]> git.proxmox.com Git - mirror_ubuntu-hirsute-kernel.git/blob - mm/internal.h
mm/page_alloc: integrate classzone_idx and high_zoneidx
[mirror_ubuntu-hirsute-kernel.git] / mm / internal.h
1 /* SPDX-License-Identifier: GPL-2.0-or-later */
2 /* internal.h: mm/ internal definitions
3 *
4 * Copyright (C) 2004 Red Hat, Inc. All Rights Reserved.
5 * Written by David Howells (dhowells@redhat.com)
6 */
7 #ifndef __MM_INTERNAL_H
8 #define __MM_INTERNAL_H
9
10 #include <linux/fs.h>
11 #include <linux/mm.h>
12 #include <linux/pagemap.h>
13 #include <linux/tracepoint-defs.h>
14
15 /*
16 * The set of flags that only affect watermark checking and reclaim
17 * behaviour. This is used by the MM to obey the caller constraints
18 * about IO, FS and watermark checking while ignoring placement
19 * hints such as HIGHMEM usage.
20 */
21 #define GFP_RECLAIM_MASK (__GFP_RECLAIM|__GFP_HIGH|__GFP_IO|__GFP_FS|\
22 __GFP_NOWARN|__GFP_RETRY_MAYFAIL|__GFP_NOFAIL|\
23 __GFP_NORETRY|__GFP_MEMALLOC|__GFP_NOMEMALLOC|\
24 __GFP_ATOMIC)
25
26 /* The GFP flags allowed during early boot */
27 #define GFP_BOOT_MASK (__GFP_BITS_MASK & ~(__GFP_RECLAIM|__GFP_IO|__GFP_FS))
28
29 /* Control allocation cpuset and node placement constraints */
30 #define GFP_CONSTRAINT_MASK (__GFP_HARDWALL|__GFP_THISNODE)
31
32 /* Do not use these with a slab allocator */
33 #define GFP_SLAB_BUG_MASK (__GFP_DMA32|__GFP_HIGHMEM|~__GFP_BITS_MASK)
34
35 void page_writeback_init(void);
36
37 vm_fault_t do_swap_page(struct vm_fault *vmf);
38
39 void free_pgtables(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
40 unsigned long floor, unsigned long ceiling);
41
42 static inline bool can_madv_lru_vma(struct vm_area_struct *vma)
43 {
44 return !(vma->vm_flags & (VM_LOCKED|VM_HUGETLB|VM_PFNMAP));
45 }
46
47 void unmap_page_range(struct mmu_gather *tlb,
48 struct vm_area_struct *vma,
49 unsigned long addr, unsigned long end,
50 struct zap_details *details);
51
52 void force_page_cache_readahead(struct address_space *, struct file *,
53 pgoff_t index, unsigned long nr_to_read);
54 void __do_page_cache_readahead(struct address_space *, struct file *,
55 pgoff_t index, unsigned long nr_to_read,
56 unsigned long lookahead_size);
57
58 /*
59 * Submit IO for the read-ahead request in file_ra_state.
60 */
61 static inline void ra_submit(struct file_ra_state *ra,
62 struct address_space *mapping, struct file *filp)
63 {
64 __do_page_cache_readahead(mapping, filp,
65 ra->start, ra->size, ra->async_size);
66 }
67
68 /**
69 * page_evictable - test whether a page is evictable
70 * @page: the page to test
71 *
72 * Test whether page is evictable--i.e., should be placed on active/inactive
73 * lists vs unevictable list.
74 *
75 * Reasons page might not be evictable:
76 * (1) page's mapping marked unevictable
77 * (2) page is part of an mlocked VMA
78 *
79 */
80 static inline bool page_evictable(struct page *page)
81 {
82 bool ret;
83
84 /* Prevent address_space of inode and swap cache from being freed */
85 rcu_read_lock();
86 ret = !mapping_unevictable(page_mapping(page)) && !PageMlocked(page);
87 rcu_read_unlock();
88 return ret;
89 }
90
91 /*
92 * Turn a non-refcounted page (->_refcount == 0) into refcounted with
93 * a count of one.
94 */
95 static inline void set_page_refcounted(struct page *page)
96 {
97 VM_BUG_ON_PAGE(PageTail(page), page);
98 VM_BUG_ON_PAGE(page_ref_count(page), page);
99 set_page_count(page, 1);
100 }
101
102 extern unsigned long highest_memmap_pfn;
103
104 /*
105 * Maximum number of reclaim retries without progress before the OOM
106 * killer is consider the only way forward.
107 */
108 #define MAX_RECLAIM_RETRIES 16
109
110 /*
111 * in mm/vmscan.c:
112 */
113 extern int isolate_lru_page(struct page *page);
114 extern void putback_lru_page(struct page *page);
115
116 /*
117 * in mm/rmap.c:
118 */
119 extern pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address);
120
121 /*
122 * in mm/page_alloc.c
123 */
124
125 /*
126 * Structure for holding the mostly immutable allocation parameters passed
127 * between functions involved in allocations, including the alloc_pages*
128 * family of functions.
129 *
130 * nodemask, migratetype and highest_zoneidx are initialized only once in
131 * __alloc_pages_nodemask() and then never change.
132 *
133 * zonelist, preferred_zone and highest_zoneidx are set first in
134 * __alloc_pages_nodemask() for the fast path, and might be later changed
135 * in __alloc_pages_slowpath(). All other functions pass the whole strucure
136 * by a const pointer.
137 */
138 struct alloc_context {
139 struct zonelist *zonelist;
140 nodemask_t *nodemask;
141 struct zoneref *preferred_zoneref;
142 int migratetype;
143
144 /*
145 * highest_zoneidx represents highest usable zone index of
146 * the allocation request. Due to the nature of the zone,
147 * memory on lower zone than the highest_zoneidx will be
148 * protected by lowmem_reserve[highest_zoneidx].
149 *
150 * highest_zoneidx is also used by reclaim/compaction to limit
151 * the target zone since higher zone than this index cannot be
152 * usable for this allocation request.
153 */
154 enum zone_type highest_zoneidx;
155 bool spread_dirty_pages;
156 };
157
158 /*
159 * Locate the struct page for both the matching buddy in our
160 * pair (buddy1) and the combined O(n+1) page they form (page).
161 *
162 * 1) Any buddy B1 will have an order O twin B2 which satisfies
163 * the following equation:
164 * B2 = B1 ^ (1 << O)
165 * For example, if the starting buddy (buddy2) is #8 its order
166 * 1 buddy is #10:
167 * B2 = 8 ^ (1 << 1) = 8 ^ 2 = 10
168 *
169 * 2) Any buddy B will have an order O+1 parent P which
170 * satisfies the following equation:
171 * P = B & ~(1 << O)
172 *
173 * Assumption: *_mem_map is contiguous at least up to MAX_ORDER
174 */
175 static inline unsigned long
176 __find_buddy_pfn(unsigned long page_pfn, unsigned int order)
177 {
178 return page_pfn ^ (1 << order);
179 }
180
181 extern struct page *__pageblock_pfn_to_page(unsigned long start_pfn,
182 unsigned long end_pfn, struct zone *zone);
183
184 static inline struct page *pageblock_pfn_to_page(unsigned long start_pfn,
185 unsigned long end_pfn, struct zone *zone)
186 {
187 if (zone->contiguous)
188 return pfn_to_page(start_pfn);
189
190 return __pageblock_pfn_to_page(start_pfn, end_pfn, zone);
191 }
192
193 extern int __isolate_free_page(struct page *page, unsigned int order);
194 extern void __putback_isolated_page(struct page *page, unsigned int order,
195 int mt);
196 extern void memblock_free_pages(struct page *page, unsigned long pfn,
197 unsigned int order);
198 extern void __free_pages_core(struct page *page, unsigned int order);
199 extern void prep_compound_page(struct page *page, unsigned int order);
200 extern void post_alloc_hook(struct page *page, unsigned int order,
201 gfp_t gfp_flags);
202 extern int user_min_free_kbytes;
203
204 extern void zone_pcp_update(struct zone *zone);
205 extern void zone_pcp_reset(struct zone *zone);
206
207 #if defined CONFIG_COMPACTION || defined CONFIG_CMA
208
209 /*
210 * in mm/compaction.c
211 */
212 /*
213 * compact_control is used to track pages being migrated and the free pages
214 * they are being migrated to during memory compaction. The free_pfn starts
215 * at the end of a zone and migrate_pfn begins at the start. Movable pages
216 * are moved to the end of a zone during a compaction run and the run
217 * completes when free_pfn <= migrate_pfn
218 */
219 struct compact_control {
220 struct list_head freepages; /* List of free pages to migrate to */
221 struct list_head migratepages; /* List of pages being migrated */
222 unsigned int nr_freepages; /* Number of isolated free pages */
223 unsigned int nr_migratepages; /* Number of pages to migrate */
224 unsigned long free_pfn; /* isolate_freepages search base */
225 unsigned long migrate_pfn; /* isolate_migratepages search base */
226 unsigned long fast_start_pfn; /* a pfn to start linear scan from */
227 struct zone *zone;
228 unsigned long total_migrate_scanned;
229 unsigned long total_free_scanned;
230 unsigned short fast_search_fail;/* failures to use free list searches */
231 short search_order; /* order to start a fast search at */
232 const gfp_t gfp_mask; /* gfp mask of a direct compactor */
233 int order; /* order a direct compactor needs */
234 int migratetype; /* migratetype of direct compactor */
235 const unsigned int alloc_flags; /* alloc flags of a direct compactor */
236 const int highest_zoneidx; /* zone index of a direct compactor */
237 enum migrate_mode mode; /* Async or sync migration mode */
238 bool ignore_skip_hint; /* Scan blocks even if marked skip */
239 bool no_set_skip_hint; /* Don't mark blocks for skipping */
240 bool ignore_block_suitable; /* Scan blocks considered unsuitable */
241 bool direct_compaction; /* False from kcompactd or /proc/... */
242 bool whole_zone; /* Whole zone should/has been scanned */
243 bool contended; /* Signal lock or sched contention */
244 bool rescan; /* Rescanning the same pageblock */
245 bool alloc_contig; /* alloc_contig_range allocation */
246 };
247
248 /*
249 * Used in direct compaction when a page should be taken from the freelists
250 * immediately when one is created during the free path.
251 */
252 struct capture_control {
253 struct compact_control *cc;
254 struct page *page;
255 };
256
257 unsigned long
258 isolate_freepages_range(struct compact_control *cc,
259 unsigned long start_pfn, unsigned long end_pfn);
260 unsigned long
261 isolate_migratepages_range(struct compact_control *cc,
262 unsigned long low_pfn, unsigned long end_pfn);
263 int find_suitable_fallback(struct free_area *area, unsigned int order,
264 int migratetype, bool only_stealable, bool *can_steal);
265
266 #endif
267
268 /*
269 * This function returns the order of a free page in the buddy system. In
270 * general, page_zone(page)->lock must be held by the caller to prevent the
271 * page from being allocated in parallel and returning garbage as the order.
272 * If a caller does not hold page_zone(page)->lock, it must guarantee that the
273 * page cannot be allocated or merged in parallel. Alternatively, it must
274 * handle invalid values gracefully, and use page_order_unsafe() below.
275 */
276 static inline unsigned int page_order(struct page *page)
277 {
278 /* PageBuddy() must be checked by the caller */
279 return page_private(page);
280 }
281
282 /*
283 * Like page_order(), but for callers who cannot afford to hold the zone lock.
284 * PageBuddy() should be checked first by the caller to minimize race window,
285 * and invalid values must be handled gracefully.
286 *
287 * READ_ONCE is used so that if the caller assigns the result into a local
288 * variable and e.g. tests it for valid range before using, the compiler cannot
289 * decide to remove the variable and inline the page_private(page) multiple
290 * times, potentially observing different values in the tests and the actual
291 * use of the result.
292 */
293 #define page_order_unsafe(page) READ_ONCE(page_private(page))
294
295 static inline bool is_cow_mapping(vm_flags_t flags)
296 {
297 return (flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
298 }
299
300 /*
301 * These three helpers classifies VMAs for virtual memory accounting.
302 */
303
304 /*
305 * Executable code area - executable, not writable, not stack
306 */
307 static inline bool is_exec_mapping(vm_flags_t flags)
308 {
309 return (flags & (VM_EXEC | VM_WRITE | VM_STACK)) == VM_EXEC;
310 }
311
312 /*
313 * Stack area - atomatically grows in one direction
314 *
315 * VM_GROWSUP / VM_GROWSDOWN VMAs are always private anonymous:
316 * do_mmap() forbids all other combinations.
317 */
318 static inline bool is_stack_mapping(vm_flags_t flags)
319 {
320 return (flags & VM_STACK) == VM_STACK;
321 }
322
323 /*
324 * Data area - private, writable, not stack
325 */
326 static inline bool is_data_mapping(vm_flags_t flags)
327 {
328 return (flags & (VM_WRITE | VM_SHARED | VM_STACK)) == VM_WRITE;
329 }
330
331 /* mm/util.c */
332 void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
333 struct vm_area_struct *prev);
334 void __vma_unlink_list(struct mm_struct *mm, struct vm_area_struct *vma);
335
336 #ifdef CONFIG_MMU
337 extern long populate_vma_page_range(struct vm_area_struct *vma,
338 unsigned long start, unsigned long end, int *nonblocking);
339 extern void munlock_vma_pages_range(struct vm_area_struct *vma,
340 unsigned long start, unsigned long end);
341 static inline void munlock_vma_pages_all(struct vm_area_struct *vma)
342 {
343 munlock_vma_pages_range(vma, vma->vm_start, vma->vm_end);
344 }
345
346 /*
347 * must be called with vma's mmap_sem held for read or write, and page locked.
348 */
349 extern void mlock_vma_page(struct page *page);
350 extern unsigned int munlock_vma_page(struct page *page);
351
352 /*
353 * Clear the page's PageMlocked(). This can be useful in a situation where
354 * we want to unconditionally remove a page from the pagecache -- e.g.,
355 * on truncation or freeing.
356 *
357 * It is legal to call this function for any page, mlocked or not.
358 * If called for a page that is still mapped by mlocked vmas, all we do
359 * is revert to lazy LRU behaviour -- semantics are not broken.
360 */
361 extern void clear_page_mlock(struct page *page);
362
363 /*
364 * mlock_migrate_page - called only from migrate_misplaced_transhuge_page()
365 * (because that does not go through the full procedure of migration ptes):
366 * to migrate the Mlocked page flag; update statistics.
367 */
368 static inline void mlock_migrate_page(struct page *newpage, struct page *page)
369 {
370 if (TestClearPageMlocked(page)) {
371 int nr_pages = hpage_nr_pages(page);
372
373 /* Holding pmd lock, no change in irq context: __mod is safe */
374 __mod_zone_page_state(page_zone(page), NR_MLOCK, -nr_pages);
375 SetPageMlocked(newpage);
376 __mod_zone_page_state(page_zone(newpage), NR_MLOCK, nr_pages);
377 }
378 }
379
380 extern pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma);
381
382 /*
383 * At what user virtual address is page expected in @vma?
384 */
385 static inline unsigned long
386 __vma_address(struct page *page, struct vm_area_struct *vma)
387 {
388 pgoff_t pgoff = page_to_pgoff(page);
389 return vma->vm_start + ((pgoff - vma->vm_pgoff) << PAGE_SHIFT);
390 }
391
392 static inline unsigned long
393 vma_address(struct page *page, struct vm_area_struct *vma)
394 {
395 unsigned long start, end;
396
397 start = __vma_address(page, vma);
398 end = start + PAGE_SIZE * (hpage_nr_pages(page) - 1);
399
400 /* page should be within @vma mapping range */
401 VM_BUG_ON_VMA(end < vma->vm_start || start >= vma->vm_end, vma);
402
403 return max(start, vma->vm_start);
404 }
405
406 static inline struct file *maybe_unlock_mmap_for_io(struct vm_fault *vmf,
407 struct file *fpin)
408 {
409 int flags = vmf->flags;
410
411 if (fpin)
412 return fpin;
413
414 /*
415 * FAULT_FLAG_RETRY_NOWAIT means we don't want to wait on page locks or
416 * anything, so we only pin the file and drop the mmap_sem if only
417 * FAULT_FLAG_ALLOW_RETRY is set, while this is the first attempt.
418 */
419 if (fault_flag_allow_retry_first(flags) &&
420 !(flags & FAULT_FLAG_RETRY_NOWAIT)) {
421 fpin = get_file(vmf->vma->vm_file);
422 up_read(&vmf->vma->vm_mm->mmap_sem);
423 }
424 return fpin;
425 }
426
427 #else /* !CONFIG_MMU */
428 static inline void clear_page_mlock(struct page *page) { }
429 static inline void mlock_vma_page(struct page *page) { }
430 static inline void mlock_migrate_page(struct page *new, struct page *old) { }
431
432 #endif /* !CONFIG_MMU */
433
434 /*
435 * Return the mem_map entry representing the 'offset' subpage within
436 * the maximally aligned gigantic page 'base'. Handle any discontiguity
437 * in the mem_map at MAX_ORDER_NR_PAGES boundaries.
438 */
439 static inline struct page *mem_map_offset(struct page *base, int offset)
440 {
441 if (unlikely(offset >= MAX_ORDER_NR_PAGES))
442 return nth_page(base, offset);
443 return base + offset;
444 }
445
446 /*
447 * Iterator over all subpages within the maximally aligned gigantic
448 * page 'base'. Handle any discontiguity in the mem_map.
449 */
450 static inline struct page *mem_map_next(struct page *iter,
451 struct page *base, int offset)
452 {
453 if (unlikely((offset & (MAX_ORDER_NR_PAGES - 1)) == 0)) {
454 unsigned long pfn = page_to_pfn(base) + offset;
455 if (!pfn_valid(pfn))
456 return NULL;
457 return pfn_to_page(pfn);
458 }
459 return iter + 1;
460 }
461
462 /* Memory initialisation debug and verification */
463 enum mminit_level {
464 MMINIT_WARNING,
465 MMINIT_VERIFY,
466 MMINIT_TRACE
467 };
468
469 #ifdef CONFIG_DEBUG_MEMORY_INIT
470
471 extern int mminit_loglevel;
472
473 #define mminit_dprintk(level, prefix, fmt, arg...) \
474 do { \
475 if (level < mminit_loglevel) { \
476 if (level <= MMINIT_WARNING) \
477 pr_warn("mminit::" prefix " " fmt, ##arg); \
478 else \
479 printk(KERN_DEBUG "mminit::" prefix " " fmt, ##arg); \
480 } \
481 } while (0)
482
483 extern void mminit_verify_pageflags_layout(void);
484 extern void mminit_verify_zonelist(void);
485 #else
486
487 static inline void mminit_dprintk(enum mminit_level level,
488 const char *prefix, const char *fmt, ...)
489 {
490 }
491
492 static inline void mminit_verify_pageflags_layout(void)
493 {
494 }
495
496 static inline void mminit_verify_zonelist(void)
497 {
498 }
499 #endif /* CONFIG_DEBUG_MEMORY_INIT */
500
501 /* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */
502 #if defined(CONFIG_SPARSEMEM)
503 extern void mminit_validate_memmodel_limits(unsigned long *start_pfn,
504 unsigned long *end_pfn);
505 #else
506 static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn,
507 unsigned long *end_pfn)
508 {
509 }
510 #endif /* CONFIG_SPARSEMEM */
511
512 #define NODE_RECLAIM_NOSCAN -2
513 #define NODE_RECLAIM_FULL -1
514 #define NODE_RECLAIM_SOME 0
515 #define NODE_RECLAIM_SUCCESS 1
516
517 #ifdef CONFIG_NUMA
518 extern int node_reclaim(struct pglist_data *, gfp_t, unsigned int);
519 #else
520 static inline int node_reclaim(struct pglist_data *pgdat, gfp_t mask,
521 unsigned int order)
522 {
523 return NODE_RECLAIM_NOSCAN;
524 }
525 #endif
526
527 extern int hwpoison_filter(struct page *p);
528
529 extern u32 hwpoison_filter_dev_major;
530 extern u32 hwpoison_filter_dev_minor;
531 extern u64 hwpoison_filter_flags_mask;
532 extern u64 hwpoison_filter_flags_value;
533 extern u64 hwpoison_filter_memcg;
534 extern u32 hwpoison_filter_enable;
535
536 extern unsigned long __must_check vm_mmap_pgoff(struct file *, unsigned long,
537 unsigned long, unsigned long,
538 unsigned long, unsigned long);
539
540 extern void set_pageblock_order(void);
541 unsigned long reclaim_clean_pages_from_list(struct zone *zone,
542 struct list_head *page_list);
543 /* The ALLOC_WMARK bits are used as an index to zone->watermark */
544 #define ALLOC_WMARK_MIN WMARK_MIN
545 #define ALLOC_WMARK_LOW WMARK_LOW
546 #define ALLOC_WMARK_HIGH WMARK_HIGH
547 #define ALLOC_NO_WATERMARKS 0x04 /* don't check watermarks at all */
548
549 /* Mask to get the watermark bits */
550 #define ALLOC_WMARK_MASK (ALLOC_NO_WATERMARKS-1)
551
552 /*
553 * Only MMU archs have async oom victim reclaim - aka oom_reaper so we
554 * cannot assume a reduced access to memory reserves is sufficient for
555 * !MMU
556 */
557 #ifdef CONFIG_MMU
558 #define ALLOC_OOM 0x08
559 #else
560 #define ALLOC_OOM ALLOC_NO_WATERMARKS
561 #endif
562
563 #define ALLOC_HARDER 0x10 /* try to alloc harder */
564 #define ALLOC_HIGH 0x20 /* __GFP_HIGH set */
565 #define ALLOC_CPUSET 0x40 /* check for correct cpuset */
566 #define ALLOC_CMA 0x80 /* allow allocations from CMA areas */
567 #ifdef CONFIG_ZONE_DMA32
568 #define ALLOC_NOFRAGMENT 0x100 /* avoid mixing pageblock types */
569 #else
570 #define ALLOC_NOFRAGMENT 0x0
571 #endif
572 #define ALLOC_KSWAPD 0x800 /* allow waking of kswapd, __GFP_KSWAPD_RECLAIM set */
573
574 enum ttu_flags;
575 struct tlbflush_unmap_batch;
576
577
578 /*
579 * only for MM internal work items which do not depend on
580 * any allocations or locks which might depend on allocations
581 */
582 extern struct workqueue_struct *mm_percpu_wq;
583
584 #ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH
585 void try_to_unmap_flush(void);
586 void try_to_unmap_flush_dirty(void);
587 void flush_tlb_batched_pending(struct mm_struct *mm);
588 #else
589 static inline void try_to_unmap_flush(void)
590 {
591 }
592 static inline void try_to_unmap_flush_dirty(void)
593 {
594 }
595 static inline void flush_tlb_batched_pending(struct mm_struct *mm)
596 {
597 }
598 #endif /* CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH */
599
600 extern const struct trace_print_flags pageflag_names[];
601 extern const struct trace_print_flags vmaflag_names[];
602 extern const struct trace_print_flags gfpflag_names[];
603
604 static inline bool is_migrate_highatomic(enum migratetype migratetype)
605 {
606 return migratetype == MIGRATE_HIGHATOMIC;
607 }
608
609 static inline bool is_migrate_highatomic_page(struct page *page)
610 {
611 return get_pageblock_migratetype(page) == MIGRATE_HIGHATOMIC;
612 }
613
614 void setup_zone_pageset(struct zone *zone);
615 extern struct page *alloc_new_node_page(struct page *page, unsigned long node);
616 #endif /* __MM_INTERNAL_H */