Commit | Line | Data |
---|---|---|
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 | 18 | #include <linux/workqueue.h> |
f0a3a24b | 19 | #include <linux/percpu-refcount.h> |
1f458cbf | 20 | |
1da177e4 | 21 | |
2e892f43 CL |
22 | /* |
23 | * Flags to pass to kmem_cache_create(). | |
124dee09 | 24 | * The ones marked DEBUG are only valid if CONFIG_DEBUG_SLAB is set. |
1da177e4 | 25 | */ |
d50112ed | 26 | /* DEBUG: Perform (expensive) checks on alloc/free */ |
4fd0b46e | 27 | #define SLAB_CONSISTENCY_CHECKS ((slab_flags_t __force)0x00000100U) |
d50112ed | 28 | /* DEBUG: Red zone objs in a cache */ |
4fd0b46e | 29 | #define SLAB_RED_ZONE ((slab_flags_t __force)0x00000400U) |
d50112ed | 30 | /* DEBUG: Poison objects */ |
4fd0b46e | 31 | #define SLAB_POISON ((slab_flags_t __force)0x00000800U) |
d50112ed | 32 | /* Align objs on cache lines */ |
4fd0b46e | 33 | #define SLAB_HWCACHE_ALIGN ((slab_flags_t __force)0x00002000U) |
d50112ed | 34 | /* Use GFP_DMA memory */ |
4fd0b46e | 35 | #define SLAB_CACHE_DMA ((slab_flags_t __force)0x00004000U) |
6d6ea1e9 NB |
36 | /* Use GFP_DMA32 memory */ |
37 | #define SLAB_CACHE_DMA32 ((slab_flags_t __force)0x00008000U) | |
d50112ed | 38 | /* DEBUG: Store the last owner for bug hunting */ |
4fd0b46e | 39 | #define SLAB_STORE_USER ((slab_flags_t __force)0x00010000U) |
d50112ed | 40 | /* Panic if kmem_cache_create() fails */ |
4fd0b46e | 41 | #define SLAB_PANIC ((slab_flags_t __force)0x00040000U) |
d7de4c1d | 42 | /* |
5f0d5a3a | 43 | * SLAB_TYPESAFE_BY_RCU - **WARNING** READ THIS! |
d7de4c1d PZ |
44 | * |
45 | * This delays freeing the SLAB page by a grace period, it does _NOT_ | |
46 | * delay object freeing. This means that if you do kmem_cache_free() | |
47 | * that memory location is free to be reused at any time. Thus it may | |
48 | * be possible to see another object there in the same RCU grace period. | |
49 | * | |
50 | * This feature only ensures the memory location backing the object | |
51 | * stays valid, the trick to using this is relying on an independent | |
52 | * object validation pass. Something like: | |
53 | * | |
54 | * rcu_read_lock() | |
55 | * again: | |
56 | * obj = lockless_lookup(key); | |
57 | * if (obj) { | |
58 | * if (!try_get_ref(obj)) // might fail for free objects | |
59 | * goto again; | |
60 | * | |
61 | * if (obj->key != key) { // not the object we expected | |
62 | * put_ref(obj); | |
63 | * goto again; | |
64 | * } | |
65 | * } | |
66 | * rcu_read_unlock(); | |
67 | * | |
68126702 JK |
68 | * This is useful if we need to approach a kernel structure obliquely, |
69 | * from its address obtained without the usual locking. We can lock | |
70 | * the structure to stabilize it and check it's still at the given address, | |
71 | * only if we can be sure that the memory has not been meanwhile reused | |
72 | * for some other kind of object (which our subsystem's lock might corrupt). | |
73 | * | |
74 | * rcu_read_lock before reading the address, then rcu_read_unlock after | |
75 | * taking the spinlock within the structure expected at that address. | |
5f0d5a3a PM |
76 | * |
77 | * Note that SLAB_TYPESAFE_BY_RCU was originally named SLAB_DESTROY_BY_RCU. | |
d7de4c1d | 78 | */ |
d50112ed | 79 | /* Defer freeing slabs to RCU */ |
4fd0b46e | 80 | #define SLAB_TYPESAFE_BY_RCU ((slab_flags_t __force)0x00080000U) |
d50112ed | 81 | /* Spread some memory over cpuset */ |
4fd0b46e | 82 | #define SLAB_MEM_SPREAD ((slab_flags_t __force)0x00100000U) |
d50112ed | 83 | /* Trace allocations and frees */ |
4fd0b46e | 84 | #define SLAB_TRACE ((slab_flags_t __force)0x00200000U) |
1da177e4 | 85 | |
30327acf TG |
86 | /* Flag to prevent checks on free */ |
87 | #ifdef CONFIG_DEBUG_OBJECTS | |
4fd0b46e | 88 | # define SLAB_DEBUG_OBJECTS ((slab_flags_t __force)0x00400000U) |
30327acf | 89 | #else |
4fd0b46e | 90 | # define SLAB_DEBUG_OBJECTS 0 |
30327acf TG |
91 | #endif |
92 | ||
d50112ed | 93 | /* Avoid kmemleak tracing */ |
4fd0b46e | 94 | #define SLAB_NOLEAKTRACE ((slab_flags_t __force)0x00800000U) |
d5cff635 | 95 | |
d50112ed | 96 | /* Fault injection mark */ |
4c13dd3b | 97 | #ifdef CONFIG_FAILSLAB |
4fd0b46e | 98 | # define SLAB_FAILSLAB ((slab_flags_t __force)0x02000000U) |
4c13dd3b | 99 | #else |
4fd0b46e | 100 | # define SLAB_FAILSLAB 0 |
4c13dd3b | 101 | #endif |
d50112ed | 102 | /* Account to memcg */ |
84c07d11 | 103 | #ifdef CONFIG_MEMCG_KMEM |
4fd0b46e | 104 | # define SLAB_ACCOUNT ((slab_flags_t __force)0x04000000U) |
230e9fc2 | 105 | #else |
4fd0b46e | 106 | # define SLAB_ACCOUNT 0 |
230e9fc2 | 107 | #endif |
2dff4405 | 108 | |
7ed2f9e6 | 109 | #ifdef CONFIG_KASAN |
4fd0b46e | 110 | #define SLAB_KASAN ((slab_flags_t __force)0x08000000U) |
7ed2f9e6 | 111 | #else |
4fd0b46e | 112 | #define SLAB_KASAN 0 |
7ed2f9e6 AP |
113 | #endif |
114 | ||
a285909f HY |
115 | /* |
116 | * Ignore user specified debugging flags. | |
117 | * Intended for caches created for self-tests so they have only flags | |
118 | * specified in the code and other flags are ignored. | |
119 | */ | |
120 | #define SLAB_NO_USER_FLAGS ((slab_flags_t __force)0x10000000U) | |
121 | ||
e12ba74d | 122 | /* The following flags affect the page allocator grouping pages by mobility */ |
d50112ed | 123 | /* Objects are reclaimable */ |
4fd0b46e | 124 | #define SLAB_RECLAIM_ACCOUNT ((slab_flags_t __force)0x00020000U) |
e12ba74d | 125 | #define SLAB_TEMPORARY SLAB_RECLAIM_ACCOUNT /* Objects are short-lived */ |
fcf8a1e4 | 126 | |
6cb8f913 CL |
127 | /* |
128 | * ZERO_SIZE_PTR will be returned for zero sized kmalloc requests. | |
129 | * | |
130 | * Dereferencing ZERO_SIZE_PTR will lead to a distinct access fault. | |
131 | * | |
132 | * ZERO_SIZE_PTR can be passed to kfree though in the same way that NULL can. | |
133 | * Both make kfree a no-op. | |
134 | */ | |
135 | #define ZERO_SIZE_PTR ((void *)16) | |
136 | ||
1d4ec7b1 | 137 | #define ZERO_OR_NULL_PTR(x) ((unsigned long)(x) <= \ |
6cb8f913 CL |
138 | (unsigned long)ZERO_SIZE_PTR) |
139 | ||
0316bec2 | 140 | #include <linux/kasan.h> |
3b0efdfa | 141 | |
88f2ef73 | 142 | struct list_lru; |
2633d7a0 | 143 | struct mem_cgroup; |
2e892f43 CL |
144 | /* |
145 | * struct kmem_cache related prototypes | |
146 | */ | |
147 | void __init kmem_cache_init(void); | |
fda90124 | 148 | bool slab_is_available(void); |
1da177e4 | 149 | |
f4957d5b AD |
150 | struct kmem_cache *kmem_cache_create(const char *name, unsigned int size, |
151 | unsigned int align, slab_flags_t flags, | |
8eb8284b DW |
152 | void (*ctor)(void *)); |
153 | struct kmem_cache *kmem_cache_create_usercopy(const char *name, | |
f4957d5b AD |
154 | unsigned int size, unsigned int align, |
155 | slab_flags_t flags, | |
7bbdb81e | 156 | unsigned int useroffset, unsigned int usersize, |
8eb8284b | 157 | void (*ctor)(void *)); |
72d67229 KC |
158 | void kmem_cache_destroy(struct kmem_cache *s); |
159 | int kmem_cache_shrink(struct kmem_cache *s); | |
2a4db7eb | 160 | |
0a31bd5f CL |
161 | /* |
162 | * Please use this macro to create slab caches. Simply specify the | |
163 | * name of the structure and maybe some flags that are listed above. | |
164 | * | |
165 | * The alignment of the struct determines object alignment. If you | |
166 | * f.e. add ____cacheline_aligned_in_smp to the struct declaration | |
167 | * then the objects will be properly aligned in SMP configurations. | |
168 | */ | |
8eb8284b DW |
169 | #define KMEM_CACHE(__struct, __flags) \ |
170 | kmem_cache_create(#__struct, sizeof(struct __struct), \ | |
171 | __alignof__(struct __struct), (__flags), NULL) | |
172 | ||
173 | /* | |
174 | * To whitelist a single field for copying to/from usercopy, use this | |
175 | * macro instead for KMEM_CACHE() above. | |
176 | */ | |
177 | #define KMEM_CACHE_USERCOPY(__struct, __flags, __field) \ | |
178 | kmem_cache_create_usercopy(#__struct, \ | |
179 | sizeof(struct __struct), \ | |
180 | __alignof__(struct __struct), (__flags), \ | |
181 | offsetof(struct __struct, __field), \ | |
182 | sizeof_field(struct __struct, __field), NULL) | |
0a31bd5f | 183 | |
34504667 CL |
184 | /* |
185 | * Common kmalloc functions provided by all allocators | |
186 | */ | |
c37495d6 | 187 | void * __must_check krealloc(const void *objp, size_t new_size, gfp_t flags) __alloc_size(2); |
72d67229 KC |
188 | void kfree(const void *objp); |
189 | void kfree_sensitive(const void *objp); | |
190 | size_t __ksize(const void *objp); | |
191 | size_t ksize(const void *objp); | |
5bb1bb35 | 192 | #ifdef CONFIG_PRINTK |
8e7f37f2 PM |
193 | bool kmem_valid_obj(void *object); |
194 | void kmem_dump_obj(void *object); | |
5bb1bb35 | 195 | #endif |
34504667 | 196 | |
c601fd69 CL |
197 | /* |
198 | * Some archs want to perform DMA into kmalloc caches and need a guaranteed | |
199 | * alignment larger than the alignment of a 64-bit integer. | |
8cf9e121 | 200 | * Setting ARCH_DMA_MINALIGN in arch headers allows that. |
c601fd69 CL |
201 | */ |
202 | #if defined(ARCH_DMA_MINALIGN) && ARCH_DMA_MINALIGN > 8 | |
203 | #define ARCH_KMALLOC_MINALIGN ARCH_DMA_MINALIGN | |
204 | #define KMALLOC_MIN_SIZE ARCH_DMA_MINALIGN | |
205 | #define KMALLOC_SHIFT_LOW ilog2(ARCH_DMA_MINALIGN) | |
206 | #else | |
207 | #define ARCH_KMALLOC_MINALIGN __alignof__(unsigned long long) | |
208 | #endif | |
209 | ||
94a58c36 RV |
210 | /* |
211 | * Setting ARCH_SLAB_MINALIGN in arch headers allows a different alignment. | |
212 | * Intended for arches that get misalignment faults even for 64 bit integer | |
213 | * aligned buffers. | |
214 | */ | |
215 | #ifndef ARCH_SLAB_MINALIGN | |
216 | #define ARCH_SLAB_MINALIGN __alignof__(unsigned long long) | |
217 | #endif | |
218 | ||
d949a815 PC |
219 | /* |
220 | * Arches can define this function if they want to decide the minimum slab | |
221 | * alignment at runtime. The value returned by the function must be a power | |
222 | * of two and >= ARCH_SLAB_MINALIGN. | |
223 | */ | |
224 | #ifndef arch_slab_minalign | |
225 | static inline unsigned int arch_slab_minalign(void) | |
226 | { | |
227 | return ARCH_SLAB_MINALIGN; | |
228 | } | |
229 | #endif | |
230 | ||
94a58c36 | 231 | /* |
154036a3 AK |
232 | * kmem_cache_alloc and friends return pointers aligned to ARCH_SLAB_MINALIGN. |
233 | * kmalloc and friends return pointers aligned to both ARCH_KMALLOC_MINALIGN | |
234 | * and ARCH_SLAB_MINALIGN, but here we only assume the former alignment. | |
94a58c36 RV |
235 | */ |
236 | #define __assume_kmalloc_alignment __assume_aligned(ARCH_KMALLOC_MINALIGN) | |
237 | #define __assume_slab_alignment __assume_aligned(ARCH_SLAB_MINALIGN) | |
238 | #define __assume_page_alignment __assume_aligned(PAGE_SIZE) | |
239 | ||
0aa817f0 | 240 | /* |
95a05b42 CL |
241 | * Kmalloc array related definitions |
242 | */ | |
243 | ||
244 | #ifdef CONFIG_SLAB | |
245 | /* | |
246 | * The largest kmalloc size supported by the SLAB allocators is | |
0aa817f0 CL |
247 | * 32 megabyte (2^25) or the maximum allocatable page order if that is |
248 | * less than 32 MB. | |
249 | * | |
250 | * WARNING: Its not easy to increase this value since the allocators have | |
251 | * to do various tricks to work around compiler limitations in order to | |
252 | * ensure proper constant folding. | |
253 | */ | |
debee076 CL |
254 | #define KMALLOC_SHIFT_HIGH ((MAX_ORDER + PAGE_SHIFT - 1) <= 25 ? \ |
255 | (MAX_ORDER + PAGE_SHIFT - 1) : 25) | |
95a05b42 | 256 | #define KMALLOC_SHIFT_MAX KMALLOC_SHIFT_HIGH |
c601fd69 | 257 | #ifndef KMALLOC_SHIFT_LOW |
95a05b42 | 258 | #define KMALLOC_SHIFT_LOW 5 |
c601fd69 | 259 | #endif |
069e2b35 CL |
260 | #endif |
261 | ||
262 | #ifdef CONFIG_SLUB | |
95a05b42 | 263 | /* |
433a91ff DH |
264 | * SLUB directly allocates requests fitting in to an order-1 page |
265 | * (PAGE_SIZE*2). Larger requests are passed to the page allocator. | |
95a05b42 CL |
266 | */ |
267 | #define KMALLOC_SHIFT_HIGH (PAGE_SHIFT + 1) | |
bb1107f7 | 268 | #define KMALLOC_SHIFT_MAX (MAX_ORDER + PAGE_SHIFT - 1) |
c601fd69 | 269 | #ifndef KMALLOC_SHIFT_LOW |
95a05b42 CL |
270 | #define KMALLOC_SHIFT_LOW 3 |
271 | #endif | |
c601fd69 | 272 | #endif |
0aa817f0 | 273 | |
069e2b35 CL |
274 | #ifdef CONFIG_SLOB |
275 | /* | |
433a91ff | 276 | * SLOB passes all requests larger than one page to the page allocator. |
069e2b35 CL |
277 | * No kmalloc array is necessary since objects of different sizes can |
278 | * be allocated from the same page. | |
279 | */ | |
069e2b35 | 280 | #define KMALLOC_SHIFT_HIGH PAGE_SHIFT |
bb1107f7 | 281 | #define KMALLOC_SHIFT_MAX (MAX_ORDER + PAGE_SHIFT - 1) |
069e2b35 CL |
282 | #ifndef KMALLOC_SHIFT_LOW |
283 | #define KMALLOC_SHIFT_LOW 3 | |
284 | #endif | |
285 | #endif | |
286 | ||
95a05b42 CL |
287 | /* Maximum allocatable size */ |
288 | #define KMALLOC_MAX_SIZE (1UL << KMALLOC_SHIFT_MAX) | |
289 | /* Maximum size for which we actually use a slab cache */ | |
290 | #define KMALLOC_MAX_CACHE_SIZE (1UL << KMALLOC_SHIFT_HIGH) | |
d7cff4de | 291 | /* Maximum order allocatable via the slab allocator */ |
95a05b42 | 292 | #define KMALLOC_MAX_ORDER (KMALLOC_SHIFT_MAX - PAGE_SHIFT) |
0aa817f0 | 293 | |
ce6a5026 CL |
294 | /* |
295 | * Kmalloc subsystem. | |
296 | */ | |
c601fd69 | 297 | #ifndef KMALLOC_MIN_SIZE |
95a05b42 | 298 | #define KMALLOC_MIN_SIZE (1 << KMALLOC_SHIFT_LOW) |
ce6a5026 CL |
299 | #endif |
300 | ||
24f870d8 JK |
301 | /* |
302 | * This restriction comes from byte sized index implementation. | |
303 | * Page size is normally 2^12 bytes and, in this case, if we want to use | |
304 | * byte sized index which can represent 2^8 entries, the size of the object | |
305 | * should be equal or greater to 2^12 / 2^8 = 2^4 = 16. | |
306 | * If minimum size of kmalloc is less than 16, we use it as minimum object | |
307 | * size and give up to use byte sized index. | |
308 | */ | |
309 | #define SLAB_OBJ_MIN_SIZE (KMALLOC_MIN_SIZE < 16 ? \ | |
310 | (KMALLOC_MIN_SIZE) : 16) | |
311 | ||
1291523f VB |
312 | /* |
313 | * Whenever changing this, take care of that kmalloc_type() and | |
314 | * create_kmalloc_caches() still work as intended. | |
494c1dfe WL |
315 | * |
316 | * KMALLOC_NORMAL can contain only unaccounted objects whereas KMALLOC_CGROUP | |
317 | * is for accounted but unreclaimable and non-dma objects. All the other | |
318 | * kmem caches can have both accounted and unaccounted objects. | |
1291523f | 319 | */ |
cc252eae VB |
320 | enum kmalloc_cache_type { |
321 | KMALLOC_NORMAL = 0, | |
494c1dfe WL |
322 | #ifndef CONFIG_ZONE_DMA |
323 | KMALLOC_DMA = KMALLOC_NORMAL, | |
324 | #endif | |
325 | #ifndef CONFIG_MEMCG_KMEM | |
326 | KMALLOC_CGROUP = KMALLOC_NORMAL, | |
327 | #else | |
328 | KMALLOC_CGROUP, | |
329 | #endif | |
1291523f | 330 | KMALLOC_RECLAIM, |
cc252eae VB |
331 | #ifdef CONFIG_ZONE_DMA |
332 | KMALLOC_DMA, | |
333 | #endif | |
334 | NR_KMALLOC_TYPES | |
335 | }; | |
336 | ||
069e2b35 | 337 | #ifndef CONFIG_SLOB |
cc252eae VB |
338 | extern struct kmem_cache * |
339 | kmalloc_caches[NR_KMALLOC_TYPES][KMALLOC_SHIFT_HIGH + 1]; | |
340 | ||
494c1dfe WL |
341 | /* |
342 | * Define gfp bits that should not be set for KMALLOC_NORMAL. | |
343 | */ | |
344 | #define KMALLOC_NOT_NORMAL_BITS \ | |
345 | (__GFP_RECLAIMABLE | \ | |
346 | (IS_ENABLED(CONFIG_ZONE_DMA) ? __GFP_DMA : 0) | \ | |
347 | (IS_ENABLED(CONFIG_MEMCG_KMEM) ? __GFP_ACCOUNT : 0)) | |
348 | ||
cc252eae VB |
349 | static __always_inline enum kmalloc_cache_type kmalloc_type(gfp_t flags) |
350 | { | |
4e45f712 VB |
351 | /* |
352 | * The most common case is KMALLOC_NORMAL, so test for it | |
494c1dfe | 353 | * with a single branch for all the relevant flags. |
4e45f712 | 354 | */ |
494c1dfe | 355 | if (likely((flags & KMALLOC_NOT_NORMAL_BITS) == 0)) |
4e45f712 | 356 | return KMALLOC_NORMAL; |
1291523f VB |
357 | |
358 | /* | |
494c1dfe WL |
359 | * At least one of the flags has to be set. Their priorities in |
360 | * decreasing order are: | |
361 | * 1) __GFP_DMA | |
362 | * 2) __GFP_RECLAIMABLE | |
363 | * 3) __GFP_ACCOUNT | |
1291523f | 364 | */ |
494c1dfe WL |
365 | if (IS_ENABLED(CONFIG_ZONE_DMA) && (flags & __GFP_DMA)) |
366 | return KMALLOC_DMA; | |
367 | if (!IS_ENABLED(CONFIG_MEMCG_KMEM) || (flags & __GFP_RECLAIMABLE)) | |
368 | return KMALLOC_RECLAIM; | |
369 | else | |
370 | return KMALLOC_CGROUP; | |
cc252eae VB |
371 | } |
372 | ||
ce6a5026 CL |
373 | /* |
374 | * Figure out which kmalloc slab an allocation of a certain size | |
375 | * belongs to. | |
376 | * 0 = zero alloc | |
377 | * 1 = 65 .. 96 bytes | |
1ed58b60 RV |
378 | * 2 = 129 .. 192 bytes |
379 | * n = 2^(n-1)+1 .. 2^n | |
588c7fa0 HY |
380 | * |
381 | * Note: __kmalloc_index() is compile-time optimized, and not runtime optimized; | |
382 | * typical usage is via kmalloc_index() and therefore evaluated at compile-time. | |
383 | * Callers where !size_is_constant should only be test modules, where runtime | |
384 | * overheads of __kmalloc_index() can be tolerated. Also see kmalloc_slab(). | |
ce6a5026 | 385 | */ |
588c7fa0 HY |
386 | static __always_inline unsigned int __kmalloc_index(size_t size, |
387 | bool size_is_constant) | |
ce6a5026 CL |
388 | { |
389 | if (!size) | |
390 | return 0; | |
391 | ||
392 | if (size <= KMALLOC_MIN_SIZE) | |
393 | return KMALLOC_SHIFT_LOW; | |
394 | ||
395 | if (KMALLOC_MIN_SIZE <= 32 && size > 64 && size <= 96) | |
396 | return 1; | |
397 | if (KMALLOC_MIN_SIZE <= 64 && size > 128 && size <= 192) | |
398 | return 2; | |
399 | if (size <= 8) return 3; | |
400 | if (size <= 16) return 4; | |
401 | if (size <= 32) return 5; | |
402 | if (size <= 64) return 6; | |
403 | if (size <= 128) return 7; | |
404 | if (size <= 256) return 8; | |
405 | if (size <= 512) return 9; | |
406 | if (size <= 1024) return 10; | |
407 | if (size <= 2 * 1024) return 11; | |
408 | if (size <= 4 * 1024) return 12; | |
409 | if (size <= 8 * 1024) return 13; | |
410 | if (size <= 16 * 1024) return 14; | |
411 | if (size <= 32 * 1024) return 15; | |
412 | if (size <= 64 * 1024) return 16; | |
413 | if (size <= 128 * 1024) return 17; | |
414 | if (size <= 256 * 1024) return 18; | |
415 | if (size <= 512 * 1024) return 19; | |
416 | if (size <= 1024 * 1024) return 20; | |
417 | if (size <= 2 * 1024 * 1024) return 21; | |
418 | if (size <= 4 * 1024 * 1024) return 22; | |
419 | if (size <= 8 * 1024 * 1024) return 23; | |
420 | if (size <= 16 * 1024 * 1024) return 24; | |
421 | if (size <= 32 * 1024 * 1024) return 25; | |
588c7fa0 | 422 | |
57b2b72a | 423 | if (!IS_ENABLED(CONFIG_PROFILE_ALL_BRANCHES) && size_is_constant) |
588c7fa0 HY |
424 | BUILD_BUG_ON_MSG(1, "unexpected size in kmalloc_index()"); |
425 | else | |
426 | BUG(); | |
ce6a5026 CL |
427 | |
428 | /* Will never be reached. Needed because the compiler may complain */ | |
429 | return -1; | |
430 | } | |
588c7fa0 | 431 | #define kmalloc_index(s) __kmalloc_index(s, true) |
069e2b35 | 432 | #endif /* !CONFIG_SLOB */ |
ce6a5026 | 433 | |
c37495d6 | 434 | void *__kmalloc(size_t size, gfp_t flags) __assume_kmalloc_alignment __alloc_size(1); |
72d67229 | 435 | void *kmem_cache_alloc(struct kmem_cache *s, gfp_t flags) __assume_slab_alignment __malloc; |
88f2ef73 MS |
436 | void *kmem_cache_alloc_lru(struct kmem_cache *s, struct list_lru *lru, |
437 | gfp_t gfpflags) __assume_slab_alignment __malloc; | |
72d67229 | 438 | void kmem_cache_free(struct kmem_cache *s, void *objp); |
f1b6eb6e | 439 | |
484748f0 | 440 | /* |
9f706d68 | 441 | * Bulk allocation and freeing operations. These are accelerated in an |
484748f0 CL |
442 | * allocator specific way to avoid taking locks repeatedly or building |
443 | * metadata structures unnecessarily. | |
444 | * | |
445 | * Note that interrupts must be enabled when calling these functions. | |
446 | */ | |
72d67229 KC |
447 | void kmem_cache_free_bulk(struct kmem_cache *s, size_t size, void **p); |
448 | int kmem_cache_alloc_bulk(struct kmem_cache *s, gfp_t flags, size_t size, void **p); | |
484748f0 | 449 | |
ca257195 JDB |
450 | /* |
451 | * Caller must not use kfree_bulk() on memory not originally allocated | |
452 | * by kmalloc(), because the SLOB allocator cannot handle this. | |
453 | */ | |
454 | static __always_inline void kfree_bulk(size_t size, void **p) | |
455 | { | |
456 | kmem_cache_free_bulk(NULL, size, p); | |
457 | } | |
458 | ||
f1b6eb6e | 459 | #ifdef CONFIG_NUMA |
c37495d6 KC |
460 | void *__kmalloc_node(size_t size, gfp_t flags, int node) __assume_kmalloc_alignment |
461 | __alloc_size(1); | |
72d67229 KC |
462 | void *kmem_cache_alloc_node(struct kmem_cache *s, gfp_t flags, int node) __assume_slab_alignment |
463 | __malloc; | |
f1b6eb6e | 464 | #else |
c37495d6 | 465 | static __always_inline __alloc_size(1) void *__kmalloc_node(size_t size, gfp_t flags, int node) |
f1b6eb6e CL |
466 | { |
467 | return __kmalloc(size, flags); | |
468 | } | |
469 | ||
470 | static __always_inline void *kmem_cache_alloc_node(struct kmem_cache *s, gfp_t flags, int node) | |
471 | { | |
472 | return kmem_cache_alloc(s, flags); | |
473 | } | |
474 | #endif | |
475 | ||
476 | #ifdef CONFIG_TRACING | |
72d67229 | 477 | extern void *kmem_cache_alloc_trace(struct kmem_cache *s, gfp_t flags, size_t size) |
c37495d6 | 478 | __assume_slab_alignment __alloc_size(3); |
f1b6eb6e CL |
479 | |
480 | #ifdef CONFIG_NUMA | |
72d67229 | 481 | extern void *kmem_cache_alloc_node_trace(struct kmem_cache *s, gfp_t gfpflags, |
c37495d6 KC |
482 | int node, size_t size) __assume_slab_alignment |
483 | __alloc_size(4); | |
f1b6eb6e | 484 | #else |
c37495d6 KC |
485 | static __always_inline __alloc_size(4) void *kmem_cache_alloc_node_trace(struct kmem_cache *s, |
486 | gfp_t gfpflags, int node, size_t size) | |
f1b6eb6e CL |
487 | { |
488 | return kmem_cache_alloc_trace(s, gfpflags, size); | |
489 | } | |
490 | #endif /* CONFIG_NUMA */ | |
491 | ||
492 | #else /* CONFIG_TRACING */ | |
c37495d6 KC |
493 | static __always_inline __alloc_size(3) void *kmem_cache_alloc_trace(struct kmem_cache *s, |
494 | gfp_t flags, size_t size) | |
f1b6eb6e | 495 | { |
0316bec2 AR |
496 | void *ret = kmem_cache_alloc(s, flags); |
497 | ||
0116523c | 498 | ret = kasan_kmalloc(s, ret, size, flags); |
0316bec2 | 499 | return ret; |
f1b6eb6e CL |
500 | } |
501 | ||
72d67229 KC |
502 | static __always_inline void *kmem_cache_alloc_node_trace(struct kmem_cache *s, gfp_t gfpflags, |
503 | int node, size_t size) | |
f1b6eb6e | 504 | { |
0316bec2 AR |
505 | void *ret = kmem_cache_alloc_node(s, gfpflags, node); |
506 | ||
0116523c | 507 | ret = kasan_kmalloc(s, ret, size, gfpflags); |
0316bec2 | 508 | return ret; |
f1b6eb6e CL |
509 | } |
510 | #endif /* CONFIG_TRACING */ | |
511 | ||
72d67229 | 512 | extern void *kmalloc_order(size_t size, gfp_t flags, unsigned int order) __assume_page_alignment |
c37495d6 | 513 | __alloc_size(1); |
f1b6eb6e CL |
514 | |
515 | #ifdef CONFIG_TRACING | |
72d67229 | 516 | extern void *kmalloc_order_trace(size_t size, gfp_t flags, unsigned int order) |
c37495d6 | 517 | __assume_page_alignment __alloc_size(1); |
f1b6eb6e | 518 | #else |
c37495d6 KC |
519 | static __always_inline __alloc_size(1) void *kmalloc_order_trace(size_t size, gfp_t flags, |
520 | unsigned int order) | |
f1b6eb6e CL |
521 | { |
522 | return kmalloc_order(size, flags, order); | |
523 | } | |
ce6a5026 CL |
524 | #endif |
525 | ||
c37495d6 | 526 | static __always_inline __alloc_size(1) void *kmalloc_large(size_t size, gfp_t flags) |
f1b6eb6e CL |
527 | { |
528 | unsigned int order = get_order(size); | |
529 | return kmalloc_order_trace(size, flags, order); | |
530 | } | |
531 | ||
532 | /** | |
533 | * kmalloc - allocate memory | |
534 | * @size: how many bytes of memory are required. | |
7e3528c3 | 535 | * @flags: the type of memory to allocate. |
f1b6eb6e CL |
536 | * |
537 | * kmalloc is the normal method of allocating memory | |
538 | * for objects smaller than page size in the kernel. | |
7e3528c3 | 539 | * |
59bb4798 VB |
540 | * The allocated object address is aligned to at least ARCH_KMALLOC_MINALIGN |
541 | * bytes. For @size of power of two bytes, the alignment is also guaranteed | |
542 | * to be at least to the size. | |
543 | * | |
01598ba6 MR |
544 | * The @flags argument may be one of the GFP flags defined at |
545 | * include/linux/gfp.h and described at | |
546 | * :ref:`Documentation/core-api/mm-api.rst <mm-api-gfp-flags>` | |
7e3528c3 | 547 | * |
01598ba6 | 548 | * The recommended usage of the @flags is described at |
2370ae4b | 549 | * :ref:`Documentation/core-api/memory-allocation.rst <memory_allocation>` |
7e3528c3 | 550 | * |
01598ba6 | 551 | * Below is a brief outline of the most useful GFP flags |
7e3528c3 | 552 | * |
01598ba6 MR |
553 | * %GFP_KERNEL |
554 | * Allocate normal kernel ram. May sleep. | |
7e3528c3 | 555 | * |
01598ba6 MR |
556 | * %GFP_NOWAIT |
557 | * Allocation will not sleep. | |
7e3528c3 | 558 | * |
01598ba6 MR |
559 | * %GFP_ATOMIC |
560 | * Allocation will not sleep. May use emergency pools. | |
7e3528c3 | 561 | * |
01598ba6 MR |
562 | * %GFP_HIGHUSER |
563 | * Allocate memory from high memory on behalf of user. | |
7e3528c3 RD |
564 | * |
565 | * Also it is possible to set different flags by OR'ing | |
566 | * in one or more of the following additional @flags: | |
567 | * | |
01598ba6 MR |
568 | * %__GFP_HIGH |
569 | * This allocation has high priority and may use emergency pools. | |
7e3528c3 | 570 | * |
01598ba6 MR |
571 | * %__GFP_NOFAIL |
572 | * Indicate that this allocation is in no way allowed to fail | |
573 | * (think twice before using). | |
7e3528c3 | 574 | * |
01598ba6 MR |
575 | * %__GFP_NORETRY |
576 | * If memory is not immediately available, | |
577 | * then give up at once. | |
7e3528c3 | 578 | * |
01598ba6 MR |
579 | * %__GFP_NOWARN |
580 | * If allocation fails, don't issue any warnings. | |
7e3528c3 | 581 | * |
01598ba6 MR |
582 | * %__GFP_RETRY_MAYFAIL |
583 | * Try really hard to succeed the allocation but fail | |
584 | * eventually. | |
f1b6eb6e | 585 | */ |
c37495d6 | 586 | static __always_inline __alloc_size(1) void *kmalloc(size_t size, gfp_t flags) |
f1b6eb6e CL |
587 | { |
588 | if (__builtin_constant_p(size)) { | |
cc252eae VB |
589 | #ifndef CONFIG_SLOB |
590 | unsigned int index; | |
591 | #endif | |
f1b6eb6e CL |
592 | if (size > KMALLOC_MAX_CACHE_SIZE) |
593 | return kmalloc_large(size, flags); | |
594 | #ifndef CONFIG_SLOB | |
cc252eae | 595 | index = kmalloc_index(size); |
f1b6eb6e | 596 | |
cc252eae VB |
597 | if (!index) |
598 | return ZERO_SIZE_PTR; | |
f1b6eb6e | 599 | |
cc252eae VB |
600 | return kmem_cache_alloc_trace( |
601 | kmalloc_caches[kmalloc_type(flags)][index], | |
602 | flags, size); | |
f1b6eb6e CL |
603 | #endif |
604 | } | |
605 | return __kmalloc(size, flags); | |
606 | } | |
607 | ||
c37495d6 | 608 | static __always_inline __alloc_size(1) void *kmalloc_node(size_t size, gfp_t flags, int node) |
f1b6eb6e CL |
609 | { |
610 | #ifndef CONFIG_SLOB | |
611 | if (__builtin_constant_p(size) && | |
cc252eae | 612 | size <= KMALLOC_MAX_CACHE_SIZE) { |
36071a27 | 613 | unsigned int i = kmalloc_index(size); |
f1b6eb6e CL |
614 | |
615 | if (!i) | |
616 | return ZERO_SIZE_PTR; | |
617 | ||
cc252eae VB |
618 | return kmem_cache_alloc_node_trace( |
619 | kmalloc_caches[kmalloc_type(flags)][i], | |
f1b6eb6e CL |
620 | flags, node, size); |
621 | } | |
622 | #endif | |
623 | return __kmalloc_node(size, flags, node); | |
624 | } | |
625 | ||
e7efa615 MO |
626 | /** |
627 | * kmalloc_array - allocate memory for an array. | |
628 | * @n: number of elements. | |
629 | * @size: element size. | |
630 | * @flags: the type of memory to allocate (see kmalloc). | |
800590f5 | 631 | */ |
c37495d6 | 632 | static inline __alloc_size(1, 2) void *kmalloc_array(size_t n, size_t size, gfp_t flags) |
1da177e4 | 633 | { |
49b7f898 KC |
634 | size_t bytes; |
635 | ||
636 | if (unlikely(check_mul_overflow(n, size, &bytes))) | |
6193a2ff | 637 | return NULL; |
91c6a05f | 638 | if (__builtin_constant_p(n) && __builtin_constant_p(size)) |
49b7f898 KC |
639 | return kmalloc(bytes, flags); |
640 | return __kmalloc(bytes, flags); | |
a8203725 XW |
641 | } |
642 | ||
f0dbd2bd BG |
643 | /** |
644 | * krealloc_array - reallocate memory for an array. | |
645 | * @p: pointer to the memory chunk to reallocate | |
646 | * @new_n: new number of elements to alloc | |
647 | * @new_size: new size of a single member of the array | |
648 | * @flags: the type of memory to allocate (see kmalloc) | |
649 | */ | |
c37495d6 KC |
650 | static inline __alloc_size(2, 3) void * __must_check krealloc_array(void *p, |
651 | size_t new_n, | |
652 | size_t new_size, | |
653 | gfp_t flags) | |
f0dbd2bd BG |
654 | { |
655 | size_t bytes; | |
656 | ||
657 | if (unlikely(check_mul_overflow(new_n, new_size, &bytes))) | |
658 | return NULL; | |
659 | ||
660 | return krealloc(p, bytes, flags); | |
661 | } | |
662 | ||
a8203725 XW |
663 | /** |
664 | * kcalloc - allocate memory for an array. The memory is set to zero. | |
665 | * @n: number of elements. | |
666 | * @size: element size. | |
667 | * @flags: the type of memory to allocate (see kmalloc). | |
668 | */ | |
c37495d6 | 669 | static inline __alloc_size(1, 2) void *kcalloc(size_t n, size_t size, gfp_t flags) |
a8203725 XW |
670 | { |
671 | return kmalloc_array(n, size, flags | __GFP_ZERO); | |
1da177e4 LT |
672 | } |
673 | ||
1d2c8eea CH |
674 | /* |
675 | * kmalloc_track_caller is a special version of kmalloc that records the | |
676 | * calling function of the routine calling it for slab leak tracking instead | |
677 | * of just the calling function (confusing, eh?). | |
678 | * It's useful when the call to kmalloc comes from a widely-used standard | |
679 | * allocator where we care about the real place the memory allocation | |
680 | * request comes from. | |
681 | */ | |
93dd04ab | 682 | extern void *__kmalloc_track_caller(size_t size, gfp_t flags, unsigned long caller); |
1d2c8eea | 683 | #define kmalloc_track_caller(size, flags) \ |
ce71e27c | 684 | __kmalloc_track_caller(size, flags, _RET_IP_) |
1da177e4 | 685 | |
c37495d6 KC |
686 | static inline __alloc_size(1, 2) void *kmalloc_array_node(size_t n, size_t size, gfp_t flags, |
687 | int node) | |
5799b255 | 688 | { |
49b7f898 KC |
689 | size_t bytes; |
690 | ||
691 | if (unlikely(check_mul_overflow(n, size, &bytes))) | |
5799b255 JT |
692 | return NULL; |
693 | if (__builtin_constant_p(n) && __builtin_constant_p(size)) | |
49b7f898 KC |
694 | return kmalloc_node(bytes, flags, node); |
695 | return __kmalloc_node(bytes, flags, node); | |
5799b255 JT |
696 | } |
697 | ||
c37495d6 | 698 | static inline __alloc_size(1, 2) void *kcalloc_node(size_t n, size_t size, gfp_t flags, int node) |
5799b255 JT |
699 | { |
700 | return kmalloc_array_node(n, size, flags | __GFP_ZERO, node); | |
701 | } | |
702 | ||
703 | ||
97e2bde4 | 704 | #ifdef CONFIG_NUMA |
72d67229 | 705 | extern void *__kmalloc_node_track_caller(size_t size, gfp_t flags, int node, |
c37495d6 | 706 | unsigned long caller) __alloc_size(1); |
8b98c169 CH |
707 | #define kmalloc_node_track_caller(size, flags, node) \ |
708 | __kmalloc_node_track_caller(size, flags, node, \ | |
ce71e27c | 709 | _RET_IP_) |
2e892f43 | 710 | |
8b98c169 | 711 | #else /* CONFIG_NUMA */ |
8b98c169 CH |
712 | |
713 | #define kmalloc_node_track_caller(size, flags, node) \ | |
714 | kmalloc_track_caller(size, flags) | |
97e2bde4 | 715 | |
dfcd3610 | 716 | #endif /* CONFIG_NUMA */ |
10cef602 | 717 | |
81cda662 CL |
718 | /* |
719 | * Shortcuts | |
720 | */ | |
721 | static inline void *kmem_cache_zalloc(struct kmem_cache *k, gfp_t flags) | |
722 | { | |
723 | return kmem_cache_alloc(k, flags | __GFP_ZERO); | |
724 | } | |
725 | ||
726 | /** | |
727 | * kzalloc - allocate memory. The memory is set to zero. | |
728 | * @size: how many bytes of memory are required. | |
729 | * @flags: the type of memory to allocate (see kmalloc). | |
730 | */ | |
c37495d6 | 731 | static inline __alloc_size(1) void *kzalloc(size_t size, gfp_t flags) |
81cda662 CL |
732 | { |
733 | return kmalloc(size, flags | __GFP_ZERO); | |
734 | } | |
735 | ||
979b0fea JL |
736 | /** |
737 | * kzalloc_node - allocate zeroed memory from a particular memory node. | |
738 | * @size: how many bytes of memory are required. | |
739 | * @flags: the type of memory to allocate (see kmalloc). | |
740 | * @node: memory node from which to allocate | |
741 | */ | |
c37495d6 | 742 | static inline __alloc_size(1) void *kzalloc_node(size_t size, gfp_t flags, int node) |
979b0fea JL |
743 | { |
744 | return kmalloc_node(size, flags | __GFP_ZERO, node); | |
745 | } | |
746 | ||
56bcf40f KC |
747 | extern void *kvmalloc_node(size_t size, gfp_t flags, int node) __alloc_size(1); |
748 | static inline __alloc_size(1) void *kvmalloc(size_t size, gfp_t flags) | |
8587ca6f MWO |
749 | { |
750 | return kvmalloc_node(size, flags, NUMA_NO_NODE); | |
751 | } | |
56bcf40f | 752 | static inline __alloc_size(1) void *kvzalloc_node(size_t size, gfp_t flags, int node) |
8587ca6f MWO |
753 | { |
754 | return kvmalloc_node(size, flags | __GFP_ZERO, node); | |
755 | } | |
56bcf40f | 756 | static inline __alloc_size(1) void *kvzalloc(size_t size, gfp_t flags) |
8587ca6f MWO |
757 | { |
758 | return kvmalloc(size, flags | __GFP_ZERO); | |
759 | } | |
760 | ||
56bcf40f | 761 | static inline __alloc_size(1, 2) void *kvmalloc_array(size_t n, size_t size, gfp_t flags) |
8587ca6f MWO |
762 | { |
763 | size_t bytes; | |
764 | ||
765 | if (unlikely(check_mul_overflow(n, size, &bytes))) | |
766 | return NULL; | |
767 | ||
768 | return kvmalloc(bytes, flags); | |
769 | } | |
770 | ||
56bcf40f | 771 | static inline __alloc_size(1, 2) void *kvcalloc(size_t n, size_t size, gfp_t flags) |
8587ca6f MWO |
772 | { |
773 | return kvmalloc_array(n, size, flags | __GFP_ZERO); | |
774 | } | |
775 | ||
56bcf40f KC |
776 | extern void *kvrealloc(const void *p, size_t oldsize, size_t newsize, gfp_t flags) |
777 | __alloc_size(3); | |
8587ca6f MWO |
778 | extern void kvfree(const void *addr); |
779 | extern void kvfree_sensitive(const void *addr, size_t len); | |
780 | ||
07f361b2 | 781 | unsigned int kmem_cache_size(struct kmem_cache *s); |
7e85ee0c PE |
782 | void __init kmem_cache_init_late(void); |
783 | ||
6731d4f1 SAS |
784 | #if defined(CONFIG_SMP) && defined(CONFIG_SLAB) |
785 | int slab_prepare_cpu(unsigned int cpu); | |
786 | int slab_dead_cpu(unsigned int cpu); | |
787 | #else | |
788 | #define slab_prepare_cpu NULL | |
789 | #define slab_dead_cpu NULL | |
790 | #endif | |
791 | ||
1da177e4 | 792 | #endif /* _LINUX_SLAB_H */ |