diff options
Diffstat (limited to 'mm/page_alloc.c')
-rw-r--r-- | mm/page_alloc.c | 62 |
1 files changed, 34 insertions, 28 deletions
diff --git a/mm/page_alloc.c b/mm/page_alloc.c index df54e2fc8ee0..791690d7d3fa 100644 --- a/mm/page_alloc.c +++ b/mm/page_alloc.c | |||
@@ -56,6 +56,7 @@ long nr_swap_pages; | |||
56 | int percpu_pagelist_fraction; | 56 | int percpu_pagelist_fraction; |
57 | 57 | ||
58 | static void fastcall free_hot_cold_page(struct page *page, int cold); | 58 | static void fastcall free_hot_cold_page(struct page *page, int cold); |
59 | static void __free_pages_ok(struct page *page, unsigned int order); | ||
59 | 60 | ||
60 | /* | 61 | /* |
61 | * results with 256, 32 in the lowmem_reserve sysctl: | 62 | * results with 256, 32 in the lowmem_reserve sysctl: |
@@ -169,20 +170,23 @@ static void bad_page(struct page *page) | |||
169 | * All pages have PG_compound set. All pages have their ->private pointing at | 170 | * All pages have PG_compound set. All pages have their ->private pointing at |
170 | * the head page (even the head page has this). | 171 | * the head page (even the head page has this). |
171 | * | 172 | * |
172 | * The first tail page's ->mapping, if non-zero, holds the address of the | 173 | * The first tail page's ->lru.next holds the address of the compound page's |
173 | * compound page's put_page() function. | 174 | * put_page() function. Its ->lru.prev holds the order of allocation. |
174 | * | 175 | * This usage means that zero-order pages may not be compound. |
175 | * The order of the allocation is stored in the first tail page's ->index | ||
176 | * This is only for debug at present. This usage means that zero-order pages | ||
177 | * may not be compound. | ||
178 | */ | 176 | */ |
177 | |||
178 | static void free_compound_page(struct page *page) | ||
179 | { | ||
180 | __free_pages_ok(page, (unsigned long)page[1].lru.prev); | ||
181 | } | ||
182 | |||
179 | static void prep_compound_page(struct page *page, unsigned long order) | 183 | static void prep_compound_page(struct page *page, unsigned long order) |
180 | { | 184 | { |
181 | int i; | 185 | int i; |
182 | int nr_pages = 1 << order; | 186 | int nr_pages = 1 << order; |
183 | 187 | ||
184 | page[1].mapping = NULL; | 188 | page[1].lru.next = (void *)free_compound_page; /* set dtor */ |
185 | page[1].index = order; | 189 | page[1].lru.prev = (void *)order; |
186 | for (i = 0; i < nr_pages; i++) { | 190 | for (i = 0; i < nr_pages; i++) { |
187 | struct page *p = page + i; | 191 | struct page *p = page + i; |
188 | 192 | ||
@@ -196,7 +200,7 @@ static void destroy_compound_page(struct page *page, unsigned long order) | |||
196 | int i; | 200 | int i; |
197 | int nr_pages = 1 << order; | 201 | int nr_pages = 1 << order; |
198 | 202 | ||
199 | if (unlikely(page[1].index != order)) | 203 | if (unlikely((unsigned long)page[1].lru.prev != order)) |
200 | bad_page(page); | 204 | bad_page(page); |
201 | 205 | ||
202 | for (i = 0; i < nr_pages; i++) { | 206 | for (i = 0; i < nr_pages; i++) { |
@@ -1011,7 +1015,7 @@ rebalance: | |||
1011 | if (page) | 1015 | if (page) |
1012 | goto got_pg; | 1016 | goto got_pg; |
1013 | 1017 | ||
1014 | out_of_memory(gfp_mask, order); | 1018 | out_of_memory(zonelist, gfp_mask, order); |
1015 | goto restart; | 1019 | goto restart; |
1016 | } | 1020 | } |
1017 | 1021 | ||
@@ -1213,18 +1217,21 @@ static void __get_page_state(struct page_state *ret, int nr, cpumask_t *cpumask) | |||
1213 | { | 1217 | { |
1214 | int cpu = 0; | 1218 | int cpu = 0; |
1215 | 1219 | ||
1216 | memset(ret, 0, sizeof(*ret)); | 1220 | memset(ret, 0, nr * sizeof(unsigned long)); |
1217 | cpus_and(*cpumask, *cpumask, cpu_online_map); | 1221 | cpus_and(*cpumask, *cpumask, cpu_online_map); |
1218 | 1222 | ||
1219 | cpu = first_cpu(*cpumask); | 1223 | cpu = first_cpu(*cpumask); |
1220 | while (cpu < NR_CPUS) { | 1224 | while (cpu < NR_CPUS) { |
1221 | unsigned long *in, *out, off; | 1225 | unsigned long *in, *out, off; |
1222 | 1226 | ||
1227 | if (!cpu_isset(cpu, *cpumask)) | ||
1228 | continue; | ||
1229 | |||
1223 | in = (unsigned long *)&per_cpu(page_states, cpu); | 1230 | in = (unsigned long *)&per_cpu(page_states, cpu); |
1224 | 1231 | ||
1225 | cpu = next_cpu(cpu, *cpumask); | 1232 | cpu = next_cpu(cpu, *cpumask); |
1226 | 1233 | ||
1227 | if (cpu < NR_CPUS) | 1234 | if (likely(cpu < NR_CPUS)) |
1228 | prefetch(&per_cpu(page_states, cpu)); | 1235 | prefetch(&per_cpu(page_states, cpu)); |
1229 | 1236 | ||
1230 | out = (unsigned long *)ret; | 1237 | out = (unsigned long *)ret; |
@@ -1534,29 +1541,29 @@ static int __initdata node_load[MAX_NUMNODES]; | |||
1534 | */ | 1541 | */ |
1535 | static int __init find_next_best_node(int node, nodemask_t *used_node_mask) | 1542 | static int __init find_next_best_node(int node, nodemask_t *used_node_mask) |
1536 | { | 1543 | { |
1537 | int i, n, val; | 1544 | int n, val; |
1538 | int min_val = INT_MAX; | 1545 | int min_val = INT_MAX; |
1539 | int best_node = -1; | 1546 | int best_node = -1; |
1540 | 1547 | ||
1541 | for_each_online_node(i) { | 1548 | /* Use the local node if we haven't already */ |
1542 | cpumask_t tmp; | 1549 | if (!node_isset(node, *used_node_mask)) { |
1550 | node_set(node, *used_node_mask); | ||
1551 | return node; | ||
1552 | } | ||
1543 | 1553 | ||
1544 | /* Start from local node */ | 1554 | for_each_online_node(n) { |
1545 | n = (node+i) % num_online_nodes(); | 1555 | cpumask_t tmp; |
1546 | 1556 | ||
1547 | /* Don't want a node to appear more than once */ | 1557 | /* Don't want a node to appear more than once */ |
1548 | if (node_isset(n, *used_node_mask)) | 1558 | if (node_isset(n, *used_node_mask)) |
1549 | continue; | 1559 | continue; |
1550 | 1560 | ||
1551 | /* Use the local node if we haven't already */ | ||
1552 | if (!node_isset(node, *used_node_mask)) { | ||
1553 | best_node = node; | ||
1554 | break; | ||
1555 | } | ||
1556 | |||
1557 | /* Use the distance array to find the distance */ | 1561 | /* Use the distance array to find the distance */ |
1558 | val = node_distance(node, n); | 1562 | val = node_distance(node, n); |
1559 | 1563 | ||
1564 | /* Penalize nodes under us ("prefer the next node") */ | ||
1565 | val += (n < node); | ||
1566 | |||
1560 | /* Give preference to headless and unused nodes */ | 1567 | /* Give preference to headless and unused nodes */ |
1561 | tmp = node_to_cpumask(n); | 1568 | tmp = node_to_cpumask(n); |
1562 | if (!cpus_empty(tmp)) | 1569 | if (!cpus_empty(tmp)) |
@@ -1799,7 +1806,7 @@ void zonetable_add(struct zone *zone, int nid, int zid, unsigned long pfn, | |||
1799 | memmap_init_zone((size), (nid), (zone), (start_pfn)) | 1806 | memmap_init_zone((size), (nid), (zone), (start_pfn)) |
1800 | #endif | 1807 | #endif |
1801 | 1808 | ||
1802 | static int __meminit zone_batchsize(struct zone *zone) | 1809 | static int __cpuinit zone_batchsize(struct zone *zone) |
1803 | { | 1810 | { |
1804 | int batch; | 1811 | int batch; |
1805 | 1812 | ||
@@ -1886,14 +1893,13 @@ static void setup_pagelist_highmark(struct per_cpu_pageset *p, | |||
1886 | * not check if the processor is online before following the pageset pointer. | 1893 | * not check if the processor is online before following the pageset pointer. |
1887 | * Other parts of the kernel may not check if the zone is available. | 1894 | * Other parts of the kernel may not check if the zone is available. |
1888 | */ | 1895 | */ |
1889 | static struct per_cpu_pageset | 1896 | static struct per_cpu_pageset boot_pageset[NR_CPUS]; |
1890 | boot_pageset[NR_CPUS]; | ||
1891 | 1897 | ||
1892 | /* | 1898 | /* |
1893 | * Dynamically allocate memory for the | 1899 | * Dynamically allocate memory for the |
1894 | * per cpu pageset array in struct zone. | 1900 | * per cpu pageset array in struct zone. |
1895 | */ | 1901 | */ |
1896 | static int __meminit process_zones(int cpu) | 1902 | static int __cpuinit process_zones(int cpu) |
1897 | { | 1903 | { |
1898 | struct zone *zone, *dzone; | 1904 | struct zone *zone, *dzone; |
1899 | 1905 | ||
@@ -1934,7 +1940,7 @@ static inline void free_zone_pagesets(int cpu) | |||
1934 | } | 1940 | } |
1935 | } | 1941 | } |
1936 | 1942 | ||
1937 | static int __meminit pageset_cpuup_callback(struct notifier_block *nfb, | 1943 | static int __cpuinit pageset_cpuup_callback(struct notifier_block *nfb, |
1938 | unsigned long action, | 1944 | unsigned long action, |
1939 | void *hcpu) | 1945 | void *hcpu) |
1940 | { | 1946 | { |