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 | ||
b84e04f1 IK |
122 | #ifdef CONFIG_KFENCE |
123 | #define SLAB_SKIP_KFENCE ((slab_flags_t __force)0x20000000U) | |
124 | #else | |
125 | #define SLAB_SKIP_KFENCE 0 | |
126 | #endif | |
127 | ||
e12ba74d | 128 | /* The following flags affect the page allocator grouping pages by mobility */ |
d50112ed | 129 | /* Objects are reclaimable */ |
4fd0b46e | 130 | #define SLAB_RECLAIM_ACCOUNT ((slab_flags_t __force)0x00020000U) |
e12ba74d | 131 | #define SLAB_TEMPORARY SLAB_RECLAIM_ACCOUNT /* Objects are short-lived */ |
fcf8a1e4 | 132 | |
6cb8f913 CL |
133 | /* |
134 | * ZERO_SIZE_PTR will be returned for zero sized kmalloc requests. | |
135 | * | |
136 | * Dereferencing ZERO_SIZE_PTR will lead to a distinct access fault. | |
137 | * | |
138 | * ZERO_SIZE_PTR can be passed to kfree though in the same way that NULL can. | |
139 | * Both make kfree a no-op. | |
140 | */ | |
141 | #define ZERO_SIZE_PTR ((void *)16) | |
142 | ||
1d4ec7b1 | 143 | #define ZERO_OR_NULL_PTR(x) ((unsigned long)(x) <= \ |
6cb8f913 CL |
144 | (unsigned long)ZERO_SIZE_PTR) |
145 | ||
0316bec2 | 146 | #include <linux/kasan.h> |
3b0efdfa | 147 | |
88f2ef73 | 148 | struct list_lru; |
2633d7a0 | 149 | struct mem_cgroup; |
2e892f43 CL |
150 | /* |
151 | * struct kmem_cache related prototypes | |
152 | */ | |
153 | void __init kmem_cache_init(void); | |
fda90124 | 154 | bool slab_is_available(void); |
1da177e4 | 155 | |
f4957d5b AD |
156 | struct kmem_cache *kmem_cache_create(const char *name, unsigned int size, |
157 | unsigned int align, slab_flags_t flags, | |
8eb8284b DW |
158 | void (*ctor)(void *)); |
159 | struct kmem_cache *kmem_cache_create_usercopy(const char *name, | |
f4957d5b AD |
160 | unsigned int size, unsigned int align, |
161 | slab_flags_t flags, | |
7bbdb81e | 162 | unsigned int useroffset, unsigned int usersize, |
8eb8284b | 163 | void (*ctor)(void *)); |
72d67229 KC |
164 | void kmem_cache_destroy(struct kmem_cache *s); |
165 | int kmem_cache_shrink(struct kmem_cache *s); | |
2a4db7eb | 166 | |
0a31bd5f CL |
167 | /* |
168 | * Please use this macro to create slab caches. Simply specify the | |
169 | * name of the structure and maybe some flags that are listed above. | |
170 | * | |
171 | * The alignment of the struct determines object alignment. If you | |
172 | * f.e. add ____cacheline_aligned_in_smp to the struct declaration | |
173 | * then the objects will be properly aligned in SMP configurations. | |
174 | */ | |
8eb8284b DW |
175 | #define KMEM_CACHE(__struct, __flags) \ |
176 | kmem_cache_create(#__struct, sizeof(struct __struct), \ | |
177 | __alignof__(struct __struct), (__flags), NULL) | |
178 | ||
179 | /* | |
180 | * To whitelist a single field for copying to/from usercopy, use this | |
181 | * macro instead for KMEM_CACHE() above. | |
182 | */ | |
183 | #define KMEM_CACHE_USERCOPY(__struct, __flags, __field) \ | |
184 | kmem_cache_create_usercopy(#__struct, \ | |
185 | sizeof(struct __struct), \ | |
186 | __alignof__(struct __struct), (__flags), \ | |
187 | offsetof(struct __struct, __field), \ | |
188 | sizeof_field(struct __struct, __field), NULL) | |
0a31bd5f | 189 | |
34504667 CL |
190 | /* |
191 | * Common kmalloc functions provided by all allocators | |
192 | */ | |
c37495d6 | 193 | void * __must_check krealloc(const void *objp, size_t new_size, gfp_t flags) __alloc_size(2); |
72d67229 KC |
194 | void kfree(const void *objp); |
195 | void kfree_sensitive(const void *objp); | |
196 | size_t __ksize(const void *objp); | |
197 | size_t ksize(const void *objp); | |
5bb1bb35 | 198 | #ifdef CONFIG_PRINTK |
8e7f37f2 PM |
199 | bool kmem_valid_obj(void *object); |
200 | void kmem_dump_obj(void *object); | |
5bb1bb35 | 201 | #endif |
34504667 | 202 | |
c601fd69 CL |
203 | /* |
204 | * Some archs want to perform DMA into kmalloc caches and need a guaranteed | |
205 | * alignment larger than the alignment of a 64-bit integer. | |
8cf9e121 | 206 | * Setting ARCH_DMA_MINALIGN in arch headers allows that. |
c601fd69 CL |
207 | */ |
208 | #if defined(ARCH_DMA_MINALIGN) && ARCH_DMA_MINALIGN > 8 | |
209 | #define ARCH_KMALLOC_MINALIGN ARCH_DMA_MINALIGN | |
210 | #define KMALLOC_MIN_SIZE ARCH_DMA_MINALIGN | |
211 | #define KMALLOC_SHIFT_LOW ilog2(ARCH_DMA_MINALIGN) | |
212 | #else | |
213 | #define ARCH_KMALLOC_MINALIGN __alignof__(unsigned long long) | |
214 | #endif | |
215 | ||
94a58c36 RV |
216 | /* |
217 | * Setting ARCH_SLAB_MINALIGN in arch headers allows a different alignment. | |
218 | * Intended for arches that get misalignment faults even for 64 bit integer | |
219 | * aligned buffers. | |
220 | */ | |
221 | #ifndef ARCH_SLAB_MINALIGN | |
222 | #define ARCH_SLAB_MINALIGN __alignof__(unsigned long long) | |
223 | #endif | |
224 | ||
d949a815 PC |
225 | /* |
226 | * Arches can define this function if they want to decide the minimum slab | |
227 | * alignment at runtime. The value returned by the function must be a power | |
228 | * of two and >= ARCH_SLAB_MINALIGN. | |
229 | */ | |
230 | #ifndef arch_slab_minalign | |
231 | static inline unsigned int arch_slab_minalign(void) | |
232 | { | |
233 | return ARCH_SLAB_MINALIGN; | |
234 | } | |
235 | #endif | |
236 | ||
94a58c36 | 237 | /* |
154036a3 AK |
238 | * kmem_cache_alloc and friends return pointers aligned to ARCH_SLAB_MINALIGN. |
239 | * kmalloc and friends return pointers aligned to both ARCH_KMALLOC_MINALIGN | |
240 | * and ARCH_SLAB_MINALIGN, but here we only assume the former alignment. | |
94a58c36 RV |
241 | */ |
242 | #define __assume_kmalloc_alignment __assume_aligned(ARCH_KMALLOC_MINALIGN) | |
243 | #define __assume_slab_alignment __assume_aligned(ARCH_SLAB_MINALIGN) | |
244 | #define __assume_page_alignment __assume_aligned(PAGE_SIZE) | |
245 | ||
0aa817f0 | 246 | /* |
95a05b42 CL |
247 | * Kmalloc array related definitions |
248 | */ | |
249 | ||
250 | #ifdef CONFIG_SLAB | |
251 | /* | |
252 | * The largest kmalloc size supported by the SLAB allocators is | |
0aa817f0 CL |
253 | * 32 megabyte (2^25) or the maximum allocatable page order if that is |
254 | * less than 32 MB. | |
255 | * | |
256 | * WARNING: Its not easy to increase this value since the allocators have | |
257 | * to do various tricks to work around compiler limitations in order to | |
258 | * ensure proper constant folding. | |
259 | */ | |
debee076 CL |
260 | #define KMALLOC_SHIFT_HIGH ((MAX_ORDER + PAGE_SHIFT - 1) <= 25 ? \ |
261 | (MAX_ORDER + PAGE_SHIFT - 1) : 25) | |
95a05b42 | 262 | #define KMALLOC_SHIFT_MAX KMALLOC_SHIFT_HIGH |
c601fd69 | 263 | #ifndef KMALLOC_SHIFT_LOW |
95a05b42 | 264 | #define KMALLOC_SHIFT_LOW 5 |
c601fd69 | 265 | #endif |
069e2b35 CL |
266 | #endif |
267 | ||
268 | #ifdef CONFIG_SLUB | |
95a05b42 | 269 | /* |
433a91ff DH |
270 | * SLUB directly allocates requests fitting in to an order-1 page |
271 | * (PAGE_SIZE*2). Larger requests are passed to the page allocator. | |
95a05b42 CL |
272 | */ |
273 | #define KMALLOC_SHIFT_HIGH (PAGE_SHIFT + 1) | |
bb1107f7 | 274 | #define KMALLOC_SHIFT_MAX (MAX_ORDER + PAGE_SHIFT - 1) |
c601fd69 | 275 | #ifndef KMALLOC_SHIFT_LOW |
95a05b42 CL |
276 | #define KMALLOC_SHIFT_LOW 3 |
277 | #endif | |
c601fd69 | 278 | #endif |
0aa817f0 | 279 | |
069e2b35 CL |
280 | #ifdef CONFIG_SLOB |
281 | /* | |
433a91ff | 282 | * SLOB passes all requests larger than one page to the page allocator. |
069e2b35 CL |
283 | * No kmalloc array is necessary since objects of different sizes can |
284 | * be allocated from the same page. | |
285 | */ | |
069e2b35 | 286 | #define KMALLOC_SHIFT_HIGH PAGE_SHIFT |
bb1107f7 | 287 | #define KMALLOC_SHIFT_MAX (MAX_ORDER + PAGE_SHIFT - 1) |
069e2b35 CL |
288 | #ifndef KMALLOC_SHIFT_LOW |
289 | #define KMALLOC_SHIFT_LOW 3 | |
290 | #endif | |
291 | #endif | |
292 | ||
95a05b42 CL |
293 | /* Maximum allocatable size */ |
294 | #define KMALLOC_MAX_SIZE (1UL << KMALLOC_SHIFT_MAX) | |
295 | /* Maximum size for which we actually use a slab cache */ | |
296 | #define KMALLOC_MAX_CACHE_SIZE (1UL << KMALLOC_SHIFT_HIGH) | |
d7cff4de | 297 | /* Maximum order allocatable via the slab allocator */ |
95a05b42 | 298 | #define KMALLOC_MAX_ORDER (KMALLOC_SHIFT_MAX - PAGE_SHIFT) |
0aa817f0 | 299 | |
ce6a5026 CL |
300 | /* |
301 | * Kmalloc subsystem. | |
302 | */ | |
c601fd69 | 303 | #ifndef KMALLOC_MIN_SIZE |
95a05b42 | 304 | #define KMALLOC_MIN_SIZE (1 << KMALLOC_SHIFT_LOW) |
ce6a5026 CL |
305 | #endif |
306 | ||
24f870d8 JK |
307 | /* |
308 | * This restriction comes from byte sized index implementation. | |
309 | * Page size is normally 2^12 bytes and, in this case, if we want to use | |
310 | * byte sized index which can represent 2^8 entries, the size of the object | |
311 | * should be equal or greater to 2^12 / 2^8 = 2^4 = 16. | |
312 | * If minimum size of kmalloc is less than 16, we use it as minimum object | |
313 | * size and give up to use byte sized index. | |
314 | */ | |
315 | #define SLAB_OBJ_MIN_SIZE (KMALLOC_MIN_SIZE < 16 ? \ | |
316 | (KMALLOC_MIN_SIZE) : 16) | |
317 | ||
1291523f VB |
318 | /* |
319 | * Whenever changing this, take care of that kmalloc_type() and | |
320 | * create_kmalloc_caches() still work as intended. | |
494c1dfe WL |
321 | * |
322 | * KMALLOC_NORMAL can contain only unaccounted objects whereas KMALLOC_CGROUP | |
323 | * is for accounted but unreclaimable and non-dma objects. All the other | |
324 | * kmem caches can have both accounted and unaccounted objects. | |
1291523f | 325 | */ |
cc252eae VB |
326 | enum kmalloc_cache_type { |
327 | KMALLOC_NORMAL = 0, | |
494c1dfe WL |
328 | #ifndef CONFIG_ZONE_DMA |
329 | KMALLOC_DMA = KMALLOC_NORMAL, | |
330 | #endif | |
331 | #ifndef CONFIG_MEMCG_KMEM | |
332 | KMALLOC_CGROUP = KMALLOC_NORMAL, | |
333 | #else | |
334 | KMALLOC_CGROUP, | |
335 | #endif | |
1291523f | 336 | KMALLOC_RECLAIM, |
cc252eae VB |
337 | #ifdef CONFIG_ZONE_DMA |
338 | KMALLOC_DMA, | |
339 | #endif | |
340 | NR_KMALLOC_TYPES | |
341 | }; | |
342 | ||
069e2b35 | 343 | #ifndef CONFIG_SLOB |
cc252eae VB |
344 | extern struct kmem_cache * |
345 | kmalloc_caches[NR_KMALLOC_TYPES][KMALLOC_SHIFT_HIGH + 1]; | |
346 | ||
494c1dfe WL |
347 | /* |
348 | * Define gfp bits that should not be set for KMALLOC_NORMAL. | |
349 | */ | |
350 | #define KMALLOC_NOT_NORMAL_BITS \ | |
351 | (__GFP_RECLAIMABLE | \ | |
352 | (IS_ENABLED(CONFIG_ZONE_DMA) ? __GFP_DMA : 0) | \ | |
353 | (IS_ENABLED(CONFIG_MEMCG_KMEM) ? __GFP_ACCOUNT : 0)) | |
354 | ||
cc252eae VB |
355 | static __always_inline enum kmalloc_cache_type kmalloc_type(gfp_t flags) |
356 | { | |
4e45f712 VB |
357 | /* |
358 | * The most common case is KMALLOC_NORMAL, so test for it | |
494c1dfe | 359 | * with a single branch for all the relevant flags. |
4e45f712 | 360 | */ |
494c1dfe | 361 | if (likely((flags & KMALLOC_NOT_NORMAL_BITS) == 0)) |
4e45f712 | 362 | return KMALLOC_NORMAL; |
1291523f VB |
363 | |
364 | /* | |
494c1dfe WL |
365 | * At least one of the flags has to be set. Their priorities in |
366 | * decreasing order are: | |
367 | * 1) __GFP_DMA | |
368 | * 2) __GFP_RECLAIMABLE | |
369 | * 3) __GFP_ACCOUNT | |
1291523f | 370 | */ |
494c1dfe WL |
371 | if (IS_ENABLED(CONFIG_ZONE_DMA) && (flags & __GFP_DMA)) |
372 | return KMALLOC_DMA; | |
373 | if (!IS_ENABLED(CONFIG_MEMCG_KMEM) || (flags & __GFP_RECLAIMABLE)) | |
374 | return KMALLOC_RECLAIM; | |
375 | else | |
376 | return KMALLOC_CGROUP; | |
cc252eae VB |
377 | } |
378 | ||
ce6a5026 CL |
379 | /* |
380 | * Figure out which kmalloc slab an allocation of a certain size | |
381 | * belongs to. | |
382 | * 0 = zero alloc | |
383 | * 1 = 65 .. 96 bytes | |
1ed58b60 RV |
384 | * 2 = 129 .. 192 bytes |
385 | * n = 2^(n-1)+1 .. 2^n | |
588c7fa0 HY |
386 | * |
387 | * Note: __kmalloc_index() is compile-time optimized, and not runtime optimized; | |
388 | * typical usage is via kmalloc_index() and therefore evaluated at compile-time. | |
389 | * Callers where !size_is_constant should only be test modules, where runtime | |
390 | * overheads of __kmalloc_index() can be tolerated. Also see kmalloc_slab(). | |
ce6a5026 | 391 | */ |
588c7fa0 HY |
392 | static __always_inline unsigned int __kmalloc_index(size_t size, |
393 | bool size_is_constant) | |
ce6a5026 CL |
394 | { |
395 | if (!size) | |
396 | return 0; | |
397 | ||
398 | if (size <= KMALLOC_MIN_SIZE) | |
399 | return KMALLOC_SHIFT_LOW; | |
400 | ||
401 | if (KMALLOC_MIN_SIZE <= 32 && size > 64 && size <= 96) | |
402 | return 1; | |
403 | if (KMALLOC_MIN_SIZE <= 64 && size > 128 && size <= 192) | |
404 | return 2; | |
405 | if (size <= 8) return 3; | |
406 | if (size <= 16) return 4; | |
407 | if (size <= 32) return 5; | |
408 | if (size <= 64) return 6; | |
409 | if (size <= 128) return 7; | |
410 | if (size <= 256) return 8; | |
411 | if (size <= 512) return 9; | |
412 | if (size <= 1024) return 10; | |
413 | if (size <= 2 * 1024) return 11; | |
414 | if (size <= 4 * 1024) return 12; | |
415 | if (size <= 8 * 1024) return 13; | |
416 | if (size <= 16 * 1024) return 14; | |
417 | if (size <= 32 * 1024) return 15; | |
418 | if (size <= 64 * 1024) return 16; | |
419 | if (size <= 128 * 1024) return 17; | |
420 | if (size <= 256 * 1024) return 18; | |
421 | if (size <= 512 * 1024) return 19; | |
422 | if (size <= 1024 * 1024) return 20; | |
423 | if (size <= 2 * 1024 * 1024) return 21; | |
424 | if (size <= 4 * 1024 * 1024) return 22; | |
425 | if (size <= 8 * 1024 * 1024) return 23; | |
426 | if (size <= 16 * 1024 * 1024) return 24; | |
427 | if (size <= 32 * 1024 * 1024) return 25; | |
588c7fa0 | 428 | |
57b2b72a | 429 | if (!IS_ENABLED(CONFIG_PROFILE_ALL_BRANCHES) && size_is_constant) |
588c7fa0 HY |
430 | BUILD_BUG_ON_MSG(1, "unexpected size in kmalloc_index()"); |
431 | else | |
432 | BUG(); | |
ce6a5026 CL |
433 | |
434 | /* Will never be reached. Needed because the compiler may complain */ | |
435 | return -1; | |
436 | } | |
588c7fa0 | 437 | #define kmalloc_index(s) __kmalloc_index(s, true) |
069e2b35 | 438 | #endif /* !CONFIG_SLOB */ |
ce6a5026 | 439 | |
c37495d6 | 440 | void *__kmalloc(size_t size, gfp_t flags) __assume_kmalloc_alignment __alloc_size(1); |
72d67229 | 441 | void *kmem_cache_alloc(struct kmem_cache *s, gfp_t flags) __assume_slab_alignment __malloc; |
88f2ef73 MS |
442 | void *kmem_cache_alloc_lru(struct kmem_cache *s, struct list_lru *lru, |
443 | gfp_t gfpflags) __assume_slab_alignment __malloc; | |
72d67229 | 444 | void kmem_cache_free(struct kmem_cache *s, void *objp); |
f1b6eb6e | 445 | |
484748f0 | 446 | /* |
9f706d68 | 447 | * Bulk allocation and freeing operations. These are accelerated in an |
484748f0 CL |
448 | * allocator specific way to avoid taking locks repeatedly or building |
449 | * metadata structures unnecessarily. | |
450 | * | |
451 | * Note that interrupts must be enabled when calling these functions. | |
452 | */ | |
72d67229 KC |
453 | void kmem_cache_free_bulk(struct kmem_cache *s, size_t size, void **p); |
454 | int kmem_cache_alloc_bulk(struct kmem_cache *s, gfp_t flags, size_t size, void **p); | |
484748f0 | 455 | |
ca257195 JDB |
456 | /* |
457 | * Caller must not use kfree_bulk() on memory not originally allocated | |
458 | * by kmalloc(), because the SLOB allocator cannot handle this. | |
459 | */ | |
460 | static __always_inline void kfree_bulk(size_t size, void **p) | |
461 | { | |
462 | kmem_cache_free_bulk(NULL, size, p); | |
463 | } | |
464 | ||
f1b6eb6e | 465 | #ifdef CONFIG_NUMA |
c37495d6 KC |
466 | void *__kmalloc_node(size_t size, gfp_t flags, int node) __assume_kmalloc_alignment |
467 | __alloc_size(1); | |
72d67229 KC |
468 | void *kmem_cache_alloc_node(struct kmem_cache *s, gfp_t flags, int node) __assume_slab_alignment |
469 | __malloc; | |
f1b6eb6e | 470 | #else |
c37495d6 | 471 | static __always_inline __alloc_size(1) void *__kmalloc_node(size_t size, gfp_t flags, int node) |
f1b6eb6e CL |
472 | { |
473 | return __kmalloc(size, flags); | |
474 | } | |
475 | ||
476 | static __always_inline void *kmem_cache_alloc_node(struct kmem_cache *s, gfp_t flags, int node) | |
477 | { | |
478 | return kmem_cache_alloc(s, flags); | |
479 | } | |
480 | #endif | |
481 | ||
482 | #ifdef CONFIG_TRACING | |
72d67229 | 483 | extern void *kmem_cache_alloc_trace(struct kmem_cache *s, gfp_t flags, size_t size) |
c37495d6 | 484 | __assume_slab_alignment __alloc_size(3); |
f1b6eb6e CL |
485 | |
486 | #ifdef CONFIG_NUMA | |
72d67229 | 487 | extern void *kmem_cache_alloc_node_trace(struct kmem_cache *s, gfp_t gfpflags, |
c37495d6 KC |
488 | int node, size_t size) __assume_slab_alignment |
489 | __alloc_size(4); | |
f1b6eb6e | 490 | #else |
c37495d6 KC |
491 | static __always_inline __alloc_size(4) void *kmem_cache_alloc_node_trace(struct kmem_cache *s, |
492 | gfp_t gfpflags, int node, size_t size) | |
f1b6eb6e CL |
493 | { |
494 | return kmem_cache_alloc_trace(s, gfpflags, size); | |
495 | } | |
496 | #endif /* CONFIG_NUMA */ | |
497 | ||
498 | #else /* CONFIG_TRACING */ | |
c37495d6 KC |
499 | static __always_inline __alloc_size(3) void *kmem_cache_alloc_trace(struct kmem_cache *s, |
500 | gfp_t flags, size_t size) | |
f1b6eb6e | 501 | { |
0316bec2 AR |
502 | void *ret = kmem_cache_alloc(s, flags); |
503 | ||
0116523c | 504 | ret = kasan_kmalloc(s, ret, size, flags); |
0316bec2 | 505 | return ret; |
f1b6eb6e CL |
506 | } |
507 | ||
72d67229 KC |
508 | static __always_inline void *kmem_cache_alloc_node_trace(struct kmem_cache *s, gfp_t gfpflags, |
509 | int node, size_t size) | |
f1b6eb6e | 510 | { |
0316bec2 AR |
511 | void *ret = kmem_cache_alloc_node(s, gfpflags, node); |
512 | ||
0116523c | 513 | ret = kasan_kmalloc(s, ret, size, gfpflags); |
0316bec2 | 514 | return ret; |
f1b6eb6e CL |
515 | } |
516 | #endif /* CONFIG_TRACING */ | |
517 | ||
72d67229 | 518 | extern void *kmalloc_order(size_t size, gfp_t flags, unsigned int order) __assume_page_alignment |
c37495d6 | 519 | __alloc_size(1); |
f1b6eb6e CL |
520 | |
521 | #ifdef CONFIG_TRACING | |
72d67229 | 522 | extern void *kmalloc_order_trace(size_t size, gfp_t flags, unsigned int order) |
c37495d6 | 523 | __assume_page_alignment __alloc_size(1); |
f1b6eb6e | 524 | #else |
c37495d6 KC |
525 | static __always_inline __alloc_size(1) void *kmalloc_order_trace(size_t size, gfp_t flags, |
526 | unsigned int order) | |
f1b6eb6e CL |
527 | { |
528 | return kmalloc_order(size, flags, order); | |
529 | } | |
ce6a5026 CL |
530 | #endif |
531 | ||
c37495d6 | 532 | static __always_inline __alloc_size(1) void *kmalloc_large(size_t size, gfp_t flags) |
f1b6eb6e CL |
533 | { |
534 | unsigned int order = get_order(size); | |
535 | return kmalloc_order_trace(size, flags, order); | |
536 | } | |
537 | ||
538 | /** | |
539 | * kmalloc - allocate memory | |
540 | * @size: how many bytes of memory are required. | |
7e3528c3 | 541 | * @flags: the type of memory to allocate. |
f1b6eb6e CL |
542 | * |
543 | * kmalloc is the normal method of allocating memory | |
544 | * for objects smaller than page size in the kernel. | |
7e3528c3 | 545 | * |
59bb4798 VB |
546 | * The allocated object address is aligned to at least ARCH_KMALLOC_MINALIGN |
547 | * bytes. For @size of power of two bytes, the alignment is also guaranteed | |
548 | * to be at least to the size. | |
549 | * | |
01598ba6 MR |
550 | * The @flags argument may be one of the GFP flags defined at |
551 | * include/linux/gfp.h and described at | |
552 | * :ref:`Documentation/core-api/mm-api.rst <mm-api-gfp-flags>` | |
7e3528c3 | 553 | * |
01598ba6 | 554 | * The recommended usage of the @flags is described at |
2370ae4b | 555 | * :ref:`Documentation/core-api/memory-allocation.rst <memory_allocation>` |
7e3528c3 | 556 | * |
01598ba6 | 557 | * Below is a brief outline of the most useful GFP flags |
7e3528c3 | 558 | * |
01598ba6 MR |
559 | * %GFP_KERNEL |
560 | * Allocate normal kernel ram. May sleep. | |
7e3528c3 | 561 | * |
01598ba6 MR |
562 | * %GFP_NOWAIT |
563 | * Allocation will not sleep. | |
7e3528c3 | 564 | * |
01598ba6 MR |
565 | * %GFP_ATOMIC |
566 | * Allocation will not sleep. May use emergency pools. | |
7e3528c3 | 567 | * |
01598ba6 MR |
568 | * %GFP_HIGHUSER |
569 | * Allocate memory from high memory on behalf of user. | |
7e3528c3 RD |
570 | * |
571 | * Also it is possible to set different flags by OR'ing | |
572 | * in one or more of the following additional @flags: | |
573 | * | |
01598ba6 MR |
574 | * %__GFP_HIGH |
575 | * This allocation has high priority and may use emergency pools. | |
7e3528c3 | 576 | * |
01598ba6 MR |
577 | * %__GFP_NOFAIL |
578 | * Indicate that this allocation is in no way allowed to fail | |
579 | * (think twice before using). | |
7e3528c3 | 580 | * |
01598ba6 MR |
581 | * %__GFP_NORETRY |
582 | * If memory is not immediately available, | |
583 | * then give up at once. | |
7e3528c3 | 584 | * |
01598ba6 MR |
585 | * %__GFP_NOWARN |
586 | * If allocation fails, don't issue any warnings. | |
7e3528c3 | 587 | * |
01598ba6 MR |
588 | * %__GFP_RETRY_MAYFAIL |
589 | * Try really hard to succeed the allocation but fail | |
590 | * eventually. | |
f1b6eb6e | 591 | */ |
c37495d6 | 592 | static __always_inline __alloc_size(1) void *kmalloc(size_t size, gfp_t flags) |
f1b6eb6e CL |
593 | { |
594 | if (__builtin_constant_p(size)) { | |
cc252eae VB |
595 | #ifndef CONFIG_SLOB |
596 | unsigned int index; | |
597 | #endif | |
f1b6eb6e CL |
598 | if (size > KMALLOC_MAX_CACHE_SIZE) |
599 | return kmalloc_large(size, flags); | |
600 | #ifndef CONFIG_SLOB | |
cc252eae | 601 | index = kmalloc_index(size); |
f1b6eb6e | 602 | |
cc252eae VB |
603 | if (!index) |
604 | return ZERO_SIZE_PTR; | |
f1b6eb6e | 605 | |
cc252eae VB |
606 | return kmem_cache_alloc_trace( |
607 | kmalloc_caches[kmalloc_type(flags)][index], | |
608 | flags, size); | |
f1b6eb6e CL |
609 | #endif |
610 | } | |
611 | return __kmalloc(size, flags); | |
612 | } | |
613 | ||
c37495d6 | 614 | static __always_inline __alloc_size(1) void *kmalloc_node(size_t size, gfp_t flags, int node) |
f1b6eb6e CL |
615 | { |
616 | #ifndef CONFIG_SLOB | |
617 | if (__builtin_constant_p(size) && | |
cc252eae | 618 | size <= KMALLOC_MAX_CACHE_SIZE) { |
36071a27 | 619 | unsigned int i = kmalloc_index(size); |
f1b6eb6e CL |
620 | |
621 | if (!i) | |
622 | return ZERO_SIZE_PTR; | |
623 | ||
cc252eae VB |
624 | return kmem_cache_alloc_node_trace( |
625 | kmalloc_caches[kmalloc_type(flags)][i], | |
f1b6eb6e CL |
626 | flags, node, size); |
627 | } | |
628 | #endif | |
629 | return __kmalloc_node(size, flags, node); | |
630 | } | |
631 | ||
e7efa615 MO |
632 | /** |
633 | * kmalloc_array - allocate memory for an array. | |
634 | * @n: number of elements. | |
635 | * @size: element size. | |
636 | * @flags: the type of memory to allocate (see kmalloc). | |
800590f5 | 637 | */ |
c37495d6 | 638 | static inline __alloc_size(1, 2) void *kmalloc_array(size_t n, size_t size, gfp_t flags) |
1da177e4 | 639 | { |
49b7f898 KC |
640 | size_t bytes; |
641 | ||
642 | if (unlikely(check_mul_overflow(n, size, &bytes))) | |
6193a2ff | 643 | return NULL; |
91c6a05f | 644 | if (__builtin_constant_p(n) && __builtin_constant_p(size)) |
49b7f898 KC |
645 | return kmalloc(bytes, flags); |
646 | return __kmalloc(bytes, flags); | |
a8203725 XW |
647 | } |
648 | ||
f0dbd2bd BG |
649 | /** |
650 | * krealloc_array - reallocate memory for an array. | |
651 | * @p: pointer to the memory chunk to reallocate | |
652 | * @new_n: new number of elements to alloc | |
653 | * @new_size: new size of a single member of the array | |
654 | * @flags: the type of memory to allocate (see kmalloc) | |
655 | */ | |
c37495d6 KC |
656 | static inline __alloc_size(2, 3) void * __must_check krealloc_array(void *p, |
657 | size_t new_n, | |
658 | size_t new_size, | |
659 | gfp_t flags) | |
f0dbd2bd BG |
660 | { |
661 | size_t bytes; | |
662 | ||
663 | if (unlikely(check_mul_overflow(new_n, new_size, &bytes))) | |
664 | return NULL; | |
665 | ||
666 | return krealloc(p, bytes, flags); | |
667 | } | |
668 | ||
a8203725 XW |
669 | /** |
670 | * kcalloc - allocate memory for an array. The memory is set to zero. | |
671 | * @n: number of elements. | |
672 | * @size: element size. | |
673 | * @flags: the type of memory to allocate (see kmalloc). | |
674 | */ | |
c37495d6 | 675 | static inline __alloc_size(1, 2) void *kcalloc(size_t n, size_t size, gfp_t flags) |
a8203725 XW |
676 | { |
677 | return kmalloc_array(n, size, flags | __GFP_ZERO); | |
1da177e4 LT |
678 | } |
679 | ||
1d2c8eea CH |
680 | /* |
681 | * kmalloc_track_caller is a special version of kmalloc that records the | |
682 | * calling function of the routine calling it for slab leak tracking instead | |
683 | * of just the calling function (confusing, eh?). | |
684 | * It's useful when the call to kmalloc comes from a widely-used standard | |
685 | * allocator where we care about the real place the memory allocation | |
686 | * request comes from. | |
687 | */ | |
93dd04ab | 688 | extern void *__kmalloc_track_caller(size_t size, gfp_t flags, unsigned long caller); |
1d2c8eea | 689 | #define kmalloc_track_caller(size, flags) \ |
ce71e27c | 690 | __kmalloc_track_caller(size, flags, _RET_IP_) |
1da177e4 | 691 | |
c37495d6 KC |
692 | static inline __alloc_size(1, 2) void *kmalloc_array_node(size_t n, size_t size, gfp_t flags, |
693 | int node) | |
5799b255 | 694 | { |
49b7f898 KC |
695 | size_t bytes; |
696 | ||
697 | if (unlikely(check_mul_overflow(n, size, &bytes))) | |
5799b255 JT |
698 | return NULL; |
699 | if (__builtin_constant_p(n) && __builtin_constant_p(size)) | |
49b7f898 KC |
700 | return kmalloc_node(bytes, flags, node); |
701 | return __kmalloc_node(bytes, flags, node); | |
5799b255 JT |
702 | } |
703 | ||
c37495d6 | 704 | static inline __alloc_size(1, 2) void *kcalloc_node(size_t n, size_t size, gfp_t flags, int node) |
5799b255 JT |
705 | { |
706 | return kmalloc_array_node(n, size, flags | __GFP_ZERO, node); | |
707 | } | |
708 | ||
709 | ||
97e2bde4 | 710 | #ifdef CONFIG_NUMA |
72d67229 | 711 | extern void *__kmalloc_node_track_caller(size_t size, gfp_t flags, int node, |
c37495d6 | 712 | unsigned long caller) __alloc_size(1); |
8b98c169 CH |
713 | #define kmalloc_node_track_caller(size, flags, node) \ |
714 | __kmalloc_node_track_caller(size, flags, node, \ | |
ce71e27c | 715 | _RET_IP_) |
2e892f43 | 716 | |
8b98c169 | 717 | #else /* CONFIG_NUMA */ |
8b98c169 CH |
718 | |
719 | #define kmalloc_node_track_caller(size, flags, node) \ | |
720 | kmalloc_track_caller(size, flags) | |
97e2bde4 | 721 | |
dfcd3610 | 722 | #endif /* CONFIG_NUMA */ |
10cef602 | 723 | |
81cda662 CL |
724 | /* |
725 | * Shortcuts | |
726 | */ | |
727 | static inline void *kmem_cache_zalloc(struct kmem_cache *k, gfp_t flags) | |
728 | { | |
729 | return kmem_cache_alloc(k, flags | __GFP_ZERO); | |
730 | } | |
731 | ||
732 | /** | |
733 | * kzalloc - allocate memory. The memory is set to zero. | |
734 | * @size: how many bytes of memory are required. | |
735 | * @flags: the type of memory to allocate (see kmalloc). | |
736 | */ | |
c37495d6 | 737 | static inline __alloc_size(1) void *kzalloc(size_t size, gfp_t flags) |
81cda662 CL |
738 | { |
739 | return kmalloc(size, flags | __GFP_ZERO); | |
740 | } | |
741 | ||
979b0fea JL |
742 | /** |
743 | * kzalloc_node - allocate zeroed memory from a particular memory node. | |
744 | * @size: how many bytes of memory are required. | |
745 | * @flags: the type of memory to allocate (see kmalloc). | |
746 | * @node: memory node from which to allocate | |
747 | */ | |
c37495d6 | 748 | static inline __alloc_size(1) void *kzalloc_node(size_t size, gfp_t flags, int node) |
979b0fea JL |
749 | { |
750 | return kmalloc_node(size, flags | __GFP_ZERO, node); | |
751 | } | |
752 | ||
56bcf40f KC |
753 | extern void *kvmalloc_node(size_t size, gfp_t flags, int node) __alloc_size(1); |
754 | static inline __alloc_size(1) void *kvmalloc(size_t size, gfp_t flags) | |
8587ca6f MWO |
755 | { |
756 | return kvmalloc_node(size, flags, NUMA_NO_NODE); | |
757 | } | |
56bcf40f | 758 | static inline __alloc_size(1) void *kvzalloc_node(size_t size, gfp_t flags, int node) |
8587ca6f MWO |
759 | { |
760 | return kvmalloc_node(size, flags | __GFP_ZERO, node); | |
761 | } | |
56bcf40f | 762 | static inline __alloc_size(1) void *kvzalloc(size_t size, gfp_t flags) |
8587ca6f MWO |
763 | { |
764 | return kvmalloc(size, flags | __GFP_ZERO); | |
765 | } | |
766 | ||
56bcf40f | 767 | static inline __alloc_size(1, 2) void *kvmalloc_array(size_t n, size_t size, gfp_t flags) |
8587ca6f MWO |
768 | { |
769 | size_t bytes; | |
770 | ||
771 | if (unlikely(check_mul_overflow(n, size, &bytes))) | |
772 | return NULL; | |
773 | ||
774 | return kvmalloc(bytes, flags); | |
775 | } | |
776 | ||
56bcf40f | 777 | static inline __alloc_size(1, 2) void *kvcalloc(size_t n, size_t size, gfp_t flags) |
8587ca6f MWO |
778 | { |
779 | return kvmalloc_array(n, size, flags | __GFP_ZERO); | |
780 | } | |
781 | ||
56bcf40f KC |
782 | extern void *kvrealloc(const void *p, size_t oldsize, size_t newsize, gfp_t flags) |
783 | __alloc_size(3); | |
8587ca6f MWO |
784 | extern void kvfree(const void *addr); |
785 | extern void kvfree_sensitive(const void *addr, size_t len); | |
786 | ||
07f361b2 | 787 | unsigned int kmem_cache_size(struct kmem_cache *s); |
7e85ee0c PE |
788 | void __init kmem_cache_init_late(void); |
789 | ||
6731d4f1 SAS |
790 | #if defined(CONFIG_SMP) && defined(CONFIG_SLAB) |
791 | int slab_prepare_cpu(unsigned int cpu); | |
792 | int slab_dead_cpu(unsigned int cpu); | |
793 | #else | |
794 | #define slab_prepare_cpu NULL | |
795 | #define slab_dead_cpu NULL | |
796 | #endif | |
797 | ||
1da177e4 | 798 | #endif /* _LINUX_SLAB_H */ |