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) |
6edf2576 FT |
32 | /* Indicate a kmalloc slab */ |
33 | #define SLAB_KMALLOC ((slab_flags_t __force)0x00001000U) | |
d50112ed | 34 | /* Align objs on cache lines */ |
4fd0b46e | 35 | #define SLAB_HWCACHE_ALIGN ((slab_flags_t __force)0x00002000U) |
d50112ed | 36 | /* Use GFP_DMA memory */ |
4fd0b46e | 37 | #define SLAB_CACHE_DMA ((slab_flags_t __force)0x00004000U) |
6d6ea1e9 NB |
38 | /* Use GFP_DMA32 memory */ |
39 | #define SLAB_CACHE_DMA32 ((slab_flags_t __force)0x00008000U) | |
d50112ed | 40 | /* DEBUG: Store the last owner for bug hunting */ |
4fd0b46e | 41 | #define SLAB_STORE_USER ((slab_flags_t __force)0x00010000U) |
d50112ed | 42 | /* Panic if kmem_cache_create() fails */ |
4fd0b46e | 43 | #define SLAB_PANIC ((slab_flags_t __force)0x00040000U) |
d7de4c1d | 44 | /* |
5f0d5a3a | 45 | * SLAB_TYPESAFE_BY_RCU - **WARNING** READ THIS! |
d7de4c1d PZ |
46 | * |
47 | * This delays freeing the SLAB page by a grace period, it does _NOT_ | |
48 | * delay object freeing. This means that if you do kmem_cache_free() | |
49 | * that memory location is free to be reused at any time. Thus it may | |
50 | * be possible to see another object there in the same RCU grace period. | |
51 | * | |
52 | * This feature only ensures the memory location backing the object | |
53 | * stays valid, the trick to using this is relying on an independent | |
54 | * object validation pass. Something like: | |
55 | * | |
56 | * rcu_read_lock() | |
57 | * again: | |
58 | * obj = lockless_lookup(key); | |
59 | * if (obj) { | |
60 | * if (!try_get_ref(obj)) // might fail for free objects | |
61 | * goto again; | |
62 | * | |
63 | * if (obj->key != key) { // not the object we expected | |
64 | * put_ref(obj); | |
65 | * goto again; | |
66 | * } | |
67 | * } | |
68 | * rcu_read_unlock(); | |
69 | * | |
68126702 JK |
70 | * This is useful if we need to approach a kernel structure obliquely, |
71 | * from its address obtained without the usual locking. We can lock | |
72 | * the structure to stabilize it and check it's still at the given address, | |
73 | * only if we can be sure that the memory has not been meanwhile reused | |
74 | * for some other kind of object (which our subsystem's lock might corrupt). | |
75 | * | |
76 | * rcu_read_lock before reading the address, then rcu_read_unlock after | |
77 | * taking the spinlock within the structure expected at that address. | |
5f0d5a3a | 78 | * |
e9f8a790 PM |
79 | * Note that it is not possible to acquire a lock within a structure |
80 | * allocated with SLAB_TYPESAFE_BY_RCU without first acquiring a reference | |
81 | * as described above. The reason is that SLAB_TYPESAFE_BY_RCU pages | |
82 | * are not zeroed before being given to the slab, which means that any | |
83 | * locks must be initialized after each and every kmem_struct_alloc(). | |
84 | * Alternatively, make the ctor passed to kmem_cache_create() initialize | |
85 | * the locks at page-allocation time, as is done in __i915_request_ctor(), | |
86 | * sighand_ctor(), and anon_vma_ctor(). Such a ctor permits readers | |
87 | * to safely acquire those ctor-initialized locks under rcu_read_lock() | |
88 | * protection. | |
89 | * | |
5f0d5a3a | 90 | * Note that SLAB_TYPESAFE_BY_RCU was originally named SLAB_DESTROY_BY_RCU. |
d7de4c1d | 91 | */ |
d50112ed | 92 | /* Defer freeing slabs to RCU */ |
4fd0b46e | 93 | #define SLAB_TYPESAFE_BY_RCU ((slab_flags_t __force)0x00080000U) |
d50112ed | 94 | /* Spread some memory over cpuset */ |
4fd0b46e | 95 | #define SLAB_MEM_SPREAD ((slab_flags_t __force)0x00100000U) |
d50112ed | 96 | /* Trace allocations and frees */ |
4fd0b46e | 97 | #define SLAB_TRACE ((slab_flags_t __force)0x00200000U) |
1da177e4 | 98 | |
30327acf TG |
99 | /* Flag to prevent checks on free */ |
100 | #ifdef CONFIG_DEBUG_OBJECTS | |
4fd0b46e | 101 | # define SLAB_DEBUG_OBJECTS ((slab_flags_t __force)0x00400000U) |
30327acf | 102 | #else |
4fd0b46e | 103 | # define SLAB_DEBUG_OBJECTS 0 |
30327acf TG |
104 | #endif |
105 | ||
d50112ed | 106 | /* Avoid kmemleak tracing */ |
4fd0b46e | 107 | #define SLAB_NOLEAKTRACE ((slab_flags_t __force)0x00800000U) |
d5cff635 | 108 | |
d50112ed | 109 | /* Fault injection mark */ |
4c13dd3b | 110 | #ifdef CONFIG_FAILSLAB |
4fd0b46e | 111 | # define SLAB_FAILSLAB ((slab_flags_t __force)0x02000000U) |
4c13dd3b | 112 | #else |
4fd0b46e | 113 | # define SLAB_FAILSLAB 0 |
4c13dd3b | 114 | #endif |
d50112ed | 115 | /* Account to memcg */ |
84c07d11 | 116 | #ifdef CONFIG_MEMCG_KMEM |
4fd0b46e | 117 | # define SLAB_ACCOUNT ((slab_flags_t __force)0x04000000U) |
230e9fc2 | 118 | #else |
4fd0b46e | 119 | # define SLAB_ACCOUNT 0 |
230e9fc2 | 120 | #endif |
2dff4405 | 121 | |
682ed089 | 122 | #ifdef CONFIG_KASAN_GENERIC |
4fd0b46e | 123 | #define SLAB_KASAN ((slab_flags_t __force)0x08000000U) |
7ed2f9e6 | 124 | #else |
4fd0b46e | 125 | #define SLAB_KASAN 0 |
7ed2f9e6 AP |
126 | #endif |
127 | ||
a285909f HY |
128 | /* |
129 | * Ignore user specified debugging flags. | |
130 | * Intended for caches created for self-tests so they have only flags | |
131 | * specified in the code and other flags are ignored. | |
132 | */ | |
133 | #define SLAB_NO_USER_FLAGS ((slab_flags_t __force)0x10000000U) | |
134 | ||
b84e04f1 IK |
135 | #ifdef CONFIG_KFENCE |
136 | #define SLAB_SKIP_KFENCE ((slab_flags_t __force)0x20000000U) | |
137 | #else | |
138 | #define SLAB_SKIP_KFENCE 0 | |
139 | #endif | |
140 | ||
e12ba74d | 141 | /* The following flags affect the page allocator grouping pages by mobility */ |
d50112ed | 142 | /* Objects are reclaimable */ |
3d97d976 | 143 | #ifndef CONFIG_SLUB_TINY |
4fd0b46e | 144 | #define SLAB_RECLAIM_ACCOUNT ((slab_flags_t __force)0x00020000U) |
3d97d976 VB |
145 | #else |
146 | #define SLAB_RECLAIM_ACCOUNT ((slab_flags_t __force)0) | |
147 | #endif | |
e12ba74d | 148 | #define SLAB_TEMPORARY SLAB_RECLAIM_ACCOUNT /* Objects are short-lived */ |
fcf8a1e4 | 149 | |
6cb8f913 CL |
150 | /* |
151 | * ZERO_SIZE_PTR will be returned for zero sized kmalloc requests. | |
152 | * | |
153 | * Dereferencing ZERO_SIZE_PTR will lead to a distinct access fault. | |
154 | * | |
155 | * ZERO_SIZE_PTR can be passed to kfree though in the same way that NULL can. | |
156 | * Both make kfree a no-op. | |
157 | */ | |
158 | #define ZERO_SIZE_PTR ((void *)16) | |
159 | ||
1d4ec7b1 | 160 | #define ZERO_OR_NULL_PTR(x) ((unsigned long)(x) <= \ |
6cb8f913 CL |
161 | (unsigned long)ZERO_SIZE_PTR) |
162 | ||
0316bec2 | 163 | #include <linux/kasan.h> |
3b0efdfa | 164 | |
88f2ef73 | 165 | struct list_lru; |
2633d7a0 | 166 | struct mem_cgroup; |
2e892f43 CL |
167 | /* |
168 | * struct kmem_cache related prototypes | |
169 | */ | |
fda90124 | 170 | bool slab_is_available(void); |
1da177e4 | 171 | |
f4957d5b AD |
172 | struct kmem_cache *kmem_cache_create(const char *name, unsigned int size, |
173 | unsigned int align, slab_flags_t flags, | |
8eb8284b DW |
174 | void (*ctor)(void *)); |
175 | struct kmem_cache *kmem_cache_create_usercopy(const char *name, | |
f4957d5b AD |
176 | unsigned int size, unsigned int align, |
177 | slab_flags_t flags, | |
7bbdb81e | 178 | unsigned int useroffset, unsigned int usersize, |
8eb8284b | 179 | void (*ctor)(void *)); |
72d67229 KC |
180 | void kmem_cache_destroy(struct kmem_cache *s); |
181 | int kmem_cache_shrink(struct kmem_cache *s); | |
2a4db7eb | 182 | |
0a31bd5f CL |
183 | /* |
184 | * Please use this macro to create slab caches. Simply specify the | |
185 | * name of the structure and maybe some flags that are listed above. | |
186 | * | |
187 | * The alignment of the struct determines object alignment. If you | |
188 | * f.e. add ____cacheline_aligned_in_smp to the struct declaration | |
189 | * then the objects will be properly aligned in SMP configurations. | |
190 | */ | |
8eb8284b DW |
191 | #define KMEM_CACHE(__struct, __flags) \ |
192 | kmem_cache_create(#__struct, sizeof(struct __struct), \ | |
193 | __alignof__(struct __struct), (__flags), NULL) | |
194 | ||
195 | /* | |
196 | * To whitelist a single field for copying to/from usercopy, use this | |
197 | * macro instead for KMEM_CACHE() above. | |
198 | */ | |
199 | #define KMEM_CACHE_USERCOPY(__struct, __flags, __field) \ | |
200 | kmem_cache_create_usercopy(#__struct, \ | |
201 | sizeof(struct __struct), \ | |
202 | __alignof__(struct __struct), (__flags), \ | |
203 | offsetof(struct __struct, __field), \ | |
204 | sizeof_field(struct __struct, __field), NULL) | |
0a31bd5f | 205 | |
34504667 CL |
206 | /* |
207 | * Common kmalloc functions provided by all allocators | |
208 | */ | |
9ed9cac1 | 209 | void * __must_check krealloc(const void *objp, size_t new_size, gfp_t flags) __realloc_size(2); |
72d67229 KC |
210 | void kfree(const void *objp); |
211 | void kfree_sensitive(const void *objp); | |
212 | size_t __ksize(const void *objp); | |
05a94065 KC |
213 | |
214 | /** | |
215 | * ksize - Report actual allocation size of associated object | |
216 | * | |
217 | * @objp: Pointer returned from a prior kmalloc()-family allocation. | |
218 | * | |
219 | * This should not be used for writing beyond the originally requested | |
220 | * allocation size. Either use krealloc() or round up the allocation size | |
221 | * with kmalloc_size_roundup() prior to allocation. If this is used to | |
222 | * access beyond the originally requested allocation size, UBSAN_BOUNDS | |
223 | * and/or FORTIFY_SOURCE may trip, since they only know about the | |
224 | * originally allocated size via the __alloc_size attribute. | |
225 | */ | |
72d67229 | 226 | size_t ksize(const void *objp); |
05a94065 | 227 | |
5bb1bb35 | 228 | #ifdef CONFIG_PRINTK |
8e7f37f2 PM |
229 | bool kmem_valid_obj(void *object); |
230 | void kmem_dump_obj(void *object); | |
5bb1bb35 | 231 | #endif |
34504667 | 232 | |
c601fd69 CL |
233 | /* |
234 | * Some archs want to perform DMA into kmalloc caches and need a guaranteed | |
235 | * alignment larger than the alignment of a 64-bit integer. | |
8cf9e121 | 236 | * Setting ARCH_DMA_MINALIGN in arch headers allows that. |
c601fd69 CL |
237 | */ |
238 | #if defined(ARCH_DMA_MINALIGN) && ARCH_DMA_MINALIGN > 8 | |
239 | #define ARCH_KMALLOC_MINALIGN ARCH_DMA_MINALIGN | |
240 | #define KMALLOC_MIN_SIZE ARCH_DMA_MINALIGN | |
241 | #define KMALLOC_SHIFT_LOW ilog2(ARCH_DMA_MINALIGN) | |
242 | #else | |
243 | #define ARCH_KMALLOC_MINALIGN __alignof__(unsigned long long) | |
244 | #endif | |
245 | ||
94a58c36 RV |
246 | /* |
247 | * Setting ARCH_SLAB_MINALIGN in arch headers allows a different alignment. | |
248 | * Intended for arches that get misalignment faults even for 64 bit integer | |
249 | * aligned buffers. | |
250 | */ | |
251 | #ifndef ARCH_SLAB_MINALIGN | |
252 | #define ARCH_SLAB_MINALIGN __alignof__(unsigned long long) | |
253 | #endif | |
254 | ||
d949a815 PC |
255 | /* |
256 | * Arches can define this function if they want to decide the minimum slab | |
257 | * alignment at runtime. The value returned by the function must be a power | |
258 | * of two and >= ARCH_SLAB_MINALIGN. | |
259 | */ | |
260 | #ifndef arch_slab_minalign | |
261 | static inline unsigned int arch_slab_minalign(void) | |
262 | { | |
263 | return ARCH_SLAB_MINALIGN; | |
264 | } | |
265 | #endif | |
266 | ||
94a58c36 | 267 | /* |
154036a3 AK |
268 | * kmem_cache_alloc and friends return pointers aligned to ARCH_SLAB_MINALIGN. |
269 | * kmalloc and friends return pointers aligned to both ARCH_KMALLOC_MINALIGN | |
270 | * and ARCH_SLAB_MINALIGN, but here we only assume the former alignment. | |
94a58c36 RV |
271 | */ |
272 | #define __assume_kmalloc_alignment __assume_aligned(ARCH_KMALLOC_MINALIGN) | |
273 | #define __assume_slab_alignment __assume_aligned(ARCH_SLAB_MINALIGN) | |
274 | #define __assume_page_alignment __assume_aligned(PAGE_SIZE) | |
275 | ||
0aa817f0 | 276 | /* |
95a05b42 CL |
277 | * Kmalloc array related definitions |
278 | */ | |
279 | ||
280 | #ifdef CONFIG_SLAB | |
281 | /* | |
d6a71648 HY |
282 | * SLAB and SLUB directly allocates requests fitting in to an order-1 page |
283 | * (PAGE_SIZE*2). Larger requests are passed to the page allocator. | |
0aa817f0 | 284 | */ |
d6a71648 | 285 | #define KMALLOC_SHIFT_HIGH (PAGE_SHIFT + 1) |
23baf831 | 286 | #define KMALLOC_SHIFT_MAX (MAX_ORDER + PAGE_SHIFT) |
c601fd69 | 287 | #ifndef KMALLOC_SHIFT_LOW |
95a05b42 | 288 | #define KMALLOC_SHIFT_LOW 5 |
c601fd69 | 289 | #endif |
069e2b35 CL |
290 | #endif |
291 | ||
292 | #ifdef CONFIG_SLUB | |
95a05b42 | 293 | #define KMALLOC_SHIFT_HIGH (PAGE_SHIFT + 1) |
23baf831 | 294 | #define KMALLOC_SHIFT_MAX (MAX_ORDER + PAGE_SHIFT) |
c601fd69 | 295 | #ifndef KMALLOC_SHIFT_LOW |
95a05b42 CL |
296 | #define KMALLOC_SHIFT_LOW 3 |
297 | #endif | |
c601fd69 | 298 | #endif |
0aa817f0 | 299 | |
95a05b42 CL |
300 | /* Maximum allocatable size */ |
301 | #define KMALLOC_MAX_SIZE (1UL << KMALLOC_SHIFT_MAX) | |
302 | /* Maximum size for which we actually use a slab cache */ | |
303 | #define KMALLOC_MAX_CACHE_SIZE (1UL << KMALLOC_SHIFT_HIGH) | |
d7cff4de | 304 | /* Maximum order allocatable via the slab allocator */ |
95a05b42 | 305 | #define KMALLOC_MAX_ORDER (KMALLOC_SHIFT_MAX - PAGE_SHIFT) |
0aa817f0 | 306 | |
ce6a5026 CL |
307 | /* |
308 | * Kmalloc subsystem. | |
309 | */ | |
c601fd69 | 310 | #ifndef KMALLOC_MIN_SIZE |
95a05b42 | 311 | #define KMALLOC_MIN_SIZE (1 << KMALLOC_SHIFT_LOW) |
ce6a5026 CL |
312 | #endif |
313 | ||
24f870d8 JK |
314 | /* |
315 | * This restriction comes from byte sized index implementation. | |
316 | * Page size is normally 2^12 bytes and, in this case, if we want to use | |
317 | * byte sized index which can represent 2^8 entries, the size of the object | |
318 | * should be equal or greater to 2^12 / 2^8 = 2^4 = 16. | |
319 | * If minimum size of kmalloc is less than 16, we use it as minimum object | |
320 | * size and give up to use byte sized index. | |
321 | */ | |
322 | #define SLAB_OBJ_MIN_SIZE (KMALLOC_MIN_SIZE < 16 ? \ | |
323 | (KMALLOC_MIN_SIZE) : 16) | |
324 | ||
1291523f VB |
325 | /* |
326 | * Whenever changing this, take care of that kmalloc_type() and | |
327 | * create_kmalloc_caches() still work as intended. | |
494c1dfe WL |
328 | * |
329 | * KMALLOC_NORMAL can contain only unaccounted objects whereas KMALLOC_CGROUP | |
330 | * is for accounted but unreclaimable and non-dma objects. All the other | |
331 | * kmem caches can have both accounted and unaccounted objects. | |
1291523f | 332 | */ |
cc252eae VB |
333 | enum kmalloc_cache_type { |
334 | KMALLOC_NORMAL = 0, | |
494c1dfe WL |
335 | #ifndef CONFIG_ZONE_DMA |
336 | KMALLOC_DMA = KMALLOC_NORMAL, | |
337 | #endif | |
338 | #ifndef CONFIG_MEMCG_KMEM | |
339 | KMALLOC_CGROUP = KMALLOC_NORMAL, | |
494c1dfe | 340 | #endif |
2f7c1c13 VB |
341 | #ifdef CONFIG_SLUB_TINY |
342 | KMALLOC_RECLAIM = KMALLOC_NORMAL, | |
343 | #else | |
1291523f | 344 | KMALLOC_RECLAIM, |
2f7c1c13 | 345 | #endif |
cc252eae VB |
346 | #ifdef CONFIG_ZONE_DMA |
347 | KMALLOC_DMA, | |
2f7c1c13 VB |
348 | #endif |
349 | #ifdef CONFIG_MEMCG_KMEM | |
350 | KMALLOC_CGROUP, | |
cc252eae VB |
351 | #endif |
352 | NR_KMALLOC_TYPES | |
353 | }; | |
354 | ||
cc252eae VB |
355 | extern struct kmem_cache * |
356 | kmalloc_caches[NR_KMALLOC_TYPES][KMALLOC_SHIFT_HIGH + 1]; | |
357 | ||
494c1dfe WL |
358 | /* |
359 | * Define gfp bits that should not be set for KMALLOC_NORMAL. | |
360 | */ | |
361 | #define KMALLOC_NOT_NORMAL_BITS \ | |
362 | (__GFP_RECLAIMABLE | \ | |
363 | (IS_ENABLED(CONFIG_ZONE_DMA) ? __GFP_DMA : 0) | \ | |
364 | (IS_ENABLED(CONFIG_MEMCG_KMEM) ? __GFP_ACCOUNT : 0)) | |
365 | ||
cc252eae VB |
366 | static __always_inline enum kmalloc_cache_type kmalloc_type(gfp_t flags) |
367 | { | |
4e45f712 VB |
368 | /* |
369 | * The most common case is KMALLOC_NORMAL, so test for it | |
494c1dfe | 370 | * with a single branch for all the relevant flags. |
4e45f712 | 371 | */ |
494c1dfe | 372 | if (likely((flags & KMALLOC_NOT_NORMAL_BITS) == 0)) |
4e45f712 | 373 | return KMALLOC_NORMAL; |
1291523f VB |
374 | |
375 | /* | |
494c1dfe WL |
376 | * At least one of the flags has to be set. Their priorities in |
377 | * decreasing order are: | |
378 | * 1) __GFP_DMA | |
379 | * 2) __GFP_RECLAIMABLE | |
380 | * 3) __GFP_ACCOUNT | |
1291523f | 381 | */ |
494c1dfe WL |
382 | if (IS_ENABLED(CONFIG_ZONE_DMA) && (flags & __GFP_DMA)) |
383 | return KMALLOC_DMA; | |
384 | if (!IS_ENABLED(CONFIG_MEMCG_KMEM) || (flags & __GFP_RECLAIMABLE)) | |
385 | return KMALLOC_RECLAIM; | |
386 | else | |
387 | return KMALLOC_CGROUP; | |
cc252eae VB |
388 | } |
389 | ||
ce6a5026 CL |
390 | /* |
391 | * Figure out which kmalloc slab an allocation of a certain size | |
392 | * belongs to. | |
393 | * 0 = zero alloc | |
394 | * 1 = 65 .. 96 bytes | |
1ed58b60 RV |
395 | * 2 = 129 .. 192 bytes |
396 | * n = 2^(n-1)+1 .. 2^n | |
588c7fa0 HY |
397 | * |
398 | * Note: __kmalloc_index() is compile-time optimized, and not runtime optimized; | |
399 | * typical usage is via kmalloc_index() and therefore evaluated at compile-time. | |
400 | * Callers where !size_is_constant should only be test modules, where runtime | |
401 | * overheads of __kmalloc_index() can be tolerated. Also see kmalloc_slab(). | |
ce6a5026 | 402 | */ |
588c7fa0 HY |
403 | static __always_inline unsigned int __kmalloc_index(size_t size, |
404 | bool size_is_constant) | |
ce6a5026 CL |
405 | { |
406 | if (!size) | |
407 | return 0; | |
408 | ||
409 | if (size <= KMALLOC_MIN_SIZE) | |
410 | return KMALLOC_SHIFT_LOW; | |
411 | ||
412 | if (KMALLOC_MIN_SIZE <= 32 && size > 64 && size <= 96) | |
413 | return 1; | |
414 | if (KMALLOC_MIN_SIZE <= 64 && size > 128 && size <= 192) | |
415 | return 2; | |
416 | if (size <= 8) return 3; | |
417 | if (size <= 16) return 4; | |
418 | if (size <= 32) return 5; | |
419 | if (size <= 64) return 6; | |
420 | if (size <= 128) return 7; | |
421 | if (size <= 256) return 8; | |
422 | if (size <= 512) return 9; | |
423 | if (size <= 1024) return 10; | |
424 | if (size <= 2 * 1024) return 11; | |
425 | if (size <= 4 * 1024) return 12; | |
426 | if (size <= 8 * 1024) return 13; | |
427 | if (size <= 16 * 1024) return 14; | |
428 | if (size <= 32 * 1024) return 15; | |
429 | if (size <= 64 * 1024) return 16; | |
430 | if (size <= 128 * 1024) return 17; | |
431 | if (size <= 256 * 1024) return 18; | |
432 | if (size <= 512 * 1024) return 19; | |
433 | if (size <= 1024 * 1024) return 20; | |
434 | if (size <= 2 * 1024 * 1024) return 21; | |
588c7fa0 | 435 | |
57b2b72a | 436 | if (!IS_ENABLED(CONFIG_PROFILE_ALL_BRANCHES) && size_is_constant) |
588c7fa0 HY |
437 | BUILD_BUG_ON_MSG(1, "unexpected size in kmalloc_index()"); |
438 | else | |
439 | BUG(); | |
ce6a5026 CL |
440 | |
441 | /* Will never be reached. Needed because the compiler may complain */ | |
442 | return -1; | |
443 | } | |
d6a71648 | 444 | static_assert(PAGE_SHIFT <= 20); |
588c7fa0 | 445 | #define kmalloc_index(s) __kmalloc_index(s, true) |
ce6a5026 | 446 | |
c37495d6 | 447 | void *__kmalloc(size_t size, gfp_t flags) __assume_kmalloc_alignment __alloc_size(1); |
838de63b VB |
448 | |
449 | /** | |
450 | * kmem_cache_alloc - Allocate an object | |
451 | * @cachep: The cache to allocate from. | |
452 | * @flags: See kmalloc(). | |
453 | * | |
454 | * Allocate an object from this cache. | |
455 | * See kmem_cache_zalloc() for a shortcut of adding __GFP_ZERO to flags. | |
456 | * | |
457 | * Return: pointer to the new object or %NULL in case of error | |
458 | */ | |
459 | void *kmem_cache_alloc(struct kmem_cache *cachep, gfp_t flags) __assume_slab_alignment __malloc; | |
88f2ef73 MS |
460 | void *kmem_cache_alloc_lru(struct kmem_cache *s, struct list_lru *lru, |
461 | gfp_t gfpflags) __assume_slab_alignment __malloc; | |
72d67229 | 462 | void kmem_cache_free(struct kmem_cache *s, void *objp); |
f1b6eb6e | 463 | |
484748f0 | 464 | /* |
9f706d68 | 465 | * Bulk allocation and freeing operations. These are accelerated in an |
484748f0 CL |
466 | * allocator specific way to avoid taking locks repeatedly or building |
467 | * metadata structures unnecessarily. | |
468 | * | |
469 | * Note that interrupts must be enabled when calling these functions. | |
470 | */ | |
72d67229 KC |
471 | void kmem_cache_free_bulk(struct kmem_cache *s, size_t size, void **p); |
472 | int kmem_cache_alloc_bulk(struct kmem_cache *s, gfp_t flags, size_t size, void **p); | |
484748f0 | 473 | |
ca257195 JDB |
474 | static __always_inline void kfree_bulk(size_t size, void **p) |
475 | { | |
476 | kmem_cache_free_bulk(NULL, size, p); | |
477 | } | |
478 | ||
c37495d6 KC |
479 | void *__kmalloc_node(size_t size, gfp_t flags, int node) __assume_kmalloc_alignment |
480 | __alloc_size(1); | |
72d67229 KC |
481 | void *kmem_cache_alloc_node(struct kmem_cache *s, gfp_t flags, int node) __assume_slab_alignment |
482 | __malloc; | |
f1b6eb6e | 483 | |
26a40990 HY |
484 | void *kmalloc_trace(struct kmem_cache *s, gfp_t flags, size_t size) |
485 | __assume_kmalloc_alignment __alloc_size(3); | |
f1b6eb6e | 486 | |
26a40990 HY |
487 | void *kmalloc_node_trace(struct kmem_cache *s, gfp_t gfpflags, |
488 | int node, size_t size) __assume_kmalloc_alignment | |
489 | __alloc_size(4); | |
e4c98d68 HY |
490 | void *kmalloc_large(size_t size, gfp_t flags) __assume_page_alignment |
491 | __alloc_size(1); | |
a0c3b940 HY |
492 | |
493 | void *kmalloc_large_node(size_t size, gfp_t flags, int node) __assume_page_alignment | |
494 | __alloc_size(1); | |
495 | ||
f1b6eb6e | 496 | /** |
838de63b | 497 | * kmalloc - allocate kernel memory |
f1b6eb6e | 498 | * @size: how many bytes of memory are required. |
838de63b | 499 | * @flags: describe the allocation context |
f1b6eb6e CL |
500 | * |
501 | * kmalloc is the normal method of allocating memory | |
502 | * for objects smaller than page size in the kernel. | |
7e3528c3 | 503 | * |
59bb4798 VB |
504 | * The allocated object address is aligned to at least ARCH_KMALLOC_MINALIGN |
505 | * bytes. For @size of power of two bytes, the alignment is also guaranteed | |
506 | * to be at least to the size. | |
507 | * | |
01598ba6 | 508 | * The @flags argument may be one of the GFP flags defined at |
e9d198f2 | 509 | * include/linux/gfp_types.h and described at |
01598ba6 | 510 | * :ref:`Documentation/core-api/mm-api.rst <mm-api-gfp-flags>` |
7e3528c3 | 511 | * |
01598ba6 | 512 | * The recommended usage of the @flags is described at |
2370ae4b | 513 | * :ref:`Documentation/core-api/memory-allocation.rst <memory_allocation>` |
7e3528c3 | 514 | * |
01598ba6 | 515 | * Below is a brief outline of the most useful GFP flags |
7e3528c3 | 516 | * |
01598ba6 MR |
517 | * %GFP_KERNEL |
518 | * Allocate normal kernel ram. May sleep. | |
7e3528c3 | 519 | * |
01598ba6 MR |
520 | * %GFP_NOWAIT |
521 | * Allocation will not sleep. | |
7e3528c3 | 522 | * |
01598ba6 MR |
523 | * %GFP_ATOMIC |
524 | * Allocation will not sleep. May use emergency pools. | |
7e3528c3 | 525 | * |
7e3528c3 RD |
526 | * Also it is possible to set different flags by OR'ing |
527 | * in one or more of the following additional @flags: | |
528 | * | |
838de63b VB |
529 | * %__GFP_ZERO |
530 | * Zero the allocated memory before returning. Also see kzalloc(). | |
531 | * | |
01598ba6 MR |
532 | * %__GFP_HIGH |
533 | * This allocation has high priority and may use emergency pools. | |
7e3528c3 | 534 | * |
01598ba6 MR |
535 | * %__GFP_NOFAIL |
536 | * Indicate that this allocation is in no way allowed to fail | |
537 | * (think twice before using). | |
7e3528c3 | 538 | * |
01598ba6 MR |
539 | * %__GFP_NORETRY |
540 | * If memory is not immediately available, | |
541 | * then give up at once. | |
7e3528c3 | 542 | * |
01598ba6 MR |
543 | * %__GFP_NOWARN |
544 | * If allocation fails, don't issue any warnings. | |
7e3528c3 | 545 | * |
01598ba6 MR |
546 | * %__GFP_RETRY_MAYFAIL |
547 | * Try really hard to succeed the allocation but fail | |
548 | * eventually. | |
f1b6eb6e | 549 | */ |
c37495d6 | 550 | static __always_inline __alloc_size(1) void *kmalloc(size_t size, gfp_t flags) |
f1b6eb6e | 551 | { |
6fa57d78 | 552 | if (__builtin_constant_p(size) && size) { |
cc252eae | 553 | unsigned int index; |
3bf01933 | 554 | |
f1b6eb6e CL |
555 | if (size > KMALLOC_MAX_CACHE_SIZE) |
556 | return kmalloc_large(size, flags); | |
f1b6eb6e | 557 | |
cc252eae | 558 | index = kmalloc_index(size); |
26a40990 | 559 | return kmalloc_trace( |
cc252eae VB |
560 | kmalloc_caches[kmalloc_type(flags)][index], |
561 | flags, size); | |
f1b6eb6e CL |
562 | } |
563 | return __kmalloc(size, flags); | |
564 | } | |
565 | ||
c37495d6 | 566 | static __always_inline __alloc_size(1) void *kmalloc_node(size_t size, gfp_t flags, int node) |
f1b6eb6e | 567 | { |
6fa57d78 | 568 | if (__builtin_constant_p(size) && size) { |
bf37d791 | 569 | unsigned int index; |
f1b6eb6e | 570 | |
bf37d791 HY |
571 | if (size > KMALLOC_MAX_CACHE_SIZE) |
572 | return kmalloc_large_node(size, flags, node); | |
573 | ||
574 | index = kmalloc_index(size); | |
26a40990 | 575 | return kmalloc_node_trace( |
bf37d791 | 576 | kmalloc_caches[kmalloc_type(flags)][index], |
26a40990 | 577 | flags, node, size); |
f1b6eb6e | 578 | } |
f1b6eb6e CL |
579 | return __kmalloc_node(size, flags, node); |
580 | } | |
581 | ||
e7efa615 MO |
582 | /** |
583 | * kmalloc_array - allocate memory for an array. | |
584 | * @n: number of elements. | |
585 | * @size: element size. | |
586 | * @flags: the type of memory to allocate (see kmalloc). | |
800590f5 | 587 | */ |
c37495d6 | 588 | static inline __alloc_size(1, 2) void *kmalloc_array(size_t n, size_t size, gfp_t flags) |
1da177e4 | 589 | { |
49b7f898 KC |
590 | size_t bytes; |
591 | ||
592 | if (unlikely(check_mul_overflow(n, size, &bytes))) | |
6193a2ff | 593 | return NULL; |
91c6a05f | 594 | if (__builtin_constant_p(n) && __builtin_constant_p(size)) |
49b7f898 KC |
595 | return kmalloc(bytes, flags); |
596 | return __kmalloc(bytes, flags); | |
a8203725 XW |
597 | } |
598 | ||
f0dbd2bd BG |
599 | /** |
600 | * krealloc_array - reallocate memory for an array. | |
601 | * @p: pointer to the memory chunk to reallocate | |
602 | * @new_n: new number of elements to alloc | |
603 | * @new_size: new size of a single member of the array | |
604 | * @flags: the type of memory to allocate (see kmalloc) | |
605 | */ | |
9ed9cac1 KC |
606 | static inline __realloc_size(2, 3) void * __must_check krealloc_array(void *p, |
607 | size_t new_n, | |
608 | size_t new_size, | |
609 | gfp_t flags) | |
f0dbd2bd BG |
610 | { |
611 | size_t bytes; | |
612 | ||
613 | if (unlikely(check_mul_overflow(new_n, new_size, &bytes))) | |
614 | return NULL; | |
615 | ||
616 | return krealloc(p, bytes, flags); | |
617 | } | |
618 | ||
a8203725 XW |
619 | /** |
620 | * kcalloc - allocate memory for an array. The memory is set to zero. | |
621 | * @n: number of elements. | |
622 | * @size: element size. | |
623 | * @flags: the type of memory to allocate (see kmalloc). | |
624 | */ | |
c37495d6 | 625 | static inline __alloc_size(1, 2) void *kcalloc(size_t n, size_t size, gfp_t flags) |
a8203725 XW |
626 | { |
627 | return kmalloc_array(n, size, flags | __GFP_ZERO); | |
1da177e4 LT |
628 | } |
629 | ||
c45248db HY |
630 | void *__kmalloc_node_track_caller(size_t size, gfp_t flags, int node, |
631 | unsigned long caller) __alloc_size(1); | |
632 | #define kmalloc_node_track_caller(size, flags, node) \ | |
633 | __kmalloc_node_track_caller(size, flags, node, \ | |
634 | _RET_IP_) | |
635 | ||
1d2c8eea CH |
636 | /* |
637 | * kmalloc_track_caller is a special version of kmalloc that records the | |
638 | * calling function of the routine calling it for slab leak tracking instead | |
639 | * of just the calling function (confusing, eh?). | |
640 | * It's useful when the call to kmalloc comes from a widely-used standard | |
641 | * allocator where we care about the real place the memory allocation | |
642 | * request comes from. | |
643 | */ | |
1d2c8eea | 644 | #define kmalloc_track_caller(size, flags) \ |
c45248db HY |
645 | __kmalloc_node_track_caller(size, flags, \ |
646 | NUMA_NO_NODE, _RET_IP_) | |
1da177e4 | 647 | |
c37495d6 KC |
648 | static inline __alloc_size(1, 2) void *kmalloc_array_node(size_t n, size_t size, gfp_t flags, |
649 | int node) | |
5799b255 | 650 | { |
49b7f898 KC |
651 | size_t bytes; |
652 | ||
653 | if (unlikely(check_mul_overflow(n, size, &bytes))) | |
5799b255 JT |
654 | return NULL; |
655 | if (__builtin_constant_p(n) && __builtin_constant_p(size)) | |
49b7f898 KC |
656 | return kmalloc_node(bytes, flags, node); |
657 | return __kmalloc_node(bytes, flags, node); | |
5799b255 JT |
658 | } |
659 | ||
c37495d6 | 660 | static inline __alloc_size(1, 2) void *kcalloc_node(size_t n, size_t size, gfp_t flags, int node) |
5799b255 JT |
661 | { |
662 | return kmalloc_array_node(n, size, flags | __GFP_ZERO, node); | |
663 | } | |
664 | ||
81cda662 CL |
665 | /* |
666 | * Shortcuts | |
667 | */ | |
668 | static inline void *kmem_cache_zalloc(struct kmem_cache *k, gfp_t flags) | |
669 | { | |
670 | return kmem_cache_alloc(k, flags | __GFP_ZERO); | |
671 | } | |
672 | ||
673 | /** | |
674 | * kzalloc - allocate memory. The memory is set to zero. | |
675 | * @size: how many bytes of memory are required. | |
676 | * @flags: the type of memory to allocate (see kmalloc). | |
677 | */ | |
c37495d6 | 678 | static inline __alloc_size(1) void *kzalloc(size_t size, gfp_t flags) |
81cda662 CL |
679 | { |
680 | return kmalloc(size, flags | __GFP_ZERO); | |
681 | } | |
682 | ||
979b0fea JL |
683 | /** |
684 | * kzalloc_node - allocate zeroed memory from a particular memory node. | |
685 | * @size: how many bytes of memory are required. | |
686 | * @flags: the type of memory to allocate (see kmalloc). | |
687 | * @node: memory node from which to allocate | |
688 | */ | |
c37495d6 | 689 | static inline __alloc_size(1) void *kzalloc_node(size_t size, gfp_t flags, int node) |
979b0fea JL |
690 | { |
691 | return kmalloc_node(size, flags | __GFP_ZERO, node); | |
692 | } | |
693 | ||
56bcf40f KC |
694 | extern void *kvmalloc_node(size_t size, gfp_t flags, int node) __alloc_size(1); |
695 | static inline __alloc_size(1) void *kvmalloc(size_t size, gfp_t flags) | |
8587ca6f MWO |
696 | { |
697 | return kvmalloc_node(size, flags, NUMA_NO_NODE); | |
698 | } | |
56bcf40f | 699 | static inline __alloc_size(1) void *kvzalloc_node(size_t size, gfp_t flags, int node) |
8587ca6f MWO |
700 | { |
701 | return kvmalloc_node(size, flags | __GFP_ZERO, node); | |
702 | } | |
56bcf40f | 703 | static inline __alloc_size(1) void *kvzalloc(size_t size, gfp_t flags) |
8587ca6f MWO |
704 | { |
705 | return kvmalloc(size, flags | __GFP_ZERO); | |
706 | } | |
707 | ||
56bcf40f | 708 | static inline __alloc_size(1, 2) void *kvmalloc_array(size_t n, size_t size, gfp_t flags) |
8587ca6f MWO |
709 | { |
710 | size_t bytes; | |
711 | ||
712 | if (unlikely(check_mul_overflow(n, size, &bytes))) | |
713 | return NULL; | |
714 | ||
715 | return kvmalloc(bytes, flags); | |
716 | } | |
717 | ||
56bcf40f | 718 | static inline __alloc_size(1, 2) void *kvcalloc(size_t n, size_t size, gfp_t flags) |
8587ca6f MWO |
719 | { |
720 | return kvmalloc_array(n, size, flags | __GFP_ZERO); | |
721 | } | |
722 | ||
56bcf40f | 723 | extern void *kvrealloc(const void *p, size_t oldsize, size_t newsize, gfp_t flags) |
9ed9cac1 | 724 | __realloc_size(3); |
8587ca6f MWO |
725 | extern void kvfree(const void *addr); |
726 | extern void kvfree_sensitive(const void *addr, size_t len); | |
727 | ||
07f361b2 | 728 | unsigned int kmem_cache_size(struct kmem_cache *s); |
05a94065 KC |
729 | |
730 | /** | |
731 | * kmalloc_size_roundup - Report allocation bucket size for the given size | |
732 | * | |
733 | * @size: Number of bytes to round up from. | |
734 | * | |
735 | * This returns the number of bytes that would be available in a kmalloc() | |
736 | * allocation of @size bytes. For example, a 126 byte request would be | |
737 | * rounded up to the next sized kmalloc bucket, 128 bytes. (This is strictly | |
738 | * for the general-purpose kmalloc()-based allocations, and is not for the | |
739 | * pre-sized kmem_cache_alloc()-based allocations.) | |
740 | * | |
741 | * Use this to kmalloc() the full bucket size ahead of time instead of using | |
742 | * ksize() to query the size after an allocation. | |
743 | */ | |
744 | size_t kmalloc_size_roundup(size_t size); | |
745 | ||
7e85ee0c PE |
746 | void __init kmem_cache_init_late(void); |
747 | ||
6731d4f1 SAS |
748 | #if defined(CONFIG_SMP) && defined(CONFIG_SLAB) |
749 | int slab_prepare_cpu(unsigned int cpu); | |
750 | int slab_dead_cpu(unsigned int cpu); | |
751 | #else | |
752 | #define slab_prepare_cpu NULL | |
753 | #define slab_dead_cpu NULL | |
754 | #endif | |
755 | ||
1da177e4 | 756 | #endif /* _LINUX_SLAB_H */ |