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> |
1b1cec4b | 16 | #include <linux/types.h> |
1f458cbf GC |
17 | #include <linux/workqueue.h> |
18 | ||
1da177e4 | 19 | |
2e892f43 CL |
20 | /* |
21 | * Flags to pass to kmem_cache_create(). | |
124dee09 | 22 | * The ones marked DEBUG are only valid if CONFIG_DEBUG_SLAB is set. |
1da177e4 | 23 | */ |
d50112ed | 24 | /* DEBUG: Perform (expensive) checks on alloc/free */ |
4fd0b46e | 25 | #define SLAB_CONSISTENCY_CHECKS ((slab_flags_t __force)0x00000100U) |
d50112ed | 26 | /* DEBUG: Red zone objs in a cache */ |
4fd0b46e | 27 | #define SLAB_RED_ZONE ((slab_flags_t __force)0x00000400U) |
d50112ed | 28 | /* DEBUG: Poison objects */ |
4fd0b46e | 29 | #define SLAB_POISON ((slab_flags_t __force)0x00000800U) |
d50112ed | 30 | /* Align objs on cache lines */ |
4fd0b46e | 31 | #define SLAB_HWCACHE_ALIGN ((slab_flags_t __force)0x00002000U) |
d50112ed | 32 | /* Use GFP_DMA memory */ |
4fd0b46e | 33 | #define SLAB_CACHE_DMA ((slab_flags_t __force)0x00004000U) |
d50112ed | 34 | /* DEBUG: Store the last owner for bug hunting */ |
4fd0b46e | 35 | #define SLAB_STORE_USER ((slab_flags_t __force)0x00010000U) |
d50112ed | 36 | /* Panic if kmem_cache_create() fails */ |
4fd0b46e | 37 | #define SLAB_PANIC ((slab_flags_t __force)0x00040000U) |
d7de4c1d | 38 | /* |
5f0d5a3a | 39 | * SLAB_TYPESAFE_BY_RCU - **WARNING** READ THIS! |
d7de4c1d PZ |
40 | * |
41 | * This delays freeing the SLAB page by a grace period, it does _NOT_ | |
42 | * delay object freeing. This means that if you do kmem_cache_free() | |
43 | * that memory location is free to be reused at any time. Thus it may | |
44 | * be possible to see another object there in the same RCU grace period. | |
45 | * | |
46 | * This feature only ensures the memory location backing the object | |
47 | * stays valid, the trick to using this is relying on an independent | |
48 | * object validation pass. Something like: | |
49 | * | |
50 | * rcu_read_lock() | |
51 | * again: | |
52 | * obj = lockless_lookup(key); | |
53 | * if (obj) { | |
54 | * if (!try_get_ref(obj)) // might fail for free objects | |
55 | * goto again; | |
56 | * | |
57 | * if (obj->key != key) { // not the object we expected | |
58 | * put_ref(obj); | |
59 | * goto again; | |
60 | * } | |
61 | * } | |
62 | * rcu_read_unlock(); | |
63 | * | |
68126702 JK |
64 | * This is useful if we need to approach a kernel structure obliquely, |
65 | * from its address obtained without the usual locking. We can lock | |
66 | * the structure to stabilize it and check it's still at the given address, | |
67 | * only if we can be sure that the memory has not been meanwhile reused | |
68 | * for some other kind of object (which our subsystem's lock might corrupt). | |
69 | * | |
70 | * rcu_read_lock before reading the address, then rcu_read_unlock after | |
71 | * taking the spinlock within the structure expected at that address. | |
5f0d5a3a PM |
72 | * |
73 | * Note that SLAB_TYPESAFE_BY_RCU was originally named SLAB_DESTROY_BY_RCU. | |
d7de4c1d | 74 | */ |
d50112ed | 75 | /* Defer freeing slabs to RCU */ |
4fd0b46e | 76 | #define SLAB_TYPESAFE_BY_RCU ((slab_flags_t __force)0x00080000U) |
d50112ed | 77 | /* Spread some memory over cpuset */ |
4fd0b46e | 78 | #define SLAB_MEM_SPREAD ((slab_flags_t __force)0x00100000U) |
d50112ed | 79 | /* Trace allocations and frees */ |
4fd0b46e | 80 | #define SLAB_TRACE ((slab_flags_t __force)0x00200000U) |
1da177e4 | 81 | |
30327acf TG |
82 | /* Flag to prevent checks on free */ |
83 | #ifdef CONFIG_DEBUG_OBJECTS | |
4fd0b46e | 84 | # define SLAB_DEBUG_OBJECTS ((slab_flags_t __force)0x00400000U) |
30327acf | 85 | #else |
4fd0b46e | 86 | # define SLAB_DEBUG_OBJECTS 0 |
30327acf TG |
87 | #endif |
88 | ||
d50112ed | 89 | /* Avoid kmemleak tracing */ |
4fd0b46e | 90 | #define SLAB_NOLEAKTRACE ((slab_flags_t __force)0x00800000U) |
d5cff635 | 91 | |
2dff4405 VN |
92 | /* Don't track use of uninitialized memory */ |
93 | #ifdef CONFIG_KMEMCHECK | |
4fd0b46e | 94 | # define SLAB_NOTRACK ((slab_flags_t __force)0x01000000U) |
2dff4405 | 95 | #else |
4fd0b46e | 96 | # define SLAB_NOTRACK 0 |
2dff4405 | 97 | #endif |
d50112ed | 98 | /* Fault injection mark */ |
4c13dd3b | 99 | #ifdef CONFIG_FAILSLAB |
4fd0b46e | 100 | # define SLAB_FAILSLAB ((slab_flags_t __force)0x02000000U) |
4c13dd3b | 101 | #else |
4fd0b46e | 102 | # define SLAB_FAILSLAB 0 |
4c13dd3b | 103 | #endif |
d50112ed | 104 | /* Account to memcg */ |
127424c8 | 105 | #if defined(CONFIG_MEMCG) && !defined(CONFIG_SLOB) |
4fd0b46e | 106 | # define SLAB_ACCOUNT ((slab_flags_t __force)0x04000000U) |
230e9fc2 | 107 | #else |
4fd0b46e | 108 | # define SLAB_ACCOUNT 0 |
230e9fc2 | 109 | #endif |
2dff4405 | 110 | |
7ed2f9e6 | 111 | #ifdef CONFIG_KASAN |
4fd0b46e | 112 | #define SLAB_KASAN ((slab_flags_t __force)0x08000000U) |
7ed2f9e6 | 113 | #else |
4fd0b46e | 114 | #define SLAB_KASAN 0 |
7ed2f9e6 AP |
115 | #endif |
116 | ||
e12ba74d | 117 | /* The following flags affect the page allocator grouping pages by mobility */ |
d50112ed | 118 | /* Objects are reclaimable */ |
4fd0b46e | 119 | #define SLAB_RECLAIM_ACCOUNT ((slab_flags_t __force)0x00020000U) |
e12ba74d | 120 | #define SLAB_TEMPORARY SLAB_RECLAIM_ACCOUNT /* Objects are short-lived */ |
6cb8f913 CL |
121 | /* |
122 | * ZERO_SIZE_PTR will be returned for zero sized kmalloc requests. | |
123 | * | |
124 | * Dereferencing ZERO_SIZE_PTR will lead to a distinct access fault. | |
125 | * | |
126 | * ZERO_SIZE_PTR can be passed to kfree though in the same way that NULL can. | |
127 | * Both make kfree a no-op. | |
128 | */ | |
129 | #define ZERO_SIZE_PTR ((void *)16) | |
130 | ||
1d4ec7b1 | 131 | #define ZERO_OR_NULL_PTR(x) ((unsigned long)(x) <= \ |
6cb8f913 CL |
132 | (unsigned long)ZERO_SIZE_PTR) |
133 | ||
f1b6eb6e | 134 | #include <linux/kmemleak.h> |
0316bec2 | 135 | #include <linux/kasan.h> |
3b0efdfa | 136 | |
2633d7a0 | 137 | struct mem_cgroup; |
2e892f43 CL |
138 | /* |
139 | * struct kmem_cache related prototypes | |
140 | */ | |
141 | void __init kmem_cache_init(void); | |
fda90124 | 142 | bool slab_is_available(void); |
1da177e4 | 143 | |
2e892f43 | 144 | struct kmem_cache *kmem_cache_create(const char *, size_t, size_t, |
d50112ed | 145 | slab_flags_t, |
51cc5068 | 146 | void (*)(void *)); |
2e892f43 CL |
147 | void kmem_cache_destroy(struct kmem_cache *); |
148 | int kmem_cache_shrink(struct kmem_cache *); | |
2a4db7eb VD |
149 | |
150 | void memcg_create_kmem_cache(struct mem_cgroup *, struct kmem_cache *); | |
151 | void memcg_deactivate_kmem_caches(struct mem_cgroup *); | |
152 | void memcg_destroy_kmem_caches(struct mem_cgroup *); | |
2e892f43 | 153 | |
0a31bd5f CL |
154 | /* |
155 | * Please use this macro to create slab caches. Simply specify the | |
156 | * name of the structure and maybe some flags that are listed above. | |
157 | * | |
158 | * The alignment of the struct determines object alignment. If you | |
159 | * f.e. add ____cacheline_aligned_in_smp to the struct declaration | |
160 | * then the objects will be properly aligned in SMP configurations. | |
161 | */ | |
162 | #define KMEM_CACHE(__struct, __flags) kmem_cache_create(#__struct,\ | |
163 | sizeof(struct __struct), __alignof__(struct __struct),\ | |
20c2df83 | 164 | (__flags), NULL) |
0a31bd5f | 165 | |
34504667 CL |
166 | /* |
167 | * Common kmalloc functions provided by all allocators | |
168 | */ | |
169 | void * __must_check __krealloc(const void *, size_t, gfp_t); | |
170 | void * __must_check krealloc(const void *, size_t, gfp_t); | |
171 | void kfree(const void *); | |
172 | void kzfree(const void *); | |
173 | size_t ksize(const void *); | |
174 | ||
f5509cc1 KC |
175 | #ifdef CONFIG_HAVE_HARDENED_USERCOPY_ALLOCATOR |
176 | const char *__check_heap_object(const void *ptr, unsigned long n, | |
177 | struct page *page); | |
178 | #else | |
179 | static inline const char *__check_heap_object(const void *ptr, | |
180 | unsigned long n, | |
181 | struct page *page) | |
182 | { | |
183 | return NULL; | |
184 | } | |
185 | #endif | |
186 | ||
c601fd69 CL |
187 | /* |
188 | * Some archs want to perform DMA into kmalloc caches and need a guaranteed | |
189 | * alignment larger than the alignment of a 64-bit integer. | |
190 | * Setting ARCH_KMALLOC_MINALIGN in arch headers allows that. | |
191 | */ | |
192 | #if defined(ARCH_DMA_MINALIGN) && ARCH_DMA_MINALIGN > 8 | |
193 | #define ARCH_KMALLOC_MINALIGN ARCH_DMA_MINALIGN | |
194 | #define KMALLOC_MIN_SIZE ARCH_DMA_MINALIGN | |
195 | #define KMALLOC_SHIFT_LOW ilog2(ARCH_DMA_MINALIGN) | |
196 | #else | |
197 | #define ARCH_KMALLOC_MINALIGN __alignof__(unsigned long long) | |
198 | #endif | |
199 | ||
94a58c36 RV |
200 | /* |
201 | * Setting ARCH_SLAB_MINALIGN in arch headers allows a different alignment. | |
202 | * Intended for arches that get misalignment faults even for 64 bit integer | |
203 | * aligned buffers. | |
204 | */ | |
205 | #ifndef ARCH_SLAB_MINALIGN | |
206 | #define ARCH_SLAB_MINALIGN __alignof__(unsigned long long) | |
207 | #endif | |
208 | ||
209 | /* | |
210 | * kmalloc and friends return ARCH_KMALLOC_MINALIGN aligned | |
211 | * pointers. kmem_cache_alloc and friends return ARCH_SLAB_MINALIGN | |
212 | * aligned pointers. | |
213 | */ | |
214 | #define __assume_kmalloc_alignment __assume_aligned(ARCH_KMALLOC_MINALIGN) | |
215 | #define __assume_slab_alignment __assume_aligned(ARCH_SLAB_MINALIGN) | |
216 | #define __assume_page_alignment __assume_aligned(PAGE_SIZE) | |
217 | ||
0aa817f0 | 218 | /* |
95a05b42 CL |
219 | * Kmalloc array related definitions |
220 | */ | |
221 | ||
222 | #ifdef CONFIG_SLAB | |
223 | /* | |
224 | * The largest kmalloc size supported by the SLAB allocators is | |
0aa817f0 CL |
225 | * 32 megabyte (2^25) or the maximum allocatable page order if that is |
226 | * less than 32 MB. | |
227 | * | |
228 | * WARNING: Its not easy to increase this value since the allocators have | |
229 | * to do various tricks to work around compiler limitations in order to | |
230 | * ensure proper constant folding. | |
231 | */ | |
debee076 CL |
232 | #define KMALLOC_SHIFT_HIGH ((MAX_ORDER + PAGE_SHIFT - 1) <= 25 ? \ |
233 | (MAX_ORDER + PAGE_SHIFT - 1) : 25) | |
95a05b42 | 234 | #define KMALLOC_SHIFT_MAX KMALLOC_SHIFT_HIGH |
c601fd69 | 235 | #ifndef KMALLOC_SHIFT_LOW |
95a05b42 | 236 | #define KMALLOC_SHIFT_LOW 5 |
c601fd69 | 237 | #endif |
069e2b35 CL |
238 | #endif |
239 | ||
240 | #ifdef CONFIG_SLUB | |
95a05b42 | 241 | /* |
433a91ff DH |
242 | * SLUB directly allocates requests fitting in to an order-1 page |
243 | * (PAGE_SIZE*2). Larger requests are passed to the page allocator. | |
95a05b42 CL |
244 | */ |
245 | #define KMALLOC_SHIFT_HIGH (PAGE_SHIFT + 1) | |
bb1107f7 | 246 | #define KMALLOC_SHIFT_MAX (MAX_ORDER + PAGE_SHIFT - 1) |
c601fd69 | 247 | #ifndef KMALLOC_SHIFT_LOW |
95a05b42 CL |
248 | #define KMALLOC_SHIFT_LOW 3 |
249 | #endif | |
c601fd69 | 250 | #endif |
0aa817f0 | 251 | |
069e2b35 CL |
252 | #ifdef CONFIG_SLOB |
253 | /* | |
433a91ff | 254 | * SLOB passes all requests larger than one page to the page allocator. |
069e2b35 CL |
255 | * No kmalloc array is necessary since objects of different sizes can |
256 | * be allocated from the same page. | |
257 | */ | |
069e2b35 | 258 | #define KMALLOC_SHIFT_HIGH PAGE_SHIFT |
bb1107f7 | 259 | #define KMALLOC_SHIFT_MAX (MAX_ORDER + PAGE_SHIFT - 1) |
069e2b35 CL |
260 | #ifndef KMALLOC_SHIFT_LOW |
261 | #define KMALLOC_SHIFT_LOW 3 | |
262 | #endif | |
263 | #endif | |
264 | ||
95a05b42 CL |
265 | /* Maximum allocatable size */ |
266 | #define KMALLOC_MAX_SIZE (1UL << KMALLOC_SHIFT_MAX) | |
267 | /* Maximum size for which we actually use a slab cache */ | |
268 | #define KMALLOC_MAX_CACHE_SIZE (1UL << KMALLOC_SHIFT_HIGH) | |
269 | /* Maximum order allocatable via the slab allocagtor */ | |
270 | #define KMALLOC_MAX_ORDER (KMALLOC_SHIFT_MAX - PAGE_SHIFT) | |
0aa817f0 | 271 | |
ce6a5026 CL |
272 | /* |
273 | * Kmalloc subsystem. | |
274 | */ | |
c601fd69 | 275 | #ifndef KMALLOC_MIN_SIZE |
95a05b42 | 276 | #define KMALLOC_MIN_SIZE (1 << KMALLOC_SHIFT_LOW) |
ce6a5026 CL |
277 | #endif |
278 | ||
24f870d8 JK |
279 | /* |
280 | * This restriction comes from byte sized index implementation. | |
281 | * Page size is normally 2^12 bytes and, in this case, if we want to use | |
282 | * byte sized index which can represent 2^8 entries, the size of the object | |
283 | * should be equal or greater to 2^12 / 2^8 = 2^4 = 16. | |
284 | * If minimum size of kmalloc is less than 16, we use it as minimum object | |
285 | * size and give up to use byte sized index. | |
286 | */ | |
287 | #define SLAB_OBJ_MIN_SIZE (KMALLOC_MIN_SIZE < 16 ? \ | |
288 | (KMALLOC_MIN_SIZE) : 16) | |
289 | ||
069e2b35 | 290 | #ifndef CONFIG_SLOB |
9425c58e CL |
291 | extern struct kmem_cache *kmalloc_caches[KMALLOC_SHIFT_HIGH + 1]; |
292 | #ifdef CONFIG_ZONE_DMA | |
293 | extern struct kmem_cache *kmalloc_dma_caches[KMALLOC_SHIFT_HIGH + 1]; | |
294 | #endif | |
295 | ||
ce6a5026 CL |
296 | /* |
297 | * Figure out which kmalloc slab an allocation of a certain size | |
298 | * belongs to. | |
299 | * 0 = zero alloc | |
300 | * 1 = 65 .. 96 bytes | |
1ed58b60 RV |
301 | * 2 = 129 .. 192 bytes |
302 | * n = 2^(n-1)+1 .. 2^n | |
ce6a5026 CL |
303 | */ |
304 | static __always_inline int kmalloc_index(size_t size) | |
305 | { | |
306 | if (!size) | |
307 | return 0; | |
308 | ||
309 | if (size <= KMALLOC_MIN_SIZE) | |
310 | return KMALLOC_SHIFT_LOW; | |
311 | ||
312 | if (KMALLOC_MIN_SIZE <= 32 && size > 64 && size <= 96) | |
313 | return 1; | |
314 | if (KMALLOC_MIN_SIZE <= 64 && size > 128 && size <= 192) | |
315 | return 2; | |
316 | if (size <= 8) return 3; | |
317 | if (size <= 16) return 4; | |
318 | if (size <= 32) return 5; | |
319 | if (size <= 64) return 6; | |
320 | if (size <= 128) return 7; | |
321 | if (size <= 256) return 8; | |
322 | if (size <= 512) return 9; | |
323 | if (size <= 1024) return 10; | |
324 | if (size <= 2 * 1024) return 11; | |
325 | if (size <= 4 * 1024) return 12; | |
326 | if (size <= 8 * 1024) return 13; | |
327 | if (size <= 16 * 1024) return 14; | |
328 | if (size <= 32 * 1024) return 15; | |
329 | if (size <= 64 * 1024) return 16; | |
330 | if (size <= 128 * 1024) return 17; | |
331 | if (size <= 256 * 1024) return 18; | |
332 | if (size <= 512 * 1024) return 19; | |
333 | if (size <= 1024 * 1024) return 20; | |
334 | if (size <= 2 * 1024 * 1024) return 21; | |
335 | if (size <= 4 * 1024 * 1024) return 22; | |
336 | if (size <= 8 * 1024 * 1024) return 23; | |
337 | if (size <= 16 * 1024 * 1024) return 24; | |
338 | if (size <= 32 * 1024 * 1024) return 25; | |
339 | if (size <= 64 * 1024 * 1024) return 26; | |
340 | BUG(); | |
341 | ||
342 | /* Will never be reached. Needed because the compiler may complain */ | |
343 | return -1; | |
344 | } | |
069e2b35 | 345 | #endif /* !CONFIG_SLOB */ |
ce6a5026 | 346 | |
48a27055 RV |
347 | void *__kmalloc(size_t size, gfp_t flags) __assume_kmalloc_alignment __malloc; |
348 | void *kmem_cache_alloc(struct kmem_cache *, gfp_t flags) __assume_slab_alignment __malloc; | |
2a4db7eb | 349 | void kmem_cache_free(struct kmem_cache *, void *); |
f1b6eb6e | 350 | |
484748f0 | 351 | /* |
9f706d68 | 352 | * Bulk allocation and freeing operations. These are accelerated in an |
484748f0 CL |
353 | * allocator specific way to avoid taking locks repeatedly or building |
354 | * metadata structures unnecessarily. | |
355 | * | |
356 | * Note that interrupts must be enabled when calling these functions. | |
357 | */ | |
358 | void kmem_cache_free_bulk(struct kmem_cache *, size_t, void **); | |
865762a8 | 359 | int kmem_cache_alloc_bulk(struct kmem_cache *, gfp_t, size_t, void **); |
484748f0 | 360 | |
ca257195 JDB |
361 | /* |
362 | * Caller must not use kfree_bulk() on memory not originally allocated | |
363 | * by kmalloc(), because the SLOB allocator cannot handle this. | |
364 | */ | |
365 | static __always_inline void kfree_bulk(size_t size, void **p) | |
366 | { | |
367 | kmem_cache_free_bulk(NULL, size, p); | |
368 | } | |
369 | ||
f1b6eb6e | 370 | #ifdef CONFIG_NUMA |
48a27055 RV |
371 | void *__kmalloc_node(size_t size, gfp_t flags, int node) __assume_kmalloc_alignment __malloc; |
372 | void *kmem_cache_alloc_node(struct kmem_cache *, gfp_t flags, int node) __assume_slab_alignment __malloc; | |
f1b6eb6e CL |
373 | #else |
374 | static __always_inline void *__kmalloc_node(size_t size, gfp_t flags, int node) | |
375 | { | |
376 | return __kmalloc(size, flags); | |
377 | } | |
378 | ||
379 | static __always_inline void *kmem_cache_alloc_node(struct kmem_cache *s, gfp_t flags, int node) | |
380 | { | |
381 | return kmem_cache_alloc(s, flags); | |
382 | } | |
383 | #endif | |
384 | ||
385 | #ifdef CONFIG_TRACING | |
48a27055 | 386 | extern void *kmem_cache_alloc_trace(struct kmem_cache *, gfp_t, size_t) __assume_slab_alignment __malloc; |
f1b6eb6e CL |
387 | |
388 | #ifdef CONFIG_NUMA | |
389 | extern void *kmem_cache_alloc_node_trace(struct kmem_cache *s, | |
390 | gfp_t gfpflags, | |
48a27055 | 391 | int node, size_t size) __assume_slab_alignment __malloc; |
f1b6eb6e CL |
392 | #else |
393 | static __always_inline void * | |
394 | kmem_cache_alloc_node_trace(struct kmem_cache *s, | |
395 | gfp_t gfpflags, | |
396 | int node, size_t size) | |
397 | { | |
398 | return kmem_cache_alloc_trace(s, gfpflags, size); | |
399 | } | |
400 | #endif /* CONFIG_NUMA */ | |
401 | ||
402 | #else /* CONFIG_TRACING */ | |
403 | static __always_inline void *kmem_cache_alloc_trace(struct kmem_cache *s, | |
404 | gfp_t flags, size_t size) | |
405 | { | |
0316bec2 AR |
406 | void *ret = kmem_cache_alloc(s, flags); |
407 | ||
505f5dcb | 408 | kasan_kmalloc(s, ret, size, flags); |
0316bec2 | 409 | return ret; |
f1b6eb6e CL |
410 | } |
411 | ||
412 | static __always_inline void * | |
413 | kmem_cache_alloc_node_trace(struct kmem_cache *s, | |
414 | gfp_t gfpflags, | |
415 | int node, size_t size) | |
416 | { | |
0316bec2 AR |
417 | void *ret = kmem_cache_alloc_node(s, gfpflags, node); |
418 | ||
505f5dcb | 419 | kasan_kmalloc(s, ret, size, gfpflags); |
0316bec2 | 420 | return ret; |
f1b6eb6e CL |
421 | } |
422 | #endif /* CONFIG_TRACING */ | |
423 | ||
48a27055 | 424 | extern void *kmalloc_order(size_t size, gfp_t flags, unsigned int order) __assume_page_alignment __malloc; |
f1b6eb6e CL |
425 | |
426 | #ifdef CONFIG_TRACING | |
48a27055 | 427 | extern void *kmalloc_order_trace(size_t size, gfp_t flags, unsigned int order) __assume_page_alignment __malloc; |
f1b6eb6e CL |
428 | #else |
429 | static __always_inline void * | |
430 | kmalloc_order_trace(size_t size, gfp_t flags, unsigned int order) | |
431 | { | |
432 | return kmalloc_order(size, flags, order); | |
433 | } | |
ce6a5026 CL |
434 | #endif |
435 | ||
f1b6eb6e CL |
436 | static __always_inline void *kmalloc_large(size_t size, gfp_t flags) |
437 | { | |
438 | unsigned int order = get_order(size); | |
439 | return kmalloc_order_trace(size, flags, order); | |
440 | } | |
441 | ||
442 | /** | |
443 | * kmalloc - allocate memory | |
444 | * @size: how many bytes of memory are required. | |
7e3528c3 | 445 | * @flags: the type of memory to allocate. |
f1b6eb6e CL |
446 | * |
447 | * kmalloc is the normal method of allocating memory | |
448 | * for objects smaller than page size in the kernel. | |
7e3528c3 RD |
449 | * |
450 | * The @flags argument may be one of: | |
451 | * | |
452 | * %GFP_USER - Allocate memory on behalf of user. May sleep. | |
453 | * | |
454 | * %GFP_KERNEL - Allocate normal kernel ram. May sleep. | |
455 | * | |
456 | * %GFP_ATOMIC - Allocation will not sleep. May use emergency pools. | |
457 | * For example, use this inside interrupt handlers. | |
458 | * | |
459 | * %GFP_HIGHUSER - Allocate pages from high memory. | |
460 | * | |
461 | * %GFP_NOIO - Do not do any I/O at all while trying to get memory. | |
462 | * | |
463 | * %GFP_NOFS - Do not make any fs calls while trying to get memory. | |
464 | * | |
465 | * %GFP_NOWAIT - Allocation will not sleep. | |
466 | * | |
e97ca8e5 | 467 | * %__GFP_THISNODE - Allocate node-local memory only. |
7e3528c3 RD |
468 | * |
469 | * %GFP_DMA - Allocation suitable for DMA. | |
470 | * Should only be used for kmalloc() caches. Otherwise, use a | |
471 | * slab created with SLAB_DMA. | |
472 | * | |
473 | * Also it is possible to set different flags by OR'ing | |
474 | * in one or more of the following additional @flags: | |
475 | * | |
476 | * %__GFP_COLD - Request cache-cold pages instead of | |
477 | * trying to return cache-warm pages. | |
478 | * | |
479 | * %__GFP_HIGH - This allocation has high priority and may use emergency pools. | |
480 | * | |
481 | * %__GFP_NOFAIL - Indicate that this allocation is in no way allowed to fail | |
482 | * (think twice before using). | |
483 | * | |
484 | * %__GFP_NORETRY - If memory is not immediately available, | |
485 | * then give up at once. | |
486 | * | |
487 | * %__GFP_NOWARN - If allocation fails, don't issue any warnings. | |
488 | * | |
dcda9b04 MH |
489 | * %__GFP_RETRY_MAYFAIL - Try really hard to succeed the allocation but fail |
490 | * eventually. | |
7e3528c3 RD |
491 | * |
492 | * There are other flags available as well, but these are not intended | |
493 | * for general use, and so are not documented here. For a full list of | |
494 | * potential flags, always refer to linux/gfp.h. | |
f1b6eb6e CL |
495 | */ |
496 | static __always_inline void *kmalloc(size_t size, gfp_t flags) | |
497 | { | |
498 | if (__builtin_constant_p(size)) { | |
499 | if (size > KMALLOC_MAX_CACHE_SIZE) | |
500 | return kmalloc_large(size, flags); | |
501 | #ifndef CONFIG_SLOB | |
502 | if (!(flags & GFP_DMA)) { | |
503 | int index = kmalloc_index(size); | |
504 | ||
505 | if (!index) | |
506 | return ZERO_SIZE_PTR; | |
507 | ||
508 | return kmem_cache_alloc_trace(kmalloc_caches[index], | |
509 | flags, size); | |
510 | } | |
511 | #endif | |
512 | } | |
513 | return __kmalloc(size, flags); | |
514 | } | |
515 | ||
ce6a5026 CL |
516 | /* |
517 | * Determine size used for the nth kmalloc cache. | |
518 | * return size or 0 if a kmalloc cache for that | |
519 | * size does not exist | |
520 | */ | |
521 | static __always_inline int kmalloc_size(int n) | |
522 | { | |
069e2b35 | 523 | #ifndef CONFIG_SLOB |
ce6a5026 CL |
524 | if (n > 2) |
525 | return 1 << n; | |
526 | ||
527 | if (n == 1 && KMALLOC_MIN_SIZE <= 32) | |
528 | return 96; | |
529 | ||
530 | if (n == 2 && KMALLOC_MIN_SIZE <= 64) | |
531 | return 192; | |
069e2b35 | 532 | #endif |
ce6a5026 CL |
533 | return 0; |
534 | } | |
ce6a5026 | 535 | |
f1b6eb6e CL |
536 | static __always_inline void *kmalloc_node(size_t size, gfp_t flags, int node) |
537 | { | |
538 | #ifndef CONFIG_SLOB | |
539 | if (__builtin_constant_p(size) && | |
23774a2f | 540 | size <= KMALLOC_MAX_CACHE_SIZE && !(flags & GFP_DMA)) { |
f1b6eb6e CL |
541 | int i = kmalloc_index(size); |
542 | ||
543 | if (!i) | |
544 | return ZERO_SIZE_PTR; | |
545 | ||
546 | return kmem_cache_alloc_node_trace(kmalloc_caches[i], | |
547 | flags, node, size); | |
548 | } | |
549 | #endif | |
550 | return __kmalloc_node(size, flags, node); | |
551 | } | |
552 | ||
f7ce3190 VD |
553 | struct memcg_cache_array { |
554 | struct rcu_head rcu; | |
555 | struct kmem_cache *entries[0]; | |
556 | }; | |
557 | ||
ba6c496e GC |
558 | /* |
559 | * This is the main placeholder for memcg-related information in kmem caches. | |
ba6c496e GC |
560 | * Both the root cache and the child caches will have it. For the root cache, |
561 | * this will hold a dynamically allocated array large enough to hold | |
f8570263 VD |
562 | * information about the currently limited memcgs in the system. To allow the |
563 | * array to be accessed without taking any locks, on relocation we free the old | |
564 | * version only after a grace period. | |
ba6c496e | 565 | * |
9eeadc8b | 566 | * Root and child caches hold different metadata. |
ba6c496e | 567 | * |
9eeadc8b TH |
568 | * @root_cache: Common to root and child caches. NULL for root, pointer to |
569 | * the root cache for children. | |
426589f5 | 570 | * |
9eeadc8b TH |
571 | * The following fields are specific to root caches. |
572 | * | |
573 | * @memcg_caches: kmemcg ID indexed table of child caches. This table is | |
574 | * used to index child cachces during allocation and cleared | |
575 | * early during shutdown. | |
576 | * | |
510ded33 TH |
577 | * @root_caches_node: List node for slab_root_caches list. |
578 | * | |
9eeadc8b TH |
579 | * @children: List of all child caches. While the child caches are also |
580 | * reachable through @memcg_caches, a child cache remains on | |
581 | * this list until it is actually destroyed. | |
582 | * | |
583 | * The following fields are specific to child caches. | |
584 | * | |
585 | * @memcg: Pointer to the memcg this cache belongs to. | |
586 | * | |
587 | * @children_node: List node for @root_cache->children list. | |
bc2791f8 TH |
588 | * |
589 | * @kmem_caches_node: List node for @memcg->kmem_caches list. | |
ba6c496e GC |
590 | */ |
591 | struct memcg_cache_params { | |
9eeadc8b | 592 | struct kmem_cache *root_cache; |
ba6c496e | 593 | union { |
9eeadc8b TH |
594 | struct { |
595 | struct memcg_cache_array __rcu *memcg_caches; | |
510ded33 | 596 | struct list_head __root_caches_node; |
9eeadc8b TH |
597 | struct list_head children; |
598 | }; | |
2633d7a0 GC |
599 | struct { |
600 | struct mem_cgroup *memcg; | |
9eeadc8b | 601 | struct list_head children_node; |
bc2791f8 | 602 | struct list_head kmem_caches_node; |
01fb58bc TH |
603 | |
604 | void (*deact_fn)(struct kmem_cache *); | |
605 | union { | |
606 | struct rcu_head deact_rcu_head; | |
607 | struct work_struct deact_work; | |
608 | }; | |
2633d7a0 | 609 | }; |
ba6c496e GC |
610 | }; |
611 | }; | |
612 | ||
2633d7a0 GC |
613 | int memcg_update_all_caches(int num_memcgs); |
614 | ||
e7efa615 MO |
615 | /** |
616 | * kmalloc_array - allocate memory for an array. | |
617 | * @n: number of elements. | |
618 | * @size: element size. | |
619 | * @flags: the type of memory to allocate (see kmalloc). | |
800590f5 | 620 | */ |
a8203725 | 621 | static inline void *kmalloc_array(size_t n, size_t size, gfp_t flags) |
1da177e4 | 622 | { |
a3860c1c | 623 | if (size != 0 && n > SIZE_MAX / size) |
6193a2ff | 624 | return NULL; |
91c6a05f AD |
625 | if (__builtin_constant_p(n) && __builtin_constant_p(size)) |
626 | return kmalloc(n * size, flags); | |
a8203725 XW |
627 | return __kmalloc(n * size, flags); |
628 | } | |
629 | ||
630 | /** | |
631 | * kcalloc - allocate memory for an array. The memory is set to zero. | |
632 | * @n: number of elements. | |
633 | * @size: element size. | |
634 | * @flags: the type of memory to allocate (see kmalloc). | |
635 | */ | |
636 | static inline void *kcalloc(size_t n, size_t size, gfp_t flags) | |
637 | { | |
638 | return kmalloc_array(n, size, flags | __GFP_ZERO); | |
1da177e4 LT |
639 | } |
640 | ||
1d2c8eea CH |
641 | /* |
642 | * kmalloc_track_caller is a special version of kmalloc that records the | |
643 | * calling function of the routine calling it for slab leak tracking instead | |
644 | * of just the calling function (confusing, eh?). | |
645 | * It's useful when the call to kmalloc comes from a widely-used standard | |
646 | * allocator where we care about the real place the memory allocation | |
647 | * request comes from. | |
648 | */ | |
ce71e27c | 649 | extern void *__kmalloc_track_caller(size_t, gfp_t, unsigned long); |
1d2c8eea | 650 | #define kmalloc_track_caller(size, flags) \ |
ce71e27c | 651 | __kmalloc_track_caller(size, flags, _RET_IP_) |
1da177e4 | 652 | |
5799b255 JT |
653 | static inline void *kmalloc_array_node(size_t n, size_t size, gfp_t flags, |
654 | int node) | |
655 | { | |
656 | if (size != 0 && n > SIZE_MAX / size) | |
657 | return NULL; | |
658 | if (__builtin_constant_p(n) && __builtin_constant_p(size)) | |
659 | return kmalloc_node(n * size, flags, node); | |
660 | return __kmalloc_node(n * size, flags, node); | |
661 | } | |
662 | ||
663 | static inline void *kcalloc_node(size_t n, size_t size, gfp_t flags, int node) | |
664 | { | |
665 | return kmalloc_array_node(n, size, flags | __GFP_ZERO, node); | |
666 | } | |
667 | ||
668 | ||
97e2bde4 | 669 | #ifdef CONFIG_NUMA |
ce71e27c | 670 | extern void *__kmalloc_node_track_caller(size_t, gfp_t, int, unsigned long); |
8b98c169 CH |
671 | #define kmalloc_node_track_caller(size, flags, node) \ |
672 | __kmalloc_node_track_caller(size, flags, node, \ | |
ce71e27c | 673 | _RET_IP_) |
2e892f43 | 674 | |
8b98c169 | 675 | #else /* CONFIG_NUMA */ |
8b98c169 CH |
676 | |
677 | #define kmalloc_node_track_caller(size, flags, node) \ | |
678 | kmalloc_track_caller(size, flags) | |
97e2bde4 | 679 | |
dfcd3610 | 680 | #endif /* CONFIG_NUMA */ |
10cef602 | 681 | |
81cda662 CL |
682 | /* |
683 | * Shortcuts | |
684 | */ | |
685 | static inline void *kmem_cache_zalloc(struct kmem_cache *k, gfp_t flags) | |
686 | { | |
687 | return kmem_cache_alloc(k, flags | __GFP_ZERO); | |
688 | } | |
689 | ||
690 | /** | |
691 | * kzalloc - allocate memory. The memory is set to zero. | |
692 | * @size: how many bytes of memory are required. | |
693 | * @flags: the type of memory to allocate (see kmalloc). | |
694 | */ | |
695 | static inline void *kzalloc(size_t size, gfp_t flags) | |
696 | { | |
697 | return kmalloc(size, flags | __GFP_ZERO); | |
698 | } | |
699 | ||
979b0fea JL |
700 | /** |
701 | * kzalloc_node - allocate zeroed memory from a particular memory node. | |
702 | * @size: how many bytes of memory are required. | |
703 | * @flags: the type of memory to allocate (see kmalloc). | |
704 | * @node: memory node from which to allocate | |
705 | */ | |
706 | static inline void *kzalloc_node(size_t size, gfp_t flags, int node) | |
707 | { | |
708 | return kmalloc_node(size, flags | __GFP_ZERO, node); | |
709 | } | |
710 | ||
07f361b2 | 711 | unsigned int kmem_cache_size(struct kmem_cache *s); |
7e85ee0c PE |
712 | void __init kmem_cache_init_late(void); |
713 | ||
6731d4f1 SAS |
714 | #if defined(CONFIG_SMP) && defined(CONFIG_SLAB) |
715 | int slab_prepare_cpu(unsigned int cpu); | |
716 | int slab_dead_cpu(unsigned int cpu); | |
717 | #else | |
718 | #define slab_prepare_cpu NULL | |
719 | #define slab_dead_cpu NULL | |
720 | #endif | |
721 | ||
1da177e4 | 722 | #endif /* _LINUX_SLAB_H */ |