1/* SPDX-License-Identifier: GPL-2.0 */ 2#ifndef _LINUX_MM_TYPES_H 3#define _LINUX_MM_TYPES_H 4 5#include <linux/mm_types_task.h> 6 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/page-flags-layout.h> 16#include <linux/workqueue.h> 17 18#include <asm/mmu.h> 19 20#ifndef AT_VECTOR_SIZE_ARCH 21#define AT_VECTOR_SIZE_ARCH 0 22#endif 23#define AT_VECTOR_SIZE (2*(AT_VECTOR_SIZE_ARCH + AT_VECTOR_SIZE_BASE + 1)) 24 25 26struct address_space; 27struct mem_cgroup; 28struct hmm; 29 30/* 31 * Each physical page in the system has a struct page associated with 32 * it to keep track of whatever it is we are using the page for at the 33 * moment. Note that we have no way to track which tasks are using 34 * a page, though if it is a pagecache page, rmap structures can tell us 35 * who is mapping it. 36 * 37 * If you allocate the page using alloc_pages(), you can use some of the 38 * space in struct page for your own purposes. The five words in the main 39 * union are available, except for bit 0 of the first word which must be 40 * kept clear. Many users use this word to store a pointer to an object 41 * which is guaranteed to be aligned. If you use the same storage as 42 * page->mapping, you must restore it to NULL before freeing the page. 43 * 44 * If your page will not be mapped to userspace, you can also use the four 45 * bytes in the mapcount union, but you must call page_mapcount_reset() 46 * before freeing it. 47 * 48 * If you want to use the refcount field, it must be used in such a way 49 * that other CPUs temporarily incrementing and then decrementing the 50 * refcount does not cause problems. On receiving the page from 51 * alloc_pages(), the refcount will be positive. 52 * 53 * If you allocate pages of order > 0, you can use some of the fields 54 * in each subpage, but you may need to restore some of their values 55 * afterwards. 56 * 57 * SLUB uses cmpxchg_double() to atomically update its freelist and 58 * counters. That requires that freelist & counters be adjacent and 59 * double-word aligned. We align all struct pages to double-word 60 * boundaries, and ensure that 'freelist' is aligned within the 61 * struct. 62 */ 63#ifdef CONFIG_HAVE_ALIGNED_STRUCT_PAGE 64#define _struct_page_alignment __aligned(2 * sizeof(unsigned long)) 65#else 66#define _struct_page_alignment 67#endif 68 69struct page { 70 unsigned long flags; /* Atomic flags, some possibly 71 * updated asynchronously */ 72 /* 73 * Five words (20/40 bytes) are available in this union. 74 * WARNING: bit 0 of the first word is used for PageTail(). That 75 * means the other users of this union MUST NOT use the bit to 76 * avoid collision and false-positive PageTail(). 77 */ 78 union { 79 struct { /* Page cache and anonymous pages */ 80 /** 81 * @lru: Pageout list, eg. active_list protected by 82 * pgdat->lru_lock. Sometimes used as a generic list 83 * by the page owner. 84 */ 85 struct list_head lru; 86 /* See page-flags.h for PAGE_MAPPING_FLAGS */ 87 struct address_space *mapping; 88 pgoff_t index; /* Our offset within mapping. */ 89 /** 90 * @private: Mapping-private opaque data. 91 * Usually used for buffer_heads if PagePrivate. 92 * Used for swp_entry_t if PageSwapCache. 93 * Indicates order in the buddy system if PageBuddy. 94 */ 95 unsigned long private; 96 }; 97 struct { /* page_pool used by netstack */ 98 /** 99 * @dma_addr: might require a 64-bit value even on 100 * 32-bit architectures. 101 */ 102 dma_addr_t dma_addr; 103 }; 104 struct { /* slab, slob and slub */ 105 union { 106 struct list_head slab_list; 107 struct { /* Partial pages */ 108 struct page *next; 109#ifdef CONFIG_64BIT 110 int pages; /* Nr of pages left */ 111 int pobjects; /* Approximate count */ 112#else 113 short int pages; 114 short int pobjects; 115#endif 116 }; 117 }; 118 struct kmem_cache *slab_cache; /* not slob */ 119 /* Double-word boundary */ 120 void *freelist; /* first free object */ 121 union { 122 void *s_mem; /* slab: first object */ 123 unsigned long counters; /* SLUB */ 124 struct { /* SLUB */ 125 unsigned inuse:16; 126 unsigned objects:15; 127 unsigned frozen:1; 128 }; 129 }; 130 }; 131 struct { /* Tail pages of compound page */ 132 unsigned long compound_head; /* Bit zero is set */ 133 134 /* First tail page only */ 135 unsigned char compound_dtor; 136 unsigned char compound_order; 137 atomic_t compound_mapcount; 138 }; 139 struct { /* Second tail page of compound page */ 140 unsigned long _compound_pad_1; /* compound_head */ 141 unsigned long _compound_pad_2; 142 struct list_head deferred_list; 143 }; 144 struct { /* Page table pages */ 145 unsigned long _pt_pad_1; /* compound_head */ 146 pgtable_t pmd_huge_pte; /* protected by page->ptl */ 147 unsigned long _pt_pad_2; /* mapping */ 148 union { 149 struct mm_struct *pt_mm; /* x86 pgds only */ 150 atomic_t pt_frag_refcount; /* powerpc */ 151 }; 152#if ALLOC_SPLIT_PTLOCKS 153 spinlock_t *ptl; 154#else 155 spinlock_t ptl; 156#endif 157 }; 158 struct { /* ZONE_DEVICE pages */ 159 /** @pgmap: Points to the hosting device page map. */ 160 struct dev_pagemap *pgmap; 161 void *zone_device_data; 162 /* 163 * ZONE_DEVICE private pages are counted as being 164 * mapped so the next 3 words hold the mapping, index, 165 * and private fields from the source anonymous or 166 * page cache page while the page is migrated to device 167 * private memory. 168 * ZONE_DEVICE MEMORY_DEVICE_FS_DAX pages also 169 * use the mapping, index, and private fields when 170 * pmem backed DAX files are mapped. 171 */ 172 }; 173 174 /** @rcu_head: You can use this to free a page by RCU. */ 175 struct rcu_head rcu_head; 176 }; 177 178 union { /* This union is 4 bytes in size. */ 179 /* 180 * If the page can be mapped to userspace, encodes the number 181 * of times this page is referenced by a page table. 182 */ 183 atomic_t _mapcount; 184 185 /* 186 * If the page is neither PageSlab nor mappable to userspace, 187 * the value stored here may help determine what this page 188 * is used for. See page-flags.h for a list of page types 189 * which are currently stored here. 190 */ 191 unsigned int page_type; 192 193 unsigned int active; /* SLAB */ 194 int units; /* SLOB */ 195 }; 196 197 /* Usage count. *DO NOT USE DIRECTLY*. See page_ref.h */ 198 atomic_t _refcount; 199 200#ifdef CONFIG_MEMCG 201 struct mem_cgroup *mem_cgroup; 202#endif 203 204 /* 205 * On machines where all RAM is mapped into kernel address space, 206 * we can simply calculate the virtual address. On machines with 207 * highmem some memory is mapped into kernel virtual memory 208 * dynamically, so we need a place to store that address. 209 * Note that this field could be 16 bits on x86 ... ;) 210 * 211 * Architectures with slow multiplication can define 212 * WANT_PAGE_VIRTUAL in asm/page.h 213 */ 214#if defined(WANT_PAGE_VIRTUAL) 215 void *virtual; /* Kernel virtual address (NULL if 216 not kmapped, ie. highmem) */ 217#endif /* WANT_PAGE_VIRTUAL */ 218 219#ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS 220 int _last_cpupid; 221#endif 222} _struct_page_alignment; 223 224/* 225 * Used for sizing the vmemmap region on some architectures 226 */ 227#define STRUCT_PAGE_MAX_SHIFT (order_base_2(sizeof(struct page))) 228 229#define PAGE_FRAG_CACHE_MAX_SIZE __ALIGN_MASK(32768, ~PAGE_MASK) 230#define PAGE_FRAG_CACHE_MAX_ORDER get_order(PAGE_FRAG_CACHE_MAX_SIZE) 231 232#define page_private(page) ((page)->private) 233#define set_page_private(page, v) ((page)->private = (v)) 234 235struct page_frag_cache { 236 void * va; 237#if (PAGE_SIZE < PAGE_FRAG_CACHE_MAX_SIZE) 238 __u16 offset; 239 __u16 size; 240#else 241 __u32 offset; 242#endif 243 /* we maintain a pagecount bias, so that we dont dirty cache line 244 * containing page->_refcount every time we allocate a fragment. 245 */ 246 unsigned int pagecnt_bias; 247 bool pfmemalloc; 248}; 249 250typedef unsigned long vm_flags_t; 251 252/* 253 * A region containing a mapping of a non-memory backed file under NOMMU 254 * conditions. These are held in a global tree and are pinned by the VMAs that 255 * map parts of them. 256 */ 257struct vm_region { 258 struct rb_node vm_rb; /* link in global region tree */ 259 vm_flags_t vm_flags; /* VMA vm_flags */ 260 unsigned long vm_start; /* start address of region */ 261 unsigned long vm_end; /* region initialised to here */ 262 unsigned long vm_top; /* region allocated to here */ 263 unsigned long vm_pgoff; /* the offset in vm_file corresponding to vm_start */ 264 struct file *vm_file; /* the backing file or NULL */ 265 266 int vm_usage; /* region usage count (access under nommu_region_sem) */ 267 bool vm_icache_flushed : 1; /* true if the icache has been flushed for 268 * this region */ 269}; 270 271#ifdef CONFIG_USERFAULTFD 272#define NULL_VM_UFFD_CTX ((struct vm_userfaultfd_ctx) { NULL, }) 273struct vm_userfaultfd_ctx { 274 struct userfaultfd_ctx *ctx; 275}; 276#else /* CONFIG_USERFAULTFD */ 277#define NULL_VM_UFFD_CTX ((struct vm_userfaultfd_ctx) {}) 278struct vm_userfaultfd_ctx {}; 279#endif /* CONFIG_USERFAULTFD */ 280 281/* 282 * This struct defines a memory VMM memory area. There is one of these 283 * per VM-area/task. A VM area is any part of the process virtual memory 284 * space that has a special rule for the page-fault handlers (ie a shared 285 * library, the executable area etc). 286 */ 287struct vm_area_struct { 288 /* The first cache line has the info for VMA tree walking. */ 289 290 unsigned long vm_start; /* Our start address within vm_mm. */ 291 unsigned long vm_end; /* The first byte after our end address 292 within vm_mm. */ 293 294 /* linked list of VM areas per task, sorted by address */ 295 struct vm_area_struct *vm_next, *vm_prev; 296 297 struct rb_node vm_rb; 298 299 /* 300 * Largest free memory gap in bytes to the left of this VMA. 301 * Either between this VMA and vma->vm_prev, or between one of the 302 * VMAs below us in the VMA rbtree and its ->vm_prev. This helps 303 * get_unmapped_area find a free area of the right size. 304 */ 305 unsigned long rb_subtree_gap; 306 307 /* Second cache line starts here. */ 308 309 struct mm_struct *vm_mm; /* The address space we belong to. */ 310 pgprot_t vm_page_prot; /* Access permissions of this VMA. */ 311 unsigned long vm_flags; /* Flags, see mm.h. */ 312 313 /* 314 * For areas with an address space and backing store, 315 * linkage into the address_space->i_mmap interval tree. 316 */ 317 struct { 318 struct rb_node rb; 319 unsigned long rb_subtree_last; 320 } shared; 321 322 /* 323 * A file's MAP_PRIVATE vma can be in both i_mmap tree and anon_vma 324 * list, after a COW of one of the file pages. A MAP_SHARED vma 325 * can only be in the i_mmap tree. An anonymous MAP_PRIVATE, stack 326 * or brk vma (with NULL file) can only be in an anon_vma list. 327 */ 328 struct list_head anon_vma_chain; /* Serialized by mmap_sem & 329 * page_table_lock */ 330 struct anon_vma *anon_vma; /* Serialized by page_table_lock */ 331 332 /* Function pointers to deal with this struct. */ 333 const struct vm_operations_struct *vm_ops; 334 335 /* Information about our backing store: */ 336 unsigned long vm_pgoff; /* Offset (within vm_file) in PAGE_SIZE 337 units */ 338 struct file * vm_file; /* File we map to (can be NULL). */ 339 void * vm_private_data; /* was vm_pte (shared mem) */ 340 341#ifdef CONFIG_SWAP 342 atomic_long_t swap_readahead_info; 343#endif 344#ifndef CONFIG_MMU 345 struct vm_region *vm_region; /* NOMMU mapping region */ 346#endif 347#ifdef CONFIG_NUMA 348 struct mempolicy *vm_policy; /* NUMA policy for the VMA */ 349#endif 350 struct vm_userfaultfd_ctx vm_userfaultfd_ctx; 351} __randomize_layout; 352 353struct core_thread { 354 struct task_struct *task; 355 struct core_thread *next; 356}; 357 358struct core_state { 359 atomic_t nr_threads; 360 struct core_thread dumper; 361 struct completion startup; 362}; 363 364struct kioctx_table; 365struct mm_struct { 366 struct { 367 struct vm_area_struct *mmap; /* list of VMAs */ 368 struct rb_root mm_rb; 369 u64 vmacache_seqnum; /* per-thread vmacache */ 370#ifdef CONFIG_MMU 371 unsigned long (*get_unmapped_area) (struct file *filp, 372 unsigned long addr, unsigned long len, 373 unsigned long pgoff, unsigned long flags); 374#endif 375 unsigned long mmap_base; /* base of mmap area */ 376 unsigned long mmap_legacy_base; /* base of mmap area in bottom-up allocations */ 377#ifdef CONFIG_HAVE_ARCH_COMPAT_MMAP_BASES 378 /* Base adresses for compatible mmap() */ 379 unsigned long mmap_compat_base; 380 unsigned long mmap_compat_legacy_base; 381#endif 382 unsigned long task_size; /* size of task vm space */ 383 unsigned long highest_vm_end; /* highest vma end address */ 384 pgd_t * pgd; 385 386 /** 387 * @mm_users: The number of users including userspace. 388 * 389 * Use mmget()/mmget_not_zero()/mmput() to modify. When this 390 * drops to 0 (i.e. when the task exits and there are no other 391 * temporary reference holders), we also release a reference on 392 * @mm_count (which may then free the &struct mm_struct if 393 * @mm_count also drops to 0). 394 */ 395 atomic_t mm_users; 396 397 /** 398 * @mm_count: The number of references to &struct mm_struct 399 * (@mm_users count as 1). 400 * 401 * Use mmgrab()/mmdrop() to modify. When this drops to 0, the 402 * &struct mm_struct is freed. 403 */ 404 atomic_t mm_count; 405 406#ifdef CONFIG_MMU 407 atomic_long_t pgtables_bytes; /* PTE page table pages */ 408#endif 409 int map_count; /* number of VMAs */ 410 411 spinlock_t page_table_lock; /* Protects page tables and some 412 * counters 413 */ 414 struct rw_semaphore mmap_sem; 415 416 struct list_head mmlist; /* List of maybe swapped mm's. These 417 * are globally strung together off 418 * init_mm.mmlist, and are protected 419 * by mmlist_lock 420 */ 421 422 423 unsigned long hiwater_rss; /* High-watermark of RSS usage */ 424 unsigned long hiwater_vm; /* High-water virtual memory usage */ 425 426 unsigned long total_vm; /* Total pages mapped */ 427 unsigned long locked_vm; /* Pages that have PG_mlocked set */ 428 atomic64_t pinned_vm; /* Refcount permanently increased */ 429 unsigned long data_vm; /* VM_WRITE & ~VM_SHARED & ~VM_STACK */ 430 unsigned long exec_vm; /* VM_EXEC & ~VM_WRITE & ~VM_STACK */ 431 unsigned long stack_vm; /* VM_STACK */ 432 unsigned long def_flags; 433 434 spinlock_t arg_lock; /* protect the below fields */ 435 unsigned long start_code, end_code, start_data, end_data; 436 unsigned long start_brk, brk, start_stack; 437 unsigned long arg_start, arg_end, env_start, env_end; 438 439 unsigned long saved_auxv[AT_VECTOR_SIZE]; /* for /proc/PID/auxv */ 440 441 /* 442 * Special counters, in some configurations protected by the 443 * page_table_lock, in other configurations by being atomic. 444 */ 445 struct mm_rss_stat rss_stat; 446 447 struct linux_binfmt *binfmt; 448 449 /* Architecture-specific MM context */ 450 mm_context_t context; 451 452 unsigned long flags; /* Must use atomic bitops to access */ 453 454 struct core_state *core_state; /* coredumping support */ 455#ifdef CONFIG_MEMBARRIER 456 atomic_t membarrier_state; 457#endif 458#ifdef CONFIG_AIO 459 spinlock_t ioctx_lock; 460 struct kioctx_table __rcu *ioctx_table; 461#endif 462#ifdef CONFIG_MEMCG 463 /* 464 * "owner" points to a task that is regarded as the canonical 465 * user/owner of this mm. All of the following must be true in 466 * order for it to be changed: 467 * 468 * current == mm->owner 469 * current->mm != mm 470 * new_owner->mm == mm 471 * new_owner->alloc_lock is held 472 */ 473 struct task_struct __rcu *owner; 474#endif 475 struct user_namespace *user_ns; 476 477 /* store ref to file /proc/<pid>/exe symlink points to */ 478 struct file __rcu *exe_file; 479#ifdef CONFIG_MMU_NOTIFIER 480 struct mmu_notifier_mm *mmu_notifier_mm; 481#endif 482#if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS 483 pgtable_t pmd_huge_pte; /* protected by page_table_lock */ 484#endif 485#ifdef CONFIG_NUMA_BALANCING 486 /* 487 * numa_next_scan is the next time that the PTEs will be marked 488 * pte_numa. NUMA hinting faults will gather statistics and 489 * migrate pages to new nodes if necessary. 490 */ 491 unsigned long numa_next_scan; 492 493 /* Restart point for scanning and setting pte_numa */ 494 unsigned long numa_scan_offset; 495 496 /* numa_scan_seq prevents two threads setting pte_numa */ 497 int numa_scan_seq; 498#endif 499 /* 500 * An operation with batched TLB flushing is going on. Anything 501 * that can move process memory needs to flush the TLB when 502 * moving a PROT_NONE or PROT_NUMA mapped page. 503 */ 504 atomic_t tlb_flush_pending; 505#ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH 506 /* See flush_tlb_batched_pending() */ 507 bool tlb_flush_batched; 508#endif 509 struct uprobes_state uprobes_state; 510#ifdef CONFIG_HUGETLB_PAGE 511 atomic_long_t hugetlb_usage; 512#endif 513 struct work_struct async_put_work; 514 515#ifdef CONFIG_HMM_MIRROR 516 /* HMM needs to track a few things per mm */ 517 struct hmm *hmm; 518#endif 519 } __randomize_layout; 520 521 /* 522 * The mm_cpumask needs to be at the end of mm_struct, because it 523 * is dynamically sized based on nr_cpu_ids. 524 */ 525 unsigned long cpu_bitmap[]; 526}; 527 528extern struct mm_struct init_mm; 529 530/* Pointer magic because the dynamic array size confuses some compilers. */ 531static inline void mm_init_cpumask(struct mm_struct *mm) 532{ 533 unsigned long cpu_bitmap = (unsigned long)mm; 534 535 cpu_bitmap += offsetof(struct mm_struct, cpu_bitmap); 536 cpumask_clear((struct cpumask *)cpu_bitmap); 537} 538 539/* Future-safe accessor for struct mm_struct's cpu_vm_mask. */ 540static inline cpumask_t *mm_cpumask(struct mm_struct *mm) 541{ 542 return (struct cpumask *)&mm->cpu_bitmap; 543} 544 545struct mmu_gather; 546extern void tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, 547 unsigned long start, unsigned long end); 548extern void tlb_finish_mmu(struct mmu_gather *tlb, 549 unsigned long start, unsigned long end); 550 551static inline void init_tlb_flush_pending(struct mm_struct *mm) 552{ 553 atomic_set(&mm->tlb_flush_pending, 0); 554} 555 556static inline void inc_tlb_flush_pending(struct mm_struct *mm) 557{ 558 atomic_inc(&mm->tlb_flush_pending); 559 /* 560 * The only time this value is relevant is when there are indeed pages 561 * to flush. And we'll only flush pages after changing them, which 562 * requires the PTL. 563 * 564 * So the ordering here is: 565 * 566 * atomic_inc(&mm->tlb_flush_pending); 567 * spin_lock(&ptl); 568 * ... 569 * set_pte_at(); 570 * spin_unlock(&ptl); 571 * 572 * spin_lock(&ptl) 573 * mm_tlb_flush_pending(); 574 * .... 575 * spin_unlock(&ptl); 576 * 577 * flush_tlb_range(); 578 * atomic_dec(&mm->tlb_flush_pending); 579 * 580 * Where the increment if constrained by the PTL unlock, it thus 581 * ensures that the increment is visible if the PTE modification is 582 * visible. After all, if there is no PTE modification, nobody cares 583 * about TLB flushes either. 584 * 585 * This very much relies on users (mm_tlb_flush_pending() and 586 * mm_tlb_flush_nested()) only caring about _specific_ PTEs (and 587 * therefore specific PTLs), because with SPLIT_PTE_PTLOCKS and RCpc 588 * locks (PPC) the unlock of one doesn't order against the lock of 589 * another PTL. 590 * 591 * The decrement is ordered by the flush_tlb_range(), such that 592 * mm_tlb_flush_pending() will not return false unless all flushes have 593 * completed. 594 */ 595} 596 597static inline void dec_tlb_flush_pending(struct mm_struct *mm) 598{ 599 /* 600 * See inc_tlb_flush_pending(). 601 * 602 * This cannot be smp_mb__before_atomic() because smp_mb() simply does 603 * not order against TLB invalidate completion, which is what we need. 604 * 605 * Therefore we must rely on tlb_flush_*() to guarantee order. 606 */ 607 atomic_dec(&mm->tlb_flush_pending); 608} 609 610static inline bool mm_tlb_flush_pending(struct mm_struct *mm) 611{ 612 /* 613 * Must be called after having acquired the PTL; orders against that 614 * PTLs release and therefore ensures that if we observe the modified 615 * PTE we must also observe the increment from inc_tlb_flush_pending(). 616 * 617 * That is, it only guarantees to return true if there is a flush 618 * pending for _this_ PTL. 619 */ 620 return atomic_read(&mm->tlb_flush_pending); 621} 622 623static inline bool mm_tlb_flush_nested(struct mm_struct *mm) 624{ 625 /* 626 * Similar to mm_tlb_flush_pending(), we must have acquired the PTL 627 * for which there is a TLB flush pending in order to guarantee 628 * we've seen both that PTE modification and the increment. 629 * 630 * (no requirement on actually still holding the PTL, that is irrelevant) 631 */ 632 return atomic_read(&mm->tlb_flush_pending) > 1; 633} 634 635struct vm_fault; 636 637/** 638 * typedef vm_fault_t - Return type for page fault handlers. 639 * 640 * Page fault handlers return a bitmask of %VM_FAULT values. 641 */ 642typedef __bitwise unsigned int vm_fault_t; 643 644/** 645 * enum vm_fault_reason - Page fault handlers return a bitmask of 646 * these values to tell the core VM what happened when handling the 647 * fault. Used to decide whether a process gets delivered SIGBUS or 648 * just gets major/minor fault counters bumped up. 649 * 650 * @VM_FAULT_OOM: Out Of Memory 651 * @VM_FAULT_SIGBUS: Bad access 652 * @VM_FAULT_MAJOR: Page read from storage 653 * @VM_FAULT_WRITE: Special case for get_user_pages 654 * @VM_FAULT_HWPOISON: Hit poisoned small page 655 * @VM_FAULT_HWPOISON_LARGE: Hit poisoned large page. Index encoded 656 * in upper bits 657 * @VM_FAULT_SIGSEGV: segmentation fault 658 * @VM_FAULT_NOPAGE: ->fault installed the pte, not return page 659 * @VM_FAULT_LOCKED: ->fault locked the returned page 660 * @VM_FAULT_RETRY: ->fault blocked, must retry 661 * @VM_FAULT_FALLBACK: huge page fault failed, fall back to small 662 * @VM_FAULT_DONE_COW: ->fault has fully handled COW 663 * @VM_FAULT_NEEDDSYNC: ->fault did not modify page tables and needs 664 * fsync() to complete (for synchronous page faults 665 * in DAX) 666 * @VM_FAULT_HINDEX_MASK: mask HINDEX value 667 * 668 */ 669enum vm_fault_reason { 670 VM_FAULT_OOM = (__force vm_fault_t)0x000001, 671 VM_FAULT_SIGBUS = (__force vm_fault_t)0x000002, 672 VM_FAULT_MAJOR = (__force vm_fault_t)0x000004, 673 VM_FAULT_WRITE = (__force vm_fault_t)0x000008, 674 VM_FAULT_HWPOISON = (__force vm_fault_t)0x000010, 675 VM_FAULT_HWPOISON_LARGE = (__force vm_fault_t)0x000020, 676 VM_FAULT_SIGSEGV = (__force vm_fault_t)0x000040, 677 VM_FAULT_NOPAGE = (__force vm_fault_t)0x000100, 678 VM_FAULT_LOCKED = (__force vm_fault_t)0x000200, 679 VM_FAULT_RETRY = (__force vm_fault_t)0x000400, 680 VM_FAULT_FALLBACK = (__force vm_fault_t)0x000800, 681 VM_FAULT_DONE_COW = (__force vm_fault_t)0x001000, 682 VM_FAULT_NEEDDSYNC = (__force vm_fault_t)0x002000, 683 VM_FAULT_HINDEX_MASK = (__force vm_fault_t)0x0f0000, 684}; 685 686/* Encode hstate index for a hwpoisoned large page */ 687#define VM_FAULT_SET_HINDEX(x) ((__force vm_fault_t)((x) << 16)) 688#define VM_FAULT_GET_HINDEX(x) (((__force unsigned int)(x) >> 16) & 0xf) 689 690#define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | \ 691 VM_FAULT_SIGSEGV | VM_FAULT_HWPOISON | \ 692 VM_FAULT_HWPOISON_LARGE | VM_FAULT_FALLBACK) 693 694#define VM_FAULT_RESULT_TRACE \ 695 { VM_FAULT_OOM, "OOM" }, \ 696 { VM_FAULT_SIGBUS, "SIGBUS" }, \ 697 { VM_FAULT_MAJOR, "MAJOR" }, \ 698 { VM_FAULT_WRITE, "WRITE" }, \ 699 { VM_FAULT_HWPOISON, "HWPOISON" }, \ 700 { VM_FAULT_HWPOISON_LARGE, "HWPOISON_LARGE" }, \ 701 { VM_FAULT_SIGSEGV, "SIGSEGV" }, \ 702 { VM_FAULT_NOPAGE, "NOPAGE" }, \ 703 { VM_FAULT_LOCKED, "LOCKED" }, \ 704 { VM_FAULT_RETRY, "RETRY" }, \ 705 { VM_FAULT_FALLBACK, "FALLBACK" }, \ 706 { VM_FAULT_DONE_COW, "DONE_COW" }, \ 707 { VM_FAULT_NEEDDSYNC, "NEEDDSYNC" } 708 709struct vm_special_mapping { 710 const char *name; /* The name, e.g. "[vdso]". */ 711 712 /* 713 * If .fault is not provided, this points to a 714 * NULL-terminated array of pages that back the special mapping. 715 * 716 * This must not be NULL unless .fault is provided. 717 */ 718 struct page **pages; 719 720 /* 721 * If non-NULL, then this is called to resolve page faults 722 * on the special mapping. If used, .pages is not checked. 723 */ 724 vm_fault_t (*fault)(const struct vm_special_mapping *sm, 725 struct vm_area_struct *vma, 726 struct vm_fault *vmf); 727 728 int (*mremap)(const struct vm_special_mapping *sm, 729 struct vm_area_struct *new_vma); 730}; 731 732enum tlb_flush_reason { 733 TLB_FLUSH_ON_TASK_SWITCH, 734 TLB_REMOTE_SHOOTDOWN, 735 TLB_LOCAL_SHOOTDOWN, 736 TLB_LOCAL_MM_SHOOTDOWN, 737 TLB_REMOTE_SEND_IPI, 738 NR_TLB_FLUSH_REASONS, 739}; 740 741 /* 742 * A swap entry has to fit into a "unsigned long", as the entry is hidden 743 * in the "index" field of the swapper address space. 744 */ 745typedef struct { 746 unsigned long val; 747} swp_entry_t; 748 749#endif /* _LINUX_MM_TYPES_H */ 750