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>
11 .\" %%%LICENSE_START(VERBATIM)
12 .\" Permission is granted to make and distribute verbatim copies of this
13 .\" manual provided the copyright notice and this permission notice are
14 .\" preserved on all copies.
16 .\" Permission is granted to copy and distribute modified versions of this
17 .\" manual under the conditions for verbatim copying, provided that the
18 .\" entire resulting derived work is distributed under the terms of a
19 .\" permission notice identical to this one.
21 .\" Since the Linux kernel and libraries are constantly changing, this
22 .\" manual page may be incorrect or out-of-date. The author(s) assume no
23 .\" responsibility for errors or omissions, or for damages resulting from
24 .\" the use of the information contained herein. The author(s) may not
25 .\" have taken the same level of care in the production of this manual,
26 .\" which is licensed free of charge, as they might when working
29 .\" Formatted or processed versions of this manual, if unaccompanied by
30 .\" the source, must acknowledge the copyright and authors of this work.
33 .\" Modified Thu Nov 11 04:19:42 MET 1999, aeb: added PR_GET_PDEATHSIG
34 .\" Modified 27 Jun 02, Michael Kerrisk
35 .\" Added PR_SET_DUMPABLE, PR_GET_DUMPABLE,
36 .\" PR_SET_KEEPCAPS, PR_GET_KEEPCAPS
37 .\" Modified 2006-08-30 Guillem Jover <guillem@hadrons.org>
38 .\" Updated Linux versions where the options where introduced.
39 .\" Added PR_SET_TIMING, PR_GET_TIMING, PR_SET_NAME, PR_GET_NAME,
40 .\" PR_SET_UNALIGN, PR_GET_UNALIGN, PR_SET_FPEMU, PR_GET_FPEMU,
41 .\" PR_SET_FPEXC, PR_GET_FPEXC
42 .\" 2008-04-29 Serge Hallyn, Document PR_CAPBSET_READ and PR_CAPBSET_DROP
43 .\" 2008-06-13 Erik Bosman, <ejbosman@cs.vu.nl>
44 .\" Document PR_GET_TSC and PR_SET_TSC.
45 .\" 2008-06-15 mtk, Document PR_SET_SECCOMP, PR_GET_SECCOMP
46 .\" 2009-10-03 Andi Kleen, document PR_MCE_KILL
47 .\" 2012-04 Cyrill Gorcunov, Document PR_SET_MM
48 .\" 2012-04-25 Michael Kerrisk, Document PR_TASK_PERF_EVENTS_DISABLE and
49 .\" PR_TASK_PERF_EVENTS_ENABLE
50 .\" 2012-09-20 Kees Cook, update PR_SET_SECCOMP for mode 2
51 .\" 2012-09-20 Kees Cook, document PR_SET_NO_NEW_PRIVS, PR_GET_NO_NEW_PRIVS
52 .\" 2012-10-25 Michael Kerrisk, Document PR_SET_TIMERSLACK and
54 .\" 2013-01-10 Kees Cook, document PR_SET_PTRACER
55 .\" 2012-02-04 Michael Kerrisk, document PR_{SET,GET}_CHILD_SUBREAPER
56 .\" 2014-11-10 Dave Hansen, document PR_MPX_{EN,DIS}ABLE_MANAGEMENT
59 .TH PRCTL 2 2021-03-22 "Linux" "Linux Programmer's Manual"
61 prctl \- operations on a process or thread
64 .B #include <sys/prctl.h>
66 .BI "int prctl(int " option ", unsigned long " arg2 ", unsigned long " arg3 ,
67 .BI " unsigned long " arg4 ", unsigned long " arg5 );
71 manipulates various aspects of the behavior
72 of the calling thread or process.
74 Note that careless use of some
76 operations can confuse the user-space run-time environment,
77 so these operations should be used with care.
80 is called with a first argument describing what to do
81 (with values defined in \fI<linux/prctl.h>\fP), and further
82 arguments with a significance depending on the first one.
83 The first argument can be:
85 .\" prctl PR_CAP_AMBIENT
87 .BR PR_CAP_AMBIENT " (since Linux 4.3)"
88 .\" commit 58319057b7847667f0c9585b9de0e8932b0fdb08
89 Reads or changes the ambient capability set of the calling thread,
90 according to the value of
92 which must be one of the following:
96 .B PR_CAP_AMBIENT_RAISE
97 The capability specified in
99 is added to the ambient set.
100 The specified capability must already be present in
101 both the permitted and the inheritable sets of the process.
102 This operation is not permitted if the
103 .B SECBIT_NO_CAP_AMBIENT_RAISE
106 .B PR_CAP_AMBIENT_LOWER
107 The capability specified in
109 is removed from the ambient set.
111 .B PR_CAP_AMBIENT_IS_SET
114 call returns 1 if the capability in
116 is in the ambient set and 0 if it is not.
118 .BR PR_CAP_AMBIENT_CLEAR_ALL
119 All capabilities will be removed from the ambient set.
120 This operation requires setting
125 In all of the above operations,
129 must be specified as 0.
131 Higher-level interfaces layered on top of the above operations are
134 library in the form of
135 .BR cap_get_ambient (3),
136 .BR cap_set_ambient (3),
138 .BR cap_reset_ambient (3).
139 .\" prctl PR_CAPBSET_READ
141 .BR PR_CAPBSET_READ " (since Linux 2.6.25)"
142 Return (as the function result) 1 if the capability specified in
144 is in the calling thread's capability bounding set,
146 (The capability constants are defined in
147 .IR <linux/capability.h> .)
148 The capability bounding set dictates
149 whether the process can receive the capability through a
150 file's permitted capability set on a subsequent call to
153 If the capability specified in
155 is not valid, then the call fails with the error
158 A higher-level interface layered on top of this operation is provided in the
160 library in the form of
161 .BR cap_get_bound (3).
162 .\" prctl PR_CAPBSET_DROP
164 .BR PR_CAPBSET_DROP " (since Linux 2.6.25)"
165 If the calling thread has the
167 capability within its user namespace, then drop the capability specified by
169 from the calling thread's capability bounding set.
170 Any children of the calling thread will inherit the newly
171 reduced bounding set.
173 The call fails with the error:
175 if the calling thread does not have the
180 does not represent a valid capability; or
182 if file capabilities are not enabled in the kernel,
183 in which case bounding sets are not supported.
185 A higher-level interface layered on top of this operation is provided in the
187 library in the form of
188 .BR cap_drop_bound (3).
189 .\" prctl PR_SET_CHILD_SUBREAPER
191 .BR PR_SET_CHILD_SUBREAPER " (since Linux 3.4)"
192 .\" commit ebec18a6d3aa1e7d84aab16225e87fd25170ec2b
196 set the "child subreaper" attribute of the calling process;
199 is zero, unset the attribute.
201 A subreaper fulfills the role of
203 for its descendant processes.
204 When a process becomes orphaned
205 (i.e., its immediate parent terminates),
206 then that process will be reparented to
207 the nearest still living ancestor subreaper.
208 Subsequently, calls to
210 in the orphaned process will now return the PID of the subreaper process,
211 and when the orphan terminates, it is the subreaper process that
214 signal and will be able to
216 on the process to discover its termination status.
218 The setting of the "child subreaper" attribute
219 is not inherited by children created by
223 The setting is preserved across
226 Establishing a subreaper process is useful in session management frameworks
227 where a hierarchical group of processes is managed by a subreaper process
228 that needs to be informed when one of the processes\(emfor example,
229 a double-forked daemon\(emterminates
230 (perhaps so that it can restart that process).
235 employ a subreaper process for similar reasons.
236 .\" prctl PR_GET_CHILD_SUBREAPER
238 .BR PR_GET_CHILD_SUBREAPER " (since Linux 3.4)"
239 Return the "child subreaper" setting of the caller,
240 in the location pointed to by
241 .IR "(int\ *) arg2" .
242 .\" prctl PR_SET_DUMPABLE
244 .BR PR_SET_DUMPABLE " (since Linux 2.3.20)"
245 Set the state of the "dumpable" attribute,
246 which determines whether core dumps are produced for the calling process
247 upon delivery of a signal whose default behavior is to produce a core dump.
249 In kernels up to and including 2.6.12,
252 .RB ( SUID_DUMP_DISABLE ,
253 process is not dumpable) or 1
254 .RB ( SUID_DUMP_USER ,
255 process is dumpable).
256 Between kernels 2.6.13 and 2.6.17,
257 .\" commit abf75a5033d4da7b8a7e92321d74021d1fcfb502
258 the value 2 was also permitted,
259 which caused any binary which normally would not be dumped
260 to be dumped readable by root only;
261 for security reasons, this feature has been removed.
262 .\" See http://marc.theaimsgroup.com/?l=linux-kernel&m=115270289030630&w=2
263 .\" Subject: Fix prctl privilege escalation (CVE-2006-2451)
264 .\" From: Marcel Holtmann <marcel () holtmann ! org>
265 .\" Date: 2006-07-12 11:12:00
266 (See also the description of
267 .I /proc/sys/fs/\:suid_dumpable
271 Normally, the "dumpable" attribute is set to 1.
272 However, it is reset to the current value contained in the file
273 .IR /proc/sys/fs/\:suid_dumpable
274 (which by default has the value 0),
275 in the following circumstances:
276 .\" See kernel/cred.c::commit_creds() (Linux 3.18 sources)
279 The process's effective user or group ID is changed.
281 The process's filesystem user or group ID is changed (see
282 .BR credentials (7)).
286 a set-user-ID or set-group-ID program, resulting in a change
287 of either the effective user ID or the effective group ID.
291 a program that has file capabilities (see
292 .BR capabilities (7)),
293 .\" See kernel/cred.c::commit_creds()
294 but only if the permitted capabilities
295 gained exceed those already permitted for the process.
296 .\" Also certain namespace operations;
299 Processes that are not dumpable can not be attached via
306 If a process is not dumpable,
307 the ownership of files in the process's
309 directory is affected as described in
311 .\" prctl PR_GET_DUMPABLE
313 .BR PR_GET_DUMPABLE " (since Linux 2.3.20)"
314 Return (as the function result) the current state of the calling
315 process's dumpable attribute.
316 .\" Since Linux 2.6.13, the dumpable flag can have the value 2,
317 .\" but in 2.6.13 PR_GET_DUMPABLE simply returns 1 if the dumpable
318 .\" flags has a nonzero value. This was fixed in 2.6.14.
319 .\" prctl PR_SET_ENDIAN
321 .BR PR_SET_ENDIAN " (since Linux 2.6.18, PowerPC only)"
322 Set the endian-ness of the calling process to the value given
323 in \fIarg2\fP, which should be one of the following:
324 .\" Respectively 0, 1, 2
326 .BR PR_ENDIAN_LITTLE ,
328 .B PR_ENDIAN_PPC_LITTLE
329 (PowerPC pseudo little endian).
330 .\" prctl PR_GET_ENDIAN
332 .BR PR_GET_ENDIAN " (since Linux 2.6.18, PowerPC only)"
333 Return the endian-ness of the calling process,
334 in the location pointed to by
335 .IR "(int\ *) arg2" .
336 .\" prctl PR_SET_FP_MODE
338 .BR PR_SET_FP_MODE " (since Linux 4.0, only on MIPS)"
339 .\" commit 9791554b45a2acc28247f66a5fd5bbc212a6b8c8
340 On the MIPS architecture,
341 user-space code can be built using an ABI which permits linking
342 with code that has more restrictive floating-point (FP) requirements.
343 For example, user-space code may be built to target the O32 FPXX ABI
344 and linked with code built for either one of the more restrictive
346 When more restrictive code is linked in,
347 the overall requirement for the process is to use the more
348 restrictive floating-point mode.
350 Because the kernel has no means of knowing in advance
351 which mode the process should be executed in,
352 and because these restrictions can
353 change over the lifetime of the process, the
355 operation is provided to allow control of the floating-point mode
358 .\" https://dmz-portal.mips.com/wiki/MIPS_O32_ABI_-_FR0_and_FR1_Interlinking
360 .I (unsigned int) arg2
361 argument is a bit mask describing the floating-point mode used:
369 mode), the 32 floating-point registers are 32 bits wide,
370 and 64-bit registers are represented as a pair of registers
371 (even- and odd- numbered,
372 with the even-numbered register containing the lower 32 bits,
373 and the odd-numbered register containing the higher 32 bits).
377 (on supported hardware),
378 the 32 floating-point registers are 64 bits wide (so called
381 Note that modern MIPS implementations (MIPS R6 and newer) support
385 Applications that use the O32 FP32 ABI can operate only when this bit is
388 or they can be used with FRE enabled, see below).
389 Applications that use the O32 FP64 ABI
390 (and the O32 FP64A ABI, which exists to
391 provide the ability to operate with existing FP32 code; see below)
392 can operate only when this bit is
395 Applications that use the O32 FPXX ABI can operate with either
401 Enable emulation of 32-bit floating-point mode.
402 When this mode is enabled,
403 it emulates 32-bit floating-point operations
404 by raising a reserved-instruction exception
405 on every instruction that uses 32-bit formats and
406 the kernel then handles the instruction in software.
407 (The problem lies in the discrepancy of handling odd-numbered registers
408 which are the high 32 bits of 64-bit registers with even numbers in
410 mode and the lower 32-bit parts of odd-numbered 64-bit registers in
413 Enabling this bit is necessary when code with the O32 FP32 ABI should operate
414 with code with compatible the O32 FPXX or O32 FP64A ABIs (which require
416 FPU mode) or when it is executed on newer hardware (MIPS R6 onwards)
419 mode support when a binary with the FP32 ABI is used.
421 Note that this mode makes sense only when the FPU is in 64-bit mode
424 Note that the use of emulation inherently has a significant performance hit
425 and should be avoided if possible.
428 In the N32/N64 ABI, 64-bit floating-point mode is always used,
429 so FPU emulation is not required and the FPU always operates in
433 This option is mainly intended for use by the dynamic linker
442 .\" prctl PR_GET_FP_MODE
444 .BR PR_GET_FP_MODE " (since Linux 4.0, only on MIPS)"
445 Return (as the function result)
446 the current floating-point mode (see the description of
451 the call returns a bit mask which represents the current floating-point mode.
460 .\" prctl PR_SET_FPEMU
462 .BR PR_SET_FPEMU " (since Linux 2.4.18, 2.5.9, only on ia64)"
463 Set floating-point emulation control bits to \fIarg2\fP.
466 to silently emulate floating-point operation accesses, or
468 to not emulate floating-point operations and send
471 .\" prctl PR_GET_FPEMU
473 .BR PR_GET_FPEMU " (since Linux 2.4.18, 2.5.9, only on ia64)"
474 Return floating-point emulation control bits,
475 in the location pointed to by
476 .IR "(int\ *) arg2" .
477 .\" prctl PR_SET_FPEXC
479 .BR PR_SET_FPEXC " (since Linux 2.4.21, 2.5.32, only on PowerPC)"
480 Set floating-point exception mode to \fIarg2\fP.
481 Pass \fBPR_FP_EXC_SW_ENABLE\fP to use FPEXC for FP exception enables,
482 \fBPR_FP_EXC_DIV\fP for floating-point divide by zero,
483 \fBPR_FP_EXC_OVF\fP for floating-point overflow,
484 \fBPR_FP_EXC_UND\fP for floating-point underflow,
485 \fBPR_FP_EXC_RES\fP for floating-point inexact result,
486 \fBPR_FP_EXC_INV\fP for floating-point invalid operation,
487 \fBPR_FP_EXC_DISABLED\fP for FP exceptions disabled,
488 \fBPR_FP_EXC_NONRECOV\fP for async nonrecoverable exception mode,
489 \fBPR_FP_EXC_ASYNC\fP for async recoverable exception mode,
490 \fBPR_FP_EXC_PRECISE\fP for precise exception mode.
491 .\" prctl PR_GET_FPEXC
493 .BR PR_GET_FPEXC " (since Linux 2.4.21, 2.5.32, only on PowerPC)"
494 Return floating-point exception mode,
495 in the location pointed to by
496 .IR "(int\ *) arg2" .
497 .\" prctl PR_SET_IO_FLUSHER
499 .BR PR_SET_IO_FLUSHER " (since Linux 5.6)"
500 If a user process is involved in the block layer or filesystem I/O path,
501 and can allocate memory while processing I/O requests it must set
503 This will put the process in the IO_FLUSHER state,
504 which allows it special treatment to make progress when allocating memory.
505 If \fIarg2\fP is 0, the process will clear the IO_FLUSHER state, and
506 the default behavior will be used.
508 The calling process must have the
518 The IO_FLUSHER state is inherited by a child process created via
520 and is preserved across
523 Examples of IO_FLUSHER applications are FUSE daemons, SCSI device
524 emulation daemons, and daemons that perform error handling like multipath
525 path recovery applications.
526 .\" prctl PR_GET_IO_FLUSHER
528 .B PR_GET_IO_FLUSHER (Since Linux 5.6)
529 Return (as the function result) the IO_FLUSHER state of the caller.
530 A value of 1 indicates that the caller is in the IO_FLUSHER state;
531 0 indicates that the caller is not in the IO_FLUSHER state.
533 The calling process must have the
543 .\" prctl PR_SET_KEEPCAPS
545 .BR PR_SET_KEEPCAPS " (since Linux 2.2.18)"
546 Set the state of the calling thread's "keep capabilities" flag.
547 The effect of this flag is described in
548 .BR capabilities (7).
550 must be either 0 (clear the flag)
552 The "keep capabilities" value will be reset to 0 on subsequent calls to
554 .\" prctl PR_GET_KEEPCAPS
556 .BR PR_GET_KEEPCAPS " (since Linux 2.2.18)"
557 Return (as the function result) the current state of the calling thread's
558 "keep capabilities" flag.
561 for a description of this flag.
562 .\" prctl PR_MCE_KILL
564 .BR PR_MCE_KILL " (since Linux 2.6.32)"
565 Set the machine check memory corruption kill policy for the calling thread.
569 .BR PR_MCE_KILL_CLEAR ,
570 clear the thread memory corruption kill policy and use the system-wide default.
571 (The system-wide default is defined by
572 .IR /proc/sys/vm/memory_failure_early_kill ;
578 .BR PR_MCE_KILL_SET ,
579 use a thread-specific memory corruption kill policy.
582 defines whether the policy is
584 .RB ( PR_MCE_KILL_EARLY ),
586 .RB ( PR_MCE_KILL_LATE ),
587 or the system-wide default
588 .RB ( PR_MCE_KILL_DEFAULT ).
589 Early kill means that the thread receives a
591 signal as soon as hardware memory corruption is detected inside
593 In late kill mode, the process is killed only when it accesses a corrupted page.
596 for more information on the
599 The policy is inherited by children.
602 arguments must be zero for future compatibility.
603 .\" prctl PR_MCE_KILL_GET
605 .BR PR_MCE_KILL_GET " (since Linux 2.6.32)"
606 Return (as the function result)
607 the current per-process machine check kill policy.
610 arguments must be zero.
613 .BR PR_SET_MM " (since Linux 3.3)"
614 .\" commit 028ee4be34a09a6d48bdf30ab991ae933a7bc036
615 Modify certain kernel memory map descriptor fields
616 of the calling process.
617 Usually these fields are set by the kernel and dynamic loader (see
619 for more information) and a regular application should not use this feature.
620 However, there are cases, such as self-modifying programs,
621 where a program might find it useful to change its own memory map.
623 The calling process must have the
628 is one of the options below, while
630 provides a new value for the option.
635 arguments must be zero if unused.
638 .\" commit 52b3694157e3aa6df871e283115652ec6f2d31e0
639 this feature is available only if the kernel is built with the
640 .BR CONFIG_CHECKPOINT_RESTORE
644 .BR PR_SET_MM_START_CODE
645 Set the address above which the program text can run.
646 The corresponding memory area must be readable and executable,
647 but not writable or shareable (see
651 for more information).
653 .BR PR_SET_MM_END_CODE
654 Set the address below which the program text can run.
655 The corresponding memory area must be readable and executable,
656 but not writable or shareable.
658 .BR PR_SET_MM_START_DATA
659 Set the address above which initialized and
660 uninitialized (bss) data are placed.
661 The corresponding memory area must be readable and writable,
662 but not executable or shareable.
664 .B PR_SET_MM_END_DATA
665 Set the address below which initialized and
666 uninitialized (bss) data are placed.
667 The corresponding memory area must be readable and writable,
668 but not executable or shareable.
670 .BR PR_SET_MM_START_STACK
671 Set the start address of the stack.
672 The corresponding memory area must be readable and writable.
674 .BR PR_SET_MM_START_BRK
675 Set the address above which the program heap can be expanded with
678 The address must be greater than the ending address of
679 the current program data segment.
680 In addition, the combined size of the resulting heap and
681 the size of the data segment can't exceed the
690 The requirements for the address are the same as for the
691 .BR PR_SET_MM_START_BRK
694 The following options are available since Linux 3.5.
695 .\" commit fe8c7f5cbf91124987106faa3bdf0c8b955c4cf7
697 .BR PR_SET_MM_ARG_START
698 Set the address above which the program command line is placed.
700 .BR PR_SET_MM_ARG_END
701 Set the address below which the program command line is placed.
703 .BR PR_SET_MM_ENV_START
704 Set the address above which the program environment is placed.
706 .BR PR_SET_MM_ENV_END
707 Set the address below which the program environment is placed.
709 The address passed with
710 .BR PR_SET_MM_ARG_START ,
711 .BR PR_SET_MM_ARG_END ,
712 .BR PR_SET_MM_ENV_START ,
714 .BR PR_SET_MM_ENV_END
715 should belong to a process stack area.
716 Thus, the corresponding memory area must be readable, writable, and
717 (depending on the kernel configuration) have the
723 Set a new auxiliary vector.
726 argument should provide the address of the vector.
729 is the size of the vector.
731 .BR PR_SET_MM_EXE_FILE
732 .\" commit b32dfe377102ce668775f8b6b1461f7ad428f8b6
735 symbolic link with a new one pointing to a new executable file
736 identified by the file descriptor provided in
739 The file descriptor should be obtained with a regular
743 To change the symbolic link, one needs to unmap all existing
744 executable memory areas, including those created by the kernel itself
745 (for example the kernel usually creates at least one executable
746 memory area for the ELF
750 In Linux 4.9 and earlier, the
751 .\" commit 3fb4afd9a504c2386b8435028d43283216bf588e
752 .BR PR_SET_MM_EXE_FILE
753 operation can be performed only once in a process's lifetime;
754 attempting to perform the operation a second time results in the error
756 This restriction was enforced for security reasons that were subsequently
758 and the restriction was removed in Linux 4.10 because some
759 user-space applications needed to perform this operation more than once.
761 The following options are available since Linux 3.18.
762 .\" commit f606b77f1a9e362451aca8f81d8f36a3a112139e
765 Provides one-shot access to all the addresses by passing in a
766 .I struct prctl_mm_map
767 (as defined in \fI<linux/prctl.h>\fP).
770 argument should provide the size of the struct.
772 This feature is available only if the kernel is built with the
773 .BR CONFIG_CHECKPOINT_RESTORE
776 .BR PR_SET_MM_MAP_SIZE
777 Returns the size of the
778 .I struct prctl_mm_map
780 This allows user space to find a compatible struct.
783 argument should be a pointer to an unsigned int.
785 This feature is available only if the kernel is built with the
786 .BR CONFIG_CHECKPOINT_RESTORE
789 .\" prctl PR_MPX_ENABLE_MANAGEMENT
791 .BR PR_MPX_ENABLE_MANAGEMENT ", " PR_MPX_DISABLE_MANAGEMENT " (since Linux 3.19, removed in Linux 5.4; only on x86)"
792 .\" commit fe3d197f84319d3bce379a9c0dc17b1f48ad358c
793 .\" See also http://lwn.net/Articles/582712/
794 .\" See also https://gcc.gnu.org/wiki/Intel%20MPX%20support%20in%20the%20GCC%20compiler
795 Enable or disable kernel management of Memory Protection eXtensions (MPX)
803 .\" commit e9d1b4f3c60997fe197bf0243cb4a41a44387a88
804 arguments must be zero.
806 MPX is a hardware-assisted mechanism for performing bounds checking on
808 It consists of a set of registers storing bounds information
809 and a set of special instruction prefixes that tell the CPU on which
810 instructions it should do bounds enforcement.
811 There is a limited number of these registers and
812 when there are more pointers than registers,
813 their contents must be "spilled" into a set of tables.
814 These tables are called "bounds tables" and the MPX
817 whether the kernel manages their allocation and freeing.
819 When management is enabled, the kernel will take over allocation
820 and freeing of the bounds tables.
821 It does this by trapping the #BR exceptions that result
822 at first use of missing bounds tables and
823 instead of delivering the exception to user space,
824 it allocates the table and populates the bounds directory
825 with the location of the new table.
826 For freeing, the kernel checks to see if bounds tables are
827 present for memory which is not allocated, and frees them if so.
829 Before enabling MPX management using
830 .BR PR_MPX_ENABLE_MANAGEMENT ,
831 the application must first have allocated a user-space buffer for
832 the bounds directory and placed the location of that directory in the
836 These calls fail if the CPU or kernel does not support MPX.
837 Kernel support for MPX is enabled via the
838 .BR CONFIG_X86_INTEL_MPX
839 configuration option.
840 You can check whether the CPU supports MPX by looking for the
842 CPUID bit, like with the following command:
846 cat /proc/cpuinfo | grep \(aq mpx \(aq
850 A thread may not switch in or out of long (64-bit) mode while MPX is
853 All threads in a process are affected by these calls.
857 inherits the state of MPX management.
860 MPX management is reset to a state as if
861 .BR PR_MPX_DISABLE_MANAGEMENT
864 For further information on Intel MPX, see the kernel source file
865 .IR Documentation/x86/intel_mpx.txt .
867 .\" commit f240652b6032b48ad7fa35c5e701cc4c8d697c0b
868 .\" See also https://lkml.kernel.org/r/20190705175321.DB42F0AD@viggo.jf.intel.com
869 Due to a lack of toolchain support,
870 .BR PR_MPX_ENABLE_MANAGEMENT " and " PR_MPX_DISABLE_MANAGEMENT
871 are not supported in Linux 5.4 and later.
872 .\" prctl PR_SET_NAME
874 .BR PR_SET_NAME " (since Linux 2.6.9)"
875 Set the name of the calling thread,
876 using the value in the location pointed to by
877 .IR "(char\ *) arg2" .
878 The name can be up to 16 bytes long,
879 .\" TASK_COMM_LEN in include/linux/sched.h
880 including the terminating null byte.
881 (If the length of the string, including the terminating null byte,
882 exceeds 16 bytes, the string is silently truncated.)
883 This is the same attribute that can be set via
884 .BR pthread_setname_np (3)
886 .BR pthread_getname_np (3).
887 The attribute is likewise accessible via
888 .IR /proc/self/task/[tid]/comm
893 is the thread ID of the calling thread, as returned by
895 .\" prctl PR_GET_NAME
897 .BR PR_GET_NAME " (since Linux 2.6.11)"
898 Return the name of the calling thread,
899 in the buffer pointed to by
900 .IR "(char\ *) arg2" .
901 The buffer should allow space for up to 16 bytes;
902 the returned string will be null-terminated.
903 .\" prctl PR_SET_NO_NEW_PRIVS
905 .BR PR_SET_NO_NEW_PRIVS " (since Linux 3.5)"
906 Set the calling thread's
908 attribute to the value in
914 promises not to grant privileges to do anything
915 that could not have been done without the
918 rendering the set-user-ID and set-group-ID mode bits,
919 and file capabilities non-functional).
922 attribute cannot be unset.
923 The setting of this attribute is inherited by children created by
931 the value of a thread's
933 attribute can be viewed via the
936 .IR /proc/[pid]/status
939 For more information, see the kernel source file
940 .IR Documentation/userspace\-api/no_new_privs.rst
941 .\" commit 40fde647ccb0ae8c11d256d271e24d385eed595b
943 .IR Documentation/prctl/no_new_privs.txt
947 .\" prctl PR_GET_NO_NEW_PRIVS
949 .BR PR_GET_NO_NEW_PRIVS " (since Linux 3.5)"
950 Return (as the function result) the value of the
952 attribute for the calling thread.
953 A value of 0 indicates the regular
956 A value of 1 indicates
958 will operate in the privilege-restricting mode described above.
959 .\" prctl PR_PAC_RESET_KEYS
960 .\" commit ba830885656414101b2f8ca88786524d4bb5e8c1
962 .BR PR_PAC_RESET_KEYS " (since Linux 5.0, only on arm64)"
963 Securely reset the thread's pointer authentication keys
964 to fresh random values generated by the kernel.
966 The set of keys to be reset is specified by
968 which must be a logical OR of zero or more of the following:
972 instruction authentication key A
975 instruction authentication key B
978 data authentication key A
981 data authentication key B
984 generic authentication \(lqA\(rq key.
986 (Yes folks, there really is no generic B key.)
989 As a special case, if
991 is zero, then all the keys are reset.
992 Since new keys could be added in future,
993 this is the recommended way to completely wipe the existing keys
994 when establishing a clean execution context.
995 Note that there is no need to use
996 .BR PR_PAC_RESET_KEYS
997 in preparation for calling
1001 resets all the pointer authentication keys.
1003 The remaining arguments
1004 .IR arg3 ", " arg4 ", and " arg5
1007 If the arguments are invalid,
1008 and in particular if
1010 contains set bits that are unrecognized
1011 or that correspond to a key not available on this platform,
1012 then the call fails with error
1016 Because the compiler or run-time environment
1017 may be using some or all of the keys,
1019 .B PR_PAC_RESET_KEYS
1020 may crash the calling process.
1021 The conditions for using it safely are complex and system-dependent.
1022 Don't use it unless you know what you are doing.
1024 For more information, see the kernel source file
1025 .I Documentation/arm64/pointer\-authentication.rst
1026 .\"commit b693d0b372afb39432e1c49ad7b3454855bc6bed
1028 .I Documentation/arm64/pointer\-authentication.txt
1030 .\" prctl PR_SET_PDEATHSIG
1032 .BR PR_SET_PDEATHSIG " (since Linux 2.1.57)"
1033 Set the parent-death signal
1034 of the calling process to \fIarg2\fP (either a signal value
1038 This is the signal that the calling process will get when its
1042 .\" https://bugzilla.kernel.org/show_bug.cgi?id=43300
1043 the "parent" in this case is considered to be the
1045 that created this process.
1046 In other words, the signal will be sent when that thread terminates
1048 .BR pthread_exit (3)),
1049 rather than after all of the threads in the parent process terminate.
1051 The parent-death signal is sent upon subsequent termination of the parent
1052 thread and also upon termination of each subreaper process
1053 (see the description of
1054 .B PR_SET_CHILD_SUBREAPER
1055 above) to which the caller is subsequently reparented.
1056 If the parent thread and all ancestor subreapers have already terminated
1058 .BR PR_SET_PDEATHSIG
1059 operation, then no parent-death signal is sent to the caller.
1061 The parent-death signal is process-directed (see
1063 and, if the child installs a handler using the
1070 argument of the handler contains the PID of the terminating parent process.
1072 The parent-death signal setting is cleared for the child of a
1075 (since Linux 2.4.36 / 2.6.23)
1076 .\" commit d2d56c5f51028cb9f3d800882eb6f4cbd3f9099f
1077 cleared when executing a set-user-ID or set-group-ID binary,
1078 or a binary that has associated capabilities (see
1079 .BR capabilities (7));
1080 otherwise, this value is preserved across
1082 The parent-death signal setting is also cleared upon changes to
1083 any of the following thread credentials:
1084 .\" FIXME capability changes can also trigger this; see
1085 .\" kernel/cred.c::commit_creds in the Linux 5.6 source.
1086 effective user ID, effective group ID, filesystem user ID,
1087 or filesystem group ID.
1088 .\" prctl PR_GET_PDEATHSIG
1090 .BR PR_GET_PDEATHSIG " (since Linux 2.3.15)"
1091 Return the current value of the parent process death signal,
1092 in the location pointed to by
1093 .IR "(int\ *) arg2" .
1094 .\" prctl PR_SET_PTRACER
1096 .BR PR_SET_PTRACER " (since Linux 3.4)"
1097 .\" commit 2d514487faf188938a4ee4fb3464eeecfbdcf8eb
1098 .\" commit bf06189e4d14641c0148bea16e9dd24943862215
1099 This is meaningful only when the Yama LSM is enabled and in mode 1
1100 ("restricted ptrace", visible via
1101 .IR /proc/sys/kernel/yama/ptrace_scope ).
1102 When a "ptracer process ID" is passed in \fIarg2\fP,
1103 the caller is declaring that the ptracer process can
1105 the calling process as if it were a direct process ancestor.
1108 operation replaces the previous "ptracer process ID".
1113 set to 0 clears the caller's "ptracer process ID".
1117 .BR PR_SET_PTRACER_ANY ,
1118 the ptrace restrictions introduced by Yama are effectively disabled for the
1121 For further information, see the kernel source file
1122 .IR Documentation/admin\-guide/LSM/Yama.rst
1123 .\" commit 90bb766440f2147486a2acc3e793d7b8348b0c22
1125 .IR Documentation/security/Yama.txt
1127 .\" prctl PR_SET_SECCOMP
1129 .BR PR_SET_SECCOMP " (since Linux 2.6.23)"
1130 .\" See http://thread.gmane.org/gmane.linux.kernel/542632
1131 .\" [PATCH 0 of 2] seccomp updates
1132 .\" andrea@cpushare.com
1133 Set the secure computing (seccomp) mode for the calling thread, to limit
1134 the available system calls.
1137 system call provides a superset of the functionality of
1138 .BR PR_SET_SECCOMP .
1140 The seccomp mode is selected via
1142 (The seccomp constants are defined in
1143 .IR <linux/seccomp.h> .)
1148 .BR SECCOMP_MODE_STRICT ,
1149 the only system calls that the thread is permitted to make are
1154 .BR exit_group (2)),
1157 Other system calls result in the delivery of a
1160 Strict secure computing mode is useful for number-crunching applications
1161 that may need to execute untrusted byte code,
1162 perhaps obtained by reading from a pipe or socket.
1163 This operation is available only
1164 if the kernel is configured with
1171 .BR SECCOMP_MODE_FILTER " (since Linux 3.5),"
1172 the system calls allowed are defined by a pointer
1173 to a Berkeley Packet Filter passed in
1175 This argument is a pointer to
1176 .IR "struct sock_fprog" ;
1177 it can be designed to filter
1178 arbitrary system calls and system call arguments.
1179 This mode is available only if the kernel is configured with
1180 .B CONFIG_SECCOMP_FILTER
1184 .BR SECCOMP_MODE_FILTER
1187 then the seccomp mode is inherited by children created by
1191 is permitted, then the seccomp mode is preserved across
1193 If the filters permit
1195 calls, then additional filters can be added;
1196 they are run in order until the first non-allow result is seen.
1198 For further information, see the kernel source file
1199 .IR Documentation/userspace\-api/seccomp_filter.rst
1200 .\" commit c061f33f35be0ccc80f4b8e0aea5dfd2ed7e01a3
1202 .IR Documentation/prctl/seccomp_filter.txt
1204 .\" prctl PR_GET_SECCOMP
1206 .BR PR_GET_SECCOMP " (since Linux 2.6.23)"
1207 Return (as the function result)
1208 the secure computing mode of the calling thread.
1209 If the caller is not in secure computing mode, this operation returns 0;
1210 if the caller is in strict secure computing mode, then the
1214 signal to be sent to the process.
1215 If the caller is in filter mode, and this system call is allowed by the
1216 seccomp filters, it returns 2; otherwise, the process is killed with a
1219 This operation is available only
1220 if the kernel is configured with
1224 Since Linux 3.8, the
1227 .IR /proc/[pid]/status
1228 file provides a method of obtaining the same information,
1229 without the risk that the process is killed; see
1231 .\" prctl PR_SET_SECUREBITS
1233 .BR PR_SET_SECUREBITS " (since Linux 2.6.26)"
1234 Set the "securebits" flags of the calling thread to the value supplied in
1237 .BR capabilities (7).
1238 .\" prctl PR_GET_SECUREBITS
1240 .BR PR_GET_SECUREBITS " (since Linux 2.6.26)"
1241 Return (as the function result)
1242 the "securebits" flags of the calling thread.
1244 .BR capabilities (7).
1245 .\" prctl PR_GET_SPECULATION_CTRL
1247 .BR PR_GET_SPECULATION_CTRL " (since Linux 4.17)"
1248 Return (as the function result)
1249 the state of the speculation misfeature specified in
1251 Currently, the only permitted value for this argument is
1252 .BR PR_SPEC_STORE_BYPASS
1253 (otherwise the call fails with the error
1256 The return value uses bits 0-3 with the following meaning:
1260 Mitigation can be controlled per thread by
1261 .BR PR_SET_SPECULATION_CTRL .
1264 The speculation feature is enabled, mitigation is disabled.
1267 The speculation feature is disabled, mitigation is enabled.
1269 .BR PR_SPEC_FORCE_DISABLE
1272 but cannot be undone.
1274 .BR PR_SPEC_DISABLE_NOEXEC " (since Linux 5.1)"
1276 .BR PR_SPEC_DISABLE ,
1277 but the state will be cleared on
1282 then the CPU is not affected by the speculation misfeature.
1286 is set, then per-thread control of the mitigation is available.
1289 for the speculation misfeature will fail.
1296 arguments must be specified as 0; otherwise the call fails with the error
1298 .\" prctl PR_SET_SPECULATION_CTRL
1300 .BR PR_SET_SPECULATION_CTRL " (since Linux 4.17)"
1301 .\" commit b617cfc858161140d69cc0b5cc211996b557a1c7
1302 .\" commit 356e4bfff2c5489e016fdb925adbf12a1e3950ee
1303 Sets the state of the speculation misfeature specified in
1305 The speculation-misfeature settings are per-thread attributes.
1312 .B PR_SPEC_STORE_BYPASS
1313 Set the state of the speculative store bypass misfeature.
1314 .\" commit 9137bb27e60e554dab694eafa4cca241fa3a694f
1316 .BR PR_SPEC_INDIRECT_BRANCH " (since Linux 4.20)"
1317 Set the state of the indirect branch speculation misfeature.
1322 does not have one of the above values,
1323 then the call fails with the error
1328 argument is used to hand in the control value,
1329 which is one of the following:
1333 The speculation feature is enabled, mitigation is disabled.
1336 The speculation feature is disabled, mitigation is enabled.
1338 .BR PR_SPEC_FORCE_DISABLE
1340 .BR PR_SPEC_DISABLE ,
1341 but cannot be undone.
1345 .BR PR_SPEC_ENABLE )
1346 with the same value for
1348 will fail with the error
1350 .\" commit 71368af9027f18fe5d1c6f372cfdff7e4bde8b48
1352 .BR PR_SPEC_DISABLE_NOEXEC " (since Linux 5.1)"
1354 .BR PR_SPEC_DISABLE ,
1355 but the state will be cleared on
1357 Currently only supported for
1360 .B PR_SPEC_STORE_BYPASS.
1363 Any unsupported value in
1365 will result in the call failing with the error
1372 arguments must be specified as 0; otherwise the call fails with the error
1375 The speculation feature can also be controlled by the
1376 .B spec_store_bypass_disable
1378 This parameter may enforce a read-only policy which will result in the
1380 call failing with the error
1382 For further details, see the kernel source file
1383 .IR Documentation/admin\-guide/kernel\-parameters.txt .
1384 .\" prctl PR_SVE_SET_VL
1385 .\" commit 2d2123bc7c7f843aa9db87720de159a049839862
1386 .\" linux-5.6/Documentation/arm64/sve.rst
1388 .BR PR_SVE_SET_VL " (since Linux 4.15, only on arm64)"
1389 Configure the thread's SVE vector length,
1393 .IR arg3 ", " arg4 ", and " arg5
1399 .B PR_SVE_VL_LEN_MASK
1400 must be set to the desired vector length in bytes.
1401 This is interpreted as an upper bound:
1402 the kernel will select the greatest available vector length
1403 that does not exceed the value specified.
1404 In particular, specifying
1407 .I <asm/sigcontext.h>)
1409 .B PR_SVE_VL_LEN_MASK
1410 bits requests the maximum supported vector length.
1412 In addition, the other bits of
1414 must be set to one of the following combinations of flags:
1418 Perform the change immediately.
1422 the vector length will be reset to the value configured in
1423 .IR /proc/sys/abi/sve_default_vector_length .
1425 .B PR_SVE_VL_INHERIT
1426 Perform the change immediately.
1429 calls will preserve the new vector length.
1431 .B PR_SVE_SET_VL_ONEXEC
1432 Defer the change, so that it is performed at the next
1437 calls will reset the vector length to the value configured in
1438 .IR /proc/sys/abi/sve_default_vector_length .
1440 .B "PR_SVE_SET_VL_ONEXEC | PR_SVE_VL_INHERIT"
1441 Defer the change, so that it is performed at the next
1446 calls will preserve the new vector length.
1450 any previously pending deferred change is canceled.
1452 The call fails with error
1454 if SVE is not supported on the platform, if
1456 is unrecognized or invalid, or the value in the bits of
1459 .B PR_SVE_VL_LEN_MASK
1460 is outside the range
1461 .BR SVE_VL_MIN .. SVE_VL_MAX
1462 or is not a multiple of 16.
1465 a nonnegative value is returned that describes the
1469 .B PR_SVE_SET_VL_ONEXEC
1472 then the configuration described by the return value
1473 will take effect at the next
1475 Otherwise, the configuration is already in effect when the
1478 In either case, the value is encoded in the same way as the return value of
1480 Note that there is no explicit flag in the return value
1482 .BR PR_SVE_SET_VL_ONEXEC .
1484 The configuration (including any pending deferred change)
1490 For more information, see the kernel source file
1491 .I Documentation/arm64/sve.rst
1492 .\"commit b693d0b372afb39432e1c49ad7b3454855bc6bed
1494 .I Documentation/arm64/sve.txt
1498 Because the compiler or run-time environment
1499 may be using SVE, using this call without the
1500 .B PR_SVE_SET_VL_ONEXEC
1501 flag may crash the calling process.
1502 The conditions for using it safely are complex and system-dependent.
1503 Don't use it unless you really know what you are doing.
1504 .\" prctl PR_SVE_GET_VL
1506 .BR PR_SVE_GET_VL " (since Linux 4.15, only on arm64)"
1507 Get the thread's current SVE vector length configuration.
1510 .IR arg2 ", " arg3 ", " arg4 ", and " arg5
1513 Provided that the kernel and platform support SVE,
1514 this operation always succeeds,
1515 returning a nonnegative value that describes the
1518 The bits corresponding to
1519 .B PR_SVE_VL_LEN_MASK
1520 contain the currently configured vector length in bytes.
1521 The bit corresponding to
1522 .B PR_SVE_VL_INHERIT
1523 indicates whether the vector length will be inherited
1527 Note that there is no way to determine whether there is
1528 a pending vector length change that has not yet taken effect.
1530 For more information, see the kernel source file
1531 .I Documentation/arm64/sve.rst
1532 .\"commit b693d0b372afb39432e1c49ad7b3454855bc6bed
1534 .I Documentation/arm64/sve.txt
1537 .\" prctl PR_SET_SYSCALL_USER_DISPATCH
1538 .\" commit 1446e1df9eb183fdf81c3f0715402f1d7595d4
1539 .BR PR_SET_SYSCALL_USER_DISPATCH " (since Linux 5.11, x86 only)"
1540 Configure the Syscall User Dispatch mechanism
1541 for the calling thread.
1542 This mechanism allows an application
1543 to selectively intercept system calls
1544 so that they can be handled within the application itself.
1545 Interception takes the form of a thread-directed
1547 signal that is delivered to the thread
1548 when it makes a system call.
1550 the system call is not executed by the kernel.
1552 To enable this mechanism,
1555 .BR PR_SYS_DISPATCH_ON .
1556 Once enabled, further system calls will be selectively intercepted,
1557 depending on a control variable provided by user space.
1562 respectively identify the
1566 of a single contiguous memory region in the process address space
1567 from where system calls are always allowed to be executed,
1568 regardless of the control variable.
1569 (Typically, this area would include the area of memory
1570 containing the C library.)
1573 points to a char-sized variable
1574 that is a fast switch to allow/block system call execution
1575 without the overhead of doing another system call
1576 to re-configure Syscall User Dispatch.
1577 This control variable can either be set to
1578 .B SYSCALL_DISPATCH_FILTER_BLOCK
1579 to block system calls from executing
1581 .B SYSCALL_DISPATCH_FILTER_ALLOW
1582 to temporarily allow them to be executed.
1583 This value is checked by the kernel
1584 on every system call entry,
1585 and any unexpected value will raise
1589 killing the application.
1591 When a system call is intercepted,
1592 the kernel sends a thread-directed
1594 signal to the triggering thread.
1595 Various fields will be set in the
1599 associated with the signal:
1607 will show the address of the system call instruction.
1612 will indicate which system call was attempted.
1616 .BR SYS_USER_DISPATCH .
1622 The program counter will be as though the system call happened
1623 (i.e., the program counter will not point to the system call instruction).
1625 When the signal handler returns to the kernel,
1626 the system call completes immediately
1627 and returns to the calling thread,
1628 without actually being executed.
1630 (i.e., when emulating the system call on user space.),
1631 the signal handler should set the system call return value
1633 by modifying the register context stored in the
1635 argument of the signal handler.
1641 for more information.
1646 .BR PR_SYS_DISPATCH_OFF ,
1647 Syscall User Dispatch is disabled for that thread.
1648 the remaining arguments must be set to 0.
1650 The setting is not preserved across
1656 For more information,
1657 see the kernel source file
1658 .IR Documentation/admin-guide/syscall-user-dispatch.rst
1659 .\" prctl PR_SET_TAGGED_ADDR_CTRL
1660 .\" commit 63f0c60379650d82250f22e4cf4137ef3dc4f43d
1662 .BR PR_SET_TAGGED_ADDR_CTRL " (since Linux 5.4, only on arm64)"
1663 Controls support for passing tagged user-space addresses to the kernel
1664 (i.e., addresses where bits 56\(em63 are not all zero).
1666 The level of support is selected by
1668 which can be one of the following:
1672 Addresses that are passed
1673 for the purpose of being dereferenced by the kernel
1676 .B PR_TAGGED_ADDR_ENABLE
1677 Addresses that are passed
1678 for the purpose of being dereferenced by the kernel
1679 may be tagged, with the exceptions summarized below.
1682 The remaining arguments
1683 .IR arg3 ", " arg4 ", and " arg5
1685 .\" Enforcement added in
1686 .\" commit 3e91ec89f527b9870fe42dcbdb74fd389d123a95
1688 On success, the mode specified in
1690 is set for the calling thread and the return value is 0.
1691 If the arguments are invalid,
1692 the mode specified in
1695 or if this feature is unsupported by the kernel
1697 .IR /proc/sys/abi/tagged_addr_disabled ,
1698 the call fails with the error
1702 .BR prctl ( PR_SET_TAGGED_ADDR_CTRL ,
1706 then all addresses passed to the kernel must be untagged.
1708 Irrespective of which mode is set,
1709 addresses passed to certain interfaces
1710 must always be untagged:
1722 (Prior to Linux 5.6 these accepted tagged addresses,
1723 but the behaviour may not be what you expect.
1726 \(oqpolymorphic\(cq interfaces
1727 that accept pointers to arbitrary types cast to a
1729 or other generic type, specifically
1734 (only certain specific
1736 options allow tagged addresses).
1739 This list of exclusions may shrink
1740 when moving from one kernel version to a later kernel version.
1741 While the kernel may make some guarantees
1742 for backwards compatibility reasons,
1743 for the purposes of new software
1744 the effect of passing tagged addresses to these interfaces
1747 The mode set by this call is inherited across
1751 The mode is reset by
1754 (i.e., tagged addresses not permitted in the user/kernel ABI).
1756 For more information, see the kernel source file
1757 .IR Documentation/arm64/tagged\-address\-abi.rst .
1760 This call is primarily intended for use by the run-time environment.
1762 .B PR_SET_TAGGED_ADDR_CTRL
1763 call elsewhere may crash the calling process.
1764 The conditions for using it safely are complex and system-dependent.
1765 Don't use it unless you know what you are doing.
1766 .\" prctl PR_GET_TAGGED_ADDR_CTRL
1767 .\" commit 63f0c60379650d82250f22e4cf4137ef3dc4f43d
1769 .BR PR_GET_TAGGED_ADDR_CTRL " (since Linux 5.4, only on arm64)"
1770 Returns the current tagged address mode
1771 for the calling thread.
1774 .IR arg2 ", " arg3 ", " arg4 ", and " arg5
1777 If the arguments are invalid
1778 or this feature is disabled or unsupported by the kernel,
1782 .BR prctl ( PR_GET_TAGGED_ADDR_CTRL ,
1786 then this feature is definitely either unsupported,
1788 .IR /proc/sys/abi/tagged_addr_disabled .
1790 all addresses passed to the kernel must be untagged.
1792 Otherwise, the call returns a nonnegative value
1793 describing the current tagged address mode,
1794 encoded in the same way as the
1797 .BR PR_SET_TAGGED_ADDR_CTRL .
1799 For more information, see the kernel source file
1800 .IR Documentation/arm64/tagged\-address\-abi.rst .
1802 .\" prctl PR_TASK_PERF_EVENTS_DISABLE
1804 .BR PR_TASK_PERF_EVENTS_DISABLE " (since Linux 2.6.31)"
1805 Disable all performance counters attached to the calling process,
1806 regardless of whether the counters were created by
1807 this process or another process.
1808 Performance counters created by the calling process for other
1809 processes are unaffected.
1810 For more information on performance counters, see the Linux kernel source file
1811 .IR tools/perf/design.txt .
1814 .BR PR_TASK_PERF_COUNTERS_DISABLE ;
1815 .\" commit 1d1c7ddbfab358445a542715551301b7fc363e28
1816 renamed (retaining the same numerical value)
1819 .\" prctl PR_TASK_PERF_EVENTS_ENABLE
1821 .BR PR_TASK_PERF_EVENTS_ENABLE " (since Linux 2.6.31)"
1823 .BR PR_TASK_PERF_EVENTS_DISABLE ;
1824 enable performance counters attached to the calling process.
1827 .BR PR_TASK_PERF_COUNTERS_ENABLE ;
1828 .\" commit 1d1c7ddbfab358445a542715551301b7fc363e28
1830 .\" commit cdd6c482c9ff9c55475ee7392ec8f672eddb7be6
1833 .\" prctl PR_SET_THP_DISABLE
1835 .BR PR_SET_THP_DISABLE " (since Linux 3.15)"
1836 .\" commit a0715cc22601e8830ace98366c0c2bd8da52af52
1837 Set the state of the "THP disable" flag for the calling thread.
1840 has a nonzero value, the flag is set, otherwise it is cleared.
1841 Setting this flag provides a method
1842 for disabling transparent huge pages
1843 for jobs where the code cannot be modified, and using a malloc hook with
1845 is not an option (i.e., statically allocated data).
1846 The setting of the "THP disable" flag is inherited by a child created via
1848 and is preserved across
1850 .\" prctl PR_GET_THP_DISABLE
1852 .BR PR_GET_THP_DISABLE " (since Linux 3.15)"
1853 Return (as the function result) the current setting of the "THP disable"
1854 flag for the calling thread:
1855 either 1, if the flag is set, or 0, if it is not.
1856 .\" prctl PR_GET_TID_ADDRESS
1858 .BR PR_GET_TID_ADDRESS " (since Linux 3.5)"
1859 .\" commit 300f786b2683f8bb1ec0afb6e1851183a479c86d
1863 .BR set_tid_address (2)
1866 .B CLONE_CHILD_CLEARTID
1867 flag, in the location pointed to by
1868 .IR "(int\ **)\ arg2" .
1869 This feature is available only if the kernel is built with the
1870 .BR CONFIG_CHECKPOINT_RESTORE
1874 system call does not have a compat implementation for
1875 the AMD64 x32 and MIPS n32 ABIs,
1876 and the kernel writes out a pointer using the kernel's pointer size,
1877 this operation expects a user-space buffer of 8 (not 4) bytes on these ABIs.
1878 .\" prctl PR_SET_TIMERSLACK
1880 .BR PR_SET_TIMERSLACK " (since Linux 2.6.28)"
1881 .\" See https://lwn.net/Articles/369549/
1882 .\" commit 6976675d94042fbd446231d1bd8b7de71a980ada
1883 Each thread has two associated timer slack values:
1884 a "default" value, and a "current" value.
1885 This operation sets the "current" timer slack value for the calling thread.
1887 is an unsigned long value, then maximum "current" value is ULONG_MAX and
1888 the minimum "current" value is 1.
1889 If the nanosecond value supplied in
1891 is greater than zero, then the "current" value is set to this value.
1895 the "current" timer slack is reset to the
1896 thread's "default" timer slack value.
1898 The "current" timer slack is used by the kernel to group timer expirations
1899 for the calling thread that are close to one another;
1900 as a consequence, timer expirations for the thread may be
1901 up to the specified number of nanoseconds late (but will never expire early).
1902 Grouping timer expirations can help reduce system power consumption
1903 by minimizing CPU wake-ups.
1905 The timer expirations affected by timer slack are those set by
1911 .BR epoll_pwait (2),
1912 .BR clock_nanosleep (2),
1916 (and thus the library functions implemented via futexes, including
1917 .\" List obtained by grepping for futex usage in glibc source
1918 .BR pthread_cond_timedwait (3),
1919 .BR pthread_mutex_timedlock (3),
1920 .BR pthread_rwlock_timedrdlock (3),
1921 .BR pthread_rwlock_timedwrlock (3),
1923 .BR sem_timedwait (3)).
1925 Timer slack is not applied to threads that are scheduled under
1926 a real-time scheduling policy (see
1927 .BR sched_setscheduler (2)).
1929 When a new thread is created,
1930 the two timer slack values are made the same as the "current" value
1931 of the creating thread.
1932 Thereafter, a thread can adjust its "current" timer slack value via
1933 .BR PR_SET_TIMERSLACK .
1934 The "default" value can't be changed.
1935 The timer slack values of
1937 (PID 1), the ancestor of all processes,
1938 are 50,000 nanoseconds (50 microseconds).
1939 The timer slack value is inherited by a child created via
1941 and is preserved across
1944 Since Linux 4.6, the "current" timer slack value of any process
1945 can be examined and changed via the file
1946 .IR /proc/[pid]/timerslack_ns .
1949 .\" prctl PR_GET_TIMERSLACK
1951 .BR PR_GET_TIMERSLACK " (since Linux 2.6.28)"
1952 Return (as the function result)
1953 the "current" timer slack value of the calling thread.
1954 .\" prctl PR_SET_TIMING
1956 .BR PR_SET_TIMING " (since Linux 2.6.0)"
1957 .\" Precisely: Linux 2.6.0-test4
1958 Set whether to use (normal, traditional) statistical process timing or
1959 accurate timestamp-based process timing, by passing
1960 .B PR_TIMING_STATISTICAL
1963 .B PR_TIMING_TIMESTAMP
1966 .B PR_TIMING_TIMESTAMP
1967 is not currently implemented
1968 (attempting to set this mode will yield the error
1970 .\" PR_TIMING_TIMESTAMP doesn't do anything in 2.6.26-rc8,
1971 .\" and looking at the patch history, it appears
1972 .\" that it never did anything.
1973 .\" prctl PR_GET_TIMING
1975 .BR PR_GET_TIMING " (since Linux 2.6.0)"
1976 .\" Precisely: Linux 2.6.0-test4
1977 Return (as the function result) which process timing method is currently
1979 .\" prctl PR_SET_TSC
1981 .BR PR_SET_TSC " (since Linux 2.6.26, x86 only)"
1982 Set the state of the flag determining whether the timestamp counter
1983 can be read by the process.
1988 to allow it to be read, or
1992 when the process tries to read the timestamp counter.
1993 .\" prctl PR_GET_TSC
1995 .BR PR_GET_TSC " (since Linux 2.6.26, x86 only)"
1996 Return the state of the flag determining whether the timestamp counter
1998 in the location pointed to by
1999 .IR "(int\ *) arg2" .
2000 .\" prctl PR_SET_UNALIGN
2003 (Only on: ia64, since Linux 2.3.48; parisc, since Linux 2.6.15;
2004 PowerPC, since Linux 2.6.18; Alpha, since Linux 2.6.22;
2005 .\" sh: 94ea5e449ae834af058ef005d16a8ad44fcf13d6
2006 .\" tile: 2f9ac29eec71a696cb0dcc5fb82c0f8d4dac28c9
2007 sh, since Linux 2.6.34; tile, since Linux 3.12)
2008 Set unaligned access control bits to \fIarg2\fP.
2010 \fBPR_UNALIGN_NOPRINT\fP to silently fix up unaligned user accesses,
2011 or \fBPR_UNALIGN_SIGBUS\fP to generate
2013 on unaligned user access.
2014 Alpha also supports an additional flag with the value
2015 of 4 and no corresponding named constant,
2016 which instructs kernel to not fix up
2017 unaligned accesses (it is analogous to providing the
2023 system call on Tru64).
2024 .\" prctl PR_GET_UNALIGN
2029 for information on versions and architectures.)
2030 Return unaligned access control bits, in the location pointed to by
2031 .IR "(unsigned int\ *) arg2" .
2034 .BR PR_CAP_AMBIENT + PR_CAP_AMBIENT_IS_SET ,
2035 .BR PR_CAPBSET_READ ,
2036 .BR PR_GET_DUMPABLE ,
2037 .BR PR_GET_FP_MODE ,
2038 .BR PR_GET_IO_FLUSHER ,
2039 .BR PR_GET_KEEPCAPS ,
2040 .BR PR_MCE_KILL_GET ,
2041 .BR PR_GET_NO_NEW_PRIVS ,
2042 .BR PR_GET_SECUREBITS ,
2043 .BR PR_GET_SPECULATION_CTRL ,
2046 .BR PR_GET_TAGGED_ADDR_CTRL ,
2047 .BR PR_GET_THP_DISABLE ,
2049 .BR PR_GET_TIMERSLACK ,
2052 return the nonnegative values described above.
2055 values return 0 on success.
2056 On error, \-1 is returned, and
2058 is set to indicate the error.
2068 .BR SECCOMP_MODE_FILTER ,
2069 but the process does not have the
2071 capability or has not set the
2073 attribute (see the discussion of
2074 .BR PR_SET_NO_NEW_PRIVS
2084 .BR PR_SET_MM_EXE_FILE ,
2085 the file is not executable.
2093 .BR PR_SET_MM_EXE_FILE ,
2094 and the file descriptor passed in
2104 .BR PR_SET_MM_EXE_FILE ,
2105 and this the second attempt to change the
2107 symbolic link, which is prohibited.
2111 is an invalid address.
2116 .BR PR_SET_SECCOMP ,
2119 .BR SECCOMP_MODE_FILTER ,
2120 the system was built with
2121 .BR CONFIG_SECCOMP_FILTER ,
2124 is an invalid address.
2129 .B PR_SET_SYSCALL_USER_DISPATCH
2132 has an invalid address.
2138 or not supported on this system.
2150 arguments were not specified as zero.
2154 is not valid value for this
2162 .BR PR_GET_SECCOMP ,
2163 and the kernel was not configured with
2164 .BR CONFIG_SECCOMP .
2169 .BR PR_SET_SECCOMP ,
2172 .BR SECCOMP_MODE_FILTER ,
2173 and the kernel was not configured with
2174 .BR CONFIG_SECCOMP_FILTER .
2180 and one of the following is true
2191 (the limit on the size of the user address space for this architecture);
2195 .BR PR_SET_MM_START_CODE ,
2196 .BR PR_SET_MM_END_CODE ,
2197 .BR PR_SET_MM_START_DATA ,
2198 .BR PR_SET_MM_END_DATA ,
2200 .BR PR_SET_MM_START_STACK ,
2201 and the permissions of the corresponding memory area are not as required;
2205 .BR PR_SET_MM_START_BRK
2210 is less than or equal to the end of the data segment
2211 or specifies a value that would cause the
2213 resource limit to be exceeded.
2223 .BR PR_SET_PTRACER_ANY ,
2224 or the PID of an existing process.
2232 is not a valid signal number.
2241 .B SUID_DUMP_DISABLE
2243 .BR SUID_DUMP_USER .
2252 .BR PR_TIMING_STATISTICAL .
2257 .BR PR_SET_NO_NEW_PRIVS
2271 .BR PR_GET_NO_NEW_PRIVS
2283 .BR PR_SET_THP_DISABLE
2294 .BR PR_GET_THP_DISABLE
2307 and an unused argument
2312 .BR PR_CAP_AMBIENT_CLEAR_ALL ,
2316 has an invalid value;
2320 .BR PR_CAP_AMBIENT_LOWER ,
2321 .BR PR_CAP_AMBIENT_RAISE ,
2323 .BR PR_CAP_AMBIENT_IS_SET
2326 does not specify a valid capability.
2331 .BR PR_GET_SPECULATION_CTRL
2333 .BR PR_SET_SPECULATION_CTRL
2334 and unused arguments to
2340 .B PR_PAC_RESET_KEYS
2341 and the arguments are invalid or unsupported.
2342 See the description of
2343 .B PR_PAC_RESET_KEYS
2350 and the arguments are invalid or unsupported,
2351 or SVE is not available on this platform.
2352 See the description of
2360 and SVE is not available on this platform.
2365 .B PR_SET_SYSCALL_USER_DISPATCH
2366 and one of the following is true:
2371 .B PR_SYS_DISPATCH_OFF
2372 and the remaining arguments are not 0;
2376 .B PR_SYS_DISPATCH_ON
2377 and the memory range specified is outside the
2378 address space of the process.
2387 .BR PR_SET_TAGGED_ADDR_CTRL
2388 and the arguments are invalid or unsupported.
2389 See the description of
2390 .B PR_SET_TAGGED_ADDR_CTRL
2396 .BR PR_GET_TAGGED_ADDR_CTRL
2397 and the arguments are invalid or unsupported.
2398 See the description of
2399 .B PR_GET_TAGGED_ADDR_CTRL
2405 .BR PR_SET_SPECULATION_CTRL
2406 the kernel or CPU does not support the requested speculation misfeature.
2411 .BR PR_MPX_ENABLE_MANAGEMENT
2413 .BR PR_MPX_DISABLE_MANAGEMENT
2414 and the kernel or the CPU does not support MPX management.
2415 Check that the kernel and processor have MPX support.
2420 .BR PR_SET_SPECULATION_CTRL
2421 implies that the control of the selected speculation misfeature is not possible.
2423 .BR PR_GET_SPECULATION_CTRL
2424 for the bit fields to determine which option is available.
2432 has an invalid or unsupported value.
2437 .BR PR_SET_SECUREBITS ,
2438 and the caller does not have the
2441 or tried to unset a "locked" flag,
2442 or tried to set a flag whose corresponding locked flag was set
2444 .BR capabilities (7)).
2449 .BR PR_SET_SPECULATION_CTRL
2450 wherein the speculation was disabled with
2451 .B PR_SPEC_FORCE_DISABLE
2452 and caller tried to enable it again.
2457 .BR PR_SET_KEEPCAPS ,
2459 .B SECBIT_KEEP_CAPS_LOCKED
2462 .BR capabilities (7)).
2467 .BR PR_CAPBSET_DROP ,
2468 and the caller does not have the
2476 and the caller does not have the
2487 .BR PR_CAP_AMBIENT_RAISE ,
2488 but either the capability specified in
2490 is not present in the process's permitted and inheritable capability sets,
2492 .B PR_CAP_AMBIENT_LOWER
2493 securebit has been set.
2498 .BR PR_SET_SPECULATION_CTRL
2502 .BR PR_SPEC_ENABLE ,
2503 .BR PR_SPEC_DISABLE ,
2504 .BR PR_SPEC_FORCE_DISABLE ,
2506 .BR PR_SPEC_DISABLE_NOEXEC .
2510 system call was introduced in Linux 2.1.57.
2511 .\" The library interface was added in glibc 2.0.6
2513 This call is Linux-specific.
2516 system call (also introduced in Linux 2.1.44
2517 as irix_prctl on the MIPS architecture),
2522 .BI "ptrdiff_t prctl(int " option ", int " arg2 ", int " arg3 );
2526 and options to get the maximum number of processes per user,
2527 get the maximum number of processors the calling process can use,
2528 find out whether a specified process is currently blocked,
2529 get or set the maximum stack size, and so on.