man2/prctl.2

   1 .\" Copyright (C) 1998 Andries Brouwer (aeb@cwi.nl)
   2 .\" and Copyright (C) 2002, 2006, 2008, 2012, 2013, 2015 Michael Kerrisk <mtk.manpages@gmail.com>
   3 .\" and Copyright Guillem Jover <guillem@hadrons.org>
   4 .\" and Copyright (C) 2010 Andi Kleen <andi@firstfloor.org>
   5 .\" and Copyright (C) 2012 Cyrill Gorcunov <gorcunov@openvz.org>
   6 .\" and Copyright (C) 2014 Dave Hansen / Intel
   7 .\" and Copyright (c) 2016 Eugene Syromyatnikov <evgsyr@gmail.com>
   8 .\" and Copyright (c) 2018 Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
   9 .\" and Copyright (c) 2020 Dave Martin <Dave.Martin@arm.com>
  10 .\"
  11 .\" SPDX-License-Identifier: Linux-man-pages-copyleft
  12 .\"
  13 .\" Modified Thu Nov 11 04:19:42 MET 1999, aeb: added PR_GET_PDEATHSIG
  14 .\" Modified 27 Jun 02, Michael Kerrisk
  15 .\"     Added PR_SET_DUMPABLE, PR_GET_DUMPABLE,
  16 .\"     PR_SET_KEEPCAPS, PR_GET_KEEPCAPS
  17 .\" Modified 2006-08-30 Guillem Jover <guillem@hadrons.org>
  18 .\"     Updated Linux versions where the options where introduced.
  19 .\"     Added PR_SET_TIMING, PR_GET_TIMING, PR_SET_NAME, PR_GET_NAME,
  20 .\"     PR_SET_UNALIGN, PR_GET_UNALIGN, PR_SET_FPEMU, PR_GET_FPEMU,
  21 .\"     PR_SET_FPEXC, PR_GET_FPEXC
  22 .\" 2008-04-29 Serge Hallyn, Document PR_CAPBSET_READ and PR_CAPBSET_DROP
  23 .\" 2008-06-13 Erik Bosman, <ejbosman@cs.vu.nl>
  24 .\"     Document PR_GET_TSC and PR_SET_TSC.
  25 .\" 2008-06-15 mtk, Document PR_SET_SECCOMP, PR_GET_SECCOMP
  26 .\" 2009-10-03 Andi Kleen, document PR_MCE_KILL
  27 .\" 2012-04 Cyrill Gorcunov, Document PR_SET_MM
  28 .\" 2012-04-25 Michael Kerrisk, Document PR_TASK_PERF_EVENTS_DISABLE and
  29 .\"                             PR_TASK_PERF_EVENTS_ENABLE
  30 .\" 2012-09-20 Kees Cook, update PR_SET_SECCOMP for mode 2
  31 .\" 2012-09-20 Kees Cook, document PR_SET_NO_NEW_PRIVS, PR_GET_NO_NEW_PRIVS
  32 .\" 2012-10-25 Michael Kerrisk, Document PR_SET_TIMERSLACK and
  33 .\"                             PR_GET_TIMERSLACK
  34 .\" 2013-01-10 Kees Cook, document PR_SET_PTRACER
  35 .\" 2012-02-04 Michael Kerrisk, document PR_{SET,GET}_CHILD_SUBREAPER
  36 .\" 2014-11-10 Dave Hansen, document PR_MPX_{EN,DIS}ABLE_MANAGEMENT
  37 .\"
  38 .\"
  39 .TH PRCTL 2 2021-03-22 "Linux" "Linux Programmer's Manual"
  40 .SH NAME
  41 prctl \- operations on a process or thread
  42 .SH LIBRARY
  43 Standard C library
  44 .RI ( libc ", " \-lc )
  45 .SH SYNOPSIS
  46 .nf
  47 .B #include <sys/prctl.h>
  48 .PP
  49 .BI "int prctl(int " option ", unsigned long " arg2 ", unsigned long " arg3 ,
  50 .BI "          unsigned long " arg4 ", unsigned long " arg5 );
  51 .fi
  52 .SH DESCRIPTION
  53 .BR prctl ()
  54 manipulates various aspects of the behavior
  55 of the calling thread or process.
  56 .PP
  57 Note that careless use of some
  58 .BR prctl ()
  59 operations can confuse the user-space run-time environment,
  60 so these operations should be used with care.
  61 .PP
  62 .BR prctl ()
  63 is called with a first argument describing what to do
  64 (with values defined in \fI<linux/prctl.h>\fP), and further
  65 arguments with a significance depending on the first one.
  66 The first argument can be:
  67 .\"
  68 .\" prctl PR_CAP_AMBIENT
  69 .TP
  70 .BR PR_CAP_AMBIENT " (since Linux 4.3)"
  71 .\" commit 58319057b7847667f0c9585b9de0e8932b0fdb08
  72 Reads or changes the ambient capability set of the calling thread,
  73 according to the value of
  74 .IR arg2 ,
  75 which must be one of the following:
  76 .RS
  77 .\"
  78 .TP
  79 .B PR_CAP_AMBIENT_RAISE
  80 The capability specified in
  81 .I arg3
  82 is added to the ambient set.
  83 The specified capability must already be present in
  84 both the permitted and the inheritable sets of the process.
  85 This operation is not permitted if the
  86 .B SECBIT_NO_CAP_AMBIENT_RAISE
  87 securebit is set.
  88 .TP
  89 .B PR_CAP_AMBIENT_LOWER
  90 The capability specified in
  91 .I arg3
  92 is removed from the ambient set.
  93 .TP
  94 .B PR_CAP_AMBIENT_IS_SET
  95 The
  96 .BR prctl ()
  97 call returns 1 if the capability in
  98 .I arg3
  99 is in the ambient set and 0 if it is not.
 100 .TP
 101 .BR PR_CAP_AMBIENT_CLEAR_ALL
 102 All capabilities will be removed from the ambient set.
 103 This operation requires setting
 104 .I arg3
 105 to zero.
 106 .RE
 107 .IP
 108 In all of the above operations,
 109 .I arg4
 110 and
 111 .I arg5
 112 must be specified as 0.
 113 .IP
 114 Higher-level interfaces layered on top of the above operations are
 115 provided in the
 116 .BR libcap (3)
 117 library in the form of
 118 .BR cap_get_ambient (3),
 119 .BR cap_set_ambient (3),
 120 and
 121 .BR cap_reset_ambient (3).
 122 .\" prctl PR_CAPBSET_READ
 123 .TP
 124 .BR PR_CAPBSET_READ " (since Linux 2.6.25)"
 125 Return (as the function result) 1 if the capability specified in
 126 .I arg2
 127 is in the calling thread's capability bounding set,
 128 or 0 if it is not.
 129 (The capability constants are defined in
 130 .IR <linux/capability.h> .)
 131 The capability bounding set dictates
 132 whether the process can receive the capability through a
 133 file's permitted capability set on a subsequent call to
 134 .BR execve (2).
 135 .IP
 136 If the capability specified in
 137 .I arg2
 138 is not valid, then the call fails with the error
 139 .BR EINVAL .
 140 .IP
 141 A higher-level interface layered on top of this operation is provided in the
 142 .BR libcap (3)
 143 library in the form of
 144 .BR cap_get_bound (3).
 145 .\" prctl PR_CAPBSET_DROP
 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 .\" prctl PR_SET_CHILD_SUBREAPER
 173 .TP
 174 .BR PR_SET_CHILD_SUBREAPER " (since Linux 3.4)"
 175 .\" commit ebec18a6d3aa1e7d84aab16225e87fd25170ec2b
 176 If
 177 .I arg2
 178 is nonzero,
 179 set the "child subreaper" attribute of the calling process;
 180 if
 181 .I arg2
 182 is zero, unset the attribute.
 183 .IP
 184 A subreaper fulfills the role of
 185 .BR init (1)
 186 for its descendant processes.
 187 When a process becomes orphaned
 188 (i.e., its immediate parent terminates),
 189 then that process will be reparented to
 190 the nearest still living ancestor subreaper.
 191 Subsequently, calls to
 192 .BR getppid (2)
 193 in the orphaned process will now return the PID of the subreaper process,
 194 and when the orphan terminates, it is the subreaper process that
 195 will receive a
 196 .BR SIGCHLD
 197 signal and will be able to
 198 .BR wait (2)
 199 on the process to discover its termination status.
 200 .IP
 201 The setting of the "child subreaper" attribute
 202 is not inherited by children created by
 203 .BR fork (2)
 204 and
 205 .BR clone (2).
 206 The setting is preserved across
 207 .BR execve (2).
 208 .IP
 209 Establishing a subreaper process is useful in session management frameworks
 210 where a hierarchical group of processes is managed by a subreaper process
 211 that needs to be informed when one of the processes\(emfor example,
 212 a double-forked daemon\(emterminates
 213 (perhaps so that it can restart that process).
 214 Some
 215 .BR init (1)
 216 frameworks (e.g.,
 217 .BR systemd (1))
 218 employ a subreaper process for similar reasons.
 219 .\" prctl PR_GET_CHILD_SUBREAPER
 220 .TP
 221 .BR PR_GET_CHILD_SUBREAPER " (since Linux 3.4)"
 222 Return the "child subreaper" setting of the caller,
 223 in the location pointed to by
 224 .IR "(int\ *) arg2" .
 225 .\" prctl PR_SET_DUMPABLE
 226 .TP
 227 .BR PR_SET_DUMPABLE " (since Linux 2.3.20)"
 228 Set the state of the "dumpable" attribute,
 229 which determines whether core dumps are produced for the calling process
 230 upon delivery of a signal whose default behavior is to produce a core dump.
 231 .IP
 232 In kernels up to and including 2.6.12,
 233 .I arg2
 234 must be either 0
 235 .RB ( SUID_DUMP_DISABLE ,
 236 process is not dumpable) or 1
 237 .RB ( SUID_DUMP_USER ,
 238 process is dumpable).
 239 Between kernels 2.6.13 and 2.6.17,
 240 .\" commit abf75a5033d4da7b8a7e92321d74021d1fcfb502
 241 the value 2 was also permitted,
 242 which caused any binary which normally would not be dumped
 243 to be dumped readable by root only;
 244 for security reasons, this feature has been removed.
 245 .\" See http://marc.theaimsgroup.com/?l=linux-kernel&m=115270289030630&w=2
 246 .\" Subject:    Fix prctl privilege escalation (CVE-2006-2451)
 247 .\" From:       Marcel Holtmann <marcel () holtmann ! org>
 248 .\" Date:       2006-07-12 11:12:00
 249 (See also the description of
 250 .I /proc/sys/fs/\:suid_dumpable
 251 in
 252 .BR proc (5).)
 253 .IP
 254 Normally, the "dumpable" attribute is set to 1.
 255 However, it is reset to the current value contained in the file
 256 .IR /proc/sys/fs/\:suid_dumpable
 257 (which by default has the value 0),
 258 in the following circumstances:
 259 .\" See kernel/cred.c::commit_creds() (Linux 3.18 sources)
 260 .RS
 261 .IP * 3
 262 The process's effective user or group ID is changed.
 263 .IP *
 264 The process's filesystem user or group ID is changed (see
 265 .BR credentials (7)).
 266 .IP *
 267 The process executes
 268 .RB ( execve (2))
 269 a set-user-ID or set-group-ID program, resulting in a change
 270 of either the effective user ID or the effective group ID.
 271 .IP *
 272 The process executes
 273 .RB ( execve (2))
 274 a program that has file capabilities (see
 275 .BR capabilities (7)),
 276 .\" See kernel/cred.c::commit_creds()
 277 but only if the permitted capabilities
 278 gained exceed those already permitted for the process.
 279 .\" Also certain namespace operations;
 280 .RE
 281 .IP
 282 Processes that are not dumpable can not be attached via
 283 .BR ptrace (2)
 284 .BR PTRACE_ATTACH ;
 285 see
 286 .BR ptrace (2)
 287 for further details.
 288 .IP
 289 If a process is not dumpable,
 290 the ownership of files in the process's
 291 .IR /proc/[pid]
 292 directory is affected as described in
 293 .BR proc (5).
 294 .\" prctl PR_GET_DUMPABLE
 295 .TP
 296 .BR PR_GET_DUMPABLE " (since Linux 2.3.20)"
 297 Return (as the function result) the current state of the calling
 298 process's dumpable attribute.
 299 .\" Since Linux 2.6.13, the dumpable flag can have the value 2,
 300 .\" but in 2.6.13 PR_GET_DUMPABLE simply returns 1 if the dumpable
 301 .\" flags has a nonzero value.  This was fixed in 2.6.14.
 302 .\" prctl PR_SET_ENDIAN
 303 .TP
 304 .BR PR_SET_ENDIAN " (since Linux 2.6.18, PowerPC only)"
 305 Set the endian-ness of the calling process to the value given
 306 in \fIarg2\fP, which should be one of the following:
 307 .\" Respectively 0, 1, 2
 308 .BR PR_ENDIAN_BIG ,
 309 .BR PR_ENDIAN_LITTLE ,
 310 or
 311 .B PR_ENDIAN_PPC_LITTLE
 312 (PowerPC pseudo little endian).
 313 .\" prctl PR_GET_ENDIAN
 314 .TP
 315 .BR PR_GET_ENDIAN " (since Linux 2.6.18, PowerPC only)"
 316 Return the endian-ness of the calling process,
 317 in the location pointed to by
 318 .IR "(int\ *) arg2" .
 319 .\" prctl PR_SET_FP_MODE
 320 .TP
 321 .BR PR_SET_FP_MODE " (since Linux 4.0, only on MIPS)"
 322 .\" commit 9791554b45a2acc28247f66a5fd5bbc212a6b8c8
 323 On the MIPS architecture,
 324 user-space code can be built using an ABI which permits linking
 325 with code that has more restrictive floating-point (FP) requirements.
 326 For example, user-space code may be built to target the O32 FPXX ABI
 327 and linked with code built for either one of the more restrictive
 328 FP32 or FP64 ABIs.
 329 When more restrictive code is linked in,
 330 the overall requirement for the process is to use the more
 331 restrictive floating-point mode.
 332 .IP
 333 Because the kernel has no means of knowing in advance
 334 which mode the process should be executed in,
 335 and because these restrictions can
 336 change over the lifetime of the process, the
 337 .B PR_SET_FP_MODE
 338 operation is provided to allow control of the floating-point mode
 339 from user space.
 340 .IP
 341 .\" https://dmz-portal.mips.com/wiki/MIPS_O32_ABI_-_FR0_and_FR1_Interlinking
 342 The
 343 .I (unsigned int) arg2
 344 argument is a bit mask describing the floating-point mode used:
 345 .RS
 346 .TP
 347 .BR PR_FP_MODE_FR
 348 When this bit is
 349 .I unset
 350 (so called
 351 .BR FR=0 " or " FR0
 352 mode), the 32 floating-point registers are 32 bits wide,
 353 and 64-bit registers are represented as a pair of registers
 354 (even- and odd- numbered,
 355 with the even-numbered register containing the lower 32 bits,
 356 and the odd-numbered register containing the higher 32 bits).
 357 .IP
 358 When this bit is
 359 .I set
 360 (on supported hardware),
 361 the 32 floating-point registers are 64 bits wide (so called
 362 .BR FR=1 " or " FR1
 363 mode).
 364 Note that modern MIPS implementations (MIPS R6 and newer) support
 365 .B FR=1
 366 mode only.
 367 .IP
 368 Applications that use the O32 FP32 ABI can operate only when this bit is
 369 .I unset
 370 .RB ( FR=0 ;
 371 or they can be used with FRE enabled, see below).
 372 Applications that use the O32 FP64 ABI
 373 (and the O32 FP64A ABI, which exists to
 374 provide the ability to operate with existing FP32 code; see below)
 375 can operate only when this bit is
 376 .I set
 377 .RB ( FR=1 ).
 378 Applications that use the O32 FPXX ABI can operate with either
 379 .BR FR=0
 380 or
 381 .BR FR=1 .
 382 .TP
 383 .BR PR_FP_MODE_FRE
 384 Enable emulation of 32-bit floating-point mode.
 385 When this mode is enabled,
 386 it emulates 32-bit floating-point operations
 387 by raising a reserved-instruction exception
 388 on every instruction that uses 32-bit formats and
 389 the kernel then handles the instruction in software.
 390 (The problem lies in the discrepancy of handling odd-numbered registers
 391 which are the high 32 bits of 64-bit registers with even numbers in
 392 .B FR=0
 393 mode and the lower 32-bit parts of odd-numbered 64-bit registers in
 394 .B FR=1
 395 mode.)
 396 Enabling this bit is necessary when code with the O32 FP32 ABI should operate
 397 with code with compatible the O32 FPXX or O32 FP64A ABIs (which require
 398 .B FR=1
 399 FPU mode) or when it is executed on newer hardware (MIPS R6 onwards)
 400 which lacks
 401 .B FR=0
 402 mode support when a binary with the FP32 ABI is used.
 403 .IP
 404 Note that this mode makes sense only when the FPU is in 64-bit mode
 405 .RB ( FR=1 ).
 406 .IP
 407 Note that the use of emulation inherently has a significant performance hit
 408 and should be avoided if possible.
 409 .RE
 410 .IP
 411 In the N32/N64 ABI, 64-bit floating-point mode is always used,
 412 so FPU emulation is not required and the FPU always operates in
 413 .B FR=1
 414 mode.
 415 .IP
 416 This option is mainly intended for use by the dynamic linker
 417 .RB ( ld.so (8)).
 418 .IP
 419 The arguments
 420 .IR arg3 ,
 421 .IR arg4 ,
 422 and
 423 .IR arg5
 424 are ignored.
 425 .\" prctl PR_GET_FP_MODE
 426 .TP
 427 .BR PR_GET_FP_MODE " (since Linux 4.0, only on MIPS)"
 428 Return (as the function result)
 429 the current floating-point mode (see the description of
 430 .B PR_SET_FP_MODE
 431 for details).
 432 .IP
 433 On success,
 434 the call returns a bit mask which represents the current floating-point mode.
 435 .IP
 436 The arguments
 437 .IR arg2 ,
 438 .IR arg3 ,
 439 .IR arg4 ,
 440 and
 441 .IR arg5
 442 are ignored.
 443 .\" prctl PR_SET_FPEMU
 444 .TP
 445 .BR PR_SET_FPEMU " (since Linux 2.4.18, 2.5.9, only on ia64)"
 446 Set floating-point emulation control bits to \fIarg2\fP.
 447 Pass
 448 .B PR_FPEMU_NOPRINT
 449 to silently emulate floating-point operation accesses, or
 450 .B PR_FPEMU_SIGFPE
 451 to not emulate floating-point operations and send
 452 .B SIGFPE
 453 instead.
 454 .\" prctl PR_GET_FPEMU
 455 .TP
 456 .BR PR_GET_FPEMU " (since Linux 2.4.18, 2.5.9, only on ia64)"
 457 Return floating-point emulation control bits,
 458 in the location pointed to by
 459 .IR "(int\ *) arg2" .
 460 .\" prctl PR_SET_FPEXC
 461 .TP
 462 .BR PR_SET_FPEXC " (since Linux 2.4.21, 2.5.32, only on PowerPC)"
 463 Set floating-point exception mode to \fIarg2\fP.
 464 Pass \fBPR_FP_EXC_SW_ENABLE\fP to use FPEXC for FP exception enables,
 465 \fBPR_FP_EXC_DIV\fP for floating-point divide by zero,
 466 \fBPR_FP_EXC_OVF\fP for floating-point overflow,
 467 \fBPR_FP_EXC_UND\fP for floating-point underflow,
 468 \fBPR_FP_EXC_RES\fP for floating-point inexact result,
 469 \fBPR_FP_EXC_INV\fP for floating-point invalid operation,
 470 \fBPR_FP_EXC_DISABLED\fP for FP exceptions disabled,
 471 \fBPR_FP_EXC_NONRECOV\fP for async nonrecoverable exception mode,
 472 \fBPR_FP_EXC_ASYNC\fP for async recoverable exception mode,
 473 \fBPR_FP_EXC_PRECISE\fP for precise exception mode.
 474 .\" prctl PR_GET_FPEXC
 475 .TP
 476 .BR PR_GET_FPEXC " (since Linux 2.4.21, 2.5.32, only on PowerPC)"
 477 Return floating-point exception mode,
 478 in the location pointed to by
 479 .IR "(int\ *) arg2" .
 480 .\" prctl PR_SET_IO_FLUSHER
 481 .TP
 482 .BR PR_SET_IO_FLUSHER " (since Linux 5.6)"
 483 If a user process is involved in the block layer or filesystem I/O path,
 484 and can allocate memory while processing I/O requests it must set
 485 \fIarg2\fP to 1.
 486 This will put the process in the IO_FLUSHER state,
 487 which allows it special treatment to make progress when allocating memory.
 488 If \fIarg2\fP is 0, the process will clear the IO_FLUSHER state, and
 489 the default behavior will be used.
 490 .IP
 491 The calling process must have the
 492 .BR CAP_SYS_RESOURCE
 493 capability.
 494 .IP
 495 .IR arg3 ,
 496 .IR arg4 ,
 497 and
 498 .IR arg5
 499 must be zero.
 500 .IP
 501 The IO_FLUSHER state is inherited by a child process created via
 502 .BR fork (2)
 503 and is preserved across
 504 .BR execve (2).
 505 .IP
 506 Examples of IO_FLUSHER applications are FUSE daemons, SCSI device
 507 emulation daemons, and daemons that perform error handling like multipath
 508 path recovery applications.
 509 .\" prctl PR_GET_IO_FLUSHER
 510 .TP
 511 .B PR_GET_IO_FLUSHER (Since Linux 5.6)
 512 Return (as the function result) the IO_FLUSHER state of the caller.
 513 A value of 1 indicates that the caller is in the IO_FLUSHER state;
 514 0 indicates that the caller is not in the IO_FLUSHER state.
 515 .IP
 516 The calling process must have the
 517 .BR CAP_SYS_RESOURCE
 518 capability.
 519 .IP
 520 .IR arg2 ,
 521 .IR arg3 ,
 522 .IR arg4 ,
 523 and
 524 .IR arg5
 525 must be zero.
 526 .\" prctl PR_SET_KEEPCAPS
 527 .TP
 528 .BR PR_SET_KEEPCAPS " (since Linux 2.2.18)"
 529 Set the state of the calling thread's "keep capabilities" flag.
 530 The effect of this flag is described in
 531 .BR capabilities (7).
 532 .I arg2
 533 must be either 0 (clear the flag)
 534 or 1 (set the flag).
 535 The "keep capabilities" value will be reset to 0 on subsequent calls to
 536 .BR execve (2).
 537 .\" prctl PR_GET_KEEPCAPS
 538 .TP
 539 .BR PR_GET_KEEPCAPS " (since Linux 2.2.18)"
 540 Return (as the function result) the current state of the calling thread's
 541 "keep capabilities" flag.
 542 See
 543 .BR capabilities (7)
 544 for a description of this flag.
 545 .\" prctl PR_MCE_KILL
 546 .TP
 547 .BR PR_MCE_KILL " (since Linux 2.6.32)"
 548 Set the machine check memory corruption kill policy for the calling thread.
 549 If
 550 .I arg2
 551 is
 552 .BR PR_MCE_KILL_CLEAR ,
 553 clear the thread memory corruption kill policy and use the system-wide default.
 554 (The system-wide default is defined by
 555 .IR /proc/sys/vm/memory_failure_early_kill ;
 556 see
 557 .BR proc (5).)
 558 If
 559 .I arg2
 560 is
 561 .BR PR_MCE_KILL_SET ,
 562 use a thread-specific memory corruption kill policy.
 563 In this case,
 564 .I arg3
 565 defines whether the policy is
 566 .I early kill
 567 .RB ( PR_MCE_KILL_EARLY ),
 568 .I late kill
 569 .RB ( PR_MCE_KILL_LATE ),
 570 or the system-wide default
 571 .RB ( PR_MCE_KILL_DEFAULT ).
 572 Early kill means that the thread receives a
 573 .B SIGBUS
 574 signal as soon as hardware memory corruption is detected inside
 575 its address space.
 576 In late kill mode, the process is killed only when it accesses a corrupted page.
 577 See
 578 .BR sigaction (2)
 579 for more information on the
 580 .BR SIGBUS
 581 signal.
 582 The policy is inherited by children.
 583 The remaining unused
 584 .BR prctl ()
 585 arguments must be zero for future compatibility.
 586 .\" prctl PR_MCE_KILL_GET
 587 .TP
 588 .BR PR_MCE_KILL_GET " (since Linux 2.6.32)"
 589 Return (as the function result)
 590 the current per-process machine check kill policy.
 591 All unused
 592 .BR prctl ()
 593 arguments must be zero.
 594 .\" prctl PR_SET_MM
 595 .TP
 596 .BR PR_SET_MM " (since Linux 3.3)"
 597 .\" commit 028ee4be34a09a6d48bdf30ab991ae933a7bc036
 598 Modify certain kernel memory map descriptor fields
 599 of the calling process.
 600 Usually these fields are set by the kernel and dynamic loader (see
 601 .BR ld.so (8)
 602 for more information) and a regular application should not use this feature.
 603 However, there are cases, such as self-modifying programs,
 604 where a program might find it useful to change its own memory map.
 605 .IP
 606 The calling process must have the
 607 .BR CAP_SYS_RESOURCE
 608 capability.
 609 The value in
 610 .I arg2
 611 is one of the options below, while
 612 .I arg3
 613 provides a new value for the option.
 614 The
 615 .I arg4
 616 and
 617 .I arg5
 618 arguments must be zero if unused.
 619 .IP
 620 Before Linux 3.10,
 621 .\" commit 52b3694157e3aa6df871e283115652ec6f2d31e0
 622 this feature is available only if the kernel is built with the
 623 .BR CONFIG_CHECKPOINT_RESTORE
 624 option enabled.
 625 .RS
 626 .TP
 627 .BR PR_SET_MM_START_CODE
 628 Set the address above which the program text can run.
 629 The corresponding memory area must be readable and executable,
 630 but not writable or shareable (see
 631 .BR mprotect (2)
 632 and
 633 .BR mmap (2)
 634 for more information).
 635 .TP
 636 .BR PR_SET_MM_END_CODE
 637 Set the address below which the program text can run.
 638 The corresponding memory area must be readable and executable,
 639 but not writable or shareable.
 640 .TP
 641 .BR PR_SET_MM_START_DATA
 642 Set the address above which initialized and
 643 uninitialized (bss) data are placed.
 644 The corresponding memory area must be readable and writable,
 645 but not executable or shareable.
 646 .TP
 647 .B PR_SET_MM_END_DATA
 648 Set the address below which initialized and
 649 uninitialized (bss) data are placed.
 650 The corresponding memory area must be readable and writable,
 651 but not executable or shareable.
 652 .TP
 653 .BR PR_SET_MM_START_STACK
 654 Set the start address of the stack.
 655 The corresponding memory area must be readable and writable.
 656 .TP
 657 .BR PR_SET_MM_START_BRK
 658 Set the address above which the program heap can be expanded with
 659 .BR brk (2)
 660 call.
 661 The address must be greater than the ending address of
 662 the current program data segment.
 663 In addition, the combined size of the resulting heap and
 664 the size of the data segment can't exceed the
 665 .BR RLIMIT_DATA
 666 resource limit (see
 667 .BR setrlimit (2)).
 668 .TP
 669 .BR PR_SET_MM_BRK
 670 Set the current
 671 .BR brk (2)
 672 value.
 673 The requirements for the address are the same as for the
 674 .BR PR_SET_MM_START_BRK
 675 option.
 676 .PP
 677 The following options are available since Linux 3.5.
 678 .\" commit fe8c7f5cbf91124987106faa3bdf0c8b955c4cf7
 679 .TP
 680 .BR PR_SET_MM_ARG_START
 681 Set the address above which the program command line is placed.
 682 .TP
 683 .BR PR_SET_MM_ARG_END
 684 Set the address below which the program command line is placed.
 685 .TP
 686 .BR PR_SET_MM_ENV_START
 687 Set the address above which the program environment is placed.
 688 .TP
 689 .BR PR_SET_MM_ENV_END
 690 Set the address below which the program environment is placed.
 691 .IP
 692 The address passed with
 693 .BR PR_SET_MM_ARG_START ,
 694 .BR PR_SET_MM_ARG_END ,
 695 .BR PR_SET_MM_ENV_START ,
 696 and
 697 .BR PR_SET_MM_ENV_END
 698 should belong to a process stack area.
 699 Thus, the corresponding memory area must be readable, writable, and
 700 (depending on the kernel configuration) have the
 701 .BR MAP_GROWSDOWN
 702 attribute set (see
 703 .BR mmap (2)).
 704 .TP
 705 .BR PR_SET_MM_AUXV
 706 Set a new auxiliary vector.
 707 The
 708 .I arg3
 709 argument should provide the address of the vector.
 710 The
 711 .I arg4
 712 is the size of the vector.
 713 .TP
 714 .BR PR_SET_MM_EXE_FILE
 715 .\" commit b32dfe377102ce668775f8b6b1461f7ad428f8b6
 716 Supersede the
 717 .IR /proc/pid/exe
 718 symbolic link with a new one pointing to a new executable file
 719 identified by the file descriptor provided in
 720 .I arg3
 721 argument.
 722 The file descriptor should be obtained with a regular
 723 .BR open (2)
 724 call.
 725 .IP
 726 To change the symbolic link, one needs to unmap all existing
 727 executable memory areas, including those created by the kernel itself
 728 (for example the kernel usually creates at least one executable
 729 memory area for the ELF
 730 .IR \.text
 731 section).
 732 .IP
 733 In Linux 4.9 and earlier, the
 734 .\" commit 3fb4afd9a504c2386b8435028d43283216bf588e
 735 .BR PR_SET_MM_EXE_FILE
 736 operation can be performed only once in a process's lifetime;
 737 attempting to perform the operation a second time results in the error
 738 .BR EPERM .
 739 This restriction was enforced for security reasons that were subsequently
 740 deemed specious,
 741 and the restriction was removed in Linux 4.10 because some
 742 user-space applications needed to perform this operation more than once.
 743 .PP
 744 The following options are available since Linux 3.18.
 745 .\" commit f606b77f1a9e362451aca8f81d8f36a3a112139e
 746 .TP
 747 .BR PR_SET_MM_MAP
 748 Provides one-shot access to all the addresses by passing in a
 749 .I struct prctl_mm_map
 750 (as defined in \fI<linux/prctl.h>\fP).
 751 The
 752 .I arg4
 753 argument should provide the size of the struct.
 754 .IP
 755 This feature is available only if the kernel is built with the
 756 .BR CONFIG_CHECKPOINT_RESTORE
 757 option enabled.
 758 .TP
 759 .BR PR_SET_MM_MAP_SIZE
 760 Returns the size of the
 761 .I struct prctl_mm_map
 762 the kernel expects.
 763 This allows user space to find a compatible struct.
 764 The
 765 .I arg4
 766 argument should be a pointer to an unsigned int.
 767 .IP
 768 This feature is available only if the kernel is built with the
 769 .BR CONFIG_CHECKPOINT_RESTORE
 770 option enabled.
 771 .RE
 772 .\" prctl PR_MPX_ENABLE_MANAGEMENT
 773 .TP
 774 .BR PR_MPX_ENABLE_MANAGEMENT ", " PR_MPX_DISABLE_MANAGEMENT " (since Linux 3.19, removed in Linux 5.4; only on x86)"
 775 .\" commit fe3d197f84319d3bce379a9c0dc17b1f48ad358c
 776 .\" See also http://lwn.net/Articles/582712/
 777 .\" See also https://gcc.gnu.org/wiki/Intel%20MPX%20support%20in%20the%20GCC%20compiler
 778 Enable or disable kernel management of Memory Protection eXtensions (MPX)
 779 bounds tables.
 780 The
 781 .IR arg2 ,
 782 .IR arg3 ,
 783 .IR arg4 ,
 784 and
 785 .IR arg5
 786 .\" commit e9d1b4f3c60997fe197bf0243cb4a41a44387a88
 787 arguments must be zero.
 788 .IP
 789 MPX is a hardware-assisted mechanism for performing bounds checking on
 790 pointers.
 791 It consists of a set of registers storing bounds information
 792 and a set of special instruction prefixes that tell the CPU on which
 793 instructions it should do bounds enforcement.
 794 There is a limited number of these registers and
 795 when there are more pointers than registers,
 796 their contents must be "spilled" into a set of tables.
 797 These tables are called "bounds tables" and the MPX
 798 .BR prctl ()
 799 operations control
 800 whether the kernel manages their allocation and freeing.
 801 .IP
 802 When management is enabled, the kernel will take over allocation
 803 and freeing of the bounds tables.
 804 It does this by trapping the #BR exceptions that result
 805 at first use of missing bounds tables and
 806 instead of delivering the exception to user space,
 807 it allocates the table and populates the bounds directory
 808 with the location of the new table.
 809 For freeing, the kernel checks to see if bounds tables are
 810 present for memory which is not allocated, and frees them if so.
 811 .IP
 812 Before enabling MPX management using
 813 .BR PR_MPX_ENABLE_MANAGEMENT ,
 814 the application must first have allocated a user-space buffer for
 815 the bounds directory and placed the location of that directory in the
 816 .I bndcfgu
 817 register.
 818 .IP
 819 These calls fail if the CPU or kernel does not support MPX.
 820 Kernel support for MPX is enabled via the
 821 .BR CONFIG_X86_INTEL_MPX
 822 configuration option.
 823 You can check whether the CPU supports MPX by looking for the
 824 .I mpx
 825 CPUID bit, like with the following command:
 826 .IP
 827 .in +4n
 828 .EX
 829 cat /proc/cpuinfo | grep \(aq mpx \(aq
 830 .EE
 831 .in
 832 .IP
 833 A thread may not switch in or out of long (64-bit) mode while MPX is
 834 enabled.
 835 .IP
 836 All threads in a process are affected by these calls.
 837 .IP
 838 The child of a
 839 .BR fork (2)
 840 inherits the state of MPX management.
 841 During
 842 .BR execve (2),
 843 MPX management is reset to a state as if
 844 .BR PR_MPX_DISABLE_MANAGEMENT
 845 had been called.
 846 .IP
 847 For further information on Intel MPX, see the kernel source file
 848 .IR Documentation/x86/intel_mpx.txt .
 849 .IP
 850 .\" commit f240652b6032b48ad7fa35c5e701cc4c8d697c0b
 851 .\" See also https://lkml.kernel.org/r/20190705175321.DB42F0AD@viggo.jf.intel.com
 852 Due to a lack of toolchain support,
 853 .BR PR_MPX_ENABLE_MANAGEMENT " and " PR_MPX_DISABLE_MANAGEMENT
 854 are not supported in Linux 5.4 and later.
 855 .\" prctl PR_SET_NAME
 856 .TP
 857 .BR PR_SET_NAME " (since Linux 2.6.9)"
 858 Set the name of the calling thread,
 859 using the value in the location pointed to by
 860 .IR "(char\ *) arg2" .
 861 The name can be up to 16 bytes long,
 862 .\" TASK_COMM_LEN in include/linux/sched.h
 863 including the terminating null byte.
 864 (If the length of the string, including the terminating null byte,
 865 exceeds 16 bytes, the string is silently truncated.)
 866 This is the same attribute that can be set via
 867 .BR pthread_setname_np (3)
 868 and retrieved using
 869 .BR pthread_getname_np (3).
 870 The attribute is likewise accessible via
 871 .IR /proc/self/task/[tid]/comm
 872 (see
 873 .BR proc (5)),
 874 where
 875 .I [tid]
 876 is the thread ID of the calling thread, as returned by
 877 .BR gettid (2).
 878 .\" prctl PR_GET_NAME
 879 .TP
 880 .BR PR_GET_NAME " (since Linux 2.6.11)"
 881 Return the name of the calling thread,
 882 in the buffer pointed to by
 883 .IR "(char\ *) arg2" .
 884 The buffer should allow space for up to 16 bytes;
 885 the returned string will be null-terminated.
 886 .\" prctl PR_SET_NO_NEW_PRIVS
 887 .TP
 888 .BR PR_SET_NO_NEW_PRIVS " (since Linux 3.5)"
 889 Set the calling thread's
 890 .I no_new_privs
 891 attribute to the value in
 892 .IR arg2 .
 893 With
 894 .I no_new_privs
 895 set to 1,
 896 .BR execve (2)
 897 promises not to grant privileges to do anything
 898 that could not have been done without the
 899 .BR execve (2)
 900 call (for example,
 901 rendering the set-user-ID and set-group-ID mode bits,
 902 and file capabilities non-functional).
 903 Once set, the
 904 .I no_new_privs
 905 attribute cannot be unset.
 906 The setting of this attribute is inherited by children created by
 907 .BR fork (2)
 908 and
 909 .BR clone (2),
 910 and preserved across
 911 .BR execve (2).
 912 .IP
 913 Since Linux 4.10,
 914 the value of a thread's
 915 .I no_new_privs
 916 attribute can be viewed via the
 917 .I NoNewPrivs
 918 field in the
 919 .IR /proc/[pid]/status
 920 file.
 921 .IP
 922 For more information, see the kernel source file
 923 .IR Documentation/userspace\-api/no_new_privs.rst
 924 .\" commit 40fde647ccb0ae8c11d256d271e24d385eed595b
 925 (or
 926 .IR Documentation/prctl/no_new_privs.txt
 927 before Linux 4.13).
 928 See also
 929 .BR seccomp (2).
 930 .\" prctl PR_GET_NO_NEW_PRIVS
 931 .TP
 932 .BR PR_GET_NO_NEW_PRIVS " (since Linux 3.5)"
 933 Return (as the function result) the value of the
 934 .I no_new_privs
 935 attribute for the calling thread.
 936 A value of 0 indicates the regular
 937 .BR execve (2)
 938 behavior.
 939 A value of 1 indicates
 940 .BR execve (2)
 941 will operate in the privilege-restricting mode described above.
 942 .\" prctl PR_PAC_RESET_KEYS
 943 .\" commit ba830885656414101b2f8ca88786524d4bb5e8c1
 944 .TP
 945 .BR PR_PAC_RESET_KEYS " (since Linux 5.0, only on arm64)"
 946 Securely reset the thread's pointer authentication keys
 947 to fresh random values generated by the kernel.
 948 .IP
 949 The set of keys to be reset is specified by
 950 .IR arg2 ,
 951 which must be a logical OR of zero or more of the following:
 952 .RS
 953 .TP
 954 .B PR_PAC_APIAKEY
 955 instruction authentication key A
 956 .TP
 957 .B PR_PAC_APIBKEY
 958 instruction authentication key B
 959 .TP
 960 .B PR_PAC_APDAKEY
 961 data authentication key A
 962 .TP
 963 .B PR_PAC_APDBKEY
 964 data authentication key B
 965 .TP
 966 .B PR_PAC_APGAKEY
 967 generic authentication \(lqA\(rq key.
 968 .IP
 969 (Yes folks, there really is no generic B key.)
 970 .RE
 971 .IP
 972 As a special case, if
 973 .I arg2
 974 is zero, then all the keys are reset.
 975 Since new keys could be added in future,
 976 this is the recommended way to completely wipe the existing keys
 977 when establishing a clean execution context.
 978 Note that there is no need to use
 979 .BR PR_PAC_RESET_KEYS
 980 in preparation for calling
 981 .BR execve (2),
 982 since
 983 .BR execve (2)
 984 resets all the pointer authentication keys.
 985 .IP
 986 The remaining arguments
 987 .IR arg3 ", " arg4 ", and " arg5
 988 must all be zero.
 989 .IP
 990 If the arguments are invalid,
 991 and in particular if
 992 .I arg2
 993 contains set bits that are unrecognized
 994 or that correspond to a key not available on this platform,
 995 then the call fails with error
 996 .BR EINVAL .
 997 .IP
 998 .B Warning:
 999 Because the compiler or run-time environment
1000 may be using some or all of the keys,
1001 a successful
1002 .B PR_PAC_RESET_KEYS
1003 may crash the calling process.
1004 The conditions for using it safely are complex and system-dependent.
1005 Don't use it unless you know what you are doing.
1006 .IP
1007 For more information, see the kernel source file
1008 .I Documentation/arm64/pointer\-authentication.rst
1009 .\"commit b693d0b372afb39432e1c49ad7b3454855bc6bed
1010 (or
1011 .I Documentation/arm64/pointer\-authentication.txt
1012 before Linux 5.3).
1013 .\" prctl PR_SET_PDEATHSIG
1014 .TP
1015 .BR PR_SET_PDEATHSIG " (since Linux 2.1.57)"
1016 Set the parent-death signal
1017 of the calling process to \fIarg2\fP (either a signal value
1018 in the range 1..\c
1019 .BR NSIG "\-1" ,
1020 or 0 to clear).
1021 This is the signal that the calling process will get when its
1022 parent dies.
1023 .IP
1024 .IR Warning :
1025 .\" https://bugzilla.kernel.org/show_bug.cgi?id=43300
1026 the "parent" in this case is considered to be the
1027 .I thread
1028 that created this process.
1029 In other words, the signal will be sent when that thread terminates
1030 (via, for example,
1031 .BR pthread_exit (3)),
1032 rather than after all of the threads in the parent process terminate.
1033 .IP
1034 The parent-death signal is sent upon subsequent termination of the parent
1035 thread and also upon termination of each subreaper process
1036 (see the description of
1037 .B PR_SET_CHILD_SUBREAPER
1038 above) to which the caller is subsequently reparented.
1039 If the parent thread and all ancestor subreapers have already terminated
1040 by the time of the
1041 .BR PR_SET_PDEATHSIG
1042 operation, then no parent-death signal is sent to the caller.
1043 .IP
1044 The parent-death signal is process-directed (see
1045 .BR signal (7))
1046 and, if the child installs a handler using the
1047 .BR sigaction (2)
1048 .B SA_SIGINFO
1049 flag, the
1050 .I si_pid
1051 field of the
1052 .I siginfo_t
1053 argument of the handler contains the PID of the terminating parent process.
1054 .IP
1055 The parent-death signal setting is cleared for the child of a
1056 .BR fork (2).
1057 It is also
1058 (since Linux 2.4.36 / 2.6.23)
1059 .\" commit d2d56c5f51028cb9f3d800882eb6f4cbd3f9099f
1060 cleared when executing a set-user-ID or set-group-ID binary,
1061 or a binary that has associated capabilities (see
1062 .BR capabilities (7));
1063 otherwise, this value is preserved across
1064 .BR execve (2).
1065 The parent-death signal setting is also cleared upon changes to
1066 any of the following thread credentials:
1067 .\" FIXME capability changes can also trigger this; see
1068 .\" kernel/cred.c::commit_creds in the Linux 5.6 source.
1069 effective user ID, effective group ID, filesystem user ID,
1070 or filesystem group ID.
1071 .\" prctl PR_GET_PDEATHSIG
1072 .TP
1073 .BR PR_GET_PDEATHSIG " (since Linux 2.3.15)"
1074 Return the current value of the parent process death signal,
1075 in the location pointed to by
1076 .IR "(int\ *) arg2" .
1077 .\" prctl PR_SET_PTRACER
1078 .TP
1079 .BR PR_SET_PTRACER " (since Linux 3.4)"
1080 .\" commit 2d514487faf188938a4ee4fb3464eeecfbdcf8eb
1081 .\" commit bf06189e4d14641c0148bea16e9dd24943862215
1082 This is meaningful only when the Yama LSM is enabled and in mode 1
1083 ("restricted ptrace", visible via
1084 .IR /proc/sys/kernel/yama/ptrace_scope ).
1085 When a "ptracer process ID" is passed in \fIarg2\fP,
1086 the caller is declaring that the ptracer process can
1087 .BR ptrace (2)
1088 the calling process as if it were a direct process ancestor.
1089 Each
1090 .B PR_SET_PTRACER
1091 operation replaces the previous "ptracer process ID".
1092 Employing
1093 .B PR_SET_PTRACER
1094 with
1095 .I arg2
1096 set to 0 clears the caller's "ptracer process ID".
1097 If
1098 .I arg2
1099 is
1100 .BR PR_SET_PTRACER_ANY ,
1101 the ptrace restrictions introduced by Yama are effectively disabled for the
1102 calling process.
1103 .IP
1104 For further information, see the kernel source file
1105 .IR Documentation/admin\-guide/LSM/Yama.rst
1106 .\" commit 90bb766440f2147486a2acc3e793d7b8348b0c22
1107 (or
1108 .IR Documentation/security/Yama.txt
1109 before Linux 4.13).
1110 .\" prctl PR_SET_SECCOMP
1111 .TP
1112 .BR PR_SET_SECCOMP " (since Linux 2.6.23)"
1113 .\" See http://thread.gmane.org/gmane.linux.kernel/542632
1114 .\" [PATCH 0 of 2] seccomp updates
1115 .\" andrea@cpushare.com
1116 Set the secure computing (seccomp) mode for the calling thread, to limit
1117 the available system calls.
1118 The more recent
1119 .BR seccomp (2)
1120 system call provides a superset of the functionality of
1121 .BR PR_SET_SECCOMP ,
1122 and is the preferred interface for new applications.
1123 .IP
1124 The seccomp mode is selected via
1125 .IR arg2 .
1126 (The seccomp constants are defined in
1127 .IR <linux/seccomp.h> .)
1128 The following values can be specified:
1129 .RS
1130 .TP
1131 .BR SECCOMP_MODE_STRICT " (since Linux 2.6.23)"
1132 See the description of
1133 .B SECCOMP_SET_MODE_STRICT
1134 in
1135 .BR seccomp (2).
1136 .IP
1137 This operation is available only
1138 if the kernel is configured with
1139 .B CONFIG_SECCOMP
1140 enabled.
1141 .TP
1142 .BR SECCOMP_MODE_FILTER " (since Linux 3.5)"
1143 The allowed system calls are defined by a pointer
1144 to a Berkeley Packet Filter passed in
1145 .IR arg3 .
1146 This argument is a pointer to
1147 .IR "struct sock_fprog" ;
1148 it can be designed to filter
1149 arbitrary system calls and system call arguments.
1150 See the description of
1151 .B SECCOMP_SET_MODE_FILTER
1152 in
1153 .BR seccomp (2).
1154 .IP
1155 This operation is available only
1156 if the kernel is configured with
1157 .B CONFIG_SECCOMP_FILTER
1158 enabled.
1159 .RE
1160 .IP
1161 For further details on seccomp filtering, see
1162 .BR seccomp (2).
1163 .\" prctl PR_GET_SECCOMP
1164 .TP
1165 .BR PR_GET_SECCOMP " (since Linux 2.6.23)"
1166 Return (as the function result)
1167 the secure computing mode of the calling thread.
1168 If the caller is not in secure computing mode, this operation returns 0;
1169 if the caller is in strict secure computing mode, then the
1170 .BR prctl ()
1171 call will cause a
1172 .B SIGKILL
1173 signal to be sent to the process.
1174 If the caller is in filter mode, and this system call is allowed by the
1175 seccomp filters, it returns 2; otherwise, the process is killed with a
1176 .BR SIGKILL
1177 signal.
1178 .IP
1179 This operation is available only
1180 if the kernel is configured with
1181 .B CONFIG_SECCOMP
1182 enabled.
1183 .IP
1184 Since Linux 3.8, the
1185 .IR Seccomp
1186 field of the
1187 .IR /proc/[pid]/status
1188 file provides a method of obtaining the same information,
1189 without the risk that the process is killed; see
1190 .BR proc (5).
1191 .\" prctl PR_SET_SECUREBITS
1192 .TP
1193 .BR PR_SET_SECUREBITS " (since Linux 2.6.26)"
1194 Set the "securebits" flags of the calling thread to the value supplied in
1195 .IR arg2 .
1196 See
1197 .BR capabilities (7).
1198 .\" prctl PR_GET_SECUREBITS
1199 .TP
1200 .BR PR_GET_SECUREBITS " (since Linux 2.6.26)"
1201 Return (as the function result)
1202 the "securebits" flags of the calling thread.
1203 See
1204 .BR capabilities (7).
1205 .\" prctl PR_GET_SPECULATION_CTRL
1206 .TP
1207 .BR PR_GET_SPECULATION_CTRL " (since Linux 4.17)"
1208 Return (as the function result)
1209 the state of the speculation misfeature specified in
1210 .IR arg2 .
1211 Currently, the only permitted value for this argument is
1212 .BR PR_SPEC_STORE_BYPASS
1213 (otherwise the call fails with the error
1214 .BR ENODEV ).
1215 .IP
1216 The return value uses bits 0-3 with the following meaning:
1217 .RS
1218 .TP
1219 .BR PR_SPEC_PRCTL
1220 Mitigation can be controlled per thread by
1221 .BR PR_SET_SPECULATION_CTRL .
1222 .TP
1223 .BR PR_SPEC_ENABLE
1224 The speculation feature is enabled, mitigation is disabled.
1225 .TP
1226 .BR PR_SPEC_DISABLE
1227 The speculation feature is disabled, mitigation is enabled.
1228 .TP
1229 .BR PR_SPEC_FORCE_DISABLE
1230 Same as
1231 .B PR_SPEC_DISABLE
1232 but cannot be undone.
1233 .TP
1234 .BR PR_SPEC_DISABLE_NOEXEC " (since Linux 5.1)"
1235 Same as
1236 .BR PR_SPEC_DISABLE ,
1237 but the state will be cleared on
1238 .BR execve (2).
1239 .RE
1240 .IP
1241 If all bits are 0,
1242 then the CPU is not affected by the speculation misfeature.
1243 .IP
1244 If
1245 .B PR_SPEC_PRCTL
1246 is set, then per-thread control of the mitigation is available.
1247 If not set,
1248 .BR prctl ()
1249 for the speculation misfeature will fail.
1250 .IP
1251 The
1252 .IR arg3 ,
1253 .IR arg4 ,
1254 and
1255 .I arg5
1256 arguments must be specified as 0; otherwise the call fails with the error
1257 .BR EINVAL .
1258 .\" prctl PR_SET_SPECULATION_CTRL
1259 .TP
1260 .BR PR_SET_SPECULATION_CTRL " (since Linux 4.17)"
1261 .\" commit b617cfc858161140d69cc0b5cc211996b557a1c7
1262 .\" commit 356e4bfff2c5489e016fdb925adbf12a1e3950ee
1263 Sets the state of the speculation misfeature specified in
1264 .IR arg2 .
1265 The speculation-misfeature settings are per-thread attributes.
1266 .IP
1267 Currently,
1268 .I arg2
1269 must be one of:
1270 .RS
1271 .TP
1272 .B PR_SPEC_STORE_BYPASS
1273 Set the state of the speculative store bypass misfeature.
1274 .\" commit 9137bb27e60e554dab694eafa4cca241fa3a694f
1275 .TP
1276 .BR PR_SPEC_INDIRECT_BRANCH " (since Linux 4.20)"
1277 Set the state of the indirect branch speculation misfeature.
1278 .RE
1279 .IP
1280 If
1281 .I arg2
1282 does not have one of the above values,
1283 then the call fails with the error
1284 .BR ENODEV .
1285 .IP
1286 The
1287 .IR arg3
1288 argument is used to hand in the control value,
1289 which is one of the following:
1290 .RS
1291 .TP
1292 .BR PR_SPEC_ENABLE
1293 The speculation feature is enabled, mitigation is disabled.
1294 .TP
1295 .BR PR_SPEC_DISABLE
1296 The speculation feature is disabled, mitigation is enabled.
1297 .TP
1298 .BR PR_SPEC_FORCE_DISABLE
1299 Same as
1300 .BR PR_SPEC_DISABLE ,
1301 but cannot be undone.
1302 A subsequent
1303 .BR prctl (\c
1304 .IR arg2 ,
1305 .BR PR_SPEC_ENABLE )
1306 with the same value for
1307 .I arg2
1308 will fail with the error
1309 .BR EPERM .
1310 .\" commit 71368af9027f18fe5d1c6f372cfdff7e4bde8b48
1311 .TP
1312 .BR PR_SPEC_DISABLE_NOEXEC " (since Linux 5.1)"
1313 Same as
1314 .BR PR_SPEC_DISABLE ,
1315 but the state will be cleared on
1316 .BR execve (2).
1317 Currently only supported for
1318 .I arg2
1319 equal to
1320 .B PR_SPEC_STORE_BYPASS.
1321 .RE
1322 .IP
1323 Any unsupported value in
1324 .IR arg3
1325 will result in the call failing with the error
1326 .BR ERANGE .
1327 .IP
1328 The
1329 .I arg4
1330 and
1331 .I arg5
1332 arguments must be specified as 0; otherwise the call fails with the error
1333 .BR EINVAL .
1334 .IP
1335 The speculation feature can also be controlled by the
1336 .B spec_store_bypass_disable
1337 boot parameter.
1338 This parameter may enforce a read-only policy which will result in the
1339 .BR prctl ()
1340 call failing with the error
1341 .BR ENXIO .
1342 For further details, see the kernel source file
1343 .IR Documentation/admin\-guide/kernel\-parameters.txt .
1344 .\" prctl PR_SVE_SET_VL
1345 .\" commit 2d2123bc7c7f843aa9db87720de159a049839862
1346 .\" linux-5.6/Documentation/arm64/sve.rst
1347 .TP
1348 .BR PR_SVE_SET_VL " (since Linux 4.15, only on arm64)"
1349 Configure the thread's SVE vector length,
1350 as specified by
1351 .IR "(int) arg2" .
1352 Arguments
1353 .IR arg3 ", " arg4 ", and " arg5
1354 are ignored.
1355 .IP
1356 The bits of
1357 .I arg2
1358 corresponding to
1359 .B PR_SVE_VL_LEN_MASK
1360 must be set to the desired vector length in bytes.
1361 This is interpreted as an upper bound:
1362 the kernel will select the greatest available vector length
1363 that does not exceed the value specified.
1364 In particular, specifying
1365 .B SVE_VL_MAX
1366 (defined in
1367 .I <asm/sigcontext.h>)
1368 for the
1369 .B PR_SVE_VL_LEN_MASK
1370 bits requests the maximum supported vector length.
1371 .IP
1372 In addition, the other bits of
1373 .I arg2
1374 must be set to one of the following combinations of flags:
1375 .RS
1376 .TP
1377 .B 0
1378 Perform the change immediately.
1379 At the next
1380 .BR execve (2)
1381 in the thread,
1382 the vector length will be reset to the value configured in
1383 .IR /proc/sys/abi/sve_default_vector_length .
1384 .TP
1385 .B PR_SVE_VL_INHERIT
1386 Perform the change immediately.
1387 Subsequent
1388 .BR execve (2)
1389 calls will preserve the new vector length.
1390 .TP
1391 .B PR_SVE_SET_VL_ONEXEC
1392 Defer the change, so that it is performed at the next
1393 .BR execve (2)
1394 in the thread.
1395 Further
1396 .BR execve (2)
1397 calls will reset the vector length to the value configured in
1398 .IR /proc/sys/abi/sve_default_vector_length .
1399 .TP
1400 .B "PR_SVE_SET_VL_ONEXEC | PR_SVE_VL_INHERIT"
1401 Defer the change, so that it is performed at the next
1402 .BR execve (2)
1403 in the thread.
1404 Further
1405 .BR execve (2)
1406 calls will preserve the new vector length.
1407 .RE
1408 .IP
1409 In all cases,
1410 any previously pending deferred change is canceled.
1411 .IP
1412 The call fails with error
1413 .B EINVAL
1414 if SVE is not supported on the platform, if
1415 .I arg2
1416 is unrecognized or invalid, or the value in the bits of
1417 .I arg2
1418 corresponding to
1419 .B PR_SVE_VL_LEN_MASK
1420 is outside the range
1421 .BR SVE_VL_MIN .. SVE_VL_MAX
1422 or is not a multiple of 16.
1423 .IP
1424 On success,
1425 a nonnegative value is returned that describes the
1426 .I selected
1427 configuration.
1428 If
1429 .B PR_SVE_SET_VL_ONEXEC
1430 was included in
1431 .IR arg2 ,
1432 then the configuration described by the return value
1433 will take effect at the next
1434 .BR execve (2).
1435 Otherwise, the configuration is already in effect when the
1436 .B PR_SVE_SET_VL
1437 call returns.
1438 In either case, the value is encoded in the same way as the return value of
1439 .BR PR_SVE_GET_VL .
1440 Note that there is no explicit flag in the return value
1441 corresponding to
1442 .BR PR_SVE_SET_VL_ONEXEC .
1443 .IP
1444 The configuration (including any pending deferred change)
1445 is inherited across
1446 .BR fork (2)
1447 and
1448 .BR clone (2).
1449 .IP
1450 For more information, see the kernel source file
1451 .I Documentation/arm64/sve.rst
1452 .\"commit b693d0b372afb39432e1c49ad7b3454855bc6bed
1453 (or
1454 .I Documentation/arm64/sve.txt
1455 before Linux 5.3).
1456 .IP
1457 .B Warning:
1458 Because the compiler or run-time environment
1459 may be using SVE, using this call without the
1460 .B PR_SVE_SET_VL_ONEXEC
1461 flag may crash the calling process.
1462 The conditions for using it safely are complex and system-dependent.
1463 Don't use it unless you really know what you are doing.
1464 .\" prctl PR_SVE_GET_VL
1465 .TP
1466 .BR PR_SVE_GET_VL " (since Linux 4.15, only on arm64)"
1467 Get the thread's current SVE vector length configuration.
1468 .IP
1469 Arguments
1470 .IR arg2 ", " arg3 ", " arg4 ", and " arg5
1471 are ignored.
1472 .IP
1473 Provided that the kernel and platform support SVE,
1474 this operation always succeeds,
1475 returning a nonnegative value that describes the
1476 .I current
1477 configuration.
1478 The bits corresponding to
1479 .B PR_SVE_VL_LEN_MASK
1480 contain the currently configured vector length in bytes.
1481 The bit corresponding to
1482 .B PR_SVE_VL_INHERIT
1483 indicates whether the vector length will be inherited
1484 across
1485 .BR execve (2).
1486 .IP
1487 Note that there is no way to determine whether there is
1488 a pending vector length change that has not yet taken effect.
1489 .IP
1490 For more information, see the kernel source file
1491 .I Documentation/arm64/sve.rst
1492 .\"commit b693d0b372afb39432e1c49ad7b3454855bc6bed
1493 (or
1494 .I Documentation/arm64/sve.txt
1495 before Linux 5.3).
1496 .TP
1497 .\" prctl PR_SET_SYSCALL_USER_DISPATCH
1498 .\" commit 1446e1df9eb183fdf81c3f0715402f1d7595d4
1499 .BR PR_SET_SYSCALL_USER_DISPATCH " (since Linux 5.11, x86 only)"
1500 Configure the Syscall User Dispatch mechanism
1501 for the calling thread.
1502 This mechanism allows an application
1503 to selectively intercept system calls
1504 so that they can be handled within the application itself.
1505 Interception takes the form of a thread-directed
1506 .B SIGSYS
1507 signal that is delivered to the thread
1508 when it makes a system call.
1509 If intercepted,
1510 the system call is not executed by the kernel.
1511 .IP
1512 To enable this mechanism,
1513 .I arg2
1514 should be set to
1515 .BR PR_SYS_DISPATCH_ON .
1516 Once enabled, further system calls will be selectively intercepted,
1517 depending on a control variable provided by user space.
1518 In this case,
1519 .I arg3
1520 and
1521 .I arg4
1522 respectively identify the
1523 .I offset
1524 and
1525 .I length
1526 of a single contiguous memory region in the process address space
1527 from where system calls are always allowed to be executed,
1528 regardless of the control variable.
1529 (Typically, this area would include the area of memory
1530 containing the C library.)
1531 .IP
1532 .I arg5
1533 points to a char-sized variable
1534 that is a fast switch to allow/block system call execution
1535 without the overhead of doing another system call
1536 to re-configure Syscall User Dispatch.
1537 This control variable can either be set to
1538 .B SYSCALL_DISPATCH_FILTER_BLOCK
1539 to block system calls from executing
1540 or to
1541 .B SYSCALL_DISPATCH_FILTER_ALLOW
1542 to temporarily allow them to be executed.
1543 This value is checked by the kernel
1544 on every system call entry,
1545 and any unexpected value will raise
1546 an uncatchable
1547 .B SIGSYS
1548 at that time,
1549 killing the application.
1550 .IP
1551 When a system call is intercepted,
1552 the kernel sends a thread-directed
1553 .B SIGSYS
1554 signal to the triggering thread.
1555 Various fields will be set in the
1556 .I siginfo_t
1557 structure (see
1558 .BR sigaction (2))
1559 associated with the signal:
1560 .RS
1561 .IP * 3
1562 .I si_signo
1563 will contain
1564 .BR SIGSYS .
1565 .IP *
1566 .IR si_call_addr
1567 will show the address of the system call instruction.
1568 .IP *
1569 .IR si_syscall
1570 and
1571 .IR si_arch
1572 will indicate which system call was attempted.
1573 .IP *
1574 .I si_code
1575 will contain
1576 .BR SYS_USER_DISPATCH .
1577 .IP *
1578 .I si_errno
1579 will be set to 0.
1580 .RE
1581 .IP
1582 The program counter will be as though the system call happened
1583 (i.e., the program counter will not point to the system call instruction).
1584 .IP
1585 When the signal handler returns to the kernel,
1586 the system call completes immediately
1587 and returns to the calling thread,
1588 without actually being executed.
1589 If necessary
1590 (i.e., when emulating the system call on user space.),
1591 the signal handler should set the system call return value
1592 to a sane value,
1593 by modifying the register context stored in the
1594 .I ucontext
1595 argument of the signal handler.
1596 See
1597 .BR sigaction (2),
1598 .BR sigreturn (2),
1599 and
1600 .BR getcontext (3)
1601 for more information.
1602 .IP
1603 If
1604 .I arg2
1605 is set to
1606 .BR PR_SYS_DISPATCH_OFF ,
1607 Syscall User Dispatch is disabled for that thread.
1608 the remaining arguments must be set to 0.
1609 .IP
1610 The setting is not preserved across
1611 .BR fork (2),
1612 .BR clone (2),
1613 or
1614 .BR execve (2).
1615 .IP
1616 For more information,
1617 see the kernel source file
1618 .IR Documentation/admin-guide/syscall-user-dispatch.rst
1619 .\" prctl PR_SET_TAGGED_ADDR_CTRL
1620 .\" commit 63f0c60379650d82250f22e4cf4137ef3dc4f43d
1621 .TP
1622 .BR PR_SET_TAGGED_ADDR_CTRL " (since Linux 5.4, only on arm64)"
1623 Controls support for passing tagged user-space addresses to the kernel
1624 (i.e., addresses where bits 56\(em63 are not all zero).
1625 .IP
1626 The level of support is selected by
1627 .IR "arg2" ,
1628 which can be one of the following:
1629 .RS
1630 .TP
1631 .B 0
1632 Addresses that are passed
1633 for the purpose of being dereferenced by the kernel
1634 must be untagged.
1635 .TP
1636 .B PR_TAGGED_ADDR_ENABLE
1637 Addresses that are passed
1638 for the purpose of being dereferenced by the kernel
1639 may be tagged, with the exceptions summarized below.
1640 .RE
1641 .IP
1642 The remaining arguments
1643 .IR arg3 ", " arg4 ", and " arg5
1644 must all be zero.
1645 .\" Enforcement added in
1646 .\" commit 3e91ec89f527b9870fe42dcbdb74fd389d123a95
1647 .IP
1648 On success, the mode specified in
1649 .I arg2
1650 is set for the calling thread and the return value is 0.
1651 If the arguments are invalid,
1652 the mode specified in
1653 .I arg2
1654 is unrecognized,
1655 or if this feature is unsupported by the kernel
1656 or disabled via
1657 .IR /proc/sys/abi/tagged_addr_disabled ,
1658 the call fails with the error
1659 .BR EINVAL .
1660 .IP
1661 In particular, if
1662 .BR prctl ( PR_SET_TAGGED_ADDR_CTRL ,
1663 0, 0, 0, 0)
1664 fails with
1665 .BR EINVAL ,
1666 then all addresses passed to the kernel must be untagged.
1667 .IP
1668 Irrespective of which mode is set,
1669 addresses passed to certain interfaces
1670 must always be untagged:
1671 .RS
1672 .IP \(bu 2
1673 .BR brk (2),
1674 .BR mmap (2),
1675 .BR shmat (2),
1676 .BR shmdt (2),
1677 and the
1678 .I new_address
1679 argument of
1680 .BR mremap (2).
1681 .IP
1682 (Prior to Linux 5.6 these accepted tagged addresses,
1683 but the behaviour may not be what you expect.
1684 Don't rely on it.)
1685 .IP \(bu
1686 \(oqpolymorphic\(cq interfaces
1687 that accept pointers to arbitrary types cast to a
1688 .I void *
1689 or other generic type, specifically
1690 .BR prctl (),
1691 .BR ioctl (2),
1692 and in general
1693 .BR setsockopt (2)
1694 (only certain specific
1695 .BR setsockopt (2)
1696 options allow tagged addresses).
1697 .RE
1698 .IP
1699 This list of exclusions may shrink
1700 when moving from one kernel version to a later kernel version.
1701 While the kernel may make some guarantees
1702 for backwards compatibility reasons,
1703 for the purposes of new software
1704 the effect of passing tagged addresses to these interfaces
1705 is unspecified.
1706 .IP
1707 The mode set by this call is inherited across
1708 .BR fork (2)
1709 and
1710 .BR clone (2).
1711 The mode is reset by
1712 .BR execve (2)
1713 to 0
1714 (i.e., tagged addresses not permitted in the user/kernel ABI).
1715 .IP
1716 For more information, see the kernel source file
1717 .IR Documentation/arm64/tagged\-address\-abi.rst .
1718 .IP
1719 .B Warning:
1720 This call is primarily intended for use by the run-time environment.
1721 A successful
1722 .B PR_SET_TAGGED_ADDR_CTRL
1723 call elsewhere may crash the calling process.
1724 The conditions for using it safely are complex and system-dependent.
1725 Don't use it unless you know what you are doing.
1726 .\" prctl PR_GET_TAGGED_ADDR_CTRL
1727 .\" commit 63f0c60379650d82250f22e4cf4137ef3dc4f43d
1728 .TP
1729 .BR PR_GET_TAGGED_ADDR_CTRL " (since Linux 5.4, only on arm64)"
1730 Returns the current tagged address mode
1731 for the calling thread.
1732 .IP
1733 Arguments
1734 .IR arg2 ", " arg3 ", " arg4 ", and " arg5
1735 must all be zero.
1736 .IP
1737 If the arguments are invalid
1738 or this feature is disabled or unsupported by the kernel,
1739 the call fails with
1740 .BR EINVAL .
1741 In particular, if
1742 .BR prctl ( PR_GET_TAGGED_ADDR_CTRL ,
1743 0, 0, 0, 0)
1744 fails with
1745 .BR EINVAL ,
1746 then this feature is definitely either unsupported,
1747 or disabled via
1748 .IR /proc/sys/abi/tagged_addr_disabled .
1749 In this case,
1750 all addresses passed to the kernel must be untagged.
1751 .IP
1752 Otherwise, the call returns a nonnegative value
1753 describing the current tagged address mode,
1754 encoded in the same way as the
1755 .I arg2
1756 argument of
1757 .BR PR_SET_TAGGED_ADDR_CTRL .
1758 .IP
1759 For more information, see the kernel source file
1760 .IR Documentation/arm64/tagged\-address\-abi.rst .
1761 .\"
1762 .\" prctl PR_TASK_PERF_EVENTS_DISABLE
1763 .TP
1764 .BR PR_TASK_PERF_EVENTS_DISABLE " (since Linux 2.6.31)"
1765 Disable all performance counters attached to the calling process,
1766 regardless of whether the counters were created by
1767 this process or another process.
1768 Performance counters created by the calling process for other
1769 processes are unaffected.
1770 For more information on performance counters, see the Linux kernel source file
1771 .IR tools/perf/design.txt .
1772 .IP
1773 Originally called
1774 .BR PR_TASK_PERF_COUNTERS_DISABLE ;
1775 .\" commit 1d1c7ddbfab358445a542715551301b7fc363e28
1776 renamed (retaining the same numerical value)
1777 in Linux 2.6.32.
1778 .\"
1779 .\" prctl PR_TASK_PERF_EVENTS_ENABLE
1780 .TP
1781 .BR PR_TASK_PERF_EVENTS_ENABLE " (since Linux 2.6.31)"
1782 The converse of
1783 .BR PR_TASK_PERF_EVENTS_DISABLE ;
1784 enable performance counters attached to the calling process.
1785 .IP
1786 Originally called
1787 .BR PR_TASK_PERF_COUNTERS_ENABLE ;
1788 .\" commit 1d1c7ddbfab358445a542715551301b7fc363e28
1789 renamed
1790 .\" commit cdd6c482c9ff9c55475ee7392ec8f672eddb7be6
1791 in Linux 2.6.32.
1792 .\"
1793 .\" prctl PR_SET_THP_DISABLE
1794 .TP
1795 .BR PR_SET_THP_DISABLE " (since Linux 3.15)"
1796 .\" commit a0715cc22601e8830ace98366c0c2bd8da52af52
1797 Set the state of the "THP disable" flag for the calling thread.
1798 If
1799 .I arg2
1800 has a nonzero value, the flag is set, otherwise it is cleared.
1801 Setting this flag provides a method
1802 for disabling transparent huge pages
1803 for jobs where the code cannot be modified, and using a malloc hook with
1804 .BR madvise (2)
1805 is not an option (i.e., statically allocated data).
1806 The setting of the "THP disable" flag is inherited by a child created via
1807 .BR fork (2)
1808 and is preserved across
1809 .BR execve (2).
1810 .\" prctl PR_GET_THP_DISABLE
1811 .TP
1812 .BR PR_GET_THP_DISABLE " (since Linux 3.15)"
1813 Return (as the function result) the current setting of the "THP disable"
1814 flag for the calling thread:
1815 either 1, if the flag is set, or 0, if it is not.
1816 .\" prctl PR_GET_TID_ADDRESS
1817 .TP
1818 .BR PR_GET_TID_ADDRESS " (since Linux 3.5)"
1819 .\" commit 300f786b2683f8bb1ec0afb6e1851183a479c86d
1820 Return the
1821 .I clear_child_tid
1822 address set by
1823 .BR set_tid_address (2)
1824 and the
1825 .BR clone (2)
1826 .B CLONE_CHILD_CLEARTID
1827 flag, in the location pointed to by
1828 .IR "(int\ **)\ arg2" .
1829 This feature is available only if the kernel is built with the
1830 .BR CONFIG_CHECKPOINT_RESTORE
1831 option enabled.
1832 Note that since the
1833 .BR prctl ()
1834 system call does not have a compat implementation for
1835 the AMD64 x32 and MIPS n32 ABIs,
1836 and the kernel writes out a pointer using the kernel's pointer size,
1837 this operation expects a user-space buffer of 8 (not 4) bytes on these ABIs.
1838 .\" prctl PR_SET_TIMERSLACK
1839 .TP
1840 .BR PR_SET_TIMERSLACK " (since Linux 2.6.28)"
1841 .\" See https://lwn.net/Articles/369549/
1842 .\" commit 6976675d94042fbd446231d1bd8b7de71a980ada
1843 Each thread has two associated timer slack values:
1844 a "default" value, and a "current" value.
1845 This operation sets the "current" timer slack value for the calling thread.
1846 .I arg2
1847 is an unsigned long value, then maximum "current" value is ULONG_MAX and
1848 the minimum "current" value is 1.
1849 If the nanosecond value supplied in
1850 .IR arg2
1851 is greater than zero, then the "current" value is set to this value.
1852 If
1853 .I arg2
1854 is equal to zero,
1855 the "current" timer slack is reset to the
1856 thread's "default" timer slack value.
1857 .IP
1858 The "current" timer slack is used by the kernel to group timer expirations
1859 for the calling thread that are close to one another;
1860 as a consequence, timer expirations for the thread may be
1861 up to the specified number of nanoseconds late (but will never expire early).
1862 Grouping timer expirations can help reduce system power consumption
1863 by minimizing CPU wake-ups.
1864 .IP
1865 The timer expirations affected by timer slack are those set by
1866 .BR select (2),
1867 .BR pselect (2),
1868 .BR poll (2),
1869 .BR ppoll (2),
1870 .BR epoll_wait (2),
1871 .BR epoll_pwait (2),
1872 .BR clock_nanosleep (2),
1873 .BR nanosleep (2),
1874 and
1875 .BR futex (2)
1876 (and thus the library functions implemented via futexes, including
1877 .\" List obtained by grepping for futex usage in glibc source
1878 .BR pthread_cond_timedwait (3),
1879 .BR pthread_mutex_timedlock (3),
1880 .BR pthread_rwlock_timedrdlock (3),
1881 .BR pthread_rwlock_timedwrlock (3),
1882 and
1883 .BR sem_timedwait (3)).
1884 .IP
1885 Timer slack is not applied to threads that are scheduled under
1886 a real-time scheduling policy (see
1887 .BR sched_setscheduler (2)).
1888 .IP
1889 When a new thread is created,
1890 the two timer slack values are made the same as the "current" value
1891 of the creating thread.
1892 Thereafter, a thread can adjust its "current" timer slack value via
1893 .BR PR_SET_TIMERSLACK .
1894 The "default" value can't be changed.
1895 The timer slack values of
1896 .IR init
1897 (PID 1), the ancestor of all processes,
1898 are 50,000 nanoseconds (50 microseconds).
1899 The timer slack value is inherited by a child created via
1900 .BR fork (2),
1901 and is preserved across
1902 .BR execve (2).
1903 .IP
1904 Since Linux 4.6, the "current" timer slack value of any process
1905 can be examined and changed via the file
1906 .IR /proc/[pid]/timerslack_ns .
1907 See
1908 .BR proc (5).
1909 .\" prctl PR_GET_TIMERSLACK
1910 .TP
1911 .BR PR_GET_TIMERSLACK " (since Linux 2.6.28)"
1912 Return (as the function result)
1913 the "current" timer slack value of the calling thread.
1914 .\" prctl PR_SET_TIMING
1915 .TP
1916 .BR PR_SET_TIMING " (since Linux 2.6.0)"
1917 .\" Precisely: Linux 2.6.0-test4
1918 Set whether to use (normal, traditional) statistical process timing or
1919 accurate timestamp-based process timing, by passing
1920 .B PR_TIMING_STATISTICAL
1921 .\" 0
1922 or
1923 .B PR_TIMING_TIMESTAMP
1924 .\" 1
1925 to \fIarg2\fP.
1926 .B PR_TIMING_TIMESTAMP
1927 is not currently implemented
1928 (attempting to set this mode will yield the error
1929 .BR EINVAL ).
1930 .\" PR_TIMING_TIMESTAMP doesn't do anything in 2.6.26-rc8,
1931 .\" and looking at the patch history, it appears
1932 .\" that it never did anything.
1933 .\" prctl PR_GET_TIMING
1934 .TP
1935 .BR PR_GET_TIMING " (since Linux 2.6.0)"
1936 .\" Precisely: Linux 2.6.0-test4
1937 Return (as the function result) which process timing method is currently
1938 in use.
1939 .\" prctl PR_SET_TSC
1940 .TP
1941 .BR PR_SET_TSC " (since Linux 2.6.26, x86 only)"
1942 Set the state of the flag determining whether the timestamp counter
1943 can be read by the process.
1944 Pass
1945 .B PR_TSC_ENABLE
1946 to
1947 .I arg2
1948 to allow it to be read, or
1949 .B PR_TSC_SIGSEGV
1950 to generate a
1951 .B SIGSEGV
1952 when the process tries to read the timestamp counter.
1953 .\" prctl PR_GET_TSC
1954 .TP
1955 .BR PR_GET_TSC " (since Linux 2.6.26, x86 only)"
1956 Return the state of the flag determining whether the timestamp counter
1957 can be read,
1958 in the location pointed to by
1959 .IR "(int\ *) arg2" .
1960 .\" prctl PR_SET_UNALIGN
1961 .TP
1962 .B PR_SET_UNALIGN
1963 (Only on: ia64, since Linux 2.3.48; parisc, since Linux 2.6.15;
1964 PowerPC, since Linux 2.6.18; Alpha, since Linux 2.6.22;
1965 .\" sh: 94ea5e449ae834af058ef005d16a8ad44fcf13d6
1966 .\" tile: 2f9ac29eec71a696cb0dcc5fb82c0f8d4dac28c9
1967 sh, since Linux 2.6.34; tile, since Linux 3.12)
1968 Set unaligned access control bits to \fIarg2\fP.
1969 Pass
1970 \fBPR_UNALIGN_NOPRINT\fP to silently fix up unaligned user accesses,
1971 or \fBPR_UNALIGN_SIGBUS\fP to generate
1972 .B SIGBUS
1973 on unaligned user access.
1974 Alpha also supports an additional flag with the value
1975 of 4 and no corresponding named constant,
1976 which instructs kernel to not fix up
1977 unaligned accesses (it is analogous to providing the
1978 .BR UAC_NOFIX
1979 flag in
1980 .BR SSI_NVPAIRS
1981 operation of the
1982 .BR setsysinfo ()
1983 system call on Tru64).
1984 .\" prctl PR_GET_UNALIGN
1985 .TP
1986 .B PR_GET_UNALIGN
1987 (See
1988 .B PR_SET_UNALIGN
1989 for information on versions and architectures.)
1990 Return unaligned access control bits, in the location pointed to by
1991 .IR "(unsigned int\ *) arg2" .
1992 .SH RETURN VALUE
1993 On success,
1994 .BR PR_CAP_AMBIENT + PR_CAP_AMBIENT_IS_SET ,
1995 .BR PR_CAPBSET_READ ,
1996 .BR PR_GET_DUMPABLE ,
1997 .BR PR_GET_FP_MODE ,
1998 .BR PR_GET_IO_FLUSHER ,
1999 .BR PR_GET_KEEPCAPS ,
2000 .BR PR_MCE_KILL_GET ,
2001 .BR PR_GET_NO_NEW_PRIVS ,
2002 .BR PR_GET_SECUREBITS ,
2003 .BR PR_GET_SPECULATION_CTRL ,
2004 .BR PR_SVE_GET_VL ,
2005 .BR PR_SVE_SET_VL ,
2006 .BR PR_GET_TAGGED_ADDR_CTRL ,
2007 .BR PR_GET_THP_DISABLE ,
2008 .BR PR_GET_TIMING ,
2009 .BR PR_GET_TIMERSLACK ,
2010 and (if it returns)
2011 .BR PR_GET_SECCOMP
2012 return the nonnegative values described above.
2013 All other
2014 .I option
2015 values return 0 on success.
2016 On error, \-1 is returned, and
2017 .I errno
2018 is set to indicate the error.
2019 .SH ERRORS
2020 .TP
2021 .B EACCES
2022 .I option
2023 is
2024 .BR PR_SET_SECCOMP
2025 and
2026 .I arg2
2027 is
2028 .BR SECCOMP_MODE_FILTER ,
2029 but the process does not have the
2030 .BR CAP_SYS_ADMIN
2031 capability or has not set the
2032 .IR no_new_privs
2033 attribute (see the discussion of
2034 .BR PR_SET_NO_NEW_PRIVS
2035 above).
2036 .TP
2037 .B EACCES
2038 .I option
2039 is
2040 .BR PR_SET_MM ,
2041 and
2042 .I arg3
2043 is
2044 .BR PR_SET_MM_EXE_FILE ,
2045 the file is not executable.
2046 .TP
2047 .B EBADF
2048 .I option
2049 is
2050 .BR PR_SET_MM ,
2051 .I arg3
2052 is
2053 .BR PR_SET_MM_EXE_FILE ,
2054 and the file descriptor passed in
2055 .I arg4
2056 is not valid.
2057 .TP
2058 .B EBUSY
2059 .I option
2060 is
2061 .BR PR_SET_MM ,
2062 .I arg3
2063 is
2064 .BR PR_SET_MM_EXE_FILE ,
2065 and this the second attempt to change the
2066 .I /proc/pid/exe
2067 symbolic link, which is prohibited.
2068 .TP
2069 .B EFAULT
2070 .I arg2
2071 is an invalid address.
2072 .TP
2073 .B EFAULT
2074 .I option
2075 is
2076 .BR PR_SET_SECCOMP ,
2077 .I arg2
2078 is
2079 .BR SECCOMP_MODE_FILTER ,
2080 the system was built with
2081 .BR CONFIG_SECCOMP_FILTER ,
2082 and
2083 .I arg3
2084 is an invalid address.
2085 .TP
2086 .B EFAULT
2087 .I option
2088 is
2089 .B PR_SET_SYSCALL_USER_DISPATCH
2090 and
2091 .I arg5
2092 has an invalid address.
2093 .TP
2094 .B EINVAL
2095 The value of
2096 .I option
2097 is not recognized,
2098 or not supported on this system.
2099 .TP
2100 .B EINVAL
2101 .I option
2102 is
2103 .BR PR_MCE_KILL
2104 or
2105 .BR PR_MCE_KILL_GET
2106 or
2107 .BR PR_SET_MM ,
2108 and unused
2109 .BR prctl ()
2110 arguments were not specified as zero.
2111 .TP
2112 .B EINVAL
2113 .I arg2
2114 is not valid value for this
2115 .IR option .
2116 .TP
2117 .B EINVAL
2118 .I option
2119 is
2120 .BR PR_SET_SECCOMP
2121 or
2122 .BR PR_GET_SECCOMP ,
2123 and the kernel was not configured with
2124 .BR CONFIG_SECCOMP .
2125 .TP
2126 .B EINVAL
2127 .I option
2128 is
2129 .BR PR_SET_SECCOMP ,
2130 .I arg2
2131 is
2132 .BR SECCOMP_MODE_FILTER ,
2133 and the kernel was not configured with
2134 .BR CONFIG_SECCOMP_FILTER .
2135 .TP
2136 .B EINVAL
2137 .I option
2138 is
2139 .BR PR_SET_MM ,
2140 and one of the following is true
2141 .RS
2142 .IP * 3
2143 .I arg4
2144 or
2145 .I arg5
2146 is nonzero;
2147 .IP *
2148 .I arg3
2149 is greater than
2150 .B TASK_SIZE
2151 (the limit on the size of the user address space for this architecture);
2152 .IP *
2153 .I arg2
2154 is
2155 .BR PR_SET_MM_START_CODE ,
2156 .BR PR_SET_MM_END_CODE ,
2157 .BR PR_SET_MM_START_DATA ,
2158 .BR PR_SET_MM_END_DATA ,
2159 or
2160 .BR PR_SET_MM_START_STACK ,
2161 and the permissions of the corresponding memory area are not as required;
2162 .IP *
2163 .I arg2
2164 is
2165 .BR PR_SET_MM_START_BRK
2166 or
2167 .BR PR_SET_MM_BRK ,
2168 and
2169 .I arg3
2170 is less than or equal to the end of the data segment
2171 or specifies a value that would cause the
2172 .B RLIMIT_DATA
2173 resource limit to be exceeded.
2174 .RE
2175 .TP
2176 .B EINVAL
2177 .I option
2178 is
2179 .BR PR_SET_PTRACER
2180 and
2181 .I arg2
2182 is not 0,
2183 .BR PR_SET_PTRACER_ANY ,
2184 or the PID of an existing process.
2185 .TP
2186 .B EINVAL
2187 .I option
2188 is
2189 .B PR_SET_PDEATHSIG
2190 and
2191 .I arg2
2192 is not a valid signal number.
2193 .TP
2194 .B EINVAL
2195 .I option
2196 is
2197 .BR PR_SET_DUMPABLE
2198 and
2199 .I arg2
2200 is neither
2201 .B SUID_DUMP_DISABLE
2202 nor
2203 .BR SUID_DUMP_USER .
2204 .TP
2205 .B EINVAL
2206 .I option
2207 is
2208 .BR PR_SET_TIMING
2209 and
2210 .I arg2
2211 is not
2212 .BR PR_TIMING_STATISTICAL .
2213 .TP
2214 .B EINVAL
2215 .I option
2216 is
2217 .BR PR_SET_NO_NEW_PRIVS
2218 and
2219 .I arg2
2220 is not equal to 1
2221 or
2222 .IR arg3 ,
2223 .IR arg4 ,
2224 or
2225 .IR arg5
2226 is nonzero.
2227 .TP
2228 .B EINVAL
2229 .I option
2230 is
2231 .BR PR_GET_NO_NEW_PRIVS
2232 and
2233 .IR arg2 ,
2234 .IR arg3 ,
2235 .IR arg4 ,
2236 or
2237 .IR arg5
2238 is nonzero.
2239 .TP
2240 .B EINVAL
2241 .I option
2242 is
2243 .BR PR_SET_THP_DISABLE
2244 and
2245 .IR arg3 ,
2246 .IR arg4 ,
2247 or
2248 .IR arg5
2249 is nonzero.
2250 .TP
2251 .B EINVAL
2252 .I option
2253 is
2254 .BR PR_GET_THP_DISABLE
2255 and
2256 .IR arg2 ,
2257 .IR arg3 ,
2258 .IR arg4 ,
2259 or
2260 .IR arg5
2261 is nonzero.
2262 .TP
2263 .B EINVAL
2264 .I option
2265 is
2266 .B PR_CAP_AMBIENT
2267 and an unused argument
2268 .RI ( arg4 ,
2269 .IR arg5 ,
2270 or,
2271 in the case of
2272 .BR PR_CAP_AMBIENT_CLEAR_ALL ,
2273 .IR arg3 )
2274 is nonzero; or
2275 .IR arg2
2276 has an invalid value;
2277 or
2278 .IR arg2
2279 is
2280 .BR PR_CAP_AMBIENT_LOWER ,
2281 .BR PR_CAP_AMBIENT_RAISE ,
2282 or
2283 .BR PR_CAP_AMBIENT_IS_SET
2284 and
2285 .IR arg3
2286 does not specify a valid capability.
2287 .TP
2288 .B EINVAL
2289 .I option
2290 was
2291 .BR PR_GET_SPECULATION_CTRL
2292 or
2293 .BR PR_SET_SPECULATION_CTRL
2294 and unused arguments to
2295 .BR prctl ()
2296 are not 0.
2297 .B EINVAL
2298 .I option
2299 is
2300 .B PR_PAC_RESET_KEYS
2301 and the arguments are invalid or unsupported.
2302 See the description of
2303 .B PR_PAC_RESET_KEYS
2304 above for details.
2305 .TP
2306 .B EINVAL
2307 .I option
2308 is
2309 .B PR_SVE_SET_VL
2310 and the arguments are invalid or unsupported,
2311 or SVE is not available on this platform.
2312 See the description of
2313 .B PR_SVE_SET_VL
2314 above for details.
2315 .TP
2316 .B EINVAL
2317 .I option
2318 is
2319 .B PR_SVE_GET_VL
2320 and SVE is not available on this platform.
2321 .TP
2322 .B EINVAL
2323 .I option
2324 is
2325 .B PR_SET_SYSCALL_USER_DISPATCH
2326 and one of the following is true:
2327 .RS
2328 .IP * 3
2329 .I arg2
2330 is
2331 .B PR_SYS_DISPATCH_OFF
2332 and the remaining arguments are not 0;
2333 .IP * 3
2334 .I arg2
2335 is
2336 .B PR_SYS_DISPATCH_ON
2337 and the memory range specified is outside the
2338 address space of the process.
2339 .IP * 3
2340 .I arg2
2341 is invalid.
2342 .RE
2343 .TP
2344 .B EINVAL
2345 .I option
2346 is
2347 .BR PR_SET_TAGGED_ADDR_CTRL
2348 and the arguments are invalid or unsupported.
2349 See the description of
2350 .B PR_SET_TAGGED_ADDR_CTRL
2351 above for details.
2352 .TP
2353 .B EINVAL
2354 .I option
2355 is
2356 .BR PR_GET_TAGGED_ADDR_CTRL
2357 and the arguments are invalid or unsupported.
2358 See the description of
2359 .B PR_GET_TAGGED_ADDR_CTRL
2360 above for details.
2361 .TP
2362 .B ENODEV
2363 .I option
2364 was
2365 .BR PR_SET_SPECULATION_CTRL
2366 the kernel or CPU does not support the requested speculation misfeature.
2367 .TP
2368 .B ENXIO
2369 .I option
2370 was
2371 .BR PR_MPX_ENABLE_MANAGEMENT
2372 or
2373 .BR PR_MPX_DISABLE_MANAGEMENT
2374 and the kernel or the CPU does not support MPX management.
2375 Check that the kernel and processor have MPX support.
2376 .TP
2377 .B ENXIO
2378 .I option
2379 was
2380 .BR PR_SET_SPECULATION_CTRL
2381 implies that the control of the selected speculation misfeature is not possible.
2382 See
2383 .BR PR_GET_SPECULATION_CTRL
2384 for the bit fields to determine which option is available.
2385 .TP
2386 .B EOPNOTSUPP
2387 .I option
2388 is
2389 .B PR_SET_FP_MODE
2390 and
2391 .I arg2
2392 has an invalid or unsupported value.
2393 .TP
2394 .B EPERM
2395 .I option
2396 is
2397 .BR PR_SET_SECUREBITS ,
2398 and the caller does not have the
2399 .B CAP_SETPCAP
2400 capability,
2401 or tried to unset a "locked" flag,
2402 or tried to set a flag whose corresponding locked flag was set
2403 (see
2404 .BR capabilities (7)).
2405 .TP
2406 .B EPERM
2407 .I option
2408 is
2409 .BR PR_SET_SPECULATION_CTRL
2410 wherein the speculation was disabled with
2411 .B PR_SPEC_FORCE_DISABLE
2412 and caller tried to enable it again.
2413 .TP
2414 .B EPERM
2415 .I option
2416 is
2417 .BR PR_SET_KEEPCAPS ,
2418 and the caller's
2419 .B SECBIT_KEEP_CAPS_LOCKED
2420 flag is set
2421 (see
2422 .BR capabilities (7)).
2423 .TP
2424 .B EPERM
2425 .I option
2426 is
2427 .BR PR_CAPBSET_DROP ,
2428 and the caller does not have the
2429 .B CAP_SETPCAP
2430 capability.
2431 .TP
2432 .B EPERM
2433 .I option
2434 is
2435 .BR PR_SET_MM ,
2436 and the caller does not have the
2437 .B CAP_SYS_RESOURCE
2438 capability.
2439 .TP
2440 .B EPERM
2441 .IR option
2442 is
2443 .BR PR_CAP_AMBIENT
2444 and
2445 .IR arg2
2446 is
2447 .BR PR_CAP_AMBIENT_RAISE ,
2448 but either the capability specified in
2449 .IR arg3
2450 is not present in the process's permitted and inheritable capability sets,
2451 or the
2452 .B PR_CAP_AMBIENT_LOWER
2453 securebit has been set.
2454 .TP
2455 .B ERANGE
2456 .I option
2457 was
2458 .BR PR_SET_SPECULATION_CTRL
2459 and
2460 .IR arg3
2461 is not
2462 .BR PR_SPEC_ENABLE ,
2463 .BR PR_SPEC_DISABLE ,
2464 .BR PR_SPEC_FORCE_DISABLE ,
2465 nor
2466 .BR PR_SPEC_DISABLE_NOEXEC .
2467 .SH VERSIONS
2468 The
2469 .BR prctl ()
2470 system call was introduced in Linux 2.1.57.
2471 .\" The library interface was added in glibc 2.0.6
2472 .SH CONFORMING TO
2473 This call is Linux-specific.
2474 IRIX has a
2475 .BR prctl ()
2476 system call (also introduced in Linux 2.1.44
2477 as irix_prctl on the MIPS architecture),
2478 with prototype
2479 .PP
2480 .in +4n
2481 .EX
2482 .BI "ptrdiff_t prctl(int " option ", int " arg2 ", int " arg3 );
2483 .EE
2484 .in
2485 .PP
2486 and options to get the maximum number of processes per user,
2487 get the maximum number of processors the calling process can use,
2488 find out whether a specified process is currently blocked,
2489 get or set the maximum stack size, and so on.
2490 .SH SEE ALSO
2491 .BR signal (2),
2492 .BR core (5)