aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
Diffstat
-rw-r--r--include/asm-s390/pgtable.h7
-rw-r--r--include/linux/mm.h6
-rw-r--r--include/linux/page-flags.h141
-rw-r--r--include/linux/vmstat.h138
-rw-r--r--mm/Makefile2
-rw-r--r--mm/page_alloc.c407
-rw-r--r--mm/vmstat.c417
7 files changed, 569 insertions, 549 deletions
diff --git a/include/asm-s390/pgtable.h b/include/asm-s390/pgtable.h
index 859b5e969826..24312387fa24 100644
--- a/include/asm-s390/pgtable.h
+++ b/include/asm-s390/pgtable.h
@@ -657,13 +657,6 @@ static inline pte_t mk_pte_phys(unsigned long physpage, pgprot_t pgprot)
657 __pte; \ 657 __pte; \
658}) 658})
659 659
660#define SetPageUptodate(_page) \
661 do { \
662 struct page *__page = (_page); \
663 if (!test_and_set_bit(PG_uptodate, &__page->flags)) \
664 page_test_and_clear_dirty(_page); \
665 } while (0)
666
667#ifdef __s390x__ 660#ifdef __s390x__
668 661
669#define pfn_pmd(pfn, pgprot) \ 662#define pfn_pmd(pfn, pgprot) \
diff --git a/include/linux/mm.h b/include/linux/mm.h
index c41a1299b8cf..75179529e399 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -36,7 +36,6 @@ extern int sysctl_legacy_va_layout;
36#include <asm/page.h> 36#include <asm/page.h>
37#include <asm/pgtable.h> 37#include <asm/pgtable.h>
38#include <asm/processor.h> 38#include <asm/processor.h>
39#include <asm/atomic.h>
40 39
41#define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n)) 40#define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
42 41
@@ -515,6 +514,11 @@ static inline void set_page_links(struct page *page, unsigned long zone,
515 set_page_section(page, pfn_to_section_nr(pfn)); 514 set_page_section(page, pfn_to_section_nr(pfn));
516} 515}
517 516
517/*
518 * Some inline functions in vmstat.h depend on page_zone()
519 */
520#include <linux/vmstat.h>
521
518#ifndef CONFIG_DISCONTIGMEM 522#ifndef CONFIG_DISCONTIGMEM
519/* The array of struct pages - for discontigmem use pgdat->lmem_map */ 523/* The array of struct pages - for discontigmem use pgdat->lmem_map */
520extern struct page *mem_map; 524extern struct page *mem_map;
diff --git a/include/linux/page-flags.h b/include/linux/page-flags.h
index 0c076d58c676..ff235c4b79ea 100644
--- a/include/linux/page-flags.h
+++ b/include/linux/page-flags.h
@@ -5,12 +5,8 @@
5#ifndef PAGE_FLAGS_H 5#ifndef PAGE_FLAGS_H
6#define PAGE_FLAGS_H 6#define PAGE_FLAGS_H
7 7
8#include <linux/percpu.h>
9#include <linux/cache.h>
10#include <linux/types.h> 8#include <linux/types.h>
11 9
12#include <asm/pgtable.h>
13
14/* 10/*
15 * Various page->flags bits: 11 * Various page->flags bits:
16 * 12 *
@@ -103,134 +99,6 @@
103#endif 99#endif
104 100
105/* 101/*
106 * Global page accounting. One instance per CPU. Only unsigned longs are
107 * allowed.
108 *
109 * - Fields can be modified with xxx_page_state and xxx_page_state_zone at
110 * any time safely (which protects the instance from modification by
111 * interrupt.
112 * - The __xxx_page_state variants can be used safely when interrupts are
113 * disabled.
114 * - The __xxx_page_state variants can be used if the field is only
115 * modified from process context and protected from preemption, or only
116 * modified from interrupt context. In this case, the field should be
117 * commented here.
118 */
119struct page_state {
120 unsigned long nr_dirty; /* Dirty writeable pages */
121 unsigned long nr_writeback; /* Pages under writeback */
122 unsigned long nr_unstable; /* NFS unstable pages */
123 unsigned long nr_page_table_pages;/* Pages used for pagetables */
124 unsigned long nr_mapped; /* mapped into pagetables.
125 * only modified from process context */
126 unsigned long nr_slab; /* In slab */
127#define GET_PAGE_STATE_LAST nr_slab
128
129 /*
130 * The below are zeroed by get_page_state(). Use get_full_page_state()
131 * to add up all these.
132 */
133 unsigned long pgpgin; /* Disk reads */
134 unsigned long pgpgout; /* Disk writes */
135 unsigned long pswpin; /* swap reads */
136 unsigned long pswpout; /* swap writes */
137
138 unsigned long pgalloc_high; /* page allocations */
139 unsigned long pgalloc_normal;
140 unsigned long pgalloc_dma32;
141 unsigned long pgalloc_dma;
142
143 unsigned long pgfree; /* page freeings */
144 unsigned long pgactivate; /* pages moved inactive->active */
145 unsigned long pgdeactivate; /* pages moved active->inactive */
146
147 unsigned long pgfault; /* faults (major+minor) */
148 unsigned long pgmajfault; /* faults (major only) */
149
150 unsigned long pgrefill_high; /* inspected in refill_inactive_zone */
151 unsigned long pgrefill_normal;
152 unsigned long pgrefill_dma32;
153 unsigned long pgrefill_dma;
154
155 unsigned long pgsteal_high; /* total highmem pages reclaimed */
156 unsigned long pgsteal_normal;
157 unsigned long pgsteal_dma32;
158 unsigned long pgsteal_dma;
159
160 unsigned long pgscan_kswapd_high;/* total highmem pages scanned */
161 unsigned long pgscan_kswapd_normal;
162 unsigned long pgscan_kswapd_dma32;
163 unsigned long pgscan_kswapd_dma;
164
165 unsigned long pgscan_direct_high;/* total highmem pages scanned */
166 unsigned long pgscan_direct_normal;
167 unsigned long pgscan_direct_dma32;
168 unsigned long pgscan_direct_dma;
169
170 unsigned long pginodesteal; /* pages reclaimed via inode freeing */
171 unsigned long slabs_scanned; /* slab objects scanned */
172 unsigned long kswapd_steal; /* pages reclaimed by kswapd */
173 unsigned long kswapd_inodesteal;/* reclaimed via kswapd inode freeing */
174 unsigned long pageoutrun; /* kswapd's calls to page reclaim */
175 unsigned long allocstall; /* direct reclaim calls */
176
177 unsigned long pgrotated; /* pages rotated to tail of the LRU */
178 unsigned long nr_bounce; /* pages for bounce buffers */
179};
180
181extern void get_page_state(struct page_state *ret);
182extern void get_page_state_node(struct page_state *ret, int node);
183extern void get_full_page_state(struct page_state *ret);
184extern unsigned long read_page_state_offset(unsigned long offset);
185extern void mod_page_state_offset(unsigned long offset, unsigned long delta);
186extern void __mod_page_state_offset(unsigned long offset, unsigned long delta);
187
188#define read_page_state(member) \
189 read_page_state_offset(offsetof(struct page_state, member))
190
191#define mod_page_state(member, delta) \
192 mod_page_state_offset(offsetof(struct page_state, member), (delta))
193
194#define __mod_page_state(member, delta) \
195 __mod_page_state_offset(offsetof(struct page_state, member), (delta))
196
197#define inc_page_state(member) mod_page_state(member, 1UL)
198#define dec_page_state(member) mod_page_state(member, 0UL - 1)
199#define add_page_state(member,delta) mod_page_state(member, (delta))
200#define sub_page_state(member,delta) mod_page_state(member, 0UL - (delta))
201
202#define __inc_page_state(member) __mod_page_state(member, 1UL)
203#define __dec_page_state(member) __mod_page_state(member, 0UL - 1)
204#define __add_page_state(member,delta) __mod_page_state(member, (delta))
205#define __sub_page_state(member,delta) __mod_page_state(member, 0UL - (delta))
206
207#define page_state(member) (*__page_state(offsetof(struct page_state, member)))
208
209#define state_zone_offset(zone, member) \
210({ \
211 unsigned offset; \
212 if (is_highmem(zone)) \
213 offset = offsetof(struct page_state, member##_high); \
214 else if (is_normal(zone)) \
215 offset = offsetof(struct page_state, member##_normal); \
216 else if (is_dma32(zone)) \
217 offset = offsetof(struct page_state, member##_dma32); \
218 else \
219 offset = offsetof(struct page_state, member##_dma); \
220 offset; \
221})
222
223#define __mod_page_state_zone(zone, member, delta) \
224 do { \
225 __mod_page_state_offset(state_zone_offset(zone, member), (delta)); \
226 } while (0)
227
228#define mod_page_state_zone(zone, member, delta) \
229 do { \
230 mod_page_state_offset(state_zone_offset(zone, member), (delta)); \
231 } while (0)
232
233/*
234 * Manipulation of page state flags 102 * Manipulation of page state flags
235 */ 103 */
236#define PageLocked(page) \ 104#define PageLocked(page) \
@@ -254,7 +122,14 @@ extern void __mod_page_state_offset(unsigned long offset, unsigned long delta);
254#define TestClearPageReferenced(page) test_and_clear_bit(PG_referenced, &(page)->flags) 122#define TestClearPageReferenced(page) test_and_clear_bit(PG_referenced, &(page)->flags)
255 123
256#define PageUptodate(page) test_bit(PG_uptodate, &(page)->flags) 124#define PageUptodate(page) test_bit(PG_uptodate, &(page)->flags)
257#ifndef SetPageUptodate 125#ifdef CONFIG_S390
126#define SetPageUptodate(_page) \
127 do { \
128 struct page *__page = (_page); \
129 if (!test_and_set_bit(PG_uptodate, &__page->flags)) \
130 page_test_and_clear_dirty(_page); \
131 } while (0)
132#else
258#define SetPageUptodate(page) set_bit(PG_uptodate, &(page)->flags) 133#define SetPageUptodate(page) set_bit(PG_uptodate, &(page)->flags)
259#endif 134#endif
260#define ClearPageUptodate(page) clear_bit(PG_uptodate, &(page)->flags) 135#define ClearPageUptodate(page) clear_bit(PG_uptodate, &(page)->flags)
diff --git a/include/linux/vmstat.h b/include/linux/vmstat.h
new file mode 100644
index 000000000000..3ca0c1989fc2
--- /dev/null
+++ b/include/linux/vmstat.h
@@ -0,0 +1,138 @@
1#ifndef _LINUX_VMSTAT_H
2#define _LINUX_VMSTAT_H
3
4#include <linux/types.h>
5#include <linux/percpu.h>
6
7/*
8 * Global page accounting. One instance per CPU. Only unsigned longs are
9 * allowed.
10 *
11 * - Fields can be modified with xxx_page_state and xxx_page_state_zone at
12 * any time safely (which protects the instance from modification by
13 * interrupt.
14 * - The __xxx_page_state variants can be used safely when interrupts are
15 * disabled.
16 * - The __xxx_page_state variants can be used if the field is only
17 * modified from process context and protected from preemption, or only
18 * modified from interrupt context. In this case, the field should be
19 * commented here.
20 */
21struct page_state {
22 unsigned long nr_dirty; /* Dirty writeable pages */
23 unsigned long nr_writeback; /* Pages under writeback */
24 unsigned long nr_unstable; /* NFS unstable pages */
25 unsigned long nr_page_table_pages;/* Pages used for pagetables */
26 unsigned long nr_mapped; /* mapped into pagetables.
27 * only modified from process context */
28 unsigned long nr_slab; /* In slab */
29#define GET_PAGE_STATE_LAST nr_slab
30
31 /*
32 * The below are zeroed by get_page_state(). Use get_full_page_state()
33 * to add up all these.
34 */
35 unsigned long pgpgin; /* Disk reads */
36 unsigned long pgpgout; /* Disk writes */
37 unsigned long pswpin; /* swap reads */
38 unsigned long pswpout; /* swap writes */
39
40 unsigned long pgalloc_high; /* page allocations */
41 unsigned long pgalloc_normal;
42 unsigned long pgalloc_dma32;
43 unsigned long pgalloc_dma;
44
45 unsigned long pgfree; /* page freeings */
46 unsigned long pgactivate; /* pages moved inactive->active */
47 unsigned long pgdeactivate; /* pages moved active->inactive */
48
49 unsigned long pgfault; /* faults (major+minor) */
50 unsigned long pgmajfault; /* faults (major only) */
51
52 unsigned long pgrefill_high; /* inspected in refill_inactive_zone */
53 unsigned long pgrefill_normal;
54 unsigned long pgrefill_dma32;
55 unsigned long pgrefill_dma;
56
57 unsigned long pgsteal_high; /* total highmem pages reclaimed */
58 unsigned long pgsteal_normal;
59 unsigned long pgsteal_dma32;
60 unsigned long pgsteal_dma;
61
62 unsigned long pgscan_kswapd_high;/* total highmem pages scanned */
63 unsigned long pgscan_kswapd_normal;
64 unsigned long pgscan_kswapd_dma32;
65 unsigned long pgscan_kswapd_dma;
66
67 unsigned long pgscan_direct_high;/* total highmem pages scanned */
68 unsigned long pgscan_direct_normal;
69 unsigned long pgscan_direct_dma32;
70 unsigned long pgscan_direct_dma;
71
72 unsigned long pginodesteal; /* pages reclaimed via inode freeing */
73 unsigned long slabs_scanned; /* slab objects scanned */
74 unsigned long kswapd_steal; /* pages reclaimed by kswapd */
75 unsigned long kswapd_inodesteal;/* reclaimed via kswapd inode freeing */
76 unsigned long pageoutrun; /* kswapd's calls to page reclaim */
77 unsigned long allocstall; /* direct reclaim calls */
78
79 unsigned long pgrotated; /* pages rotated to tail of the LRU */
80 unsigned long nr_bounce; /* pages for bounce buffers */
81};
82
83extern void get_page_state(struct page_state *ret);
84extern void get_page_state_node(struct page_state *ret, int node);
85extern void get_full_page_state(struct page_state *ret);
86extern unsigned long read_page_state_offset(unsigned long offset);
87extern void mod_page_state_offset(unsigned long offset, unsigned long delta);
88extern void __mod_page_state_offset(unsigned long offset, unsigned long delta);
89
90#define read_page_state(member) \
91 read_page_state_offset(offsetof(struct page_state, member))
92
93#define mod_page_state(member, delta) \
94 mod_page_state_offset(offsetof(struct page_state, member), (delta))
95
96#define __mod_page_state(member, delta) \
97 __mod_page_state_offset(offsetof(struct page_state, member), (delta))
98
99#define inc_page_state(member) mod_page_state(member, 1UL)
100#define dec_page_state(member) mod_page_state(member, 0UL - 1)
101#define add_page_state(member,delta) mod_page_state(member, (delta))
102#define sub_page_state(member,delta) mod_page_state(member, 0UL - (delta))
103
104#define __inc_page_state(member) __mod_page_state(member, 1UL)
105#define __dec_page_state(member) __mod_page_state(member, 0UL - 1)
106#define __add_page_state(member,delta) __mod_page_state(member, (delta))
107#define __sub_page_state(member,delta) __mod_page_state(member, 0UL - (delta))
108
109#define page_state(member) (*__page_state(offsetof(struct page_state, member)))
110
111#define state_zone_offset(zone, member) \
112({ \
113 unsigned offset; \
114 if (is_highmem(zone)) \
115 offset = offsetof(struct page_state, member##_high); \
116 else if (is_normal(zone)) \
117 offset = offsetof(struct page_state, member##_normal); \
118 else if (is_dma32(zone)) \
119 offset = offsetof(struct page_state, member##_dma32); \
120 else \
121 offset = offsetof(struct page_state, member##_dma); \
122 offset; \
123})
124
125#define __mod_page_state_zone(zone, member, delta) \
126 do { \
127 __mod_page_state_offset(state_zone_offset(zone, member), (delta)); \
128 } while (0)
129
130#define mod_page_state_zone(zone, member, delta) \
131 do { \
132 mod_page_state_offset(state_zone_offset(zone, member), (delta)); \
133 } while (0)
134
135DECLARE_PER_CPU(struct page_state, page_states);
136
137#endif /* _LINUX_VMSTAT_H */
138
diff --git a/mm/Makefile b/mm/Makefile
index 0b8f73f2ed16..9dd824c11eeb 100644
--- a/mm/Makefile
+++ b/mm/Makefile
@@ -10,7 +10,7 @@ mmu-$(CONFIG_MMU) := fremap.o highmem.o madvise.o memory.o mincore.o \
10obj-y := bootmem.o filemap.o mempool.o oom_kill.o fadvise.o \ 10obj-y := bootmem.o filemap.o mempool.o oom_kill.o fadvise.o \
11 page_alloc.o page-writeback.o pdflush.o \ 11 page_alloc.o page-writeback.o pdflush.o \
12 readahead.o swap.o truncate.o vmscan.o \ 12 readahead.o swap.o truncate.o vmscan.o \
13 prio_tree.o util.o mmzone.o $(mmu-y) 13 prio_tree.o util.o mmzone.o vmstat.o $(mmu-y)
14 14
15obj-$(CONFIG_SWAP) += page_io.o swap_state.o swapfile.o thrash.o 15obj-$(CONFIG_SWAP) += page_io.o swap_state.o swapfile.o thrash.o
16obj-$(CONFIG_HUGETLBFS) += hugetlb.o 16obj-$(CONFIG_HUGETLBFS) += hugetlb.o
diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index 084a2de7e52a..87dc1297fe39 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -1231,141 +1231,6 @@ static void show_node(struct zone *zone)
1231#define show_node(zone) do { } while (0) 1231#define show_node(zone) do { } while (0)
1232#endif 1232#endif
1233 1233
1234/*
1235 * Accumulate the page_state information across all CPUs.
1236 * The result is unavoidably approximate - it can change
1237 * during and after execution of this function.
1238 */
1239static DEFINE_PER_CPU(struct page_state, page_states) = {0};
1240
1241atomic_t nr_pagecache = ATOMIC_INIT(0);
1242EXPORT_SYMBOL(nr_pagecache);
1243#ifdef CONFIG_SMP
1244DEFINE_PER_CPU(long, nr_pagecache_local) = 0;
1245#endif
1246
1247static void __get_page_state(struct page_state *ret, int nr, cpumask_t *cpumask)
1248{
1249 unsigned cpu;
1250
1251 memset(ret, 0, nr * sizeof(unsigned long));
1252 cpus_and(*cpumask, *cpumask, cpu_online_map);
1253
1254 for_each_cpu_mask(cpu, *cpumask) {
1255 unsigned long *in;
1256 unsigned long *out;
1257 unsigned off;
1258 unsigned next_cpu;
1259
1260 in = (unsigned long *)&per_cpu(page_states, cpu);
1261
1262 next_cpu = next_cpu(cpu, *cpumask);
1263 if (likely(next_cpu < NR_CPUS))
1264 prefetch(&per_cpu(page_states, next_cpu));
1265
1266 out = (unsigned long *)ret;
1267 for (off = 0; off < nr; off++)
1268 *out++ += *in++;
1269 }
1270}
1271
1272void get_page_state_node(struct page_state *ret, int node)
1273{
1274 int nr;
1275 cpumask_t mask = node_to_cpumask(node);
1276
1277 nr = offsetof(struct page_state, GET_PAGE_STATE_LAST);
1278 nr /= sizeof(unsigned long);
1279
1280 __get_page_state(ret, nr+1, &mask);
1281}
1282
1283void get_page_state(struct page_state *ret)
1284{
1285 int nr;
1286 cpumask_t mask = CPU_MASK_ALL;
1287
1288 nr = offsetof(struct page_state, GET_PAGE_STATE_LAST);
1289 nr /= sizeof(unsigned long);
1290
1291 __get_page_state(ret, nr + 1, &mask);
1292}
1293
1294void get_full_page_state(struct page_state *ret)
1295{
1296 cpumask_t mask = CPU_MASK_ALL;
1297
1298 __get_page_state(ret, sizeof(*ret) / sizeof(unsigned long), &mask);
1299}
1300
1301unsigned long read_page_state_offset(unsigned long offset)
1302{
1303 unsigned long ret = 0;
1304 int cpu;
1305
1306 for_each_online_cpu(cpu) {
1307 unsigned long in;
1308
1309 in = (unsigned long)&per_cpu(page_states, cpu) + offset;
1310 ret += *((unsigned long *)in);
1311 }
1312 return ret;
1313}
1314
1315void __mod_page_state_offset(unsigned long offset, unsigned long delta)
1316{
1317 void *ptr;
1318
1319 ptr = &__get_cpu_var(page_states);
1320 *(unsigned long *)(ptr + offset) += delta;
1321}
1322EXPORT_SYMBOL(__mod_page_state_offset);
1323
1324void mod_page_state_offset(unsigned long offset, unsigned long delta)
1325{
1326 unsigned long flags;
1327 void *ptr;
1328
1329 local_irq_save(flags);
1330 ptr = &__get_cpu_var(page_states);
1331 *(unsigned long *)(ptr + offset) += delta;
1332 local_irq_restore(flags);
1333}
1334EXPORT_SYMBOL(mod_page_state_offset);
1335
1336void __get_zone_counts(unsigned long *active, unsigned long *inactive,
1337 unsigned long *free, struct pglist_data *pgdat)
1338{
1339 struct zone *zones = pgdat->node_zones;
1340 int i;
1341
1342 *active = 0;
1343 *inactive = 0;
1344 *free = 0;
1345 for (i = 0; i < MAX_NR_ZONES; i++) {
1346 *active += zones[i].nr_active;
1347 *inactive += zones[i].nr_inactive;
1348 *free += zones[i].free_pages;
1349 }
1350}
1351
1352void get_zone_counts(unsigned long *active,
1353 unsigned long *inactive, unsigned long *free)
1354{
1355 struct pglist_data *pgdat;
1356
1357 *active = 0;
1358 *inactive = 0;
1359 *free = 0;
1360 for_each_online_pgdat(pgdat) {
1361 unsigned long l, m, n;
1362 __get_zone_counts(&l, &m, &n, pgdat);
1363 *active += l;
1364 *inactive += m;
1365 *free += n;
1366 }
1367}
1368
1369void si_meminfo(struct sysinfo *val) 1234void si_meminfo(struct sysinfo *val)
1370{ 1235{
1371 val->totalram = totalram_pages; 1236 val->totalram = totalram_pages;
@@ -2253,278 +2118,6 @@ void __init free_area_init(unsigned long *zones_size)
2253 __pa(PAGE_OFFSET) >> PAGE_SHIFT, NULL); 2118 __pa(PAGE_OFFSET) >> PAGE_SHIFT, NULL);
2254} 2119}
2255 2120
2256#ifdef CONFIG_PROC_FS
2257
2258#include <linux/seq_file.h>
2259
2260static void *frag_start(struct seq_file *m, loff_t *pos)
2261{
2262 pg_data_t *pgdat;
2263 loff_t node = *pos;
2264 for (pgdat = first_online_pgdat();
2265 pgdat && node;
2266 pgdat = next_online_pgdat(pgdat))
2267 --node;
2268
2269 return pgdat;
2270}
2271
2272static void *frag_next(struct seq_file *m, void *arg, loff_t *pos)
2273{
2274 pg_data_t *pgdat = (pg_data_t *)arg;
2275
2276 (*pos)++;
2277 return next_online_pgdat(pgdat);
2278}
2279
2280static void frag_stop(struct seq_file *m, void *arg)
2281{
2282}
2283
2284/*
2285 * This walks the free areas for each zone.
2286 */
2287static int frag_show(struct seq_file *m, void *arg)
2288{
2289 pg_data_t *pgdat = (pg_data_t *)arg;
2290 struct zone *zone;
2291 struct zone *node_zones = pgdat->node_zones;
2292 unsigned long flags;
2293 int order;
2294
2295 for (zone = node_zones; zone - node_zones < MAX_NR_ZONES; ++zone) {
2296 if (!populated_zone(zone))
2297 continue;
2298
2299 spin_lock_irqsave(&zone->lock, flags);
2300 seq_printf(m, "Node %d, zone %8s ", pgdat->node_id, zone->name);
2301 for (order = 0; order < MAX_ORDER; ++order)
2302 seq_printf(m, "%6lu ", zone->free_area[order].nr_free);
2303 spin_unlock_irqrestore(&zone->lock, flags);
2304 seq_putc(m, '\n');
2305 }
2306 return 0;
2307}
2308
2309struct seq_operations fragmentation_op = {
2310 .start = frag_start,
2311 .next = frag_next,
2312 .stop = frag_stop,
2313 .show = frag_show,
2314};
2315
2316/*
2317 * Output information about zones in @pgdat.
2318 */
2319static int zoneinfo_show(struct seq_file *m, void *arg)
2320{
2321 pg_data_t *pgdat = arg;
2322 struct zone *zone;
2323 struct zone *node_zones = pgdat->node_zones;
2324 unsigned long flags;
2325
2326 for (zone = node_zones; zone - node_zones < MAX_NR_ZONES; zone++) {
2327 int i;
2328
2329 if (!populated_zone(zone))
2330 continue;
2331
2332 spin_lock_irqsave(&zone->lock, flags);
2333 seq_printf(m, "Node %d, zone %8s", pgdat->node_id, zone->name);
2334 seq_printf(m,
2335 "\n pages free %lu"
2336 "\n min %lu"
2337 "\n low %lu"
2338 "\n high %lu"
2339 "\n active %lu"
2340 "\n inactive %lu"
2341 "\n scanned %lu (a: %lu i: %lu)"
2342 "\n spanned %lu"
2343 "\n present %lu",
2344 zone->free_pages,
2345 zone->pages_min,
2346 zone->pages_low,
2347 zone->pages_high,
2348 zone->nr_active,
2349 zone->nr_inactive,
2350 zone->pages_scanned,
2351 zone->nr_scan_active, zone->nr_scan_inactive,
2352 zone->spanned_pages,
2353 zone->present_pages);
2354 seq_printf(m,
2355 "\n protection: (%lu",
2356 zone->lowmem_reserve[0]);
2357 for (i = 1; i < ARRAY_SIZE(zone->lowmem_reserve); i++)
2358 seq_printf(m, ", %lu", zone->lowmem_reserve[i]);
2359 seq_printf(m,
2360 ")"
2361 "\n pagesets");
2362 for_each_online_cpu(i) {
2363 struct per_cpu_pageset *pageset;
2364 int j;
2365
2366 pageset = zone_pcp(zone, i);
2367 for (j = 0; j < ARRAY_SIZE(pageset->pcp); j++) {
2368 if (pageset->pcp[j].count)
2369 break;
2370 }
2371 if (j == ARRAY_SIZE(pageset->pcp))
2372 continue;
2373 for (j = 0; j < ARRAY_SIZE(pageset->pcp); j++) {
2374 seq_printf(m,
2375 "\n cpu: %i pcp: %i"
2376 "\n count: %i"
2377 "\n high: %i"
2378 "\n batch: %i",
2379 i, j,
2380 pageset->pcp[j].count,
2381 pageset->pcp[j].high,
2382 pageset->pcp[j].batch);
2383 }
2384#ifdef CONFIG_NUMA
2385 seq_printf(m,
2386 "\n numa_hit: %lu"
2387 "\n numa_miss: %lu"
2388 "\n numa_foreign: %lu"
2389 "\n interleave_hit: %lu"
2390 "\n local_node: %lu"
2391 "\n other_node: %lu",
2392 pageset->numa_hit,
2393 pageset->numa_miss,
2394 pageset->numa_foreign,
2395 pageset->interleave_hit,
2396 pageset->local_node,
2397 pageset->other_node);
2398#endif
2399 }
2400 seq_printf(m,
2401 "\n all_unreclaimable: %u"
2402 "\n prev_priority: %i"
2403 "\n temp_priority: %i"
2404 "\n start_pfn: %lu",
2405 zone->all_unreclaimable,
2406 zone->prev_priority,
2407 zone->temp_priority,
2408 zone->zone_start_pfn);
2409 spin_unlock_irqrestore(&zone->lock, flags);
2410 seq_putc(m, '\n');
2411 }
2412 return 0;
2413}
2414
2415struct seq_operations zoneinfo_op = {
2416 .start = frag_start, /* iterate over all zones. The same as in
2417 * fragmentation. */
2418 .next = frag_next,
2419 .stop = frag_stop,
2420 .show = zoneinfo_show,
2421};
2422
2423static char *vmstat_text[] = {
2424 "nr_dirty",
2425 "nr_writeback",
2426 "nr_unstable",
2427 "nr_page_table_pages",
2428 "nr_mapped",
2429 "nr_slab",
2430
2431 "pgpgin",
2432 "pgpgout",
2433 "pswpin",
2434 "pswpout",
2435
2436 "pgalloc_high",
2437 "pgalloc_normal",
2438 "pgalloc_dma32",
2439 "pgalloc_dma",
2440
2441 "pgfree",
2442 "pgactivate",
2443 "pgdeactivate",
2444
2445 "pgfault",
2446 "pgmajfault",
2447
2448 "pgrefill_high",
2449 "pgrefill_normal",
2450 "pgrefill_dma32",
2451 "pgrefill_dma",
2452
2453 "pgsteal_high",
2454 "pgsteal_normal",
2455 "pgsteal_dma32",
2456 "pgsteal_dma",
2457
2458 "pgscan_kswapd_high",
2459 "pgscan_kswapd_normal",
2460 "pgscan_kswapd_dma32",
2461 "pgscan_kswapd_dma",
2462
2463 "pgscan_direct_high",
2464 "pgscan_direct_normal",
2465 "pgscan_direct_dma32",
2466 "pgscan_direct_dma",
2467
2468 "pginodesteal",
2469 "slabs_scanned",
2470 "kswapd_steal",
2471 "kswapd_inodesteal",
2472 "pageoutrun",
2473 "allocstall",
2474
2475 "pgrotated",
2476 "nr_bounce",
2477};
2478
2479static void *vmstat_start(struct seq_file *m, loff_t *pos)
2480{
2481 struct page_state *ps;
2482
2483 if (*pos >= ARRAY_SIZE(vmstat_text))
2484 return NULL;
2485
2486 ps = kmalloc(sizeof(*ps), GFP_KERNEL);
2487 m->private = ps;
2488 if (!ps)
2489 return ERR_PTR(-ENOMEM);
2490 get_full_page_state(ps);
2491 ps->pgpgin /= 2; /* sectors -> kbytes */
2492 ps->pgpgout /= 2;
2493 return (unsigned long *)ps + *pos;
2494}
2495
2496static void *vmstat_next(struct seq_file *m, void *arg, loff_t *pos)
2497{
2498 (*pos)++;
2499 if (*pos >= ARRAY_SIZE(vmstat_text))
2500 return NULL;
2501 return (unsigned long *)m->private + *pos;
2502}
2503
2504static int vmstat_show(struct seq_file *m, void *arg)
2505{
2506 unsigned long *l = arg;
2507 unsigned long off = l - (unsigned long *)m->private;
2508
2509 seq_printf(m, "%s %lu\n", vmstat_text[off], *l);
2510 return 0;
2511}
2512
2513static void vmstat_stop(struct seq_file *m, void *arg)
2514{
2515 kfree(m->private);
2516 m->private = NULL;
2517}
2518
2519struct seq_operations vmstat_op = {
2520 .start = vmstat_start,
2521 .next = vmstat_next,
2522 .stop = vmstat_stop,
2523 .show = vmstat_show,
2524};
2525
2526#endif /* CONFIG_PROC_FS */
2527
2528#ifdef CONFIG_HOTPLUG_CPU 2121#ifdef CONFIG_HOTPLUG_CPU
2529static int page_alloc_cpu_notify(struct notifier_block *self, 2122static int page_alloc_cpu_notify(struct notifier_block *self,
2530 unsigned long action, void *hcpu) 2123 unsigned long action, void *hcpu)
diff --git a/mm/vmstat.c b/mm/vmstat.c
new file mode 100644
index 000000000000..ad456202ff1a
--- /dev/null
+++ b/mm/vmstat.c
@@ -0,0 +1,417 @@
1/*
2 * linux/mm/vmstat.c
3 *
4 * Manages VM statistics
5 * Copyright (C) 1991, 1992, 1993, 1994 Linus Torvalds
6 */
7
8#include <linux/config.h>
9#include <linux/mm.h>
10
11/*
12 * Accumulate the page_state information across all CPUs.
13 * The result is unavoidably approximate - it can change
14 * during and after execution of this function.
15 */
16DEFINE_PER_CPU(struct page_state, page_states) = {0};
17
18atomic_t nr_pagecache = ATOMIC_INIT(0);
19EXPORT_SYMBOL(nr_pagecache);
20#ifdef CONFIG_SMP
21DEFINE_PER_CPU(long, nr_pagecache_local) = 0;
22#endif
23
24static void __get_page_state(struct page_state *ret, int nr, cpumask_t *cpumask)
25{
26 unsigned cpu;
27
28 memset(ret, 0, nr * sizeof(unsigned long));
29 cpus_and(*cpumask, *cpumask, cpu_online_map);
30
31 for_each_cpu_mask(cpu, *cpumask) {
32 unsigned long *in;
33 unsigned long *out;
34 unsigned off;
35 unsigned next_cpu;
36
37 in = (unsigned long *)&per_cpu(page_states, cpu);
38
39 next_cpu = next_cpu(cpu, *cpumask);
40 if (likely(next_cpu < NR_CPUS))
41 prefetch(&per_cpu(page_states, next_cpu));
42
43 out = (unsigned long *)ret;
44 for (off = 0; off < nr; off++)
45 *out++ += *in++;
46 }
47}
48
49void get_page_state_node(struct page_state *ret, int node)
50{
51 int nr;
52 cpumask_t mask = node_to_cpumask(node);
53
54 nr = offsetof(struct page_state, GET_PAGE_STATE_LAST);
55 nr /= sizeof(unsigned long);
56
57 __get_page_state(ret, nr+1, &mask);
58}
59
60void get_page_state(struct page_state *ret)
61{
62 int nr;
63 cpumask_t mask = CPU_MASK_ALL;
64
65 nr = offsetof(struct page_state, GET_PAGE_STATE_LAST);
66 nr /= sizeof(unsigned long);
67
68 __get_page_state(ret, nr + 1, &mask);
69}
70
71void get_full_page_state(struct page_state *ret)
72{
73 cpumask_t mask = CPU_MASK_ALL;
74
75 __get_page_state(ret, sizeof(*ret) / sizeof(unsigned long), &mask);
76}
77
78unsigned long read_page_state_offset(unsigned long offset)
79{
80 unsigned long ret = 0;
81 int cpu;
82
83 for_each_online_cpu(cpu) {
84 unsigned long in;
85
86 in = (unsigned long)&per_cpu(page_states, cpu) + offset;
87 ret += *((unsigned long *)in);
88 }
89 return ret;
90}
91
92void __mod_page_state_offset(unsigned long offset, unsigned long delta)
93{
94 void *ptr;
95
96 ptr = &__get_cpu_var(page_states);
97 *(unsigned long *)(ptr + offset) += delta;
98}
99EXPORT_SYMBOL(__mod_page_state_offset);
100
101void mod_page_state_offset(unsigned long offset, unsigned long delta)
102{
103 unsigned long flags;
104 void *ptr;
105
106 local_irq_save(flags);
107 ptr = &__get_cpu_var(page_states);
108 *(unsigned long *)(ptr + offset) += delta;
109 local_irq_restore(flags);
110}
111EXPORT_SYMBOL(mod_page_state_offset);
112
113void __get_zone_counts(unsigned long *active, unsigned long *inactive,
114 unsigned long *free, struct pglist_data *pgdat)
115{
116 struct zone *zones = pgdat->node_zones;
117 int i;
118
119 *active = 0;
120 *inactive = 0;
121 *free = 0;
122 for (i = 0; i < MAX_NR_ZONES; i++) {
123 *active += zones[i].nr_active;
124 *inactive += zones[i].nr_inactive;
125 *free += zones[i].free_pages;
126 }
127}
128
129void get_zone_counts(unsigned long *active,
130 unsigned long *inactive, unsigned long *free)
131{
132 struct pglist_data *pgdat;
133
134 *active = 0;
135 *inactive = 0;
136 *free = 0;
137 for_each_online_pgdat(pgdat) {
138 unsigned long l, m, n;
139 __get_zone_counts(&l, &m, &n, pgdat);
140 *active += l;
141 *inactive += m;
142 *free += n;
143 }
144}
145
146#ifdef CONFIG_PROC_FS
147
148#include <linux/seq_file.h>
149
150static void *frag_start(struct seq_file *m, loff_t *pos)
151{
152 pg_data_t *pgdat;
153 loff_t node = *pos;
154 for (pgdat = first_online_pgdat();
155 pgdat && node;
156 pgdat = next_online_pgdat(pgdat))
157 --node;
158
159 return pgdat;
160}
161
162static void *frag_next(struct seq_file *m, void *arg, loff_t *pos)
163{
164 pg_data_t *pgdat = (pg_data_t *)arg;
165
166 (*pos)++;
167 return next_online_pgdat(pgdat);
168}
169
170static void frag_stop(struct seq_file *m, void *arg)
171{
172}
173
174/*
175 * This walks the free areas for each zone.
176 */
177static int frag_show(struct seq_file *m, void *arg)
178{
179 pg_data_t *pgdat = (pg_data_t *)arg;
180 struct zone *zone;
181 struct zone *node_zones = pgdat->node_zones;
182 unsigned long flags;
183 int order;
184
185 for (zone = node_zones; zone - node_zones < MAX_NR_ZONES; ++zone) {
186 if (!populated_zone(zone))
187 continue;
188
189 spin_lock_irqsave(&zone->lock, flags);
190 seq_printf(m, "Node %d, zone %8s ", pgdat->node_id, zone->name);
191 for (order = 0; order < MAX_ORDER; ++order)
192 seq_printf(m, "%6lu ", zone->free_area[order].nr_free);
193 spin_unlock_irqrestore(&zone->lock, flags);
194 seq_putc(m, '\n');
195 }
196 return 0;
197}
198
199struct seq_operations fragmentation_op = {
200 .start = frag_start,
201 .next = frag_next,
202 .stop = frag_stop,
203 .show = frag_show,
204};
205
206static char *vmstat_text[] = {
207 "nr_dirty",
208 "nr_writeback",
209 "nr_unstable",
210 "nr_page_table_pages",
211 "nr_mapped",
212 "nr_slab",
213
214 "pgpgin",
215 "pgpgout",
216 "pswpin",
217 "pswpout",
218
219 "pgalloc_high",
220 "pgalloc_normal",
221 "pgalloc_dma32",
222 "pgalloc_dma",
223
224 "pgfree",
225 "pgactivate",
226 "pgdeactivate",
227
228 "pgfault",
229 "pgmajfault",
230
231 "pgrefill_high",
232 "pgrefill_normal",
233 "pgrefill_dma32",
234 "pgrefill_dma",
235
236 "pgsteal_high",
237 "pgsteal_normal",
238 "pgsteal_dma32",
239 "pgsteal_dma",
240
241 "pgscan_kswapd_high",
242 "pgscan_kswapd_normal",
243 "pgscan_kswapd_dma32",
244 "pgscan_kswapd_dma",
245
246 "pgscan_direct_high",
247 "pgscan_direct_normal",
248 "pgscan_direct_dma32",
249 "pgscan_direct_dma",
250
251 "pginodesteal",
252 "slabs_scanned",
253 "kswapd_steal",
254 "kswapd_inodesteal",
255 "pageoutrun",
256 "allocstall",
257
258 "pgrotated",
259 "nr_bounce",
260};
261
262/*
263 * Output information about zones in @pgdat.
264 */
265static int zoneinfo_show(struct seq_file *m, void *arg)
266{
267 pg_data_t *pgdat = arg;
268 struct zone *zone;
269 struct zone *node_zones = pgdat->node_zones;
270 unsigned long flags;
271
272 for (zone = node_zones; zone - node_zones < MAX_NR_ZONES; zone++) {
273 int i;
274
275 if (!populated_zone(zone))
276 continue;
277
278 spin_lock_irqsave(&zone->lock, flags);
279 seq_printf(m, "Node %d, zone %8s", pgdat->node_id, zone->name);
280 seq_printf(m,
281 "\n pages free %lu"
282 "\n min %lu"
283 "\n low %lu"
284 "\n high %lu"
285 "\n active %lu"
286 "\n inactive %lu"
287 "\n scanned %lu (a: %lu i: %lu)"
288 "\n spanned %lu"
289 "\n present %lu",
290 zone->free_pages,
291 zone->pages_min,
292 zone->pages_low,
293 zone->pages_high,
294 zone->nr_active,
295 zone->nr_inactive,
296 zone->pages_scanned,
297 zone->nr_scan_active, zone->nr_scan_inactive,
298 zone->spanned_pages,
299 zone->present_pages);
300 seq_printf(m,
301 "\n protection: (%lu",
302 zone->lowmem_reserve[0]);
303 for (i = 1; i < ARRAY_SIZE(zone->lowmem_reserve); i++)
304 seq_printf(m, ", %lu", zone->lowmem_reserve[i]);
305 seq_printf(m,
306 ")"
307 "\n pagesets");
308 for_each_online_cpu(i) {
309 struct per_cpu_pageset *pageset;
310 int j;
311
312 pageset = zone_pcp(zone, i);
313 for (j = 0; j < ARRAY_SIZE(pageset->pcp); j++) {
314 if (pageset->pcp[j].count)
315 break;
316 }
317 if (j == ARRAY_SIZE(pageset->pcp))
318 continue;
319 for (j = 0; j < ARRAY_SIZE(pageset->pcp); j++) {
320 seq_printf(m,
321 "\n cpu: %i pcp: %i"
322 "\n count: %i"
323 "\n high: %i"
324 "\n batch: %i",
325 i, j,
326 pageset->pcp[j].count,
327 pageset->pcp[j].high,
328 pageset->pcp[j].batch);
329 }
330#ifdef CONFIG_NUMA
331 seq_printf(m,
332 "\n numa_hit: %lu"
333 "\n numa_miss: %lu"
334 "\n numa_foreign: %lu"
335 "\n interleave_hit: %lu"
336 "\n local_node: %lu"
337 "\n other_node: %lu",
338 pageset->numa_hit,
339 pageset->numa_miss,
340 pageset->numa_foreign,
341 pageset->interleave_hit,
342 pageset->local_node,
343 pageset->other_node);
344#endif
345 }
346 seq_printf(m,
347 "\n all_unreclaimable: %u"
348 "\n prev_priority: %i"
349 "\n temp_priority: %i"
350 "\n start_pfn: %lu",
351 zone->all_unreclaimable,
352 zone->prev_priority,
353 zone->temp_priority,
354 zone->zone_start_pfn);
355 spin_unlock_irqrestore(&zone->lock, flags);
356 seq_putc(m, '\n');
357 }
358 return 0;
359}
360
361struct seq_operations zoneinfo_op = {
362 .start = frag_start, /* iterate over all zones. The same as in
363 * fragmentation. */
364 .next = frag_next,
365 .stop = frag_stop,
366 .show = zoneinfo_show,
367};
368
369static void *vmstat_start(struct seq_file *m, loff_t *pos)
370{
371 struct page_state *ps;
372
373 if (*pos >= ARRAY_SIZE(vmstat_text))
374 return NULL;
375
376 ps = kmalloc(sizeof(*ps), GFP_KERNEL);
377 m->private = ps;
378 if (!ps)
379 return ERR_PTR(-ENOMEM);
380 get_full_page_state(ps);
381 ps->pgpgin /= 2; /* sectors -> kbytes */
382 ps->pgpgout /= 2;
383 return (unsigned long *)ps + *pos;
384}
385
386static void *vmstat_next(struct seq_file *m, void *arg, loff_t *pos)
387{
388 (*pos)++;
389 if (*pos >= ARRAY_SIZE(vmstat_text))
390 return NULL;
391 return (unsigned long *)m->private + *pos;
392}
393
394static int vmstat_show(struct seq_file *m, void *arg)
395{
396 unsigned long *l = arg;
397 unsigned long off = l - (unsigned long *)m->private;
398
399 seq_printf(m, "%s %lu\n", vmstat_text[off], *l);
400 return 0;
401}
402
403static void vmstat_stop(struct seq_file *m, void *arg)
404{
405 kfree(m->private);
406 m->private = NULL;
407}
408
409struct seq_operations vmstat_op = {
410 .start = vmstat_start,
411 .next = vmstat_next,
412 .stop = vmstat_stop,
413 .show = vmstat_show,
414};
415
416#endif /* CONFIG_PROC_FS */
417
generated by cgit v1.2.2 (git 2.25.0) at 2024-10-09 03:15:40 -0400