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 page = compound_head(page);
362 if (!PageActive(page) && !PageUnevictable(page) &&
363 PageReferenced(page)) {
366 * If the page is on the LRU, queue it for activation via
367 * activate_page_pvecs. Otherwise, assume the page is on a
368 * pagevec, mark it active and it'll be moved to the active
369 * LRU on the next drain.
374 __lru_cache_activate_page(page);
375 ClearPageReferenced(page);
376 if (page_is_file_cache(page))
377 workingset_activation(page);
378 } else if (!PageReferenced(page)) {
379 SetPageReferenced(page);
381 if (page_is_idle(page))
382 clear_page_idle(page);
384 EXPORT_SYMBOL(mark_page_accessed);
386 static void __lru_cache_add(struct page *page)
388 struct pagevec *pvec = &get_cpu_var(lru_add_pvec);
390 page_cache_get(page);
391 if (!pagevec_space(pvec))
392 __pagevec_lru_add(pvec);
393 pagevec_add(pvec, page);
394 put_cpu_var(lru_add_pvec);
398 * lru_cache_add: add a page to the page lists
399 * @page: the page to add
401 void lru_cache_add_anon(struct page *page)
403 if (PageActive(page))
404 ClearPageActive(page);
405 __lru_cache_add(page);
408 void lru_cache_add_file(struct page *page)
410 if (PageActive(page))
411 ClearPageActive(page);
412 __lru_cache_add(page);
414 EXPORT_SYMBOL(lru_cache_add_file);
417 * lru_cache_add - add a page to a page list
418 * @page: the page to be added to the LRU.
420 * Queue the page for addition to the LRU via pagevec. The decision on whether
421 * to add the page to the [in]active [file|anon] list is deferred until the
422 * pagevec is drained. This gives a chance for the caller of lru_cache_add()
423 * have the page added to the active list using mark_page_accessed().
425 void lru_cache_add(struct page *page)
427 VM_BUG_ON_PAGE(PageActive(page) && PageUnevictable(page), page);
428 VM_BUG_ON_PAGE(PageLRU(page), page);
429 __lru_cache_add(page);
433 * add_page_to_unevictable_list - add a page to the unevictable list
434 * @page: the page to be added to the unevictable list
436 * Add page directly to its zone's unevictable list. To avoid races with
437 * tasks that might be making the page evictable, through eg. munlock,
438 * munmap or exit, while it's not on the lru, we want to add the page
439 * while it's locked or otherwise "invisible" to other tasks. This is
440 * difficult to do when using the pagevec cache, so bypass that.
442 void add_page_to_unevictable_list(struct page *page)
444 struct zone *zone = page_zone(page);
445 struct lruvec *lruvec;
447 spin_lock_irq(&zone->lru_lock);
448 lruvec = mem_cgroup_page_lruvec(page, zone);
449 ClearPageActive(page);
450 SetPageUnevictable(page);
452 add_page_to_lru_list(page, lruvec, LRU_UNEVICTABLE);
453 spin_unlock_irq(&zone->lru_lock);
457 * lru_cache_add_active_or_unevictable
458 * @page: the page to be added to LRU
459 * @vma: vma in which page is mapped for determining reclaimability
461 * Place @page on the active or unevictable LRU list, depending on its
462 * evictability. Note that if the page is not evictable, it goes
463 * directly back onto it's zone's unevictable list, it does NOT use a
466 void lru_cache_add_active_or_unevictable(struct page *page,
467 struct vm_area_struct *vma)
469 VM_BUG_ON_PAGE(PageLRU(page), page);
471 if (likely((vma->vm_flags & (VM_LOCKED | VM_SPECIAL)) != VM_LOCKED)) {
477 if (!TestSetPageMlocked(page)) {
479 * We use the irq-unsafe __mod_zone_page_stat because this
480 * counter is not modified from interrupt context, and the pte
481 * lock is held(spinlock), which implies preemption disabled.
483 __mod_zone_page_state(page_zone(page), NR_MLOCK,
484 hpage_nr_pages(page));
485 count_vm_event(UNEVICTABLE_PGMLOCKED);
487 add_page_to_unevictable_list(page);
491 * If the page can not be invalidated, it is moved to the
492 * inactive list to speed up its reclaim. It is moved to the
493 * head of the list, rather than the tail, to give the flusher
494 * threads some time to write it out, as this is much more
495 * effective than the single-page writeout from reclaim.
497 * If the page isn't page_mapped and dirty/writeback, the page
498 * could reclaim asap using PG_reclaim.
500 * 1. active, mapped page -> none
501 * 2. active, dirty/writeback page -> inactive, head, PG_reclaim
502 * 3. inactive, mapped page -> none
503 * 4. inactive, dirty/writeback page -> inactive, head, PG_reclaim
504 * 5. inactive, clean -> inactive, tail
507 * In 4, why it moves inactive's head, the VM expects the page would
508 * be write it out by flusher threads as this is much more effective
509 * than the single-page writeout from reclaim.
511 static void lru_deactivate_file_fn(struct page *page, struct lruvec *lruvec,
520 if (PageUnevictable(page))
523 /* Some processes are using the page */
524 if (page_mapped(page))
527 active = PageActive(page);
528 file = page_is_file_cache(page);
529 lru = page_lru_base_type(page);
531 del_page_from_lru_list(page, lruvec, lru + active);
532 ClearPageActive(page);
533 ClearPageReferenced(page);
534 add_page_to_lru_list(page, lruvec, lru);
536 if (PageWriteback(page) || PageDirty(page)) {
538 * PG_reclaim could be raced with end_page_writeback
539 * It can make readahead confusing. But race window
540 * is _really_ small and it's non-critical problem.
542 SetPageReclaim(page);
545 * The page's writeback ends up during pagevec
546 * We moves tha page into tail of inactive.
548 list_move_tail(&page->lru, &lruvec->lists[lru]);
549 __count_vm_event(PGROTATED);
553 __count_vm_event(PGDEACTIVATE);
554 update_page_reclaim_stat(lruvec, file, 0);
558 * Drain pages out of the cpu's pagevecs.
559 * Either "cpu" is the current CPU, and preemption has already been
560 * disabled; or "cpu" is being hot-unplugged, and is already dead.
562 void lru_add_drain_cpu(int cpu)
564 struct pagevec *pvec = &per_cpu(lru_add_pvec, cpu);
566 if (pagevec_count(pvec))
567 __pagevec_lru_add(pvec);
569 pvec = &per_cpu(lru_rotate_pvecs, cpu);
570 if (pagevec_count(pvec)) {
573 /* No harm done if a racing interrupt already did this */
574 local_irq_save(flags);
575 pagevec_move_tail(pvec);
576 local_irq_restore(flags);
579 pvec = &per_cpu(lru_deactivate_file_pvecs, cpu);
580 if (pagevec_count(pvec))
581 pagevec_lru_move_fn(pvec, lru_deactivate_file_fn, NULL);
583 activate_page_drain(cpu);
587 * deactivate_file_page - forcefully deactivate a file page
588 * @page: page to deactivate
590 * This function hints the VM that @page is a good reclaim candidate,
591 * for example if its invalidation fails due to the page being dirty
592 * or under writeback.
594 void deactivate_file_page(struct page *page)
597 * In a workload with many unevictable page such as mprotect,
598 * unevictable page deactivation for accelerating reclaim is pointless.
600 if (PageUnevictable(page))
603 if (likely(get_page_unless_zero(page))) {
604 struct pagevec *pvec = &get_cpu_var(lru_deactivate_file_pvecs);
606 if (!pagevec_add(pvec, page))
607 pagevec_lru_move_fn(pvec, lru_deactivate_file_fn, NULL);
608 put_cpu_var(lru_deactivate_file_pvecs);
612 void lru_add_drain(void)
614 lru_add_drain_cpu(get_cpu());
618 static void lru_add_drain_per_cpu(struct work_struct *dummy)
623 static DEFINE_PER_CPU(struct work_struct, lru_add_drain_work);
625 void lru_add_drain_all(void)
627 static DEFINE_MUTEX(lock);
628 static struct cpumask has_work;
633 cpumask_clear(&has_work);
635 for_each_online_cpu(cpu) {
636 struct work_struct *work = &per_cpu(lru_add_drain_work, cpu);
638 if (pagevec_count(&per_cpu(lru_add_pvec, cpu)) ||
639 pagevec_count(&per_cpu(lru_rotate_pvecs, cpu)) ||
640 pagevec_count(&per_cpu(lru_deactivate_file_pvecs, cpu)) ||
641 need_activate_page_drain(cpu)) {
642 INIT_WORK(work, lru_add_drain_per_cpu);
643 schedule_work_on(cpu, work);
644 cpumask_set_cpu(cpu, &has_work);
648 for_each_cpu(cpu, &has_work)
649 flush_work(&per_cpu(lru_add_drain_work, cpu));
656 * release_pages - batched page_cache_release()
657 * @pages: array of pages to release
658 * @nr: number of pages
659 * @cold: whether the pages are cache cold
661 * Decrement the reference count on all the pages in @pages. If it
662 * fell to zero, remove the page from the LRU and free it.
664 void release_pages(struct page **pages, int nr, bool cold)
667 LIST_HEAD(pages_to_free);
668 struct zone *zone = NULL;
669 struct lruvec *lruvec;
670 unsigned long uninitialized_var(flags);
671 unsigned int uninitialized_var(lock_batch);
673 for (i = 0; i < nr; i++) {
674 struct page *page = pages[i];
677 * Make sure the IRQ-safe lock-holding time does not get
678 * excessive with a continuous string of pages from the
679 * same zone. The lock is held only if zone != NULL.
681 if (zone && ++lock_batch == SWAP_CLUSTER_MAX) {
682 spin_unlock_irqrestore(&zone->lru_lock, flags);
686 page = compound_head(page);
687 if (!put_page_testzero(page))
690 if (PageCompound(page)) {
692 spin_unlock_irqrestore(&zone->lru_lock, flags);
695 __put_compound_page(page);
700 struct zone *pagezone = page_zone(page);
702 if (pagezone != zone) {
704 spin_unlock_irqrestore(&zone->lru_lock,
708 spin_lock_irqsave(&zone->lru_lock, flags);
711 lruvec = mem_cgroup_page_lruvec(page, zone);
712 VM_BUG_ON_PAGE(!PageLRU(page), page);
713 __ClearPageLRU(page);
714 del_page_from_lru_list(page, lruvec, page_off_lru(page));
717 /* Clear Active bit in case of parallel mark_page_accessed */
718 __ClearPageActive(page);
720 list_add(&page->lru, &pages_to_free);
723 spin_unlock_irqrestore(&zone->lru_lock, flags);
725 mem_cgroup_uncharge_list(&pages_to_free);
726 free_hot_cold_page_list(&pages_to_free, cold);
728 EXPORT_SYMBOL(release_pages);
731 * The pages which we're about to release may be in the deferred lru-addition
732 * queues. That would prevent them from really being freed right now. That's
733 * OK from a correctness point of view but is inefficient - those pages may be
734 * cache-warm and we want to give them back to the page allocator ASAP.
736 * So __pagevec_release() will drain those queues here. __pagevec_lru_add()
737 * and __pagevec_lru_add_active() call release_pages() directly to avoid
740 void __pagevec_release(struct pagevec *pvec)
743 release_pages(pvec->pages, pagevec_count(pvec), pvec->cold);
744 pagevec_reinit(pvec);
746 EXPORT_SYMBOL(__pagevec_release);
748 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
749 /* used by __split_huge_page_refcount() */
750 void lru_add_page_tail(struct page *page, struct page *page_tail,
751 struct lruvec *lruvec, struct list_head *list)
755 VM_BUG_ON_PAGE(!PageHead(page), page);
756 VM_BUG_ON_PAGE(PageCompound(page_tail), page);
757 VM_BUG_ON_PAGE(PageLRU(page_tail), page);
758 VM_BUG_ON(NR_CPUS != 1 &&
759 !spin_is_locked(&lruvec_zone(lruvec)->lru_lock));
762 SetPageLRU(page_tail);
764 if (likely(PageLRU(page)))
765 list_add_tail(&page_tail->lru, &page->lru);
767 /* page reclaim is reclaiming a huge page */
769 list_add_tail(&page_tail->lru, list);
771 struct list_head *list_head;
773 * Head page has not yet been counted, as an hpage,
774 * so we must account for each subpage individually.
776 * Use the standard add function to put page_tail on the list,
777 * but then correct its position so they all end up in order.
779 add_page_to_lru_list(page_tail, lruvec, page_lru(page_tail));
780 list_head = page_tail->lru.prev;
781 list_move_tail(&page_tail->lru, list_head);
784 if (!PageUnevictable(page))
785 update_page_reclaim_stat(lruvec, file, PageActive(page_tail));
787 #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
789 static void __pagevec_lru_add_fn(struct page *page, struct lruvec *lruvec,
792 int file = page_is_file_cache(page);
793 int active = PageActive(page);
794 enum lru_list lru = page_lru(page);
796 VM_BUG_ON_PAGE(PageLRU(page), page);
799 add_page_to_lru_list(page, lruvec, lru);
800 update_page_reclaim_stat(lruvec, file, active);
801 trace_mm_lru_insertion(page, lru);
805 * Add the passed pages to the LRU, then drop the caller's refcount
806 * on them. Reinitialises the caller's pagevec.
808 void __pagevec_lru_add(struct pagevec *pvec)
810 pagevec_lru_move_fn(pvec, __pagevec_lru_add_fn, NULL);
812 EXPORT_SYMBOL(__pagevec_lru_add);
815 * pagevec_lookup_entries - gang pagecache lookup
816 * @pvec: Where the resulting entries are placed
817 * @mapping: The address_space to search
818 * @start: The starting entry index
819 * @nr_entries: The maximum number of entries
820 * @indices: The cache indices corresponding to the entries in @pvec
822 * pagevec_lookup_entries() will search for and return a group of up
823 * to @nr_entries pages and shadow entries in the mapping. All
824 * entries are placed in @pvec. pagevec_lookup_entries() takes a
825 * reference against actual pages in @pvec.
827 * The search returns a group of mapping-contiguous entries with
828 * ascending indexes. There may be holes in the indices due to
829 * not-present entries.
831 * pagevec_lookup_entries() returns the number of entries which were
834 unsigned pagevec_lookup_entries(struct pagevec *pvec,
835 struct address_space *mapping,
836 pgoff_t start, unsigned nr_pages,
839 pvec->nr = find_get_entries(mapping, start, nr_pages,
840 pvec->pages, indices);
841 return pagevec_count(pvec);
845 * pagevec_remove_exceptionals - pagevec exceptionals pruning
846 * @pvec: The pagevec to prune
848 * pagevec_lookup_entries() fills both pages and exceptional radix
849 * tree entries into the pagevec. This function prunes all
850 * exceptionals from @pvec without leaving holes, so that it can be
851 * passed on to page-only pagevec operations.
853 void pagevec_remove_exceptionals(struct pagevec *pvec)
857 for (i = 0, j = 0; i < pagevec_count(pvec); i++) {
858 struct page *page = pvec->pages[i];
859 if (!radix_tree_exceptional_entry(page))
860 pvec->pages[j++] = page;
866 * pagevec_lookup - gang pagecache lookup
867 * @pvec: Where the resulting pages are placed
868 * @mapping: The address_space to search
869 * @start: The starting page index
870 * @nr_pages: The maximum number of pages
872 * pagevec_lookup() will search for and return a group of up to @nr_pages pages
873 * in the mapping. The pages are placed in @pvec. pagevec_lookup() takes a
874 * reference against the pages in @pvec.
876 * The search returns a group of mapping-contiguous pages with ascending
877 * indexes. There may be holes in the indices due to not-present pages.
879 * pagevec_lookup() returns the number of pages which were found.
881 unsigned pagevec_lookup(struct pagevec *pvec, struct address_space *mapping,
882 pgoff_t start, unsigned nr_pages)
884 pvec->nr = find_get_pages(mapping, start, nr_pages, pvec->pages);
885 return pagevec_count(pvec);
887 EXPORT_SYMBOL(pagevec_lookup);
889 unsigned pagevec_lookup_tag(struct pagevec *pvec, struct address_space *mapping,
890 pgoff_t *index, int tag, unsigned nr_pages)
892 pvec->nr = find_get_pages_tag(mapping, index, tag,
893 nr_pages, pvec->pages);
894 return pagevec_count(pvec);
896 EXPORT_SYMBOL(pagevec_lookup_tag);
899 * Perform any setup for the swap system
901 void __init swap_setup(void)
903 unsigned long megs = totalram_pages >> (20 - PAGE_SHIFT);
907 for (i = 0; i < MAX_SWAPFILES; i++)
908 spin_lock_init(&swapper_spaces[i].tree_lock);
911 /* Use a smaller cluster for small-memory machines */
917 * Right now other parts of the system means that we
918 * _really_ don't want to cluster much more