man2/prctl.2

   1 .\" Copyright (C) 1998 Andries Brouwer (aeb@cwi.nl)
   2 .\" and Copyright (C) 2002, 2006, 2008, 2012, 2013 Michael Kerrisk <mtk.manpages@gmail.com>
   3 .\" and Copyright Guillem Jover <guillem@hadrons.org>
   4 .\" and Copyright (C) 2014 Dave Hansen / Intel
   5 .\"
   6 .\" %%%LICENSE_START(VERBATIM)
   7 .\" Permission is granted to make and distribute verbatim copies of this
   8 .\" manual provided the copyright notice and this permission notice are
   9 .\" preserved on all copies.
  10 .\"
  11 .\" Permission is granted to copy and distribute modified versions of this
  12 .\" manual under the conditions for verbatim copying, provided that the
  13 .\" entire resulting derived work is distributed under the terms of a
  14 .\" permission notice identical to this one.
  15 .\"
  16 .\" Since the Linux kernel and libraries are constantly changing, this
  17 .\" manual page may be incorrect or out-of-date.  The author(s) assume no
  18 .\" responsibility for errors or omissions, or for damages resulting from
  19 .\" the use of the information contained herein.  The author(s) may not
  20 .\" have taken the same level of care in the production of this manual,
  21 .\" which is licensed free of charge, as they might when working
  22 .\" professionally.
  23 .\"
  24 .\" Formatted or processed versions of this manual, if unaccompanied by
  25 .\" the source, must acknowledge the copyright and authors of this work.
  26 .\" %%%LICENSE_END
  27 .\"
  28 .\" Modified Thu Nov 11 04:19:42 MET 1999, aeb: added PR_GET_PDEATHSIG
  29 .\" Modified 27 Jun 02, Michael Kerrisk
  30 .\"     Added PR_SET_DUMPABLE, PR_GET_DUMPABLE,
  31 .\"     PR_SET_KEEPCAPS, PR_GET_KEEPCAPS
  32 .\" Modified 2006-08-30 Guillem Jover <guillem@hadrons.org>
  33 .\"     Updated Linux versions where the options where introduced.
  34 .\"     Added PR_SET_TIMING, PR_GET_TIMING, PR_SET_NAME, PR_GET_NAME,
  35 .\"     PR_SET_UNALIGN, PR_GET_UNALIGN, PR_SET_FPEMU, PR_GET_FPEMU,
  36 .\"     PR_SET_FPEXC, PR_GET_FPEXC
  37 .\" 2008-04-29 Serge Hallyn, Document PR_CAPBSET_READ and PR_CAPBSET_DROP
  38 .\" 2008-06-13 Erik Bosman, <ejbosman@cs.vu.nl>
  39 .\"     Document PR_GET_TSC and PR_SET_TSC.
  40 .\" 2008-06-15 mtk, Document PR_SET_SECCOMP, PR_GET_SECCOMP
  41 .\" 2009-10-03 Andi Kleen, document PR_MCE_KILL
  42 .\" 2012-04 Cyrill Gorcunov, Document PR_SET_MM
  43 .\" 2012-04-25 Michael Kerrisk, Document PR_TASK_PERF_EVENTS_DISABLE and
  44 .\"                             PR_TASK_PERF_EVENTS_ENABLE
  45 .\" 2012-09-20 Kees Cook, update PR_SET_SECCOMP for mode 2
  46 .\" 2012-09-20 Kees Cook, document PR_SET_NO_NEW_PRIVS, PR_GET_NO_NEW_PRIVS
  47 .\" 2012-10-25 Michael Kerrisk, Document PR_SET_TIMERSLACK and
  48 .\"                             PR_GET_TIMERSLACK
  49 .\" 2013-01-10 Kees Cook, document PR_SET_PTRACER
  50 .\" 2012-02-04 Michael Kerrisk, document PR_{SET,GET}_CHILD_SUBREAPER
  51 .\" 2014-11-10 Dave Hansen, document PR_MPX_{EN,DIS}ABLE_MANAGEMENT
  52 .\"
  53 .\"
  54 .TH PRCTL 2 2020-04-11 "Linux" "Linux Programmer's Manual"
  55 .SH NAME
  56 prctl \- operations on a process
  57 .SH SYNOPSIS
  58 .nf
  59 .B #include <sys/prctl.h>
  60 .PP
  61 .BI "int prctl(int " option ", unsigned long " arg2 ", unsigned long " arg3 ,
  62 .BI "          unsigned long " arg4 ", unsigned long " arg5 );
  63 .fi
  64 .SH DESCRIPTION
  65 .BR prctl ()
  66 is called with a first argument describing what to do
  67 (with values defined in \fI<linux/prctl.h>\fP), and further
  68 arguments with a significance depending on the first one.
  69 The first argument can be:
  70 .\"
  71 .TP
  72 .BR PR_CAP_AMBIENT " (since Linux 4.3)"
  73 .\" commit 58319057b7847667f0c9585b9de0e8932b0fdb08
  74 Reads or changes the ambient capability set of the calling thread,
  75 according to the value of
  76 .IR arg2 ,
  77 which must be one of the following:
  78 .RS
  79 .\"
  80 .TP
  81 .B PR_CAP_AMBIENT_RAISE
  82 The capability specified in
  83 .I arg3
  84 is added to the ambient set.
  85 The specified capability must already be present in
  86 both the permitted and the inheritable sets of the process.
  87 This operation is not permitted if the
  88 .B SECBIT_NO_CAP_AMBIENT_RAISE
  89 securebit is set.
  90 .TP
  91 .B PR_CAP_AMBIENT_LOWER
  92 The capability specified in
  93 .I arg3
  94 is removed from the ambient set.
  95 .TP
  96 .B PR_CAP_AMBIENT_IS_SET
  97 The
  98 .BR prctl ()
  99 call returns 1 if the capability in
 100 .I arg3
 101 is in the ambient set and 0 if it is not.
 102 .TP
 103 .BR PR_CAP_AMBIENT_CLEAR_ALL
 104 All capabilities will be removed from the ambient set.
 105 This operation requires setting
 106 .I arg3
 107 to zero.
 108 .RE
 109 .IP
 110 In all of the above operations,
 111 .I arg4
 112 and
 113 .I arg5
 114 must be specified as 0.
 115 .IP
 116 Higher-level interfaces layered on top of the above operations are
 117 provided in the
 118 .BR libcap (3)
 119 library in the form of
 120 .BR cap_get_ambient (3),
 121 .BR cap_set_ambient (3),
 122 and
 123 .BR cap_reset_ambient (3).
 124 .TP
 125 .BR PR_CAPBSET_READ " (since Linux 2.6.25)"
 126 Return (as the function result) 1 if the capability specified in
 127 .I arg2
 128 is in the calling thread's capability bounding set,
 129 or 0 if it is not.
 130 (The capability constants are defined in
 131 .IR <linux/capability.h> .)
 132 The capability bounding set dictates
 133 whether the process can receive the capability through a
 134 file's permitted capability set on a subsequent call to
 135 .BR execve (2).
 136 .IP
 137 If the capability specified in
 138 .I arg2
 139 is not valid, then the call fails with the error
 140 .BR EINVAL .
 141 .IP
 142 A higher-level interface layered on top of this operation is provided in the
 143 .BR libcap (3)
 144 library in the form of
 145 .BR cap_get_bound (3).
 146 .TP
 147 .BR PR_CAPBSET_DROP " (since Linux 2.6.25)"
 148 If the calling thread has the
 149 .B CAP_SETPCAP
 150 capability within its user namespace, then drop the capability specified by
 151 .I arg2
 152 from the calling thread's capability bounding set.
 153 Any children of the calling thread will inherit the newly
 154 reduced bounding set.
 155 .IP
 156 The call fails with the error:
 157 .B EPERM
 158 if the calling thread does not have the
 159 .BR CAP_SETPCAP ;
 160 .BR EINVAL
 161 if
 162 .I arg2
 163 does not represent a valid capability; or
 164 .BR EINVAL
 165 if file capabilities are not enabled in the kernel,
 166 in which case bounding sets are not supported.
 167 .IP
 168 A higher-level interface layered on top of this operation is provided in the
 169 .BR libcap (3)
 170 library in the form of
 171 .BR cap_drop_bound (3).
 172 .TP
 173 .BR PR_SET_CHILD_SUBREAPER " (since Linux 3.4)"
 174 .\" commit ebec18a6d3aa1e7d84aab16225e87fd25170ec2b
 175 If
 176 .I arg2
 177 is nonzero,
 178 set the "child subreaper" attribute of the calling process;
 179 if
 180 .I arg2
 181 is zero, unset the attribute.
 182 .IP
 183 A subreaper fulfills the role of
 184 .BR init (1)
 185 for its descendant processes.
 186 When a process becomes orphaned
 187 (i.e., its immediate parent terminates),
 188 then that process will be reparented to
 189 the nearest still living ancestor subreaper.
 190 Subsequently, calls to
 191 .BR getppid ()
 192 in the orphaned process will now return the PID of the subreaper process,
 193 and when the orphan terminates, it is the subreaper process that
 194 will receive a
 195 .BR SIGCHLD
 196 signal and will be able to
 197 .BR wait (2)
 198 on the process to discover its termination status.
 199 .IP
 200 The setting of the "child subreaper" attribute
 201 is not inherited by children created by
 202 .BR fork (2)
 203 and
 204 .BR clone (2).
 205 The setting is preserved across
 206 .BR execve (2).
 207 .IP
 208 Establishing a subreaper process is useful in session management frameworks
 209 where a hierarchical group of processes is managed by a subreaper process
 210 that needs to be informed when one of the processes\(emfor example,
 211 a double-forked daemon\(emterminates
 212 (perhaps so that it can restart that process).
 213 Some
 214 .BR init (1)
 215 frameworks (e.g.,
 216 .BR systemd (1))
 217 employ a subreaper process for similar reasons.
 218 .TP
 219 .BR PR_GET_CHILD_SUBREAPER " (since Linux 3.4)"
 220 Return the "child subreaper" setting of the caller,
 221 in the location pointed to by
 222 .IR "(int\ *) arg2" .
 223 .TP
 224 .BR PR_SET_DUMPABLE " (since Linux 2.3.20)"
 225 Set the state of the "dumpable" attribute,
 226 which determines whether core dumps are produced for the calling process
 227 upon delivery of a signal whose default behavior is to produce a core dump.
 228 .IP
 229 In kernels up to and including 2.6.12,
 230 .I arg2
 231 must be either 0
 232 .RB ( SUID_DUMP_DISABLE ,
 233 process is not dumpable) or 1
 234 .RB ( SUID_DUMP_USER ,
 235 process is dumpable).
 236 Between kernels 2.6.13 and 2.6.17,
 237 .\" commit abf75a5033d4da7b8a7e92321d74021d1fcfb502
 238 the value 2 was also permitted,
 239 which caused any binary which normally would not be dumped
 240 to be dumped readable by root only;
 241 for security reasons, this feature has been removed.
 242 .\" See http://marc.theaimsgroup.com/?l=linux-kernel&m=115270289030630&w=2
 243 .\" Subject:    Fix prctl privilege escalation (CVE-2006-2451)
 244 .\" From:       Marcel Holtmann <marcel () holtmann ! org>
 245 .\" Date:       2006-07-12 11:12:00
 246 (See also the description of
 247 .I /proc/sys/fs/\:suid_dumpable
 248 in
 249 .BR proc (5).)
 250 .IP
 251 Normally, the "dumpable" attribute is set to 1.
 252 However, it is reset to the current value contained in the file
 253 .IR /proc/sys/fs/\:suid_dumpable
 254 (which by default has the value 0),
 255 in the following circumstances:
 256 .\" See kernel/cred.c::commit_creds() (Linux 3.18 sources)
 257 .RS
 258 .IP * 3
 259 The process's effective user or group ID is changed.
 260 .IP *
 261 The process's filesystem user or group ID is changed (see
 262 .BR credentials (7)).
 263 .IP *
 264 The process executes
 265 .RB ( execve (2))
 266 a set-user-ID or set-group-ID program, resulting in a change
 267 of either the effective user ID or the effective group ID.
 268 .IP *
 269 The process executes
 270 .RB ( execve (2))
 271 a program that has file capabilities (see
 272 .BR capabilities (7)),
 273 .\" See kernel/cred.c::commit_creds()
 274 but only if the permitted capabilities
 275 gained exceed those already permitted for the process.
 276 .\" Also certain namespace operations;
 277 .RE
 278 .IP
 279 Processes that are not dumpable can not be attached via
 280 .BR ptrace (2)
 281 .BR PTRACE_ATTACH ;
 282 see
 283 .BR ptrace (2)
 284 for further details.
 285 .IP
 286 If a process is not dumpable,
 287 the ownership of files in the process's
 288 .IR /proc/[pid]
 289 directory is affected as described in
 290 .BR proc (5).
 291 .TP
 292 .BR PR_GET_DUMPABLE " (since Linux 2.3.20)"
 293 Return (as the function result) the current state of the calling
 294 process's dumpable attribute.
 295 .\" Since Linux 2.6.13, the dumpable flag can have the value 2,
 296 .\" but in 2.6.13 PR_GET_DUMPABLE simply returns 1 if the dumpable
 297 .\" flags has a nonzero value.  This was fixed in 2.6.14.
 298 .TP
 299 .BR PR_SET_ENDIAN " (since Linux 2.6.18, PowerPC only)"
 300 Set the endian-ness of the calling process to the value given
 301 in \fIarg2\fP, which should be one of the following:
 302 .\" Respectively 0, 1, 2
 303 .BR PR_ENDIAN_BIG ,
 304 .BR PR_ENDIAN_LITTLE ,
 305 or
 306 .B PR_ENDIAN_PPC_LITTLE
 307 (PowerPC pseudo little endian).
 308 .TP
 309 .BR PR_GET_ENDIAN " (since Linux 2.6.18, PowerPC only)"
 310 Return the endian-ness of the calling process,
 311 in the location pointed to by
 312 .IR "(int\ *) arg2" .
 313 .TP
 314 .BR PR_SET_FP_MODE " (since Linux 4.0, only on MIPS)"
 315 .\" commit 9791554b45a2acc28247f66a5fd5bbc212a6b8c8
 316 On the MIPS architecture,
 317 user-space code can be built using an ABI which permits linking
 318 with code that has more restrictive floating-point (FP) requirements.
 319 For example, user-space code may be built to target the O32 FPXX ABI
 320 and linked with code built for either one of the more restrictive
 321 FP32 or FP64 ABIs.
 322 When more restrictive code is linked in,
 323 the overall requirement for the process is to use the more
 324 restrictive floating-point mode.
 325 .IP
 326 Because the kernel has no means of knowing in advance
 327 which mode the process should be executed in,
 328 and because these restrictions can
 329 change over the lifetime of the process, the
 330 .B PR_SET_FP_MODE
 331 operation is provided to allow control of the floating-point mode
 332 from user space.
 333 .IP
 334 .\" https://dmz-portal.mips.com/wiki/MIPS_O32_ABI_-_FR0_and_FR1_Interlinking
 335 The
 336 .I (unsigned int) arg2
 337 argument is a bit mask describing the floating-point mode used:
 338 .RS
 339 .TP
 340 .BR PR_FP_MODE_FR
 341 When this bit is
 342 .I unset
 343 (so called
 344 .BR FR=0 " or " FR0
 345 mode), the 32 floating-point registers are 32 bits wide,
 346 and 64-bit registers are represented as a pair of registers
 347 (even- and odd- numbered,
 348 with the even-numbered register containing the lower 32 bits,
 349 and the odd-numbered register containing the higher 32 bits).
 350 .IP
 351 When this bit is
 352 .I set
 353 (on supported hardware),
 354 the 32 floating-point registers are 64 bits wide (so called
 355 .BR FR=1 " or " FR1
 356 mode).
 357 Note that modern MIPS implementations (MIPS R6 and newer) support
 358 .B FR=1
 359 mode only.
 360 .IP
 361 .IP
 362 Applications that use the O32 FP32 ABI can operate only when this bit is
 363 .I unset
 364 .RB ( FR=0 ;
 365 or they can be used with FRE enabled, see below).
 366 Applications that use the O32 FP64 ABI
 367 (and the O32 FP64A ABI, which exists to
 368 provide the ability to operate with existing FP32 code; see below)
 369 can operate only when this bit is
 370 .I set
 371 .RB ( FR=1 ).
 372 Applications that use the O32 FPXX ABI can operate with either
 373 .BR FR=0
 374 or
 375 .BR FR=1 .
 376 .TP
 377 .BR PR_FP_MODE_FRE
 378 Enable emulation of 32-bit floating-point mode.
 379 When this mode is enabled,
 380 it emulates 32-bit floating-point operations
 381 by raising a reserved-instruction exception
 382 on every instruction that uses 32-bit formats and
 383 the kernel then handles the instruction in software.
 384 (The problem lies in the discrepancy of handling odd-numbered registers
 385 which are the high 32 bits of 64-bit registers with even numbers in
 386 .B FR=0
 387 mode and the lower 32-bit parts of odd-numbered 64-bit registers in
 388 .B FR=1
 389 mode.)
 390 Enabling this bit is necessary when code with the O32 FP32 ABI should operate
 391 with code with compatible the O32 FPXX or O32 FP64A ABIs (which require
 392 .B FR=1
 393 FPU mode) or when it is executed on newer hardware (MIPS R6 onwards)
 394 which lacks
 395 .B FR=0
 396 mode support when a binary with the FP32 ABI is used.
 397 .IP
 398 Note that this mode makes sense only when the FPU is in 64-bit mode
 399 .RB ( FR=1 ).
 400 .IP
 401 Note that the use of emulation inherently has a significant performance hit
 402 and should be avoided if possible.
 403 .RE
 404 .IP
 405 In the N32/N64 ABI, 64-bit floating-point mode is always used,
 406 so FPU emulation is not required and the FPU always operates in
 407 .B FR=1
 408 mode.
 409 .IP
 410 This option is mainly intended for use by the dynamic linker
 411 .RB ( ld.so (8)).
 412 .IP
 413 The arguments
 414 .IR arg3 ,
 415 .IR arg4 ,
 416 and
 417 .IR arg5
 418 are ignored.
 419 .TP
 420 .BR PR_GET_FP_MODE " (since Linux 4.0, only on MIPS)"
 421 Return (as the function result)
 422 the current floating-point mode (see the description of
 423 .B PR_SET_FP_MODE
 424 for details).
 425 .IP
 426 On success,
 427 the call returns a bit mask which represents the current floating-point mode.
 428 .IP
 429 The arguments
 430 .IR arg2 ,
 431 .IR arg3 ,
 432 .IR arg4 ,
 433 and
 434 .IR arg5
 435 are ignored.
 436 .TP
 437 .BR PR_SET_FPEMU " (since Linux 2.4.18, 2.5.9, only on ia64)"
 438 Set floating-point emulation control bits to \fIarg2\fP.
 439 Pass
 440 .B PR_FPEMU_NOPRINT
 441 to silently emulate floating-point operation accesses, or
 442 .B PR_FPEMU_SIGFPE
 443 to not emulate floating-point operations and send
 444 .B SIGFPE
 445 instead.
 446 .TP
 447 .BR PR_GET_FPEMU " (since Linux 2.4.18, 2.5.9, only on ia64)"
 448 Return floating-point emulation control bits,
 449 in the location pointed to by
 450 .IR "(int\ *) arg2" .
 451 .TP
 452 .BR PR_SET_FPEXC " (since Linux 2.4.21, 2.5.32, only on PowerPC)"
 453 Set floating-point exception mode to \fIarg2\fP.
 454 Pass \fBPR_FP_EXC_SW_ENABLE\fP to use FPEXC for FP exception enables,
 455 \fBPR_FP_EXC_DIV\fP for floating-point divide by zero,
 456 \fBPR_FP_EXC_OVF\fP for floating-point overflow,
 457 \fBPR_FP_EXC_UND\fP for floating-point underflow,
 458 \fBPR_FP_EXC_RES\fP for floating-point inexact result,
 459 \fBPR_FP_EXC_INV\fP for floating-point invalid operation,
 460 \fBPR_FP_EXC_DISABLED\fP for FP exceptions disabled,
 461 \fBPR_FP_EXC_NONRECOV\fP for async nonrecoverable exception mode,
 462 \fBPR_FP_EXC_ASYNC\fP for async recoverable exception mode,
 463 \fBPR_FP_EXC_PRECISE\fP for precise exception mode.
 464 .TP
 465 .BR PR_GET_FPEXC " (since Linux 2.4.21, 2.5.32, only on PowerPC)"
 466 Return floating-point exception mode,
 467 in the location pointed to by
 468 .IR "(int\ *) arg2" .
 469 .TP
 470 .BR PR_SET_KEEPCAPS " (since Linux 2.2.18)"
 471 Set the state of the calling thread's "keep capabilities" flag.
 472 The effect of this flag is described in
 473 .BR capabilities (7).
 474 .I arg2
 475 must be either 0 (clear the flag)
 476 or 1 (set the flag).
 477 The "keep capabilities" value will be reset to 0 on subsequent calls to
 478 .BR execve (2).
 479 .TP
 480 .BR PR_GET_KEEPCAPS " (since Linux 2.2.18)"
 481 Return (as the function result) the current state of the calling thread's
 482 "keep capabilities" flag.
 483 See
 484 .BR capabilities (7)
 485 for a description of this flag.
 486 .TP
 487 .BR PR_MCE_KILL " (since Linux 2.6.32)"
 488 Set the machine check memory corruption kill policy for the calling thread.
 489 If
 490 .I arg2
 491 is
 492 .BR PR_MCE_KILL_CLEAR ,
 493 clear the thread memory corruption kill policy and use the system-wide default.
 494 (The system-wide default is defined by
 495 .IR /proc/sys/vm/memory_failure_early_kill ;
 496 see
 497 .BR proc (5).)
 498 If
 499 .I arg2
 500 is
 501 .BR PR_MCE_KILL_SET ,
 502 use a thread-specific memory corruption kill policy.
 503 In this case,
 504 .I arg3
 505 defines whether the policy is
 506 .I early kill
 507 .RB ( PR_MCE_KILL_EARLY ),
 508 .I late kill
 509 .RB ( PR_MCE_KILL_LATE ),
 510 or the system-wide default
 511 .RB ( PR_MCE_KILL_DEFAULT ).
 512 Early kill means that the thread receives a
 513 .B SIGBUS
 514 signal as soon as hardware memory corruption is detected inside
 515 its address space.
 516 In late kill mode, the process is killed only when it accesses a corrupted page.
 517 See
 518 .BR sigaction (2)
 519 for more information on the
 520 .BR SIGBUS
 521 signal.
 522 The policy is inherited by children.
 523 The remaining unused
 524 .BR prctl ()
 525 arguments must be zero for future compatibility.
 526 .TP
 527 .BR PR_MCE_KILL_GET " (since Linux 2.6.32)"
 528 Return (as the function result)
 529 the current per-process machine check kill policy.
 530 All unused
 531 .BR prctl ()
 532 arguments must be zero.
 533 .TP
 534 .BR PR_SET_MM " (since Linux 3.3)"
 535 .\" commit 028ee4be34a09a6d48bdf30ab991ae933a7bc036
 536 Modify certain kernel memory map descriptor fields
 537 of the calling process.
 538 Usually these fields are set by the kernel and dynamic loader (see
 539 .BR ld.so (8)
 540 for more information) and a regular application should not use this feature.
 541 However, there are cases, such as self-modifying programs,
 542 where a program might find it useful to change its own memory map.
 543 .IP
 544 The calling process must have the
 545 .BR CAP_SYS_RESOURCE
 546 capability.
 547 The value in
 548 .I arg2
 549 is one of the options below, while
 550 .I arg3
 551 provides a new value for the option.
 552 The
 553 .I arg4
 554 and
 555 .I arg5
 556 arguments must be zero if unused.
 557 .IP
 558 Before Linux 3.10,
 559 .\" commit 52b3694157e3aa6df871e283115652ec6f2d31e0
 560 this feature is available only if the kernel is built with the
 561 .BR CONFIG_CHECKPOINT_RESTORE
 562 option enabled.
 563 .RS
 564 .TP
 565 .BR PR_SET_MM_START_CODE
 566 Set the address above which the program text can run.
 567 The corresponding memory area must be readable and executable,
 568 but not writable or shareable (see
 569 .BR mprotect (2)
 570 and
 571 .BR mmap (2)
 572 for more information).
 573 .TP
 574 .BR PR_SET_MM_END_CODE
 575 Set the address below which the program text can run.
 576 The corresponding memory area must be readable and executable,
 577 but not writable or shareable.
 578 .TP
 579 .BR PR_SET_MM_START_DATA
 580 Set the address above which initialized and
 581 uninitialized (bss) data are placed.
 582 The corresponding memory area must be readable and writable,
 583 but not executable or shareable.
 584 .TP
 585 .B PR_SET_MM_END_DATA
 586 Set the address below which initialized and
 587 uninitialized (bss) data are placed.
 588 The corresponding memory area must be readable and writable,
 589 but not executable or shareable.
 590 .TP
 591 .BR PR_SET_MM_START_STACK
 592 Set the start address of the stack.
 593 The corresponding memory area must be readable and writable.
 594 .TP
 595 .BR PR_SET_MM_START_BRK
 596 Set the address above which the program heap can be expanded with
 597 .BR brk (2)
 598 call.
 599 The address must be greater than the ending address of
 600 the current program data segment.
 601 In addition, the combined size of the resulting heap and
 602 the size of the data segment can't exceed the
 603 .BR RLIMIT_DATA
 604 resource limit (see
 605 .BR setrlimit (2)).
 606 .TP
 607 .BR PR_SET_MM_BRK
 608 Set the current
 609 .BR brk (2)
 610 value.
 611 The requirements for the address are the same as for the
 612 .BR PR_SET_MM_START_BRK
 613 option.
 614 .PP
 615 The following options are available since Linux 3.5.
 616 .\" commit fe8c7f5cbf91124987106faa3bdf0c8b955c4cf7
 617 .TP
 618 .BR PR_SET_MM_ARG_START
 619 Set the address above which the program command line is placed.
 620 .TP
 621 .BR PR_SET_MM_ARG_END
 622 Set the address below which the program command line is placed.
 623 .TP
 624 .BR PR_SET_MM_ENV_START
 625 Set the address above which the program environment is placed.
 626 .TP
 627 .BR PR_SET_MM_ENV_END
 628 Set the address below which the program environment is placed.
 629 .IP
 630 The address passed with
 631 .BR PR_SET_MM_ARG_START ,
 632 .BR PR_SET_MM_ARG_END ,
 633 .BR PR_SET_MM_ENV_START ,
 634 and
 635 .BR PR_SET_MM_ENV_END
 636 should belong to a process stack area.
 637 Thus, the corresponding memory area must be readable, writable, and
 638 (depending on the kernel configuration) have the
 639 .BR MAP_GROWSDOWN
 640 attribute set (see
 641 .BR mmap (2)).
 642 .TP
 643 .BR PR_SET_MM_AUXV
 644 Set a new auxiliary vector.
 645 The
 646 .I arg3
 647 argument should provide the address of the vector.
 648 The
 649 .I arg4
 650 is the size of the vector.
 651 .TP
 652 .BR PR_SET_MM_EXE_FILE
 653 .\" commit b32dfe377102ce668775f8b6b1461f7ad428f8b6
 654 Supersede the
 655 .IR /proc/pid/exe
 656 symbolic link with a new one pointing to a new executable file
 657 identified by the file descriptor provided in
 658 .I arg3
 659 argument.
 660 The file descriptor should be obtained with a regular
 661 .BR open (2)
 662 call.
 663 .IP
 664 To change the symbolic link, one needs to unmap all existing
 665 executable memory areas, including those created by the kernel itself
 666 (for example the kernel usually creates at least one executable
 667 memory area for the ELF
 668 .IR \.text
 669 section).
 670 .IP
 671 In Linux 4.9 and earlier, the
 672 .\" commit 3fb4afd9a504c2386b8435028d43283216bf588e
 673 .BR PR_SET_MM_EXE_FILE
 674 operation can be performed only once in a process's lifetime;
 675 attempting to perform the operation a second time results in the error
 676 .BR EPERM .
 677 This restriction was enforced for security reasons that were subsequently
 678 deemed specious,
 679 and the restriction was removed in Linux 4.10 because some
 680 user-space applications needed to perform this operation more than once.
 681 .PP
 682 The following options are available since Linux 3.18.
 683 .\" commit f606b77f1a9e362451aca8f81d8f36a3a112139e
 684 .TP
 685 .BR PR_SET_MM_MAP
 686 Provides one-shot access to all the addresses by passing in a
 687 .I struct prctl_mm_map
 688 (as defined in \fI<linux/prctl.h>\fP).
 689 The
 690 .I arg4
 691 argument should provide the size of the struct.
 692 .IP
 693 This feature is available only if the kernel is built with the
 694 .BR CONFIG_CHECKPOINT_RESTORE
 695 option enabled.
 696 .TP
 697 .BR PR_SET_MM_MAP_SIZE
 698 Returns the size of the
 699 .I struct prctl_mm_map
 700 the kernel expects.
 701 This allows user space to find a compatible struct.
 702 The
 703 .I arg4
 704 argument should be a pointer to an unsigned int.
 705 .IP
 706 This feature is available only if the kernel is built with the
 707 .BR CONFIG_CHECKPOINT_RESTORE
 708 option enabled.
 709 .RE
 710 .TP
 711 .BR PR_MPX_ENABLE_MANAGEMENT ", " PR_MPX_DISABLE_MANAGEMENT " (since Linux 3.19) "
 712 .\" commit fe3d197f84319d3bce379a9c0dc17b1f48ad358c
 713 .\" See also http://lwn.net/Articles/582712/
 714 .\" See also https://gcc.gnu.org/wiki/Intel%20MPX%20support%20in%20the%20GCC%20compiler
 715 Enable or disable kernel management of Memory Protection eXtensions (MPX)
 716 bounds tables.
 717 The
 718 .IR arg2 ,
 719 .IR arg3 ,
 720 .IR arg4 ,
 721 and
 722 .IR arg5
 723 .\" commit e9d1b4f3c60997fe197bf0243cb4a41a44387a88
 724 arguments must be zero.
 725 .IP
 726 MPX is a hardware-assisted mechanism for performing bounds checking on
 727 pointers.
 728 It consists of a set of registers storing bounds information
 729 and a set of special instruction prefixes that tell the CPU on which
 730 instructions it should do bounds enforcement.
 731 There is a limited number of these registers and
 732 when there are more pointers than registers,
 733 their contents must be "spilled" into a set of tables.
 734 These tables are called "bounds tables" and the MPX
 735 .BR prctl ()
 736 operations control
 737 whether the kernel manages their allocation and freeing.
 738 .IP
 739 When management is enabled, the kernel will take over allocation
 740 and freeing of the bounds tables.
 741 It does this by trapping the #BR exceptions that result
 742 at first use of missing bounds tables and
 743 instead of delivering the exception to user space,
 744 it allocates the table and populates the bounds directory
 745 with the location of the new table.
 746 For freeing, the kernel checks to see if bounds tables are
 747 present for memory which is not allocated, and frees them if so.
 748 .IP
 749 Before enabling MPX management using
 750 .BR PR_MPX_ENABLE_MANAGEMENT ,
 751 the application must first have allocated a user-space buffer for
 752 the bounds directory and placed the location of that directory in the
 753 .I bndcfgu
 754 register.
 755 .IP
 756 These calls fail if the CPU or kernel does not support MPX.
 757 Kernel support for MPX is enabled via the
 758 .BR CONFIG_X86_INTEL_MPX
 759 configuration option.
 760 You can check whether the CPU supports MPX by looking for the 'mpx'
 761 CPUID bit, like with the following command:
 762 .IP
 763 .in +4n
 764 .EX
 765 cat /proc/cpuinfo | grep ' mpx '
 766 .EE
 767 .in
 768 .IP
 769 A thread may not switch in or out of long (64-bit) mode while MPX is
 770 enabled.
 771 .IP
 772 All threads in a process are affected by these calls.
 773 .IP
 774 The child of a
 775 .BR fork (2)
 776 inherits the state of MPX management.
 777 During
 778 .BR execve (2),
 779 MPX management is reset to a state as if
 780 .BR PR_MPX_DISABLE_MANAGEMENT
 781 had been called.
 782 .IP
 783 For further information on Intel MPX, see the kernel source file
 784 .IR Documentation/x86/intel_mpx.txt .
 785 .TP
 786 .BR PR_SET_NAME " (since Linux 2.6.9)"
 787 Set the name of the calling thread,
 788 using the value in the location pointed to by
 789 .IR "(char\ *) arg2" .
 790 The name can be up to 16 bytes long,
 791 .\" TASK_COMM_LEN in include/linux/sched.h
 792 including the terminating null byte.
 793 (If the length of the string, including the terminating null byte,
 794 exceeds 16 bytes, the string is silently truncated.)
 795 This is the same attribute that can be set via
 796 .BR pthread_setname_np (3)
 797 and retrieved using
 798 .BR pthread_getname_np (3).
 799 The attribute is likewise accessible via
 800 .IR /proc/self/task/[tid]/comm ,
 801 where
 802 .I tid
 803 is the name of the calling thread.
 804 .TP
 805 .BR PR_GET_NAME " (since Linux 2.6.11)"
 806 Return the name of the calling thread,
 807 in the buffer pointed to by
 808 .IR "(char\ *) arg2" .
 809 The buffer should allow space for up to 16 bytes;
 810 the returned string will be null-terminated.
 811 .TP
 812 .BR PR_SET_NO_NEW_PRIVS " (since Linux 3.5)"
 813 Set the calling thread's
 814 .I no_new_privs
 815 attribute to the value in
 816 .IR arg2 .
 817 With
 818 .I no_new_privs
 819 set to 1,
 820 .BR execve (2)
 821 promises not to grant privileges to do anything
 822 that could not have been done without the
 823 .BR execve (2)
 824 call (for example,
 825 rendering the set-user-ID and set-group-ID mode bits,
 826 and file capabilities non-functional).
 827 Once set, the
 828 .I no_new_privs
 829 attribute cannot be unset.
 830 The setting of this attribute is inherited by children created by
 831 .BR fork (2)
 832 and
 833 .BR clone (2),
 834 and preserved across
 835 .BR execve (2).
 836 .IP
 837 Since Linux 4.10,
 838 the value of a thread's
 839 .I no_new_privs
 840 attribute can be viewed via the
 841 .I NoNewPrivs
 842 field in the
 843 .IR /proc/[pid]/status
 844 file.
 845 .IP
 846 For more information, see the kernel source file
 847 .IR Documentation/userspace\-api/no_new_privs.rst
 848 .\" commit 40fde647ccb0ae8c11d256d271e24d385eed595b
 849 (or
 850 .IR Documentation/prctl/no_new_privs.txt
 851 before Linux 4.13).
 852 See also
 853 .BR seccomp (2).
 854 .TP
 855 .BR PR_GET_NO_NEW_PRIVS " (since Linux 3.5)"
 856 Return (as the function result) the value of the
 857 .I no_new_privs
 858 attribute for the calling thread.
 859 A value of 0 indicates the regular
 860 .BR execve (2)
 861 behavior.
 862 A value of 1 indicates
 863 .BR execve (2)
 864 will operate in the privilege-restricting mode described above.
 865 .TP
 866 .BR PR_SET_PDEATHSIG " (since Linux 2.1.57)"
 867 Set the parent-death signal
 868 of the calling process to \fIarg2\fP (either a signal value
 869 in the range 1..maxsig, or 0 to clear).
 870 This is the signal that the calling process will get when its
 871 parent dies.
 872 .IP
 873 .IR Warning :
 874 .\" https://bugzilla.kernel.org/show_bug.cgi?id=43300
 875 the "parent" in this case is considered to be the
 876 .I thread
 877 that created this process.
 878 In other words, the signal will be sent when that thread terminates
 879 (via, for example,
 880 .BR pthread_exit (3)),
 881 rather than after all of the threads in the parent process terminate.
 882 .IP
 883 The parent-death signal is sent upon subsequent termination of the parent
 884 thread and also upon termination of each subreaper process
 885 (see the description of
 886 .B PR_SET_CHILD_SUBREAPER
 887 above) to which the caller is subsequently reparented.
 888 If the parent thread and all ancestor subreapers have already terminated
 889 by the time of the
 890 .BR PR_SET_PDEATHSIG
 891 operation, then no parent-death signal is sent to the caller.
 892 .IP
 893 The parent-death signal is process-directed (see
 894 .BR signal (7))
 895 and, if the child installs a handler using the
 896 .BR sigaction (2)
 897 .B SA_SIGINFO
 898 flag, the
 899 .I si_pid
 900 field of the
 901 .I siginfo_t
 902 argument of the handler contains the PID of the terminating parent process.
 903 .IP
 904 The parent-death signal setting is cleared for the child of a
 905 .BR fork (2).
 906 It is also
 907 (since Linux 2.4.36 / 2.6.23)
 908 .\" commit d2d56c5f51028cb9f3d800882eb6f4cbd3f9099f
 909 cleared when executing a set-user-ID or set-group-ID binary,
 910 or a binary that has associated capabilities (see
 911 .BR capabilities (7));
 912 otherwise, this value is preserved across
 913 .BR execve (2).
 914 .TP
 915 .BR PR_GET_PDEATHSIG " (since Linux 2.3.15)"
 916 Return the current value of the parent process death signal,
 917 in the location pointed to by
 918 .IR "(int\ *) arg2" .
 919 .TP
 920 .BR PR_SET_PTRACER " (since Linux 3.4)"
 921 .\" commit 2d514487faf188938a4ee4fb3464eeecfbdcf8eb
 922 .\" commit bf06189e4d14641c0148bea16e9dd24943862215
 923 This is meaningful only when the Yama LSM is enabled and in mode 1
 924 ("restricted ptrace", visible via
 925 .IR /proc/sys/kernel/yama/ptrace_scope ).
 926 When a "ptracer process ID" is passed in \fIarg2\fP,
 927 the caller is declaring that the ptracer process can
 928 .BR ptrace (2)
 929 the calling process as if it were a direct process ancestor.
 930 Each
 931 .B PR_SET_PTRACER
 932 operation replaces the previous "ptracer process ID".
 933 Employing
 934 .B PR_SET_PTRACER
 935 with
 936 .I arg2
 937 set to 0 clears the caller's "ptracer process ID".
 938 If
 939 .I arg2
 940 is
 941 .BR PR_SET_PTRACER_ANY ,
 942 the ptrace restrictions introduced by Yama are effectively disabled for the
 943 calling process.
 944 .IP
 945 For further information, see the kernel source file
 946 .IR Documentation/admin\-guide/LSM/Yama.rst
 947 .\" commit 90bb766440f2147486a2acc3e793d7b8348b0c22
 948 (or
 949 .IR Documentation/security/Yama.txt
 950 before Linux 4.13).
 951 .TP
 952 .BR PR_SET_SECCOMP " (since Linux 2.6.23)"
 953 .\" See http://thread.gmane.org/gmane.linux.kernel/542632
 954 .\" [PATCH 0 of 2] seccomp updates
 955 .\" andrea@cpushare.com
 956 Set the secure computing (seccomp) mode for the calling thread, to limit
 957 the available system calls.
 958 The more recent
 959 .BR seccomp (2)
 960 system call provides a superset of the functionality of
 961 .BR PR_SET_SECCOMP .
 962 .IP
 963 The seccomp mode is selected via
 964 .IR arg2 .
 965 (The seccomp constants are defined in
 966 .IR <linux/seccomp.h> .)
 967 .IP
 968 With
 969 .IR arg2
 970 set to
 971 .BR SECCOMP_MODE_STRICT ,
 972 the only system calls that the thread is permitted to make are
 973 .BR read (2),
 974 .BR write (2),
 975 .BR _exit (2)
 976 (but not
 977 .BR exit_group (2)),
 978 and
 979 .BR sigreturn (2).
 980 Other system calls result in the delivery of a
 981 .BR SIGKILL
 982 signal.
 983 Strict secure computing mode is useful for number-crunching applications
 984 that may need to execute untrusted byte code,
 985 perhaps obtained by reading from a pipe or socket.
 986 This operation is available only
 987 if the kernel is configured with
 988 .B CONFIG_SECCOMP
 989 enabled.
 990 .IP
 991 With
 992 .IR arg2
 993 set to
 994 .BR SECCOMP_MODE_FILTER " (since Linux 3.5),"
 995 the system calls allowed are defined by a pointer
 996 to a Berkeley Packet Filter passed in
 997 .IR arg3 .
 998 This argument is a pointer to
 999 .IR "struct sock_fprog" ;
1000 it can be designed to filter
1001 arbitrary system calls and system call arguments.
1002 This mode is available only if the kernel is configured with
1003 .B CONFIG_SECCOMP_FILTER
1004 enabled.
1005 .IP
1006 If
1007 .BR SECCOMP_MODE_FILTER
1008 filters permit
1009 .BR fork (2),
1010 then the seccomp mode is inherited by children created by
1011 .BR fork (2);
1012 if
1013 .BR execve (2)
1014 is permitted, then the seccomp mode is preserved across
1015 .BR execve (2).
1016 If the filters permit
1017 .BR prctl ()
1018 calls, then additional filters can be added;
1019 they are run in order until the first non-allow result is seen.
1020 .IP
1021 For further information, see the kernel source file
1022 .IR Documentation/userspace\-api/seccomp_filter.rst
1023 .\" commit c061f33f35be0ccc80f4b8e0aea5dfd2ed7e01a3
1024 (or
1025 .IR Documentation/prctl/seccomp_filter.txt
1026 before Linux 4.13).
1027 .TP
1028 .BR PR_GET_SECCOMP " (since Linux 2.6.23)"
1029 Return (as the function result)
1030 the secure computing mode of the calling thread.
1031 If the caller is not in secure computing mode, this operation returns 0;
1032 if the caller is in strict secure computing mode, then the
1033 .BR prctl ()
1034 call will cause a
1035 .B SIGKILL
1036 signal to be sent to the process.
1037 If the caller is in filter mode, and this system call is allowed by the
1038 seccomp filters, it returns 2; otherwise, the process is killed with a
1039 .BR SIGKILL
1040 signal.
1041 This operation is available only
1042 if the kernel is configured with
1043 .B CONFIG_SECCOMP
1044 enabled.
1045 .IP
1046 Since Linux 3.8, the
1047 .IR Seccomp
1048 field of the
1049 .IR /proc/[pid]/status
1050 file provides a method of obtaining the same information,
1051 without the risk that the process is killed; see
1052 .BR proc (5).
1053 .TP
1054 .BR PR_SET_SECUREBITS " (since Linux 2.6.26)"
1055 Set the "securebits" flags of the calling thread to the value supplied in
1056 .IR arg2 .
1057 See
1058 .BR capabilities (7).
1059 .TP
1060 .BR PR_GET_SECUREBITS " (since Linux 2.6.26)"
1061 Return (as the function result)
1062 the "securebits" flags of the calling thread.
1063 See
1064 .BR capabilities (7).
1065 .TP
1066 .BR PR_GET_SPECULATION_CTRL " (since Linux 4.17)"
1067 Return (as the function result)
1068 the state of the speculation misfeature specified in
1069 .IR arg2 .
1070 Currently, the only permitted value for this argument is
1071 .BR PR_SPEC_STORE_BYPASS
1072 (otherwise the call fails with the error
1073 .BR ENODEV ).
1074 .IP
1075 The return value uses bits 0-3 with the following meaning:
1076 .RS
1077 .TP
1078 .BR PR_SPEC_PRCTL
1079 Mitigation can be controlled per thread by
1080 .B PR_SET_SPECULATION_CTRL
1081 .TP
1082 .BR PR_SPEC_ENABLE
1083 The speculation feature is enabled, mitigation is disabled.
1084 .TP
1085 .BR PR_SPEC_DISABLE
1086 The speculation feature is disabled, mitigation is enabled
1087 .TP
1088 .BR PR_SPEC_FORCE_DISABLE
1089 Same as
1090 .B PR_SPEC_DISABLE
1091 but cannot be undone.
1092 .RE
1093 .IP
1094 If all bits are 0,
1095 then the CPU is not affected by the speculation misfeature.
1096 .IP
1097 If
1098 .B PR_SPEC_PRCTL
1099 is set, then per-thread control of the mitigation is available.
1100 If not set,
1101 .BR prctl ()
1102 for the speculation misfeature will fail.
1103 .IP
1104 The
1105 .IR arg3 ,
1106 .IR arg4 ,
1107 and
1108 .I arg5
1109 arguments must be specified as 0; otherwise the call fails with the error
1110 .BR EINVAL .
1111 .TP
1112 .BR PR_SET_SPECULATION_CTRL " (since Linux 4.17)"
1113 .\" commit b617cfc858161140d69cc0b5cc211996b557a1c7
1114 .\" commit 356e4bfff2c5489e016fdb925adbf12a1e3950ee
1115 Sets the state of the speculation misfeature specified in
1116 .IR arg2 .
1117 Currently, the only permitted value for this argument is
1118 .B PR_SPEC_STORE_BYPASS
1119 (otherwise the call fails with the error
1120 .BR ENODEV ).
1121 This setting is a per-thread attribute.
1122 The
1123 .IR arg3
1124 argument is used to hand in the control value,
1125 which is one of the following:
1126 .RS
1127 .TP
1128 .BR PR_SPEC_ENABLE
1129 The speculation feature is enabled, mitigation is disabled.
1130 .TP
1131 .BR PR_SPEC_DISABLE
1132 The speculation feature is disabled, mitigation is enabled
1133 .TP
1134 .BR PR_SPEC_FORCE_DISABLE
1135 Same as
1136 .B PR_SPEC_DISABLE
1137 but cannot be undone.
1138 A subsequent
1139 .B
1140 prctl(..., PR_SPEC_ENABLE)
1141 will fail with the error
1142 .BR EPERM .
1143 .RE
1144 .IP
1145 Any other value in
1146 .IR arg3
1147 will result in the call failing with the error
1148 .BR ERANGE .
1149 .IP
1150 The
1151 .I arg4
1152 and
1153 .I arg5
1154 arguments must be specified as 0; otherwise the call fails with the error
1155 .BR EINVAL .
1156 .IP
1157 The speculation feature can also be controlled by the
1158 .B spec_store_bypass_disable
1159 boot parameter.
1160 This parameter may enforce a read-only policy which will result in the
1161 .BR prctl ()
1162 call failing with the error
1163 .BR ENXIO .
1164 For further details, see the kernel source file
1165 .IR Documentation/admin-guide/kernel-parameters.txt .
1166 .TP
1167 .BR PR_SET_THP_DISABLE " (since Linux 3.15)"
1168 .\" commit a0715cc22601e8830ace98366c0c2bd8da52af52
1169 Set the state of the "THP disable" flag for the calling thread.
1170 If
1171 .I arg2
1172 has a nonzero value, the flag is set, otherwise it is cleared.
1173 Setting this flag provides a method
1174 for disabling transparent huge pages
1175 for jobs where the code cannot be modified, and using a malloc hook with
1176 .BR madvise (2)
1177 is not an option (i.e., statically allocated data).
1178 The setting of the "THP disable" flag is inherited by a child created via
1179 .BR fork (2)
1180 and is preserved across
1181 .BR execve (2).
1182 .\"
1183 .TP
1184 .BR PR_TASK_PERF_EVENTS_DISABLE " (since Linux 2.6.31)"
1185 Disable all performance counters attached to the calling process,
1186 regardless of whether the counters were created by
1187 this process or another process.
1188 Performance counters created by the calling process for other
1189 processes are unaffected.
1190 For more information on performance counters, see the Linux kernel source file
1191 .IR tools/perf/design.txt .
1192 .IP
1193 Originally called
1194 .BR PR_TASK_PERF_COUNTERS_DISABLE ;
1195 .\" commit 1d1c7ddbfab358445a542715551301b7fc363e28
1196 renamed (retaining the same numerical value)
1197 in Linux 2.6.32.
1198 .\"
1199 .TP
1200 .BR PR_TASK_PERF_EVENTS_ENABLE " (since Linux 2.6.31)"
1201 The converse of
1202 .BR PR_TASK_PERF_EVENTS_DISABLE ;
1203 enable performance counters attached to the calling process.
1204 .IP
1205 Originally called
1206 .BR PR_TASK_PERF_COUNTERS_ENABLE ;
1207 .\" commit 1d1c7ddbfab358445a542715551301b7fc363e28
1208 renamed
1209 .\" commit cdd6c482c9ff9c55475ee7392ec8f672eddb7be6
1210 in Linux 2.6.32.
1211 .\"
1212 .TP
1213 .BR PR_GET_THP_DISABLE " (since Linux 3.15)"
1214 Return (as the function result) the current setting of the "THP disable"
1215 flag for the calling thread:
1216 either 1, if the flag is set, or 0, if it is not.
1217 .TP
1218 .BR PR_GET_TID_ADDRESS " (since Linux 3.5)"
1219 .\" commit 300f786b2683f8bb1ec0afb6e1851183a479c86d
1220 Return the
1221 .I clear_child_tid
1222 address set by
1223 .BR set_tid_address (2)
1224 and the
1225 .BR clone (2)
1226 .B CLONE_CHILD_CLEARTID
1227 flag, in the location pointed to by
1228 .IR "(int\ **)\ arg2" .
1229 This feature is available only if the kernel is built with the
1230 .BR CONFIG_CHECKPOINT_RESTORE
1231 option enabled.
1232 Note that since the
1233 .BR prctl ()
1234 system call does not have a compat implementation for
1235 the AMD64 x32 and MIPS n32 ABIs,
1236 and the kernel writes out a pointer using the kernel's pointer size,
1237 this operation expects a user-space buffer of 8 (not 4) bytes on these ABIs.
1238 .TP
1239 .BR PR_SET_TIMERSLACK " (since Linux 2.6.28)"
1240 .\" See https://lwn.net/Articles/369549/
1241 .\" commit 6976675d94042fbd446231d1bd8b7de71a980ada
1242 Each thread has two associated timer slack values:
1243 a "default" value, and a "current" value.
1244 This operation sets the "current" timer slack value for the calling thread.
1245 .I arg2
1246 is an unsigned long value, then maximum "current" value is ULONG_MAX and
1247 the minimum "current" value is 1.
1248 If the nanosecond value supplied in
1249 .IR arg2
1250 is greater than zero, then the "current" value is set to this value.
1251 If
1252 .I arg2
1253 is equal to zero,
1254 the "current" timer slack is reset to the
1255 thread's "default" timer slack value.
1256 .IP
1257 The "current" timer slack is used by the kernel to group timer expirations
1258 for the calling thread that are close to one another;
1259 as a consequence, timer expirations for the thread may be
1260 up to the specified number of nanoseconds late (but will never expire early).
1261 Grouping timer expirations can help reduce system power consumption
1262 by minimizing CPU wake-ups.
1263 .IP
1264 The timer expirations affected by timer slack are those set by
1265 .BR select (2),
1266 .BR pselect (2),
1267 .BR poll (2),
1268 .BR ppoll (2),
1269 .BR epoll_wait (2),
1270 .BR epoll_pwait (2),
1271 .BR clock_nanosleep (2),
1272 .BR nanosleep (2),
1273 and
1274 .BR futex (2)
1275 (and thus the library functions implemented via futexes, including
1276 .\" List obtained by grepping for futex usage in glibc source
1277 .BR pthread_cond_timedwait (3),
1278 .BR pthread_mutex_timedlock (3),
1279 .BR pthread_rwlock_timedrdlock (3),
1280 .BR pthread_rwlock_timedwrlock (3),
1281 and
1282 .BR sem_timedwait (3)).
1283 .IP
1284 Timer slack is not applied to threads that are scheduled under
1285 a real-time scheduling policy (see
1286 .BR sched_setscheduler (2)).
1287 .IP
1288 When a new thread is created,
1289 the two timer slack values are made the same as the "current" value
1290 of the creating thread.
1291 Thereafter, a thread can adjust its "current" timer slack value via
1292 .BR PR_SET_TIMERSLACK .
1293 The "default" value can't be changed.
1294 The timer slack values of
1295 .IR init
1296 (PID 1), the ancestor of all processes,
1297 are 50,000 nanoseconds (50 microseconds).
1298 The timer slack value is inherited by a child created via
1299 .BR fork (2),
1300 and is preserved across
1301 .BR execve (2).
1302 .IP
1303 Since Linux 4.6, the "current" timer slack value of any process
1304 can be examined and changed via the file
1305 .IR /proc/[pid]/timerslack_ns .
1306 See
1307 .BR proc (5).
1308 .TP
1309 .BR PR_GET_TIMERSLACK " (since Linux 2.6.28)"
1310 Return (as the function result)
1311 the "current" timer slack value of the calling thread.
1312 .TP
1313 .BR PR_SET_TIMING " (since Linux 2.6.0)"
1314 .\" Precisely: Linux 2.6.0-test4
1315 Set whether to use (normal, traditional) statistical process timing or
1316 accurate timestamp-based process timing, by passing
1317 .B PR_TIMING_STATISTICAL
1318 .\" 0
1319 or
1320 .B PR_TIMING_TIMESTAMP
1321 .\" 1
1322 to \fIarg2\fP.
1323 .B PR_TIMING_TIMESTAMP
1324 is not currently implemented
1325 (attempting to set this mode will yield the error
1326 .BR EINVAL ).
1327 .\" PR_TIMING_TIMESTAMP doesn't do anything in 2.6.26-rc8,
1328 .\" and looking at the patch history, it appears
1329 .\" that it never did anything.
1330 .TP
1331 .BR PR_GET_TIMING " (since Linux 2.6.0)"
1332 .\" Precisely: Linux 2.6.0-test4
1333 Return (as the function result) which process timing method is currently
1334 in use.
1335 .TP
1336 .BR PR_SET_TSC " (since Linux 2.6.26, x86 only)"
1337 Set the state of the flag determining whether the timestamp counter
1338 can be read by the process.
1339 Pass
1340 .B PR_TSC_ENABLE
1341 to
1342 .I arg2
1343 to allow it to be read, or
1344 .B PR_TSC_SIGSEGV
1345 to generate a
1346 .B SIGSEGV
1347 when the process tries to read the timestamp counter.
1348 .TP
1349 .BR PR_GET_TSC " (since Linux 2.6.26, x86 only)"
1350 Return the state of the flag determining whether the timestamp counter
1351 can be read,
1352 in the location pointed to by
1353 .IR "(int\ *) arg2" .
1354 .TP
1355 .B PR_SET_UNALIGN
1356 (Only on: ia64, since Linux 2.3.48; parisc, since Linux 2.6.15;
1357 PowerPC, since Linux 2.6.18; Alpha, since Linux 2.6.22;
1358 .\" sh: 94ea5e449ae834af058ef005d16a8ad44fcf13d6
1359 .\" tile: 2f9ac29eec71a696cb0dcc5fb82c0f8d4dac28c9
1360 sh, since Linux 2.6.34; tile, since Linux 3.12)
1361 Set unaligned access control bits to \fIarg2\fP.
1362 Pass
1363 \fBPR_UNALIGN_NOPRINT\fP to silently fix up unaligned user accesses,
1364 or \fBPR_UNALIGN_SIGBUS\fP to generate
1365 .B SIGBUS
1366 on unaligned user access.
1367 Alpha also supports an additional flag with the value
1368 of 4 and no corresponding named constant,
1369 which instructs kernel to not fix up
1370 unaligned accesses (it is analogous to providing the
1371 .BR UAC_NOFIX
1372 flag in
1373 .BR SSI_NVPAIRS
1374 operation of the
1375 .BR setsysinfo ()
1376 system call on Tru64).
1377 .TP
1378 .B PR_GET_UNALIGN
1379 (See
1380 .B PR_SET_UNALIGN
1381 for information on versions and architectures.)
1382 Return unaligned access control bits, in the location pointed to by
1383 .IR "(unsigned int\ *) arg2" .
1384 .TP
1385 .BR PR_SET_IO_FLUSHER " (since Linux 5.6)"
1386 If a user process is involved in the block layer or filesystem I/O path,
1387 and can allocate memory while processing I/O requests it must set
1388 \fIarg2\fP to 1.
1389 This will put the process in the IO_FLUSHER state,
1390 which allows it special treatment to make progress when allocating memory.
1391 If \fIarg2\fP is 0, the process will clear the IO_FLUSHER state, and
1392 the default behavior will be used.
1393 .IP
1394 The calling process must have the
1395 .BR CAP_SYS_RESOURCE
1396 capability.
1397 .IP
1398 .IR arg3 ,
1399 .IR arg4 ,
1400 and
1401 .IR arg5
1402 must be zero.
1403 .IP
1404 The IO_FLUSHER state is inherited by a child process created via
1405 .BR fork (2)
1406 and is preserved across
1407 .BR execve (2).
1408 .IP
1409 Examples of IO_FLUSHER applications are FUSE daemons, SCSI device
1410 emulation daemons, and daemons that perform error handling like multipath
1411 path recovery applications.
1412 .TP
1413 .B PR_GET_IO_FLUSHER (Since Linux 5.6)
1414 Return (as the function result) the IO_FLUSHER state of the caller.
1415 A value of 1 indicates that the caller is in the IO_FLUSHER state;
1416 0 indicates that the caller is not in the IO_FLUSHER state.
1417 .IP
1418 The calling process must have the
1419 .BR CAP_SYS_RESOURCE
1420 capability.
1421 .IP
1422 .IR arg2 ,
1423 .IR arg3 ,
1424 .IR arg4 ,
1425 and
1426 .IR arg5
1427 must be zero.
1428 .SH RETURN VALUE
1429 On success,
1430 .BR PR_GET_DUMPABLE ,
1431 .BR PR_GET_FP_MODE ,
1432 .BR PR_GET_KEEPCAPS ,
1433 .BR PR_GET_NO_NEW_PRIVS ,
1434 .BR PR_GET_THP_DISABLE ,
1435 .BR PR_CAPBSET_READ ,
1436 .BR PR_GET_TIMING ,
1437 .BR PR_GET_TIMERSLACK ,
1438 .BR PR_GET_SECUREBITS ,
1439 .BR PR_GET_SPECULATION_CTRL ,
1440 .BR PR_MCE_KILL_GET ,
1441 .BR PR_CAP_AMBIENT + PR_CAP_AMBIENT_IS_SET ,
1442 .BR PR_GET_IO_FLUSHER ,
1443 and (if it returns)
1444 .BR PR_GET_SECCOMP
1445 return the nonnegative values described above.
1446 All other
1447 .I option
1448 values return 0 on success.
1449 On error, \-1 is returned, and
1450 .I errno
1451 is set appropriately.
1452 .SH ERRORS
1453 .TP
1454 .B EACCES
1455 .I option
1456 is
1457 .BR PR_SET_SECCOMP
1458 and
1459 .I arg2
1460 is
1461 .BR SECCOMP_MODE_FILTER ,
1462 but the process does not have the
1463 .BR CAP_SYS_ADMIN
1464 capability or has not set the
1465 .IR no_new_privs
1466 attribute (see the discussion of
1467 .BR PR_SET_NO_NEW_PRIVS
1468 above).
1469 .TP
1470 .B EACCES
1471 .I option
1472 is
1473 .BR PR_SET_MM ,
1474 and
1475 .I arg3
1476 is
1477 .BR PR_SET_MM_EXE_FILE ,
1478 the file is not executable.
1479 .TP
1480 .B EBADF
1481 .I option
1482 is
1483 .BR PR_SET_MM ,
1484 .I arg3
1485 is
1486 .BR PR_SET_MM_EXE_FILE ,
1487 and the file descriptor passed in
1488 .I arg4
1489 is not valid.
1490 .TP
1491 .B EBUSY
1492 .I option
1493 is
1494 .BR PR_SET_MM ,
1495 .I arg3
1496 is
1497 .BR PR_SET_MM_EXE_FILE ,
1498 and this the second attempt to change the
1499 .I /proc/pid/exe
1500 symbolic link, which is prohibited.
1501 .TP
1502 .B EFAULT
1503 .I arg2
1504 is an invalid address.
1505 .TP
1506 .B EFAULT
1507 .I option
1508 is
1509 .BR PR_SET_SECCOMP ,
1510 .I arg2
1511 is
1512 .BR SECCOMP_MODE_FILTER ,
1513 the system was built with
1514 .BR CONFIG_SECCOMP_FILTER ,
1515 and
1516 .I arg3
1517 is an invalid address.
1518 .TP
1519 .B EINVAL
1520 The value of
1521 .I option
1522 is not recognized.
1523 .TP
1524 .B EINVAL
1525 .I option
1526 is
1527 .BR PR_MCE_KILL
1528 or
1529 .BR PR_MCE_KILL_GET
1530 or
1531 .BR PR_SET_MM ,
1532 and unused
1533 .BR prctl ()
1534 arguments were not specified as zero.
1535 .TP
1536 .B EINVAL
1537 .I arg2
1538 is not valid value for this
1539 .IR option .
1540 .TP
1541 .B EINVAL
1542 .I option
1543 is
1544 .BR PR_SET_SECCOMP
1545 or
1546 .BR PR_GET_SECCOMP ,
1547 and the kernel was not configured with
1548 .BR CONFIG_SECCOMP .
1549 .TP
1550 .B EINVAL
1551 .I option
1552 is
1553 .BR PR_SET_SECCOMP ,
1554 .I arg2
1555 is
1556 .BR SECCOMP_MODE_FILTER ,
1557 and the kernel was not configured with
1558 .BR CONFIG_SECCOMP_FILTER .
1559 .TP
1560 .B EINVAL
1561 .I option
1562 is
1563 .BR PR_SET_MM ,
1564 and one of the following is true
1565 .RS
1566 .IP * 3
1567 .I arg4
1568 or
1569 .I arg5
1570 is nonzero;
1571 .IP *
1572 .I arg3
1573 is greater than
1574 .B TASK_SIZE
1575 (the limit on the size of the user address space for this architecture);
1576 .IP *
1577 .I arg2
1578 is
1579 .BR PR_SET_MM_START_CODE ,
1580 .BR PR_SET_MM_END_CODE ,
1581 .BR PR_SET_MM_START_DATA ,
1582 .BR PR_SET_MM_END_DATA ,
1583 or
1584 .BR PR_SET_MM_START_STACK ,
1585 and the permissions of the corresponding memory area are not as required;
1586 .IP *
1587 .I arg2
1588 is
1589 .BR PR_SET_MM_START_BRK
1590 or
1591 .BR PR_SET_MM_BRK ,
1592 and
1593 .I arg3
1594 is less than or equal to the end of the data segment
1595 or specifies a value that would cause the
1596 .B RLIMIT_DATA
1597 resource limit to be exceeded.
1598 .RE
1599 .TP
1600 .B EINVAL
1601 .I option
1602 is
1603 .BR PR_SET_PTRACER
1604 and
1605 .I arg2
1606 is not 0,
1607 .BR PR_SET_PTRACER_ANY ,
1608 or the PID of an existing process.
1609 .TP
1610 .B EINVAL
1611 .I option
1612 is
1613 .B PR_SET_PDEATHSIG
1614 and
1615 .I arg2
1616 is not a valid signal number.
1617 .TP
1618 .B EINVAL
1619 .I option
1620 is
1621 .BR PR_SET_DUMPABLE
1622 and
1623 .I arg2
1624 is neither
1625 .B SUID_DUMP_DISABLE
1626 nor
1627 .BR SUID_DUMP_USER .
1628 .TP
1629 .B EINVAL
1630 .I option
1631 is
1632 .BR PR_SET_TIMING
1633 and
1634 .I arg2
1635 is not
1636 .BR PR_TIMING_STATISTICAL .
1637 .TP
1638 .B EINVAL
1639 .I option
1640 is
1641 .BR PR_SET_NO_NEW_PRIVS
1642 and
1643 .I arg2
1644 is not equal to 1
1645 or
1646 .IR arg3 ,
1647 .IR arg4 ,
1648 or
1649 .IR arg5
1650 is nonzero.
1651 .TP
1652 .B EINVAL
1653 .I option
1654 is
1655 .BR PR_GET_NO_NEW_PRIVS
1656 and
1657 .IR arg2 ,
1658 .IR arg3 ,
1659 .IR arg4 ,
1660 or
1661 .IR arg5
1662 is nonzero.
1663 .TP
1664 .B EINVAL
1665 .I option
1666 is
1667 .BR PR_SET_THP_DISABLE
1668 and
1669 .IR arg3 ,
1670 .IR arg4 ,
1671 or
1672 .IR arg5
1673 is nonzero.
1674 .TP
1675 .B EINVAL
1676 .I option
1677 is
1678 .BR PR_GET_THP_DISABLE
1679 and
1680 .IR arg2 ,
1681 .IR arg3 ,
1682 .IR arg4 ,
1683 or
1684 .IR arg5
1685 is nonzero.
1686 .TP
1687 .B EINVAL
1688 .I option
1689 is
1690 .B PR_CAP_AMBIENT
1691 and an unused argument
1692 .RI ( arg4 ,
1693 .IR arg5 ,
1694 or,
1695 in the case of
1696 .BR PR_CAP_AMBIENT_CLEAR_ALL ,
1697 .IR arg3 )
1698 is nonzero; or
1699 .IR arg2
1700 has an invalid value;
1701 or
1702 .IR arg2
1703 is
1704 .BR PR_CAP_AMBIENT_LOWER ,
1705 .BR PR_CAP_AMBIENT_RAISE ,
1706 or
1707 .BR PR_CAP_AMBIENT_IS_SET
1708 and
1709 .IR arg3
1710 does not specify a valid capability.
1711 .TP
1712 .B EINVAL
1713 .I option
1714 was
1715 .BR PR_GET_SPECULATION_CTRL
1716 or
1717 .BR PR_SET_SPECULATION_CTRL
1718 and unused arguments to
1719 .BR prctl ()
1720 are not 0.
1721 .TP
1722 .B ENODEV
1723 .I option
1724 was
1725 .BR PR_SET_SPECULATION_CTRL
1726 the kernel or CPU does not support the requested speculation misfeature.
1727 .TP
1728 .B ENXIO
1729 .I option
1730 was
1731 .BR PR_MPX_ENABLE_MANAGEMENT
1732 or
1733 .BR PR_MPX_DISABLE_MANAGEMENT
1734 and the kernel or the CPU does not support MPX management.
1735 Check that the kernel and processor have MPX support.
1736 .TP
1737 .B ENXIO
1738 .I option
1739 was
1740 .BR PR_SET_SPECULATION_CTRL
1741 implies that the control of the selected speculation misfeature is not possible.
1742 See
1743 .BR PR_GET_SPECULATION_CTRL
1744 for the bit fields to determine which option is available.
1745 .TP
1746 .B EOPNOTSUPP
1747 .I option
1748 is
1749 .B PR_SET_FP_MODE
1750 and
1751 .I arg2
1752 has an invalid or unsupported value.
1753 .TP
1754 .B EPERM
1755 .I option
1756 is
1757 .BR PR_SET_SECUREBITS ,
1758 and the caller does not have the
1759 .B CAP_SETPCAP
1760 capability,
1761 or tried to unset a "locked" flag,
1762 or tried to set a flag whose corresponding locked flag was set
1763 (see
1764 .BR capabilities (7)).
1765 .TP
1766 .B EPERM
1767 .I option
1768 is
1769 .BR PR_SET_SPECULATION_CTRL
1770 wherein the speculation was disabled with
1771 .B PR_SPEC_FORCE_DISABLE
1772 and caller tried to enable it again.
1773 .TP
1774 .B EPERM
1775 .I option
1776 is
1777 .BR PR_SET_KEEPCAPS ,
1778 and the caller's
1779 .B SECBIT_KEEP_CAPS_LOCKED
1780 flag is set
1781 (see
1782 .BR capabilities (7)).
1783 .TP
1784 .B EPERM
1785 .I option
1786 is
1787 .BR PR_CAPBSET_DROP ,
1788 and the caller does not have the
1789 .B CAP_SETPCAP
1790 capability.
1791 .TP
1792 .B EPERM
1793 .I option
1794 is
1795 .BR PR_SET_MM ,
1796 and the caller does not have the
1797 .B CAP_SYS_RESOURCE
1798 capability.
1799 .TP
1800 .B EPERM
1801 .IR option
1802 is
1803 .BR PR_CAP_AMBIENT
1804 and
1805 .IR arg2
1806 is
1807 .BR PR_CAP_AMBIENT_RAISE ,
1808 but either the capability specified in
1809 .IR arg3
1810 is not present in the process's permitted and inheritable capability sets,
1811 or the
1812 .B PR_CAP_AMBIENT_LOWER
1813 securebit has been set.
1814 .TP
1815 .B ERANGE
1816 .I option
1817 was
1818 .BR PR_SET_SPECULATION_CTRL
1819 and
1820 .IR arg3
1821 is neither
1822 .BR PR_SPEC_ENABLE ,
1823 .BR PR_SPEC_DISABLE ,
1824 nor
1825 .BR PR_SPEC_FORCE_DISABLE .
1826 .SH VERSIONS
1827 The
1828 .BR prctl ()
1829 system call was introduced in Linux 2.1.57.
1830 .\" The library interface was added in glibc 2.0.6
1831 .SH CONFORMING TO
1832 This call is Linux-specific.
1833 IRIX has a
1834 .BR prctl ()
1835 system call (also introduced in Linux 2.1.44
1836 as irix_prctl on the MIPS architecture),
1837 with prototype
1838 .PP
1839 .in +4n
1840 .EX
1841 .BI "ptrdiff_t prctl(int " option ", int " arg2 ", int " arg3 );
1842 .EE
1843 .in
1844 .PP
1845 and options to get the maximum number of processes per user,
1846 get the maximum number of processors the calling process can use,
1847 find out whether a specified process is currently blocked,
1848 get or set the maximum stack size, and so on.
1849 .SH SEE ALSO
1850 .BR signal (2),
1851 .BR core (5)