]>
Commit | Line | Data |
---|---|---|
b2441318 | 1 | // SPDX-License-Identifier: GPL-2.0 |
1da177e4 | 2 | /* |
10fbcf4c | 3 | * Basic Node interface support |
1da177e4 LT |
4 | */ |
5 | ||
1da177e4 LT |
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> |
6e259e7d | 11 | #include <linux/notifier.h> |
1da177e4 LT |
12 | #include <linux/node.h> |
13 | #include <linux/hugetlb.h> | |
ed4a6d7f | 14 | #include <linux/compaction.h> |
1da177e4 LT |
15 | #include <linux/cpumask.h> |
16 | #include <linux/topology.h> | |
17 | #include <linux/nodemask.h> | |
76b67ed9 | 18 | #include <linux/cpu.h> |
bde631a5 | 19 | #include <linux/device.h> |
af936a16 | 20 | #include <linux/swap.h> |
18e5b539 | 21 | #include <linux/slab.h> |
1da177e4 | 22 | |
10fbcf4c | 23 | static struct bus_type node_subsys = { |
af5ca3f4 | 24 | .name = "node", |
10fbcf4c | 25 | .dev_name = "node", |
1da177e4 LT |
26 | }; |
27 | ||
28 | ||
5aaba363 | 29 | static ssize_t node_read_cpumap(struct device *dev, bool list, char *buf) |
1da177e4 | 30 | { |
064f0e93 ZL |
31 | ssize_t n; |
32 | cpumask_var_t mask; | |
1da177e4 | 33 | struct node *node_dev = to_node(dev); |
1da177e4 | 34 | |
39106dcf MT |
35 | /* 2008/04/07: buf currently PAGE_SIZE, need 9 chars per 32 bits. */ |
36 | BUILD_BUG_ON((NR_CPUS/32 * 9) > (PAGE_SIZE-1)); | |
1da177e4 | 37 | |
064f0e93 ZL |
38 | if (!alloc_cpumask_var(&mask, GFP_KERNEL)) |
39 | return 0; | |
40 | ||
41 | cpumask_and(mask, cpumask_of_node(node_dev->dev.id), cpu_online_mask); | |
42 | n = cpumap_print_to_pagebuf(list, buf, mask); | |
43 | free_cpumask_var(mask); | |
44 | ||
45 | return n; | |
1da177e4 LT |
46 | } |
47 | ||
10fbcf4c KS |
48 | static inline ssize_t node_read_cpumask(struct device *dev, |
49 | struct device_attribute *attr, char *buf) | |
39106dcf | 50 | { |
5aaba363 | 51 | return node_read_cpumap(dev, false, buf); |
39106dcf | 52 | } |
10fbcf4c KS |
53 | static inline ssize_t node_read_cpulist(struct device *dev, |
54 | struct device_attribute *attr, char *buf) | |
39106dcf | 55 | { |
5aaba363 | 56 | return node_read_cpumap(dev, true, buf); |
39106dcf MT |
57 | } |
58 | ||
10fbcf4c KS |
59 | static DEVICE_ATTR(cpumap, S_IRUGO, node_read_cpumask, NULL); |
60 | static DEVICE_ATTR(cpulist, S_IRUGO, node_read_cpulist, NULL); | |
1da177e4 LT |
61 | |
62 | #define K(x) ((x) << (PAGE_SHIFT - 10)) | |
10fbcf4c KS |
63 | static ssize_t node_read_meminfo(struct device *dev, |
64 | struct device_attribute *attr, char *buf) | |
1da177e4 LT |
65 | { |
66 | int n; | |
67 | int nid = dev->id; | |
599d0c95 | 68 | struct pglist_data *pgdat = NODE_DATA(nid); |
1da177e4 | 69 | struct sysinfo i; |
1da177e4 LT |
70 | |
71 | si_meminfo_node(&i, nid); | |
7ee92255 | 72 | n = sprintf(buf, |
4f98a2fe RR |
73 | "Node %d MemTotal: %8lu kB\n" |
74 | "Node %d MemFree: %8lu kB\n" | |
75 | "Node %d MemUsed: %8lu kB\n" | |
76 | "Node %d Active: %8lu kB\n" | |
77 | "Node %d Inactive: %8lu kB\n" | |
78 | "Node %d Active(anon): %8lu kB\n" | |
79 | "Node %d Inactive(anon): %8lu kB\n" | |
80 | "Node %d Active(file): %8lu kB\n" | |
81 | "Node %d Inactive(file): %8lu kB\n" | |
5344b7e6 | 82 | "Node %d Unevictable: %8lu kB\n" |
7ee92255 KM |
83 | "Node %d Mlocked: %8lu kB\n", |
84 | nid, K(i.totalram), | |
85 | nid, K(i.freeram), | |
86 | nid, K(i.totalram - i.freeram), | |
599d0c95 MG |
87 | nid, K(node_page_state(pgdat, NR_ACTIVE_ANON) + |
88 | node_page_state(pgdat, NR_ACTIVE_FILE)), | |
89 | nid, K(node_page_state(pgdat, NR_INACTIVE_ANON) + | |
90 | node_page_state(pgdat, NR_INACTIVE_FILE)), | |
91 | nid, K(node_page_state(pgdat, NR_ACTIVE_ANON)), | |
92 | nid, K(node_page_state(pgdat, NR_INACTIVE_ANON)), | |
93 | nid, K(node_page_state(pgdat, NR_ACTIVE_FILE)), | |
94 | nid, K(node_page_state(pgdat, NR_INACTIVE_FILE)), | |
95 | nid, K(node_page_state(pgdat, NR_UNEVICTABLE)), | |
75ef7184 | 96 | nid, K(sum_zone_node_page_state(nid, NR_MLOCK))); |
7ee92255 | 97 | |
182e8e23 | 98 | #ifdef CONFIG_HIGHMEM |
7ee92255 | 99 | n += sprintf(buf + n, |
4f98a2fe RR |
100 | "Node %d HighTotal: %8lu kB\n" |
101 | "Node %d HighFree: %8lu kB\n" | |
102 | "Node %d LowTotal: %8lu kB\n" | |
7ee92255 KM |
103 | "Node %d LowFree: %8lu kB\n", |
104 | nid, K(i.totalhigh), | |
105 | nid, K(i.freehigh), | |
106 | nid, K(i.totalram - i.totalhigh), | |
107 | nid, K(i.freeram - i.freehigh)); | |
182e8e23 | 108 | #endif |
7ee92255 | 109 | n += sprintf(buf + n, |
4f98a2fe RR |
110 | "Node %d Dirty: %8lu kB\n" |
111 | "Node %d Writeback: %8lu kB\n" | |
112 | "Node %d FilePages: %8lu kB\n" | |
113 | "Node %d Mapped: %8lu kB\n" | |
114 | "Node %d AnonPages: %8lu kB\n" | |
4b02108a | 115 | "Node %d Shmem: %8lu kB\n" |
c6a7f572 | 116 | "Node %d KernelStack: %8lu kB\n" |
4f98a2fe RR |
117 | "Node %d PageTables: %8lu kB\n" |
118 | "Node %d NFS_Unstable: %8lu kB\n" | |
119 | "Node %d Bounce: %8lu kB\n" | |
120 | "Node %d WritebackTmp: %8lu kB\n" | |
121 | "Node %d Slab: %8lu kB\n" | |
122 | "Node %d SReclaimable: %8lu kB\n" | |
05b258e9 DR |
123 | "Node %d SUnreclaim: %8lu kB\n" |
124 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE | |
125 | "Node %d AnonHugePages: %8lu kB\n" | |
65c45377 KS |
126 | "Node %d ShmemHugePages: %8lu kB\n" |
127 | "Node %d ShmemPmdMapped: %8lu kB\n" | |
05b258e9 DR |
128 | #endif |
129 | , | |
11fb9989 MG |
130 | nid, K(node_page_state(pgdat, NR_FILE_DIRTY)), |
131 | nid, K(node_page_state(pgdat, NR_WRITEBACK)), | |
132 | nid, K(node_page_state(pgdat, NR_FILE_PAGES)), | |
50658e2e | 133 | nid, K(node_page_state(pgdat, NR_FILE_MAPPED)), |
4b9d0fab | 134 | nid, K(node_page_state(pgdat, NR_ANON_MAPPED)), |
cc7452b6 | 135 | nid, K(i.sharedram), |
d30dd8be | 136 | nid, sum_zone_node_page_state(nid, NR_KERNEL_STACK_KB), |
75ef7184 | 137 | nid, K(sum_zone_node_page_state(nid, NR_PAGETABLE)), |
11fb9989 | 138 | nid, K(node_page_state(pgdat, NR_UNSTABLE_NFS)), |
75ef7184 | 139 | nid, K(sum_zone_node_page_state(nid, NR_BOUNCE)), |
11fb9989 | 140 | nid, K(node_page_state(pgdat, NR_WRITEBACK_TEMP)), |
385386cf JW |
141 | nid, K(node_page_state(pgdat, NR_SLAB_RECLAIMABLE) + |
142 | node_page_state(pgdat, NR_SLAB_UNRECLAIMABLE)), | |
143 | nid, K(node_page_state(pgdat, NR_SLAB_RECLAIMABLE)), | |
05b258e9 | 144 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
385386cf | 145 | nid, K(node_page_state(pgdat, NR_SLAB_UNRECLAIMABLE)), |
11fb9989 | 146 | nid, K(node_page_state(pgdat, NR_ANON_THPS) * |
65c45377 | 147 | HPAGE_PMD_NR), |
11fb9989 | 148 | nid, K(node_page_state(pgdat, NR_SHMEM_THPS) * |
65c45377 | 149 | HPAGE_PMD_NR), |
11fb9989 | 150 | nid, K(node_page_state(pgdat, NR_SHMEM_PMDMAPPED) * |
65c45377 | 151 | HPAGE_PMD_NR)); |
91a13c28 | 152 | #else |
385386cf | 153 | nid, K(node_page_state(pgdat, 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 | |
10fbcf4c | 160 | static DEVICE_ATTR(meminfo, S_IRUGO, node_read_meminfo, NULL); |
1da177e4 | 161 | |
10fbcf4c KS |
162 | static ssize_t node_read_numastat(struct device *dev, |
163 | struct device_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", | |
3a321d2a KW |
172 | sum_zone_numa_state(dev->id, NUMA_HIT), |
173 | sum_zone_numa_state(dev->id, NUMA_MISS), | |
174 | sum_zone_numa_state(dev->id, NUMA_FOREIGN), | |
175 | sum_zone_numa_state(dev->id, NUMA_INTERLEAVE_HIT), | |
176 | sum_zone_numa_state(dev->id, NUMA_LOCAL), | |
177 | sum_zone_numa_state(dev->id, NUMA_OTHER)); | |
1da177e4 | 178 | } |
10fbcf4c | 179 | static DEVICE_ATTR(numastat, S_IRUGO, node_read_numastat, NULL); |
1da177e4 | 180 | |
10fbcf4c KS |
181 | static ssize_t node_read_vmstat(struct device *dev, |
182 | struct device_attribute *attr, char *buf) | |
2ac39037 MR |
183 | { |
184 | int nid = dev->id; | |
75ef7184 | 185 | struct pglist_data *pgdat = NODE_DATA(nid); |
fa25c503 KM |
186 | int i; |
187 | int n = 0; | |
188 | ||
189 | for (i = 0; i < NR_VM_ZONE_STAT_ITEMS; i++) | |
190 | n += sprintf(buf+n, "%s %lu\n", vmstat_text[i], | |
75ef7184 MG |
191 | sum_zone_node_page_state(nid, i)); |
192 | ||
3a321d2a KW |
193 | #ifdef CONFIG_NUMA |
194 | for (i = 0; i < NR_VM_NUMA_STAT_ITEMS; i++) | |
75ef7184 MG |
195 | n += sprintf(buf+n, "%s %lu\n", |
196 | vmstat_text[i + NR_VM_ZONE_STAT_ITEMS], | |
3a321d2a KW |
197 | sum_zone_numa_state(nid, i)); |
198 | #endif | |
199 | ||
200 | for (i = 0; i < NR_VM_NODE_STAT_ITEMS; i++) | |
201 | n += sprintf(buf+n, "%s %lu\n", | |
202 | vmstat_text[i + NR_VM_ZONE_STAT_ITEMS + | |
203 | NR_VM_NUMA_STAT_ITEMS], | |
75ef7184 | 204 | node_page_state(pgdat, i)); |
fa25c503 KM |
205 | |
206 | return n; | |
2ac39037 | 207 | } |
10fbcf4c | 208 | static DEVICE_ATTR(vmstat, S_IRUGO, node_read_vmstat, NULL); |
2ac39037 | 209 | |
10fbcf4c | 210 | static ssize_t node_read_distance(struct device *dev, |
518d3f38 | 211 | struct device_attribute *attr, char *buf) |
1da177e4 LT |
212 | { |
213 | int nid = dev->id; | |
214 | int len = 0; | |
215 | int i; | |
216 | ||
12ee3c0a DR |
217 | /* |
218 | * buf is currently PAGE_SIZE in length and each node needs 4 chars | |
219 | * at the most (distance + space or newline). | |
220 | */ | |
221 | BUILD_BUG_ON(MAX_NUMNODES * 4 > PAGE_SIZE); | |
1da177e4 LT |
222 | |
223 | for_each_online_node(i) | |
224 | len += sprintf(buf + len, "%s%d", i ? " " : "", node_distance(nid, i)); | |
225 | ||
226 | len += sprintf(buf + len, "\n"); | |
227 | return len; | |
228 | } | |
10fbcf4c | 229 | static DEVICE_ATTR(distance, S_IRUGO, node_read_distance, NULL); |
1da177e4 | 230 | |
3c9b8aaf TI |
231 | static struct attribute *node_dev_attrs[] = { |
232 | &dev_attr_cpumap.attr, | |
233 | &dev_attr_cpulist.attr, | |
234 | &dev_attr_meminfo.attr, | |
235 | &dev_attr_numastat.attr, | |
236 | &dev_attr_distance.attr, | |
237 | &dev_attr_vmstat.attr, | |
238 | NULL | |
239 | }; | |
7ca7ec40 | 240 | ATTRIBUTE_GROUPS(node_dev); |
3c9b8aaf | 241 | |
9a305230 LS |
242 | #ifdef CONFIG_HUGETLBFS |
243 | /* | |
244 | * hugetlbfs per node attributes registration interface: | |
245 | * When/if hugetlb[fs] subsystem initializes [sometime after this module], | |
4faf8d95 LS |
246 | * it will register its per node attributes for all online nodes with |
247 | * memory. It will also call register_hugetlbfs_with_node(), below, to | |
9a305230 LS |
248 | * register its attribute registration functions with this node driver. |
249 | * Once these hooks have been initialized, the node driver will call into | |
250 | * the hugetlb module to [un]register attributes for hot-plugged nodes. | |
251 | */ | |
252 | static node_registration_func_t __hugetlb_register_node; | |
253 | static node_registration_func_t __hugetlb_unregister_node; | |
254 | ||
39da08cb | 255 | static inline bool hugetlb_register_node(struct node *node) |
9a305230 | 256 | { |
4faf8d95 | 257 | if (__hugetlb_register_node && |
8cebfcd0 | 258 | node_state(node->dev.id, N_MEMORY)) { |
9a305230 | 259 | __hugetlb_register_node(node); |
39da08cb LS |
260 | return true; |
261 | } | |
262 | return false; | |
9a305230 LS |
263 | } |
264 | ||
265 | static inline void hugetlb_unregister_node(struct node *node) | |
266 | { | |
267 | if (__hugetlb_unregister_node) | |
268 | __hugetlb_unregister_node(node); | |
269 | } | |
270 | ||
271 | void register_hugetlbfs_with_node(node_registration_func_t doregister, | |
272 | node_registration_func_t unregister) | |
273 | { | |
274 | __hugetlb_register_node = doregister; | |
275 | __hugetlb_unregister_node = unregister; | |
276 | } | |
277 | #else | |
278 | static inline void hugetlb_register_node(struct node *node) {} | |
279 | ||
280 | static inline void hugetlb_unregister_node(struct node *node) {} | |
281 | #endif | |
282 | ||
8c7b5b4e YI |
283 | static void node_device_release(struct device *dev) |
284 | { | |
285 | struct node *node = to_node(dev); | |
286 | ||
287 | #if defined(CONFIG_MEMORY_HOTPLUG_SPARSE) && defined(CONFIG_HUGETLBFS) | |
288 | /* | |
289 | * We schedule the work only when a memory section is | |
290 | * onlined/offlined on this node. When we come here, | |
291 | * all the memory on this node has been offlined, | |
292 | * so we won't enqueue new work to this work. | |
293 | * | |
294 | * The work is using node->node_work, so we should | |
295 | * flush work before freeing the memory. | |
296 | */ | |
297 | flush_work(&node->node_work); | |
298 | #endif | |
299 | kfree(node); | |
300 | } | |
1da177e4 LT |
301 | |
302 | /* | |
405ae7d3 | 303 | * register_node - Setup a sysfs device for a node. |
1da177e4 LT |
304 | * @num - Node number to use when creating the device. |
305 | * | |
306 | * Initialize and register the node device. | |
307 | */ | |
a7be6e5a | 308 | static int register_node(struct node *node, int num) |
1da177e4 LT |
309 | { |
310 | int error; | |
311 | ||
10fbcf4c KS |
312 | node->dev.id = num; |
313 | node->dev.bus = &node_subsys; | |
8c7b5b4e | 314 | node->dev.release = node_device_release; |
7ca7ec40 | 315 | node->dev.groups = node_dev_groups; |
10fbcf4c | 316 | error = device_register(&node->dev); |
1da177e4 LT |
317 | |
318 | if (!error){ | |
9a305230 | 319 | hugetlb_register_node(node); |
ed4a6d7f MG |
320 | |
321 | compaction_register_node(node); | |
1da177e4 LT |
322 | } |
323 | return error; | |
324 | } | |
325 | ||
4b45099b KT |
326 | /** |
327 | * unregister_node - unregister a node device | |
328 | * @node: node going away | |
329 | * | |
330 | * Unregisters a node device @node. All the devices on the node must be | |
331 | * unregistered before calling this function. | |
332 | */ | |
333 | void unregister_node(struct node *node) | |
334 | { | |
4faf8d95 | 335 | hugetlb_unregister_node(node); /* no-op, if memoryless node */ |
af936a16 | 336 | |
10fbcf4c | 337 | device_unregister(&node->dev); |
4b45099b | 338 | } |
1da177e4 | 339 | |
8732794b | 340 | struct node *node_devices[MAX_NUMNODES]; |
0fc44159 | 341 | |
76b67ed9 KH |
342 | /* |
343 | * register cpu under node | |
344 | */ | |
345 | int register_cpu_under_node(unsigned int cpu, unsigned int nid) | |
346 | { | |
1830794a | 347 | int ret; |
8a25a2fd | 348 | struct device *obj; |
76b67ed9 | 349 | |
f8246f31 AC |
350 | if (!node_online(nid)) |
351 | return 0; | |
352 | ||
8a25a2fd | 353 | obj = get_cpu_device(cpu); |
f8246f31 AC |
354 | if (!obj) |
355 | return 0; | |
356 | ||
8732794b | 357 | ret = sysfs_create_link(&node_devices[nid]->dev.kobj, |
f8246f31 AC |
358 | &obj->kobj, |
359 | kobject_name(&obj->kobj)); | |
1830794a AC |
360 | if (ret) |
361 | return ret; | |
362 | ||
363 | return sysfs_create_link(&obj->kobj, | |
8732794b WC |
364 | &node_devices[nid]->dev.kobj, |
365 | kobject_name(&node_devices[nid]->dev.kobj)); | |
76b67ed9 KH |
366 | } |
367 | ||
368 | int unregister_cpu_under_node(unsigned int cpu, unsigned int nid) | |
369 | { | |
8a25a2fd | 370 | struct device *obj; |
b9d52dad AC |
371 | |
372 | if (!node_online(nid)) | |
373 | return 0; | |
374 | ||
8a25a2fd | 375 | obj = get_cpu_device(cpu); |
b9d52dad AC |
376 | if (!obj) |
377 | return 0; | |
378 | ||
8732794b | 379 | sysfs_remove_link(&node_devices[nid]->dev.kobj, |
b9d52dad | 380 | kobject_name(&obj->kobj)); |
1830794a | 381 | sysfs_remove_link(&obj->kobj, |
8732794b | 382 | kobject_name(&node_devices[nid]->dev.kobj)); |
b9d52dad | 383 | |
76b67ed9 KH |
384 | return 0; |
385 | } | |
386 | ||
c04fc586 | 387 | #ifdef CONFIG_MEMORY_HOTPLUG_SPARSE |
bd721ea7 | 388 | static int __ref get_nid_for_pfn(unsigned long pfn) |
c04fc586 | 389 | { |
c04fc586 GH |
390 | if (!pfn_valid_within(pfn)) |
391 | return -1; | |
3a80a7fa | 392 | #ifdef CONFIG_DEFERRED_STRUCT_PAGE_INIT |
8cdde385 | 393 | if (system_state < SYSTEM_RUNNING) |
3a80a7fa MG |
394 | return early_pfn_to_nid(pfn); |
395 | #endif | |
c04fc586 GH |
396 | return pfn_to_nid(pfn); |
397 | } | |
398 | ||
399 | /* register memory section under specified node if it spans that node */ | |
400 | int register_mem_sect_under_node(struct memory_block *mem_blk, int nid) | |
401 | { | |
dee5d0d5 | 402 | int ret; |
c04fc586 GH |
403 | unsigned long pfn, sect_start_pfn, sect_end_pfn; |
404 | ||
405 | if (!mem_blk) | |
406 | return -EFAULT; | |
407 | if (!node_online(nid)) | |
408 | return 0; | |
d3360164 NF |
409 | |
410 | sect_start_pfn = section_nr_to_pfn(mem_blk->start_section_nr); | |
411 | sect_end_pfn = section_nr_to_pfn(mem_blk->end_section_nr); | |
412 | sect_end_pfn += PAGES_PER_SECTION - 1; | |
c04fc586 GH |
413 | for (pfn = sect_start_pfn; pfn <= sect_end_pfn; pfn++) { |
414 | int page_nid; | |
415 | ||
04697858 YL |
416 | /* |
417 | * memory block could have several absent sections from start. | |
418 | * skip pfn range from absent section | |
419 | */ | |
420 | if (!pfn_present(pfn)) { | |
421 | pfn = round_down(pfn + PAGES_PER_SECTION, | |
422 | PAGES_PER_SECTION) - 1; | |
423 | continue; | |
424 | } | |
425 | ||
c04fc586 GH |
426 | page_nid = get_nid_for_pfn(pfn); |
427 | if (page_nid < 0) | |
428 | continue; | |
429 | if (page_nid != nid) | |
430 | continue; | |
8732794b | 431 | ret = sysfs_create_link_nowarn(&node_devices[nid]->dev.kobj, |
10fbcf4c KS |
432 | &mem_blk->dev.kobj, |
433 | kobject_name(&mem_blk->dev.kobj)); | |
dee5d0d5 AC |
434 | if (ret) |
435 | return ret; | |
436 | ||
10fbcf4c | 437 | return sysfs_create_link_nowarn(&mem_blk->dev.kobj, |
8732794b WC |
438 | &node_devices[nid]->dev.kobj, |
439 | kobject_name(&node_devices[nid]->dev.kobj)); | |
c04fc586 GH |
440 | } |
441 | /* mem section does not span the specified node */ | |
442 | return 0; | |
443 | } | |
444 | ||
445 | /* unregister memory section under all nodes that it spans */ | |
d3360164 NF |
446 | int unregister_mem_sect_under_nodes(struct memory_block *mem_blk, |
447 | unsigned long phys_index) | |
c04fc586 | 448 | { |
9ae49fab | 449 | NODEMASK_ALLOC(nodemask_t, unlinked_nodes, GFP_KERNEL); |
c04fc586 GH |
450 | unsigned long pfn, sect_start_pfn, sect_end_pfn; |
451 | ||
9ae49fab DR |
452 | if (!mem_blk) { |
453 | NODEMASK_FREE(unlinked_nodes); | |
c04fc586 | 454 | return -EFAULT; |
9ae49fab DR |
455 | } |
456 | if (!unlinked_nodes) | |
457 | return -ENOMEM; | |
458 | nodes_clear(*unlinked_nodes); | |
d3360164 NF |
459 | |
460 | sect_start_pfn = section_nr_to_pfn(phys_index); | |
c04fc586 GH |
461 | sect_end_pfn = sect_start_pfn + PAGES_PER_SECTION - 1; |
462 | for (pfn = sect_start_pfn; pfn <= sect_end_pfn; pfn++) { | |
47504980 | 463 | int nid; |
c04fc586 GH |
464 | |
465 | nid = get_nid_for_pfn(pfn); | |
466 | if (nid < 0) | |
467 | continue; | |
468 | if (!node_online(nid)) | |
469 | continue; | |
9ae49fab | 470 | if (node_test_and_set(nid, *unlinked_nodes)) |
c04fc586 | 471 | continue; |
8732794b | 472 | sysfs_remove_link(&node_devices[nid]->dev.kobj, |
10fbcf4c KS |
473 | kobject_name(&mem_blk->dev.kobj)); |
474 | sysfs_remove_link(&mem_blk->dev.kobj, | |
8732794b | 475 | kobject_name(&node_devices[nid]->dev.kobj)); |
c04fc586 | 476 | } |
9ae49fab | 477 | NODEMASK_FREE(unlinked_nodes); |
c04fc586 GH |
478 | return 0; |
479 | } | |
480 | ||
9037a993 | 481 | int link_mem_sections(int nid, unsigned long start_pfn, unsigned long nr_pages) |
c04fc586 | 482 | { |
9037a993 | 483 | unsigned long end_pfn = start_pfn + nr_pages; |
c04fc586 | 484 | unsigned long pfn; |
63d027a6 | 485 | struct memory_block *mem_blk = NULL; |
c04fc586 GH |
486 | int err = 0; |
487 | ||
488 | for (pfn = start_pfn; pfn < end_pfn; pfn += PAGES_PER_SECTION) { | |
489 | unsigned long section_nr = pfn_to_section_nr(pfn); | |
490 | struct mem_section *mem_sect; | |
c04fc586 GH |
491 | int ret; |
492 | ||
493 | if (!present_section_nr(section_nr)) | |
494 | continue; | |
495 | mem_sect = __nr_to_section(section_nr); | |
321bf4ed YL |
496 | |
497 | /* same memblock ? */ | |
498 | if (mem_blk) | |
499 | if ((section_nr >= mem_blk->start_section_nr) && | |
500 | (section_nr <= mem_blk->end_section_nr)) | |
501 | continue; | |
502 | ||
63d027a6 | 503 | mem_blk = find_memory_block_hinted(mem_sect, mem_blk); |
321bf4ed | 504 | |
c04fc586 GH |
505 | ret = register_mem_sect_under_node(mem_blk, nid); |
506 | if (!err) | |
507 | err = ret; | |
508 | ||
509 | /* discard ref obtained in find_memory_block() */ | |
c04fc586 | 510 | } |
63d027a6 RH |
511 | |
512 | if (mem_blk) | |
10fbcf4c | 513 | kobject_put(&mem_blk->dev.kobj); |
c04fc586 GH |
514 | return err; |
515 | } | |
4faf8d95 | 516 | |
39da08cb | 517 | #ifdef CONFIG_HUGETLBFS |
4faf8d95 LS |
518 | /* |
519 | * Handle per node hstate attribute [un]registration on transistions | |
520 | * to/from memoryless state. | |
521 | */ | |
39da08cb LS |
522 | static void node_hugetlb_work(struct work_struct *work) |
523 | { | |
524 | struct node *node = container_of(work, struct node, node_work); | |
525 | ||
526 | /* | |
527 | * We only get here when a node transitions to/from memoryless state. | |
528 | * We can detect which transition occurred by examining whether the | |
529 | * node has memory now. hugetlb_register_node() already check this | |
530 | * so we try to register the attributes. If that fails, then the | |
531 | * node has transitioned to memoryless, try to unregister the | |
532 | * attributes. | |
533 | */ | |
534 | if (!hugetlb_register_node(node)) | |
535 | hugetlb_unregister_node(node); | |
536 | } | |
537 | ||
538 | static void init_node_hugetlb_work(int nid) | |
539 | { | |
8732794b | 540 | INIT_WORK(&node_devices[nid]->node_work, node_hugetlb_work); |
39da08cb | 541 | } |
4faf8d95 LS |
542 | |
543 | static int node_memory_callback(struct notifier_block *self, | |
544 | unsigned long action, void *arg) | |
545 | { | |
546 | struct memory_notify *mnb = arg; | |
547 | int nid = mnb->status_change_nid; | |
548 | ||
549 | switch (action) { | |
39da08cb LS |
550 | case MEM_ONLINE: |
551 | case MEM_OFFLINE: | |
552 | /* | |
553 | * offload per node hstate [un]registration to a work thread | |
554 | * when transitioning to/from memoryless state. | |
555 | */ | |
4faf8d95 | 556 | if (nid != NUMA_NO_NODE) |
8732794b | 557 | schedule_work(&node_devices[nid]->node_work); |
4faf8d95 | 558 | break; |
39da08cb | 559 | |
4faf8d95 LS |
560 | case MEM_GOING_ONLINE: |
561 | case MEM_GOING_OFFLINE: | |
562 | case MEM_CANCEL_ONLINE: | |
563 | case MEM_CANCEL_OFFLINE: | |
564 | default: | |
565 | break; | |
566 | } | |
567 | ||
568 | return NOTIFY_OK; | |
569 | } | |
39da08cb | 570 | #endif /* CONFIG_HUGETLBFS */ |
9037a993 | 571 | #endif /* CONFIG_MEMORY_HOTPLUG_SPARSE */ |
4faf8d95 | 572 | |
39da08cb LS |
573 | #if !defined(CONFIG_MEMORY_HOTPLUG_SPARSE) || \ |
574 | !defined(CONFIG_HUGETLBFS) | |
4faf8d95 LS |
575 | static inline int node_memory_callback(struct notifier_block *self, |
576 | unsigned long action, void *arg) | |
577 | { | |
578 | return NOTIFY_OK; | |
579 | } | |
39da08cb LS |
580 | |
581 | static void init_node_hugetlb_work(int nid) { } | |
582 | ||
583 | #endif | |
c04fc586 | 584 | |
9037a993 | 585 | int __register_one_node(int nid) |
0fc44159 | 586 | { |
9037a993 | 587 | int error; |
76b67ed9 | 588 | int cpu; |
0fc44159 | 589 | |
9037a993 MH |
590 | node_devices[nid] = kzalloc(sizeof(struct node), GFP_KERNEL); |
591 | if (!node_devices[nid]) | |
592 | return -ENOMEM; | |
c04fc586 | 593 | |
a7be6e5a | 594 | error = register_node(node_devices[nid], nid); |
39da08cb | 595 | |
9037a993 MH |
596 | /* link cpu under this node */ |
597 | for_each_present_cpu(cpu) { | |
598 | if (cpu_to_node(cpu) == nid) | |
599 | register_cpu_under_node(cpu, nid); | |
0fc44159 YG |
600 | } |
601 | ||
9037a993 MH |
602 | /* initialize work queue for memory hot plug */ |
603 | init_node_hugetlb_work(nid); | |
0fc44159 | 604 | |
9037a993 | 605 | return error; |
0fc44159 YG |
606 | } |
607 | ||
608 | void unregister_one_node(int nid) | |
609 | { | |
92d585ef XQ |
610 | if (!node_devices[nid]) |
611 | return; | |
612 | ||
8732794b | 613 | unregister_node(node_devices[nid]); |
8732794b | 614 | node_devices[nid] = NULL; |
0fc44159 YG |
615 | } |
616 | ||
bde631a5 LS |
617 | /* |
618 | * node states attributes | |
619 | */ | |
620 | ||
621 | static ssize_t print_nodes_state(enum node_states state, char *buf) | |
622 | { | |
623 | int n; | |
624 | ||
f799b1a7 TH |
625 | n = scnprintf(buf, PAGE_SIZE - 1, "%*pbl", |
626 | nodemask_pr_args(&node_states[state])); | |
f6238818 RO |
627 | buf[n++] = '\n'; |
628 | buf[n] = '\0'; | |
bde631a5 LS |
629 | return n; |
630 | } | |
631 | ||
b15f562f | 632 | struct node_attr { |
10fbcf4c | 633 | struct device_attribute attr; |
b15f562f AK |
634 | enum node_states state; |
635 | }; | |
bde631a5 | 636 | |
10fbcf4c KS |
637 | static ssize_t show_node_state(struct device *dev, |
638 | struct device_attribute *attr, char *buf) | |
bde631a5 | 639 | { |
b15f562f AK |
640 | struct node_attr *na = container_of(attr, struct node_attr, attr); |
641 | return print_nodes_state(na->state, buf); | |
bde631a5 LS |
642 | } |
643 | ||
b15f562f | 644 | #define _NODE_ATTR(name, state) \ |
10fbcf4c | 645 | { __ATTR(name, 0444, show_node_state, NULL), state } |
bde631a5 | 646 | |
b15f562f | 647 | static struct node_attr node_state_attr[] = { |
fcf07d22 LJ |
648 | [N_POSSIBLE] = _NODE_ATTR(possible, N_POSSIBLE), |
649 | [N_ONLINE] = _NODE_ATTR(online, N_ONLINE), | |
650 | [N_NORMAL_MEMORY] = _NODE_ATTR(has_normal_memory, N_NORMAL_MEMORY), | |
bde631a5 | 651 | #ifdef CONFIG_HIGHMEM |
fcf07d22 | 652 | [N_HIGH_MEMORY] = _NODE_ATTR(has_high_memory, N_HIGH_MEMORY), |
20b2f52b | 653 | #endif |
20b2f52b | 654 | [N_MEMORY] = _NODE_ATTR(has_memory, N_MEMORY), |
fcf07d22 | 655 | [N_CPU] = _NODE_ATTR(has_cpu, N_CPU), |
bde631a5 LS |
656 | }; |
657 | ||
10fbcf4c | 658 | static struct attribute *node_state_attrs[] = { |
fcf07d22 LJ |
659 | &node_state_attr[N_POSSIBLE].attr.attr, |
660 | &node_state_attr[N_ONLINE].attr.attr, | |
661 | &node_state_attr[N_NORMAL_MEMORY].attr.attr, | |
3701cde6 | 662 | #ifdef CONFIG_HIGHMEM |
fcf07d22 | 663 | &node_state_attr[N_HIGH_MEMORY].attr.attr, |
20b2f52b | 664 | #endif |
20b2f52b | 665 | &node_state_attr[N_MEMORY].attr.attr, |
fcf07d22 | 666 | &node_state_attr[N_CPU].attr.attr, |
3701cde6 AK |
667 | NULL |
668 | }; | |
bde631a5 | 669 | |
10fbcf4c KS |
670 | static struct attribute_group memory_root_attr_group = { |
671 | .attrs = node_state_attrs, | |
672 | }; | |
673 | ||
674 | static const struct attribute_group *cpu_root_attr_groups[] = { | |
675 | &memory_root_attr_group, | |
676 | NULL, | |
677 | }; | |
678 | ||
4faf8d95 | 679 | #define NODE_CALLBACK_PRI 2 /* lower than SLAB */ |
4b45099b | 680 | static int __init register_node_type(void) |
1da177e4 | 681 | { |
bde631a5 LS |
682 | int ret; |
683 | ||
3701cde6 AK |
684 | BUILD_BUG_ON(ARRAY_SIZE(node_state_attr) != NR_NODE_STATES); |
685 | BUILD_BUG_ON(ARRAY_SIZE(node_state_attrs)-1 != NR_NODE_STATES); | |
686 | ||
10fbcf4c | 687 | ret = subsys_system_register(&node_subsys, cpu_root_attr_groups); |
4faf8d95 | 688 | if (!ret) { |
6e259e7d AM |
689 | static struct notifier_block node_memory_callback_nb = { |
690 | .notifier_call = node_memory_callback, | |
691 | .priority = NODE_CALLBACK_PRI, | |
692 | }; | |
693 | register_hotmemory_notifier(&node_memory_callback_nb); | |
4faf8d95 | 694 | } |
bde631a5 LS |
695 | |
696 | /* | |
697 | * Note: we're not going to unregister the node class if we fail | |
698 | * to register the node state class attribute files. | |
699 | */ | |
700 | return ret; | |
1da177e4 LT |
701 | } |
702 | postcore_initcall(register_node_type); |