Root/mm/slob.c

1/*
2 * SLOB Allocator: Simple List Of Blocks
3 *
4 * Matt Mackall <mpm@selenic.com> 12/30/03
5 *
6 * NUMA support by Paul Mundt, 2007.
7 *
8 * How SLOB works:
9 *
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.
14 *
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.
20 *
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.
26 *
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.
32 * These objects are detected in kfree() because PageSlab()
33 * is false for them.
34 *
35 * SLAB is emulated on top of SLOB by simply calling constructors and
36 * destructors for every SLAB allocation. Objects are returned with the
37 * 4-byte alignment unless the SLAB_HWCACHE_ALIGN flag is set, in which
38 * case the low-level allocator will fragment blocks to create the proper
39 * alignment. Again, objects of page-size or greater are allocated by
40 * calling alloc_pages(). As SLAB objects know their size, no separate
41 * size bookkeeping is necessary and there is essentially no allocation
42 * space overhead, and compound pages aren't needed for multi-page
43 * allocations.
44 *
45 * NUMA support in SLOB is fairly simplistic, pushing most of the real
46 * logic down to the page allocator, and simply doing the node accounting
47 * on the upper levels. In the event that a node id is explicitly
48 * provided, alloc_pages_exact_node() with the specified node id is used
49 * instead. The common case (or when the node id isn't explicitly provided)
50 * will default to the current node, as per numa_node_id().
51 *
52 * Node aware pages are still inserted in to the global freelist, and
53 * these are scanned for by matching against the node id encoded in the
54 * page flags. As a result, block allocations that can be satisfied from
55 * the freelist will only be done so on pages residing on the same node,
56 * in order to prevent random node placement.
57 */
58
59#include <linux/kernel.h>
60#include <linux/slab.h>
61
62#include <linux/mm.h>
63#include <linux/swap.h> /* struct reclaim_state */
64#include <linux/cache.h>
65#include <linux/init.h>
66#include <linux/export.h>
67#include <linux/rcupdate.h>
68#include <linux/list.h>
69#include <linux/kmemleak.h>
70
71#include <trace/events/kmem.h>
72
73#include <linux/atomic.h>
74
75#include "slab.h"
76/*
77 * slob_block has a field 'units', which indicates size of block if +ve,
78 * or offset of next block if -ve (in SLOB_UNITs).
79 *
80 * Free blocks of size 1 unit simply contain the offset of the next block.
81 * Those with larger size contain their size in the first SLOB_UNIT of
82 * memory, and the offset of the next free block in the second SLOB_UNIT.
83 */
84#if PAGE_SIZE <= (32767 * 2)
85typedef s16 slobidx_t;
86#else
87typedef s32 slobidx_t;
88#endif
89
90struct slob_block {
91    slobidx_t units;
92};
93typedef struct slob_block slob_t;
94
95/*
96 * All partially free slob pages go on these lists.
97 */
98#define SLOB_BREAK1 256
99#define SLOB_BREAK2 1024
100static LIST_HEAD(free_slob_small);
101static LIST_HEAD(free_slob_medium);
102static LIST_HEAD(free_slob_large);
103
104/*
105 * slob_page_free: true for pages on free_slob_pages list.
106 */
107static inline int slob_page_free(struct page *sp)
108{
109    return PageSlobFree(sp);
110}
111
112static void set_slob_page_free(struct page *sp, struct list_head *list)
113{
114    list_add(&sp->list, list);
115    __SetPageSlobFree(sp);
116}
117
118static inline void clear_slob_page_free(struct page *sp)
119{
120    list_del(&sp->list);
121    __ClearPageSlobFree(sp);
122}
123
124#define SLOB_UNIT sizeof(slob_t)
125#define SLOB_UNITS(size) DIV_ROUND_UP(size, SLOB_UNIT)
126
127/*
128 * struct slob_rcu is inserted at the tail of allocated slob blocks, which
129 * were created with a SLAB_DESTROY_BY_RCU slab. slob_rcu is used to free
130 * the block using call_rcu.
131 */
132struct slob_rcu {
133    struct rcu_head head;
134    int size;
135};
136
137/*
138 * slob_lock protects all slob allocator structures.
139 */
140static DEFINE_SPINLOCK(slob_lock);
141
142/*
143 * Encode the given size and next info into a free slob block s.
144 */
145static void set_slob(slob_t *s, slobidx_t size, slob_t *next)
146{
147    slob_t *base = (slob_t *)((unsigned long)s & PAGE_MASK);
148    slobidx_t offset = next - base;
149
150    if (size > 1) {
151        s[0].units = size;
152        s[1].units = offset;
153    } else
154        s[0].units = -offset;
155}
156
157/*
158 * Return the size of a slob block.
159 */
160static slobidx_t slob_units(slob_t *s)
161{
162    if (s->units > 0)
163        return s->units;
164    return 1;
165}
166
167/*
168 * Return the next free slob block pointer after this one.
169 */
170static slob_t *slob_next(slob_t *s)
171{
172    slob_t *base = (slob_t *)((unsigned long)s & PAGE_MASK);
173    slobidx_t next;
174
175    if (s[0].units < 0)
176        next = -s[0].units;
177    else
178        next = s[1].units;
179    return base+next;
180}
181
182/*
183 * Returns true if s is the last free block in its page.
184 */
185static int slob_last(slob_t *s)
186{
187    return !((unsigned long)slob_next(s) & ~PAGE_MASK);
188}
189
190static void *slob_new_pages(gfp_t gfp, int order, int node)
191{
192    void *page;
193
194#ifdef CONFIG_NUMA
195    if (node != NUMA_NO_NODE)
196        page = alloc_pages_exact_node(node, gfp, order);
197    else
198#endif
199        page = alloc_pages(gfp, order);
200
201    if (!page)
202        return NULL;
203
204    return page_address(page);
205}
206
207static void slob_free_pages(void *b, int order)
208{
209    if (current->reclaim_state)
210        current->reclaim_state->reclaimed_slab += 1 << order;
211    free_pages((unsigned long)b, order);
212}
213
214/*
215 * Allocate a slob block within a given slob_page sp.
216 */
217static void *slob_page_alloc(struct page *sp, size_t size, int align)
218{
219    slob_t *prev, *cur, *aligned = NULL;
220    int delta = 0, units = SLOB_UNITS(size);
221
222    for (prev = NULL, cur = sp->freelist; ; prev = cur, cur = slob_next(cur)) {
223        slobidx_t avail = slob_units(cur);
224
225        if (align) {
226            aligned = (slob_t *)ALIGN((unsigned long)cur, align);
227            delta = aligned - cur;
228        }
229        if (avail >= units + delta) { /* room enough? */
230            slob_t *next;
231
232            if (delta) { /* need to fragment head to align? */
233                next = slob_next(cur);
234                set_slob(aligned, avail - delta, next);
235                set_slob(cur, delta, aligned);
236                prev = cur;
237                cur = aligned;
238                avail = slob_units(cur);
239            }
240
241            next = slob_next(cur);
242            if (avail == units) { /* exact fit? unlink. */
243                if (prev)
244                    set_slob(prev, slob_units(prev), next);
245                else
246                    sp->freelist = next;
247            } else { /* fragment */
248                if (prev)
249                    set_slob(prev, slob_units(prev), cur + units);
250                else
251                    sp->freelist = cur + units;
252                set_slob(cur + units, avail - units, next);
253            }
254
255            sp->units -= units;
256            if (!sp->units)
257                clear_slob_page_free(sp);
258            return cur;
259        }
260        if (slob_last(cur))
261            return NULL;
262    }
263}
264
265/*
266 * slob_alloc: entry point into the slob allocator.
267 */
268static void *slob_alloc(size_t size, gfp_t gfp, int align, int node)
269{
270    struct page *sp;
271    struct list_head *prev;
272    struct list_head *slob_list;
273    slob_t *b = NULL;
274    unsigned long flags;
275
276    if (size < SLOB_BREAK1)
277        slob_list = &free_slob_small;
278    else if (size < SLOB_BREAK2)
279        slob_list = &free_slob_medium;
280    else
281        slob_list = &free_slob_large;
282
283    spin_lock_irqsave(&slob_lock, flags);
284    /* Iterate through each partially free page, try to find room */
285    list_for_each_entry(sp, slob_list, list) {
286#ifdef CONFIG_NUMA
287        /*
288         * If there's a node specification, search for a partial
289         * page with a matching node id in the freelist.
290         */
291        if (node != NUMA_NO_NODE && page_to_nid(sp) != node)
292            continue;
293#endif
294        /* Enough room on this page? */
295        if (sp->units < SLOB_UNITS(size))
296            continue;
297
298        /* Attempt to alloc */
299        prev = sp->list.prev;
300        b = slob_page_alloc(sp, size, align);
301        if (!b)
302            continue;
303
304        /* Improve fragment distribution and reduce our average
305         * search time by starting our next search here. (see
306         * Knuth vol 1, sec 2.5, pg 449) */
307        if (prev != slob_list->prev &&
308                slob_list->next != prev->next)
309            list_move_tail(slob_list, prev->next);
310        break;
311    }
312    spin_unlock_irqrestore(&slob_lock, flags);
313
314    /* Not enough space: must allocate a new page */
315    if (!b) {
316        b = slob_new_pages(gfp & ~__GFP_ZERO, 0, node);
317        if (!b)
318            return NULL;
319        sp = virt_to_page(b);
320        __SetPageSlab(sp);
321
322        spin_lock_irqsave(&slob_lock, flags);
323        sp->units = SLOB_UNITS(PAGE_SIZE);
324        sp->freelist = b;
325        INIT_LIST_HEAD(&sp->list);
326        set_slob(b, SLOB_UNITS(PAGE_SIZE), b + SLOB_UNITS(PAGE_SIZE));
327        set_slob_page_free(sp, slob_list);
328        b = slob_page_alloc(sp, size, align);
329        BUG_ON(!b);
330        spin_unlock_irqrestore(&slob_lock, flags);
331    }
332    if (unlikely((gfp & __GFP_ZERO) && b))
333        memset(b, 0, size);
334    return b;
335}
336
337/*
338 * slob_free: entry point into the slob allocator.
339 */
340static void slob_free(void *block, int size)
341{
342    struct page *sp;
343    slob_t *prev, *next, *b = (slob_t *)block;
344    slobidx_t units;
345    unsigned long flags;
346    struct list_head *slob_list;
347
348    if (unlikely(ZERO_OR_NULL_PTR(block)))
349        return;
350    BUG_ON(!size);
351
352    sp = virt_to_page(block);
353    units = SLOB_UNITS(size);
354
355    spin_lock_irqsave(&slob_lock, flags);
356
357    if (sp->units + units == SLOB_UNITS(PAGE_SIZE)) {
358        /* Go directly to page allocator. Do not pass slob allocator */
359        if (slob_page_free(sp))
360            clear_slob_page_free(sp);
361        spin_unlock_irqrestore(&slob_lock, flags);
362        __ClearPageSlab(sp);
363        page_mapcount_reset(sp);
364        slob_free_pages(b, 0);
365        return;
366    }
367
368    if (!slob_page_free(sp)) {
369        /* This slob page is about to become partially free. Easy! */
370        sp->units = units;
371        sp->freelist = b;
372        set_slob(b, units,
373            (void *)((unsigned long)(b +
374                    SLOB_UNITS(PAGE_SIZE)) & PAGE_MASK));
375        if (size < SLOB_BREAK1)
376            slob_list = &free_slob_small;
377        else if (size < SLOB_BREAK2)
378            slob_list = &free_slob_medium;
379        else
380            slob_list = &free_slob_large;
381        set_slob_page_free(sp, slob_list);
382        goto out;
383    }
384
385    /*
386     * Otherwise the page is already partially free, so find reinsertion
387     * point.
388     */
389    sp->units += units;
390
391    if (b < (slob_t *)sp->freelist) {
392        if (b + units == sp->freelist) {
393            units += slob_units(sp->freelist);
394            sp->freelist = slob_next(sp->freelist);
395        }
396        set_slob(b, units, sp->freelist);
397        sp->freelist = b;
398    } else {
399        prev = sp->freelist;
400        next = slob_next(prev);
401        while (b > next) {
402            prev = next;
403            next = slob_next(prev);
404        }
405
406        if (!slob_last(prev) && b + units == next) {
407            units += slob_units(next);
408            set_slob(b, units, slob_next(next));
409        } else
410            set_slob(b, units, next);
411
412        if (prev + slob_units(prev) == b) {
413            units = slob_units(b) + slob_units(prev);
414            set_slob(prev, units, slob_next(b));
415        } else
416            set_slob(prev, slob_units(prev), b);
417    }
418out:
419    spin_unlock_irqrestore(&slob_lock, flags);
420}
421
422/*
423 * End of slob allocator proper. Begin kmem_cache_alloc and kmalloc frontend.
424 */
425
426static __always_inline void *
427__do_kmalloc_node(size_t size, gfp_t gfp, int node, unsigned long caller)
428{
429    unsigned int *m;
430    int align = max_t(size_t, ARCH_KMALLOC_MINALIGN, ARCH_SLAB_MINALIGN);
431    void *ret;
432
433    gfp &= gfp_allowed_mask;
434
435    lockdep_trace_alloc(gfp);
436
437    if (size < PAGE_SIZE - align) {
438        if (!size)
439            return ZERO_SIZE_PTR;
440
441        m = slob_alloc(size + align, gfp, align, node);
442
443        if (!m)
444            return NULL;
445        *m = size;
446        ret = (void *)m + align;
447
448        trace_kmalloc_node(caller, ret,
449                   size, size + align, gfp, node);
450    } else {
451        unsigned int order = get_order(size);
452
453        if (likely(order))
454            gfp |= __GFP_COMP;
455        ret = slob_new_pages(gfp, order, node);
456
457        trace_kmalloc_node(caller, ret,
458                   size, PAGE_SIZE << order, gfp, node);
459    }
460
461    kmemleak_alloc(ret, size, 1, gfp);
462    return ret;
463}
464
465void *__kmalloc_node(size_t size, gfp_t gfp, int node)
466{
467    return __do_kmalloc_node(size, gfp, node, _RET_IP_);
468}
469EXPORT_SYMBOL(__kmalloc_node);
470
471#ifdef CONFIG_TRACING
472void *__kmalloc_track_caller(size_t size, gfp_t gfp, unsigned long caller)
473{
474    return __do_kmalloc_node(size, gfp, NUMA_NO_NODE, caller);
475}
476
477#ifdef CONFIG_NUMA
478void *__kmalloc_node_track_caller(size_t size, gfp_t gfp,
479                    int node, unsigned long caller)
480{
481    return __do_kmalloc_node(size, gfp, node, caller);
482}
483#endif
484#endif
485
486void kfree(const void *block)
487{
488    struct page *sp;
489
490    trace_kfree(_RET_IP_, block);
491
492    if (unlikely(ZERO_OR_NULL_PTR(block)))
493        return;
494    kmemleak_free(block);
495
496    sp = virt_to_page(block);
497    if (PageSlab(sp)) {
498        int align = max_t(size_t, ARCH_KMALLOC_MINALIGN, ARCH_SLAB_MINALIGN);
499        unsigned int *m = (unsigned int *)(block - align);
500        slob_free(m, *m + align);
501    } else
502        __free_pages(sp, compound_order(sp));
503}
504EXPORT_SYMBOL(kfree);
505
506/* can't use ksize for kmem_cache_alloc memory, only kmalloc */
507size_t ksize(const void *block)
508{
509    struct page *sp;
510    int align;
511    unsigned int *m;
512
513    BUG_ON(!block);
514    if (unlikely(block == ZERO_SIZE_PTR))
515        return 0;
516
517    sp = virt_to_page(block);
518    if (unlikely(!PageSlab(sp)))
519        return PAGE_SIZE << compound_order(sp);
520
521    align = max_t(size_t, ARCH_KMALLOC_MINALIGN, ARCH_SLAB_MINALIGN);
522    m = (unsigned int *)(block - align);
523    return SLOB_UNITS(*m) * SLOB_UNIT;
524}
525EXPORT_SYMBOL(ksize);
526
527int __kmem_cache_create(struct kmem_cache *c, unsigned long flags)
528{
529    if (flags & SLAB_DESTROY_BY_RCU) {
530        /* leave room for rcu footer at the end of object */
531        c->size += sizeof(struct slob_rcu);
532    }
533    c->flags = flags;
534    return 0;
535}
536
537void *kmem_cache_alloc_node(struct kmem_cache *c, gfp_t flags, int node)
538{
539    void *b;
540
541    flags &= gfp_allowed_mask;
542
543    lockdep_trace_alloc(flags);
544
545    if (c->size < PAGE_SIZE) {
546        b = slob_alloc(c->size, flags, c->align, node);
547        trace_kmem_cache_alloc_node(_RET_IP_, b, c->object_size,
548                        SLOB_UNITS(c->size) * SLOB_UNIT,
549                        flags, node);
550    } else {
551        b = slob_new_pages(flags, get_order(c->size), node);
552        trace_kmem_cache_alloc_node(_RET_IP_, b, c->object_size,
553                        PAGE_SIZE << get_order(c->size),
554                        flags, node);
555    }
556
557    if (b && c->ctor)
558        c->ctor(b);
559
560    kmemleak_alloc_recursive(b, c->size, 1, c->flags, flags);
561    return b;
562}
563EXPORT_SYMBOL(kmem_cache_alloc_node);
564
565static void __kmem_cache_free(void *b, int size)
566{
567    if (size < PAGE_SIZE)
568        slob_free(b, size);
569    else
570        slob_free_pages(b, get_order(size));
571}
572
573static void kmem_rcu_free(struct rcu_head *head)
574{
575    struct slob_rcu *slob_rcu = (struct slob_rcu *)head;
576    void *b = (void *)slob_rcu - (slob_rcu->size - sizeof(struct slob_rcu));
577
578    __kmem_cache_free(b, slob_rcu->size);
579}
580
581void kmem_cache_free(struct kmem_cache *c, void *b)
582{
583    kmemleak_free_recursive(b, c->flags);
584    if (unlikely(c->flags & SLAB_DESTROY_BY_RCU)) {
585        struct slob_rcu *slob_rcu;
586        slob_rcu = b + (c->size - sizeof(struct slob_rcu));
587        slob_rcu->size = c->size;
588        call_rcu(&slob_rcu->head, kmem_rcu_free);
589    } else {
590        __kmem_cache_free(b, c->size);
591    }
592
593    trace_kmem_cache_free(_RET_IP_, b);
594}
595EXPORT_SYMBOL(kmem_cache_free);
596
597int __kmem_cache_shutdown(struct kmem_cache *c)
598{
599    /* No way to check for remaining objects */
600    return 0;
601}
602
603int kmem_cache_shrink(struct kmem_cache *d)
604{
605    return 0;
606}
607EXPORT_SYMBOL(kmem_cache_shrink);
608
609struct kmem_cache kmem_cache_boot = {
610    .name = "kmem_cache",
611    .size = sizeof(struct kmem_cache),
612    .flags = SLAB_PANIC,
613    .align = ARCH_KMALLOC_MINALIGN,
614};
615
616void __init kmem_cache_init(void)
617{
618    kmem_cache = &kmem_cache_boot;
619    slab_state = UP;
620}
621
622void __init kmem_cache_init_late(void)
623{
624    slab_state = FULL;
625}
626

Archive Download this file



interactive