1 /* SPDX-License-Identifier: GPL-2.0 */
18 #include <linux/pageblock-flags.h>
19 #include <linux/page-flags-layout.h>
22 #include <linux/page-flags.h>
26 /* Free memory management - zoned buddy allocator.  */
32 #define MAX_ORDER_NR_PAGES (1 << (MAX_ORDER - 1))
55 	 * The way to use it is to change migratetype of a range of
58 	 * is that a range of pageblocks must be aligned to
60 	 * a single pageblock.
92 #define MIGRATETYPE_MASK ((1UL << PB_migratetype_bits) - 1)
105 	return list_first_entry_or_null(&area->free_list[migratetype],  in get_page_from_free_area()
111 	return list_empty(&area->free_list[migratetype]);  in free_area_empty()
117  * Add a wild amount of padding here to ensure data fall into separate
119  * consumption is not a concern here.
202 	NR_KERNEL_MISC_RECLAIMABLE,	/* reclaimable non-slab kernel pages */
241 	 * Global and per-node slab counters track slab pages.  in vmstat_item_in_bytes()
245 	 * Per-memcg and per-lruvec counters track memory, consumed  in vmstat_item_in_bytes()
247 	 * byte-precise.  in vmstat_item_in_bytes()
293 					 * backed by a congested BDI
302 	 * These track the cost of reclaiming one LRU - file or anon -
308 	/* Non-resident age, driven by LRU movement */
352 #define NR_PCP_ORDER_MASK ((1<<NR_PCP_ORDER_WIDTH) - 1)
354 #define min_wmark_pages(z) (z->_watermark[WMARK_MIN] + z->watermark_boost)  argument
355 #define low_wmark_pages(z) (z->_watermark[WMARK_LOW] + z->watermark_boost)  argument
356 #define high_wmark_pages(z) (z->_watermark[WMARK_HIGH] + z->watermark_boost)  argument
357 #define wmark_pages(z, i) (z->_watermark[i] + z->watermark_boost)  argument
369 	/* Lists of pages, one per migrate type stored on the pcp-lists */
381 	 * on demand. Use a large type to avoid the overhead of
401 	 * DMA addressing constraints. This distinction is important as a 32bit
402 	 * DMA mask is assumed when ZONE_DMA32 is defined. Some 64-bit
420 	 * A memory area that is only addressable by the kernel through
433 	 * likely to succeed, and to locally limit unmovable allocations - e.g.,
436 	 * 1. Pinned pages: (long-term) pinning of movable pages might
438 	 *    pinning long-term pages in ZONE_MOVABLE. When pages are pinned and
443 	 *    to a different zone. When migration fails - pinning fails.
455 	 *    buddy (e.g., via XEN-balloon, Hyper-V balloon, virtio-mem). The
457 	 *    some cases (virtio-mem), such pages can be skipped during
461 	 *    of memory unplug in virtio-mem).
464 	 *    on different platforms may end up in a movable zone. ZERO_PAGE(0)
466 	 * 7. Memory-hotplug: when using memmap_on_memory and onlining the
469 	 *    self-stored in the range, but they are treated as movable when
491 	/* Read-mostly fields */
525 	 * Flags for a pageblock_nr_pages block. See pageblock-flags.h.
537 	 * 	spanned_pages = zone_end_pfn - zone_start_pfn;
541 	 *	present_pages = spanned_pages - absent_pages(pages in holes);
550 	 *	managed_pages = present_pages - reserved_pages;
557 	 * (present_pages - managed_pages). And managed_pages should be used
564 	 * It is a seqlock because it has to be read outside of zone->lock,
568 	 * The span_seq lock is declared along with zone->lock because it is
569 	 * frequently read in proximity to zone->lock.  It's good to
570 	 * give them a chance of being in the same cacheline.
574 	 * present_pages should get_online_mems() to get a stable value.
592 	 * of pageblock. Protected by zone->lock.
604 	/* Write-intensive fields used from the page allocator */
616 	/* Write-intensive fields used by compaction and vmstats. */
621 	 * when reading the number of free pages to avoid per-cpu counter
680 	return (unsigned long)atomic_long_read(&zone->managed_pages);  in zone_managed_pages()
686 	return zone->cma_pages;  in zone_cma_pages()
694 	return zone->zone_start_pfn + zone->spanned_pages;  in zone_end_pfn()
699 	return zone->zone_start_pfn <= pfn && pfn < zone_end_pfn(zone);  in zone_spans_pfn()
704 	return zone->initialized;  in zone_is_initialized()
709 	return zone->spanned_pages == 0;  in zone_is_empty()
713  * Return true if [start_pfn, start_pfn + nr_pages) range has a non-empty
722 	    start_pfn + nr_pages <= zone->zone_start_pfn)  in zone_intersects()
730  * go. A value of 12 for DEF_PRIORITY implies that we will scan 1/4096th of the
735 /* Maximum number of zones on a zonelist */
743 	 * restrict the allocations to a single node for __GFP_THISNODE.
751  * This struct contains information about a zone in a zonelist. It is stored
756 	int zone_idx;		/* zone_idx(zoneref->zone) */
760  * One allocation request operates on a zonelist. A zonelist
761  * is a list of zones, the first one is the 'goal' of the
767  * a struct zoneref are
769  * zonelist_zone()	- Return the struct zone * for an entry in _zonerefs
770  * zonelist_zone_idx()	- Return the index of the zone for an entry
771  * zonelist_node_idx()	- Return the index of the node for an entry
793  * On NUMA machines, each NUMA node would have a pg_data_t to describe
794  * it's memory layout. On UMA machines there is a single pglist_data which
797  * Memory statistics and page replacement data structures are maintained on a
798  * per-zone basis.
826 	 * Also synchronizes pgdat->first_deferred_pfn during deferred page
833 	 * Nests above zone->lock and zone->span_seqlock
859 	 * This is a per-node reserve of pages that are not available
872 	/* Write-intensive fields used by page reclaim */
900 	/* Per-node vmstats */
905 #define node_present_pages(nid)	(NODE_DATA(nid)->node_present_pages)
906 #define node_spanned_pages(nid)	(NODE_DATA(nid)->node_spanned_pages)
908 #define pgdat_page_nr(pgdat, pagenr)	((pgdat)->node_mem_map + (pagenr))
910 #define pgdat_page_nr(pgdat, pagenr)	pfn_to_page((pgdat)->node_start_pfn + (pagenr))
914 #define node_start_pfn(nid)	(NODE_DATA(nid)->node_start_pfn)
919 	return pgdat->node_start_pfn + pgdat->node_spanned_pages;  in pgdat_end_pfn()
924 	return !pgdat->node_start_pfn && !pgdat->node_spanned_pages;  in pgdat_is_empty()
932 bool __zone_watermark_ok(struct zone *z, unsigned int order, unsigned long mark,
935 bool zone_watermark_ok(struct zone *z, unsigned int order,
938 bool zone_watermark_ok_safe(struct zone *z, unsigned int order,
957 	return lruvec->pgdat;  in lruvec_pgdat()
972 #define zone_idx(zone)		((zone) - (zone)->zone_pgdat->node_zones)
987  * Returns true if a zone has pages managed by the buddy allocator.
997 /* Returns true if a zone has memory */
1000 	return zone->present_pages;  in populated_zone()
1006 	return zone->node;  in zone_to_nid()
1011 	zone->node = nid;  in zone_set_nid()
1035  * is_highmem - helper function to quickly check if a struct zone is a
1037  *              to ZONE_{DMA/NORMAL/HIGHMEM/etc} in general code to a minimum.
1039  * Return: 1 for a highmem zone, 0 otherwise
1092  * for_each_online_pgdat - helper macro to iterate over all online nodes
1093  * @pgdat: pointer to a pg_data_t variable
1100  * for_each_zone - helper macro to iterate over all memory zones
1107 	for (zone = (first_online_pgdat())->node_zones; \
1112 	for (zone = (first_online_pgdat())->node_zones; \
1121 	return zoneref->zone;  in zonelist_zone()
1126 	return zoneref->zone_idx;  in zonelist_zone_idx()
1131 	return zone_to_nid(zoneref->zone);  in zonelist_node_idx()
1134 struct zoneref *__next_zones_zonelist(struct zoneref *z,
1139 …xt_zones_zonelist - Returns the next zone at or below highest_zoneidx within the allowed nodemask …
1140  * @z: The cursor used as a starting point for the search
1144  * This function returns the next zone at or below a given zone index that is
1145  * within the allowed nodemask using a cursor as the starting point for the
1146  * search. The zoneref returned is a cursor that represents the current zone
1151  * nodemask using a cursor within a zonelist as a starting point
1153 static __always_inline struct zoneref *next_zones_zonelist(struct zoneref *z,  in next_zones_zonelist()  argument
1157 	if (likely(!nodes && zonelist_zone_idx(z) <= highest_zoneidx))  in next_zones_zonelist()
1158 		return z;  in next_zones_zonelist()
1159 	return __next_zones_zonelist(z, highest_zoneidx, nodes);  in next_zones_zonelist()
1163 …* first_zones_zonelist - Returns the first zone at or below highest_zoneidx within the allowed nod…
1164  * @zonelist: The zonelist to search for a suitable zone
1168  * This function returns the first zone at or below a given zone index that is
1169  * within the allowed nodemask. The zoneref returned is a cursor that can be
1173  * When no eligible zone is found, zoneref->zone is NULL (zoneref itself is
1183 	return next_zones_zonelist(zonelist->_zonerefs,  in first_zones_zonelist()
1188 …ch_zone_zonelist_nodemask - helper macro to iterate over valid zones in a zonelist at or below a g…
1190  * @z: The current pointer within zonelist->_zonerefs being iterated
1195  * This iterator iterates though all zones at or below a given zone index and
1196  * within a given nodemask
1198 #define for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, nodemask) \  argument
1199 	for (z = first_zones_zonelist(zlist, highidx, nodemask), zone = zonelist_zone(z);	\
1201 		z = next_zones_zonelist(++z, highidx, nodemask),	\
1202 			zone = zonelist_zone(z))
1204 #define for_next_zone_zonelist_nodemask(zone, z, highidx, nodemask) \  argument
1205 	for (zone = z->zone;	\
1207 		z = next_zones_zonelist(++z, highidx, nodemask),	\
1208 			zone = zonelist_zone(z))
1212 …* for_each_zone_zonelist - helper macro to iterate over valid zones in a zonelist at or below a gi…
1214  * @z: The current pointer within zonelist->zones being iterated
1218  * This iterator iterates though all zones at or below a given zone index.
1220 #define for_each_zone_zonelist(zone, z, zlist, highidx) \  argument
1221 	for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, NULL)
1238 #define PFN_SECTION_SHIFT	(SECTION_SIZE_BITS - PAGE_SHIFT)
1243 #define PAGE_SECTION_MASK	(~(PAGES_PER_SECTION-1))
1246 	((1UL << (PFN_SECTION_SHIFT - pageblock_order)) * NR_PAGEBLOCK_BITS)
1248 #if (MAX_ORDER - 1 + PAGE_SHIFT) > SECTION_SIZE_BITS
1261 #define SECTION_ALIGN_UP(pfn)	(((pfn) + PAGES_PER_SECTION - 1) & PAGE_SECTION_MASK)
1267 #define PFN_SUBSECTION_SHIFT (SUBSECTION_SHIFT - PAGE_SHIFT)
1269 #define PAGE_SUBSECTION_MASK (~(PAGES_PER_SUBSECTION-1))
1274 #define SUBSECTIONS_PER_SECTION (1UL << (SECTION_SIZE_BITS - SUBSECTION_SHIFT))
1294 	 * This is, logically, a pointer to an array of struct
1302 	 * Making it a UL at least makes someone do a cast
1317 	 * WARNING: mem_section must be a power-of-2 in size for the
1330 #define SECTION_ROOT_MASK	(SECTIONS_PER_ROOT - 1)
1340 	return ms->usage->pageblock_flags;  in section_to_usemap()
1357  * a little bit of information.  The pointer is calculated
1358  * as mem_map - section_nr_to_pfn(pnum).  The result is
1360  *   1. All mem_map arrays are page-aligned.
1362  *      lowest bits.  PFN_SECTION_SHIFT is arch-specific
1363  *      (equal SECTION_SIZE_BITS - PAGE_SHIFT), and the
1374 #define SECTION_MAP_MASK		(~(SECTION_MAP_LAST_BIT-1))
1379 	unsigned long map = section->section_mem_map;  in __section_mem_map_addr()
1386 	return (section && (section->section_mem_map & SECTION_MARKED_PRESENT));  in present_section()
1396 	return (section && (section->section_mem_map & SECTION_HAS_MEM_MAP));  in valid_section()
1401 	return (section && (section->section_mem_map & SECTION_IS_EARLY));  in early_section()
1411 	return (section && (section->section_mem_map & SECTION_IS_ONLINE));  in online_section()
1418 	return section && ((section->section_mem_map & flags) == flags);  in online_device_section()
1448 	return test_bit(idx, ms->usage->subsection_map);  in pfn_section_valid()
1459  * pfn_valid - check if there is a valid memory map entry for a PFN
1462  * Check if there is a valid memory map entry aka struct page for the @pfn.
1465  * represent a hole or an unusable page frame.
1477 	 * match a valid pfn.  in pfn_valid()
1489 	 * the entire section-sized span.  in pfn_valid()
1509 	return -1;  in next_present_section_nr()