINALIGN
 201 #define KMALLOC_MIN_SIZE ARCH_DMA_MINALIGN
 202 #define KMALLOC_SHIFT_LOW ilog2(ARCH_DMA_MINALIGN)
 203 #else
 204 #define ARCH_KMALLOC_MINALIGN __alignof__(unsigned long long)
 205 #endif
 206
 207 /*
 208  * Setting ARCH_SLAB_MINALIGN in arch headers allows a different alignment.
 209  * Intended for arches that get misalignment faults even for 64 bit integer
 210  * aligned buffers.
 211  */
 212 #ifndef ARCH_SLAB_MINALIGN
 213 #define ARCH_SLAB_MINALIGN __alignof__(unsigned long long)
 214 #endif
 215
 216 /*
 217  * kmalloc and friends return ARCH_KMALLOC_MINALIGN aligned
 218  * pointers. kmem_cache_alloc and friends return ARCH_SLAB_MINALIGN
 219  * aligned pointers.
 220  */
 221 #define __assume_kmalloc_alignment __assume_aligned(ARCH_KMALLOC_MINALIGN)
 222 #define __assume_slab_alignment __assume_aligned(ARCH_SLAB_MINALIGN)
 223 #define __assume_page_alignment __assume_aligned(PAGE_SIZE)
 224
 225 /*
 226  * Kmalloc array related definitions
 227  */
 228
 229 #ifdef CONFIG_SLAB
 230 /*
 231  * The largest kmalloc size supported by the SLAB allocators is
 232  * 32 megabyte (2^25) or the maximum allocatable page order if that is
 233  * less than 32 MB.
 234  *
 235  * WARNING: Its not easy to increase this value since the allocators have
 236  * to do various tricks to work around compiler limitations in order to
 237  * ensure proper constant folding.
 238  */
 239 #define KMALLOC_SHIFT_HIGH      ((MAX_ORDER + PAGE_SHIFT - 1) <= 25 ? \
 240                                 (MAX_ORDER + PAGE_SHIFT - 1) : 25)
 241 #define KMALLOC_SHIFT_MAX       KMALLOC_SHIFT_HIGH
 242 #ifndef KMALLOC_SHIFT_LOW
 243 #define KMALLOC_SHIFT_LOW       5
 244 #endif
 245 #endif
 246
 247 #ifdef CONFIG_SLUB
 248 /*
 249  * SLUB directly allocates requests fitting in to an order-1 page
 250  * (PAGE_SIZE*2).  Larger requests are passed to the page allocator.
 251  */
 252 #define KMALLOC_SHIFT_HIGH      (PAGE_SHIFT + 1)
 253 #define KMALLOC_SHIFT_MAX       (MAX_ORDER + PAGE_SHIFT - 1)
 254 #ifndef KMALLOC_SHIFT_LOW
 255 #define KMALLOC_SHIFT_LOW       3
 256 #endif
 257 #endif
 258
 259 #ifdef CONFIG_SLOB
 260 /*
 261  * SLOB passes all requests larger than one page to the page allocator.
 262  * No kmalloc array is necessary since objects of different sizes can
 263  * be allocated from the same page.
 264  */
 265 #define KMALLOC_SHIFT_HIGH      PAGE_SHIFT
 266 #define KMALLOC_SHIFT_MAX       (MAX_ORDER + PAGE_SHIFT - 1)
 267 #ifndef KMALLOC_SHIFT_LOW
 268 #define KMALLOC_SHIFT_LOW       3
 269 #endif
 270 #endif
 271
 272 /* Maximum allocatable size */
 273 #define KMALLOC_MAX_SIZE        (1UL << KMALLOC_SHIFT_MAX)
 274 /* Maximum size for which we actually use a slab cache */
 275 #define KMALLOC_MAX_CACHE_SIZE  (1UL << KMALLOC_SHIFT_HIGH)
 276 /* Maximum order allocatable via the slab allocagtor */
 277 #define KMALLOC_MAX_ORDER       (KMALLOC_SHIFT_MAX - PAGE_SHIFT)
 278
 279 /*
 280  * Kmalloc subsystem.
 281  */
 282 #ifndef KMALLOC_MIN_SIZE
 283 #define KMALLOC_MIN_SIZE (1 << KMALLOC_SHIFT_LOW)
 284 #endif
 285
 286 /*
 287  * This restriction comes from byte sized index implementation.
 288  * Page size is normally 2^12 bytes and, in this case, if we want to use
 289  * byte sized index which can represent 2^8 entries, the size of the object
 290  * should be equal or greater to 2^12 / 2^8 = 2^4 = 16.
 291  * If minimum size of kmalloc is less than 16, we use it as minimum object
 292  * size and give up to use byte sized index.
 293  */
 294 #define SLAB_OBJ_MIN_SIZE      (KMALLOC_MIN_SIZE < 16 ? \
 295                                (KMALLOC_MIN_SIZE) : 16)
 296
 297 #ifndef CONFIG_SLOB
 298 extern struct kmem_cache *kmalloc_caches[KMALLOC_SHIFT_HIGH + 1];
 299 #ifdef CONFIG_ZONE_DMA
 300 extern struct kmem_cache *kmalloc_dma_caches[KMALLOC_SHIFT_HIGH + 1];
 301 #endif
 302
 303 /*
 304  * Figure out which kmalloc slab an allocation of a certain size
 305  * belongs to.
 306  * 0 = zero alloc
 307  * 1 =  65 .. 96 bytes
 308  * 2 = 129 .. 192 bytes
 309  * n = 2^(n-1)+1 .. 2^n
 310  */
 311 static __always_inline int kmalloc_index(size_t size)
 312 {
 313         if (!size)
 314                 return 0;
 315
 316         if (size <= KMALLOC_MIN_SIZE)
 317                 return KMALLOC_SHIFT_LOW;
 318
 319         if (KMALLOC_MIN_SIZE <= 32 && size > 64 && size <= 96)
 320                 return 1;
 321         if (KMALLOC_MIN_SIZE <= 64 && size > 128 && size <= 192)
 322                 return 2;
 323         if (size <=          8) return 3;
 324         if (size <=         16) return 4;
 325         if (size <=         32) return 5;
 326         if (size <=         64) return 6;
 327         if (size <=        128) return 7;
 328         if (size <=        256) return 8;
 329         if (size <=        512) return 9;
 330         if (size <=       1024) return 10;
 331         if (size <=   2 * 1024) return 11;
 332         if (size <=   4 * 1024) return 12;
 333         if (size <=   8 * 1024) return 13;
 334         if (size <=  16 * 1024) return 14;
 335         if (size <=  32 * 1024) return 15;
 336         if (size <=  64 * 1024) return 16;
 337         if (size <= 128 * 1024) return 17;
 338         if (size <= 256 * 1024) return 18;
 339         if (size <= 512 * 1024) return 19;
 340         if (size <= 1024 * 1024) return 20;
 341         if (size <=  2 * 1024 * 1024) return 21;
 342         if (size <=  4 * 1024 * 1024) return 22;
 343         if (size <=  8 * 1024 * 1024) return 23;
 344         if (size <=  16 * 1024 * 1024) return 24;
 345         if (size <=  32 * 1024 * 1024) return 25;
 346         if (size <=  64 * 1024 * 1024) return 26;
 347         BUG();
 348
 349         /* Will never be reached. Needed because the compiler may complain */
 350         return -1;
 351 }
 352 #endif /* !CONFIG_SLOB */
 353
 354 void *__kmalloc(size_t size, gfp_t flags) __assume_kmalloc_alignment __malloc;
 355 void *kmem_cache_alloc(struct kmem_cache *, gfp_t flags) __assume_slab_alignment __malloc;
 356 void kmem_cache_free(struct kmem_cache *, void *);
 357
 358 /*
 359  * Bulk allocation and freeing operations. These are accelerated in an
 360  * allocator specific way to avoid taking locks repeatedly or building
 361  * metadata structures unnecessarily.
 362  *
 363  * Note that interrupts must be enabled when calling these functions.
 364  */
 365 void kmem_cache_free_bulk(struct kmem_cache *, size_t, void **);
 366 int kmem_cache_alloc_bulk(struct kmem_cache *, gfp_t, size_t, void **);
 367
 368 /*
 369  * Caller must not use kfree_bulk() on memory not originally allocated
 370  * by kmalloc(), because the SLOB allocator cannot handle this.
 371  */
 372 static __always_inline void kfree_bulk(size_t size, void **p)
 373 {
 374         kmem_cache_free_bulk(NULL, size, p);
 375 }
 376
 377 #ifdef CONFIG_NUMA
 378 void *__kmalloc_node(size_t size, gfp_t flags, int node) __assume_kmalloc_alignment __malloc;
 379 void *kmem_cache_alloc_node(struct kmem_cache *, gfp_t flags, int node) __assume_slab_alignment __malloc;
 380 #else
 381 static __always_inline void *__kmalloc_node(size_t size, gfp_t flags, int node)
 382 {
 383         return __kmalloc(size, flags);
 384 }
 385
 386 static __always_inline void *kmem_cache_alloc_node(struct kmem_cache *s, gfp_t flags, int node)
 387 {
 388         return kmem_cache_alloc(s, flags);
 389 }
 390 #endif
 391
 392 #ifdef CONFIG_TRACING
 393 extern void *kmem_cache_alloc_trace(struct kmem_cache *, gfp_t, size_t) __assume_slab_alignment __malloc;
 394
 395 #ifdef CONFIG_NUMA
 396 extern void *kmem_cache_alloc_node_trace(struct kmem_cache *s,
 397                                            gfp_t gfpflags,
 398                                            int node, size_t size) __assume_slab_alignment __malloc;
 399 #else
 400 static __always_inline void *
 401 kmem_cache_alloc_node_trace(struct kmem_cache *s,
 402                               gfp_t gfpflags,
 403                               int node, size_t size)
 404 {
 405         return kmem_cache_alloc_trace(s, gfpflags, size);
 406 }
 407 #endif /* CONFIG_NUMA */
 408
 409 #else /* CONFIG_TRACING */
 410 static __always_inline void *kmem_cache_alloc_trace(struct kmem_cache *s,
 411                 gfp_t flags, size_t size)
 412 {
 413         void *ret = kmem_cache_alloc(s, flags);
 414
 415         kasan_kmalloc(s, ret, size, flags);
 416         return ret;
 417 }
 418
 419 static __always_inline void *
 420 kmem_cache_alloc_node_trace(struct kmem_cache *s,
 421                               gfp_t gfpflags,
 422                               int node, size_t size)
 423 {
 424         void *ret = kmem_cache_alloc_node(s, gfpflags, node);
 425
 426         kasan_kmalloc(s, ret, size, gfpflags);
 427         return ret;
 428 }
 429 #endif /* CONFIG_TRACING */
 430
 431 extern void *kmalloc_order(size_t size, gfp_t flags, unsigned int order) __assume_page_alignment __malloc;
 432
 433 #ifdef CONFIG_TRACING
 434 extern void *kmalloc_order_trace(size_t size, gfp_t flags, unsigned int order) __assume_page_alignment __malloc;
 435 #else
 436 static __always_inline void *
 437 kmalloc_order_trace(size_t size, gfp_t flags, unsigned int order)
 438 {
 439         return kmalloc_order(size, flags, order);
 440 }
 441 #endif
 442
 443 static __always_inline void *kmalloc_large(size_t size, gfp_t flags)
 444 {
 445         unsigned int order = get_order(size);
 446         return kmalloc_order_trace(size, flags, order);
 447 }
 448
 449 /**
 450  * kmalloc - allocate memory
 451  * @size: how many bytes of memory are required.
 452  * @flags: the type of memory to allocate.
 453  *
 454  * kmalloc is the normal method of allocating memory
 455  * for objects smaller than page size in the kernel.
 456  *
 457  * The @flags argument may be one of:
 458  *
 459  * %GFP_USER - Allocate memory on behalf of user.  May sleep.
 460  *
 461  * %GFP_KERNEL - Allocate normal kernel ram.  May sleep.
 462  *
 463  * %GFP_ATOMIC - Allocation will not sleep.  May use emergency pools.
 464  *   For example, use this inside interrupt handlers.
 465  *
 466  * %GFP_HIGHUSER - Allocate pages from high memory.
 467  *
 468  * %GFP_NOIO - Do not do any I/O at all while trying to get memory.
 469  *
 470  * %GFP_NOFS - Do not make any fs calls while trying to get memory.
 471  *
 472  * %GFP_NOWAIT - Allocation will not sleep.
 473  *
 474  * %__GFP_THISNODE - Allocate node-local memory only.
 475  *
 476  * %GFP_DMA - Allocation suitable for DMA.
 477  *   Should only be used for kmalloc() caches. Otherwise, use a
 478  *   slab created with SLAB_DMA.
 479  *
 480  * Also it is possible to set different flags by OR'ing
 481  * in one or more of the following additional @flags:
 482  *
 483  * %__GFP_HIGH - This allocation has high priority and may use emergency pools.
 484  *
 485  * %__GFP_NOFAIL - Indicate that this allocation is in no way allowed to fail
 486  *   (think twice before using).
 487  *
 488  * %__GFP_NORETRY - If memory is not immediately available,
 489  *   then give up at once.
 490  *
 491  * %__GFP_NOWARN - If allocation fails, don't issue any warnings.
 492  *
 493  * %__GFP_RETRY_MAYFAIL - Try really hard to succeed the allocation but fail
 494  *   eventually.
 495  *
 496  * There are other flags available as well, but these are not intended
 497  * for general use, and so are not documented here. For a full list of
 498  * potential flags, always refer to linux/gfp.h.
 499  */
 500 static __always_inline void *kmalloc(size_t size, gfp_t flags)
 501 {
 502         if (__builtin_constant_p(size)) {
 503                 if (size > KMALLOC_MAX_CACHE_SIZE)
 504                         return kmalloc_large(size, flags);
 505 #ifndef CONFIG_SLOB
 506                 if (!(flags & GFP_DMA)) {
 507                         int index = kmalloc_index(size);
 508
 509                         if (!index)
 510                                 return ZERO_SIZE_PTR;
 511
 512                         return kmem_cache_alloc_trace(kmalloc_caches[index],
 513                                         flags, size);
 514                 }
 515 #endif
 516         }
 517         return __kmalloc(size, flags);
 518 }
 519
 520 /*
 521  * Determine size used for the nth kmalloc cache.
 522  * return size or 0 if a kmalloc cache for that
 523  * size does not exist
 524  */
 525 static __always_inline int kmalloc_size(int n)
 526 {
 527 #ifndef CONFIG_SLOB
 528         if (n > 2)
 529                 return 1 << n;
 530
 531         if (n == 1 && KMALLOC_MIN_SIZE <= 32)
 532                 return 96;
 533
 534         if (n == 2 && KMALLOC_MIN_SIZE <= 64)
 535                 return 192;
 536 #endif
 537         return 0;
 538 }
 539
 540 static __always_inline void *kmalloc_node(size_t size, gfp_t flags, int node)
 541 {
 542 #ifndef CONFIG_SLOB
 543         if (__builtin_constant_p(size) &&
 544                 size <= KMALLOC_MAX_CACHE_SIZE && !(flags & GFP_DMA)) {
 545                 int i = kmalloc_index(size);
 546
 547                 if (!i)
 548                         return ZERO_SIZE_PTR;
 549
 550                 return kmem_cache_alloc_node_trace(kmalloc_caches[i],
 551                                                 flags, node, size);
 552         }
 553 #endif
 554         return __kmalloc_node(size, flags, node);
 555 }
 556
 557 struct memcg_cache_array {
 558         struct rcu_head rcu;
 559         struct kmem_cache *entries[0];
 560 };
 561
 562 /*
 563  * This is the main placeholder for memcg-related information in kmem caches.
 564  * Both the root cache and the child caches will have it. For the root cache,
 565  * this will hold a dynamically allocated array large enough to hold
 566  * information about the currently limited memcgs in the system. To allow the
 567  * array to be accessed without taking any locks, on relocation we free the old
 568  * version only after a grace period.
 569  *
 570  * Root and child caches hold different metadata.
 571  *
 572  * @root_cache: Common to root and child caches.  NULL for root, pointer to
 573  *              the root cache for children.
 574  *
 575  * The following fields are specific to root caches.
 576  *
 577  * @memcg_caches: kmemcg ID indexed table of child caches.  This table is
 578  *              used to index child cachces during allocation and cleared
 579  *              early during shutdown.
 580  *
 581  * @root_caches_node: List node for slab_root_caches list.
 582  *
 583  * @children:   List of all child caches.  While the child caches are also
 584  *              reachable through @memcg_caches, a child cache remains on
 585  *              this list until it is actually destroyed.
 586  *
 587  * The following fields are specific to child caches.
 588  *
 589  * @memcg:      Pointer to the memcg this cache belongs to.
 590  *
 591  * @children_node: List node for @root_cache->children list.
 592  *
 593  * @kmem_caches_node: List node for @memcg->kmem_caches list.
 594  */
 595 struct memcg_cache_params {
 596         struct kmem_cache *root_cache;
 597         union {
 598                 struct {
 599                         struct memcg_cache_array __rcu *memcg_caches;
 600                         struct list_head __root_caches_node;
 601                         struct list_head children;
 602                 };
 603                 struct {
 604                         struct mem_cgroup *memcg;
 605                         struct list_head children_node;
 606                         struct list_head kmem_caches_node;
 607
 608                         void (*deact_fn)(struct kmem_cache *);
 609                         union {
 610                                 struct rcu_head deact_rcu_head;
 611                                 struct work_struct deact_work;
 612                         };
 613                 };
 614         };
 615 };
 616
 617 int memcg_update_all_caches(int num_memcgs);
 618
 619 /**
 620  * kmalloc_array - allocate memory for an array.
 621  * @n: number of elements.
 622  * @size: element size.
 623  * @flags: the type of memory to allocate (see kmalloc).
 624  */
 625 static inline void *kmalloc_array(size_t n, size_t size, gfp_t flags)
 626 {
 627         if (size != 0 && n > SIZE_MAX / size)
 628                 return NULL;
 629         if (__builtin_constant_p(n) && __builtin_constant_p(size))
 630                 return kmalloc(n * size, flags);
 631         return __kmalloc(n * size, flags);
 632 }
 633
 634 /**
 635  * kcalloc - allocate memory for an array. The memory is set to zero.
 636  * @n: number of elements.
 637  * @size: element size.
 638  * @flags: the type of memory to allocate (see kmalloc).
 639  */
 640 static inline void *kcalloc(size_t n, size_t size, gfp_t flags)
 641 {
 642         return kmalloc_array(n, size, flags | __GFP_ZERO);
 643 }
 644
 645 /*
 646  * kmalloc_track_caller is a special version of kmalloc that records the
 647  * calling function of the routine calling it for slab leak tracking instead
 648  * of just the calling function (confusing, eh?).
 649  * It's useful when the call to kmalloc comes from a widely-used standard
 650  * allocator where we care about the real place the memory allocation
 651  * request comes from.
 652  */
 653 extern void *__kmalloc_track_caller(size_t, gfp_t, unsigned long);
 654 #define kmalloc_track_caller(size, flags) \
 655         __kmalloc_track_caller(size, flags, _RET_IP_)
 656
 657 static inline void *kmalloc_array_node(size_t n, size_t size, gfp_t flags,
 658                                        int node)
 659 {
 660         if (size != 0 && n > SIZE_MAX / size)
 661                 return NULL;
 662         if (__builtin_constant_p(n) && __builtin_constant_p(size))
 663                 return kmalloc_node(n * size, flags, node);
 664         return __kmalloc_node(n * size, flags, node);
 665 }
 666
 667 static inline void *kcalloc_node(size_t n, size_t size, gfp_t flags, int node)
 668 {
 669         return kmalloc_array_node(n, size, flags | __GFP_ZERO, node);
 670 }
 671
 672
 673 #ifdef CONFIG_NUMA
 674 extern void *__kmalloc_node_track_caller(size_t, gfp_t, int, unsigned long);
 675 #define kmalloc_node_track_caller(size, flags, node) \
 676         __kmalloc_node_track_caller(size, flags, node, \
 677                         _RET_IP_)
 678
 679 #else /* CONFIG_NUMA */
 680
 681 #define kmalloc_node_track_caller(size, flags, node) \
 682         kmalloc_track_caller(size, flags)
 683
 684 #endif /* CONFIG_NUMA */
 685
 686 /*
 687  * Shortcuts
 688  */
 689 static inline void *kmem_cache_zalloc(struct kmem_cache *k, gfp_t flags)
 690 {
 691         return kmem_cache_alloc(k, flags | __GFP_ZERO);
 692 }
 693
 694 /**
 695  * kzalloc - allocate memory. The memory is set to zero.
 696  * @size: how many bytes of memory are required.
 697  * @flags: the type of memory to allocate (see kmalloc).
 698  */
 699 static inline void *kzalloc(size_t size, gfp_t flags)
 700 {
 701         return kmalloc(size, flags | __GFP_ZERO);
 702 }
 703
 704 /**
 705  * kzalloc_node - allocate zeroed memory from a particular memory node.
 706  * @size: how many bytes of memory are required.
 707  * @flags: the type of memory to allocate (see kmalloc).
 708  * @node: memory node from which to allocate
 709  */
 710 static inline void *kzalloc_node(size_t size, gfp_t flags, int node)
 711 {
 712         return kmalloc_node(size, flags | __GFP_ZERO, node);
 713 }
 714
 715 unsigned int kmem_cache_size(struct kmem_cache *s);
 716 void __init kmem_cache_init_late(void);
 717
 718 #if defined(CONFIG_SMP) && defined(CONFIG_SLAB)
 719 int slab_prepare_cpu(unsigned int cpu);
 720 int slab_dead_cpu(unsigned int cpu);
 721 #else
 722 #define slab_prepare_cpu        NULL
 723 #define slab_dead_cpu           NULL
 724 #endif
 725
 726 #endif  /* _LINUX_SLAB_H */