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 :doc:`index`.
14 ------------------------------------------------------------------------------
16 This file contains documentation for the sysctl files in
17 ``/proc/sys/kernel/`` and is valid for Linux kernel version 2.2.
19 The files in this directory can be used to tune and monitor
20 miscellaneous and general things in the operation of the Linux
21 kernel. Since some of the files *can* be used to screw up your
22 system, it is advisable to read both documentation and source
23 before actually making adjustments.
25 Currently, these files might (depending on your configuration)
26 show up in ``/proc/sys/kernel``:
36 highwater lowwater frequency
38 If BSD-style process accounting is enabled these values control
39 its behaviour. If free space on filesystem where the log lives
40 goes below ``lowwater``% accounting suspends. If free space gets
41 above ``highwater``% accounting resumes. ``frequency`` determines
42 how often do we check the amount of free space (value is in
49 That is, suspend accounting if free space drops below 2%; resume it
50 if it increases to at least 4%; consider information about amount of
51 free space valid for 30 seconds.
57 See :doc:`/power/video`. This allows the video resume mode to be set,
58 in a similar fashion to the ``acpi_sleep`` kernel parameter, by
59 combining the following values:
71 This variable has no effect and may be removed in future kernel
72 releases. Reading it always returns 0.
73 Up to Linux 3.17, it enabled/disabled automatic recomputing of
75 upon memory add/remove or upon IPC namespace creation/removal.
76 Echoing "1" into this file enabled msgmni automatic recomputing.
77 Echoing "0" turned it off. The default value was 1.
80 bootloader_type (x86 only)
81 ==========================
83 This gives the bootloader type number as indicated by the bootloader,
84 shifted left by 4, and OR'd with the low four bits of the bootloader
85 version. The reason for this encoding is that this used to match the
86 ``type_of_loader`` field in the kernel header; the encoding is kept for
87 backwards compatibility. That is, if the full bootloader type number
88 is 0x15 and the full version number is 0x234, this file will contain
89 the value 340 = 0x154.
91 See the ``type_of_loader`` and ``ext_loader_type`` fields in
92 :doc:`/x86/boot` for additional information.
95 bootloader_version (x86 only)
96 =============================
98 The complete bootloader version number. In the example above, this
99 file will contain the value 564 = 0x234.
101 See the ``type_of_loader`` and ``ext_loader_ver`` fields in
102 :doc:`/x86/boot` for additional information.
108 Controls whether the kernel should collect statistics on BPF programs
109 (total time spent running, number of times run...). Enabling
110 statistics causes a slight reduction in performance on each program
111 run. The statistics can be seen using ``bpftool``.
113 = ===================================
114 0 Don't collect statistics (default).
115 1 Collect statistics.
116 = ===================================
122 This is the pid which will be signalled on reboot (notably, by
123 Ctrl-Alt-Delete). Writing a value to this file which doesn't
124 correspond to a running process will result in ``-ESRCH``.
126 See also `ctrl-alt-del`_.
132 Highest valid capability of the running kernel. Exports
133 ``CAP_LAST_CAP`` from the kernel.
139 ``core_pattern`` is used to specify a core dumpfile pattern name.
141 * max length 127 characters; default value is "core"
142 * ``core_pattern`` is used as a pattern template for the output
143 filename; certain string patterns (beginning with '%') are
144 substituted with their actual values.
145 * backward compatibility with ``core_uses_pid``:
147 If ``core_pattern`` does not include "%p" (default does not)
148 and ``core_uses_pid`` is set, then .PID will be appended to
151 * corename format specifiers
153 ======== ==========================================
154 %<NUL> '%' is dropped
157 %P global pid (init PID namespace)
159 %I global tid (init PID namespace)
160 %u uid (in initial user namespace)
161 %g gid (in initial user namespace)
162 %d dump mode, matches ``PR_SET_DUMPABLE`` and
163 ``/proc/sys/fs/suid_dumpable``
167 %e executable filename (may be shortened)
169 %c maximum size of core file by resource limit RLIMIT_CORE
170 %<OTHER> both are dropped
171 ======== ==========================================
173 * If the first character of the pattern is a '|', the kernel will treat
174 the rest of the pattern as a command to run. The core dump will be
175 written to the standard input of that program instead of to a file.
181 This sysctl is only applicable when `core_pattern`_ is configured to
182 pipe core files to a user space helper (when the first character of
183 ``core_pattern`` is a '|', see above).
184 When collecting cores via a pipe to an application, it is occasionally
185 useful for the collecting application to gather data about the
186 crashing process from its ``/proc/pid`` directory.
187 In order to do this safely, the kernel must wait for the collecting
188 process to exit, so as not to remove the crashing processes proc files
190 This in turn creates the possibility that a misbehaving userspace
191 collecting process can block the reaping of a crashed process simply
193 This sysctl defends against that.
194 It defines how many concurrent crashing processes may be piped to user
195 space applications in parallel.
196 If this value is exceeded, then those crashing processes above that
197 value are noted via the kernel log and their cores are skipped.
198 0 is a special value, indicating that unlimited processes may be
199 captured in parallel, but that no waiting will take place (i.e. the
200 collecting process is not guaranteed access to ``/proc/<crashing
202 This value defaults to 0.
208 The default coredump filename is "core". By setting
209 ``core_uses_pid`` to 1, the coredump filename becomes core.PID.
210 If `core_pattern`_ does not include "%p" (default does not)
211 and ``core_uses_pid`` is set, then .PID will be appended to
218 When the value in this file is 0, ctrl-alt-del is trapped and
219 sent to the ``init(1)`` program to handle a graceful restart.
220 When, however, the value is > 0, Linux's reaction to a Vulcan
221 Nerve Pinch (tm) will be an immediate reboot, without even
222 syncing its dirty buffers.
225 when a program (like dosemu) has the keyboard in 'raw'
226 mode, the ctrl-alt-del is intercepted by the program before it
227 ever reaches the kernel tty layer, and it's up to the program
228 to decide what to do with it.
234 This toggle indicates whether unprivileged users are prevented
235 from using ``dmesg(8)`` to view messages from the kernel's log
237 When ``dmesg_restrict`` is set to 0 there are no restrictions.
238 When ``dmesg_restrict`` is set to 1, users must have
239 ``CAP_SYSLOG`` to use ``dmesg(8)``.
241 The kernel config option ``CONFIG_SECURITY_DMESG_RESTRICT`` sets the
242 default value of ``dmesg_restrict``.
245 domainname & hostname
246 =====================
248 These files can be used to set the NIS/YP domainname and the
249 hostname of your box in exactly the same way as the commands
250 domainname and hostname, i.e.::
252 # echo "darkstar" > /proc/sys/kernel/hostname
253 # echo "mydomain" > /proc/sys/kernel/domainname
255 has the same effect as::
257 # hostname "darkstar"
258 # domainname "mydomain"
260 Note, however, that the classic darkstar.frop.org has the
261 hostname "darkstar" and DNS (Internet Domain Name Server)
262 domainname "frop.org", not to be confused with the NIS (Network
263 Information Service) or YP (Yellow Pages) domainname. These two
264 domain names are in general different. For a detailed discussion
265 see the ``hostname(1)`` man page.
271 See :doc:`/driver-api/firmware/fallback-mechanisms`.
273 The entries in this directory allow the firmware loader helper
274 fallback to be controlled:
276 * ``force_sysfs_fallback``, when set to 1, forces the use of the
278 * ``ignore_sysfs_fallback``, when set to 1, ignores any fallback.
284 Determines whether ``ftrace_dump()`` should be called on an oops (or
285 kernel panic). This will output the contents of the ftrace buffers to
286 the console. This is very useful for capturing traces that lead to
287 crashes and outputting them to a serial console.
289 = ===================================================
290 0 Disabled (default).
291 1 Dump buffers of all CPUs.
292 2 Dump the buffer of the CPU that triggered the oops.
293 = ===================================================
296 ftrace_enabled, stack_tracer_enabled
297 ====================================
299 See :doc:`/trace/ftrace`.
302 hardlockup_all_cpu_backtrace
303 ============================
305 This value controls the hard lockup detector behavior when a hard
306 lockup condition is detected as to whether or not to gather further
307 debug information. If enabled, arch-specific all-CPU stack dumping
310 = ============================================
311 0 Do nothing. This is the default behavior.
312 1 On detection capture more debug information.
313 = ============================================
319 This parameter can be used to control whether the kernel panics
320 when a hard lockup is detected.
322 = ===========================
323 0 Don't panic on hard lockup.
324 1 Panic on hard lockup.
325 = ===========================
327 See :doc:`/admin-guide/lockup-watchdogs` for more information.
328 This can also be set using the nmi_watchdog kernel parameter.
334 Path for the hotplug policy agent.
335 Default value is "``/sbin/hotplug``".
338 hung_task_all_cpu_backtrace
339 ===========================
341 If this option is set, the kernel will send an NMI to all CPUs to dump
342 their backtraces when a hung task is detected. This file shows up if
343 CONFIG_DETECT_HUNG_TASK and CONFIG_SMP are enabled.
345 0: Won't show all CPUs backtraces when a hung task is detected.
346 This is the default behavior.
348 1: Will non-maskably interrupt all CPUs and dump their backtraces when
349 a hung task is detected.
355 Controls the kernel's behavior when a hung task is detected.
356 This file shows up if ``CONFIG_DETECT_HUNG_TASK`` is enabled.
358 = =================================================
359 0 Continue operation. This is the default behavior.
361 = =================================================
364 hung_task_check_count
365 =====================
367 The upper bound on the number of tasks that are checked.
368 This file shows up if ``CONFIG_DETECT_HUNG_TASK`` is enabled.
371 hung_task_timeout_secs
372 ======================
374 When a task in D state did not get scheduled
375 for more than this value report a warning.
376 This file shows up if ``CONFIG_DETECT_HUNG_TASK`` is enabled.
378 0 means infinite timeout, no checking is done.
380 Possible values to set are in range {0:``LONG_MAX``/``HZ``}.
383 hung_task_check_interval_secs
384 =============================
386 Hung task check interval. If hung task checking is enabled
387 (see `hung_task_timeout_secs`_), the check is done every
388 ``hung_task_check_interval_secs`` seconds.
389 This file shows up if ``CONFIG_DETECT_HUNG_TASK`` is enabled.
391 0 (default) means use ``hung_task_timeout_secs`` as checking
394 Possible values to set are in range {0:``LONG_MAX``/``HZ``}.
400 The maximum number of warnings to report. During a check interval
401 if a hung task is detected, this value is decreased by 1.
402 When this value reaches 0, no more warnings will be reported.
403 This file shows up if ``CONFIG_DETECT_HUNG_TASK`` is enabled.
405 -1: report an infinite number of warnings.
408 hyperv_record_panic_msg
409 =======================
411 Controls whether the panic kmsg data should be reported to Hyper-V.
413 = =========================================================
414 0 Do not report panic kmsg data.
415 1 Report the panic kmsg data. This is the default behavior.
416 = =========================================================
419 ignore-unaligned-usertrap
420 =========================
422 On architectures where unaligned accesses cause traps, and where this
423 feature is supported (``CONFIG_SYSCTL_ARCH_UNALIGN_NO_WARN``;
424 currently, ``arc`` and ``ia64``), controls whether all unaligned traps
427 = =============================================================
428 0 Log all unaligned accesses.
429 1 Only warn the first time a process traps. This is the default
431 = =============================================================
433 See also `unaligned-trap`_ and `unaligned-dump-stack`_. On ``ia64``,
434 this allows system administrators to override the
435 ``IA64_THREAD_UAC_NOPRINT`` ``prctl`` and avoid logs being flooded.
441 A toggle indicating if the ``kexec_load`` syscall has been disabled.
442 This value defaults to 0 (false: ``kexec_load`` enabled), but can be
443 set to 1 (true: ``kexec_load`` disabled).
444 Once true, kexec can no longer be used, and the toggle cannot be set
446 This allows a kexec image to be loaded before disabling the syscall,
447 allowing a system to set up (and later use) an image without it being
449 Generally used together with the `modules_disabled`_ sysctl.
455 This toggle indicates whether restrictions are placed on
456 exposing kernel addresses via ``/proc`` and other interfaces.
458 When ``kptr_restrict`` is set to 0 (the default) the address is hashed
460 (This is the equivalent to %p.)
462 When ``kptr_restrict`` is set to 1, kernel pointers printed using the
463 %pK format specifier will be replaced with 0s unless the user has
464 ``CAP_SYSLOG`` and effective user and group ids are equal to the real
466 This is because %pK checks are done at read() time rather than open()
467 time, so if permissions are elevated between the open() and the read()
468 (e.g via a setuid binary) then %pK will not leak kernel pointers to
470 Note, this is a temporary solution only.
471 The correct long-term solution is to do the permission checks at
473 Consider removing world read permissions from files that use %pK, and
474 using `dmesg_restrict`_ to protect against uses of %pK in ``dmesg(8)``
475 if leaking kernel pointer values to unprivileged users is a concern.
477 When ``kptr_restrict`` is set to 2, kernel pointers printed using
478 %pK will be replaced with 0s regardless of privileges.
484 The full path to the usermode helper for autoloading kernel modules,
485 by default "/sbin/modprobe". This binary is executed when the kernel
486 requests a module. For example, if userspace passes an unknown
487 filesystem type to mount(), then the kernel will automatically request
488 the corresponding filesystem module by executing this usermode helper.
489 This usermode helper should insert the needed module into the kernel.
491 This sysctl only affects module autoloading. It has no effect on the
492 ability to explicitly insert modules.
494 This sysctl can be used to debug module loading requests::
496 echo '#! /bin/sh' > /tmp/modprobe
497 echo 'echo "$@" >> /tmp/modprobe.log' >> /tmp/modprobe
498 echo 'exec /sbin/modprobe "$@"' >> /tmp/modprobe
499 chmod a+x /tmp/modprobe
500 echo /tmp/modprobe > /proc/sys/kernel/modprobe
502 Alternatively, if this sysctl is set to the empty string, then module
503 autoloading is completely disabled. The kernel will not try to
504 execute a usermode helper at all, nor will it call the
505 kernel_module_request LSM hook.
507 If CONFIG_STATIC_USERMODEHELPER=y is set in the kernel configuration,
508 then the configured static usermode helper overrides this sysctl,
509 except that the empty string is still accepted to completely disable
510 module autoloading as described above.
515 A toggle value indicating if modules are allowed to be loaded
516 in an otherwise modular kernel. This toggle defaults to off
517 (0), but can be set true (1). Once true, modules can be
518 neither loaded nor unloaded, and the toggle cannot be set back
519 to false. Generally used with the `kexec_load_disabled`_ toggle.
524 msgmax, msgmnb, and msgmni
525 ==========================
527 ``msgmax`` is the maximum size of an IPC message, in bytes. 8192 by
528 default (``MSGMAX``).
530 ``msgmnb`` is the maximum size of an IPC queue, in bytes. 16384 by
531 default (``MSGMNB``).
533 ``msgmni`` is the maximum number of IPC queues. 32000 by default
537 msg_next_id, sem_next_id, and shm_next_id (System V IPC)
538 ========================================================
540 These three toggles allows to specify desired id for next allocated IPC
541 object: message, semaphore or shared memory respectively.
543 By default they are equal to -1, which means generic allocation logic.
544 Possible values to set are in range {0:``INT_MAX``}.
547 1) kernel doesn't guarantee, that new object will have desired id. So,
548 it's up to userspace, how to handle an object with "wrong" id.
549 2) Toggle with non-default value will be set back to -1 by kernel after
550 successful IPC object allocation. If an IPC object allocation syscall
551 fails, it is undefined if the value remains unmodified or is reset to -1.
557 Maximum number of supplementary groups, _i.e._ the maximum size which
558 ``setgroups`` will accept. Exports ``NGROUPS_MAX`` from the kernel.
565 This parameter can be used to control the NMI watchdog
566 (i.e. the hard lockup detector) on x86 systems.
568 = =================================
569 0 Disable the hard lockup detector.
570 1 Enable the hard lockup detector.
571 = =================================
573 The hard lockup detector monitors each CPU for its ability to respond to
574 timer interrupts. The mechanism utilizes CPU performance counter registers
575 that are programmed to generate Non-Maskable Interrupts (NMIs) periodically
576 while a CPU is busy. Hence, the alternative name 'NMI watchdog'.
578 The NMI watchdog is disabled by default if the kernel is running as a guest
579 in a KVM virtual machine. This default can be overridden by adding::
583 to the guest kernel command line (see :doc:`/admin-guide/kernel-parameters`).
589 Enables/disables automatic page fault based NUMA memory
590 balancing. Memory is moved automatically to nodes
591 that access it often.
593 Enables/disables automatic NUMA memory balancing. On NUMA machines, there
594 is a performance penalty if remote memory is accessed by a CPU. When this
595 feature is enabled the kernel samples what task thread is accessing memory
596 by periodically unmapping pages and later trapping a page fault. At the
597 time of the page fault, it is determined if the data being accessed should
598 be migrated to a local memory node.
600 The unmapping of pages and trapping faults incur additional overhead that
601 ideally is offset by improved memory locality but there is no universal
602 guarantee. If the target workload is already bound to NUMA nodes then this
603 feature should be disabled. Otherwise, if the system overhead from the
604 feature is too high then the rate the kernel samples for NUMA hinting
605 faults may be controlled by the `numa_balancing_scan_period_min_ms,
606 numa_balancing_scan_delay_ms, numa_balancing_scan_period_max_ms,
607 numa_balancing_scan_size_mb`_, and numa_balancing_settle_count sysctls.
610 numa_balancing_scan_period_min_ms, numa_balancing_scan_delay_ms, numa_balancing_scan_period_max_ms, numa_balancing_scan_size_mb
611 ===============================================================================================================================
614 Automatic NUMA balancing scans tasks address space and unmaps pages to
615 detect if pages are properly placed or if the data should be migrated to a
616 memory node local to where the task is running. Every "scan delay" the task
617 scans the next "scan size" number of pages in its address space. When the
618 end of the address space is reached the scanner restarts from the beginning.
620 In combination, the "scan delay" and "scan size" determine the scan rate.
621 When "scan delay" decreases, the scan rate increases. The scan delay and
622 hence the scan rate of every task is adaptive and depends on historical
623 behaviour. If pages are properly placed then the scan delay increases,
624 otherwise the scan delay decreases. The "scan size" is not adaptive but
625 the higher the "scan size", the higher the scan rate.
627 Higher scan rates incur higher system overhead as page faults must be
628 trapped and potentially data must be migrated. However, the higher the scan
629 rate, the more quickly a tasks memory is migrated to a local node if the
630 workload pattern changes and minimises performance impact due to remote
631 memory accesses. These sysctls control the thresholds for scan delays and
632 the number of pages scanned.
634 ``numa_balancing_scan_period_min_ms`` is the minimum time in milliseconds to
635 scan a tasks virtual memory. It effectively controls the maximum scanning
638 ``numa_balancing_scan_delay_ms`` is the starting "scan delay" used for a task
639 when it initially forks.
641 ``numa_balancing_scan_period_max_ms`` is the maximum time in milliseconds to
642 scan a tasks virtual memory. It effectively controls the minimum scanning
645 ``numa_balancing_scan_size_mb`` is how many megabytes worth of pages are
646 scanned for a given scan.
649 oops_all_cpu_backtrace
650 ======================
652 If this option is set, the kernel will send an NMI to all CPUs to dump
653 their backtraces when an oops event occurs. It should be used as a last
654 resort in case a panic cannot be triggered (to protect VMs running, for
655 example) or kdump can't be collected. This file shows up if CONFIG_SMP
658 0: Won't show all CPUs backtraces when an oops is detected.
659 This is the default behavior.
661 1: Will non-maskably interrupt all CPUs and dump their backtraces when
662 an oops event is detected.
665 osrelease, ostype & version
666 ===========================
675 #5 Wed Feb 25 21:49:24 MET 1998
677 The files ``osrelease`` and ``ostype`` should be clear enough.
679 needs a little more clarification however. The '#5' means that
680 this is the fifth kernel built from this source base and the
681 date behind it indicates the time the kernel was built.
682 The only way to tune these values is to rebuild the kernel :-)
685 overflowgid & overflowuid
686 =========================
688 if your architecture did not always support 32-bit UIDs (i.e. arm,
689 i386, m68k, sh, and sparc32), a fixed UID and GID will be returned to
690 applications that use the old 16-bit UID/GID system calls, if the
691 actual UID or GID would exceed 65535.
693 These sysctls allow you to change the value of the fixed UID and GID.
694 The default is 65534.
700 The value in this file determines the behaviour of the kernel on a
703 * if zero, the kernel will loop forever;
704 * if negative, the kernel will reboot immediately;
705 * if positive, the kernel will reboot after the corresponding number
708 When you use the software watchdog, the recommended setting is 60.
714 Controls the kernel's behavior when a CPU receives an NMI caused by
717 = ==================================================================
718 0 Try to continue operation (default).
719 1 Panic immediately. The IO error triggered an NMI. This indicates a
720 serious system condition which could result in IO data corruption.
721 Rather than continuing, panicking might be a better choice. Some
722 servers issue this sort of NMI when the dump button is pushed,
723 and you can use this option to take a crash dump.
724 = ==================================================================
730 Controls the kernel's behaviour when an oops or BUG is encountered.
732 = ===================================================================
733 0 Try to continue operation.
734 1 Panic immediately. If the `panic` sysctl is also non-zero then the
735 machine will be rebooted.
736 = ===================================================================
739 panic_on_stackoverflow
740 ======================
742 Controls the kernel's behavior when detecting the overflows of
743 kernel, IRQ and exception stacks except a user stack.
744 This file shows up if ``CONFIG_DEBUG_STACKOVERFLOW`` is enabled.
746 = ==========================
747 0 Try to continue operation.
749 = ==========================
752 panic_on_unrecovered_nmi
753 ========================
755 The default Linux behaviour on an NMI of either memory or unknown is
756 to continue operation. For many environments such as scientific
757 computing it is preferable that the box is taken out and the error
758 dealt with than an uncorrected parity/ECC error get propagated.
760 A small number of systems do generate NMIs for bizarre random reasons
761 such as power management so the default is off. That sysctl works like
762 the existing panic controls already in that directory.
768 Calls panic() in the WARN() path when set to 1. This is useful to avoid
769 a kernel rebuild when attempting to kdump at the location of a WARN().
771 = ================================================
772 0 Only WARN(), default behaviour.
773 1 Call panic() after printing out WARN() location.
774 = ================================================
780 Bitmask for printing system info when panic happens. User can chose
781 combination of the following bits:
783 ===== ============================================
784 bit 0 print all tasks info
785 bit 1 print system memory info
786 bit 2 print timer info
787 bit 3 print locks info if ``CONFIG_LOCKDEP`` is on
788 bit 4 print ftrace buffer
789 ===== ============================================
791 So for example to print tasks and memory info on panic, user can::
793 echo 3 > /proc/sys/kernel/panic_print
799 When set to 1, calls panic() after RCU stall detection messages. This
800 is useful to define the root cause of RCU stalls using a vmcore.
802 = ============================================================
803 0 Do not panic() when RCU stall takes place, default behavior.
804 1 panic() after printing RCU stall messages.
805 = ============================================================
808 perf_cpu_time_max_percent
809 =========================
811 Hints to the kernel how much CPU time it should be allowed to
812 use to handle perf sampling events. If the perf subsystem
813 is informed that its samples are exceeding this limit, it
814 will drop its sampling frequency to attempt to reduce its CPU
817 Some perf sampling happens in NMIs. If these samples
818 unexpectedly take too long to execute, the NMIs can become
819 stacked up next to each other so much that nothing else is
822 ===== ========================================================
823 0 Disable the mechanism. Do not monitor or correct perf's
824 sampling rate no matter how CPU time it takes.
826 1-100 Attempt to throttle perf's sample rate to this
827 percentage of CPU. Note: the kernel calculates an
828 "expected" length of each sample event. 100 here means
829 100% of that expected length. Even if this is set to
830 100, you may still see sample throttling if this
831 length is exceeded. Set to 0 if you truly do not care
832 how much CPU is consumed.
833 ===== ========================================================
839 Controls use of the performance events system by unprivileged
840 users (without CAP_PERFMON). The default value is 2.
842 For backward compatibility reasons access to system performance
843 monitoring and observability remains open for CAP_SYS_ADMIN
844 privileged processes but CAP_SYS_ADMIN usage for secure system
845 performance monitoring and observability operations is discouraged
846 with respect to CAP_PERFMON use cases.
848 === ==================================================================
849 -1 Allow use of (almost) all events by all users.
851 Ignore mlock limit after perf_event_mlock_kb without
854 >=0 Disallow ftrace function tracepoint by users without
857 Disallow raw tracepoint access by users without ``CAP_PERFMON``.
859 >=1 Disallow CPU event access by users without ``CAP_PERFMON``.
861 >=2 Disallow kernel profiling by users without ``CAP_PERFMON``.
862 === ==================================================================
868 Controls maximum number of stack frames to copy for (``attr.sample_type &
869 PERF_SAMPLE_CALLCHAIN``) configured events, for instance, when using
870 '``perf record -g``' or '``perf trace --call-graph fp``'.
872 This can only be done when no events are in use that have callchains
873 enabled, otherwise writing to this file will return ``-EBUSY``.
875 The default value is 127.
881 Control size of per-cpu ring buffer not counted agains mlock limit.
883 The default value is 512 + 1 page
886 perf_event_max_contexts_per_stack
887 =================================
889 Controls maximum number of stack frame context entries for
890 (``attr.sample_type & PERF_SAMPLE_CALLCHAIN``) configured events, for
891 instance, when using '``perf record -g``' or '``perf trace --call-graph fp``'.
893 This can only be done when no events are in use that have callchains
894 enabled, otherwise writing to this file will return ``-EBUSY``.
896 The default value is 8.
902 PID allocation wrap value. When the kernel's next PID value
903 reaches this value, it wraps back to a minimum PID value.
904 PIDs of value ``pid_max`` or larger are not allocated.
910 The last pid allocated in the current (the one task using this sysctl
911 lives in) pid namespace. When selecting a pid for a next task on fork
912 kernel tries to allocate a number starting from this one.
915 powersave-nap (PPC only)
916 ========================
918 If set, Linux-PPC will use the 'nap' mode of powersaving,
919 otherwise the 'doze' mode will be used.
922 ==============================================================
927 The four values in printk denote: ``console_loglevel``,
928 ``default_message_loglevel``, ``minimum_console_loglevel`` and
929 ``default_console_loglevel`` respectively.
931 These values influence printk() behavior when printing or
932 logging error messages. See '``man 2 syslog``' for more info on
933 the different loglevels.
935 ======================== =====================================
936 console_loglevel messages with a higher priority than
937 this will be printed to the console
938 default_message_loglevel messages without an explicit priority
939 will be printed with this priority
940 minimum_console_loglevel minimum (highest) value to which
941 console_loglevel can be set
942 default_console_loglevel default value for console_loglevel
943 ======================== =====================================
949 Delay each printk message in ``printk_delay`` milliseconds
951 Value from 0 - 10000 is allowed.
957 Some warning messages are rate limited. ``printk_ratelimit`` specifies
958 the minimum length of time between these messages (in seconds).
959 The default value is 5 seconds.
961 A value of 0 will disable rate limiting.
964 printk_ratelimit_burst
965 ======================
967 While long term we enforce one message per `printk_ratelimit`_
968 seconds, we do allow a burst of messages to pass through.
969 ``printk_ratelimit_burst`` specifies the number of messages we can
970 send before ratelimiting kicks in.
972 The default value is 10 messages.
978 Control the logging to ``/dev/kmsg`` from userspace:
980 ========= =============================================
981 ratelimit default, ratelimited
982 on unlimited logging to /dev/kmsg from userspace
983 off logging to /dev/kmsg disabled
984 ========= =============================================
986 The kernel command line parameter ``printk.devkmsg=`` overrides this and is
987 a one-time setting until next reboot: once set, it cannot be changed by
988 this sysctl interface anymore.
990 ==============================================================
996 See Documentation/filesystems/devpts.rst.
1002 This is a directory, with the following entries:
1004 * ``boot_id``: a UUID generated the first time this is retrieved, and
1005 unvarying after that;
1007 * ``entropy_avail``: the pool's entropy count, in bits;
1009 * ``poolsize``: the entropy pool size, in bits;
1011 * ``urandom_min_reseed_secs``: obsolete (used to determine the minimum
1012 number of seconds between urandom pool reseeding).
1014 * ``uuid``: a UUID generated every time this is retrieved (this can
1015 thus be used to generate UUIDs at will);
1017 * ``write_wakeup_threshold``: when the entropy count drops below this
1018 (as a number of bits), processes waiting to write to ``/dev/random``
1021 If ``drivers/char/random.c`` is built with ``ADD_INTERRUPT_BENCH``
1022 defined, these additional entries are present:
1024 * ``add_interrupt_avg_cycles``: the average number of cycles between
1025 interrupts used to feed the pool;
1027 * ``add_interrupt_avg_deviation``: the standard deviation seen on the
1028 number of cycles between interrupts used to feed the pool.
1034 This option can be used to select the type of process address
1035 space randomization that is used in the system, for architectures
1036 that support this feature.
1038 == ===========================================================================
1039 0 Turn the process address space randomization off. This is the
1040 default for architectures that do not support this feature anyways,
1041 and kernels that are booted with the "norandmaps" parameter.
1043 1 Make the addresses of mmap base, stack and VDSO page randomized.
1044 This, among other things, implies that shared libraries will be
1045 loaded to random addresses. Also for PIE-linked binaries, the
1046 location of code start is randomized. This is the default if the
1047 ``CONFIG_COMPAT_BRK`` option is enabled.
1049 2 Additionally enable heap randomization. This is the default if
1050 ``CONFIG_COMPAT_BRK`` is disabled.
1052 There are a few legacy applications out there (such as some ancient
1053 versions of libc.so.5 from 1996) that assume that brk area starts
1054 just after the end of the code+bss. These applications break when
1055 start of the brk area is randomized. There are however no known
1056 non-legacy applications that would be broken this way, so for most
1057 systems it is safe to choose full randomization.
1059 Systems with ancient and/or broken binaries should be configured
1060 with ``CONFIG_COMPAT_BRK`` enabled, which excludes the heap from process
1061 address space randomization.
1062 == ===========================================================================
1068 See :doc:`/admin-guide/initrd`.
1071 reboot-cmd (SPARC only)
1072 =======================
1074 ??? This seems to be a way to give an argument to the Sparc
1075 ROM/Flash boot loader. Maybe to tell it what to do after
1082 Enables/disables Energy Aware Scheduling (EAS). EAS starts
1083 automatically on platforms where it can run (that is,
1084 platforms with asymmetric CPU topologies and having an Energy
1085 Model available). If your platform happens to meet the
1086 requirements for EAS but you do not want to use it, change
1093 Enables/disables scheduler statistics. Enabling this feature
1094 incurs a small amount of overhead in the scheduler but is
1095 useful for debugging and performance tuning.
1097 sched_util_clamp_min:
1098 =====================
1100 Max allowed *minimum* utilization.
1102 Default value is 1024, which is the maximum possible value.
1104 It means that any requested uclamp.min value cannot be greater than
1105 sched_util_clamp_min, i.e., it is restricted to the range
1106 [0:sched_util_clamp_min].
1108 sched_util_clamp_max:
1109 =====================
1111 Max allowed *maximum* utilization.
1113 Default value is 1024, which is the maximum possible value.
1115 It means that any requested uclamp.max value cannot be greater than
1116 sched_util_clamp_max, i.e., it is restricted to the range
1117 [0:sched_util_clamp_max].
1119 sched_util_clamp_min_rt_default:
1120 ================================
1122 By default Linux is tuned for performance. Which means that RT tasks always run
1123 at the highest frequency and most capable (highest capacity) CPU (in
1124 heterogeneous systems).
1126 Uclamp achieves this by setting the requested uclamp.min of all RT tasks to
1127 1024 by default, which effectively boosts the tasks to run at the highest
1128 frequency and biases them to run on the biggest CPU.
1130 This knob allows admins to change the default behavior when uclamp is being
1131 used. In battery powered devices particularly, running at the maximum
1132 capacity and frequency will increase energy consumption and shorten the battery
1135 This knob is only effective for RT tasks which the user hasn't modified their
1136 requested uclamp.min value via sched_setattr() syscall.
1138 This knob will not escape the range constraint imposed by sched_util_clamp_min
1143 sched_util_clamp_min_rt_default = 800
1144 sched_util_clamp_min = 600
1146 Then the boost will be clamped to 600 because 800 is outside of the permissible
1147 range of [0:600]. This could happen for instance if a powersave mode will
1148 restrict all boosts temporarily by modifying sched_util_clamp_min. As soon as
1149 this restriction is lifted, the requested sched_util_clamp_min_rt_default
1155 See :doc:`/userspace-api/seccomp_filter`.
1161 This file shows the size of the generic SCSI (sg) buffer.
1162 You can't tune it just yet, but you could change it on
1163 compile time by editing ``include/scsi/sg.h`` and changing
1164 the value of ``SG_BIG_BUFF``.
1166 There shouldn't be any reason to change this value. If
1167 you can come up with one, you probably know what you
1174 This parameter sets the total amount of shared memory pages that
1175 can be used system wide. Hence, ``shmall`` should always be at least
1176 ``ceil(shmmax/PAGE_SIZE)``.
1178 If you are not sure what the default ``PAGE_SIZE`` is on your Linux
1179 system, you can run the following command::
1187 This value can be used to query and set the run time limit
1188 on the maximum shared memory segment size that can be created.
1189 Shared memory segments up to 1Gb are now supported in the
1190 kernel. This value defaults to ``SHMMAX``.
1196 This value determines the maximum number of shared memory segments.
1197 4096 by default (``SHMMNI``).
1203 Linux lets you set resource limits, including how much memory one
1204 process can consume, via ``setrlimit(2)``. Unfortunately, shared memory
1205 segments are allowed to exist without association with any process, and
1206 thus might not be counted against any resource limits. If enabled,
1207 shared memory segments are automatically destroyed when their attach
1208 count becomes zero after a detach or a process termination. It will
1209 also destroy segments that were created, but never attached to, on exit
1210 from the process. The only use left for ``IPC_RMID`` is to immediately
1211 destroy an unattached segment. Of course, this breaks the way things are
1212 defined, so some applications might stop working. Note that this
1213 feature will do you no good unless you also configure your resource
1214 limits (in particular, ``RLIMIT_AS`` and ``RLIMIT_NPROC``). Most systems don't
1217 Note that if you change this from 0 to 1, already created segments
1218 without users and with a dead originative process will be destroyed.
1221 sysctl_writes_strict
1222 ====================
1224 Control how file position affects the behavior of updating sysctl values
1225 via the ``/proc/sys`` interface:
1227 == ======================================================================
1228 -1 Legacy per-write sysctl value handling, with no printk warnings.
1229 Each write syscall must fully contain the sysctl value to be
1230 written, and multiple writes on the same sysctl file descriptor
1231 will rewrite the sysctl value, regardless of file position.
1232 0 Same behavior as above, but warn about processes that perform writes
1233 to a sysctl file descriptor when the file position is not 0.
1234 1 (default) Respect file position when writing sysctl strings. Multiple
1235 writes will append to the sysctl value buffer. Anything past the max
1236 length of the sysctl value buffer will be ignored. Writes to numeric
1237 sysctl entries must always be at file position 0 and the value must
1238 be fully contained in the buffer sent in the write syscall.
1239 == ======================================================================
1242 softlockup_all_cpu_backtrace
1243 ============================
1245 This value controls the soft lockup detector thread's behavior
1246 when a soft lockup condition is detected as to whether or not
1247 to gather further debug information. If enabled, each cpu will
1248 be issued an NMI and instructed to capture stack trace.
1250 This feature is only applicable for architectures which support
1253 = ============================================
1254 0 Do nothing. This is the default behavior.
1255 1 On detection capture more debug information.
1256 = ============================================
1262 This parameter can be used to control whether the kernel panics
1263 when a soft lockup is detected.
1265 = ============================================
1266 0 Don't panic on soft lockup.
1267 1 Panic on soft lockup.
1268 = ============================================
1270 This can also be set using the softlockup_panic kernel parameter.
1276 This parameter can be used to control the soft lockup detector.
1278 = =================================
1279 0 Disable the soft lockup detector.
1280 1 Enable the soft lockup detector.
1281 = =================================
1283 The soft lockup detector monitors CPUs for threads that are hogging the CPUs
1284 without rescheduling voluntarily, and thus prevent the 'watchdog/N' threads
1285 from running. The mechanism depends on the CPUs ability to respond to timer
1286 interrupts which are needed for the 'watchdog/N' threads to be woken up by
1287 the watchdog timer function, otherwise the NMI watchdog — if enabled — can
1288 detect a hard lockup condition.
1294 This parameter can be used to control kernel stack erasing at the end
1295 of syscalls for kernels built with ``CONFIG_GCC_PLUGIN_STACKLEAK``.
1297 That erasing reduces the information which kernel stack leak bugs
1298 can reveal and blocks some uninitialized stack variable attacks.
1299 The tradeoff is the performance impact: on a single CPU system kernel
1300 compilation sees a 1% slowdown, other systems and workloads may vary.
1302 = ====================================================================
1303 0 Kernel stack erasing is disabled, STACKLEAK_METRICS are not updated.
1304 1 Kernel stack erasing is enabled (default), it is performed before
1305 returning to the userspace at the end of syscalls.
1306 = ====================================================================
1314 = ====================================
1315 0 Stop-A has no effect.
1316 1 Stop-A breaks to the PROM (default).
1317 = ====================================
1319 Stop-A is always enabled on a panic, so that the user can return to
1326 See :doc:`/admin-guide/sysrq`.
1332 Non-zero if the kernel has been tainted. Numeric values, which can be
1333 ORed together. The letters are seen in "Tainted" line of Oops reports.
1335 ====== ===== ==============================================================
1336 1 `(P)` proprietary module was loaded
1337 2 `(F)` module was force loaded
1338 4 `(S)` SMP kernel oops on an officially SMP incapable processor
1339 8 `(R)` module was force unloaded
1340 16 `(M)` processor reported a Machine Check Exception (MCE)
1341 32 `(B)` bad page referenced or some unexpected page flags
1342 64 `(U)` taint requested by userspace application
1343 128 `(D)` kernel died recently, i.e. there was an OOPS or BUG
1344 256 `(A)` an ACPI table was overridden by user
1345 512 `(W)` kernel issued warning
1346 1024 `(C)` staging driver was loaded
1347 2048 `(I)` workaround for bug in platform firmware applied
1348 4096 `(O)` externally-built ("out-of-tree") module was loaded
1349 8192 `(E)` unsigned module was loaded
1350 16384 `(L)` soft lockup occurred
1351 32768 `(K)` kernel has been live patched
1352 65536 `(X)` Auxiliary taint, defined and used by for distros
1353 131072 `(T)` The kernel was built with the struct randomization plugin
1354 ====== ===== ==============================================================
1356 See :doc:`/admin-guide/tainted-kernels` for more information.
1359 writes to this sysctl interface will fail with ``EINVAL`` if the kernel is
1360 booted with the command line option ``panic_on_taint=<bitmask>,nousertaint``
1361 and any of the ORed together values being written to ``tainted`` match with
1362 the bitmask declared on panic_on_taint.
1363 See :doc:`/admin-guide/kernel-parameters` for more details on that particular
1364 kernel command line option and its optional ``nousertaint`` switch.
1369 This value controls the maximum number of threads that can be created
1372 During initialization the kernel sets this value such that even if the
1373 maximum number of threads is created, the thread structures occupy only
1374 a part (1/8th) of the available RAM pages.
1376 The minimum value that can be written to ``threads-max`` is 1.
1378 The maximum value that can be written to ``threads-max`` is given by the
1379 constant ``FUTEX_TID_MASK`` (0x3fffffff).
1381 If a value outside of this range is written to ``threads-max`` an
1382 ``EINVAL`` error occurs.
1388 When set, disables tracing (see :doc:`/trace/ftrace`) when a
1395 When tracepoints are sent to printk() (enabled by the ``tp_printk``
1396 boot parameter), this entry provides runtime control::
1398 echo 0 > /proc/sys/kernel/tracepoint_printk
1400 will stop tracepoints from being sent to printk(), and::
1402 echo 1 > /proc/sys/kernel/tracepoint_printk
1404 will send them to printk() again.
1406 This only works if the kernel was booted with ``tp_printk`` enabled.
1408 See :doc:`/admin-guide/kernel-parameters` and
1409 :doc:`/trace/boottime-trace`.
1412 .. _unaligned-dump-stack:
1414 unaligned-dump-stack (ia64)
1415 ===========================
1417 When logging unaligned accesses, controls whether the stack is
1420 = ===================================================
1421 0 Do not dump the stack. This is the default setting.
1423 = ===================================================
1425 See also `ignore-unaligned-usertrap`_.
1431 On architectures where unaligned accesses cause traps, and where this
1432 feature is supported (``CONFIG_SYSCTL_ARCH_UNALIGN_ALLOW``; currently,
1433 ``arc`` and ``parisc``), controls whether unaligned traps are caught
1434 and emulated (instead of failing).
1436 = ========================================================
1437 0 Do not emulate unaligned accesses.
1438 1 Emulate unaligned accesses. This is the default setting.
1439 = ========================================================
1441 See also `ignore-unaligned-usertrap`_.
1447 The value in this file affects behavior of handling NMI. When the
1448 value is non-zero, unknown NMI is trapped and then panic occurs. At
1449 that time, kernel debugging information is displayed on console.
1451 NMI switch that most IA32 servers have fires unknown NMI up, for
1452 example. If a system hangs up, try pressing the NMI switch.
1455 unprivileged_bpf_disabled
1456 =========================
1458 Writing 1 to this entry will disable unprivileged calls to ``bpf()``;
1459 once disabled, calling ``bpf()`` without ``CAP_SYS_ADMIN`` will return
1462 Once set, this can't be cleared.
1468 This parameter can be used to disable or enable the soft lockup detector
1469 *and* the NMI watchdog (i.e. the hard lockup detector) at the same time.
1471 = ==============================
1472 0 Disable both lockup detectors.
1473 1 Enable both lockup detectors.
1474 = ==============================
1476 The soft lockup detector and the NMI watchdog can also be disabled or
1477 enabled individually, using the ``soft_watchdog`` and ``nmi_watchdog``
1479 If the ``watchdog`` parameter is read, for example by executing::
1481 cat /proc/sys/kernel/watchdog
1483 the output of this command (0 or 1) shows the logical OR of
1484 ``soft_watchdog`` and ``nmi_watchdog``.
1490 This value can be used to control on which cpus the watchdog may run.
1491 The default cpumask is all possible cores, but if ``NO_HZ_FULL`` is
1492 enabled in the kernel config, and cores are specified with the
1493 ``nohz_full=`` boot argument, those cores are excluded by default.
1494 Offline cores can be included in this mask, and if the core is later
1495 brought online, the watchdog will be started based on the mask value.
1497 Typically this value would only be touched in the ``nohz_full`` case
1498 to re-enable cores that by default were not running the watchdog,
1499 if a kernel lockup was suspected on those cores.
1501 The argument value is the standard cpulist format for cpumasks,
1502 so for example to enable the watchdog on cores 0, 2, 3, and 4 you
1505 echo 0,2-4 > /proc/sys/kernel/watchdog_cpumask
1511 This value can be used to control the frequency of hrtimer and NMI
1512 events and the soft and hard lockup thresholds. The default threshold
1515 The softlockup threshold is (``2 * watchdog_thresh``). Setting this
1516 tunable to zero will disable lockup detection altogether.