]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - include/linux/slab.h
net: rtnetlink: validate IFLA_MTU attribute in rtnl_create_link()
[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)
26f42db9
NB
34/* Use GFP_DMA32 memory */
35#define SLAB_CACHE_DMA32 ((slab_flags_t __force)0x00008000U)
d50112ed 36/* DEBUG: Store the last owner for bug hunting */
4fd0b46e 37#define SLAB_STORE_USER ((slab_flags_t __force)0x00010000U)
d50112ed 38/* Panic if kmem_cache_create() fails */
4fd0b46e 39#define SLAB_PANIC ((slab_flags_t __force)0x00040000U)
d7de4c1d 40/*
5f0d5a3a 41 * SLAB_TYPESAFE_BY_RCU - **WARNING** READ THIS!
d7de4c1d
PZ
42 *
43 * This delays freeing the SLAB page by a grace period, it does _NOT_
44 * delay object freeing. This means that if you do kmem_cache_free()
45 * that memory location is free to be reused at any time. Thus it may
46 * be possible to see another object there in the same RCU grace period.
47 *
48 * This feature only ensures the memory location backing the object
49 * stays valid, the trick to using this is relying on an independent
50 * object validation pass. Something like:
51 *
52 * rcu_read_lock()
53 * again:
54 * obj = lockless_lookup(key);
55 * if (obj) {
56 * if (!try_get_ref(obj)) // might fail for free objects
57 * goto again;
58 *
59 * if (obj->key != key) { // not the object we expected
60 * put_ref(obj);
61 * goto again;
62 * }
63 * }
64 * rcu_read_unlock();
65 *
68126702
JK
66 * This is useful if we need to approach a kernel structure obliquely,
67 * from its address obtained without the usual locking. We can lock
68 * the structure to stabilize it and check it's still at the given address,
69 * only if we can be sure that the memory has not been meanwhile reused
70 * for some other kind of object (which our subsystem's lock might corrupt).
71 *
72 * rcu_read_lock before reading the address, then rcu_read_unlock after
73 * taking the spinlock within the structure expected at that address.
5f0d5a3a
PM
74 *
75 * Note that SLAB_TYPESAFE_BY_RCU was originally named SLAB_DESTROY_BY_RCU.
d7de4c1d 76 */
d50112ed 77/* Defer freeing slabs to RCU */
4fd0b46e 78#define SLAB_TYPESAFE_BY_RCU ((slab_flags_t __force)0x00080000U)
d50112ed 79/* Spread some memory over cpuset */
4fd0b46e 80#define SLAB_MEM_SPREAD ((slab_flags_t __force)0x00100000U)
d50112ed 81/* Trace allocations and frees */
4fd0b46e 82#define SLAB_TRACE ((slab_flags_t __force)0x00200000U)
1da177e4 83
30327acf
TG
84/* Flag to prevent checks on free */
85#ifdef CONFIG_DEBUG_OBJECTS
4fd0b46e 86# define SLAB_DEBUG_OBJECTS ((slab_flags_t __force)0x00400000U)
30327acf 87#else
4fd0b46e 88# define SLAB_DEBUG_OBJECTS 0
30327acf
TG
89#endif
90
d50112ed 91/* Avoid kmemleak tracing */
4fd0b46e 92#define SLAB_NOLEAKTRACE ((slab_flags_t __force)0x00800000U)
d5cff635 93
d50112ed 94/* Fault injection mark */
4c13dd3b 95#ifdef CONFIG_FAILSLAB
4fd0b46e 96# define SLAB_FAILSLAB ((slab_flags_t __force)0x02000000U)
4c13dd3b 97#else
4fd0b46e 98# define SLAB_FAILSLAB 0
4c13dd3b 99#endif
d50112ed 100/* Account to memcg */
127424c8 101#if defined(CONFIG_MEMCG) && !defined(CONFIG_SLOB)
4fd0b46e 102# define SLAB_ACCOUNT ((slab_flags_t __force)0x04000000U)
230e9fc2 103#else
4fd0b46e 104# define SLAB_ACCOUNT 0
230e9fc2 105#endif
2dff4405 106
7ed2f9e6 107#ifdef CONFIG_KASAN
4fd0b46e 108#define SLAB_KASAN ((slab_flags_t __force)0x08000000U)
7ed2f9e6 109#else
4fd0b46e 110#define SLAB_KASAN 0
7ed2f9e6
AP
111#endif
112
e12ba74d 113/* The following flags affect the page allocator grouping pages by mobility */
d50112ed 114/* Objects are reclaimable */
4fd0b46e 115#define SLAB_RECLAIM_ACCOUNT ((slab_flags_t __force)0x00020000U)
e12ba74d 116#define SLAB_TEMPORARY SLAB_RECLAIM_ACCOUNT /* Objects are short-lived */
6cb8f913
CL
117/*
118 * ZERO_SIZE_PTR will be returned for zero sized kmalloc requests.
119 *
120 * Dereferencing ZERO_SIZE_PTR will lead to a distinct access fault.
121 *
122 * ZERO_SIZE_PTR can be passed to kfree though in the same way that NULL can.
123 * Both make kfree a no-op.
124 */
125#define ZERO_SIZE_PTR ((void *)16)
126
1d4ec7b1 127#define ZERO_OR_NULL_PTR(x) ((unsigned long)(x) <= \
6cb8f913
CL
128 (unsigned long)ZERO_SIZE_PTR)
129
f1b6eb6e 130#include <linux/kmemleak.h>
0316bec2 131#include <linux/kasan.h>
3b0efdfa 132
2633d7a0 133struct mem_cgroup;
2e892f43
CL
134/*
135 * struct kmem_cache related prototypes
136 */
137void __init kmem_cache_init(void);
fda90124 138bool slab_is_available(void);
1da177e4 139
2e892f43 140struct kmem_cache *kmem_cache_create(const char *, size_t, size_t,
d50112ed 141 slab_flags_t,
51cc5068 142 void (*)(void *));
2e892f43
CL
143void kmem_cache_destroy(struct kmem_cache *);
144int kmem_cache_shrink(struct kmem_cache *);
2a4db7eb
VD
145
146void memcg_create_kmem_cache(struct mem_cgroup *, struct kmem_cache *);
147void memcg_deactivate_kmem_caches(struct mem_cgroup *);
148void memcg_destroy_kmem_caches(struct mem_cgroup *);
2e892f43 149
0a31bd5f
CL
150/*
151 * Please use this macro to create slab caches. Simply specify the
152 * name of the structure and maybe some flags that are listed above.
153 *
154 * The alignment of the struct determines object alignment. If you
155 * f.e. add ____cacheline_aligned_in_smp to the struct declaration
156 * then the objects will be properly aligned in SMP configurations.
157 */
158#define KMEM_CACHE(__struct, __flags) kmem_cache_create(#__struct,\
159 sizeof(struct __struct), __alignof__(struct __struct),\
20c2df83 160 (__flags), NULL)
0a31bd5f 161
34504667
CL
162/*
163 * Common kmalloc functions provided by all allocators
164 */
165void * __must_check __krealloc(const void *, size_t, gfp_t);
166void * __must_check krealloc(const void *, size_t, gfp_t);
167void kfree(const void *);
168void kzfree(const void *);
169size_t ksize(const void *);
170
f5509cc1
KC
171#ifdef CONFIG_HAVE_HARDENED_USERCOPY_ALLOCATOR
172const char *__check_heap_object(const void *ptr, unsigned long n,
173 struct page *page);
174#else
175static inline const char *__check_heap_object(const void *ptr,
176 unsigned long n,
177 struct page *page)
178{
179 return NULL;
180}
181#endif
182
c601fd69
CL
183/*
184 * Some archs want to perform DMA into kmalloc caches and need a guaranteed
185 * alignment larger than the alignment of a 64-bit integer.
186 * Setting ARCH_KMALLOC_MINALIGN in arch headers allows that.
187 */
188#if defined(ARCH_DMA_MINALIGN) && ARCH_DMA_MINALIGN > 8
189#define ARCH_KMALLOC_MINALIGN ARCH_DMA_MINALIGN
190#define KMALLOC_MIN_SIZE ARCH_DMA_MINALIGN
191#define KMALLOC_SHIFT_LOW ilog2(ARCH_DMA_MINALIGN)
192#else
193#define ARCH_KMALLOC_MINALIGN __alignof__(unsigned long long)
194#endif
195
94a58c36
RV
196/*
197 * Setting ARCH_SLAB_MINALIGN in arch headers allows a different alignment.
198 * Intended for arches that get misalignment faults even for 64 bit integer
199 * aligned buffers.
200 */
201#ifndef ARCH_SLAB_MINALIGN
202#define ARCH_SLAB_MINALIGN __alignof__(unsigned long long)
203#endif
204
205/*
206 * kmalloc and friends return ARCH_KMALLOC_MINALIGN aligned
207 * pointers. kmem_cache_alloc and friends return ARCH_SLAB_MINALIGN
208 * aligned pointers.
209 */
210#define __assume_kmalloc_alignment __assume_aligned(ARCH_KMALLOC_MINALIGN)
211#define __assume_slab_alignment __assume_aligned(ARCH_SLAB_MINALIGN)
212#define __assume_page_alignment __assume_aligned(PAGE_SIZE)
213
0aa817f0 214/*
95a05b42
CL
215 * Kmalloc array related definitions
216 */
217
218#ifdef CONFIG_SLAB
219/*
220 * The largest kmalloc size supported by the SLAB allocators is
0aa817f0
CL
221 * 32 megabyte (2^25) or the maximum allocatable page order if that is
222 * less than 32 MB.
223 *
224 * WARNING: Its not easy to increase this value since the allocators have
225 * to do various tricks to work around compiler limitations in order to
226 * ensure proper constant folding.
227 */
debee076
CL
228#define KMALLOC_SHIFT_HIGH ((MAX_ORDER + PAGE_SHIFT - 1) <= 25 ? \
229 (MAX_ORDER + PAGE_SHIFT - 1) : 25)
95a05b42 230#define KMALLOC_SHIFT_MAX KMALLOC_SHIFT_HIGH
c601fd69 231#ifndef KMALLOC_SHIFT_LOW
95a05b42 232#define KMALLOC_SHIFT_LOW 5
c601fd69 233#endif
069e2b35
CL
234#endif
235
236#ifdef CONFIG_SLUB
95a05b42 237/*
433a91ff
DH
238 * SLUB directly allocates requests fitting in to an order-1 page
239 * (PAGE_SIZE*2). Larger requests are passed to the page allocator.
95a05b42
CL
240 */
241#define KMALLOC_SHIFT_HIGH (PAGE_SHIFT + 1)
bb1107f7 242#define KMALLOC_SHIFT_MAX (MAX_ORDER + PAGE_SHIFT - 1)
c601fd69 243#ifndef KMALLOC_SHIFT_LOW
95a05b42
CL
244#define KMALLOC_SHIFT_LOW 3
245#endif
c601fd69 246#endif
0aa817f0 247
069e2b35
CL
248#ifdef CONFIG_SLOB
249/*
433a91ff 250 * SLOB passes all requests larger than one page to the page allocator.
069e2b35
CL
251 * No kmalloc array is necessary since objects of different sizes can
252 * be allocated from the same page.
253 */
069e2b35 254#define KMALLOC_SHIFT_HIGH PAGE_SHIFT
bb1107f7 255#define KMALLOC_SHIFT_MAX (MAX_ORDER + PAGE_SHIFT - 1)
069e2b35
CL
256#ifndef KMALLOC_SHIFT_LOW
257#define KMALLOC_SHIFT_LOW 3
258#endif
259#endif
260
95a05b42
CL
261/* Maximum allocatable size */
262#define KMALLOC_MAX_SIZE (1UL << KMALLOC_SHIFT_MAX)
263/* Maximum size for which we actually use a slab cache */
264#define KMALLOC_MAX_CACHE_SIZE (1UL << KMALLOC_SHIFT_HIGH)
265/* Maximum order allocatable via the slab allocagtor */
266#define KMALLOC_MAX_ORDER (KMALLOC_SHIFT_MAX - PAGE_SHIFT)
0aa817f0 267
ce6a5026
CL
268/*
269 * Kmalloc subsystem.
270 */
c601fd69 271#ifndef KMALLOC_MIN_SIZE
95a05b42 272#define KMALLOC_MIN_SIZE (1 << KMALLOC_SHIFT_LOW)
ce6a5026
CL
273#endif
274
24f870d8
JK
275/*
276 * This restriction comes from byte sized index implementation.
277 * Page size is normally 2^12 bytes and, in this case, if we want to use
278 * byte sized index which can represent 2^8 entries, the size of the object
279 * should be equal or greater to 2^12 / 2^8 = 2^4 = 16.
280 * If minimum size of kmalloc is less than 16, we use it as minimum object
281 * size and give up to use byte sized index.
282 */
283#define SLAB_OBJ_MIN_SIZE (KMALLOC_MIN_SIZE < 16 ? \
284 (KMALLOC_MIN_SIZE) : 16)
285
069e2b35 286#ifndef CONFIG_SLOB
9425c58e
CL
287extern struct kmem_cache *kmalloc_caches[KMALLOC_SHIFT_HIGH + 1];
288#ifdef CONFIG_ZONE_DMA
289extern struct kmem_cache *kmalloc_dma_caches[KMALLOC_SHIFT_HIGH + 1];
290#endif
291
ce6a5026
CL
292/*
293 * Figure out which kmalloc slab an allocation of a certain size
294 * belongs to.
295 * 0 = zero alloc
296 * 1 = 65 .. 96 bytes
1ed58b60
RV
297 * 2 = 129 .. 192 bytes
298 * n = 2^(n-1)+1 .. 2^n
ce6a5026
CL
299 */
300static __always_inline int kmalloc_index(size_t size)
301{
302 if (!size)
303 return 0;
304
305 if (size <= KMALLOC_MIN_SIZE)
306 return KMALLOC_SHIFT_LOW;
307
308 if (KMALLOC_MIN_SIZE <= 32 && size > 64 && size <= 96)
309 return 1;
310 if (KMALLOC_MIN_SIZE <= 64 && size > 128 && size <= 192)
311 return 2;
312 if (size <= 8) return 3;
313 if (size <= 16) return 4;
314 if (size <= 32) return 5;
315 if (size <= 64) return 6;
316 if (size <= 128) return 7;
317 if (size <= 256) return 8;
318 if (size <= 512) return 9;
319 if (size <= 1024) return 10;
320 if (size <= 2 * 1024) return 11;
321 if (size <= 4 * 1024) return 12;
322 if (size <= 8 * 1024) return 13;
323 if (size <= 16 * 1024) return 14;
324 if (size <= 32 * 1024) return 15;
325 if (size <= 64 * 1024) return 16;
326 if (size <= 128 * 1024) return 17;
327 if (size <= 256 * 1024) return 18;
328 if (size <= 512 * 1024) return 19;
329 if (size <= 1024 * 1024) return 20;
330 if (size <= 2 * 1024 * 1024) return 21;
331 if (size <= 4 * 1024 * 1024) return 22;
332 if (size <= 8 * 1024 * 1024) return 23;
333 if (size <= 16 * 1024 * 1024) return 24;
334 if (size <= 32 * 1024 * 1024) return 25;
335 if (size <= 64 * 1024 * 1024) return 26;
336 BUG();
337
338 /* Will never be reached. Needed because the compiler may complain */
339 return -1;
340}
069e2b35 341#endif /* !CONFIG_SLOB */
ce6a5026 342
48a27055
RV
343void *__kmalloc(size_t size, gfp_t flags) __assume_kmalloc_alignment __malloc;
344void *kmem_cache_alloc(struct kmem_cache *, gfp_t flags) __assume_slab_alignment __malloc;
2a4db7eb 345void kmem_cache_free(struct kmem_cache *, void *);
f1b6eb6e 346
484748f0 347/*
9f706d68 348 * Bulk allocation and freeing operations. These are accelerated in an
484748f0
CL
349 * allocator specific way to avoid taking locks repeatedly or building
350 * metadata structures unnecessarily.
351 *
352 * Note that interrupts must be enabled when calling these functions.
353 */
354void kmem_cache_free_bulk(struct kmem_cache *, size_t, void **);
865762a8 355int kmem_cache_alloc_bulk(struct kmem_cache *, gfp_t, size_t, void **);
484748f0 356
ca257195
JDB
357/*
358 * Caller must not use kfree_bulk() on memory not originally allocated
359 * by kmalloc(), because the SLOB allocator cannot handle this.
360 */
361static __always_inline void kfree_bulk(size_t size, void **p)
362{
363 kmem_cache_free_bulk(NULL, size, p);
364}
365
f1b6eb6e 366#ifdef CONFIG_NUMA
48a27055
RV
367void *__kmalloc_node(size_t size, gfp_t flags, int node) __assume_kmalloc_alignment __malloc;
368void *kmem_cache_alloc_node(struct kmem_cache *, gfp_t flags, int node) __assume_slab_alignment __malloc;
f1b6eb6e
CL
369#else
370static __always_inline void *__kmalloc_node(size_t size, gfp_t flags, int node)
371{
372 return __kmalloc(size, flags);
373}
374
375static __always_inline void *kmem_cache_alloc_node(struct kmem_cache *s, gfp_t flags, int node)
376{
377 return kmem_cache_alloc(s, flags);
378}
379#endif
380
381#ifdef CONFIG_TRACING
48a27055 382extern void *kmem_cache_alloc_trace(struct kmem_cache *, gfp_t, size_t) __assume_slab_alignment __malloc;
f1b6eb6e
CL
383
384#ifdef CONFIG_NUMA
385extern void *kmem_cache_alloc_node_trace(struct kmem_cache *s,
386 gfp_t gfpflags,
48a27055 387 int node, size_t size) __assume_slab_alignment __malloc;
f1b6eb6e
CL
388#else
389static __always_inline void *
390kmem_cache_alloc_node_trace(struct kmem_cache *s,
391 gfp_t gfpflags,
392 int node, size_t size)
393{
394 return kmem_cache_alloc_trace(s, gfpflags, size);
395}
396#endif /* CONFIG_NUMA */
397
398#else /* CONFIG_TRACING */
399static __always_inline void *kmem_cache_alloc_trace(struct kmem_cache *s,
400 gfp_t flags, size_t size)
401{
0316bec2
AR
402 void *ret = kmem_cache_alloc(s, flags);
403
505f5dcb 404 kasan_kmalloc(s, ret, size, flags);
0316bec2 405 return ret;
f1b6eb6e
CL
406}
407
408static __always_inline void *
409kmem_cache_alloc_node_trace(struct kmem_cache *s,
410 gfp_t gfpflags,
411 int node, size_t size)
412{
0316bec2
AR
413 void *ret = kmem_cache_alloc_node(s, gfpflags, node);
414
505f5dcb 415 kasan_kmalloc(s, ret, size, gfpflags);
0316bec2 416 return ret;
f1b6eb6e
CL
417}
418#endif /* CONFIG_TRACING */
419
48a27055 420extern void *kmalloc_order(size_t size, gfp_t flags, unsigned int order) __assume_page_alignment __malloc;
f1b6eb6e
CL
421
422#ifdef CONFIG_TRACING
48a27055 423extern void *kmalloc_order_trace(size_t size, gfp_t flags, unsigned int order) __assume_page_alignment __malloc;
f1b6eb6e
CL
424#else
425static __always_inline void *
426kmalloc_order_trace(size_t size, gfp_t flags, unsigned int order)
427{
428 return kmalloc_order(size, flags, order);
429}
ce6a5026
CL
430#endif
431
f1b6eb6e
CL
432static __always_inline void *kmalloc_large(size_t size, gfp_t flags)
433{
434 unsigned int order = get_order(size);
435 return kmalloc_order_trace(size, flags, order);
436}
437
438/**
439 * kmalloc - allocate memory
440 * @size: how many bytes of memory are required.
7e3528c3 441 * @flags: the type of memory to allocate.
f1b6eb6e
CL
442 *
443 * kmalloc is the normal method of allocating memory
444 * for objects smaller than page size in the kernel.
7e3528c3
RD
445 *
446 * The @flags argument may be one of:
447 *
448 * %GFP_USER - Allocate memory on behalf of user. May sleep.
449 *
450 * %GFP_KERNEL - Allocate normal kernel ram. May sleep.
451 *
452 * %GFP_ATOMIC - Allocation will not sleep. May use emergency pools.
453 * For example, use this inside interrupt handlers.
454 *
455 * %GFP_HIGHUSER - Allocate pages from high memory.
456 *
457 * %GFP_NOIO - Do not do any I/O at all while trying to get memory.
458 *
459 * %GFP_NOFS - Do not make any fs calls while trying to get memory.
460 *
461 * %GFP_NOWAIT - Allocation will not sleep.
462 *
e97ca8e5 463 * %__GFP_THISNODE - Allocate node-local memory only.
7e3528c3
RD
464 *
465 * %GFP_DMA - Allocation suitable for DMA.
466 * Should only be used for kmalloc() caches. Otherwise, use a
467 * slab created with SLAB_DMA.
468 *
469 * Also it is possible to set different flags by OR'ing
470 * in one or more of the following additional @flags:
471 *
7e3528c3
RD
472 * %__GFP_HIGH - This allocation has high priority and may use emergency pools.
473 *
474 * %__GFP_NOFAIL - Indicate that this allocation is in no way allowed to fail
475 * (think twice before using).
476 *
477 * %__GFP_NORETRY - If memory is not immediately available,
478 * then give up at once.
479 *
480 * %__GFP_NOWARN - If allocation fails, don't issue any warnings.
481 *
dcda9b04
MH
482 * %__GFP_RETRY_MAYFAIL - Try really hard to succeed the allocation but fail
483 * eventually.
7e3528c3
RD
484 *
485 * There are other flags available as well, but these are not intended
486 * for general use, and so are not documented here. For a full list of
487 * potential flags, always refer to linux/gfp.h.
f1b6eb6e
CL
488 */
489static __always_inline void *kmalloc(size_t size, gfp_t flags)
490{
491 if (__builtin_constant_p(size)) {
492 if (size > KMALLOC_MAX_CACHE_SIZE)
493 return kmalloc_large(size, flags);
494#ifndef CONFIG_SLOB
495 if (!(flags & GFP_DMA)) {
496 int index = kmalloc_index(size);
497
498 if (!index)
499 return ZERO_SIZE_PTR;
500
501 return kmem_cache_alloc_trace(kmalloc_caches[index],
502 flags, size);
503 }
504#endif
505 }
506 return __kmalloc(size, flags);
507}
508
ce6a5026
CL
509/*
510 * Determine size used for the nth kmalloc cache.
511 * return size or 0 if a kmalloc cache for that
512 * size does not exist
513 */
514static __always_inline int kmalloc_size(int n)
515{
069e2b35 516#ifndef CONFIG_SLOB
ce6a5026
CL
517 if (n > 2)
518 return 1 << n;
519
520 if (n == 1 && KMALLOC_MIN_SIZE <= 32)
521 return 96;
522
523 if (n == 2 && KMALLOC_MIN_SIZE <= 64)
524 return 192;
069e2b35 525#endif
ce6a5026
CL
526 return 0;
527}
ce6a5026 528
f1b6eb6e
CL
529static __always_inline void *kmalloc_node(size_t size, gfp_t flags, int node)
530{
531#ifndef CONFIG_SLOB
532 if (__builtin_constant_p(size) &&
23774a2f 533 size <= KMALLOC_MAX_CACHE_SIZE && !(flags & GFP_DMA)) {
f1b6eb6e
CL
534 int i = kmalloc_index(size);
535
536 if (!i)
537 return ZERO_SIZE_PTR;
538
539 return kmem_cache_alloc_node_trace(kmalloc_caches[i],
540 flags, node, size);
541 }
542#endif
543 return __kmalloc_node(size, flags, node);
544}
545
f7ce3190
VD
546struct memcg_cache_array {
547 struct rcu_head rcu;
548 struct kmem_cache *entries[0];
549};
550
ba6c496e
GC
551/*
552 * This is the main placeholder for memcg-related information in kmem caches.
ba6c496e
GC
553 * Both the root cache and the child caches will have it. For the root cache,
554 * this will hold a dynamically allocated array large enough to hold
f8570263
VD
555 * information about the currently limited memcgs in the system. To allow the
556 * array to be accessed without taking any locks, on relocation we free the old
557 * version only after a grace period.
ba6c496e 558 *
9eeadc8b 559 * Root and child caches hold different metadata.
ba6c496e 560 *
9eeadc8b
TH
561 * @root_cache: Common to root and child caches. NULL for root, pointer to
562 * the root cache for children.
426589f5 563 *
9eeadc8b
TH
564 * The following fields are specific to root caches.
565 *
566 * @memcg_caches: kmemcg ID indexed table of child caches. This table is
567 * used to index child cachces during allocation and cleared
568 * early during shutdown.
569 *
510ded33
TH
570 * @root_caches_node: List node for slab_root_caches list.
571 *
9eeadc8b
TH
572 * @children: List of all child caches. While the child caches are also
573 * reachable through @memcg_caches, a child cache remains on
574 * this list until it is actually destroyed.
575 *
576 * The following fields are specific to child caches.
577 *
578 * @memcg: Pointer to the memcg this cache belongs to.
579 *
580 * @children_node: List node for @root_cache->children list.
bc2791f8
TH
581 *
582 * @kmem_caches_node: List node for @memcg->kmem_caches list.
ba6c496e
GC
583 */
584struct memcg_cache_params {
9eeadc8b 585 struct kmem_cache *root_cache;
ba6c496e 586 union {
9eeadc8b
TH
587 struct {
588 struct memcg_cache_array __rcu *memcg_caches;
510ded33 589 struct list_head __root_caches_node;
9eeadc8b
TH
590 struct list_head children;
591 };
2633d7a0
GC
592 struct {
593 struct mem_cgroup *memcg;
9eeadc8b 594 struct list_head children_node;
bc2791f8 595 struct list_head kmem_caches_node;
01fb58bc
TH
596
597 void (*deact_fn)(struct kmem_cache *);
598 union {
599 struct rcu_head deact_rcu_head;
600 struct work_struct deact_work;
601 };
2633d7a0 602 };
ba6c496e
GC
603 };
604};
605
2633d7a0
GC
606int memcg_update_all_caches(int num_memcgs);
607
e7efa615
MO
608/**
609 * kmalloc_array - allocate memory for an array.
610 * @n: number of elements.
611 * @size: element size.
612 * @flags: the type of memory to allocate (see kmalloc).
800590f5 613 */
a8203725 614static inline void *kmalloc_array(size_t n, size_t size, gfp_t flags)
1da177e4 615{
a3860c1c 616 if (size != 0 && n > SIZE_MAX / size)
6193a2ff 617 return NULL;
91c6a05f
AD
618 if (__builtin_constant_p(n) && __builtin_constant_p(size))
619 return kmalloc(n * size, flags);
a8203725
XW
620 return __kmalloc(n * size, flags);
621}
622
623/**
624 * kcalloc - allocate memory for an array. The memory is set to zero.
625 * @n: number of elements.
626 * @size: element size.
627 * @flags: the type of memory to allocate (see kmalloc).
628 */
629static inline void *kcalloc(size_t n, size_t size, gfp_t flags)
630{
631 return kmalloc_array(n, size, flags | __GFP_ZERO);
1da177e4
LT
632}
633
1d2c8eea
CH
634/*
635 * kmalloc_track_caller is a special version of kmalloc that records the
636 * calling function of the routine calling it for slab leak tracking instead
637 * of just the calling function (confusing, eh?).
638 * It's useful when the call to kmalloc comes from a widely-used standard
639 * allocator where we care about the real place the memory allocation
640 * request comes from.
641 */
ce71e27c 642extern void *__kmalloc_track_caller(size_t, gfp_t, unsigned long);
1d2c8eea 643#define kmalloc_track_caller(size, flags) \
ce71e27c 644 __kmalloc_track_caller(size, flags, _RET_IP_)
1da177e4 645
5799b255
JT
646static inline void *kmalloc_array_node(size_t n, size_t size, gfp_t flags,
647 int node)
648{
649 if (size != 0 && n > SIZE_MAX / size)
650 return NULL;
651 if (__builtin_constant_p(n) && __builtin_constant_p(size))
652 return kmalloc_node(n * size, flags, node);
653 return __kmalloc_node(n * size, flags, node);
654}
655
656static inline void *kcalloc_node(size_t n, size_t size, gfp_t flags, int node)
657{
658 return kmalloc_array_node(n, size, flags | __GFP_ZERO, node);
659}
660
661
97e2bde4 662#ifdef CONFIG_NUMA
ce71e27c 663extern void *__kmalloc_node_track_caller(size_t, gfp_t, int, unsigned long);
8b98c169
CH
664#define kmalloc_node_track_caller(size, flags, node) \
665 __kmalloc_node_track_caller(size, flags, node, \
ce71e27c 666 _RET_IP_)
2e892f43 667
8b98c169 668#else /* CONFIG_NUMA */
8b98c169
CH
669
670#define kmalloc_node_track_caller(size, flags, node) \
671 kmalloc_track_caller(size, flags)
97e2bde4 672
dfcd3610 673#endif /* CONFIG_NUMA */
10cef602 674
81cda662
CL
675/*
676 * Shortcuts
677 */
678static inline void *kmem_cache_zalloc(struct kmem_cache *k, gfp_t flags)
679{
680 return kmem_cache_alloc(k, flags | __GFP_ZERO);
681}
682
683/**
684 * kzalloc - allocate memory. The memory is set to zero.
685 * @size: how many bytes of memory are required.
686 * @flags: the type of memory to allocate (see kmalloc).
687 */
688static inline void *kzalloc(size_t size, gfp_t flags)
689{
690 return kmalloc(size, flags | __GFP_ZERO);
691}
692
979b0fea
JL
693/**
694 * kzalloc_node - allocate zeroed memory from a particular memory node.
695 * @size: how many bytes of memory are required.
696 * @flags: the type of memory to allocate (see kmalloc).
697 * @node: memory node from which to allocate
698 */
699static inline void *kzalloc_node(size_t size, gfp_t flags, int node)
700{
701 return kmalloc_node(size, flags | __GFP_ZERO, node);
702}
703
07f361b2 704unsigned int kmem_cache_size(struct kmem_cache *s);
7e85ee0c
PE
705void __init kmem_cache_init_late(void);
706
6731d4f1
SAS
707#if defined(CONFIG_SMP) && defined(CONFIG_SLAB)
708int slab_prepare_cpu(unsigned int cpu);
709int slab_dead_cpu(unsigned int cpu);
710#else
711#define slab_prepare_cpu NULL
712#define slab_dead_cpu NULL
713#endif
714
1da177e4 715#endif /* _LINUX_SLAB_H */