include/linux/memory_hotplug.h

   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_block;
  15 struct resource;
  16 struct vmem_altmap;
  17
  18 #ifdef CONFIG_MEMORY_HOTPLUG
  19 struct page *pfn_to_online_page(unsigned long pfn);
  20
  21 /*
  22  * Types for free bootmem stored in page->lru.next. These have to be in
  23  * some random range in unsigned long space for debugging purposes.
  24  */
  25 enum {
  26         MEMORY_HOTPLUG_MIN_BOOTMEM_TYPE = 12,
  27         SECTION_INFO = MEMORY_HOTPLUG_MIN_BOOTMEM_TYPE,
  28         MIX_SECTION_INFO,
  29         NODE_INFO,
  30         MEMORY_HOTPLUG_MAX_BOOTMEM_TYPE = NODE_INFO,
  31 };
  32
  33 /* Types for control the zone type of onlined and offlined memory */
  34 enum {
  35         /* Offline the memory. */
  36         MMOP_OFFLINE = 0,
  37         /* Online the memory. Zone depends, see default_zone_for_pfn(). */
  38         MMOP_ONLINE,
  39         /* Online the memory to ZONE_NORMAL. */
  40         MMOP_ONLINE_KERNEL,
  41         /* Online the memory to ZONE_MOVABLE. */
  42         MMOP_ONLINE_MOVABLE,
  43 };
  44
  45 /* Flags for add_memory() and friends to specify memory hotplug details. */
  46 typedef int __bitwise mhp_t;
  47
  48 /* No special request */
  49 #define MHP_NONE                ((__force mhp_t)0)
  50 /*
  51  * Allow merging of the added System RAM resource with adjacent,
  52  * mergeable resources. After a successful call to add_memory_resource()
  53  * with this flag set, the resource pointer must no longer be used as it
  54  * might be stale, or the resource might have changed.
  55  */
  56 #define MHP_MERGE_RESOURCE      ((__force mhp_t)BIT(0))
  57
  58 /*
  59  * We want memmap (struct page array) to be self contained.
  60  * To do so, we will use the beginning of the hot-added range to build
  61  * the page tables for the memmap array that describes the entire range.
  62  * Only selected architectures support it with SPARSE_VMEMMAP.
  63  */
  64 #define MHP_MEMMAP_ON_MEMORY   ((__force mhp_t)BIT(1))
  65
  66 /*
  67  * Extended parameters for memory hotplug:
  68  * altmap: alternative allocator for memmap array (optional)
  69  * pgprot: page protection flags to apply to newly created page tables
  70  *      (required)
  71  */
  72 struct mhp_params {
  73         struct vmem_altmap *altmap;
  74         pgprot_t pgprot;
  75 };
  76
  77 bool mhp_range_allowed(u64 start, u64 size, bool need_mapping);
  78 struct range mhp_get_pluggable_range(bool need_mapping);
  79
  80 /*
  81  * Zone resizing functions
  82  *
  83  * Note: any attempt to resize a zone should has pgdat_resize_lock()
  84  * zone_span_writelock() both held. This ensure the size of a zone
  85  * can't be changed while pgdat_resize_lock() held.
  86  */
  87 static inline unsigned zone_span_seqbegin(struct zone *zone)
  88 {
  89         return read_seqbegin(&zone->span_seqlock);
  90 }
  91 static inline int zone_span_seqretry(struct zone *zone, unsigned iv)
  92 {
  93         return read_seqretry(&zone->span_seqlock, iv);
  94 }
  95 static inline void zone_span_writelock(struct zone *zone)
  96 {
  97         write_seqlock(&zone->span_seqlock);
  98 }
  99 static inline void zone_span_writeunlock(struct zone *zone)
 100 {
 101         write_sequnlock(&zone->span_seqlock);
 102 }
 103 static inline void zone_seqlock_init(struct zone *zone)
 104 {
 105         seqlock_init(&zone->span_seqlock);
 106 }
 107 extern int zone_grow_free_lists(struct zone *zone, unsigned long new_nr_pages);
 108 extern int zone_grow_waitqueues(struct zone *zone, unsigned long nr_pages);
 109 extern int add_one_highpage(struct page *page, int pfn, int bad_ppro);
 110 extern void adjust_present_page_count(struct zone *zone, long nr_pages);
 111 /* VM interface that may be used by firmware interface */
 112 extern int mhp_init_memmap_on_memory(unsigned long pfn, unsigned long nr_pages,
 113                                      struct zone *zone);
 114 extern void mhp_deinit_memmap_on_memory(unsigned long pfn, unsigned long nr_pages);
 115 extern int online_pages(unsigned long pfn, unsigned long nr_pages,
 116                         struct zone *zone);
 117 extern struct zone *test_pages_in_a_zone(unsigned long start_pfn,
 118                                          unsigned long end_pfn);
 119 extern void __offline_isolated_pages(unsigned long start_pfn,
 120                                      unsigned long end_pfn);
 121
 122 typedef void (*online_page_callback_t)(struct page *page, unsigned int order);
 123
 124 extern void generic_online_page(struct page *page, unsigned int order);
 125 extern int set_online_page_callback(online_page_callback_t callback);
 126 extern int restore_online_page_callback(online_page_callback_t callback);
 127
 128 extern int try_online_node(int nid);
 129
 130 extern int arch_add_memory(int nid, u64 start, u64 size,
 131                            struct mhp_params *params);
 132 extern u64 max_mem_size;
 133
 134 extern int mhp_online_type_from_str(const char *str);
 135
 136 /* Default online_type (MMOP_*) when new memory blocks are added. */
 137 extern int mhp_default_online_type;
 138 /* If movable_node boot option specified */
 139 extern bool movable_node_enabled;
 140 static inline bool movable_node_is_enabled(void)
 141 {
 142         return movable_node_enabled;
 143 }
 144
 145 extern void arch_remove_memory(int nid, u64 start, u64 size,
 146                                struct vmem_altmap *altmap);
 147 extern void __remove_pages(unsigned long start_pfn, unsigned long nr_pages,
 148                            struct vmem_altmap *altmap);
 149
 150 /* reasonably generic interface to expand the physical pages */
 151 extern int __add_pages(int nid, unsigned long start_pfn, unsigned long nr_pages,
 152                        struct mhp_params *params);
 153
 154 #ifndef CONFIG_ARCH_HAS_ADD_PAGES
 155 static inline int add_pages(int nid, unsigned long start_pfn,
 156                 unsigned long nr_pages, struct mhp_params *params)
 157 {
 158         return __add_pages(nid, start_pfn, nr_pages, params);
 159 }
 160 #else /* ARCH_HAS_ADD_PAGES */
 161 int add_pages(int nid, unsigned long start_pfn, unsigned long nr_pages,
 162               struct mhp_params *params);
 163 #endif /* ARCH_HAS_ADD_PAGES */
 164
 165 #ifdef CONFIG_HAVE_ARCH_NODEDATA_EXTENSION
 166 /*
 167  * For supporting node-hotadd, we have to allocate a new pgdat.
 168  *
 169  * If an arch has generic style NODE_DATA(),
 170  * node_data[nid] = kzalloc() works well. But it depends on the architecture.
 171  *
 172  * In general, generic_alloc_nodedata() is used.
 173  * Now, arch_free_nodedata() is just defined for error path of node_hot_add.
 174  *
 175  */
 176 extern pg_data_t *arch_alloc_nodedata(int nid);
 177 extern void arch_free_nodedata(pg_data_t *pgdat);
 178 extern void arch_refresh_nodedata(int nid, pg_data_t *pgdat);
 179
 180 #else /* CONFIG_HAVE_ARCH_NODEDATA_EXTENSION */
 181
 182 #define arch_alloc_nodedata(nid)        generic_alloc_nodedata(nid)
 183 #define arch_free_nodedata(pgdat)       generic_free_nodedata(pgdat)
 184
 185 #ifdef CONFIG_NUMA
 186 /*
 187  * If ARCH_HAS_NODEDATA_EXTENSION=n, this func is used to allocate pgdat.
 188  * XXX: kmalloc_node() can't work well to get new node's memory at this time.
 189  *      Because, pgdat for the new node is not allocated/initialized yet itself.
 190  *      To use new node's memory, more consideration will be necessary.
 191  */
 192 #define generic_alloc_nodedata(nid)                             \
 193 ({                                                              \
 194         kzalloc(sizeof(pg_data_t), GFP_KERNEL);                 \
 195 })
 196 /*
 197  * This definition is just for error path in node hotadd.
 198  * For node hotremove, we have to replace this.
 199  */
 200 #define generic_free_nodedata(pgdat)    kfree(pgdat)
 201
 202 extern pg_data_t *node_data[];
 203 static inline void arch_refresh_nodedata(int nid, pg_data_t *pgdat)
 204 {
 205         node_data[nid] = pgdat;
 206 }
 207
 208 #else /* !CONFIG_NUMA */
 209
 210 /* never called */
 211 static inline pg_data_t *generic_alloc_nodedata(int nid)
 212 {
 213         BUG();
 214         return NULL;
 215 }
 216 static inline void generic_free_nodedata(pg_data_t *pgdat)
 217 {
 218 }
 219 static inline void arch_refresh_nodedata(int nid, pg_data_t *pgdat)
 220 {
 221 }
 222 #endif /* CONFIG_NUMA */
 223 #endif /* CONFIG_HAVE_ARCH_NODEDATA_EXTENSION */
 224
 225 #ifdef CONFIG_HAVE_BOOTMEM_INFO_NODE
 226 extern void __init register_page_bootmem_info_node(struct pglist_data *pgdat);
 227 #else
 228 static inline void register_page_bootmem_info_node(struct pglist_data *pgdat)
 229 {
 230 }
 231 #endif
 232 extern void put_page_bootmem(struct page *page);
 233 extern void get_page_bootmem(unsigned long ingo, struct page *page,
 234                              unsigned long type);
 235
 236 void get_online_mems(void);
 237 void put_online_mems(void);
 238
 239 void mem_hotplug_begin(void);
 240 void mem_hotplug_done(void);
 241
 242 #else /* ! CONFIG_MEMORY_HOTPLUG */
 243 #define pfn_to_online_page(pfn)                 \
 244 ({                                              \
 245         struct page *___page = NULL;            \
 246         if (pfn_valid(pfn))                     \
 247                 ___page = pfn_to_page(pfn);     \
 248         ___page;                                \
 249  })
 250
 251 static inline unsigned zone_span_seqbegin(struct zone *zone)
 252 {
 253         return 0;
 254 }
 255 static inline int zone_span_seqretry(struct zone *zone, unsigned iv)
 256 {
 257         return 0;
 258 }
 259 static inline void zone_span_writelock(struct zone *zone) {}
 260 static inline void zone_span_writeunlock(struct zone *zone) {}
 261 static inline void zone_seqlock_init(struct zone *zone) {}
 262
 263 static inline void register_page_bootmem_info_node(struct pglist_data *pgdat)
 264 {
 265 }
 266
 267 static inline int try_online_node(int nid)
 268 {
 269         return 0;
 270 }
 271
 272 static inline void get_online_mems(void) {}
 273 static inline void put_online_mems(void) {}
 274
 275 static inline void mem_hotplug_begin(void) {}
 276 static inline void mem_hotplug_done(void) {}
 277
 278 static inline bool movable_node_is_enabled(void)
 279 {
 280         return false;
 281 }
 282 #endif /* ! CONFIG_MEMORY_HOTPLUG */
 283
 284 /*
 285  * Keep this declaration outside CONFIG_MEMORY_HOTPLUG as some
 286  * platforms might override and use arch_get_mappable_range()
 287  * for internal non memory hotplug purposes.
 288  */
 289 struct range arch_get_mappable_range(void);
 290
 291 #if defined(CONFIG_MEMORY_HOTPLUG) || defined(CONFIG_DEFERRED_STRUCT_PAGE_INIT)
 292 /*
 293  * pgdat resizing functions
 294  */
 295 static inline
 296 void pgdat_resize_lock(struct pglist_data *pgdat, unsigned long *flags)
 297 {
 298         spin_lock_irqsave(&pgdat->node_size_lock, *flags);
 299 }
 300 static inline
 301 void pgdat_resize_unlock(struct pglist_data *pgdat, unsigned long *flags)
 302 {
 303         spin_unlock_irqrestore(&pgdat->node_size_lock, *flags);
 304 }
 305 static inline
 306 void pgdat_resize_init(struct pglist_data *pgdat)
 307 {
 308         spin_lock_init(&pgdat->node_size_lock);
 309 }
 310 #else /* !(CONFIG_MEMORY_HOTPLUG || CONFIG_DEFERRED_STRUCT_PAGE_INIT) */
 311 /*
 312  * Stub functions for when hotplug is off
 313  */
 314 static inline void pgdat_resize_lock(struct pglist_data *p, unsigned long *f) {}
 315 static inline void pgdat_resize_unlock(struct pglist_data *p, unsigned long *f) {}
 316 static inline void pgdat_resize_init(struct pglist_data *pgdat) {}
 317 #endif /* !(CONFIG_MEMORY_HOTPLUG || CONFIG_DEFERRED_STRUCT_PAGE_INIT) */
 318
 319 #ifdef CONFIG_MEMORY_HOTREMOVE
 320
 321 extern void try_offline_node(int nid);
 322 extern int offline_pages(unsigned long start_pfn, unsigned long nr_pages);
 323 extern int remove_memory(int nid, u64 start, u64 size);
 324 extern void __remove_memory(int nid, u64 start, u64 size);
 325 extern int offline_and_remove_memory(int nid, u64 start, u64 size);
 326
 327 #else
 328 static inline void try_offline_node(int nid) {}
 329
 330 static inline int offline_pages(unsigned long start_pfn, unsigned long nr_pages)
 331 {
 332         return -EINVAL;
 333 }
 334
 335 static inline int remove_memory(int nid, u64 start, u64 size)
 336 {
 337         return -EBUSY;
 338 }
 339
 340 static inline void __remove_memory(int nid, u64 start, u64 size) {}
 341 #endif /* CONFIG_MEMORY_HOTREMOVE */
 342
 343 extern void set_zone_contiguous(struct zone *zone);
 344 extern void clear_zone_contiguous(struct zone *zone);
 345
 346 #ifdef CONFIG_MEMORY_HOTPLUG
 347 extern void __ref free_area_init_core_hotplug(int nid);
 348 extern int __add_memory(int nid, u64 start, u64 size, mhp_t mhp_flags);
 349 extern int add_memory(int nid, u64 start, u64 size, mhp_t mhp_flags);
 350 extern int add_memory_resource(int nid, struct resource *resource,
 351                                mhp_t mhp_flags);
 352 extern int add_memory_driver_managed(int nid, u64 start, u64 size,
 353                                      const char *resource_name,
 354                                      mhp_t mhp_flags);
 355 extern void move_pfn_range_to_zone(struct zone *zone, unsigned long start_pfn,
 356                                    unsigned long nr_pages,
 357                                    struct vmem_altmap *altmap, int migratetype);
 358 extern void remove_pfn_range_from_zone(struct zone *zone,
 359                                        unsigned long start_pfn,
 360                                        unsigned long nr_pages);
 361 extern bool is_memblock_offlined(struct memory_block *mem);
 362 extern int sparse_add_section(int nid, unsigned long pfn,
 363                 unsigned long nr_pages, struct vmem_altmap *altmap);
 364 extern void sparse_remove_section(struct mem_section *ms,
 365                 unsigned long pfn, unsigned long nr_pages,
 366                 unsigned long map_offset, struct vmem_altmap *altmap);
 367 extern struct page *sparse_decode_mem_map(unsigned long coded_mem_map,
 368                                           unsigned long pnum);
 369 extern struct zone *zone_for_pfn_range(int online_type, int nid, unsigned start_pfn,
 370                 unsigned long nr_pages);
 371 extern int arch_create_linear_mapping(int nid, u64 start, u64 size,
 372                                       struct mhp_params *params);
 373 void arch_remove_linear_mapping(u64 start, u64 size);
 374 extern bool mhp_supports_memmap_on_memory(unsigned long size);
 375 #endif /* CONFIG_MEMORY_HOTPLUG */
 376
 377 #endif /* __LINUX_MEMORY_HOTPLUG_H */