Commit | Line | Data |
---|---|---|
1da177e4 | 1 | /* |
10fbcf4c | 2 | * Basic Node interface support |
1da177e4 LT |
3 | */ |
4 | ||
1da177e4 LT |
5 | #include <linux/module.h> |
6 | #include <linux/init.h> | |
7 | #include <linux/mm.h> | |
c04fc586 | 8 | #include <linux/memory.h> |
fa25c503 | 9 | #include <linux/vmstat.h> |
6e259e7d | 10 | #include <linux/notifier.h> |
1da177e4 LT |
11 | #include <linux/node.h> |
12 | #include <linux/hugetlb.h> | |
ed4a6d7f | 13 | #include <linux/compaction.h> |
1da177e4 LT |
14 | #include <linux/cpumask.h> |
15 | #include <linux/topology.h> | |
16 | #include <linux/nodemask.h> | |
76b67ed9 | 17 | #include <linux/cpu.h> |
bde631a5 | 18 | #include <linux/device.h> |
af936a16 | 19 | #include <linux/swap.h> |
18e5b539 | 20 | #include <linux/slab.h> |
1da177e4 | 21 | |
10fbcf4c | 22 | static struct bus_type node_subsys = { |
af5ca3f4 | 23 | .name = "node", |
10fbcf4c | 24 | .dev_name = "node", |
1da177e4 LT |
25 | }; |
26 | ||
27 | ||
10fbcf4c | 28 | static ssize_t node_read_cpumap(struct device *dev, int type, char *buf) |
1da177e4 LT |
29 | { |
30 | struct node *node_dev = to_node(dev); | |
10fbcf4c | 31 | const struct cpumask *mask = cpumask_of_node(node_dev->dev.id); |
1da177e4 LT |
32 | int len; |
33 | ||
39106dcf MT |
34 | /* 2008/04/07: buf currently PAGE_SIZE, need 9 chars per 32 bits. */ |
35 | BUILD_BUG_ON((NR_CPUS/32 * 9) > (PAGE_SIZE-1)); | |
1da177e4 | 36 | |
39106dcf | 37 | len = type? |
29c0177e RR |
38 | cpulist_scnprintf(buf, PAGE_SIZE-2, mask) : |
39 | cpumask_scnprintf(buf, PAGE_SIZE-2, mask); | |
c5f59f08 MT |
40 | buf[len++] = '\n'; |
41 | buf[len] = '\0'; | |
1da177e4 LT |
42 | return len; |
43 | } | |
44 | ||
10fbcf4c KS |
45 | static inline ssize_t node_read_cpumask(struct device *dev, |
46 | struct device_attribute *attr, char *buf) | |
39106dcf MT |
47 | { |
48 | return node_read_cpumap(dev, 0, buf); | |
49 | } | |
10fbcf4c KS |
50 | static inline ssize_t node_read_cpulist(struct device *dev, |
51 | struct device_attribute *attr, char *buf) | |
39106dcf MT |
52 | { |
53 | return node_read_cpumap(dev, 1, buf); | |
54 | } | |
55 | ||
10fbcf4c KS |
56 | static DEVICE_ATTR(cpumap, S_IRUGO, node_read_cpumask, NULL); |
57 | static DEVICE_ATTR(cpulist, S_IRUGO, node_read_cpulist, NULL); | |
1da177e4 LT |
58 | |
59 | #define K(x) ((x) << (PAGE_SHIFT - 10)) | |
10fbcf4c KS |
60 | static ssize_t node_read_meminfo(struct device *dev, |
61 | struct device_attribute *attr, char *buf) | |
1da177e4 LT |
62 | { |
63 | int n; | |
64 | int nid = dev->id; | |
65 | struct sysinfo i; | |
1da177e4 LT |
66 | |
67 | si_meminfo_node(&i, nid); | |
7ee92255 | 68 | n = sprintf(buf, |
4f98a2fe RR |
69 | "Node %d MemTotal: %8lu kB\n" |
70 | "Node %d MemFree: %8lu kB\n" | |
71 | "Node %d MemUsed: %8lu kB\n" | |
72 | "Node %d Active: %8lu kB\n" | |
73 | "Node %d Inactive: %8lu kB\n" | |
74 | "Node %d Active(anon): %8lu kB\n" | |
75 | "Node %d Inactive(anon): %8lu kB\n" | |
76 | "Node %d Active(file): %8lu kB\n" | |
77 | "Node %d Inactive(file): %8lu kB\n" | |
5344b7e6 | 78 | "Node %d Unevictable: %8lu kB\n" |
7ee92255 KM |
79 | "Node %d Mlocked: %8lu kB\n", |
80 | nid, K(i.totalram), | |
81 | nid, K(i.freeram), | |
82 | nid, K(i.totalram - i.freeram), | |
83 | nid, K(node_page_state(nid, NR_ACTIVE_ANON) + | |
84 | node_page_state(nid, NR_ACTIVE_FILE)), | |
85 | nid, K(node_page_state(nid, NR_INACTIVE_ANON) + | |
86 | node_page_state(nid, NR_INACTIVE_FILE)), | |
87 | nid, K(node_page_state(nid, NR_ACTIVE_ANON)), | |
88 | nid, K(node_page_state(nid, NR_INACTIVE_ANON)), | |
89 | nid, K(node_page_state(nid, NR_ACTIVE_FILE)), | |
90 | nid, K(node_page_state(nid, NR_INACTIVE_FILE)), | |
91 | nid, K(node_page_state(nid, NR_UNEVICTABLE)), | |
92 | nid, K(node_page_state(nid, NR_MLOCK))); | |
93 | ||
182e8e23 | 94 | #ifdef CONFIG_HIGHMEM |
7ee92255 | 95 | n += sprintf(buf + n, |
4f98a2fe RR |
96 | "Node %d HighTotal: %8lu kB\n" |
97 | "Node %d HighFree: %8lu kB\n" | |
98 | "Node %d LowTotal: %8lu kB\n" | |
7ee92255 KM |
99 | "Node %d LowFree: %8lu kB\n", |
100 | nid, K(i.totalhigh), | |
101 | nid, K(i.freehigh), | |
102 | nid, K(i.totalram - i.totalhigh), | |
103 | nid, K(i.freeram - i.freehigh)); | |
182e8e23 | 104 | #endif |
7ee92255 | 105 | n += sprintf(buf + n, |
4f98a2fe RR |
106 | "Node %d Dirty: %8lu kB\n" |
107 | "Node %d Writeback: %8lu kB\n" | |
108 | "Node %d FilePages: %8lu kB\n" | |
109 | "Node %d Mapped: %8lu kB\n" | |
110 | "Node %d AnonPages: %8lu kB\n" | |
4b02108a | 111 | "Node %d Shmem: %8lu kB\n" |
c6a7f572 | 112 | "Node %d KernelStack: %8lu kB\n" |
4f98a2fe RR |
113 | "Node %d PageTables: %8lu kB\n" |
114 | "Node %d NFS_Unstable: %8lu kB\n" | |
115 | "Node %d Bounce: %8lu kB\n" | |
116 | "Node %d WritebackTmp: %8lu kB\n" | |
117 | "Node %d Slab: %8lu kB\n" | |
118 | "Node %d SReclaimable: %8lu kB\n" | |
05b258e9 DR |
119 | "Node %d SUnreclaim: %8lu kB\n" |
120 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE | |
121 | "Node %d AnonHugePages: %8lu kB\n" | |
122 | #endif | |
123 | , | |
b1e7a8fd | 124 | nid, K(node_page_state(nid, NR_FILE_DIRTY)), |
ce866b34 | 125 | nid, K(node_page_state(nid, NR_WRITEBACK)), |
347ce434 | 126 | nid, K(node_page_state(nid, NR_FILE_PAGES)), |
65ba55f5 | 127 | nid, K(node_page_state(nid, NR_FILE_MAPPED)), |
91a13c28 | 128 | nid, K(node_page_state(nid, NR_ANON_PAGES)), |
cc7452b6 | 129 | nid, K(i.sharedram), |
c6a7f572 KM |
130 | nid, node_page_state(nid, NR_KERNEL_STACK) * |
131 | THREAD_SIZE / 1024, | |
df849a15 | 132 | nid, K(node_page_state(nid, NR_PAGETABLE)), |
fd39fc85 | 133 | nid, K(node_page_state(nid, NR_UNSTABLE_NFS)), |
d2c5e30c | 134 | nid, K(node_page_state(nid, NR_BOUNCE)), |
fc3ba692 | 135 | nid, K(node_page_state(nid, NR_WRITEBACK_TEMP)), |
972d1a7b CL |
136 | nid, K(node_page_state(nid, NR_SLAB_RECLAIMABLE) + |
137 | node_page_state(nid, NR_SLAB_UNRECLAIMABLE)), | |
138 | nid, K(node_page_state(nid, NR_SLAB_RECLAIMABLE)), | |
05b258e9 | 139 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
91a13c28 | 140 | nid, K(node_page_state(nid, NR_SLAB_UNRECLAIMABLE)) |
05b258e9 DR |
141 | , nid, |
142 | K(node_page_state(nid, NR_ANON_TRANSPARENT_HUGEPAGES) * | |
91a13c28 CS |
143 | HPAGE_PMD_NR)); |
144 | #else | |
145 | nid, K(node_page_state(nid, NR_SLAB_UNRECLAIMABLE))); | |
05b258e9 | 146 | #endif |
1da177e4 LT |
147 | n += hugetlb_report_node_meminfo(nid, buf + n); |
148 | return n; | |
149 | } | |
150 | ||
151 | #undef K | |
10fbcf4c | 152 | static DEVICE_ATTR(meminfo, S_IRUGO, node_read_meminfo, NULL); |
1da177e4 | 153 | |
10fbcf4c KS |
154 | static ssize_t node_read_numastat(struct device *dev, |
155 | struct device_attribute *attr, char *buf) | |
1da177e4 | 156 | { |
1da177e4 LT |
157 | return sprintf(buf, |
158 | "numa_hit %lu\n" | |
159 | "numa_miss %lu\n" | |
160 | "numa_foreign %lu\n" | |
161 | "interleave_hit %lu\n" | |
162 | "local_node %lu\n" | |
163 | "other_node %lu\n", | |
ca889e6c CL |
164 | node_page_state(dev->id, NUMA_HIT), |
165 | node_page_state(dev->id, NUMA_MISS), | |
166 | node_page_state(dev->id, NUMA_FOREIGN), | |
167 | node_page_state(dev->id, NUMA_INTERLEAVE_HIT), | |
168 | node_page_state(dev->id, NUMA_LOCAL), | |
169 | node_page_state(dev->id, NUMA_OTHER)); | |
1da177e4 | 170 | } |
10fbcf4c | 171 | static DEVICE_ATTR(numastat, S_IRUGO, node_read_numastat, NULL); |
1da177e4 | 172 | |
10fbcf4c KS |
173 | static ssize_t node_read_vmstat(struct device *dev, |
174 | struct device_attribute *attr, char *buf) | |
2ac39037 MR |
175 | { |
176 | int nid = dev->id; | |
fa25c503 KM |
177 | int i; |
178 | int n = 0; | |
179 | ||
180 | for (i = 0; i < NR_VM_ZONE_STAT_ITEMS; i++) | |
181 | n += sprintf(buf+n, "%s %lu\n", vmstat_text[i], | |
182 | node_page_state(nid, i)); | |
183 | ||
184 | return n; | |
2ac39037 | 185 | } |
10fbcf4c | 186 | static DEVICE_ATTR(vmstat, S_IRUGO, node_read_vmstat, NULL); |
2ac39037 | 187 | |
10fbcf4c KS |
188 | static ssize_t node_read_distance(struct device *dev, |
189 | struct device_attribute *attr, char * buf) | |
1da177e4 LT |
190 | { |
191 | int nid = dev->id; | |
192 | int len = 0; | |
193 | int i; | |
194 | ||
12ee3c0a DR |
195 | /* |
196 | * buf is currently PAGE_SIZE in length and each node needs 4 chars | |
197 | * at the most (distance + space or newline). | |
198 | */ | |
199 | BUILD_BUG_ON(MAX_NUMNODES * 4 > PAGE_SIZE); | |
1da177e4 LT |
200 | |
201 | for_each_online_node(i) | |
202 | len += sprintf(buf + len, "%s%d", i ? " " : "", node_distance(nid, i)); | |
203 | ||
204 | len += sprintf(buf + len, "\n"); | |
205 | return len; | |
206 | } | |
10fbcf4c | 207 | static DEVICE_ATTR(distance, S_IRUGO, node_read_distance, NULL); |
1da177e4 | 208 | |
9a305230 LS |
209 | #ifdef CONFIG_HUGETLBFS |
210 | /* | |
211 | * hugetlbfs per node attributes registration interface: | |
212 | * When/if hugetlb[fs] subsystem initializes [sometime after this module], | |
4faf8d95 LS |
213 | * it will register its per node attributes for all online nodes with |
214 | * memory. It will also call register_hugetlbfs_with_node(), below, to | |
9a305230 LS |
215 | * register its attribute registration functions with this node driver. |
216 | * Once these hooks have been initialized, the node driver will call into | |
217 | * the hugetlb module to [un]register attributes for hot-plugged nodes. | |
218 | */ | |
219 | static node_registration_func_t __hugetlb_register_node; | |
220 | static node_registration_func_t __hugetlb_unregister_node; | |
221 | ||
39da08cb | 222 | static inline bool hugetlb_register_node(struct node *node) |
9a305230 | 223 | { |
4faf8d95 | 224 | if (__hugetlb_register_node && |
8cebfcd0 | 225 | node_state(node->dev.id, N_MEMORY)) { |
9a305230 | 226 | __hugetlb_register_node(node); |
39da08cb LS |
227 | return true; |
228 | } | |
229 | return false; | |
9a305230 LS |
230 | } |
231 | ||
232 | static inline void hugetlb_unregister_node(struct node *node) | |
233 | { | |
234 | if (__hugetlb_unregister_node) | |
235 | __hugetlb_unregister_node(node); | |
236 | } | |
237 | ||
238 | void register_hugetlbfs_with_node(node_registration_func_t doregister, | |
239 | node_registration_func_t unregister) | |
240 | { | |
241 | __hugetlb_register_node = doregister; | |
242 | __hugetlb_unregister_node = unregister; | |
243 | } | |
244 | #else | |
245 | static inline void hugetlb_register_node(struct node *node) {} | |
246 | ||
247 | static inline void hugetlb_unregister_node(struct node *node) {} | |
248 | #endif | |
249 | ||
8c7b5b4e YI |
250 | static void node_device_release(struct device *dev) |
251 | { | |
252 | struct node *node = to_node(dev); | |
253 | ||
254 | #if defined(CONFIG_MEMORY_HOTPLUG_SPARSE) && defined(CONFIG_HUGETLBFS) | |
255 | /* | |
256 | * We schedule the work only when a memory section is | |
257 | * onlined/offlined on this node. When we come here, | |
258 | * all the memory on this node has been offlined, | |
259 | * so we won't enqueue new work to this work. | |
260 | * | |
261 | * The work is using node->node_work, so we should | |
262 | * flush work before freeing the memory. | |
263 | */ | |
264 | flush_work(&node->node_work); | |
265 | #endif | |
266 | kfree(node); | |
267 | } | |
1da177e4 LT |
268 | |
269 | /* | |
405ae7d3 | 270 | * register_node - Setup a sysfs device for a node. |
1da177e4 LT |
271 | * @num - Node number to use when creating the device. |
272 | * | |
273 | * Initialize and register the node device. | |
274 | */ | |
fa264375 | 275 | static int register_node(struct node *node, int num, struct node *parent) |
1da177e4 LT |
276 | { |
277 | int error; | |
278 | ||
10fbcf4c KS |
279 | node->dev.id = num; |
280 | node->dev.bus = &node_subsys; | |
8c7b5b4e | 281 | node->dev.release = node_device_release; |
10fbcf4c | 282 | error = device_register(&node->dev); |
1da177e4 LT |
283 | |
284 | if (!error){ | |
10fbcf4c KS |
285 | device_create_file(&node->dev, &dev_attr_cpumap); |
286 | device_create_file(&node->dev, &dev_attr_cpulist); | |
287 | device_create_file(&node->dev, &dev_attr_meminfo); | |
288 | device_create_file(&node->dev, &dev_attr_numastat); | |
289 | device_create_file(&node->dev, &dev_attr_distance); | |
290 | device_create_file(&node->dev, &dev_attr_vmstat); | |
af936a16 | 291 | |
9a305230 | 292 | hugetlb_register_node(node); |
ed4a6d7f MG |
293 | |
294 | compaction_register_node(node); | |
1da177e4 LT |
295 | } |
296 | return error; | |
297 | } | |
298 | ||
4b45099b KT |
299 | /** |
300 | * unregister_node - unregister a node device | |
301 | * @node: node going away | |
302 | * | |
303 | * Unregisters a node device @node. All the devices on the node must be | |
304 | * unregistered before calling this function. | |
305 | */ | |
306 | void unregister_node(struct node *node) | |
307 | { | |
10fbcf4c KS |
308 | device_remove_file(&node->dev, &dev_attr_cpumap); |
309 | device_remove_file(&node->dev, &dev_attr_cpulist); | |
310 | device_remove_file(&node->dev, &dev_attr_meminfo); | |
311 | device_remove_file(&node->dev, &dev_attr_numastat); | |
312 | device_remove_file(&node->dev, &dev_attr_distance); | |
313 | device_remove_file(&node->dev, &dev_attr_vmstat); | |
4b45099b | 314 | |
4faf8d95 | 315 | hugetlb_unregister_node(node); /* no-op, if memoryless node */ |
af936a16 | 316 | |
10fbcf4c | 317 | device_unregister(&node->dev); |
4b45099b | 318 | } |
1da177e4 | 319 | |
8732794b | 320 | struct node *node_devices[MAX_NUMNODES]; |
0fc44159 | 321 | |
76b67ed9 KH |
322 | /* |
323 | * register cpu under node | |
324 | */ | |
325 | int register_cpu_under_node(unsigned int cpu, unsigned int nid) | |
326 | { | |
1830794a | 327 | int ret; |
8a25a2fd | 328 | struct device *obj; |
76b67ed9 | 329 | |
f8246f31 AC |
330 | if (!node_online(nid)) |
331 | return 0; | |
332 | ||
8a25a2fd | 333 | obj = get_cpu_device(cpu); |
f8246f31 AC |
334 | if (!obj) |
335 | return 0; | |
336 | ||
8732794b | 337 | ret = sysfs_create_link(&node_devices[nid]->dev.kobj, |
f8246f31 AC |
338 | &obj->kobj, |
339 | kobject_name(&obj->kobj)); | |
1830794a AC |
340 | if (ret) |
341 | return ret; | |
342 | ||
343 | return sysfs_create_link(&obj->kobj, | |
8732794b WC |
344 | &node_devices[nid]->dev.kobj, |
345 | kobject_name(&node_devices[nid]->dev.kobj)); | |
76b67ed9 KH |
346 | } |
347 | ||
348 | int unregister_cpu_under_node(unsigned int cpu, unsigned int nid) | |
349 | { | |
8a25a2fd | 350 | struct device *obj; |
b9d52dad AC |
351 | |
352 | if (!node_online(nid)) | |
353 | return 0; | |
354 | ||
8a25a2fd | 355 | obj = get_cpu_device(cpu); |
b9d52dad AC |
356 | if (!obj) |
357 | return 0; | |
358 | ||
8732794b | 359 | sysfs_remove_link(&node_devices[nid]->dev.kobj, |
b9d52dad | 360 | kobject_name(&obj->kobj)); |
1830794a | 361 | sysfs_remove_link(&obj->kobj, |
8732794b | 362 | kobject_name(&node_devices[nid]->dev.kobj)); |
b9d52dad | 363 | |
76b67ed9 KH |
364 | return 0; |
365 | } | |
366 | ||
c04fc586 GH |
367 | #ifdef CONFIG_MEMORY_HOTPLUG_SPARSE |
368 | #define page_initialized(page) (page->lru.next) | |
369 | ||
370 | static int get_nid_for_pfn(unsigned long pfn) | |
371 | { | |
372 | struct page *page; | |
373 | ||
374 | if (!pfn_valid_within(pfn)) | |
375 | return -1; | |
376 | page = pfn_to_page(pfn); | |
377 | if (!page_initialized(page)) | |
378 | return -1; | |
379 | return pfn_to_nid(pfn); | |
380 | } | |
381 | ||
382 | /* register memory section under specified node if it spans that node */ | |
383 | int register_mem_sect_under_node(struct memory_block *mem_blk, int nid) | |
384 | { | |
dee5d0d5 | 385 | int ret; |
c04fc586 GH |
386 | unsigned long pfn, sect_start_pfn, sect_end_pfn; |
387 | ||
388 | if (!mem_blk) | |
389 | return -EFAULT; | |
390 | if (!node_online(nid)) | |
391 | return 0; | |
d3360164 NF |
392 | |
393 | sect_start_pfn = section_nr_to_pfn(mem_blk->start_section_nr); | |
394 | sect_end_pfn = section_nr_to_pfn(mem_blk->end_section_nr); | |
395 | sect_end_pfn += PAGES_PER_SECTION - 1; | |
c04fc586 GH |
396 | for (pfn = sect_start_pfn; pfn <= sect_end_pfn; pfn++) { |
397 | int page_nid; | |
398 | ||
399 | page_nid = get_nid_for_pfn(pfn); | |
400 | if (page_nid < 0) | |
401 | continue; | |
402 | if (page_nid != nid) | |
403 | continue; | |
8732794b | 404 | ret = sysfs_create_link_nowarn(&node_devices[nid]->dev.kobj, |
10fbcf4c KS |
405 | &mem_blk->dev.kobj, |
406 | kobject_name(&mem_blk->dev.kobj)); | |
dee5d0d5 AC |
407 | if (ret) |
408 | return ret; | |
409 | ||
10fbcf4c | 410 | return sysfs_create_link_nowarn(&mem_blk->dev.kobj, |
8732794b WC |
411 | &node_devices[nid]->dev.kobj, |
412 | kobject_name(&node_devices[nid]->dev.kobj)); | |
c04fc586 GH |
413 | } |
414 | /* mem section does not span the specified node */ | |
415 | return 0; | |
416 | } | |
417 | ||
418 | /* unregister memory section under all nodes that it spans */ | |
d3360164 NF |
419 | int unregister_mem_sect_under_nodes(struct memory_block *mem_blk, |
420 | unsigned long phys_index) | |
c04fc586 | 421 | { |
9ae49fab | 422 | NODEMASK_ALLOC(nodemask_t, unlinked_nodes, GFP_KERNEL); |
c04fc586 GH |
423 | unsigned long pfn, sect_start_pfn, sect_end_pfn; |
424 | ||
9ae49fab DR |
425 | if (!mem_blk) { |
426 | NODEMASK_FREE(unlinked_nodes); | |
c04fc586 | 427 | return -EFAULT; |
9ae49fab DR |
428 | } |
429 | if (!unlinked_nodes) | |
430 | return -ENOMEM; | |
431 | nodes_clear(*unlinked_nodes); | |
d3360164 NF |
432 | |
433 | sect_start_pfn = section_nr_to_pfn(phys_index); | |
c04fc586 GH |
434 | sect_end_pfn = sect_start_pfn + PAGES_PER_SECTION - 1; |
435 | for (pfn = sect_start_pfn; pfn <= sect_end_pfn; pfn++) { | |
47504980 | 436 | int nid; |
c04fc586 GH |
437 | |
438 | nid = get_nid_for_pfn(pfn); | |
439 | if (nid < 0) | |
440 | continue; | |
441 | if (!node_online(nid)) | |
442 | continue; | |
9ae49fab | 443 | if (node_test_and_set(nid, *unlinked_nodes)) |
c04fc586 | 444 | continue; |
8732794b | 445 | sysfs_remove_link(&node_devices[nid]->dev.kobj, |
10fbcf4c KS |
446 | kobject_name(&mem_blk->dev.kobj)); |
447 | sysfs_remove_link(&mem_blk->dev.kobj, | |
8732794b | 448 | kobject_name(&node_devices[nid]->dev.kobj)); |
c04fc586 | 449 | } |
9ae49fab | 450 | NODEMASK_FREE(unlinked_nodes); |
c04fc586 GH |
451 | return 0; |
452 | } | |
453 | ||
454 | static int link_mem_sections(int nid) | |
455 | { | |
456 | unsigned long start_pfn = NODE_DATA(nid)->node_start_pfn; | |
457 | unsigned long end_pfn = start_pfn + NODE_DATA(nid)->node_spanned_pages; | |
458 | unsigned long pfn; | |
63d027a6 | 459 | struct memory_block *mem_blk = NULL; |
c04fc586 GH |
460 | int err = 0; |
461 | ||
462 | for (pfn = start_pfn; pfn < end_pfn; pfn += PAGES_PER_SECTION) { | |
463 | unsigned long section_nr = pfn_to_section_nr(pfn); | |
464 | struct mem_section *mem_sect; | |
c04fc586 GH |
465 | int ret; |
466 | ||
467 | if (!present_section_nr(section_nr)) | |
468 | continue; | |
469 | mem_sect = __nr_to_section(section_nr); | |
321bf4ed YL |
470 | |
471 | /* same memblock ? */ | |
472 | if (mem_blk) | |
473 | if ((section_nr >= mem_blk->start_section_nr) && | |
474 | (section_nr <= mem_blk->end_section_nr)) | |
475 | continue; | |
476 | ||
63d027a6 | 477 | mem_blk = find_memory_block_hinted(mem_sect, mem_blk); |
321bf4ed | 478 | |
c04fc586 GH |
479 | ret = register_mem_sect_under_node(mem_blk, nid); |
480 | if (!err) | |
481 | err = ret; | |
482 | ||
483 | /* discard ref obtained in find_memory_block() */ | |
c04fc586 | 484 | } |
63d027a6 RH |
485 | |
486 | if (mem_blk) | |
10fbcf4c | 487 | kobject_put(&mem_blk->dev.kobj); |
c04fc586 GH |
488 | return err; |
489 | } | |
4faf8d95 | 490 | |
39da08cb | 491 | #ifdef CONFIG_HUGETLBFS |
4faf8d95 LS |
492 | /* |
493 | * Handle per node hstate attribute [un]registration on transistions | |
494 | * to/from memoryless state. | |
495 | */ | |
39da08cb LS |
496 | static void node_hugetlb_work(struct work_struct *work) |
497 | { | |
498 | struct node *node = container_of(work, struct node, node_work); | |
499 | ||
500 | /* | |
501 | * We only get here when a node transitions to/from memoryless state. | |
502 | * We can detect which transition occurred by examining whether the | |
503 | * node has memory now. hugetlb_register_node() already check this | |
504 | * so we try to register the attributes. If that fails, then the | |
505 | * node has transitioned to memoryless, try to unregister the | |
506 | * attributes. | |
507 | */ | |
508 | if (!hugetlb_register_node(node)) | |
509 | hugetlb_unregister_node(node); | |
510 | } | |
511 | ||
512 | static void init_node_hugetlb_work(int nid) | |
513 | { | |
8732794b | 514 | INIT_WORK(&node_devices[nid]->node_work, node_hugetlb_work); |
39da08cb | 515 | } |
4faf8d95 LS |
516 | |
517 | static int node_memory_callback(struct notifier_block *self, | |
518 | unsigned long action, void *arg) | |
519 | { | |
520 | struct memory_notify *mnb = arg; | |
521 | int nid = mnb->status_change_nid; | |
522 | ||
523 | switch (action) { | |
39da08cb LS |
524 | case MEM_ONLINE: |
525 | case MEM_OFFLINE: | |
526 | /* | |
527 | * offload per node hstate [un]registration to a work thread | |
528 | * when transitioning to/from memoryless state. | |
529 | */ | |
4faf8d95 | 530 | if (nid != NUMA_NO_NODE) |
8732794b | 531 | schedule_work(&node_devices[nid]->node_work); |
4faf8d95 | 532 | break; |
39da08cb | 533 | |
4faf8d95 LS |
534 | case MEM_GOING_ONLINE: |
535 | case MEM_GOING_OFFLINE: | |
536 | case MEM_CANCEL_ONLINE: | |
537 | case MEM_CANCEL_OFFLINE: | |
538 | default: | |
539 | break; | |
540 | } | |
541 | ||
542 | return NOTIFY_OK; | |
543 | } | |
39da08cb LS |
544 | #endif /* CONFIG_HUGETLBFS */ |
545 | #else /* !CONFIG_MEMORY_HOTPLUG_SPARSE */ | |
546 | ||
c04fc586 | 547 | static int link_mem_sections(int nid) { return 0; } |
39da08cb | 548 | #endif /* CONFIG_MEMORY_HOTPLUG_SPARSE */ |
4faf8d95 | 549 | |
39da08cb LS |
550 | #if !defined(CONFIG_MEMORY_HOTPLUG_SPARSE) || \ |
551 | !defined(CONFIG_HUGETLBFS) | |
4faf8d95 LS |
552 | static inline int node_memory_callback(struct notifier_block *self, |
553 | unsigned long action, void *arg) | |
554 | { | |
555 | return NOTIFY_OK; | |
556 | } | |
39da08cb LS |
557 | |
558 | static void init_node_hugetlb_work(int nid) { } | |
559 | ||
560 | #endif | |
c04fc586 | 561 | |
0fc44159 YG |
562 | int register_one_node(int nid) |
563 | { | |
564 | int error = 0; | |
76b67ed9 | 565 | int cpu; |
0fc44159 YG |
566 | |
567 | if (node_online(nid)) { | |
568 | int p_node = parent_node(nid); | |
569 | struct node *parent = NULL; | |
570 | ||
571 | if (p_node != nid) | |
8732794b | 572 | parent = node_devices[p_node]; |
0fc44159 | 573 | |
8732794b WC |
574 | node_devices[nid] = kzalloc(sizeof(struct node), GFP_KERNEL); |
575 | if (!node_devices[nid]) | |
576 | return -ENOMEM; | |
577 | ||
578 | error = register_node(node_devices[nid], nid, parent); | |
76b67ed9 KH |
579 | |
580 | /* link cpu under this node */ | |
581 | for_each_present_cpu(cpu) { | |
582 | if (cpu_to_node(cpu) == nid) | |
583 | register_cpu_under_node(cpu, nid); | |
584 | } | |
c04fc586 GH |
585 | |
586 | /* link memory sections under this node */ | |
587 | error = link_mem_sections(nid); | |
39da08cb LS |
588 | |
589 | /* initialize work queue for memory hot plug */ | |
590 | init_node_hugetlb_work(nid); | |
0fc44159 YG |
591 | } |
592 | ||
593 | return error; | |
594 | ||
595 | } | |
596 | ||
597 | void unregister_one_node(int nid) | |
598 | { | |
92d585ef XQ |
599 | if (!node_devices[nid]) |
600 | return; | |
601 | ||
8732794b | 602 | unregister_node(node_devices[nid]); |
8732794b | 603 | node_devices[nid] = NULL; |
0fc44159 YG |
604 | } |
605 | ||
bde631a5 LS |
606 | /* |
607 | * node states attributes | |
608 | */ | |
609 | ||
610 | static ssize_t print_nodes_state(enum node_states state, char *buf) | |
611 | { | |
612 | int n; | |
613 | ||
f6238818 RO |
614 | n = nodelist_scnprintf(buf, PAGE_SIZE-2, node_states[state]); |
615 | buf[n++] = '\n'; | |
616 | buf[n] = '\0'; | |
bde631a5 LS |
617 | return n; |
618 | } | |
619 | ||
b15f562f | 620 | struct node_attr { |
10fbcf4c | 621 | struct device_attribute attr; |
b15f562f AK |
622 | enum node_states state; |
623 | }; | |
bde631a5 | 624 | |
10fbcf4c KS |
625 | static ssize_t show_node_state(struct device *dev, |
626 | struct device_attribute *attr, char *buf) | |
bde631a5 | 627 | { |
b15f562f AK |
628 | struct node_attr *na = container_of(attr, struct node_attr, attr); |
629 | return print_nodes_state(na->state, buf); | |
bde631a5 LS |
630 | } |
631 | ||
b15f562f | 632 | #define _NODE_ATTR(name, state) \ |
10fbcf4c | 633 | { __ATTR(name, 0444, show_node_state, NULL), state } |
bde631a5 | 634 | |
b15f562f | 635 | static struct node_attr node_state_attr[] = { |
fcf07d22 LJ |
636 | [N_POSSIBLE] = _NODE_ATTR(possible, N_POSSIBLE), |
637 | [N_ONLINE] = _NODE_ATTR(online, N_ONLINE), | |
638 | [N_NORMAL_MEMORY] = _NODE_ATTR(has_normal_memory, N_NORMAL_MEMORY), | |
bde631a5 | 639 | #ifdef CONFIG_HIGHMEM |
fcf07d22 | 640 | [N_HIGH_MEMORY] = _NODE_ATTR(has_high_memory, N_HIGH_MEMORY), |
20b2f52b LJ |
641 | #endif |
642 | #ifdef CONFIG_MOVABLE_NODE | |
643 | [N_MEMORY] = _NODE_ATTR(has_memory, N_MEMORY), | |
bde631a5 | 644 | #endif |
fcf07d22 | 645 | [N_CPU] = _NODE_ATTR(has_cpu, N_CPU), |
bde631a5 LS |
646 | }; |
647 | ||
10fbcf4c | 648 | static struct attribute *node_state_attrs[] = { |
fcf07d22 LJ |
649 | &node_state_attr[N_POSSIBLE].attr.attr, |
650 | &node_state_attr[N_ONLINE].attr.attr, | |
651 | &node_state_attr[N_NORMAL_MEMORY].attr.attr, | |
3701cde6 | 652 | #ifdef CONFIG_HIGHMEM |
fcf07d22 | 653 | &node_state_attr[N_HIGH_MEMORY].attr.attr, |
20b2f52b LJ |
654 | #endif |
655 | #ifdef CONFIG_MOVABLE_NODE | |
656 | &node_state_attr[N_MEMORY].attr.attr, | |
3701cde6 | 657 | #endif |
fcf07d22 | 658 | &node_state_attr[N_CPU].attr.attr, |
3701cde6 AK |
659 | NULL |
660 | }; | |
bde631a5 | 661 | |
10fbcf4c KS |
662 | static struct attribute_group memory_root_attr_group = { |
663 | .attrs = node_state_attrs, | |
664 | }; | |
665 | ||
666 | static const struct attribute_group *cpu_root_attr_groups[] = { | |
667 | &memory_root_attr_group, | |
668 | NULL, | |
669 | }; | |
670 | ||
4faf8d95 | 671 | #define NODE_CALLBACK_PRI 2 /* lower than SLAB */ |
4b45099b | 672 | static int __init register_node_type(void) |
1da177e4 | 673 | { |
bde631a5 LS |
674 | int ret; |
675 | ||
3701cde6 AK |
676 | BUILD_BUG_ON(ARRAY_SIZE(node_state_attr) != NR_NODE_STATES); |
677 | BUILD_BUG_ON(ARRAY_SIZE(node_state_attrs)-1 != NR_NODE_STATES); | |
678 | ||
10fbcf4c | 679 | ret = subsys_system_register(&node_subsys, cpu_root_attr_groups); |
4faf8d95 | 680 | if (!ret) { |
6e259e7d AM |
681 | static struct notifier_block node_memory_callback_nb = { |
682 | .notifier_call = node_memory_callback, | |
683 | .priority = NODE_CALLBACK_PRI, | |
684 | }; | |
685 | register_hotmemory_notifier(&node_memory_callback_nb); | |
4faf8d95 | 686 | } |
bde631a5 LS |
687 | |
688 | /* | |
689 | * Note: we're not going to unregister the node class if we fail | |
690 | * to register the node state class attribute files. | |
691 | */ | |
692 | return ret; | |
1da177e4 LT |
693 | } |
694 | postcore_initcall(register_node_type); |