kernel/power/snapshot.c

/* [<][>][^][v][top][bottom][index][help] */
This source file includes following definitions.
enable_restore_image_protection
hibernate_restore_protection_begin
hibernate_restore_protection_end
hibernate_restore_protect_page
hibernate_restore_unprotect_page
hibernate_restore_protection_begin
hibernate_restore_protection_end
hibernate_restore_protect_page
hibernate_restore_unprotect_page
hibernate_reserved_size_init
hibernate_image_size_init
get_image_page
__get_safe_page
get_safe_page
alloc_image_page
recycle_safe_page
free_image_page
free_list_of_pages
chain_init
chain_alloc
alloc_rtree_node
add_rtree_block
create_zone_bm_rtree
free_zone_bm_rtree
memory_bm_position_reset
free_mem_extents
create_mem_extents
memory_bm_create
memory_bm_free
memory_bm_find_bit
memory_bm_set_bit
mem_bm_set_bit_check
memory_bm_clear_bit
memory_bm_clear_current
memory_bm_test_bit
memory_bm_pfn_present
rtree_next_node
memory_bm_next_pfn
recycle_zone_bm_rtree
memory_bm_recycle
__register_nosave_region
swsusp_set_page_free
swsusp_page_is_free
swsusp_unset_page_free
swsusp_set_page_forbidden
swsusp_page_is_forbidden
swsusp_unset_page_forbidden
mark_nosave_pages
create_basic_memory_bitmaps
free_basic_memory_bitmaps
clear_free_pages
snapshot_additional_pages
count_free_highmem_pages
saveable_highmem_page
count_highmem_pages
saveable_highmem_page
saveable_page
count_data_pages
do_copy_page
safe_copy_page
page_is_saveable
copy_data_page
copy_data_page
copy_data_pages
swsusp_free
preallocate_image_pages
preallocate_image_memory
preallocate_image_highmem
__fraction
preallocate_highmem_fraction
preallocate_image_highmem
preallocate_highmem_fraction
free_unnecessary_pages
minimum_image_size
hibernate_preallocate_memory
count_pages_for_highmem
count_pages_for_highmem
enough_free_mem
get_highmem_buffer
alloc_highmem_pages
get_highmem_buffer
alloc_highmem_pages
swsusp_alloc
swsusp_save
init_header_complete
check_image_kernel
snapshot_get_image_size
init_header
pack_pfns
snapshot_read_next
duplicate_memory_bitmap
mark_unsafe_pages
check_header
load_header
unpack_orig_pfns
count_highmem_image_pages
prepare_highmem_image
get_highmem_page_buffer
copy_last_highmem_page
last_highmem_page_copied
free_highmem_data
count_highmem_image_pages
prepare_highmem_image
get_highmem_page_buffer
copy_last_highmem_page
last_highmem_page_copied
free_highmem_data
get_buffer
snapshot_write_next
snapshot_write_finalize
snapshot_image_loaded
swap_two_pages_data
restore_highmem
   1 // SPDX-License-Identifier: GPL-2.0-only
   2 /*
   3  * linux/kernel/power/snapshot.c
   4  *
   5  * This file provides system snapshot/restore functionality for swsusp.
   6  *
   7  * Copyright (C) 1998-2005 Pavel Machek <pavel@ucw.cz>
   8  * Copyright (C) 2006 Rafael J. Wysocki <rjw@sisk.pl>
   9  */
  10 
  11 #define pr_fmt(fmt) "PM: " fmt
  12 
  13 #include <linux/version.h>
  14 #include <linux/module.h>
  15 #include <linux/mm.h>
  16 #include <linux/suspend.h>
  17 #include <linux/delay.h>
  18 #include <linux/bitops.h>
  19 #include <linux/spinlock.h>
  20 #include <linux/kernel.h>
  21 #include <linux/pm.h>
  22 #include <linux/device.h>
  23 #include <linux/init.h>
  24 #include <linux/memblock.h>
  25 #include <linux/nmi.h>
  26 #include <linux/syscalls.h>
  27 #include <linux/console.h>
  28 #include <linux/highmem.h>
  29 #include <linux/list.h>
  30 #include <linux/slab.h>
  31 #include <linux/compiler.h>
  32 #include <linux/ktime.h>
  33 #include <linux/set_memory.h>
  34 
  35 #include <linux/uaccess.h>
  36 #include <asm/mmu_context.h>
  37 #include <asm/pgtable.h>
  38 #include <asm/tlbflush.h>
  39 #include <asm/io.h>
  40 
  41 #include "power.h"
  42 
  43 #if defined(CONFIG_STRICT_KERNEL_RWX) && defined(CONFIG_ARCH_HAS_SET_MEMORY)
  44 static bool hibernate_restore_protection;
  45 static bool hibernate_restore_protection_active;
  46 
  47 void enable_restore_image_protection(void)
  48 {
  49         hibernate_restore_protection = true;
  50 }
  51 
  52 static inline void hibernate_restore_protection_begin(void)
  53 {
  54         hibernate_restore_protection_active = hibernate_restore_protection;
  55 }
  56 
  57 static inline void hibernate_restore_protection_end(void)
  58 {
  59         hibernate_restore_protection_active = false;
  60 }
  61 
  62 static inline void hibernate_restore_protect_page(void *page_address)
  63 {
  64         if (hibernate_restore_protection_active)
  65                 set_memory_ro((unsigned long)page_address, 1);
  66 }
  67 
  68 static inline void hibernate_restore_unprotect_page(void *page_address)
  69 {
  70         if (hibernate_restore_protection_active)
  71                 set_memory_rw((unsigned long)page_address, 1);
  72 }
  73 #else
  74 static inline void hibernate_restore_protection_begin(void) {}
  75 static inline void hibernate_restore_protection_end(void) {}
  76 static inline void hibernate_restore_protect_page(void *page_address) {}
  77 static inline void hibernate_restore_unprotect_page(void *page_address) {}
  78 #endif /* CONFIG_STRICT_KERNEL_RWX  && CONFIG_ARCH_HAS_SET_MEMORY */
  79 
  80 static int swsusp_page_is_free(struct page *);
  81 static void swsusp_set_page_forbidden(struct page *);
  82 static void swsusp_unset_page_forbidden(struct page *);
  83 
  84 /*
  85  * Number of bytes to reserve for memory allocations made by device drivers
  86  * from their ->freeze() and ->freeze_noirq() callbacks so that they don't
  87  * cause image creation to fail (tunable via /sys/power/reserved_size).
  88  */
  89 unsigned long reserved_size;
  90 
  91 void __init hibernate_reserved_size_init(void)
  92 {
  93         reserved_size = SPARE_PAGES * PAGE_SIZE;
  94 }
  95 
  96 /*
  97  * Preferred image size in bytes (tunable via /sys/power/image_size).
  98  * When it is set to N, swsusp will do its best to ensure the image
  99  * size will not exceed N bytes, but if that is impossible, it will
 100  * try to create the smallest image possible.
 101  */
 102 unsigned long image_size;
 103 
 104 void __init hibernate_image_size_init(void)
 105 {
 106         image_size = ((totalram_pages() * 2) / 5) * PAGE_SIZE;
 107 }
 108 
 109 /*
 110  * List of PBEs needed for restoring the pages that were allocated before
 111  * the suspend and included in the suspend image, but have also been
 112  * allocated by the "resume" kernel, so their contents cannot be written
 113  * directly to their "original" page frames.
 114  */
 115 struct pbe *restore_pblist;
 116 
 117 /* struct linked_page is used to build chains of pages */
 118 
 119 #define LINKED_PAGE_DATA_SIZE   (PAGE_SIZE - sizeof(void *))
 120 
 121 struct linked_page {
 122         struct linked_page *next;
 123         char data[LINKED_PAGE_DATA_SIZE];
 124 } __packed;
 125 
 126 /*
 127  * List of "safe" pages (ie. pages that were not used by the image kernel
 128  * before hibernation) that may be used as temporary storage for image kernel
 129  * memory contents.
 130  */
 131 static struct linked_page *safe_pages_list;
 132 
 133 /* Pointer to an auxiliary buffer (1 page) */
 134 static void *buffer;
 135 
 136 #define PG_ANY          0
 137 #define PG_SAFE         1
 138 #define PG_UNSAFE_CLEAR 1
 139 #define PG_UNSAFE_KEEP  0
 140 
 141 static unsigned int allocated_unsafe_pages;
 142 
 143 /**
 144  * get_image_page - Allocate a page for a hibernation image.
 145  * @gfp_mask: GFP mask for the allocation.
 146  * @safe_needed: Get pages that were not used before hibernation (restore only)
 147  *
 148  * During image restoration, for storing the PBE list and the image data, we can
 149  * only use memory pages that do not conflict with the pages used before
 150  * hibernation.  The "unsafe" pages have PageNosaveFree set and we count them
 151  * using allocated_unsafe_pages.
 152  *
 153  * Each allocated image page is marked as PageNosave and PageNosaveFree so that
 154  * swsusp_free() can release it.
 155  */
 156 static void *get_image_page(gfp_t gfp_mask, int safe_needed)
 157 {
 158         void *res;
 159 
 160         res = (void *)get_zeroed_page(gfp_mask);
 161         if (safe_needed)
 162                 while (res && swsusp_page_is_free(virt_to_page(res))) {
 163                         /* The page is unsafe, mark it for swsusp_free() */
 164                         swsusp_set_page_forbidden(virt_to_page(res));
 165                         allocated_unsafe_pages++;
 166                         res = (void *)get_zeroed_page(gfp_mask);
 167                 }
 168         if (res) {
 169                 swsusp_set_page_forbidden(virt_to_page(res));
 170                 swsusp_set_page_free(virt_to_page(res));
 171         }
 172         return res;
 173 }
 174 
 175 static void *__get_safe_page(gfp_t gfp_mask)
 176 {
 177         if (safe_pages_list) {
 178                 void *ret = safe_pages_list;
 179 
 180                 safe_pages_list = safe_pages_list->next;
 181                 memset(ret, 0, PAGE_SIZE);
 182                 return ret;
 183         }
 184         return get_image_page(gfp_mask, PG_SAFE);
 185 }
 186 
 187 unsigned long get_safe_page(gfp_t gfp_mask)
 188 {
 189         return (unsigned long)__get_safe_page(gfp_mask);
 190 }
 191 
 192 static struct page *alloc_image_page(gfp_t gfp_mask)
 193 {
 194         struct page *page;
 195 
 196         page = alloc_page(gfp_mask);
 197         if (page) {
 198                 swsusp_set_page_forbidden(page);
 199                 swsusp_set_page_free(page);
 200         }
 201         return page;
 202 }
 203 
 204 static void recycle_safe_page(void *page_address)
 205 {
 206         struct linked_page *lp = page_address;
 207 
 208         lp->next = safe_pages_list;
 209         safe_pages_list = lp;
 210 }
 211 
 212 /**
 213  * free_image_page - Free a page allocated for hibernation image.
 214  * @addr: Address of the page to free.
 215  * @clear_nosave_free: If set, clear the PageNosaveFree bit for the page.
 216  *
 217  * The page to free should have been allocated by get_image_page() (page flags
 218  * set by it are affected).
 219  */
 220 static inline void free_image_page(void *addr, int clear_nosave_free)
 221 {
 222         struct page *page;
 223 
 224         BUG_ON(!virt_addr_valid(addr));
 225 
 226         page = virt_to_page(addr);
 227 
 228         swsusp_unset_page_forbidden(page);
 229         if (clear_nosave_free)
 230                 swsusp_unset_page_free(page);
 231 
 232         __free_page(page);
 233 }
 234 
 235 static inline void free_list_of_pages(struct linked_page *list,
 236                                       int clear_page_nosave)
 237 {
 238         while (list) {
 239                 struct linked_page *lp = list->next;
 240 
 241                 free_image_page(list, clear_page_nosave);
 242                 list = lp;
 243         }
 244 }
 245 
 246 /*
 247  * struct chain_allocator is used for allocating small objects out of
 248  * a linked list of pages called 'the chain'.
 249  *
 250  * The chain grows each time when there is no room for a new object in
 251  * the current page.  The allocated objects cannot be freed individually.
 252  * It is only possible to free them all at once, by freeing the entire
 253  * chain.
 254  *
 255  * NOTE: The chain allocator may be inefficient if the allocated objects
 256  * are not much smaller than PAGE_SIZE.
 257  */
 258 struct chain_allocator {
 259         struct linked_page *chain;      /* the chain */
 260         unsigned int used_space;        /* total size of objects allocated out
 261                                            of the current page */
 262         gfp_t gfp_mask;         /* mask for allocating pages */
 263         int safe_needed;        /* if set, only "safe" pages are allocated */
 264 };
 265 
 266 static void chain_init(struct chain_allocator *ca, gfp_t gfp_mask,
 267                        int safe_needed)
 268 {
 269         ca->chain = NULL;
 270         ca->used_space = LINKED_PAGE_DATA_SIZE;
 271         ca->gfp_mask = gfp_mask;
 272         ca->safe_needed = safe_needed;
 273 }
 274 
 275 static void *chain_alloc(struct chain_allocator *ca, unsigned int size)
 276 {
 277         void *ret;
 278 
 279         if (LINKED_PAGE_DATA_SIZE - ca->used_space < size) {
 280                 struct linked_page *lp;
 281 
 282                 lp = ca->safe_needed ? __get_safe_page(ca->gfp_mask) :
 283                                         get_image_page(ca->gfp_mask, PG_ANY);
 284                 if (!lp)
 285                         return NULL;
 286 
 287                 lp->next = ca->chain;
 288                 ca->chain = lp;
 289                 ca->used_space = 0;
 290         }
 291         ret = ca->chain->data + ca->used_space;
 292         ca->used_space += size;
 293         return ret;
 294 }
 295 
 296 /**
 297  * Data types related to memory bitmaps.
 298  *
 299  * Memory bitmap is a structure consiting of many linked lists of
 300  * objects.  The main list's elements are of type struct zone_bitmap
 301  * and each of them corresonds to one zone.  For each zone bitmap
 302  * object there is a list of objects of type struct bm_block that
 303  * represent each blocks of bitmap in which information is stored.
 304  *
 305  * struct memory_bitmap contains a pointer to the main list of zone
 306  * bitmap objects, a struct bm_position used for browsing the bitmap,
 307  * and a pointer to the list of pages used for allocating all of the
 308  * zone bitmap objects and bitmap block objects.
 309  *
 310  * NOTE: It has to be possible to lay out the bitmap in memory
 311  * using only allocations of order 0.  Additionally, the bitmap is
 312  * designed to work with arbitrary number of zones (this is over the
 313  * top for now, but let's avoid making unnecessary assumptions ;-).
 314  *
 315  * struct zone_bitmap contains a pointer to a list of bitmap block
 316  * objects and a pointer to the bitmap block object that has been
 317  * most recently used for setting bits.  Additionally, it contains the
 318  * PFNs that correspond to the start and end of the represented zone.
 319  *
 320  * struct bm_block contains a pointer to the memory page in which
 321  * information is stored (in the form of a block of bitmap)
 322  * It also contains the pfns that correspond to the start and end of
 323  * the represented memory area.
 324  *
 325  * The memory bitmap is organized as a radix tree to guarantee fast random
 326  * access to the bits. There is one radix tree for each zone (as returned
 327  * from create_mem_extents).
 328  *
 329  * One radix tree is represented by one struct mem_zone_bm_rtree. There are
 330  * two linked lists for the nodes of the tree, one for the inner nodes and
 331  * one for the leave nodes. The linked leave nodes are used for fast linear
 332  * access of the memory bitmap.
 333  *
 334  * The struct rtree_node represents one node of the radix tree.
 335  */
 336 
 337 #define BM_END_OF_MAP   (~0UL)
 338 
 339 #define BM_BITS_PER_BLOCK       (PAGE_SIZE * BITS_PER_BYTE)
 340 #define BM_BLOCK_SHIFT          (PAGE_SHIFT + 3)
 341 #define BM_BLOCK_MASK           ((1UL << BM_BLOCK_SHIFT) - 1)
 342 
 343 /*
 344  * struct rtree_node is a wrapper struct to link the nodes
 345  * of the rtree together for easy linear iteration over
 346  * bits and easy freeing
 347  */
 348 struct rtree_node {
 349         struct list_head list;
 350         unsigned long *data;
 351 };
 352 
 353 /*
 354  * struct mem_zone_bm_rtree represents a bitmap used for one
 355  * populated memory zone.
 356  */
 357 struct mem_zone_bm_rtree {
 358         struct list_head list;          /* Link Zones together         */
 359         struct list_head nodes;         /* Radix Tree inner nodes      */
 360         struct list_head leaves;        /* Radix Tree leaves           */
 361         unsigned long start_pfn;        /* Zone start page frame       */
 362         unsigned long end_pfn;          /* Zone end page frame + 1     */
 363         struct rtree_node *rtree;       /* Radix Tree Root             */
 364         int levels;                     /* Number of Radix Tree Levels */
 365         unsigned int blocks;            /* Number of Bitmap Blocks     */
 366 };
 367 
 368 /* strcut bm_position is used for browsing memory bitmaps */
 369 
 370 struct bm_position {
 371         struct mem_zone_bm_rtree *zone;
 372         struct rtree_node *node;
 373         unsigned long node_pfn;
 374         int node_bit;
 375 };
 376 
 377 struct memory_bitmap {
 378         struct list_head zones;
 379         struct linked_page *p_list;     /* list of pages used to store zone
 380                                            bitmap objects and bitmap block
 381                                            objects */
 382         struct bm_position cur; /* most recently used bit position */
 383 };
 384 
 385 /* Functions that operate on memory bitmaps */
 386 
 387 #define BM_ENTRIES_PER_LEVEL    (PAGE_SIZE / sizeof(unsigned long))
 388 #if BITS_PER_LONG == 32
 389 #define BM_RTREE_LEVEL_SHIFT    (PAGE_SHIFT - 2)
 390 #else
 391 #define BM_RTREE_LEVEL_SHIFT    (PAGE_SHIFT - 3)
 392 #endif
 393 #define BM_RTREE_LEVEL_MASK     ((1UL << BM_RTREE_LEVEL_SHIFT) - 1)
 394 
 395 /**
 396  * alloc_rtree_node - Allocate a new node and add it to the radix tree.
 397  *
 398  * This function is used to allocate inner nodes as well as the
 399  * leave nodes of the radix tree. It also adds the node to the
 400  * corresponding linked list passed in by the *list parameter.
 401  */
 402 static struct rtree_node *alloc_rtree_node(gfp_t gfp_mask, int safe_needed,
 403                                            struct chain_allocator *ca,
 404                                            struct list_head *list)
 405 {
 406         struct rtree_node *node;
 407 
 408         node = chain_alloc(ca, sizeof(struct rtree_node));
 409         if (!node)
 410                 return NULL;
 411 
 412         node->data = get_image_page(gfp_mask, safe_needed);
 413         if (!node->data)
 414                 return NULL;
 415 
 416         list_add_tail(&node->list, list);
 417 
 418         return node;
 419 }
 420 
 421 /**
 422  * add_rtree_block - Add a new leave node to the radix tree.
 423  *
 424  * The leave nodes need to be allocated in order to keep the leaves
 425  * linked list in order. This is guaranteed by the zone->blocks
 426  * counter.
 427  */
 428 static int add_rtree_block(struct mem_zone_bm_rtree *zone, gfp_t gfp_mask,
 429                            int safe_needed, struct chain_allocator *ca)
 430 {
 431         struct rtree_node *node, *block, **dst;
 432         unsigned int levels_needed, block_nr;
 433         int i;
 434 
 435         block_nr = zone->blocks;
 436         levels_needed = 0;
 437 
 438         /* How many levels do we need for this block nr? */
 439         while (block_nr) {
 440                 levels_needed += 1;
 441                 block_nr >>= BM_RTREE_LEVEL_SHIFT;
 442         }
 443 
 444         /* Make sure the rtree has enough levels */
 445         for (i = zone->levels; i < levels_needed; i++) {
 446                 node = alloc_rtree_node(gfp_mask, safe_needed, ca,
 447                                         &zone->nodes);
 448                 if (!node)
 449                         return -ENOMEM;
 450 
 451                 node->data[0] = (unsigned long)zone->rtree;
 452                 zone->rtree = node;
 453                 zone->levels += 1;
 454         }
 455 
 456         /* Allocate new block */
 457         block = alloc_rtree_node(gfp_mask, safe_needed, ca, &zone->leaves);
 458         if (!block)
 459                 return -ENOMEM;
 460 
 461         /* Now walk the rtree to insert the block */
 462         node = zone->rtree;
 463         dst = &zone->rtree;
 464         block_nr = zone->blocks;
 465         for (i = zone->levels; i > 0; i--) {
 466                 int index;
 467 
 468                 if (!node) {
 469                         node = alloc_rtree_node(gfp_mask, safe_needed, ca,
 470                                                 &zone->nodes);
 471                         if (!node)
 472                                 return -ENOMEM;
 473                         *dst = node;
 474                 }
 475 
 476                 index = block_nr >> ((i - 1) * BM_RTREE_LEVEL_SHIFT);
 477                 index &= BM_RTREE_LEVEL_MASK;
 478                 dst = (struct rtree_node **)&((*dst)->data[index]);
 479                 node = *dst;
 480         }
 481 
 482         zone->blocks += 1;
 483         *dst = block;
 484 
 485         return 0;
 486 }
 487 
 488 static void free_zone_bm_rtree(struct mem_zone_bm_rtree *zone,
 489                                int clear_nosave_free);
 490 
 491 /**
 492  * create_zone_bm_rtree - Create a radix tree for one zone.
 493  *
 494  * Allocated the mem_zone_bm_rtree structure and initializes it.
 495  * This function also allocated and builds the radix tree for the
 496  * zone.
 497  */
 498 static struct mem_zone_bm_rtree *create_zone_bm_rtree(gfp_t gfp_mask,
 499                                                       int safe_needed,
 500                                                       struct chain_allocator *ca,
 501                                                       unsigned long start,
 502                                                       unsigned long end)
 503 {
 504         struct mem_zone_bm_rtree *zone;
 505         unsigned int i, nr_blocks;
 506         unsigned long pages;
 507 
 508         pages = end - start;
 509         zone  = chain_alloc(ca, sizeof(struct mem_zone_bm_rtree));
 510         if (!zone)
 511                 return NULL;
 512 
 513         INIT_LIST_HEAD(&zone->nodes);
 514         INIT_LIST_HEAD(&zone->leaves);
 515         zone->start_pfn = start;
 516         zone->end_pfn = end;
 517         nr_blocks = DIV_ROUND_UP(pages, BM_BITS_PER_BLOCK);
 518 
 519         for (i = 0; i < nr_blocks; i++) {
 520                 if (add_rtree_block(zone, gfp_mask, safe_needed, ca)) {
 521                         free_zone_bm_rtree(zone, PG_UNSAFE_CLEAR);
 522                         return NULL;
 523                 }
 524         }
 525 
 526         return zone;
 527 }
 528 
 529 /**
 530  * free_zone_bm_rtree - Free the memory of the radix tree.
 531  *
 532  * Free all node pages of the radix tree. The mem_zone_bm_rtree
 533  * structure itself is not freed here nor are the rtree_node
 534  * structs.
 535  */
 536 static void free_zone_bm_rtree(struct mem_zone_bm_rtree *zone,
 537                                int clear_nosave_free)
 538 {
 539         struct rtree_node *node;
 540 
 541         list_for_each_entry(node, &zone->nodes, list)
 542                 free_image_page(node->data, clear_nosave_free);
 543 
 544         list_for_each_entry(node, &zone->leaves, list)
 545                 free_image_page(node->data, clear_nosave_free);
 546 }
 547 
 548 static void memory_bm_position_reset(struct memory_bitmap *bm)
 549 {
 550         bm->cur.zone = list_entry(bm->zones.next, struct mem_zone_bm_rtree,
 551                                   list);
 552         bm->cur.node = list_entry(bm->cur.zone->leaves.next,
 553                                   struct rtree_node, list);
 554         bm->cur.node_pfn = 0;
 555         bm->cur.node_bit = 0;
 556 }
 557 
 558 static void memory_bm_free(struct memory_bitmap *bm, int clear_nosave_free);
 559 
 560 struct mem_extent {
 561         struct list_head hook;
 562         unsigned long start;
 563         unsigned long end;
 564 };
 565 
 566 /**
 567  * free_mem_extents - Free a list of memory extents.
 568  * @list: List of extents to free.
 569  */
 570 static void free_mem_extents(struct list_head *list)
 571 {
 572         struct mem_extent *ext, *aux;
 573 
 574         list_for_each_entry_safe(ext, aux, list, hook) {
 575                 list_del(&ext->hook);
 576                 kfree(ext);
 577         }
 578 }
 579 
 580 /**
 581  * create_mem_extents - Create a list of memory extents.
 582  * @list: List to put the extents into.
 583  * @gfp_mask: Mask to use for memory allocations.
 584  *
 585  * The extents represent contiguous ranges of PFNs.
 586  */
 587 static int create_mem_extents(struct list_head *list, gfp_t gfp_mask)
 588 {
 589         struct zone *zone;
 590 
 591         INIT_LIST_HEAD(list);
 592 
 593         for_each_populated_zone(zone) {
 594                 unsigned long zone_start, zone_end;
 595                 struct mem_extent *ext, *cur, *aux;
 596 
 597                 zone_start = zone->zone_start_pfn;
 598                 zone_end = zone_end_pfn(zone);
 599 
 600                 list_for_each_entry(ext, list, hook)
 601                         if (zone_start <= ext->end)
 602                                 break;
 603 
 604                 if (&ext->hook == list || zone_end < ext->start) {
 605                         /* New extent is necessary */
 606                         struct mem_extent *new_ext;
 607 
 608                         new_ext = kzalloc(sizeof(struct mem_extent), gfp_mask);
 609                         if (!new_ext) {
 610                                 free_mem_extents(list);
 611                                 return -ENOMEM;
 612                         }
 613                         new_ext->start = zone_start;
 614                         new_ext->end = zone_end;
 615                         list_add_tail(&new_ext->hook, &ext->hook);
 616                         continue;
 617                 }
 618 
 619                 /* Merge this zone's range of PFNs with the existing one */
 620                 if (zone_start < ext->start)
 621                         ext->start = zone_start;
 622                 if (zone_end > ext->end)
 623                         ext->end = zone_end;
 624 
 625                 /* More merging may be possible */
 626                 cur = ext;
 627                 list_for_each_entry_safe_continue(cur, aux, list, hook) {
 628                         if (zone_end < cur->start)
 629                                 break;
 630                         if (zone_end < cur->end)
 631                                 ext->end = cur->end;
 632                         list_del(&cur->hook);
 633                         kfree(cur);
 634                 }
 635         }
 636 
 637         return 0;
 638 }
 639 
 640 /**
 641  * memory_bm_create - Allocate memory for a memory bitmap.
 642  */
 643 static int memory_bm_create(struct memory_bitmap *bm, gfp_t gfp_mask,
 644                             int safe_needed)
 645 {
 646         struct chain_allocator ca;
 647         struct list_head mem_extents;
 648         struct mem_extent *ext;
 649         int error;
 650 
 651         chain_init(&ca, gfp_mask, safe_needed);
 652         INIT_LIST_HEAD(&bm->zones);
 653 
 654         error = create_mem_extents(&mem_extents, gfp_mask);
 655         if (error)
 656                 return error;
 657 
 658         list_for_each_entry(ext, &mem_extents, hook) {
 659                 struct mem_zone_bm_rtree *zone;
 660 
 661                 zone = create_zone_bm_rtree(gfp_mask, safe_needed, &ca,
 662                                             ext->start, ext->end);
 663                 if (!zone) {
 664                         error = -ENOMEM;
 665                         goto Error;
 666                 }
 667                 list_add_tail(&zone->list, &bm->zones);
 668         }
 669 
 670         bm->p_list = ca.chain;
 671         memory_bm_position_reset(bm);
 672  Exit:
 673         free_mem_extents(&mem_extents);
 674         return error;
 675 
 676  Error:
 677         bm->p_list = ca.chain;
 678         memory_bm_free(bm, PG_UNSAFE_CLEAR);
 679         goto Exit;
 680 }
 681 
 682 /**
 683  * memory_bm_free - Free memory occupied by the memory bitmap.
 684  * @bm: Memory bitmap.
 685  */
 686 static void memory_bm_free(struct memory_bitmap *bm, int clear_nosave_free)
 687 {
 688         struct mem_zone_bm_rtree *zone;
 689 
 690         list_for_each_entry(zone, &bm->zones, list)
 691                 free_zone_bm_rtree(zone, clear_nosave_free);
 692 
 693         free_list_of_pages(bm->p_list, clear_nosave_free);
 694 
 695         INIT_LIST_HEAD(&bm->zones);
 696 }
 697 
 698 /**
 699  * memory_bm_find_bit - Find the bit for a given PFN in a memory bitmap.
 700  *
 701  * Find the bit in memory bitmap @bm that corresponds to the given PFN.
 702  * The cur.zone, cur.block and cur.node_pfn members of @bm are updated.
 703  *
 704  * Walk the radix tree to find the page containing the bit that represents @pfn
 705  * and return the position of the bit in @addr and @bit_nr.
 706  */
 707 static int memory_bm_find_bit(struct memory_bitmap *bm, unsigned long pfn,
 708                               void **addr, unsigned int *bit_nr)
 709 {
 710         struct mem_zone_bm_rtree *curr, *zone;
 711         struct rtree_node *node;
 712         int i, block_nr;
 713 
 714         zone = bm->cur.zone;
 715 
 716         if (pfn >= zone->start_pfn && pfn < zone->end_pfn)
 717                 goto zone_found;
 718 
 719         zone = NULL;
 720 
 721         /* Find the right zone */
 722         list_for_each_entry(curr, &bm->zones, list) {
 723                 if (pfn >= curr->start_pfn && pfn < curr->end_pfn) {
 724                         zone = curr;
 725                         break;
 726                 }
 727         }
 728 
 729         if (!zone)
 730                 return -EFAULT;
 731 
 732 zone_found:
 733         /*
 734          * We have found the zone. Now walk the radix tree to find the leaf node
 735          * for our PFN.
 736          */
 737 
 738         /*
 739          * If the zone we wish to scan is the the current zone and the
 740          * pfn falls into the current node then we do not need to walk
 741          * the tree.
 742          */
 743         node = bm->cur.node;
 744         if (zone == bm->cur.zone &&
 745             ((pfn - zone->start_pfn) & ~BM_BLOCK_MASK) == bm->cur.node_pfn)
 746                 goto node_found;
 747 
 748         node      = zone->rtree;
 749         block_nr  = (pfn - zone->start_pfn) >> BM_BLOCK_SHIFT;
 750 
 751         for (i = zone->levels; i > 0; i--) {
 752                 int index;
 753 
 754                 index = block_nr >> ((i - 1) * BM_RTREE_LEVEL_SHIFT);
 755                 index &= BM_RTREE_LEVEL_MASK;
 756                 BUG_ON(node->data[index] == 0);
 757                 node = (struct rtree_node *)node->data[index];
 758         }
 759 
 760 node_found:
 761         /* Update last position */
 762         bm->cur.zone = zone;
 763         bm->cur.node = node;
 764         bm->cur.node_pfn = (pfn - zone->start_pfn) & ~BM_BLOCK_MASK;
 765 
 766         /* Set return values */
 767         *addr = node->data;
 768         *bit_nr = (pfn - zone->start_pfn) & BM_BLOCK_MASK;
 769 
 770         return 0;
 771 }
 772 
 773 static void memory_bm_set_bit(struct memory_bitmap *bm, unsigned long pfn)
 774 {
 775         void *addr;
 776         unsigned int bit;
 777         int error;
 778 
 779         error = memory_bm_find_bit(bm, pfn, &addr, &bit);
 780         BUG_ON(error);
 781         set_bit(bit, addr);
 782 }
 783 
 784 static int mem_bm_set_bit_check(struct memory_bitmap *bm, unsigned long pfn)
 785 {
 786         void *addr;
 787         unsigned int bit;
 788         int error;
 789 
 790         error = memory_bm_find_bit(bm, pfn, &addr, &bit);
 791         if (!error)
 792                 set_bit(bit, addr);
 793 
 794         return error;
 795 }
 796 
 797 static void memory_bm_clear_bit(struct memory_bitmap *bm, unsigned long pfn)
 798 {
 799         void *addr;
 800         unsigned int bit;
 801         int error;
 802 
 803         error = memory_bm_find_bit(bm, pfn, &addr, &bit);
 804         BUG_ON(error);
 805         clear_bit(bit, addr);
 806 }
 807 
 808 static void memory_bm_clear_current(struct memory_bitmap *bm)
 809 {
 810         int bit;
 811 
 812         bit = max(bm->cur.node_bit - 1, 0);
 813         clear_bit(bit, bm->cur.node->data);
 814 }
 815 
 816 static int memory_bm_test_bit(struct memory_bitmap *bm, unsigned long pfn)
 817 {
 818         void *addr;
 819         unsigned int bit;
 820         int error;
 821 
 822         error = memory_bm_find_bit(bm, pfn, &addr, &bit);
 823         BUG_ON(error);
 824         return test_bit(bit, addr);
 825 }
 826 
 827 static bool memory_bm_pfn_present(struct memory_bitmap *bm, unsigned long pfn)
 828 {
 829         void *addr;
 830         unsigned int bit;
 831 
 832         return !memory_bm_find_bit(bm, pfn, &addr, &bit);
 833 }
 834 
 835 /*
 836  * rtree_next_node - Jump to the next leaf node.
 837  *
 838  * Set the position to the beginning of the next node in the
 839  * memory bitmap. This is either the next node in the current
 840  * zone's radix tree or the first node in the radix tree of the
 841  * next zone.
 842  *
 843  * Return true if there is a next node, false otherwise.
 844  */
 845 static bool rtree_next_node(struct memory_bitmap *bm)
 846 {
 847         if (!list_is_last(&bm->cur.node->list, &bm->cur.zone->leaves)) {
 848                 bm->cur.node = list_entry(bm->cur.node->list.next,
 849                                           struct rtree_node, list);
 850                 bm->cur.node_pfn += BM_BITS_PER_BLOCK;
 851                 bm->cur.node_bit  = 0;
 852                 touch_softlockup_watchdog();
 853                 return true;
 854         }
 855 
 856         /* No more nodes, goto next zone */
 857         if (!list_is_last(&bm->cur.zone->list, &bm->zones)) {
 858                 bm->cur.zone = list_entry(bm->cur.zone->list.next,
 859                                   struct mem_zone_bm_rtree, list);
 860                 bm->cur.node = list_entry(bm->cur.zone->leaves.next,
 861                                           struct rtree_node, list);
 862                 bm->cur.node_pfn = 0;
 863                 bm->cur.node_bit = 0;
 864                 return true;
 865         }
 866 
 867         /* No more zones */
 868         return false;
 869 }
 870 
 871 /**
 872  * memory_bm_rtree_next_pfn - Find the next set bit in a memory bitmap.
 873  * @bm: Memory bitmap.
 874  *
 875  * Starting from the last returned position this function searches for the next
 876  * set bit in @bm and returns the PFN represented by it.  If no more bits are
 877  * set, BM_END_OF_MAP is returned.
 878  *
 879  * It is required to run memory_bm_position_reset() before the first call to
 880  * this function for the given memory bitmap.
 881  */
 882 static unsigned long memory_bm_next_pfn(struct memory_bitmap *bm)
 883 {
 884         unsigned long bits, pfn, pages;
 885         int bit;
 886 
 887         do {
 888                 pages     = bm->cur.zone->end_pfn - bm->cur.zone->start_pfn;
 889                 bits      = min(pages - bm->cur.node_pfn, BM_BITS_PER_BLOCK);
 890                 bit       = find_next_bit(bm->cur.node->data, bits,
 891                                           bm->cur.node_bit);
 892                 if (bit < bits) {
 893                         pfn = bm->cur.zone->start_pfn + bm->cur.node_pfn + bit;
 894                         bm->cur.node_bit = bit + 1;
 895                         return pfn;
 896                 }
 897         } while (rtree_next_node(bm));
 898 
 899         return BM_END_OF_MAP;
 900 }
 901 
 902 /*
 903  * This structure represents a range of page frames the contents of which
 904  * should not be saved during hibernation.
 905  */
 906 struct nosave_region {
 907         struct list_head list;
 908         unsigned long start_pfn;
 909         unsigned long end_pfn;
 910 };
 911 
 912 static LIST_HEAD(nosave_regions);
 913 
 914 static void recycle_zone_bm_rtree(struct mem_zone_bm_rtree *zone)
 915 {
 916         struct rtree_node *node;
 917 
 918         list_for_each_entry(node, &zone->nodes, list)
 919                 recycle_safe_page(node->data);
 920 
 921         list_for_each_entry(node, &zone->leaves, list)
 922                 recycle_safe_page(node->data);
 923 }
 924 
 925 static void memory_bm_recycle(struct memory_bitmap *bm)
 926 {
 927         struct mem_zone_bm_rtree *zone;
 928         struct linked_page *p_list;
 929 
 930         list_for_each_entry(zone, &bm->zones, list)
 931                 recycle_zone_bm_rtree(zone);
 932 
 933         p_list = bm->p_list;
 934         while (p_list) {
 935                 struct linked_page *lp = p_list;
 936 
 937                 p_list = lp->next;
 938                 recycle_safe_page(lp);
 939         }
 940 }
 941 
 942 /**
 943  * register_nosave_region - Register a region of unsaveable memory.
 944  *
 945  * Register a range of page frames the contents of which should not be saved
 946  * during hibernation (to be used in the early initialization code).
 947  */
 948 void __init __register_nosave_region(unsigned long start_pfn,
 949                                      unsigned long end_pfn, int use_kmalloc)
 950 {
 951         struct nosave_region *region;
 952 
 953         if (start_pfn >= end_pfn)
 954                 return;
 955 
 956         if (!list_empty(&nosave_regions)) {
 957                 /* Try to extend the previous region (they should be sorted) */
 958                 region = list_entry(nosave_regions.prev,
 959                                         struct nosave_region, list);
 960                 if (region->end_pfn == start_pfn) {
 961                         region->end_pfn = end_pfn;
 962                         goto Report;
 963                 }
 964         }
 965         if (use_kmalloc) {
 966                 /* During init, this shouldn't fail */
 967                 region = kmalloc(sizeof(struct nosave_region), GFP_KERNEL);
 968                 BUG_ON(!region);
 969         } else {
 970                 /* This allocation cannot fail */
 971                 region = memblock_alloc(sizeof(struct nosave_region),
 972                                         SMP_CACHE_BYTES);
 973                 if (!region)
 974                         panic("%s: Failed to allocate %zu bytes\n", __func__,
 975                               sizeof(struct nosave_region));
 976         }
 977         region->start_pfn = start_pfn;
 978         region->end_pfn = end_pfn;
 979         list_add_tail(&region->list, &nosave_regions);
 980  Report:
 981         pr_info("Registered nosave memory: [mem %#010llx-%#010llx]\n",
 982                 (unsigned long long) start_pfn << PAGE_SHIFT,
 983                 ((unsigned long long) end_pfn << PAGE_SHIFT) - 1);
 984 }
 985 
 986 /*
 987  * Set bits in this map correspond to the page frames the contents of which
 988  * should not be saved during the suspend.
 989  */
 990 static struct memory_bitmap *forbidden_pages_map;
 991 
 992 /* Set bits in this map correspond to free page frames. */
 993 static struct memory_bitmap *free_pages_map;
 994 
 995 /*
 996  * Each page frame allocated for creating the image is marked by setting the
 997  * corresponding bits in forbidden_pages_map and free_pages_map simultaneously
 998  */
 999 
1000 void swsusp_set_page_free(struct page *page)
1001 {
1002         if (free_pages_map)
1003                 memory_bm_set_bit(free_pages_map, page_to_pfn(page));
1004 }
1005 
1006 static int swsusp_page_is_free(struct page *page)
1007 {
1008         return free_pages_map ?
1009                 memory_bm_test_bit(free_pages_map, page_to_pfn(page)) : 0;
1010 }
1011 
1012 void swsusp_unset_page_free(struct page *page)
1013 {
1014         if (free_pages_map)
1015                 memory_bm_clear_bit(free_pages_map, page_to_pfn(page));
1016 }
1017 
1018 static void swsusp_set_page_forbidden(struct page *page)
1019 {
1020         if (forbidden_pages_map)
1021                 memory_bm_set_bit(forbidden_pages_map, page_to_pfn(page));
1022 }
1023 
1024 int swsusp_page_is_forbidden(struct page *page)
1025 {
1026         return forbidden_pages_map ?
1027                 memory_bm_test_bit(forbidden_pages_map, page_to_pfn(page)) : 0;
1028 }
1029 
1030 static void swsusp_unset_page_forbidden(struct page *page)
1031 {
1032         if (forbidden_pages_map)
1033                 memory_bm_clear_bit(forbidden_pages_map, page_to_pfn(page));
1034 }
1035 
1036 /**
1037  * mark_nosave_pages - Mark pages that should not be saved.
1038  * @bm: Memory bitmap.
1039  *
1040  * Set the bits in @bm that correspond to the page frames the contents of which
1041  * should not be saved.
1042  */
1043 static void mark_nosave_pages(struct memory_bitmap *bm)
1044 {
1045         struct nosave_region *region;
1046 
1047         if (list_empty(&nosave_regions))
1048                 return;
1049 
1050         list_for_each_entry(region, &nosave_regions, list) {
1051                 unsigned long pfn;
1052 
1053                 pr_debug("Marking nosave pages: [mem %#010llx-%#010llx]\n",
1054                          (unsigned long long) region->start_pfn << PAGE_SHIFT,
1055                          ((unsigned long long) region->end_pfn << PAGE_SHIFT)
1056                                 - 1);
1057 
1058                 for (pfn = region->start_pfn; pfn < region->end_pfn; pfn++)
1059                         if (pfn_valid(pfn)) {
1060                                 /*
1061                                  * It is safe to ignore the result of
1062                                  * mem_bm_set_bit_check() here, since we won't
1063                                  * touch the PFNs for which the error is
1064                                  * returned anyway.
1065                                  */
1066                                 mem_bm_set_bit_check(bm, pfn);
1067                         }
1068         }
1069 }
1070 
1071 /**
1072  * create_basic_memory_bitmaps - Create bitmaps to hold basic page information.
1073  *
1074  * Create bitmaps needed for marking page frames that should not be saved and
1075  * free page frames.  The forbidden_pages_map and free_pages_map pointers are
1076  * only modified if everything goes well, because we don't want the bits to be
1077  * touched before both bitmaps are set up.
1078  */
1079 int create_basic_memory_bitmaps(void)
1080 {
1081         struct memory_bitmap *bm1, *bm2;
1082         int error = 0;
1083 
1084         if (forbidden_pages_map && free_pages_map)
1085                 return 0;
1086         else
1087                 BUG_ON(forbidden_pages_map || free_pages_map);
1088 
1089         bm1 = kzalloc(sizeof(struct memory_bitmap), GFP_KERNEL);
1090         if (!bm1)
1091                 return -ENOMEM;
1092 
1093         error = memory_bm_create(bm1, GFP_KERNEL, PG_ANY);
1094         if (error)
1095                 goto Free_first_object;
1096 
1097         bm2 = kzalloc(sizeof(struct memory_bitmap), GFP_KERNEL);
1098         if (!bm2)
1099                 goto Free_first_bitmap;
1100 
1101         error = memory_bm_create(bm2, GFP_KERNEL, PG_ANY);
1102         if (error)
1103                 goto Free_second_object;
1104 
1105         forbidden_pages_map = bm1;
1106         free_pages_map = bm2;
1107         mark_nosave_pages(forbidden_pages_map);
1108 
1109         pr_debug("Basic memory bitmaps created\n");
1110 
1111         return 0;
1112 
1113  Free_second_object:
1114         kfree(bm2);
1115  Free_first_bitmap:
1116         memory_bm_free(bm1, PG_UNSAFE_CLEAR);
1117  Free_first_object:
1118         kfree(bm1);
1119         return -ENOMEM;
1120 }
1121 
1122 /**
1123  * free_basic_memory_bitmaps - Free memory bitmaps holding basic information.
1124  *
1125  * Free memory bitmaps allocated by create_basic_memory_bitmaps().  The
1126  * auxiliary pointers are necessary so that the bitmaps themselves are not
1127  * referred to while they are being freed.
1128  */
1129 void free_basic_memory_bitmaps(void)
1130 {
1131         struct memory_bitmap *bm1, *bm2;
1132 
1133         if (WARN_ON(!(forbidden_pages_map && free_pages_map)))
1134                 return;
1135 
1136         bm1 = forbidden_pages_map;
1137         bm2 = free_pages_map;
1138         forbidden_pages_map = NULL;
1139         free_pages_map = NULL;
1140         memory_bm_free(bm1, PG_UNSAFE_CLEAR);
1141         kfree(bm1);
1142         memory_bm_free(bm2, PG_UNSAFE_CLEAR);
1143         kfree(bm2);
1144 
1145         pr_debug("Basic memory bitmaps freed\n");
1146 }
1147 
1148 void clear_free_pages(void)
1149 {
1150         struct memory_bitmap *bm = free_pages_map;
1151         unsigned long pfn;
1152 
1153         if (WARN_ON(!(free_pages_map)))
1154                 return;
1155 
1156         if (IS_ENABLED(CONFIG_PAGE_POISONING_ZERO) || want_init_on_free()) {
1157                 memory_bm_position_reset(bm);
1158                 pfn = memory_bm_next_pfn(bm);
1159                 while (pfn != BM_END_OF_MAP) {
1160                         if (pfn_valid(pfn))
1161                                 clear_highpage(pfn_to_page(pfn));
1162 
1163                         pfn = memory_bm_next_pfn(bm);
1164                 }
1165                 memory_bm_position_reset(bm);
1166                 pr_info("free pages cleared after restore\n");
1167         }
1168 }
1169 
1170 /**
1171  * snapshot_additional_pages - Estimate the number of extra pages needed.
1172  * @zone: Memory zone to carry out the computation for.
1173  *
1174  * Estimate the number of additional pages needed for setting up a hibernation
1175  * image data structures for @zone (usually, the returned value is greater than
1176  * the exact number).
1177  */
1178 unsigned int snapshot_additional_pages(struct zone *zone)
1179 {
1180         unsigned int rtree, nodes;
1181 
1182         rtree = nodes = DIV_ROUND_UP(zone->spanned_pages, BM_BITS_PER_BLOCK);
1183         rtree += DIV_ROUND_UP(rtree * sizeof(struct rtree_node),
1184                               LINKED_PAGE_DATA_SIZE);
1185         while (nodes > 1) {
1186                 nodes = DIV_ROUND_UP(nodes, BM_ENTRIES_PER_LEVEL);
1187                 rtree += nodes;
1188         }
1189 
1190         return 2 * rtree;
1191 }
1192 
1193 #ifdef CONFIG_HIGHMEM
1194 /**
1195  * count_free_highmem_pages - Compute the total number of free highmem pages.
1196  *
1197  * The returned number is system-wide.
1198  */
1199 static unsigned int count_free_highmem_pages(void)
1200 {
1201         struct zone *zone;
1202         unsigned int cnt = 0;
1203 
1204         for_each_populated_zone(zone)
1205                 if (is_highmem(zone))
1206                         cnt += zone_page_state(zone, NR_FREE_PAGES);
1207 
1208         return cnt;
1209 }
1210 
1211 /**
1212  * saveable_highmem_page - Check if a highmem page is saveable.
1213  *
1214  * Determine whether a highmem page should be included in a hibernation image.
1215  *
1216  * We should save the page if it isn't Nosave or NosaveFree, or Reserved,
1217  * and it isn't part of a free chunk of pages.
1218  */
1219 static struct page *saveable_highmem_page(struct zone *zone, unsigned long pfn)
1220 {
1221         struct page *page;
1222 
1223         if (!pfn_valid(pfn))
1224                 return NULL;
1225 
1226         page = pfn_to_online_page(pfn);
1227         if (!page || page_zone(page) != zone)
1228                 return NULL;
1229 
1230         BUG_ON(!PageHighMem(page));
1231 
1232         if (swsusp_page_is_forbidden(page) ||  swsusp_page_is_free(page))
1233                 return NULL;
1234 
1235         if (PageReserved(page) || PageOffline(page))
1236                 return NULL;
1237 
1238         if (page_is_guard(page))
1239                 return NULL;
1240 
1241         return page;
1242 }
1243 
1244 /**
1245  * count_highmem_pages - Compute the total number of saveable highmem pages.
1246  */
1247 static unsigned int count_highmem_pages(void)
1248 {
1249         struct zone *zone;
1250         unsigned int n = 0;
1251 
1252         for_each_populated_zone(zone) {
1253                 unsigned long pfn, max_zone_pfn;
1254 
1255                 if (!is_highmem(zone))
1256                         continue;
1257 
1258                 mark_free_pages(zone);
1259                 max_zone_pfn = zone_end_pfn(zone);
1260                 for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++)
1261                         if (saveable_highmem_page(zone, pfn))
1262                                 n++;
1263         }
1264         return n;
1265 }
1266 #else
1267 static inline void *saveable_highmem_page(struct zone *z, unsigned long p)
1268 {
1269         return NULL;
1270 }
1271 #endif /* CONFIG_HIGHMEM */
1272 
1273 /**
1274  * saveable_page - Check if the given page is saveable.
1275  *
1276  * Determine whether a non-highmem page should be included in a hibernation
1277  * image.
1278  *
1279  * We should save the page if it isn't Nosave, and is not in the range
1280  * of pages statically defined as 'unsaveable', and it isn't part of
1281  * a free chunk of pages.
1282  */
1283 static struct page *saveable_page(struct zone *zone, unsigned long pfn)
1284 {
1285         struct page *page;
1286 
1287         if (!pfn_valid(pfn))
1288                 return NULL;
1289 
1290         page = pfn_to_online_page(pfn);
1291         if (!page || page_zone(page) != zone)
1292                 return NULL;
1293 
1294         BUG_ON(PageHighMem(page));
1295 
1296         if (swsusp_page_is_forbidden(page) || swsusp_page_is_free(page))
1297                 return NULL;
1298 
1299         if (PageOffline(page))
1300                 return NULL;
1301 
1302         if (PageReserved(page)
1303             && (!kernel_page_present(page) || pfn_is_nosave(pfn)))
1304                 return NULL;
1305 
1306         if (page_is_guard(page))
1307                 return NULL;
1308 
1309         return page;
1310 }
1311 
1312 /**
1313  * count_data_pages - Compute the total number of saveable non-highmem pages.
1314  */
1315 static unsigned int count_data_pages(void)
1316 {
1317         struct zone *zone;
1318         unsigned long pfn, max_zone_pfn;
1319         unsigned int n = 0;
1320 
1321         for_each_populated_zone(zone) {
1322                 if (is_highmem(zone))
1323                         continue;
1324 
1325                 mark_free_pages(zone);
1326                 max_zone_pfn = zone_end_pfn(zone);
1327                 for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++)
1328                         if (saveable_page(zone, pfn))
1329                                 n++;
1330         }
1331         return n;
1332 }
1333 
1334 /*
1335  * This is needed, because copy_page and memcpy are not usable for copying
1336  * task structs.
1337  */
1338 static inline void do_copy_page(long *dst, long *src)
1339 {
1340         int n;
1341 
1342         for (n = PAGE_SIZE / sizeof(long); n; n--)
1343                 *dst++ = *src++;
1344 }
1345 
1346 /**
1347  * safe_copy_page - Copy a page in a safe way.
1348  *
1349  * Check if the page we are going to copy is marked as present in the kernel
1350  * page tables. This always is the case if CONFIG_DEBUG_PAGEALLOC or
1351  * CONFIG_ARCH_HAS_SET_DIRECT_MAP is not set. In that case kernel_page_present()
1352  * always returns 'true'.
1353  */
1354 static void safe_copy_page(void *dst, struct page *s_page)
1355 {
1356         if (kernel_page_present(s_page)) {
1357                 do_copy_page(dst, page_address(s_page));
1358         } else {
1359                 kernel_map_pages(s_page, 1, 1);
1360                 do_copy_page(dst, page_address(s_page));
1361                 kernel_map_pages(s_page, 1, 0);
1362         }
1363 }
1364 
1365 #ifdef CONFIG_HIGHMEM
1366 static inline struct page *page_is_saveable(struct zone *zone, unsigned long pfn)
1367 {
1368         return is_highmem(zone) ?
1369                 saveable_highmem_page(zone, pfn) : saveable_page(zone, pfn);
1370 }
1371 
1372 static void copy_data_page(unsigned long dst_pfn, unsigned long src_pfn)
1373 {
1374         struct page *s_page, *d_page;
1375         void *src, *dst;
1376 
1377         s_page = pfn_to_page(src_pfn);
1378         d_page = pfn_to_page(dst_pfn);
1379         if (PageHighMem(s_page)) {
1380                 src = kmap_atomic(s_page);
1381                 dst = kmap_atomic(d_page);
1382                 do_copy_page(dst, src);
1383                 kunmap_atomic(dst);
1384                 kunmap_atomic(src);
1385         } else {
1386                 if (PageHighMem(d_page)) {
1387                         /*
1388                          * The page pointed to by src may contain some kernel
1389                          * data modified by kmap_atomic()
1390                          */
1391                         safe_copy_page(buffer, s_page);
1392                         dst = kmap_atomic(d_page);
1393                         copy_page(dst, buffer);
1394                         kunmap_atomic(dst);
1395                 } else {
1396                         safe_copy_page(page_address(d_page), s_page);
1397                 }
1398         }
1399 }
1400 #else
1401 #define page_is_saveable(zone, pfn)     saveable_page(zone, pfn)
1402 
1403 static inline void copy_data_page(unsigned long dst_pfn, unsigned long src_pfn)
1404 {
1405         safe_copy_page(page_address(pfn_to_page(dst_pfn)),
1406                                 pfn_to_page(src_pfn));
1407 }
1408 #endif /* CONFIG_HIGHMEM */
1409 
1410 static void copy_data_pages(struct memory_bitmap *copy_bm,
1411                             struct memory_bitmap *orig_bm)
1412 {
1413         struct zone *zone;
1414         unsigned long pfn;
1415 
1416         for_each_populated_zone(zone) {
1417                 unsigned long max_zone_pfn;
1418 
1419                 mark_free_pages(zone);
1420                 max_zone_pfn = zone_end_pfn(zone);
1421                 for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++)
1422                         if (page_is_saveable(zone, pfn))
1423                                 memory_bm_set_bit(orig_bm, pfn);
1424         }
1425         memory_bm_position_reset(orig_bm);
1426         memory_bm_position_reset(copy_bm);
1427         for(;;) {
1428                 pfn = memory_bm_next_pfn(orig_bm);
1429                 if (unlikely(pfn == BM_END_OF_MAP))
1430                         break;
1431                 copy_data_page(memory_bm_next_pfn(copy_bm), pfn);
1432         }
1433 }
1434 
1435 /* Total number of image pages */
1436 static unsigned int nr_copy_pages;
1437 /* Number of pages needed for saving the original pfns of the image pages */
1438 static unsigned int nr_meta_pages;
1439 /*
1440  * Numbers of normal and highmem page frames allocated for hibernation image
1441  * before suspending devices.
1442  */
1443 static unsigned int alloc_normal, alloc_highmem;
1444 /*
1445  * Memory bitmap used for marking saveable pages (during hibernation) or
1446  * hibernation image pages (during restore)
1447  */
1448 static struct memory_bitmap orig_bm;
1449 /*
1450  * Memory bitmap used during hibernation for marking allocated page frames that
1451  * will contain copies of saveable pages.  During restore it is initially used
1452  * for marking hibernation image pages, but then the set bits from it are
1453  * duplicated in @orig_bm and it is released.  On highmem systems it is next
1454  * used for marking "safe" highmem pages, but it has to be reinitialized for
1455  * this purpose.
1456  */
1457 static struct memory_bitmap copy_bm;
1458 
1459 /**
1460  * swsusp_free - Free pages allocated for hibernation image.
1461  *
1462  * Image pages are alocated before snapshot creation, so they need to be
1463  * released after resume.
1464  */
1465 void swsusp_free(void)
1466 {
1467         unsigned long fb_pfn, fr_pfn;
1468 
1469         if (!forbidden_pages_map || !free_pages_map)
1470                 goto out;
1471 
1472         memory_bm_position_reset(forbidden_pages_map);
1473         memory_bm_position_reset(free_pages_map);
1474 
1475 loop:
1476         fr_pfn = memory_bm_next_pfn(free_pages_map);
1477         fb_pfn = memory_bm_next_pfn(forbidden_pages_map);
1478 
1479         /*
1480          * Find the next bit set in both bitmaps. This is guaranteed to
1481          * terminate when fb_pfn == fr_pfn == BM_END_OF_MAP.
1482          */
1483         do {
1484                 if (fb_pfn < fr_pfn)
1485                         fb_pfn = memory_bm_next_pfn(forbidden_pages_map);
1486                 if (fr_pfn < fb_pfn)
1487                         fr_pfn = memory_bm_next_pfn(free_pages_map);
1488         } while (fb_pfn != fr_pfn);
1489 
1490         if (fr_pfn != BM_END_OF_MAP && pfn_valid(fr_pfn)) {
1491                 struct page *page = pfn_to_page(fr_pfn);
1492 
1493                 memory_bm_clear_current(forbidden_pages_map);
1494                 memory_bm_clear_current(free_pages_map);
1495                 hibernate_restore_unprotect_page(page_address(page));
1496                 __free_page(page);
1497                 goto loop;
1498         }
1499 
1500 out:
1501         nr_copy_pages = 0;
1502         nr_meta_pages = 0;
1503         restore_pblist = NULL;
1504         buffer = NULL;
1505         alloc_normal = 0;
1506         alloc_highmem = 0;
1507         hibernate_restore_protection_end();
1508 }
1509 
1510 /* Helper functions used for the shrinking of memory. */
1511 
1512 #define GFP_IMAGE       (GFP_KERNEL | __GFP_NOWARN)
1513 
1514 /**
1515  * preallocate_image_pages - Allocate a number of pages for hibernation image.
1516  * @nr_pages: Number of page frames to allocate.
1517  * @mask: GFP flags to use for the allocation.
1518  *
1519  * Return value: Number of page frames actually allocated
1520  */
1521 static unsigned long preallocate_image_pages(unsigned long nr_pages, gfp_t mask)
1522 {
1523         unsigned long nr_alloc = 0;
1524 
1525         while (nr_pages > 0) {
1526                 struct page *page;
1527 
1528                 page = alloc_image_page(mask);
1529                 if (!page)
1530                         break;
1531                 memory_bm_set_bit(&copy_bm, page_to_pfn(page));
1532                 if (PageHighMem(page))
1533                         alloc_highmem++;
1534                 else
1535                         alloc_normal++;
1536                 nr_pages--;
1537                 nr_alloc++;
1538         }
1539 
1540         return nr_alloc;
1541 }
1542 
1543 static unsigned long preallocate_image_memory(unsigned long nr_pages,
1544                                               unsigned long avail_normal)
1545 {
1546         unsigned long alloc;
1547 
1548         if (avail_normal <= alloc_normal)
1549                 return 0;
1550 
1551         alloc = avail_normal - alloc_normal;
1552         if (nr_pages < alloc)
1553                 alloc = nr_pages;
1554 
1555         return preallocate_image_pages(alloc, GFP_IMAGE);
1556 }
1557 
1558 #ifdef CONFIG_HIGHMEM
1559 static unsigned long preallocate_image_highmem(unsigned long nr_pages)
1560 {
1561         return preallocate_image_pages(nr_pages, GFP_IMAGE | __GFP_HIGHMEM);
1562 }
1563 
1564 /**
1565  *  __fraction - Compute (an approximation of) x * (multiplier / base).
1566  */
1567 static unsigned long __fraction(u64 x, u64 multiplier, u64 base)
1568 {
1569         x *= multiplier;
1570         do_div(x, base);
1571         return (unsigned long)x;
1572 }
1573 
1574 static unsigned long preallocate_highmem_fraction(unsigned long nr_pages,
1575                                                   unsigned long highmem,
1576                                                   unsigned long total)
1577 {
1578         unsigned long alloc = __fraction(nr_pages, highmem, total);
1579 
1580         return preallocate_image_pages(alloc, GFP_IMAGE | __GFP_HIGHMEM);
1581 }
1582 #else /* CONFIG_HIGHMEM */
1583 static inline unsigned long preallocate_image_highmem(unsigned long nr_pages)
1584 {
1585         return 0;
1586 }
1587 
1588 static inline unsigned long preallocate_highmem_fraction(unsigned long nr_pages,
1589                                                          unsigned long highmem,
1590                                                          unsigned long total)
1591 {
1592         return 0;
1593 }
1594 #endif /* CONFIG_HIGHMEM */
1595 
1596 /**
1597  * free_unnecessary_pages - Release preallocated pages not needed for the image.
1598  */
1599 static unsigned long free_unnecessary_pages(void)
1600 {
1601         unsigned long save, to_free_normal, to_free_highmem, free;
1602 
1603         save = count_data_pages();
1604         if (alloc_normal >= save) {
1605                 to_free_normal = alloc_normal - save;
1606                 save = 0;
1607         } else {
1608                 to_free_normal = 0;
1609                 save -= alloc_normal;
1610         }
1611         save += count_highmem_pages();
1612         if (alloc_highmem >= save) {
1613                 to_free_highmem = alloc_highmem - save;
1614         } else {
1615                 to_free_highmem = 0;
1616                 save -= alloc_highmem;
1617                 if (to_free_normal > save)
1618                         to_free_normal -= save;
1619                 else
1620                         to_free_normal = 0;
1621         }
1622         free = to_free_normal + to_free_highmem;
1623 
1624         memory_bm_position_reset(&copy_bm);
1625 
1626         while (to_free_normal > 0 || to_free_highmem > 0) {
1627                 unsigned long pfn = memory_bm_next_pfn(&copy_bm);
1628                 struct page *page = pfn_to_page(pfn);
1629 
1630                 if (PageHighMem(page)) {
1631                         if (!to_free_highmem)
1632                                 continue;
1633                         to_free_highmem--;
1634                         alloc_highmem--;
1635                 } else {
1636                         if (!to_free_normal)
1637                                 continue;
1638                         to_free_normal--;
1639                         alloc_normal--;
1640                 }
1641                 memory_bm_clear_bit(&copy_bm, pfn);
1642                 swsusp_unset_page_forbidden(page);
1643                 swsusp_unset_page_free(page);
1644                 __free_page(page);
1645         }
1646 
1647         return free;
1648 }
1649 
1650 /**
1651  * minimum_image_size - Estimate the minimum acceptable size of an image.
1652  * @saveable: Number of saveable pages in the system.
1653  *
1654  * We want to avoid attempting to free too much memory too hard, so estimate the
1655  * minimum acceptable size of a hibernation image to use as the lower limit for
1656  * preallocating memory.
1657  *
1658  * We assume that the minimum image size should be proportional to
1659  *
1660  * [number of saveable pages] - [number of pages that can be freed in theory]
1661  *
1662  * where the second term is the sum of (1) reclaimable slab pages, (2) active
1663  * and (3) inactive anonymous pages, (4) active and (5) inactive file pages.
1664  */
1665 static unsigned long minimum_image_size(unsigned long saveable)
1666 {
1667         unsigned long size;
1668 
1669         size = global_node_page_state(NR_SLAB_RECLAIMABLE)
1670                 + global_node_page_state(NR_ACTIVE_ANON)
1671                 + global_node_page_state(NR_INACTIVE_ANON)
1672                 + global_node_page_state(NR_ACTIVE_FILE)
1673                 + global_node_page_state(NR_INACTIVE_FILE);
1674 
1675         return saveable <= size ? 0 : saveable - size;
1676 }
1677 
1678 /**
1679  * hibernate_preallocate_memory - Preallocate memory for hibernation image.
1680  *
1681  * To create a hibernation image it is necessary to make a copy of every page
1682  * frame in use.  We also need a number of page frames to be free during
1683  * hibernation for allocations made while saving the image and for device
1684  * drivers, in case they need to allocate memory from their hibernation
1685  * callbacks (these two numbers are given by PAGES_FOR_IO (which is a rough
1686  * estimate) and reserverd_size divided by PAGE_SIZE (which is tunable through
1687  * /sys/power/reserved_size, respectively).  To make this happen, we compute the
1688  * total number of available page frames and allocate at least
1689  *
1690  * ([page frames total] + PAGES_FOR_IO + [metadata pages]) / 2
1691  *  + 2 * DIV_ROUND_UP(reserved_size, PAGE_SIZE)
1692  *
1693  * of them, which corresponds to the maximum size of a hibernation image.
1694  *
1695  * If image_size is set below the number following from the above formula,
1696  * the preallocation of memory is continued until the total number of saveable
1697  * pages in the system is below the requested image size or the minimum
1698  * acceptable image size returned by minimum_image_size(), whichever is greater.
1699  */
1700 int hibernate_preallocate_memory(void)
1701 {
1702         struct zone *zone;
1703         unsigned long saveable, size, max_size, count, highmem, pages = 0;
1704         unsigned long alloc, save_highmem, pages_highmem, avail_normal;
1705         ktime_t start, stop;
1706         int error;
1707 
1708         pr_info("Preallocating image memory... ");
1709         start = ktime_get();
1710 
1711         error = memory_bm_create(&orig_bm, GFP_IMAGE, PG_ANY);
1712         if (error)
1713                 goto err_out;
1714 
1715         error = memory_bm_create(&copy_bm, GFP_IMAGE, PG_ANY);
1716         if (error)
1717                 goto err_out;
1718 
1719         alloc_normal = 0;
1720         alloc_highmem = 0;
1721 
1722         /* Count the number of saveable data pages. */
1723         save_highmem = count_highmem_pages();
1724         saveable = count_data_pages();
1725 
1726         /*
1727          * Compute the total number of page frames we can use (count) and the
1728          * number of pages needed for image metadata (size).
1729          */
1730         count = saveable;
1731         saveable += save_highmem;
1732         highmem = save_highmem;
1733         size = 0;
1734         for_each_populated_zone(zone) {
1735                 size += snapshot_additional_pages(zone);
1736                 if (is_highmem(zone))
1737                         highmem += zone_page_state(zone, NR_FREE_PAGES);
1738                 else
1739                         count += zone_page_state(zone, NR_FREE_PAGES);
1740         }
1741         avail_normal = count;
1742         count += highmem;
1743         count -= totalreserve_pages;
1744 
1745         /* Add number of pages required for page keys (s390 only). */
1746         size += page_key_additional_pages(saveable);
1747 
1748         /* Compute the maximum number of saveable pages to leave in memory. */
1749         max_size = (count - (size + PAGES_FOR_IO)) / 2
1750                         - 2 * DIV_ROUND_UP(reserved_size, PAGE_SIZE);
1751         /* Compute the desired number of image pages specified by image_size. */
1752         size = DIV_ROUND_UP(image_size, PAGE_SIZE);
1753         if (size > max_size)
1754                 size = max_size;
1755         /*
1756          * If the desired number of image pages is at least as large as the
1757          * current number of saveable pages in memory, allocate page frames for
1758          * the image and we're done.
1759          */
1760         if (size >= saveable) {
1761                 pages = preallocate_image_highmem(save_highmem);
1762                 pages += preallocate_image_memory(saveable - pages, avail_normal);
1763                 goto out;
1764         }
1765 
1766         /* Estimate the minimum size of the image. */
1767         pages = minimum_image_size(saveable);
1768         /*
1769          * To avoid excessive pressure on the normal zone, leave room in it to
1770          * accommodate an image of the minimum size (unless it's already too
1771          * small, in which case don't preallocate pages from it at all).
1772          */
1773         if (avail_normal > pages)
1774                 avail_normal -= pages;
1775         else
1776                 avail_normal = 0;
1777         if (size < pages)
1778                 size = min_t(unsigned long, pages, max_size);
1779 
1780         /*
1781          * Let the memory management subsystem know that we're going to need a
1782          * large number of page frames to allocate and make it free some memory.
1783          * NOTE: If this is not done, performance will be hurt badly in some
1784          * test cases.
1785          */
1786         shrink_all_memory(saveable - size);
1787 
1788         /*
1789          * The number of saveable pages in memory was too high, so apply some
1790          * pressure to decrease it.  First, make room for the largest possible
1791          * image and fail if that doesn't work.  Next, try to decrease the size
1792          * of the image as much as indicated by 'size' using allocations from
1793          * highmem and non-highmem zones separately.
1794          */
1795         pages_highmem = preallocate_image_highmem(highmem / 2);
1796         alloc = count - max_size;
1797         if (alloc > pages_highmem)
1798                 alloc -= pages_highmem;
1799         else
1800                 alloc = 0;
1801         pages = preallocate_image_memory(alloc, avail_normal);
1802         if (pages < alloc) {
1803                 /* We have exhausted non-highmem pages, try highmem. */
1804                 alloc -= pages;
1805                 pages += pages_highmem;
1806                 pages_highmem = preallocate_image_highmem(alloc);
1807                 if (pages_highmem < alloc)
1808                         goto err_out;
1809                 pages += pages_highmem;
1810                 /*
1811                  * size is the desired number of saveable pages to leave in
1812                  * memory, so try to preallocate (all memory - size) pages.
1813                  */
1814                 alloc = (count - pages) - size;
1815                 pages += preallocate_image_highmem(alloc);
1816         } else {
1817                 /*
1818                  * There are approximately max_size saveable pages at this point
1819                  * and we want to reduce this number down to size.
1820                  */
1821                 alloc = max_size - size;
1822                 size = preallocate_highmem_fraction(alloc, highmem, count);
1823                 pages_highmem += size;
1824                 alloc -= size;
1825                 size = preallocate_image_memory(alloc, avail_normal);
1826                 pages_highmem += preallocate_image_highmem(alloc - size);
1827                 pages += pages_highmem + size;
1828         }
1829 
1830         /*
1831          * We only need as many page frames for the image as there are saveable
1832          * pages in memory, but we have allocated more.  Release the excessive
1833          * ones now.
1834          */
1835         pages -= free_unnecessary_pages();
1836 
1837  out:
1838         stop = ktime_get();
1839         pr_cont("done (allocated %lu pages)\n", pages);
1840         swsusp_show_speed(start, stop, pages, "Allocated");
1841 
1842         return 0;
1843 
1844  err_out:
1845         pr_cont("\n");
1846         swsusp_free();
1847         return -ENOMEM;
1848 }
1849 
1850 #ifdef CONFIG_HIGHMEM
1851 /**
1852  * count_pages_for_highmem - Count non-highmem pages needed for copying highmem.
1853  *
1854  * Compute the number of non-highmem pages that will be necessary for creating
1855  * copies of highmem pages.
1856  */
1857 static unsigned int count_pages_for_highmem(unsigned int nr_highmem)
1858 {
1859         unsigned int free_highmem = count_free_highmem_pages() + alloc_highmem;
1860 
1861         if (free_highmem >= nr_highmem)
1862                 nr_highmem = 0;
1863         else
1864                 nr_highmem -= free_highmem;
1865 
1866         return nr_highmem;
1867 }
1868 #else
1869 static unsigned int count_pages_for_highmem(unsigned int nr_highmem) { return 0; }
1870 #endif /* CONFIG_HIGHMEM */
1871 
1872 /**
1873  * enough_free_mem - Check if there is enough free memory for the image.
1874  */
1875 static int enough_free_mem(unsigned int nr_pages, unsigned int nr_highmem)
1876 {
1877         struct zone *zone;
1878         unsigned int free = alloc_normal;
1879 
1880         for_each_populated_zone(zone)
1881                 if (!is_highmem(zone))
1882                         free += zone_page_state(zone, NR_FREE_PAGES);
1883 
1884         nr_pages += count_pages_for_highmem(nr_highmem);
1885         pr_debug("Normal pages needed: %u + %u, available pages: %u\n",
1886                  nr_pages, PAGES_FOR_IO, free);
1887 
1888         return free > nr_pages + PAGES_FOR_IO;
1889 }
1890 
1891 #ifdef CONFIG_HIGHMEM
1892 /**
1893  * get_highmem_buffer - Allocate a buffer for highmem pages.
1894  *
1895  * If there are some highmem pages in the hibernation image, we may need a
1896  * buffer to copy them and/or load their data.
1897  */
1898 static inline int get_highmem_buffer(int safe_needed)
1899 {
1900         buffer = get_image_page(GFP_ATOMIC, safe_needed);
1901         return buffer ? 0 : -ENOMEM;
1902 }
1903 
1904 /**
1905  * alloc_highmem_image_pages - Allocate some highmem pages for the image.
1906  *
1907  * Try to allocate as many pages as needed, but if the number of free highmem
1908  * pages is less than that, allocate them all.
1909  */
1910 static inline unsigned int alloc_highmem_pages(struct memory_bitmap *bm,
1911                                                unsigned int nr_highmem)
1912 {
1913         unsigned int to_alloc = count_free_highmem_pages();
1914 
1915         if (to_alloc > nr_highmem)
1916                 to_alloc = nr_highmem;
1917 
1918         nr_highmem -= to_alloc;
1919         while (to_alloc-- > 0) {
1920                 struct page *page;
1921 
1922                 page = alloc_image_page(__GFP_HIGHMEM|__GFP_KSWAPD_RECLAIM);
1923                 memory_bm_set_bit(bm, page_to_pfn(page));
1924         }
1925         return nr_highmem;
1926 }
1927 #else
1928 static inline int get_highmem_buffer(int safe_needed) { return 0; }
1929 
1930 static inline unsigned int alloc_highmem_pages(struct memory_bitmap *bm,
1931                                                unsigned int n) { return 0; }
1932 #endif /* CONFIG_HIGHMEM */
1933 
1934 /**
1935  * swsusp_alloc - Allocate memory for hibernation image.
1936  *
1937  * We first try to allocate as many highmem pages as there are
1938  * saveable highmem pages in the system.  If that fails, we allocate
1939  * non-highmem pages for the copies of the remaining highmem ones.
1940  *
1941  * In this approach it is likely that the copies of highmem pages will
1942  * also be located in the high memory, because of the way in which
1943  * copy_data_pages() works.
1944  */
1945 static int swsusp_alloc(struct memory_bitmap *copy_bm,
1946                         unsigned int nr_pages, unsigned int nr_highmem)
1947 {
1948         if (nr_highmem > 0) {
1949                 if (get_highmem_buffer(PG_ANY))
1950                         goto err_out;
1951                 if (nr_highmem > alloc_highmem) {
1952                         nr_highmem -= alloc_highmem;
1953                         nr_pages += alloc_highmem_pages(copy_bm, nr_highmem);
1954                 }
1955         }
1956         if (nr_pages > alloc_normal) {
1957                 nr_pages -= alloc_normal;
1958                 while (nr_pages-- > 0) {
1959                         struct page *page;
1960 
1961                         page = alloc_image_page(GFP_ATOMIC);
1962                         if (!page)
1963                                 goto err_out;
1964                         memory_bm_set_bit(copy_bm, page_to_pfn(page));
1965                 }
1966         }
1967 
1968         return 0;
1969 
1970  err_out:
1971         swsusp_free();
1972         return -ENOMEM;
1973 }
1974 
1975 asmlinkage __visible int swsusp_save(void)
1976 {
1977         unsigned int nr_pages, nr_highmem;
1978 
1979         pr_info("Creating hibernation image:\n");
1980 
1981         drain_local_pages(NULL);
1982         nr_pages = count_data_pages();
1983         nr_highmem = count_highmem_pages();
1984         pr_info("Need to copy %u pages\n", nr_pages + nr_highmem);
1985 
1986         if (!enough_free_mem(nr_pages, nr_highmem)) {
1987                 pr_err("Not enough free memory\n");
1988                 return -ENOMEM;
1989         }
1990 
1991         if (swsusp_alloc(&copy_bm, nr_pages, nr_highmem)) {
1992                 pr_err("Memory allocation failed\n");
1993                 return -ENOMEM;
1994         }
1995 
1996         /*
1997          * During allocating of suspend pagedir, new cold pages may appear.
1998          * Kill them.
1999          */
2000         drain_local_pages(NULL);
2001         copy_data_pages(&copy_bm, &orig_bm);
2002 
2003         /*
2004          * End of critical section. From now on, we can write to memory,
2005          * but we should not touch disk. This specially means we must _not_
2006          * touch swap space! Except we must write out our image of course.
2007          */
2008 
2009         nr_pages += nr_highmem;
2010         nr_copy_pages = nr_pages;
2011         nr_meta_pages = DIV_ROUND_UP(nr_pages * sizeof(long), PAGE_SIZE);
2012 
2013         pr_info("Hibernation image created (%d pages copied)\n", nr_pages);
2014 
2015         return 0;
2016 }
2017 
2018 #ifndef CONFIG_ARCH_HIBERNATION_HEADER
2019 static int init_header_complete(struct swsusp_info *info)
2020 {
2021         memcpy(&info->uts, init_utsname(), sizeof(struct new_utsname));
2022         info->version_code = LINUX_VERSION_CODE;
2023         return 0;
2024 }
2025 
2026 static char *check_image_kernel(struct swsusp_info *info)
2027 {
2028         if (info->version_code != LINUX_VERSION_CODE)
2029                 return "kernel version";
2030         if (strcmp(info->uts.sysname,init_utsname()->sysname))
2031                 return "system type";
2032         if (strcmp(info->uts.release,init_utsname()->release))
2033                 return "kernel release";
2034         if (strcmp(info->uts.version,init_utsname()->version))
2035                 return "version";
2036         if (strcmp(info->uts.machine,init_utsname()->machine))
2037                 return "machine";
2038         return NULL;
2039 }
2040 #endif /* CONFIG_ARCH_HIBERNATION_HEADER */
2041 
2042 unsigned long snapshot_get_image_size(void)
2043 {
2044         return nr_copy_pages + nr_meta_pages + 1;
2045 }
2046 
2047 static int init_header(struct swsusp_info *info)
2048 {
2049         memset(info, 0, sizeof(struct swsusp_info));
2050         info->num_physpages = get_num_physpages();
2051         info->image_pages = nr_copy_pages;
2052         info->pages = snapshot_get_image_size();
2053         info->size = info->pages;
2054         info->size <<= PAGE_SHIFT;
2055         return init_header_complete(info);
2056 }
2057 
2058 /**
2059  * pack_pfns - Prepare PFNs for saving.
2060  * @bm: Memory bitmap.
2061  * @buf: Memory buffer to store the PFNs in.
2062  *
2063  * PFNs corresponding to set bits in @bm are stored in the area of memory
2064  * pointed to by @buf (1 page at a time).
2065  */
2066 static inline void pack_pfns(unsigned long *buf, struct memory_bitmap *bm)
2067 {
2068         int j;
2069 
2070         for (j = 0; j < PAGE_SIZE / sizeof(long); j++) {
2071                 buf[j] = memory_bm_next_pfn(bm);
2072                 if (unlikely(buf[j] == BM_END_OF_MAP))
2073                         break;
2074                 /* Save page key for data page (s390 only). */
2075                 page_key_read(buf + j);
2076         }
2077 }
2078 
2079 /**
2080  * snapshot_read_next - Get the address to read the next image page from.
2081  * @handle: Snapshot handle to be used for the reading.
2082  *
2083  * On the first call, @handle should point to a zeroed snapshot_handle
2084  * structure.  The structure gets populated then and a pointer to it should be
2085  * passed to this function every next time.
2086  *
2087  * On success, the function returns a positive number.  Then, the caller
2088  * is allowed to read up to the returned number of bytes from the memory
2089  * location computed by the data_of() macro.
2090  *
2091  * The function returns 0 to indicate the end of the data stream condition,
2092  * and negative numbers are returned on errors.  If that happens, the structure
2093  * pointed to by @handle is not updated and should not be used any more.
2094  */
2095 int snapshot_read_next(struct snapshot_handle *handle)
2096 {
2097         if (handle->cur > nr_meta_pages + nr_copy_pages)
2098                 return 0;
2099 
2100         if (!buffer) {
2101                 /* This makes the buffer be freed by swsusp_free() */
2102                 buffer = get_image_page(GFP_ATOMIC, PG_ANY);
2103                 if (!buffer)
2104                         return -ENOMEM;
2105         }
2106         if (!handle->cur) {
2107                 int error;
2108 
2109                 error = init_header((struct swsusp_info *)buffer);
2110                 if (error)
2111                         return error;
2112                 handle->buffer = buffer;
2113                 memory_bm_position_reset(&orig_bm);
2114                 memory_bm_position_reset(&copy_bm);
2115         } else if (handle->cur <= nr_meta_pages) {
2116                 clear_page(buffer);
2117                 pack_pfns(buffer, &orig_bm);
2118         } else {
2119                 struct page *page;
2120 
2121                 page = pfn_to_page(memory_bm_next_pfn(&copy_bm));
2122                 if (PageHighMem(page)) {
2123                         /*
2124                          * Highmem pages are copied to the buffer,
2125                          * because we can't return with a kmapped
2126                          * highmem page (we may not be called again).
2127                          */
2128                         void *kaddr;
2129 
2130                         kaddr = kmap_atomic(page);
2131                         copy_page(buffer, kaddr);
2132                         kunmap_atomic(kaddr);
2133                         handle->buffer = buffer;
2134                 } else {
2135                         handle->buffer = page_address(page);
2136                 }
2137         }
2138         handle->cur++;
2139         return PAGE_SIZE;
2140 }
2141 
2142 static void duplicate_memory_bitmap(struct memory_bitmap *dst,
2143                                     struct memory_bitmap *src)
2144 {
2145         unsigned long pfn;
2146 
2147         memory_bm_position_reset(src);
2148         pfn = memory_bm_next_pfn(src);
2149         while (pfn != BM_END_OF_MAP) {
2150                 memory_bm_set_bit(dst, pfn);
2151                 pfn = memory_bm_next_pfn(src);
2152         }
2153 }
2154 
2155 /**
2156  * mark_unsafe_pages - Mark pages that were used before hibernation.
2157  *
2158  * Mark the pages that cannot be used for storing the image during restoration,
2159  * because they conflict with the pages that had been used before hibernation.
2160  */
2161 static void mark_unsafe_pages(struct memory_bitmap *bm)
2162 {
2163         unsigned long pfn;
2164 
2165         /* Clear the "free"/"unsafe" bit for all PFNs */
2166         memory_bm_position_reset(free_pages_map);
2167         pfn = memory_bm_next_pfn(free_pages_map);
2168         while (pfn != BM_END_OF_MAP) {
2169                 memory_bm_clear_current(free_pages_map);
2170                 pfn = memory_bm_next_pfn(free_pages_map);
2171         }
2172 
2173         /* Mark pages that correspond to the "original" PFNs as "unsafe" */
2174         duplicate_memory_bitmap(free_pages_map, bm);
2175 
2176         allocated_unsafe_pages = 0;
2177 }
2178 
2179 static int check_header(struct swsusp_info *info)
2180 {
2181         char *reason;
2182 
2183         reason = check_image_kernel(info);
2184         if (!reason && info->num_physpages != get_num_physpages())
2185                 reason = "memory size";
2186         if (reason) {
2187                 pr_err("Image mismatch: %s\n", reason);
2188                 return -EPERM;
2189         }
2190         return 0;
2191 }
2192 
2193 /**
2194  * load header - Check the image header and copy the data from it.
2195  */
2196 static int load_header(struct swsusp_info *info)
2197 {
2198         int error;
2199 
2200         restore_pblist = NULL;
2201         error = check_header(info);
2202         if (!error) {
2203                 nr_copy_pages = info->image_pages;
2204                 nr_meta_pages = info->pages - info->image_pages - 1;
2205         }
2206         return error;
2207 }
2208 
2209 /**
2210  * unpack_orig_pfns - Set bits corresponding to given PFNs in a memory bitmap.
2211  * @bm: Memory bitmap.
2212  * @buf: Area of memory containing the PFNs.
2213  *
2214  * For each element of the array pointed to by @buf (1 page at a time), set the
2215  * corresponding bit in @bm.
2216  */
2217 static int unpack_orig_pfns(unsigned long *buf, struct memory_bitmap *bm)
2218 {
2219         int j;
2220 
2221         for (j = 0; j < PAGE_SIZE / sizeof(long); j++) {
2222                 if (unlikely(buf[j] == BM_END_OF_MAP))
2223                         break;
2224 
2225                 /* Extract and buffer page key for data page (s390 only). */
2226                 page_key_memorize(buf + j);
2227 
2228                 if (pfn_valid(buf[j]) && memory_bm_pfn_present(bm, buf[j]))
2229                         memory_bm_set_bit(bm, buf[j]);
2230                 else
2231                         return -EFAULT;
2232         }
2233 
2234         return 0;
2235 }
2236 
2237 #ifdef CONFIG_HIGHMEM
2238 /*
2239  * struct highmem_pbe is used for creating the list of highmem pages that
2240  * should be restored atomically during the resume from disk, because the page
2241  * frames they have occupied before the suspend are in use.
2242  */
2243 struct highmem_pbe {
2244         struct page *copy_page; /* data is here now */
2245         struct page *orig_page; /* data was here before the suspend */
2246         struct highmem_pbe *next;
2247 };
2248 
2249 /*
2250  * List of highmem PBEs needed for restoring the highmem pages that were
2251  * allocated before the suspend and included in the suspend image, but have
2252  * also been allocated by the "resume" kernel, so their contents cannot be
2253  * written directly to their "original" page frames.
2254  */
2255 static struct highmem_pbe *highmem_pblist;
2256 
2257 /**
2258  * count_highmem_image_pages - Compute the number of highmem pages in the image.
2259  * @bm: Memory bitmap.
2260  *
2261  * The bits in @bm that correspond to image pages are assumed to be set.
2262  */
2263 static unsigned int count_highmem_image_pages(struct memory_bitmap *bm)
2264 {
2265         unsigned long pfn;
2266         unsigned int cnt = 0;
2267 
2268         memory_bm_position_reset(bm);
2269         pfn = memory_bm_next_pfn(bm);
2270         while (pfn != BM_END_OF_MAP) {
2271                 if (PageHighMem(pfn_to_page(pfn)))
2272                         cnt++;
2273 
2274                 pfn = memory_bm_next_pfn(bm);
2275         }
2276         return cnt;
2277 }
2278 
2279 static unsigned int safe_highmem_pages;
2280 
2281 static struct memory_bitmap *safe_highmem_bm;
2282 
2283 /**
2284  * prepare_highmem_image - Allocate memory for loading highmem data from image.
2285  * @bm: Pointer to an uninitialized memory bitmap structure.
2286  * @nr_highmem_p: Pointer to the number of highmem image pages.
2287  *
2288  * Try to allocate as many highmem pages as there are highmem image pages
2289  * (@nr_highmem_p points to the variable containing the number of highmem image
2290  * pages).  The pages that are "safe" (ie. will not be overwritten when the
2291  * hibernation image is restored entirely) have the corresponding bits set in
2292  * @bm (it must be unitialized).
2293  *
2294  * NOTE: This function should not be called if there are no highmem image pages.
2295  */
2296 static int prepare_highmem_image(struct memory_bitmap *bm,
2297                                  unsigned int *nr_highmem_p)
2298 {
2299         unsigned int to_alloc;
2300 
2301         if (memory_bm_create(bm, GFP_ATOMIC, PG_SAFE))
2302                 return -ENOMEM;
2303 
2304         if (get_highmem_buffer(PG_SAFE))
2305                 return -ENOMEM;
2306 
2307         to_alloc = count_free_highmem_pages();
2308         if (to_alloc > *nr_highmem_p)
2309                 to_alloc = *nr_highmem_p;
2310         else
2311                 *nr_highmem_p = to_alloc;
2312 
2313         safe_highmem_pages = 0;
2314         while (to_alloc-- > 0) {
2315                 struct page *page;
2316 
2317                 page = alloc_page(__GFP_HIGHMEM);
2318                 if (!swsusp_page_is_free(page)) {
2319                         /* The page is "safe", set its bit the bitmap */
2320                         memory_bm_set_bit(bm, page_to_pfn(page));
2321                         safe_highmem_pages++;
2322                 }
2323                 /* Mark the page as allocated */
2324                 swsusp_set_page_forbidden(page);
2325                 swsusp_set_page_free(page);
2326         }
2327         memory_bm_position_reset(bm);
2328         safe_highmem_bm = bm;
2329         return 0;
2330 }
2331 
2332 static struct page *last_highmem_page;
2333 
2334 /**
2335  * get_highmem_page_buffer - Prepare a buffer to store a highmem image page.
2336  *
2337  * For a given highmem image page get a buffer that suspend_write_next() should
2338  * return to its caller to write to.
2339  *
2340  * If the page is to be saved to its "original" page frame or a copy of
2341  * the page is to be made in the highmem, @buffer is returned.  Otherwise,
2342  * the copy of the page is to be made in normal memory, so the address of
2343  * the copy is returned.
2344  *
2345  * If @buffer is returned, the caller of suspend_write_next() will write
2346  * the page's contents to @buffer, so they will have to be copied to the
2347  * right location on the next call to suspend_write_next() and it is done
2348  * with the help of copy_last_highmem_page().  For this purpose, if
2349  * @buffer is returned, @last_highmem_page is set to the page to which
2350  * the data will have to be copied from @buffer.
2351  */
2352 static void *get_highmem_page_buffer(struct page *page,
2353                                      struct chain_allocator *ca)
2354 {
2355         struct highmem_pbe *pbe;
2356         void *kaddr;
2357 
2358         if (swsusp_page_is_forbidden(page) && swsusp_page_is_free(page)) {
2359                 /*
2360                  * We have allocated the "original" page frame and we can
2361                  * use it directly to store the loaded page.
2362                  */
2363                 last_highmem_page = page;
2364                 return buffer;
2365         }
2366         /*
2367          * The "original" page frame has not been allocated and we have to
2368          * use a "safe" page frame to store the loaded page.
2369          */
2370         pbe = chain_alloc(ca, sizeof(struct highmem_pbe));
2371         if (!pbe) {
2372                 swsusp_free();
2373                 return ERR_PTR(-ENOMEM);
2374         }
2375         pbe->orig_page = page;
2376         if (safe_highmem_pages > 0) {
2377                 struct page *tmp;
2378 
2379                 /* Copy of the page will be stored in high memory */
2380                 kaddr = buffer;
2381                 tmp = pfn_to_page(memory_bm_next_pfn(safe_highmem_bm));
2382                 safe_highmem_pages--;
2383                 last_highmem_page = tmp;
2384                 pbe->copy_page = tmp;
2385         } else {
2386                 /* Copy of the page will be stored in normal memory */
2387                 kaddr = safe_pages_list;
2388                 safe_pages_list = safe_pages_list->next;
2389                 pbe->copy_page = virt_to_page(kaddr);
2390         }
2391         pbe->next = highmem_pblist;
2392         highmem_pblist = pbe;
2393         return kaddr;
2394 }
2395 
2396 /**
2397  * copy_last_highmem_page - Copy most the most recent highmem image page.
2398  *
2399  * Copy the contents of a highmem image from @buffer, where the caller of
2400  * snapshot_write_next() has stored them, to the right location represented by
2401  * @last_highmem_page .
2402  */
2403 static void copy_last_highmem_page(void)
2404 {
2405         if (last_highmem_page) {
2406                 void *dst;
2407 
2408                 dst = kmap_atomic(last_highmem_page);
2409                 copy_page(dst, buffer);
2410                 kunmap_atomic(dst);
2411                 last_highmem_page = NULL;
2412         }
2413 }
2414 
2415 static inline int last_highmem_page_copied(void)
2416 {
2417         return !last_highmem_page;
2418 }
2419 
2420 static inline void free_highmem_data(void)
2421 {
2422         if (safe_highmem_bm)
2423                 memory_bm_free(safe_highmem_bm, PG_UNSAFE_CLEAR);
2424 
2425         if (buffer)
2426                 free_image_page(buffer, PG_UNSAFE_CLEAR);
2427 }
2428 #else
2429 static unsigned int count_highmem_image_pages(struct memory_bitmap *bm) { return 0; }
2430 
2431 static inline int prepare_highmem_image(struct memory_bitmap *bm,
2432                                         unsigned int *nr_highmem_p) { return 0; }
2433 
2434 static inline void *get_highmem_page_buffer(struct page *page,
2435                                             struct chain_allocator *ca)
2436 {
2437         return ERR_PTR(-EINVAL);
2438 }
2439 
2440 static inline void copy_last_highmem_page(void) {}
2441 static inline int last_highmem_page_copied(void) { return 1; }
2442 static inline void free_highmem_data(void) {}
2443 #endif /* CONFIG_HIGHMEM */
2444 
2445 #define PBES_PER_LINKED_PAGE    (LINKED_PAGE_DATA_SIZE / sizeof(struct pbe))
2446 
2447 /**
2448  * prepare_image - Make room for loading hibernation image.
2449  * @new_bm: Unitialized memory bitmap structure.
2450  * @bm: Memory bitmap with unsafe pages marked.
2451  *
2452  * Use @bm to mark the pages that will be overwritten in the process of
2453  * restoring the system memory state from the suspend image ("unsafe" pages)
2454  * and allocate memory for the image.
2455  *
2456  * The idea is to allocate a new memory bitmap first and then allocate
2457  * as many pages as needed for image data, but without specifying what those
2458  * pages will be used for just yet.  Instead, we mark them all as allocated and
2459  * create a lists of "safe" pages to be used later.  On systems with high
2460  * memory a list of "safe" highmem pages is created too.
2461  */
2462 static int prepare_image(struct memory_bitmap *new_bm, struct memory_bitmap *bm)
2463 {
2464         unsigned int nr_pages, nr_highmem;
2465         struct linked_page *lp;
2466         int error;
2467 
2468         /* If there is no highmem, the buffer will not be necessary */
2469         free_image_page(buffer, PG_UNSAFE_CLEAR);
2470         buffer = NULL;
2471 
2472         nr_highmem = count_highmem_image_pages(bm);
2473         mark_unsafe_pages(bm);
2474 
2475         error = memory_bm_create(new_bm, GFP_ATOMIC, PG_SAFE);
2476         if (error)
2477                 goto Free;
2478 
2479         duplicate_memory_bitmap(new_bm, bm);
2480         memory_bm_free(bm, PG_UNSAFE_KEEP);
2481         if (nr_highmem > 0) {
2482                 error = prepare_highmem_image(bm, &nr_highmem);
2483                 if (error)
2484                         goto Free;
2485         }
2486         /*
2487          * Reserve some safe pages for potential later use.
2488          *
2489          * NOTE: This way we make sure there will be enough safe pages for the
2490          * chain_alloc() in get_buffer().  It is a bit wasteful, but
2491          * nr_copy_pages cannot be greater than 50% of the memory anyway.
2492          *
2493          * nr_copy_pages cannot be less than allocated_unsafe_pages too.
2494          */
2495         nr_pages = nr_copy_pages - nr_highmem - allocated_unsafe_pages;
2496         nr_pages = DIV_ROUND_UP(nr_pages, PBES_PER_LINKED_PAGE);
2497         while (nr_pages > 0) {
2498                 lp = get_image_page(GFP_ATOMIC, PG_SAFE);
2499                 if (!lp) {
2500                         error = -ENOMEM;
2501                         goto Free;
2502                 }
2503                 lp->next = safe_pages_list;
2504                 safe_pages_list = lp;
2505                 nr_pages--;
2506         }
2507         /* Preallocate memory for the image */
2508         nr_pages = nr_copy_pages - nr_highmem - allocated_unsafe_pages;
2509         while (nr_pages > 0) {
2510                 lp = (struct linked_page *)get_zeroed_page(GFP_ATOMIC);
2511                 if (!lp) {
2512                         error = -ENOMEM;
2513                         goto Free;
2514                 }
2515                 if (!swsusp_page_is_free(virt_to_page(lp))) {
2516                         /* The page is "safe", add it to the list */
2517                         lp->next = safe_pages_list;
2518                         safe_pages_list = lp;
2519                 }
2520                 /* Mark the page as allocated */
2521                 swsusp_set_page_forbidden(virt_to_page(lp));
2522                 swsusp_set_page_free(virt_to_page(lp));
2523                 nr_pages--;
2524         }
2525         return 0;
2526 
2527  Free:
2528         swsusp_free();
2529         return error;
2530 }
2531 
2532 /**
2533  * get_buffer - Get the address to store the next image data page.
2534  *
2535  * Get the address that snapshot_write_next() should return to its caller to
2536  * write to.
2537  */
2538 static void *get_buffer(struct memory_bitmap *bm, struct chain_allocator *ca)
2539 {
2540         struct pbe *pbe;
2541         struct page *page;
2542         unsigned long pfn = memory_bm_next_pfn(bm);
2543 
2544         if (pfn == BM_END_OF_MAP)
2545                 return ERR_PTR(-EFAULT);
2546 
2547         page = pfn_to_page(pfn);
2548         if (PageHighMem(page))
2549                 return get_highmem_page_buffer(page, ca);
2550 
2551         if (swsusp_page_is_forbidden(page) && swsusp_page_is_free(page))
2552                 /*
2553                  * We have allocated the "original" page frame and we can
2554                  * use it directly to store the loaded page.
2555                  */
2556                 return page_address(page);
2557 
2558         /*
2559          * The "original" page frame has not been allocated and we have to
2560          * use a "safe" page frame to store the loaded page.
2561          */
2562         pbe = chain_alloc(ca, sizeof(struct pbe));
2563         if (!pbe) {
2564                 swsusp_free();
2565                 return ERR_PTR(-ENOMEM);
2566         }
2567         pbe->orig_address = page_address(page);
2568         pbe->address = safe_pages_list;
2569         safe_pages_list = safe_pages_list->next;
2570         pbe->next = restore_pblist;
2571         restore_pblist = pbe;
2572         return pbe->address;
2573 }
2574 
2575 /**
2576  * snapshot_write_next - Get the address to store the next image page.
2577  * @handle: Snapshot handle structure to guide the writing.
2578  *
2579  * On the first call, @handle should point to a zeroed snapshot_handle
2580  * structure.  The structure gets populated then and a pointer to it should be
2581  * passed to this function every next time.
2582  *
2583  * On success, the function returns a positive number.  Then, the caller
2584  * is allowed to write up to the returned number of bytes to the memory
2585  * location computed by the data_of() macro.
2586  *
2587  * The function returns 0 to indicate the "end of file" condition.  Negative
2588  * numbers are returned on errors, in which cases the structure pointed to by
2589  * @handle is not updated and should not be used any more.
2590  */
2591 int snapshot_write_next(struct snapshot_handle *handle)
2592 {
2593         static struct chain_allocator ca;
2594         int error = 0;
2595 
2596         /* Check if we have already loaded the entire image */
2597         if (handle->cur > 1 && handle->cur > nr_meta_pages + nr_copy_pages)
2598                 return 0;
2599 
2600         handle->sync_read = 1;
2601 
2602         if (!handle->cur) {
2603                 if (!buffer)
2604                         /* This makes the buffer be freed by swsusp_free() */
2605                         buffer = get_image_page(GFP_ATOMIC, PG_ANY);
2606 
2607                 if (!buffer)
2608                         return -ENOMEM;
2609 
2610                 handle->buffer = buffer;
2611         } else if (handle->cur == 1) {
2612                 error = load_header(buffer);
2613                 if (error)
2614                         return error;
2615 
2616                 safe_pages_list = NULL;
2617 
2618                 error = memory_bm_create(&copy_bm, GFP_ATOMIC, PG_ANY);
2619                 if (error)
2620                         return error;
2621 
2622                 /* Allocate buffer for page keys. */
2623                 error = page_key_alloc(nr_copy_pages);
2624                 if (error)
2625                         return error;
2626 
2627                 hibernate_restore_protection_begin();
2628         } else if (handle->cur <= nr_meta_pages + 1) {
2629                 error = unpack_orig_pfns(buffer, &copy_bm);
2630                 if (error)
2631                         return error;
2632 
2633                 if (handle->cur == nr_meta_pages + 1) {
2634                         error = prepare_image(&orig_bm, &copy_bm);
2635                         if (error)
2636                                 return error;
2637 
2638                         chain_init(&ca, GFP_ATOMIC, PG_SAFE);
2639                         memory_bm_position_reset(&orig_bm);
2640                         restore_pblist = NULL;
2641                         handle->buffer = get_buffer(&orig_bm, &ca);
2642                         handle->sync_read = 0;
2643                         if (IS_ERR(handle->buffer))
2644                                 return PTR_ERR(handle->buffer);
2645                 }
2646         } else {
2647                 copy_last_highmem_page();
2648                 /* Restore page key for data page (s390 only). */
2649                 page_key_write(handle->buffer);
2650                 hibernate_restore_protect_page(handle->buffer);
2651                 handle->buffer = get_buffer(&orig_bm, &ca);
2652                 if (IS_ERR(handle->buffer))
2653                         return PTR_ERR(handle->buffer);
2654                 if (handle->buffer != buffer)
2655                         handle->sync_read = 0;
2656         }
2657         handle->cur++;
2658         return PAGE_SIZE;
2659 }
2660 
2661 /**
2662  * snapshot_write_finalize - Complete the loading of a hibernation image.
2663  *
2664  * Must be called after the last call to snapshot_write_next() in case the last
2665  * page in the image happens to be a highmem page and its contents should be
2666  * stored in highmem.  Additionally, it recycles bitmap memory that's not
2667  * necessary any more.
2668  */
2669 void snapshot_write_finalize(struct snapshot_handle *handle)
2670 {
2671         copy_last_highmem_page();
2672         /* Restore page key for data page (s390 only). */
2673         page_key_write(handle->buffer);
2674         page_key_free();
2675         hibernate_restore_protect_page(handle->buffer);
2676         /* Do that only if we have loaded the image entirely */
2677         if (handle->cur > 1 && handle->cur > nr_meta_pages + nr_copy_pages) {
2678                 memory_bm_recycle(&orig_bm);
2679                 free_highmem_data();
2680         }
2681 }
2682 
2683 int snapshot_image_loaded(struct snapshot_handle *handle)
2684 {
2685         return !(!nr_copy_pages || !last_highmem_page_copied() ||
2686                         handle->cur <= nr_meta_pages + nr_copy_pages);
2687 }
2688 
2689 #ifdef CONFIG_HIGHMEM
2690 /* Assumes that @buf is ready and points to a "safe" page */
2691 static inline void swap_two_pages_data(struct page *p1, struct page *p2,
2692                                        void *buf)
2693 {
2694         void *kaddr1, *kaddr2;
2695 
2696         kaddr1 = kmap_atomic(p1);
2697         kaddr2 = kmap_atomic(p2);
2698         copy_page(buf, kaddr1);
2699         copy_page(kaddr1, kaddr2);
2700         copy_page(kaddr2, buf);
2701         kunmap_atomic(kaddr2);
2702         kunmap_atomic(kaddr1);
2703 }
2704 
2705 /**
2706  * restore_highmem - Put highmem image pages into their original locations.
2707  *
2708  * For each highmem page that was in use before hibernation and is included in
2709  * the image, and also has been allocated by the "restore" kernel, swap its
2710  * current contents with the previous (ie. "before hibernation") ones.
2711  *
2712  * If the restore eventually fails, we can call this function once again and
2713  * restore the highmem state as seen by the restore kernel.
2714  */
2715 int restore_highmem(void)
2716 {
2717         struct highmem_pbe *pbe = highmem_pblist;
2718         void *buf;
2719 
2720         if (!pbe)
2721                 return 0;
2722 
2723         buf = get_image_page(GFP_ATOMIC, PG_SAFE);
2724         if (!buf)
2725                 return -ENOMEM;
2726 
2727         while (pbe) {
2728                 swap_two_pages_data(pbe->copy_page, pbe->orig_page, buf);
2729                 pbe = pbe->next;
2730         }
2731         free_image_page(buf, PG_UNSAFE_CLEAR);
2732         return 0;
2733 }
2734 #endif /* CONFIG_HIGHMEM */
/* [<][>][^][v][top][bottom][index][help] */
root/kernel/power/snapshot.c

DEFINITIONS
