4 * Copyright (C) 1991, 1992, 1993, 1994 Linus Torvalds
8 * This file contains the default values for the operation of the
9 * Linux VM subsystem. Fine-tuning documentation can be found in
10 * Documentation/sysctl/vm.txt.
12 * Swap aging added 23.2.95, Stephen Tweedie.
13 * Buffermem limits added 12.3.98, Rik van Riel.
17 #include <linux/sched.h>
18 #include <linux/kernel_stat.h>
19 #include <linux/swap.h>
20 #include <linux/mman.h>
21 #include <linux/pagemap.h>
22 #include <linux/pagevec.h>
23 #include <linux/init.h>
24 #include <linux/export.h>
25 #include <linux/mm_inline.h>
26 #include <linux/percpu_counter.h>
27 #include <linux/percpu.h>
28 #include <linux/cpu.h>
29 #include <linux/notifier.h>
30 #include <linux/backing-dev.h>
31 #include <linux/memcontrol.h>
32 #include <linux/gfp.h>
33 #include <linux/uio.h>
34 #include <linux/hugetlb.h>
35 #include <linux/page_idle.h>
39 #define CREATE_TRACE_POINTS
40 #include <trace/events/pagemap.h>
42 /* How many pages do we try to swap or page in/out together? */
45 static DEFINE_PER_CPU(struct pagevec
, lru_add_pvec
);
46 static DEFINE_PER_CPU(struct pagevec
, lru_rotate_pvecs
);
47 static DEFINE_PER_CPU(struct pagevec
, lru_deactivate_file_pvecs
);
50 * This path almost never happens for VM activity - pages are normally
51 * freed via pagevecs. But it gets used by networking.
53 static void __page_cache_release(struct page
*page
)
56 struct zone
*zone
= page_zone(page
);
57 struct lruvec
*lruvec
;
60 spin_lock_irqsave(&zone
->lru_lock
, flags
);
61 lruvec
= mem_cgroup_page_lruvec(page
, zone
);
62 VM_BUG_ON_PAGE(!PageLRU(page
), page
);
64 del_page_from_lru_list(page
, lruvec
, page_off_lru(page
));
65 spin_unlock_irqrestore(&zone
->lru_lock
, flags
);
67 mem_cgroup_uncharge(page
);
70 static void __put_single_page(struct page
*page
)
72 __page_cache_release(page
);
73 free_hot_cold_page(page
, false);
76 static void __put_compound_page(struct page
*page
)
78 compound_page_dtor
*dtor
;
81 * __page_cache_release() is supposed to be called for thp, not for
82 * hugetlb. This is because hugetlb page does never have PageLRU set
83 * (it's never listed to any LRU lists) and no memcg routines should
84 * be called for hugetlb (it has a separate hugetlb_cgroup.)
87 __page_cache_release(page
);
88 dtor
= get_compound_page_dtor(page
);
92 void __put_page(struct page
*page
)
94 if (unlikely(PageCompound(page
)))
95 __put_compound_page(page
);
97 __put_single_page(page
);
99 EXPORT_SYMBOL(__put_page
);
102 * put_pages_list() - release a list of pages
103 * @pages: list of pages threaded on page->lru
105 * Release a list of pages which are strung together on page.lru. Currently
106 * used by read_cache_pages() and related error recovery code.
108 void put_pages_list(struct list_head
*pages
)
110 while (!list_empty(pages
)) {
113 victim
= list_entry(pages
->prev
, struct page
, lru
);
114 list_del(&victim
->lru
);
115 page_cache_release(victim
);
118 EXPORT_SYMBOL(put_pages_list
);
121 * get_kernel_pages() - pin kernel pages in memory
122 * @kiov: An array of struct kvec structures
123 * @nr_segs: number of segments to pin
124 * @write: pinning for read/write, currently ignored
125 * @pages: array that receives pointers to the pages pinned.
126 * Should be at least nr_segs long.
128 * Returns number of pages pinned. This may be fewer than the number
129 * requested. If nr_pages is 0 or negative, returns 0. If no pages
130 * were pinned, returns -errno. Each page returned must be released
131 * with a put_page() call when it is finished with.
133 int get_kernel_pages(const struct kvec
*kiov
, int nr_segs
, int write
,
138 for (seg
= 0; seg
< nr_segs
; seg
++) {
139 if (WARN_ON(kiov
[seg
].iov_len
!= PAGE_SIZE
))
142 pages
[seg
] = kmap_to_page(kiov
[seg
].iov_base
);
143 page_cache_get(pages
[seg
]);
148 EXPORT_SYMBOL_GPL(get_kernel_pages
);
151 * get_kernel_page() - pin a kernel page in memory
152 * @start: starting kernel address
153 * @write: pinning for read/write, currently ignored
154 * @pages: array that receives pointer to the page pinned.
155 * Must be at least nr_segs long.
157 * Returns 1 if page is pinned. If the page was not pinned, returns
158 * -errno. The page returned must be released with a put_page() call
159 * when it is finished with.
161 int get_kernel_page(unsigned long start
, int write
, struct page
**pages
)
163 const struct kvec kiov
= {
164 .iov_base
= (void *)start
,
168 return get_kernel_pages(&kiov
, 1, write
, pages
);
170 EXPORT_SYMBOL_GPL(get_kernel_page
);
172 static void pagevec_lru_move_fn(struct pagevec
*pvec
,
173 void (*move_fn
)(struct page
*page
, struct lruvec
*lruvec
, void *arg
),
177 struct zone
*zone
= NULL
;
178 struct lruvec
*lruvec
;
179 unsigned long flags
= 0;
181 for (i
= 0; i
< pagevec_count(pvec
); i
++) {
182 struct page
*page
= pvec
->pages
[i
];
183 struct zone
*pagezone
= page_zone(page
);
185 if (pagezone
!= zone
) {
187 spin_unlock_irqrestore(&zone
->lru_lock
, flags
);
189 spin_lock_irqsave(&zone
->lru_lock
, flags
);
192 lruvec
= mem_cgroup_page_lruvec(page
, zone
);
193 (*move_fn
)(page
, lruvec
, arg
);
196 spin_unlock_irqrestore(&zone
->lru_lock
, flags
);
197 release_pages(pvec
->pages
, pvec
->nr
, pvec
->cold
);
198 pagevec_reinit(pvec
);
201 static void pagevec_move_tail_fn(struct page
*page
, struct lruvec
*lruvec
,
206 if (PageLRU(page
) && !PageActive(page
) && !PageUnevictable(page
)) {
207 enum lru_list lru
= page_lru_base_type(page
);
208 list_move_tail(&page
->lru
, &lruvec
->lists
[lru
]);
214 * pagevec_move_tail() must be called with IRQ disabled.
215 * Otherwise this may cause nasty races.
217 static void pagevec_move_tail(struct pagevec
*pvec
)
221 pagevec_lru_move_fn(pvec
, pagevec_move_tail_fn
, &pgmoved
);
222 __count_vm_events(PGROTATED
, pgmoved
);
226 * Writeback is about to end against a page which has been marked for immediate
227 * reclaim. If it still appears to be reclaimable, move it to the tail of the
230 void rotate_reclaimable_page(struct page
*page
)
232 if (!PageLocked(page
) && !PageDirty(page
) && !PageActive(page
) &&
233 !PageUnevictable(page
) && PageLRU(page
)) {
234 struct pagevec
*pvec
;
237 page_cache_get(page
);
238 local_irq_save(flags
);
239 pvec
= this_cpu_ptr(&lru_rotate_pvecs
);
240 if (!pagevec_add(pvec
, page
))
241 pagevec_move_tail(pvec
);
242 local_irq_restore(flags
);
246 static void update_page_reclaim_stat(struct lruvec
*lruvec
,
247 int file
, int rotated
)
249 struct zone_reclaim_stat
*reclaim_stat
= &lruvec
->reclaim_stat
;
251 reclaim_stat
->recent_scanned
[file
]++;
253 reclaim_stat
->recent_rotated
[file
]++;
256 static void __activate_page(struct page
*page
, struct lruvec
*lruvec
,
259 if (PageLRU(page
) && !PageActive(page
) && !PageUnevictable(page
)) {
260 int file
= page_is_file_cache(page
);
261 int lru
= page_lru_base_type(page
);
263 del_page_from_lru_list(page
, lruvec
, lru
);
266 add_page_to_lru_list(page
, lruvec
, lru
);
267 trace_mm_lru_activate(page
);
269 __count_vm_event(PGACTIVATE
);
270 update_page_reclaim_stat(lruvec
, file
, 1);
275 static DEFINE_PER_CPU(struct pagevec
, activate_page_pvecs
);
277 static void activate_page_drain(int cpu
)
279 struct pagevec
*pvec
= &per_cpu(activate_page_pvecs
, cpu
);
281 if (pagevec_count(pvec
))
282 pagevec_lru_move_fn(pvec
, __activate_page
, NULL
);
285 static bool need_activate_page_drain(int cpu
)
287 return pagevec_count(&per_cpu(activate_page_pvecs
, cpu
)) != 0;
290 void activate_page(struct page
*page
)
292 if (PageLRU(page
) && !PageActive(page
) && !PageUnevictable(page
)) {
293 struct pagevec
*pvec
= &get_cpu_var(activate_page_pvecs
);
295 page_cache_get(page
);
296 if (!pagevec_add(pvec
, page
))
297 pagevec_lru_move_fn(pvec
, __activate_page
, NULL
);
298 put_cpu_var(activate_page_pvecs
);
303 static inline void activate_page_drain(int cpu
)
307 static bool need_activate_page_drain(int cpu
)
312 void activate_page(struct page
*page
)
314 struct zone
*zone
= page_zone(page
);
316 spin_lock_irq(&zone
->lru_lock
);
317 __activate_page(page
, mem_cgroup_page_lruvec(page
, zone
), NULL
);
318 spin_unlock_irq(&zone
->lru_lock
);
322 static void __lru_cache_activate_page(struct page
*page
)
324 struct pagevec
*pvec
= &get_cpu_var(lru_add_pvec
);
328 * Search backwards on the optimistic assumption that the page being
329 * activated has just been added to this pagevec. Note that only
330 * the local pagevec is examined as a !PageLRU page could be in the
331 * process of being released, reclaimed, migrated or on a remote
332 * pagevec that is currently being drained. Furthermore, marking
333 * a remote pagevec's page PageActive potentially hits a race where
334 * a page is marked PageActive just after it is added to the inactive
335 * list causing accounting errors and BUG_ON checks to trigger.
337 for (i
= pagevec_count(pvec
) - 1; i
>= 0; i
--) {
338 struct page
*pagevec_page
= pvec
->pages
[i
];
340 if (pagevec_page
== page
) {
346 put_cpu_var(lru_add_pvec
);
350 * Mark a page as having seen activity.
352 * inactive,unreferenced -> inactive,referenced
353 * inactive,referenced -> active,unreferenced
354 * active,unreferenced -> active,referenced
356 * When a newly allocated page is not yet visible, so safe for non-atomic ops,
357 * __SetPageReferenced(page) may be substituted for mark_page_accessed(page).
359 void mark_page_accessed(struct page
*page
)
361 if (!PageActive(page
) && !PageUnevictable(page
) &&
362 PageReferenced(page
)) {
365 * If the page is on the LRU, queue it for activation via
366 * activate_page_pvecs. Otherwise, assume the page is on a
367 * pagevec, mark it active and it'll be moved to the active
368 * LRU on the next drain.
373 __lru_cache_activate_page(page
);
374 ClearPageReferenced(page
);
375 if (page_is_file_cache(page
))
376 workingset_activation(page
);
377 } else if (!PageReferenced(page
)) {
378 SetPageReferenced(page
);
380 if (page_is_idle(page
))
381 clear_page_idle(page
);
383 EXPORT_SYMBOL(mark_page_accessed
);
385 static void __lru_cache_add(struct page
*page
)
387 struct pagevec
*pvec
= &get_cpu_var(lru_add_pvec
);
389 page_cache_get(page
);
390 if (!pagevec_space(pvec
))
391 __pagevec_lru_add(pvec
);
392 pagevec_add(pvec
, page
);
393 put_cpu_var(lru_add_pvec
);
397 * lru_cache_add: add a page to the page lists
398 * @page: the page to add
400 void lru_cache_add_anon(struct page
*page
)
402 if (PageActive(page
))
403 ClearPageActive(page
);
404 __lru_cache_add(page
);
407 void lru_cache_add_file(struct page
*page
)
409 if (PageActive(page
))
410 ClearPageActive(page
);
411 __lru_cache_add(page
);
413 EXPORT_SYMBOL(lru_cache_add_file
);
416 * lru_cache_add - add a page to a page list
417 * @page: the page to be added to the LRU.
419 * Queue the page for addition to the LRU via pagevec. The decision on whether
420 * to add the page to the [in]active [file|anon] list is deferred until the
421 * pagevec is drained. This gives a chance for the caller of lru_cache_add()
422 * have the page added to the active list using mark_page_accessed().
424 void lru_cache_add(struct page
*page
)
426 VM_BUG_ON_PAGE(PageActive(page
) && PageUnevictable(page
), page
);
427 VM_BUG_ON_PAGE(PageLRU(page
), page
);
428 __lru_cache_add(page
);
432 * add_page_to_unevictable_list - add a page to the unevictable list
433 * @page: the page to be added to the unevictable list
435 * Add page directly to its zone's unevictable list. To avoid races with
436 * tasks that might be making the page evictable, through eg. munlock,
437 * munmap or exit, while it's not on the lru, we want to add the page
438 * while it's locked or otherwise "invisible" to other tasks. This is
439 * difficult to do when using the pagevec cache, so bypass that.
441 void add_page_to_unevictable_list(struct page
*page
)
443 struct zone
*zone
= page_zone(page
);
444 struct lruvec
*lruvec
;
446 spin_lock_irq(&zone
->lru_lock
);
447 lruvec
= mem_cgroup_page_lruvec(page
, zone
);
448 ClearPageActive(page
);
449 SetPageUnevictable(page
);
451 add_page_to_lru_list(page
, lruvec
, LRU_UNEVICTABLE
);
452 spin_unlock_irq(&zone
->lru_lock
);
456 * lru_cache_add_active_or_unevictable
457 * @page: the page to be added to LRU
458 * @vma: vma in which page is mapped for determining reclaimability
460 * Place @page on the active or unevictable LRU list, depending on its
461 * evictability. Note that if the page is not evictable, it goes
462 * directly back onto it's zone's unevictable list, it does NOT use a
465 void lru_cache_add_active_or_unevictable(struct page
*page
,
466 struct vm_area_struct
*vma
)
468 VM_BUG_ON_PAGE(PageLRU(page
), page
);
470 if (likely((vma
->vm_flags
& (VM_LOCKED
| VM_SPECIAL
)) != VM_LOCKED
)) {
476 if (!TestSetPageMlocked(page
)) {
478 * We use the irq-unsafe __mod_zone_page_stat because this
479 * counter is not modified from interrupt context, and the pte
480 * lock is held(spinlock), which implies preemption disabled.
482 __mod_zone_page_state(page_zone(page
), NR_MLOCK
,
483 hpage_nr_pages(page
));
484 count_vm_event(UNEVICTABLE_PGMLOCKED
);
486 add_page_to_unevictable_list(page
);
490 * If the page can not be invalidated, it is moved to the
491 * inactive list to speed up its reclaim. It is moved to the
492 * head of the list, rather than the tail, to give the flusher
493 * threads some time to write it out, as this is much more
494 * effective than the single-page writeout from reclaim.
496 * If the page isn't page_mapped and dirty/writeback, the page
497 * could reclaim asap using PG_reclaim.
499 * 1. active, mapped page -> none
500 * 2. active, dirty/writeback page -> inactive, head, PG_reclaim
501 * 3. inactive, mapped page -> none
502 * 4. inactive, dirty/writeback page -> inactive, head, PG_reclaim
503 * 5. inactive, clean -> inactive, tail
506 * In 4, why it moves inactive's head, the VM expects the page would
507 * be write it out by flusher threads as this is much more effective
508 * than the single-page writeout from reclaim.
510 static void lru_deactivate_file_fn(struct page
*page
, struct lruvec
*lruvec
,
519 if (PageUnevictable(page
))
522 /* Some processes are using the page */
523 if (page_mapped(page
))
526 active
= PageActive(page
);
527 file
= page_is_file_cache(page
);
528 lru
= page_lru_base_type(page
);
530 del_page_from_lru_list(page
, lruvec
, lru
+ active
);
531 ClearPageActive(page
);
532 ClearPageReferenced(page
);
533 add_page_to_lru_list(page
, lruvec
, lru
);
535 if (PageWriteback(page
) || PageDirty(page
)) {
537 * PG_reclaim could be raced with end_page_writeback
538 * It can make readahead confusing. But race window
539 * is _really_ small and it's non-critical problem.
541 SetPageReclaim(page
);
544 * The page's writeback ends up during pagevec
545 * We moves tha page into tail of inactive.
547 list_move_tail(&page
->lru
, &lruvec
->lists
[lru
]);
548 __count_vm_event(PGROTATED
);
552 __count_vm_event(PGDEACTIVATE
);
553 update_page_reclaim_stat(lruvec
, file
, 0);
557 * Drain pages out of the cpu's pagevecs.
558 * Either "cpu" is the current CPU, and preemption has already been
559 * disabled; or "cpu" is being hot-unplugged, and is already dead.
561 void lru_add_drain_cpu(int cpu
)
563 struct pagevec
*pvec
= &per_cpu(lru_add_pvec
, cpu
);
565 if (pagevec_count(pvec
))
566 __pagevec_lru_add(pvec
);
568 pvec
= &per_cpu(lru_rotate_pvecs
, cpu
);
569 if (pagevec_count(pvec
)) {
572 /* No harm done if a racing interrupt already did this */
573 local_irq_save(flags
);
574 pagevec_move_tail(pvec
);
575 local_irq_restore(flags
);
578 pvec
= &per_cpu(lru_deactivate_file_pvecs
, cpu
);
579 if (pagevec_count(pvec
))
580 pagevec_lru_move_fn(pvec
, lru_deactivate_file_fn
, NULL
);
582 activate_page_drain(cpu
);
586 * deactivate_file_page - forcefully deactivate a file page
587 * @page: page to deactivate
589 * This function hints the VM that @page is a good reclaim candidate,
590 * for example if its invalidation fails due to the page being dirty
591 * or under writeback.
593 void deactivate_file_page(struct page
*page
)
596 * In a workload with many unevictable page such as mprotect,
597 * unevictable page deactivation for accelerating reclaim is pointless.
599 if (PageUnevictable(page
))
602 if (likely(get_page_unless_zero(page
))) {
603 struct pagevec
*pvec
= &get_cpu_var(lru_deactivate_file_pvecs
);
605 if (!pagevec_add(pvec
, page
))
606 pagevec_lru_move_fn(pvec
, lru_deactivate_file_fn
, NULL
);
607 put_cpu_var(lru_deactivate_file_pvecs
);
611 void lru_add_drain(void)
613 lru_add_drain_cpu(get_cpu());
617 static void lru_add_drain_per_cpu(struct work_struct
*dummy
)
622 static DEFINE_PER_CPU(struct work_struct
, lru_add_drain_work
);
624 void lru_add_drain_all(void)
626 static DEFINE_MUTEX(lock
);
627 static struct cpumask has_work
;
632 cpumask_clear(&has_work
);
634 for_each_online_cpu(cpu
) {
635 struct work_struct
*work
= &per_cpu(lru_add_drain_work
, cpu
);
637 if (pagevec_count(&per_cpu(lru_add_pvec
, cpu
)) ||
638 pagevec_count(&per_cpu(lru_rotate_pvecs
, cpu
)) ||
639 pagevec_count(&per_cpu(lru_deactivate_file_pvecs
, cpu
)) ||
640 need_activate_page_drain(cpu
)) {
641 INIT_WORK(work
, lru_add_drain_per_cpu
);
642 schedule_work_on(cpu
, work
);
643 cpumask_set_cpu(cpu
, &has_work
);
647 for_each_cpu(cpu
, &has_work
)
648 flush_work(&per_cpu(lru_add_drain_work
, cpu
));
655 * release_pages - batched page_cache_release()
656 * @pages: array of pages to release
657 * @nr: number of pages
658 * @cold: whether the pages are cache cold
660 * Decrement the reference count on all the pages in @pages. If it
661 * fell to zero, remove the page from the LRU and free it.
663 void release_pages(struct page
**pages
, int nr
, bool cold
)
666 LIST_HEAD(pages_to_free
);
667 struct zone
*zone
= NULL
;
668 struct lruvec
*lruvec
;
669 unsigned long uninitialized_var(flags
);
670 unsigned int uninitialized_var(lock_batch
);
672 for (i
= 0; i
< nr
; i
++) {
673 struct page
*page
= pages
[i
];
676 * Make sure the IRQ-safe lock-holding time does not get
677 * excessive with a continuous string of pages from the
678 * same zone. The lock is held only if zone != NULL.
680 if (zone
&& ++lock_batch
== SWAP_CLUSTER_MAX
) {
681 spin_unlock_irqrestore(&zone
->lru_lock
, flags
);
685 page
= compound_head(page
);
686 if (!put_page_testzero(page
))
689 if (PageCompound(page
)) {
691 spin_unlock_irqrestore(&zone
->lru_lock
, flags
);
694 __put_compound_page(page
);
699 struct zone
*pagezone
= page_zone(page
);
701 if (pagezone
!= zone
) {
703 spin_unlock_irqrestore(&zone
->lru_lock
,
707 spin_lock_irqsave(&zone
->lru_lock
, flags
);
710 lruvec
= mem_cgroup_page_lruvec(page
, zone
);
711 VM_BUG_ON_PAGE(!PageLRU(page
), page
);
712 __ClearPageLRU(page
);
713 del_page_from_lru_list(page
, lruvec
, page_off_lru(page
));
716 /* Clear Active bit in case of parallel mark_page_accessed */
717 __ClearPageActive(page
);
719 list_add(&page
->lru
, &pages_to_free
);
722 spin_unlock_irqrestore(&zone
->lru_lock
, flags
);
724 mem_cgroup_uncharge_list(&pages_to_free
);
725 free_hot_cold_page_list(&pages_to_free
, cold
);
727 EXPORT_SYMBOL(release_pages
);
730 * The pages which we're about to release may be in the deferred lru-addition
731 * queues. That would prevent them from really being freed right now. That's
732 * OK from a correctness point of view but is inefficient - those pages may be
733 * cache-warm and we want to give them back to the page allocator ASAP.
735 * So __pagevec_release() will drain those queues here. __pagevec_lru_add()
736 * and __pagevec_lru_add_active() call release_pages() directly to avoid
739 void __pagevec_release(struct pagevec
*pvec
)
742 release_pages(pvec
->pages
, pagevec_count(pvec
), pvec
->cold
);
743 pagevec_reinit(pvec
);
745 EXPORT_SYMBOL(__pagevec_release
);
747 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
748 /* used by __split_huge_page_refcount() */
749 void lru_add_page_tail(struct page
*page
, struct page
*page_tail
,
750 struct lruvec
*lruvec
, struct list_head
*list
)
754 VM_BUG_ON_PAGE(!PageHead(page
), page
);
755 VM_BUG_ON_PAGE(PageCompound(page_tail
), page
);
756 VM_BUG_ON_PAGE(PageLRU(page_tail
), page
);
757 VM_BUG_ON(NR_CPUS
!= 1 &&
758 !spin_is_locked(&lruvec_zone(lruvec
)->lru_lock
));
761 SetPageLRU(page_tail
);
763 if (likely(PageLRU(page
)))
764 list_add_tail(&page_tail
->lru
, &page
->lru
);
766 /* page reclaim is reclaiming a huge page */
768 list_add_tail(&page_tail
->lru
, list
);
770 struct list_head
*list_head
;
772 * Head page has not yet been counted, as an hpage,
773 * so we must account for each subpage individually.
775 * Use the standard add function to put page_tail on the list,
776 * but then correct its position so they all end up in order.
778 add_page_to_lru_list(page_tail
, lruvec
, page_lru(page_tail
));
779 list_head
= page_tail
->lru
.prev
;
780 list_move_tail(&page_tail
->lru
, list_head
);
783 if (!PageUnevictable(page
))
784 update_page_reclaim_stat(lruvec
, file
, PageActive(page_tail
));
786 #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
788 static void __pagevec_lru_add_fn(struct page
*page
, struct lruvec
*lruvec
,
791 int file
= page_is_file_cache(page
);
792 int active
= PageActive(page
);
793 enum lru_list lru
= page_lru(page
);
795 VM_BUG_ON_PAGE(PageLRU(page
), page
);
798 add_page_to_lru_list(page
, lruvec
, lru
);
799 update_page_reclaim_stat(lruvec
, file
, active
);
800 trace_mm_lru_insertion(page
, lru
);
804 * Add the passed pages to the LRU, then drop the caller's refcount
805 * on them. Reinitialises the caller's pagevec.
807 void __pagevec_lru_add(struct pagevec
*pvec
)
809 pagevec_lru_move_fn(pvec
, __pagevec_lru_add_fn
, NULL
);
811 EXPORT_SYMBOL(__pagevec_lru_add
);
814 * pagevec_lookup_entries - gang pagecache lookup
815 * @pvec: Where the resulting entries are placed
816 * @mapping: The address_space to search
817 * @start: The starting entry index
818 * @nr_entries: The maximum number of entries
819 * @indices: The cache indices corresponding to the entries in @pvec
821 * pagevec_lookup_entries() will search for and return a group of up
822 * to @nr_entries pages and shadow entries in the mapping. All
823 * entries are placed in @pvec. pagevec_lookup_entries() takes a
824 * reference against actual pages in @pvec.
826 * The search returns a group of mapping-contiguous entries with
827 * ascending indexes. There may be holes in the indices due to
828 * not-present entries.
830 * pagevec_lookup_entries() returns the number of entries which were
833 unsigned pagevec_lookup_entries(struct pagevec
*pvec
,
834 struct address_space
*mapping
,
835 pgoff_t start
, unsigned nr_pages
,
838 pvec
->nr
= find_get_entries(mapping
, start
, nr_pages
,
839 pvec
->pages
, indices
);
840 return pagevec_count(pvec
);
844 * pagevec_remove_exceptionals - pagevec exceptionals pruning
845 * @pvec: The pagevec to prune
847 * pagevec_lookup_entries() fills both pages and exceptional radix
848 * tree entries into the pagevec. This function prunes all
849 * exceptionals from @pvec without leaving holes, so that it can be
850 * passed on to page-only pagevec operations.
852 void pagevec_remove_exceptionals(struct pagevec
*pvec
)
856 for (i
= 0, j
= 0; i
< pagevec_count(pvec
); i
++) {
857 struct page
*page
= pvec
->pages
[i
];
858 if (!radix_tree_exceptional_entry(page
))
859 pvec
->pages
[j
++] = page
;
865 * pagevec_lookup - gang pagecache lookup
866 * @pvec: Where the resulting pages are placed
867 * @mapping: The address_space to search
868 * @start: The starting page index
869 * @nr_pages: The maximum number of pages
871 * pagevec_lookup() will search for and return a group of up to @nr_pages pages
872 * in the mapping. The pages are placed in @pvec. pagevec_lookup() takes a
873 * reference against the pages in @pvec.
875 * The search returns a group of mapping-contiguous pages with ascending
876 * indexes. There may be holes in the indices due to not-present pages.
878 * pagevec_lookup() returns the number of pages which were found.
880 unsigned pagevec_lookup(struct pagevec
*pvec
, struct address_space
*mapping
,
881 pgoff_t start
, unsigned nr_pages
)
883 pvec
->nr
= find_get_pages(mapping
, start
, nr_pages
, pvec
->pages
);
884 return pagevec_count(pvec
);
886 EXPORT_SYMBOL(pagevec_lookup
);
888 unsigned pagevec_lookup_tag(struct pagevec
*pvec
, struct address_space
*mapping
,
889 pgoff_t
*index
, int tag
, unsigned nr_pages
)
891 pvec
->nr
= find_get_pages_tag(mapping
, index
, tag
,
892 nr_pages
, pvec
->pages
);
893 return pagevec_count(pvec
);
895 EXPORT_SYMBOL(pagevec_lookup_tag
);
898 * Perform any setup for the swap system
900 void __init
swap_setup(void)
902 unsigned long megs
= totalram_pages
>> (20 - PAGE_SHIFT
);
906 for (i
= 0; i
< MAX_SWAPFILES
; i
++)
907 spin_lock_init(&swapper_spaces
[i
].tree_lock
);
910 /* Use a smaller cluster for small-memory machines */
916 * Right now other parts of the system means that we
917 * _really_ don't want to cluster much more