blk-throttle: fix lockdep warning of "cgroup_mutex or RCU read lock required!"
[linux-block.git] / mm / show_mem.c
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  * Generic show_mem() implementation
4  *
5  * Copyright (C) 2008 Johannes Weiner <hannes@saeurebad.de>
6  */
7
8 #include <linux/blkdev.h>
9 #include <linux/cma.h>
10 #include <linux/cpuset.h>
11 #include <linux/highmem.h>
12 #include <linux/hugetlb.h>
13 #include <linux/mm.h>
14 #include <linux/mmzone.h>
15 #include <linux/swap.h>
16 #include <linux/vmstat.h>
17
18 #include "internal.h"
19 #include "swap.h"
20
21 atomic_long_t _totalram_pages __read_mostly;
22 EXPORT_SYMBOL(_totalram_pages);
23 unsigned long totalreserve_pages __read_mostly;
24 unsigned long totalcma_pages __read_mostly;
25
26 static inline void show_node(struct zone *zone)
27 {
28         if (IS_ENABLED(CONFIG_NUMA))
29                 printk("Node %d ", zone_to_nid(zone));
30 }
31
32 long si_mem_available(void)
33 {
34         long available;
35         unsigned long pagecache;
36         unsigned long wmark_low = 0;
37         unsigned long reclaimable;
38         struct zone *zone;
39
40         for_each_zone(zone)
41                 wmark_low += low_wmark_pages(zone);
42
43         /*
44          * Estimate the amount of memory available for userspace allocations,
45          * without causing swapping or OOM.
46          */
47         available = global_zone_page_state(NR_FREE_PAGES) - totalreserve_pages;
48
49         /*
50          * Not all the page cache can be freed, otherwise the system will
51          * start swapping or thrashing. Assume at least half of the page
52          * cache, or the low watermark worth of cache, needs to stay.
53          */
54         pagecache = global_node_page_state(NR_ACTIVE_FILE) +
55                 global_node_page_state(NR_INACTIVE_FILE);
56         pagecache -= min(pagecache / 2, wmark_low);
57         available += pagecache;
58
59         /*
60          * Part of the reclaimable slab and other kernel memory consists of
61          * items that are in use, and cannot be freed. Cap this estimate at the
62          * low watermark.
63          */
64         reclaimable = global_node_page_state_pages(NR_SLAB_RECLAIMABLE_B) +
65                 global_node_page_state(NR_KERNEL_MISC_RECLAIMABLE);
66         reclaimable -= min(reclaimable / 2, wmark_low);
67         available += reclaimable;
68
69         if (available < 0)
70                 available = 0;
71         return available;
72 }
73 EXPORT_SYMBOL_GPL(si_mem_available);
74
75 void si_meminfo(struct sysinfo *val)
76 {
77         val->totalram = totalram_pages();
78         val->sharedram = global_node_page_state(NR_SHMEM);
79         val->freeram = global_zone_page_state(NR_FREE_PAGES);
80         val->bufferram = nr_blockdev_pages();
81         val->totalhigh = totalhigh_pages();
82         val->freehigh = nr_free_highpages();
83         val->mem_unit = PAGE_SIZE;
84 }
85
86 EXPORT_SYMBOL(si_meminfo);
87
88 #ifdef CONFIG_NUMA
89 void si_meminfo_node(struct sysinfo *val, int nid)
90 {
91         int zone_type;          /* needs to be signed */
92         unsigned long managed_pages = 0;
93         unsigned long managed_highpages = 0;
94         unsigned long free_highpages = 0;
95         pg_data_t *pgdat = NODE_DATA(nid);
96
97         for (zone_type = 0; zone_type < MAX_NR_ZONES; zone_type++)
98                 managed_pages += zone_managed_pages(&pgdat->node_zones[zone_type]);
99         val->totalram = managed_pages;
100         val->sharedram = node_page_state(pgdat, NR_SHMEM);
101         val->freeram = sum_zone_node_page_state(nid, NR_FREE_PAGES);
102 #ifdef CONFIG_HIGHMEM
103         for (zone_type = 0; zone_type < MAX_NR_ZONES; zone_type++) {
104                 struct zone *zone = &pgdat->node_zones[zone_type];
105
106                 if (is_highmem(zone)) {
107                         managed_highpages += zone_managed_pages(zone);
108                         free_highpages += zone_page_state(zone, NR_FREE_PAGES);
109                 }
110         }
111         val->totalhigh = managed_highpages;
112         val->freehigh = free_highpages;
113 #else
114         val->totalhigh = managed_highpages;
115         val->freehigh = free_highpages;
116 #endif
117         val->mem_unit = PAGE_SIZE;
118 }
119 #endif
120
121 /*
122  * Determine whether the node should be displayed or not, depending on whether
123  * SHOW_MEM_FILTER_NODES was passed to show_free_areas().
124  */
125 static bool show_mem_node_skip(unsigned int flags, int nid, nodemask_t *nodemask)
126 {
127         if (!(flags & SHOW_MEM_FILTER_NODES))
128                 return false;
129
130         /*
131          * no node mask - aka implicit memory numa policy. Do not bother with
132          * the synchronization - read_mems_allowed_begin - because we do not
133          * have to be precise here.
134          */
135         if (!nodemask)
136                 nodemask = &cpuset_current_mems_allowed;
137
138         return !node_isset(nid, *nodemask);
139 }
140
141 static void show_migration_types(unsigned char type)
142 {
143         static const char types[MIGRATE_TYPES] = {
144                 [MIGRATE_UNMOVABLE]     = 'U',
145                 [MIGRATE_MOVABLE]       = 'M',
146                 [MIGRATE_RECLAIMABLE]   = 'E',
147                 [MIGRATE_HIGHATOMIC]    = 'H',
148 #ifdef CONFIG_CMA
149                 [MIGRATE_CMA]           = 'C',
150 #endif
151 #ifdef CONFIG_MEMORY_ISOLATION
152                 [MIGRATE_ISOLATE]       = 'I',
153 #endif
154         };
155         char tmp[MIGRATE_TYPES + 1];
156         char *p = tmp;
157         int i;
158
159         for (i = 0; i < MIGRATE_TYPES; i++) {
160                 if (type & (1 << i))
161                         *p++ = types[i];
162         }
163
164         *p = '\0';
165         printk(KERN_CONT "(%s) ", tmp);
166 }
167
168 static bool node_has_managed_zones(pg_data_t *pgdat, int max_zone_idx)
169 {
170         int zone_idx;
171         for (zone_idx = 0; zone_idx <= max_zone_idx; zone_idx++)
172                 if (zone_managed_pages(pgdat->node_zones + zone_idx))
173                         return true;
174         return false;
175 }
176
177 /*
178  * Show free area list (used inside shift_scroll-lock stuff)
179  * We also calculate the percentage fragmentation. We do this by counting the
180  * memory on each free list with the exception of the first item on the list.
181  *
182  * Bits in @filter:
183  * SHOW_MEM_FILTER_NODES: suppress nodes that are not allowed by current's
184  *   cpuset.
185  */
186 static void show_free_areas(unsigned int filter, nodemask_t *nodemask, int max_zone_idx)
187 {
188         unsigned long free_pcp = 0;
189         int cpu, nid;
190         struct zone *zone;
191         pg_data_t *pgdat;
192
193         for_each_populated_zone(zone) {
194                 if (zone_idx(zone) > max_zone_idx)
195                         continue;
196                 if (show_mem_node_skip(filter, zone_to_nid(zone), nodemask))
197                         continue;
198
199                 for_each_online_cpu(cpu)
200                         free_pcp += per_cpu_ptr(zone->per_cpu_pageset, cpu)->count;
201         }
202
203         printk("active_anon:%lu inactive_anon:%lu isolated_anon:%lu\n"
204                 " active_file:%lu inactive_file:%lu isolated_file:%lu\n"
205                 " unevictable:%lu dirty:%lu writeback:%lu\n"
206                 " slab_reclaimable:%lu slab_unreclaimable:%lu\n"
207                 " mapped:%lu shmem:%lu pagetables:%lu\n"
208                 " sec_pagetables:%lu bounce:%lu\n"
209                 " kernel_misc_reclaimable:%lu\n"
210                 " free:%lu free_pcp:%lu free_cma:%lu\n",
211                 global_node_page_state(NR_ACTIVE_ANON),
212                 global_node_page_state(NR_INACTIVE_ANON),
213                 global_node_page_state(NR_ISOLATED_ANON),
214                 global_node_page_state(NR_ACTIVE_FILE),
215                 global_node_page_state(NR_INACTIVE_FILE),
216                 global_node_page_state(NR_ISOLATED_FILE),
217                 global_node_page_state(NR_UNEVICTABLE),
218                 global_node_page_state(NR_FILE_DIRTY),
219                 global_node_page_state(NR_WRITEBACK),
220                 global_node_page_state_pages(NR_SLAB_RECLAIMABLE_B),
221                 global_node_page_state_pages(NR_SLAB_UNRECLAIMABLE_B),
222                 global_node_page_state(NR_FILE_MAPPED),
223                 global_node_page_state(NR_SHMEM),
224                 global_node_page_state(NR_PAGETABLE),
225                 global_node_page_state(NR_SECONDARY_PAGETABLE),
226                 global_zone_page_state(NR_BOUNCE),
227                 global_node_page_state(NR_KERNEL_MISC_RECLAIMABLE),
228                 global_zone_page_state(NR_FREE_PAGES),
229                 free_pcp,
230                 global_zone_page_state(NR_FREE_CMA_PAGES));
231
232         for_each_online_pgdat(pgdat) {
233                 if (show_mem_node_skip(filter, pgdat->node_id, nodemask))
234                         continue;
235                 if (!node_has_managed_zones(pgdat, max_zone_idx))
236                         continue;
237
238                 printk("Node %d"
239                         " active_anon:%lukB"
240                         " inactive_anon:%lukB"
241                         " active_file:%lukB"
242                         " inactive_file:%lukB"
243                         " unevictable:%lukB"
244                         " isolated(anon):%lukB"
245                         " isolated(file):%lukB"
246                         " mapped:%lukB"
247                         " dirty:%lukB"
248                         " writeback:%lukB"
249                         " shmem:%lukB"
250 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
251                         " shmem_thp:%lukB"
252                         " shmem_pmdmapped:%lukB"
253                         " anon_thp:%lukB"
254 #endif
255                         " writeback_tmp:%lukB"
256                         " kernel_stack:%lukB"
257 #ifdef CONFIG_SHADOW_CALL_STACK
258                         " shadow_call_stack:%lukB"
259 #endif
260                         " pagetables:%lukB"
261                         " sec_pagetables:%lukB"
262                         " all_unreclaimable? %s"
263                         "\n",
264                         pgdat->node_id,
265                         K(node_page_state(pgdat, NR_ACTIVE_ANON)),
266                         K(node_page_state(pgdat, NR_INACTIVE_ANON)),
267                         K(node_page_state(pgdat, NR_ACTIVE_FILE)),
268                         K(node_page_state(pgdat, NR_INACTIVE_FILE)),
269                         K(node_page_state(pgdat, NR_UNEVICTABLE)),
270                         K(node_page_state(pgdat, NR_ISOLATED_ANON)),
271                         K(node_page_state(pgdat, NR_ISOLATED_FILE)),
272                         K(node_page_state(pgdat, NR_FILE_MAPPED)),
273                         K(node_page_state(pgdat, NR_FILE_DIRTY)),
274                         K(node_page_state(pgdat, NR_WRITEBACK)),
275                         K(node_page_state(pgdat, NR_SHMEM)),
276 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
277                         K(node_page_state(pgdat, NR_SHMEM_THPS)),
278                         K(node_page_state(pgdat, NR_SHMEM_PMDMAPPED)),
279                         K(node_page_state(pgdat, NR_ANON_THPS)),
280 #endif
281                         K(node_page_state(pgdat, NR_WRITEBACK_TEMP)),
282                         node_page_state(pgdat, NR_KERNEL_STACK_KB),
283 #ifdef CONFIG_SHADOW_CALL_STACK
284                         node_page_state(pgdat, NR_KERNEL_SCS_KB),
285 #endif
286                         K(node_page_state(pgdat, NR_PAGETABLE)),
287                         K(node_page_state(pgdat, NR_SECONDARY_PAGETABLE)),
288                         pgdat->kswapd_failures >= MAX_RECLAIM_RETRIES ?
289                                 "yes" : "no");
290         }
291
292         for_each_populated_zone(zone) {
293                 int i;
294
295                 if (zone_idx(zone) > max_zone_idx)
296                         continue;
297                 if (show_mem_node_skip(filter, zone_to_nid(zone), nodemask))
298                         continue;
299
300                 free_pcp = 0;
301                 for_each_online_cpu(cpu)
302                         free_pcp += per_cpu_ptr(zone->per_cpu_pageset, cpu)->count;
303
304                 show_node(zone);
305                 printk(KERN_CONT
306                         "%s"
307                         " free:%lukB"
308                         " boost:%lukB"
309                         " min:%lukB"
310                         " low:%lukB"
311                         " high:%lukB"
312                         " reserved_highatomic:%luKB"
313                         " active_anon:%lukB"
314                         " inactive_anon:%lukB"
315                         " active_file:%lukB"
316                         " inactive_file:%lukB"
317                         " unevictable:%lukB"
318                         " writepending:%lukB"
319                         " present:%lukB"
320                         " managed:%lukB"
321                         " mlocked:%lukB"
322                         " bounce:%lukB"
323                         " free_pcp:%lukB"
324                         " local_pcp:%ukB"
325                         " free_cma:%lukB"
326                         "\n",
327                         zone->name,
328                         K(zone_page_state(zone, NR_FREE_PAGES)),
329                         K(zone->watermark_boost),
330                         K(min_wmark_pages(zone)),
331                         K(low_wmark_pages(zone)),
332                         K(high_wmark_pages(zone)),
333                         K(zone->nr_reserved_highatomic),
334                         K(zone_page_state(zone, NR_ZONE_ACTIVE_ANON)),
335                         K(zone_page_state(zone, NR_ZONE_INACTIVE_ANON)),
336                         K(zone_page_state(zone, NR_ZONE_ACTIVE_FILE)),
337                         K(zone_page_state(zone, NR_ZONE_INACTIVE_FILE)),
338                         K(zone_page_state(zone, NR_ZONE_UNEVICTABLE)),
339                         K(zone_page_state(zone, NR_ZONE_WRITE_PENDING)),
340                         K(zone->present_pages),
341                         K(zone_managed_pages(zone)),
342                         K(zone_page_state(zone, NR_MLOCK)),
343                         K(zone_page_state(zone, NR_BOUNCE)),
344                         K(free_pcp),
345                         K(this_cpu_read(zone->per_cpu_pageset->count)),
346                         K(zone_page_state(zone, NR_FREE_CMA_PAGES)));
347                 printk("lowmem_reserve[]:");
348                 for (i = 0; i < MAX_NR_ZONES; i++)
349                         printk(KERN_CONT " %ld", zone->lowmem_reserve[i]);
350                 printk(KERN_CONT "\n");
351         }
352
353         for_each_populated_zone(zone) {
354                 unsigned int order;
355                 unsigned long nr[MAX_ORDER + 1], flags, total = 0;
356                 unsigned char types[MAX_ORDER + 1];
357
358                 if (zone_idx(zone) > max_zone_idx)
359                         continue;
360                 if (show_mem_node_skip(filter, zone_to_nid(zone), nodemask))
361                         continue;
362                 show_node(zone);
363                 printk(KERN_CONT "%s: ", zone->name);
364
365                 spin_lock_irqsave(&zone->lock, flags);
366                 for (order = 0; order <= MAX_ORDER; order++) {
367                         struct free_area *area = &zone->free_area[order];
368                         int type;
369
370                         nr[order] = area->nr_free;
371                         total += nr[order] << order;
372
373                         types[order] = 0;
374                         for (type = 0; type < MIGRATE_TYPES; type++) {
375                                 if (!free_area_empty(area, type))
376                                         types[order] |= 1 << type;
377                         }
378                 }
379                 spin_unlock_irqrestore(&zone->lock, flags);
380                 for (order = 0; order <= MAX_ORDER; order++) {
381                         printk(KERN_CONT "%lu*%lukB ",
382                                nr[order], K(1UL) << order);
383                         if (nr[order])
384                                 show_migration_types(types[order]);
385                 }
386                 printk(KERN_CONT "= %lukB\n", K(total));
387         }
388
389         for_each_online_node(nid) {
390                 if (show_mem_node_skip(filter, nid, nodemask))
391                         continue;
392                 hugetlb_show_meminfo_node(nid);
393         }
394
395         printk("%ld total pagecache pages\n", global_node_page_state(NR_FILE_PAGES));
396
397         show_swap_cache_info();
398 }
399
400 void __show_mem(unsigned int filter, nodemask_t *nodemask, int max_zone_idx)
401 {
402         unsigned long total = 0, reserved = 0, highmem = 0;
403         struct zone *zone;
404
405         printk("Mem-Info:\n");
406         show_free_areas(filter, nodemask, max_zone_idx);
407
408         for_each_populated_zone(zone) {
409
410                 total += zone->present_pages;
411                 reserved += zone->present_pages - zone_managed_pages(zone);
412
413                 if (is_highmem(zone))
414                         highmem += zone->present_pages;
415         }
416
417         printk("%lu pages RAM\n", total);
418         printk("%lu pages HighMem/MovableOnly\n", highmem);
419         printk("%lu pages reserved\n", reserved);
420 #ifdef CONFIG_CMA
421         printk("%lu pages cma reserved\n", totalcma_pages);
422 #endif
423 #ifdef CONFIG_MEMORY_FAILURE
424         printk("%lu pages hwpoisoned\n", atomic_long_read(&num_poisoned_pages));
425 #endif
426 }