1 ===================================
2 Documentation for /proc/sys/kernel/
3 ===================================
4
5 .. See scripts/check-sysctl-docs to keep this
6
7
8 Copyright (c) 1998, 1999, Rik van Riel <riel@n
9
10 Copyright (c) 2009, Shen Feng<shen@cn.fu
11
12 For general info and legal blurb, please look
13 Documentation/admin-guide/sysctl/index.rst.
14
15 ----------------------------------------------
16
17 This file contains documentation for the sysct
18 ``/proc/sys/kernel/``.
19
20 The files in this directory can be used to tun
21 miscellaneous and general things in the operat
22 kernel. Since some of the files *can* be used
23 system, it is advisable to read both documenta
24 before actually making adjustments.
25
26 Currently, these files might (depending on you
27 show up in ``/proc/sys/kernel``:
28
29 .. contents:: :local:
30
31
32 acct
33 ====
34
35 ::
36
37 highwater lowwater frequency
38
39 If BSD-style process accounting is enabled the
40 its behaviour. If free space on filesystem whe
41 goes below ``lowwater``\ % accounting suspends
42 above ``highwater``\ % accounting resumes. ``f
43 how often do we check the amount of free space
44 seconds). Default:
45
46 ::
47
48 4 2 30
49
50 That is, suspend accounting if free space drop
51 if it increases to at least 4%; consider infor
52 free space valid for 30 seconds.
53
54
55 acpi_video_flags
56 ================
57
58 See Documentation/power/video.rst. This allows
59 in a similar fashion to the ``acpi_sleep`` ker
60 combining the following values:
61
62 = =======
63 1 s3_bios
64 2 s3_mode
65 4 s3_beep
66 = =======
67
68 arch
69 ====
70
71 The machine hardware name, the same output as
72 (e.g. ``x86_64`` or ``aarch64``).
73
74 auto_msgmni
75 ===========
76
77 This variable has no effect and may be removed
78 releases. Reading it always returns 0.
79 Up to Linux 3.17, it enabled/disabled automati
80 `msgmni`_
81 upon memory add/remove or upon IPC namespace c
82 Echoing "1" into this file enabled msgmni auto
83 Echoing "0" turned it off. The default value w
84
85
86 bootloader_type (x86 only)
87 ==========================
88
89 This gives the bootloader type number as indic
90 shifted left by 4, and OR'd with the low four
91 version. The reason for this encoding is that
92 ``type_of_loader`` field in the kernel header;
93 backwards compatibility. That is, if the full
94 is 0x15 and the full version number is 0x234,
95 the value 340 = 0x154.
96
97 See the ``type_of_loader`` and ``ext_loader_ty
98 Documentation/arch/x86/boot.rst for additional
99
100
101 bootloader_version (x86 only)
102 =============================
103
104 The complete bootloader version number. In th
105 file will contain the value 564 = 0x234.
106
107 See the ``type_of_loader`` and ``ext_loader_ve
108 Documentation/arch/x86/boot.rst for additional
109
110
111 bpf_stats_enabled
112 =================
113
114 Controls whether the kernel should collect sta
115 (total time spent running, number of times run
116 statistics causes a slight reduction in perfor
117 run. The statistics can be seen using ``bpftoo
118
119 = ===================================
120 0 Don't collect statistics (default).
121 1 Collect statistics.
122 = ===================================
123
124
125 cad_pid
126 =======
127
128 This is the pid which will be signalled on reb
129 Ctrl-Alt-Delete). Writing a value to this file
130 correspond to a running process will result in
131
132 See also `ctrl-alt-del`_.
133
134
135 cap_last_cap
136 ============
137
138 Highest valid capability of the running kernel
139 ``CAP_LAST_CAP`` from the kernel.
140
141
142 .. _core_pattern:
143
144 core_pattern
145 ============
146
147 ``core_pattern`` is used to specify a core dum
148
149 * max length 127 characters; default value is
150 * ``core_pattern`` is used as a pattern templa
151 filename; certain string patterns (beginning
152 substituted with their actual values.
153 * backward compatibility with ``core_uses_pid`
154
155 If ``core_pattern`` does not include "
156 and ``core_uses_pid`` is set, then .PI
157 the filename.
158
159 * corename format specifiers
160
161 ======== ======================
162 %<NUL> '%' is dropped
163 %% output one '%'
164 %p pid
165 %P global pid (init PID n
166 %i tid
167 %I global tid (init PID n
168 %u uid (in initial user n
169 %g gid (in initial user n
170 %d dump mode, matches ``P
171 ``/proc/sys/fs/suid_du
172 %s signal number
173 %t UNIX time of dump
174 %h hostname
175 %e executable filename (m
176 %f executable filename
177 %E executable path
178 %c maximum size of core f
179 %C CPU the task ran on
180 %<OTHER> both are dropped
181 ======== ======================
182
183 * If the first character of the pattern is a '
184 the rest of the pattern as a command to run.
185 written to the standard input of that progra
186
187
188 core_pipe_limit
189 ===============
190
191 This sysctl is only applicable when `core_patt
192 pipe core files to a user space helper (when t
193 ``core_pattern`` is a '|', see above).
194 When collecting cores via a pipe to an applica
195 useful for the collecting application to gathe
196 crashing process from its ``/proc/pid`` direct
197 In order to do this safely, the kernel must wa
198 process to exit, so as not to remove the crash
199 prematurely.
200 This in turn creates the possibility that a mi
201 collecting process can block the reaping of a
202 by never exiting.
203 This sysctl defends against that.
204 It defines how many concurrent crashing proces
205 space applications in parallel.
206 If this value is exceeded, then those crashing
207 value are noted via the kernel log and their c
208 0 is a special value, indicating that unlimite
209 captured in parallel, but that no waiting will
210 collecting process is not guaranteed access to
211 pid>/``).
212 This value defaults to 0.
213
214
215 core_uses_pid
216 =============
217
218 The default coredump filename is "core". By s
219 ``core_uses_pid`` to 1, the coredump filename
220 If `core_pattern`_ does not include "%p" (defa
221 and ``core_uses_pid`` is set, then .PID will b
222 the filename.
223
224
225 ctrl-alt-del
226 ============
227
228 When the value in this file is 0, ctrl-alt-del
229 sent to the ``init(1)`` program to handle a gr
230 When, however, the value is > 0, Linux's react
231 Nerve Pinch (tm) will be an immediate reboot,
232 syncing its dirty buffers.
233
234 Note:
235 when a program (like dosemu) has the keyboar
236 mode, the ctrl-alt-del is intercepted by the
237 ever reaches the kernel tty layer, and it's
238 to decide what to do with it.
239
240
241 dmesg_restrict
242 ==============
243
244 This toggle indicates whether unprivileged use
245 from using ``dmesg(8)`` to view messages from
246 buffer.
247 When ``dmesg_restrict`` is set to 0 there are
248 When ``dmesg_restrict`` is set to 1, users mus
249 ``CAP_SYSLOG`` to use ``dmesg(8)``.
250
251 The kernel config option ``CONFIG_SECURITY_DME
252 default value of ``dmesg_restrict``.
253
254
255 domainname & hostname
256 =====================
257
258 These files can be used to set the NIS/YP doma
259 hostname of your box in exactly the same way a
260 domainname and hostname, i.e.::
261
262 # echo "darkstar" > /proc/sys/kernel/h
263 # echo "mydomain" > /proc/sys/kernel/d
264
265 has the same effect as::
266
267 # hostname "darkstar"
268 # domainname "mydomain"
269
270 Note, however, that the classic darkstar.frop.
271 hostname "darkstar" and DNS (Internet Domain N
272 domainname "frop.org", not to be confused with
273 Information Service) or YP (Yellow Pages) doma
274 domain names are in general different. For a d
275 see the ``hostname(1)`` man page.
276
277
278 firmware_config
279 ===============
280
281 See Documentation/driver-api/firmware/fallback
282
283 The entries in this directory allow the firmwa
284 fallback to be controlled:
285
286 * ``force_sysfs_fallback``, when set to 1, for
287 fallback;
288 * ``ignore_sysfs_fallback``, when set to 1, ig
289
290
291 ftrace_dump_on_oops
292 ===================
293
294 Determines whether ``ftrace_dump()`` should be
295 kernel panic). This will output the contents o
296 the console. This is very useful for capturin
297 crashes and outputting them to a serial consol
298
299 ======================= ======================
300 0 Disabled (default).
301 1 Dump buffers of all CP
302 2(orig_cpu) Dump the buffer of the
303 oops.
304 <instance> Dump the specific inst
305 <instance>=2(orig_cpu) Dump the specific inst
306 that triggered the oop
307 ======================= ======================
308
309 Multiple instance dump is also supported, and
310 by commas. If global buffer also needs to be d
311 the dump mode (1/2/orig_cpu) first for global
312
313 So for example to dump "foo" and "bar" instanc
314 user can::
315
316 echo "foo,bar" > /proc/sys/kernel/ftrace_dum
317
318 To dump global buffer and "foo" instance buffe
319 CPUs along with the "bar" instance buffer on C
320 oops, user can::
321
322 echo "1,foo,bar=2" > /proc/sys/kernel/ftrace
323
324 ftrace_enabled, stack_tracer_enabled
325 ====================================
326
327 See Documentation/trace/ftrace.rst.
328
329
330 hardlockup_all_cpu_backtrace
331 ============================
332
333 This value controls the hard lockup detector b
334 lockup condition is detected as to whether or
335 debug information. If enabled, arch-specific a
336 will be initiated.
337
338 = ============================================
339 0 Do nothing. This is the default behavior.
340 1 On detection capture more debug information.
341 = ============================================
342
343
344 hardlockup_panic
345 ================
346
347 This parameter can be used to control whether
348 when a hard lockup is detected.
349
350 = ===========================
351 0 Don't panic on hard lockup.
352 1 Panic on hard lockup.
353 = ===========================
354
355 See Documentation/admin-guide/lockup-watchdogs
356 This can also be set using the nmi_watchdog ke
357
358
359 hotplug
360 =======
361
362 Path for the hotplug policy agent.
363 Default value is ``CONFIG_UEVENT_HELPER_PATH``
364 to the empty string.
365
366 This file only exists when ``CONFIG_UEVENT_HEL
367 modern systems rely exclusively on the netlink
368 don't need this.
369
370
371 hung_task_all_cpu_backtrace
372 ===========================
373
374 If this option is set, the kernel will send an
375 their backtraces when a hung task is detected.
376 CONFIG_DETECT_HUNG_TASK and CONFIG_SMP are ena
377
378 0: Won't show all CPUs backtraces when a hung
379 This is the default behavior.
380
381 1: Will non-maskably interrupt all CPUs and du
382 a hung task is detected.
383
384
385 hung_task_panic
386 ===============
387
388 Controls the kernel's behavior when a hung tas
389 This file shows up if ``CONFIG_DETECT_HUNG_TAS
390
391 = ============================================
392 0 Continue operation. This is the default beha
393 1 Panic immediately.
394 = ============================================
395
396
397 hung_task_check_count
398 =====================
399
400 The upper bound on the number of tasks that ar
401 This file shows up if ``CONFIG_DETECT_HUNG_TAS
402
403
404 hung_task_timeout_secs
405 ======================
406
407 When a task in D state did not get scheduled
408 for more than this value report a warning.
409 This file shows up if ``CONFIG_DETECT_HUNG_TAS
410
411 0 means infinite timeout, no checking is done.
412
413 Possible values to set are in range {0:``LONG_
414
415
416 hung_task_check_interval_secs
417 =============================
418
419 Hung task check interval. If hung task checkin
420 (see `hung_task_timeout_secs`_), the check is
421 ``hung_task_check_interval_secs`` seconds.
422 This file shows up if ``CONFIG_DETECT_HUNG_TAS
423
424 0 (default) means use ``hung_task_timeout_secs
425 interval.
426
427 Possible values to set are in range {0:``LONG_
428
429
430 hung_task_warnings
431 ==================
432
433 The maximum number of warnings to report. Duri
434 if a hung task is detected, this value is decr
435 When this value reaches 0, no more warnings wi
436 This file shows up if ``CONFIG_DETECT_HUNG_TAS
437
438 -1: report an infinite number of warnings.
439
440
441 hyperv_record_panic_msg
442 =======================
443
444 Controls whether the panic kmsg data should be
445
446 = ============================================
447 0 Do not report panic kmsg data.
448 1 Report the panic kmsg data. This is the defa
449 = ============================================
450
451
452 ignore-unaligned-usertrap
453 =========================
454
455 On architectures where unaligned accesses caus
456 feature is supported (``CONFIG_SYSCTL_ARCH_UNA
457 currently, ``arc``, ``parisc`` and ``loongarch
458 unaligned traps are logged.
459
460 = ============================================
461 0 Log all unaligned accesses.
462 1 Only warn the first time a process traps. Th
463 setting.
464 = ============================================
465
466 See also `unaligned-trap`_.
467
468 io_uring_disabled
469 =================
470
471 Prevents all processes from creating new io_ur
472 shrinks the kernel's attack surface.
473
474 = ============================================
475 0 All processes can create io_uring instances
476 default setting.
477 1 io_uring creation is disabled (io_uring_setu
478 -EPERM) for unprivileged processes not in th
479 Existing io_uring instances can still be use
480 documentation for io_uring_group for more in
481 2 io_uring creation is disabled for all proces
482 always fails with -EPERM. Existing io_uring
483 used.
484 = ============================================
485
486
487 io_uring_group
488 ==============
489
490 When io_uring_disabled is set to 1, a process
491 privileged (CAP_SYS_ADMIN) or be in the io_uri
492 to create an io_uring instance. If io_uring_g
493 default), only processes with the CAP_SYS_ADMI
494 io_uring instances.
495
496
497 kexec_load_disabled
498 ===================
499
500 A toggle indicating if the syscalls ``kexec_lo
501 ``kexec_file_load`` have been disabled.
502 This value defaults to 0 (false: ``kexec_*load
503 set to 1 (true: ``kexec_*load`` disabled).
504 Once true, kexec can no longer be used, and th
505 back to false.
506 This allows a kexec image to be loaded before
507 allowing a system to set up (and later use) an
508 altered.
509 Generally used together with the `modules_disa
510
511 kexec_load_limit_panic
512 ======================
513
514 This parameter specifies a limit to the number
515 ``kexec_load`` and ``kexec_file_load`` can be
516 image. It can only be set with a more restrict
517 current one.
518
519 == ===========================================
520 -1 Unlimited calls to kexec. This is the defau
521 N Number of calls left.
522 == ===========================================
523
524 kexec_load_limit_reboot
525 =======================
526
527 Similar functionality as ``kexec_load_limit_pa
528 image.
529
530 kptr_restrict
531 =============
532
533 This toggle indicates whether restrictions are
534 exposing kernel addresses via ``/proc`` and ot
535
536 When ``kptr_restrict`` is set to 0 (the defaul
537 before printing.
538 (This is the equivalent to %p.)
539
540 When ``kptr_restrict`` is set to 1, kernel poi
541 %pK format specifier will be replaced with 0s
542 ``CAP_SYSLOG`` and effective user and group id
543 ids.
544 This is because %pK checks are done at read()
545 time, so if permissions are elevated between t
546 (e.g via a setuid binary) then %pK will not le
547 unprivileged users.
548 Note, this is a temporary solution only.
549 The correct long-term solution is to do the pe
550 open() time.
551 Consider removing world read permissions from
552 using `dmesg_restrict`_ to protect against use
553 if leaking kernel pointer values to unprivileg
554
555 When ``kptr_restrict`` is set to 2, kernel poi
556 %pK will be replaced with 0s regardless of pri
557
558
559 modprobe
560 ========
561
562 The full path to the usermode helper for autol
563 by default ``CONFIG_MODPROBE_PATH``, which in
564 "/sbin/modprobe". This binary is executed whe
565 module. For example, if userspace passes an u
566 to mount(), then the kernel will automatically
567 corresponding filesystem module by executing t
568 This usermode helper should insert the needed
569
570 This sysctl only affects module autoloading.
571 ability to explicitly insert modules.
572
573 This sysctl can be used to debug module loadin
574
575 echo '#! /bin/sh' > /tmp/modprobe
576 echo 'echo "$@" >> /tmp/modprobe.log' >> /
577 echo 'exec /sbin/modprobe "$@"' >> /tmp/mo
578 chmod a+x /tmp/modprobe
579 echo /tmp/modprobe > /proc/sys/kernel/modp
580
581 Alternatively, if this sysctl is set to the em
582 autoloading is completely disabled. The kerne
583 execute a usermode helper at all, nor will it
584 kernel_module_request LSM hook.
585
586 If CONFIG_STATIC_USERMODEHELPER=y is set in th
587 then the configured static usermode helper ove
588 except that the empty string is still accepted
589 module autoloading as described above.
590
591 modules_disabled
592 ================
593
594 A toggle value indicating if modules are allow
595 in an otherwise modular kernel. This toggle d
596 (0), but can be set true (1). Once true, modu
597 neither loaded nor unloaded, and the toggle ca
598 to false. Generally used with the `kexec_load
599
600
601 .. _msgmni:
602
603 msgmax, msgmnb, and msgmni
604 ==========================
605
606 ``msgmax`` is the maximum size of an IPC messa
607 default (``MSGMAX``).
608
609 ``msgmnb`` is the maximum size of an IPC queue
610 default (``MSGMNB``).
611
612 ``msgmni`` is the maximum number of IPC queues
613 (``MSGMNI``).
614
615 All of these parameters are set per ipc namesp
616 in POSIX message queues is limited by ``RLIMIT
617 respected hierarchically in the each user name
618
619 msg_next_id, sem_next_id, and shm_next_id (Sys
620 ==============================================
621
622 These three toggles allows to specify desired
623 object: message, semaphore or shared memory re
624
625 By default they are equal to -1, which means g
626 Possible values to set are in range {0:``INT_M
627
628 Notes:
629 1) kernel doesn't guarantee, that new object
630 it's up to userspace, how to handle an ob
631 2) Toggle with non-default value will be set
632 successful IPC object allocation. If an I
633 fails, it is undefined if the value remai
634
635
636 ngroups_max
637 ===========
638
639 Maximum number of supplementary groups, _i.e._
640 ``setgroups`` will accept. Exports ``NGROUPS_M
641
642
643
644 nmi_watchdog
645 ============
646
647 This parameter can be used to control the NMI
648 (i.e. the hard lockup detector) on x86 systems
649
650 = =================================
651 0 Disable the hard lockup detector.
652 1 Enable the hard lockup detector.
653 = =================================
654
655 The hard lockup detector monitors each CPU for
656 timer interrupts. The mechanism utilizes CPU p
657 that are programmed to generate Non-Maskable I
658 while a CPU is busy. Hence, the alternative na
659
660 The NMI watchdog is disabled by default if the
661 in a KVM virtual machine. This default can be
662
663 nmi_watchdog=1
664
665 to the guest kernel command line (see
666 Documentation/admin-guide/kernel-parameters.rs
667
668
669 nmi_wd_lpm_factor (PPC only)
670 ============================
671
672 Factor to apply to the NMI watchdog timeout (o
673 set to 1). This factor represents the percenta
674 ``watchdog_thresh`` when calculating the NMI w
675 LPM. The soft lockup timeout is not impacted.
676
677 A value of 0 means no change. The default valu
678 watchdog is set to 30s (based on ``watchdog_th
679
680
681 numa_balancing
682 ==============
683
684 Enables/disables and configures automatic page
685 balancing. Memory is moved automatically to n
686 The value to set can be the result of ORing th
687
688 = =================================
689 0 NUMA_BALANCING_DISABLED
690 1 NUMA_BALANCING_NORMAL
691 2 NUMA_BALANCING_MEMORY_TIERING
692 = =================================
693
694 Or NUMA_BALANCING_NORMAL to optimize page plac
695 NUMA nodes to reduce remote accessing. On NUM
696 performance penalty if remote memory is access
697 feature is enabled the kernel samples what tas
698 memory by periodically unmapping pages and lat
699 fault. At the time of the page fault, it is de
700 being accessed should be migrated to a local m
701
702 The unmapping of pages and trapping faults inc
703 ideally is offset by improved memory locality
704 guarantee. If the target workload is already b
705 feature should be disabled.
706
707 Or NUMA_BALANCING_MEMORY_TIERING to optimize p
708 different types of memory (represented as diff
709 place the hot pages in the fast memory. This
710 unmapping and page fault too.
711
712 numa_balancing_promote_rate_limit_MBps
713 ======================================
714
715 Too high promotion/demotion throughput between
716 may hurt application latency. This can be use
717 promotion throughput. The per-node max promot
718 will be limited to be no more than the set val
719
720 A rule of thumb is to set this to less than 1/
721 write bandwidth.
722
723 oops_all_cpu_backtrace
724 ======================
725
726 If this option is set, the kernel will send an
727 their backtraces when an oops event occurs. It
728 resort in case a panic cannot be triggered (to
729 example) or kdump can't be collected. This fil
730 is enabled.
731
732 0: Won't show all CPUs backtraces when an oops
733 This is the default behavior.
734
735 1: Will non-maskably interrupt all CPUs and du
736 an oops event is detected.
737
738
739 oops_limit
740 ==========
741
742 Number of kernel oopses after which the kernel
743 ``panic_on_oops`` is not set. Setting this to
744 the count. Setting this to 1 has the same eff
745 ``panic_on_oops=1``. The default value is 1000
746
747
748 osrelease, ostype & version
749 ===========================
750
751 ::
752
753 # cat osrelease
754 2.1.88
755 # cat ostype
756 Linux
757 # cat version
758 #5 Wed Feb 25 21:49:24 MET 1998
759
760 The files ``osrelease`` and ``ostype`` should
761 ``version``
762 needs a little more clarification however. The
763 this is the fifth kernel built from this sourc
764 date behind it indicates the time the kernel w
765 The only way to tune these values is to rebuil
766
767
768 overflowgid & overflowuid
769 =========================
770
771 if your architecture did not always support 32
772 i386, m68k, sh, and sparc32), a fixed UID and
773 applications that use the old 16-bit UID/GID s
774 actual UID or GID would exceed 65535.
775
776 These sysctls allow you to change the value of
777 The default is 65534.
778
779
780 panic
781 =====
782
783 The value in this file determines the behaviou
784 panic:
785
786 * if zero, the kernel will loop forever;
787 * if negative, the kernel will reboot immediat
788 * if positive, the kernel will reboot after th
789 of seconds.
790
791 When you use the software watchdog, the recomm
792
793
794 panic_on_io_nmi
795 ===============
796
797 Controls the kernel's behavior when a CPU rece
798 an IO error.
799
800 = ============================================
801 0 Try to continue operation (default).
802 1 Panic immediately. The IO error triggered an
803 serious system condition which could result
804 Rather than continuing, panicking might be a
805 servers issue this sort of NMI when the dump
806 and you can use this option to take a crash
807 = ============================================
808
809
810 panic_on_oops
811 =============
812
813 Controls the kernel's behaviour when an oops o
814
815 = ============================================
816 0 Try to continue operation.
817 1 Panic immediately. If the `panic` sysctl is
818 machine will be rebooted.
819 = ============================================
820
821
822 panic_on_stackoverflow
823 ======================
824
825 Controls the kernel's behavior when detecting
826 kernel, IRQ and exception stacks except a user
827 This file shows up if ``CONFIG_DEBUG_STACKOVER
828
829 = ==========================
830 0 Try to continue operation.
831 1 Panic immediately.
832 = ==========================
833
834
835 panic_on_unrecovered_nmi
836 ========================
837
838 The default Linux behaviour on an NMI of eithe
839 to continue operation. For many environments s
840 computing it is preferable that the box is tak
841 dealt with than an uncorrected parity/ECC erro
842
843 A small number of systems do generate NMIs for
844 such as power management so the default is off
845 the existing panic controls already in that di
846
847
848 panic_on_warn
849 =============
850
851 Calls panic() in the WARN() path when set to 1
852 a kernel rebuild when attempting to kdump at t
853
854 = ============================================
855 0 Only WARN(), default behaviour.
856 1 Call panic() after printing out WARN() locat
857 = ============================================
858
859
860 panic_print
861 ===========
862
863 Bitmask for printing system info when panic ha
864 combination of the following bits:
865
866 ===== =======================================
867 bit 0 print all tasks info
868 bit 1 print system memory info
869 bit 2 print timer info
870 bit 3 print locks info if ``CONFIG_LOCKDEP``
871 bit 4 print ftrace buffer
872 bit 5 print all printk messages in buffer
873 bit 6 print all CPUs backtrace (if available
874 bit 7 print only tasks in uninterruptible (bl
875 ===== =======================================
876
877 So for example to print tasks and memory info
878
879 echo 3 > /proc/sys/kernel/panic_print
880
881
882 panic_on_rcu_stall
883 ==================
884
885 When set to 1, calls panic() after RCU stall d
886 is useful to define the root cause of RCU stal
887
888 = ============================================
889 0 Do not panic() when RCU stall takes place, d
890 1 panic() after printing RCU stall messages.
891 = ============================================
892
893 max_rcu_stall_to_panic
894 ======================
895
896 When ``panic_on_rcu_stall`` is set to 1, this
897 number of times that RCU can stall before pani
898
899 When ``panic_on_rcu_stall`` is set to 0, this
900
901 perf_cpu_time_max_percent
902 =========================
903
904 Hints to the kernel how much CPU time it shoul
905 use to handle perf sampling events. If the pe
906 is informed that its samples are exceeding thi
907 will drop its sampling frequency to attempt to
908 usage.
909
910 Some perf sampling happens in NMIs. If these
911 unexpectedly take too long to execute, the NMI
912 stacked up next to each other so much that not
913 allowed to execute.
914
915 ===== ========================================
916 0 Disable the mechanism. Do not monitor o
917 sampling rate no matter how CPU time it
918
919 1-100 Attempt to throttle perf's sample rate t
920 percentage of CPU. Note: the kernel cal
921 "expected" length of each sample event.
922 100% of that expected length. Even if t
923 100, you may still see sample throttling
924 length is exceeded. Set to 0 if you tru
925 how much CPU is consumed.
926 ===== ========================================
927
928
929 perf_event_paranoid
930 ===================
931
932 Controls use of the performance events system
933 users (without CAP_PERFMON). The default valu
934
935 For backward compatibility reasons access to s
936 monitoring and observability remains open for
937 privileged processes but CAP_SYS_ADMIN usage f
938 performance monitoring and observability opera
939 with respect to CAP_PERFMON use cases.
940
941 === =========================================
942 -1 Allow use of (almost) all events by all u
943
944 Ignore mlock limit after perf_event_mlock
945 ``CAP_IPC_LOCK``.
946
947 >=0 Disallow ftrace function tracepoint by us
948 ``CAP_PERFMON``.
949
950 Disallow raw tracepoint access by users w
951
952 >=1 Disallow CPU event access by users withou
953
954 >=2 Disallow kernel profiling by users withou
955 === =========================================
956
957
958 perf_event_max_stack
959 ====================
960
961 Controls maximum number of stack frames to cop
962 PERF_SAMPLE_CALLCHAIN``) configured events, fo
963 '``perf record -g``' or '``perf trace --call-g
964
965 This can only be done when no events are in us
966 enabled, otherwise writing to this file will r
967
968 The default value is 127.
969
970
971 perf_event_mlock_kb
972 ===================
973
974 Control size of per-cpu ring buffer not counte
975
976 The default value is 512 + 1 page
977
978
979 perf_event_max_contexts_per_stack
980 =================================
981
982 Controls maximum number of stack frame context
983 (``attr.sample_type & PERF_SAMPLE_CALLCHAIN``)
984 instance, when using '``perf record -g``' or '
985
986 This can only be done when no events are in us
987 enabled, otherwise writing to this file will r
988
989 The default value is 8.
990
991
992 perf_user_access (arm64 and riscv only)
993 =======================================
994
995 Controls user space access for reading perf ev
996
997 arm64
998 =====
999
1000 The default value is 0 (access disabled).
1001
1002 When set to 1, user space can read performanc
1003 directly.
1004
1005 See Documentation/arch/arm64/perf.rst for mor
1006
1007 riscv
1008 =====
1009
1010 When set to 0, user space access is disabled.
1011
1012 The default value is 1, user space can read p
1013 registers through perf, any direct access wit
1014 an illegal instruction.
1015
1016 When set to 2, which enables legacy mode (use
1017 and insret CSRs only). Note that this legacy
1018 removed once all user space applications are
1019
1020 Note that the time CSR is always directly acc
1021
1022 pid_max
1023 =======
1024
1025 PID allocation wrap value. When the kernel's
1026 reaches this value, it wraps back to a minimu
1027 PIDs of value ``pid_max`` or larger are not a
1028
1029
1030 ns_last_pid
1031 ===========
1032
1033 The last pid allocated in the current (the on
1034 lives in) pid namespace. When selecting a pid
1035 kernel tries to allocate a number starting fr
1036
1037
1038 powersave-nap (PPC only)
1039 ========================
1040
1041 If set, Linux-PPC will use the 'nap' mode of
1042 otherwise the 'doze' mode will be used.
1043
1044
1045 =============================================
1046
1047 printk
1048 ======
1049
1050 The four values in printk denote: ``console_l
1051 ``default_message_loglevel``, ``minimum_conso
1052 ``default_console_loglevel`` respectively.
1053
1054 These values influence printk() behavior when
1055 logging error messages. See '``man 2 syslog``
1056 the different loglevels.
1057
1058 ======================== ====================
1059 console_loglevel messages with a high
1060 this will be printed
1061 default_message_loglevel messages without an
1062 will be printed with
1063 minimum_console_loglevel minimum (highest) va
1064 console_loglevel can
1065 default_console_loglevel default value for co
1066 ======================== ====================
1067
1068
1069 printk_delay
1070 ============
1071
1072 Delay each printk message in ``printk_delay``
1073
1074 Value from 0 - 10000 is allowed.
1075
1076
1077 printk_ratelimit
1078 ================
1079
1080 Some warning messages are rate limited. ``pri
1081 the minimum length of time between these mess
1082 The default value is 5 seconds.
1083
1084 A value of 0 will disable rate limiting.
1085
1086
1087 printk_ratelimit_burst
1088 ======================
1089
1090 While long term we enforce one message per `p
1091 seconds, we do allow a burst of messages to p
1092 ``printk_ratelimit_burst`` specifies the numb
1093 send before ratelimiting kicks in.
1094
1095 The default value is 10 messages.
1096
1097
1098 printk_devkmsg
1099 ==============
1100
1101 Control the logging to ``/dev/kmsg`` from use
1102
1103 ========= ===================================
1104 ratelimit default, ratelimited
1105 on unlimited logging to /dev/kmsg from
1106 off logging to /dev/kmsg disabled
1107 ========= ===================================
1108
1109 The kernel command line parameter ``printk.de
1110 a one-time setting until next reboot: once se
1111 this sysctl interface anymore.
1112
1113 =============================================
1114
1115
1116 pty
1117 ===
1118
1119 See Documentation/filesystems/devpts.rst.
1120
1121
1122 random
1123 ======
1124
1125 This is a directory, with the following entri
1126
1127 * ``boot_id``: a UUID generated the first tim
1128 unvarying after that;
1129
1130 * ``uuid``: a UUID generated every time this
1131 thus be used to generate UUIDs at will);
1132
1133 * ``entropy_avail``: the pool's entropy count
1134
1135 * ``poolsize``: the entropy pool size, in bit
1136
1137 * ``urandom_min_reseed_secs``: obsolete (used
1138 number of seconds between urandom pool rese
1139 writable for compatibility purposes, but wr
1140 on any RNG behavior;
1141
1142 * ``write_wakeup_threshold``: when the entrop
1143 (as a number of bits), processes waiting to
1144 are woken up. This file is writable for com
1145 writing to it has no effect on any RNG beha
1146
1147
1148 randomize_va_space
1149 ==================
1150
1151 This option can be used to select the type of
1152 space randomization that is used in the syste
1153 that support this feature.
1154
1155 == =========================================
1156 0 Turn the process address space randomizat
1157 default for architectures that do not sup
1158 and kernels that are booted with the "nor
1159
1160 1 Make the addresses of mmap base, stack an
1161 This, among other things, implies that sh
1162 loaded to random addresses. Also for PIE
1163 location of code start is randomized. Th
1164 ``CONFIG_COMPAT_BRK`` option is enabled.
1165
1166 2 Additionally enable heap randomization.
1167 ``CONFIG_COMPAT_BRK`` is disabled.
1168
1169 There are a few legacy applications out t
1170 versions of libc.so.5 from 1996) that ass
1171 just after the end of the code+bss. Thes
1172 start of the brk area is randomized. The
1173 non-legacy applications that would be bro
1174 systems it is safe to choose full randomi
1175
1176 Systems with ancient and/or broken binari
1177 with ``CONFIG_COMPAT_BRK`` enabled, which
1178 address space randomization.
1179 == =========================================
1180
1181
1182 real-root-dev
1183 =============
1184
1185 See Documentation/admin-guide/initrd.rst.
1186
1187
1188 reboot-cmd (SPARC only)
1189 =======================
1190
1191 ??? This seems to be a way to give an argumen
1192 ROM/Flash boot loader. Maybe to tell it what
1193 rebooting. ???
1194
1195
1196 sched_energy_aware
1197 ==================
1198
1199 Enables/disables Energy Aware Scheduling (EAS
1200 automatically on platforms where it can run (
1201 platforms with asymmetric CPU topologies and
1202 Model available). If your platform happens to
1203 requirements for EAS but you do not want to u
1204 this value to 0. On Non-EAS platforms, write
1205 read doesn't return anything.
1206
1207 task_delayacct
1208 ===============
1209
1210 Enables/disables task delay accounting (see
1211 Documentation/accounting/delay-accounting.rst
1212 a small amount of overhead in the scheduler b
1213 and performance tuning. It is required by som
1214
1215 sched_schedstats
1216 ================
1217
1218 Enables/disables scheduler statistics. Enabli
1219 incurs a small amount of overhead in the sche
1220 useful for debugging and performance tuning.
1221
1222 sched_util_clamp_min
1223 ====================
1224
1225 Max allowed *minimum* utilization.
1226
1227 Default value is 1024, which is the maximum p
1228
1229 It means that any requested uclamp.min value
1230 sched_util_clamp_min, i.e., it is restricted
1231 [0:sched_util_clamp_min].
1232
1233 sched_util_clamp_max
1234 ====================
1235
1236 Max allowed *maximum* utilization.
1237
1238 Default value is 1024, which is the maximum p
1239
1240 It means that any requested uclamp.max value
1241 sched_util_clamp_max, i.e., it is restricted
1242 [0:sched_util_clamp_max].
1243
1244 sched_util_clamp_min_rt_default
1245 ===============================
1246
1247 By default Linux is tuned for performance. Wh
1248 at the highest frequency and most capable (hi
1249 heterogeneous systems).
1250
1251 Uclamp achieves this by setting the requested
1252 1024 by default, which effectively boosts the
1253 frequency and biases them to run on the bigge
1254
1255 This knob allows admins to change the default
1256 used. In battery powered devices particularly
1257 capacity and frequency will increase energy c
1258 life.
1259
1260 This knob is only effective for RT tasks whic
1261 requested uclamp.min value via sched_setattr(
1262
1263 This knob will not escape the range constrain
1264 defined above.
1265
1266 For example if
1267
1268 sched_util_clamp_min_rt_default = 800
1269 sched_util_clamp_min = 600
1270
1271 Then the boost will be clamped to 600 because
1272 range of [0:600]. This could happen for insta
1273 restrict all boosts temporarily by modifying
1274 this restriction is lifted, the requested sch
1275 will take effect.
1276
1277 seccomp
1278 =======
1279
1280 See Documentation/userspace-api/seccomp_filte
1281
1282
1283 sg-big-buff
1284 ===========
1285
1286 This file shows the size of the generic SCSI
1287 You can't tune it just yet, but you could cha
1288 compile time by editing ``include/scsi/sg.h``
1289 the value of ``SG_BIG_BUFF``.
1290
1291 There shouldn't be any reason to change this
1292 you can come up with one, you probably know w
1293 are doing anyway :)
1294
1295
1296 shmall
1297 ======
1298
1299 This parameter sets the total amount of share
1300 inside ipc namespace. The shared memory pages
1301 namespace separately and is not inherited. He
1302 least ``ceil(shmmax/PAGE_SIZE)``.
1303
1304 If you are not sure what the default ``PAGE_S
1305 system, you can run the following command::
1306
1307 # getconf PAGE_SIZE
1308
1309 To reduce or disable the ability to allocate
1310 new ipc namespace, set this parameter to the
1311 creation of a new ipc namespace in the curren
1312 be used.
1313
1314 shmmax
1315 ======
1316
1317 This value can be used to query and set the r
1318 on the maximum shared memory segment size tha
1319 Shared memory segments up to 1Gb are now supp
1320 kernel. This value defaults to ``SHMMAX``.
1321
1322
1323 shmmni
1324 ======
1325
1326 This value determines the maximum number of s
1327 4096 by default (``SHMMNI``).
1328
1329
1330 shm_rmid_forced
1331 ===============
1332
1333 Linux lets you set resource limits, including
1334 process can consume, via ``setrlimit(2)``. U
1335 segments are allowed to exist without associa
1336 thus might not be counted against any resourc
1337 shared memory segments are automatically dest
1338 count becomes zero after a detach or a proces
1339 also destroy segments that were created, but
1340 from the process. The only use left for ``IP
1341 destroy an unattached segment. Of course, th
1342 defined, so some applications might stop work
1343 feature will do you no good unless you also c
1344 limits (in particular, ``RLIMIT_AS`` and ``RL
1345 need this.
1346
1347 Note that if you change this from 0 to 1, alr
1348 without users and with a dead originative pro
1349
1350
1351 sysctl_writes_strict
1352 ====================
1353
1354 Control how file position affects the behavio
1355 via the ``/proc/sys`` interface:
1356
1357 == ======================================
1358 -1 Legacy per-write sysctl value handling
1359 Each write syscall must fully contain
1360 written, and multiple writes on the sa
1361 will rewrite the sysctl value, regardl
1362 0 Same behavior as above, but warn about
1363 to a sysctl file descriptor when the f
1364 1 (default) Respect file position when w
1365 writes will append to the sysctl value
1366 length of the sysctl value buffer will
1367 sysctl entries must always be at file
1368 be fully contained in the buffer sent
1369 == ======================================
1370
1371
1372 softlockup_all_cpu_backtrace
1373 ============================
1374
1375 This value controls the soft lockup detector
1376 when a soft lockup condition is detected as t
1377 to gather further debug information. If enabl
1378 be issued an NMI and instructed to capture st
1379
1380 This feature is only applicable for architect
1381 NMI.
1382
1383 = ===========================================
1384 0 Do nothing. This is the default behavior.
1385 1 On detection capture more debug information
1386 = ===========================================
1387
1388
1389 softlockup_panic
1390 =================
1391
1392 This parameter can be used to control whether
1393 when a soft lockup is detected.
1394
1395 = ===========================================
1396 0 Don't panic on soft lockup.
1397 1 Panic on soft lockup.
1398 = ===========================================
1399
1400 This can also be set using the softlockup_pan
1401
1402
1403 soft_watchdog
1404 =============
1405
1406 This parameter can be used to control the sof
1407
1408 = =================================
1409 0 Disable the soft lockup detector.
1410 1 Enable the soft lockup detector.
1411 = =================================
1412
1413 The soft lockup detector monitors CPUs for th
1414 without rescheduling voluntarily, and thus pr
1415 from running, causing the watchdog work fail
1416 on the CPUs ability to respond to timer inter
1417 watchdog work to be queued by the watchdog ti
1418 watchdog — if enabled — can detect a hard
1419
1420
1421 split_lock_mitigate (x86 only)
1422 ==============================
1423
1424 On x86, each "split lock" imposes a system-wi
1425 systems, large numbers of split locks from un
1426 denials of service to well-behaved and potent
1427
1428 The kernel mitigates these bad users by detec
1429 penalties: forcing them to wait and only allo
1430 locks at a time.
1431
1432 These mitigations can make those bad applicat
1433 split_lock_mitigate=0 may restore some applic
1434 increase system exposure to denial of service
1435
1436 = ===========================================
1437 0 Disable the mitigation mode - just warns th
1438 and exposes the system to denials of servic
1439 1 Enable the mitigation mode (this is the def
1440 lockers with intentional performance degrad
1441 = ===========================================
1442
1443
1444 stack_erasing
1445 =============
1446
1447 This parameter can be used to control kernel
1448 of syscalls for kernels built with ``CONFIG_G
1449
1450 That erasing reduces the information which ke
1451 can reveal and blocks some uninitialized stac
1452 The tradeoff is the performance impact: on a
1453 compilation sees a 1% slowdown, other systems
1454
1455 = ===========================================
1456 0 Kernel stack erasing is disabled, STACKLEAK
1457 1 Kernel stack erasing is enabled (default),
1458 returning to the userspace at the end of sy
1459 = ===========================================
1460
1461
1462 stop-a (SPARC only)
1463 ===================
1464
1465 Controls Stop-A:
1466
1467 = ====================================
1468 0 Stop-A has no effect.
1469 1 Stop-A breaks to the PROM (default).
1470 = ====================================
1471
1472 Stop-A is always enabled on a panic, so that
1473 the boot PROM.
1474
1475
1476 sysrq
1477 =====
1478
1479 See Documentation/admin-guide/sysrq.rst.
1480
1481
1482 tainted
1483 =======
1484
1485 Non-zero if the kernel has been tainted. Nume
1486 ORed together. The letters are seen in "Taint
1487
1488 ====== ===== ==============================
1489 1 `(P)` proprietary module was loaded
1490 2 `(F)` module was force loaded
1491 4 `(S)` kernel running on an out of sp
1492 8 `(R)` module was force unloaded
1493 16 `(M)` processor reported a Machine C
1494 32 `(B)` bad page referenced or some un
1495 64 `(U)` taint requested by userspace a
1496 128 `(D)` kernel died recently, i.e. the
1497 256 `(A)` an ACPI table was overridden b
1498 512 `(W)` kernel issued warning
1499 1024 `(C)` staging driver was loaded
1500 2048 `(I)` workaround for bug in platform
1501 4096 `(O)` externally-built ("out-of-tree
1502 8192 `(E)` unsigned module was loaded
1503 16384 `(L)` soft lockup occurred
1504 32768 `(K)` kernel has been live patched
1505 65536 `(X)` Auxiliary taint, defined and u
1506 131072 `(T)` The kernel was built with the
1507 ====== ===== ==============================
1508
1509 See Documentation/admin-guide/tainted-kernels
1510
1511 Note:
1512 writes to this sysctl interface will fail w
1513 booted with the command line option ``panic
1514 and any of the ORed together values being w
1515 the bitmask declared on panic_on_taint.
1516 See Documentation/admin-guide/kernel-parame
1517 that particular kernel command line option
1518 ``nousertaint`` switch.
1519
1520 threads-max
1521 ===========
1522
1523 This value controls the maximum number of thr
1524 using ``fork()``.
1525
1526 During initialization the kernel sets this va
1527 maximum number of threads is created, the thr
1528 a part (1/8th) of the available RAM pages.
1529
1530 The minimum value that can be written to ``th
1531
1532 The maximum value that can be written to ``th
1533 constant ``FUTEX_TID_MASK`` (0x3fffffff).
1534
1535 If a value outside of this range is written t
1536 ``EINVAL`` error occurs.
1537
1538
1539 traceoff_on_warning
1540 ===================
1541
1542 When set, disables tracing (see Documentation
1543 ``WARN()`` is hit.
1544
1545
1546 tracepoint_printk
1547 =================
1548
1549 When tracepoints are sent to printk() (enable
1550 boot parameter), this entry provides runtime
1551
1552 echo 0 > /proc/sys/kernel/tracepoint_prin
1553
1554 will stop tracepoints from being sent to prin
1555
1556 echo 1 > /proc/sys/kernel/tracepoint_prin
1557
1558 will send them to printk() again.
1559
1560 This only works if the kernel was booted with
1561
1562 See Documentation/admin-guide/kernel-paramete
1563 Documentation/trace/boottime-trace.rst.
1564
1565
1566 unaligned-trap
1567 ==============
1568
1569 On architectures where unaligned accesses cau
1570 feature is supported (``CONFIG_SYSCTL_ARCH_UN
1571 ``arc``, ``parisc`` and ``loongarch``), contr
1572 are caught and emulated (instead of failing).
1573
1574 = ===========================================
1575 0 Do not emulate unaligned accesses.
1576 1 Emulate unaligned accesses. This is the def
1577 = ===========================================
1578
1579 See also `ignore-unaligned-usertrap`_.
1580
1581
1582 unknown_nmi_panic
1583 =================
1584
1585 The value in this file affects behavior of ha
1586 value is non-zero, unknown NMI is trapped and
1587 that time, kernel debugging information is di
1588
1589 NMI switch that most IA32 servers have fires
1590 example. If a system hangs up, try pressing
1591
1592
1593 unprivileged_bpf_disabled
1594 =========================
1595
1596 Writing 1 to this entry will disable unprivil
1597 once disabled, calling ``bpf()`` without ``CA
1598 will return ``-EPERM``. Once set to 1, this c
1599 running kernel anymore.
1600
1601 Writing 2 to this entry will also disable unp
1602 however, an admin can still change this setti
1603 writing 0 or 1 to this entry.
1604
1605 If ``BPF_UNPRIV_DEFAULT_OFF`` is enabled in t
1606 entry will default to 2 instead of 0.
1607
1608 = ===========================================
1609 0 Unprivileged calls to ``bpf()`` are enabled
1610 1 Unprivileged calls to ``bpf()`` are disable
1611 2 Unprivileged calls to ``bpf()`` are disable
1612 = ===========================================
1613
1614
1615 warn_limit
1616 ==========
1617
1618 Number of kernel warnings after which the ker
1619 ``panic_on_warn`` is not set. Setting this to
1620 the warning count. Setting this to 1 has the
1621 ``panic_on_warn=1``. The default value is 0.
1622
1623
1624 watchdog
1625 ========
1626
1627 This parameter can be used to disable or enab
1628 *and* the NMI watchdog (i.e. the hard lockup
1629
1630 = ==============================
1631 0 Disable both lockup detectors.
1632 1 Enable both lockup detectors.
1633 = ==============================
1634
1635 The soft lockup detector and the NMI watchdog
1636 enabled individually, using the ``soft_watchd
1637 parameters.
1638 If the ``watchdog`` parameter is read, for ex
1639
1640 cat /proc/sys/kernel/watchdog
1641
1642 the output of this command (0 or 1) shows the
1643 ``soft_watchdog`` and ``nmi_watchdog``.
1644
1645
1646 watchdog_cpumask
1647 ================
1648
1649 This value can be used to control on which cp
1650 The default cpumask is all possible cores, bu
1651 enabled in the kernel config, and cores are s
1652 ``nohz_full=`` boot argument, those cores are
1653 Offline cores can be included in this mask, a
1654 brought online, the watchdog will be started
1655
1656 Typically this value would only be touched in
1657 to re-enable cores that by default were not r
1658 if a kernel lockup was suspected on those cor
1659
1660 The argument value is the standard cpulist fo
1661 so for example to enable the watchdog on core
1662 might say::
1663
1664 echo 0,2-4 > /proc/sys/kernel/watchdog_cpum
1665
1666
1667 watchdog_thresh
1668 ===============
1669
1670 This value can be used to control the frequen
1671 events and the soft and hard lockup threshold
1672 is 10 seconds.
1673
1674 The softlockup threshold is (``2 * watchdog_t
1675 tunable to zero will disable lockup detection
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.