1b2441318SGreg Kroah-Hartman /* SPDX-License-Identifier: GPL-2.0 */
25b99cd0eSHeiko Carstens #ifndef _LINUX_MM_TYPES_H
35b99cd0eSHeiko Carstens #define _LINUX_MM_TYPES_H
45b99cd0eSHeiko Carstens
52e58f173SIngo Molnar #include <linux/mm_types_task.h>
62e58f173SIngo Molnar
74f9a58d7SOlaf Hering #include <linux/auxvec.h>
878db3412SSuren Baghdasaryan #include <linux/kref.h>
95b99cd0eSHeiko Carstens #include <linux/list.h>
105b99cd0eSHeiko Carstens #include <linux/spinlock.h>
11c92ff1bdSMartin Schwidefsky #include <linux/rbtree.h>
12d4af56c5SLiam R. Howlett #include <linux/maple_tree.h>
13c92ff1bdSMartin Schwidefsky #include <linux/rwsem.h>
14c92ff1bdSMartin Schwidefsky #include <linux/completion.h>
15cddb8a5cSAndrea Arcangeli #include <linux/cpumask.h>
16d4b3b638SSrikar Dronamraju #include <linux/uprobes.h>
178d491de6SThomas Gleixner #include <linux/rcupdate.h>
18bbeae5b0SPeter Zijlstra #include <linux/page-flags-layout.h>
19ec8d7c14SMichal Hocko #include <linux/workqueue.h>
2057efa1feSJason Gunthorpe #include <linux/seqlock.h>
21f1a79412SShakeel Butt #include <linux/percpu_counter.h>
222e58f173SIngo Molnar
23c92ff1bdSMartin Schwidefsky #include <asm/mmu.h>
245b99cd0eSHeiko Carstens
254f9a58d7SOlaf Hering #ifndef AT_VECTOR_SIZE_ARCH
264f9a58d7SOlaf Hering #define AT_VECTOR_SIZE_ARCH 0
274f9a58d7SOlaf Hering #endif
284f9a58d7SOlaf Hering #define AT_VECTOR_SIZE (2*(AT_VECTOR_SIZE_ARCH + AT_VECTOR_SIZE_BASE + 1))
294f9a58d7SOlaf Hering
3082e69a12SFenghua Yu #define INIT_PASID 0
311c8f4220SSouptick Joarder
325b99cd0eSHeiko Carstens struct address_space;
331306a85aSJohannes Weiner struct mem_cgroup;
345b99cd0eSHeiko Carstens
355b99cd0eSHeiko Carstens /*
365b99cd0eSHeiko Carstens * Each physical page in the system has a struct page associated with
375b99cd0eSHeiko Carstens * it to keep track of whatever it is we are using the page for at the
385b99cd0eSHeiko Carstens * moment. Note that we have no way to track which tasks are using
395b99cd0eSHeiko Carstens * a page, though if it is a pagecache page, rmap structures can tell us
4097b4a671SMatthew Wilcox * who is mapping it.
41be50015dSMatthew Wilcox *
4297b4a671SMatthew Wilcox * If you allocate the page using alloc_pages(), you can use some of the
4397b4a671SMatthew Wilcox * space in struct page for your own purposes. The five words in the main
4497b4a671SMatthew Wilcox * union are available, except for bit 0 of the first word which must be
4597b4a671SMatthew Wilcox * kept clear. Many users use this word to store a pointer to an object
4697b4a671SMatthew Wilcox * which is guaranteed to be aligned. If you use the same storage as
4797b4a671SMatthew Wilcox * page->mapping, you must restore it to NULL before freeing the page.
4897b4a671SMatthew Wilcox *
4997b4a671SMatthew Wilcox * If your page will not be mapped to userspace, you can also use the four
5097b4a671SMatthew Wilcox * bytes in the mapcount union, but you must call page_mapcount_reset()
5197b4a671SMatthew Wilcox * before freeing it.
5297b4a671SMatthew Wilcox *
5397b4a671SMatthew Wilcox * If you want to use the refcount field, it must be used in such a way
5497b4a671SMatthew Wilcox * that other CPUs temporarily incrementing and then decrementing the
55be50015dSMatthew Wilcox * refcount does not cause problems. On receiving the page from
56be50015dSMatthew Wilcox * alloc_pages(), the refcount will be positive.
57be50015dSMatthew Wilcox *
5897b4a671SMatthew Wilcox * If you allocate pages of order > 0, you can use some of the fields
5997b4a671SMatthew Wilcox * in each subpage, but you may need to restore some of their values
6097b4a671SMatthew Wilcox * afterwards.
61fc9bb8c7SChristoph Lameter *
62d122019bSMatthew Wilcox (Oracle) * SLUB uses cmpxchg_double() to atomically update its freelist and counters.
63d122019bSMatthew Wilcox (Oracle) * That requires that freelist & counters in struct slab be adjacent and
64d122019bSMatthew Wilcox (Oracle) * double-word aligned. Because struct slab currently just reinterprets the
65d122019bSMatthew Wilcox (Oracle) * bits of struct page, we align all struct pages to double-word boundaries,
66d122019bSMatthew Wilcox (Oracle) * and ensure that 'freelist' is aligned within struct slab.
675b99cd0eSHeiko Carstens */
68e20df2c6SMatthew Wilcox #ifdef CONFIG_HAVE_ALIGNED_STRUCT_PAGE
69e20df2c6SMatthew Wilcox #define _struct_page_alignment __aligned(2 * sizeof(unsigned long))
70e20df2c6SMatthew Wilcox #else
7170fb4fdfSLinus Torvalds #define _struct_page_alignment __aligned(sizeof(unsigned long))
727d27a04bSMatthew Wilcox #endif
73e20df2c6SMatthew Wilcox
745b99cd0eSHeiko Carstens struct page {
755b99cd0eSHeiko Carstens unsigned long flags; /* Atomic flags, some possibly
765b99cd0eSHeiko Carstens * updated asynchronously */
77b7ccc7f8SMatthew Wilcox /*
784da1984eSMatthew Wilcox * Five words (20/40 bytes) are available in this union.
794da1984eSMatthew Wilcox * WARNING: bit 0 of the first word is used for PageTail(). That
804da1984eSMatthew Wilcox * means the other users of this union MUST NOT use the bit to
81b7ccc7f8SMatthew Wilcox * avoid collision and false-positive PageTail().
82b7ccc7f8SMatthew Wilcox */
83b7ccc7f8SMatthew Wilcox union {
844da1984eSMatthew Wilcox struct { /* Page cache and anonymous pages */
854da1984eSMatthew Wilcox /**
864da1984eSMatthew Wilcox * @lru: Pageout list, eg. active_list protected by
8715b44736SHugh Dickins * lruvec->lru_lock. Sometimes used as a generic list
88b7ccc7f8SMatthew Wilcox * by the page owner.
89b7ccc7f8SMatthew Wilcox */
9007ca7606SHugh Dickins union {
914da1984eSMatthew Wilcox struct list_head lru;
92bf75f200SMel Gorman
9307ca7606SHugh Dickins /* Or, for the Unevictable "LRU list" slot */
9407ca7606SHugh Dickins struct {
9507ca7606SHugh Dickins /* Always even, to negate PageTail */
9607ca7606SHugh Dickins void *__filler;
9707ca7606SHugh Dickins /* Count page's or folio's mlocks */
9807ca7606SHugh Dickins unsigned int mlock_count;
9907ca7606SHugh Dickins };
100bf75f200SMel Gorman
101bf75f200SMel Gorman /* Or, free page */
102bf75f200SMel Gorman struct list_head buddy_list;
103bf75f200SMel Gorman struct list_head pcp_list;
10407ca7606SHugh Dickins };
10566a6ffd2SMatthew Wilcox /* See page-flags.h for PAGE_MAPPING_FLAGS */
106b26435a0SMatthew Wilcox struct address_space *mapping;
10716900426SShiyang Ruan union {
108fc9bb8c7SChristoph Lameter pgoff_t index; /* Our offset within mapping. */
10916900426SShiyang Ruan unsigned long share; /* share count for fsdax */
11016900426SShiyang Ruan };
11166a6ffd2SMatthew Wilcox /**
11266a6ffd2SMatthew Wilcox * @private: Mapping-private opaque data.
11366a6ffd2SMatthew Wilcox * Usually used for buffer_heads if PagePrivate.
11466a6ffd2SMatthew Wilcox * Used for swp_entry_t if PageSwapCache.
11566a6ffd2SMatthew Wilcox * Indicates order in the buddy system if PageBuddy.
1167d27a04bSMatthew Wilcox */
1177d27a04bSMatthew Wilcox unsigned long private;
11866a6ffd2SMatthew Wilcox };
119c25fff71SJesper Dangaard Brouer struct { /* page_pool used by netstack */
120c25fff71SJesper Dangaard Brouer /**
121c07aea3eSMatteo Croce * @pp_magic: magic value to avoid recycling non
122c07aea3eSMatteo Croce * page_pool allocated pages.
123c07aea3eSMatteo Croce */
124c07aea3eSMatteo Croce unsigned long pp_magic;
125c07aea3eSMatteo Croce struct page_pool *pp;
126c07aea3eSMatteo Croce unsigned long _pp_mapping_pad;
1270e9d2a0aSYunsheng Lin unsigned long dma_addr;
128f915b75bSYunsheng Lin union {
129f915b75bSYunsheng Lin /**
130f915b75bSYunsheng Lin * dma_addr_upper: might require a 64-bit
131f915b75bSYunsheng Lin * value on 32-bit architectures.
132f915b75bSYunsheng Lin */
133f915b75bSYunsheng Lin unsigned long dma_addr_upper;
134f915b75bSYunsheng Lin /**
135f915b75bSYunsheng Lin * For frag page support, not supported in
136f915b75bSYunsheng Lin * 32-bit architectures with 64-bit DMA.
137f915b75bSYunsheng Lin */
1380e9d2a0aSYunsheng Lin atomic_long_t pp_frag_count;
1390e9d2a0aSYunsheng Lin };
140f915b75bSYunsheng Lin };
1414da1984eSMatthew Wilcox struct { /* Tail pages of compound page */
1424da1984eSMatthew Wilcox unsigned long compound_head; /* Bit zero is set */
143dad6a5ebSHugh Dickins };
14450e7fbc3SMatthew Wilcox struct { /* ZONE_DEVICE pages */
14550e7fbc3SMatthew Wilcox /** @pgmap: Points to the hosting device page map. */
14650e7fbc3SMatthew Wilcox struct dev_pagemap *pgmap;
1478a164fefSChristoph Hellwig void *zone_device_data;
14876470ccdSRalph Campbell /*
14976470ccdSRalph Campbell * ZONE_DEVICE private pages are counted as being
15076470ccdSRalph Campbell * mapped so the next 3 words hold the mapping, index,
15176470ccdSRalph Campbell * and private fields from the source anonymous or
15276470ccdSRalph Campbell * page cache page while the page is migrated to device
15376470ccdSRalph Campbell * private memory.
15476470ccdSRalph Campbell * ZONE_DEVICE MEMORY_DEVICE_FS_DAX pages also
15576470ccdSRalph Campbell * use the mapping, index, and private fields when
15676470ccdSRalph Campbell * pmem backed DAX files are mapped.
15776470ccdSRalph Campbell */
15850e7fbc3SMatthew Wilcox };
1594da1984eSMatthew Wilcox
1604da1984eSMatthew Wilcox /** @rcu_head: You can use this to free a page by RCU. */
1614da1984eSMatthew Wilcox struct rcu_head rcu_head;
16266a6ffd2SMatthew Wilcox };
1637d27a04bSMatthew Wilcox
164b21999daSMatthew Wilcox union { /* This union is 4 bytes in size. */
165b21999daSMatthew Wilcox /*
166b21999daSMatthew Wilcox * If the page can be mapped to userspace, encodes the number
167b21999daSMatthew Wilcox * of times this page is referenced by a page table.
168b21999daSMatthew Wilcox */
169b21999daSMatthew Wilcox atomic_t _mapcount;
170b21999daSMatthew Wilcox
1717d27a04bSMatthew Wilcox /*
1726e292b9bSMatthew Wilcox * If the page is neither PageSlab nor mappable to userspace,
1736e292b9bSMatthew Wilcox * the value stored here may help determine what this page
1746e292b9bSMatthew Wilcox * is used for. See page-flags.h for a list of page types
1756e292b9bSMatthew Wilcox * which are currently stored here.
1766e292b9bSMatthew Wilcox */
1776e292b9bSMatthew Wilcox unsigned int page_type;
178b21999daSMatthew Wilcox };
179013e8963SChristoph Lameter
180b21999daSMatthew Wilcox /* Usage count. *DO NOT USE DIRECTLY*. See page_ref.h */
1810139aa7bSJoonsoo Kim atomic_t _refcount;
182fc9bb8c7SChristoph Lameter
1831306a85aSJohannes Weiner #ifdef CONFIG_MEMCG
184bcfe06bfSRoman Gushchin unsigned long memcg_data;
1851306a85aSJohannes Weiner #endif
1861306a85aSJohannes Weiner
1875b99cd0eSHeiko Carstens /*
1885b99cd0eSHeiko Carstens * On machines where all RAM is mapped into kernel address space,
1895b99cd0eSHeiko Carstens * we can simply calculate the virtual address. On machines with
1905b99cd0eSHeiko Carstens * highmem some memory is mapped into kernel virtual memory
1915b99cd0eSHeiko Carstens * dynamically, so we need a place to store that address.
1925b99cd0eSHeiko Carstens * Note that this field could be 16 bits on x86 ... ;)
1935b99cd0eSHeiko Carstens *
1945b99cd0eSHeiko Carstens * Architectures with slow multiplication can define
1955b99cd0eSHeiko Carstens * WANT_PAGE_VIRTUAL in asm/page.h
1965b99cd0eSHeiko Carstens */
1975b99cd0eSHeiko Carstens #if defined(WANT_PAGE_VIRTUAL)
1985b99cd0eSHeiko Carstens void *virtual; /* Kernel virtual address (NULL if
1995b99cd0eSHeiko Carstens not kmapped, ie. highmem) */
2005b99cd0eSHeiko Carstens #endif /* WANT_PAGE_VIRTUAL */
201dfec072eSVegard Nossum
202f80be457SAlexander Potapenko #ifdef CONFIG_KMSAN
203f80be457SAlexander Potapenko /*
204f80be457SAlexander Potapenko * KMSAN metadata for this page:
205f80be457SAlexander Potapenko * - shadow page: every bit indicates whether the corresponding
206f80be457SAlexander Potapenko * bit of the original page is initialized (0) or not (1);
207f80be457SAlexander Potapenko * - origin page: every 4 bytes contain an id of the stack trace
208f80be457SAlexander Potapenko * where the uninitialized value was created.
209f80be457SAlexander Potapenko */
210f80be457SAlexander Potapenko struct page *kmsan_shadow;
211f80be457SAlexander Potapenko struct page *kmsan_origin;
212f80be457SAlexander Potapenko #endif
213f80be457SAlexander Potapenko
21490572890SPeter Zijlstra #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
21590572890SPeter Zijlstra int _last_cpupid;
21657e0a030SMel Gorman #endif
217e20df2c6SMatthew Wilcox } _struct_page_alignment;
2185b99cd0eSHeiko Carstens
21970fb4fdfSLinus Torvalds /*
22070fb4fdfSLinus Torvalds * struct encoded_page - a nonexistent type marking this pointer
22170fb4fdfSLinus Torvalds *
22270fb4fdfSLinus Torvalds * An 'encoded_page' pointer is a pointer to a regular 'struct page', but
22370fb4fdfSLinus Torvalds * with the low bits of the pointer indicating extra context-dependent
22470fb4fdfSLinus Torvalds * information. Not super-common, but happens in mmu_gather and mlock
22570fb4fdfSLinus Torvalds * handling, and this acts as a type system check on that use.
22670fb4fdfSLinus Torvalds *
22770fb4fdfSLinus Torvalds * We only really have two guaranteed bits in general, although you could
22870fb4fdfSLinus Torvalds * play with 'struct page' alignment (see CONFIG_HAVE_ALIGNED_STRUCT_PAGE)
22970fb4fdfSLinus Torvalds * for more.
23070fb4fdfSLinus Torvalds *
23170fb4fdfSLinus Torvalds * Use the supplied helper functions to endcode/decode the pointer and bits.
23270fb4fdfSLinus Torvalds */
23370fb4fdfSLinus Torvalds struct encoded_page;
23470fb4fdfSLinus Torvalds #define ENCODE_PAGE_BITS 3ul
encode_page(struct page * page,unsigned long flags)23570fb4fdfSLinus Torvalds static __always_inline struct encoded_page *encode_page(struct page *page, unsigned long flags)
23670fb4fdfSLinus Torvalds {
23770fb4fdfSLinus Torvalds BUILD_BUG_ON(flags > ENCODE_PAGE_BITS);
23870fb4fdfSLinus Torvalds return (struct encoded_page *)(flags | (unsigned long)page);
23970fb4fdfSLinus Torvalds }
24070fb4fdfSLinus Torvalds
encoded_page_flags(struct encoded_page * page)24170fb4fdfSLinus Torvalds static inline unsigned long encoded_page_flags(struct encoded_page *page)
24270fb4fdfSLinus Torvalds {
24370fb4fdfSLinus Torvalds return ENCODE_PAGE_BITS & (unsigned long)page;
24470fb4fdfSLinus Torvalds }
24570fb4fdfSLinus Torvalds
encoded_page_ptr(struct encoded_page * page)24670fb4fdfSLinus Torvalds static inline struct page *encoded_page_ptr(struct encoded_page *page)
24770fb4fdfSLinus Torvalds {
24870fb4fdfSLinus Torvalds return (struct page *)(~ENCODE_PAGE_BITS & (unsigned long)page);
24970fb4fdfSLinus Torvalds }
25070fb4fdfSLinus Torvalds
251*85a13334SMatthew Wilcox /*
252*85a13334SMatthew Wilcox * A swap entry has to fit into a "unsigned long", as the entry is hidden
253*85a13334SMatthew Wilcox * in the "index" field of the swapper address space.
254*85a13334SMatthew Wilcox */
255*85a13334SMatthew Wilcox typedef struct {
256*85a13334SMatthew Wilcox unsigned long val;
257*85a13334SMatthew Wilcox } swp_entry_t;
258*85a13334SMatthew Wilcox
2597b230db3SMatthew Wilcox (Oracle) /**
2607b230db3SMatthew Wilcox (Oracle) * struct folio - Represents a contiguous set of bytes.
2617b230db3SMatthew Wilcox (Oracle) * @flags: Identical to the page flags.
2627b230db3SMatthew Wilcox (Oracle) * @lru: Least Recently Used list; tracks how recently this folio was used.
263334f6f53SMatthew Wilcox (Oracle) * @mlock_count: Number of times this folio has been pinned by mlock().
2647b230db3SMatthew Wilcox (Oracle) * @mapping: The file this page belongs to, or refers to the anon_vma for
2657b230db3SMatthew Wilcox (Oracle) * anonymous memory.
2667b230db3SMatthew Wilcox (Oracle) * @index: Offset within the file, in units of pages. For anonymous memory,
2677b230db3SMatthew Wilcox (Oracle) * this is the index from the beginning of the mmap.
2687b230db3SMatthew Wilcox (Oracle) * @private: Filesystem per-folio data (see folio_attach_private()).
269*85a13334SMatthew Wilcox * @swap: Used for swp_entry_t if folio_test_swapcache().
2707b230db3SMatthew Wilcox (Oracle) * @_mapcount: Do not access this member directly. Use folio_mapcount() to
2717b230db3SMatthew Wilcox (Oracle) * find out how many times this folio is mapped by userspace.
2727b230db3SMatthew Wilcox (Oracle) * @_refcount: Do not access this member directly. Use folio_ref_count()
2737b230db3SMatthew Wilcox (Oracle) * to find how many references there are to this folio.
2747b230db3SMatthew Wilcox (Oracle) * @memcg_data: Memory Control Group data.
275b14224fbSMatthew Wilcox (Oracle) * @_entire_mapcount: Do not use directly, call folio_entire_mapcount().
276eec20426SMatthew Wilcox (Oracle) * @_nr_pages_mapped: Do not use directly, call folio_mapcount().
277379708ffSMatthew Wilcox (Oracle) * @_pincount: Do not use directly, call folio_maybe_dma_pinned().
278379708ffSMatthew Wilcox (Oracle) * @_folio_nr_pages: Do not use directly, call folio_nr_pages().
279dad6a5ebSHugh Dickins * @_hugetlb_subpool: Do not use directly, use accessor in hugetlb.h.
280dad6a5ebSHugh Dickins * @_hugetlb_cgroup: Do not use directly, use accessor in hugetlb_cgroup.h.
281dad6a5ebSHugh Dickins * @_hugetlb_cgroup_rsvd: Do not use directly, use accessor in hugetlb_cgroup.h.
282dad6a5ebSHugh Dickins * @_hugetlb_hwpoison: Do not use directly, call raw_hwp_list_head().
2834375a553SMatthew Wilcox (Oracle) * @_deferred_list: Folios to be split under memory pressure.
2847b230db3SMatthew Wilcox (Oracle) *
2857b230db3SMatthew Wilcox (Oracle) * A folio is a physically, virtually and logically contiguous set
2867b230db3SMatthew Wilcox (Oracle) * of bytes. It is a power-of-two in size, and it is aligned to that
2877b230db3SMatthew Wilcox (Oracle) * same power-of-two. It is at least as large as %PAGE_SIZE. If it is
2887b230db3SMatthew Wilcox (Oracle) * in the page cache, it is at a file offset which is a multiple of that
2897b230db3SMatthew Wilcox (Oracle) * power-of-two. It may be mapped into userspace at an address which is
2907b230db3SMatthew Wilcox (Oracle) * at an arbitrary page offset, but its kernel virtual address is aligned
2917b230db3SMatthew Wilcox (Oracle) * to its size.
2927b230db3SMatthew Wilcox (Oracle) */
2937b230db3SMatthew Wilcox (Oracle) struct folio {
2947b230db3SMatthew Wilcox (Oracle) /* private: don't document the anon union */
2957b230db3SMatthew Wilcox (Oracle) union {
2967b230db3SMatthew Wilcox (Oracle) struct {
2977b230db3SMatthew Wilcox (Oracle) /* public: */
2987b230db3SMatthew Wilcox (Oracle) unsigned long flags;
29907ca7606SHugh Dickins union {
3007b230db3SMatthew Wilcox (Oracle) struct list_head lru;
301334f6f53SMatthew Wilcox (Oracle) /* private: avoid cluttering the output */
30207ca7606SHugh Dickins struct {
30307ca7606SHugh Dickins void *__filler;
304334f6f53SMatthew Wilcox (Oracle) /* public: */
30507ca7606SHugh Dickins unsigned int mlock_count;
306334f6f53SMatthew Wilcox (Oracle) /* private: */
30707ca7606SHugh Dickins };
308334f6f53SMatthew Wilcox (Oracle) /* public: */
30907ca7606SHugh Dickins };
3107b230db3SMatthew Wilcox (Oracle) struct address_space *mapping;
3117b230db3SMatthew Wilcox (Oracle) pgoff_t index;
312*85a13334SMatthew Wilcox union {
3137b230db3SMatthew Wilcox (Oracle) void *private;
314*85a13334SMatthew Wilcox swp_entry_t swap;
315*85a13334SMatthew Wilcox };
3167b230db3SMatthew Wilcox (Oracle) atomic_t _mapcount;
3177b230db3SMatthew Wilcox (Oracle) atomic_t _refcount;
3187b230db3SMatthew Wilcox (Oracle) #ifdef CONFIG_MEMCG
3197b230db3SMatthew Wilcox (Oracle) unsigned long memcg_data;
3207b230db3SMatthew Wilcox (Oracle) #endif
3217b230db3SMatthew Wilcox (Oracle) /* private: the union with struct page is transitional */
3227b230db3SMatthew Wilcox (Oracle) };
3237b230db3SMatthew Wilcox (Oracle) struct page page;
3247b230db3SMatthew Wilcox (Oracle) };
325dad6a5ebSHugh Dickins union {
326dad6a5ebSHugh Dickins struct {
327379708ffSMatthew Wilcox (Oracle) unsigned long _flags_1;
328dad6a5ebSHugh Dickins unsigned long _head_1;
329ebc1baf5SMatthew Wilcox (Oracle) unsigned long _folio_avail;
330a8d55327SMatthew Wilcox (Oracle) /* public: */
331b14224fbSMatthew Wilcox (Oracle) atomic_t _entire_mapcount;
332eec20426SMatthew Wilcox (Oracle) atomic_t _nr_pages_mapped;
333379708ffSMatthew Wilcox (Oracle) atomic_t _pincount;
334379708ffSMatthew Wilcox (Oracle) #ifdef CONFIG_64BIT
335379708ffSMatthew Wilcox (Oracle) unsigned int _folio_nr_pages;
336b10ff04dSMatthew Wilcox (Oracle) #endif
337cfeed8ffSDavid Hildenbrand /* private: the union with struct page is transitional */
338dad6a5ebSHugh Dickins };
339dad6a5ebSHugh Dickins struct page __page_1;
340dad6a5ebSHugh Dickins };
341dad6a5ebSHugh Dickins union {
342dad6a5ebSHugh Dickins struct {
343dad6a5ebSHugh Dickins unsigned long _flags_2;
344dad6a5ebSHugh Dickins unsigned long _head_2;
345a8d55327SMatthew Wilcox (Oracle) /* public: */
346dad6a5ebSHugh Dickins void *_hugetlb_subpool;
347dad6a5ebSHugh Dickins void *_hugetlb_cgroup;
348dad6a5ebSHugh Dickins void *_hugetlb_cgroup_rsvd;
349dad6a5ebSHugh Dickins void *_hugetlb_hwpoison;
350a8d55327SMatthew Wilcox (Oracle) /* private: the union with struct page is transitional */
351dad6a5ebSHugh Dickins };
3524375a553SMatthew Wilcox (Oracle) struct {
3534375a553SMatthew Wilcox (Oracle) unsigned long _flags_2a;
3544375a553SMatthew Wilcox (Oracle) unsigned long _head_2a;
3554375a553SMatthew Wilcox (Oracle) /* public: */
3564375a553SMatthew Wilcox (Oracle) struct list_head _deferred_list;
3574375a553SMatthew Wilcox (Oracle) /* private: the union with struct page is transitional */
358dad6a5ebSHugh Dickins };
359dad6a5ebSHugh Dickins struct page __page_2;
360dad6a5ebSHugh Dickins };
3617b230db3SMatthew Wilcox (Oracle) };
3627b230db3SMatthew Wilcox (Oracle)
3637b230db3SMatthew Wilcox (Oracle) #define FOLIO_MATCH(pg, fl) \
3647b230db3SMatthew Wilcox (Oracle) static_assert(offsetof(struct page, pg) == offsetof(struct folio, fl))
3657b230db3SMatthew Wilcox (Oracle) FOLIO_MATCH(flags, flags);
3667b230db3SMatthew Wilcox (Oracle) FOLIO_MATCH(lru, lru);
367536f4217SWei Yang FOLIO_MATCH(mapping, mapping);
3687b230db3SMatthew Wilcox (Oracle) FOLIO_MATCH(compound_head, lru);
3697b230db3SMatthew Wilcox (Oracle) FOLIO_MATCH(index, index);
3707b230db3SMatthew Wilcox (Oracle) FOLIO_MATCH(private, private);
3717b230db3SMatthew Wilcox (Oracle) FOLIO_MATCH(_mapcount, _mapcount);
3727b230db3SMatthew Wilcox (Oracle) FOLIO_MATCH(_refcount, _refcount);
3737b230db3SMatthew Wilcox (Oracle) #ifdef CONFIG_MEMCG
3747b230db3SMatthew Wilcox (Oracle) FOLIO_MATCH(memcg_data, memcg_data);
3757b230db3SMatthew Wilcox (Oracle) #endif
3767b230db3SMatthew Wilcox (Oracle) #undef FOLIO_MATCH
377379708ffSMatthew Wilcox (Oracle) #define FOLIO_MATCH(pg, fl) \
378379708ffSMatthew Wilcox (Oracle) static_assert(offsetof(struct folio, fl) == \
379379708ffSMatthew Wilcox (Oracle) offsetof(struct page, pg) + sizeof(struct page))
380379708ffSMatthew Wilcox (Oracle) FOLIO_MATCH(flags, _flags_1);
381dad6a5ebSHugh Dickins FOLIO_MATCH(compound_head, _head_1);
382379708ffSMatthew Wilcox (Oracle) #undef FOLIO_MATCH
383dad6a5ebSHugh Dickins #define FOLIO_MATCH(pg, fl) \
384dad6a5ebSHugh Dickins static_assert(offsetof(struct folio, fl) == \
385dad6a5ebSHugh Dickins offsetof(struct page, pg) + 2 * sizeof(struct page))
386dad6a5ebSHugh Dickins FOLIO_MATCH(flags, _flags_2);
387dad6a5ebSHugh Dickins FOLIO_MATCH(compound_head, _head_2);
388b10ff04dSMatthew Wilcox (Oracle) FOLIO_MATCH(flags, _flags_2a);
389b10ff04dSMatthew Wilcox (Oracle) FOLIO_MATCH(compound_head, _head_2a);
390dad6a5ebSHugh Dickins #undef FOLIO_MATCH
3917b230db3SMatthew Wilcox (Oracle)
3929a35de4fSVishal Moola (Oracle) /**
3939a35de4fSVishal Moola (Oracle) * struct ptdesc - Memory descriptor for page tables.
3949a35de4fSVishal Moola (Oracle) * @__page_flags: Same as page flags. Unused for page tables.
3959a35de4fSVishal Moola (Oracle) * @pt_rcu_head: For freeing page table pages.
3969a35de4fSVishal Moola (Oracle) * @pt_list: List of used page tables. Used for s390 and x86.
3979a35de4fSVishal Moola (Oracle) * @_pt_pad_1: Padding that aliases with page's compound head.
3989a35de4fSVishal Moola (Oracle) * @pmd_huge_pte: Protected by ptdesc->ptl, used for THPs.
3999a35de4fSVishal Moola (Oracle) * @__page_mapping: Aliases with page->mapping. Unused for page tables.
4009a35de4fSVishal Moola (Oracle) * @pt_mm: Used for x86 pgds.
4019a35de4fSVishal Moola (Oracle) * @pt_frag_refcount: For fragmented page table tracking. Powerpc and s390 only.
4029a35de4fSVishal Moola (Oracle) * @_pt_pad_2: Padding to ensure proper alignment.
4039a35de4fSVishal Moola (Oracle) * @ptl: Lock for the page table.
4049a35de4fSVishal Moola (Oracle) * @__page_type: Same as page->page_type. Unused for page tables.
4059a35de4fSVishal Moola (Oracle) * @_refcount: Same as page refcount. Used for s390 page tables.
4069a35de4fSVishal Moola (Oracle) * @pt_memcg_data: Memcg data. Tracked for page tables here.
4079a35de4fSVishal Moola (Oracle) *
4089a35de4fSVishal Moola (Oracle) * This struct overlays struct page for now. Do not modify without a good
4099a35de4fSVishal Moola (Oracle) * understanding of the issues.
4109a35de4fSVishal Moola (Oracle) */
4119a35de4fSVishal Moola (Oracle) struct ptdesc {
4129a35de4fSVishal Moola (Oracle) unsigned long __page_flags;
4139a35de4fSVishal Moola (Oracle)
4149a35de4fSVishal Moola (Oracle) union {
4159a35de4fSVishal Moola (Oracle) struct rcu_head pt_rcu_head;
4169a35de4fSVishal Moola (Oracle) struct list_head pt_list;
4179a35de4fSVishal Moola (Oracle) struct {
4189a35de4fSVishal Moola (Oracle) unsigned long _pt_pad_1;
4199a35de4fSVishal Moola (Oracle) pgtable_t pmd_huge_pte;
4209a35de4fSVishal Moola (Oracle) };
4219a35de4fSVishal Moola (Oracle) };
4229a35de4fSVishal Moola (Oracle) unsigned long __page_mapping;
4239a35de4fSVishal Moola (Oracle)
4249a35de4fSVishal Moola (Oracle) union {
4259a35de4fSVishal Moola (Oracle) struct mm_struct *pt_mm;
4269a35de4fSVishal Moola (Oracle) atomic_t pt_frag_refcount;
4279a35de4fSVishal Moola (Oracle) };
4289a35de4fSVishal Moola (Oracle)
4299a35de4fSVishal Moola (Oracle) union {
4309a35de4fSVishal Moola (Oracle) unsigned long _pt_pad_2;
4319a35de4fSVishal Moola (Oracle) #if ALLOC_SPLIT_PTLOCKS
4329a35de4fSVishal Moola (Oracle) spinlock_t *ptl;
4339a35de4fSVishal Moola (Oracle) #else
4349a35de4fSVishal Moola (Oracle) spinlock_t ptl;
4359a35de4fSVishal Moola (Oracle) #endif
4369a35de4fSVishal Moola (Oracle) };
4379a35de4fSVishal Moola (Oracle) unsigned int __page_type;
4389a35de4fSVishal Moola (Oracle) atomic_t _refcount;
4399a35de4fSVishal Moola (Oracle) #ifdef CONFIG_MEMCG
4409a35de4fSVishal Moola (Oracle) unsigned long pt_memcg_data;
4419a35de4fSVishal Moola (Oracle) #endif
4429a35de4fSVishal Moola (Oracle) };
4439a35de4fSVishal Moola (Oracle)
4449a35de4fSVishal Moola (Oracle) #define TABLE_MATCH(pg, pt) \
4459a35de4fSVishal Moola (Oracle) static_assert(offsetof(struct page, pg) == offsetof(struct ptdesc, pt))
4469a35de4fSVishal Moola (Oracle) TABLE_MATCH(flags, __page_flags);
4479a35de4fSVishal Moola (Oracle) TABLE_MATCH(compound_head, pt_list);
4489a35de4fSVishal Moola (Oracle) TABLE_MATCH(compound_head, _pt_pad_1);
4499a35de4fSVishal Moola (Oracle) TABLE_MATCH(mapping, __page_mapping);
4509a35de4fSVishal Moola (Oracle) TABLE_MATCH(rcu_head, pt_rcu_head);
4519a35de4fSVishal Moola (Oracle) TABLE_MATCH(page_type, __page_type);
4529a35de4fSVishal Moola (Oracle) TABLE_MATCH(_refcount, _refcount);
4539a35de4fSVishal Moola (Oracle) #ifdef CONFIG_MEMCG
4549a35de4fSVishal Moola (Oracle) TABLE_MATCH(memcg_data, pt_memcg_data);
4559a35de4fSVishal Moola (Oracle) #endif
4569a35de4fSVishal Moola (Oracle) #undef TABLE_MATCH
4579a35de4fSVishal Moola (Oracle) static_assert(sizeof(struct ptdesc) <= sizeof(struct page));
4589a35de4fSVishal Moola (Oracle)
459bf2d4334SVishal Moola (Oracle) #define ptdesc_page(pt) (_Generic((pt), \
460bf2d4334SVishal Moola (Oracle) const struct ptdesc *: (const struct page *)(pt), \
461bf2d4334SVishal Moola (Oracle) struct ptdesc *: (struct page *)(pt)))
462bf2d4334SVishal Moola (Oracle)
463bf2d4334SVishal Moola (Oracle) #define ptdesc_folio(pt) (_Generic((pt), \
464bf2d4334SVishal Moola (Oracle) const struct ptdesc *: (const struct folio *)(pt), \
465bf2d4334SVishal Moola (Oracle) struct ptdesc *: (struct folio *)(pt)))
466bf2d4334SVishal Moola (Oracle)
467bf2d4334SVishal Moola (Oracle) #define page_ptdesc(p) (_Generic((p), \
468bf2d4334SVishal Moola (Oracle) const struct page *: (const struct ptdesc *)(p), \
469bf2d4334SVishal Moola (Oracle) struct page *: (struct ptdesc *)(p)))
470bf2d4334SVishal Moola (Oracle)
471d1402fc7SLogan Gunthorpe /*
472d1402fc7SLogan Gunthorpe * Used for sizing the vmemmap region on some architectures
473d1402fc7SLogan Gunthorpe */
474d1402fc7SLogan Gunthorpe #define STRUCT_PAGE_MAX_SHIFT (order_base_2(sizeof(struct page)))
475d1402fc7SLogan Gunthorpe
476b63ae8caSAlexander Duyck #define PAGE_FRAG_CACHE_MAX_SIZE __ALIGN_MASK(32768, ~PAGE_MASK)
477b63ae8caSAlexander Duyck #define PAGE_FRAG_CACHE_MAX_ORDER get_order(PAGE_FRAG_CACHE_MAX_SIZE)
478b63ae8caSAlexander Duyck
47985d0a2edSMatthew Wilcox (Oracle) /*
48085d0a2edSMatthew Wilcox (Oracle) * page_private can be used on tail pages. However, PagePrivate is only
48185d0a2edSMatthew Wilcox (Oracle) * checked by the VM on the head page. So page_private on the tail pages
48285d0a2edSMatthew Wilcox (Oracle) * should be used for data that's ancillary to the head page (eg attaching
48385d0a2edSMatthew Wilcox (Oracle) * buffer heads to tail pages after attaching buffer heads to the head page)
48485d0a2edSMatthew Wilcox (Oracle) */
485b03641afSDan Williams #define page_private(page) ((page)->private)
48660e65a6fSGuoqing Jiang
set_page_private(struct page * page,unsigned long private)48760e65a6fSGuoqing Jiang static inline void set_page_private(struct page *page, unsigned long private)
48860e65a6fSGuoqing Jiang {
48960e65a6fSGuoqing Jiang page->private = private;
49060e65a6fSGuoqing Jiang }
491b03641afSDan Williams
folio_get_private(struct folio * folio)49285d0a2edSMatthew Wilcox (Oracle) static inline void *folio_get_private(struct folio *folio)
49385d0a2edSMatthew Wilcox (Oracle) {
49485d0a2edSMatthew Wilcox (Oracle) return folio->private;
49585d0a2edSMatthew Wilcox (Oracle) }
49685d0a2edSMatthew Wilcox (Oracle)
497b63ae8caSAlexander Duyck struct page_frag_cache {
498b63ae8caSAlexander Duyck void * va;
499b63ae8caSAlexander Duyck #if (PAGE_SIZE < PAGE_FRAG_CACHE_MAX_SIZE)
500b63ae8caSAlexander Duyck __u16 offset;
501b63ae8caSAlexander Duyck __u16 size;
502b63ae8caSAlexander Duyck #else
503b63ae8caSAlexander Duyck __u32 offset;
504b63ae8caSAlexander Duyck #endif
505b63ae8caSAlexander Duyck /* we maintain a pagecount bias, so that we dont dirty cache line
5060139aa7bSJoonsoo Kim * containing page->_refcount every time we allocate a fragment.
507b63ae8caSAlexander Duyck */
508b63ae8caSAlexander Duyck unsigned int pagecnt_bias;
509b63ae8caSAlexander Duyck bool pfmemalloc;
510b63ae8caSAlexander Duyck };
511b63ae8caSAlexander Duyck
51264b990d2SKirill A. Shutemov typedef unsigned long vm_flags_t;
513ca16d140SKOSAKI Motohiro
514c92ff1bdSMartin Schwidefsky /*
5158feae131SDavid Howells * A region containing a mapping of a non-memory backed file under NOMMU
5168feae131SDavid Howells * conditions. These are held in a global tree and are pinned by the VMAs that
5178feae131SDavid Howells * map parts of them.
5188feae131SDavid Howells */
5198feae131SDavid Howells struct vm_region {
5208feae131SDavid Howells struct rb_node vm_rb; /* link in global region tree */
521ca16d140SKOSAKI Motohiro vm_flags_t vm_flags; /* VMA vm_flags */
5228feae131SDavid Howells unsigned long vm_start; /* start address of region */
5238feae131SDavid Howells unsigned long vm_end; /* region initialised to here */
524dd8632a1SPaul Mundt unsigned long vm_top; /* region allocated to here */
5258feae131SDavid Howells unsigned long vm_pgoff; /* the offset in vm_file corresponding to vm_start */
5268feae131SDavid Howells struct file *vm_file; /* the backing file or NULL */
5278feae131SDavid Howells
5281e2ae599SDavid Howells int vm_usage; /* region usage count (access under nommu_region_sem) */
529cfe79c00SMike Frysinger bool vm_icache_flushed : 1; /* true if the icache has been flushed for
530cfe79c00SMike Frysinger * this region */
5318feae131SDavid Howells };
5328feae131SDavid Howells
533745f234bSAndrea Arcangeli #ifdef CONFIG_USERFAULTFD
534745f234bSAndrea Arcangeli #define NULL_VM_UFFD_CTX ((struct vm_userfaultfd_ctx) { NULL, })
535745f234bSAndrea Arcangeli struct vm_userfaultfd_ctx {
536745f234bSAndrea Arcangeli struct userfaultfd_ctx *ctx;
537745f234bSAndrea Arcangeli };
538745f234bSAndrea Arcangeli #else /* CONFIG_USERFAULTFD */
539745f234bSAndrea Arcangeli #define NULL_VM_UFFD_CTX ((struct vm_userfaultfd_ctx) {})
540745f234bSAndrea Arcangeli struct vm_userfaultfd_ctx {};
541745f234bSAndrea Arcangeli #endif /* CONFIG_USERFAULTFD */
542745f234bSAndrea Arcangeli
54378db3412SSuren Baghdasaryan struct anon_vma_name {
54478db3412SSuren Baghdasaryan struct kref kref;
54578db3412SSuren Baghdasaryan /* The name needs to be at the end because it is dynamically sized. */
54678db3412SSuren Baghdasaryan char name[];
54778db3412SSuren Baghdasaryan };
54878db3412SSuren Baghdasaryan
549c7f8f31cSSuren Baghdasaryan struct vma_lock {
550c7f8f31cSSuren Baghdasaryan struct rw_semaphore lock;
551c7f8f31cSSuren Baghdasaryan };
552c7f8f31cSSuren Baghdasaryan
553ef6a22b7SMel Gorman struct vma_numab_state {
554ef6a22b7SMel Gorman unsigned long next_scan;
55520f58648SRaghavendra K T unsigned long next_pid_reset;
55620f58648SRaghavendra K T unsigned long access_pids[2];
557ef6a22b7SMel Gorman };
558ef6a22b7SMel Gorman
5598feae131SDavid Howells /*
560552657b7Schenqiwu * This struct describes a virtual memory area. There is one of these
561c92ff1bdSMartin Schwidefsky * per VM-area/task. A VM area is any part of the process virtual memory
562c92ff1bdSMartin Schwidefsky * space that has a special rule for the page-fault handlers (ie a shared
563c92ff1bdSMartin Schwidefsky * library, the executable area etc).
564c92ff1bdSMartin Schwidefsky */
565c92ff1bdSMartin Schwidefsky struct vm_area_struct {
566e4c6bfd2SRik van Riel /* The first cache line has the info for VMA tree walking. */
567e4c6bfd2SRik van Riel
56820cce633SMichel Lespinasse union {
56920cce633SMichel Lespinasse struct {
57020cce633SMichel Lespinasse /* VMA covers [vm_start; vm_end) addresses within mm */
57120cce633SMichel Lespinasse unsigned long vm_start;
57220cce633SMichel Lespinasse unsigned long vm_end;
57320cce633SMichel Lespinasse };
57420cce633SMichel Lespinasse #ifdef CONFIG_PER_VMA_LOCK
57520cce633SMichel Lespinasse struct rcu_head vm_rcu; /* Used for deferred freeing. */
57620cce633SMichel Lespinasse #endif
57720cce633SMichel Lespinasse };
578c92ff1bdSMartin Schwidefsky
579e4c6bfd2SRik van Riel struct mm_struct *vm_mm; /* The address space we belong to. */
58028d8b812SLorenzo Stoakes pgprot_t vm_page_prot; /* Access permissions of this VMA. */
581bc292ab0SSuren Baghdasaryan
582bc292ab0SSuren Baghdasaryan /*
583bc292ab0SSuren Baghdasaryan * Flags, see mm.h.
584bc292ab0SSuren Baghdasaryan * To modify use vm_flags_{init|reset|set|clear|mod} functions.
585bc292ab0SSuren Baghdasaryan */
586bc292ab0SSuren Baghdasaryan union {
587bc292ab0SSuren Baghdasaryan const vm_flags_t vm_flags;
588bc292ab0SSuren Baghdasaryan vm_flags_t __private __vm_flags;
589bc292ab0SSuren Baghdasaryan };
590e4c6bfd2SRik van Riel
5915e31275cSSuren Baghdasaryan #ifdef CONFIG_PER_VMA_LOCK
592b1f02b95SJann Horn /*
593b1f02b95SJann Horn * Can only be written (using WRITE_ONCE()) while holding both:
594b1f02b95SJann Horn * - mmap_lock (in write mode)
595b1f02b95SJann Horn * - vm_lock->lock (in write mode)
596b1f02b95SJann Horn * Can be read reliably while holding one of:
597b1f02b95SJann Horn * - mmap_lock (in read or write mode)
598b1f02b95SJann Horn * - vm_lock->lock (in read or write mode)
599b1f02b95SJann Horn * Can be read unreliably (using READ_ONCE()) for pessimistic bailout
600b1f02b95SJann Horn * while holding nothing (except RCU to keep the VMA struct allocated).
601b1f02b95SJann Horn *
602b1f02b95SJann Horn * This sequence counter is explicitly allowed to overflow; sequence
603b1f02b95SJann Horn * counter reuse can only lead to occasional unnecessary use of the
604b1f02b95SJann Horn * slowpath.
605b1f02b95SJann Horn */
6065e31275cSSuren Baghdasaryan int vm_lock_seq;
607c7f8f31cSSuren Baghdasaryan struct vma_lock *vm_lock;
608457f67beSSuren Baghdasaryan
609457f67beSSuren Baghdasaryan /* Flag to indicate areas detached from the mm->mm_mt tree */
610457f67beSSuren Baghdasaryan bool detached;
6115e31275cSSuren Baghdasaryan #endif
6125e31275cSSuren Baghdasaryan
613d3737187SMichel Lespinasse /*
614c92ff1bdSMartin Schwidefsky * For areas with an address space and backing store,
61527ba0644SKirill A. Shutemov * linkage into the address_space->i_mmap interval tree.
6169a10064fSColin Cross *
617c92ff1bdSMartin Schwidefsky */
618c92ff1bdSMartin Schwidefsky struct {
6196b2dbba8SMichel Lespinasse struct rb_node rb;
6206b2dbba8SMichel Lespinasse unsigned long rb_subtree_last;
621c92ff1bdSMartin Schwidefsky } shared;
622c92ff1bdSMartin Schwidefsky
623c92ff1bdSMartin Schwidefsky /*
624c92ff1bdSMartin Schwidefsky * A file's MAP_PRIVATE vma can be in both i_mmap tree and anon_vma
625c92ff1bdSMartin Schwidefsky * list, after a COW of one of the file pages. A MAP_SHARED vma
626c92ff1bdSMartin Schwidefsky * can only be in the i_mmap tree. An anonymous MAP_PRIVATE, stack
627c92ff1bdSMartin Schwidefsky * or brk vma (with NULL file) can only be in an anon_vma list.
628c92ff1bdSMartin Schwidefsky */
629c1e8d7c6SMichel Lespinasse struct list_head anon_vma_chain; /* Serialized by mmap_lock &
6305beb4930SRik van Riel * page_table_lock */
631c92ff1bdSMartin Schwidefsky struct anon_vma *anon_vma; /* Serialized by page_table_lock */
632c92ff1bdSMartin Schwidefsky
633c92ff1bdSMartin Schwidefsky /* Function pointers to deal with this struct. */
634f0f37e2fSAlexey Dobriyan const struct vm_operations_struct *vm_ops;
635c92ff1bdSMartin Schwidefsky
636c92ff1bdSMartin Schwidefsky /* Information about our backing store: */
637c92ff1bdSMartin Schwidefsky unsigned long vm_pgoff; /* Offset (within vm_file) in PAGE_SIZE
638ea1754a0SKirill A. Shutemov units */
639c92ff1bdSMartin Schwidefsky struct file * vm_file; /* File we map to (can be NULL). */
640c92ff1bdSMartin Schwidefsky void * vm_private_data; /* was vm_pte (shared mem) */
641c92ff1bdSMartin Schwidefsky
642d09e8ca6SPasha Tatashin #ifdef CONFIG_ANON_VMA_NAME
643d09e8ca6SPasha Tatashin /*
644d09e8ca6SPasha Tatashin * For private and shared anonymous mappings, a pointer to a null
645d09e8ca6SPasha Tatashin * terminated string containing the name given to the vma, or NULL if
6468651a137SLorenzo Stoakes * unnamed. Serialized by mmap_lock. Use anon_vma_name to access.
647d09e8ca6SPasha Tatashin */
648d09e8ca6SPasha Tatashin struct anon_vma_name *anon_name;
649d09e8ca6SPasha Tatashin #endif
650219f8a2eSAlexey Dobriyan #ifdef CONFIG_SWAP
651ec560175SHuang Ying atomic_long_t swap_readahead_info;
652219f8a2eSAlexey Dobriyan #endif
653c92ff1bdSMartin Schwidefsky #ifndef CONFIG_MMU
6548feae131SDavid Howells struct vm_region *vm_region; /* NOMMU mapping region */
655c92ff1bdSMartin Schwidefsky #endif
656c92ff1bdSMartin Schwidefsky #ifdef CONFIG_NUMA
657c92ff1bdSMartin Schwidefsky struct mempolicy *vm_policy; /* NUMA policy for the VMA */
658c92ff1bdSMartin Schwidefsky #endif
659ef6a22b7SMel Gorman #ifdef CONFIG_NUMA_BALANCING
660ef6a22b7SMel Gorman struct vma_numab_state *numab_state; /* NUMA Balancing state */
661ef6a22b7SMel Gorman #endif
662745f234bSAndrea Arcangeli struct vm_userfaultfd_ctx vm_userfaultfd_ctx;
6633859a271SKees Cook } __randomize_layout;
664c92ff1bdSMartin Schwidefsky
665223baf9dSMathieu Desnoyers #ifdef CONFIG_SCHED_MM_CID
666223baf9dSMathieu Desnoyers struct mm_cid {
667223baf9dSMathieu Desnoyers u64 time;
668223baf9dSMathieu Desnoyers int cid;
669223baf9dSMathieu Desnoyers };
670223baf9dSMathieu Desnoyers #endif
671223baf9dSMathieu Desnoyers
672db446a08SBenjamin LaHaise struct kioctx_table;
673c92ff1bdSMartin Schwidefsky struct mm_struct {
674c1a2f7f0SRik van Riel struct {
675c1753fd0SMathieu Desnoyers /*
676c1753fd0SMathieu Desnoyers * Fields which are often written to are placed in a separate
677c1753fd0SMathieu Desnoyers * cache line.
678c1753fd0SMathieu Desnoyers */
679c1753fd0SMathieu Desnoyers struct {
680c1753fd0SMathieu Desnoyers /**
681c1753fd0SMathieu Desnoyers * @mm_count: The number of references to &struct
682c1753fd0SMathieu Desnoyers * mm_struct (@mm_users count as 1).
683c1753fd0SMathieu Desnoyers *
684c1753fd0SMathieu Desnoyers * Use mmgrab()/mmdrop() to modify. When this drops to
685c1753fd0SMathieu Desnoyers * 0, the &struct mm_struct is freed.
686c1753fd0SMathieu Desnoyers */
687c1753fd0SMathieu Desnoyers atomic_t mm_count;
688c1753fd0SMathieu Desnoyers } ____cacheline_aligned_in_smp;
689c1753fd0SMathieu Desnoyers
690d4af56c5SLiam R. Howlett struct maple_tree mm_mt;
691efc1a3b1SDavid Howells #ifdef CONFIG_MMU
692c92ff1bdSMartin Schwidefsky unsigned long (*get_unmapped_area) (struct file *filp,
693c92ff1bdSMartin Schwidefsky unsigned long addr, unsigned long len,
694c92ff1bdSMartin Schwidefsky unsigned long pgoff, unsigned long flags);
695efc1a3b1SDavid Howells #endif
696c92ff1bdSMartin Schwidefsky unsigned long mmap_base; /* base of mmap area */
69741aacc1eSRadu Caragea unsigned long mmap_legacy_base; /* base of mmap area in bottom-up allocations */
6981b028f78SDmitry Safonov #ifdef CONFIG_HAVE_ARCH_COMPAT_MMAP_BASES
699041711ceSZhen Lei /* Base addresses for compatible mmap() */
7001b028f78SDmitry Safonov unsigned long mmap_compat_base;
7011b028f78SDmitry Safonov unsigned long mmap_compat_legacy_base;
7021b028f78SDmitry Safonov #endif
703c92ff1bdSMartin Schwidefsky unsigned long task_size; /* size of task vm space */
704c92ff1bdSMartin Schwidefsky pgd_t * pgd;
705b279ddc3SVegard Nossum
706227a4aadSMathieu Desnoyers #ifdef CONFIG_MEMBARRIER
707227a4aadSMathieu Desnoyers /**
708227a4aadSMathieu Desnoyers * @membarrier_state: Flags controlling membarrier behavior.
709227a4aadSMathieu Desnoyers *
710227a4aadSMathieu Desnoyers * This field is close to @pgd to hopefully fit in the same
711227a4aadSMathieu Desnoyers * cache-line, which needs to be touched by switch_mm().
712227a4aadSMathieu Desnoyers */
713227a4aadSMathieu Desnoyers atomic_t membarrier_state;
714227a4aadSMathieu Desnoyers #endif
715227a4aadSMathieu Desnoyers
716b279ddc3SVegard Nossum /**
717b279ddc3SVegard Nossum * @mm_users: The number of users including userspace.
718b279ddc3SVegard Nossum *
719c1a2f7f0SRik van Riel * Use mmget()/mmget_not_zero()/mmput() to modify. When this
720c1a2f7f0SRik van Riel * drops to 0 (i.e. when the task exits and there are no other
721c1a2f7f0SRik van Riel * temporary reference holders), we also release a reference on
722c1a2f7f0SRik van Riel * @mm_count (which may then free the &struct mm_struct if
723c1a2f7f0SRik van Riel * @mm_count also drops to 0).
724b279ddc3SVegard Nossum */
725b279ddc3SVegard Nossum atomic_t mm_users;
726b279ddc3SVegard Nossum
727af7f588dSMathieu Desnoyers #ifdef CONFIG_SCHED_MM_CID
728af7f588dSMathieu Desnoyers /**
729223baf9dSMathieu Desnoyers * @pcpu_cid: Per-cpu current cid.
730af7f588dSMathieu Desnoyers *
731223baf9dSMathieu Desnoyers * Keep track of the currently allocated mm_cid for each cpu.
732223baf9dSMathieu Desnoyers * The per-cpu mm_cid values are serialized by their respective
733223baf9dSMathieu Desnoyers * runqueue locks.
734af7f588dSMathieu Desnoyers */
735223baf9dSMathieu Desnoyers struct mm_cid __percpu *pcpu_cid;
736223baf9dSMathieu Desnoyers /*
737223baf9dSMathieu Desnoyers * @mm_cid_next_scan: Next mm_cid scan (in jiffies).
738223baf9dSMathieu Desnoyers *
739223baf9dSMathieu Desnoyers * When the next mm_cid scan is due (in jiffies).
740223baf9dSMathieu Desnoyers */
741223baf9dSMathieu Desnoyers unsigned long mm_cid_next_scan;
742af7f588dSMathieu Desnoyers #endif
743c4812909SKirill A. Shutemov #ifdef CONFIG_MMU
7443783e172SKele Huang atomic_long_t pgtables_bytes; /* size of all page tables */
745b4e98d9aSKirill A. Shutemov #endif
746c92ff1bdSMartin Schwidefsky int map_count; /* number of VMAs */
747481b4bb5SRichard Kennedy
748c1a2f7f0SRik van Riel spinlock_t page_table_lock; /* Protects page tables and some
749c1a2f7f0SRik van Riel * counters
750c1a2f7f0SRik van Riel */
7512e302543SFeng Tang /*
7522e302543SFeng Tang * With some kernel config, the current mmap_lock's offset
7532e302543SFeng Tang * inside 'mm_struct' is at 0x120, which is very optimal, as
7542e302543SFeng Tang * its two hot fields 'count' and 'owner' sit in 2 different
7552e302543SFeng Tang * cachelines, and when mmap_lock is highly contended, both
7562e302543SFeng Tang * of the 2 fields will be accessed frequently, current layout
7572e302543SFeng Tang * will help to reduce cache bouncing.
7582e302543SFeng Tang *
7592e302543SFeng Tang * So please be careful with adding new fields before
7602e302543SFeng Tang * mmap_lock, which can easily push the 2 fields into one
7612e302543SFeng Tang * cacheline.
7622e302543SFeng Tang */
763da1c55f1SMichel Lespinasse struct rw_semaphore mmap_lock;
764c92ff1bdSMartin Schwidefsky
765c1a2f7f0SRik van Riel struct list_head mmlist; /* List of maybe swapped mm's. These
766c1a2f7f0SRik van Riel * are globally strung together off
767c1a2f7f0SRik van Riel * init_mm.mmlist, and are protected
768c92ff1bdSMartin Schwidefsky * by mmlist_lock
769c92ff1bdSMartin Schwidefsky */
7705e31275cSSuren Baghdasaryan #ifdef CONFIG_PER_VMA_LOCK
771b1f02b95SJann Horn /*
772b1f02b95SJann Horn * This field has lock-like semantics, meaning it is sometimes
773b1f02b95SJann Horn * accessed with ACQUIRE/RELEASE semantics.
774b1f02b95SJann Horn * Roughly speaking, incrementing the sequence number is
775b1f02b95SJann Horn * equivalent to releasing locks on VMAs; reading the sequence
776b1f02b95SJann Horn * number can be part of taking a read lock on a VMA.
777b1f02b95SJann Horn *
778b1f02b95SJann Horn * Can be modified under write mmap_lock using RELEASE
779b1f02b95SJann Horn * semantics.
780b1f02b95SJann Horn * Can be read with no other protection when holding write
781b1f02b95SJann Horn * mmap_lock.
782b1f02b95SJann Horn * Can be read with ACQUIRE semantics if not holding write
783b1f02b95SJann Horn * mmap_lock.
784b1f02b95SJann Horn */
7855e31275cSSuren Baghdasaryan int mm_lock_seq;
7865e31275cSSuren Baghdasaryan #endif
787c92ff1bdSMartin Schwidefsky
788c92ff1bdSMartin Schwidefsky
789c92ff1bdSMartin Schwidefsky unsigned long hiwater_rss; /* High-watermark of RSS usage */
790c92ff1bdSMartin Schwidefsky unsigned long hiwater_vm; /* High-water virtual memory usage */
791c92ff1bdSMartin Schwidefsky
792e10d59f2SChristoph Lameter unsigned long total_vm; /* Total pages mapped */
793e10d59f2SChristoph Lameter unsigned long locked_vm; /* Pages that have PG_mlocked set */
79470f8a3caSDavidlohr Bueso atomic64_t pinned_vm; /* Refcount permanently increased */
79530bdbb78SKonstantin Khlebnikov unsigned long data_vm; /* VM_WRITE & ~VM_SHARED & ~VM_STACK */
79630bdbb78SKonstantin Khlebnikov unsigned long exec_vm; /* VM_EXEC & ~VM_WRITE & ~VM_STACK */
79730bdbb78SKonstantin Khlebnikov unsigned long stack_vm; /* VM_STACK */
798e10d59f2SChristoph Lameter unsigned long def_flags;
79988aa7cc6SYang Shi
8002e302543SFeng Tang /**
8012e302543SFeng Tang * @write_protect_seq: Locked when any thread is write
8022e302543SFeng Tang * protecting pages mapped by this mm to enforce a later COW,
8032e302543SFeng Tang * for instance during page table copying for fork().
8042e302543SFeng Tang */
8052e302543SFeng Tang seqcount_t write_protect_seq;
8062e302543SFeng Tang
80788aa7cc6SYang Shi spinlock_t arg_lock; /* protect the below fields */
8082e302543SFeng Tang
809c92ff1bdSMartin Schwidefsky unsigned long start_code, end_code, start_data, end_data;
810c92ff1bdSMartin Schwidefsky unsigned long start_brk, brk, start_stack;
811c92ff1bdSMartin Schwidefsky unsigned long arg_start, arg_end, env_start, env_end;
812c92ff1bdSMartin Schwidefsky
813c92ff1bdSMartin Schwidefsky unsigned long saved_auxv[AT_VECTOR_SIZE]; /* for /proc/PID/auxv */
814c92ff1bdSMartin Schwidefsky
815f1a79412SShakeel Butt struct percpu_counter rss_stat[NR_MM_COUNTERS];
816d559db08SKAMEZAWA Hiroyuki
817801460d0SHiroshi Shimamoto struct linux_binfmt *binfmt;
818801460d0SHiroshi Shimamoto
819c92ff1bdSMartin Schwidefsky /* Architecture-specific MM context */
820c92ff1bdSMartin Schwidefsky mm_context_t context;
821c92ff1bdSMartin Schwidefsky
822c1a2f7f0SRik van Riel unsigned long flags; /* Must use atomic bitops to access */
823c92ff1bdSMartin Schwidefsky
824858f0993SAlexey Dobriyan #ifdef CONFIG_AIO
825abf137ddSJens Axboe spinlock_t ioctx_lock;
826db446a08SBenjamin LaHaise struct kioctx_table __rcu *ioctx_table;
827858f0993SAlexey Dobriyan #endif
828f98bafa0SOleg Nesterov #ifdef CONFIG_MEMCG
8294cd1a8fcSKOSAKI Motohiro /*
8304cd1a8fcSKOSAKI Motohiro * "owner" points to a task that is regarded as the canonical
8314cd1a8fcSKOSAKI Motohiro * user/owner of this mm. All of the following must be true in
8324cd1a8fcSKOSAKI Motohiro * order for it to be changed:
8334cd1a8fcSKOSAKI Motohiro *
8344cd1a8fcSKOSAKI Motohiro * current == mm->owner
8354cd1a8fcSKOSAKI Motohiro * current->mm != mm
8364cd1a8fcSKOSAKI Motohiro * new_owner->mm == mm
8374cd1a8fcSKOSAKI Motohiro * new_owner->alloc_lock is held
8384cd1a8fcSKOSAKI Motohiro */
8394d2deb40SArnd Bergmann struct task_struct __rcu *owner;
84078fb7466SPavel Emelianov #endif
841bfedb589SEric W. Biederman struct user_namespace *user_ns;
842925d1c40SMatt Helsley
843925d1c40SMatt Helsley /* store ref to file /proc/<pid>/exe symlink points to */
84490f31d0eSKonstantin Khlebnikov struct file __rcu *exe_file;
845cddb8a5cSAndrea Arcangeli #ifdef CONFIG_MMU_NOTIFIER
846984cfe4eSJason Gunthorpe struct mmu_notifier_subscriptions *notifier_subscriptions;
847cddb8a5cSAndrea Arcangeli #endif
848e009bb30SKirill A. Shutemov #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS
849e7a00c45SAndrea Arcangeli pgtable_t pmd_huge_pte; /* protected by page_table_lock */
850e7a00c45SAndrea Arcangeli #endif
851cbee9f88SPeter Zijlstra #ifdef CONFIG_NUMA_BALANCING
852cbee9f88SPeter Zijlstra /*
8537014887aSDavid Hildenbrand * numa_next_scan is the next time that PTEs will be remapped
8547014887aSDavid Hildenbrand * PROT_NONE to trigger NUMA hinting faults; such faults gather
8557014887aSDavid Hildenbrand * statistics and migrate pages to new nodes if necessary.
856cbee9f88SPeter Zijlstra */
857cbee9f88SPeter Zijlstra unsigned long numa_next_scan;
858cbee9f88SPeter Zijlstra
8597014887aSDavid Hildenbrand /* Restart point for scanning and remapping PTEs. */
8606e5fb223SPeter Zijlstra unsigned long numa_scan_offset;
8616e5fb223SPeter Zijlstra
8627014887aSDavid Hildenbrand /* numa_scan_seq prevents two threads remapping PTEs. */
863cbee9f88SPeter Zijlstra int numa_scan_seq;
864cbee9f88SPeter Zijlstra #endif
86520841405SRik van Riel /*
866c1a2f7f0SRik van Riel * An operation with batched TLB flushing is going on. Anything
867c1a2f7f0SRik van Riel * that can move process memory needs to flush the TLB when
8687014887aSDavid Hildenbrand * moving a PROT_NONE mapped page.
86920841405SRik van Riel */
87016af97dcSNadav Amit atomic_t tlb_flush_pending;
8713ea27719SMel Gorman #ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH
8723ea27719SMel Gorman /* See flush_tlb_batched_pending() */
8735ee2fa2fSHuang Ying atomic_t tlb_flush_batched;
8743ea27719SMel Gorman #endif
875d4b3b638SSrikar Dronamraju struct uprobes_state uprobes_state;
8768d491de6SThomas Gleixner #ifdef CONFIG_PREEMPT_RT
8778d491de6SThomas Gleixner struct rcu_head delayed_drop;
8788d491de6SThomas Gleixner #endif
8795d317b2bSNaoya Horiguchi #ifdef CONFIG_HUGETLB_PAGE
8805d317b2bSNaoya Horiguchi atomic_long_t hugetlb_usage;
8815d317b2bSNaoya Horiguchi #endif
882ec8d7c14SMichal Hocko struct work_struct async_put_work;
88352ad9bc6SFenghua Yu
8847a853c2dSFenghua Yu #ifdef CONFIG_IOMMU_SVA
88552ad9bc6SFenghua Yu u32 pasid;
88652ad9bc6SFenghua Yu #endif
88776093853Sxu xin #ifdef CONFIG_KSM
88876093853Sxu xin /*
88976093853Sxu xin * Represent how many pages of this process are involved in KSM
8906080d19fSxu xin * merging (not including ksm_zero_pages).
89176093853Sxu xin */
89276093853Sxu xin unsigned long ksm_merging_pages;
893cb4df4caSxu xin /*
894cb4df4caSxu xin * Represent how many pages are checked for ksm merging
895cb4df4caSxu xin * including merged and not merged.
896cb4df4caSxu xin */
897cb4df4caSxu xin unsigned long ksm_rmap_items;
8986080d19fSxu xin /*
8996080d19fSxu xin * Represent how many empty pages are merged with kernel zero
9006080d19fSxu xin * pages when enabling KSM use_zero_pages.
9016080d19fSxu xin */
9026080d19fSxu xin unsigned long ksm_zero_pages;
9036080d19fSxu xin #endif /* CONFIG_KSM */
904bd74fdaeSYu Zhao #ifdef CONFIG_LRU_GEN
905bd74fdaeSYu Zhao struct {
906bd74fdaeSYu Zhao /* this mm_struct is on lru_gen_mm_list */
907bd74fdaeSYu Zhao struct list_head list;
908bd74fdaeSYu Zhao /*
909bd74fdaeSYu Zhao * Set when switching to this mm_struct, as a hint of
910bd74fdaeSYu Zhao * whether it has been used since the last time per-node
911bd74fdaeSYu Zhao * page table walkers cleared the corresponding bits.
912bd74fdaeSYu Zhao */
913bd74fdaeSYu Zhao unsigned long bitmap;
914bd74fdaeSYu Zhao #ifdef CONFIG_MEMCG
915bd74fdaeSYu Zhao /* points to the memcg of "owner" above */
916bd74fdaeSYu Zhao struct mem_cgroup *memcg;
917bd74fdaeSYu Zhao #endif
918bd74fdaeSYu Zhao } lru_gen;
919bd74fdaeSYu Zhao #endif /* CONFIG_LRU_GEN */
9203859a271SKees Cook } __randomize_layout;
921c92ff1bdSMartin Schwidefsky
922c1a2f7f0SRik van Riel /*
923c1a2f7f0SRik van Riel * The mm_cpumask needs to be at the end of mm_struct, because it
924c1a2f7f0SRik van Riel * is dynamically sized based on nr_cpu_ids.
925c1a2f7f0SRik van Riel */
926c1a2f7f0SRik van Riel unsigned long cpu_bitmap[];
927c1a2f7f0SRik van Riel };
928c1a2f7f0SRik van Riel
9293dd44325SLiam R. Howlett #define MM_MT_FLAGS (MT_FLAGS_ALLOC_RANGE | MT_FLAGS_LOCK_EXTERN | \
9303dd44325SLiam R. Howlett MT_FLAGS_USE_RCU)
931abe722a1SIngo Molnar extern struct mm_struct init_mm;
932abe722a1SIngo Molnar
933c1a2f7f0SRik van Riel /* Pointer magic because the dynamic array size confuses some compilers. */
mm_init_cpumask(struct mm_struct * mm)9346345d24dSLinus Torvalds static inline void mm_init_cpumask(struct mm_struct *mm)
9356345d24dSLinus Torvalds {
936c1a2f7f0SRik van Riel unsigned long cpu_bitmap = (unsigned long)mm;
937c1a2f7f0SRik van Riel
938c1a2f7f0SRik van Riel cpu_bitmap += offsetof(struct mm_struct, cpu_bitmap);
939c1a2f7f0SRik van Riel cpumask_clear((struct cpumask *)cpu_bitmap);
9406345d24dSLinus Torvalds }
9416345d24dSLinus Torvalds
94245e575abSRusty Russell /* Future-safe accessor for struct mm_struct's cpu_vm_mask. */
mm_cpumask(struct mm_struct * mm)943de03c72cSKOSAKI Motohiro static inline cpumask_t *mm_cpumask(struct mm_struct *mm)
944de03c72cSKOSAKI Motohiro {
945c1a2f7f0SRik van Riel return (struct cpumask *)&mm->cpu_bitmap;
946de03c72cSKOSAKI Motohiro }
94745e575abSRusty Russell
948bd74fdaeSYu Zhao #ifdef CONFIG_LRU_GEN
949bd74fdaeSYu Zhao
950bd74fdaeSYu Zhao struct lru_gen_mm_list {
951bd74fdaeSYu Zhao /* mm_struct list for page table walkers */
952bd74fdaeSYu Zhao struct list_head fifo;
953bd74fdaeSYu Zhao /* protects the list above */
954bd74fdaeSYu Zhao spinlock_t lock;
955bd74fdaeSYu Zhao };
956bd74fdaeSYu Zhao
957bd74fdaeSYu Zhao void lru_gen_add_mm(struct mm_struct *mm);
958bd74fdaeSYu Zhao void lru_gen_del_mm(struct mm_struct *mm);
959bd74fdaeSYu Zhao #ifdef CONFIG_MEMCG
960bd74fdaeSYu Zhao void lru_gen_migrate_mm(struct mm_struct *mm);
961bd74fdaeSYu Zhao #endif
962bd74fdaeSYu Zhao
lru_gen_init_mm(struct mm_struct * mm)963bd74fdaeSYu Zhao static inline void lru_gen_init_mm(struct mm_struct *mm)
964bd74fdaeSYu Zhao {
965bd74fdaeSYu Zhao INIT_LIST_HEAD(&mm->lru_gen.list);
966bd74fdaeSYu Zhao mm->lru_gen.bitmap = 0;
967bd74fdaeSYu Zhao #ifdef CONFIG_MEMCG
968bd74fdaeSYu Zhao mm->lru_gen.memcg = NULL;
969bd74fdaeSYu Zhao #endif
970bd74fdaeSYu Zhao }
971bd74fdaeSYu Zhao
lru_gen_use_mm(struct mm_struct * mm)972bd74fdaeSYu Zhao static inline void lru_gen_use_mm(struct mm_struct *mm)
973bd74fdaeSYu Zhao {
974bd74fdaeSYu Zhao /*
975bd74fdaeSYu Zhao * When the bitmap is set, page reclaim knows this mm_struct has been
976bd74fdaeSYu Zhao * used since the last time it cleared the bitmap. So it might be worth
977bd74fdaeSYu Zhao * walking the page tables of this mm_struct to clear the accessed bit.
978bd74fdaeSYu Zhao */
979bd74fdaeSYu Zhao WRITE_ONCE(mm->lru_gen.bitmap, -1);
980bd74fdaeSYu Zhao }
981bd74fdaeSYu Zhao
982bd74fdaeSYu Zhao #else /* !CONFIG_LRU_GEN */
983bd74fdaeSYu Zhao
lru_gen_add_mm(struct mm_struct * mm)984bd74fdaeSYu Zhao static inline void lru_gen_add_mm(struct mm_struct *mm)
985bd74fdaeSYu Zhao {
986bd74fdaeSYu Zhao }
987bd74fdaeSYu Zhao
lru_gen_del_mm(struct mm_struct * mm)988bd74fdaeSYu Zhao static inline void lru_gen_del_mm(struct mm_struct *mm)
989bd74fdaeSYu Zhao {
990bd74fdaeSYu Zhao }
991bd74fdaeSYu Zhao
992bd74fdaeSYu Zhao #ifdef CONFIG_MEMCG
lru_gen_migrate_mm(struct mm_struct * mm)993bd74fdaeSYu Zhao static inline void lru_gen_migrate_mm(struct mm_struct *mm)
994bd74fdaeSYu Zhao {
995bd74fdaeSYu Zhao }
996bd74fdaeSYu Zhao #endif
997bd74fdaeSYu Zhao
lru_gen_init_mm(struct mm_struct * mm)998bd74fdaeSYu Zhao static inline void lru_gen_init_mm(struct mm_struct *mm)
999bd74fdaeSYu Zhao {
1000bd74fdaeSYu Zhao }
1001bd74fdaeSYu Zhao
lru_gen_use_mm(struct mm_struct * mm)1002bd74fdaeSYu Zhao static inline void lru_gen_use_mm(struct mm_struct *mm)
1003bd74fdaeSYu Zhao {
1004bd74fdaeSYu Zhao }
1005bd74fdaeSYu Zhao
1006bd74fdaeSYu Zhao #endif /* CONFIG_LRU_GEN */
1007bd74fdaeSYu Zhao
1008f39af059SMatthew Wilcox (Oracle) struct vma_iterator {
1009f39af059SMatthew Wilcox (Oracle) struct ma_state mas;
1010f39af059SMatthew Wilcox (Oracle) };
1011f39af059SMatthew Wilcox (Oracle)
1012f39af059SMatthew Wilcox (Oracle) #define VMA_ITERATOR(name, __mm, __addr) \
1013f39af059SMatthew Wilcox (Oracle) struct vma_iterator name = { \
1014f39af059SMatthew Wilcox (Oracle) .mas = { \
1015f39af059SMatthew Wilcox (Oracle) .tree = &(__mm)->mm_mt, \
1016f39af059SMatthew Wilcox (Oracle) .index = __addr, \
1017f39af059SMatthew Wilcox (Oracle) .node = MAS_START, \
1018f39af059SMatthew Wilcox (Oracle) }, \
1019f39af059SMatthew Wilcox (Oracle) }
1020f39af059SMatthew Wilcox (Oracle)
vma_iter_init(struct vma_iterator * vmi,struct mm_struct * mm,unsigned long addr)1021f39af059SMatthew Wilcox (Oracle) static inline void vma_iter_init(struct vma_iterator *vmi,
1022f39af059SMatthew Wilcox (Oracle) struct mm_struct *mm, unsigned long addr)
1023f39af059SMatthew Wilcox (Oracle) {
1024b62b633eSLiam R. Howlett mas_init(&vmi->mas, &mm->mm_mt, addr);
1025f39af059SMatthew Wilcox (Oracle) }
1026f39af059SMatthew Wilcox (Oracle)
1027af7f588dSMathieu Desnoyers #ifdef CONFIG_SCHED_MM_CID
1028223baf9dSMathieu Desnoyers
1029223baf9dSMathieu Desnoyers enum mm_cid_state {
1030223baf9dSMathieu Desnoyers MM_CID_UNSET = -1U, /* Unset state has lazy_put flag set. */
1031223baf9dSMathieu Desnoyers MM_CID_LAZY_PUT = (1U << 31),
1032223baf9dSMathieu Desnoyers };
1033223baf9dSMathieu Desnoyers
mm_cid_is_unset(int cid)1034223baf9dSMathieu Desnoyers static inline bool mm_cid_is_unset(int cid)
1035223baf9dSMathieu Desnoyers {
1036223baf9dSMathieu Desnoyers return cid == MM_CID_UNSET;
1037223baf9dSMathieu Desnoyers }
1038223baf9dSMathieu Desnoyers
mm_cid_is_lazy_put(int cid)1039223baf9dSMathieu Desnoyers static inline bool mm_cid_is_lazy_put(int cid)
1040223baf9dSMathieu Desnoyers {
1041223baf9dSMathieu Desnoyers return !mm_cid_is_unset(cid) && (cid & MM_CID_LAZY_PUT);
1042223baf9dSMathieu Desnoyers }
1043223baf9dSMathieu Desnoyers
mm_cid_is_valid(int cid)1044223baf9dSMathieu Desnoyers static inline bool mm_cid_is_valid(int cid)
1045223baf9dSMathieu Desnoyers {
1046223baf9dSMathieu Desnoyers return !(cid & MM_CID_LAZY_PUT);
1047223baf9dSMathieu Desnoyers }
1048223baf9dSMathieu Desnoyers
mm_cid_set_lazy_put(int cid)1049223baf9dSMathieu Desnoyers static inline int mm_cid_set_lazy_put(int cid)
1050223baf9dSMathieu Desnoyers {
1051223baf9dSMathieu Desnoyers return cid | MM_CID_LAZY_PUT;
1052223baf9dSMathieu Desnoyers }
1053223baf9dSMathieu Desnoyers
mm_cid_clear_lazy_put(int cid)1054223baf9dSMathieu Desnoyers static inline int mm_cid_clear_lazy_put(int cid)
1055223baf9dSMathieu Desnoyers {
1056223baf9dSMathieu Desnoyers return cid & ~MM_CID_LAZY_PUT;
1057223baf9dSMathieu Desnoyers }
1058223baf9dSMathieu Desnoyers
1059af7f588dSMathieu Desnoyers /* Accessor for struct mm_struct's cidmask. */
mm_cidmask(struct mm_struct * mm)1060af7f588dSMathieu Desnoyers static inline cpumask_t *mm_cidmask(struct mm_struct *mm)
1061af7f588dSMathieu Desnoyers {
1062af7f588dSMathieu Desnoyers unsigned long cid_bitmap = (unsigned long)mm;
1063af7f588dSMathieu Desnoyers
1064af7f588dSMathieu Desnoyers cid_bitmap += offsetof(struct mm_struct, cpu_bitmap);
1065af7f588dSMathieu Desnoyers /* Skip cpu_bitmap */
1066af7f588dSMathieu Desnoyers cid_bitmap += cpumask_size();
1067af7f588dSMathieu Desnoyers return (struct cpumask *)cid_bitmap;
1068af7f588dSMathieu Desnoyers }
1069af7f588dSMathieu Desnoyers
mm_init_cid(struct mm_struct * mm)1070af7f588dSMathieu Desnoyers static inline void mm_init_cid(struct mm_struct *mm)
1071af7f588dSMathieu Desnoyers {
1072223baf9dSMathieu Desnoyers int i;
1073223baf9dSMathieu Desnoyers
1074223baf9dSMathieu Desnoyers for_each_possible_cpu(i) {
1075223baf9dSMathieu Desnoyers struct mm_cid *pcpu_cid = per_cpu_ptr(mm->pcpu_cid, i);
1076223baf9dSMathieu Desnoyers
1077223baf9dSMathieu Desnoyers pcpu_cid->cid = MM_CID_UNSET;
1078223baf9dSMathieu Desnoyers pcpu_cid->time = 0;
1079223baf9dSMathieu Desnoyers }
1080af7f588dSMathieu Desnoyers cpumask_clear(mm_cidmask(mm));
1081af7f588dSMathieu Desnoyers }
1082af7f588dSMathieu Desnoyers
mm_alloc_cid(struct mm_struct * mm)1083223baf9dSMathieu Desnoyers static inline int mm_alloc_cid(struct mm_struct *mm)
1084223baf9dSMathieu Desnoyers {
1085223baf9dSMathieu Desnoyers mm->pcpu_cid = alloc_percpu(struct mm_cid);
1086223baf9dSMathieu Desnoyers if (!mm->pcpu_cid)
1087223baf9dSMathieu Desnoyers return -ENOMEM;
1088223baf9dSMathieu Desnoyers mm_init_cid(mm);
1089223baf9dSMathieu Desnoyers return 0;
1090223baf9dSMathieu Desnoyers }
1091223baf9dSMathieu Desnoyers
mm_destroy_cid(struct mm_struct * mm)1092223baf9dSMathieu Desnoyers static inline void mm_destroy_cid(struct mm_struct *mm)
1093223baf9dSMathieu Desnoyers {
1094223baf9dSMathieu Desnoyers free_percpu(mm->pcpu_cid);
1095223baf9dSMathieu Desnoyers mm->pcpu_cid = NULL;
1096223baf9dSMathieu Desnoyers }
1097223baf9dSMathieu Desnoyers
mm_cid_size(void)1098af7f588dSMathieu Desnoyers static inline unsigned int mm_cid_size(void)
1099af7f588dSMathieu Desnoyers {
1100af7f588dSMathieu Desnoyers return cpumask_size();
1101af7f588dSMathieu Desnoyers }
1102af7f588dSMathieu Desnoyers #else /* CONFIG_SCHED_MM_CID */
mm_init_cid(struct mm_struct * mm)1103af7f588dSMathieu Desnoyers static inline void mm_init_cid(struct mm_struct *mm) { }
mm_alloc_cid(struct mm_struct * mm)1104223baf9dSMathieu Desnoyers static inline int mm_alloc_cid(struct mm_struct *mm) { return 0; }
mm_destroy_cid(struct mm_struct * mm)1105223baf9dSMathieu Desnoyers static inline void mm_destroy_cid(struct mm_struct *mm) { }
mm_cid_size(void)1106af7f588dSMathieu Desnoyers static inline unsigned int mm_cid_size(void)
1107af7f588dSMathieu Desnoyers {
1108af7f588dSMathieu Desnoyers return 0;
1109af7f588dSMathieu Desnoyers }
1110af7f588dSMathieu Desnoyers #endif /* CONFIG_SCHED_MM_CID */
1111af7f588dSMathieu Desnoyers
111256236a59SMinchan Kim struct mmu_gather;
1113a72afd87SWill Deacon extern void tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm);
1114d8b45053SWill Deacon extern void tlb_gather_mmu_fullmm(struct mmu_gather *tlb, struct mm_struct *mm);
1115ae8eba8bSWill Deacon extern void tlb_finish_mmu(struct mmu_gather *tlb);
111656236a59SMinchan Kim
1117f872f540SAndy Lutomirski struct vm_fault;
1118f872f540SAndy Lutomirski
11193d353901SSouptick Joarder /**
11203d353901SSouptick Joarder * typedef vm_fault_t - Return type for page fault handlers.
11213d353901SSouptick Joarder *
11223d353901SSouptick Joarder * Page fault handlers return a bitmask of %VM_FAULT values.
11233d353901SSouptick Joarder */
11243d353901SSouptick Joarder typedef __bitwise unsigned int vm_fault_t;
11253d353901SSouptick Joarder
11263d353901SSouptick Joarder /**
11273d353901SSouptick Joarder * enum vm_fault_reason - Page fault handlers return a bitmask of
11283d353901SSouptick Joarder * these values to tell the core VM what happened when handling the
11293d353901SSouptick Joarder * fault. Used to decide whether a process gets delivered SIGBUS or
11303d353901SSouptick Joarder * just gets major/minor fault counters bumped up.
11313d353901SSouptick Joarder *
11323d353901SSouptick Joarder * @VM_FAULT_OOM: Out Of Memory
11333d353901SSouptick Joarder * @VM_FAULT_SIGBUS: Bad access
11343d353901SSouptick Joarder * @VM_FAULT_MAJOR: Page read from storage
11353d353901SSouptick Joarder * @VM_FAULT_HWPOISON: Hit poisoned small page
11363d353901SSouptick Joarder * @VM_FAULT_HWPOISON_LARGE: Hit poisoned large page. Index encoded
11373d353901SSouptick Joarder * in upper bits
11383d353901SSouptick Joarder * @VM_FAULT_SIGSEGV: segmentation fault
11393d353901SSouptick Joarder * @VM_FAULT_NOPAGE: ->fault installed the pte, not return page
11403d353901SSouptick Joarder * @VM_FAULT_LOCKED: ->fault locked the returned page
11413d353901SSouptick Joarder * @VM_FAULT_RETRY: ->fault blocked, must retry
11423d353901SSouptick Joarder * @VM_FAULT_FALLBACK: huge page fault failed, fall back to small
11433d353901SSouptick Joarder * @VM_FAULT_DONE_COW: ->fault has fully handled COW
11443d353901SSouptick Joarder * @VM_FAULT_NEEDDSYNC: ->fault did not modify page tables and needs
11453d353901SSouptick Joarder * fsync() to complete (for synchronous page faults
11463d353901SSouptick Joarder * in DAX)
1147d9272525SPeter Xu * @VM_FAULT_COMPLETED: ->fault completed, meanwhile mmap lock released
11483d353901SSouptick Joarder * @VM_FAULT_HINDEX_MASK: mask HINDEX value
11493d353901SSouptick Joarder *
11503d353901SSouptick Joarder */
11513d353901SSouptick Joarder enum vm_fault_reason {
11523d353901SSouptick Joarder VM_FAULT_OOM = (__force vm_fault_t)0x000001,
11533d353901SSouptick Joarder VM_FAULT_SIGBUS = (__force vm_fault_t)0x000002,
11543d353901SSouptick Joarder VM_FAULT_MAJOR = (__force vm_fault_t)0x000004,
11553d353901SSouptick Joarder VM_FAULT_HWPOISON = (__force vm_fault_t)0x000010,
11563d353901SSouptick Joarder VM_FAULT_HWPOISON_LARGE = (__force vm_fault_t)0x000020,
11573d353901SSouptick Joarder VM_FAULT_SIGSEGV = (__force vm_fault_t)0x000040,
11583d353901SSouptick Joarder VM_FAULT_NOPAGE = (__force vm_fault_t)0x000100,
11593d353901SSouptick Joarder VM_FAULT_LOCKED = (__force vm_fault_t)0x000200,
11603d353901SSouptick Joarder VM_FAULT_RETRY = (__force vm_fault_t)0x000400,
11613d353901SSouptick Joarder VM_FAULT_FALLBACK = (__force vm_fault_t)0x000800,
11623d353901SSouptick Joarder VM_FAULT_DONE_COW = (__force vm_fault_t)0x001000,
11633d353901SSouptick Joarder VM_FAULT_NEEDDSYNC = (__force vm_fault_t)0x002000,
1164d9272525SPeter Xu VM_FAULT_COMPLETED = (__force vm_fault_t)0x004000,
11653d353901SSouptick Joarder VM_FAULT_HINDEX_MASK = (__force vm_fault_t)0x0f0000,
11663d353901SSouptick Joarder };
11673d353901SSouptick Joarder
11683d353901SSouptick Joarder /* Encode hstate index for a hwpoisoned large page */
11693d353901SSouptick Joarder #define VM_FAULT_SET_HINDEX(x) ((__force vm_fault_t)((x) << 16))
1170fcae96ffSJann Horn #define VM_FAULT_GET_HINDEX(x) (((__force unsigned int)(x) >> 16) & 0xf)
11713d353901SSouptick Joarder
11723d353901SSouptick Joarder #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | \
11733d353901SSouptick Joarder VM_FAULT_SIGSEGV | VM_FAULT_HWPOISON | \
11743d353901SSouptick Joarder VM_FAULT_HWPOISON_LARGE | VM_FAULT_FALLBACK)
11753d353901SSouptick Joarder
11763d353901SSouptick Joarder #define VM_FAULT_RESULT_TRACE \
11773d353901SSouptick Joarder { VM_FAULT_OOM, "OOM" }, \
11783d353901SSouptick Joarder { VM_FAULT_SIGBUS, "SIGBUS" }, \
11793d353901SSouptick Joarder { VM_FAULT_MAJOR, "MAJOR" }, \
11803d353901SSouptick Joarder { VM_FAULT_HWPOISON, "HWPOISON" }, \
11813d353901SSouptick Joarder { VM_FAULT_HWPOISON_LARGE, "HWPOISON_LARGE" }, \
11823d353901SSouptick Joarder { VM_FAULT_SIGSEGV, "SIGSEGV" }, \
11833d353901SSouptick Joarder { VM_FAULT_NOPAGE, "NOPAGE" }, \
11843d353901SSouptick Joarder { VM_FAULT_LOCKED, "LOCKED" }, \
11853d353901SSouptick Joarder { VM_FAULT_RETRY, "RETRY" }, \
11863d353901SSouptick Joarder { VM_FAULT_FALLBACK, "FALLBACK" }, \
11873d353901SSouptick Joarder { VM_FAULT_DONE_COW, "DONE_COW" }, \
11887a32b58bSSuren Baghdasaryan { VM_FAULT_NEEDDSYNC, "NEEDDSYNC" }, \
11897a32b58bSSuren Baghdasaryan { VM_FAULT_COMPLETED, "COMPLETED" }
11903d353901SSouptick Joarder
1191f872f540SAndy Lutomirski struct vm_special_mapping {
1192f872f540SAndy Lutomirski const char *name; /* The name, e.g. "[vdso]". */
1193f872f540SAndy Lutomirski
1194f872f540SAndy Lutomirski /*
1195f872f540SAndy Lutomirski * If .fault is not provided, this points to a
1196f872f540SAndy Lutomirski * NULL-terminated array of pages that back the special mapping.
1197f872f540SAndy Lutomirski *
1198f872f540SAndy Lutomirski * This must not be NULL unless .fault is provided.
1199f872f540SAndy Lutomirski */
1200a62c34bdSAndy Lutomirski struct page **pages;
1201f872f540SAndy Lutomirski
1202f872f540SAndy Lutomirski /*
1203f872f540SAndy Lutomirski * If non-NULL, then this is called to resolve page faults
1204f872f540SAndy Lutomirski * on the special mapping. If used, .pages is not checked.
1205f872f540SAndy Lutomirski */
1206b3ec9f33SSouptick Joarder vm_fault_t (*fault)(const struct vm_special_mapping *sm,
1207f872f540SAndy Lutomirski struct vm_area_struct *vma,
1208f872f540SAndy Lutomirski struct vm_fault *vmf);
1209b059a453SDmitry Safonov
1210b059a453SDmitry Safonov int (*mremap)(const struct vm_special_mapping *sm,
1211b059a453SDmitry Safonov struct vm_area_struct *new_vma);
1212a62c34bdSAndy Lutomirski };
1213a62c34bdSAndy Lutomirski
1214d17d8f9dSDave Hansen enum tlb_flush_reason {
1215d17d8f9dSDave Hansen TLB_FLUSH_ON_TASK_SWITCH,
1216d17d8f9dSDave Hansen TLB_REMOTE_SHOOTDOWN,
1217d17d8f9dSDave Hansen TLB_LOCAL_SHOOTDOWN,
1218d17d8f9dSDave Hansen TLB_LOCAL_MM_SHOOTDOWN,
12195b74283aSMel Gorman TLB_REMOTE_SEND_IPI,
1220d17d8f9dSDave Hansen NR_TLB_FLUSH_REASONS,
1221d17d8f9dSDave Hansen };
1222d17d8f9dSDave Hansen
122336090defSArnd Bergmann /**
122436090defSArnd Bergmann * enum fault_flag - Fault flag definitions.
122536090defSArnd Bergmann * @FAULT_FLAG_WRITE: Fault was a write fault.
122636090defSArnd Bergmann * @FAULT_FLAG_MKWRITE: Fault was mkwrite of existing PTE.
122736090defSArnd Bergmann * @FAULT_FLAG_ALLOW_RETRY: Allow to retry the fault if blocked.
122836090defSArnd Bergmann * @FAULT_FLAG_RETRY_NOWAIT: Don't drop mmap_lock and wait when retrying.
122936090defSArnd Bergmann * @FAULT_FLAG_KILLABLE: The fault task is in SIGKILL killable region.
123036090defSArnd Bergmann * @FAULT_FLAG_TRIED: The fault has been tried once.
123136090defSArnd Bergmann * @FAULT_FLAG_USER: The fault originated in userspace.
123236090defSArnd Bergmann * @FAULT_FLAG_REMOTE: The fault is not for current task/mm.
123336090defSArnd Bergmann * @FAULT_FLAG_INSTRUCTION: The fault was during an instruction fetch.
123436090defSArnd Bergmann * @FAULT_FLAG_INTERRUPTIBLE: The fault can be interrupted by non-fatal signals.
12358d6a0ac0SDavid Hildenbrand * @FAULT_FLAG_UNSHARE: The fault is an unsharing request to break COW in a
12368d6a0ac0SDavid Hildenbrand * COW mapping, making sure that an exclusive anon page is
12378d6a0ac0SDavid Hildenbrand * mapped after the fault.
1238f46f2adeSPeter Xu * @FAULT_FLAG_ORIG_PTE_VALID: whether the fault has vmf->orig_pte cached.
1239f46f2adeSPeter Xu * We should only access orig_pte if this flag set.
124055324e46SSuren Baghdasaryan * @FAULT_FLAG_VMA_LOCK: The fault is handled under VMA lock.
124136090defSArnd Bergmann *
124236090defSArnd Bergmann * About @FAULT_FLAG_ALLOW_RETRY and @FAULT_FLAG_TRIED: we can specify
124336090defSArnd Bergmann * whether we would allow page faults to retry by specifying these two
124436090defSArnd Bergmann * fault flags correctly. Currently there can be three legal combinations:
124536090defSArnd Bergmann *
124636090defSArnd Bergmann * (a) ALLOW_RETRY and !TRIED: this means the page fault allows retry, and
124736090defSArnd Bergmann * this is the first try
124836090defSArnd Bergmann *
124936090defSArnd Bergmann * (b) ALLOW_RETRY and TRIED: this means the page fault allows retry, and
125036090defSArnd Bergmann * we've already tried at least once
125136090defSArnd Bergmann *
125236090defSArnd Bergmann * (c) !ALLOW_RETRY and !TRIED: this means the page fault does not allow retry
125336090defSArnd Bergmann *
125436090defSArnd Bergmann * The unlisted combination (!ALLOW_RETRY && TRIED) is illegal and should never
125536090defSArnd Bergmann * be used. Note that page faults can be allowed to retry for multiple times,
125636090defSArnd Bergmann * in which case we'll have an initial fault with flags (a) then later on
125736090defSArnd Bergmann * continuous faults with flags (b). We should always try to detect pending
125836090defSArnd Bergmann * signals before a retry to make sure the continuous page faults can still be
125936090defSArnd Bergmann * interrupted if necessary.
1260c89357e2SDavid Hildenbrand *
1261c89357e2SDavid Hildenbrand * The combination FAULT_FLAG_WRITE|FAULT_FLAG_UNSHARE is illegal.
1262c89357e2SDavid Hildenbrand * FAULT_FLAG_UNSHARE is ignored and treated like an ordinary read fault when
12638d6a0ac0SDavid Hildenbrand * applied to mappings that are not COW mappings.
126436090defSArnd Bergmann */
126536090defSArnd Bergmann enum fault_flag {
126636090defSArnd Bergmann FAULT_FLAG_WRITE = 1 << 0,
126736090defSArnd Bergmann FAULT_FLAG_MKWRITE = 1 << 1,
126836090defSArnd Bergmann FAULT_FLAG_ALLOW_RETRY = 1 << 2,
126936090defSArnd Bergmann FAULT_FLAG_RETRY_NOWAIT = 1 << 3,
127036090defSArnd Bergmann FAULT_FLAG_KILLABLE = 1 << 4,
127136090defSArnd Bergmann FAULT_FLAG_TRIED = 1 << 5,
127236090defSArnd Bergmann FAULT_FLAG_USER = 1 << 6,
127336090defSArnd Bergmann FAULT_FLAG_REMOTE = 1 << 7,
127436090defSArnd Bergmann FAULT_FLAG_INSTRUCTION = 1 << 8,
127536090defSArnd Bergmann FAULT_FLAG_INTERRUPTIBLE = 1 << 9,
1276c89357e2SDavid Hildenbrand FAULT_FLAG_UNSHARE = 1 << 10,
1277f46f2adeSPeter Xu FAULT_FLAG_ORIG_PTE_VALID = 1 << 11,
127855324e46SSuren Baghdasaryan FAULT_FLAG_VMA_LOCK = 1 << 12,
127936090defSArnd Bergmann };
128036090defSArnd Bergmann
128105e90bd0SPeter Xu typedef unsigned int __bitwise zap_flags_t;
128205e90bd0SPeter Xu
1283b5054174SDavid Howells /*
1284b5054174SDavid Howells * FOLL_PIN and FOLL_LONGTERM may be used in various combinations with each
1285b5054174SDavid Howells * other. Here is what they mean, and how to use them:
1286b5054174SDavid Howells *
1287b5054174SDavid Howells *
1288b5054174SDavid Howells * FIXME: For pages which are part of a filesystem, mappings are subject to the
1289b5054174SDavid Howells * lifetime enforced by the filesystem and we need guarantees that longterm
1290b5054174SDavid Howells * users like RDMA and V4L2 only establish mappings which coordinate usage with
1291b5054174SDavid Howells * the filesystem. Ideas for this coordination include revoking the longterm
1292b5054174SDavid Howells * pin, delaying writeback, bounce buffer page writeback, etc. As FS DAX was
1293b5054174SDavid Howells * added after the problem with filesystems was found FS DAX VMAs are
1294b5054174SDavid Howells * specifically failed. Filesystem pages are still subject to bugs and use of
1295b5054174SDavid Howells * FOLL_LONGTERM should be avoided on those pages.
1296b5054174SDavid Howells *
1297b5054174SDavid Howells * In the CMA case: long term pins in a CMA region would unnecessarily fragment
1298b5054174SDavid Howells * that region. And so, CMA attempts to migrate the page before pinning, when
1299b5054174SDavid Howells * FOLL_LONGTERM is specified.
1300b5054174SDavid Howells *
1301b5054174SDavid Howells * FOLL_PIN indicates that a special kind of tracking (not just page->_refcount,
1302b5054174SDavid Howells * but an additional pin counting system) will be invoked. This is intended for
1303b5054174SDavid Howells * anything that gets a page reference and then touches page data (for example,
1304b5054174SDavid Howells * Direct IO). This lets the filesystem know that some non-file-system entity is
1305b5054174SDavid Howells * potentially changing the pages' data. In contrast to FOLL_GET (whose pages
1306b5054174SDavid Howells * are released via put_page()), FOLL_PIN pages must be released, ultimately, by
1307b5054174SDavid Howells * a call to unpin_user_page().
1308b5054174SDavid Howells *
1309b5054174SDavid Howells * FOLL_PIN is similar to FOLL_GET: both of these pin pages. They use different
1310b5054174SDavid Howells * and separate refcounting mechanisms, however, and that means that each has
1311b5054174SDavid Howells * its own acquire and release mechanisms:
1312b5054174SDavid Howells *
1313b5054174SDavid Howells * FOLL_GET: get_user_pages*() to acquire, and put_page() to release.
1314b5054174SDavid Howells *
1315b5054174SDavid Howells * FOLL_PIN: pin_user_pages*() to acquire, and unpin_user_pages to release.
1316b5054174SDavid Howells *
1317b5054174SDavid Howells * FOLL_PIN and FOLL_GET are mutually exclusive for a given function call.
1318b5054174SDavid Howells * (The underlying pages may experience both FOLL_GET-based and FOLL_PIN-based
1319b5054174SDavid Howells * calls applied to them, and that's perfectly OK. This is a constraint on the
1320b5054174SDavid Howells * callers, not on the pages.)
1321b5054174SDavid Howells *
1322b5054174SDavid Howells * FOLL_PIN should be set internally by the pin_user_pages*() APIs, never
1323b5054174SDavid Howells * directly by the caller. That's in order to help avoid mismatches when
1324b5054174SDavid Howells * releasing pages: get_user_pages*() pages must be released via put_page(),
1325b5054174SDavid Howells * while pin_user_pages*() pages must be released via unpin_user_page().
1326b5054174SDavid Howells *
1327b5054174SDavid Howells * Please see Documentation/core-api/pin_user_pages.rst for more information.
1328b5054174SDavid Howells */
1329b5054174SDavid Howells
13302c224108SJason Gunthorpe enum {
13312c224108SJason Gunthorpe /* check pte is writable */
13322c224108SJason Gunthorpe FOLL_WRITE = 1 << 0,
13332c224108SJason Gunthorpe /* do get_page on page */
13342c224108SJason Gunthorpe FOLL_GET = 1 << 1,
13352c224108SJason Gunthorpe /* give error on hole if it would be zero */
13362c224108SJason Gunthorpe FOLL_DUMP = 1 << 2,
13372c224108SJason Gunthorpe /* get_user_pages read/write w/o permission */
13382c224108SJason Gunthorpe FOLL_FORCE = 1 << 3,
13392c224108SJason Gunthorpe /*
13402c224108SJason Gunthorpe * if a disk transfer is needed, start the IO and return without waiting
13412c224108SJason Gunthorpe * upon it
13422c224108SJason Gunthorpe */
13432c224108SJason Gunthorpe FOLL_NOWAIT = 1 << 4,
13442c224108SJason Gunthorpe /* do not fault in pages */
13452c224108SJason Gunthorpe FOLL_NOFAULT = 1 << 5,
13462c224108SJason Gunthorpe /* check page is hwpoisoned */
13472c224108SJason Gunthorpe FOLL_HWPOISON = 1 << 6,
13482c224108SJason Gunthorpe /* don't do file mappings */
13492c224108SJason Gunthorpe FOLL_ANON = 1 << 7,
13502c224108SJason Gunthorpe /*
13512c224108SJason Gunthorpe * FOLL_LONGTERM indicates that the page will be held for an indefinite
13522c224108SJason Gunthorpe * time period _often_ under userspace control. This is in contrast to
13532c224108SJason Gunthorpe * iov_iter_get_pages(), whose usages are transient.
13542c224108SJason Gunthorpe */
13552c224108SJason Gunthorpe FOLL_LONGTERM = 1 << 8,
13562c224108SJason Gunthorpe /* split huge pmd before returning */
13572c224108SJason Gunthorpe FOLL_SPLIT_PMD = 1 << 9,
13582c224108SJason Gunthorpe /* allow returning PCI P2PDMA pages */
13592c224108SJason Gunthorpe FOLL_PCI_P2PDMA = 1 << 10,
13602c224108SJason Gunthorpe /* allow interrupts from generic signals */
13612c224108SJason Gunthorpe FOLL_INTERRUPTIBLE = 1 << 11,
1362d74943a2SDavid Hildenbrand /*
1363d74943a2SDavid Hildenbrand * Always honor (trigger) NUMA hinting faults.
1364d74943a2SDavid Hildenbrand *
1365d74943a2SDavid Hildenbrand * FOLL_WRITE implicitly honors NUMA hinting faults because a
1366d74943a2SDavid Hildenbrand * PROT_NONE-mapped page is not writable (exceptions with FOLL_FORCE
1367d74943a2SDavid Hildenbrand * apply). get_user_pages_fast_only() always implicitly honors NUMA
1368d74943a2SDavid Hildenbrand * hinting faults.
1369d74943a2SDavid Hildenbrand */
1370d74943a2SDavid Hildenbrand FOLL_HONOR_NUMA_FAULT = 1 << 12,
13712c224108SJason Gunthorpe
13722c224108SJason Gunthorpe /* See also internal only FOLL flags in mm/internal.h */
13732c224108SJason Gunthorpe };
1374b5054174SDavid Howells
13755b99cd0eSHeiko Carstens #endif /* _LINUX_MM_TYPES_H */
1376