1 ===================================
2 Documentation for /proc/sys/kernel/
3 ===================================
5 .. See scripts/check-sysctl-docs to keep this up to date
8 Copyright (c) 1998, 1999, Rik van Riel <riel@nl.linux.org>
10 Copyright (c) 2009, Shen Feng<shen@cn.fujitsu.com>
12 For general info and legal blurb, please look in
13 Documentation/admin-guide/sysctl/index.rst.
15 ------------------------------------------------------------------------------
17 This file contains documentation for the sysctl files in
18 ``/proc/sys/kernel/``.
20 The files in this directory can be used to tune and monitor
21 miscellaneous and general things in the operation of the Linux
22 kernel. Since some of the files *can* be used to screw up your
23 system, it is advisable to read both documentation and source
24 before actually making adjustments.
26 Currently, these files might (depending on your configuration)
27 show up in ``/proc/sys/kernel``:
37 highwater lowwater frequency
39 If BSD-style process accounting is enabled these values control
40 its behaviour. If free space on filesystem where the log lives
41 goes below ``lowwater``% accounting suspends. If free space gets
42 above ``highwater``% accounting resumes. ``frequency`` determines
43 how often do we check the amount of free space (value is in
50 That is, suspend accounting if free space drops below 2%; resume it
51 if it increases to at least 4%; consider information about amount of
52 free space valid for 30 seconds.
58 See Documentation/power/video.rst. This allows the video resume mode to be set,
59 in a similar fashion to the ``acpi_sleep`` kernel parameter, by
60 combining the following values:
72 This variable has no effect and may be removed in future kernel
73 releases. Reading it always returns 0.
74 Up to Linux 3.17, it enabled/disabled automatic recomputing of
76 upon memory add/remove or upon IPC namespace creation/removal.
77 Echoing "1" into this file enabled msgmni automatic recomputing.
78 Echoing "0" turned it off. The default value was 1.
81 bootloader_type (x86 only)
82 ==========================
84 This gives the bootloader type number as indicated by the bootloader,
85 shifted left by 4, and OR'd with the low four bits of the bootloader
86 version. The reason for this encoding is that this used to match the
87 ``type_of_loader`` field in the kernel header; the encoding is kept for
88 backwards compatibility. That is, if the full bootloader type number
89 is 0x15 and the full version number is 0x234, this file will contain
90 the value 340 = 0x154.
92 See the ``type_of_loader`` and ``ext_loader_type`` fields in
93 Documentation/x86/boot.rst for additional information.
96 bootloader_version (x86 only)
97 =============================
99 The complete bootloader version number. In the example above, this
100 file will contain the value 564 = 0x234.
102 See the ``type_of_loader`` and ``ext_loader_ver`` fields in
103 Documentation/x86/boot.rst for additional information.
109 Controls whether the kernel should collect statistics on BPF programs
110 (total time spent running, number of times run...). Enabling
111 statistics causes a slight reduction in performance on each program
112 run. The statistics can be seen using ``bpftool``.
114 = ===================================
115 0 Don't collect statistics (default).
116 1 Collect statistics.
117 = ===================================
123 This is the pid which will be signalled on reboot (notably, by
124 Ctrl-Alt-Delete). Writing a value to this file which doesn't
125 correspond to a running process will result in ``-ESRCH``.
127 See also `ctrl-alt-del`_.
133 Highest valid capability of the running kernel. Exports
134 ``CAP_LAST_CAP`` from the kernel.
140 ``core_pattern`` is used to specify a core dumpfile pattern name.
142 * max length 127 characters; default value is "core"
143 * ``core_pattern`` is used as a pattern template for the output
144 filename; certain string patterns (beginning with '%') are
145 substituted with their actual values.
146 * backward compatibility with ``core_uses_pid``:
148 If ``core_pattern`` does not include "%p" (default does not)
149 and ``core_uses_pid`` is set, then .PID will be appended to
152 * corename format specifiers
154 ======== ==========================================
155 %<NUL> '%' is dropped
158 %P global pid (init PID namespace)
160 %I global tid (init PID namespace)
161 %u uid (in initial user namespace)
162 %g gid (in initial user namespace)
163 %d dump mode, matches ``PR_SET_DUMPABLE`` and
164 ``/proc/sys/fs/suid_dumpable``
168 %e executable filename (may be shortened, could be changed by prctl etc)
169 %f executable filename
171 %c maximum size of core file by resource limit RLIMIT_CORE
172 %<OTHER> both are dropped
173 ======== ==========================================
175 * If the first character of the pattern is a '|', the kernel will treat
176 the rest of the pattern as a command to run. The core dump will be
177 written to the standard input of that program instead of to a file.
183 This sysctl is only applicable when `core_pattern`_ is configured to
184 pipe core files to a user space helper (when the first character of
185 ``core_pattern`` is a '|', see above).
186 When collecting cores via a pipe to an application, it is occasionally
187 useful for the collecting application to gather data about the
188 crashing process from its ``/proc/pid`` directory.
189 In order to do this safely, the kernel must wait for the collecting
190 process to exit, so as not to remove the crashing processes proc files
192 This in turn creates the possibility that a misbehaving userspace
193 collecting process can block the reaping of a crashed process simply
195 This sysctl defends against that.
196 It defines how many concurrent crashing processes may be piped to user
197 space applications in parallel.
198 If this value is exceeded, then those crashing processes above that
199 value are noted via the kernel log and their cores are skipped.
200 0 is a special value, indicating that unlimited processes may be
201 captured in parallel, but that no waiting will take place (i.e. the
202 collecting process is not guaranteed access to ``/proc/<crashing
204 This value defaults to 0.
210 The default coredump filename is "core". By setting
211 ``core_uses_pid`` to 1, the coredump filename becomes core.PID.
212 If `core_pattern`_ does not include "%p" (default does not)
213 and ``core_uses_pid`` is set, then .PID will be appended to
220 When the value in this file is 0, ctrl-alt-del is trapped and
221 sent to the ``init(1)`` program to handle a graceful restart.
222 When, however, the value is > 0, Linux's reaction to a Vulcan
223 Nerve Pinch (tm) will be an immediate reboot, without even
224 syncing its dirty buffers.
227 when a program (like dosemu) has the keyboard in 'raw'
228 mode, the ctrl-alt-del is intercepted by the program before it
229 ever reaches the kernel tty layer, and it's up to the program
230 to decide what to do with it.
236 This toggle indicates whether unprivileged users are prevented
237 from using ``dmesg(8)`` to view messages from the kernel's log
239 When ``dmesg_restrict`` is set to 0 there are no restrictions.
240 When ``dmesg_restrict`` is set to 1, users must have
241 ``CAP_SYSLOG`` to use ``dmesg(8)``.
243 The kernel config option ``CONFIG_SECURITY_DMESG_RESTRICT`` sets the
244 default value of ``dmesg_restrict``.
247 domainname & hostname
248 =====================
250 These files can be used to set the NIS/YP domainname and the
251 hostname of your box in exactly the same way as the commands
252 domainname and hostname, i.e.::
254 # echo "darkstar" > /proc/sys/kernel/hostname
255 # echo "mydomain" > /proc/sys/kernel/domainname
257 has the same effect as::
259 # hostname "darkstar"
260 # domainname "mydomain"
262 Note, however, that the classic darkstar.frop.org has the
263 hostname "darkstar" and DNS (Internet Domain Name Server)
264 domainname "frop.org", not to be confused with the NIS (Network
265 Information Service) or YP (Yellow Pages) domainname. These two
266 domain names are in general different. For a detailed discussion
267 see the ``hostname(1)`` man page.
273 See Documentation/driver-api/firmware/fallback-mechanisms.rst.
275 The entries in this directory allow the firmware loader helper
276 fallback to be controlled:
278 * ``force_sysfs_fallback``, when set to 1, forces the use of the
280 * ``ignore_sysfs_fallback``, when set to 1, ignores any fallback.
286 Determines whether ``ftrace_dump()`` should be called on an oops (or
287 kernel panic). This will output the contents of the ftrace buffers to
288 the console. This is very useful for capturing traces that lead to
289 crashes and outputting them to a serial console.
291 = ===================================================
292 0 Disabled (default).
293 1 Dump buffers of all CPUs.
294 2 Dump the buffer of the CPU that triggered the oops.
295 = ===================================================
298 ftrace_enabled, stack_tracer_enabled
299 ====================================
301 See Documentation/trace/ftrace.rst.
304 hardlockup_all_cpu_backtrace
305 ============================
307 This value controls the hard lockup detector behavior when a hard
308 lockup condition is detected as to whether or not to gather further
309 debug information. If enabled, arch-specific all-CPU stack dumping
312 = ============================================
313 0 Do nothing. This is the default behavior.
314 1 On detection capture more debug information.
315 = ============================================
321 This parameter can be used to control whether the kernel panics
322 when a hard lockup is detected.
324 = ===========================
325 0 Don't panic on hard lockup.
326 1 Panic on hard lockup.
327 = ===========================
329 See Documentation/admin-guide/lockup-watchdogs.rst for more information.
330 This can also be set using the nmi_watchdog kernel parameter.
336 Path for the hotplug policy agent.
337 Default value is ``CONFIG_UEVENT_HELPER_PATH``, which in turn defaults
340 This file only exists when ``CONFIG_UEVENT_HELPER`` is enabled. Most
341 modern systems rely exclusively on the netlink-based uevent source and
345 hung_task_all_cpu_backtrace
346 ===========================
348 If this option is set, the kernel will send an NMI to all CPUs to dump
349 their backtraces when a hung task is detected. This file shows up if
350 CONFIG_DETECT_HUNG_TASK and CONFIG_SMP are enabled.
352 0: Won't show all CPUs backtraces when a hung task is detected.
353 This is the default behavior.
355 1: Will non-maskably interrupt all CPUs and dump their backtraces when
356 a hung task is detected.
362 Controls the kernel's behavior when a hung task is detected.
363 This file shows up if ``CONFIG_DETECT_HUNG_TASK`` is enabled.
365 = =================================================
366 0 Continue operation. This is the default behavior.
368 = =================================================
371 hung_task_check_count
372 =====================
374 The upper bound on the number of tasks that are checked.
375 This file shows up if ``CONFIG_DETECT_HUNG_TASK`` is enabled.
378 hung_task_timeout_secs
379 ======================
381 When a task in D state did not get scheduled
382 for more than this value report a warning.
383 This file shows up if ``CONFIG_DETECT_HUNG_TASK`` is enabled.
385 0 means infinite timeout, no checking is done.
387 Possible values to set are in range {0:``LONG_MAX``/``HZ``}.
390 hung_task_check_interval_secs
391 =============================
393 Hung task check interval. If hung task checking is enabled
394 (see `hung_task_timeout_secs`_), the check is done every
395 ``hung_task_check_interval_secs`` seconds.
396 This file shows up if ``CONFIG_DETECT_HUNG_TASK`` is enabled.
398 0 (default) means use ``hung_task_timeout_secs`` as checking
401 Possible values to set are in range {0:``LONG_MAX``/``HZ``}.
407 The maximum number of warnings to report. During a check interval
408 if a hung task is detected, this value is decreased by 1.
409 When this value reaches 0, no more warnings will be reported.
410 This file shows up if ``CONFIG_DETECT_HUNG_TASK`` is enabled.
412 -1: report an infinite number of warnings.
415 hyperv_record_panic_msg
416 =======================
418 Controls whether the panic kmsg data should be reported to Hyper-V.
420 = =========================================================
421 0 Do not report panic kmsg data.
422 1 Report the panic kmsg data. This is the default behavior.
423 = =========================================================
426 ignore-unaligned-usertrap
427 =========================
429 On architectures where unaligned accesses cause traps, and where this
430 feature is supported (``CONFIG_SYSCTL_ARCH_UNALIGN_NO_WARN``;
431 currently, ``arc`` and ``ia64``), controls whether all unaligned traps
434 = =============================================================
435 0 Log all unaligned accesses.
436 1 Only warn the first time a process traps. This is the default
438 = =============================================================
440 See also `unaligned-trap`_ and `unaligned-dump-stack`_. On ``ia64``,
441 this allows system administrators to override the
442 ``IA64_THREAD_UAC_NOPRINT`` ``prctl`` and avoid logs being flooded.
448 A toggle indicating if the ``kexec_load`` syscall has been disabled.
449 This value defaults to 0 (false: ``kexec_load`` enabled), but can be
450 set to 1 (true: ``kexec_load`` disabled).
451 Once true, kexec can no longer be used, and the toggle cannot be set
453 This allows a kexec image to be loaded before disabling the syscall,
454 allowing a system to set up (and later use) an image without it being
456 Generally used together with the `modules_disabled`_ sysctl.
462 This toggle indicates whether restrictions are placed on
463 exposing kernel addresses via ``/proc`` and other interfaces.
465 When ``kptr_restrict`` is set to 0 (the default) the address is hashed
467 (This is the equivalent to %p.)
469 When ``kptr_restrict`` is set to 1, kernel pointers printed using the
470 %pK format specifier will be replaced with 0s unless the user has
471 ``CAP_SYSLOG`` and effective user and group ids are equal to the real
473 This is because %pK checks are done at read() time rather than open()
474 time, so if permissions are elevated between the open() and the read()
475 (e.g via a setuid binary) then %pK will not leak kernel pointers to
477 Note, this is a temporary solution only.
478 The correct long-term solution is to do the permission checks at
480 Consider removing world read permissions from files that use %pK, and
481 using `dmesg_restrict`_ to protect against uses of %pK in ``dmesg(8)``
482 if leaking kernel pointer values to unprivileged users is a concern.
484 When ``kptr_restrict`` is set to 2, kernel pointers printed using
485 %pK will be replaced with 0s regardless of privileges.
491 The full path to the usermode helper for autoloading kernel modules,
492 by default ``CONFIG_MODPROBE_PATH``, which in turn defaults to
493 "/sbin/modprobe". This binary is executed when the kernel requests a
494 module. For example, if userspace passes an unknown filesystem type
495 to mount(), then the kernel will automatically request the
496 corresponding filesystem module by executing this usermode helper.
497 This usermode helper should insert the needed module into the kernel.
499 This sysctl only affects module autoloading. It has no effect on the
500 ability to explicitly insert modules.
502 This sysctl can be used to debug module loading requests::
504 echo '#! /bin/sh' > /tmp/modprobe
505 echo 'echo "$@" >> /tmp/modprobe.log' >> /tmp/modprobe
506 echo 'exec /sbin/modprobe "$@"' >> /tmp/modprobe
507 chmod a+x /tmp/modprobe
508 echo /tmp/modprobe > /proc/sys/kernel/modprobe
510 Alternatively, if this sysctl is set to the empty string, then module
511 autoloading is completely disabled. The kernel will not try to
512 execute a usermode helper at all, nor will it call the
513 kernel_module_request LSM hook.
515 If CONFIG_STATIC_USERMODEHELPER=y is set in the kernel configuration,
516 then the configured static usermode helper overrides this sysctl,
517 except that the empty string is still accepted to completely disable
518 module autoloading as described above.
523 A toggle value indicating if modules are allowed to be loaded
524 in an otherwise modular kernel. This toggle defaults to off
525 (0), but can be set true (1). Once true, modules can be
526 neither loaded nor unloaded, and the toggle cannot be set back
527 to false. Generally used with the `kexec_load_disabled`_ toggle.
532 msgmax, msgmnb, and msgmni
533 ==========================
535 ``msgmax`` is the maximum size of an IPC message, in bytes. 8192 by
536 default (``MSGMAX``).
538 ``msgmnb`` is the maximum size of an IPC queue, in bytes. 16384 by
539 default (``MSGMNB``).
541 ``msgmni`` is the maximum number of IPC queues. 32000 by default
545 msg_next_id, sem_next_id, and shm_next_id (System V IPC)
546 ========================================================
548 These three toggles allows to specify desired id for next allocated IPC
549 object: message, semaphore or shared memory respectively.
551 By default they are equal to -1, which means generic allocation logic.
552 Possible values to set are in range {0:``INT_MAX``}.
555 1) kernel doesn't guarantee, that new object will have desired id. So,
556 it's up to userspace, how to handle an object with "wrong" id.
557 2) Toggle with non-default value will be set back to -1 by kernel after
558 successful IPC object allocation. If an IPC object allocation syscall
559 fails, it is undefined if the value remains unmodified or is reset to -1.
565 Maximum number of supplementary groups, _i.e._ the maximum size which
566 ``setgroups`` will accept. Exports ``NGROUPS_MAX`` from the kernel.
573 This parameter can be used to control the NMI watchdog
574 (i.e. the hard lockup detector) on x86 systems.
576 = =================================
577 0 Disable the hard lockup detector.
578 1 Enable the hard lockup detector.
579 = =================================
581 The hard lockup detector monitors each CPU for its ability to respond to
582 timer interrupts. The mechanism utilizes CPU performance counter registers
583 that are programmed to generate Non-Maskable Interrupts (NMIs) periodically
584 while a CPU is busy. Hence, the alternative name 'NMI watchdog'.
586 The NMI watchdog is disabled by default if the kernel is running as a guest
587 in a KVM virtual machine. This default can be overridden by adding::
591 to the guest kernel command line (see
592 Documentation/admin-guide/kernel-parameters.rst).
598 Enables/disables automatic page fault based NUMA memory
599 balancing. Memory is moved automatically to nodes
600 that access it often.
602 Enables/disables automatic NUMA memory balancing. On NUMA machines, there
603 is a performance penalty if remote memory is accessed by a CPU. When this
604 feature is enabled the kernel samples what task thread is accessing memory
605 by periodically unmapping pages and later trapping a page fault. At the
606 time of the page fault, it is determined if the data being accessed should
607 be migrated to a local memory node.
609 The unmapping of pages and trapping faults incur additional overhead that
610 ideally is offset by improved memory locality but there is no universal
611 guarantee. If the target workload is already bound to NUMA nodes then this
612 feature should be disabled. Otherwise, if the system overhead from the
613 feature is too high then the rate the kernel samples for NUMA hinting
614 faults may be controlled by the `numa_balancing_scan_period_min_ms,
615 numa_balancing_scan_delay_ms, numa_balancing_scan_period_max_ms,
616 numa_balancing_scan_size_mb`_, and numa_balancing_settle_count sysctls.
619 numa_balancing_scan_period_min_ms, numa_balancing_scan_delay_ms, numa_balancing_scan_period_max_ms, numa_balancing_scan_size_mb
620 ===============================================================================================================================
623 Automatic NUMA balancing scans tasks address space and unmaps pages to
624 detect if pages are properly placed or if the data should be migrated to a
625 memory node local to where the task is running. Every "scan delay" the task
626 scans the next "scan size" number of pages in its address space. When the
627 end of the address space is reached the scanner restarts from the beginning.
629 In combination, the "scan delay" and "scan size" determine the scan rate.
630 When "scan delay" decreases, the scan rate increases. The scan delay and
631 hence the scan rate of every task is adaptive and depends on historical
632 behaviour. If pages are properly placed then the scan delay increases,
633 otherwise the scan delay decreases. The "scan size" is not adaptive but
634 the higher the "scan size", the higher the scan rate.
636 Higher scan rates incur higher system overhead as page faults must be
637 trapped and potentially data must be migrated. However, the higher the scan
638 rate, the more quickly a tasks memory is migrated to a local node if the
639 workload pattern changes and minimises performance impact due to remote
640 memory accesses. These sysctls control the thresholds for scan delays and
641 the number of pages scanned.
643 ``numa_balancing_scan_period_min_ms`` is the minimum time in milliseconds to
644 scan a tasks virtual memory. It effectively controls the maximum scanning
647 ``numa_balancing_scan_delay_ms`` is the starting "scan delay" used for a task
648 when it initially forks.
650 ``numa_balancing_scan_period_max_ms`` is the maximum time in milliseconds to
651 scan a tasks virtual memory. It effectively controls the minimum scanning
654 ``numa_balancing_scan_size_mb`` is how many megabytes worth of pages are
655 scanned for a given scan.
658 oops_all_cpu_backtrace
659 ======================
661 If this option is set, the kernel will send an NMI to all CPUs to dump
662 their backtraces when an oops event occurs. It should be used as a last
663 resort in case a panic cannot be triggered (to protect VMs running, for
664 example) or kdump can't be collected. This file shows up if CONFIG_SMP
667 0: Won't show all CPUs backtraces when an oops is detected.
668 This is the default behavior.
670 1: Will non-maskably interrupt all CPUs and dump their backtraces when
671 an oops event is detected.
674 osrelease, ostype & version
675 ===========================
684 #5 Wed Feb 25 21:49:24 MET 1998
686 The files ``osrelease`` and ``ostype`` should be clear enough.
688 needs a little more clarification however. The '#5' means that
689 this is the fifth kernel built from this source base and the
690 date behind it indicates the time the kernel was built.
691 The only way to tune these values is to rebuild the kernel :-)
694 overflowgid & overflowuid
695 =========================
697 if your architecture did not always support 32-bit UIDs (i.e. arm,
698 i386, m68k, sh, and sparc32), a fixed UID and GID will be returned to
699 applications that use the old 16-bit UID/GID system calls, if the
700 actual UID or GID would exceed 65535.
702 These sysctls allow you to change the value of the fixed UID and GID.
703 The default is 65534.
709 The value in this file determines the behaviour of the kernel on a
712 * if zero, the kernel will loop forever;
713 * if negative, the kernel will reboot immediately;
714 * if positive, the kernel will reboot after the corresponding number
717 When you use the software watchdog, the recommended setting is 60.
723 Controls the kernel's behavior when a CPU receives an NMI caused by
726 = ==================================================================
727 0 Try to continue operation (default).
728 1 Panic immediately. The IO error triggered an NMI. This indicates a
729 serious system condition which could result in IO data corruption.
730 Rather than continuing, panicking might be a better choice. Some
731 servers issue this sort of NMI when the dump button is pushed,
732 and you can use this option to take a crash dump.
733 = ==================================================================
739 Controls the kernel's behaviour when an oops or BUG is encountered.
741 = ===================================================================
742 0 Try to continue operation.
743 1 Panic immediately. If the `panic` sysctl is also non-zero then the
744 machine will be rebooted.
745 = ===================================================================
748 panic_on_stackoverflow
749 ======================
751 Controls the kernel's behavior when detecting the overflows of
752 kernel, IRQ and exception stacks except a user stack.
753 This file shows up if ``CONFIG_DEBUG_STACKOVERFLOW`` is enabled.
755 = ==========================
756 0 Try to continue operation.
758 = ==========================
761 panic_on_unrecovered_nmi
762 ========================
764 The default Linux behaviour on an NMI of either memory or unknown is
765 to continue operation. For many environments such as scientific
766 computing it is preferable that the box is taken out and the error
767 dealt with than an uncorrected parity/ECC error get propagated.
769 A small number of systems do generate NMIs for bizarre random reasons
770 such as power management so the default is off. That sysctl works like
771 the existing panic controls already in that directory.
777 Calls panic() in the WARN() path when set to 1. This is useful to avoid
778 a kernel rebuild when attempting to kdump at the location of a WARN().
780 = ================================================
781 0 Only WARN(), default behaviour.
782 1 Call panic() after printing out WARN() location.
783 = ================================================
789 Bitmask for printing system info when panic happens. User can chose
790 combination of the following bits:
792 ===== ============================================
793 bit 0 print all tasks info
794 bit 1 print system memory info
795 bit 2 print timer info
796 bit 3 print locks info if ``CONFIG_LOCKDEP`` is on
797 bit 4 print ftrace buffer
798 bit 5 print all printk messages in buffer
799 ===== ============================================
801 So for example to print tasks and memory info on panic, user can::
803 echo 3 > /proc/sys/kernel/panic_print
809 When set to 1, calls panic() after RCU stall detection messages. This
810 is useful to define the root cause of RCU stalls using a vmcore.
812 = ============================================================
813 0 Do not panic() when RCU stall takes place, default behavior.
814 1 panic() after printing RCU stall messages.
815 = ============================================================
818 perf_cpu_time_max_percent
819 =========================
821 Hints to the kernel how much CPU time it should be allowed to
822 use to handle perf sampling events. If the perf subsystem
823 is informed that its samples are exceeding this limit, it
824 will drop its sampling frequency to attempt to reduce its CPU
827 Some perf sampling happens in NMIs. If these samples
828 unexpectedly take too long to execute, the NMIs can become
829 stacked up next to each other so much that nothing else is
832 ===== ========================================================
833 0 Disable the mechanism. Do not monitor or correct perf's
834 sampling rate no matter how CPU time it takes.
836 1-100 Attempt to throttle perf's sample rate to this
837 percentage of CPU. Note: the kernel calculates an
838 "expected" length of each sample event. 100 here means
839 100% of that expected length. Even if this is set to
840 100, you may still see sample throttling if this
841 length is exceeded. Set to 0 if you truly do not care
842 how much CPU is consumed.
843 ===== ========================================================
849 Controls use of the performance events system by unprivileged
850 users (without CAP_PERFMON). The default value is 2.
852 For backward compatibility reasons access to system performance
853 monitoring and observability remains open for CAP_SYS_ADMIN
854 privileged processes but CAP_SYS_ADMIN usage for secure system
855 performance monitoring and observability operations is discouraged
856 with respect to CAP_PERFMON use cases.
858 === ==================================================================
859 -1 Allow use of (almost) all events by all users.
861 Ignore mlock limit after perf_event_mlock_kb without
864 >=0 Disallow ftrace function tracepoint by users without
867 Disallow raw tracepoint access by users without ``CAP_PERFMON``.
869 >=1 Disallow CPU event access by users without ``CAP_PERFMON``.
871 >=2 Disallow kernel profiling by users without ``CAP_PERFMON``.
872 === ==================================================================
878 Controls maximum number of stack frames to copy for (``attr.sample_type &
879 PERF_SAMPLE_CALLCHAIN``) configured events, for instance, when using
880 '``perf record -g``' or '``perf trace --call-graph fp``'.
882 This can only be done when no events are in use that have callchains
883 enabled, otherwise writing to this file will return ``-EBUSY``.
885 The default value is 127.
891 Control size of per-cpu ring buffer not counted against mlock limit.
893 The default value is 512 + 1 page
896 perf_event_max_contexts_per_stack
897 =================================
899 Controls maximum number of stack frame context entries for
900 (``attr.sample_type & PERF_SAMPLE_CALLCHAIN``) configured events, for
901 instance, when using '``perf record -g``' or '``perf trace --call-graph fp``'.
903 This can only be done when no events are in use that have callchains
904 enabled, otherwise writing to this file will return ``-EBUSY``.
906 The default value is 8.
912 PID allocation wrap value. When the kernel's next PID value
913 reaches this value, it wraps back to a minimum PID value.
914 PIDs of value ``pid_max`` or larger are not allocated.
920 The last pid allocated in the current (the one task using this sysctl
921 lives in) pid namespace. When selecting a pid for a next task on fork
922 kernel tries to allocate a number starting from this one.
925 powersave-nap (PPC only)
926 ========================
928 If set, Linux-PPC will use the 'nap' mode of powersaving,
929 otherwise the 'doze' mode will be used.
932 ==============================================================
937 The four values in printk denote: ``console_loglevel``,
938 ``default_message_loglevel``, ``minimum_console_loglevel`` and
939 ``default_console_loglevel`` respectively.
941 These values influence printk() behavior when printing or
942 logging error messages. See '``man 2 syslog``' for more info on
943 the different loglevels.
945 ======================== =====================================
946 console_loglevel messages with a higher priority than
947 this will be printed to the console
948 default_message_loglevel messages without an explicit priority
949 will be printed with this priority
950 minimum_console_loglevel minimum (highest) value to which
951 console_loglevel can be set
952 default_console_loglevel default value for console_loglevel
953 ======================== =====================================
959 Delay each printk message in ``printk_delay`` milliseconds
961 Value from 0 - 10000 is allowed.
967 Some warning messages are rate limited. ``printk_ratelimit`` specifies
968 the minimum length of time between these messages (in seconds).
969 The default value is 5 seconds.
971 A value of 0 will disable rate limiting.
974 printk_ratelimit_burst
975 ======================
977 While long term we enforce one message per `printk_ratelimit`_
978 seconds, we do allow a burst of messages to pass through.
979 ``printk_ratelimit_burst`` specifies the number of messages we can
980 send before ratelimiting kicks in.
982 The default value is 10 messages.
988 Control the logging to ``/dev/kmsg`` from userspace:
990 ========= =============================================
991 ratelimit default, ratelimited
992 on unlimited logging to /dev/kmsg from userspace
993 off logging to /dev/kmsg disabled
994 ========= =============================================
996 The kernel command line parameter ``printk.devkmsg=`` overrides this and is
997 a one-time setting until next reboot: once set, it cannot be changed by
998 this sysctl interface anymore.
1000 ==============================================================
1006 See Documentation/filesystems/devpts.rst.
1012 This is a directory, with the following entries:
1014 * ``boot_id``: a UUID generated the first time this is retrieved, and
1015 unvarying after that;
1017 * ``uuid``: a UUID generated every time this is retrieved (this can
1018 thus be used to generate UUIDs at will);
1020 * ``entropy_avail``: the pool's entropy count, in bits;
1022 * ``poolsize``: the entropy pool size, in bits;
1024 * ``urandom_min_reseed_secs``: obsolete (used to determine the minimum
1025 number of seconds between urandom pool reseeding). This file is
1026 writable for compatibility purposes, but writing to it has no effect
1027 on any RNG behavior;
1029 * ``write_wakeup_threshold``: when the entropy count drops below this
1030 (as a number of bits), processes waiting to write to ``/dev/random``
1031 are woken up. This file is writable for compatibility purposes, but
1032 writing to it has no effect on any RNG behavior.
1038 This option can be used to select the type of process address
1039 space randomization that is used in the system, for architectures
1040 that support this feature.
1042 == ===========================================================================
1043 0 Turn the process address space randomization off. This is the
1044 default for architectures that do not support this feature anyways,
1045 and kernels that are booted with the "norandmaps" parameter.
1047 1 Make the addresses of mmap base, stack and VDSO page randomized.
1048 This, among other things, implies that shared libraries will be
1049 loaded to random addresses. Also for PIE-linked binaries, the
1050 location of code start is randomized. This is the default if the
1051 ``CONFIG_COMPAT_BRK`` option is enabled.
1053 2 Additionally enable heap randomization. This is the default if
1054 ``CONFIG_COMPAT_BRK`` is disabled.
1056 There are a few legacy applications out there (such as some ancient
1057 versions of libc.so.5 from 1996) that assume that brk area starts
1058 just after the end of the code+bss. These applications break when
1059 start of the brk area is randomized. There are however no known
1060 non-legacy applications that would be broken this way, so for most
1061 systems it is safe to choose full randomization.
1063 Systems with ancient and/or broken binaries should be configured
1064 with ``CONFIG_COMPAT_BRK`` enabled, which excludes the heap from process
1065 address space randomization.
1066 == ===========================================================================
1072 See Documentation/admin-guide/initrd.rst.
1075 reboot-cmd (SPARC only)
1076 =======================
1078 ??? This seems to be a way to give an argument to the Sparc
1079 ROM/Flash boot loader. Maybe to tell it what to do after
1086 Enables/disables Energy Aware Scheduling (EAS). EAS starts
1087 automatically on platforms where it can run (that is,
1088 platforms with asymmetric CPU topologies and having an Energy
1089 Model available). If your platform happens to meet the
1090 requirements for EAS but you do not want to use it, change
1096 Enables/disables task delay accounting (see
1097 Documentation/accounting/delay-accounting.rst. Enabling this feature incurs
1098 a small amount of overhead in the scheduler but is useful for debugging
1099 and performance tuning. It is required by some tools such as iotop.
1104 Enables/disables scheduler statistics. Enabling this feature
1105 incurs a small amount of overhead in the scheduler but is
1106 useful for debugging and performance tuning.
1108 sched_util_clamp_min
1109 ====================
1111 Max allowed *minimum* utilization.
1113 Default value is 1024, which is the maximum possible value.
1115 It means that any requested uclamp.min value cannot be greater than
1116 sched_util_clamp_min, i.e., it is restricted to the range
1117 [0:sched_util_clamp_min].
1119 sched_util_clamp_max
1120 ====================
1122 Max allowed *maximum* utilization.
1124 Default value is 1024, which is the maximum possible value.
1126 It means that any requested uclamp.max value cannot be greater than
1127 sched_util_clamp_max, i.e., it is restricted to the range
1128 [0:sched_util_clamp_max].
1130 sched_util_clamp_min_rt_default
1131 ===============================
1133 By default Linux is tuned for performance. Which means that RT tasks always run
1134 at the highest frequency and most capable (highest capacity) CPU (in
1135 heterogeneous systems).
1137 Uclamp achieves this by setting the requested uclamp.min of all RT tasks to
1138 1024 by default, which effectively boosts the tasks to run at the highest
1139 frequency and biases them to run on the biggest CPU.
1141 This knob allows admins to change the default behavior when uclamp is being
1142 used. In battery powered devices particularly, running at the maximum
1143 capacity and frequency will increase energy consumption and shorten the battery
1146 This knob is only effective for RT tasks which the user hasn't modified their
1147 requested uclamp.min value via sched_setattr() syscall.
1149 This knob will not escape the range constraint imposed by sched_util_clamp_min
1154 sched_util_clamp_min_rt_default = 800
1155 sched_util_clamp_min = 600
1157 Then the boost will be clamped to 600 because 800 is outside of the permissible
1158 range of [0:600]. This could happen for instance if a powersave mode will
1159 restrict all boosts temporarily by modifying sched_util_clamp_min. As soon as
1160 this restriction is lifted, the requested sched_util_clamp_min_rt_default
1166 See Documentation/userspace-api/seccomp_filter.rst.
1172 This file shows the size of the generic SCSI (sg) buffer.
1173 You can't tune it just yet, but you could change it on
1174 compile time by editing ``include/scsi/sg.h`` and changing
1175 the value of ``SG_BIG_BUFF``.
1177 There shouldn't be any reason to change this value. If
1178 you can come up with one, you probably know what you
1185 This parameter sets the total amount of shared memory pages that
1186 can be used system wide. Hence, ``shmall`` should always be at least
1187 ``ceil(shmmax/PAGE_SIZE)``.
1189 If you are not sure what the default ``PAGE_SIZE`` is on your Linux
1190 system, you can run the following command::
1198 This value can be used to query and set the run time limit
1199 on the maximum shared memory segment size that can be created.
1200 Shared memory segments up to 1Gb are now supported in the
1201 kernel. This value defaults to ``SHMMAX``.
1207 This value determines the maximum number of shared memory segments.
1208 4096 by default (``SHMMNI``).
1214 Linux lets you set resource limits, including how much memory one
1215 process can consume, via ``setrlimit(2)``. Unfortunately, shared memory
1216 segments are allowed to exist without association with any process, and
1217 thus might not be counted against any resource limits. If enabled,
1218 shared memory segments are automatically destroyed when their attach
1219 count becomes zero after a detach or a process termination. It will
1220 also destroy segments that were created, but never attached to, on exit
1221 from the process. The only use left for ``IPC_RMID`` is to immediately
1222 destroy an unattached segment. Of course, this breaks the way things are
1223 defined, so some applications might stop working. Note that this
1224 feature will do you no good unless you also configure your resource
1225 limits (in particular, ``RLIMIT_AS`` and ``RLIMIT_NPROC``). Most systems don't
1228 Note that if you change this from 0 to 1, already created segments
1229 without users and with a dead originative process will be destroyed.
1232 sysctl_writes_strict
1233 ====================
1235 Control how file position affects the behavior of updating sysctl values
1236 via the ``/proc/sys`` interface:
1238 == ======================================================================
1239 -1 Legacy per-write sysctl value handling, with no printk warnings.
1240 Each write syscall must fully contain the sysctl value to be
1241 written, and multiple writes on the same sysctl file descriptor
1242 will rewrite the sysctl value, regardless of file position.
1243 0 Same behavior as above, but warn about processes that perform writes
1244 to a sysctl file descriptor when the file position is not 0.
1245 1 (default) Respect file position when writing sysctl strings. Multiple
1246 writes will append to the sysctl value buffer. Anything past the max
1247 length of the sysctl value buffer will be ignored. Writes to numeric
1248 sysctl entries must always be at file position 0 and the value must
1249 be fully contained in the buffer sent in the write syscall.
1250 == ======================================================================
1253 softlockup_all_cpu_backtrace
1254 ============================
1256 This value controls the soft lockup detector thread's behavior
1257 when a soft lockup condition is detected as to whether or not
1258 to gather further debug information. If enabled, each cpu will
1259 be issued an NMI and instructed to capture stack trace.
1261 This feature is only applicable for architectures which support
1264 = ============================================
1265 0 Do nothing. This is the default behavior.
1266 1 On detection capture more debug information.
1267 = ============================================
1273 This parameter can be used to control whether the kernel panics
1274 when a soft lockup is detected.
1276 = ============================================
1277 0 Don't panic on soft lockup.
1278 1 Panic on soft lockup.
1279 = ============================================
1281 This can also be set using the softlockup_panic kernel parameter.
1287 This parameter can be used to control the soft lockup detector.
1289 = =================================
1290 0 Disable the soft lockup detector.
1291 1 Enable the soft lockup detector.
1292 = =================================
1294 The soft lockup detector monitors CPUs for threads that are hogging the CPUs
1295 without rescheduling voluntarily, and thus prevent the 'migration/N' threads
1296 from running, causing the watchdog work fail to execute. The mechanism depends
1297 on the CPUs ability to respond to timer interrupts which are needed for the
1298 watchdog work to be queued by the watchdog timer function, otherwise the NMI
1299 watchdog — if enabled — can detect a hard lockup condition.
1305 This parameter can be used to control kernel stack erasing at the end
1306 of syscalls for kernels built with ``CONFIG_GCC_PLUGIN_STACKLEAK``.
1308 That erasing reduces the information which kernel stack leak bugs
1309 can reveal and blocks some uninitialized stack variable attacks.
1310 The tradeoff is the performance impact: on a single CPU system kernel
1311 compilation sees a 1% slowdown, other systems and workloads may vary.
1313 = ====================================================================
1314 0 Kernel stack erasing is disabled, STACKLEAK_METRICS are not updated.
1315 1 Kernel stack erasing is enabled (default), it is performed before
1316 returning to the userspace at the end of syscalls.
1317 = ====================================================================
1325 = ====================================
1326 0 Stop-A has no effect.
1327 1 Stop-A breaks to the PROM (default).
1328 = ====================================
1330 Stop-A is always enabled on a panic, so that the user can return to
1337 See Documentation/admin-guide/sysrq.rst.
1343 Non-zero if the kernel has been tainted. Numeric values, which can be
1344 ORed together. The letters are seen in "Tainted" line of Oops reports.
1346 ====== ===== ==============================================================
1347 1 `(P)` proprietary module was loaded
1348 2 `(F)` module was force loaded
1349 4 `(S)` kernel running on an out of specification system
1350 8 `(R)` module was force unloaded
1351 16 `(M)` processor reported a Machine Check Exception (MCE)
1352 32 `(B)` bad page referenced or some unexpected page flags
1353 64 `(U)` taint requested by userspace application
1354 128 `(D)` kernel died recently, i.e. there was an OOPS or BUG
1355 256 `(A)` an ACPI table was overridden by user
1356 512 `(W)` kernel issued warning
1357 1024 `(C)` staging driver was loaded
1358 2048 `(I)` workaround for bug in platform firmware applied
1359 4096 `(O)` externally-built ("out-of-tree") module was loaded
1360 8192 `(E)` unsigned module was loaded
1361 16384 `(L)` soft lockup occurred
1362 32768 `(K)` kernel has been live patched
1363 65536 `(X)` Auxiliary taint, defined and used by for distros
1364 131072 `(T)` The kernel was built with the struct randomization plugin
1365 ====== ===== ==============================================================
1367 See Documentation/admin-guide/tainted-kernels.rst for more information.
1370 writes to this sysctl interface will fail with ``EINVAL`` if the kernel is
1371 booted with the command line option ``panic_on_taint=<bitmask>,nousertaint``
1372 and any of the ORed together values being written to ``tainted`` match with
1373 the bitmask declared on panic_on_taint.
1374 See Documentation/admin-guide/kernel-parameters.rst for more details on
1375 that particular kernel command line option and its optional
1376 ``nousertaint`` switch.
1381 This value controls the maximum number of threads that can be created
1384 During initialization the kernel sets this value such that even if the
1385 maximum number of threads is created, the thread structures occupy only
1386 a part (1/8th) of the available RAM pages.
1388 The minimum value that can be written to ``threads-max`` is 1.
1390 The maximum value that can be written to ``threads-max`` is given by the
1391 constant ``FUTEX_TID_MASK`` (0x3fffffff).
1393 If a value outside of this range is written to ``threads-max`` an
1394 ``EINVAL`` error occurs.
1400 When set, disables tracing (see Documentation/trace/ftrace.rst) when a
1407 When tracepoints are sent to printk() (enabled by the ``tp_printk``
1408 boot parameter), this entry provides runtime control::
1410 echo 0 > /proc/sys/kernel/tracepoint_printk
1412 will stop tracepoints from being sent to printk(), and::
1414 echo 1 > /proc/sys/kernel/tracepoint_printk
1416 will send them to printk() again.
1418 This only works if the kernel was booted with ``tp_printk`` enabled.
1420 See Documentation/admin-guide/kernel-parameters.rst and
1421 Documentation/trace/boottime-trace.rst.
1424 .. _unaligned-dump-stack:
1426 unaligned-dump-stack (ia64)
1427 ===========================
1429 When logging unaligned accesses, controls whether the stack is
1432 = ===================================================
1433 0 Do not dump the stack. This is the default setting.
1435 = ===================================================
1437 See also `ignore-unaligned-usertrap`_.
1443 On architectures where unaligned accesses cause traps, and where this
1444 feature is supported (``CONFIG_SYSCTL_ARCH_UNALIGN_ALLOW``; currently,
1445 ``arc`` and ``parisc``), controls whether unaligned traps are caught
1446 and emulated (instead of failing).
1448 = ========================================================
1449 0 Do not emulate unaligned accesses.
1450 1 Emulate unaligned accesses. This is the default setting.
1451 = ========================================================
1453 See also `ignore-unaligned-usertrap`_.
1459 The value in this file affects behavior of handling NMI. When the
1460 value is non-zero, unknown NMI is trapped and then panic occurs. At
1461 that time, kernel debugging information is displayed on console.
1463 NMI switch that most IA32 servers have fires unknown NMI up, for
1464 example. If a system hangs up, try pressing the NMI switch.
1467 unprivileged_bpf_disabled
1468 =========================
1470 Writing 1 to this entry will disable unprivileged calls to ``bpf()``;
1471 once disabled, calling ``bpf()`` without ``CAP_SYS_ADMIN`` or ``CAP_BPF``
1472 will return ``-EPERM``. Once set to 1, this can't be cleared from the
1473 running kernel anymore.
1475 Writing 2 to this entry will also disable unprivileged calls to ``bpf()``,
1476 however, an admin can still change this setting later on, if needed, by
1477 writing 0 or 1 to this entry.
1479 If ``BPF_UNPRIV_DEFAULT_OFF`` is enabled in the kernel config, then this
1480 entry will default to 2 instead of 0.
1482 = =============================================================
1483 0 Unprivileged calls to ``bpf()`` are enabled
1484 1 Unprivileged calls to ``bpf()`` are disabled without recovery
1485 2 Unprivileged calls to ``bpf()`` are disabled
1486 = =============================================================
1491 This parameter can be used to disable or enable the soft lockup detector
1492 *and* the NMI watchdog (i.e. the hard lockup detector) at the same time.
1494 = ==============================
1495 0 Disable both lockup detectors.
1496 1 Enable both lockup detectors.
1497 = ==============================
1499 The soft lockup detector and the NMI watchdog can also be disabled or
1500 enabled individually, using the ``soft_watchdog`` and ``nmi_watchdog``
1502 If the ``watchdog`` parameter is read, for example by executing::
1504 cat /proc/sys/kernel/watchdog
1506 the output of this command (0 or 1) shows the logical OR of
1507 ``soft_watchdog`` and ``nmi_watchdog``.
1513 This value can be used to control on which cpus the watchdog may run.
1514 The default cpumask is all possible cores, but if ``NO_HZ_FULL`` is
1515 enabled in the kernel config, and cores are specified with the
1516 ``nohz_full=`` boot argument, those cores are excluded by default.
1517 Offline cores can be included in this mask, and if the core is later
1518 brought online, the watchdog will be started based on the mask value.
1520 Typically this value would only be touched in the ``nohz_full`` case
1521 to re-enable cores that by default were not running the watchdog,
1522 if a kernel lockup was suspected on those cores.
1524 The argument value is the standard cpulist format for cpumasks,
1525 so for example to enable the watchdog on cores 0, 2, 3, and 4 you
1528 echo 0,2-4 > /proc/sys/kernel/watchdog_cpumask
1534 This value can be used to control the frequency of hrtimer and NMI
1535 events and the soft and hard lockup thresholds. The default threshold
1538 The softlockup threshold is (``2 * watchdog_thresh``). Setting this
1539 tunable to zero will disable lockup detection altogether.