ASoC: Blackfin: keep better track of SPORT configuration state
[linux-2.6/linux-2.6-openrd.git] / mm / slob.c
bloba2d4ab32198d85b9773926c27ba7fdd07adfbfe7
1 /*
2 * SLOB Allocator: Simple List Of Blocks
4 * Matt Mackall <mpm@selenic.com> 12/30/03
6 * NUMA support by Paul Mundt, 2007.
8 * How SLOB works:
10 * The core of SLOB is a traditional K&R style heap allocator, with
11 * support for returning aligned objects. The granularity of this
12 * allocator is as little as 2 bytes, however typically most architectures
13 * will require 4 bytes on 32-bit and 8 bytes on 64-bit.
15 * The slob heap is a set of linked list of pages from alloc_pages(),
16 * and within each page, there is a singly-linked list of free blocks
17 * (slob_t). The heap is grown on demand. To reduce fragmentation,
18 * heap pages are segregated into three lists, with objects less than
19 * 256 bytes, objects less than 1024 bytes, and all other objects.
21 * Allocation from heap involves first searching for a page with
22 * sufficient free blocks (using a next-fit-like approach) followed by
23 * a first-fit scan of the page. Deallocation inserts objects back
24 * into the free list in address order, so this is effectively an
25 * address-ordered first fit.
27 * Above this is an implementation of kmalloc/kfree. Blocks returned
28 * from kmalloc are prepended with a 4-byte header with the kmalloc size.
29 * If kmalloc is asked for objects of PAGE_SIZE or larger, it calls
30 * alloc_pages() directly, allocating compound pages so the page order
31 * does not have to be separately tracked, and also stores the exact
32 * allocation size in page->private so that it can be used to accurately
33 * provide ksize(). These objects are detected in kfree() because slob_page()
34 * is false for them.
36 * SLAB is emulated on top of SLOB by simply calling constructors and
37 * destructors for every SLAB allocation. Objects are returned with the
38 * 4-byte alignment unless the SLAB_HWCACHE_ALIGN flag is set, in which
39 * case the low-level allocator will fragment blocks to create the proper
40 * alignment. Again, objects of page-size or greater are allocated by
41 * calling alloc_pages(). As SLAB objects know their size, no separate
42 * size bookkeeping is necessary and there is essentially no allocation
43 * space overhead, and compound pages aren't needed for multi-page
44 * allocations.
46 * NUMA support in SLOB is fairly simplistic, pushing most of the real
47 * logic down to the page allocator, and simply doing the node accounting
48 * on the upper levels. In the event that a node id is explicitly
49 * provided, alloc_pages_node() with the specified node id is used
50 * instead. The common case (or when the node id isn't explicitly provided)
51 * will default to the current node, as per numa_node_id().
53 * Node aware pages are still inserted in to the global freelist, and
54 * these are scanned for by matching against the node id encoded in the
55 * page flags. As a result, block allocations that can be satisfied from
56 * the freelist will only be done so on pages residing on the same node,
57 * in order to prevent random node placement.
60 #include <linux/kernel.h>
61 #include <linux/slab.h>
62 #include <linux/mm.h>
63 #include <linux/cache.h>
64 #include <linux/init.h>
65 #include <linux/module.h>
66 #include <linux/rcupdate.h>
67 #include <linux/list.h>
68 #include <trace/kmemtrace.h>
69 #include <asm/atomic.h>
72 * slob_block has a field 'units', which indicates size of block if +ve,
73 * or offset of next block if -ve (in SLOB_UNITs).
75 * Free blocks of size 1 unit simply contain the offset of the next block.
76 * Those with larger size contain their size in the first SLOB_UNIT of
77 * memory, and the offset of the next free block in the second SLOB_UNIT.
79 #if PAGE_SIZE <= (32767 * 2)
80 typedef s16 slobidx_t;
81 #else
82 typedef s32 slobidx_t;
83 #endif
85 struct slob_block {
86 slobidx_t units;
88 typedef struct slob_block slob_t;
91 * We use struct page fields to manage some slob allocation aspects,
92 * however to avoid the horrible mess in include/linux/mm_types.h, we'll
93 * just define our own struct page type variant here.
95 struct slob_page {
96 union {
97 struct {
98 unsigned long flags; /* mandatory */
99 atomic_t _count; /* mandatory */
100 slobidx_t units; /* free units left in page */
101 unsigned long pad[2];
102 slob_t *free; /* first free slob_t in page */
103 struct list_head list; /* linked list of free pages */
105 struct page page;
108 static inline void struct_slob_page_wrong_size(void)
109 { BUILD_BUG_ON(sizeof(struct slob_page) != sizeof(struct page)); }
112 * free_slob_page: call before a slob_page is returned to the page allocator.
114 static inline void free_slob_page(struct slob_page *sp)
116 reset_page_mapcount(&sp->page);
117 sp->page.mapping = NULL;
121 * All partially free slob pages go on these lists.
123 #define SLOB_BREAK1 256
124 #define SLOB_BREAK2 1024
125 static LIST_HEAD(free_slob_small);
126 static LIST_HEAD(free_slob_medium);
127 static LIST_HEAD(free_slob_large);
130 * is_slob_page: True for all slob pages (false for bigblock pages)
132 static inline int is_slob_page(struct slob_page *sp)
134 return PageSlobPage((struct page *)sp);
137 static inline void set_slob_page(struct slob_page *sp)
139 __SetPageSlobPage((struct page *)sp);
142 static inline void clear_slob_page(struct slob_page *sp)
144 __ClearPageSlobPage((struct page *)sp);
147 static inline struct slob_page *slob_page(const void *addr)
149 return (struct slob_page *)virt_to_page(addr);
153 * slob_page_free: true for pages on free_slob_pages list.
155 static inline int slob_page_free(struct slob_page *sp)
157 return PageSlobFree((struct page *)sp);
160 static void set_slob_page_free(struct slob_page *sp, struct list_head *list)
162 list_add(&sp->list, list);
163 __SetPageSlobFree((struct page *)sp);
166 static inline void clear_slob_page_free(struct slob_page *sp)
168 list_del(&sp->list);
169 __ClearPageSlobFree((struct page *)sp);
172 #define SLOB_UNIT sizeof(slob_t)
173 #define SLOB_UNITS(size) (((size) + SLOB_UNIT - 1)/SLOB_UNIT)
174 #define SLOB_ALIGN L1_CACHE_BYTES
177 * struct slob_rcu is inserted at the tail of allocated slob blocks, which
178 * were created with a SLAB_DESTROY_BY_RCU slab. slob_rcu is used to free
179 * the block using call_rcu.
181 struct slob_rcu {
182 struct rcu_head head;
183 int size;
187 * slob_lock protects all slob allocator structures.
189 static DEFINE_SPINLOCK(slob_lock);
192 * Encode the given size and next info into a free slob block s.
194 static void set_slob(slob_t *s, slobidx_t size, slob_t *next)
196 slob_t *base = (slob_t *)((unsigned long)s & PAGE_MASK);
197 slobidx_t offset = next - base;
199 if (size > 1) {
200 s[0].units = size;
201 s[1].units = offset;
202 } else
203 s[0].units = -offset;
207 * Return the size of a slob block.
209 static slobidx_t slob_units(slob_t *s)
211 if (s->units > 0)
212 return s->units;
213 return 1;
217 * Return the next free slob block pointer after this one.
219 static slob_t *slob_next(slob_t *s)
221 slob_t *base = (slob_t *)((unsigned long)s & PAGE_MASK);
222 slobidx_t next;
224 if (s[0].units < 0)
225 next = -s[0].units;
226 else
227 next = s[1].units;
228 return base+next;
232 * Returns true if s is the last free block in its page.
234 static int slob_last(slob_t *s)
236 return !((unsigned long)slob_next(s) & ~PAGE_MASK);
239 static void *slob_new_pages(gfp_t gfp, int order, int node)
241 void *page;
243 #ifdef CONFIG_NUMA
244 if (node != -1)
245 page = alloc_pages_node(node, gfp, order);
246 else
247 #endif
248 page = alloc_pages(gfp, order);
250 if (!page)
251 return NULL;
253 return page_address(page);
256 static void slob_free_pages(void *b, int order)
258 free_pages((unsigned long)b, order);
262 * Allocate a slob block within a given slob_page sp.
264 static void *slob_page_alloc(struct slob_page *sp, size_t size, int align)
266 slob_t *prev, *cur, *aligned = NULL;
267 int delta = 0, units = SLOB_UNITS(size);
269 for (prev = NULL, cur = sp->free; ; prev = cur, cur = slob_next(cur)) {
270 slobidx_t avail = slob_units(cur);
272 if (align) {
273 aligned = (slob_t *)ALIGN((unsigned long)cur, align);
274 delta = aligned - cur;
276 if (avail >= units + delta) { /* room enough? */
277 slob_t *next;
279 if (delta) { /* need to fragment head to align? */
280 next = slob_next(cur);
281 set_slob(aligned, avail - delta, next);
282 set_slob(cur, delta, aligned);
283 prev = cur;
284 cur = aligned;
285 avail = slob_units(cur);
288 next = slob_next(cur);
289 if (avail == units) { /* exact fit? unlink. */
290 if (prev)
291 set_slob(prev, slob_units(prev), next);
292 else
293 sp->free = next;
294 } else { /* fragment */
295 if (prev)
296 set_slob(prev, slob_units(prev), cur + units);
297 else
298 sp->free = cur + units;
299 set_slob(cur + units, avail - units, next);
302 sp->units -= units;
303 if (!sp->units)
304 clear_slob_page_free(sp);
305 return cur;
307 if (slob_last(cur))
308 return NULL;
313 * slob_alloc: entry point into the slob allocator.
315 static void *slob_alloc(size_t size, gfp_t gfp, int align, int node)
317 struct slob_page *sp;
318 struct list_head *prev;
319 struct list_head *slob_list;
320 slob_t *b = NULL;
321 unsigned long flags;
323 if (size < SLOB_BREAK1)
324 slob_list = &free_slob_small;
325 else if (size < SLOB_BREAK2)
326 slob_list = &free_slob_medium;
327 else
328 slob_list = &free_slob_large;
330 spin_lock_irqsave(&slob_lock, flags);
331 /* Iterate through each partially free page, try to find room */
332 list_for_each_entry(sp, slob_list, list) {
333 #ifdef CONFIG_NUMA
335 * If there's a node specification, search for a partial
336 * page with a matching node id in the freelist.
338 if (node != -1 && page_to_nid(&sp->page) != node)
339 continue;
340 #endif
341 /* Enough room on this page? */
342 if (sp->units < SLOB_UNITS(size))
343 continue;
345 /* Attempt to alloc */
346 prev = sp->list.prev;
347 b = slob_page_alloc(sp, size, align);
348 if (!b)
349 continue;
351 /* Improve fragment distribution and reduce our average
352 * search time by starting our next search here. (see
353 * Knuth vol 1, sec 2.5, pg 449) */
354 if (prev != slob_list->prev &&
355 slob_list->next != prev->next)
356 list_move_tail(slob_list, prev->next);
357 break;
359 spin_unlock_irqrestore(&slob_lock, flags);
361 /* Not enough space: must allocate a new page */
362 if (!b) {
363 b = slob_new_pages(gfp & ~__GFP_ZERO, 0, node);
364 if (!b)
365 return NULL;
366 sp = slob_page(b);
367 set_slob_page(sp);
369 spin_lock_irqsave(&slob_lock, flags);
370 sp->units = SLOB_UNITS(PAGE_SIZE);
371 sp->free = b;
372 INIT_LIST_HEAD(&sp->list);
373 set_slob(b, SLOB_UNITS(PAGE_SIZE), b + SLOB_UNITS(PAGE_SIZE));
374 set_slob_page_free(sp, slob_list);
375 b = slob_page_alloc(sp, size, align);
376 BUG_ON(!b);
377 spin_unlock_irqrestore(&slob_lock, flags);
379 if (unlikely((gfp & __GFP_ZERO) && b))
380 memset(b, 0, size);
381 return b;
385 * slob_free: entry point into the slob allocator.
387 static void slob_free(void *block, int size)
389 struct slob_page *sp;
390 slob_t *prev, *next, *b = (slob_t *)block;
391 slobidx_t units;
392 unsigned long flags;
394 if (unlikely(ZERO_OR_NULL_PTR(block)))
395 return;
396 BUG_ON(!size);
398 sp = slob_page(block);
399 units = SLOB_UNITS(size);
401 spin_lock_irqsave(&slob_lock, flags);
403 if (sp->units + units == SLOB_UNITS(PAGE_SIZE)) {
404 /* Go directly to page allocator. Do not pass slob allocator */
405 if (slob_page_free(sp))
406 clear_slob_page_free(sp);
407 spin_unlock_irqrestore(&slob_lock, flags);
408 clear_slob_page(sp);
409 free_slob_page(sp);
410 free_page((unsigned long)b);
411 return;
414 if (!slob_page_free(sp)) {
415 /* This slob page is about to become partially free. Easy! */
416 sp->units = units;
417 sp->free = b;
418 set_slob(b, units,
419 (void *)((unsigned long)(b +
420 SLOB_UNITS(PAGE_SIZE)) & PAGE_MASK));
421 set_slob_page_free(sp, &free_slob_small);
422 goto out;
426 * Otherwise the page is already partially free, so find reinsertion
427 * point.
429 sp->units += units;
431 if (b < sp->free) {
432 if (b + units == sp->free) {
433 units += slob_units(sp->free);
434 sp->free = slob_next(sp->free);
436 set_slob(b, units, sp->free);
437 sp->free = b;
438 } else {
439 prev = sp->free;
440 next = slob_next(prev);
441 while (b > next) {
442 prev = next;
443 next = slob_next(prev);
446 if (!slob_last(prev) && b + units == next) {
447 units += slob_units(next);
448 set_slob(b, units, slob_next(next));
449 } else
450 set_slob(b, units, next);
452 if (prev + slob_units(prev) == b) {
453 units = slob_units(b) + slob_units(prev);
454 set_slob(prev, units, slob_next(b));
455 } else
456 set_slob(prev, slob_units(prev), b);
458 out:
459 spin_unlock_irqrestore(&slob_lock, flags);
463 * End of slob allocator proper. Begin kmem_cache_alloc and kmalloc frontend.
466 #ifndef ARCH_KMALLOC_MINALIGN
467 #define ARCH_KMALLOC_MINALIGN __alignof__(unsigned long)
468 #endif
470 #ifndef ARCH_SLAB_MINALIGN
471 #define ARCH_SLAB_MINALIGN __alignof__(unsigned long)
472 #endif
474 void *__kmalloc_node(size_t size, gfp_t gfp, int node)
476 unsigned int *m;
477 int align = max(ARCH_KMALLOC_MINALIGN, ARCH_SLAB_MINALIGN);
478 void *ret;
480 lockdep_trace_alloc(gfp);
482 if (size < PAGE_SIZE - align) {
483 if (!size)
484 return ZERO_SIZE_PTR;
486 m = slob_alloc(size + align, gfp, align, node);
488 if (!m)
489 return NULL;
490 *m = size;
491 ret = (void *)m + align;
493 trace_kmalloc_node(_RET_IP_, ret,
494 size, size + align, gfp, node);
495 } else {
496 unsigned int order = get_order(size);
498 ret = slob_new_pages(gfp | __GFP_COMP, get_order(size), node);
499 if (ret) {
500 struct page *page;
501 page = virt_to_page(ret);
502 page->private = size;
505 trace_kmalloc_node(_RET_IP_, ret,
506 size, PAGE_SIZE << order, gfp, node);
509 return ret;
511 EXPORT_SYMBOL(__kmalloc_node);
513 void kfree(const void *block)
515 struct slob_page *sp;
517 trace_kfree(_RET_IP_, block);
519 if (unlikely(ZERO_OR_NULL_PTR(block)))
520 return;
522 sp = slob_page(block);
523 if (is_slob_page(sp)) {
524 int align = max(ARCH_KMALLOC_MINALIGN, ARCH_SLAB_MINALIGN);
525 unsigned int *m = (unsigned int *)(block - align);
526 slob_free(m, *m + align);
527 } else
528 put_page(&sp->page);
530 EXPORT_SYMBOL(kfree);
532 /* can't use ksize for kmem_cache_alloc memory, only kmalloc */
533 size_t ksize(const void *block)
535 struct slob_page *sp;
537 BUG_ON(!block);
538 if (unlikely(block == ZERO_SIZE_PTR))
539 return 0;
541 sp = slob_page(block);
542 if (is_slob_page(sp)) {
543 int align = max(ARCH_KMALLOC_MINALIGN, ARCH_SLAB_MINALIGN);
544 unsigned int *m = (unsigned int *)(block - align);
545 return SLOB_UNITS(*m) * SLOB_UNIT;
546 } else
547 return sp->page.private;
549 EXPORT_SYMBOL(ksize);
551 struct kmem_cache {
552 unsigned int size, align;
553 unsigned long flags;
554 const char *name;
555 void (*ctor)(void *);
558 struct kmem_cache *kmem_cache_create(const char *name, size_t size,
559 size_t align, unsigned long flags, void (*ctor)(void *))
561 struct kmem_cache *c;
563 c = slob_alloc(sizeof(struct kmem_cache),
564 GFP_KERNEL, ARCH_KMALLOC_MINALIGN, -1);
566 if (c) {
567 c->name = name;
568 c->size = size;
569 if (flags & SLAB_DESTROY_BY_RCU) {
570 /* leave room for rcu footer at the end of object */
571 c->size += sizeof(struct slob_rcu);
573 c->flags = flags;
574 c->ctor = ctor;
575 /* ignore alignment unless it's forced */
576 c->align = (flags & SLAB_HWCACHE_ALIGN) ? SLOB_ALIGN : 0;
577 if (c->align < ARCH_SLAB_MINALIGN)
578 c->align = ARCH_SLAB_MINALIGN;
579 if (c->align < align)
580 c->align = align;
581 } else if (flags & SLAB_PANIC)
582 panic("Cannot create slab cache %s\n", name);
584 return c;
586 EXPORT_SYMBOL(kmem_cache_create);
588 void kmem_cache_destroy(struct kmem_cache *c)
590 slob_free(c, sizeof(struct kmem_cache));
592 EXPORT_SYMBOL(kmem_cache_destroy);
594 void *kmem_cache_alloc_node(struct kmem_cache *c, gfp_t flags, int node)
596 void *b;
598 if (c->size < PAGE_SIZE) {
599 b = slob_alloc(c->size, flags, c->align, node);
600 trace_kmem_cache_alloc_node(_RET_IP_, b, c->size,
601 SLOB_UNITS(c->size) * SLOB_UNIT,
602 flags, node);
603 } else {
604 b = slob_new_pages(flags, get_order(c->size), node);
605 trace_kmem_cache_alloc_node(_RET_IP_, b, c->size,
606 PAGE_SIZE << get_order(c->size),
607 flags, node);
610 if (c->ctor)
611 c->ctor(b);
613 return b;
615 EXPORT_SYMBOL(kmem_cache_alloc_node);
617 static void __kmem_cache_free(void *b, int size)
619 if (size < PAGE_SIZE)
620 slob_free(b, size);
621 else
622 slob_free_pages(b, get_order(size));
625 static void kmem_rcu_free(struct rcu_head *head)
627 struct slob_rcu *slob_rcu = (struct slob_rcu *)head;
628 void *b = (void *)slob_rcu - (slob_rcu->size - sizeof(struct slob_rcu));
630 __kmem_cache_free(b, slob_rcu->size);
633 void kmem_cache_free(struct kmem_cache *c, void *b)
635 if (unlikely(c->flags & SLAB_DESTROY_BY_RCU)) {
636 struct slob_rcu *slob_rcu;
637 slob_rcu = b + (c->size - sizeof(struct slob_rcu));
638 INIT_RCU_HEAD(&slob_rcu->head);
639 slob_rcu->size = c->size;
640 call_rcu(&slob_rcu->head, kmem_rcu_free);
641 } else {
642 __kmem_cache_free(b, c->size);
645 trace_kmem_cache_free(_RET_IP_, b);
647 EXPORT_SYMBOL(kmem_cache_free);
649 unsigned int kmem_cache_size(struct kmem_cache *c)
651 return c->size;
653 EXPORT_SYMBOL(kmem_cache_size);
655 const char *kmem_cache_name(struct kmem_cache *c)
657 return c->name;
659 EXPORT_SYMBOL(kmem_cache_name);
661 int kmem_cache_shrink(struct kmem_cache *d)
663 return 0;
665 EXPORT_SYMBOL(kmem_cache_shrink);
667 int kmem_ptr_validate(struct kmem_cache *a, const void *b)
669 return 0;
672 static unsigned int slob_ready __read_mostly;
674 int slab_is_available(void)
676 return slob_ready;
679 void __init kmem_cache_init(void)
681 slob_ready = 1;