]>
Commit | Line | Data |
---|---|---|
1da177e4 | 1 | /* |
2e892f43 CL |
2 | * Written by Mark Hemment, 1996 (markhe@nextd.demon.co.uk). |
3 | * | |
cde53535 | 4 | * (C) SGI 2006, Christoph Lameter |
2e892f43 CL |
5 | * Cleaned up and restructured to ease the addition of alternative |
6 | * implementations of SLAB allocators. | |
f1b6eb6e CL |
7 | * (C) Linux Foundation 2008-2013 |
8 | * Unified interface for all slab allocators | |
1da177e4 LT |
9 | */ |
10 | ||
11 | #ifndef _LINUX_SLAB_H | |
12 | #define _LINUX_SLAB_H | |
13 | ||
1b1cec4b | 14 | #include <linux/gfp.h> |
1b1cec4b | 15 | #include <linux/types.h> |
1f458cbf GC |
16 | #include <linux/workqueue.h> |
17 | ||
1da177e4 | 18 | |
2e892f43 CL |
19 | /* |
20 | * Flags to pass to kmem_cache_create(). | |
124dee09 | 21 | * The ones marked DEBUG are only valid if CONFIG_DEBUG_SLAB is set. |
1da177e4 | 22 | */ |
becfda68 | 23 | #define SLAB_CONSISTENCY_CHECKS 0x00000100UL /* DEBUG: Perform (expensive) checks on alloc/free */ |
55935a34 CL |
24 | #define SLAB_RED_ZONE 0x00000400UL /* DEBUG: Red zone objs in a cache */ |
25 | #define SLAB_POISON 0x00000800UL /* DEBUG: Poison objects */ | |
26 | #define SLAB_HWCACHE_ALIGN 0x00002000UL /* Align objs on cache lines */ | |
2e892f43 | 27 | #define SLAB_CACHE_DMA 0x00004000UL /* Use GFP_DMA memory */ |
2e892f43 | 28 | #define SLAB_STORE_USER 0x00010000UL /* DEBUG: Store the last owner for bug hunting */ |
2e892f43 | 29 | #define SLAB_PANIC 0x00040000UL /* Panic if kmem_cache_create() fails */ |
d7de4c1d PZ |
30 | /* |
31 | * SLAB_DESTROY_BY_RCU - **WARNING** READ THIS! | |
32 | * | |
33 | * This delays freeing the SLAB page by a grace period, it does _NOT_ | |
34 | * delay object freeing. This means that if you do kmem_cache_free() | |
35 | * that memory location is free to be reused at any time. Thus it may | |
36 | * be possible to see another object there in the same RCU grace period. | |
37 | * | |
38 | * This feature only ensures the memory location backing the object | |
39 | * stays valid, the trick to using this is relying on an independent | |
40 | * object validation pass. Something like: | |
41 | * | |
42 | * rcu_read_lock() | |
43 | * again: | |
44 | * obj = lockless_lookup(key); | |
45 | * if (obj) { | |
46 | * if (!try_get_ref(obj)) // might fail for free objects | |
47 | * goto again; | |
48 | * | |
49 | * if (obj->key != key) { // not the object we expected | |
50 | * put_ref(obj); | |
51 | * goto again; | |
52 | * } | |
53 | * } | |
54 | * rcu_read_unlock(); | |
55 | * | |
68126702 JK |
56 | * This is useful if we need to approach a kernel structure obliquely, |
57 | * from its address obtained without the usual locking. We can lock | |
58 | * the structure to stabilize it and check it's still at the given address, | |
59 | * only if we can be sure that the memory has not been meanwhile reused | |
60 | * for some other kind of object (which our subsystem's lock might corrupt). | |
61 | * | |
62 | * rcu_read_lock before reading the address, then rcu_read_unlock after | |
63 | * taking the spinlock within the structure expected at that address. | |
d7de4c1d | 64 | */ |
2e892f43 | 65 | #define SLAB_DESTROY_BY_RCU 0x00080000UL /* Defer freeing slabs to RCU */ |
101a5001 | 66 | #define SLAB_MEM_SPREAD 0x00100000UL /* Spread some memory over cpuset */ |
81819f0f | 67 | #define SLAB_TRACE 0x00200000UL /* Trace allocations and frees */ |
1da177e4 | 68 | |
30327acf TG |
69 | /* Flag to prevent checks on free */ |
70 | #ifdef CONFIG_DEBUG_OBJECTS | |
71 | # define SLAB_DEBUG_OBJECTS 0x00400000UL | |
72 | #else | |
73 | # define SLAB_DEBUG_OBJECTS 0x00000000UL | |
74 | #endif | |
75 | ||
d5cff635 CM |
76 | #define SLAB_NOLEAKTRACE 0x00800000UL /* Avoid kmemleak tracing */ |
77 | ||
2dff4405 VN |
78 | /* Don't track use of uninitialized memory */ |
79 | #ifdef CONFIG_KMEMCHECK | |
80 | # define SLAB_NOTRACK 0x01000000UL | |
81 | #else | |
82 | # define SLAB_NOTRACK 0x00000000UL | |
83 | #endif | |
4c13dd3b DM |
84 | #ifdef CONFIG_FAILSLAB |
85 | # define SLAB_FAILSLAB 0x02000000UL /* Fault injection mark */ | |
86 | #else | |
87 | # define SLAB_FAILSLAB 0x00000000UL | |
88 | #endif | |
127424c8 | 89 | #if defined(CONFIG_MEMCG) && !defined(CONFIG_SLOB) |
230e9fc2 VD |
90 | # define SLAB_ACCOUNT 0x04000000UL /* Account to memcg */ |
91 | #else | |
92 | # define SLAB_ACCOUNT 0x00000000UL | |
93 | #endif | |
2dff4405 | 94 | |
7ed2f9e6 AP |
95 | #ifdef CONFIG_KASAN |
96 | #define SLAB_KASAN 0x08000000UL | |
97 | #else | |
98 | #define SLAB_KASAN 0x00000000UL | |
99 | #endif | |
100 | ||
e12ba74d MG |
101 | /* The following flags affect the page allocator grouping pages by mobility */ |
102 | #define SLAB_RECLAIM_ACCOUNT 0x00020000UL /* Objects are reclaimable */ | |
103 | #define SLAB_TEMPORARY SLAB_RECLAIM_ACCOUNT /* Objects are short-lived */ | |
6cb8f913 CL |
104 | /* |
105 | * ZERO_SIZE_PTR will be returned for zero sized kmalloc requests. | |
106 | * | |
107 | * Dereferencing ZERO_SIZE_PTR will lead to a distinct access fault. | |
108 | * | |
109 | * ZERO_SIZE_PTR can be passed to kfree though in the same way that NULL can. | |
110 | * Both make kfree a no-op. | |
111 | */ | |
112 | #define ZERO_SIZE_PTR ((void *)16) | |
113 | ||
1d4ec7b1 | 114 | #define ZERO_OR_NULL_PTR(x) ((unsigned long)(x) <= \ |
6cb8f913 CL |
115 | (unsigned long)ZERO_SIZE_PTR) |
116 | ||
f1b6eb6e | 117 | #include <linux/kmemleak.h> |
0316bec2 | 118 | #include <linux/kasan.h> |
3b0efdfa | 119 | |
2633d7a0 | 120 | struct mem_cgroup; |
2e892f43 CL |
121 | /* |
122 | * struct kmem_cache related prototypes | |
123 | */ | |
124 | void __init kmem_cache_init(void); | |
fda90124 | 125 | bool slab_is_available(void); |
1da177e4 | 126 | |
2e892f43 | 127 | struct kmem_cache *kmem_cache_create(const char *, size_t, size_t, |
ebe29738 | 128 | unsigned long, |
51cc5068 | 129 | void (*)(void *)); |
2e892f43 CL |
130 | void kmem_cache_destroy(struct kmem_cache *); |
131 | int kmem_cache_shrink(struct kmem_cache *); | |
2a4db7eb VD |
132 | |
133 | void memcg_create_kmem_cache(struct mem_cgroup *, struct kmem_cache *); | |
134 | void memcg_deactivate_kmem_caches(struct mem_cgroup *); | |
135 | void memcg_destroy_kmem_caches(struct mem_cgroup *); | |
2e892f43 | 136 | |
0a31bd5f CL |
137 | /* |
138 | * Please use this macro to create slab caches. Simply specify the | |
139 | * name of the structure and maybe some flags that are listed above. | |
140 | * | |
141 | * The alignment of the struct determines object alignment. If you | |
142 | * f.e. add ____cacheline_aligned_in_smp to the struct declaration | |
143 | * then the objects will be properly aligned in SMP configurations. | |
144 | */ | |
145 | #define KMEM_CACHE(__struct, __flags) kmem_cache_create(#__struct,\ | |
146 | sizeof(struct __struct), __alignof__(struct __struct),\ | |
20c2df83 | 147 | (__flags), NULL) |
0a31bd5f | 148 | |
34504667 CL |
149 | /* |
150 | * Common kmalloc functions provided by all allocators | |
151 | */ | |
152 | void * __must_check __krealloc(const void *, size_t, gfp_t); | |
153 | void * __must_check krealloc(const void *, size_t, gfp_t); | |
154 | void kfree(const void *); | |
155 | void kzfree(const void *); | |
156 | size_t ksize(const void *); | |
157 | ||
c601fd69 CL |
158 | /* |
159 | * Some archs want to perform DMA into kmalloc caches and need a guaranteed | |
160 | * alignment larger than the alignment of a 64-bit integer. | |
161 | * Setting ARCH_KMALLOC_MINALIGN in arch headers allows that. | |
162 | */ | |
163 | #if defined(ARCH_DMA_MINALIGN) && ARCH_DMA_MINALIGN > 8 | |
164 | #define ARCH_KMALLOC_MINALIGN ARCH_DMA_MINALIGN | |
165 | #define KMALLOC_MIN_SIZE ARCH_DMA_MINALIGN | |
166 | #define KMALLOC_SHIFT_LOW ilog2(ARCH_DMA_MINALIGN) | |
167 | #else | |
168 | #define ARCH_KMALLOC_MINALIGN __alignof__(unsigned long long) | |
169 | #endif | |
170 | ||
94a58c36 RV |
171 | /* |
172 | * Setting ARCH_SLAB_MINALIGN in arch headers allows a different alignment. | |
173 | * Intended for arches that get misalignment faults even for 64 bit integer | |
174 | * aligned buffers. | |
175 | */ | |
176 | #ifndef ARCH_SLAB_MINALIGN | |
177 | #define ARCH_SLAB_MINALIGN __alignof__(unsigned long long) | |
178 | #endif | |
179 | ||
180 | /* | |
181 | * kmalloc and friends return ARCH_KMALLOC_MINALIGN aligned | |
182 | * pointers. kmem_cache_alloc and friends return ARCH_SLAB_MINALIGN | |
183 | * aligned pointers. | |
184 | */ | |
185 | #define __assume_kmalloc_alignment __assume_aligned(ARCH_KMALLOC_MINALIGN) | |
186 | #define __assume_slab_alignment __assume_aligned(ARCH_SLAB_MINALIGN) | |
187 | #define __assume_page_alignment __assume_aligned(PAGE_SIZE) | |
188 | ||
0aa817f0 | 189 | /* |
95a05b42 CL |
190 | * Kmalloc array related definitions |
191 | */ | |
192 | ||
193 | #ifdef CONFIG_SLAB | |
194 | /* | |
195 | * The largest kmalloc size supported by the SLAB allocators is | |
0aa817f0 CL |
196 | * 32 megabyte (2^25) or the maximum allocatable page order if that is |
197 | * less than 32 MB. | |
198 | * | |
199 | * WARNING: Its not easy to increase this value since the allocators have | |
200 | * to do various tricks to work around compiler limitations in order to | |
201 | * ensure proper constant folding. | |
202 | */ | |
debee076 CL |
203 | #define KMALLOC_SHIFT_HIGH ((MAX_ORDER + PAGE_SHIFT - 1) <= 25 ? \ |
204 | (MAX_ORDER + PAGE_SHIFT - 1) : 25) | |
95a05b42 | 205 | #define KMALLOC_SHIFT_MAX KMALLOC_SHIFT_HIGH |
c601fd69 | 206 | #ifndef KMALLOC_SHIFT_LOW |
95a05b42 | 207 | #define KMALLOC_SHIFT_LOW 5 |
c601fd69 | 208 | #endif |
069e2b35 CL |
209 | #endif |
210 | ||
211 | #ifdef CONFIG_SLUB | |
95a05b42 | 212 | /* |
433a91ff DH |
213 | * SLUB directly allocates requests fitting in to an order-1 page |
214 | * (PAGE_SIZE*2). Larger requests are passed to the page allocator. | |
95a05b42 CL |
215 | */ |
216 | #define KMALLOC_SHIFT_HIGH (PAGE_SHIFT + 1) | |
217 | #define KMALLOC_SHIFT_MAX (MAX_ORDER + PAGE_SHIFT) | |
c601fd69 | 218 | #ifndef KMALLOC_SHIFT_LOW |
95a05b42 CL |
219 | #define KMALLOC_SHIFT_LOW 3 |
220 | #endif | |
c601fd69 | 221 | #endif |
0aa817f0 | 222 | |
069e2b35 CL |
223 | #ifdef CONFIG_SLOB |
224 | /* | |
433a91ff | 225 | * SLOB passes all requests larger than one page to the page allocator. |
069e2b35 CL |
226 | * No kmalloc array is necessary since objects of different sizes can |
227 | * be allocated from the same page. | |
228 | */ | |
069e2b35 | 229 | #define KMALLOC_SHIFT_HIGH PAGE_SHIFT |
433a91ff | 230 | #define KMALLOC_SHIFT_MAX 30 |
069e2b35 CL |
231 | #ifndef KMALLOC_SHIFT_LOW |
232 | #define KMALLOC_SHIFT_LOW 3 | |
233 | #endif | |
234 | #endif | |
235 | ||
95a05b42 CL |
236 | /* Maximum allocatable size */ |
237 | #define KMALLOC_MAX_SIZE (1UL << KMALLOC_SHIFT_MAX) | |
238 | /* Maximum size for which we actually use a slab cache */ | |
239 | #define KMALLOC_MAX_CACHE_SIZE (1UL << KMALLOC_SHIFT_HIGH) | |
240 | /* Maximum order allocatable via the slab allocagtor */ | |
241 | #define KMALLOC_MAX_ORDER (KMALLOC_SHIFT_MAX - PAGE_SHIFT) | |
0aa817f0 | 242 | |
ce6a5026 CL |
243 | /* |
244 | * Kmalloc subsystem. | |
245 | */ | |
c601fd69 | 246 | #ifndef KMALLOC_MIN_SIZE |
95a05b42 | 247 | #define KMALLOC_MIN_SIZE (1 << KMALLOC_SHIFT_LOW) |
ce6a5026 CL |
248 | #endif |
249 | ||
24f870d8 JK |
250 | /* |
251 | * This restriction comes from byte sized index implementation. | |
252 | * Page size is normally 2^12 bytes and, in this case, if we want to use | |
253 | * byte sized index which can represent 2^8 entries, the size of the object | |
254 | * should be equal or greater to 2^12 / 2^8 = 2^4 = 16. | |
255 | * If minimum size of kmalloc is less than 16, we use it as minimum object | |
256 | * size and give up to use byte sized index. | |
257 | */ | |
258 | #define SLAB_OBJ_MIN_SIZE (KMALLOC_MIN_SIZE < 16 ? \ | |
259 | (KMALLOC_MIN_SIZE) : 16) | |
260 | ||
069e2b35 | 261 | #ifndef CONFIG_SLOB |
9425c58e CL |
262 | extern struct kmem_cache *kmalloc_caches[KMALLOC_SHIFT_HIGH + 1]; |
263 | #ifdef CONFIG_ZONE_DMA | |
264 | extern struct kmem_cache *kmalloc_dma_caches[KMALLOC_SHIFT_HIGH + 1]; | |
265 | #endif | |
266 | ||
ce6a5026 CL |
267 | /* |
268 | * Figure out which kmalloc slab an allocation of a certain size | |
269 | * belongs to. | |
270 | * 0 = zero alloc | |
271 | * 1 = 65 .. 96 bytes | |
1ed58b60 RV |
272 | * 2 = 129 .. 192 bytes |
273 | * n = 2^(n-1)+1 .. 2^n | |
ce6a5026 CL |
274 | */ |
275 | static __always_inline int kmalloc_index(size_t size) | |
276 | { | |
277 | if (!size) | |
278 | return 0; | |
279 | ||
280 | if (size <= KMALLOC_MIN_SIZE) | |
281 | return KMALLOC_SHIFT_LOW; | |
282 | ||
283 | if (KMALLOC_MIN_SIZE <= 32 && size > 64 && size <= 96) | |
284 | return 1; | |
285 | if (KMALLOC_MIN_SIZE <= 64 && size > 128 && size <= 192) | |
286 | return 2; | |
287 | if (size <= 8) return 3; | |
288 | if (size <= 16) return 4; | |
289 | if (size <= 32) return 5; | |
290 | if (size <= 64) return 6; | |
291 | if (size <= 128) return 7; | |
292 | if (size <= 256) return 8; | |
293 | if (size <= 512) return 9; | |
294 | if (size <= 1024) return 10; | |
295 | if (size <= 2 * 1024) return 11; | |
296 | if (size <= 4 * 1024) return 12; | |
297 | if (size <= 8 * 1024) return 13; | |
298 | if (size <= 16 * 1024) return 14; | |
299 | if (size <= 32 * 1024) return 15; | |
300 | if (size <= 64 * 1024) return 16; | |
301 | if (size <= 128 * 1024) return 17; | |
302 | if (size <= 256 * 1024) return 18; | |
303 | if (size <= 512 * 1024) return 19; | |
304 | if (size <= 1024 * 1024) return 20; | |
305 | if (size <= 2 * 1024 * 1024) return 21; | |
306 | if (size <= 4 * 1024 * 1024) return 22; | |
307 | if (size <= 8 * 1024 * 1024) return 23; | |
308 | if (size <= 16 * 1024 * 1024) return 24; | |
309 | if (size <= 32 * 1024 * 1024) return 25; | |
310 | if (size <= 64 * 1024 * 1024) return 26; | |
311 | BUG(); | |
312 | ||
313 | /* Will never be reached. Needed because the compiler may complain */ | |
314 | return -1; | |
315 | } | |
069e2b35 | 316 | #endif /* !CONFIG_SLOB */ |
ce6a5026 | 317 | |
94a58c36 RV |
318 | void *__kmalloc(size_t size, gfp_t flags) __assume_kmalloc_alignment; |
319 | void *kmem_cache_alloc(struct kmem_cache *, gfp_t flags) __assume_slab_alignment; | |
2a4db7eb | 320 | void kmem_cache_free(struct kmem_cache *, void *); |
f1b6eb6e | 321 | |
484748f0 | 322 | /* |
9f706d68 | 323 | * Bulk allocation and freeing operations. These are accelerated in an |
484748f0 CL |
324 | * allocator specific way to avoid taking locks repeatedly or building |
325 | * metadata structures unnecessarily. | |
326 | * | |
327 | * Note that interrupts must be enabled when calling these functions. | |
328 | */ | |
329 | void kmem_cache_free_bulk(struct kmem_cache *, size_t, void **); | |
865762a8 | 330 | int kmem_cache_alloc_bulk(struct kmem_cache *, gfp_t, size_t, void **); |
484748f0 | 331 | |
ca257195 JDB |
332 | /* |
333 | * Caller must not use kfree_bulk() on memory not originally allocated | |
334 | * by kmalloc(), because the SLOB allocator cannot handle this. | |
335 | */ | |
336 | static __always_inline void kfree_bulk(size_t size, void **p) | |
337 | { | |
338 | kmem_cache_free_bulk(NULL, size, p); | |
339 | } | |
340 | ||
f1b6eb6e | 341 | #ifdef CONFIG_NUMA |
94a58c36 RV |
342 | void *__kmalloc_node(size_t size, gfp_t flags, int node) __assume_kmalloc_alignment; |
343 | void *kmem_cache_alloc_node(struct kmem_cache *, gfp_t flags, int node) __assume_slab_alignment; | |
f1b6eb6e CL |
344 | #else |
345 | static __always_inline void *__kmalloc_node(size_t size, gfp_t flags, int node) | |
346 | { | |
347 | return __kmalloc(size, flags); | |
348 | } | |
349 | ||
350 | static __always_inline void *kmem_cache_alloc_node(struct kmem_cache *s, gfp_t flags, int node) | |
351 | { | |
352 | return kmem_cache_alloc(s, flags); | |
353 | } | |
354 | #endif | |
355 | ||
356 | #ifdef CONFIG_TRACING | |
94a58c36 | 357 | extern void *kmem_cache_alloc_trace(struct kmem_cache *, gfp_t, size_t) __assume_slab_alignment; |
f1b6eb6e CL |
358 | |
359 | #ifdef CONFIG_NUMA | |
360 | extern void *kmem_cache_alloc_node_trace(struct kmem_cache *s, | |
361 | gfp_t gfpflags, | |
94a58c36 | 362 | int node, size_t size) __assume_slab_alignment; |
f1b6eb6e CL |
363 | #else |
364 | static __always_inline void * | |
365 | kmem_cache_alloc_node_trace(struct kmem_cache *s, | |
366 | gfp_t gfpflags, | |
367 | int node, size_t size) | |
368 | { | |
369 | return kmem_cache_alloc_trace(s, gfpflags, size); | |
370 | } | |
371 | #endif /* CONFIG_NUMA */ | |
372 | ||
373 | #else /* CONFIG_TRACING */ | |
374 | static __always_inline void *kmem_cache_alloc_trace(struct kmem_cache *s, | |
375 | gfp_t flags, size_t size) | |
376 | { | |
0316bec2 AR |
377 | void *ret = kmem_cache_alloc(s, flags); |
378 | ||
379 | kasan_kmalloc(s, ret, size); | |
380 | return ret; | |
f1b6eb6e CL |
381 | } |
382 | ||
383 | static __always_inline void * | |
384 | kmem_cache_alloc_node_trace(struct kmem_cache *s, | |
385 | gfp_t gfpflags, | |
386 | int node, size_t size) | |
387 | { | |
0316bec2 AR |
388 | void *ret = kmem_cache_alloc_node(s, gfpflags, node); |
389 | ||
390 | kasan_kmalloc(s, ret, size); | |
391 | return ret; | |
f1b6eb6e CL |
392 | } |
393 | #endif /* CONFIG_TRACING */ | |
394 | ||
94a58c36 | 395 | extern void *kmalloc_order(size_t size, gfp_t flags, unsigned int order) __assume_page_alignment; |
f1b6eb6e CL |
396 | |
397 | #ifdef CONFIG_TRACING | |
94a58c36 | 398 | extern void *kmalloc_order_trace(size_t size, gfp_t flags, unsigned int order) __assume_page_alignment; |
f1b6eb6e CL |
399 | #else |
400 | static __always_inline void * | |
401 | kmalloc_order_trace(size_t size, gfp_t flags, unsigned int order) | |
402 | { | |
403 | return kmalloc_order(size, flags, order); | |
404 | } | |
ce6a5026 CL |
405 | #endif |
406 | ||
f1b6eb6e CL |
407 | static __always_inline void *kmalloc_large(size_t size, gfp_t flags) |
408 | { | |
409 | unsigned int order = get_order(size); | |
410 | return kmalloc_order_trace(size, flags, order); | |
411 | } | |
412 | ||
413 | /** | |
414 | * kmalloc - allocate memory | |
415 | * @size: how many bytes of memory are required. | |
7e3528c3 | 416 | * @flags: the type of memory to allocate. |
f1b6eb6e CL |
417 | * |
418 | * kmalloc is the normal method of allocating memory | |
419 | * for objects smaller than page size in the kernel. | |
7e3528c3 RD |
420 | * |
421 | * The @flags argument may be one of: | |
422 | * | |
423 | * %GFP_USER - Allocate memory on behalf of user. May sleep. | |
424 | * | |
425 | * %GFP_KERNEL - Allocate normal kernel ram. May sleep. | |
426 | * | |
427 | * %GFP_ATOMIC - Allocation will not sleep. May use emergency pools. | |
428 | * For example, use this inside interrupt handlers. | |
429 | * | |
430 | * %GFP_HIGHUSER - Allocate pages from high memory. | |
431 | * | |
432 | * %GFP_NOIO - Do not do any I/O at all while trying to get memory. | |
433 | * | |
434 | * %GFP_NOFS - Do not make any fs calls while trying to get memory. | |
435 | * | |
436 | * %GFP_NOWAIT - Allocation will not sleep. | |
437 | * | |
e97ca8e5 | 438 | * %__GFP_THISNODE - Allocate node-local memory only. |
7e3528c3 RD |
439 | * |
440 | * %GFP_DMA - Allocation suitable for DMA. | |
441 | * Should only be used for kmalloc() caches. Otherwise, use a | |
442 | * slab created with SLAB_DMA. | |
443 | * | |
444 | * Also it is possible to set different flags by OR'ing | |
445 | * in one or more of the following additional @flags: | |
446 | * | |
447 | * %__GFP_COLD - Request cache-cold pages instead of | |
448 | * trying to return cache-warm pages. | |
449 | * | |
450 | * %__GFP_HIGH - This allocation has high priority and may use emergency pools. | |
451 | * | |
452 | * %__GFP_NOFAIL - Indicate that this allocation is in no way allowed to fail | |
453 | * (think twice before using). | |
454 | * | |
455 | * %__GFP_NORETRY - If memory is not immediately available, | |
456 | * then give up at once. | |
457 | * | |
458 | * %__GFP_NOWARN - If allocation fails, don't issue any warnings. | |
459 | * | |
460 | * %__GFP_REPEAT - If allocation fails initially, try once more before failing. | |
461 | * | |
462 | * There are other flags available as well, but these are not intended | |
463 | * for general use, and so are not documented here. For a full list of | |
464 | * potential flags, always refer to linux/gfp.h. | |
f1b6eb6e CL |
465 | */ |
466 | static __always_inline void *kmalloc(size_t size, gfp_t flags) | |
467 | { | |
468 | if (__builtin_constant_p(size)) { | |
469 | if (size > KMALLOC_MAX_CACHE_SIZE) | |
470 | return kmalloc_large(size, flags); | |
471 | #ifndef CONFIG_SLOB | |
472 | if (!(flags & GFP_DMA)) { | |
473 | int index = kmalloc_index(size); | |
474 | ||
475 | if (!index) | |
476 | return ZERO_SIZE_PTR; | |
477 | ||
478 | return kmem_cache_alloc_trace(kmalloc_caches[index], | |
479 | flags, size); | |
480 | } | |
481 | #endif | |
482 | } | |
483 | return __kmalloc(size, flags); | |
484 | } | |
485 | ||
ce6a5026 CL |
486 | /* |
487 | * Determine size used for the nth kmalloc cache. | |
488 | * return size or 0 if a kmalloc cache for that | |
489 | * size does not exist | |
490 | */ | |
491 | static __always_inline int kmalloc_size(int n) | |
492 | { | |
069e2b35 | 493 | #ifndef CONFIG_SLOB |
ce6a5026 CL |
494 | if (n > 2) |
495 | return 1 << n; | |
496 | ||
497 | if (n == 1 && KMALLOC_MIN_SIZE <= 32) | |
498 | return 96; | |
499 | ||
500 | if (n == 2 && KMALLOC_MIN_SIZE <= 64) | |
501 | return 192; | |
069e2b35 | 502 | #endif |
ce6a5026 CL |
503 | return 0; |
504 | } | |
ce6a5026 | 505 | |
f1b6eb6e CL |
506 | static __always_inline void *kmalloc_node(size_t size, gfp_t flags, int node) |
507 | { | |
508 | #ifndef CONFIG_SLOB | |
509 | if (__builtin_constant_p(size) && | |
23774a2f | 510 | size <= KMALLOC_MAX_CACHE_SIZE && !(flags & GFP_DMA)) { |
f1b6eb6e CL |
511 | int i = kmalloc_index(size); |
512 | ||
513 | if (!i) | |
514 | return ZERO_SIZE_PTR; | |
515 | ||
516 | return kmem_cache_alloc_node_trace(kmalloc_caches[i], | |
517 | flags, node, size); | |
518 | } | |
519 | #endif | |
520 | return __kmalloc_node(size, flags, node); | |
521 | } | |
522 | ||
f7ce3190 VD |
523 | struct memcg_cache_array { |
524 | struct rcu_head rcu; | |
525 | struct kmem_cache *entries[0]; | |
526 | }; | |
527 | ||
ba6c496e GC |
528 | /* |
529 | * This is the main placeholder for memcg-related information in kmem caches. | |
ba6c496e GC |
530 | * Both the root cache and the child caches will have it. For the root cache, |
531 | * this will hold a dynamically allocated array large enough to hold | |
f8570263 VD |
532 | * information about the currently limited memcgs in the system. To allow the |
533 | * array to be accessed without taking any locks, on relocation we free the old | |
534 | * version only after a grace period. | |
ba6c496e GC |
535 | * |
536 | * Child caches will hold extra metadata needed for its operation. Fields are: | |
537 | * | |
538 | * @memcg: pointer to the memcg this cache belongs to | |
2633d7a0 | 539 | * @root_cache: pointer to the global, root cache, this cache was derived from |
426589f5 VD |
540 | * |
541 | * Both root and child caches of the same kind are linked into a list chained | |
542 | * through @list. | |
ba6c496e GC |
543 | */ |
544 | struct memcg_cache_params { | |
545 | bool is_root_cache; | |
426589f5 | 546 | struct list_head list; |
ba6c496e | 547 | union { |
f7ce3190 | 548 | struct memcg_cache_array __rcu *memcg_caches; |
2633d7a0 GC |
549 | struct { |
550 | struct mem_cgroup *memcg; | |
2633d7a0 GC |
551 | struct kmem_cache *root_cache; |
552 | }; | |
ba6c496e GC |
553 | }; |
554 | }; | |
555 | ||
2633d7a0 GC |
556 | int memcg_update_all_caches(int num_memcgs); |
557 | ||
e7efa615 MO |
558 | /** |
559 | * kmalloc_array - allocate memory for an array. | |
560 | * @n: number of elements. | |
561 | * @size: element size. | |
562 | * @flags: the type of memory to allocate (see kmalloc). | |
800590f5 | 563 | */ |
a8203725 | 564 | static inline void *kmalloc_array(size_t n, size_t size, gfp_t flags) |
1da177e4 | 565 | { |
a3860c1c | 566 | if (size != 0 && n > SIZE_MAX / size) |
6193a2ff | 567 | return NULL; |
a8203725 XW |
568 | return __kmalloc(n * size, flags); |
569 | } | |
570 | ||
571 | /** | |
572 | * kcalloc - allocate memory for an array. The memory is set to zero. | |
573 | * @n: number of elements. | |
574 | * @size: element size. | |
575 | * @flags: the type of memory to allocate (see kmalloc). | |
576 | */ | |
577 | static inline void *kcalloc(size_t n, size_t size, gfp_t flags) | |
578 | { | |
579 | return kmalloc_array(n, size, flags | __GFP_ZERO); | |
1da177e4 LT |
580 | } |
581 | ||
1d2c8eea CH |
582 | /* |
583 | * kmalloc_track_caller is a special version of kmalloc that records the | |
584 | * calling function of the routine calling it for slab leak tracking instead | |
585 | * of just the calling function (confusing, eh?). | |
586 | * It's useful when the call to kmalloc comes from a widely-used standard | |
587 | * allocator where we care about the real place the memory allocation | |
588 | * request comes from. | |
589 | */ | |
ce71e27c | 590 | extern void *__kmalloc_track_caller(size_t, gfp_t, unsigned long); |
1d2c8eea | 591 | #define kmalloc_track_caller(size, flags) \ |
ce71e27c | 592 | __kmalloc_track_caller(size, flags, _RET_IP_) |
1da177e4 | 593 | |
97e2bde4 | 594 | #ifdef CONFIG_NUMA |
ce71e27c | 595 | extern void *__kmalloc_node_track_caller(size_t, gfp_t, int, unsigned long); |
8b98c169 CH |
596 | #define kmalloc_node_track_caller(size, flags, node) \ |
597 | __kmalloc_node_track_caller(size, flags, node, \ | |
ce71e27c | 598 | _RET_IP_) |
2e892f43 | 599 | |
8b98c169 | 600 | #else /* CONFIG_NUMA */ |
8b98c169 CH |
601 | |
602 | #define kmalloc_node_track_caller(size, flags, node) \ | |
603 | kmalloc_track_caller(size, flags) | |
97e2bde4 | 604 | |
dfcd3610 | 605 | #endif /* CONFIG_NUMA */ |
10cef602 | 606 | |
81cda662 CL |
607 | /* |
608 | * Shortcuts | |
609 | */ | |
610 | static inline void *kmem_cache_zalloc(struct kmem_cache *k, gfp_t flags) | |
611 | { | |
612 | return kmem_cache_alloc(k, flags | __GFP_ZERO); | |
613 | } | |
614 | ||
615 | /** | |
616 | * kzalloc - allocate memory. The memory is set to zero. | |
617 | * @size: how many bytes of memory are required. | |
618 | * @flags: the type of memory to allocate (see kmalloc). | |
619 | */ | |
620 | static inline void *kzalloc(size_t size, gfp_t flags) | |
621 | { | |
622 | return kmalloc(size, flags | __GFP_ZERO); | |
623 | } | |
624 | ||
979b0fea JL |
625 | /** |
626 | * kzalloc_node - allocate zeroed memory from a particular memory node. | |
627 | * @size: how many bytes of memory are required. | |
628 | * @flags: the type of memory to allocate (see kmalloc). | |
629 | * @node: memory node from which to allocate | |
630 | */ | |
631 | static inline void *kzalloc_node(size_t size, gfp_t flags, int node) | |
632 | { | |
633 | return kmalloc_node(size, flags | __GFP_ZERO, node); | |
634 | } | |
635 | ||
07f361b2 | 636 | unsigned int kmem_cache_size(struct kmem_cache *s); |
7e85ee0c PE |
637 | void __init kmem_cache_init_late(void); |
638 | ||
1da177e4 | 639 | #endif /* _LINUX_SLAB_H */ |