xref: /linux-6.15/include/linux/memory_hotplug.h (revision fc5dfebc)
1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef __LINUX_MEMORY_HOTPLUG_H
3 #define __LINUX_MEMORY_HOTPLUG_H
4 
5 #include <linux/mmzone.h>
6 #include <linux/spinlock.h>
7 #include <linux/notifier.h>
8 #include <linux/bug.h>
9 
10 struct page;
11 struct zone;
12 struct pglist_data;
13 struct mem_section;
14 struct memory_group;
15 struct resource;
16 struct vmem_altmap;
17 struct dev_pagemap;
18 
19 #ifdef CONFIG_HAVE_ARCH_NODEDATA_EXTENSION
20 /*
21  * For supporting node-hotadd, we have to allocate a new pgdat.
22  *
23  * If an arch has generic style NODE_DATA(),
24  * node_data[nid] = kzalloc() works well. But it depends on the architecture.
25  *
26  * In general, generic_alloc_nodedata() is used.
27  *
28  */
29 extern pg_data_t *arch_alloc_nodedata(int nid);
30 extern void arch_refresh_nodedata(int nid, pg_data_t *pgdat);
31 
32 #else /* CONFIG_HAVE_ARCH_NODEDATA_EXTENSION */
33 
34 #define arch_alloc_nodedata(nid)	generic_alloc_nodedata(nid)
35 
36 #ifdef CONFIG_NUMA
37 /*
38  * XXX: node aware allocation can't work well to get new node's memory at this time.
39  *	Because, pgdat for the new node is not allocated/initialized yet itself.
40  *	To use new node's memory, more consideration will be necessary.
41  */
42 #define generic_alloc_nodedata(nid)				\
43 ({								\
44 	memblock_alloc(sizeof(*pgdat), SMP_CACHE_BYTES);	\
45 })
46 /*
47  * This definition is just for error path in node hotadd.
48  * For node hotremove, we have to replace this.
49  */
50 #define generic_free_nodedata(pgdat)	kfree(pgdat)
51 
52 extern pg_data_t *node_data[];
53 static inline void arch_refresh_nodedata(int nid, pg_data_t *pgdat)
54 {
55 	node_data[nid] = pgdat;
56 }
57 
58 #else /* !CONFIG_NUMA */
59 
60 /* never called */
61 static inline pg_data_t *generic_alloc_nodedata(int nid)
62 {
63 	BUG();
64 	return NULL;
65 }
66 static inline void generic_free_nodedata(pg_data_t *pgdat)
67 {
68 }
69 static inline void arch_refresh_nodedata(int nid, pg_data_t *pgdat)
70 {
71 }
72 #endif /* CONFIG_NUMA */
73 #endif /* CONFIG_HAVE_ARCH_NODEDATA_EXTENSION */
74 
75 #ifdef CONFIG_MEMORY_HOTPLUG
76 struct page *pfn_to_online_page(unsigned long pfn);
77 
78 /* Types for control the zone type of onlined and offlined memory */
79 enum {
80 	/* Offline the memory. */
81 	MMOP_OFFLINE = 0,
82 	/* Online the memory. Zone depends, see default_zone_for_pfn(). */
83 	MMOP_ONLINE,
84 	/* Online the memory to ZONE_NORMAL. */
85 	MMOP_ONLINE_KERNEL,
86 	/* Online the memory to ZONE_MOVABLE. */
87 	MMOP_ONLINE_MOVABLE,
88 };
89 
90 /* Flags for add_memory() and friends to specify memory hotplug details. */
91 typedef int __bitwise mhp_t;
92 
93 /* No special request */
94 #define MHP_NONE		((__force mhp_t)0)
95 /*
96  * Allow merging of the added System RAM resource with adjacent,
97  * mergeable resources. After a successful call to add_memory_resource()
98  * with this flag set, the resource pointer must no longer be used as it
99  * might be stale, or the resource might have changed.
100  */
101 #define MHP_MERGE_RESOURCE	((__force mhp_t)BIT(0))
102 
103 /*
104  * We want memmap (struct page array) to be self contained.
105  * To do so, we will use the beginning of the hot-added range to build
106  * the page tables for the memmap array that describes the entire range.
107  * Only selected architectures support it with SPARSE_VMEMMAP.
108  */
109 #define MHP_MEMMAP_ON_MEMORY   ((__force mhp_t)BIT(1))
110 /*
111  * The nid field specifies a memory group id (mgid) instead. The memory group
112  * implies the node id (nid).
113  */
114 #define MHP_NID_IS_MGID		((__force mhp_t)BIT(2))
115 
116 /*
117  * Extended parameters for memory hotplug:
118  * altmap: alternative allocator for memmap array (optional)
119  * pgprot: page protection flags to apply to newly created page tables
120  *	(required)
121  */
122 struct mhp_params {
123 	struct vmem_altmap *altmap;
124 	pgprot_t pgprot;
125 	struct dev_pagemap *pgmap;
126 };
127 
128 bool mhp_range_allowed(u64 start, u64 size, bool need_mapping);
129 struct range mhp_get_pluggable_range(bool need_mapping);
130 
131 /*
132  * Zone resizing functions
133  *
134  * Note: any attempt to resize a zone should has pgdat_resize_lock()
135  * zone_span_writelock() both held. This ensure the size of a zone
136  * can't be changed while pgdat_resize_lock() held.
137  */
138 static inline unsigned zone_span_seqbegin(struct zone *zone)
139 {
140 	return read_seqbegin(&zone->span_seqlock);
141 }
142 static inline int zone_span_seqretry(struct zone *zone, unsigned iv)
143 {
144 	return read_seqretry(&zone->span_seqlock, iv);
145 }
146 static inline void zone_span_writelock(struct zone *zone)
147 {
148 	write_seqlock(&zone->span_seqlock);
149 }
150 static inline void zone_span_writeunlock(struct zone *zone)
151 {
152 	write_sequnlock(&zone->span_seqlock);
153 }
154 static inline void zone_seqlock_init(struct zone *zone)
155 {
156 	seqlock_init(&zone->span_seqlock);
157 }
158 extern void adjust_present_page_count(struct page *page,
159 				      struct memory_group *group,
160 				      long nr_pages);
161 /* VM interface that may be used by firmware interface */
162 extern int mhp_init_memmap_on_memory(unsigned long pfn, unsigned long nr_pages,
163 				     struct zone *zone);
164 extern void mhp_deinit_memmap_on_memory(unsigned long pfn, unsigned long nr_pages);
165 extern int online_pages(unsigned long pfn, unsigned long nr_pages,
166 			struct zone *zone, struct memory_group *group);
167 extern void __offline_isolated_pages(unsigned long start_pfn,
168 				     unsigned long end_pfn);
169 
170 typedef void (*online_page_callback_t)(struct page *page, unsigned int order);
171 
172 extern void generic_online_page(struct page *page, unsigned int order);
173 extern int set_online_page_callback(online_page_callback_t callback);
174 extern int restore_online_page_callback(online_page_callback_t callback);
175 
176 extern int try_online_node(int nid);
177 
178 extern int arch_add_memory(int nid, u64 start, u64 size,
179 			   struct mhp_params *params);
180 extern u64 max_mem_size;
181 
182 extern int mhp_online_type_from_str(const char *str);
183 
184 /* Default online_type (MMOP_*) when new memory blocks are added. */
185 extern int mhp_default_online_type;
186 /* If movable_node boot option specified */
187 extern bool movable_node_enabled;
188 static inline bool movable_node_is_enabled(void)
189 {
190 	return movable_node_enabled;
191 }
192 
193 extern void arch_remove_memory(u64 start, u64 size, struct vmem_altmap *altmap);
194 extern void __remove_pages(unsigned long start_pfn, unsigned long nr_pages,
195 			   struct vmem_altmap *altmap);
196 
197 /* reasonably generic interface to expand the physical pages */
198 extern int __add_pages(int nid, unsigned long start_pfn, unsigned long nr_pages,
199 		       struct mhp_params *params);
200 
201 #ifndef CONFIG_ARCH_HAS_ADD_PAGES
202 static inline int add_pages(int nid, unsigned long start_pfn,
203 		unsigned long nr_pages, struct mhp_params *params)
204 {
205 	return __add_pages(nid, start_pfn, nr_pages, params);
206 }
207 #else /* ARCH_HAS_ADD_PAGES */
208 int add_pages(int nid, unsigned long start_pfn, unsigned long nr_pages,
209 	      struct mhp_params *params);
210 #endif /* ARCH_HAS_ADD_PAGES */
211 
212 void get_online_mems(void);
213 void put_online_mems(void);
214 
215 void mem_hotplug_begin(void);
216 void mem_hotplug_done(void);
217 
218 /* See kswapd_is_running() */
219 static inline void pgdat_kswapd_lock(pg_data_t *pgdat)
220 {
221 	mutex_lock(&pgdat->kswapd_lock);
222 }
223 
224 static inline void pgdat_kswapd_unlock(pg_data_t *pgdat)
225 {
226 	mutex_unlock(&pgdat->kswapd_lock);
227 }
228 
229 static inline void pgdat_kswapd_lock_init(pg_data_t *pgdat)
230 {
231 	mutex_init(&pgdat->kswapd_lock);
232 }
233 
234 #else /* ! CONFIG_MEMORY_HOTPLUG */
235 #define pfn_to_online_page(pfn)			\
236 ({						\
237 	struct page *___page = NULL;		\
238 	if (pfn_valid(pfn))			\
239 		___page = pfn_to_page(pfn);	\
240 	___page;				\
241  })
242 
243 static inline unsigned zone_span_seqbegin(struct zone *zone)
244 {
245 	return 0;
246 }
247 static inline int zone_span_seqretry(struct zone *zone, unsigned iv)
248 {
249 	return 0;
250 }
251 static inline void zone_span_writelock(struct zone *zone) {}
252 static inline void zone_span_writeunlock(struct zone *zone) {}
253 static inline void zone_seqlock_init(struct zone *zone) {}
254 
255 static inline int try_online_node(int nid)
256 {
257 	return 0;
258 }
259 
260 static inline void get_online_mems(void) {}
261 static inline void put_online_mems(void) {}
262 
263 static inline void mem_hotplug_begin(void) {}
264 static inline void mem_hotplug_done(void) {}
265 
266 static inline bool movable_node_is_enabled(void)
267 {
268 	return false;
269 }
270 
271 static inline void pgdat_kswapd_lock(pg_data_t *pgdat) {}
272 static inline void pgdat_kswapd_unlock(pg_data_t *pgdat) {}
273 static inline void pgdat_kswapd_lock_init(pg_data_t *pgdat) {}
274 #endif /* ! CONFIG_MEMORY_HOTPLUG */
275 
276 /*
277  * Keep this declaration outside CONFIG_MEMORY_HOTPLUG as some
278  * platforms might override and use arch_get_mappable_range()
279  * for internal non memory hotplug purposes.
280  */
281 struct range arch_get_mappable_range(void);
282 
283 #if defined(CONFIG_MEMORY_HOTPLUG) || defined(CONFIG_DEFERRED_STRUCT_PAGE_INIT)
284 /*
285  * pgdat resizing functions
286  */
287 static inline
288 void pgdat_resize_lock(struct pglist_data *pgdat, unsigned long *flags)
289 {
290 	spin_lock_irqsave(&pgdat->node_size_lock, *flags);
291 }
292 static inline
293 void pgdat_resize_unlock(struct pglist_data *pgdat, unsigned long *flags)
294 {
295 	spin_unlock_irqrestore(&pgdat->node_size_lock, *flags);
296 }
297 static inline
298 void pgdat_resize_init(struct pglist_data *pgdat)
299 {
300 	spin_lock_init(&pgdat->node_size_lock);
301 }
302 #else /* !(CONFIG_MEMORY_HOTPLUG || CONFIG_DEFERRED_STRUCT_PAGE_INIT) */
303 /*
304  * Stub functions for when hotplug is off
305  */
306 static inline void pgdat_resize_lock(struct pglist_data *p, unsigned long *f) {}
307 static inline void pgdat_resize_unlock(struct pglist_data *p, unsigned long *f) {}
308 static inline void pgdat_resize_init(struct pglist_data *pgdat) {}
309 #endif /* !(CONFIG_MEMORY_HOTPLUG || CONFIG_DEFERRED_STRUCT_PAGE_INIT) */
310 
311 #ifdef CONFIG_MEMORY_HOTREMOVE
312 
313 extern void try_offline_node(int nid);
314 extern int offline_pages(unsigned long start_pfn, unsigned long nr_pages,
315 			 struct zone *zone, struct memory_group *group);
316 extern int remove_memory(u64 start, u64 size);
317 extern void __remove_memory(u64 start, u64 size);
318 extern int offline_and_remove_memory(u64 start, u64 size);
319 
320 #else
321 static inline void try_offline_node(int nid) {}
322 
323 static inline int offline_pages(unsigned long start_pfn, unsigned long nr_pages,
324 				struct zone *zone, struct memory_group *group)
325 {
326 	return -EINVAL;
327 }
328 
329 static inline int remove_memory(u64 start, u64 size)
330 {
331 	return -EBUSY;
332 }
333 
334 static inline void __remove_memory(u64 start, u64 size) {}
335 #endif /* CONFIG_MEMORY_HOTREMOVE */
336 
337 extern void set_zone_contiguous(struct zone *zone);
338 extern void clear_zone_contiguous(struct zone *zone);
339 
340 #ifdef CONFIG_MEMORY_HOTPLUG
341 extern void __ref free_area_init_core_hotplug(struct pglist_data *pgdat);
342 extern int __add_memory(int nid, u64 start, u64 size, mhp_t mhp_flags);
343 extern int add_memory(int nid, u64 start, u64 size, mhp_t mhp_flags);
344 extern int add_memory_resource(int nid, struct resource *resource,
345 			       mhp_t mhp_flags);
346 extern int add_memory_driver_managed(int nid, u64 start, u64 size,
347 				     const char *resource_name,
348 				     mhp_t mhp_flags);
349 extern void move_pfn_range_to_zone(struct zone *zone, unsigned long start_pfn,
350 				   unsigned long nr_pages,
351 				   struct vmem_altmap *altmap, int migratetype);
352 extern void remove_pfn_range_from_zone(struct zone *zone,
353 				       unsigned long start_pfn,
354 				       unsigned long nr_pages);
355 extern int sparse_add_section(int nid, unsigned long pfn,
356 		unsigned long nr_pages, struct vmem_altmap *altmap,
357 		struct dev_pagemap *pgmap);
358 extern void sparse_remove_section(struct mem_section *ms,
359 		unsigned long pfn, unsigned long nr_pages,
360 		unsigned long map_offset, struct vmem_altmap *altmap);
361 extern struct page *sparse_decode_mem_map(unsigned long coded_mem_map,
362 					  unsigned long pnum);
363 extern struct zone *zone_for_pfn_range(int online_type, int nid,
364 		struct memory_group *group, unsigned long start_pfn,
365 		unsigned long nr_pages);
366 extern int arch_create_linear_mapping(int nid, u64 start, u64 size,
367 				      struct mhp_params *params);
368 void arch_remove_linear_mapping(u64 start, u64 size);
369 extern bool mhp_supports_memmap_on_memory(unsigned long size);
370 #endif /* CONFIG_MEMORY_HOTPLUG */
371 
372 #endif /* __LINUX_MEMORY_HOTPLUG_H */
373