Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1 | #ifndef _LINUX_MM_H |
| 2 | #define _LINUX_MM_H |
| 3 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4 | #include <linux/errno.h> |
| 5 | |
| 6 | #ifdef __KERNEL__ |
| 7 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 8 | #include <linux/gfp.h> |
| 9 | #include <linux/list.h> |
| 10 | #include <linux/mmzone.h> |
| 11 | #include <linux/rbtree.h> |
| 12 | #include <linux/prio_tree.h> |
Ingo Molnar | 9a11b49a | 2006-07-03 00:24:33 -0700 | [diff] [blame] | 13 | #include <linux/debug_locks.h> |
Heiko Carstens | 5b99cd0 | 2006-09-27 01:50:01 -0700 | [diff] [blame] | 14 | #include <linux/mm_types.h> |
Yinghai Lu | 0867721 | 2010-02-10 01:20:20 -0800 | [diff] [blame] | 15 | #include <linux/range.h> |
Chris Metcalf | c6f6b59 | 2010-05-24 14:32:53 -0700 | [diff] [blame] | 16 | #include <linux/pfn.h> |
Andrea Arcangeli | e9da73d | 2011-01-13 15:46:32 -0800 | [diff] [blame] | 17 | #include <linux/bit_spinlock.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 18 | |
| 19 | struct mempolicy; |
| 20 | struct anon_vma; |
Alexey Dobriyan | 4e950f6 | 2007-07-30 02:36:13 +0400 | [diff] [blame] | 21 | struct file_ra_state; |
Alexey Dobriyan | e8edc6e | 2007-05-21 01:22:52 +0400 | [diff] [blame] | 22 | struct user_struct; |
Alexey Dobriyan | 4e950f6 | 2007-07-30 02:36:13 +0400 | [diff] [blame] | 23 | struct writeback_control; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 24 | |
| 25 | #ifndef CONFIG_DISCONTIGMEM /* Don't use mapnrs, do it properly */ |
| 26 | extern unsigned long max_mapnr; |
| 27 | #endif |
| 28 | |
| 29 | extern unsigned long num_physpages; |
Jan Beulich | 4481374 | 2009-09-21 17:03:05 -0700 | [diff] [blame] | 30 | extern unsigned long totalram_pages; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 31 | extern void * high_memory; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 32 | extern int page_cluster; |
| 33 | |
| 34 | #ifdef CONFIG_SYSCTL |
| 35 | extern int sysctl_legacy_va_layout; |
| 36 | #else |
| 37 | #define sysctl_legacy_va_layout 0 |
| 38 | #endif |
| 39 | |
| 40 | #include <asm/page.h> |
| 41 | #include <asm/pgtable.h> |
| 42 | #include <asm/processor.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 43 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 44 | #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n)) |
| 45 | |
Andrea Righi | 27ac792 | 2008-07-23 21:28:13 -0700 | [diff] [blame] | 46 | /* to align the pointer to the (next) page boundary */ |
| 47 | #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE) |
| 48 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 49 | /* |
| 50 | * Linux kernel virtual memory manager primitives. |
| 51 | * The idea being to have a "virtual" mm in the same way |
| 52 | * we have a virtual fs - giving a cleaner interface to the |
| 53 | * mm details, and allowing different kinds of memory mappings |
| 54 | * (from shared memory to executable loading to arbitrary |
| 55 | * mmap() functions). |
| 56 | */ |
| 57 | |
Christoph Lameter | c43692e | 2006-12-06 20:32:48 -0800 | [diff] [blame] | 58 | extern struct kmem_cache *vm_area_cachep; |
| 59 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 60 | #ifndef CONFIG_MMU |
David Howells | 8feae13 | 2009-01-08 12:04:47 +0000 | [diff] [blame] | 61 | extern struct rb_root nommu_region_tree; |
| 62 | extern struct rw_semaphore nommu_region_sem; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 63 | |
| 64 | extern unsigned int kobjsize(const void *objp); |
| 65 | #endif |
| 66 | |
| 67 | /* |
Hugh Dickins | 605d928 | 2008-08-16 11:07:21 +0100 | [diff] [blame] | 68 | * vm_flags in vm_area_struct, see mm_types.h. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 69 | */ |
| 70 | #define VM_READ 0x00000001 /* currently active flags */ |
| 71 | #define VM_WRITE 0x00000002 |
| 72 | #define VM_EXEC 0x00000004 |
| 73 | #define VM_SHARED 0x00000008 |
| 74 | |
Paolo 'Blaisorblade' Giarrusso | 7e2cff4 | 2005-09-21 09:55:39 -0700 | [diff] [blame] | 75 | /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 76 | #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */ |
| 77 | #define VM_MAYWRITE 0x00000020 |
| 78 | #define VM_MAYEXEC 0x00000040 |
| 79 | #define VM_MAYSHARE 0x00000080 |
| 80 | |
| 81 | #define VM_GROWSDOWN 0x00000100 /* general info on the segment */ |
Luck, Tony | 8ca3eb0 | 2010-08-24 11:44:18 -0700 | [diff] [blame] | 82 | #if defined(CONFIG_STACK_GROWSUP) || defined(CONFIG_IA64) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 83 | #define VM_GROWSUP 0x00000200 |
Luck, Tony | 8ca3eb0 | 2010-08-24 11:44:18 -0700 | [diff] [blame] | 84 | #else |
| 85 | #define VM_GROWSUP 0x00000000 |
Andrea Arcangeli | a664b2d | 2011-01-13 15:47:17 -0800 | [diff] [blame] | 86 | #define VM_NOHUGEPAGE 0x00000200 /* MADV_NOHUGEPAGE marked this vma */ |
Luck, Tony | 8ca3eb0 | 2010-08-24 11:44:18 -0700 | [diff] [blame] | 87 | #endif |
Linus Torvalds | 6aab341 | 2005-11-28 14:34:23 -0800 | [diff] [blame] | 88 | #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 89 | #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */ |
| 90 | |
| 91 | #define VM_EXECUTABLE 0x00001000 |
| 92 | #define VM_LOCKED 0x00002000 |
| 93 | #define VM_IO 0x00004000 /* Memory mapped I/O or similar */ |
| 94 | |
| 95 | /* Used by sys_madvise() */ |
| 96 | #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */ |
| 97 | #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */ |
| 98 | |
| 99 | #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */ |
| 100 | #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */ |
Hugh Dickins | 0b14c17 | 2005-11-21 21:32:15 -0800 | [diff] [blame] | 101 | #define VM_RESERVED 0x00080000 /* Count as reserved_vm like IO */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 102 | #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */ |
Andy Whitcroft | cdfd432 | 2008-07-23 21:27:28 -0700 | [diff] [blame] | 103 | #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 104 | #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */ |
| 105 | #define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */ |
Johannes Weiner | f2d6bfe | 2011-01-13 15:47:01 -0800 | [diff] [blame] | 106 | #ifndef CONFIG_TRANSPARENT_HUGEPAGE |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 107 | #define VM_MAPPED_COPY 0x01000000 /* T if mapped copy of data (nommu mmap) */ |
Johannes Weiner | f2d6bfe | 2011-01-13 15:47:01 -0800 | [diff] [blame] | 108 | #else |
| 109 | #define VM_HUGEPAGE 0x01000000 /* MADV_HUGEPAGE marked this vma */ |
| 110 | #endif |
Pallipadi, Venkatesh | 895791d | 2009-03-13 16:35:44 -0700 | [diff] [blame] | 111 | #define VM_INSERTPAGE 0x02000000 /* The vma has had "vm_insert_page()" done on it */ |
Roland McGrath | e5b97dd | 2007-01-26 00:56:48 -0800 | [diff] [blame] | 112 | #define VM_ALWAYSDUMP 0x04000000 /* Always include in core dumps */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 113 | |
Nick Piggin | d0217ac | 2007-07-19 01:47:03 -0700 | [diff] [blame] | 114 | #define VM_CAN_NONLINEAR 0x08000000 /* Has ->fault & does nonlinear pages */ |
Jared Hulbert | b379d79 | 2008-04-28 02:12:58 -0700 | [diff] [blame] | 115 | #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */ |
Dave Kleikamp | aba46c5 | 2008-07-08 00:28:52 +1000 | [diff] [blame] | 116 | #define VM_SAO 0x20000000 /* Strong Access Ordering (powerpc) */ |
Pallipadi, Venkatesh | 895791d | 2009-03-13 16:35:44 -0700 | [diff] [blame] | 117 | #define VM_PFN_AT_MMAP 0x40000000 /* PFNMAP vma that is fully mapped at mmap time */ |
Hugh Dickins | f8af4da | 2009-09-21 17:01:57 -0700 | [diff] [blame] | 118 | #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */ |
Nick Piggin | d00806b | 2007-07-19 01:46:57 -0700 | [diff] [blame] | 119 | |
Mel Gorman | a8bef8ff | 2010-05-24 14:32:24 -0700 | [diff] [blame] | 120 | /* Bits set in the VMA until the stack is in its final location */ |
| 121 | #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ) |
| 122 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 123 | #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */ |
| 124 | #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS |
| 125 | #endif |
| 126 | |
| 127 | #ifdef CONFIG_STACK_GROWSUP |
| 128 | #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT) |
| 129 | #else |
| 130 | #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT) |
| 131 | #endif |
| 132 | |
| 133 | #define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ) |
| 134 | #define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK |
| 135 | #define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK)) |
| 136 | #define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ) |
| 137 | #define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ) |
| 138 | |
| 139 | /* |
Andrea Arcangeli | 78f11a2 | 2011-04-27 15:26:45 -0700 | [diff] [blame] | 140 | * Special vmas that are non-mergable, non-mlock()able. |
| 141 | * Note: mm/huge_memory.c VM_NO_THP depends on this definition. |
Nick Piggin | b291f00 | 2008-10-18 20:26:44 -0700 | [diff] [blame] | 142 | */ |
| 143 | #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_RESERVED | VM_PFNMAP) |
| 144 | |
| 145 | /* |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 146 | * mapping from the currently active vm_flags protection bits (the |
| 147 | * low four bits) to a page protection mask.. |
| 148 | */ |
| 149 | extern pgprot_t protection_map[16]; |
| 150 | |
Nick Piggin | d0217ac | 2007-07-19 01:47:03 -0700 | [diff] [blame] | 151 | #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */ |
| 152 | #define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */ |
Nick Piggin | c2ec175 | 2009-03-31 15:23:21 -0700 | [diff] [blame] | 153 | #define FAULT_FLAG_MKWRITE 0x04 /* Fault was mkwrite of existing pte */ |
Michel Lespinasse | d065bd8 | 2010-10-26 14:21:57 -0700 | [diff] [blame] | 154 | #define FAULT_FLAG_ALLOW_RETRY 0x08 /* Retry fault if blocking */ |
Gleb Natapov | 318b275 | 2011-03-22 16:30:51 -0700 | [diff] [blame] | 155 | #define FAULT_FLAG_RETRY_NOWAIT 0x10 /* Don't drop mmap_sem and wait when retrying */ |
Nick Piggin | d0217ac | 2007-07-19 01:47:03 -0700 | [diff] [blame] | 156 | |
venkatesh.pallipadi@intel.com | 6bd9cd5 | 2008-12-19 13:47:26 -0800 | [diff] [blame] | 157 | /* |
| 158 | * This interface is used by x86 PAT code to identify a pfn mapping that is |
| 159 | * linear over entire vma. This is to optimize PAT code that deals with |
| 160 | * marking the physical region with a particular prot. This is not for generic |
| 161 | * mm use. Note also that this check will not work if the pfn mapping is |
| 162 | * linear for a vma starting at physical address 0. In which case PAT code |
| 163 | * falls back to slow path of reserving physical range page by page. |
| 164 | */ |
venkatesh.pallipadi@intel.com | 3c8bb73 | 2008-12-18 11:41:27 -0800 | [diff] [blame] | 165 | static inline int is_linear_pfn_mapping(struct vm_area_struct *vma) |
| 166 | { |
Pallipadi, Venkatesh | 895791d | 2009-03-13 16:35:44 -0700 | [diff] [blame] | 167 | return (vma->vm_flags & VM_PFN_AT_MMAP); |
venkatesh.pallipadi@intel.com | 3c8bb73 | 2008-12-18 11:41:27 -0800 | [diff] [blame] | 168 | } |
| 169 | |
| 170 | static inline int is_pfn_mapping(struct vm_area_struct *vma) |
| 171 | { |
| 172 | return (vma->vm_flags & VM_PFNMAP); |
| 173 | } |
Nick Piggin | d0217ac | 2007-07-19 01:47:03 -0700 | [diff] [blame] | 174 | |
Nick Piggin | 54cb882 | 2007-07-19 01:46:59 -0700 | [diff] [blame] | 175 | /* |
Nick Piggin | d0217ac | 2007-07-19 01:47:03 -0700 | [diff] [blame] | 176 | * vm_fault is filled by the the pagefault handler and passed to the vma's |
Nick Piggin | 83c5407 | 2007-07-19 01:47:05 -0700 | [diff] [blame] | 177 | * ->fault function. The vma's ->fault is responsible for returning a bitmask |
| 178 | * of VM_FAULT_xxx flags that give details about how the fault was handled. |
Nick Piggin | 54cb882 | 2007-07-19 01:46:59 -0700 | [diff] [blame] | 179 | * |
Nick Piggin | d0217ac | 2007-07-19 01:47:03 -0700 | [diff] [blame] | 180 | * pgoff should be used in favour of virtual_address, if possible. If pgoff |
| 181 | * is used, one may set VM_CAN_NONLINEAR in the vma->vm_flags to get nonlinear |
| 182 | * mapping support. |
Nick Piggin | 54cb882 | 2007-07-19 01:46:59 -0700 | [diff] [blame] | 183 | */ |
Nick Piggin | d0217ac | 2007-07-19 01:47:03 -0700 | [diff] [blame] | 184 | struct vm_fault { |
| 185 | unsigned int flags; /* FAULT_FLAG_xxx flags */ |
| 186 | pgoff_t pgoff; /* Logical page offset based on vma */ |
| 187 | void __user *virtual_address; /* Faulting virtual address */ |
| 188 | |
| 189 | struct page *page; /* ->fault handlers should return a |
Nick Piggin | 83c5407 | 2007-07-19 01:47:05 -0700 | [diff] [blame] | 190 | * page here, unless VM_FAULT_NOPAGE |
Nick Piggin | d0217ac | 2007-07-19 01:47:03 -0700 | [diff] [blame] | 191 | * is set (which is also implied by |
Nick Piggin | 83c5407 | 2007-07-19 01:47:05 -0700 | [diff] [blame] | 192 | * VM_FAULT_ERROR). |
Nick Piggin | d0217ac | 2007-07-19 01:47:03 -0700 | [diff] [blame] | 193 | */ |
Nick Piggin | 54cb882 | 2007-07-19 01:46:59 -0700 | [diff] [blame] | 194 | }; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 195 | |
| 196 | /* |
| 197 | * These are the virtual MM functions - opening of an area, closing and |
| 198 | * unmapping it (needed to keep files on disk up-to-date etc), pointer |
| 199 | * to the functions called when a no-page or a wp-page exception occurs. |
| 200 | */ |
| 201 | struct vm_operations_struct { |
| 202 | void (*open)(struct vm_area_struct * area); |
| 203 | void (*close)(struct vm_area_struct * area); |
Nick Piggin | d0217ac | 2007-07-19 01:47:03 -0700 | [diff] [blame] | 204 | int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf); |
David Howells | 9637a5e | 2006-06-23 02:03:43 -0700 | [diff] [blame] | 205 | |
| 206 | /* notification that a previously read-only page is about to become |
| 207 | * writable, if an error is returned it will cause a SIGBUS */ |
Nick Piggin | c2ec175 | 2009-03-31 15:23:21 -0700 | [diff] [blame] | 208 | int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf); |
Rik van Riel | 28b2ee2 | 2008-07-23 21:27:05 -0700 | [diff] [blame] | 209 | |
| 210 | /* called by access_process_vm when get_user_pages() fails, typically |
| 211 | * for use by special VMAs that can switch between memory and hardware |
| 212 | */ |
| 213 | int (*access)(struct vm_area_struct *vma, unsigned long addr, |
| 214 | void *buf, int len, int write); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 215 | #ifdef CONFIG_NUMA |
Lee Schermerhorn | a6020ed | 2008-04-28 02:13:14 -0700 | [diff] [blame] | 216 | /* |
| 217 | * set_policy() op must add a reference to any non-NULL @new mempolicy |
| 218 | * to hold the policy upon return. Caller should pass NULL @new to |
| 219 | * remove a policy and fall back to surrounding context--i.e. do not |
| 220 | * install a MPOL_DEFAULT policy, nor the task or system default |
| 221 | * mempolicy. |
| 222 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 223 | int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new); |
Lee Schermerhorn | a6020ed | 2008-04-28 02:13:14 -0700 | [diff] [blame] | 224 | |
| 225 | /* |
| 226 | * get_policy() op must add reference [mpol_get()] to any policy at |
| 227 | * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure |
| 228 | * in mm/mempolicy.c will do this automatically. |
| 229 | * get_policy() must NOT add a ref if the policy at (vma,addr) is not |
| 230 | * marked as MPOL_SHARED. vma policies are protected by the mmap_sem. |
| 231 | * If no [shared/vma] mempolicy exists at the addr, get_policy() op |
| 232 | * must return NULL--i.e., do not "fallback" to task or system default |
| 233 | * policy. |
| 234 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 235 | struct mempolicy *(*get_policy)(struct vm_area_struct *vma, |
| 236 | unsigned long addr); |
Christoph Lameter | 7b2259b | 2006-06-25 05:46:48 -0700 | [diff] [blame] | 237 | int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from, |
| 238 | const nodemask_t *to, unsigned long flags); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 239 | #endif |
| 240 | }; |
| 241 | |
| 242 | struct mmu_gather; |
| 243 | struct inode; |
| 244 | |
Andrew Morton | 349aef0 | 2006-01-08 01:04:36 -0800 | [diff] [blame] | 245 | #define page_private(page) ((page)->private) |
| 246 | #define set_page_private(page, v) ((page)->private = (v)) |
Hugh Dickins | 4c21e2f | 2005-10-29 18:16:40 -0700 | [diff] [blame] | 247 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 248 | /* |
| 249 | * FIXME: take this include out, include page-flags.h in |
| 250 | * files which need it (119 of them) |
| 251 | */ |
| 252 | #include <linux/page-flags.h> |
Andrea Arcangeli | 71e3aac | 2011-01-13 15:46:52 -0800 | [diff] [blame] | 253 | #include <linux/huge_mm.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 254 | |
| 255 | /* |
| 256 | * Methods to modify the page usage count. |
| 257 | * |
| 258 | * What counts for a page usage: |
| 259 | * - cache mapping (page->mapping) |
| 260 | * - private data (page->private) |
| 261 | * - page mapped in a task's page tables, each mapping |
| 262 | * is counted separately |
| 263 | * |
| 264 | * Also, many kernel routines increase the page count before a critical |
| 265 | * routine so they can be sure the page doesn't go away from under them. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 266 | */ |
| 267 | |
| 268 | /* |
Nick Piggin | da6052f | 2006-09-25 23:31:35 -0700 | [diff] [blame] | 269 | * Drop a ref, return true if the refcount fell to zero (the page has no users) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 270 | */ |
Nick Piggin | 7c8ee9a | 2006-03-22 00:08:03 -0800 | [diff] [blame] | 271 | static inline int put_page_testzero(struct page *page) |
| 272 | { |
Nick Piggin | 725d704 | 2006-09-25 23:30:55 -0700 | [diff] [blame] | 273 | VM_BUG_ON(atomic_read(&page->_count) == 0); |
Nick Piggin | 8dc04ef | 2006-03-22 00:08:03 -0800 | [diff] [blame] | 274 | return atomic_dec_and_test(&page->_count); |
Nick Piggin | 7c8ee9a | 2006-03-22 00:08:03 -0800 | [diff] [blame] | 275 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 276 | |
| 277 | /* |
Nick Piggin | 7c8ee9a | 2006-03-22 00:08:03 -0800 | [diff] [blame] | 278 | * Try to grab a ref unless the page has a refcount of zero, return false if |
| 279 | * that is the case. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 280 | */ |
Nick Piggin | 7c8ee9a | 2006-03-22 00:08:03 -0800 | [diff] [blame] | 281 | static inline int get_page_unless_zero(struct page *page) |
| 282 | { |
Nick Piggin | 8dc04ef | 2006-03-22 00:08:03 -0800 | [diff] [blame] | 283 | return atomic_inc_not_zero(&page->_count); |
Nick Piggin | 7c8ee9a | 2006-03-22 00:08:03 -0800 | [diff] [blame] | 284 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 285 | |
Wu Fengguang | 53df8fd | 2010-01-27 11:06:39 +0800 | [diff] [blame] | 286 | extern int page_is_ram(unsigned long pfn); |
| 287 | |
Christoph Lameter | 48667e7 | 2008-02-04 22:28:31 -0800 | [diff] [blame] | 288 | /* Support for virtually mapped pages */ |
Christoph Lameter | b3bdda0 | 2008-02-04 22:28:32 -0800 | [diff] [blame] | 289 | struct page *vmalloc_to_page(const void *addr); |
| 290 | unsigned long vmalloc_to_pfn(const void *addr); |
Christoph Lameter | 48667e7 | 2008-02-04 22:28:31 -0800 | [diff] [blame] | 291 | |
Paul Mundt | 0738c4b | 2008-03-12 16:51:31 +0900 | [diff] [blame] | 292 | /* |
| 293 | * Determine if an address is within the vmalloc range |
| 294 | * |
| 295 | * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there |
| 296 | * is no special casing required. |
| 297 | */ |
Christoph Lameter | 9e2779f | 2008-02-04 22:28:34 -0800 | [diff] [blame] | 298 | static inline int is_vmalloc_addr(const void *x) |
| 299 | { |
Paul Mundt | 0738c4b | 2008-03-12 16:51:31 +0900 | [diff] [blame] | 300 | #ifdef CONFIG_MMU |
Christoph Lameter | 9e2779f | 2008-02-04 22:28:34 -0800 | [diff] [blame] | 301 | unsigned long addr = (unsigned long)x; |
| 302 | |
| 303 | return addr >= VMALLOC_START && addr < VMALLOC_END; |
Paul Mundt | 0738c4b | 2008-03-12 16:51:31 +0900 | [diff] [blame] | 304 | #else |
| 305 | return 0; |
David Howells | 8ca3ed8 | 2008-02-23 15:23:37 -0800 | [diff] [blame] | 306 | #endif |
Paul Mundt | 0738c4b | 2008-03-12 16:51:31 +0900 | [diff] [blame] | 307 | } |
KAMEZAWA Hiroyuki | 81ac3ad | 2009-09-22 16:45:49 -0700 | [diff] [blame] | 308 | #ifdef CONFIG_MMU |
| 309 | extern int is_vmalloc_or_module_addr(const void *x); |
| 310 | #else |
David Howells | 934831d | 2009-09-24 12:33:32 +0100 | [diff] [blame] | 311 | static inline int is_vmalloc_or_module_addr(const void *x) |
KAMEZAWA Hiroyuki | 81ac3ad | 2009-09-22 16:45:49 -0700 | [diff] [blame] | 312 | { |
| 313 | return 0; |
| 314 | } |
| 315 | #endif |
Christoph Lameter | 9e2779f | 2008-02-04 22:28:34 -0800 | [diff] [blame] | 316 | |
Andrea Arcangeli | e9da73d | 2011-01-13 15:46:32 -0800 | [diff] [blame] | 317 | static inline void compound_lock(struct page *page) |
| 318 | { |
| 319 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
| 320 | bit_spin_lock(PG_compound_lock, &page->flags); |
| 321 | #endif |
| 322 | } |
| 323 | |
| 324 | static inline void compound_unlock(struct page *page) |
| 325 | { |
| 326 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
| 327 | bit_spin_unlock(PG_compound_lock, &page->flags); |
| 328 | #endif |
| 329 | } |
| 330 | |
| 331 | static inline unsigned long compound_lock_irqsave(struct page *page) |
| 332 | { |
| 333 | unsigned long uninitialized_var(flags); |
| 334 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
| 335 | local_irq_save(flags); |
| 336 | compound_lock(page); |
| 337 | #endif |
| 338 | return flags; |
| 339 | } |
| 340 | |
| 341 | static inline void compound_unlock_irqrestore(struct page *page, |
| 342 | unsigned long flags) |
| 343 | { |
| 344 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
| 345 | compound_unlock(page); |
| 346 | local_irq_restore(flags); |
| 347 | #endif |
| 348 | } |
| 349 | |
Christoph Lameter | d85f338 | 2007-05-06 14:49:39 -0700 | [diff] [blame] | 350 | static inline struct page *compound_head(struct page *page) |
| 351 | { |
Christoph Lameter | 6d77795 | 2007-05-06 14:49:40 -0700 | [diff] [blame] | 352 | if (unlikely(PageTail(page))) |
Christoph Lameter | d85f338 | 2007-05-06 14:49:39 -0700 | [diff] [blame] | 353 | return page->first_page; |
| 354 | return page; |
| 355 | } |
| 356 | |
Hugh Dickins | 4c21e2f | 2005-10-29 18:16:40 -0700 | [diff] [blame] | 357 | static inline int page_count(struct page *page) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 358 | { |
Christoph Lameter | d85f338 | 2007-05-06 14:49:39 -0700 | [diff] [blame] | 359 | return atomic_read(&compound_head(page)->_count); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 360 | } |
| 361 | |
| 362 | static inline void get_page(struct page *page) |
| 363 | { |
Andrea Arcangeli | 9180706 | 2011-01-13 15:46:32 -0800 | [diff] [blame] | 364 | /* |
| 365 | * Getting a normal page or the head of a compound page |
| 366 | * requires to already have an elevated page->_count. Only if |
| 367 | * we're getting a tail page, the elevated page->_count is |
| 368 | * required only in the head page, so for tail pages the |
| 369 | * bugcheck only verifies that the page->_count isn't |
| 370 | * negative. |
| 371 | */ |
| 372 | VM_BUG_ON(atomic_read(&page->_count) < !PageTail(page)); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 373 | atomic_inc(&page->_count); |
Andrea Arcangeli | 9180706 | 2011-01-13 15:46:32 -0800 | [diff] [blame] | 374 | /* |
| 375 | * Getting a tail page will elevate both the head and tail |
| 376 | * page->_count(s). |
| 377 | */ |
| 378 | if (unlikely(PageTail(page))) { |
| 379 | /* |
| 380 | * This is safe only because |
| 381 | * __split_huge_page_refcount can't run under |
| 382 | * get_page(). |
| 383 | */ |
| 384 | VM_BUG_ON(atomic_read(&page->first_page->_count) <= 0); |
| 385 | atomic_inc(&page->first_page->_count); |
| 386 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 387 | } |
| 388 | |
Christoph Lameter | b49af68 | 2007-05-06 14:49:41 -0700 | [diff] [blame] | 389 | static inline struct page *virt_to_head_page(const void *x) |
| 390 | { |
| 391 | struct page *page = virt_to_page(x); |
| 392 | return compound_head(page); |
| 393 | } |
| 394 | |
Nick Piggin | 7835e98 | 2006-03-22 00:08:40 -0800 | [diff] [blame] | 395 | /* |
| 396 | * Setup the page count before being freed into the page allocator for |
| 397 | * the first time (boot or memory hotplug) |
| 398 | */ |
| 399 | static inline void init_page_count(struct page *page) |
| 400 | { |
| 401 | atomic_set(&page->_count, 1); |
| 402 | } |
| 403 | |
Andrea Arcangeli | 5f24ce5 | 2011-01-13 15:47:00 -0800 | [diff] [blame] | 404 | /* |
| 405 | * PageBuddy() indicate that the page is free and in the buddy system |
| 406 | * (see mm/page_alloc.c). |
Andrea Arcangeli | ef2b4b9 | 2011-03-18 00:16:35 +0100 | [diff] [blame] | 407 | * |
| 408 | * PAGE_BUDDY_MAPCOUNT_VALUE must be <= -2 but better not too close to |
| 409 | * -2 so that an underflow of the page_mapcount() won't be mistaken |
| 410 | * for a genuine PAGE_BUDDY_MAPCOUNT_VALUE. -128 can be created very |
| 411 | * efficiently by most CPU architectures. |
Andrea Arcangeli | 5f24ce5 | 2011-01-13 15:47:00 -0800 | [diff] [blame] | 412 | */ |
Andrea Arcangeli | ef2b4b9 | 2011-03-18 00:16:35 +0100 | [diff] [blame] | 413 | #define PAGE_BUDDY_MAPCOUNT_VALUE (-128) |
| 414 | |
Andrea Arcangeli | 5f24ce5 | 2011-01-13 15:47:00 -0800 | [diff] [blame] | 415 | static inline int PageBuddy(struct page *page) |
| 416 | { |
Andrea Arcangeli | ef2b4b9 | 2011-03-18 00:16:35 +0100 | [diff] [blame] | 417 | return atomic_read(&page->_mapcount) == PAGE_BUDDY_MAPCOUNT_VALUE; |
Andrea Arcangeli | 5f24ce5 | 2011-01-13 15:47:00 -0800 | [diff] [blame] | 418 | } |
| 419 | |
| 420 | static inline void __SetPageBuddy(struct page *page) |
| 421 | { |
| 422 | VM_BUG_ON(atomic_read(&page->_mapcount) != -1); |
Andrea Arcangeli | ef2b4b9 | 2011-03-18 00:16:35 +0100 | [diff] [blame] | 423 | atomic_set(&page->_mapcount, PAGE_BUDDY_MAPCOUNT_VALUE); |
Andrea Arcangeli | 5f24ce5 | 2011-01-13 15:47:00 -0800 | [diff] [blame] | 424 | } |
| 425 | |
| 426 | static inline void __ClearPageBuddy(struct page *page) |
| 427 | { |
| 428 | VM_BUG_ON(!PageBuddy(page)); |
| 429 | atomic_set(&page->_mapcount, -1); |
| 430 | } |
| 431 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 432 | void put_page(struct page *page); |
Alexander Zarochentsev | 1d7ea73 | 2006-08-13 23:24:27 -0700 | [diff] [blame] | 433 | void put_pages_list(struct list_head *pages); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 434 | |
Nick Piggin | 8dfcc9b | 2006-03-22 00:08:05 -0800 | [diff] [blame] | 435 | void split_page(struct page *page, unsigned int order); |
Mel Gorman | 748446b | 2010-05-24 14:32:27 -0700 | [diff] [blame] | 436 | int split_free_page(struct page *page); |
Nick Piggin | 8dfcc9b | 2006-03-22 00:08:05 -0800 | [diff] [blame] | 437 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 438 | /* |
Andy Whitcroft | 33f2ef8 | 2006-12-06 20:33:32 -0800 | [diff] [blame] | 439 | * Compound pages have a destructor function. Provide a |
| 440 | * prototype for that function and accessor functions. |
| 441 | * These are _only_ valid on the head of a PG_compound page. |
| 442 | */ |
| 443 | typedef void compound_page_dtor(struct page *); |
| 444 | |
| 445 | static inline void set_compound_page_dtor(struct page *page, |
| 446 | compound_page_dtor *dtor) |
| 447 | { |
| 448 | page[1].lru.next = (void *)dtor; |
| 449 | } |
| 450 | |
| 451 | static inline compound_page_dtor *get_compound_page_dtor(struct page *page) |
| 452 | { |
| 453 | return (compound_page_dtor *)page[1].lru.next; |
| 454 | } |
| 455 | |
Christoph Lameter | d85f338 | 2007-05-06 14:49:39 -0700 | [diff] [blame] | 456 | static inline int compound_order(struct page *page) |
| 457 | { |
Christoph Lameter | 6d77795 | 2007-05-06 14:49:40 -0700 | [diff] [blame] | 458 | if (!PageHead(page)) |
Christoph Lameter | d85f338 | 2007-05-06 14:49:39 -0700 | [diff] [blame] | 459 | return 0; |
| 460 | return (unsigned long)page[1].lru.prev; |
| 461 | } |
| 462 | |
Andrea Arcangeli | 37c2ac7 | 2011-01-13 15:47:16 -0800 | [diff] [blame] | 463 | static inline int compound_trans_order(struct page *page) |
| 464 | { |
| 465 | int order; |
| 466 | unsigned long flags; |
| 467 | |
| 468 | if (!PageHead(page)) |
| 469 | return 0; |
| 470 | |
| 471 | flags = compound_lock_irqsave(page); |
| 472 | order = compound_order(page); |
| 473 | compound_unlock_irqrestore(page, flags); |
| 474 | return order; |
| 475 | } |
| 476 | |
Christoph Lameter | d85f338 | 2007-05-06 14:49:39 -0700 | [diff] [blame] | 477 | static inline void set_compound_order(struct page *page, unsigned long order) |
| 478 | { |
| 479 | page[1].lru.prev = (void *)order; |
| 480 | } |
| 481 | |
Michal Simek | 3dece37 | 2011-01-21 08:49:56 +0100 | [diff] [blame] | 482 | #ifdef CONFIG_MMU |
Andy Whitcroft | 33f2ef8 | 2006-12-06 20:33:32 -0800 | [diff] [blame] | 483 | /* |
Andrea Arcangeli | 14fd403 | 2011-01-13 15:46:37 -0800 | [diff] [blame] | 484 | * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when |
| 485 | * servicing faults for write access. In the normal case, do always want |
| 486 | * pte_mkwrite. But get_user_pages can cause write faults for mappings |
| 487 | * that do not have writing enabled, when used by access_process_vm. |
| 488 | */ |
| 489 | static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma) |
| 490 | { |
| 491 | if (likely(vma->vm_flags & VM_WRITE)) |
| 492 | pte = pte_mkwrite(pte); |
| 493 | return pte; |
| 494 | } |
Michal Simek | 3dece37 | 2011-01-21 08:49:56 +0100 | [diff] [blame] | 495 | #endif |
Andrea Arcangeli | 14fd403 | 2011-01-13 15:46:37 -0800 | [diff] [blame] | 496 | |
| 497 | /* |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 498 | * Multiple processes may "see" the same page. E.g. for untouched |
| 499 | * mappings of /dev/null, all processes see the same page full of |
| 500 | * zeroes, and text pages of executables and shared libraries have |
| 501 | * only one copy in memory, at most, normally. |
| 502 | * |
| 503 | * For the non-reserved pages, page_count(page) denotes a reference count. |
Paolo 'Blaisorblade' Giarrusso | 7e871b6 | 2005-09-21 09:55:38 -0700 | [diff] [blame] | 504 | * page_count() == 0 means the page is free. page->lru is then used for |
| 505 | * freelist management in the buddy allocator. |
Nick Piggin | da6052f | 2006-09-25 23:31:35 -0700 | [diff] [blame] | 506 | * page_count() > 0 means the page has been allocated. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 507 | * |
Nick Piggin | da6052f | 2006-09-25 23:31:35 -0700 | [diff] [blame] | 508 | * Pages are allocated by the slab allocator in order to provide memory |
| 509 | * to kmalloc and kmem_cache_alloc. In this case, the management of the |
| 510 | * page, and the fields in 'struct page' are the responsibility of mm/slab.c |
| 511 | * unless a particular usage is carefully commented. (the responsibility of |
| 512 | * freeing the kmalloc memory is the caller's, of course). |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 513 | * |
Nick Piggin | da6052f | 2006-09-25 23:31:35 -0700 | [diff] [blame] | 514 | * A page may be used by anyone else who does a __get_free_page(). |
| 515 | * In this case, page_count still tracks the references, and should only |
| 516 | * be used through the normal accessor functions. The top bits of page->flags |
| 517 | * and page->virtual store page management information, but all other fields |
| 518 | * are unused and could be used privately, carefully. The management of this |
| 519 | * page is the responsibility of the one who allocated it, and those who have |
| 520 | * subsequently been given references to it. |
| 521 | * |
| 522 | * The other pages (we may call them "pagecache pages") are completely |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 523 | * managed by the Linux memory manager: I/O, buffers, swapping etc. |
| 524 | * The following discussion applies only to them. |
| 525 | * |
Nick Piggin | da6052f | 2006-09-25 23:31:35 -0700 | [diff] [blame] | 526 | * A pagecache page contains an opaque `private' member, which belongs to the |
| 527 | * page's address_space. Usually, this is the address of a circular list of |
| 528 | * the page's disk buffers. PG_private must be set to tell the VM to call |
| 529 | * into the filesystem to release these pages. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 530 | * |
Nick Piggin | da6052f | 2006-09-25 23:31:35 -0700 | [diff] [blame] | 531 | * A page may belong to an inode's memory mapping. In this case, page->mapping |
| 532 | * is the pointer to the inode, and page->index is the file offset of the page, |
| 533 | * in units of PAGE_CACHE_SIZE. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 534 | * |
Nick Piggin | da6052f | 2006-09-25 23:31:35 -0700 | [diff] [blame] | 535 | * If pagecache pages are not associated with an inode, they are said to be |
| 536 | * anonymous pages. These may become associated with the swapcache, and in that |
| 537 | * case PG_swapcache is set, and page->private is an offset into the swapcache. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 538 | * |
Nick Piggin | da6052f | 2006-09-25 23:31:35 -0700 | [diff] [blame] | 539 | * In either case (swapcache or inode backed), the pagecache itself holds one |
| 540 | * reference to the page. Setting PG_private should also increment the |
| 541 | * refcount. The each user mapping also has a reference to the page. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 542 | * |
Nick Piggin | da6052f | 2006-09-25 23:31:35 -0700 | [diff] [blame] | 543 | * The pagecache pages are stored in a per-mapping radix tree, which is |
| 544 | * rooted at mapping->page_tree, and indexed by offset. |
| 545 | * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space |
| 546 | * lists, we instead now tag pages as dirty/writeback in the radix tree. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 547 | * |
Nick Piggin | da6052f | 2006-09-25 23:31:35 -0700 | [diff] [blame] | 548 | * All pagecache pages may be subject to I/O: |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 549 | * - inode pages may need to be read from disk, |
| 550 | * - inode pages which have been modified and are MAP_SHARED may need |
Nick Piggin | da6052f | 2006-09-25 23:31:35 -0700 | [diff] [blame] | 551 | * to be written back to the inode on disk, |
| 552 | * - anonymous pages (including MAP_PRIVATE file mappings) which have been |
| 553 | * modified may need to be swapped out to swap space and (later) to be read |
| 554 | * back into memory. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 555 | */ |
| 556 | |
| 557 | /* |
| 558 | * The zone field is never updated after free_area_init_core() |
| 559 | * sets it, so none of the operations on it need to be atomic. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 560 | */ |
Dave Hansen | 348f8b6 | 2005-06-23 00:07:40 -0700 | [diff] [blame] | 561 | |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 562 | |
| 563 | /* |
| 564 | * page->flags layout: |
| 565 | * |
| 566 | * There are three possibilities for how page->flags get |
| 567 | * laid out. The first is for the normal case, without |
| 568 | * sparsemem. The second is for sparsemem when there is |
| 569 | * plenty of space for node and section. The last is when |
| 570 | * we have run out of space and have to fall back to an |
| 571 | * alternate (slower) way of determining the node. |
| 572 | * |
Christoph Lameter | 308c05e | 2008-04-28 02:12:43 -0700 | [diff] [blame] | 573 | * No sparsemem or sparsemem vmemmap: | NODE | ZONE | ... | FLAGS | |
| 574 | * classic sparse with space for node:| SECTION | NODE | ZONE | ... | FLAGS | |
| 575 | * classic sparse no space for node: | SECTION | ZONE | ... | FLAGS | |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 576 | */ |
Christoph Lameter | 308c05e | 2008-04-28 02:12:43 -0700 | [diff] [blame] | 577 | #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP) |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 578 | #define SECTIONS_WIDTH SECTIONS_SHIFT |
| 579 | #else |
| 580 | #define SECTIONS_WIDTH 0 |
| 581 | #endif |
| 582 | |
| 583 | #define ZONES_WIDTH ZONES_SHIFT |
| 584 | |
Christoph Lameter | 9223b41 | 2008-04-28 02:12:48 -0700 | [diff] [blame] | 585 | #if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= BITS_PER_LONG - NR_PAGEFLAGS |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 586 | #define NODES_WIDTH NODES_SHIFT |
| 587 | #else |
Christoph Lameter | 308c05e | 2008-04-28 02:12:43 -0700 | [diff] [blame] | 588 | #ifdef CONFIG_SPARSEMEM_VMEMMAP |
| 589 | #error "Vmemmap: No space for nodes field in page flags" |
| 590 | #endif |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 591 | #define NODES_WIDTH 0 |
| 592 | #endif |
| 593 | |
| 594 | /* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */ |
Andi Kleen | 07808b7 | 2005-11-05 17:25:53 +0100 | [diff] [blame] | 595 | #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH) |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 596 | #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH) |
| 597 | #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH) |
| 598 | |
| 599 | /* |
| 600 | * We are going to use the flags for the page to node mapping if its in |
| 601 | * there. This includes the case where there is no node, so it is implicit. |
| 602 | */ |
Christoph Lameter | 89689ae | 2006-12-06 20:31:45 -0800 | [diff] [blame] | 603 | #if !(NODES_WIDTH > 0 || NODES_SHIFT == 0) |
| 604 | #define NODE_NOT_IN_PAGE_FLAGS |
| 605 | #endif |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 606 | |
| 607 | #ifndef PFN_SECTION_SHIFT |
| 608 | #define PFN_SECTION_SHIFT 0 |
| 609 | #endif |
Dave Hansen | 348f8b6 | 2005-06-23 00:07:40 -0700 | [diff] [blame] | 610 | |
| 611 | /* |
Lucas De Marchi | 25985ed | 2011-03-30 22:57:33 -0300 | [diff] [blame] | 612 | * Define the bit shifts to access each section. For non-existent |
Dave Hansen | 348f8b6 | 2005-06-23 00:07:40 -0700 | [diff] [blame] | 613 | * sections we define the shift as 0; that plus a 0 mask ensures |
| 614 | * the compiler will optimise away reference to them. |
| 615 | */ |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 616 | #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0)) |
| 617 | #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0)) |
| 618 | #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0)) |
Dave Hansen | 348f8b6 | 2005-06-23 00:07:40 -0700 | [diff] [blame] | 619 | |
Will Deacon | bce54bb | 2010-10-26 14:21:37 -0700 | [diff] [blame] | 620 | /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */ |
| 621 | #ifdef NODE_NOT_IN_PAGE_FLAGS |
Christoph Lameter | 89689ae | 2006-12-06 20:31:45 -0800 | [diff] [blame] | 622 | #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT) |
Andy Whitcroft | bd8029b | 2007-02-10 01:43:14 -0800 | [diff] [blame] | 623 | #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \ |
| 624 | SECTIONS_PGOFF : ZONES_PGOFF) |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 625 | #else |
Christoph Lameter | 89689ae | 2006-12-06 20:31:45 -0800 | [diff] [blame] | 626 | #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT) |
Andy Whitcroft | bd8029b | 2007-02-10 01:43:14 -0800 | [diff] [blame] | 627 | #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \ |
| 628 | NODES_PGOFF : ZONES_PGOFF) |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 629 | #endif |
Christoph Lameter | 89689ae | 2006-12-06 20:31:45 -0800 | [diff] [blame] | 630 | |
Andy Whitcroft | bd8029b | 2007-02-10 01:43:14 -0800 | [diff] [blame] | 631 | #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0)) |
Dave Hansen | 348f8b6 | 2005-06-23 00:07:40 -0700 | [diff] [blame] | 632 | |
Christoph Lameter | 9223b41 | 2008-04-28 02:12:48 -0700 | [diff] [blame] | 633 | #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS |
| 634 | #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS |
Dave Hansen | 348f8b6 | 2005-06-23 00:07:40 -0700 | [diff] [blame] | 635 | #endif |
| 636 | |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 637 | #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1) |
| 638 | #define NODES_MASK ((1UL << NODES_WIDTH) - 1) |
| 639 | #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1) |
Christoph Lameter | 89689ae | 2006-12-06 20:31:45 -0800 | [diff] [blame] | 640 | #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1) |
Dave Hansen | 348f8b6 | 2005-06-23 00:07:40 -0700 | [diff] [blame] | 641 | |
Christoph Lameter | 2f1b624 | 2006-09-25 23:31:13 -0700 | [diff] [blame] | 642 | static inline enum zone_type page_zonenum(struct page *page) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 643 | { |
Dave Hansen | 348f8b6 | 2005-06-23 00:07:40 -0700 | [diff] [blame] | 644 | return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 645 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 646 | |
Christoph Lameter | 89689ae | 2006-12-06 20:31:45 -0800 | [diff] [blame] | 647 | /* |
| 648 | * The identification function is only used by the buddy allocator for |
| 649 | * determining if two pages could be buddies. We are not really |
| 650 | * identifying a zone since we could be using a the section number |
| 651 | * id if we have not node id available in page flags. |
| 652 | * We guarantee only that it will return the same value for two |
| 653 | * combinable pages in a zone. |
| 654 | */ |
Andy Whitcroft | cb2b95e | 2006-06-23 02:03:01 -0700 | [diff] [blame] | 655 | static inline int page_zone_id(struct page *page) |
| 656 | { |
Christoph Lameter | 89689ae | 2006-12-06 20:31:45 -0800 | [diff] [blame] | 657 | return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 658 | } |
| 659 | |
Andy Whitcroft | 25ba77c | 2006-12-06 20:33:03 -0800 | [diff] [blame] | 660 | static inline int zone_to_nid(struct zone *zone) |
Christoph Lameter | 89fa302 | 2006-09-25 23:31:55 -0700 | [diff] [blame] | 661 | { |
Christoph Lameter | d5f541e | 2006-09-27 01:50:08 -0700 | [diff] [blame] | 662 | #ifdef CONFIG_NUMA |
| 663 | return zone->node; |
| 664 | #else |
| 665 | return 0; |
| 666 | #endif |
Christoph Lameter | 89fa302 | 2006-09-25 23:31:55 -0700 | [diff] [blame] | 667 | } |
| 668 | |
Christoph Lameter | 89689ae | 2006-12-06 20:31:45 -0800 | [diff] [blame] | 669 | #ifdef NODE_NOT_IN_PAGE_FLAGS |
Andy Whitcroft | 25ba77c | 2006-12-06 20:33:03 -0800 | [diff] [blame] | 670 | extern int page_to_nid(struct page *page); |
Christoph Lameter | 89689ae | 2006-12-06 20:31:45 -0800 | [diff] [blame] | 671 | #else |
Andy Whitcroft | 25ba77c | 2006-12-06 20:33:03 -0800 | [diff] [blame] | 672 | static inline int page_to_nid(struct page *page) |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 673 | { |
Christoph Lameter | 89689ae | 2006-12-06 20:31:45 -0800 | [diff] [blame] | 674 | return (page->flags >> NODES_PGSHIFT) & NODES_MASK; |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 675 | } |
Christoph Lameter | 89689ae | 2006-12-06 20:31:45 -0800 | [diff] [blame] | 676 | #endif |
| 677 | |
| 678 | static inline struct zone *page_zone(struct page *page) |
| 679 | { |
| 680 | return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)]; |
| 681 | } |
| 682 | |
Christoph Lameter | 308c05e | 2008-04-28 02:12:43 -0700 | [diff] [blame] | 683 | #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP) |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 684 | static inline unsigned long page_to_section(struct page *page) |
| 685 | { |
| 686 | return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK; |
| 687 | } |
Christoph Lameter | 308c05e | 2008-04-28 02:12:43 -0700 | [diff] [blame] | 688 | #endif |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 689 | |
Christoph Lameter | 2f1b624 | 2006-09-25 23:31:13 -0700 | [diff] [blame] | 690 | static inline void set_page_zone(struct page *page, enum zone_type zone) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 691 | { |
Dave Hansen | 348f8b6 | 2005-06-23 00:07:40 -0700 | [diff] [blame] | 692 | page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT); |
| 693 | page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT; |
| 694 | } |
Christoph Lameter | 2f1b624 | 2006-09-25 23:31:13 -0700 | [diff] [blame] | 695 | |
Dave Hansen | 348f8b6 | 2005-06-23 00:07:40 -0700 | [diff] [blame] | 696 | static inline void set_page_node(struct page *page, unsigned long node) |
| 697 | { |
| 698 | page->flags &= ~(NODES_MASK << NODES_PGSHIFT); |
| 699 | page->flags |= (node & NODES_MASK) << NODES_PGSHIFT; |
| 700 | } |
Christoph Lameter | 89689ae | 2006-12-06 20:31:45 -0800 | [diff] [blame] | 701 | |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 702 | static inline void set_page_section(struct page *page, unsigned long section) |
| 703 | { |
| 704 | page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT); |
| 705 | page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT; |
| 706 | } |
Dave Hansen | 348f8b6 | 2005-06-23 00:07:40 -0700 | [diff] [blame] | 707 | |
Christoph Lameter | 2f1b624 | 2006-09-25 23:31:13 -0700 | [diff] [blame] | 708 | static inline void set_page_links(struct page *page, enum zone_type zone, |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 709 | unsigned long node, unsigned long pfn) |
Dave Hansen | 348f8b6 | 2005-06-23 00:07:40 -0700 | [diff] [blame] | 710 | { |
| 711 | set_page_zone(page, zone); |
| 712 | set_page_node(page, node); |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 713 | set_page_section(page, pfn_to_section_nr(pfn)); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 714 | } |
| 715 | |
Christoph Lameter | f6ac235 | 2006-06-30 01:55:32 -0700 | [diff] [blame] | 716 | /* |
| 717 | * Some inline functions in vmstat.h depend on page_zone() |
| 718 | */ |
| 719 | #include <linux/vmstat.h> |
| 720 | |
Ingo Molnar | 652050a | 2006-01-14 13:21:30 -0800 | [diff] [blame] | 721 | static __always_inline void *lowmem_page_address(struct page *page) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 722 | { |
Chris Metcalf | c6f6b59 | 2010-05-24 14:32:53 -0700 | [diff] [blame] | 723 | return __va(PFN_PHYS(page_to_pfn(page))); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 724 | } |
| 725 | |
| 726 | #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL) |
| 727 | #define HASHED_PAGE_VIRTUAL |
| 728 | #endif |
| 729 | |
| 730 | #if defined(WANT_PAGE_VIRTUAL) |
| 731 | #define page_address(page) ((page)->virtual) |
| 732 | #define set_page_address(page, address) \ |
| 733 | do { \ |
| 734 | (page)->virtual = (address); \ |
| 735 | } while(0) |
| 736 | #define page_address_init() do { } while(0) |
| 737 | #endif |
| 738 | |
| 739 | #if defined(HASHED_PAGE_VIRTUAL) |
| 740 | void *page_address(struct page *page); |
| 741 | void set_page_address(struct page *page, void *virtual); |
| 742 | void page_address_init(void); |
| 743 | #endif |
| 744 | |
| 745 | #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL) |
| 746 | #define page_address(page) lowmem_page_address(page) |
| 747 | #define set_page_address(page, address) do { } while(0) |
| 748 | #define page_address_init() do { } while(0) |
| 749 | #endif |
| 750 | |
| 751 | /* |
| 752 | * On an anonymous page mapped into a user virtual memory area, |
| 753 | * page->mapping points to its anon_vma, not to a struct address_space; |
Hugh Dickins | 3ca7b3c | 2009-12-14 17:58:57 -0800 | [diff] [blame] | 754 | * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h. |
| 755 | * |
| 756 | * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled, |
| 757 | * the PAGE_MAPPING_KSM bit may be set along with the PAGE_MAPPING_ANON bit; |
| 758 | * and then page->mapping points, not to an anon_vma, but to a private |
| 759 | * structure which KSM associates with that merged page. See ksm.h. |
| 760 | * |
| 761 | * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is currently never used. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 762 | * |
| 763 | * Please note that, confusingly, "page_mapping" refers to the inode |
| 764 | * address_space which maps the page from disk; whereas "page_mapped" |
| 765 | * refers to user virtual address space into which the page is mapped. |
| 766 | */ |
| 767 | #define PAGE_MAPPING_ANON 1 |
Hugh Dickins | 3ca7b3c | 2009-12-14 17:58:57 -0800 | [diff] [blame] | 768 | #define PAGE_MAPPING_KSM 2 |
| 769 | #define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_KSM) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 770 | |
| 771 | extern struct address_space swapper_space; |
| 772 | static inline struct address_space *page_mapping(struct page *page) |
| 773 | { |
| 774 | struct address_space *mapping = page->mapping; |
| 775 | |
Christoph Lameter | b5fab14e | 2007-07-17 04:03:33 -0700 | [diff] [blame] | 776 | VM_BUG_ON(PageSlab(page)); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 777 | if (unlikely(PageSwapCache(page))) |
| 778 | mapping = &swapper_space; |
Steven Rostedt | e20e877 | 2011-01-13 15:46:17 -0800 | [diff] [blame] | 779 | else if ((unsigned long)mapping & PAGE_MAPPING_ANON) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 780 | mapping = NULL; |
| 781 | return mapping; |
| 782 | } |
| 783 | |
Hugh Dickins | 3ca7b3c | 2009-12-14 17:58:57 -0800 | [diff] [blame] | 784 | /* Neutral page->mapping pointer to address_space or anon_vma or other */ |
| 785 | static inline void *page_rmapping(struct page *page) |
| 786 | { |
| 787 | return (void *)((unsigned long)page->mapping & ~PAGE_MAPPING_FLAGS); |
| 788 | } |
| 789 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 790 | static inline int PageAnon(struct page *page) |
| 791 | { |
| 792 | return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0; |
| 793 | } |
| 794 | |
| 795 | /* |
| 796 | * Return the pagecache index of the passed page. Regular pagecache pages |
| 797 | * use ->index whereas swapcache pages use ->private |
| 798 | */ |
| 799 | static inline pgoff_t page_index(struct page *page) |
| 800 | { |
| 801 | if (unlikely(PageSwapCache(page))) |
Hugh Dickins | 4c21e2f | 2005-10-29 18:16:40 -0700 | [diff] [blame] | 802 | return page_private(page); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 803 | return page->index; |
| 804 | } |
| 805 | |
| 806 | /* |
| 807 | * The atomic page->_mapcount, like _count, starts from -1: |
| 808 | * so that transitions both from it and to it can be tracked, |
| 809 | * using atomic_inc_and_test and atomic_add_negative(-1). |
| 810 | */ |
| 811 | static inline void reset_page_mapcount(struct page *page) |
| 812 | { |
| 813 | atomic_set(&(page)->_mapcount, -1); |
| 814 | } |
| 815 | |
| 816 | static inline int page_mapcount(struct page *page) |
| 817 | { |
| 818 | return atomic_read(&(page)->_mapcount) + 1; |
| 819 | } |
| 820 | |
| 821 | /* |
| 822 | * Return true if this page is mapped into pagetables. |
| 823 | */ |
| 824 | static inline int page_mapped(struct page *page) |
| 825 | { |
| 826 | return atomic_read(&(page)->_mapcount) >= 0; |
| 827 | } |
| 828 | |
| 829 | /* |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 830 | * Different kinds of faults, as returned by handle_mm_fault(). |
| 831 | * Used to decide whether a process gets delivered SIGBUS or |
| 832 | * just gets major/minor fault counters bumped up. |
| 833 | */ |
Nick Piggin | d0217ac | 2007-07-19 01:47:03 -0700 | [diff] [blame] | 834 | |
Nick Piggin | 83c5407 | 2007-07-19 01:47:05 -0700 | [diff] [blame] | 835 | #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */ |
Nick Piggin | d0217ac | 2007-07-19 01:47:03 -0700 | [diff] [blame] | 836 | |
Nick Piggin | 83c5407 | 2007-07-19 01:47:05 -0700 | [diff] [blame] | 837 | #define VM_FAULT_OOM 0x0001 |
| 838 | #define VM_FAULT_SIGBUS 0x0002 |
| 839 | #define VM_FAULT_MAJOR 0x0004 |
| 840 | #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */ |
Andi Kleen | aa50d3a | 2010-10-06 21:45:00 +0200 | [diff] [blame] | 841 | #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */ |
| 842 | #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */ |
Nick Piggin | f33ea7f | 2005-08-03 20:24:01 +1000 | [diff] [blame] | 843 | |
Nick Piggin | 83c5407 | 2007-07-19 01:47:05 -0700 | [diff] [blame] | 844 | #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */ |
| 845 | #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */ |
Michel Lespinasse | d065bd8 | 2010-10-26 14:21:57 -0700 | [diff] [blame] | 846 | #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 847 | |
Andi Kleen | aa50d3a | 2010-10-06 21:45:00 +0200 | [diff] [blame] | 848 | #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */ |
| 849 | |
| 850 | #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_HWPOISON | \ |
| 851 | VM_FAULT_HWPOISON_LARGE) |
| 852 | |
| 853 | /* Encode hstate index for a hwpoisoned large page */ |
| 854 | #define VM_FAULT_SET_HINDEX(x) ((x) << 12) |
| 855 | #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf) |
Nick Piggin | d0217ac | 2007-07-19 01:47:03 -0700 | [diff] [blame] | 856 | |
Nick Piggin | 1c0fe6e | 2009-01-06 14:38:59 -0800 | [diff] [blame] | 857 | /* |
| 858 | * Can be called by the pagefault handler when it gets a VM_FAULT_OOM. |
| 859 | */ |
| 860 | extern void pagefault_out_of_memory(void); |
| 861 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 862 | #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK) |
| 863 | |
David Rientjes | ddd588b | 2011-03-22 16:30:46 -0700 | [diff] [blame] | 864 | /* |
David Rientjes | b2b755b | 2011-03-24 15:18:15 -0700 | [diff] [blame] | 865 | * Flags passed to show_mem() and __show_free_areas() to suppress output in |
David Rientjes | ddd588b | 2011-03-22 16:30:46 -0700 | [diff] [blame] | 866 | * various contexts. |
| 867 | */ |
| 868 | #define SHOW_MEM_FILTER_NODES (0x0001u) /* filter disallowed nodes */ |
| 869 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 870 | extern void show_free_areas(void); |
David Rientjes | ddd588b | 2011-03-22 16:30:46 -0700 | [diff] [blame] | 871 | extern void __show_free_areas(unsigned int flags); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 872 | |
Hugh Dickins | 3f96b79 | 2009-09-21 17:03:37 -0700 | [diff] [blame] | 873 | int shmem_lock(struct file *file, int lock, struct user_struct *user); |
Sergei Trofimovich | 168f5ac | 2009-06-16 15:33:02 -0700 | [diff] [blame] | 874 | struct file *shmem_file_setup(const char *name, loff_t size, unsigned long flags); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 875 | int shmem_zero_setup(struct vm_area_struct *); |
| 876 | |
David Howells | b0e1519 | 2006-01-06 00:11:42 -0800 | [diff] [blame] | 877 | #ifndef CONFIG_MMU |
| 878 | extern unsigned long shmem_get_unmapped_area(struct file *file, |
| 879 | unsigned long addr, |
| 880 | unsigned long len, |
| 881 | unsigned long pgoff, |
| 882 | unsigned long flags); |
| 883 | #endif |
| 884 | |
Alexey Dobriyan | e8edc6e | 2007-05-21 01:22:52 +0400 | [diff] [blame] | 885 | extern int can_do_mlock(void); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 886 | extern int user_shm_lock(size_t, struct user_struct *); |
| 887 | extern void user_shm_unlock(size_t, struct user_struct *); |
| 888 | |
| 889 | /* |
| 890 | * Parameter block passed down to zap_pte_range in exceptional cases. |
| 891 | */ |
| 892 | struct zap_details { |
| 893 | struct vm_area_struct *nonlinear_vma; /* Check page->index if set */ |
| 894 | struct address_space *check_mapping; /* Check page->mapping if set */ |
| 895 | pgoff_t first_index; /* Lowest page->index to unmap */ |
| 896 | pgoff_t last_index; /* Highest page->index to unmap */ |
| 897 | spinlock_t *i_mmap_lock; /* For unmap_mapping_range: */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 898 | unsigned long truncate_count; /* Compare vm_truncate_count */ |
| 899 | }; |
| 900 | |
Nick Piggin | 7e67513 | 2008-04-28 02:13:00 -0700 | [diff] [blame] | 901 | struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr, |
| 902 | pte_t pte); |
| 903 | |
Jack Steiner | c627f9c | 2008-07-29 22:33:53 -0700 | [diff] [blame] | 904 | int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address, |
| 905 | unsigned long size); |
Hugh Dickins | ee39b37 | 2005-04-19 13:29:15 -0700 | [diff] [blame] | 906 | unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 907 | unsigned long size, struct zap_details *); |
Hugh Dickins | 508034a | 2005-10-29 18:16:30 -0700 | [diff] [blame] | 908 | unsigned long unmap_vmas(struct mmu_gather **tlb, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 909 | struct vm_area_struct *start_vma, unsigned long start_addr, |
| 910 | unsigned long end_addr, unsigned long *nr_accounted, |
| 911 | struct zap_details *); |
Matt Mackall | e647309 | 2008-02-04 22:29:01 -0800 | [diff] [blame] | 912 | |
| 913 | /** |
| 914 | * mm_walk - callbacks for walk_page_range |
| 915 | * @pgd_entry: if set, called for each non-empty PGD (top-level) entry |
| 916 | * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry |
| 917 | * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry |
Dave Hansen | 0331932 | 2011-03-22 16:32:56 -0700 | [diff] [blame] | 918 | * this handler is required to be able to handle |
| 919 | * pmd_trans_huge() pmds. They may simply choose to |
| 920 | * split_huge_page() instead of handling it explicitly. |
Matt Mackall | e647309 | 2008-02-04 22:29:01 -0800 | [diff] [blame] | 921 | * @pte_entry: if set, called for each non-empty PTE (4th-level) entry |
| 922 | * @pte_hole: if set, called for each hole at all levels |
Naoya Horiguchi | 5dc3764 | 2009-12-14 18:00:01 -0800 | [diff] [blame] | 923 | * @hugetlb_entry: if set, called for each hugetlb entry |
Matt Mackall | e647309 | 2008-02-04 22:29:01 -0800 | [diff] [blame] | 924 | * |
| 925 | * (see walk_page_range for more details) |
| 926 | */ |
| 927 | struct mm_walk { |
Dave Hansen | 2165009 | 2008-06-12 15:21:47 -0700 | [diff] [blame] | 928 | int (*pgd_entry)(pgd_t *, unsigned long, unsigned long, struct mm_walk *); |
| 929 | int (*pud_entry)(pud_t *, unsigned long, unsigned long, struct mm_walk *); |
| 930 | int (*pmd_entry)(pmd_t *, unsigned long, unsigned long, struct mm_walk *); |
| 931 | int (*pte_entry)(pte_t *, unsigned long, unsigned long, struct mm_walk *); |
| 932 | int (*pte_hole)(unsigned long, unsigned long, struct mm_walk *); |
Naoya Horiguchi | 116354d | 2010-04-06 14:35:04 -0700 | [diff] [blame] | 933 | int (*hugetlb_entry)(pte_t *, unsigned long, |
| 934 | unsigned long, unsigned long, struct mm_walk *); |
Dave Hansen | 2165009 | 2008-06-12 15:21:47 -0700 | [diff] [blame] | 935 | struct mm_struct *mm; |
| 936 | void *private; |
Matt Mackall | e647309 | 2008-02-04 22:29:01 -0800 | [diff] [blame] | 937 | }; |
| 938 | |
Dave Hansen | 2165009 | 2008-06-12 15:21:47 -0700 | [diff] [blame] | 939 | int walk_page_range(unsigned long addr, unsigned long end, |
| 940 | struct mm_walk *walk); |
Jan Beulich | 42b7772 | 2008-07-23 21:27:10 -0700 | [diff] [blame] | 941 | void free_pgd_range(struct mmu_gather *tlb, unsigned long addr, |
Hugh Dickins | 3bf5ee9 | 2005-04-19 13:29:16 -0700 | [diff] [blame] | 942 | unsigned long end, unsigned long floor, unsigned long ceiling); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 943 | int copy_page_range(struct mm_struct *dst, struct mm_struct *src, |
| 944 | struct vm_area_struct *vma); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 945 | void unmap_mapping_range(struct address_space *mapping, |
| 946 | loff_t const holebegin, loff_t const holelen, int even_cows); |
Johannes Weiner | 3b6748e | 2009-06-16 15:32:35 -0700 | [diff] [blame] | 947 | int follow_pfn(struct vm_area_struct *vma, unsigned long address, |
| 948 | unsigned long *pfn); |
venkatesh.pallipadi@intel.com | d87fe66 | 2008-12-19 13:47:27 -0800 | [diff] [blame] | 949 | int follow_phys(struct vm_area_struct *vma, unsigned long address, |
| 950 | unsigned int flags, unsigned long *prot, resource_size_t *phys); |
Rik van Riel | 28b2ee2 | 2008-07-23 21:27:05 -0700 | [diff] [blame] | 951 | int generic_access_phys(struct vm_area_struct *vma, unsigned long addr, |
| 952 | void *buf, int len, int write); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 953 | |
| 954 | static inline void unmap_shared_mapping_range(struct address_space *mapping, |
| 955 | loff_t const holebegin, loff_t const holelen) |
| 956 | { |
| 957 | unmap_mapping_range(mapping, holebegin, holelen, 0); |
| 958 | } |
| 959 | |
npiggin@suse.de | 25d9e2d | 2009-08-21 02:35:05 +1000 | [diff] [blame] | 960 | extern void truncate_pagecache(struct inode *inode, loff_t old, loff_t new); |
Christoph Hellwig | 2c27c65 | 2010-06-04 11:30:04 +0200 | [diff] [blame] | 961 | extern void truncate_setsize(struct inode *inode, loff_t newsize); |
npiggin@suse.de | 25d9e2d | 2009-08-21 02:35:05 +1000 | [diff] [blame] | 962 | extern int vmtruncate(struct inode *inode, loff_t offset); |
| 963 | extern int vmtruncate_range(struct inode *inode, loff_t offset, loff_t end); |
Nick Piggin | f33ea7f | 2005-08-03 20:24:01 +1000 | [diff] [blame] | 964 | |
Nick Piggin | 750b498 | 2009-09-16 11:50:12 +0200 | [diff] [blame] | 965 | int truncate_inode_page(struct address_space *mapping, struct page *page); |
Andi Kleen | 2571873 | 2009-09-16 11:50:13 +0200 | [diff] [blame] | 966 | int generic_error_remove_page(struct address_space *mapping, struct page *page); |
Nick Piggin | 750b498 | 2009-09-16 11:50:12 +0200 | [diff] [blame] | 967 | |
Wu Fengguang | 83f7866 | 2009-09-16 11:50:13 +0200 | [diff] [blame] | 968 | int invalidate_inode_page(struct page *page); |
| 969 | |
David Howells | 7ee1dd3 | 2006-01-06 00:11:44 -0800 | [diff] [blame] | 970 | #ifdef CONFIG_MMU |
Nick Piggin | 83c5407 | 2007-07-19 01:47:05 -0700 | [diff] [blame] | 971 | extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma, |
Linus Torvalds | d06063c | 2009-04-10 09:01:23 -0700 | [diff] [blame] | 972 | unsigned long address, unsigned int flags); |
David Howells | 7ee1dd3 | 2006-01-06 00:11:44 -0800 | [diff] [blame] | 973 | #else |
| 974 | static inline int handle_mm_fault(struct mm_struct *mm, |
| 975 | struct vm_area_struct *vma, unsigned long address, |
Linus Torvalds | d06063c | 2009-04-10 09:01:23 -0700 | [diff] [blame] | 976 | unsigned int flags) |
David Howells | 7ee1dd3 | 2006-01-06 00:11:44 -0800 | [diff] [blame] | 977 | { |
| 978 | /* should never happen if there's no MMU */ |
| 979 | BUG(); |
| 980 | return VM_FAULT_SIGBUS; |
| 981 | } |
| 982 | #endif |
Nick Piggin | f33ea7f | 2005-08-03 20:24:01 +1000 | [diff] [blame] | 983 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 984 | extern int make_pages_present(unsigned long addr, unsigned long end); |
| 985 | extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write); |
Stephen Wilson | 5ddd36b | 2011-03-13 15:49:20 -0400 | [diff] [blame] | 986 | extern int access_remote_vm(struct mm_struct *mm, unsigned long addr, |
| 987 | void *buf, int len, int write); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 988 | |
Huang Ying | 0014bd9 | 2011-01-30 11:15:47 +0800 | [diff] [blame] | 989 | int __get_user_pages(struct task_struct *tsk, struct mm_struct *mm, |
| 990 | unsigned long start, int len, unsigned int foll_flags, |
| 991 | struct page **pages, struct vm_area_struct **vmas, |
| 992 | int *nonblocking); |
Nick Piggin | d2bf6be | 2009-06-16 15:31:39 -0700 | [diff] [blame] | 993 | int get_user_pages(struct task_struct *tsk, struct mm_struct *mm, |
Peter Zijlstra | 9d73777 | 2009-06-25 11:58:55 +0200 | [diff] [blame] | 994 | unsigned long start, int nr_pages, int write, int force, |
Nick Piggin | d2bf6be | 2009-06-16 15:31:39 -0700 | [diff] [blame] | 995 | struct page **pages, struct vm_area_struct **vmas); |
| 996 | int get_user_pages_fast(unsigned long start, int nr_pages, int write, |
| 997 | struct page **pages); |
Hugh Dickins | f3e8fcc | 2009-09-21 17:03:25 -0700 | [diff] [blame] | 998 | struct page *get_dump_page(unsigned long addr); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 999 | |
David Howells | cf9a2ae | 2006-08-29 19:05:54 +0100 | [diff] [blame] | 1000 | extern int try_to_release_page(struct page * page, gfp_t gfp_mask); |
| 1001 | extern void do_invalidatepage(struct page *page, unsigned long offset); |
| 1002 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1003 | int __set_page_dirty_nobuffers(struct page *page); |
Ken Chen | 7671932 | 2007-02-10 01:43:15 -0800 | [diff] [blame] | 1004 | int __set_page_dirty_no_writeback(struct page *page); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1005 | int redirty_page_for_writepage(struct writeback_control *wbc, |
| 1006 | struct page *page); |
Edward Shishkin | e3a7cca | 2009-03-31 15:19:39 -0700 | [diff] [blame] | 1007 | void account_page_dirtied(struct page *page, struct address_space *mapping); |
Michael Rubin | f629d1c | 2010-10-26 14:21:33 -0700 | [diff] [blame] | 1008 | void account_page_writeback(struct page *page); |
Harvey Harrison | b3c9752 | 2008-02-13 15:03:15 -0800 | [diff] [blame] | 1009 | int set_page_dirty(struct page *page); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1010 | int set_page_dirty_lock(struct page *page); |
| 1011 | int clear_page_dirty_for_io(struct page *page); |
| 1012 | |
Stefan Bader | 39aa3cb | 2010-08-31 15:52:27 +0200 | [diff] [blame] | 1013 | /* Is the vma a continuation of the stack vma above it? */ |
| 1014 | static inline int vma_stack_continue(struct vm_area_struct *vma, unsigned long addr) |
| 1015 | { |
| 1016 | return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN); |
| 1017 | } |
| 1018 | |
Ollie Wild | b6a2fea | 2007-07-19 01:48:16 -0700 | [diff] [blame] | 1019 | extern unsigned long move_page_tables(struct vm_area_struct *vma, |
| 1020 | unsigned long old_addr, struct vm_area_struct *new_vma, |
| 1021 | unsigned long new_addr, unsigned long len); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1022 | extern unsigned long do_mremap(unsigned long addr, |
| 1023 | unsigned long old_len, unsigned long new_len, |
| 1024 | unsigned long flags, unsigned long new_addr); |
Ollie Wild | b6a2fea | 2007-07-19 01:48:16 -0700 | [diff] [blame] | 1025 | extern int mprotect_fixup(struct vm_area_struct *vma, |
| 1026 | struct vm_area_struct **pprev, unsigned long start, |
| 1027 | unsigned long end, unsigned long newflags); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1028 | |
Nick Piggin | 21cc199 | 2008-07-25 19:45:22 -0700 | [diff] [blame] | 1029 | /* |
Peter Zijlstra | 465a454 | 2009-06-15 12:31:37 +0200 | [diff] [blame] | 1030 | * doesn't attempt to fault and will return short. |
| 1031 | */ |
| 1032 | int __get_user_pages_fast(unsigned long start, int nr_pages, int write, |
| 1033 | struct page **pages); |
KAMEZAWA Hiroyuki | d559db0 | 2010-03-05 13:41:39 -0800 | [diff] [blame] | 1034 | /* |
| 1035 | * per-process(per-mm_struct) statistics. |
| 1036 | */ |
KAMEZAWA Hiroyuki | 34e5523 | 2010-03-05 13:41:40 -0800 | [diff] [blame] | 1037 | #if defined(SPLIT_RSS_COUNTING) |
KAMEZAWA Hiroyuki | d559db0 | 2010-03-05 13:41:39 -0800 | [diff] [blame] | 1038 | /* |
| 1039 | * The mm counters are not protected by its page_table_lock, |
| 1040 | * so must be incremented atomically. |
| 1041 | */ |
| 1042 | static inline void set_mm_counter(struct mm_struct *mm, int member, long value) |
| 1043 | { |
| 1044 | atomic_long_set(&mm->rss_stat.count[member], value); |
| 1045 | } |
| 1046 | |
KAMEZAWA Hiroyuki | 34e5523 | 2010-03-05 13:41:40 -0800 | [diff] [blame] | 1047 | unsigned long get_mm_counter(struct mm_struct *mm, int member); |
KAMEZAWA Hiroyuki | d559db0 | 2010-03-05 13:41:39 -0800 | [diff] [blame] | 1048 | |
| 1049 | static inline void add_mm_counter(struct mm_struct *mm, int member, long value) |
| 1050 | { |
| 1051 | atomic_long_add(value, &mm->rss_stat.count[member]); |
| 1052 | } |
| 1053 | |
| 1054 | static inline void inc_mm_counter(struct mm_struct *mm, int member) |
| 1055 | { |
| 1056 | atomic_long_inc(&mm->rss_stat.count[member]); |
| 1057 | } |
| 1058 | |
| 1059 | static inline void dec_mm_counter(struct mm_struct *mm, int member) |
| 1060 | { |
| 1061 | atomic_long_dec(&mm->rss_stat.count[member]); |
| 1062 | } |
| 1063 | |
| 1064 | #else /* !USE_SPLIT_PTLOCKS */ |
| 1065 | /* |
| 1066 | * The mm counters are protected by its page_table_lock, |
| 1067 | * so can be incremented directly. |
| 1068 | */ |
| 1069 | static inline void set_mm_counter(struct mm_struct *mm, int member, long value) |
| 1070 | { |
| 1071 | mm->rss_stat.count[member] = value; |
| 1072 | } |
| 1073 | |
| 1074 | static inline unsigned long get_mm_counter(struct mm_struct *mm, int member) |
| 1075 | { |
| 1076 | return mm->rss_stat.count[member]; |
| 1077 | } |
| 1078 | |
| 1079 | static inline void add_mm_counter(struct mm_struct *mm, int member, long value) |
| 1080 | { |
| 1081 | mm->rss_stat.count[member] += value; |
| 1082 | } |
| 1083 | |
| 1084 | static inline void inc_mm_counter(struct mm_struct *mm, int member) |
| 1085 | { |
| 1086 | mm->rss_stat.count[member]++; |
| 1087 | } |
| 1088 | |
| 1089 | static inline void dec_mm_counter(struct mm_struct *mm, int member) |
| 1090 | { |
| 1091 | mm->rss_stat.count[member]--; |
| 1092 | } |
| 1093 | |
| 1094 | #endif /* !USE_SPLIT_PTLOCKS */ |
| 1095 | |
| 1096 | static inline unsigned long get_mm_rss(struct mm_struct *mm) |
| 1097 | { |
| 1098 | return get_mm_counter(mm, MM_FILEPAGES) + |
| 1099 | get_mm_counter(mm, MM_ANONPAGES); |
| 1100 | } |
| 1101 | |
| 1102 | static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm) |
| 1103 | { |
| 1104 | return max(mm->hiwater_rss, get_mm_rss(mm)); |
| 1105 | } |
| 1106 | |
| 1107 | static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm) |
| 1108 | { |
| 1109 | return max(mm->hiwater_vm, mm->total_vm); |
| 1110 | } |
| 1111 | |
| 1112 | static inline void update_hiwater_rss(struct mm_struct *mm) |
| 1113 | { |
| 1114 | unsigned long _rss = get_mm_rss(mm); |
| 1115 | |
| 1116 | if ((mm)->hiwater_rss < _rss) |
| 1117 | (mm)->hiwater_rss = _rss; |
| 1118 | } |
| 1119 | |
| 1120 | static inline void update_hiwater_vm(struct mm_struct *mm) |
| 1121 | { |
| 1122 | if (mm->hiwater_vm < mm->total_vm) |
| 1123 | mm->hiwater_vm = mm->total_vm; |
| 1124 | } |
| 1125 | |
| 1126 | static inline void setmax_mm_hiwater_rss(unsigned long *maxrss, |
| 1127 | struct mm_struct *mm) |
| 1128 | { |
| 1129 | unsigned long hiwater_rss = get_mm_hiwater_rss(mm); |
| 1130 | |
| 1131 | if (*maxrss < hiwater_rss) |
| 1132 | *maxrss = hiwater_rss; |
| 1133 | } |
| 1134 | |
KAMEZAWA Hiroyuki | 53bddb4 | 2010-03-10 15:20:38 -0800 | [diff] [blame] | 1135 | #if defined(SPLIT_RSS_COUNTING) |
KAMEZAWA Hiroyuki | 34e5523 | 2010-03-05 13:41:40 -0800 | [diff] [blame] | 1136 | void sync_mm_rss(struct task_struct *task, struct mm_struct *mm); |
KAMEZAWA Hiroyuki | 53bddb4 | 2010-03-10 15:20:38 -0800 | [diff] [blame] | 1137 | #else |
| 1138 | static inline void sync_mm_rss(struct task_struct *task, struct mm_struct *mm) |
| 1139 | { |
| 1140 | } |
| 1141 | #endif |
Peter Zijlstra | 465a454 | 2009-06-15 12:31:37 +0200 | [diff] [blame] | 1142 | |
| 1143 | /* |
Rusty Russell | 8e1f936 | 2007-07-17 04:03:17 -0700 | [diff] [blame] | 1144 | * A callback you can register to apply pressure to ageable caches. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1145 | * |
Rusty Russell | 8e1f936 | 2007-07-17 04:03:17 -0700 | [diff] [blame] | 1146 | * 'shrink' is passed a count 'nr_to_scan' and a 'gfpmask'. It should |
| 1147 | * look through the least-recently-used 'nr_to_scan' entries and |
| 1148 | * attempt to free them up. It should return the number of objects |
| 1149 | * which remain in the cache. If it returns -1, it means it cannot do |
| 1150 | * any scanning at this time (eg. there is a risk of deadlock). |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1151 | * |
Rusty Russell | 8e1f936 | 2007-07-17 04:03:17 -0700 | [diff] [blame] | 1152 | * The 'gfpmask' refers to the allocation we are currently trying to |
| 1153 | * fulfil. |
| 1154 | * |
| 1155 | * Note that 'shrink' will be passed nr_to_scan == 0 when the VM is |
| 1156 | * querying the cache size, so a fastpath for that case is appropriate. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1157 | */ |
Rusty Russell | 8e1f936 | 2007-07-17 04:03:17 -0700 | [diff] [blame] | 1158 | struct shrinker { |
Dave Chinner | 7f8275d | 2010-07-19 14:56:17 +1000 | [diff] [blame] | 1159 | int (*shrink)(struct shrinker *, int nr_to_scan, gfp_t gfp_mask); |
Rusty Russell | 8e1f936 | 2007-07-17 04:03:17 -0700 | [diff] [blame] | 1160 | int seeks; /* seeks to recreate an obj */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1161 | |
Rusty Russell | 8e1f936 | 2007-07-17 04:03:17 -0700 | [diff] [blame] | 1162 | /* These are for internal use */ |
| 1163 | struct list_head list; |
| 1164 | long nr; /* objs pending delete */ |
| 1165 | }; |
| 1166 | #define DEFAULT_SEEKS 2 /* A good number if you don't know better. */ |
| 1167 | extern void register_shrinker(struct shrinker *); |
| 1168 | extern void unregister_shrinker(struct shrinker *); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1169 | |
Alexey Dobriyan | 4e950f6 | 2007-07-30 02:36:13 +0400 | [diff] [blame] | 1170 | int vma_wants_writenotify(struct vm_area_struct *vma); |
Peter Zijlstra | d08b385 | 2006-09-25 23:30:57 -0700 | [diff] [blame] | 1171 | |
Namhyung Kim | 25ca1d6 | 2010-10-26 14:21:59 -0700 | [diff] [blame] | 1172 | extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr, |
| 1173 | spinlock_t **ptl); |
| 1174 | static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr, |
| 1175 | spinlock_t **ptl) |
| 1176 | { |
| 1177 | pte_t *ptep; |
| 1178 | __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl)); |
| 1179 | return ptep; |
| 1180 | } |
Linus Torvalds | c9cfcdd | 2005-11-29 14:03:14 -0800 | [diff] [blame] | 1181 | |
Nick Piggin | 5f22df0 | 2007-05-06 14:49:02 -0700 | [diff] [blame] | 1182 | #ifdef __PAGETABLE_PUD_FOLDED |
| 1183 | static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, |
| 1184 | unsigned long address) |
| 1185 | { |
| 1186 | return 0; |
| 1187 | } |
| 1188 | #else |
Hugh Dickins | 1bb3630 | 2005-10-29 18:16:22 -0700 | [diff] [blame] | 1189 | int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address); |
Nick Piggin | 5f22df0 | 2007-05-06 14:49:02 -0700 | [diff] [blame] | 1190 | #endif |
| 1191 | |
| 1192 | #ifdef __PAGETABLE_PMD_FOLDED |
| 1193 | static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud, |
| 1194 | unsigned long address) |
| 1195 | { |
| 1196 | return 0; |
| 1197 | } |
| 1198 | #else |
Hugh Dickins | 1bb3630 | 2005-10-29 18:16:22 -0700 | [diff] [blame] | 1199 | int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address); |
Nick Piggin | 5f22df0 | 2007-05-06 14:49:02 -0700 | [diff] [blame] | 1200 | #endif |
| 1201 | |
Andrea Arcangeli | 8ac1f83 | 2011-01-13 15:46:43 -0800 | [diff] [blame] | 1202 | int __pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma, |
| 1203 | pmd_t *pmd, unsigned long address); |
Hugh Dickins | 1bb3630 | 2005-10-29 18:16:22 -0700 | [diff] [blame] | 1204 | int __pte_alloc_kernel(pmd_t *pmd, unsigned long address); |
| 1205 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1206 | /* |
| 1207 | * The following ifdef needed to get the 4level-fixup.h header to work. |
| 1208 | * Remove it when 4level-fixup.h has been removed. |
| 1209 | */ |
Hugh Dickins | 1bb3630 | 2005-10-29 18:16:22 -0700 | [diff] [blame] | 1210 | #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1211 | static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address) |
| 1212 | { |
Hugh Dickins | 1bb3630 | 2005-10-29 18:16:22 -0700 | [diff] [blame] | 1213 | return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))? |
| 1214 | NULL: pud_offset(pgd, address); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1215 | } |
| 1216 | |
| 1217 | static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address) |
| 1218 | { |
Hugh Dickins | 1bb3630 | 2005-10-29 18:16:22 -0700 | [diff] [blame] | 1219 | return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))? |
| 1220 | NULL: pmd_offset(pud, address); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1221 | } |
Hugh Dickins | 1bb3630 | 2005-10-29 18:16:22 -0700 | [diff] [blame] | 1222 | #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */ |
| 1223 | |
Jeremy Fitzhardinge | f7d0b92 | 2008-09-09 15:43:22 -0700 | [diff] [blame] | 1224 | #if USE_SPLIT_PTLOCKS |
Hugh Dickins | 4c21e2f | 2005-10-29 18:16:40 -0700 | [diff] [blame] | 1225 | /* |
| 1226 | * We tuck a spinlock to guard each pagetable page into its struct page, |
| 1227 | * at page->private, with BUILD_BUG_ON to make sure that this will not |
| 1228 | * overflow into the next struct page (as it might with DEBUG_SPINLOCK). |
| 1229 | * When freeing, reset page->mapping so free_pages_check won't complain. |
| 1230 | */ |
Andrew Morton | 349aef0 | 2006-01-08 01:04:36 -0800 | [diff] [blame] | 1231 | #define __pte_lockptr(page) &((page)->ptl) |
Hugh Dickins | 4c21e2f | 2005-10-29 18:16:40 -0700 | [diff] [blame] | 1232 | #define pte_lock_init(_page) do { \ |
| 1233 | spin_lock_init(__pte_lockptr(_page)); \ |
| 1234 | } while (0) |
| 1235 | #define pte_lock_deinit(page) ((page)->mapping = NULL) |
| 1236 | #define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));}) |
Jeremy Fitzhardinge | f7d0b92 | 2008-09-09 15:43:22 -0700 | [diff] [blame] | 1237 | #else /* !USE_SPLIT_PTLOCKS */ |
Hugh Dickins | 4c21e2f | 2005-10-29 18:16:40 -0700 | [diff] [blame] | 1238 | /* |
| 1239 | * We use mm->page_table_lock to guard all pagetable pages of the mm. |
| 1240 | */ |
| 1241 | #define pte_lock_init(page) do {} while (0) |
| 1242 | #define pte_lock_deinit(page) do {} while (0) |
| 1243 | #define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;}) |
Jeremy Fitzhardinge | f7d0b92 | 2008-09-09 15:43:22 -0700 | [diff] [blame] | 1244 | #endif /* USE_SPLIT_PTLOCKS */ |
Hugh Dickins | 4c21e2f | 2005-10-29 18:16:40 -0700 | [diff] [blame] | 1245 | |
Martin Schwidefsky | 2f569af | 2008-02-08 04:22:04 -0800 | [diff] [blame] | 1246 | static inline void pgtable_page_ctor(struct page *page) |
| 1247 | { |
| 1248 | pte_lock_init(page); |
| 1249 | inc_zone_page_state(page, NR_PAGETABLE); |
| 1250 | } |
| 1251 | |
| 1252 | static inline void pgtable_page_dtor(struct page *page) |
| 1253 | { |
| 1254 | pte_lock_deinit(page); |
| 1255 | dec_zone_page_state(page, NR_PAGETABLE); |
| 1256 | } |
| 1257 | |
Hugh Dickins | c74df32 | 2005-10-29 18:16:23 -0700 | [diff] [blame] | 1258 | #define pte_offset_map_lock(mm, pmd, address, ptlp) \ |
| 1259 | ({ \ |
Hugh Dickins | 4c21e2f | 2005-10-29 18:16:40 -0700 | [diff] [blame] | 1260 | spinlock_t *__ptl = pte_lockptr(mm, pmd); \ |
Hugh Dickins | c74df32 | 2005-10-29 18:16:23 -0700 | [diff] [blame] | 1261 | pte_t *__pte = pte_offset_map(pmd, address); \ |
| 1262 | *(ptlp) = __ptl; \ |
| 1263 | spin_lock(__ptl); \ |
| 1264 | __pte; \ |
| 1265 | }) |
| 1266 | |
| 1267 | #define pte_unmap_unlock(pte, ptl) do { \ |
| 1268 | spin_unlock(ptl); \ |
| 1269 | pte_unmap(pte); \ |
| 1270 | } while (0) |
| 1271 | |
Andrea Arcangeli | 8ac1f83 | 2011-01-13 15:46:43 -0800 | [diff] [blame] | 1272 | #define pte_alloc_map(mm, vma, pmd, address) \ |
| 1273 | ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma, \ |
| 1274 | pmd, address))? \ |
| 1275 | NULL: pte_offset_map(pmd, address)) |
Hugh Dickins | 1bb3630 | 2005-10-29 18:16:22 -0700 | [diff] [blame] | 1276 | |
Hugh Dickins | c74df32 | 2005-10-29 18:16:23 -0700 | [diff] [blame] | 1277 | #define pte_alloc_map_lock(mm, pmd, address, ptlp) \ |
Andrea Arcangeli | 8ac1f83 | 2011-01-13 15:46:43 -0800 | [diff] [blame] | 1278 | ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL, \ |
| 1279 | pmd, address))? \ |
Hugh Dickins | c74df32 | 2005-10-29 18:16:23 -0700 | [diff] [blame] | 1280 | NULL: pte_offset_map_lock(mm, pmd, address, ptlp)) |
| 1281 | |
Hugh Dickins | 1bb3630 | 2005-10-29 18:16:22 -0700 | [diff] [blame] | 1282 | #define pte_alloc_kernel(pmd, address) \ |
Andrea Arcangeli | 8ac1f83 | 2011-01-13 15:46:43 -0800 | [diff] [blame] | 1283 | ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \ |
Hugh Dickins | 1bb3630 | 2005-10-29 18:16:22 -0700 | [diff] [blame] | 1284 | NULL: pte_offset_kernel(pmd, address)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1285 | |
| 1286 | extern void free_area_init(unsigned long * zones_size); |
Johannes Weiner | 9109fb7 | 2008-07-23 21:27:20 -0700 | [diff] [blame] | 1287 | extern void free_area_init_node(int nid, unsigned long * zones_size, |
| 1288 | unsigned long zone_start_pfn, unsigned long *zholes_size); |
Mel Gorman | c713216 | 2006-09-27 01:49:43 -0700 | [diff] [blame] | 1289 | #ifdef CONFIG_ARCH_POPULATES_NODE_MAP |
| 1290 | /* |
| 1291 | * With CONFIG_ARCH_POPULATES_NODE_MAP set, an architecture may initialise its |
| 1292 | * zones, allocate the backing mem_map and account for memory holes in a more |
| 1293 | * architecture independent manner. This is a substitute for creating the |
| 1294 | * zone_sizes[] and zholes_size[] arrays and passing them to |
| 1295 | * free_area_init_node() |
| 1296 | * |
| 1297 | * An architecture is expected to register range of page frames backed by |
| 1298 | * physical memory with add_active_range() before calling |
| 1299 | * free_area_init_nodes() passing in the PFN each zone ends at. At a basic |
| 1300 | * usage, an architecture is expected to do something like |
| 1301 | * |
| 1302 | * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn, |
| 1303 | * max_highmem_pfn}; |
| 1304 | * for_each_valid_physical_page_range() |
| 1305 | * add_active_range(node_id, start_pfn, end_pfn) |
| 1306 | * free_area_init_nodes(max_zone_pfns); |
| 1307 | * |
| 1308 | * If the architecture guarantees that there are no holes in the ranges |
| 1309 | * registered with add_active_range(), free_bootmem_active_regions() |
| 1310 | * will call free_bootmem_node() for each registered physical page range. |
| 1311 | * Similarly sparse_memory_present_with_active_regions() calls |
| 1312 | * memory_present() for each range when SPARSEMEM is enabled. |
| 1313 | * |
| 1314 | * See mm/page_alloc.c for more information on each function exposed by |
| 1315 | * CONFIG_ARCH_POPULATES_NODE_MAP |
| 1316 | */ |
| 1317 | extern void free_area_init_nodes(unsigned long *max_zone_pfn); |
| 1318 | extern void add_active_range(unsigned int nid, unsigned long start_pfn, |
| 1319 | unsigned long end_pfn); |
Yinghai Lu | cc1050b | 2008-06-13 19:08:52 -0700 | [diff] [blame] | 1320 | extern void remove_active_range(unsigned int nid, unsigned long start_pfn, |
| 1321 | unsigned long end_pfn); |
Mel Gorman | c713216 | 2006-09-27 01:49:43 -0700 | [diff] [blame] | 1322 | extern void remove_all_active_ranges(void); |
Yinghai Lu | 3299625 | 2009-12-15 17:59:02 -0800 | [diff] [blame] | 1323 | void sort_node_map(void); |
| 1324 | unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn, |
| 1325 | unsigned long end_pfn); |
Mel Gorman | c713216 | 2006-09-27 01:49:43 -0700 | [diff] [blame] | 1326 | extern unsigned long absent_pages_in_range(unsigned long start_pfn, |
| 1327 | unsigned long end_pfn); |
| 1328 | extern void get_pfn_range_for_nid(unsigned int nid, |
| 1329 | unsigned long *start_pfn, unsigned long *end_pfn); |
| 1330 | extern unsigned long find_min_pfn_with_active_regions(void); |
Mel Gorman | c713216 | 2006-09-27 01:49:43 -0700 | [diff] [blame] | 1331 | extern void free_bootmem_with_active_regions(int nid, |
| 1332 | unsigned long max_low_pfn); |
Yinghai Lu | 0867721 | 2010-02-10 01:20:20 -0800 | [diff] [blame] | 1333 | int add_from_early_node_map(struct range *range, int az, |
| 1334 | int nr_range, int nid); |
Yinghai Lu | edbe7d2 | 2010-08-25 13:39:16 -0700 | [diff] [blame] | 1335 | u64 __init find_memory_core_early(int nid, u64 size, u64 align, |
| 1336 | u64 goal, u64 limit); |
Yinghai Lu | d52d53b | 2008-06-16 20:10:55 -0700 | [diff] [blame] | 1337 | typedef int (*work_fn_t)(unsigned long, unsigned long, void *); |
Yinghai Lu | b5bc6c0 | 2008-06-14 18:32:52 -0700 | [diff] [blame] | 1338 | extern void work_with_active_regions(int nid, work_fn_t work_fn, void *data); |
Mel Gorman | c713216 | 2006-09-27 01:49:43 -0700 | [diff] [blame] | 1339 | extern void sparse_memory_present_with_active_regions(int nid); |
Mel Gorman | c713216 | 2006-09-27 01:49:43 -0700 | [diff] [blame] | 1340 | #endif /* CONFIG_ARCH_POPULATES_NODE_MAP */ |
KAMEZAWA Hiroyuki | f2dbcfa | 2009-02-18 14:48:32 -0800 | [diff] [blame] | 1341 | |
| 1342 | #if !defined(CONFIG_ARCH_POPULATES_NODE_MAP) && \ |
| 1343 | !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID) |
| 1344 | static inline int __early_pfn_to_nid(unsigned long pfn) |
| 1345 | { |
| 1346 | return 0; |
| 1347 | } |
| 1348 | #else |
| 1349 | /* please see mm/page_alloc.c */ |
| 1350 | extern int __meminit early_pfn_to_nid(unsigned long pfn); |
| 1351 | #ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID |
| 1352 | /* there is a per-arch backend function. */ |
| 1353 | extern int __meminit __early_pfn_to_nid(unsigned long pfn); |
| 1354 | #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */ |
| 1355 | #endif |
| 1356 | |
Mel Gorman | 0e0b864 | 2006-09-27 01:49:56 -0700 | [diff] [blame] | 1357 | extern void set_dma_reserve(unsigned long new_dma_reserve); |
Dave Hansen | a2f3aa02 | 2007-01-10 23:15:30 -0800 | [diff] [blame] | 1358 | extern void memmap_init_zone(unsigned long, int, unsigned long, |
| 1359 | unsigned long, enum memmap_context); |
Minchan Kim | bc75d33 | 2009-06-16 15:32:48 -0700 | [diff] [blame] | 1360 | extern void setup_per_zone_wmarks(void); |
Minchan Kim | 96cb4df | 2009-06-16 15:32:49 -0700 | [diff] [blame] | 1361 | extern void calculate_zone_inactive_ratio(struct zone *zone); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1362 | extern void mem_init(void); |
David Howells | 8feae13 | 2009-01-08 12:04:47 +0000 | [diff] [blame] | 1363 | extern void __init mmap_init(void); |
David Rientjes | b2b755b | 2011-03-24 15:18:15 -0700 | [diff] [blame] | 1364 | extern void show_mem(unsigned int flags); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1365 | extern void si_meminfo(struct sysinfo * val); |
| 1366 | extern void si_meminfo_node(struct sysinfo *val, int nid); |
Mike Travis | 3461b0a | 2008-05-12 21:21:13 +0200 | [diff] [blame] | 1367 | extern int after_bootmem; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1368 | |
Christoph Lameter | e7c8d5c | 2005-06-21 17:14:47 -0700 | [diff] [blame] | 1369 | extern void setup_per_cpu_pageset(void); |
Christoph Lameter | e7c8d5c | 2005-06-21 17:14:47 -0700 | [diff] [blame] | 1370 | |
Shaohua Li | 112067f | 2009-09-21 17:01:16 -0700 | [diff] [blame] | 1371 | extern void zone_pcp_update(struct zone *zone); |
| 1372 | |
David Howells | 8feae13 | 2009-01-08 12:04:47 +0000 | [diff] [blame] | 1373 | /* nommu.c */ |
David Howells | 33e5d769 | 2009-04-02 16:56:32 -0700 | [diff] [blame] | 1374 | extern atomic_long_t mmap_pages_allocated; |
David Howells | 7e66087 | 2010-01-15 17:01:39 -0800 | [diff] [blame] | 1375 | extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t); |
David Howells | 8feae13 | 2009-01-08 12:04:47 +0000 | [diff] [blame] | 1376 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1377 | /* prio_tree.c */ |
| 1378 | void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old); |
| 1379 | void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *); |
| 1380 | void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *); |
| 1381 | struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma, |
| 1382 | struct prio_tree_iter *iter); |
| 1383 | |
| 1384 | #define vma_prio_tree_foreach(vma, iter, root, begin, end) \ |
| 1385 | for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \ |
| 1386 | (vma = vma_prio_tree_next(vma, iter)); ) |
| 1387 | |
| 1388 | static inline void vma_nonlinear_insert(struct vm_area_struct *vma, |
| 1389 | struct list_head *list) |
| 1390 | { |
| 1391 | vma->shared.vm_set.parent = NULL; |
| 1392 | list_add_tail(&vma->shared.vm_set.list, list); |
| 1393 | } |
| 1394 | |
| 1395 | /* mmap.c */ |
Alan Cox | 34b4e4a | 2007-08-22 14:01:28 -0700 | [diff] [blame] | 1396 | extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin); |
Rik van Riel | 5beb493 | 2010-03-05 13:42:07 -0800 | [diff] [blame] | 1397 | extern int vma_adjust(struct vm_area_struct *vma, unsigned long start, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1398 | unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert); |
| 1399 | extern struct vm_area_struct *vma_merge(struct mm_struct *, |
| 1400 | struct vm_area_struct *prev, unsigned long addr, unsigned long end, |
| 1401 | unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t, |
| 1402 | struct mempolicy *); |
| 1403 | extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *); |
| 1404 | extern int split_vma(struct mm_struct *, |
| 1405 | struct vm_area_struct *, unsigned long addr, int new_below); |
| 1406 | extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *); |
| 1407 | extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *, |
| 1408 | struct rb_node **, struct rb_node *); |
Hugh Dickins | a8fb561 | 2005-10-29 18:15:57 -0700 | [diff] [blame] | 1409 | extern void unlink_file_vma(struct vm_area_struct *); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1410 | extern struct vm_area_struct *copy_vma(struct vm_area_struct **, |
| 1411 | unsigned long addr, unsigned long len, pgoff_t pgoff); |
| 1412 | extern void exit_mmap(struct mm_struct *); |
Matt Helsley | 925d1c4 | 2008-04-29 01:01:36 -0700 | [diff] [blame] | 1413 | |
Andrea Arcangeli | 7906d00 | 2008-07-28 15:46:26 -0700 | [diff] [blame] | 1414 | extern int mm_take_all_locks(struct mm_struct *mm); |
| 1415 | extern void mm_drop_all_locks(struct mm_struct *mm); |
| 1416 | |
Matt Helsley | 925d1c4 | 2008-04-29 01:01:36 -0700 | [diff] [blame] | 1417 | #ifdef CONFIG_PROC_FS |
| 1418 | /* From fs/proc/base.c. callers must _not_ hold the mm's exe_file_lock */ |
| 1419 | extern void added_exe_file_vma(struct mm_struct *mm); |
| 1420 | extern void removed_exe_file_vma(struct mm_struct *mm); |
| 1421 | #else |
| 1422 | static inline void added_exe_file_vma(struct mm_struct *mm) |
| 1423 | {} |
| 1424 | |
| 1425 | static inline void removed_exe_file_vma(struct mm_struct *mm) |
| 1426 | {} |
| 1427 | #endif /* CONFIG_PROC_FS */ |
| 1428 | |
akpm@osdl.org | 119f657 | 2005-05-01 08:58:35 -0700 | [diff] [blame] | 1429 | extern int may_expand_vm(struct mm_struct *mm, unsigned long npages); |
Roland McGrath | fa5dc22 | 2007-02-08 14:20:41 -0800 | [diff] [blame] | 1430 | extern int install_special_mapping(struct mm_struct *mm, |
| 1431 | unsigned long addr, unsigned long len, |
| 1432 | unsigned long flags, struct page **pages); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1433 | |
| 1434 | extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long); |
| 1435 | |
| 1436 | extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr, |
| 1437 | unsigned long len, unsigned long prot, |
| 1438 | unsigned long flag, unsigned long pgoff); |
Miklos Szeredi | 0165ab4 | 2007-07-15 23:38:26 -0700 | [diff] [blame] | 1439 | extern unsigned long mmap_region(struct file *file, unsigned long addr, |
| 1440 | unsigned long len, unsigned long flags, |
Mel Gorman | 5a6fe12 | 2009-02-10 14:02:27 +0000 | [diff] [blame] | 1441 | unsigned int vm_flags, unsigned long pgoff); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1442 | |
| 1443 | static inline unsigned long do_mmap(struct file *file, unsigned long addr, |
| 1444 | unsigned long len, unsigned long prot, |
| 1445 | unsigned long flag, unsigned long offset) |
| 1446 | { |
| 1447 | unsigned long ret = -EINVAL; |
| 1448 | if ((offset + PAGE_ALIGN(len)) < offset) |
| 1449 | goto out; |
| 1450 | if (!(offset & ~PAGE_MASK)) |
| 1451 | ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT); |
| 1452 | out: |
| 1453 | return ret; |
| 1454 | } |
| 1455 | |
| 1456 | extern int do_munmap(struct mm_struct *, unsigned long, size_t); |
| 1457 | |
| 1458 | extern unsigned long do_brk(unsigned long, unsigned long); |
| 1459 | |
| 1460 | /* filemap.c */ |
| 1461 | extern unsigned long page_unuse(struct page *); |
| 1462 | extern void truncate_inode_pages(struct address_space *, loff_t); |
Hans Reiser | d733907 | 2006-01-06 00:10:36 -0800 | [diff] [blame] | 1463 | extern void truncate_inode_pages_range(struct address_space *, |
| 1464 | loff_t lstart, loff_t lend); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1465 | |
| 1466 | /* generic vm_area_ops exported for stackable file systems */ |
Nick Piggin | d0217ac | 2007-07-19 01:47:03 -0700 | [diff] [blame] | 1467 | extern int filemap_fault(struct vm_area_struct *, struct vm_fault *); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1468 | |
| 1469 | /* mm/page-writeback.c */ |
| 1470 | int write_one_page(struct page *page, int wait); |
Nick Piggin | 1cf6e7d | 2009-02-18 14:48:18 -0800 | [diff] [blame] | 1471 | void task_dirty_inc(struct task_struct *tsk); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1472 | |
| 1473 | /* readahead.c */ |
| 1474 | #define VM_MAX_READAHEAD 128 /* kbytes */ |
| 1475 | #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1476 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1477 | int force_page_cache_readahead(struct address_space *mapping, struct file *filp, |
Andrew Morton | 7361f4d | 2005-11-07 00:59:28 -0800 | [diff] [blame] | 1478 | pgoff_t offset, unsigned long nr_to_read); |
Rusty Russell | cf914a7 | 2007-07-19 01:48:08 -0700 | [diff] [blame] | 1479 | |
| 1480 | void page_cache_sync_readahead(struct address_space *mapping, |
| 1481 | struct file_ra_state *ra, |
| 1482 | struct file *filp, |
| 1483 | pgoff_t offset, |
| 1484 | unsigned long size); |
| 1485 | |
| 1486 | void page_cache_async_readahead(struct address_space *mapping, |
| 1487 | struct file_ra_state *ra, |
| 1488 | struct file *filp, |
| 1489 | struct page *pg, |
| 1490 | pgoff_t offset, |
| 1491 | unsigned long size); |
| 1492 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1493 | unsigned long max_sane_readahead(unsigned long nr); |
Wu Fengguang | d30a110 | 2009-06-16 15:31:30 -0700 | [diff] [blame] | 1494 | unsigned long ra_submit(struct file_ra_state *ra, |
| 1495 | struct address_space *mapping, |
| 1496 | struct file *filp); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1497 | |
| 1498 | /* Do stack extension */ |
Hugh Dickins | 46dea3d | 2005-10-29 18:16:20 -0700 | [diff] [blame] | 1499 | extern int expand_stack(struct vm_area_struct *vma, unsigned long address); |
Luck, Tony | 8ca3eb0 | 2010-08-24 11:44:18 -0700 | [diff] [blame] | 1500 | #if VM_GROWSUP |
Hugh Dickins | 46dea3d | 2005-10-29 18:16:20 -0700 | [diff] [blame] | 1501 | extern int expand_upwards(struct vm_area_struct *vma, unsigned long address); |
Luck, Tony | 8ca3eb0 | 2010-08-24 11:44:18 -0700 | [diff] [blame] | 1502 | #else |
| 1503 | #define expand_upwards(vma, address) do { } while (0) |
Matthew Wilcox | 9ab8851 | 2005-11-18 16:16:42 -0500 | [diff] [blame] | 1504 | #endif |
Ollie Wild | b6a2fea | 2007-07-19 01:48:16 -0700 | [diff] [blame] | 1505 | extern int expand_stack_downwards(struct vm_area_struct *vma, |
| 1506 | unsigned long address); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1507 | |
| 1508 | /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */ |
| 1509 | extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr); |
| 1510 | extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr, |
| 1511 | struct vm_area_struct **pprev); |
| 1512 | |
| 1513 | /* Look up the first VMA which intersects the interval start_addr..end_addr-1, |
| 1514 | NULL if none. Assume start_addr < end_addr. */ |
| 1515 | static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr) |
| 1516 | { |
| 1517 | struct vm_area_struct * vma = find_vma(mm,start_addr); |
| 1518 | |
| 1519 | if (vma && end_addr <= vma->vm_start) |
| 1520 | vma = NULL; |
| 1521 | return vma; |
| 1522 | } |
| 1523 | |
| 1524 | static inline unsigned long vma_pages(struct vm_area_struct *vma) |
| 1525 | { |
| 1526 | return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT; |
| 1527 | } |
| 1528 | |
David Howells | bad849b | 2010-08-26 16:00:34 +0100 | [diff] [blame] | 1529 | #ifdef CONFIG_MMU |
Hugh Dickins | 804af2c | 2006-07-26 21:39:49 +0100 | [diff] [blame] | 1530 | pgprot_t vm_get_page_prot(unsigned long vm_flags); |
David Howells | bad849b | 2010-08-26 16:00:34 +0100 | [diff] [blame] | 1531 | #else |
| 1532 | static inline pgprot_t vm_get_page_prot(unsigned long vm_flags) |
| 1533 | { |
| 1534 | return __pgprot(0); |
| 1535 | } |
| 1536 | #endif |
| 1537 | |
Hugh Dickins | deceb6c | 2005-10-29 18:16:33 -0700 | [diff] [blame] | 1538 | struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr); |
Hugh Dickins | deceb6c | 2005-10-29 18:16:33 -0700 | [diff] [blame] | 1539 | int remap_pfn_range(struct vm_area_struct *, unsigned long addr, |
| 1540 | unsigned long pfn, unsigned long size, pgprot_t); |
Linus Torvalds | a145dd4 | 2005-11-30 09:35:19 -0800 | [diff] [blame] | 1541 | int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *); |
Nick Piggin | e0dc0d8 | 2007-02-12 00:51:36 -0800 | [diff] [blame] | 1542 | int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr, |
| 1543 | unsigned long pfn); |
Nick Piggin | 423bad6 | 2008-04-28 02:13:01 -0700 | [diff] [blame] | 1544 | int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr, |
| 1545 | unsigned long pfn); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1546 | |
Linus Torvalds | 6aab341 | 2005-11-28 14:34:23 -0800 | [diff] [blame] | 1547 | struct page *follow_page(struct vm_area_struct *, unsigned long address, |
Hugh Dickins | deceb6c | 2005-10-29 18:16:33 -0700 | [diff] [blame] | 1548 | unsigned int foll_flags); |
| 1549 | #define FOLL_WRITE 0x01 /* check pte is writable */ |
| 1550 | #define FOLL_TOUCH 0x02 /* mark page accessed */ |
| 1551 | #define FOLL_GET 0x04 /* do get_page on page */ |
Hugh Dickins | 8e4b9a60 | 2009-09-21 17:03:26 -0700 | [diff] [blame] | 1552 | #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */ |
Hugh Dickins | 58fa879 | 2009-09-21 17:03:31 -0700 | [diff] [blame] | 1553 | #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */ |
Gleb Natapov | 318b275 | 2011-03-22 16:30:51 -0700 | [diff] [blame] | 1554 | #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO |
| 1555 | * and return without waiting upon it */ |
Michel Lespinasse | 110d74a | 2011-01-13 15:46:11 -0800 | [diff] [blame] | 1556 | #define FOLL_MLOCK 0x40 /* mark page as mlocked */ |
Andrea Arcangeli | 500d65d | 2011-01-13 15:46:55 -0800 | [diff] [blame] | 1557 | #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */ |
Huang Ying | 69ebb83 | 2011-01-30 11:15:48 +0800 | [diff] [blame] | 1558 | #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1559 | |
Martin Schwidefsky | 2f569af | 2008-02-08 04:22:04 -0800 | [diff] [blame] | 1560 | typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr, |
Jeremy Fitzhardinge | aee16b3 | 2007-05-06 14:48:54 -0700 | [diff] [blame] | 1561 | void *data); |
| 1562 | extern int apply_to_page_range(struct mm_struct *mm, unsigned long address, |
| 1563 | unsigned long size, pte_fn_t fn, void *data); |
| 1564 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1565 | #ifdef CONFIG_PROC_FS |
Hugh Dickins | ab50b8e | 2005-10-29 18:15:56 -0700 | [diff] [blame] | 1566 | void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1567 | #else |
Hugh Dickins | ab50b8e | 2005-10-29 18:15:56 -0700 | [diff] [blame] | 1568 | static inline void vm_stat_account(struct mm_struct *mm, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1569 | unsigned long flags, struct file *file, long pages) |
| 1570 | { |
| 1571 | } |
| 1572 | #endif /* CONFIG_PROC_FS */ |
| 1573 | |
Ingo Molnar | 12d6f21 | 2008-01-30 13:33:58 +0100 | [diff] [blame] | 1574 | #ifdef CONFIG_DEBUG_PAGEALLOC |
| 1575 | extern int debug_pagealloc_enabled; |
| 1576 | |
| 1577 | extern void kernel_map_pages(struct page *page, int numpages, int enable); |
| 1578 | |
| 1579 | static inline void enable_debug_pagealloc(void) |
| 1580 | { |
| 1581 | debug_pagealloc_enabled = 1; |
| 1582 | } |
Rafael J. Wysocki | 8a235ef | 2008-02-20 01:47:44 +0100 | [diff] [blame] | 1583 | #ifdef CONFIG_HIBERNATION |
| 1584 | extern bool kernel_page_present(struct page *page); |
| 1585 | #endif /* CONFIG_HIBERNATION */ |
Ingo Molnar | 12d6f21 | 2008-01-30 13:33:58 +0100 | [diff] [blame] | 1586 | #else |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1587 | static inline void |
Nick Piggin | 9858db5 | 2006-10-11 01:21:30 -0700 | [diff] [blame] | 1588 | kernel_map_pages(struct page *page, int numpages, int enable) {} |
Ingo Molnar | 12d6f21 | 2008-01-30 13:33:58 +0100 | [diff] [blame] | 1589 | static inline void enable_debug_pagealloc(void) |
| 1590 | { |
| 1591 | } |
Rafael J. Wysocki | 8a235ef | 2008-02-20 01:47:44 +0100 | [diff] [blame] | 1592 | #ifdef CONFIG_HIBERNATION |
| 1593 | static inline bool kernel_page_present(struct page *page) { return true; } |
| 1594 | #endif /* CONFIG_HIBERNATION */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1595 | #endif |
| 1596 | |
Stephen Wilson | 31db58b | 2011-03-13 15:49:15 -0400 | [diff] [blame] | 1597 | extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1598 | #ifdef __HAVE_ARCH_GATE_AREA |
Stephen Wilson | cae5d39 | 2011-03-13 15:49:17 -0400 | [diff] [blame] | 1599 | int in_gate_area_no_mm(unsigned long addr); |
Stephen Wilson | 83b964b | 2011-03-13 15:49:16 -0400 | [diff] [blame] | 1600 | int in_gate_area(struct mm_struct *mm, unsigned long addr); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1601 | #else |
Stephen Wilson | cae5d39 | 2011-03-13 15:49:17 -0400 | [diff] [blame] | 1602 | int in_gate_area_no_mm(unsigned long addr); |
| 1603 | #define in_gate_area(mm, addr) ({(void)mm; in_gate_area_no_mm(addr);}) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1604 | #endif /* __HAVE_ARCH_GATE_AREA */ |
| 1605 | |
Alexey Dobriyan | 8d65af7 | 2009-09-23 15:57:19 -0700 | [diff] [blame] | 1606 | int drop_caches_sysctl_handler(struct ctl_table *, int, |
Andrew Morton | 9d0243b | 2006-01-08 01:00:39 -0800 | [diff] [blame] | 1607 | void __user *, size_t *, loff_t *); |
Andrew Morton | 69e0594 | 2006-03-22 00:08:19 -0800 | [diff] [blame] | 1608 | unsigned long shrink_slab(unsigned long scanned, gfp_t gfp_mask, |
Andrew Morton | 9d0243b | 2006-01-08 01:00:39 -0800 | [diff] [blame] | 1609 | unsigned long lru_pages); |
Andrew Morton | 9d0243b | 2006-01-08 01:00:39 -0800 | [diff] [blame] | 1610 | |
Luke Yang | 7a9166e | 2006-02-20 18:28:07 -0800 | [diff] [blame] | 1611 | #ifndef CONFIG_MMU |
| 1612 | #define randomize_va_space 0 |
| 1613 | #else |
Andi Kleen | a62eaf1 | 2006-02-16 23:41:58 +0100 | [diff] [blame] | 1614 | extern int randomize_va_space; |
Luke Yang | 7a9166e | 2006-02-20 18:28:07 -0800 | [diff] [blame] | 1615 | #endif |
Andi Kleen | a62eaf1 | 2006-02-16 23:41:58 +0100 | [diff] [blame] | 1616 | |
Sam Ravnborg | 045e72a | 2007-07-26 10:41:13 -0700 | [diff] [blame] | 1617 | const char * arch_vma_name(struct vm_area_struct *vma); |
Andi Kleen | 0325291 | 2008-01-30 13:33:18 +0100 | [diff] [blame] | 1618 | void print_vma_addr(char *prefix, unsigned long rip); |
Ingo Molnar | e6e5494 | 2006-06-27 02:53:50 -0700 | [diff] [blame] | 1619 | |
Yinghai Lu | 9bdac91 | 2010-02-10 01:20:22 -0800 | [diff] [blame] | 1620 | void sparse_mem_maps_populate_node(struct page **map_map, |
| 1621 | unsigned long pnum_begin, |
| 1622 | unsigned long pnum_end, |
| 1623 | unsigned long map_count, |
| 1624 | int nodeid); |
| 1625 | |
Yasunori Goto | 98f3cfc | 2007-10-16 01:26:14 -0700 | [diff] [blame] | 1626 | struct page *sparse_mem_map_populate(unsigned long pnum, int nid); |
Andy Whitcroft | 29c7111 | 2007-10-16 01:24:14 -0700 | [diff] [blame] | 1627 | pgd_t *vmemmap_pgd_populate(unsigned long addr, int node); |
| 1628 | pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node); |
| 1629 | pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node); |
| 1630 | pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node); |
Christoph Lameter | 8f6aac4 | 2007-10-16 01:24:13 -0700 | [diff] [blame] | 1631 | void *vmemmap_alloc_block(unsigned long size, int node); |
Yinghai Lu | 9bdac91 | 2010-02-10 01:20:22 -0800 | [diff] [blame] | 1632 | void *vmemmap_alloc_block_buf(unsigned long size, int node); |
Christoph Lameter | 8f6aac4 | 2007-10-16 01:24:13 -0700 | [diff] [blame] | 1633 | void vmemmap_verify(pte_t *, int, unsigned long, unsigned long); |
Andy Whitcroft | 29c7111 | 2007-10-16 01:24:14 -0700 | [diff] [blame] | 1634 | int vmemmap_populate_basepages(struct page *start_page, |
| 1635 | unsigned long pages, int node); |
| 1636 | int vmemmap_populate(struct page *start_page, unsigned long pages, int node); |
Yinghai Lu | c2b91e2 | 2008-04-12 01:19:24 -0700 | [diff] [blame] | 1637 | void vmemmap_populate_print_last(void); |
Christoph Lameter | 8f6aac4 | 2007-10-16 01:24:13 -0700 | [diff] [blame] | 1638 | |
Andi Kleen | 6a46079 | 2009-09-16 11:50:15 +0200 | [diff] [blame] | 1639 | |
Andi Kleen | 82ba011 | 2009-12-16 12:19:57 +0100 | [diff] [blame] | 1640 | enum mf_flags { |
| 1641 | MF_COUNT_INCREASED = 1 << 0, |
| 1642 | }; |
Andi Kleen | 6a46079 | 2009-09-16 11:50:15 +0200 | [diff] [blame] | 1643 | extern void memory_failure(unsigned long pfn, int trapno); |
Andi Kleen | 82ba011 | 2009-12-16 12:19:57 +0100 | [diff] [blame] | 1644 | extern int __memory_failure(unsigned long pfn, int trapno, int flags); |
Wu Fengguang | 847ce40 | 2009-12-16 12:19:58 +0100 | [diff] [blame] | 1645 | extern int unpoison_memory(unsigned long pfn); |
Andi Kleen | 6a46079 | 2009-09-16 11:50:15 +0200 | [diff] [blame] | 1646 | extern int sysctl_memory_failure_early_kill; |
| 1647 | extern int sysctl_memory_failure_recovery; |
Andi Kleen | facb601 | 2009-12-16 12:20:00 +0100 | [diff] [blame] | 1648 | extern void shake_page(struct page *p, int access); |
Andi Kleen | 6a46079 | 2009-09-16 11:50:15 +0200 | [diff] [blame] | 1649 | extern atomic_long_t mce_bad_pages; |
Andi Kleen | facb601 | 2009-12-16 12:20:00 +0100 | [diff] [blame] | 1650 | extern int soft_offline_page(struct page *page, int flags); |
Andi Kleen | 6a46079 | 2009-09-16 11:50:15 +0200 | [diff] [blame] | 1651 | |
Wu Fengguang | 718a382 | 2010-03-10 15:20:43 -0800 | [diff] [blame] | 1652 | extern void dump_page(struct page *page); |
| 1653 | |
Andrea Arcangeli | 47ad847 | 2011-01-13 15:46:47 -0800 | [diff] [blame] | 1654 | #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS) |
| 1655 | extern void clear_huge_page(struct page *page, |
| 1656 | unsigned long addr, |
| 1657 | unsigned int pages_per_huge_page); |
| 1658 | extern void copy_user_huge_page(struct page *dst, struct page *src, |
| 1659 | unsigned long addr, struct vm_area_struct *vma, |
| 1660 | unsigned int pages_per_huge_page); |
| 1661 | #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */ |
| 1662 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1663 | #endif /* __KERNEL__ */ |
| 1664 | #endif /* _LINUX_MM_H */ |