1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef _LINUX_MM_TYPES_H
3 #define _LINUX_MM_TYPES_H
5 #include <linux/mm_types_task.h>
7 #include <linux/auxvec.h>
8 #include <linux/list.h>
9 #include <linux/spinlock.h>
10 #include <linux/rbtree.h>
11 #include <linux/rwsem.h>
12 #include <linux/completion.h>
13 #include <linux/cpumask.h>
14 #include <linux/uprobes.h>
15 #include <linux/rcupdate.h>
16 #include <linux/page-flags-layout.h>
17 #include <linux/workqueue.h>
18 #include <linux/seqlock.h>
22 #ifndef AT_VECTOR_SIZE_ARCH
23 #define AT_VECTOR_SIZE_ARCH 0
25 #define AT_VECTOR_SIZE (2*(AT_VECTOR_SIZE_ARCH + AT_VECTOR_SIZE_BASE + 1))
33 * Each physical page in the system has a struct page associated with
34 * it to keep track of whatever it is we are using the page for at the
35 * moment. Note that we have no way to track which tasks are using
36 * a page, though if it is a pagecache page, rmap structures can tell us
39 * If you allocate the page using alloc_pages(), you can use some of the
40 * space in struct page for your own purposes. The five words in the main
41 * union are available, except for bit 0 of the first word which must be
42 * kept clear. Many users use this word to store a pointer to an object
43 * which is guaranteed to be aligned. If you use the same storage as
44 * page->mapping, you must restore it to NULL before freeing the page.
46 * If your page will not be mapped to userspace, you can also use the four
47 * bytes in the mapcount union, but you must call page_mapcount_reset()
50 * If you want to use the refcount field, it must be used in such a way
51 * that other CPUs temporarily incrementing and then decrementing the
52 * refcount does not cause problems. On receiving the page from
53 * alloc_pages(), the refcount will be positive.
55 * If you allocate pages of order > 0, you can use some of the fields
56 * in each subpage, but you may need to restore some of their values
59 * SLUB uses cmpxchg_double() to atomically update its freelist and counters.
60 * That requires that freelist & counters in struct slab be adjacent and
61 * double-word aligned. Because struct slab currently just reinterprets the
62 * bits of struct page, we align all struct pages to double-word boundaries,
63 * and ensure that 'freelist' is aligned within struct slab.
65 #ifdef CONFIG_HAVE_ALIGNED_STRUCT_PAGE
66 #define _struct_page_alignment __aligned(2 * sizeof(unsigned long))
68 #define _struct_page_alignment
72 unsigned long flags; /* Atomic flags, some possibly
73 * updated asynchronously */
75 * Five words (20/40 bytes) are available in this union.
76 * WARNING: bit 0 of the first word is used for PageTail(). That
77 * means the other users of this union MUST NOT use the bit to
78 * avoid collision and false-positive PageTail().
81 struct { /* Page cache and anonymous pages */
83 * @lru: Pageout list, eg. active_list protected by
84 * lruvec->lru_lock. Sometimes used as a generic list
88 /* See page-flags.h for PAGE_MAPPING_FLAGS */
89 struct address_space *mapping;
90 pgoff_t index; /* Our offset within mapping. */
92 * @private: Mapping-private opaque data.
93 * Usually used for buffer_heads if PagePrivate.
94 * Used for swp_entry_t if PageSwapCache.
95 * Indicates order in the buddy system if PageBuddy.
97 unsigned long private;
99 struct { /* page_pool used by netstack */
101 * @pp_magic: magic value to avoid recycling non
102 * page_pool allocated pages.
104 unsigned long pp_magic;
105 struct page_pool *pp;
106 unsigned long _pp_mapping_pad;
107 unsigned long dma_addr;
110 * dma_addr_upper: might require a 64-bit
111 * value on 32-bit architectures.
113 unsigned long dma_addr_upper;
115 * For frag page support, not supported in
116 * 32-bit architectures with 64-bit DMA.
118 atomic_long_t pp_frag_count;
121 struct { /* slab, slob and slub */
123 struct list_head slab_list;
124 struct { /* Partial pages */
127 int pages; /* Nr of pages left */
133 struct kmem_cache *slab_cache; /* not slob */
134 /* Double-word boundary */
135 void *freelist; /* first free object */
137 void *s_mem; /* slab: first object */
138 unsigned long counters; /* SLUB */
146 struct { /* Tail pages of compound page */
147 unsigned long compound_head; /* Bit zero is set */
149 /* First tail page only */
150 unsigned char compound_dtor;
151 unsigned char compound_order;
152 atomic_t compound_mapcount;
153 unsigned int compound_nr; /* 1 << compound_order */
155 struct { /* Second tail page of compound page */
156 unsigned long _compound_pad_1; /* compound_head */
157 atomic_t hpage_pinned_refcount;
158 /* For both global and memcg */
159 struct list_head deferred_list;
161 struct { /* Page table pages */
162 unsigned long _pt_pad_1; /* compound_head */
163 pgtable_t pmd_huge_pte; /* protected by page->ptl */
164 unsigned long _pt_pad_2; /* mapping */
166 struct mm_struct *pt_mm; /* x86 pgds only */
167 atomic_t pt_frag_refcount; /* powerpc */
169 #if ALLOC_SPLIT_PTLOCKS
175 struct { /* ZONE_DEVICE pages */
176 /** @pgmap: Points to the hosting device page map. */
177 struct dev_pagemap *pgmap;
178 void *zone_device_data;
180 * ZONE_DEVICE private pages are counted as being
181 * mapped so the next 3 words hold the mapping, index,
182 * and private fields from the source anonymous or
183 * page cache page while the page is migrated to device
185 * ZONE_DEVICE MEMORY_DEVICE_FS_DAX pages also
186 * use the mapping, index, and private fields when
187 * pmem backed DAX files are mapped.
191 /** @rcu_head: You can use this to free a page by RCU. */
192 struct rcu_head rcu_head;
195 union { /* This union is 4 bytes in size. */
197 * If the page can be mapped to userspace, encodes the number
198 * of times this page is referenced by a page table.
203 * If the page is neither PageSlab nor mappable to userspace,
204 * the value stored here may help determine what this page
205 * is used for. See page-flags.h for a list of page types
206 * which are currently stored here.
208 unsigned int page_type;
210 unsigned int active; /* SLAB */
211 int units; /* SLOB */
214 /* Usage count. *DO NOT USE DIRECTLY*. See page_ref.h */
218 unsigned long memcg_data;
222 * On machines where all RAM is mapped into kernel address space,
223 * we can simply calculate the virtual address. On machines with
224 * highmem some memory is mapped into kernel virtual memory
225 * dynamically, so we need a place to store that address.
226 * Note that this field could be 16 bits on x86 ... ;)
228 * Architectures with slow multiplication can define
229 * WANT_PAGE_VIRTUAL in asm/page.h
231 #if defined(WANT_PAGE_VIRTUAL)
232 void *virtual; /* Kernel virtual address (NULL if
233 not kmapped, ie. highmem) */
234 #endif /* WANT_PAGE_VIRTUAL */
236 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
239 } _struct_page_alignment;
242 * struct folio - Represents a contiguous set of bytes.
243 * @flags: Identical to the page flags.
244 * @lru: Least Recently Used list; tracks how recently this folio was used.
245 * @mapping: The file this page belongs to, or refers to the anon_vma for
247 * @index: Offset within the file, in units of pages. For anonymous memory,
248 * this is the index from the beginning of the mmap.
249 * @private: Filesystem per-folio data (see folio_attach_private()).
250 * Used for swp_entry_t if folio_test_swapcache().
251 * @_mapcount: Do not access this member directly. Use folio_mapcount() to
252 * find out how many times this folio is mapped by userspace.
253 * @_refcount: Do not access this member directly. Use folio_ref_count()
254 * to find how many references there are to this folio.
255 * @memcg_data: Memory Control Group data.
257 * A folio is a physically, virtually and logically contiguous set
258 * of bytes. It is a power-of-two in size, and it is aligned to that
259 * same power-of-two. It is at least as large as %PAGE_SIZE. If it is
260 * in the page cache, it is at a file offset which is a multiple of that
261 * power-of-two. It may be mapped into userspace at an address which is
262 * at an arbitrary page offset, but its kernel virtual address is aligned
266 /* private: don't document the anon union */
271 struct list_head lru;
272 struct address_space *mapping;
278 unsigned long memcg_data;
280 /* private: the union with struct page is transitional */
286 static_assert(sizeof(struct page) == sizeof(struct folio));
287 #define FOLIO_MATCH(pg, fl) \
288 static_assert(offsetof(struct page, pg) == offsetof(struct folio, fl))
289 FOLIO_MATCH(flags, flags);
290 FOLIO_MATCH(lru, lru);
291 FOLIO_MATCH(compound_head, lru);
292 FOLIO_MATCH(index, index);
293 FOLIO_MATCH(private, private);
294 FOLIO_MATCH(_mapcount, _mapcount);
295 FOLIO_MATCH(_refcount, _refcount);
297 FOLIO_MATCH(memcg_data, memcg_data);
301 static inline atomic_t *folio_mapcount_ptr(struct folio *folio)
303 struct page *tail = &folio->page + 1;
304 return &tail->compound_mapcount;
307 static inline atomic_t *compound_mapcount_ptr(struct page *page)
309 return &page[1].compound_mapcount;
312 static inline atomic_t *compound_pincount_ptr(struct page *page)
314 return &page[2].hpage_pinned_refcount;
318 * Used for sizing the vmemmap region on some architectures
320 #define STRUCT_PAGE_MAX_SHIFT (order_base_2(sizeof(struct page)))
322 #define PAGE_FRAG_CACHE_MAX_SIZE __ALIGN_MASK(32768, ~PAGE_MASK)
323 #define PAGE_FRAG_CACHE_MAX_ORDER get_order(PAGE_FRAG_CACHE_MAX_SIZE)
326 * page_private can be used on tail pages. However, PagePrivate is only
327 * checked by the VM on the head page. So page_private on the tail pages
328 * should be used for data that's ancillary to the head page (eg attaching
329 * buffer heads to tail pages after attaching buffer heads to the head page)
331 #define page_private(page) ((page)->private)
333 static inline void set_page_private(struct page *page, unsigned long private)
335 page->private = private;
338 static inline void *folio_get_private(struct folio *folio)
340 return folio->private;
343 struct page_frag_cache {
345 #if (PAGE_SIZE < PAGE_FRAG_CACHE_MAX_SIZE)
351 /* we maintain a pagecount bias, so that we dont dirty cache line
352 * containing page->_refcount every time we allocate a fragment.
354 unsigned int pagecnt_bias;
358 typedef unsigned long vm_flags_t;
361 * A region containing a mapping of a non-memory backed file under NOMMU
362 * conditions. These are held in a global tree and are pinned by the VMAs that
366 struct rb_node vm_rb; /* link in global region tree */
367 vm_flags_t vm_flags; /* VMA vm_flags */
368 unsigned long vm_start; /* start address of region */
369 unsigned long vm_end; /* region initialised to here */
370 unsigned long vm_top; /* region allocated to here */
371 unsigned long vm_pgoff; /* the offset in vm_file corresponding to vm_start */
372 struct file *vm_file; /* the backing file or NULL */
374 int vm_usage; /* region usage count (access under nommu_region_sem) */
375 bool vm_icache_flushed : 1; /* true if the icache has been flushed for
379 #ifdef CONFIG_USERFAULTFD
380 #define NULL_VM_UFFD_CTX ((struct vm_userfaultfd_ctx) { NULL, })
381 struct vm_userfaultfd_ctx {
382 struct userfaultfd_ctx *ctx;
384 #else /* CONFIG_USERFAULTFD */
385 #define NULL_VM_UFFD_CTX ((struct vm_userfaultfd_ctx) {})
386 struct vm_userfaultfd_ctx {};
387 #endif /* CONFIG_USERFAULTFD */
390 * This struct describes a virtual memory area. There is one of these
391 * per VM-area/task. A VM area is any part of the process virtual memory
392 * space that has a special rule for the page-fault handlers (ie a shared
393 * library, the executable area etc).
395 struct vm_area_struct {
396 /* The first cache line has the info for VMA tree walking. */
398 unsigned long vm_start; /* Our start address within vm_mm. */
399 unsigned long vm_end; /* The first byte after our end address
402 /* linked list of VM areas per task, sorted by address */
403 struct vm_area_struct *vm_next, *vm_prev;
405 struct rb_node vm_rb;
408 * Largest free memory gap in bytes to the left of this VMA.
409 * Either between this VMA and vma->vm_prev, or between one of the
410 * VMAs below us in the VMA rbtree and its ->vm_prev. This helps
411 * get_unmapped_area find a free area of the right size.
413 unsigned long rb_subtree_gap;
415 /* Second cache line starts here. */
417 struct mm_struct *vm_mm; /* The address space we belong to. */
420 * Access permissions of this VMA.
421 * See vmf_insert_mixed_prot() for discussion.
423 pgprot_t vm_page_prot;
424 unsigned long vm_flags; /* Flags, see mm.h. */
427 * For areas with an address space and backing store,
428 * linkage into the address_space->i_mmap interval tree.
432 unsigned long rb_subtree_last;
436 * A file's MAP_PRIVATE vma can be in both i_mmap tree and anon_vma
437 * list, after a COW of one of the file pages. A MAP_SHARED vma
438 * can only be in the i_mmap tree. An anonymous MAP_PRIVATE, stack
439 * or brk vma (with NULL file) can only be in an anon_vma list.
441 struct list_head anon_vma_chain; /* Serialized by mmap_lock &
443 struct anon_vma *anon_vma; /* Serialized by page_table_lock */
445 /* Function pointers to deal with this struct. */
446 const struct vm_operations_struct *vm_ops;
448 /* Information about our backing store: */
449 unsigned long vm_pgoff; /* Offset (within vm_file) in PAGE_SIZE
451 struct file * vm_file; /* File we map to (can be NULL). */
452 void * vm_private_data; /* was vm_pte (shared mem) */
455 atomic_long_t swap_readahead_info;
458 struct vm_region *vm_region; /* NOMMU mapping region */
461 struct mempolicy *vm_policy; /* NUMA policy for the VMA */
463 struct vm_userfaultfd_ctx vm_userfaultfd_ctx;
464 } __randomize_layout;
469 struct vm_area_struct *mmap; /* list of VMAs */
470 struct rb_root mm_rb;
471 u64 vmacache_seqnum; /* per-thread vmacache */
473 unsigned long (*get_unmapped_area) (struct file *filp,
474 unsigned long addr, unsigned long len,
475 unsigned long pgoff, unsigned long flags);
477 unsigned long mmap_base; /* base of mmap area */
478 unsigned long mmap_legacy_base; /* base of mmap area in bottom-up allocations */
479 #ifdef CONFIG_HAVE_ARCH_COMPAT_MMAP_BASES
480 /* Base addresses for compatible mmap() */
481 unsigned long mmap_compat_base;
482 unsigned long mmap_compat_legacy_base;
484 unsigned long task_size; /* size of task vm space */
485 unsigned long highest_vm_end; /* highest vma end address */
488 #ifdef CONFIG_MEMBARRIER
490 * @membarrier_state: Flags controlling membarrier behavior.
492 * This field is close to @pgd to hopefully fit in the same
493 * cache-line, which needs to be touched by switch_mm().
495 atomic_t membarrier_state;
499 * @mm_users: The number of users including userspace.
501 * Use mmget()/mmget_not_zero()/mmput() to modify. When this
502 * drops to 0 (i.e. when the task exits and there are no other
503 * temporary reference holders), we also release a reference on
504 * @mm_count (which may then free the &struct mm_struct if
505 * @mm_count also drops to 0).
510 * @mm_count: The number of references to &struct mm_struct
511 * (@mm_users count as 1).
513 * Use mmgrab()/mmdrop() to modify. When this drops to 0, the
514 * &struct mm_struct is freed.
519 atomic_long_t pgtables_bytes; /* PTE page table pages */
521 int map_count; /* number of VMAs */
523 spinlock_t page_table_lock; /* Protects page tables and some
527 * With some kernel config, the current mmap_lock's offset
528 * inside 'mm_struct' is at 0x120, which is very optimal, as
529 * its two hot fields 'count' and 'owner' sit in 2 different
530 * cachelines, and when mmap_lock is highly contended, both
531 * of the 2 fields will be accessed frequently, current layout
532 * will help to reduce cache bouncing.
534 * So please be careful with adding new fields before
535 * mmap_lock, which can easily push the 2 fields into one
538 struct rw_semaphore mmap_lock;
540 struct list_head mmlist; /* List of maybe swapped mm's. These
541 * are globally strung together off
542 * init_mm.mmlist, and are protected
547 unsigned long hiwater_rss; /* High-watermark of RSS usage */
548 unsigned long hiwater_vm; /* High-water virtual memory usage */
550 unsigned long total_vm; /* Total pages mapped */
551 unsigned long locked_vm; /* Pages that have PG_mlocked set */
552 atomic64_t pinned_vm; /* Refcount permanently increased */
553 unsigned long data_vm; /* VM_WRITE & ~VM_SHARED & ~VM_STACK */
554 unsigned long exec_vm; /* VM_EXEC & ~VM_WRITE & ~VM_STACK */
555 unsigned long stack_vm; /* VM_STACK */
556 unsigned long def_flags;
559 * @write_protect_seq: Locked when any thread is write
560 * protecting pages mapped by this mm to enforce a later COW,
561 * for instance during page table copying for fork().
563 seqcount_t write_protect_seq;
565 spinlock_t arg_lock; /* protect the below fields */
567 unsigned long start_code, end_code, start_data, end_data;
568 unsigned long start_brk, brk, start_stack;
569 unsigned long arg_start, arg_end, env_start, env_end;
571 unsigned long saved_auxv[AT_VECTOR_SIZE]; /* for /proc/PID/auxv */
574 * Special counters, in some configurations protected by the
575 * page_table_lock, in other configurations by being atomic.
577 struct mm_rss_stat rss_stat;
579 struct linux_binfmt *binfmt;
581 /* Architecture-specific MM context */
582 mm_context_t context;
584 unsigned long flags; /* Must use atomic bitops to access */
587 spinlock_t ioctx_lock;
588 struct kioctx_table __rcu *ioctx_table;
592 * "owner" points to a task that is regarded as the canonical
593 * user/owner of this mm. All of the following must be true in
594 * order for it to be changed:
596 * current == mm->owner
598 * new_owner->mm == mm
599 * new_owner->alloc_lock is held
601 struct task_struct __rcu *owner;
603 struct user_namespace *user_ns;
605 /* store ref to file /proc/<pid>/exe symlink points to */
606 struct file __rcu *exe_file;
607 #ifdef CONFIG_MMU_NOTIFIER
608 struct mmu_notifier_subscriptions *notifier_subscriptions;
610 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS
611 pgtable_t pmd_huge_pte; /* protected by page_table_lock */
613 #ifdef CONFIG_NUMA_BALANCING
615 * numa_next_scan is the next time that the PTEs will be marked
616 * pte_numa. NUMA hinting faults will gather statistics and
617 * migrate pages to new nodes if necessary.
619 unsigned long numa_next_scan;
621 /* Restart point for scanning and setting pte_numa */
622 unsigned long numa_scan_offset;
624 /* numa_scan_seq prevents two threads setting pte_numa */
628 * An operation with batched TLB flushing is going on. Anything
629 * that can move process memory needs to flush the TLB when
630 * moving a PROT_NONE or PROT_NUMA mapped page.
632 atomic_t tlb_flush_pending;
633 #ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH
634 /* See flush_tlb_batched_pending() */
635 bool tlb_flush_batched;
637 struct uprobes_state uprobes_state;
638 #ifdef CONFIG_PREEMPT_RT
639 struct rcu_head delayed_drop;
641 #ifdef CONFIG_HUGETLB_PAGE
642 atomic_long_t hugetlb_usage;
644 struct work_struct async_put_work;
646 #ifdef CONFIG_IOMMU_SUPPORT
649 } __randomize_layout;
652 * The mm_cpumask needs to be at the end of mm_struct, because it
653 * is dynamically sized based on nr_cpu_ids.
655 unsigned long cpu_bitmap[];
658 extern struct mm_struct init_mm;
660 /* Pointer magic because the dynamic array size confuses some compilers. */
661 static inline void mm_init_cpumask(struct mm_struct *mm)
663 unsigned long cpu_bitmap = (unsigned long)mm;
665 cpu_bitmap += offsetof(struct mm_struct, cpu_bitmap);
666 cpumask_clear((struct cpumask *)cpu_bitmap);
669 /* Future-safe accessor for struct mm_struct's cpu_vm_mask. */
670 static inline cpumask_t *mm_cpumask(struct mm_struct *mm)
672 return (struct cpumask *)&mm->cpu_bitmap;
676 extern void tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm);
677 extern void tlb_gather_mmu_fullmm(struct mmu_gather *tlb, struct mm_struct *mm);
678 extern void tlb_finish_mmu(struct mmu_gather *tlb);
680 static inline void init_tlb_flush_pending(struct mm_struct *mm)
682 atomic_set(&mm->tlb_flush_pending, 0);
685 static inline void inc_tlb_flush_pending(struct mm_struct *mm)
687 atomic_inc(&mm->tlb_flush_pending);
689 * The only time this value is relevant is when there are indeed pages
690 * to flush. And we'll only flush pages after changing them, which
693 * So the ordering here is:
695 * atomic_inc(&mm->tlb_flush_pending);
702 * mm_tlb_flush_pending();
707 * atomic_dec(&mm->tlb_flush_pending);
709 * Where the increment if constrained by the PTL unlock, it thus
710 * ensures that the increment is visible if the PTE modification is
711 * visible. After all, if there is no PTE modification, nobody cares
712 * about TLB flushes either.
714 * This very much relies on users (mm_tlb_flush_pending() and
715 * mm_tlb_flush_nested()) only caring about _specific_ PTEs (and
716 * therefore specific PTLs), because with SPLIT_PTE_PTLOCKS and RCpc
717 * locks (PPC) the unlock of one doesn't order against the lock of
720 * The decrement is ordered by the flush_tlb_range(), such that
721 * mm_tlb_flush_pending() will not return false unless all flushes have
726 static inline void dec_tlb_flush_pending(struct mm_struct *mm)
729 * See inc_tlb_flush_pending().
731 * This cannot be smp_mb__before_atomic() because smp_mb() simply does
732 * not order against TLB invalidate completion, which is what we need.
734 * Therefore we must rely on tlb_flush_*() to guarantee order.
736 atomic_dec(&mm->tlb_flush_pending);
739 static inline bool mm_tlb_flush_pending(struct mm_struct *mm)
742 * Must be called after having acquired the PTL; orders against that
743 * PTLs release and therefore ensures that if we observe the modified
744 * PTE we must also observe the increment from inc_tlb_flush_pending().
746 * That is, it only guarantees to return true if there is a flush
747 * pending for _this_ PTL.
749 return atomic_read(&mm->tlb_flush_pending);
752 static inline bool mm_tlb_flush_nested(struct mm_struct *mm)
755 * Similar to mm_tlb_flush_pending(), we must have acquired the PTL
756 * for which there is a TLB flush pending in order to guarantee
757 * we've seen both that PTE modification and the increment.
759 * (no requirement on actually still holding the PTL, that is irrelevant)
761 return atomic_read(&mm->tlb_flush_pending) > 1;
767 * typedef vm_fault_t - Return type for page fault handlers.
769 * Page fault handlers return a bitmask of %VM_FAULT values.
771 typedef __bitwise unsigned int vm_fault_t;
774 * enum vm_fault_reason - Page fault handlers return a bitmask of
775 * these values to tell the core VM what happened when handling the
776 * fault. Used to decide whether a process gets delivered SIGBUS or
777 * just gets major/minor fault counters bumped up.
779 * @VM_FAULT_OOM: Out Of Memory
780 * @VM_FAULT_SIGBUS: Bad access
781 * @VM_FAULT_MAJOR: Page read from storage
782 * @VM_FAULT_WRITE: Special case for get_user_pages
783 * @VM_FAULT_HWPOISON: Hit poisoned small page
784 * @VM_FAULT_HWPOISON_LARGE: Hit poisoned large page. Index encoded
786 * @VM_FAULT_SIGSEGV: segmentation fault
787 * @VM_FAULT_NOPAGE: ->fault installed the pte, not return page
788 * @VM_FAULT_LOCKED: ->fault locked the returned page
789 * @VM_FAULT_RETRY: ->fault blocked, must retry
790 * @VM_FAULT_FALLBACK: huge page fault failed, fall back to small
791 * @VM_FAULT_DONE_COW: ->fault has fully handled COW
792 * @VM_FAULT_NEEDDSYNC: ->fault did not modify page tables and needs
793 * fsync() to complete (for synchronous page faults
795 * @VM_FAULT_HINDEX_MASK: mask HINDEX value
798 enum vm_fault_reason {
799 VM_FAULT_OOM = (__force vm_fault_t)0x000001,
800 VM_FAULT_SIGBUS = (__force vm_fault_t)0x000002,
801 VM_FAULT_MAJOR = (__force vm_fault_t)0x000004,
802 VM_FAULT_WRITE = (__force vm_fault_t)0x000008,
803 VM_FAULT_HWPOISON = (__force vm_fault_t)0x000010,
804 VM_FAULT_HWPOISON_LARGE = (__force vm_fault_t)0x000020,
805 VM_FAULT_SIGSEGV = (__force vm_fault_t)0x000040,
806 VM_FAULT_NOPAGE = (__force vm_fault_t)0x000100,
807 VM_FAULT_LOCKED = (__force vm_fault_t)0x000200,
808 VM_FAULT_RETRY = (__force vm_fault_t)0x000400,
809 VM_FAULT_FALLBACK = (__force vm_fault_t)0x000800,
810 VM_FAULT_DONE_COW = (__force vm_fault_t)0x001000,
811 VM_FAULT_NEEDDSYNC = (__force vm_fault_t)0x002000,
812 VM_FAULT_HINDEX_MASK = (__force vm_fault_t)0x0f0000,
815 /* Encode hstate index for a hwpoisoned large page */
816 #define VM_FAULT_SET_HINDEX(x) ((__force vm_fault_t)((x) << 16))
817 #define VM_FAULT_GET_HINDEX(x) (((__force unsigned int)(x) >> 16) & 0xf)
819 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | \
820 VM_FAULT_SIGSEGV | VM_FAULT_HWPOISON | \
821 VM_FAULT_HWPOISON_LARGE | VM_FAULT_FALLBACK)
823 #define VM_FAULT_RESULT_TRACE \
824 { VM_FAULT_OOM, "OOM" }, \
825 { VM_FAULT_SIGBUS, "SIGBUS" }, \
826 { VM_FAULT_MAJOR, "MAJOR" }, \
827 { VM_FAULT_WRITE, "WRITE" }, \
828 { VM_FAULT_HWPOISON, "HWPOISON" }, \
829 { VM_FAULT_HWPOISON_LARGE, "HWPOISON_LARGE" }, \
830 { VM_FAULT_SIGSEGV, "SIGSEGV" }, \
831 { VM_FAULT_NOPAGE, "NOPAGE" }, \
832 { VM_FAULT_LOCKED, "LOCKED" }, \
833 { VM_FAULT_RETRY, "RETRY" }, \
834 { VM_FAULT_FALLBACK, "FALLBACK" }, \
835 { VM_FAULT_DONE_COW, "DONE_COW" }, \
836 { VM_FAULT_NEEDDSYNC, "NEEDDSYNC" }
838 struct vm_special_mapping {
839 const char *name; /* The name, e.g. "[vdso]". */
842 * If .fault is not provided, this points to a
843 * NULL-terminated array of pages that back the special mapping.
845 * This must not be NULL unless .fault is provided.
850 * If non-NULL, then this is called to resolve page faults
851 * on the special mapping. If used, .pages is not checked.
853 vm_fault_t (*fault)(const struct vm_special_mapping *sm,
854 struct vm_area_struct *vma,
855 struct vm_fault *vmf);
857 int (*mremap)(const struct vm_special_mapping *sm,
858 struct vm_area_struct *new_vma);
861 enum tlb_flush_reason {
862 TLB_FLUSH_ON_TASK_SWITCH,
863 TLB_REMOTE_SHOOTDOWN,
865 TLB_LOCAL_MM_SHOOTDOWN,
867 NR_TLB_FLUSH_REASONS,
871 * A swap entry has to fit into a "unsigned long", as the entry is hidden
872 * in the "index" field of the swapper address space.
878 #endif /* _LINUX_MM_TYPES_H */