Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * drivers/base/node.c - basic Node class support | |
3 | */ | |
4 | ||
5 | #include <linux/sysdev.h> | |
6 | #include <linux/module.h> | |
7 | #include <linux/init.h> | |
8 | #include <linux/mm.h> | |
c04fc586 | 9 | #include <linux/memory.h> |
fa25c503 | 10 | #include <linux/vmstat.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 | |
3701cde6 AK |
22 | static struct sysdev_class_attribute *node_state_attrs[]; |
23 | ||
1da177e4 | 24 | static struct sysdev_class node_class = { |
af5ca3f4 | 25 | .name = "node", |
3701cde6 | 26 | .attrs = node_state_attrs, |
1da177e4 LT |
27 | }; |
28 | ||
29 | ||
39106dcf | 30 | static ssize_t node_read_cpumap(struct sys_device *dev, int type, char *buf) |
1da177e4 LT |
31 | { |
32 | struct node *node_dev = to_node(dev); | |
a70f7302 | 33 | const struct cpumask *mask = cpumask_of_node(node_dev->sysdev.id); |
1da177e4 LT |
34 | int len; |
35 | ||
39106dcf MT |
36 | /* 2008/04/07: buf currently PAGE_SIZE, need 9 chars per 32 bits. */ |
37 | BUILD_BUG_ON((NR_CPUS/32 * 9) > (PAGE_SIZE-1)); | |
1da177e4 | 38 | |
39106dcf | 39 | len = type? |
29c0177e RR |
40 | cpulist_scnprintf(buf, PAGE_SIZE-2, mask) : |
41 | cpumask_scnprintf(buf, PAGE_SIZE-2, mask); | |
c5f59f08 MT |
42 | buf[len++] = '\n'; |
43 | buf[len] = '\0'; | |
1da177e4 LT |
44 | return len; |
45 | } | |
46 | ||
4a0b2b4d AK |
47 | static inline ssize_t node_read_cpumask(struct sys_device *dev, |
48 | struct sysdev_attribute *attr, char *buf) | |
39106dcf MT |
49 | { |
50 | return node_read_cpumap(dev, 0, buf); | |
51 | } | |
4a0b2b4d AK |
52 | static inline ssize_t node_read_cpulist(struct sys_device *dev, |
53 | struct sysdev_attribute *attr, char *buf) | |
39106dcf MT |
54 | { |
55 | return node_read_cpumap(dev, 1, buf); | |
56 | } | |
57 | ||
58 | static SYSDEV_ATTR(cpumap, S_IRUGO, node_read_cpumask, NULL); | |
59 | static SYSDEV_ATTR(cpulist, S_IRUGO, node_read_cpulist, NULL); | |
1da177e4 LT |
60 | |
61 | #define K(x) ((x) << (PAGE_SHIFT - 10)) | |
4a0b2b4d AK |
62 | static ssize_t node_read_meminfo(struct sys_device * dev, |
63 | struct sysdev_attribute *attr, char * buf) | |
1da177e4 LT |
64 | { |
65 | int n; | |
66 | int nid = dev->id; | |
67 | struct sysinfo i; | |
1da177e4 LT |
68 | |
69 | si_meminfo_node(&i, nid); | |
7ee92255 | 70 | n = sprintf(buf, |
4f98a2fe RR |
71 | "Node %d MemTotal: %8lu kB\n" |
72 | "Node %d MemFree: %8lu kB\n" | |
73 | "Node %d MemUsed: %8lu kB\n" | |
74 | "Node %d Active: %8lu kB\n" | |
75 | "Node %d Inactive: %8lu kB\n" | |
76 | "Node %d Active(anon): %8lu kB\n" | |
77 | "Node %d Inactive(anon): %8lu kB\n" | |
78 | "Node %d Active(file): %8lu kB\n" | |
79 | "Node %d Inactive(file): %8lu kB\n" | |
5344b7e6 | 80 | "Node %d Unevictable: %8lu kB\n" |
7ee92255 KM |
81 | "Node %d Mlocked: %8lu kB\n", |
82 | nid, K(i.totalram), | |
83 | nid, K(i.freeram), | |
84 | nid, K(i.totalram - i.freeram), | |
85 | nid, K(node_page_state(nid, NR_ACTIVE_ANON) + | |
86 | node_page_state(nid, NR_ACTIVE_FILE)), | |
87 | nid, K(node_page_state(nid, NR_INACTIVE_ANON) + | |
88 | node_page_state(nid, NR_INACTIVE_FILE)), | |
89 | nid, K(node_page_state(nid, NR_ACTIVE_ANON)), | |
90 | nid, K(node_page_state(nid, NR_INACTIVE_ANON)), | |
91 | nid, K(node_page_state(nid, NR_ACTIVE_FILE)), | |
92 | nid, K(node_page_state(nid, NR_INACTIVE_FILE)), | |
93 | nid, K(node_page_state(nid, NR_UNEVICTABLE)), | |
94 | nid, K(node_page_state(nid, NR_MLOCK))); | |
95 | ||
182e8e23 | 96 | #ifdef CONFIG_HIGHMEM |
7ee92255 | 97 | n += sprintf(buf + n, |
4f98a2fe RR |
98 | "Node %d HighTotal: %8lu kB\n" |
99 | "Node %d HighFree: %8lu kB\n" | |
100 | "Node %d LowTotal: %8lu kB\n" | |
7ee92255 KM |
101 | "Node %d LowFree: %8lu kB\n", |
102 | nid, K(i.totalhigh), | |
103 | nid, K(i.freehigh), | |
104 | nid, K(i.totalram - i.totalhigh), | |
105 | nid, K(i.freeram - i.freehigh)); | |
182e8e23 | 106 | #endif |
7ee92255 | 107 | n += sprintf(buf + n, |
4f98a2fe RR |
108 | "Node %d Dirty: %8lu kB\n" |
109 | "Node %d Writeback: %8lu kB\n" | |
110 | "Node %d FilePages: %8lu kB\n" | |
111 | "Node %d Mapped: %8lu kB\n" | |
112 | "Node %d AnonPages: %8lu kB\n" | |
4b02108a | 113 | "Node %d Shmem: %8lu kB\n" |
c6a7f572 | 114 | "Node %d KernelStack: %8lu kB\n" |
4f98a2fe RR |
115 | "Node %d PageTables: %8lu kB\n" |
116 | "Node %d NFS_Unstable: %8lu kB\n" | |
117 | "Node %d Bounce: %8lu kB\n" | |
118 | "Node %d WritebackTmp: %8lu kB\n" | |
119 | "Node %d Slab: %8lu kB\n" | |
120 | "Node %d SReclaimable: %8lu kB\n" | |
05b258e9 DR |
121 | "Node %d SUnreclaim: %8lu kB\n" |
122 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE | |
123 | "Node %d AnonHugePages: %8lu kB\n" | |
124 | #endif | |
125 | , | |
b1e7a8fd | 126 | nid, K(node_page_state(nid, NR_FILE_DIRTY)), |
ce866b34 | 127 | nid, K(node_page_state(nid, NR_WRITEBACK)), |
347ce434 | 128 | nid, K(node_page_state(nid, NR_FILE_PAGES)), |
65ba55f5 | 129 | nid, K(node_page_state(nid, NR_FILE_MAPPED)), |
05b258e9 | 130 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
91a13c28 | 131 | nid, K(node_page_state(nid, NR_ANON_PAGES) |
05b258e9 | 132 | + node_page_state(nid, NR_ANON_TRANSPARENT_HUGEPAGES) * |
91a13c28 CS |
133 | HPAGE_PMD_NR), |
134 | #else | |
135 | nid, K(node_page_state(nid, NR_ANON_PAGES)), | |
05b258e9 | 136 | #endif |
4b02108a | 137 | nid, K(node_page_state(nid, NR_SHMEM)), |
c6a7f572 KM |
138 | nid, node_page_state(nid, NR_KERNEL_STACK) * |
139 | THREAD_SIZE / 1024, | |
df849a15 | 140 | nid, K(node_page_state(nid, NR_PAGETABLE)), |
fd39fc85 | 141 | nid, K(node_page_state(nid, NR_UNSTABLE_NFS)), |
d2c5e30c | 142 | nid, K(node_page_state(nid, NR_BOUNCE)), |
fc3ba692 | 143 | nid, K(node_page_state(nid, NR_WRITEBACK_TEMP)), |
972d1a7b CL |
144 | nid, K(node_page_state(nid, NR_SLAB_RECLAIMABLE) + |
145 | node_page_state(nid, NR_SLAB_UNRECLAIMABLE)), | |
146 | nid, K(node_page_state(nid, NR_SLAB_RECLAIMABLE)), | |
05b258e9 | 147 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
91a13c28 | 148 | nid, K(node_page_state(nid, NR_SLAB_UNRECLAIMABLE)) |
05b258e9 DR |
149 | , nid, |
150 | K(node_page_state(nid, NR_ANON_TRANSPARENT_HUGEPAGES) * | |
91a13c28 CS |
151 | HPAGE_PMD_NR)); |
152 | #else | |
153 | nid, K(node_page_state(nid, NR_SLAB_UNRECLAIMABLE))); | |
05b258e9 | 154 | #endif |
1da177e4 LT |
155 | n += hugetlb_report_node_meminfo(nid, buf + n); |
156 | return n; | |
157 | } | |
158 | ||
159 | #undef K | |
160 | static SYSDEV_ATTR(meminfo, S_IRUGO, node_read_meminfo, NULL); | |
161 | ||
4a0b2b4d AK |
162 | static ssize_t node_read_numastat(struct sys_device * dev, |
163 | struct sysdev_attribute *attr, char * buf) | |
1da177e4 | 164 | { |
1da177e4 LT |
165 | return sprintf(buf, |
166 | "numa_hit %lu\n" | |
167 | "numa_miss %lu\n" | |
168 | "numa_foreign %lu\n" | |
169 | "interleave_hit %lu\n" | |
170 | "local_node %lu\n" | |
171 | "other_node %lu\n", | |
ca889e6c CL |
172 | node_page_state(dev->id, NUMA_HIT), |
173 | node_page_state(dev->id, NUMA_MISS), | |
174 | node_page_state(dev->id, NUMA_FOREIGN), | |
175 | node_page_state(dev->id, NUMA_INTERLEAVE_HIT), | |
176 | node_page_state(dev->id, NUMA_LOCAL), | |
177 | node_page_state(dev->id, NUMA_OTHER)); | |
1da177e4 LT |
178 | } |
179 | static SYSDEV_ATTR(numastat, S_IRUGO, node_read_numastat, NULL); | |
180 | ||
2ac39037 MR |
181 | static ssize_t node_read_vmstat(struct sys_device *dev, |
182 | struct sysdev_attribute *attr, char *buf) | |
183 | { | |
184 | int nid = dev->id; | |
fa25c503 KM |
185 | int i; |
186 | int n = 0; | |
187 | ||
188 | for (i = 0; i < NR_VM_ZONE_STAT_ITEMS; i++) | |
189 | n += sprintf(buf+n, "%s %lu\n", vmstat_text[i], | |
190 | node_page_state(nid, i)); | |
191 | ||
192 | return n; | |
2ac39037 MR |
193 | } |
194 | static SYSDEV_ATTR(vmstat, S_IRUGO, node_read_vmstat, NULL); | |
195 | ||
4a0b2b4d AK |
196 | static ssize_t node_read_distance(struct sys_device * dev, |
197 | struct sysdev_attribute *attr, char * buf) | |
1da177e4 LT |
198 | { |
199 | int nid = dev->id; | |
200 | int len = 0; | |
201 | int i; | |
202 | ||
12ee3c0a DR |
203 | /* |
204 | * buf is currently PAGE_SIZE in length and each node needs 4 chars | |
205 | * at the most (distance + space or newline). | |
206 | */ | |
207 | BUILD_BUG_ON(MAX_NUMNODES * 4 > PAGE_SIZE); | |
1da177e4 LT |
208 | |
209 | for_each_online_node(i) | |
210 | len += sprintf(buf + len, "%s%d", i ? " " : "", node_distance(nid, i)); | |
211 | ||
212 | len += sprintf(buf + len, "\n"); | |
213 | return len; | |
214 | } | |
215 | static SYSDEV_ATTR(distance, S_IRUGO, node_read_distance, NULL); | |
216 | ||
9a305230 LS |
217 | #ifdef CONFIG_HUGETLBFS |
218 | /* | |
219 | * hugetlbfs per node attributes registration interface: | |
220 | * When/if hugetlb[fs] subsystem initializes [sometime after this module], | |
4faf8d95 LS |
221 | * it will register its per node attributes for all online nodes with |
222 | * memory. It will also call register_hugetlbfs_with_node(), below, to | |
9a305230 LS |
223 | * register its attribute registration functions with this node driver. |
224 | * Once these hooks have been initialized, the node driver will call into | |
225 | * the hugetlb module to [un]register attributes for hot-plugged nodes. | |
226 | */ | |
227 | static node_registration_func_t __hugetlb_register_node; | |
228 | static node_registration_func_t __hugetlb_unregister_node; | |
229 | ||
39da08cb | 230 | static inline bool hugetlb_register_node(struct node *node) |
9a305230 | 231 | { |
4faf8d95 | 232 | if (__hugetlb_register_node && |
39da08cb | 233 | node_state(node->sysdev.id, N_HIGH_MEMORY)) { |
9a305230 | 234 | __hugetlb_register_node(node); |
39da08cb LS |
235 | return true; |
236 | } | |
237 | return false; | |
9a305230 LS |
238 | } |
239 | ||
240 | static inline void hugetlb_unregister_node(struct node *node) | |
241 | { | |
242 | if (__hugetlb_unregister_node) | |
243 | __hugetlb_unregister_node(node); | |
244 | } | |
245 | ||
246 | void register_hugetlbfs_with_node(node_registration_func_t doregister, | |
247 | node_registration_func_t unregister) | |
248 | { | |
249 | __hugetlb_register_node = doregister; | |
250 | __hugetlb_unregister_node = unregister; | |
251 | } | |
252 | #else | |
253 | static inline void hugetlb_register_node(struct node *node) {} | |
254 | ||
255 | static inline void hugetlb_unregister_node(struct node *node) {} | |
256 | #endif | |
257 | ||
1da177e4 LT |
258 | |
259 | /* | |
405ae7d3 | 260 | * register_node - Setup a sysfs device for a node. |
1da177e4 LT |
261 | * @num - Node number to use when creating the device. |
262 | * | |
263 | * Initialize and register the node device. | |
264 | */ | |
4b45099b | 265 | int register_node(struct node *node, int num, struct node *parent) |
1da177e4 LT |
266 | { |
267 | int error; | |
268 | ||
269 | node->sysdev.id = num; | |
270 | node->sysdev.cls = &node_class; | |
271 | error = sysdev_register(&node->sysdev); | |
272 | ||
273 | if (!error){ | |
274 | sysdev_create_file(&node->sysdev, &attr_cpumap); | |
39106dcf | 275 | sysdev_create_file(&node->sysdev, &attr_cpulist); |
1da177e4 LT |
276 | sysdev_create_file(&node->sysdev, &attr_meminfo); |
277 | sysdev_create_file(&node->sysdev, &attr_numastat); | |
278 | sysdev_create_file(&node->sysdev, &attr_distance); | |
2ac39037 | 279 | sysdev_create_file(&node->sysdev, &attr_vmstat); |
af936a16 LS |
280 | |
281 | scan_unevictable_register_node(node); | |
4faf8d95 | 282 | |
9a305230 | 283 | hugetlb_register_node(node); |
ed4a6d7f MG |
284 | |
285 | compaction_register_node(node); | |
1da177e4 LT |
286 | } |
287 | return error; | |
288 | } | |
289 | ||
4b45099b KT |
290 | /** |
291 | * unregister_node - unregister a node device | |
292 | * @node: node going away | |
293 | * | |
294 | * Unregisters a node device @node. All the devices on the node must be | |
295 | * unregistered before calling this function. | |
296 | */ | |
297 | void unregister_node(struct node *node) | |
298 | { | |
299 | sysdev_remove_file(&node->sysdev, &attr_cpumap); | |
39106dcf | 300 | sysdev_remove_file(&node->sysdev, &attr_cpulist); |
4b45099b KT |
301 | sysdev_remove_file(&node->sysdev, &attr_meminfo); |
302 | sysdev_remove_file(&node->sysdev, &attr_numastat); | |
303 | sysdev_remove_file(&node->sysdev, &attr_distance); | |
2ac39037 | 304 | sysdev_remove_file(&node->sysdev, &attr_vmstat); |
4b45099b | 305 | |
af936a16 | 306 | scan_unevictable_unregister_node(node); |
4faf8d95 | 307 | hugetlb_unregister_node(node); /* no-op, if memoryless node */ |
af936a16 | 308 | |
4b45099b KT |
309 | sysdev_unregister(&node->sysdev); |
310 | } | |
1da177e4 | 311 | |
0fc44159 YG |
312 | struct node node_devices[MAX_NUMNODES]; |
313 | ||
76b67ed9 KH |
314 | /* |
315 | * register cpu under node | |
316 | */ | |
317 | int register_cpu_under_node(unsigned int cpu, unsigned int nid) | |
318 | { | |
1830794a | 319 | int ret; |
f8246f31 | 320 | struct sys_device *obj; |
76b67ed9 | 321 | |
f8246f31 AC |
322 | if (!node_online(nid)) |
323 | return 0; | |
324 | ||
325 | obj = get_cpu_sysdev(cpu); | |
326 | if (!obj) | |
327 | return 0; | |
328 | ||
1830794a | 329 | ret = sysfs_create_link(&node_devices[nid].sysdev.kobj, |
f8246f31 AC |
330 | &obj->kobj, |
331 | kobject_name(&obj->kobj)); | |
1830794a AC |
332 | if (ret) |
333 | return ret; | |
334 | ||
335 | return sysfs_create_link(&obj->kobj, | |
336 | &node_devices[nid].sysdev.kobj, | |
337 | kobject_name(&node_devices[nid].sysdev.kobj)); | |
76b67ed9 KH |
338 | } |
339 | ||
340 | int unregister_cpu_under_node(unsigned int cpu, unsigned int nid) | |
341 | { | |
b9d52dad AC |
342 | struct sys_device *obj; |
343 | ||
344 | if (!node_online(nid)) | |
345 | return 0; | |
346 | ||
347 | obj = get_cpu_sysdev(cpu); | |
348 | if (!obj) | |
349 | return 0; | |
350 | ||
351 | sysfs_remove_link(&node_devices[nid].sysdev.kobj, | |
352 | kobject_name(&obj->kobj)); | |
1830794a AC |
353 | sysfs_remove_link(&obj->kobj, |
354 | kobject_name(&node_devices[nid].sysdev.kobj)); | |
b9d52dad | 355 | |
76b67ed9 KH |
356 | return 0; |
357 | } | |
358 | ||
c04fc586 GH |
359 | #ifdef CONFIG_MEMORY_HOTPLUG_SPARSE |
360 | #define page_initialized(page) (page->lru.next) | |
361 | ||
362 | static int get_nid_for_pfn(unsigned long pfn) | |
363 | { | |
364 | struct page *page; | |
365 | ||
366 | if (!pfn_valid_within(pfn)) | |
367 | return -1; | |
368 | page = pfn_to_page(pfn); | |
369 | if (!page_initialized(page)) | |
370 | return -1; | |
371 | return pfn_to_nid(pfn); | |
372 | } | |
373 | ||
374 | /* register memory section under specified node if it spans that node */ | |
375 | int register_mem_sect_under_node(struct memory_block *mem_blk, int nid) | |
376 | { | |
dee5d0d5 | 377 | int ret; |
c04fc586 GH |
378 | unsigned long pfn, sect_start_pfn, sect_end_pfn; |
379 | ||
380 | if (!mem_blk) | |
381 | return -EFAULT; | |
382 | if (!node_online(nid)) | |
383 | return 0; | |
d3360164 NF |
384 | |
385 | sect_start_pfn = section_nr_to_pfn(mem_blk->start_section_nr); | |
386 | sect_end_pfn = section_nr_to_pfn(mem_blk->end_section_nr); | |
387 | sect_end_pfn += PAGES_PER_SECTION - 1; | |
c04fc586 GH |
388 | for (pfn = sect_start_pfn; pfn <= sect_end_pfn; pfn++) { |
389 | int page_nid; | |
390 | ||
391 | page_nid = get_nid_for_pfn(pfn); | |
392 | if (page_nid < 0) | |
393 | continue; | |
394 | if (page_nid != nid) | |
395 | continue; | |
dee5d0d5 | 396 | ret = sysfs_create_link_nowarn(&node_devices[nid].sysdev.kobj, |
c04fc586 GH |
397 | &mem_blk->sysdev.kobj, |
398 | kobject_name(&mem_blk->sysdev.kobj)); | |
dee5d0d5 AC |
399 | if (ret) |
400 | return ret; | |
401 | ||
402 | return sysfs_create_link_nowarn(&mem_blk->sysdev.kobj, | |
403 | &node_devices[nid].sysdev.kobj, | |
404 | kobject_name(&node_devices[nid].sysdev.kobj)); | |
c04fc586 GH |
405 | } |
406 | /* mem section does not span the specified node */ | |
407 | return 0; | |
408 | } | |
409 | ||
410 | /* unregister memory section under all nodes that it spans */ | |
d3360164 NF |
411 | int unregister_mem_sect_under_nodes(struct memory_block *mem_blk, |
412 | unsigned long phys_index) | |
c04fc586 | 413 | { |
9ae49fab | 414 | NODEMASK_ALLOC(nodemask_t, unlinked_nodes, GFP_KERNEL); |
c04fc586 GH |
415 | unsigned long pfn, sect_start_pfn, sect_end_pfn; |
416 | ||
9ae49fab DR |
417 | if (!mem_blk) { |
418 | NODEMASK_FREE(unlinked_nodes); | |
c04fc586 | 419 | return -EFAULT; |
9ae49fab DR |
420 | } |
421 | if (!unlinked_nodes) | |
422 | return -ENOMEM; | |
423 | nodes_clear(*unlinked_nodes); | |
d3360164 NF |
424 | |
425 | sect_start_pfn = section_nr_to_pfn(phys_index); | |
c04fc586 GH |
426 | sect_end_pfn = sect_start_pfn + PAGES_PER_SECTION - 1; |
427 | for (pfn = sect_start_pfn; pfn <= sect_end_pfn; pfn++) { | |
47504980 | 428 | int nid; |
c04fc586 GH |
429 | |
430 | nid = get_nid_for_pfn(pfn); | |
431 | if (nid < 0) | |
432 | continue; | |
433 | if (!node_online(nid)) | |
434 | continue; | |
9ae49fab | 435 | if (node_test_and_set(nid, *unlinked_nodes)) |
c04fc586 GH |
436 | continue; |
437 | sysfs_remove_link(&node_devices[nid].sysdev.kobj, | |
438 | kobject_name(&mem_blk->sysdev.kobj)); | |
dee5d0d5 AC |
439 | sysfs_remove_link(&mem_blk->sysdev.kobj, |
440 | kobject_name(&node_devices[nid].sysdev.kobj)); | |
c04fc586 | 441 | } |
9ae49fab | 442 | NODEMASK_FREE(unlinked_nodes); |
c04fc586 GH |
443 | return 0; |
444 | } | |
445 | ||
446 | static int link_mem_sections(int nid) | |
447 | { | |
448 | unsigned long start_pfn = NODE_DATA(nid)->node_start_pfn; | |
449 | unsigned long end_pfn = start_pfn + NODE_DATA(nid)->node_spanned_pages; | |
450 | unsigned long pfn; | |
63d027a6 | 451 | struct memory_block *mem_blk = NULL; |
c04fc586 GH |
452 | int err = 0; |
453 | ||
454 | for (pfn = start_pfn; pfn < end_pfn; pfn += PAGES_PER_SECTION) { | |
455 | unsigned long section_nr = pfn_to_section_nr(pfn); | |
456 | struct mem_section *mem_sect; | |
c04fc586 GH |
457 | int ret; |
458 | ||
459 | if (!present_section_nr(section_nr)) | |
460 | continue; | |
461 | mem_sect = __nr_to_section(section_nr); | |
63d027a6 | 462 | mem_blk = find_memory_block_hinted(mem_sect, mem_blk); |
c04fc586 GH |
463 | ret = register_mem_sect_under_node(mem_blk, nid); |
464 | if (!err) | |
465 | err = ret; | |
466 | ||
467 | /* discard ref obtained in find_memory_block() */ | |
c04fc586 | 468 | } |
63d027a6 RH |
469 | |
470 | if (mem_blk) | |
471 | kobject_put(&mem_blk->sysdev.kobj); | |
c04fc586 GH |
472 | return err; |
473 | } | |
4faf8d95 | 474 | |
39da08cb | 475 | #ifdef CONFIG_HUGETLBFS |
4faf8d95 LS |
476 | /* |
477 | * Handle per node hstate attribute [un]registration on transistions | |
478 | * to/from memoryless state. | |
479 | */ | |
39da08cb LS |
480 | static void node_hugetlb_work(struct work_struct *work) |
481 | { | |
482 | struct node *node = container_of(work, struct node, node_work); | |
483 | ||
484 | /* | |
485 | * We only get here when a node transitions to/from memoryless state. | |
486 | * We can detect which transition occurred by examining whether the | |
487 | * node has memory now. hugetlb_register_node() already check this | |
488 | * so we try to register the attributes. If that fails, then the | |
489 | * node has transitioned to memoryless, try to unregister the | |
490 | * attributes. | |
491 | */ | |
492 | if (!hugetlb_register_node(node)) | |
493 | hugetlb_unregister_node(node); | |
494 | } | |
495 | ||
496 | static void init_node_hugetlb_work(int nid) | |
497 | { | |
498 | INIT_WORK(&node_devices[nid].node_work, node_hugetlb_work); | |
499 | } | |
4faf8d95 LS |
500 | |
501 | static int node_memory_callback(struct notifier_block *self, | |
502 | unsigned long action, void *arg) | |
503 | { | |
504 | struct memory_notify *mnb = arg; | |
505 | int nid = mnb->status_change_nid; | |
506 | ||
507 | switch (action) { | |
39da08cb LS |
508 | case MEM_ONLINE: |
509 | case MEM_OFFLINE: | |
510 | /* | |
511 | * offload per node hstate [un]registration to a work thread | |
512 | * when transitioning to/from memoryless state. | |
513 | */ | |
4faf8d95 | 514 | if (nid != NUMA_NO_NODE) |
39da08cb | 515 | schedule_work(&node_devices[nid].node_work); |
4faf8d95 | 516 | break; |
39da08cb | 517 | |
4faf8d95 LS |
518 | case MEM_GOING_ONLINE: |
519 | case MEM_GOING_OFFLINE: | |
520 | case MEM_CANCEL_ONLINE: | |
521 | case MEM_CANCEL_OFFLINE: | |
522 | default: | |
523 | break; | |
524 | } | |
525 | ||
526 | return NOTIFY_OK; | |
527 | } | |
39da08cb LS |
528 | #endif /* CONFIG_HUGETLBFS */ |
529 | #else /* !CONFIG_MEMORY_HOTPLUG_SPARSE */ | |
530 | ||
c04fc586 | 531 | static int link_mem_sections(int nid) { return 0; } |
39da08cb | 532 | #endif /* CONFIG_MEMORY_HOTPLUG_SPARSE */ |
4faf8d95 | 533 | |
39da08cb LS |
534 | #if !defined(CONFIG_MEMORY_HOTPLUG_SPARSE) || \ |
535 | !defined(CONFIG_HUGETLBFS) | |
4faf8d95 LS |
536 | static inline int node_memory_callback(struct notifier_block *self, |
537 | unsigned long action, void *arg) | |
538 | { | |
539 | return NOTIFY_OK; | |
540 | } | |
39da08cb LS |
541 | |
542 | static void init_node_hugetlb_work(int nid) { } | |
543 | ||
544 | #endif | |
c04fc586 | 545 | |
0fc44159 YG |
546 | int register_one_node(int nid) |
547 | { | |
548 | int error = 0; | |
76b67ed9 | 549 | int cpu; |
0fc44159 YG |
550 | |
551 | if (node_online(nid)) { | |
552 | int p_node = parent_node(nid); | |
553 | struct node *parent = NULL; | |
554 | ||
555 | if (p_node != nid) | |
556 | parent = &node_devices[p_node]; | |
557 | ||
558 | error = register_node(&node_devices[nid], nid, parent); | |
76b67ed9 KH |
559 | |
560 | /* link cpu under this node */ | |
561 | for_each_present_cpu(cpu) { | |
562 | if (cpu_to_node(cpu) == nid) | |
563 | register_cpu_under_node(cpu, nid); | |
564 | } | |
c04fc586 GH |
565 | |
566 | /* link memory sections under this node */ | |
567 | error = link_mem_sections(nid); | |
39da08cb LS |
568 | |
569 | /* initialize work queue for memory hot plug */ | |
570 | init_node_hugetlb_work(nid); | |
0fc44159 YG |
571 | } |
572 | ||
573 | return error; | |
574 | ||
575 | } | |
576 | ||
577 | void unregister_one_node(int nid) | |
578 | { | |
579 | unregister_node(&node_devices[nid]); | |
580 | } | |
581 | ||
bde631a5 LS |
582 | /* |
583 | * node states attributes | |
584 | */ | |
585 | ||
586 | static ssize_t print_nodes_state(enum node_states state, char *buf) | |
587 | { | |
588 | int n; | |
589 | ||
590 | n = nodelist_scnprintf(buf, PAGE_SIZE, node_states[state]); | |
591 | if (n > 0 && PAGE_SIZE > n + 1) { | |
592 | *(buf + n++) = '\n'; | |
593 | *(buf + n++) = '\0'; | |
594 | } | |
595 | return n; | |
596 | } | |
597 | ||
b15f562f AK |
598 | struct node_attr { |
599 | struct sysdev_class_attribute attr; | |
600 | enum node_states state; | |
601 | }; | |
bde631a5 | 602 | |
b15f562f AK |
603 | static ssize_t show_node_state(struct sysdev_class *class, |
604 | struct sysdev_class_attribute *attr, char *buf) | |
bde631a5 | 605 | { |
b15f562f AK |
606 | struct node_attr *na = container_of(attr, struct node_attr, attr); |
607 | return print_nodes_state(na->state, buf); | |
bde631a5 LS |
608 | } |
609 | ||
b15f562f AK |
610 | #define _NODE_ATTR(name, state) \ |
611 | { _SYSDEV_CLASS_ATTR(name, 0444, show_node_state, NULL), state } | |
bde631a5 | 612 | |
b15f562f AK |
613 | static struct node_attr node_state_attr[] = { |
614 | _NODE_ATTR(possible, N_POSSIBLE), | |
615 | _NODE_ATTR(online, N_ONLINE), | |
616 | _NODE_ATTR(has_normal_memory, N_NORMAL_MEMORY), | |
617 | _NODE_ATTR(has_cpu, N_CPU), | |
bde631a5 | 618 | #ifdef CONFIG_HIGHMEM |
b15f562f | 619 | _NODE_ATTR(has_high_memory, N_HIGH_MEMORY), |
bde631a5 | 620 | #endif |
bde631a5 LS |
621 | }; |
622 | ||
3701cde6 AK |
623 | static struct sysdev_class_attribute *node_state_attrs[] = { |
624 | &node_state_attr[0].attr, | |
625 | &node_state_attr[1].attr, | |
626 | &node_state_attr[2].attr, | |
627 | &node_state_attr[3].attr, | |
628 | #ifdef CONFIG_HIGHMEM | |
629 | &node_state_attr[4].attr, | |
630 | #endif | |
631 | NULL | |
632 | }; | |
bde631a5 | 633 | |
4faf8d95 | 634 | #define NODE_CALLBACK_PRI 2 /* lower than SLAB */ |
4b45099b | 635 | static int __init register_node_type(void) |
1da177e4 | 636 | { |
bde631a5 LS |
637 | int ret; |
638 | ||
3701cde6 AK |
639 | BUILD_BUG_ON(ARRAY_SIZE(node_state_attr) != NR_NODE_STATES); |
640 | BUILD_BUG_ON(ARRAY_SIZE(node_state_attrs)-1 != NR_NODE_STATES); | |
641 | ||
bde631a5 | 642 | ret = sysdev_class_register(&node_class); |
4faf8d95 | 643 | if (!ret) { |
4faf8d95 LS |
644 | hotplug_memory_notifier(node_memory_callback, |
645 | NODE_CALLBACK_PRI); | |
646 | } | |
bde631a5 LS |
647 | |
648 | /* | |
649 | * Note: we're not going to unregister the node class if we fail | |
650 | * to register the node state class attribute files. | |
651 | */ | |
652 | return ret; | |
1da177e4 LT |
653 | } |
654 | postcore_initcall(register_node_type); |