mm/slob.c

   1 /*
   2  * SLOB Allocator: Simple List Of Blocks
   3  *
   4  * Matt Mackall <mpm@selenic.com> 12/30/03
   5  *
   6  * How SLOB works:
   7  *
   8  * The core of SLOB is a traditional K&R style heap allocator, with
   9  * support for returning aligned objects. The granularity of this
  10  * allocator is 8 bytes on x86, though it's perhaps possible to reduce
  11  * this to 4 if it's deemed worth the effort. The slob heap is a
  12  * singly-linked list of pages from __get_free_page, grown on demand
  13  * and allocation from the heap is currently first-fit.
  14  *
  15  * Above this is an implementation of kmalloc/kfree. Blocks returned
  16  * from kmalloc are 8-byte aligned and prepended with a 8-byte header.
  17  * If kmalloc is asked for objects of PAGE_SIZE or larger, it calls
  18  * __get_free_pages directly so that it can return page-aligned blocks
  19  * and keeps a linked list of such pages and their orders. These
  20  * objects are detected in kfree() by their page alignment.
  21  *
  22  * SLAB is emulated on top of SLOB by simply calling constructors and
  23  * destructors for every SLAB allocation. Objects are returned with
  24  * the 8-byte alignment unless the SLAB_MUST_HWCACHE_ALIGN flag is
  25  * set, in which case the low-level allocator will fragment blocks to
  26  * create the proper alignment. Again, objects of page-size or greater
  27  * are allocated by calling __get_free_pages. As SLAB objects know
  28  * their size, no separate size bookkeeping is necessary and there is
  29  * essentially no allocation space overhead.
  30  */
  31
  32 #include <linux/slab.h>
  33 #include <linux/mm.h>
  34 #include <linux/cache.h>
  35 #include <linux/init.h>
  36 #include <linux/module.h>
  37 #include <linux/timer.h>
  38
  39 struct slob_block {
  40         int units;
  41         struct slob_block *next;
  42 };
  43 typedef struct slob_block slob_t;
  44
  45 #define SLOB_UNIT sizeof(slob_t)
  46 #define SLOB_UNITS(size) (((size) + SLOB_UNIT - 1)/SLOB_UNIT)
  47 #define SLOB_ALIGN L1_CACHE_BYTES
  48
  49 struct bigblock {
  50         int order;
  51         void *pages;
  52         struct bigblock *next;
  53 };
  54 typedef struct bigblock bigblock_t;
  55
  56 static slob_t arena = { .next = &arena, .units = 1 };
  57 static slob_t *slobfree = &arena;
  58 static bigblock_t *bigblocks;
  59 static DEFINE_SPINLOCK(slob_lock);
  60 static DEFINE_SPINLOCK(block_lock);
  61
  62 static void slob_free(void *b, int size);
  63
  64 static void *slob_alloc(size_t size, gfp_t gfp, int align)
  65 {
  66         slob_t *prev, *cur, *aligned = 0;
  67         int delta = 0, units = SLOB_UNITS(size);
  68         unsigned long flags;
  69
  70         spin_lock_irqsave(&slob_lock, flags);
  71         prev = slobfree;
  72         for (cur = prev->next; ; prev = cur, cur = cur->next) {
  73                 if (align) {
  74                         aligned = (slob_t *)ALIGN((unsigned long)cur, align);
  75                         delta = aligned - cur;
  76                 }
  77                 if (cur->units >= units + delta) { /* room enough? */
  78                         if (delta) { /* need to fragment head to align? */
  79                                 aligned->units = cur->units - delta;
  80                                 aligned->next = cur->next;
  81                                 cur->next = aligned;
  82                                 cur->units = delta;
  83                                 prev = cur;
  84                                 cur = aligned;
  85                         }
  86
  87                         if (cur->units == units) /* exact fit? */
  88                                 prev->next = cur->next; /* unlink */
  89                         else { /* fragment */
  90                                 prev->next = cur + units;
  91                                 prev->next->units = cur->units - units;
  92                                 prev->next->next = cur->next;
  93                                 cur->units = units;
  94                         }
  95
  96                         slobfree = prev;
  97                         spin_unlock_irqrestore(&slob_lock, flags);
  98                         return cur;
  99                 }
 100                 if (cur == slobfree) {
 101                         spin_unlock_irqrestore(&slob_lock, flags);
 102
 103                         if (size == PAGE_SIZE) /* trying to shrink arena? */
 104                                 return 0;
 105
 106                         cur = (slob_t *)__get_free_page(gfp);
 107                         if (!cur)
 108                                 return 0;
 109
 110                         slob_free(cur, PAGE_SIZE);
 111                         spin_lock_irqsave(&slob_lock, flags);
 112                         cur = slobfree;
 113                 }
 114         }
 115 }
 116
 117 static void slob_free(void *block, int size)
 118 {
 119         slob_t *cur, *b = (slob_t *)block;
 120         unsigned long flags;
 121
 122         if (!block)
 123                 return;
 124
 125         if (size)
 126                 b->units = SLOB_UNITS(size);
 127
 128         /* Find reinsertion point */
 129         spin_lock_irqsave(&slob_lock, flags);
 130         for (cur = slobfree; !(b > cur && b < cur->next); cur = cur->next)
 131                 if (cur >= cur->next && (b > cur || b < cur->next))
 132                         break;
 133
 134         if (b + b->units == cur->next) {
 135                 b->units += cur->next->units;
 136                 b->next = cur->next->next;
 137         } else
 138                 b->next = cur->next;
 139
 140         if (cur + cur->units == b) {
 141                 cur->units += b->units;
 142                 cur->next = b->next;
 143         } else
 144                 cur->next = b;
 145
 146         slobfree = cur;
 147
 148         spin_unlock_irqrestore(&slob_lock, flags);
 149 }
 150
 151 static int FASTCALL(find_order(int size));
 152 static int fastcall find_order(int size)
 153 {
 154         int order = 0;
 155         for ( ; size > 4096 ; size >>=1)
 156                 order++;
 157         return order;
 158 }
 159
 160 void *__kmalloc(size_t size, gfp_t gfp)
 161 {
 162         slob_t *m;
 163         bigblock_t *bb;
 164         unsigned long flags;
 165
 166         if (size < PAGE_SIZE - SLOB_UNIT) {
 167                 m = slob_alloc(size + SLOB_UNIT, gfp, 0);
 168                 return m ? (void *)(m + 1) : 0;
 169         }
 170
 171         bb = slob_alloc(sizeof(bigblock_t), gfp, 0);
 172         if (!bb)
 173                 return 0;
 174
 175         bb->order = find_order(size);
 176         bb->pages = (void *)__get_free_pages(gfp, bb->order);
 177
 178         if (bb->pages) {
 179                 spin_lock_irqsave(&block_lock, flags);
 180                 bb->next = bigblocks;
 181                 bigblocks = bb;
 182                 spin_unlock_irqrestore(&block_lock, flags);
 183                 return bb->pages;
 184         }
 185
 186         slob_free(bb, sizeof(bigblock_t));
 187         return 0;
 188 }
 189 EXPORT_SYMBOL(__kmalloc);
 190
 191 void kfree(const void *block)
 192 {
 193         bigblock_t *bb, **last = &bigblocks;
 194         unsigned long flags;
 195
 196         if (!block)
 197                 return;
 198
 199         if (!((unsigned long)block & (PAGE_SIZE-1))) {
 200                 /* might be on the big block list */
 201                 spin_lock_irqsave(&block_lock, flags);
 202                 for (bb = bigblocks; bb; last = &bb->next, bb = bb->next) {
 203                         if (bb->pages == block) {
 204                                 *last = bb->next;
 205                                 spin_unlock_irqrestore(&block_lock, flags);
 206                                 free_pages((unsigned long)block, bb->order);
 207                                 slob_free(bb, sizeof(bigblock_t));
 208                                 return;
 209                         }
 210                 }
 211                 spin_unlock_irqrestore(&block_lock, flags);
 212         }
 213
 214         slob_free((slob_t *)block - 1, 0);
 215         return;
 216 }
 217
 218 EXPORT_SYMBOL(kfree);
 219
 220 unsigned int ksize(const void *block)
 221 {
 222         bigblock_t *bb;
 223         unsigned long flags;
 224
 225         if (!block)
 226                 return 0;
 227
 228         if (!((unsigned long)block & (PAGE_SIZE-1))) {
 229                 spin_lock_irqsave(&block_lock, flags);
 230                 for (bb = bigblocks; bb; bb = bb->next)
 231                         if (bb->pages == block) {
 232                                 spin_unlock_irqrestore(&slob_lock, flags);
 233                                 return PAGE_SIZE << bb->order;
 234                         }
 235                 spin_unlock_irqrestore(&block_lock, flags);
 236         }
 237
 238         return ((slob_t *)block - 1)->units * SLOB_UNIT;
 239 }
 240
 241 struct kmem_cache {
 242         unsigned int size, align;
 243         const char *name;
 244         void (*ctor)(void *, struct kmem_cache *, unsigned long);
 245         void (*dtor)(void *, struct kmem_cache *, unsigned long);
 246 };
 247
 248 struct kmem_cache *kmem_cache_create(const char *name, size_t size,
 249         size_t align, unsigned long flags,
 250         void (*ctor)(void*, struct kmem_cache *, unsigned long),
 251         void (*dtor)(void*, struct kmem_cache *, unsigned long))
 252 {
 253         struct kmem_cache *c;
 254
 255         c = slob_alloc(sizeof(struct kmem_cache), flags, 0);
 256
 257         if (c) {
 258                 c->name = name;
 259                 c->size = size;
 260                 c->ctor = ctor;
 261                 c->dtor = dtor;
 262                 /* ignore alignment unless it's forced */
 263                 c->align = (flags & SLAB_MUST_HWCACHE_ALIGN) ? SLOB_ALIGN : 0;
 264                 if (c->align < align)
 265                         c->align = align;
 266         }
 267
 268         return c;
 269 }
 270 EXPORT_SYMBOL(kmem_cache_create);
 271
 272 void kmem_cache_destroy(struct kmem_cache *c)
 273 {
 274         slob_free(c, sizeof(struct kmem_cache));
 275 }
 276 EXPORT_SYMBOL(kmem_cache_destroy);
 277
 278 void *kmem_cache_alloc(struct kmem_cache *c, gfp_t flags)
 279 {
 280         void *b;
 281
 282         if (c->size < PAGE_SIZE)
 283                 b = slob_alloc(c->size, flags, c->align);
 284         else
 285                 b = (void *)__get_free_pages(flags, find_order(c->size));
 286
 287         if (c->ctor)
 288                 c->ctor(b, c, SLAB_CTOR_CONSTRUCTOR);
 289
 290         return b;
 291 }
 292 EXPORT_SYMBOL(kmem_cache_alloc);
 293
 294 void *kmem_cache_zalloc(struct kmem_cache *c, gfp_t flags)
 295 {
 296         void *ret = kmem_cache_alloc(c, flags);
 297         if (ret)
 298                 memset(ret, 0, c->size);
 299
 300         return ret;
 301 }
 302 EXPORT_SYMBOL(kmem_cache_zalloc);
 303
 304 void kmem_cache_free(struct kmem_cache *c, void *b)
 305 {
 306         if (c->dtor)
 307                 c->dtor(b, c, 0);
 308
 309         if (c->size < PAGE_SIZE)
 310                 slob_free(b, c->size);
 311         else
 312                 free_pages((unsigned long)b, find_order(c->size));
 313 }
 314 EXPORT_SYMBOL(kmem_cache_free);
 315
 316 unsigned int kmem_cache_size(struct kmem_cache *c)
 317 {
 318         return c->size;
 319 }
 320 EXPORT_SYMBOL(kmem_cache_size);
 321
 322 const char *kmem_cache_name(struct kmem_cache *c)
 323 {
 324         return c->name;
 325 }
 326 EXPORT_SYMBOL(kmem_cache_name);
 327
 328 static struct timer_list slob_timer = TIMER_INITIALIZER(
 329         (void (*)(unsigned long))kmem_cache_init, 0, 0);
 330
 331 int kmem_cache_shrink(struct kmem_cache *d)
 332 {
 333         return 0;
 334 }
 335 EXPORT_SYMBOL(kmem_cache_shrink);
 336
 337 int kmem_ptr_validate(struct kmem_cache *a, const void *b)
 338 {
 339         return 0;
 340 }
 341
 342 void kmem_cache_init(void)
 343 {
 344         void *p = slob_alloc(PAGE_SIZE, 0, PAGE_SIZE-1);
 345
 346         if (p)
 347                 free_page((unsigned long)p);
 348
 349         mod_timer(&slob_timer, jiffies + HZ);
 350 }