man2/madvise.2

   1 .\" Copyright (C) 2001 David Gómez <davidge@jazzfree.com>
   2 .\"
   3 .\" %%%LICENSE_START(VERBATIM)
   4 .\" Permission is granted to make and distribute verbatim copies of this
   5 .\" manual provided the copyright notice and this permission notice are
   6 .\" preserved on all copies.
   7 .\"
   8 .\" Permission is granted to copy and distribute modified versions of this
   9 .\" manual under the conditions for verbatim copying, provided that the
  10 .\" entire resulting derived work is distributed under the terms of a
  11 .\" permission notice identical to this one.
  12 .\"
  13 .\" Since the Linux kernel and libraries are constantly changing, this
  14 .\" manual page may be incorrect or out-of-date.  The author(s) assume no
  15 .\" responsibility for errors or omissions, or for damages resulting from
  16 .\" the use of the information contained herein.  The author(s) may not
  17 .\" have taken the same level of care in the production of this manual,
  18 .\" which is licensed free of charge, as they might when working
  19 .\" professionally.
  20 .\"
  21 .\" Formatted or processed versions of this manual, if unaccompanied by
  22 .\" the source, must acknowledge the copyright and authors of this work.
  23 .\" %%%LICENSE_END
  24 .\"
  25 .\" Based on comments from mm/filemap.c. Last modified on 10-06-2001
  26 .\" Modified, 25 Feb 2002, Michael Kerrisk, <mtk.manpages@gmail.com>
  27 .\"     Added notes on MADV_DONTNEED
  28 .\" 2010-06-19, mtk, Added documentation of MADV_MERGEABLE and
  29 .\"     MADV_UNMERGEABLE
  30 .\" 2010-06-15, Andi Kleen, Add documentation of MADV_HWPOISON.
  31 .\" 2010-06-19, Andi Kleen, Add documentation of MADV_SOFT_OFFLINE.
  32 .\" 2011-09-18, Doug Goldstein <cardoe@cardoe.com>
  33 .\"     Document MADV_HUGEPAGE and MADV_NOHUGEPAGE
  34 .\"
  35 .TH MADVISE 2 2014-12-31 "Linux" "Linux Programmer's Manual"
  36 .SH NAME
  37 madvise \- give advice about use of memory
  38 .SH SYNOPSIS
  39 .B #include <sys/mman.h>
  40 .sp
  41 .BI "int madvise(void *" addr ", size_t " length ", int " advice );
  42 .sp
  43 .in -4n
  44 Feature Test Macro Requirements for glibc (see
  45 .BR feature_test_macros (7)):
  46 .in
  47 .sp
  48 .BR madvise ():
  49 _BSD_SOURCE
  50 .SH DESCRIPTION
  51 The
  52 .BR madvise ()
  53 system call is used to give advice or directions to the kernel
  54 about the address range beginning at address
  55 .I addr
  56 and with size
  57 .I length
  58 bytes.
  59 Initially, the system call supported a set of "conventional"
  60 .I advice
  61 values, which are also available on several other implementations.
  62 (Note, though, that
  63 .BR madvise ()
  64 is not specified in POSIX.)
  65 Subsequently, a number of Linux-specific
  66 .IR advice
  67 values have been added.
  68 .\"
  69 .\" ======================================================================
  70 .\"
  71 .SS Conventional advice values
  72 The
  73 .I advice
  74 values listed below
  75 allow an application to tell the kernel how it expects to use
  76 some mapped or shared memory areas, so that the kernel can choose
  77 appropriate read-ahead and caching techniques.
  78 These
  79 .I advice
  80 values do not influence the semantics of the application
  81 (except in the case of
  82 .BR MADV_DONTNEED ),
  83 but may influence its performance.
  84 All of the
  85 .I advice
  86 values listed here have analogs in the POSIX-specified
  87 .BR posix_madvise (3)
  88 function, and the values have the same meanings, with the exception of
  89 .BR MADV_DONTNEED .
  90 .LP
  91 The advice is indicated in the
  92 .I advice
  93 argument, which is one of the following:
  94 .TP
  95 .B MADV_NORMAL
  96 No special treatment.
  97 This is the default.
  98 .TP
  99 .B MADV_RANDOM
 100 Expect page references in random order.
 101 (Hence, read ahead may be less useful than normally.)
 102 .TP
 103 .B MADV_SEQUENTIAL
 104 Expect page references in sequential order.
 105 (Hence, pages in the given range can be aggressively read ahead,
 106 and may be freed soon after they are accessed.)
 107 .TP
 108 .B MADV_WILLNEED
 109 Expect access in the near future.
 110 (Hence, it might be a good idea to read some pages ahead.)
 111 .TP
 112 .B MADV_DONTNEED
 113 Do not expect access in the near future.
 114 (For the time being, the application is finished with the given range,
 115 so the kernel can free resources associated with it.)
 116
 117 After a successful
 118 .B MADV_DONTNEED
 119 operation,
 120 the semantics of memory access in the specified region are changed:
 121 subsequent accesses of pages in the range will succeed, but will result
 122 in either reloading of the memory contents from the underlying mapped file
 123 (for shared file mappings, shared anonymous mappings,
 124 and shmem-based techniques such as System V shared memory segments)
 125 or zero-fill-on-demand pages for anonymous private mappings.
 126
 127 .B MADV_DONTNEED
 128 cannot be applied to locked pages or Huge TLB pages.
 129 .\"
 130 .\" ======================================================================
 131 .\"
 132 .SS Linux-specific advice values
 133 The following Linux-specific
 134 .I advice
 135 values have no counterparts in the POSIX-specified
 136 .BR posix_madvise (3),
 137 and may or may not have counterparts in the
 138 .BR madvise ()
 139 interface available on other implementations.
 140 Note that some of these operations change the semantics of memory accesses.
 141 .TP
 142 .BR MADV_REMOVE " (since Linux 2.6.16)"
 143 .\" commit f6b3ec238d12c8cc6cc71490c6e3127988460349
 144 Free up a given range of pages
 145 and its associated backing store.
 146 This is equivalent to punching a hole in the corresponding byte
 147 range of the backing store (see
 148 .BR fallocate (2)).
 149 Subsequent accesses in the specified address range will see
 150 bytes containing zero.
 151 .\" Databases want to use this feature to drop a section of their
 152 .\" bufferpool (shared memory segments) - without writing back to
 153 .\" disk/swap space.  This feature is also useful for supporting
 154 .\" hot-plug memory on UML.
 155
 156 The specified address range must be mapped shared and writable.
 157 This flag cannot be applied to locked pages or Huge TLB pages.
 158
 159 In the initial implementation, only shmfs/tmpfs supported
 160 .BR MADV_REMOVE ;
 161 but since Linux 3.5,
 162 .\" commit 3f31d07571eeea18a7d34db9af21d2285b807a17
 163 any filesystem which supports the
 164 .BR fallocate (2)
 165 .BR FALLOC_FL_PUNCH_HOLE
 166 mode also supports
 167 .BR MADV_REMOVE .
 168 Other filesystems fail with the error
 169 .BR EOPNOTSUPP .
 170 .TP
 171 .BR MADV_DONTFORK " (since Linux 2.6.16)"
 172 .\" commit f822566165dd46ff5de9bf895cfa6c51f53bb0c4
 173 .\" See http://lwn.net/Articles/171941/
 174 Do not make the pages in this range available to the child after a
 175 .BR fork (2).
 176 This is useful to prevent copy-on-write semantics from changing
 177 the physical location of a page if the parent writes to it after a
 178 .BR fork (2).
 179 (Such page relocations cause problems for hardware that
 180 DMAs into the page.)
 181 .\" [PATCH] madvise MADV_DONTFORK/MADV_DOFORK
 182 .\" Currently, copy-on-write may change the physical address of
 183 .\" a page even if the user requested that the page is pinned in
 184 .\" memory (either by mlock or by get_user_pages).  This happens
 185 .\" if the process forks meanwhile, and the parent writes to that
 186 .\" page.  As a result, the page is orphaned: in case of
 187 .\" get_user_pages, the application will never see any data hardware
 188 .\" DMA's into this page after the COW.  In case of mlock'd memory,
 189 .\" the parent is not getting the realtime/security benefits of mlock.
 190 .\"
 191 .\" In particular, this affects the Infiniband modules which do DMA from
 192 .\" and into user pages all the time.
 193 .\"
 194 .\" This patch adds madvise options to control whether memory range is
 195 .\" inherited across fork. Useful e.g. for when hardware is doing DMA
 196 .\" from/into these pages.  Could also be useful to an application
 197 .\" wanting to speed up its forks by cutting large areas out of
 198 .\" consideration.
 199 .\"
 200 .\" SEE ALSO: http://lwn.net/Articles/171941/
 201 .\" "Tweaks to madvise() and posix_fadvise()", 14 Feb 2006
 202 .TP
 203 .BR MADV_DOFORK " (since Linux 2.6.16)"
 204 Undo the effect of
 205 .BR MADV_DONTFORK ,
 206 restoring the default behavior, whereby a mapping is inherited across
 207 .BR fork (2).
 208 .TP
 209 .BR MADV_HWPOISON " (since Linux 2.6.32)
 210 .\" commit 9893e49d64a4874ea67849ee2cfbf3f3d6817573
 211 Poison a page and handle it like a hardware memory corruption.
 212 This operation is available only for privileged
 213 .RB ( CAP_SYS_ADMIN )
 214 processes.
 215 This operation may result in the calling process receiving a
 216 .B SIGBUS
 217 and the page being unmapped.
 218
 219 This feature is intended for testing of memory error-handling code;
 220 it is available only if the kernel was configured with
 221 .BR CONFIG_MEMORY_FAILURE .
 222 .TP
 223 .BR MADV_SOFT_OFFLINE " (since Linux 2.6.33)
 224 .\" commit afcf938ee0aac4ef95b1a23bac704c6fbeb26de6
 225 Soft offline the pages in the range specified by
 226 .I addr
 227 and
 228 .IR length .
 229 The memory of each page in the specified range is preserved
 230 (i.e., when next accessed, the same content will be visible,
 231 but in a new physical page frame),
 232 and the original page is offlined
 233 (i.e., no longer used, and taken out of normal memory management).
 234 The effect of the
 235 .B MADV_SOFT_OFFLINE
 236 operation is invisible to (i.e., does not change the semantics of)
 237 the calling process.
 238
 239 This feature is intended for testing of memory error-handling code;
 240 it is available only if the kernel was configured with
 241 .BR CONFIG_MEMORY_FAILURE .
 242 .TP
 243 .BR MADV_MERGEABLE " (since Linux 2.6.32)"
 244 .\" commit f8af4da3b4c14e7267c4ffb952079af3912c51c5
 245 Enable Kernel Samepage Merging (KSM) for the pages in the range specified by
 246 .I addr
 247 and
 248 .IR length .
 249 The kernel regularly scans those areas of user memory that have
 250 been marked as mergeable,
 251 looking for pages with identical content.
 252 These are replaced by a single write-protected page (which is automatically
 253 copied if a process later wants to update the content of the page).
 254 KSM merges only private anonymous pages (see
 255 .BR mmap (2)).
 256
 257 The KSM feature is intended for applications that generate many
 258 instances of the same data (e.g., virtualization systems such as KVM).
 259 It can consume a lot of processing power; use with care.
 260 See the Linux kernel source file
 261 .I Documentation/vm/ksm.txt
 262 for more details.
 263
 264 The
 265 .BR MADV_MERGEABLE
 266 and
 267 .BR MADV_UNMERGEABLE
 268 operations are available only if the kernel was configured with
 269 .BR CONFIG_KSM .
 270 .TP
 271 .BR MADV_UNMERGEABLE " (since Linux 2.6.32)"
 272 Undo the effect of an earlier
 273 .BR MADV_MERGEABLE
 274 operation on the specified address range;
 275 KSM unmerges whatever pages it had merged in the address range specified by
 276 .IR addr
 277 and
 278 .IR length .
 279 .TP
 280 .BR MADV_HUGEPAGE " (since Linux 2.6.38)"
 281 .\" commit 0af4e98b6b095c74588af04872f83d333c958c32
 282 .\" http://lwn.net/Articles/358904/
 283 .\" https://lwn.net/Articles/423584/
 284 Enable Transparent Huge Pages (THP) for pages in the range specified by
 285 .I addr
 286 and
 287 .IR length .
 288 Currently, Transparent Huge Pages work only with private anonymous pages (see
 289 .BR mmap (2)).
 290 The kernel will regularly scan the areas marked as huge page candidates
 291 to replace them with huge pages.
 292 The kernel will also allocate huge pages directly when the region is
 293 naturally aligned to the huge page size (see
 294 .BR posix_memalign (2)).
 295
 296 This feature is primarily aimed at applications that use large mappings of
 297 data and access large regions of that memory at a time (e.g., virtualization
 298 systems such as QEMU).
 299 It can very easily waste memory (e.g., a 2MB mapping that only ever accesses
 300 1 byte will result in 2MB of wired memory instead of one 4KB page).
 301 See the Linux kernel source file
 302 .I Documentation/vm/transhuge.txt
 303 for more details.
 304
 305 The
 306 .BR MADV_HUGEPAGE
 307 and
 308 .BR MADV_NOHUGEPAGE
 309 operations are available only if the kernel was configured with
 310 .BR CONFIG_TRANSPARENT_HUGEPAGE .
 311 .TP
 312 .BR MADV_NOHUGEPAGE " (since Linux 2.6.38)"
 313 Ensures that memory in the address range specified by
 314 .IR addr
 315 and
 316 .IR length
 317 will not be collapsed into huge pages.
 318 .TP
 319 .BR MADV_DONTDUMP " (since Linux 3.4)"
 320 .\" commit 909af768e88867016f427264ae39d27a57b6a8ed
 321 .\" commit accb61fe7bb0f5c2a4102239e4981650f9048519
 322 Exclude from a core dump those pages in the range specified by
 323 .I addr
 324 and
 325 .IR length .
 326 This is useful in applications that have large areas of memory
 327 that are known not to be useful in a core dump.
 328 The effect of
 329 .BR MADV_DONTDUMP
 330 takes precedence over the bit mask that is set via the
 331 .I /proc/PID/coredump_filter
 332 file (see
 333 .BR core (5)).
 334 .TP
 335 .BR MADV_DODUMP " (since Linux 3.4)"
 336 Undo the effect of an earlier
 337 .BR MADV_DONTDUMP .
 338 .SH RETURN VALUE
 339 On success,
 340 .BR madvise ()
 341 returns zero.
 342 On error, it returns \-1 and
 343 .I errno
 344 is set appropriately.
 345 .SH ERRORS
 346 .TP
 347 .B EACCES
 348 .I advice
 349 is
 350 .BR MADV_REMOVE ,
 351 but the specified address range is not a shared writable mapping.
 352 .TP
 353 .B EAGAIN
 354 A kernel resource was temporarily unavailable.
 355 .TP
 356 .B EBADF
 357 The map exists, but the area maps something that isn't a file.
 358 .TP
 359 .B EINVAL
 360 .I addr
 361 is not page-aligned or
 362 .I length
 363 is negative.
 364 .\" .I length
 365 .\" is zero,
 366 .TP
 367 .B EINVAL
 368 .I advice
 369 is not a valid.
 370 .TP
 371 .B EINVAL
 372 .I advice
 373 is
 374 .B MADV_DONTNEED
 375 or
 376 .BR MADV_REMOVE
 377 and the specified address range includes locked or Huge TLB pages.
 378 .TP
 379 .B EINVAL
 380 .I advice
 381 is
 382 .BR MADV_MERGEABLE
 383 or
 384 .BR MADV_UNMERGEABLE ,
 385 but the kernel was not configured with
 386 .BR CONFIG_KSM .
 387 .TP
 388 .B EIO
 389 (for
 390 .BR MADV_WILLNEED )
 391 Paging in this area would exceed the process's
 392 maximum resident set size.
 393 .TP
 394 .B ENOMEM
 395 (for
 396 .BR MADV_WILLNEED )
 397 Not enough memory: paging in failed.
 398 .TP
 399 .B ENOMEM
 400 Addresses in the specified range are not currently
 401 mapped, or are outside the address space of the process.
 402 .TP
 403 .B EPERM
 404 .I advice
 405 is
 406 .BR MADV_HWPOISON ,
 407 but the caller does not have the
 408 .B CAP_SYS_ADMIN
 409 capability.
 410 .SH VERSIONS
 411 Since Linux 3.18,
 412 .\" commit d3ac21cacc24790eb45d735769f35753f5b56ceb
 413 support for this system call is optional,
 414 depending on the setting of the
 415 .B CONFIG_ADVISE_SYSCALLS
 416 configuration option.
 417 .SH CONFORMING TO
 418 .BR madvise ()
 419 is not specified by any standards.
 420 Versions of this system call, implementing a wide variety of
 421 .I advice
 422 values, exist on many other implementations.
 423 Other implementations typically implement at least the flags listed
 424 above under
 425 .IR "Conventional advice flags" ,
 426 albeit with some variation in semantics.
 427
 428 POSIX.1-2001 describes
 429 .BR posix_madvise (3)
 430 with constants
 431 .BR POSIX_MADV_NORMAL ,
 432 .BR POSIX_MADV_RANDOM ,
 433 .BR POSIX_MADV_SEQUENTIAL ,
 434 .BR POSIX_MADV_WILLNEED ,
 435 and
 436 .BR POSIX_MADV_DONTNEED ,
 437 and so on, with behavior close to the similarly named flags listed above.
 438 (POSIX.1-2008 adds a further flag,
 439 .BR POSIX_MADV_NOREUSE ,
 440 that has no analog in
 441 .BR madvise (2).)
 442 .SH NOTES
 443 .SS Linux notes
 444 The Linux implementation requires that the address
 445 .I addr
 446 be page-aligned, and allows
 447 .I length
 448 to be zero.
 449 If there are some parts of the specified address range
 450 that are not mapped, the Linux version of
 451 .BR madvise ()
 452 ignores them and applies the call to the rest (but returns
 453 .B ENOMEM
 454 from the system call, as it should).
 455 .\" .SH HISTORY
 456 .\" The
 457 .\" .BR madvise ()
 458 .\" function first appeared in 4.4BSD.
 459 .SH SEE ALSO
 460 .BR getrlimit (2),
 461 .BR mincore (2),
 462 .BR mmap (2),
 463 .BR mprotect (2),
 464 .BR msync (2),
 465 .BR munmap (2),
 466 .BR posix_fadvise (2),
 467 .BR prctl (2),
 468 .BR core (5)