Linux 2.6.38-rc8
[linux-2.6-block.git] / include / linux / slub_def.h
CommitLineData
81819f0f
CL
1#ifndef _LINUX_SLUB_DEF_H
2#define _LINUX_SLUB_DEF_H
3
4/*
5 * SLUB : A Slab allocator without object queues.
6 *
cde53535 7 * (C) 2007 SGI, Christoph Lameter
81819f0f
CL
8 */
9#include <linux/types.h>
10#include <linux/gfp.h>
11#include <linux/workqueue.h>
12#include <linux/kobject.h>
13
4a92379b 14#include <linux/kmemleak.h>
039ca4e7 15
8ff12cfc
CL
16enum stat_item {
17 ALLOC_FASTPATH, /* Allocation from cpu slab */
18 ALLOC_SLOWPATH, /* Allocation by getting a new cpu slab */
19 FREE_FASTPATH, /* Free to cpu slub */
20 FREE_SLOWPATH, /* Freeing not to cpu slab */
21 FREE_FROZEN, /* Freeing to frozen slab */
22 FREE_ADD_PARTIAL, /* Freeing moves slab to partial list */
23 FREE_REMOVE_PARTIAL, /* Freeing removes last object */
24 ALLOC_FROM_PARTIAL, /* Cpu slab acquired from partial list */
25 ALLOC_SLAB, /* Cpu slab acquired from page allocator */
26 ALLOC_REFILL, /* Refill cpu slab from slab freelist */
27 FREE_SLAB, /* Slab freed to the page allocator */
28 CPUSLAB_FLUSH, /* Abandoning of the cpu slab */
29 DEACTIVATE_FULL, /* Cpu slab was full when deactivated */
30 DEACTIVATE_EMPTY, /* Cpu slab was empty when deactivated */
31 DEACTIVATE_TO_HEAD, /* Cpu slab was moved to the head of partials */
32 DEACTIVATE_TO_TAIL, /* Cpu slab was moved to the tail of partials */
33 DEACTIVATE_REMOTE_FREES,/* Slab contained remotely freed objects */
65c3376a 34 ORDER_FALLBACK, /* Number of times fallback was necessary */
8ff12cfc
CL
35 NR_SLUB_STAT_ITEMS };
36
dfb4f096 37struct kmem_cache_cpu {
da89b79e
CL
38 void **freelist; /* Pointer to first free per cpu object */
39 struct page *page; /* The slab from which we are allocating */
40 int node; /* The node of the page (or -1 for debug) */
8ff12cfc
CL
41#ifdef CONFIG_SLUB_STATS
42 unsigned stat[NR_SLUB_STAT_ITEMS];
43#endif
4c93c355 44};
dfb4f096 45
81819f0f
CL
46struct kmem_cache_node {
47 spinlock_t list_lock; /* Protect partial list and nr_partial */
48 unsigned long nr_partial;
81819f0f 49 struct list_head partial;
0c710013 50#ifdef CONFIG_SLUB_DEBUG
0f389ec6 51 atomic_long_t nr_slabs;
205ab99d 52 atomic_long_t total_objects;
643b1138 53 struct list_head full;
0c710013 54#endif
81819f0f
CL
55};
56
834f3d11
CL
57/*
58 * Word size structure that can be atomically updated or read and that
59 * contains both the order and the number of objects that a slab of the
60 * given order would contain.
61 */
62struct kmem_cache_order_objects {
63 unsigned long x;
64};
65
81819f0f
CL
66/*
67 * Slab cache management.
68 */
69struct kmem_cache {
1b5ad248 70 struct kmem_cache_cpu __percpu *cpu_slab;
81819f0f
CL
71 /* Used for retriving partial slabs etc */
72 unsigned long flags;
73 int size; /* The size of an object including meta data */
74 int objsize; /* The size of an object without meta data */
75 int offset; /* Free pointer offset. */
834f3d11 76 struct kmem_cache_order_objects oo;
81819f0f 77
81819f0f 78 /* Allocation and freeing of slabs */
205ab99d 79 struct kmem_cache_order_objects max;
65c3376a 80 struct kmem_cache_order_objects min;
b7a49f0d 81 gfp_t allocflags; /* gfp flags to use on each alloc */
81819f0f 82 int refcount; /* Refcount for slab cache destroy */
51cc5068 83 void (*ctor)(void *);
81819f0f
CL
84 int inuse; /* Offset to metadata */
85 int align; /* Alignment */
3b89d7d8 86 unsigned long min_partial;
81819f0f
CL
87 const char *name; /* Name (only for display!) */
88 struct list_head list; /* List of slab caches */
ab4d5ed5 89#ifdef CONFIG_SYSFS
81819f0f 90 struct kobject kobj; /* For sysfs */
0c710013 91#endif
81819f0f
CL
92
93#ifdef CONFIG_NUMA
9824601e
CL
94 /*
95 * Defragmentation by allocating from a remote node.
96 */
97 int remote_node_defrag_ratio;
81819f0f 98#endif
7340cc84 99 struct kmem_cache_node *node[MAX_NUMNODES];
81819f0f
CL
100};
101
102/*
103 * Kmalloc subsystem.
104 */
a6eb9fe1
FT
105#if defined(ARCH_DMA_MINALIGN) && ARCH_DMA_MINALIGN > 8
106#define KMALLOC_MIN_SIZE ARCH_DMA_MINALIGN
4b356be0
CL
107#else
108#define KMALLOC_MIN_SIZE 8
109#endif
110
111#define KMALLOC_SHIFT_LOW ilog2(KMALLOC_MIN_SIZE)
81819f0f 112
a6eb9fe1
FT
113#ifdef ARCH_DMA_MINALIGN
114#define ARCH_KMALLOC_MINALIGN ARCH_DMA_MINALIGN
115#else
4581ced3
DW
116#define ARCH_KMALLOC_MINALIGN __alignof__(unsigned long long)
117#endif
118
119#ifndef ARCH_SLAB_MINALIGN
120#define ARCH_SLAB_MINALIGN __alignof__(unsigned long long)
121#endif
122
ffadd4d0
CL
123/*
124 * Maximum kmalloc object size handled by SLUB. Larger object allocations
125 * are passed through to the page allocator. The page allocator "fastpath"
126 * is relatively slow so we need this value sufficiently high so that
127 * performance critical objects are allocated through the SLUB fastpath.
128 *
129 * This should be dropped to PAGE_SIZE / 2 once the page allocator
130 * "fastpath" becomes competitive with the slab allocator fastpaths.
131 */
51735a7c 132#define SLUB_MAX_SIZE (2 * PAGE_SIZE)
ffadd4d0 133
51735a7c 134#define SLUB_PAGE_SHIFT (PAGE_SHIFT + 2)
ffadd4d0 135
756dee75
CL
136#ifdef CONFIG_ZONE_DMA
137#define SLUB_DMA __GFP_DMA
756dee75
CL
138#else
139/* Disable DMA functionality */
140#define SLUB_DMA (__force gfp_t)0
756dee75
CL
141#endif
142
81819f0f
CL
143/*
144 * We keep the general caches in an array of slab caches that are used for
145 * 2^x bytes of allocations.
146 */
51df1142 147extern struct kmem_cache *kmalloc_caches[SLUB_PAGE_SHIFT];
81819f0f
CL
148
149/*
150 * Sorry that the following has to be that ugly but some versions of GCC
151 * have trouble with constant propagation and loops.
152 */
aa137f9d 153static __always_inline int kmalloc_index(size_t size)
81819f0f 154{
272c1d21
CL
155 if (!size)
156 return 0;
614410d5 157
4b356be0
CL
158 if (size <= KMALLOC_MIN_SIZE)
159 return KMALLOC_SHIFT_LOW;
160
acdfcd04 161 if (KMALLOC_MIN_SIZE <= 32 && size > 64 && size <= 96)
81819f0f 162 return 1;
acdfcd04 163 if (KMALLOC_MIN_SIZE <= 64 && size > 128 && size <= 192)
81819f0f
CL
164 return 2;
165 if (size <= 8) return 3;
166 if (size <= 16) return 4;
167 if (size <= 32) return 5;
168 if (size <= 64) return 6;
169 if (size <= 128) return 7;
170 if (size <= 256) return 8;
171 if (size <= 512) return 9;
172 if (size <= 1024) return 10;
173 if (size <= 2 * 1024) return 11;
6446faa2 174 if (size <= 4 * 1024) return 12;
aadb4bc4
CL
175/*
176 * The following is only needed to support architectures with a larger page
177 * size than 4k.
178 */
81819f0f
CL
179 if (size <= 8 * 1024) return 13;
180 if (size <= 16 * 1024) return 14;
181 if (size <= 32 * 1024) return 15;
182 if (size <= 64 * 1024) return 16;
183 if (size <= 128 * 1024) return 17;
184 if (size <= 256 * 1024) return 18;
aadb4bc4 185 if (size <= 512 * 1024) return 19;
81819f0f 186 if (size <= 1024 * 1024) return 20;
81819f0f 187 if (size <= 2 * 1024 * 1024) return 21;
81819f0f
CL
188 return -1;
189
190/*
191 * What we really wanted to do and cannot do because of compiler issues is:
192 * int i;
193 * for (i = KMALLOC_SHIFT_LOW; i <= KMALLOC_SHIFT_HIGH; i++)
194 * if (size <= (1 << i))
195 * return i;
196 */
197}
198
199/*
200 * Find the slab cache for a given combination of allocation flags and size.
201 *
202 * This ought to end up with a global pointer to the right cache
203 * in kmalloc_caches.
204 */
aa137f9d 205static __always_inline struct kmem_cache *kmalloc_slab(size_t size)
81819f0f
CL
206{
207 int index = kmalloc_index(size);
208
209 if (index == 0)
210 return NULL;
211
51df1142 212 return kmalloc_caches[index];
81819f0f
CL
213}
214
6193a2ff
PM
215void *kmem_cache_alloc(struct kmem_cache *, gfp_t);
216void *__kmalloc(size_t size, gfp_t flags);
217
4a92379b
RK
218static __always_inline void *
219kmalloc_order(size_t size, gfp_t flags, unsigned int order)
220{
221 void *ret = (void *) __get_free_pages(flags | __GFP_COMP, order);
222 kmemleak_alloc(ret, size, 1, flags);
223 return ret;
224}
225
0f24f128 226#ifdef CONFIG_TRACING
4a92379b
RK
227extern void *
228kmem_cache_alloc_trace(struct kmem_cache *s, gfp_t gfpflags, size_t size);
229extern void *kmalloc_order_trace(size_t size, gfp_t flags, unsigned int order);
5b882be4
EGM
230#else
231static __always_inline void *
4a92379b 232kmem_cache_alloc_trace(struct kmem_cache *s, gfp_t gfpflags, size_t size)
5b882be4
EGM
233{
234 return kmem_cache_alloc(s, gfpflags);
235}
4a92379b
RK
236
237static __always_inline void *
238kmalloc_order_trace(size_t size, gfp_t flags, unsigned int order)
239{
240 return kmalloc_order(size, flags, order);
241}
5b882be4
EGM
242#endif
243
eada35ef
PE
244static __always_inline void *kmalloc_large(size_t size, gfp_t flags)
245{
5b882be4 246 unsigned int order = get_order(size);
4a92379b 247 return kmalloc_order_trace(size, flags, order);
eada35ef
PE
248}
249
aa137f9d 250static __always_inline void *kmalloc(size_t size, gfp_t flags)
81819f0f 251{
aadb4bc4 252 if (__builtin_constant_p(size)) {
ffadd4d0 253 if (size > SLUB_MAX_SIZE)
eada35ef 254 return kmalloc_large(size, flags);
81819f0f 255
aadb4bc4
CL
256 if (!(flags & SLUB_DMA)) {
257 struct kmem_cache *s = kmalloc_slab(size);
258
259 if (!s)
260 return ZERO_SIZE_PTR;
81819f0f 261
4a92379b 262 return kmem_cache_alloc_trace(s, flags, size);
aadb4bc4
CL
263 }
264 }
265 return __kmalloc(size, flags);
81819f0f
CL
266}
267
81819f0f 268#ifdef CONFIG_NUMA
6193a2ff
PM
269void *__kmalloc_node(size_t size, gfp_t flags, int node);
270void *kmem_cache_alloc_node(struct kmem_cache *, gfp_t flags, int node);
81819f0f 271
0f24f128 272#ifdef CONFIG_TRACING
4a92379b 273extern void *kmem_cache_alloc_node_trace(struct kmem_cache *s,
5b882be4 274 gfp_t gfpflags,
4a92379b 275 int node, size_t size);
5b882be4
EGM
276#else
277static __always_inline void *
4a92379b 278kmem_cache_alloc_node_trace(struct kmem_cache *s,
5b882be4 279 gfp_t gfpflags,
4a92379b 280 int node, size_t size)
5b882be4
EGM
281{
282 return kmem_cache_alloc_node(s, gfpflags, node);
283}
284#endif
285
aa137f9d 286static __always_inline void *kmalloc_node(size_t size, gfp_t flags, int node)
81819f0f 287{
aadb4bc4 288 if (__builtin_constant_p(size) &&
ffadd4d0 289 size <= SLUB_MAX_SIZE && !(flags & SLUB_DMA)) {
aadb4bc4 290 struct kmem_cache *s = kmalloc_slab(size);
81819f0f
CL
291
292 if (!s)
272c1d21 293 return ZERO_SIZE_PTR;
81819f0f 294
4a92379b 295 return kmem_cache_alloc_node_trace(s, flags, node, size);
aadb4bc4
CL
296 }
297 return __kmalloc_node(size, flags, node);
81819f0f
CL
298}
299#endif
300
301#endif /* _LINUX_SLUB_DEF_H */