Heiko Carstens | 5b99cd0 | 2006-09-27 01:50:01 -0700 | [diff] [blame] | 1 | #ifndef _LINUX_MM_TYPES_H |
| 2 | #define _LINUX_MM_TYPES_H |
| 3 | |
Olaf Hering | 4f9a58d | 2007-10-16 23:30:12 -0700 | [diff] [blame] | 4 | #include <linux/auxvec.h> |
Heiko Carstens | 5b99cd0 | 2006-09-27 01:50:01 -0700 | [diff] [blame] | 5 | #include <linux/types.h> |
| 6 | #include <linux/threads.h> |
| 7 | #include <linux/list.h> |
| 8 | #include <linux/spinlock.h> |
Martin Schwidefsky | c92ff1b | 2007-10-16 01:24:43 -0700 | [diff] [blame] | 9 | #include <linux/rbtree.h> |
| 10 | #include <linux/rwsem.h> |
| 11 | #include <linux/completion.h> |
Andrea Arcangeli | cddb8a5 | 2008-07-28 15:46:29 -0700 | [diff] [blame] | 12 | #include <linux/cpumask.h> |
Srikar Dronamraju | d4b3b63 | 2012-03-30 23:56:31 +0530 | [diff] [blame] | 13 | #include <linux/uprobes.h> |
Peter Zijlstra | bbeae5b | 2013-02-22 16:34:30 -0800 | [diff] [blame] | 14 | #include <linux/page-flags-layout.h> |
Martin Schwidefsky | c92ff1b | 2007-10-16 01:24:43 -0700 | [diff] [blame] | 15 | #include <asm/page.h> |
| 16 | #include <asm/mmu.h> |
Heiko Carstens | 5b99cd0 | 2006-09-27 01:50:01 -0700 | [diff] [blame] | 17 | |
Olaf Hering | 4f9a58d | 2007-10-16 23:30:12 -0700 | [diff] [blame] | 18 | #ifndef AT_VECTOR_SIZE_ARCH |
| 19 | #define AT_VECTOR_SIZE_ARCH 0 |
| 20 | #endif |
| 21 | #define AT_VECTOR_SIZE (2*(AT_VECTOR_SIZE_ARCH + AT_VECTOR_SIZE_BASE + 1)) |
| 22 | |
Heiko Carstens | 5b99cd0 | 2006-09-27 01:50:01 -0700 | [diff] [blame] | 23 | struct address_space; |
Johannes Weiner | 1306a85 | 2014-12-10 15:44:52 -0800 | [diff] [blame] | 24 | struct mem_cgroup; |
Heiko Carstens | 5b99cd0 | 2006-09-27 01:50:01 -0700 | [diff] [blame] | 25 | |
Kirill A. Shutemov | 57c1ffc | 2013-11-14 14:30:45 -0800 | [diff] [blame] | 26 | #define USE_SPLIT_PTE_PTLOCKS (NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS) |
Kirill A. Shutemov | e009bb3 | 2013-11-14 14:31:07 -0800 | [diff] [blame] | 27 | #define USE_SPLIT_PMD_PTLOCKS (USE_SPLIT_PTE_PTLOCKS && \ |
| 28 | IS_ENABLED(CONFIG_ARCH_ENABLE_SPLIT_PMD_PTLOCK)) |
Kirill A. Shutemov | 597d795 | 2013-12-20 13:35:58 +0200 | [diff] [blame] | 29 | #define ALLOC_SPLIT_PTLOCKS (SPINLOCK_SIZE > BITS_PER_LONG/8) |
Jeremy Fitzhardinge | f7d0b92 | 2008-09-09 15:43:22 -0700 | [diff] [blame] | 30 | |
Kirill A. Shutemov | e4b294c | 2015-02-11 15:24:46 -0800 | [diff] [blame] | 31 | typedef void compound_page_dtor(struct page *); |
| 32 | |
Heiko Carstens | 5b99cd0 | 2006-09-27 01:50:01 -0700 | [diff] [blame] | 33 | /* |
| 34 | * Each physical page in the system has a struct page associated with |
| 35 | * it to keep track of whatever it is we are using the page for at the |
| 36 | * moment. Note that we have no way to track which tasks are using |
| 37 | * a page, though if it is a pagecache page, rmap structures can tell us |
| 38 | * who is mapping it. |
Christoph Lameter | fc9bb8c | 2011-06-01 12:25:48 -0500 | [diff] [blame] | 39 | * |
| 40 | * The objects in struct page are organized in double word blocks in |
| 41 | * order to allows us to use atomic double word operations on portions |
| 42 | * of struct page. That is currently only used by slub but the arrangement |
| 43 | * allows the use of atomic double word operations on the flags/mapping |
| 44 | * and lru list pointers also. |
Heiko Carstens | 5b99cd0 | 2006-09-27 01:50:01 -0700 | [diff] [blame] | 45 | */ |
| 46 | struct page { |
Christoph Lameter | fc9bb8c | 2011-06-01 12:25:48 -0500 | [diff] [blame] | 47 | /* First double word block */ |
Heiko Carstens | 5b99cd0 | 2006-09-27 01:50:01 -0700 | [diff] [blame] | 48 | unsigned long flags; /* Atomic flags, some possibly |
| 49 | * updated asynchronously */ |
Joonsoo Kim | 8456a64 | 2013-10-24 10:07:49 +0900 | [diff] [blame] | 50 | union { |
| 51 | struct address_space *mapping; /* If low bit clear, points to |
| 52 | * inode address_space, or NULL. |
| 53 | * If page mapped as anonymous |
| 54 | * memory, low bit is set, and |
| 55 | * it points to anon_vma object: |
| 56 | * see PAGE_MAPPING_ANON below. |
| 57 | */ |
| 58 | void *s_mem; /* slab first object */ |
| 59 | }; |
| 60 | |
Christoph Lameter | fc9bb8c | 2011-06-01 12:25:48 -0500 | [diff] [blame] | 61 | /* Second double word */ |
Christoph Lameter | 013e896 | 2011-07-14 12:48:14 -0500 | [diff] [blame] | 62 | struct { |
| 63 | union { |
Christoph Lameter | fc9bb8c | 2011-06-01 12:25:48 -0500 | [diff] [blame] | 64 | pgoff_t index; /* Our offset within mapping. */ |
Joonsoo Kim | 8456a64 | 2013-10-24 10:07:49 +0900 | [diff] [blame] | 65 | void *freelist; /* sl[aou]b first free object */ |
Mel Gorman | 072bb0a | 2012-07-31 16:43:58 -0700 | [diff] [blame] | 66 | bool pfmemalloc; /* If set by the page allocator, |
Mel Gorman | b37f1dd | 2012-07-31 16:44:03 -0700 | [diff] [blame] | 67 | * ALLOC_NO_WATERMARKS was set |
Mel Gorman | 072bb0a | 2012-07-31 16:43:58 -0700 | [diff] [blame] | 68 | * and the low watermark was not |
| 69 | * met implying that the system |
| 70 | * is under some pressure. The |
| 71 | * caller should try ensure |
| 72 | * this page is only used to |
| 73 | * free other pages. |
| 74 | */ |
Christoph Lameter | 013e896 | 2011-07-14 12:48:14 -0500 | [diff] [blame] | 75 | }; |
| 76 | |
| 77 | union { |
Pravin B Shelar | abca7c4 | 2012-06-20 12:52:56 -0700 | [diff] [blame] | 78 | #if defined(CONFIG_HAVE_CMPXCHG_DOUBLE) && \ |
| 79 | defined(CONFIG_HAVE_ALIGNED_STRUCT_PAGE) |
Christoph Lameter | 013e896 | 2011-07-14 12:48:14 -0500 | [diff] [blame] | 80 | /* Used for cmpxchg_double in slub */ |
| 81 | unsigned long counters; |
Pravin B Shelar | abca7c4 | 2012-06-20 12:52:56 -0700 | [diff] [blame] | 82 | #else |
| 83 | /* |
| 84 | * Keep _count separate from slub cmpxchg_double data. |
| 85 | * As the rest of the double word is protected by |
| 86 | * slab_lock but _count is not. |
| 87 | */ |
| 88 | unsigned counters; |
| 89 | #endif |
Christoph Lameter | 013e896 | 2011-07-14 12:48:14 -0500 | [diff] [blame] | 90 | |
| 91 | struct { |
| 92 | |
| 93 | union { |
Andrea Arcangeli | 70b50f9 | 2011-11-02 13:36:59 -0700 | [diff] [blame] | 94 | /* |
| 95 | * Count of ptes mapped in |
| 96 | * mms, to show when page is |
| 97 | * mapped & limit reverse map |
| 98 | * searches. |
| 99 | * |
| 100 | * Used also for tail pages |
| 101 | * refcounting instead of |
| 102 | * _count. Tail pages cannot |
| 103 | * be mapped and keeping the |
| 104 | * tail page _count zero at |
| 105 | * all times guarantees |
| 106 | * get_page_unless_zero() will |
| 107 | * never succeed on tail |
| 108 | * pages. |
| 109 | */ |
| 110 | atomic_t _mapcount; |
Christoph Lameter | fc9bb8c | 2011-06-01 12:25:48 -0500 | [diff] [blame] | 111 | |
Christoph Lameter | b8c24c4 | 2012-06-13 10:24:52 -0500 | [diff] [blame] | 112 | struct { /* SLUB */ |
Christoph Lameter | 013e896 | 2011-07-14 12:48:14 -0500 | [diff] [blame] | 113 | unsigned inuse:16; |
| 114 | unsigned objects:15; |
| 115 | unsigned frozen:1; |
| 116 | }; |
Christoph Lameter | b8c24c4 | 2012-06-13 10:24:52 -0500 | [diff] [blame] | 117 | int units; /* SLOB */ |
Pekka Enberg | 3adf004 | 2011-07-18 15:16:55 +0300 | [diff] [blame] | 118 | }; |
Christoph Lameter | 013e896 | 2011-07-14 12:48:14 -0500 | [diff] [blame] | 119 | atomic_t _count; /* Usage count, see below. */ |
Christoph Lameter | fc9bb8c | 2011-06-01 12:25:48 -0500 | [diff] [blame] | 120 | }; |
Joonsoo Kim | 8456a64 | 2013-10-24 10:07:49 +0900 | [diff] [blame] | 121 | unsigned int active; /* SLAB */ |
Christoph Lameter | 39b2646 | 2008-04-14 19:11:30 +0300 | [diff] [blame] | 122 | }; |
Christoph Lameter | 81819f0 | 2007-05-06 14:49:36 -0700 | [diff] [blame] | 123 | }; |
Christoph Lameter | fc9bb8c | 2011-06-01 12:25:48 -0500 | [diff] [blame] | 124 | |
| 125 | /* Third double word block */ |
Christoph Lameter | 49e2258 | 2011-08-09 16:12:27 -0500 | [diff] [blame] | 126 | union { |
| 127 | struct list_head lru; /* Pageout list, eg. active_list |
Christoph Lameter | fc9bb8c | 2011-06-01 12:25:48 -0500 | [diff] [blame] | 128 | * protected by zone->lru_lock ! |
Dave Hansen | 34bf6ef | 2014-04-08 13:44:27 -0700 | [diff] [blame] | 129 | * Can be used as a generic list |
| 130 | * by the page owner. |
Christoph Lameter | fc9bb8c | 2011-06-01 12:25:48 -0500 | [diff] [blame] | 131 | */ |
Christoph Lameter | 49e2258 | 2011-08-09 16:12:27 -0500 | [diff] [blame] | 132 | struct { /* slub per cpu partial pages */ |
| 133 | struct page *next; /* Next partial slab */ |
| 134 | #ifdef CONFIG_64BIT |
| 135 | int pages; /* Nr of partial slabs left */ |
| 136 | int pobjects; /* Approximate # of objects */ |
| 137 | #else |
| 138 | short int pages; |
| 139 | short int pobjects; |
| 140 | #endif |
| 141 | }; |
Christoph Lameter | b8c24c4 | 2012-06-13 10:24:52 -0500 | [diff] [blame] | 142 | |
Glauber Costa | 1b4f59e3 | 2012-10-22 18:05:36 +0400 | [diff] [blame] | 143 | struct slab *slab_page; /* slab fields */ |
Joonsoo Kim | 6812670 | 2013-10-24 10:07:42 +0900 | [diff] [blame] | 144 | struct rcu_head rcu_head; /* Used by SLAB |
| 145 | * when destroying via RCU |
| 146 | */ |
Kirill A. Shutemov | e4b294c | 2015-02-11 15:24:46 -0800 | [diff] [blame] | 147 | /* First tail page of compound page */ |
| 148 | struct { |
| 149 | compound_page_dtor *compound_dtor; |
| 150 | unsigned long compound_order; |
| 151 | }; |
| 152 | |
Kirill A. Shutemov | 7aa555b | 2013-11-21 14:32:11 -0800 | [diff] [blame] | 153 | #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && USE_SPLIT_PMD_PTLOCKS |
| 154 | pgtable_t pmd_huge_pte; /* protected by page->ptl */ |
| 155 | #endif |
Christoph Lameter | 49e2258 | 2011-08-09 16:12:27 -0500 | [diff] [blame] | 156 | }; |
Christoph Lameter | fc9bb8c | 2011-06-01 12:25:48 -0500 | [diff] [blame] | 157 | |
| 158 | /* Remainder is not double word aligned */ |
Heiko Carstens | 5b99cd0 | 2006-09-27 01:50:01 -0700 | [diff] [blame] | 159 | union { |
Heiko Carstens | 5b99cd0 | 2006-09-27 01:50:01 -0700 | [diff] [blame] | 160 | unsigned long private; /* Mapping-private opaque data: |
| 161 | * usually used for buffer_heads |
| 162 | * if PagePrivate set; used for |
| 163 | * swp_entry_t if PageSwapCache; |
| 164 | * indicates order in the buddy |
| 165 | * system if PG_buddy is set. |
| 166 | */ |
Kirill A. Shutemov | 57c1ffc | 2013-11-14 14:30:45 -0800 | [diff] [blame] | 167 | #if USE_SPLIT_PTE_PTLOCKS |
Kirill A. Shutemov | 597d795 | 2013-12-20 13:35:58 +0200 | [diff] [blame] | 168 | #if ALLOC_SPLIT_PTLOCKS |
Peter Zijlstra | 539edb5 | 2013-11-14 14:31:52 -0800 | [diff] [blame] | 169 | spinlock_t *ptl; |
| 170 | #else |
| 171 | spinlock_t ptl; |
| 172 | #endif |
Heiko Carstens | 5b99cd0 | 2006-09-27 01:50:01 -0700 | [diff] [blame] | 173 | #endif |
Glauber Costa | 1b4f59e3 | 2012-10-22 18:05:36 +0400 | [diff] [blame] | 174 | struct kmem_cache *slab_cache; /* SL[AU]B: Pointer to slab */ |
Christoph Lameter | fc9bb8c | 2011-06-01 12:25:48 -0500 | [diff] [blame] | 175 | struct page *first_page; /* Compound tail pages */ |
Heiko Carstens | 5b99cd0 | 2006-09-27 01:50:01 -0700 | [diff] [blame] | 176 | }; |
Christoph Lameter | fc9bb8c | 2011-06-01 12:25:48 -0500 | [diff] [blame] | 177 | |
Johannes Weiner | 1306a85 | 2014-12-10 15:44:52 -0800 | [diff] [blame] | 178 | #ifdef CONFIG_MEMCG |
| 179 | struct mem_cgroup *mem_cgroup; |
| 180 | #endif |
| 181 | |
Heiko Carstens | 5b99cd0 | 2006-09-27 01:50:01 -0700 | [diff] [blame] | 182 | /* |
| 183 | * On machines where all RAM is mapped into kernel address space, |
| 184 | * we can simply calculate the virtual address. On machines with |
| 185 | * highmem some memory is mapped into kernel virtual memory |
| 186 | * dynamically, so we need a place to store that address. |
| 187 | * Note that this field could be 16 bits on x86 ... ;) |
| 188 | * |
| 189 | * Architectures with slow multiplication can define |
| 190 | * WANT_PAGE_VIRTUAL in asm/page.h |
| 191 | */ |
| 192 | #if defined(WANT_PAGE_VIRTUAL) |
| 193 | void *virtual; /* Kernel virtual address (NULL if |
| 194 | not kmapped, ie. highmem) */ |
| 195 | #endif /* WANT_PAGE_VIRTUAL */ |
Vegard Nossum | dfec072 | 2008-04-04 00:51:41 +0200 | [diff] [blame] | 196 | |
| 197 | #ifdef CONFIG_KMEMCHECK |
| 198 | /* |
| 199 | * kmemcheck wants to track the status of each byte in a page; this |
| 200 | * is a pointer to such a status block. NULL if not tracked. |
| 201 | */ |
| 202 | void *shadow; |
| 203 | #endif |
Mel Gorman | 57e0a03 | 2012-11-12 09:06:20 +0000 | [diff] [blame] | 204 | |
Peter Zijlstra | 9057289 | 2013-10-07 11:29:20 +0100 | [diff] [blame] | 205 | #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS |
| 206 | int _last_cpupid; |
Mel Gorman | 57e0a03 | 2012-11-12 09:06:20 +0000 | [diff] [blame] | 207 | #endif |
Christoph Lameter | fc9bb8c | 2011-06-01 12:25:48 -0500 | [diff] [blame] | 208 | } |
| 209 | /* |
Heiko Carstens | 43570fd | 2012-01-12 17:17:27 -0800 | [diff] [blame] | 210 | * The struct page can be forced to be double word aligned so that atomic ops |
| 211 | * on double words work. The SLUB allocator can make use of such a feature. |
Christoph Lameter | fc9bb8c | 2011-06-01 12:25:48 -0500 | [diff] [blame] | 212 | */ |
Heiko Carstens | 43570fd | 2012-01-12 17:17:27 -0800 | [diff] [blame] | 213 | #ifdef CONFIG_HAVE_ALIGNED_STRUCT_PAGE |
| 214 | __aligned(2 * sizeof(unsigned long)) |
Christoph Lameter | fc9bb8c | 2011-06-01 12:25:48 -0500 | [diff] [blame] | 215 | #endif |
| 216 | ; |
Heiko Carstens | 5b99cd0 | 2006-09-27 01:50:01 -0700 | [diff] [blame] | 217 | |
Ian Campbell | 30d3c12 | 2011-10-20 04:58:32 -0400 | [diff] [blame] | 218 | struct page_frag { |
| 219 | struct page *page; |
| 220 | #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536) |
| 221 | __u32 offset; |
| 222 | __u32 size; |
| 223 | #else |
| 224 | __u16 offset; |
| 225 | __u16 size; |
| 226 | #endif |
| 227 | }; |
| 228 | |
KOSAKI Motohiro | ca16d14 | 2011-05-26 19:16:19 +0900 | [diff] [blame] | 229 | typedef unsigned long __nocast vm_flags_t; |
| 230 | |
Martin Schwidefsky | c92ff1b | 2007-10-16 01:24:43 -0700 | [diff] [blame] | 231 | /* |
David Howells | 8feae13 | 2009-01-08 12:04:47 +0000 | [diff] [blame] | 232 | * A region containing a mapping of a non-memory backed file under NOMMU |
| 233 | * conditions. These are held in a global tree and are pinned by the VMAs that |
| 234 | * map parts of them. |
| 235 | */ |
| 236 | struct vm_region { |
| 237 | struct rb_node vm_rb; /* link in global region tree */ |
KOSAKI Motohiro | ca16d14 | 2011-05-26 19:16:19 +0900 | [diff] [blame] | 238 | vm_flags_t vm_flags; /* VMA vm_flags */ |
David Howells | 8feae13 | 2009-01-08 12:04:47 +0000 | [diff] [blame] | 239 | unsigned long vm_start; /* start address of region */ |
| 240 | unsigned long vm_end; /* region initialised to here */ |
Paul Mundt | dd8632a | 2009-01-08 12:04:47 +0000 | [diff] [blame] | 241 | unsigned long vm_top; /* region allocated to here */ |
David Howells | 8feae13 | 2009-01-08 12:04:47 +0000 | [diff] [blame] | 242 | unsigned long vm_pgoff; /* the offset in vm_file corresponding to vm_start */ |
| 243 | struct file *vm_file; /* the backing file or NULL */ |
| 244 | |
David Howells | 1e2ae59 | 2010-01-15 17:01:33 -0800 | [diff] [blame] | 245 | int vm_usage; /* region usage count (access under nommu_region_sem) */ |
Mike Frysinger | cfe79c0 | 2010-01-06 17:23:23 +0000 | [diff] [blame] | 246 | bool vm_icache_flushed : 1; /* true if the icache has been flushed for |
| 247 | * this region */ |
David Howells | 8feae13 | 2009-01-08 12:04:47 +0000 | [diff] [blame] | 248 | }; |
| 249 | |
| 250 | /* |
Martin Schwidefsky | c92ff1b | 2007-10-16 01:24:43 -0700 | [diff] [blame] | 251 | * This struct defines a memory VMM memory area. There is one of these |
| 252 | * per VM-area/task. A VM area is any part of the process virtual memory |
| 253 | * space that has a special rule for the page-fault handlers (ie a shared |
| 254 | * library, the executable area etc). |
| 255 | */ |
| 256 | struct vm_area_struct { |
Rik van Riel | e4c6bfd | 2012-12-11 16:01:44 -0800 | [diff] [blame] | 257 | /* The first cache line has the info for VMA tree walking. */ |
| 258 | |
Martin Schwidefsky | c92ff1b | 2007-10-16 01:24:43 -0700 | [diff] [blame] | 259 | unsigned long vm_start; /* Our start address within vm_mm. */ |
| 260 | unsigned long vm_end; /* The first byte after our end address |
| 261 | within vm_mm. */ |
| 262 | |
| 263 | /* linked list of VM areas per task, sorted by address */ |
Linus Torvalds | 297c5ee | 2010-08-20 16:24:55 -0700 | [diff] [blame] | 264 | struct vm_area_struct *vm_next, *vm_prev; |
Martin Schwidefsky | c92ff1b | 2007-10-16 01:24:43 -0700 | [diff] [blame] | 265 | |
Martin Schwidefsky | c92ff1b | 2007-10-16 01:24:43 -0700 | [diff] [blame] | 266 | struct rb_node vm_rb; |
| 267 | |
| 268 | /* |
Michel Lespinasse | d373718 | 2012-12-11 16:01:38 -0800 | [diff] [blame] | 269 | * Largest free memory gap in bytes to the left of this VMA. |
| 270 | * Either between this VMA and vma->vm_prev, or between one of the |
| 271 | * VMAs below us in the VMA rbtree and its ->vm_prev. This helps |
| 272 | * get_unmapped_area find a free area of the right size. |
| 273 | */ |
| 274 | unsigned long rb_subtree_gap; |
| 275 | |
Rik van Riel | e4c6bfd | 2012-12-11 16:01:44 -0800 | [diff] [blame] | 276 | /* Second cache line starts here. */ |
| 277 | |
| 278 | struct mm_struct *vm_mm; /* The address space we belong to. */ |
| 279 | pgprot_t vm_page_prot; /* Access permissions of this VMA. */ |
| 280 | unsigned long vm_flags; /* Flags, see mm.h. */ |
| 281 | |
Michel Lespinasse | d373718 | 2012-12-11 16:01:38 -0800 | [diff] [blame] | 282 | /* |
Martin Schwidefsky | c92ff1b | 2007-10-16 01:24:43 -0700 | [diff] [blame] | 283 | * For areas with an address space and backing store, |
Kirill A. Shutemov | 27ba064 | 2015-02-10 14:09:59 -0800 | [diff] [blame] | 284 | * linkage into the address_space->i_mmap interval tree. |
Martin Schwidefsky | c92ff1b | 2007-10-16 01:24:43 -0700 | [diff] [blame] | 285 | */ |
Kirill A. Shutemov | ac51b93 | 2015-02-10 14:10:02 -0800 | [diff] [blame] | 286 | struct { |
| 287 | struct rb_node rb; |
| 288 | unsigned long rb_subtree_last; |
Martin Schwidefsky | c92ff1b | 2007-10-16 01:24:43 -0700 | [diff] [blame] | 289 | } shared; |
| 290 | |
| 291 | /* |
| 292 | * A file's MAP_PRIVATE vma can be in both i_mmap tree and anon_vma |
| 293 | * list, after a COW of one of the file pages. A MAP_SHARED vma |
| 294 | * can only be in the i_mmap tree. An anonymous MAP_PRIVATE, stack |
| 295 | * or brk vma (with NULL file) can only be in an anon_vma list. |
| 296 | */ |
Rik van Riel | 5beb493 | 2010-03-05 13:42:07 -0800 | [diff] [blame] | 297 | struct list_head anon_vma_chain; /* Serialized by mmap_sem & |
| 298 | * page_table_lock */ |
Martin Schwidefsky | c92ff1b | 2007-10-16 01:24:43 -0700 | [diff] [blame] | 299 | struct anon_vma *anon_vma; /* Serialized by page_table_lock */ |
| 300 | |
| 301 | /* Function pointers to deal with this struct. */ |
Alexey Dobriyan | f0f37e2 | 2009-09-27 22:29:37 +0400 | [diff] [blame] | 302 | const struct vm_operations_struct *vm_ops; |
Martin Schwidefsky | c92ff1b | 2007-10-16 01:24:43 -0700 | [diff] [blame] | 303 | |
| 304 | /* Information about our backing store: */ |
| 305 | unsigned long vm_pgoff; /* Offset (within vm_file) in PAGE_SIZE |
| 306 | units, *not* PAGE_CACHE_SIZE */ |
| 307 | struct file * vm_file; /* File we map to (can be NULL). */ |
| 308 | void * vm_private_data; /* was vm_pte (shared mem) */ |
Martin Schwidefsky | c92ff1b | 2007-10-16 01:24:43 -0700 | [diff] [blame] | 309 | |
| 310 | #ifndef CONFIG_MMU |
David Howells | 8feae13 | 2009-01-08 12:04:47 +0000 | [diff] [blame] | 311 | struct vm_region *vm_region; /* NOMMU mapping region */ |
Martin Schwidefsky | c92ff1b | 2007-10-16 01:24:43 -0700 | [diff] [blame] | 312 | #endif |
| 313 | #ifdef CONFIG_NUMA |
| 314 | struct mempolicy *vm_policy; /* NUMA policy for the VMA */ |
| 315 | #endif |
| 316 | }; |
| 317 | |
Oleg Nesterov | b564daf | 2008-07-25 01:47:44 -0700 | [diff] [blame] | 318 | struct core_thread { |
| 319 | struct task_struct *task; |
| 320 | struct core_thread *next; |
| 321 | }; |
| 322 | |
Oleg Nesterov | 32ecb1f | 2008-07-25 01:47:41 -0700 | [diff] [blame] | 323 | struct core_state { |
Oleg Nesterov | c5f1cc8 | 2008-07-25 01:47:42 -0700 | [diff] [blame] | 324 | atomic_t nr_threads; |
Oleg Nesterov | b564daf | 2008-07-25 01:47:44 -0700 | [diff] [blame] | 325 | struct core_thread dumper; |
Oleg Nesterov | 32ecb1f | 2008-07-25 01:47:41 -0700 | [diff] [blame] | 326 | struct completion startup; |
| 327 | }; |
| 328 | |
KAMEZAWA Hiroyuki | d559db0 | 2010-03-05 13:41:39 -0800 | [diff] [blame] | 329 | enum { |
| 330 | MM_FILEPAGES, |
| 331 | MM_ANONPAGES, |
KAMEZAWA Hiroyuki | b084d43 | 2010-03-05 13:41:42 -0800 | [diff] [blame] | 332 | MM_SWAPENTS, |
KAMEZAWA Hiroyuki | d559db0 | 2010-03-05 13:41:39 -0800 | [diff] [blame] | 333 | NR_MM_COUNTERS |
| 334 | }; |
| 335 | |
Kirill A. Shutemov | 57c1ffc | 2013-11-14 14:30:45 -0800 | [diff] [blame] | 336 | #if USE_SPLIT_PTE_PTLOCKS && defined(CONFIG_MMU) |
KAMEZAWA Hiroyuki | 34e5523 | 2010-03-05 13:41:40 -0800 | [diff] [blame] | 337 | #define SPLIT_RSS_COUNTING |
KAMEZAWA Hiroyuki | 34e5523 | 2010-03-05 13:41:40 -0800 | [diff] [blame] | 338 | /* per-thread cached information, */ |
| 339 | struct task_rss_stat { |
| 340 | int events; /* for synchronization threshold */ |
| 341 | int count[NR_MM_COUNTERS]; |
| 342 | }; |
Kirill A. Shutemov | 57c1ffc | 2013-11-14 14:30:45 -0800 | [diff] [blame] | 343 | #endif /* USE_SPLIT_PTE_PTLOCKS */ |
Matt Fleming | 172703b | 2011-05-24 17:12:36 -0700 | [diff] [blame] | 344 | |
KAMEZAWA Hiroyuki | d559db0 | 2010-03-05 13:41:39 -0800 | [diff] [blame] | 345 | struct mm_rss_stat { |
Matt Fleming | 172703b | 2011-05-24 17:12:36 -0700 | [diff] [blame] | 346 | atomic_long_t count[NR_MM_COUNTERS]; |
KAMEZAWA Hiroyuki | d559db0 | 2010-03-05 13:41:39 -0800 | [diff] [blame] | 347 | }; |
KAMEZAWA Hiroyuki | d559db0 | 2010-03-05 13:41:39 -0800 | [diff] [blame] | 348 | |
Benjamin LaHaise | db446a0 | 2013-07-30 12:54:40 -0400 | [diff] [blame] | 349 | struct kioctx_table; |
Martin Schwidefsky | c92ff1b | 2007-10-16 01:24:43 -0700 | [diff] [blame] | 350 | struct mm_struct { |
Davidlohr Bueso | 615d6e8 | 2014-04-07 15:37:25 -0700 | [diff] [blame] | 351 | struct vm_area_struct *mmap; /* list of VMAs */ |
Martin Schwidefsky | c92ff1b | 2007-10-16 01:24:43 -0700 | [diff] [blame] | 352 | struct rb_root mm_rb; |
Davidlohr Bueso | 615d6e8 | 2014-04-07 15:37:25 -0700 | [diff] [blame] | 353 | u32 vmacache_seqnum; /* per-thread vmacache */ |
David Howells | efc1a3b | 2010-01-15 17:01:35 -0800 | [diff] [blame] | 354 | #ifdef CONFIG_MMU |
Martin Schwidefsky | c92ff1b | 2007-10-16 01:24:43 -0700 | [diff] [blame] | 355 | unsigned long (*get_unmapped_area) (struct file *filp, |
| 356 | unsigned long addr, unsigned long len, |
| 357 | unsigned long pgoff, unsigned long flags); |
David Howells | efc1a3b | 2010-01-15 17:01:35 -0800 | [diff] [blame] | 358 | #endif |
Martin Schwidefsky | c92ff1b | 2007-10-16 01:24:43 -0700 | [diff] [blame] | 359 | unsigned long mmap_base; /* base of mmap area */ |
Radu Caragea | 41aacc1 | 2013-08-21 20:55:59 +0300 | [diff] [blame] | 360 | unsigned long mmap_legacy_base; /* base of mmap area in bottom-up allocations */ |
Martin Schwidefsky | c92ff1b | 2007-10-16 01:24:43 -0700 | [diff] [blame] | 361 | unsigned long task_size; /* size of task vm space */ |
Michel Lespinasse | d373718 | 2012-12-11 16:01:38 -0800 | [diff] [blame] | 362 | unsigned long highest_vm_end; /* highest vma end address */ |
Martin Schwidefsky | c92ff1b | 2007-10-16 01:24:43 -0700 | [diff] [blame] | 363 | pgd_t * pgd; |
| 364 | atomic_t mm_users; /* How many users with user space? */ |
| 365 | atomic_t mm_count; /* How many references to "struct mm_struct" (users count as 1) */ |
Kirill A. Shutemov | dc6c9a3 | 2015-02-11 15:26:50 -0800 | [diff] [blame] | 366 | atomic_long_t nr_ptes; /* PTE page table pages */ |
Kirill A. Shutemov | 5a3fbef | 2015-04-14 15:46:21 -0700 | [diff] [blame] | 367 | #if CONFIG_PGTABLE_LEVELS > 2 |
Kirill A. Shutemov | dc6c9a3 | 2015-02-11 15:26:50 -0800 | [diff] [blame] | 368 | atomic_long_t nr_pmds; /* PMD page table pages */ |
Kirill A. Shutemov | 5a3fbef | 2015-04-14 15:46:21 -0700 | [diff] [blame] | 369 | #endif |
Martin Schwidefsky | c92ff1b | 2007-10-16 01:24:43 -0700 | [diff] [blame] | 370 | int map_count; /* number of VMAs */ |
Richard Kennedy | 481b4bb | 2011-03-22 16:32:50 -0700 | [diff] [blame] | 371 | |
Martin Schwidefsky | c92ff1b | 2007-10-16 01:24:43 -0700 | [diff] [blame] | 372 | spinlock_t page_table_lock; /* Protects page tables and some counters */ |
Richard Kennedy | 481b4bb | 2011-03-22 16:32:50 -0700 | [diff] [blame] | 373 | struct rw_semaphore mmap_sem; |
Martin Schwidefsky | c92ff1b | 2007-10-16 01:24:43 -0700 | [diff] [blame] | 374 | |
| 375 | struct list_head mmlist; /* List of maybe swapped mm's. These are globally strung |
| 376 | * together off init_mm.mmlist, and are protected |
| 377 | * by mmlist_lock |
| 378 | */ |
| 379 | |
Martin Schwidefsky | c92ff1b | 2007-10-16 01:24:43 -0700 | [diff] [blame] | 380 | |
| 381 | unsigned long hiwater_rss; /* High-watermark of RSS usage */ |
| 382 | unsigned long hiwater_vm; /* High-water virtual memory usage */ |
| 383 | |
Christoph Lameter | e10d59f | 2011-10-31 17:07:34 -0700 | [diff] [blame] | 384 | unsigned long total_vm; /* Total pages mapped */ |
| 385 | unsigned long locked_vm; /* Pages that have PG_mlocked set */ |
| 386 | unsigned long pinned_vm; /* Refcount permanently increased */ |
| 387 | unsigned long shared_vm; /* Shared pages (files) */ |
| 388 | unsigned long exec_vm; /* VM_EXEC & ~VM_WRITE */ |
| 389 | unsigned long stack_vm; /* VM_GROWSUP/DOWN */ |
Christoph Lameter | e10d59f | 2011-10-31 17:07:34 -0700 | [diff] [blame] | 390 | unsigned long def_flags; |
Martin Schwidefsky | c92ff1b | 2007-10-16 01:24:43 -0700 | [diff] [blame] | 391 | unsigned long start_code, end_code, start_data, end_data; |
| 392 | unsigned long start_brk, brk, start_stack; |
| 393 | unsigned long arg_start, arg_end, env_start, env_end; |
| 394 | |
| 395 | unsigned long saved_auxv[AT_VECTOR_SIZE]; /* for /proc/PID/auxv */ |
| 396 | |
KAMEZAWA Hiroyuki | d559db0 | 2010-03-05 13:41:39 -0800 | [diff] [blame] | 397 | /* |
| 398 | * Special counters, in some configurations protected by the |
| 399 | * page_table_lock, in other configurations by being atomic. |
| 400 | */ |
| 401 | struct mm_rss_stat rss_stat; |
| 402 | |
Hiroshi Shimamoto | 801460d | 2009-09-23 15:57:41 -0700 | [diff] [blame] | 403 | struct linux_binfmt *binfmt; |
| 404 | |
Linus Torvalds | 6345d24 | 2011-05-29 11:32:28 -0700 | [diff] [blame] | 405 | cpumask_var_t cpu_vm_mask_var; |
| 406 | |
Martin Schwidefsky | c92ff1b | 2007-10-16 01:24:43 -0700 | [diff] [blame] | 407 | /* Architecture-specific MM context */ |
| 408 | mm_context_t context; |
| 409 | |
Martin Schwidefsky | c92ff1b | 2007-10-16 01:24:43 -0700 | [diff] [blame] | 410 | unsigned long flags; /* Must use atomic bitops to access the bits */ |
| 411 | |
Oleg Nesterov | a94e2d4 | 2008-07-25 01:47:46 -0700 | [diff] [blame] | 412 | struct core_state *core_state; /* coredumping support */ |
Alexey Dobriyan | 858f099 | 2009-09-23 15:57:32 -0700 | [diff] [blame] | 413 | #ifdef CONFIG_AIO |
Benjamin LaHaise | db446a0 | 2013-07-30 12:54:40 -0400 | [diff] [blame] | 414 | spinlock_t ioctx_lock; |
| 415 | struct kioctx_table __rcu *ioctx_table; |
Alexey Dobriyan | 858f099 | 2009-09-23 15:57:32 -0700 | [diff] [blame] | 416 | #endif |
Oleg Nesterov | f98bafa | 2014-06-04 16:07:34 -0700 | [diff] [blame] | 417 | #ifdef CONFIG_MEMCG |
KOSAKI Motohiro | 4cd1a8f | 2008-05-12 14:02:31 -0700 | [diff] [blame] | 418 | /* |
| 419 | * "owner" points to a task that is regarded as the canonical |
| 420 | * user/owner of this mm. All of the following must be true in |
| 421 | * order for it to be changed: |
| 422 | * |
| 423 | * current == mm->owner |
| 424 | * current->mm != mm |
| 425 | * new_owner->mm == mm |
| 426 | * new_owner->alloc_lock is held |
| 427 | */ |
Arnd Bergmann | 4d2deb4 | 2010-02-24 20:01:56 +0100 | [diff] [blame] | 428 | struct task_struct __rcu *owner; |
Pavel Emelianov | 78fb746 | 2008-02-07 00:13:51 -0800 | [diff] [blame] | 429 | #endif |
Matt Helsley | 925d1c4 | 2008-04-29 01:01:36 -0700 | [diff] [blame] | 430 | |
Matt Helsley | 925d1c4 | 2008-04-29 01:01:36 -0700 | [diff] [blame] | 431 | /* store ref to file /proc/<pid>/exe symlink points to */ |
Konstantin Khlebnikov | 90f31d0 | 2015-04-16 12:47:56 -0700 | [diff] [blame] | 432 | struct file __rcu *exe_file; |
Andrea Arcangeli | cddb8a5 | 2008-07-28 15:46:29 -0700 | [diff] [blame] | 433 | #ifdef CONFIG_MMU_NOTIFIER |
| 434 | struct mmu_notifier_mm *mmu_notifier_mm; |
| 435 | #endif |
Kirill A. Shutemov | e009bb3 | 2013-11-14 14:31:07 -0800 | [diff] [blame] | 436 | #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS |
Andrea Arcangeli | e7a00c4 | 2011-01-13 15:46:45 -0800 | [diff] [blame] | 437 | pgtable_t pmd_huge_pte; /* protected by page_table_lock */ |
| 438 | #endif |
Linus Torvalds | 6345d24 | 2011-05-29 11:32:28 -0700 | [diff] [blame] | 439 | #ifdef CONFIG_CPUMASK_OFFSTACK |
| 440 | struct cpumask cpumask_allocation; |
| 441 | #endif |
Peter Zijlstra | cbee9f8 | 2012-10-25 14:16:43 +0200 | [diff] [blame] | 442 | #ifdef CONFIG_NUMA_BALANCING |
| 443 | /* |
Mel Gorman | 34f0315 | 2013-02-22 16:34:25 -0800 | [diff] [blame] | 444 | * numa_next_scan is the next time that the PTEs will be marked |
| 445 | * pte_numa. NUMA hinting faults will gather statistics and migrate |
| 446 | * pages to new nodes if necessary. |
Peter Zijlstra | cbee9f8 | 2012-10-25 14:16:43 +0200 | [diff] [blame] | 447 | */ |
| 448 | unsigned long numa_next_scan; |
| 449 | |
Peter Zijlstra | 6e5fb22 | 2012-10-25 14:16:45 +0200 | [diff] [blame] | 450 | /* Restart point for scanning and setting pte_numa */ |
| 451 | unsigned long numa_scan_offset; |
| 452 | |
Peter Zijlstra | cbee9f8 | 2012-10-25 14:16:43 +0200 | [diff] [blame] | 453 | /* numa_scan_seq prevents two threads setting pte_numa */ |
| 454 | int numa_scan_seq; |
| 455 | #endif |
Rik van Riel | 2084140 | 2013-12-18 17:08:44 -0800 | [diff] [blame] | 456 | #if defined(CONFIG_NUMA_BALANCING) || defined(CONFIG_COMPACTION) |
| 457 | /* |
| 458 | * An operation with batched TLB flushing is going on. Anything that |
| 459 | * can move process memory needs to flush the TLB when moving a |
| 460 | * PROT_NONE or PROT_NUMA mapped page. |
| 461 | */ |
| 462 | bool tlb_flush_pending; |
| 463 | #endif |
Srikar Dronamraju | d4b3b63 | 2012-03-30 23:56:31 +0530 | [diff] [blame] | 464 | struct uprobes_state uprobes_state; |
Dave Hansen | fe3d197 | 2014-11-14 07:18:29 -0800 | [diff] [blame] | 465 | #ifdef CONFIG_X86_INTEL_MPX |
| 466 | /* address of the bounds directory */ |
| 467 | void __user *bd_addr; |
| 468 | #endif |
Martin Schwidefsky | c92ff1b | 2007-10-16 01:24:43 -0700 | [diff] [blame] | 469 | }; |
| 470 | |
Linus Torvalds | 6345d24 | 2011-05-29 11:32:28 -0700 | [diff] [blame] | 471 | static inline void mm_init_cpumask(struct mm_struct *mm) |
| 472 | { |
| 473 | #ifdef CONFIG_CPUMASK_OFFSTACK |
| 474 | mm->cpu_vm_mask_var = &mm->cpumask_allocation; |
| 475 | #endif |
Vladimir Davydov | 41f727f | 2014-08-08 14:21:56 -0700 | [diff] [blame] | 476 | cpumask_clear(mm->cpu_vm_mask_var); |
Linus Torvalds | 6345d24 | 2011-05-29 11:32:28 -0700 | [diff] [blame] | 477 | } |
| 478 | |
Rusty Russell | 45e575a | 2009-03-12 14:35:44 -0600 | [diff] [blame] | 479 | /* Future-safe accessor for struct mm_struct's cpu_vm_mask. */ |
KOSAKI Motohiro | de03c72 | 2011-05-24 17:12:15 -0700 | [diff] [blame] | 480 | static inline cpumask_t *mm_cpumask(struct mm_struct *mm) |
| 481 | { |
| 482 | return mm->cpu_vm_mask_var; |
| 483 | } |
Rusty Russell | 45e575a | 2009-03-12 14:35:44 -0600 | [diff] [blame] | 484 | |
Rik van Riel | 2084140 | 2013-12-18 17:08:44 -0800 | [diff] [blame] | 485 | #if defined(CONFIG_NUMA_BALANCING) || defined(CONFIG_COMPACTION) |
| 486 | /* |
| 487 | * Memory barriers to keep this state in sync are graciously provided by |
| 488 | * the page table locks, outside of which no page table modifications happen. |
| 489 | * The barriers below prevent the compiler from re-ordering the instructions |
| 490 | * around the memory barriers that are already present in the code. |
| 491 | */ |
| 492 | static inline bool mm_tlb_flush_pending(struct mm_struct *mm) |
| 493 | { |
| 494 | barrier(); |
| 495 | return mm->tlb_flush_pending; |
| 496 | } |
| 497 | static inline void set_tlb_flush_pending(struct mm_struct *mm) |
| 498 | { |
| 499 | mm->tlb_flush_pending = true; |
Mel Gorman | af2c140 | 2013-12-18 17:08:45 -0800 | [diff] [blame] | 500 | |
| 501 | /* |
| 502 | * Guarantee that the tlb_flush_pending store does not leak into the |
| 503 | * critical section updating the page tables |
| 504 | */ |
| 505 | smp_mb__before_spinlock(); |
Rik van Riel | 2084140 | 2013-12-18 17:08:44 -0800 | [diff] [blame] | 506 | } |
| 507 | /* Clearing is done after a TLB flush, which also provides a barrier. */ |
| 508 | static inline void clear_tlb_flush_pending(struct mm_struct *mm) |
| 509 | { |
| 510 | barrier(); |
| 511 | mm->tlb_flush_pending = false; |
| 512 | } |
| 513 | #else |
| 514 | static inline bool mm_tlb_flush_pending(struct mm_struct *mm) |
| 515 | { |
| 516 | return false; |
| 517 | } |
| 518 | static inline void set_tlb_flush_pending(struct mm_struct *mm) |
| 519 | { |
| 520 | } |
| 521 | static inline void clear_tlb_flush_pending(struct mm_struct *mm) |
| 522 | { |
| 523 | } |
| 524 | #endif |
| 525 | |
Andy Lutomirski | a62c34b | 2014-05-19 15:58:33 -0700 | [diff] [blame] | 526 | struct vm_special_mapping |
| 527 | { |
| 528 | const char *name; |
| 529 | struct page **pages; |
| 530 | }; |
| 531 | |
Dave Hansen | d17d8f9 | 2014-07-31 08:40:59 -0700 | [diff] [blame] | 532 | enum tlb_flush_reason { |
| 533 | TLB_FLUSH_ON_TASK_SWITCH, |
| 534 | TLB_REMOTE_SHOOTDOWN, |
| 535 | TLB_LOCAL_SHOOTDOWN, |
| 536 | TLB_LOCAL_MM_SHOOTDOWN, |
| 537 | NR_TLB_FLUSH_REASONS, |
| 538 | }; |
| 539 | |
Tejun Heo | bd6dace | 2014-12-12 16:55:35 -0800 | [diff] [blame] | 540 | /* |
| 541 | * A swap entry has to fit into a "unsigned long", as the entry is hidden |
| 542 | * in the "index" field of the swapper address space. |
| 543 | */ |
| 544 | typedef struct { |
| 545 | unsigned long val; |
| 546 | } swp_entry_t; |
| 547 | |
Heiko Carstens | 5b99cd0 | 2006-09-27 01:50:01 -0700 | [diff] [blame] | 548 | #endif /* _LINUX_MM_TYPES_H */ |