userfaultfd: allow get_mempolicy(MPOL_F_NODE|MPOL_F_ADDR) to trigger userfaults
[linux-block.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>
49b7f898 16#include <linux/overflow.h>
1b1cec4b 17#include <linux/types.h>
1f458cbf
GC
18#include <linux/workqueue.h>
19
1da177e4 20
2e892f43
CL
21/*
22 * Flags to pass to kmem_cache_create().
124dee09 23 * The ones marked DEBUG are only valid if CONFIG_DEBUG_SLAB is set.
1da177e4 24 */
d50112ed 25/* DEBUG: Perform (expensive) checks on alloc/free */
4fd0b46e 26#define SLAB_CONSISTENCY_CHECKS ((slab_flags_t __force)0x00000100U)
d50112ed 27/* DEBUG: Red zone objs in a cache */
4fd0b46e 28#define SLAB_RED_ZONE ((slab_flags_t __force)0x00000400U)
d50112ed 29/* DEBUG: Poison objects */
4fd0b46e 30#define SLAB_POISON ((slab_flags_t __force)0x00000800U)
d50112ed 31/* Align objs on cache lines */
4fd0b46e 32#define SLAB_HWCACHE_ALIGN ((slab_flags_t __force)0x00002000U)
d50112ed 33/* Use GFP_DMA memory */
4fd0b46e 34#define SLAB_CACHE_DMA ((slab_flags_t __force)0x00004000U)
d50112ed 35/* DEBUG: Store the last owner for bug hunting */
4fd0b46e 36#define SLAB_STORE_USER ((slab_flags_t __force)0x00010000U)
d50112ed 37/* Panic if kmem_cache_create() fails */
4fd0b46e 38#define SLAB_PANIC ((slab_flags_t __force)0x00040000U)
d7de4c1d 39/*
5f0d5a3a 40 * SLAB_TYPESAFE_BY_RCU - **WARNING** READ THIS!
d7de4c1d
PZ
41 *
42 * This delays freeing the SLAB page by a grace period, it does _NOT_
43 * delay object freeing. This means that if you do kmem_cache_free()
44 * that memory location is free to be reused at any time. Thus it may
45 * be possible to see another object there in the same RCU grace period.
46 *
47 * This feature only ensures the memory location backing the object
48 * stays valid, the trick to using this is relying on an independent
49 * object validation pass. Something like:
50 *
51 * rcu_read_lock()
52 * again:
53 * obj = lockless_lookup(key);
54 * if (obj) {
55 * if (!try_get_ref(obj)) // might fail for free objects
56 * goto again;
57 *
58 * if (obj->key != key) { // not the object we expected
59 * put_ref(obj);
60 * goto again;
61 * }
62 * }
63 * rcu_read_unlock();
64 *
68126702
JK
65 * This is useful if we need to approach a kernel structure obliquely,
66 * from its address obtained without the usual locking. We can lock
67 * the structure to stabilize it and check it's still at the given address,
68 * only if we can be sure that the memory has not been meanwhile reused
69 * for some other kind of object (which our subsystem's lock might corrupt).
70 *
71 * rcu_read_lock before reading the address, then rcu_read_unlock after
72 * taking the spinlock within the structure expected at that address.
5f0d5a3a
PM
73 *
74 * Note that SLAB_TYPESAFE_BY_RCU was originally named SLAB_DESTROY_BY_RCU.
d7de4c1d 75 */
d50112ed 76/* Defer freeing slabs to RCU */
4fd0b46e 77#define SLAB_TYPESAFE_BY_RCU ((slab_flags_t __force)0x00080000U)
d50112ed 78/* Spread some memory over cpuset */
4fd0b46e 79#define SLAB_MEM_SPREAD ((slab_flags_t __force)0x00100000U)
d50112ed 80/* Trace allocations and frees */
4fd0b46e 81#define SLAB_TRACE ((slab_flags_t __force)0x00200000U)
1da177e4 82
30327acf
TG
83/* Flag to prevent checks on free */
84#ifdef CONFIG_DEBUG_OBJECTS
4fd0b46e 85# define SLAB_DEBUG_OBJECTS ((slab_flags_t __force)0x00400000U)
30327acf 86#else
4fd0b46e 87# define SLAB_DEBUG_OBJECTS 0
30327acf
TG
88#endif
89
d50112ed 90/* Avoid kmemleak tracing */
4fd0b46e 91#define SLAB_NOLEAKTRACE ((slab_flags_t __force)0x00800000U)
d5cff635 92
d50112ed 93/* Fault injection mark */
4c13dd3b 94#ifdef CONFIG_FAILSLAB
4fd0b46e 95# define SLAB_FAILSLAB ((slab_flags_t __force)0x02000000U)
4c13dd3b 96#else
4fd0b46e 97# define SLAB_FAILSLAB 0
4c13dd3b 98#endif
d50112ed 99/* Account to memcg */
84c07d11 100#ifdef CONFIG_MEMCG_KMEM
4fd0b46e 101# define SLAB_ACCOUNT ((slab_flags_t __force)0x04000000U)
230e9fc2 102#else
4fd0b46e 103# define SLAB_ACCOUNT 0
230e9fc2 104#endif
2dff4405 105
7ed2f9e6 106#ifdef CONFIG_KASAN
4fd0b46e 107#define SLAB_KASAN ((slab_flags_t __force)0x08000000U)
7ed2f9e6 108#else
4fd0b46e 109#define SLAB_KASAN 0
7ed2f9e6
AP
110#endif
111
e12ba74d 112/* The following flags affect the page allocator grouping pages by mobility */
d50112ed 113/* Objects are reclaimable */
4fd0b46e 114#define SLAB_RECLAIM_ACCOUNT ((slab_flags_t __force)0x00020000U)
e12ba74d 115#define SLAB_TEMPORARY SLAB_RECLAIM_ACCOUNT /* Objects are short-lived */
6cb8f913
CL
116/*
117 * ZERO_SIZE_PTR will be returned for zero sized kmalloc requests.
118 *
119 * Dereferencing ZERO_SIZE_PTR will lead to a distinct access fault.
120 *
121 * ZERO_SIZE_PTR can be passed to kfree though in the same way that NULL can.
122 * Both make kfree a no-op.
123 */
124#define ZERO_SIZE_PTR ((void *)16)
125
1d4ec7b1 126#define ZERO_OR_NULL_PTR(x) ((unsigned long)(x) <= \
6cb8f913
CL
127 (unsigned long)ZERO_SIZE_PTR)
128
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
2d891fbc
KC
138extern bool usercopy_fallback;
139
f4957d5b
AD
140struct kmem_cache *kmem_cache_create(const char *name, unsigned int size,
141 unsigned int align, slab_flags_t flags,
8eb8284b
DW
142 void (*ctor)(void *));
143struct kmem_cache *kmem_cache_create_usercopy(const char *name,
f4957d5b
AD
144 unsigned int size, unsigned int align,
145 slab_flags_t flags,
7bbdb81e 146 unsigned int useroffset, unsigned int usersize,
8eb8284b 147 void (*ctor)(void *));
2e892f43
CL
148void kmem_cache_destroy(struct kmem_cache *);
149int kmem_cache_shrink(struct kmem_cache *);
2a4db7eb
VD
150
151void memcg_create_kmem_cache(struct mem_cgroup *, struct kmem_cache *);
152void memcg_deactivate_kmem_caches(struct mem_cgroup *);
153void memcg_destroy_kmem_caches(struct mem_cgroup *);
2e892f43 154
0a31bd5f
CL
155/*
156 * Please use this macro to create slab caches. Simply specify the
157 * name of the structure and maybe some flags that are listed above.
158 *
159 * The alignment of the struct determines object alignment. If you
160 * f.e. add ____cacheline_aligned_in_smp to the struct declaration
161 * then the objects will be properly aligned in SMP configurations.
162 */
8eb8284b
DW
163#define KMEM_CACHE(__struct, __flags) \
164 kmem_cache_create(#__struct, sizeof(struct __struct), \
165 __alignof__(struct __struct), (__flags), NULL)
166
167/*
168 * To whitelist a single field for copying to/from usercopy, use this
169 * macro instead for KMEM_CACHE() above.
170 */
171#define KMEM_CACHE_USERCOPY(__struct, __flags, __field) \
172 kmem_cache_create_usercopy(#__struct, \
173 sizeof(struct __struct), \
174 __alignof__(struct __struct), (__flags), \
175 offsetof(struct __struct, __field), \
176 sizeof_field(struct __struct, __field), NULL)
0a31bd5f 177
34504667
CL
178/*
179 * Common kmalloc functions provided by all allocators
180 */
181void * __must_check __krealloc(const void *, size_t, gfp_t);
182void * __must_check krealloc(const void *, size_t, gfp_t);
183void kfree(const void *);
184void kzfree(const void *);
185size_t ksize(const void *);
186
f5509cc1 187#ifdef CONFIG_HAVE_HARDENED_USERCOPY_ALLOCATOR
f4e6e289
KC
188void __check_heap_object(const void *ptr, unsigned long n, struct page *page,
189 bool to_user);
f5509cc1 190#else
f4e6e289
KC
191static inline void __check_heap_object(const void *ptr, unsigned long n,
192 struct page *page, bool to_user) { }
f5509cc1
KC
193#endif
194
c601fd69
CL
195/*
196 * Some archs want to perform DMA into kmalloc caches and need a guaranteed
197 * alignment larger than the alignment of a 64-bit integer.
198 * Setting ARCH_KMALLOC_MINALIGN in arch headers allows that.
199 */
200#if defined(ARCH_DMA_MINALIGN) && ARCH_DMA_MINALIGN > 8
201#define ARCH_KMALLOC_MINALIGN ARCH_DMA_MINALIGN
202#define KMALLOC_MIN_SIZE ARCH_DMA_MINALIGN
203#define KMALLOC_SHIFT_LOW ilog2(ARCH_DMA_MINALIGN)
204#else
205#define ARCH_KMALLOC_MINALIGN __alignof__(unsigned long long)
206#endif
207
94a58c36
RV
208/*
209 * Setting ARCH_SLAB_MINALIGN in arch headers allows a different alignment.
210 * Intended for arches that get misalignment faults even for 64 bit integer
211 * aligned buffers.
212 */
213#ifndef ARCH_SLAB_MINALIGN
214#define ARCH_SLAB_MINALIGN __alignof__(unsigned long long)
215#endif
216
217/*
218 * kmalloc and friends return ARCH_KMALLOC_MINALIGN aligned
219 * pointers. kmem_cache_alloc and friends return ARCH_SLAB_MINALIGN
220 * aligned pointers.
221 */
222#define __assume_kmalloc_alignment __assume_aligned(ARCH_KMALLOC_MINALIGN)
223#define __assume_slab_alignment __assume_aligned(ARCH_SLAB_MINALIGN)
224#define __assume_page_alignment __assume_aligned(PAGE_SIZE)
225
0aa817f0 226/*
95a05b42
CL
227 * Kmalloc array related definitions
228 */
229
230#ifdef CONFIG_SLAB
231/*
232 * The largest kmalloc size supported by the SLAB allocators is
0aa817f0
CL
233 * 32 megabyte (2^25) or the maximum allocatable page order if that is
234 * less than 32 MB.
235 *
236 * WARNING: Its not easy to increase this value since the allocators have
237 * to do various tricks to work around compiler limitations in order to
238 * ensure proper constant folding.
239 */
debee076
CL
240#define KMALLOC_SHIFT_HIGH ((MAX_ORDER + PAGE_SHIFT - 1) <= 25 ? \
241 (MAX_ORDER + PAGE_SHIFT - 1) : 25)
95a05b42 242#define KMALLOC_SHIFT_MAX KMALLOC_SHIFT_HIGH
c601fd69 243#ifndef KMALLOC_SHIFT_LOW
95a05b42 244#define KMALLOC_SHIFT_LOW 5
c601fd69 245#endif
069e2b35
CL
246#endif
247
248#ifdef CONFIG_SLUB
95a05b42 249/*
433a91ff
DH
250 * SLUB directly allocates requests fitting in to an order-1 page
251 * (PAGE_SIZE*2). Larger requests are passed to the page allocator.
95a05b42
CL
252 */
253#define KMALLOC_SHIFT_HIGH (PAGE_SHIFT + 1)
bb1107f7 254#define KMALLOC_SHIFT_MAX (MAX_ORDER + PAGE_SHIFT - 1)
c601fd69 255#ifndef KMALLOC_SHIFT_LOW
95a05b42
CL
256#define KMALLOC_SHIFT_LOW 3
257#endif
c601fd69 258#endif
0aa817f0 259
069e2b35
CL
260#ifdef CONFIG_SLOB
261/*
433a91ff 262 * SLOB passes all requests larger than one page to the page allocator.
069e2b35
CL
263 * No kmalloc array is necessary since objects of different sizes can
264 * be allocated from the same page.
265 */
069e2b35 266#define KMALLOC_SHIFT_HIGH PAGE_SHIFT
bb1107f7 267#define KMALLOC_SHIFT_MAX (MAX_ORDER + PAGE_SHIFT - 1)
069e2b35
CL
268#ifndef KMALLOC_SHIFT_LOW
269#define KMALLOC_SHIFT_LOW 3
270#endif
271#endif
272
95a05b42
CL
273/* Maximum allocatable size */
274#define KMALLOC_MAX_SIZE (1UL << KMALLOC_SHIFT_MAX)
275/* Maximum size for which we actually use a slab cache */
276#define KMALLOC_MAX_CACHE_SIZE (1UL << KMALLOC_SHIFT_HIGH)
277/* Maximum order allocatable via the slab allocagtor */
278#define KMALLOC_MAX_ORDER (KMALLOC_SHIFT_MAX - PAGE_SHIFT)
0aa817f0 279
ce6a5026
CL
280/*
281 * Kmalloc subsystem.
282 */
c601fd69 283#ifndef KMALLOC_MIN_SIZE
95a05b42 284#define KMALLOC_MIN_SIZE (1 << KMALLOC_SHIFT_LOW)
ce6a5026
CL
285#endif
286
24f870d8
JK
287/*
288 * This restriction comes from byte sized index implementation.
289 * Page size is normally 2^12 bytes and, in this case, if we want to use
290 * byte sized index which can represent 2^8 entries, the size of the object
291 * should be equal or greater to 2^12 / 2^8 = 2^4 = 16.
292 * If minimum size of kmalloc is less than 16, we use it as minimum object
293 * size and give up to use byte sized index.
294 */
295#define SLAB_OBJ_MIN_SIZE (KMALLOC_MIN_SIZE < 16 ? \
296 (KMALLOC_MIN_SIZE) : 16)
297
069e2b35 298#ifndef CONFIG_SLOB
9425c58e
CL
299extern struct kmem_cache *kmalloc_caches[KMALLOC_SHIFT_HIGH + 1];
300#ifdef CONFIG_ZONE_DMA
301extern struct kmem_cache *kmalloc_dma_caches[KMALLOC_SHIFT_HIGH + 1];
302#endif
303
ce6a5026
CL
304/*
305 * Figure out which kmalloc slab an allocation of a certain size
306 * belongs to.
307 * 0 = zero alloc
308 * 1 = 65 .. 96 bytes
1ed58b60
RV
309 * 2 = 129 .. 192 bytes
310 * n = 2^(n-1)+1 .. 2^n
ce6a5026 311 */
36071a27 312static __always_inline unsigned int kmalloc_index(size_t size)
ce6a5026
CL
313{
314 if (!size)
315 return 0;
316
317 if (size <= KMALLOC_MIN_SIZE)
318 return KMALLOC_SHIFT_LOW;
319
320 if (KMALLOC_MIN_SIZE <= 32 && size > 64 && size <= 96)
321 return 1;
322 if (KMALLOC_MIN_SIZE <= 64 && size > 128 && size <= 192)
323 return 2;
324 if (size <= 8) return 3;
325 if (size <= 16) return 4;
326 if (size <= 32) return 5;
327 if (size <= 64) return 6;
328 if (size <= 128) return 7;
329 if (size <= 256) return 8;
330 if (size <= 512) return 9;
331 if (size <= 1024) return 10;
332 if (size <= 2 * 1024) return 11;
333 if (size <= 4 * 1024) return 12;
334 if (size <= 8 * 1024) return 13;
335 if (size <= 16 * 1024) return 14;
336 if (size <= 32 * 1024) return 15;
337 if (size <= 64 * 1024) return 16;
338 if (size <= 128 * 1024) return 17;
339 if (size <= 256 * 1024) return 18;
340 if (size <= 512 * 1024) return 19;
341 if (size <= 1024 * 1024) return 20;
342 if (size <= 2 * 1024 * 1024) return 21;
343 if (size <= 4 * 1024 * 1024) return 22;
344 if (size <= 8 * 1024 * 1024) return 23;
345 if (size <= 16 * 1024 * 1024) return 24;
346 if (size <= 32 * 1024 * 1024) return 25;
347 if (size <= 64 * 1024 * 1024) return 26;
348 BUG();
349
350 /* Will never be reached. Needed because the compiler may complain */
351 return -1;
352}
069e2b35 353#endif /* !CONFIG_SLOB */
ce6a5026 354
48a27055
RV
355void *__kmalloc(size_t size, gfp_t flags) __assume_kmalloc_alignment __malloc;
356void *kmem_cache_alloc(struct kmem_cache *, gfp_t flags) __assume_slab_alignment __malloc;
2a4db7eb 357void kmem_cache_free(struct kmem_cache *, void *);
f1b6eb6e 358
484748f0 359/*
9f706d68 360 * Bulk allocation and freeing operations. These are accelerated in an
484748f0
CL
361 * allocator specific way to avoid taking locks repeatedly or building
362 * metadata structures unnecessarily.
363 *
364 * Note that interrupts must be enabled when calling these functions.
365 */
366void kmem_cache_free_bulk(struct kmem_cache *, size_t, void **);
865762a8 367int kmem_cache_alloc_bulk(struct kmem_cache *, gfp_t, size_t, void **);
484748f0 368
ca257195
JDB
369/*
370 * Caller must not use kfree_bulk() on memory not originally allocated
371 * by kmalloc(), because the SLOB allocator cannot handle this.
372 */
373static __always_inline void kfree_bulk(size_t size, void **p)
374{
375 kmem_cache_free_bulk(NULL, size, p);
376}
377
f1b6eb6e 378#ifdef CONFIG_NUMA
48a27055
RV
379void *__kmalloc_node(size_t size, gfp_t flags, int node) __assume_kmalloc_alignment __malloc;
380void *kmem_cache_alloc_node(struct kmem_cache *, gfp_t flags, int node) __assume_slab_alignment __malloc;
f1b6eb6e
CL
381#else
382static __always_inline void *__kmalloc_node(size_t size, gfp_t flags, int node)
383{
384 return __kmalloc(size, flags);
385}
386
387static __always_inline void *kmem_cache_alloc_node(struct kmem_cache *s, gfp_t flags, int node)
388{
389 return kmem_cache_alloc(s, flags);
390}
391#endif
392
393#ifdef CONFIG_TRACING
48a27055 394extern void *kmem_cache_alloc_trace(struct kmem_cache *, gfp_t, size_t) __assume_slab_alignment __malloc;
f1b6eb6e
CL
395
396#ifdef CONFIG_NUMA
397extern void *kmem_cache_alloc_node_trace(struct kmem_cache *s,
398 gfp_t gfpflags,
48a27055 399 int node, size_t size) __assume_slab_alignment __malloc;
f1b6eb6e
CL
400#else
401static __always_inline void *
402kmem_cache_alloc_node_trace(struct kmem_cache *s,
403 gfp_t gfpflags,
404 int node, size_t size)
405{
406 return kmem_cache_alloc_trace(s, gfpflags, size);
407}
408#endif /* CONFIG_NUMA */
409
410#else /* CONFIG_TRACING */
411static __always_inline void *kmem_cache_alloc_trace(struct kmem_cache *s,
412 gfp_t flags, size_t size)
413{
0316bec2
AR
414 void *ret = kmem_cache_alloc(s, flags);
415
505f5dcb 416 kasan_kmalloc(s, ret, size, flags);
0316bec2 417 return ret;
f1b6eb6e
CL
418}
419
420static __always_inline void *
421kmem_cache_alloc_node_trace(struct kmem_cache *s,
422 gfp_t gfpflags,
423 int node, size_t size)
424{
0316bec2
AR
425 void *ret = kmem_cache_alloc_node(s, gfpflags, node);
426
505f5dcb 427 kasan_kmalloc(s, ret, size, gfpflags);
0316bec2 428 return ret;
f1b6eb6e
CL
429}
430#endif /* CONFIG_TRACING */
431
48a27055 432extern void *kmalloc_order(size_t size, gfp_t flags, unsigned int order) __assume_page_alignment __malloc;
f1b6eb6e
CL
433
434#ifdef CONFIG_TRACING
48a27055 435extern void *kmalloc_order_trace(size_t size, gfp_t flags, unsigned int order) __assume_page_alignment __malloc;
f1b6eb6e
CL
436#else
437static __always_inline void *
438kmalloc_order_trace(size_t size, gfp_t flags, unsigned int order)
439{
440 return kmalloc_order(size, flags, order);
441}
ce6a5026
CL
442#endif
443
f1b6eb6e
CL
444static __always_inline void *kmalloc_large(size_t size, gfp_t flags)
445{
446 unsigned int order = get_order(size);
447 return kmalloc_order_trace(size, flags, order);
448}
449
450/**
451 * kmalloc - allocate memory
452 * @size: how many bytes of memory are required.
7e3528c3 453 * @flags: the type of memory to allocate.
f1b6eb6e
CL
454 *
455 * kmalloc is the normal method of allocating memory
456 * for objects smaller than page size in the kernel.
7e3528c3
RD
457 *
458 * The @flags argument may be one of:
459 *
460 * %GFP_USER - Allocate memory on behalf of user. May sleep.
461 *
462 * %GFP_KERNEL - Allocate normal kernel ram. May sleep.
463 *
464 * %GFP_ATOMIC - Allocation will not sleep. May use emergency pools.
465 * For example, use this inside interrupt handlers.
466 *
467 * %GFP_HIGHUSER - Allocate pages from high memory.
468 *
469 * %GFP_NOIO - Do not do any I/O at all while trying to get memory.
470 *
471 * %GFP_NOFS - Do not make any fs calls while trying to get memory.
472 *
473 * %GFP_NOWAIT - Allocation will not sleep.
474 *
e97ca8e5 475 * %__GFP_THISNODE - Allocate node-local memory only.
7e3528c3
RD
476 *
477 * %GFP_DMA - Allocation suitable for DMA.
478 * Should only be used for kmalloc() caches. Otherwise, use a
479 * slab created with SLAB_DMA.
480 *
481 * Also it is possible to set different flags by OR'ing
482 * in one or more of the following additional @flags:
483 *
7e3528c3
RD
484 * %__GFP_HIGH - This allocation has high priority and may use emergency pools.
485 *
486 * %__GFP_NOFAIL - Indicate that this allocation is in no way allowed to fail
487 * (think twice before using).
488 *
489 * %__GFP_NORETRY - If memory is not immediately available,
490 * then give up at once.
491 *
492 * %__GFP_NOWARN - If allocation fails, don't issue any warnings.
493 *
dcda9b04
MH
494 * %__GFP_RETRY_MAYFAIL - Try really hard to succeed the allocation but fail
495 * eventually.
7e3528c3
RD
496 *
497 * There are other flags available as well, but these are not intended
498 * for general use, and so are not documented here. For a full list of
499 * potential flags, always refer to linux/gfp.h.
f1b6eb6e
CL
500 */
501static __always_inline void *kmalloc(size_t size, gfp_t flags)
502{
503 if (__builtin_constant_p(size)) {
504 if (size > KMALLOC_MAX_CACHE_SIZE)
505 return kmalloc_large(size, flags);
506#ifndef CONFIG_SLOB
507 if (!(flags & GFP_DMA)) {
36071a27 508 unsigned int index = kmalloc_index(size);
f1b6eb6e
CL
509
510 if (!index)
511 return ZERO_SIZE_PTR;
512
513 return kmem_cache_alloc_trace(kmalloc_caches[index],
514 flags, size);
515 }
516#endif
517 }
518 return __kmalloc(size, flags);
519}
520
ce6a5026
CL
521/*
522 * Determine size used for the nth kmalloc cache.
523 * return size or 0 if a kmalloc cache for that
524 * size does not exist
525 */
0be70327 526static __always_inline unsigned int kmalloc_size(unsigned int n)
ce6a5026 527{
069e2b35 528#ifndef CONFIG_SLOB
ce6a5026 529 if (n > 2)
0be70327 530 return 1U << n;
ce6a5026
CL
531
532 if (n == 1 && KMALLOC_MIN_SIZE <= 32)
533 return 96;
534
535 if (n == 2 && KMALLOC_MIN_SIZE <= 64)
536 return 192;
069e2b35 537#endif
ce6a5026
CL
538 return 0;
539}
ce6a5026 540
f1b6eb6e
CL
541static __always_inline void *kmalloc_node(size_t size, gfp_t flags, int node)
542{
543#ifndef CONFIG_SLOB
544 if (__builtin_constant_p(size) &&
23774a2f 545 size <= KMALLOC_MAX_CACHE_SIZE && !(flags & GFP_DMA)) {
36071a27 546 unsigned int i = kmalloc_index(size);
f1b6eb6e
CL
547
548 if (!i)
549 return ZERO_SIZE_PTR;
550
551 return kmem_cache_alloc_node_trace(kmalloc_caches[i],
552 flags, node, size);
553 }
554#endif
555 return __kmalloc_node(size, flags, node);
556}
557
f7ce3190
VD
558struct memcg_cache_array {
559 struct rcu_head rcu;
560 struct kmem_cache *entries[0];
561};
562
ba6c496e
GC
563/*
564 * This is the main placeholder for memcg-related information in kmem caches.
ba6c496e
GC
565 * Both the root cache and the child caches will have it. For the root cache,
566 * this will hold a dynamically allocated array large enough to hold
f8570263
VD
567 * information about the currently limited memcgs in the system. To allow the
568 * array to be accessed without taking any locks, on relocation we free the old
569 * version only after a grace period.
ba6c496e 570 *
9eeadc8b 571 * Root and child caches hold different metadata.
ba6c496e 572 *
9eeadc8b
TH
573 * @root_cache: Common to root and child caches. NULL for root, pointer to
574 * the root cache for children.
426589f5 575 *
9eeadc8b
TH
576 * The following fields are specific to root caches.
577 *
578 * @memcg_caches: kmemcg ID indexed table of child caches. This table is
579 * used to index child cachces during allocation and cleared
580 * early during shutdown.
581 *
510ded33
TH
582 * @root_caches_node: List node for slab_root_caches list.
583 *
9eeadc8b
TH
584 * @children: List of all child caches. While the child caches are also
585 * reachable through @memcg_caches, a child cache remains on
586 * this list until it is actually destroyed.
587 *
588 * The following fields are specific to child caches.
589 *
590 * @memcg: Pointer to the memcg this cache belongs to.
591 *
592 * @children_node: List node for @root_cache->children list.
bc2791f8
TH
593 *
594 * @kmem_caches_node: List node for @memcg->kmem_caches list.
ba6c496e
GC
595 */
596struct memcg_cache_params {
9eeadc8b 597 struct kmem_cache *root_cache;
ba6c496e 598 union {
9eeadc8b
TH
599 struct {
600 struct memcg_cache_array __rcu *memcg_caches;
510ded33 601 struct list_head __root_caches_node;
9eeadc8b 602 struct list_head children;
92ee383f 603 bool dying;
9eeadc8b 604 };
2633d7a0
GC
605 struct {
606 struct mem_cgroup *memcg;
9eeadc8b 607 struct list_head children_node;
bc2791f8 608 struct list_head kmem_caches_node;
01fb58bc
TH
609
610 void (*deact_fn)(struct kmem_cache *);
611 union {
612 struct rcu_head deact_rcu_head;
613 struct work_struct deact_work;
614 };
2633d7a0 615 };
ba6c496e
GC
616 };
617};
618
2633d7a0
GC
619int memcg_update_all_caches(int num_memcgs);
620
e7efa615
MO
621/**
622 * kmalloc_array - allocate memory for an array.
623 * @n: number of elements.
624 * @size: element size.
625 * @flags: the type of memory to allocate (see kmalloc).
800590f5 626 */
a8203725 627static inline void *kmalloc_array(size_t n, size_t size, gfp_t flags)
1da177e4 628{
49b7f898
KC
629 size_t bytes;
630
631 if (unlikely(check_mul_overflow(n, size, &bytes)))
6193a2ff 632 return NULL;
91c6a05f 633 if (__builtin_constant_p(n) && __builtin_constant_p(size))
49b7f898
KC
634 return kmalloc(bytes, flags);
635 return __kmalloc(bytes, flags);
a8203725
XW
636}
637
638/**
639 * kcalloc - allocate memory for an array. The memory is set to zero.
640 * @n: number of elements.
641 * @size: element size.
642 * @flags: the type of memory to allocate (see kmalloc).
643 */
644static inline void *kcalloc(size_t n, size_t size, gfp_t flags)
645{
646 return kmalloc_array(n, size, flags | __GFP_ZERO);
1da177e4
LT
647}
648
1d2c8eea
CH
649/*
650 * kmalloc_track_caller is a special version of kmalloc that records the
651 * calling function of the routine calling it for slab leak tracking instead
652 * of just the calling function (confusing, eh?).
653 * It's useful when the call to kmalloc comes from a widely-used standard
654 * allocator where we care about the real place the memory allocation
655 * request comes from.
656 */
ce71e27c 657extern void *__kmalloc_track_caller(size_t, gfp_t, unsigned long);
1d2c8eea 658#define kmalloc_track_caller(size, flags) \
ce71e27c 659 __kmalloc_track_caller(size, flags, _RET_IP_)
1da177e4 660
5799b255
JT
661static inline void *kmalloc_array_node(size_t n, size_t size, gfp_t flags,
662 int node)
663{
49b7f898
KC
664 size_t bytes;
665
666 if (unlikely(check_mul_overflow(n, size, &bytes)))
5799b255
JT
667 return NULL;
668 if (__builtin_constant_p(n) && __builtin_constant_p(size))
49b7f898
KC
669 return kmalloc_node(bytes, flags, node);
670 return __kmalloc_node(bytes, flags, node);
5799b255
JT
671}
672
673static inline void *kcalloc_node(size_t n, size_t size, gfp_t flags, int node)
674{
675 return kmalloc_array_node(n, size, flags | __GFP_ZERO, node);
676}
677
678
97e2bde4 679#ifdef CONFIG_NUMA
ce71e27c 680extern void *__kmalloc_node_track_caller(size_t, gfp_t, int, unsigned long);
8b98c169
CH
681#define kmalloc_node_track_caller(size, flags, node) \
682 __kmalloc_node_track_caller(size, flags, node, \
ce71e27c 683 _RET_IP_)
2e892f43 684
8b98c169 685#else /* CONFIG_NUMA */
8b98c169
CH
686
687#define kmalloc_node_track_caller(size, flags, node) \
688 kmalloc_track_caller(size, flags)
97e2bde4 689
dfcd3610 690#endif /* CONFIG_NUMA */
10cef602 691
81cda662
CL
692/*
693 * Shortcuts
694 */
695static inline void *kmem_cache_zalloc(struct kmem_cache *k, gfp_t flags)
696{
697 return kmem_cache_alloc(k, flags | __GFP_ZERO);
698}
699
700/**
701 * kzalloc - allocate memory. The memory is set to zero.
702 * @size: how many bytes of memory are required.
703 * @flags: the type of memory to allocate (see kmalloc).
704 */
705static inline void *kzalloc(size_t size, gfp_t flags)
706{
707 return kmalloc(size, flags | __GFP_ZERO);
708}
709
979b0fea
JL
710/**
711 * kzalloc_node - allocate zeroed memory from a particular memory node.
712 * @size: how many bytes of memory are required.
713 * @flags: the type of memory to allocate (see kmalloc).
714 * @node: memory node from which to allocate
715 */
716static inline void *kzalloc_node(size_t size, gfp_t flags, int node)
717{
718 return kmalloc_node(size, flags | __GFP_ZERO, node);
719}
720
07f361b2 721unsigned int kmem_cache_size(struct kmem_cache *s);
7e85ee0c
PE
722void __init kmem_cache_init_late(void);
723
6731d4f1
SAS
724#if defined(CONFIG_SMP) && defined(CONFIG_SLAB)
725int slab_prepare_cpu(unsigned int cpu);
726int slab_dead_cpu(unsigned int cpu);
727#else
728#define slab_prepare_cpu NULL
729#define slab_dead_cpu NULL
730#endif
731
1da177e4 732#endif /* _LINUX_SLAB_H */