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 | ||
5aaba363 | 28 | static ssize_t node_read_cpumap(struct device *dev, bool list, 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 | 32 | |
39106dcf MT |
33 | /* 2008/04/07: buf currently PAGE_SIZE, need 9 chars per 32 bits. */ |
34 | BUILD_BUG_ON((NR_CPUS/32 * 9) > (PAGE_SIZE-1)); | |
1da177e4 | 35 | |
5aaba363 | 36 | return cpumap_print_to_pagebuf(list, buf, mask); |
1da177e4 LT |
37 | } |
38 | ||
10fbcf4c KS |
39 | static inline ssize_t node_read_cpumask(struct device *dev, |
40 | struct device_attribute *attr, char *buf) | |
39106dcf | 41 | { |
5aaba363 | 42 | return node_read_cpumap(dev, false, buf); |
39106dcf | 43 | } |
10fbcf4c KS |
44 | static inline ssize_t node_read_cpulist(struct device *dev, |
45 | struct device_attribute *attr, char *buf) | |
39106dcf | 46 | { |
5aaba363 | 47 | return node_read_cpumap(dev, true, buf); |
39106dcf MT |
48 | } |
49 | ||
10fbcf4c KS |
50 | static DEVICE_ATTR(cpumap, S_IRUGO, node_read_cpumask, NULL); |
51 | static DEVICE_ATTR(cpulist, S_IRUGO, node_read_cpulist, NULL); | |
1da177e4 LT |
52 | |
53 | #define K(x) ((x) << (PAGE_SHIFT - 10)) | |
10fbcf4c KS |
54 | static ssize_t node_read_meminfo(struct device *dev, |
55 | struct device_attribute *attr, char *buf) | |
1da177e4 LT |
56 | { |
57 | int n; | |
58 | int nid = dev->id; | |
59 | struct sysinfo i; | |
1da177e4 LT |
60 | |
61 | si_meminfo_node(&i, nid); | |
7ee92255 | 62 | n = sprintf(buf, |
4f98a2fe RR |
63 | "Node %d MemTotal: %8lu kB\n" |
64 | "Node %d MemFree: %8lu kB\n" | |
65 | "Node %d MemUsed: %8lu kB\n" | |
66 | "Node %d Active: %8lu kB\n" | |
67 | "Node %d Inactive: %8lu kB\n" | |
68 | "Node %d Active(anon): %8lu kB\n" | |
69 | "Node %d Inactive(anon): %8lu kB\n" | |
70 | "Node %d Active(file): %8lu kB\n" | |
71 | "Node %d Inactive(file): %8lu kB\n" | |
5344b7e6 | 72 | "Node %d Unevictable: %8lu kB\n" |
7ee92255 KM |
73 | "Node %d Mlocked: %8lu kB\n", |
74 | nid, K(i.totalram), | |
75 | nid, K(i.freeram), | |
76 | nid, K(i.totalram - i.freeram), | |
77 | nid, K(node_page_state(nid, NR_ACTIVE_ANON) + | |
78 | node_page_state(nid, NR_ACTIVE_FILE)), | |
79 | nid, K(node_page_state(nid, NR_INACTIVE_ANON) + | |
80 | node_page_state(nid, NR_INACTIVE_FILE)), | |
81 | nid, K(node_page_state(nid, NR_ACTIVE_ANON)), | |
82 | nid, K(node_page_state(nid, NR_INACTIVE_ANON)), | |
83 | nid, K(node_page_state(nid, NR_ACTIVE_FILE)), | |
84 | nid, K(node_page_state(nid, NR_INACTIVE_FILE)), | |
85 | nid, K(node_page_state(nid, NR_UNEVICTABLE)), | |
86 | nid, K(node_page_state(nid, NR_MLOCK))); | |
87 | ||
182e8e23 | 88 | #ifdef CONFIG_HIGHMEM |
7ee92255 | 89 | n += sprintf(buf + n, |
4f98a2fe RR |
90 | "Node %d HighTotal: %8lu kB\n" |
91 | "Node %d HighFree: %8lu kB\n" | |
92 | "Node %d LowTotal: %8lu kB\n" | |
7ee92255 KM |
93 | "Node %d LowFree: %8lu kB\n", |
94 | nid, K(i.totalhigh), | |
95 | nid, K(i.freehigh), | |
96 | nid, K(i.totalram - i.totalhigh), | |
97 | nid, K(i.freeram - i.freehigh)); | |
182e8e23 | 98 | #endif |
7ee92255 | 99 | n += sprintf(buf + n, |
4f98a2fe RR |
100 | "Node %d Dirty: %8lu kB\n" |
101 | "Node %d Writeback: %8lu kB\n" | |
102 | "Node %d FilePages: %8lu kB\n" | |
103 | "Node %d Mapped: %8lu kB\n" | |
104 | "Node %d AnonPages: %8lu kB\n" | |
4b02108a | 105 | "Node %d Shmem: %8lu kB\n" |
c6a7f572 | 106 | "Node %d KernelStack: %8lu kB\n" |
4f98a2fe RR |
107 | "Node %d PageTables: %8lu kB\n" |
108 | "Node %d NFS_Unstable: %8lu kB\n" | |
109 | "Node %d Bounce: %8lu kB\n" | |
110 | "Node %d WritebackTmp: %8lu kB\n" | |
111 | "Node %d Slab: %8lu kB\n" | |
112 | "Node %d SReclaimable: %8lu kB\n" | |
05b258e9 DR |
113 | "Node %d SUnreclaim: %8lu kB\n" |
114 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE | |
115 | "Node %d AnonHugePages: %8lu kB\n" | |
116 | #endif | |
117 | , | |
b1e7a8fd | 118 | nid, K(node_page_state(nid, NR_FILE_DIRTY)), |
ce866b34 | 119 | nid, K(node_page_state(nid, NR_WRITEBACK)), |
347ce434 | 120 | nid, K(node_page_state(nid, NR_FILE_PAGES)), |
65ba55f5 | 121 | nid, K(node_page_state(nid, NR_FILE_MAPPED)), |
91a13c28 | 122 | nid, K(node_page_state(nid, NR_ANON_PAGES)), |
cc7452b6 | 123 | nid, K(i.sharedram), |
c6a7f572 KM |
124 | nid, node_page_state(nid, NR_KERNEL_STACK) * |
125 | THREAD_SIZE / 1024, | |
df849a15 | 126 | nid, K(node_page_state(nid, NR_PAGETABLE)), |
fd39fc85 | 127 | nid, K(node_page_state(nid, NR_UNSTABLE_NFS)), |
d2c5e30c | 128 | nid, K(node_page_state(nid, NR_BOUNCE)), |
fc3ba692 | 129 | nid, K(node_page_state(nid, NR_WRITEBACK_TEMP)), |
972d1a7b CL |
130 | nid, K(node_page_state(nid, NR_SLAB_RECLAIMABLE) + |
131 | node_page_state(nid, NR_SLAB_UNRECLAIMABLE)), | |
132 | nid, K(node_page_state(nid, NR_SLAB_RECLAIMABLE)), | |
05b258e9 | 133 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
91a13c28 | 134 | nid, K(node_page_state(nid, NR_SLAB_UNRECLAIMABLE)) |
05b258e9 DR |
135 | , nid, |
136 | K(node_page_state(nid, NR_ANON_TRANSPARENT_HUGEPAGES) * | |
91a13c28 CS |
137 | HPAGE_PMD_NR)); |
138 | #else | |
139 | nid, K(node_page_state(nid, NR_SLAB_UNRECLAIMABLE))); | |
05b258e9 | 140 | #endif |
1da177e4 LT |
141 | n += hugetlb_report_node_meminfo(nid, buf + n); |
142 | return n; | |
143 | } | |
144 | ||
145 | #undef K | |
10fbcf4c | 146 | static DEVICE_ATTR(meminfo, S_IRUGO, node_read_meminfo, NULL); |
1da177e4 | 147 | |
10fbcf4c KS |
148 | static ssize_t node_read_numastat(struct device *dev, |
149 | struct device_attribute *attr, char *buf) | |
1da177e4 | 150 | { |
1da177e4 LT |
151 | return sprintf(buf, |
152 | "numa_hit %lu\n" | |
153 | "numa_miss %lu\n" | |
154 | "numa_foreign %lu\n" | |
155 | "interleave_hit %lu\n" | |
156 | "local_node %lu\n" | |
157 | "other_node %lu\n", | |
ca889e6c CL |
158 | node_page_state(dev->id, NUMA_HIT), |
159 | node_page_state(dev->id, NUMA_MISS), | |
160 | node_page_state(dev->id, NUMA_FOREIGN), | |
161 | node_page_state(dev->id, NUMA_INTERLEAVE_HIT), | |
162 | node_page_state(dev->id, NUMA_LOCAL), | |
163 | node_page_state(dev->id, NUMA_OTHER)); | |
1da177e4 | 164 | } |
10fbcf4c | 165 | static DEVICE_ATTR(numastat, S_IRUGO, node_read_numastat, NULL); |
1da177e4 | 166 | |
10fbcf4c KS |
167 | static ssize_t node_read_vmstat(struct device *dev, |
168 | struct device_attribute *attr, char *buf) | |
2ac39037 MR |
169 | { |
170 | int nid = dev->id; | |
fa25c503 KM |
171 | int i; |
172 | int n = 0; | |
173 | ||
174 | for (i = 0; i < NR_VM_ZONE_STAT_ITEMS; i++) | |
175 | n += sprintf(buf+n, "%s %lu\n", vmstat_text[i], | |
176 | node_page_state(nid, i)); | |
177 | ||
178 | return n; | |
2ac39037 | 179 | } |
10fbcf4c | 180 | static DEVICE_ATTR(vmstat, S_IRUGO, node_read_vmstat, NULL); |
2ac39037 | 181 | |
10fbcf4c KS |
182 | static ssize_t node_read_distance(struct device *dev, |
183 | struct device_attribute *attr, char * buf) | |
1da177e4 LT |
184 | { |
185 | int nid = dev->id; | |
186 | int len = 0; | |
187 | int i; | |
188 | ||
12ee3c0a DR |
189 | /* |
190 | * buf is currently PAGE_SIZE in length and each node needs 4 chars | |
191 | * at the most (distance + space or newline). | |
192 | */ | |
193 | BUILD_BUG_ON(MAX_NUMNODES * 4 > PAGE_SIZE); | |
1da177e4 LT |
194 | |
195 | for_each_online_node(i) | |
196 | len += sprintf(buf + len, "%s%d", i ? " " : "", node_distance(nid, i)); | |
197 | ||
198 | len += sprintf(buf + len, "\n"); | |
199 | return len; | |
200 | } | |
10fbcf4c | 201 | static DEVICE_ATTR(distance, S_IRUGO, node_read_distance, NULL); |
1da177e4 | 202 | |
9a305230 LS |
203 | #ifdef CONFIG_HUGETLBFS |
204 | /* | |
205 | * hugetlbfs per node attributes registration interface: | |
206 | * When/if hugetlb[fs] subsystem initializes [sometime after this module], | |
4faf8d95 LS |
207 | * it will register its per node attributes for all online nodes with |
208 | * memory. It will also call register_hugetlbfs_with_node(), below, to | |
9a305230 LS |
209 | * register its attribute registration functions with this node driver. |
210 | * Once these hooks have been initialized, the node driver will call into | |
211 | * the hugetlb module to [un]register attributes for hot-plugged nodes. | |
212 | */ | |
213 | static node_registration_func_t __hugetlb_register_node; | |
214 | static node_registration_func_t __hugetlb_unregister_node; | |
215 | ||
39da08cb | 216 | static inline bool hugetlb_register_node(struct node *node) |
9a305230 | 217 | { |
4faf8d95 | 218 | if (__hugetlb_register_node && |
8cebfcd0 | 219 | node_state(node->dev.id, N_MEMORY)) { |
9a305230 | 220 | __hugetlb_register_node(node); |
39da08cb LS |
221 | return true; |
222 | } | |
223 | return false; | |
9a305230 LS |
224 | } |
225 | ||
226 | static inline void hugetlb_unregister_node(struct node *node) | |
227 | { | |
228 | if (__hugetlb_unregister_node) | |
229 | __hugetlb_unregister_node(node); | |
230 | } | |
231 | ||
232 | void register_hugetlbfs_with_node(node_registration_func_t doregister, | |
233 | node_registration_func_t unregister) | |
234 | { | |
235 | __hugetlb_register_node = doregister; | |
236 | __hugetlb_unregister_node = unregister; | |
237 | } | |
238 | #else | |
239 | static inline void hugetlb_register_node(struct node *node) {} | |
240 | ||
241 | static inline void hugetlb_unregister_node(struct node *node) {} | |
242 | #endif | |
243 | ||
8c7b5b4e YI |
244 | static void node_device_release(struct device *dev) |
245 | { | |
246 | struct node *node = to_node(dev); | |
247 | ||
248 | #if defined(CONFIG_MEMORY_HOTPLUG_SPARSE) && defined(CONFIG_HUGETLBFS) | |
249 | /* | |
250 | * We schedule the work only when a memory section is | |
251 | * onlined/offlined on this node. When we come here, | |
252 | * all the memory on this node has been offlined, | |
253 | * so we won't enqueue new work to this work. | |
254 | * | |
255 | * The work is using node->node_work, so we should | |
256 | * flush work before freeing the memory. | |
257 | */ | |
258 | flush_work(&node->node_work); | |
259 | #endif | |
260 | kfree(node); | |
261 | } | |
1da177e4 LT |
262 | |
263 | /* | |
405ae7d3 | 264 | * register_node - Setup a sysfs device for a node. |
1da177e4 LT |
265 | * @num - Node number to use when creating the device. |
266 | * | |
267 | * Initialize and register the node device. | |
268 | */ | |
fa264375 | 269 | static int register_node(struct node *node, int num, struct node *parent) |
1da177e4 LT |
270 | { |
271 | int error; | |
272 | ||
10fbcf4c KS |
273 | node->dev.id = num; |
274 | node->dev.bus = &node_subsys; | |
8c7b5b4e | 275 | node->dev.release = node_device_release; |
10fbcf4c | 276 | error = device_register(&node->dev); |
1da177e4 LT |
277 | |
278 | if (!error){ | |
10fbcf4c KS |
279 | device_create_file(&node->dev, &dev_attr_cpumap); |
280 | device_create_file(&node->dev, &dev_attr_cpulist); | |
281 | device_create_file(&node->dev, &dev_attr_meminfo); | |
282 | device_create_file(&node->dev, &dev_attr_numastat); | |
283 | device_create_file(&node->dev, &dev_attr_distance); | |
284 | device_create_file(&node->dev, &dev_attr_vmstat); | |
af936a16 | 285 | |
9a305230 | 286 | hugetlb_register_node(node); |
ed4a6d7f MG |
287 | |
288 | compaction_register_node(node); | |
1da177e4 LT |
289 | } |
290 | return error; | |
291 | } | |
292 | ||
4b45099b KT |
293 | /** |
294 | * unregister_node - unregister a node device | |
295 | * @node: node going away | |
296 | * | |
297 | * Unregisters a node device @node. All the devices on the node must be | |
298 | * unregistered before calling this function. | |
299 | */ | |
300 | void unregister_node(struct node *node) | |
301 | { | |
10fbcf4c KS |
302 | device_remove_file(&node->dev, &dev_attr_cpumap); |
303 | device_remove_file(&node->dev, &dev_attr_cpulist); | |
304 | device_remove_file(&node->dev, &dev_attr_meminfo); | |
305 | device_remove_file(&node->dev, &dev_attr_numastat); | |
306 | device_remove_file(&node->dev, &dev_attr_distance); | |
307 | device_remove_file(&node->dev, &dev_attr_vmstat); | |
4b45099b | 308 | |
4faf8d95 | 309 | hugetlb_unregister_node(node); /* no-op, if memoryless node */ |
af936a16 | 310 | |
10fbcf4c | 311 | device_unregister(&node->dev); |
4b45099b | 312 | } |
1da177e4 | 313 | |
8732794b | 314 | struct node *node_devices[MAX_NUMNODES]; |
0fc44159 | 315 | |
76b67ed9 KH |
316 | /* |
317 | * register cpu under node | |
318 | */ | |
319 | int register_cpu_under_node(unsigned int cpu, unsigned int nid) | |
320 | { | |
1830794a | 321 | int ret; |
8a25a2fd | 322 | struct device *obj; |
76b67ed9 | 323 | |
f8246f31 AC |
324 | if (!node_online(nid)) |
325 | return 0; | |
326 | ||
8a25a2fd | 327 | obj = get_cpu_device(cpu); |
f8246f31 AC |
328 | if (!obj) |
329 | return 0; | |
330 | ||
8732794b | 331 | ret = sysfs_create_link(&node_devices[nid]->dev.kobj, |
f8246f31 AC |
332 | &obj->kobj, |
333 | kobject_name(&obj->kobj)); | |
1830794a AC |
334 | if (ret) |
335 | return ret; | |
336 | ||
337 | return sysfs_create_link(&obj->kobj, | |
8732794b WC |
338 | &node_devices[nid]->dev.kobj, |
339 | kobject_name(&node_devices[nid]->dev.kobj)); | |
76b67ed9 KH |
340 | } |
341 | ||
342 | int unregister_cpu_under_node(unsigned int cpu, unsigned int nid) | |
343 | { | |
8a25a2fd | 344 | struct device *obj; |
b9d52dad AC |
345 | |
346 | if (!node_online(nid)) | |
347 | return 0; | |
348 | ||
8a25a2fd | 349 | obj = get_cpu_device(cpu); |
b9d52dad AC |
350 | if (!obj) |
351 | return 0; | |
352 | ||
8732794b | 353 | sysfs_remove_link(&node_devices[nid]->dev.kobj, |
b9d52dad | 354 | kobject_name(&obj->kobj)); |
1830794a | 355 | sysfs_remove_link(&obj->kobj, |
8732794b | 356 | kobject_name(&node_devices[nid]->dev.kobj)); |
b9d52dad | 357 | |
76b67ed9 KH |
358 | return 0; |
359 | } | |
360 | ||
c04fc586 GH |
361 | #ifdef CONFIG_MEMORY_HOTPLUG_SPARSE |
362 | #define page_initialized(page) (page->lru.next) | |
363 | ||
364 | static int get_nid_for_pfn(unsigned long pfn) | |
365 | { | |
366 | struct page *page; | |
367 | ||
368 | if (!pfn_valid_within(pfn)) | |
369 | return -1; | |
370 | page = pfn_to_page(pfn); | |
371 | if (!page_initialized(page)) | |
372 | return -1; | |
373 | return pfn_to_nid(pfn); | |
374 | } | |
375 | ||
376 | /* register memory section under specified node if it spans that node */ | |
377 | int register_mem_sect_under_node(struct memory_block *mem_blk, int nid) | |
378 | { | |
dee5d0d5 | 379 | int ret; |
c04fc586 GH |
380 | unsigned long pfn, sect_start_pfn, sect_end_pfn; |
381 | ||
382 | if (!mem_blk) | |
383 | return -EFAULT; | |
384 | if (!node_online(nid)) | |
385 | return 0; | |
d3360164 NF |
386 | |
387 | sect_start_pfn = section_nr_to_pfn(mem_blk->start_section_nr); | |
388 | sect_end_pfn = section_nr_to_pfn(mem_blk->end_section_nr); | |
389 | sect_end_pfn += PAGES_PER_SECTION - 1; | |
c04fc586 GH |
390 | for (pfn = sect_start_pfn; pfn <= sect_end_pfn; pfn++) { |
391 | int page_nid; | |
392 | ||
393 | page_nid = get_nid_for_pfn(pfn); | |
394 | if (page_nid < 0) | |
395 | continue; | |
396 | if (page_nid != nid) | |
397 | continue; | |
8732794b | 398 | ret = sysfs_create_link_nowarn(&node_devices[nid]->dev.kobj, |
10fbcf4c KS |
399 | &mem_blk->dev.kobj, |
400 | kobject_name(&mem_blk->dev.kobj)); | |
dee5d0d5 AC |
401 | if (ret) |
402 | return ret; | |
403 | ||
10fbcf4c | 404 | return sysfs_create_link_nowarn(&mem_blk->dev.kobj, |
8732794b WC |
405 | &node_devices[nid]->dev.kobj, |
406 | kobject_name(&node_devices[nid]->dev.kobj)); | |
c04fc586 GH |
407 | } |
408 | /* mem section does not span the specified node */ | |
409 | return 0; | |
410 | } | |
411 | ||
412 | /* unregister memory section under all nodes that it spans */ | |
d3360164 NF |
413 | int unregister_mem_sect_under_nodes(struct memory_block *mem_blk, |
414 | unsigned long phys_index) | |
c04fc586 | 415 | { |
9ae49fab | 416 | NODEMASK_ALLOC(nodemask_t, unlinked_nodes, GFP_KERNEL); |
c04fc586 GH |
417 | unsigned long pfn, sect_start_pfn, sect_end_pfn; |
418 | ||
9ae49fab DR |
419 | if (!mem_blk) { |
420 | NODEMASK_FREE(unlinked_nodes); | |
c04fc586 | 421 | return -EFAULT; |
9ae49fab DR |
422 | } |
423 | if (!unlinked_nodes) | |
424 | return -ENOMEM; | |
425 | nodes_clear(*unlinked_nodes); | |
d3360164 NF |
426 | |
427 | sect_start_pfn = section_nr_to_pfn(phys_index); | |
c04fc586 GH |
428 | sect_end_pfn = sect_start_pfn + PAGES_PER_SECTION - 1; |
429 | for (pfn = sect_start_pfn; pfn <= sect_end_pfn; pfn++) { | |
47504980 | 430 | int nid; |
c04fc586 GH |
431 | |
432 | nid = get_nid_for_pfn(pfn); | |
433 | if (nid < 0) | |
434 | continue; | |
435 | if (!node_online(nid)) | |
436 | continue; | |
9ae49fab | 437 | if (node_test_and_set(nid, *unlinked_nodes)) |
c04fc586 | 438 | continue; |
8732794b | 439 | sysfs_remove_link(&node_devices[nid]->dev.kobj, |
10fbcf4c KS |
440 | kobject_name(&mem_blk->dev.kobj)); |
441 | sysfs_remove_link(&mem_blk->dev.kobj, | |
8732794b | 442 | kobject_name(&node_devices[nid]->dev.kobj)); |
c04fc586 | 443 | } |
9ae49fab | 444 | NODEMASK_FREE(unlinked_nodes); |
c04fc586 GH |
445 | return 0; |
446 | } | |
447 | ||
448 | static int link_mem_sections(int nid) | |
449 | { | |
450 | unsigned long start_pfn = NODE_DATA(nid)->node_start_pfn; | |
451 | unsigned long end_pfn = start_pfn + NODE_DATA(nid)->node_spanned_pages; | |
452 | unsigned long pfn; | |
63d027a6 | 453 | struct memory_block *mem_blk = NULL; |
c04fc586 GH |
454 | int err = 0; |
455 | ||
456 | for (pfn = start_pfn; pfn < end_pfn; pfn += PAGES_PER_SECTION) { | |
457 | unsigned long section_nr = pfn_to_section_nr(pfn); | |
458 | struct mem_section *mem_sect; | |
c04fc586 GH |
459 | int ret; |
460 | ||
461 | if (!present_section_nr(section_nr)) | |
462 | continue; | |
463 | mem_sect = __nr_to_section(section_nr); | |
321bf4ed YL |
464 | |
465 | /* same memblock ? */ | |
466 | if (mem_blk) | |
467 | if ((section_nr >= mem_blk->start_section_nr) && | |
468 | (section_nr <= mem_blk->end_section_nr)) | |
469 | continue; | |
470 | ||
63d027a6 | 471 | mem_blk = find_memory_block_hinted(mem_sect, mem_blk); |
321bf4ed | 472 | |
c04fc586 GH |
473 | ret = register_mem_sect_under_node(mem_blk, nid); |
474 | if (!err) | |
475 | err = ret; | |
476 | ||
477 | /* discard ref obtained in find_memory_block() */ | |
c04fc586 | 478 | } |
63d027a6 RH |
479 | |
480 | if (mem_blk) | |
10fbcf4c | 481 | kobject_put(&mem_blk->dev.kobj); |
c04fc586 GH |
482 | return err; |
483 | } | |
4faf8d95 | 484 | |
39da08cb | 485 | #ifdef CONFIG_HUGETLBFS |
4faf8d95 LS |
486 | /* |
487 | * Handle per node hstate attribute [un]registration on transistions | |
488 | * to/from memoryless state. | |
489 | */ | |
39da08cb LS |
490 | static void node_hugetlb_work(struct work_struct *work) |
491 | { | |
492 | struct node *node = container_of(work, struct node, node_work); | |
493 | ||
494 | /* | |
495 | * We only get here when a node transitions to/from memoryless state. | |
496 | * We can detect which transition occurred by examining whether the | |
497 | * node has memory now. hugetlb_register_node() already check this | |
498 | * so we try to register the attributes. If that fails, then the | |
499 | * node has transitioned to memoryless, try to unregister the | |
500 | * attributes. | |
501 | */ | |
502 | if (!hugetlb_register_node(node)) | |
503 | hugetlb_unregister_node(node); | |
504 | } | |
505 | ||
506 | static void init_node_hugetlb_work(int nid) | |
507 | { | |
8732794b | 508 | INIT_WORK(&node_devices[nid]->node_work, node_hugetlb_work); |
39da08cb | 509 | } |
4faf8d95 LS |
510 | |
511 | static int node_memory_callback(struct notifier_block *self, | |
512 | unsigned long action, void *arg) | |
513 | { | |
514 | struct memory_notify *mnb = arg; | |
515 | int nid = mnb->status_change_nid; | |
516 | ||
517 | switch (action) { | |
39da08cb LS |
518 | case MEM_ONLINE: |
519 | case MEM_OFFLINE: | |
520 | /* | |
521 | * offload per node hstate [un]registration to a work thread | |
522 | * when transitioning to/from memoryless state. | |
523 | */ | |
4faf8d95 | 524 | if (nid != NUMA_NO_NODE) |
8732794b | 525 | schedule_work(&node_devices[nid]->node_work); |
4faf8d95 | 526 | break; |
39da08cb | 527 | |
4faf8d95 LS |
528 | case MEM_GOING_ONLINE: |
529 | case MEM_GOING_OFFLINE: | |
530 | case MEM_CANCEL_ONLINE: | |
531 | case MEM_CANCEL_OFFLINE: | |
532 | default: | |
533 | break; | |
534 | } | |
535 | ||
536 | return NOTIFY_OK; | |
537 | } | |
39da08cb LS |
538 | #endif /* CONFIG_HUGETLBFS */ |
539 | #else /* !CONFIG_MEMORY_HOTPLUG_SPARSE */ | |
540 | ||
c04fc586 | 541 | static int link_mem_sections(int nid) { return 0; } |
39da08cb | 542 | #endif /* CONFIG_MEMORY_HOTPLUG_SPARSE */ |
4faf8d95 | 543 | |
39da08cb LS |
544 | #if !defined(CONFIG_MEMORY_HOTPLUG_SPARSE) || \ |
545 | !defined(CONFIG_HUGETLBFS) | |
4faf8d95 LS |
546 | static inline int node_memory_callback(struct notifier_block *self, |
547 | unsigned long action, void *arg) | |
548 | { | |
549 | return NOTIFY_OK; | |
550 | } | |
39da08cb LS |
551 | |
552 | static void init_node_hugetlb_work(int nid) { } | |
553 | ||
554 | #endif | |
c04fc586 | 555 | |
0fc44159 YG |
556 | int register_one_node(int nid) |
557 | { | |
558 | int error = 0; | |
76b67ed9 | 559 | int cpu; |
0fc44159 YG |
560 | |
561 | if (node_online(nid)) { | |
562 | int p_node = parent_node(nid); | |
563 | struct node *parent = NULL; | |
564 | ||
565 | if (p_node != nid) | |
8732794b | 566 | parent = node_devices[p_node]; |
0fc44159 | 567 | |
8732794b WC |
568 | node_devices[nid] = kzalloc(sizeof(struct node), GFP_KERNEL); |
569 | if (!node_devices[nid]) | |
570 | return -ENOMEM; | |
571 | ||
572 | error = register_node(node_devices[nid], nid, parent); | |
76b67ed9 KH |
573 | |
574 | /* link cpu under this node */ | |
575 | for_each_present_cpu(cpu) { | |
576 | if (cpu_to_node(cpu) == nid) | |
577 | register_cpu_under_node(cpu, nid); | |
578 | } | |
c04fc586 GH |
579 | |
580 | /* link memory sections under this node */ | |
581 | error = link_mem_sections(nid); | |
39da08cb LS |
582 | |
583 | /* initialize work queue for memory hot plug */ | |
584 | init_node_hugetlb_work(nid); | |
0fc44159 YG |
585 | } |
586 | ||
587 | return error; | |
588 | ||
589 | } | |
590 | ||
591 | void unregister_one_node(int nid) | |
592 | { | |
92d585ef XQ |
593 | if (!node_devices[nid]) |
594 | return; | |
595 | ||
8732794b | 596 | unregister_node(node_devices[nid]); |
8732794b | 597 | node_devices[nid] = NULL; |
0fc44159 YG |
598 | } |
599 | ||
bde631a5 LS |
600 | /* |
601 | * node states attributes | |
602 | */ | |
603 | ||
604 | static ssize_t print_nodes_state(enum node_states state, char *buf) | |
605 | { | |
606 | int n; | |
607 | ||
f799b1a7 TH |
608 | n = scnprintf(buf, PAGE_SIZE - 1, "%*pbl", |
609 | nodemask_pr_args(&node_states[state])); | |
f6238818 RO |
610 | buf[n++] = '\n'; |
611 | buf[n] = '\0'; | |
bde631a5 LS |
612 | return n; |
613 | } | |
614 | ||
b15f562f | 615 | struct node_attr { |
10fbcf4c | 616 | struct device_attribute attr; |
b15f562f AK |
617 | enum node_states state; |
618 | }; | |
bde631a5 | 619 | |
10fbcf4c KS |
620 | static ssize_t show_node_state(struct device *dev, |
621 | struct device_attribute *attr, char *buf) | |
bde631a5 | 622 | { |
b15f562f AK |
623 | struct node_attr *na = container_of(attr, struct node_attr, attr); |
624 | return print_nodes_state(na->state, buf); | |
bde631a5 LS |
625 | } |
626 | ||
b15f562f | 627 | #define _NODE_ATTR(name, state) \ |
10fbcf4c | 628 | { __ATTR(name, 0444, show_node_state, NULL), state } |
bde631a5 | 629 | |
b15f562f | 630 | static struct node_attr node_state_attr[] = { |
fcf07d22 LJ |
631 | [N_POSSIBLE] = _NODE_ATTR(possible, N_POSSIBLE), |
632 | [N_ONLINE] = _NODE_ATTR(online, N_ONLINE), | |
633 | [N_NORMAL_MEMORY] = _NODE_ATTR(has_normal_memory, N_NORMAL_MEMORY), | |
bde631a5 | 634 | #ifdef CONFIG_HIGHMEM |
fcf07d22 | 635 | [N_HIGH_MEMORY] = _NODE_ATTR(has_high_memory, N_HIGH_MEMORY), |
20b2f52b LJ |
636 | #endif |
637 | #ifdef CONFIG_MOVABLE_NODE | |
638 | [N_MEMORY] = _NODE_ATTR(has_memory, N_MEMORY), | |
bde631a5 | 639 | #endif |
fcf07d22 | 640 | [N_CPU] = _NODE_ATTR(has_cpu, N_CPU), |
bde631a5 LS |
641 | }; |
642 | ||
10fbcf4c | 643 | static struct attribute *node_state_attrs[] = { |
fcf07d22 LJ |
644 | &node_state_attr[N_POSSIBLE].attr.attr, |
645 | &node_state_attr[N_ONLINE].attr.attr, | |
646 | &node_state_attr[N_NORMAL_MEMORY].attr.attr, | |
3701cde6 | 647 | #ifdef CONFIG_HIGHMEM |
fcf07d22 | 648 | &node_state_attr[N_HIGH_MEMORY].attr.attr, |
20b2f52b LJ |
649 | #endif |
650 | #ifdef CONFIG_MOVABLE_NODE | |
651 | &node_state_attr[N_MEMORY].attr.attr, | |
3701cde6 | 652 | #endif |
fcf07d22 | 653 | &node_state_attr[N_CPU].attr.attr, |
3701cde6 AK |
654 | NULL |
655 | }; | |
bde631a5 | 656 | |
10fbcf4c KS |
657 | static struct attribute_group memory_root_attr_group = { |
658 | .attrs = node_state_attrs, | |
659 | }; | |
660 | ||
661 | static const struct attribute_group *cpu_root_attr_groups[] = { | |
662 | &memory_root_attr_group, | |
663 | NULL, | |
664 | }; | |
665 | ||
4faf8d95 | 666 | #define NODE_CALLBACK_PRI 2 /* lower than SLAB */ |
4b45099b | 667 | static int __init register_node_type(void) |
1da177e4 | 668 | { |
bde631a5 LS |
669 | int ret; |
670 | ||
3701cde6 AK |
671 | BUILD_BUG_ON(ARRAY_SIZE(node_state_attr) != NR_NODE_STATES); |
672 | BUILD_BUG_ON(ARRAY_SIZE(node_state_attrs)-1 != NR_NODE_STATES); | |
673 | ||
10fbcf4c | 674 | ret = subsys_system_register(&node_subsys, cpu_root_attr_groups); |
4faf8d95 | 675 | if (!ret) { |
6e259e7d AM |
676 | static struct notifier_block node_memory_callback_nb = { |
677 | .notifier_call = node_memory_callback, | |
678 | .priority = NODE_CALLBACK_PRI, | |
679 | }; | |
680 | register_hotmemory_notifier(&node_memory_callback_nb); | |
4faf8d95 | 681 | } |
bde631a5 LS |
682 | |
683 | /* | |
684 | * Note: we're not going to unregister the node class if we fail | |
685 | * to register the node state class attribute files. | |
686 | */ | |
687 | return ret; | |
1da177e4 LT |
688 | } |
689 | postcore_initcall(register_node_type); |