]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - include/linux/slab.h
ceph: quota: add initial infrastructure to support cephfs quotas
[mirror_ubuntu-bionic-kernel.git] / include / linux / slab.h
CommitLineData
b2441318 1/* SPDX-License-Identifier: GPL-2.0 */
1da177e4 2/*
2e892f43
CL
3 * Written by Mark Hemment, 1996 (markhe@nextd.demon.co.uk).
4 *
cde53535 5 * (C) SGI 2006, Christoph Lameter
2e892f43
CL
6 * Cleaned up and restructured to ease the addition of alternative
7 * implementations of SLAB allocators.
f1b6eb6e
CL
8 * (C) Linux Foundation 2008-2013
9 * Unified interface for all slab allocators
1da177e4
LT
10 */
11
12#ifndef _LINUX_SLAB_H
13#define _LINUX_SLAB_H
14
1b1cec4b 15#include <linux/gfp.h>
1b1cec4b 16#include <linux/types.h>
1f458cbf
GC
17#include <linux/workqueue.h>
18
1da177e4 19
2e892f43
CL
20/*
21 * Flags to pass to kmem_cache_create().
124dee09 22 * The ones marked DEBUG are only valid if CONFIG_DEBUG_SLAB is set.
1da177e4 23 */
d50112ed 24/* DEBUG: Perform (expensive) checks on alloc/free */
4fd0b46e 25#define SLAB_CONSISTENCY_CHECKS ((slab_flags_t __force)0x00000100U)
d50112ed 26/* DEBUG: Red zone objs in a cache */
4fd0b46e 27#define SLAB_RED_ZONE ((slab_flags_t __force)0x00000400U)
d50112ed 28/* DEBUG: Poison objects */
4fd0b46e 29#define SLAB_POISON ((slab_flags_t __force)0x00000800U)
d50112ed 30/* Align objs on cache lines */
4fd0b46e 31#define SLAB_HWCACHE_ALIGN ((slab_flags_t __force)0x00002000U)
d50112ed 32/* Use GFP_DMA memory */
4fd0b46e 33#define SLAB_CACHE_DMA ((slab_flags_t __force)0x00004000U)
d50112ed 34/* DEBUG: Store the last owner for bug hunting */
4fd0b46e 35#define SLAB_STORE_USER ((slab_flags_t __force)0x00010000U)
d50112ed 36/* Panic if kmem_cache_create() fails */
4fd0b46e 37#define SLAB_PANIC ((slab_flags_t __force)0x00040000U)
d7de4c1d 38/*
5f0d5a3a 39 * SLAB_TYPESAFE_BY_RCU - **WARNING** READ THIS!
d7de4c1d
PZ
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 *
68126702
JK
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.
5f0d5a3a
PM
72 *
73 * Note that SLAB_TYPESAFE_BY_RCU was originally named SLAB_DESTROY_BY_RCU.
d7de4c1d 74 */
d50112ed 75/* Defer freeing slabs to RCU */
4fd0b46e 76#define SLAB_TYPESAFE_BY_RCU ((slab_flags_t __force)0x00080000U)
d50112ed 77/* Spread some memory over cpuset */
4fd0b46e 78#define SLAB_MEM_SPREAD ((slab_flags_t __force)0x00100000U)
d50112ed 79/* Trace allocations and frees */
4fd0b46e 80#define SLAB_TRACE ((slab_flags_t __force)0x00200000U)
1da177e4 81
30327acf
TG
82/* Flag to prevent checks on free */
83#ifdef CONFIG_DEBUG_OBJECTS
4fd0b46e 84# define SLAB_DEBUG_OBJECTS ((slab_flags_t __force)0x00400000U)
30327acf 85#else
4fd0b46e 86# define SLAB_DEBUG_OBJECTS 0
30327acf
TG
87#endif
88
d50112ed 89/* Avoid kmemleak tracing */
4fd0b46e 90#define SLAB_NOLEAKTRACE ((slab_flags_t __force)0x00800000U)
d5cff635 91
d50112ed 92/* Fault injection mark */
4c13dd3b 93#ifdef CONFIG_FAILSLAB
4fd0b46e 94# define SLAB_FAILSLAB ((slab_flags_t __force)0x02000000U)
4c13dd3b 95#else
4fd0b46e 96# define SLAB_FAILSLAB 0
4c13dd3b 97#endif
d50112ed 98/* Account to memcg */
127424c8 99#if defined(CONFIG_MEMCG) && !defined(CONFIG_SLOB)
4fd0b46e 100# define SLAB_ACCOUNT ((slab_flags_t __force)0x04000000U)
230e9fc2 101#else
4fd0b46e 102# define SLAB_ACCOUNT 0
230e9fc2 103#endif
2dff4405 104
7ed2f9e6 105#ifdef CONFIG_KASAN
4fd0b46e 106#define SLAB_KASAN ((slab_flags_t __force)0x08000000U)
7ed2f9e6 107#else
4fd0b46e 108#define SLAB_KASAN 0
7ed2f9e6
AP
109#endif
110
e12ba74d 111/* The following flags affect the page allocator grouping pages by mobility */
d50112ed 112/* Objects are reclaimable */
4fd0b46e 113#define SLAB_RECLAIM_ACCOUNT ((slab_flags_t __force)0x00020000U)
e12ba74d 114#define SLAB_TEMPORARY SLAB_RECLAIM_ACCOUNT /* Objects are short-lived */
6cb8f913
CL
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
1d4ec7b1 125#define ZERO_OR_NULL_PTR(x) ((unsigned long)(x) <= \
6cb8f913
CL
126 (unsigned long)ZERO_SIZE_PTR)
127
f1b6eb6e 128#include <linux/kmemleak.h>
0316bec2 129#include <linux/kasan.h>
3b0efdfa 130
2633d7a0 131struct mem_cgroup;
2e892f43
CL
132/*
133 * struct kmem_cache related prototypes
134 */
135void __init kmem_cache_init(void);
fda90124 136bool slab_is_available(void);
1da177e4 137
2e892f43 138struct kmem_cache *kmem_cache_create(const char *, size_t, size_t,
d50112ed 139 slab_flags_t,
51cc5068 140 void (*)(void *));
2e892f43
CL
141void kmem_cache_destroy(struct kmem_cache *);
142int kmem_cache_shrink(struct kmem_cache *);
2a4db7eb
VD
143
144void memcg_create_kmem_cache(struct mem_cgroup *, struct kmem_cache *);
145void memcg_deactivate_kmem_caches(struct mem_cgroup *);
146void memcg_destroy_kmem_caches(struct mem_cgroup *);
2e892f43 147
0a31bd5f
CL
148/*
149 * Please use this macro to create slab caches. Simply specify the
150 * name of the structure and maybe some flags that are listed above.
151 *
152 * The alignment of the struct determines object alignment. If you
153 * f.e. add ____cacheline_aligned_in_smp to the struct declaration
154 * then the objects will be properly aligned in SMP configurations.
155 */
156#define KMEM_CACHE(__struct, __flags) kmem_cache_create(#__struct,\
157 sizeof(struct __struct), __alignof__(struct __struct),\
20c2df83 158 (__flags), NULL)
0a31bd5f 159
34504667
CL
160/*
161 * Common kmalloc functions provided by all allocators
162 */
163void * __must_check __krealloc(const void *, size_t, gfp_t);
164void * __must_check krealloc(const void *, size_t, gfp_t);
165void kfree(const void *);
166void kzfree(const void *);
167size_t ksize(const void *);
168
f5509cc1
KC
169#ifdef CONFIG_HAVE_HARDENED_USERCOPY_ALLOCATOR
170const char *__check_heap_object(const void *ptr, unsigned long n,
171 struct page *page);
172#else
173static inline const char *__check_heap_object(const void *ptr,
174 unsigned long n,
175 struct page *page)
176{
177 return NULL;
178}
179#endif
180
c601fd69
CL
181/*
182 * Some archs want to perform DMA into kmalloc caches and need a guaranteed
183 * alignment larger than the alignment of a 64-bit integer.
184 * Setting ARCH_KMALLOC_MINALIGN in arch headers allows that.
185 */
186#if defined(ARCH_DMA_MINALIGN) && ARCH_DMA_MINALIGN > 8
187#define ARCH_KMALLOC_MINALIGN ARCH_DMA_MINALIGN
188#define KMALLOC_MIN_SIZE ARCH_DMA_MINALIGN
189#define KMALLOC_SHIFT_LOW ilog2(ARCH_DMA_MINALIGN)
190#else
191#define ARCH_KMALLOC_MINALIGN __alignof__(unsigned long long)
192#endif
193
94a58c36
RV
194/*
195 * Setting ARCH_SLAB_MINALIGN in arch headers allows a different alignment.
196 * Intended for arches that get misalignment faults even for 64 bit integer
197 * aligned buffers.
198 */
199#ifndef ARCH_SLAB_MINALIGN
200#define ARCH_SLAB_MINALIGN __alignof__(unsigned long long)
201#endif
202
203/*
204 * kmalloc and friends return ARCH_KMALLOC_MINALIGN aligned
205 * pointers. kmem_cache_alloc and friends return ARCH_SLAB_MINALIGN
206 * aligned pointers.
207 */
208#define __assume_kmalloc_alignment __assume_aligned(ARCH_KMALLOC_MINALIGN)
209#define __assume_slab_alignment __assume_aligned(ARCH_SLAB_MINALIGN)
210#define __assume_page_alignment __assume_aligned(PAGE_SIZE)
211
0aa817f0 212/*
95a05b42
CL
213 * Kmalloc array related definitions
214 */
215
216#ifdef CONFIG_SLAB
217/*
218 * The largest kmalloc size supported by the SLAB allocators is
0aa817f0
CL
219 * 32 megabyte (2^25) or the maximum allocatable page order if that is
220 * less than 32 MB.
221 *
222 * WARNING: Its not easy to increase this value since the allocators have
223 * to do various tricks to work around compiler limitations in order to
224 * ensure proper constant folding.
225 */
debee076
CL
226#define KMALLOC_SHIFT_HIGH ((MAX_ORDER + PAGE_SHIFT - 1) <= 25 ? \
227 (MAX_ORDER + PAGE_SHIFT - 1) : 25)
95a05b42 228#define KMALLOC_SHIFT_MAX KMALLOC_SHIFT_HIGH
c601fd69 229#ifndef KMALLOC_SHIFT_LOW
95a05b42 230#define KMALLOC_SHIFT_LOW 5
c601fd69 231#endif
069e2b35
CL
232#endif
233
234#ifdef CONFIG_SLUB
95a05b42 235/*
433a91ff
DH
236 * SLUB directly allocates requests fitting in to an order-1 page
237 * (PAGE_SIZE*2). Larger requests are passed to the page allocator.
95a05b42
CL
238 */
239#define KMALLOC_SHIFT_HIGH (PAGE_SHIFT + 1)
bb1107f7 240#define KMALLOC_SHIFT_MAX (MAX_ORDER + PAGE_SHIFT - 1)
c601fd69 241#ifndef KMALLOC_SHIFT_LOW
95a05b42
CL
242#define KMALLOC_SHIFT_LOW 3
243#endif
c601fd69 244#endif
0aa817f0 245
069e2b35
CL
246#ifdef CONFIG_SLOB
247/*
433a91ff 248 * SLOB passes all requests larger than one page to the page allocator.
069e2b35
CL
249 * No kmalloc array is necessary since objects of different sizes can
250 * be allocated from the same page.
251 */
069e2b35 252#define KMALLOC_SHIFT_HIGH PAGE_SHIFT
bb1107f7 253#define KMALLOC_SHIFT_MAX (MAX_ORDER + PAGE_SHIFT - 1)
069e2b35
CL
254#ifndef KMALLOC_SHIFT_LOW
255#define KMALLOC_SHIFT_LOW 3
256#endif
257#endif
258
95a05b42
CL
259/* Maximum allocatable size */
260#define KMALLOC_MAX_SIZE (1UL << KMALLOC_SHIFT_MAX)
261/* Maximum size for which we actually use a slab cache */
262#define KMALLOC_MAX_CACHE_SIZE (1UL << KMALLOC_SHIFT_HIGH)
263/* Maximum order allocatable via the slab allocagtor */
264#define KMALLOC_MAX_ORDER (KMALLOC_SHIFT_MAX - PAGE_SHIFT)
0aa817f0 265
ce6a5026
CL
266/*
267 * Kmalloc subsystem.
268 */
c601fd69 269#ifndef KMALLOC_MIN_SIZE
95a05b42 270#define KMALLOC_MIN_SIZE (1 << KMALLOC_SHIFT_LOW)
ce6a5026
CL
271#endif
272
24f870d8
JK
273/*
274 * This restriction comes from byte sized index implementation.
275 * Page size is normally 2^12 bytes and, in this case, if we want to use
276 * byte sized index which can represent 2^8 entries, the size of the object
277 * should be equal or greater to 2^12 / 2^8 = 2^4 = 16.
278 * If minimum size of kmalloc is less than 16, we use it as minimum object
279 * size and give up to use byte sized index.
280 */
281#define SLAB_OBJ_MIN_SIZE (KMALLOC_MIN_SIZE < 16 ? \
282 (KMALLOC_MIN_SIZE) : 16)
283
069e2b35 284#ifndef CONFIG_SLOB
9425c58e
CL
285extern struct kmem_cache *kmalloc_caches[KMALLOC_SHIFT_HIGH + 1];
286#ifdef CONFIG_ZONE_DMA
287extern struct kmem_cache *kmalloc_dma_caches[KMALLOC_SHIFT_HIGH + 1];
288#endif
289
ce6a5026
CL
290/*
291 * Figure out which kmalloc slab an allocation of a certain size
292 * belongs to.
293 * 0 = zero alloc
294 * 1 = 65 .. 96 bytes
1ed58b60
RV
295 * 2 = 129 .. 192 bytes
296 * n = 2^(n-1)+1 .. 2^n
ce6a5026
CL
297 */
298static __always_inline int kmalloc_index(size_t size)
299{
300 if (!size)
301 return 0;
302
303 if (size <= KMALLOC_MIN_SIZE)
304 return KMALLOC_SHIFT_LOW;
305
306 if (KMALLOC_MIN_SIZE <= 32 && size > 64 && size <= 96)
307 return 1;
308 if (KMALLOC_MIN_SIZE <= 64 && size > 128 && size <= 192)
309 return 2;
310 if (size <= 8) return 3;
311 if (size <= 16) return 4;
312 if (size <= 32) return 5;
313 if (size <= 64) return 6;
314 if (size <= 128) return 7;
315 if (size <= 256) return 8;
316 if (size <= 512) return 9;
317 if (size <= 1024) return 10;
318 if (size <= 2 * 1024) return 11;
319 if (size <= 4 * 1024) return 12;
320 if (size <= 8 * 1024) return 13;
321 if (size <= 16 * 1024) return 14;
322 if (size <= 32 * 1024) return 15;
323 if (size <= 64 * 1024) return 16;
324 if (size <= 128 * 1024) return 17;
325 if (size <= 256 * 1024) return 18;
326 if (size <= 512 * 1024) return 19;
327 if (size <= 1024 * 1024) return 20;
328 if (size <= 2 * 1024 * 1024) return 21;
329 if (size <= 4 * 1024 * 1024) return 22;
330 if (size <= 8 * 1024 * 1024) return 23;
331 if (size <= 16 * 1024 * 1024) return 24;
332 if (size <= 32 * 1024 * 1024) return 25;
333 if (size <= 64 * 1024 * 1024) return 26;
334 BUG();
335
336 /* Will never be reached. Needed because the compiler may complain */
337 return -1;
338}
069e2b35 339#endif /* !CONFIG_SLOB */
ce6a5026 340
48a27055
RV
341void *__kmalloc(size_t size, gfp_t flags) __assume_kmalloc_alignment __malloc;
342void *kmem_cache_alloc(struct kmem_cache *, gfp_t flags) __assume_slab_alignment __malloc;
2a4db7eb 343void kmem_cache_free(struct kmem_cache *, void *);
f1b6eb6e 344
484748f0 345/*
9f706d68 346 * Bulk allocation and freeing operations. These are accelerated in an
484748f0
CL
347 * allocator specific way to avoid taking locks repeatedly or building
348 * metadata structures unnecessarily.
349 *
350 * Note that interrupts must be enabled when calling these functions.
351 */
352void kmem_cache_free_bulk(struct kmem_cache *, size_t, void **);
865762a8 353int kmem_cache_alloc_bulk(struct kmem_cache *, gfp_t, size_t, void **);
484748f0 354
ca257195
JDB
355/*
356 * Caller must not use kfree_bulk() on memory not originally allocated
357 * by kmalloc(), because the SLOB allocator cannot handle this.
358 */
359static __always_inline void kfree_bulk(size_t size, void **p)
360{
361 kmem_cache_free_bulk(NULL, size, p);
362}
363
f1b6eb6e 364#ifdef CONFIG_NUMA
48a27055
RV
365void *__kmalloc_node(size_t size, gfp_t flags, int node) __assume_kmalloc_alignment __malloc;
366void *kmem_cache_alloc_node(struct kmem_cache *, gfp_t flags, int node) __assume_slab_alignment __malloc;
f1b6eb6e
CL
367#else
368static __always_inline void *__kmalloc_node(size_t size, gfp_t flags, int node)
369{
370 return __kmalloc(size, flags);
371}
372
373static __always_inline void *kmem_cache_alloc_node(struct kmem_cache *s, gfp_t flags, int node)
374{
375 return kmem_cache_alloc(s, flags);
376}
377#endif
378
379#ifdef CONFIG_TRACING
48a27055 380extern void *kmem_cache_alloc_trace(struct kmem_cache *, gfp_t, size_t) __assume_slab_alignment __malloc;
f1b6eb6e
CL
381
382#ifdef CONFIG_NUMA
383extern void *kmem_cache_alloc_node_trace(struct kmem_cache *s,
384 gfp_t gfpflags,
48a27055 385 int node, size_t size) __assume_slab_alignment __malloc;
f1b6eb6e
CL
386#else
387static __always_inline void *
388kmem_cache_alloc_node_trace(struct kmem_cache *s,
389 gfp_t gfpflags,
390 int node, size_t size)
391{
392 return kmem_cache_alloc_trace(s, gfpflags, size);
393}
394#endif /* CONFIG_NUMA */
395
396#else /* CONFIG_TRACING */
397static __always_inline void *kmem_cache_alloc_trace(struct kmem_cache *s,
398 gfp_t flags, size_t size)
399{
0316bec2
AR
400 void *ret = kmem_cache_alloc(s, flags);
401
505f5dcb 402 kasan_kmalloc(s, ret, size, flags);
0316bec2 403 return ret;
f1b6eb6e
CL
404}
405
406static __always_inline void *
407kmem_cache_alloc_node_trace(struct kmem_cache *s,
408 gfp_t gfpflags,
409 int node, size_t size)
410{
0316bec2
AR
411 void *ret = kmem_cache_alloc_node(s, gfpflags, node);
412
505f5dcb 413 kasan_kmalloc(s, ret, size, gfpflags);
0316bec2 414 return ret;
f1b6eb6e
CL
415}
416#endif /* CONFIG_TRACING */
417
48a27055 418extern void *kmalloc_order(size_t size, gfp_t flags, unsigned int order) __assume_page_alignment __malloc;
f1b6eb6e
CL
419
420#ifdef CONFIG_TRACING
48a27055 421extern void *kmalloc_order_trace(size_t size, gfp_t flags, unsigned int order) __assume_page_alignment __malloc;
f1b6eb6e
CL
422#else
423static __always_inline void *
424kmalloc_order_trace(size_t size, gfp_t flags, unsigned int order)
425{
426 return kmalloc_order(size, flags, order);
427}
ce6a5026
CL
428#endif
429
f1b6eb6e
CL
430static __always_inline void *kmalloc_large(size_t size, gfp_t flags)
431{
432 unsigned int order = get_order(size);
433 return kmalloc_order_trace(size, flags, order);
434}
435
436/**
437 * kmalloc - allocate memory
438 * @size: how many bytes of memory are required.
7e3528c3 439 * @flags: the type of memory to allocate.
f1b6eb6e
CL
440 *
441 * kmalloc is the normal method of allocating memory
442 * for objects smaller than page size in the kernel.
7e3528c3
RD
443 *
444 * The @flags argument may be one of:
445 *
446 * %GFP_USER - Allocate memory on behalf of user. May sleep.
447 *
448 * %GFP_KERNEL - Allocate normal kernel ram. May sleep.
449 *
450 * %GFP_ATOMIC - Allocation will not sleep. May use emergency pools.
451 * For example, use this inside interrupt handlers.
452 *
453 * %GFP_HIGHUSER - Allocate pages from high memory.
454 *
455 * %GFP_NOIO - Do not do any I/O at all while trying to get memory.
456 *
457 * %GFP_NOFS - Do not make any fs calls while trying to get memory.
458 *
459 * %GFP_NOWAIT - Allocation will not sleep.
460 *
e97ca8e5 461 * %__GFP_THISNODE - Allocate node-local memory only.
7e3528c3
RD
462 *
463 * %GFP_DMA - Allocation suitable for DMA.
464 * Should only be used for kmalloc() caches. Otherwise, use a
465 * slab created with SLAB_DMA.
466 *
467 * Also it is possible to set different flags by OR'ing
468 * in one or more of the following additional @flags:
469 *
7e3528c3
RD
470 * %__GFP_HIGH - This allocation has high priority and may use emergency pools.
471 *
472 * %__GFP_NOFAIL - Indicate that this allocation is in no way allowed to fail
473 * (think twice before using).
474 *
475 * %__GFP_NORETRY - If memory is not immediately available,
476 * then give up at once.
477 *
478 * %__GFP_NOWARN - If allocation fails, don't issue any warnings.
479 *
dcda9b04
MH
480 * %__GFP_RETRY_MAYFAIL - Try really hard to succeed the allocation but fail
481 * eventually.
7e3528c3
RD
482 *
483 * There are other flags available as well, but these are not intended
484 * for general use, and so are not documented here. For a full list of
485 * potential flags, always refer to linux/gfp.h.
f1b6eb6e
CL
486 */
487static __always_inline void *kmalloc(size_t size, gfp_t flags)
488{
489 if (__builtin_constant_p(size)) {
490 if (size > KMALLOC_MAX_CACHE_SIZE)
491 return kmalloc_large(size, flags);
492#ifndef CONFIG_SLOB
493 if (!(flags & GFP_DMA)) {
494 int index = kmalloc_index(size);
495
496 if (!index)
497 return ZERO_SIZE_PTR;
498
499 return kmem_cache_alloc_trace(kmalloc_caches[index],
500 flags, size);
501 }
502#endif
503 }
504 return __kmalloc(size, flags);
505}
506
ce6a5026
CL
507/*
508 * Determine size used for the nth kmalloc cache.
509 * return size or 0 if a kmalloc cache for that
510 * size does not exist
511 */
512static __always_inline int kmalloc_size(int n)
513{
069e2b35 514#ifndef CONFIG_SLOB
ce6a5026
CL
515 if (n > 2)
516 return 1 << n;
517
518 if (n == 1 && KMALLOC_MIN_SIZE <= 32)
519 return 96;
520
521 if (n == 2 && KMALLOC_MIN_SIZE <= 64)
522 return 192;
069e2b35 523#endif
ce6a5026
CL
524 return 0;
525}
ce6a5026 526
f1b6eb6e
CL
527static __always_inline void *kmalloc_node(size_t size, gfp_t flags, int node)
528{
529#ifndef CONFIG_SLOB
530 if (__builtin_constant_p(size) &&
23774a2f 531 size <= KMALLOC_MAX_CACHE_SIZE && !(flags & GFP_DMA)) {
f1b6eb6e
CL
532 int i = kmalloc_index(size);
533
534 if (!i)
535 return ZERO_SIZE_PTR;
536
537 return kmem_cache_alloc_node_trace(kmalloc_caches[i],
538 flags, node, size);
539 }
540#endif
541 return __kmalloc_node(size, flags, node);
542}
543
f7ce3190
VD
544struct memcg_cache_array {
545 struct rcu_head rcu;
546 struct kmem_cache *entries[0];
547};
548
ba6c496e
GC
549/*
550 * This is the main placeholder for memcg-related information in kmem caches.
ba6c496e
GC
551 * Both the root cache and the child caches will have it. For the root cache,
552 * this will hold a dynamically allocated array large enough to hold
f8570263
VD
553 * information about the currently limited memcgs in the system. To allow the
554 * array to be accessed without taking any locks, on relocation we free the old
555 * version only after a grace period.
ba6c496e 556 *
9eeadc8b 557 * Root and child caches hold different metadata.
ba6c496e 558 *
9eeadc8b
TH
559 * @root_cache: Common to root and child caches. NULL for root, pointer to
560 * the root cache for children.
426589f5 561 *
9eeadc8b
TH
562 * The following fields are specific to root caches.
563 *
564 * @memcg_caches: kmemcg ID indexed table of child caches. This table is
565 * used to index child cachces during allocation and cleared
566 * early during shutdown.
567 *
510ded33
TH
568 * @root_caches_node: List node for slab_root_caches list.
569 *
9eeadc8b
TH
570 * @children: List of all child caches. While the child caches are also
571 * reachable through @memcg_caches, a child cache remains on
572 * this list until it is actually destroyed.
573 *
574 * The following fields are specific to child caches.
575 *
576 * @memcg: Pointer to the memcg this cache belongs to.
577 *
578 * @children_node: List node for @root_cache->children list.
bc2791f8
TH
579 *
580 * @kmem_caches_node: List node for @memcg->kmem_caches list.
ba6c496e
GC
581 */
582struct memcg_cache_params {
9eeadc8b 583 struct kmem_cache *root_cache;
ba6c496e 584 union {
9eeadc8b
TH
585 struct {
586 struct memcg_cache_array __rcu *memcg_caches;
510ded33 587 struct list_head __root_caches_node;
9eeadc8b
TH
588 struct list_head children;
589 };
2633d7a0
GC
590 struct {
591 struct mem_cgroup *memcg;
9eeadc8b 592 struct list_head children_node;
bc2791f8 593 struct list_head kmem_caches_node;
01fb58bc
TH
594
595 void (*deact_fn)(struct kmem_cache *);
596 union {
597 struct rcu_head deact_rcu_head;
598 struct work_struct deact_work;
599 };
2633d7a0 600 };
ba6c496e
GC
601 };
602};
603
2633d7a0
GC
604int memcg_update_all_caches(int num_memcgs);
605
e7efa615
MO
606/**
607 * kmalloc_array - allocate memory for an array.
608 * @n: number of elements.
609 * @size: element size.
610 * @flags: the type of memory to allocate (see kmalloc).
800590f5 611 */
a8203725 612static inline void *kmalloc_array(size_t n, size_t size, gfp_t flags)
1da177e4 613{
a3860c1c 614 if (size != 0 && n > SIZE_MAX / size)
6193a2ff 615 return NULL;
91c6a05f
AD
616 if (__builtin_constant_p(n) && __builtin_constant_p(size))
617 return kmalloc(n * size, flags);
a8203725
XW
618 return __kmalloc(n * size, flags);
619}
620
621/**
622 * kcalloc - allocate memory for an array. The memory is set to zero.
623 * @n: number of elements.
624 * @size: element size.
625 * @flags: the type of memory to allocate (see kmalloc).
626 */
627static inline void *kcalloc(size_t n, size_t size, gfp_t flags)
628{
629 return kmalloc_array(n, size, flags | __GFP_ZERO);
1da177e4
LT
630}
631
1d2c8eea
CH
632/*
633 * kmalloc_track_caller is a special version of kmalloc that records the
634 * calling function of the routine calling it for slab leak tracking instead
635 * of just the calling function (confusing, eh?).
636 * It's useful when the call to kmalloc comes from a widely-used standard
637 * allocator where we care about the real place the memory allocation
638 * request comes from.
639 */
ce71e27c 640extern void *__kmalloc_track_caller(size_t, gfp_t, unsigned long);
1d2c8eea 641#define kmalloc_track_caller(size, flags) \
ce71e27c 642 __kmalloc_track_caller(size, flags, _RET_IP_)
1da177e4 643
5799b255
JT
644static inline void *kmalloc_array_node(size_t n, size_t size, gfp_t flags,
645 int node)
646{
647 if (size != 0 && n > SIZE_MAX / size)
648 return NULL;
649 if (__builtin_constant_p(n) && __builtin_constant_p(size))
650 return kmalloc_node(n * size, flags, node);
651 return __kmalloc_node(n * size, flags, node);
652}
653
654static inline void *kcalloc_node(size_t n, size_t size, gfp_t flags, int node)
655{
656 return kmalloc_array_node(n, size, flags | __GFP_ZERO, node);
657}
658
659
97e2bde4 660#ifdef CONFIG_NUMA
ce71e27c 661extern void *__kmalloc_node_track_caller(size_t, gfp_t, int, unsigned long);
8b98c169
CH
662#define kmalloc_node_track_caller(size, flags, node) \
663 __kmalloc_node_track_caller(size, flags, node, \
ce71e27c 664 _RET_IP_)
2e892f43 665
8b98c169 666#else /* CONFIG_NUMA */
8b98c169
CH
667
668#define kmalloc_node_track_caller(size, flags, node) \
669 kmalloc_track_caller(size, flags)
97e2bde4 670
dfcd3610 671#endif /* CONFIG_NUMA */
10cef602 672
81cda662
CL
673/*
674 * Shortcuts
675 */
676static inline void *kmem_cache_zalloc(struct kmem_cache *k, gfp_t flags)
677{
678 return kmem_cache_alloc(k, flags | __GFP_ZERO);
679}
680
681/**
682 * kzalloc - allocate memory. The memory is set to zero.
683 * @size: how many bytes of memory are required.
684 * @flags: the type of memory to allocate (see kmalloc).
685 */
686static inline void *kzalloc(size_t size, gfp_t flags)
687{
688 return kmalloc(size, flags | __GFP_ZERO);
689}
690
979b0fea
JL
691/**
692 * kzalloc_node - allocate zeroed memory from a particular memory node.
693 * @size: how many bytes of memory are required.
694 * @flags: the type of memory to allocate (see kmalloc).
695 * @node: memory node from which to allocate
696 */
697static inline void *kzalloc_node(size_t size, gfp_t flags, int node)
698{
699 return kmalloc_node(size, flags | __GFP_ZERO, node);
700}
701
07f361b2 702unsigned int kmem_cache_size(struct kmem_cache *s);
7e85ee0c
PE
703void __init kmem_cache_init_late(void);
704
6731d4f1
SAS
705#if defined(CONFIG_SMP) && defined(CONFIG_SLAB)
706int slab_prepare_cpu(unsigned int cpu);
707int slab_dead_cpu(unsigned int cpu);
708#else
709#define slab_prepare_cpu NULL
710#define slab_dead_cpu NULL
711#endif
712
1da177e4 713#endif /* _LINUX_SLAB_H */