diff options
Diffstat (limited to 'mm')
-rw-r--r-- | mm/page_alloc.c | 51 | ||||
-rw-r--r-- | mm/vmscan.c | 39 | ||||
-rw-r--r-- | mm/vmstat.c | 6 |
3 files changed, 50 insertions, 46 deletions
diff --git a/mm/page_alloc.c b/mm/page_alloc.c index 8485735fc690..abe26003124d 100644 --- a/mm/page_alloc.c +++ b/mm/page_alloc.c | |||
@@ -1150,10 +1150,15 @@ failed: | |||
1150 | return NULL; | 1150 | return NULL; |
1151 | } | 1151 | } |
1152 | 1152 | ||
1153 | #define ALLOC_NO_WATERMARKS 0x01 /* don't check watermarks at all */ | 1153 | /* The ALLOC_WMARK bits are used as an index to zone->watermark */ |
1154 | #define ALLOC_WMARK_MIN 0x02 /* use pages_min watermark */ | 1154 | #define ALLOC_WMARK_MIN WMARK_MIN |
1155 | #define ALLOC_WMARK_LOW 0x04 /* use pages_low watermark */ | 1155 | #define ALLOC_WMARK_LOW WMARK_LOW |
1156 | #define ALLOC_WMARK_HIGH 0x08 /* use pages_high watermark */ | 1156 | #define ALLOC_WMARK_HIGH WMARK_HIGH |
1157 | #define ALLOC_NO_WATERMARKS 0x04 /* don't check watermarks at all */ | ||
1158 | |||
1159 | /* Mask to get the watermark bits */ | ||
1160 | #define ALLOC_WMARK_MASK (ALLOC_NO_WATERMARKS-1) | ||
1161 | |||
1157 | #define ALLOC_HARDER 0x10 /* try to alloc harder */ | 1162 | #define ALLOC_HARDER 0x10 /* try to alloc harder */ |
1158 | #define ALLOC_HIGH 0x20 /* __GFP_HIGH set */ | 1163 | #define ALLOC_HIGH 0x20 /* __GFP_HIGH set */ |
1159 | #define ALLOC_CPUSET 0x40 /* check for correct cpuset */ | 1164 | #define ALLOC_CPUSET 0x40 /* check for correct cpuset */ |
@@ -1440,14 +1445,10 @@ zonelist_scan: | |||
1440 | !cpuset_zone_allowed_softwall(zone, gfp_mask)) | 1445 | !cpuset_zone_allowed_softwall(zone, gfp_mask)) |
1441 | goto try_next_zone; | 1446 | goto try_next_zone; |
1442 | 1447 | ||
1448 | BUILD_BUG_ON(ALLOC_NO_WATERMARKS < NR_WMARK); | ||
1443 | if (!(alloc_flags & ALLOC_NO_WATERMARKS)) { | 1449 | if (!(alloc_flags & ALLOC_NO_WATERMARKS)) { |
1444 | unsigned long mark; | 1450 | unsigned long mark; |
1445 | if (alloc_flags & ALLOC_WMARK_MIN) | 1451 | mark = zone->watermark[alloc_flags & ALLOC_WMARK_MASK]; |
1446 | mark = zone->pages_min; | ||
1447 | else if (alloc_flags & ALLOC_WMARK_LOW) | ||
1448 | mark = zone->pages_low; | ||
1449 | else | ||
1450 | mark = zone->pages_high; | ||
1451 | if (!zone_watermark_ok(zone, order, mark, | 1452 | if (!zone_watermark_ok(zone, order, mark, |
1452 | classzone_idx, alloc_flags)) { | 1453 | classzone_idx, alloc_flags)) { |
1453 | if (!zone_reclaim_mode || | 1454 | if (!zone_reclaim_mode || |
@@ -1959,7 +1960,7 @@ static unsigned int nr_free_zone_pages(int offset) | |||
1959 | 1960 | ||
1960 | for_each_zone_zonelist(zone, z, zonelist, offset) { | 1961 | for_each_zone_zonelist(zone, z, zonelist, offset) { |
1961 | unsigned long size = zone->present_pages; | 1962 | unsigned long size = zone->present_pages; |
1962 | unsigned long high = zone->pages_high; | 1963 | unsigned long high = high_wmark_pages(zone); |
1963 | if (size > high) | 1964 | if (size > high) |
1964 | sum += size - high; | 1965 | sum += size - high; |
1965 | } | 1966 | } |
@@ -2096,9 +2097,9 @@ void show_free_areas(void) | |||
2096 | "\n", | 2097 | "\n", |
2097 | zone->name, | 2098 | zone->name, |
2098 | K(zone_page_state(zone, NR_FREE_PAGES)), | 2099 | K(zone_page_state(zone, NR_FREE_PAGES)), |
2099 | K(zone->pages_min), | 2100 | K(min_wmark_pages(zone)), |
2100 | K(zone->pages_low), | 2101 | K(low_wmark_pages(zone)), |
2101 | K(zone->pages_high), | 2102 | K(high_wmark_pages(zone)), |
2102 | K(zone_page_state(zone, NR_ACTIVE_ANON)), | 2103 | K(zone_page_state(zone, NR_ACTIVE_ANON)), |
2103 | K(zone_page_state(zone, NR_INACTIVE_ANON)), | 2104 | K(zone_page_state(zone, NR_INACTIVE_ANON)), |
2104 | K(zone_page_state(zone, NR_ACTIVE_FILE)), | 2105 | K(zone_page_state(zone, NR_ACTIVE_FILE)), |
@@ -2702,8 +2703,8 @@ static inline unsigned long wait_table_bits(unsigned long size) | |||
2702 | 2703 | ||
2703 | /* | 2704 | /* |
2704 | * Mark a number of pageblocks as MIGRATE_RESERVE. The number | 2705 | * Mark a number of pageblocks as MIGRATE_RESERVE. The number |
2705 | * of blocks reserved is based on zone->pages_min. The memory within the | 2706 | * of blocks reserved is based on min_wmark_pages(zone). The memory within |
2706 | * reserve will tend to store contiguous free pages. Setting min_free_kbytes | 2707 | * the reserve will tend to store contiguous free pages. Setting min_free_kbytes |
2707 | * higher will lead to a bigger reserve which will get freed as contiguous | 2708 | * higher will lead to a bigger reserve which will get freed as contiguous |
2708 | * blocks as reclaim kicks in | 2709 | * blocks as reclaim kicks in |
2709 | */ | 2710 | */ |
@@ -2716,7 +2717,7 @@ static void setup_zone_migrate_reserve(struct zone *zone) | |||
2716 | /* Get the start pfn, end pfn and the number of blocks to reserve */ | 2717 | /* Get the start pfn, end pfn and the number of blocks to reserve */ |
2717 | start_pfn = zone->zone_start_pfn; | 2718 | start_pfn = zone->zone_start_pfn; |
2718 | end_pfn = start_pfn + zone->spanned_pages; | 2719 | end_pfn = start_pfn + zone->spanned_pages; |
2719 | reserve = roundup(zone->pages_min, pageblock_nr_pages) >> | 2720 | reserve = roundup(min_wmark_pages(zone), pageblock_nr_pages) >> |
2720 | pageblock_order; | 2721 | pageblock_order; |
2721 | 2722 | ||
2722 | for (pfn = start_pfn; pfn < end_pfn; pfn += pageblock_nr_pages) { | 2723 | for (pfn = start_pfn; pfn < end_pfn; pfn += pageblock_nr_pages) { |
@@ -4319,8 +4320,8 @@ static void calculate_totalreserve_pages(void) | |||
4319 | max = zone->lowmem_reserve[j]; | 4320 | max = zone->lowmem_reserve[j]; |
4320 | } | 4321 | } |
4321 | 4322 | ||
4322 | /* we treat pages_high as reserved pages. */ | 4323 | /* we treat the high watermark as reserved pages. */ |
4323 | max += zone->pages_high; | 4324 | max += high_wmark_pages(zone); |
4324 | 4325 | ||
4325 | if (max > zone->present_pages) | 4326 | if (max > zone->present_pages) |
4326 | max = zone->present_pages; | 4327 | max = zone->present_pages; |
@@ -4400,7 +4401,7 @@ void setup_per_zone_pages_min(void) | |||
4400 | * need highmem pages, so cap pages_min to a small | 4401 | * need highmem pages, so cap pages_min to a small |
4401 | * value here. | 4402 | * value here. |
4402 | * | 4403 | * |
4403 | * The (pages_high-pages_low) and (pages_low-pages_min) | 4404 | * The WMARK_HIGH-WMARK_LOW and (WMARK_LOW-WMARK_MIN) |
4404 | * deltas controls asynch page reclaim, and so should | 4405 | * deltas controls asynch page reclaim, and so should |
4405 | * not be capped for highmem. | 4406 | * not be capped for highmem. |
4406 | */ | 4407 | */ |
@@ -4411,17 +4412,17 @@ void setup_per_zone_pages_min(void) | |||
4411 | min_pages = SWAP_CLUSTER_MAX; | 4412 | min_pages = SWAP_CLUSTER_MAX; |
4412 | if (min_pages > 128) | 4413 | if (min_pages > 128) |
4413 | min_pages = 128; | 4414 | min_pages = 128; |
4414 | zone->pages_min = min_pages; | 4415 | zone->watermark[WMARK_MIN] = min_pages; |
4415 | } else { | 4416 | } else { |
4416 | /* | 4417 | /* |
4417 | * If it's a lowmem zone, reserve a number of pages | 4418 | * If it's a lowmem zone, reserve a number of pages |
4418 | * proportionate to the zone's size. | 4419 | * proportionate to the zone's size. |
4419 | */ | 4420 | */ |
4420 | zone->pages_min = tmp; | 4421 | zone->watermark[WMARK_MIN] = tmp; |
4421 | } | 4422 | } |
4422 | 4423 | ||
4423 | zone->pages_low = zone->pages_min + (tmp >> 2); | 4424 | zone->watermark[WMARK_LOW] = min_wmark_pages(zone) + (tmp >> 2); |
4424 | zone->pages_high = zone->pages_min + (tmp >> 1); | 4425 | zone->watermark[WMARK_HIGH] = min_wmark_pages(zone) + (tmp >> 1); |
4425 | setup_zone_migrate_reserve(zone); | 4426 | setup_zone_migrate_reserve(zone); |
4426 | spin_unlock_irqrestore(&zone->lock, flags); | 4427 | spin_unlock_irqrestore(&zone->lock, flags); |
4427 | } | 4428 | } |
@@ -4566,7 +4567,7 @@ int sysctl_min_slab_ratio_sysctl_handler(ctl_table *table, int write, | |||
4566 | * whenever sysctl_lowmem_reserve_ratio changes. | 4567 | * whenever sysctl_lowmem_reserve_ratio changes. |
4567 | * | 4568 | * |
4568 | * The reserve ratio obviously has absolutely no relation with the | 4569 | * The reserve ratio obviously has absolutely no relation with the |
4569 | * pages_min watermarks. The lowmem reserve ratio can only make sense | 4570 | * minimum watermarks. The lowmem reserve ratio can only make sense |
4570 | * if in function of the boot time zone sizes. | 4571 | * if in function of the boot time zone sizes. |
4571 | */ | 4572 | */ |
4572 | int lowmem_reserve_ratio_sysctl_handler(ctl_table *table, int write, | 4573 | int lowmem_reserve_ratio_sysctl_handler(ctl_table *table, int write, |
diff --git a/mm/vmscan.c b/mm/vmscan.c index a6b7d14812e6..e5245d051647 100644 --- a/mm/vmscan.c +++ b/mm/vmscan.c | |||
@@ -1401,7 +1401,7 @@ static void get_scan_ratio(struct zone *zone, struct scan_control *sc, | |||
1401 | free = zone_page_state(zone, NR_FREE_PAGES); | 1401 | free = zone_page_state(zone, NR_FREE_PAGES); |
1402 | /* If we have very few page cache pages, | 1402 | /* If we have very few page cache pages, |
1403 | force-scan anon pages. */ | 1403 | force-scan anon pages. */ |
1404 | if (unlikely(file + free <= zone->pages_high)) { | 1404 | if (unlikely(file + free <= high_wmark_pages(zone))) { |
1405 | percent[0] = 100; | 1405 | percent[0] = 100; |
1406 | percent[1] = 0; | 1406 | percent[1] = 0; |
1407 | return; | 1407 | return; |
@@ -1533,11 +1533,13 @@ static void shrink_zone(int priority, struct zone *zone, | |||
1533 | * try to reclaim pages from zones which will satisfy the caller's allocation | 1533 | * try to reclaim pages from zones which will satisfy the caller's allocation |
1534 | * request. | 1534 | * request. |
1535 | * | 1535 | * |
1536 | * We reclaim from a zone even if that zone is over pages_high. Because: | 1536 | * We reclaim from a zone even if that zone is over high_wmark_pages(zone). |
1537 | * Because: | ||
1537 | * a) The caller may be trying to free *extra* pages to satisfy a higher-order | 1538 | * a) The caller may be trying to free *extra* pages to satisfy a higher-order |
1538 | * allocation or | 1539 | * allocation or |
1539 | * b) The zones may be over pages_high but they must go *over* pages_high to | 1540 | * b) The target zone may be at high_wmark_pages(zone) but the lower zones |
1540 | * satisfy the `incremental min' zone defense algorithm. | 1541 | * must go *over* high_wmark_pages(zone) to satisfy the `incremental min' |
1542 | * zone defense algorithm. | ||
1541 | * | 1543 | * |
1542 | * If a zone is deemed to be full of pinned pages then just give it a light | 1544 | * If a zone is deemed to be full of pinned pages then just give it a light |
1543 | * scan then give up on it. | 1545 | * scan then give up on it. |
@@ -1743,7 +1745,7 @@ unsigned long try_to_free_mem_cgroup_pages(struct mem_cgroup *mem_cont, | |||
1743 | 1745 | ||
1744 | /* | 1746 | /* |
1745 | * For kswapd, balance_pgdat() will work across all this node's zones until | 1747 | * For kswapd, balance_pgdat() will work across all this node's zones until |
1746 | * they are all at pages_high. | 1748 | * they are all at high_wmark_pages(zone). |
1747 | * | 1749 | * |
1748 | * Returns the number of pages which were actually freed. | 1750 | * Returns the number of pages which were actually freed. |
1749 | * | 1751 | * |
@@ -1756,11 +1758,11 @@ unsigned long try_to_free_mem_cgroup_pages(struct mem_cgroup *mem_cont, | |||
1756 | * the zone for when the problem goes away. | 1758 | * the zone for when the problem goes away. |
1757 | * | 1759 | * |
1758 | * kswapd scans the zones in the highmem->normal->dma direction. It skips | 1760 | * kswapd scans the zones in the highmem->normal->dma direction. It skips |
1759 | * zones which have free_pages > pages_high, but once a zone is found to have | 1761 | * zones which have free_pages > high_wmark_pages(zone), but once a zone is |
1760 | * free_pages <= pages_high, we scan that zone and the lower zones regardless | 1762 | * found to have free_pages <= high_wmark_pages(zone), we scan that zone and the |
1761 | * of the number of free pages in the lower zones. This interoperates with | 1763 | * lower zones regardless of the number of free pages in the lower zones. This |
1762 | * the page allocator fallback scheme to ensure that aging of pages is balanced | 1764 | * interoperates with the page allocator fallback scheme to ensure that aging |
1763 | * across the zones. | 1765 | * of pages is balanced across the zones. |
1764 | */ | 1766 | */ |
1765 | static unsigned long balance_pgdat(pg_data_t *pgdat, int order) | 1767 | static unsigned long balance_pgdat(pg_data_t *pgdat, int order) |
1766 | { | 1768 | { |
@@ -1781,7 +1783,8 @@ static unsigned long balance_pgdat(pg_data_t *pgdat, int order) | |||
1781 | }; | 1783 | }; |
1782 | /* | 1784 | /* |
1783 | * temp_priority is used to remember the scanning priority at which | 1785 | * temp_priority is used to remember the scanning priority at which |
1784 | * this zone was successfully refilled to free_pages == pages_high. | 1786 | * this zone was successfully refilled to |
1787 | * free_pages == high_wmark_pages(zone). | ||
1785 | */ | 1788 | */ |
1786 | int temp_priority[MAX_NR_ZONES]; | 1789 | int temp_priority[MAX_NR_ZONES]; |
1787 | 1790 | ||
@@ -1826,8 +1829,8 @@ loop_again: | |||
1826 | shrink_active_list(SWAP_CLUSTER_MAX, zone, | 1829 | shrink_active_list(SWAP_CLUSTER_MAX, zone, |
1827 | &sc, priority, 0); | 1830 | &sc, priority, 0); |
1828 | 1831 | ||
1829 | if (!zone_watermark_ok(zone, order, zone->pages_high, | 1832 | if (!zone_watermark_ok(zone, order, |
1830 | 0, 0)) { | 1833 | high_wmark_pages(zone), 0, 0)) { |
1831 | end_zone = i; | 1834 | end_zone = i; |
1832 | break; | 1835 | break; |
1833 | } | 1836 | } |
@@ -1861,8 +1864,8 @@ loop_again: | |||
1861 | priority != DEF_PRIORITY) | 1864 | priority != DEF_PRIORITY) |
1862 | continue; | 1865 | continue; |
1863 | 1866 | ||
1864 | if (!zone_watermark_ok(zone, order, zone->pages_high, | 1867 | if (!zone_watermark_ok(zone, order, |
1865 | end_zone, 0)) | 1868 | high_wmark_pages(zone), end_zone, 0)) |
1866 | all_zones_ok = 0; | 1869 | all_zones_ok = 0; |
1867 | temp_priority[i] = priority; | 1870 | temp_priority[i] = priority; |
1868 | sc.nr_scanned = 0; | 1871 | sc.nr_scanned = 0; |
@@ -1871,8 +1874,8 @@ loop_again: | |||
1871 | * We put equal pressure on every zone, unless one | 1874 | * We put equal pressure on every zone, unless one |
1872 | * zone has way too many pages free already. | 1875 | * zone has way too many pages free already. |
1873 | */ | 1876 | */ |
1874 | if (!zone_watermark_ok(zone, order, 8*zone->pages_high, | 1877 | if (!zone_watermark_ok(zone, order, |
1875 | end_zone, 0)) | 1878 | 8*high_wmark_pages(zone), end_zone, 0)) |
1876 | shrink_zone(priority, zone, &sc); | 1879 | shrink_zone(priority, zone, &sc); |
1877 | reclaim_state->reclaimed_slab = 0; | 1880 | reclaim_state->reclaimed_slab = 0; |
1878 | nr_slab = shrink_slab(sc.nr_scanned, GFP_KERNEL, | 1881 | nr_slab = shrink_slab(sc.nr_scanned, GFP_KERNEL, |
@@ -2038,7 +2041,7 @@ void wakeup_kswapd(struct zone *zone, int order) | |||
2038 | return; | 2041 | return; |
2039 | 2042 | ||
2040 | pgdat = zone->zone_pgdat; | 2043 | pgdat = zone->zone_pgdat; |
2041 | if (zone_watermark_ok(zone, order, zone->pages_low, 0, 0)) | 2044 | if (zone_watermark_ok(zone, order, low_wmark_pages(zone), 0, 0)) |
2042 | return; | 2045 | return; |
2043 | if (pgdat->kswapd_max_order < order) | 2046 | if (pgdat->kswapd_max_order < order) |
2044 | pgdat->kswapd_max_order = order; | 2047 | pgdat->kswapd_max_order = order; |
diff --git a/mm/vmstat.c b/mm/vmstat.c index 74d66dba0cbe..415110772c73 100644 --- a/mm/vmstat.c +++ b/mm/vmstat.c | |||
@@ -714,9 +714,9 @@ static void zoneinfo_show_print(struct seq_file *m, pg_data_t *pgdat, | |||
714 | "\n spanned %lu" | 714 | "\n spanned %lu" |
715 | "\n present %lu", | 715 | "\n present %lu", |
716 | zone_page_state(zone, NR_FREE_PAGES), | 716 | zone_page_state(zone, NR_FREE_PAGES), |
717 | zone->pages_min, | 717 | min_wmark_pages(zone), |
718 | zone->pages_low, | 718 | low_wmark_pages(zone), |
719 | zone->pages_high, | 719 | high_wmark_pages(zone), |
720 | zone->pages_scanned, | 720 | zone->pages_scanned, |
721 | zone->lru[LRU_ACTIVE_ANON].nr_scan, | 721 | zone->lru[LRU_ACTIVE_ANON].nr_scan, |
722 | zone->lru[LRU_INACTIVE_ANON].nr_scan, | 722 | zone->lru[LRU_INACTIVE_ANON].nr_scan, |