include/linux/mm.h

   1 #ifndef _LINUX_MM_H
   2 #define _LINUX_MM_H
   3
   4 #include <linux/sched.h>
   5 #include <linux/errno.h>
   6
   7 #ifdef __KERNEL__
   8
   9 #include <linux/string.h>
  10
  11 extern unsigned long max_mapnr;
  12 extern unsigned long num_physpages;
  13 extern void * high_memory;
  14 extern int page_cluster;
  15
  16 #include <asm/page.h>
  17 #include <asm/atomic.h>
  18
  19 /*
  20  * Linux kernel virtual memory manager primitives.
  21  * The idea being to have a "virtual" mm in the same way
  22  * we have a virtual fs - giving a cleaner interface to the
  23  * mm details, and allowing different kinds of memory mappings
  24  * (from shared memory to executable loading to arbitrary
  25  * mmap() functions).
  26  */
  27
  28 /*
  29  * This struct defines a memory VMM memory area. There is one of these
  30  * per VM-area/task.  A VM area is any part of the process virtual memory
  31  * space that has a special rule for the page-fault handlers (ie a shared
  32  * library, the executable area etc).
  33  */
  34 struct vm_area_struct {
  35         struct mm_struct * vm_mm;       /* VM area parameters */
  36         unsigned long vm_start;
  37         unsigned long vm_end;
  38
  39         /* linked list of VM areas per task, sorted by address */
  40         struct vm_area_struct *vm_next;
  41
  42         pgprot_t vm_page_prot;
  43         unsigned short vm_flags;
  44
  45         /* AVL tree of VM areas per task, sorted by address */
  46         short vm_avl_height;
  47         struct vm_area_struct * vm_avl_left;
  48         struct vm_area_struct * vm_avl_right;
  49
  50         /* For areas with inode, the list inode->i_mmap, for shm areas,
  51          * the list of attaches, otherwise unused.
  52          */
  53         struct vm_area_struct *vm_next_share;
  54         struct vm_area_struct **vm_pprev_share;
  55
  56         struct vm_operations_struct * vm_ops;
  57         unsigned long vm_offset;
  58         struct file * vm_file;
  59         unsigned long vm_pte;                   /* shared mem */
  60 };
  61
  62 /*
  63  * vm_flags..
  64  */
  65 #define VM_READ         0x0001  /* currently active flags */
  66 #define VM_WRITE        0x0002
  67 #define VM_EXEC         0x0004
  68 #define VM_SHARED       0x0008
  69
  70 #define VM_MAYREAD      0x0010  /* limits for mprotect() etc */
  71 #define VM_MAYWRITE     0x0020
  72 #define VM_MAYEXEC      0x0040
  73 #define VM_MAYSHARE     0x0080
  74
  75 #define VM_GROWSDOWN    0x0100  /* general info on the segment */
  76 #define VM_GROWSUP      0x0200
  77 #define VM_SHM          0x0400  /* shared memory area, don't swap out */
  78 #define VM_DENYWRITE    0x0800  /* ETXTBSY on write attempts.. */
  79
  80 #define VM_EXECUTABLE   0x1000
  81 #define VM_LOCKED       0x2000
  82 #define VM_IO           0x4000  /* Memory mapped I/O or similar */
  83
  84 #define VM_STACK_FLAGS  0x0177
  85
  86 /*
  87  * mapping from the currently active vm_flags protection bits (the
  88  * low four bits) to a page protection mask..
  89  */
  90 extern pgprot_t protection_map[16];
  91
  92
  93 /*
  94  * These are the virtual MM functions - opening of an area, closing and
  95  * unmapping it (needed to keep files on disk up-to-date etc), pointer
  96  * to the functions called when a no-page or a wp-page exception occurs.
  97  */
  98 struct vm_operations_struct {
  99         void (*open)(struct vm_area_struct * area);
 100         void (*close)(struct vm_area_struct * area);
 101         void (*unmap)(struct vm_area_struct *area, unsigned long, size_t);
 102         void (*protect)(struct vm_area_struct *area, unsigned long, size_t, unsigned int newprot);
 103         int (*sync)(struct vm_area_struct *area, unsigned long, size_t, unsigned int flags);
 104         void (*advise)(struct vm_area_struct *area, unsigned long, size_t, unsigned int advise);
 105         unsigned long (*nopage)(struct vm_area_struct * area, unsigned long address, int write_access);
 106         unsigned long (*wppage)(struct vm_area_struct * area, unsigned long address,
 107                 unsigned long page);
 108         int (*swapout)(struct vm_area_struct *, struct page *);
 109         pte_t (*swapin)(struct vm_area_struct *, unsigned long, unsigned long);
 110 };
 111
 112 /*
 113  * Try to keep the most commonly accessed fields in single cache lines
 114  * here (16 bytes or greater).  This ordering should be particularly
 115  * beneficial on 32-bit processors.
 116  *
 117  * The first line is data used in page cache lookup, the second line
 118  * is used for linear searches (eg. clock algorithm scans).
 119  */
 120 typedef struct page {
 121         /* these must be first (free area handling) */
 122         struct page *next;
 123         struct page *prev;
 124         struct inode *inode;
 125         unsigned long offset;
 126         struct page *next_hash;
 127         atomic_t count;
 128         unsigned long flags;    /* atomic flags, some possibly updated asynchronously */
 129         wait_queue_head_t wait;
 130         struct page **pprev_hash;
 131         struct buffer_head * buffers;
 132 } mem_map_t;
 133
 134 /* Page flag bit values */
 135 #define PG_locked                0
 136 #define PG_error                 1
 137 #define PG_referenced            2
 138 #define PG_dirty                 3
 139 #define PG_uptodate              4
 140 #define PG_free_after            5
 141 #define PG_decr_after            6
 142 #define PG_swap_unlock_after     7
 143 #define PG_DMA                   8
 144 #define PG_Slab                  9
 145 #define PG_swap_cache           10
 146 #define PG_skip                 11
 147 #define PG_reserved             31
 148
 149 /* Make it prettier to test the above... */
 150 #define PageLocked(page)        (test_bit(PG_locked, &(page)->flags))
 151 #define PageError(page)         (test_bit(PG_error, &(page)->flags))
 152 #define PageReferenced(page)    (test_bit(PG_referenced, &(page)->flags))
 153 #define PageDirty(page)         (test_bit(PG_dirty, &(page)->flags))
 154 #define PageUptodate(page)      (test_bit(PG_uptodate, &(page)->flags))
 155 #define PageFreeAfter(page)     (test_bit(PG_free_after, &(page)->flags))
 156 #define PageDecrAfter(page)     (test_bit(PG_decr_after, &(page)->flags))
 157 #define PageSwapUnlockAfter(page) (test_bit(PG_swap_unlock_after, &(page)->flags))
 158 #define PageDMA(page)           (test_bit(PG_DMA, &(page)->flags))
 159 #define PageSlab(page)          (test_bit(PG_Slab, &(page)->flags))
 160 #define PageSwapCache(page)     (test_bit(PG_swap_cache, &(page)->flags))
 161 #define PageReserved(page)      (test_bit(PG_reserved, &(page)->flags))
 162
 163 #define PageSetSlab(page)       (set_bit(PG_Slab, &(page)->flags))
 164 #define PageSetSwapCache(page)  (set_bit(PG_swap_cache, &(page)->flags))
 165
 166 #define PageTestandSetDirty(page)       \
 167                         (test_and_set_bit(PG_dirty, &(page)->flags))
 168 #define PageTestandSetSwapCache(page)   \
 169                         (test_and_set_bit(PG_swap_cache, &(page)->flags))
 170
 171 #define PageClearSlab(page)     (clear_bit(PG_Slab, &(page)->flags))
 172 #define PageClearSwapCache(page)(clear_bit(PG_swap_cache, &(page)->flags))
 173
 174 #define PageTestandClearDirty(page) \
 175                         (test_and_clear_bit(PG_dirty, &(page)->flags))
 176 #define PageTestandClearSwapCache(page) \
 177                         (test_and_clear_bit(PG_swap_cache, &(page)->flags))
 178
 179 /*
 180  * Various page->flags bits:
 181  *
 182  * PG_reserved is set for a page which must never be accessed (which
 183  * may not even be present).
 184  *
 185  * PG_DMA is set for those pages which lie in the range of
 186  * physical addresses capable of carrying DMA transfers.
 187  *
 188  * Multiple processes may "see" the same page. E.g. for untouched
 189  * mappings of /dev/null, all processes see the same page full of
 190  * zeroes, and text pages of executables and shared libraries have
 191  * only one copy in memory, at most, normally.
 192  *
 193  * For the non-reserved pages, page->count denotes a reference count.
 194  *   page->count == 0 means the page is free.
 195  *   page->count == 1 means the page is used for exactly one purpose
 196  *   (e.g. a private data page of one process).
 197  *
 198  * A page may be used for kmalloc() or anyone else who does a
 199  * get_free_page(). In this case the page->count is at least 1, and
 200  * all other fields are unused but should be 0 or NULL. The
 201  * management of this page is the responsibility of the one who uses
 202  * it.
 203  *
 204  * The other pages (we may call them "process pages") are completely
 205  * managed by the Linux memory manager: I/O, buffers, swapping etc.
 206  * The following discussion applies only to them.
 207  *
 208  * A page may belong to an inode's memory mapping. In this case,
 209  * page->inode is the pointer to the inode, and page->offset is the
 210  * file offset of the page (not necessarily a multiple of PAGE_SIZE).
 211  *
 212  * A page may have buffers allocated to it. In this case,
 213  * page->buffers is a circular list of these buffer heads. Else,
 214  * page->buffers == NULL.
 215  *
 216  * For pages belonging to inodes, the page->count is the number of
 217  * attaches, plus 1 if buffers are allocated to the page.
 218  *
 219  * All pages belonging to an inode make up a doubly linked list
 220  * inode->i_pages, using the fields page->next and page->prev. (These
 221  * fields are also used for freelist management when page->count==0.)
 222  * There is also a hash table mapping (inode,offset) to the page
 223  * in memory if present. The lists for this hash table use the fields
 224  * page->next_hash and page->pprev_hash.
 225  *
 226  * All process pages can do I/O:
 227  * - inode pages may need to be read from disk,
 228  * - inode pages which have been modified and are MAP_SHARED may need
 229  *   to be written to disk,
 230  * - private pages which have been modified may need to be swapped out
 231  *   to swap space and (later) to be read back into memory.
 232  * During disk I/O, PG_locked is used. This bit is set before I/O
 233  * and reset when I/O completes. page->wait is a wait queue of all
 234  * tasks waiting for the I/O on this page to complete.
 235  * PG_uptodate tells whether the page's contents is valid.
 236  * When a read completes, the page becomes uptodate, unless a disk I/O
 237  * error happened.
 238  * When a write completes, and PG_free_after is set, the page is
 239  * freed without any further delay.
 240  *
 241  * For choosing which pages to swap out, inode pages carry a
 242  * PG_referenced bit, which is set any time the system accesses
 243  * that page through the (inode,offset) hash table.
 244  *
 245  * PG_skip is used on sparc/sparc64 architectures to "skip" certain
 246  * parts of the address space.
 247  *
 248  * PG_error is set to indicate that an I/O error occurred on this page.
 249  */
 250
 251 extern mem_map_t * mem_map;
 252
 253 /*
 254  * This is timing-critical - most of the time in getting a new page
 255  * goes to clearing the page. If you want a page without the clearing
 256  * overhead, just use __get_free_page() directly..
 257  */
 258 #define __get_free_page(gfp_mask) __get_free_pages((gfp_mask),0)
 259 #define __get_dma_pages(gfp_mask, order) __get_free_pages((gfp_mask) | GFP_DMA,(order))
 260 extern unsigned long FASTCALL(__get_free_pages(int gfp_mask, unsigned long gfp_order));
 261
 262 extern inline unsigned long get_free_page(int gfp_mask)
 263 {
 264         unsigned long page;
 265
 266         page = __get_free_page(gfp_mask);
 267         if (page)
 268                 clear_page(page);
 269         return page;
 270 }
 271
 272 extern int low_on_memory;
 273
 274 /* memory.c & swap.c*/
 275
 276 #define free_page(addr) free_pages((addr),0)
 277 extern void FASTCALL(free_pages(unsigned long addr, unsigned long order));
 278 extern void FASTCALL(__free_page(struct page *));
 279
 280 extern void show_free_areas(void);
 281 extern unsigned long put_dirty_page(struct task_struct * tsk,unsigned long page,
 282         unsigned long address);
 283
 284 extern void free_page_tables(struct mm_struct * mm);
 285 extern void clear_page_tables(struct mm_struct *, unsigned long, int);
 286 extern int new_page_tables(struct task_struct * tsk);
 287
 288 extern void zap_page_range(struct mm_struct *mm, unsigned long address, unsigned long size);
 289 extern int copy_page_range(struct mm_struct *dst, struct mm_struct *src, struct vm_area_struct *vma);
 290 extern int remap_page_range(unsigned long from, unsigned long to, unsigned long size, pgprot_t prot);
 291 extern int zeromap_page_range(unsigned long from, unsigned long size, pgprot_t prot);
 292
 293 extern void vmtruncate(struct inode * inode, unsigned long offset);
 294 extern int handle_mm_fault(struct task_struct *tsk,struct vm_area_struct *vma, unsigned long address, int write_access);
 295 extern void make_pages_present(unsigned long addr, unsigned long end);
 296
 297 extern int pgt_cache_water[2];
 298 extern int check_pgt_cache(void);
 299
 300 extern unsigned long paging_init(unsigned long start_mem, unsigned long end_mem);
 301 extern void mem_init(unsigned long start_mem, unsigned long end_mem);
 302 extern void show_mem(void);
 303 extern void oom(struct task_struct * tsk);
 304 extern void si_meminfo(struct sysinfo * val);
 305
 306 /* mmap.c */
 307 extern void vma_init(void);
 308 extern void merge_segments(struct mm_struct *, unsigned long, unsigned long);
 309 extern void insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
 310 extern void build_mmap_avl(struct mm_struct *);
 311 extern void exit_mmap(struct mm_struct *);
 312 extern unsigned long get_unmapped_area(unsigned long, unsigned long);
 313
 314 extern unsigned long do_mmap(struct file *, unsigned long, unsigned long,
 315         unsigned long, unsigned long, unsigned long);
 316 extern int do_munmap(unsigned long, size_t);
 317
 318 /* filemap.c */
 319 extern void remove_inode_page(struct page *);
 320 extern unsigned long page_unuse(struct page *);
 321 extern int shrink_mmap(int, int);
 322 extern void truncate_inode_pages(struct inode *, unsigned long);
 323 extern unsigned long get_cached_page(struct inode *, unsigned long, int);
 324 extern void put_cached_page(unsigned long);
 325
 326 /*
 327  * GFP bitmasks..
 328  */
 329 #define __GFP_WAIT      0x01
 330 #define __GFP_LOW       0x02
 331 #define __GFP_MED       0x04
 332 #define __GFP_HIGH      0x08
 333 #define __GFP_IO        0x10
 334 #define __GFP_SWAP      0x20
 335
 336 #define __GFP_DMA       0x80
 337
 338 #define GFP_BUFFER      (__GFP_LOW | __GFP_WAIT)
 339 #define GFP_ATOMIC      (__GFP_HIGH)
 340 #define GFP_USER        (__GFP_LOW | __GFP_WAIT | __GFP_IO)
 341 #define GFP_KERNEL      (__GFP_MED | __GFP_WAIT | __GFP_IO)
 342 #define GFP_NFS         (__GFP_HIGH | __GFP_WAIT | __GFP_IO)
 343 #define GFP_KSWAPD      (__GFP_IO | __GFP_SWAP)
 344
 345 /* Flag - indicates that the buffer will be suitable for DMA.  Ignored on some
 346    platforms, used as appropriate on others */
 347
 348 #define GFP_DMA         __GFP_DMA
 349
 350 /* vma is the first one with  address < vma->vm_end,
 351  * and even  address < vma->vm_start. Have to extend vma. */
 352 static inline int expand_stack(struct vm_area_struct * vma, unsigned long address)
 353 {
 354         unsigned long grow;
 355
 356         address &= PAGE_MASK;
 357         grow = vma->vm_start - address;
 358         if (vma->vm_end - address
 359             > (unsigned long) current->rlim[RLIMIT_STACK].rlim_cur ||
 360             (vma->vm_mm->total_vm << PAGE_SHIFT) + grow
 361             > (unsigned long) current->rlim[RLIMIT_AS].rlim_cur)
 362                 return -ENOMEM;
 363         vma->vm_start = address;
 364         vma->vm_offset -= grow;
 365         vma->vm_mm->total_vm += grow >> PAGE_SHIFT;
 366         if (vma->vm_flags & VM_LOCKED)
 367                 vma->vm_mm->locked_vm += grow >> PAGE_SHIFT;
 368         return 0;
 369 }
 370
 371 /* Look up the first VMA which satisfies  addr < vm_end,  NULL if none. */
 372 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
 373
 374 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
 375    NULL if none.  Assume start_addr < end_addr. */
 376 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
 377 {
 378         struct vm_area_struct * vma = find_vma(mm,start_addr);
 379
 380         if (vma && end_addr <= vma->vm_start)
 381                 vma = NULL;
 382         return vma;
 383 }
 384
 385 #define buffer_under_min()      ((buffermem >> PAGE_SHIFT) * 100 < \
 386                                 buffer_mem.min_percent * num_physpages)
 387 #define pgcache_under_min()     (page_cache_size * 100 < \
 388                                 page_cache.min_percent * num_physpages)
 389
 390 #endif /* __KERNEL__ */
 391
 392 #endif