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/``.
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, could be changed by prctl etc)
168 %f executable filename
170 %c maximum size of core file by resource limit RLIMIT_CORE
171 %<OTHER> both are dropped
172 ======== ==========================================
174 * If the first character of the pattern is a '|', the kernel will treat
175 the rest of the pattern as a command to run. The core dump will be
176 written to the standard input of that program instead of to a file.
182 This sysctl is only applicable when `core_pattern`_ is configured to
183 pipe core files to a user space helper (when the first character of
184 ``core_pattern`` is a '|', see above).
185 When collecting cores via a pipe to an application, it is occasionally
186 useful for the collecting application to gather data about the
187 crashing process from its ``/proc/pid`` directory.
188 In order to do this safely, the kernel must wait for the collecting
189 process to exit, so as not to remove the crashing processes proc files
191 This in turn creates the possibility that a misbehaving userspace
192 collecting process can block the reaping of a crashed process simply
194 This sysctl defends against that.
195 It defines how many concurrent crashing processes may be piped to user
196 space applications in parallel.
197 If this value is exceeded, then those crashing processes above that
198 value are noted via the kernel log and their cores are skipped.
199 0 is a special value, indicating that unlimited processes may be
200 captured in parallel, but that no waiting will take place (i.e. the
201 collecting process is not guaranteed access to ``/proc/<crashing
203 This value defaults to 0.
209 The default coredump filename is "core". By setting
210 ``core_uses_pid`` to 1, the coredump filename becomes core.PID.
211 If `core_pattern`_ does not include "%p" (default does not)
212 and ``core_uses_pid`` is set, then .PID will be appended to
219 When the value in this file is 0, ctrl-alt-del is trapped and
220 sent to the ``init(1)`` program to handle a graceful restart.
221 When, however, the value is > 0, Linux's reaction to a Vulcan
222 Nerve Pinch (tm) will be an immediate reboot, without even
223 syncing its dirty buffers.
226 when a program (like dosemu) has the keyboard in 'raw'
227 mode, the ctrl-alt-del is intercepted by the program before it
228 ever reaches the kernel tty layer, and it's up to the program
229 to decide what to do with it.
235 This toggle indicates whether unprivileged users are prevented
236 from using ``dmesg(8)`` to view messages from the kernel's log
238 When ``dmesg_restrict`` is set to 0 there are no restrictions.
239 When ``dmesg_restrict`` is set to 1, users must have
240 ``CAP_SYSLOG`` to use ``dmesg(8)``.
242 The kernel config option ``CONFIG_SECURITY_DMESG_RESTRICT`` sets the
243 default value of ``dmesg_restrict``.
246 domainname & hostname
247 =====================
249 These files can be used to set the NIS/YP domainname and the
250 hostname of your box in exactly the same way as the commands
251 domainname and hostname, i.e.::
253 # echo "darkstar" > /proc/sys/kernel/hostname
254 # echo "mydomain" > /proc/sys/kernel/domainname
256 has the same effect as::
258 # hostname "darkstar"
259 # domainname "mydomain"
261 Note, however, that the classic darkstar.frop.org has the
262 hostname "darkstar" and DNS (Internet Domain Name Server)
263 domainname "frop.org", not to be confused with the NIS (Network
264 Information Service) or YP (Yellow Pages) domainname. These two
265 domain names are in general different. For a detailed discussion
266 see the ``hostname(1)`` man page.
272 See :doc:`/driver-api/firmware/fallback-mechanisms`.
274 The entries in this directory allow the firmware loader helper
275 fallback to be controlled:
277 * ``force_sysfs_fallback``, when set to 1, forces the use of the
279 * ``ignore_sysfs_fallback``, when set to 1, ignores any fallback.
285 Determines whether ``ftrace_dump()`` should be called on an oops (or
286 kernel panic). This will output the contents of the ftrace buffers to
287 the console. This is very useful for capturing traces that lead to
288 crashes and outputting them to a serial console.
290 = ===================================================
291 0 Disabled (default).
292 1 Dump buffers of all CPUs.
293 2 Dump the buffer of the CPU that triggered the oops.
294 = ===================================================
297 ftrace_enabled, stack_tracer_enabled
298 ====================================
300 See :doc:`/trace/ftrace`.
303 hardlockup_all_cpu_backtrace
304 ============================
306 This value controls the hard lockup detector behavior when a hard
307 lockup condition is detected as to whether or not to gather further
308 debug information. If enabled, arch-specific all-CPU stack dumping
311 = ============================================
312 0 Do nothing. This is the default behavior.
313 1 On detection capture more debug information.
314 = ============================================
320 This parameter can be used to control whether the kernel panics
321 when a hard lockup is detected.
323 = ===========================
324 0 Don't panic on hard lockup.
325 1 Panic on hard lockup.
326 = ===========================
328 See :doc:`/admin-guide/lockup-watchdogs` for more information.
329 This can also be set using the nmi_watchdog kernel parameter.
335 Path for the hotplug policy agent.
336 Default value is "``/sbin/hotplug``".
339 hung_task_all_cpu_backtrace
340 ===========================
342 If this option is set, the kernel will send an NMI to all CPUs to dump
343 their backtraces when a hung task is detected. This file shows up if
344 CONFIG_DETECT_HUNG_TASK and CONFIG_SMP are enabled.
346 0: Won't show all CPUs backtraces when a hung task is detected.
347 This is the default behavior.
349 1: Will non-maskably interrupt all CPUs and dump their backtraces when
350 a hung task is detected.
356 Controls the kernel's behavior when a hung task is detected.
357 This file shows up if ``CONFIG_DETECT_HUNG_TASK`` is enabled.
359 = =================================================
360 0 Continue operation. This is the default behavior.
362 = =================================================
365 hung_task_check_count
366 =====================
368 The upper bound on the number of tasks that are checked.
369 This file shows up if ``CONFIG_DETECT_HUNG_TASK`` is enabled.
372 hung_task_timeout_secs
373 ======================
375 When a task in D state did not get scheduled
376 for more than this value report a warning.
377 This file shows up if ``CONFIG_DETECT_HUNG_TASK`` is enabled.
379 0 means infinite timeout, no checking is done.
381 Possible values to set are in range {0:``LONG_MAX``/``HZ``}.
384 hung_task_check_interval_secs
385 =============================
387 Hung task check interval. If hung task checking is enabled
388 (see `hung_task_timeout_secs`_), the check is done every
389 ``hung_task_check_interval_secs`` seconds.
390 This file shows up if ``CONFIG_DETECT_HUNG_TASK`` is enabled.
392 0 (default) means use ``hung_task_timeout_secs`` as checking
395 Possible values to set are in range {0:``LONG_MAX``/``HZ``}.
401 The maximum number of warnings to report. During a check interval
402 if a hung task is detected, this value is decreased by 1.
403 When this value reaches 0, no more warnings will be reported.
404 This file shows up if ``CONFIG_DETECT_HUNG_TASK`` is enabled.
406 -1: report an infinite number of warnings.
409 hyperv_record_panic_msg
410 =======================
412 Controls whether the panic kmsg data should be reported to Hyper-V.
414 = =========================================================
415 0 Do not report panic kmsg data.
416 1 Report the panic kmsg data. This is the default behavior.
417 = =========================================================
420 ignore-unaligned-usertrap
421 =========================
423 On architectures where unaligned accesses cause traps, and where this
424 feature is supported (``CONFIG_SYSCTL_ARCH_UNALIGN_NO_WARN``;
425 currently, ``arc`` and ``ia64``), controls whether all unaligned traps
428 = =============================================================
429 0 Log all unaligned accesses.
430 1 Only warn the first time a process traps. This is the default
432 = =============================================================
434 See also `unaligned-trap`_ and `unaligned-dump-stack`_. On ``ia64``,
435 this allows system administrators to override the
436 ``IA64_THREAD_UAC_NOPRINT`` ``prctl`` and avoid logs being flooded.
442 A toggle indicating if the ``kexec_load`` syscall has been disabled.
443 This value defaults to 0 (false: ``kexec_load`` enabled), but can be
444 set to 1 (true: ``kexec_load`` disabled).
445 Once true, kexec can no longer be used, and the toggle cannot be set
447 This allows a kexec image to be loaded before disabling the syscall,
448 allowing a system to set up (and later use) an image without it being
450 Generally used together with the `modules_disabled`_ sysctl.
456 This toggle indicates whether restrictions are placed on
457 exposing kernel addresses via ``/proc`` and other interfaces.
459 When ``kptr_restrict`` is set to 0 (the default) the address is hashed
461 (This is the equivalent to %p.)
463 When ``kptr_restrict`` is set to 1, kernel pointers printed using the
464 %pK format specifier will be replaced with 0s unless the user has
465 ``CAP_SYSLOG`` and effective user and group ids are equal to the real
467 This is because %pK checks are done at read() time rather than open()
468 time, so if permissions are elevated between the open() and the read()
469 (e.g via a setuid binary) then %pK will not leak kernel pointers to
471 Note, this is a temporary solution only.
472 The correct long-term solution is to do the permission checks at
474 Consider removing world read permissions from files that use %pK, and
475 using `dmesg_restrict`_ to protect against uses of %pK in ``dmesg(8)``
476 if leaking kernel pointer values to unprivileged users is a concern.
478 When ``kptr_restrict`` is set to 2, kernel pointers printed using
479 %pK will be replaced with 0s regardless of privileges.
485 The full path to the usermode helper for autoloading kernel modules,
486 by default "/sbin/modprobe". This binary is executed when the kernel
487 requests a module. For example, if userspace passes an unknown
488 filesystem type to mount(), then the kernel will automatically request
489 the corresponding filesystem module by executing this usermode helper.
490 This usermode helper should insert the needed module into the kernel.
492 This sysctl only affects module autoloading. It has no effect on the
493 ability to explicitly insert modules.
495 This sysctl can be used to debug module loading requests::
497 echo '#! /bin/sh' > /tmp/modprobe
498 echo 'echo "$@" >> /tmp/modprobe.log' >> /tmp/modprobe
499 echo 'exec /sbin/modprobe "$@"' >> /tmp/modprobe
500 chmod a+x /tmp/modprobe
501 echo /tmp/modprobe > /proc/sys/kernel/modprobe
503 Alternatively, if this sysctl is set to the empty string, then module
504 autoloading is completely disabled. The kernel will not try to
505 execute a usermode helper at all, nor will it call the
506 kernel_module_request LSM hook.
508 If CONFIG_STATIC_USERMODEHELPER=y is set in the kernel configuration,
509 then the configured static usermode helper overrides this sysctl,
510 except that the empty string is still accepted to completely disable
511 module autoloading as described above.
516 A toggle value indicating if modules are allowed to be loaded
517 in an otherwise modular kernel. This toggle defaults to off
518 (0), but can be set true (1). Once true, modules can be
519 neither loaded nor unloaded, and the toggle cannot be set back
520 to false. Generally used with the `kexec_load_disabled`_ toggle.
525 msgmax, msgmnb, and msgmni
526 ==========================
528 ``msgmax`` is the maximum size of an IPC message, in bytes. 8192 by
529 default (``MSGMAX``).
531 ``msgmnb`` is the maximum size of an IPC queue, in bytes. 16384 by
532 default (``MSGMNB``).
534 ``msgmni`` is the maximum number of IPC queues. 32000 by default
538 msg_next_id, sem_next_id, and shm_next_id (System V IPC)
539 ========================================================
541 These three toggles allows to specify desired id for next allocated IPC
542 object: message, semaphore or shared memory respectively.
544 By default they are equal to -1, which means generic allocation logic.
545 Possible values to set are in range {0:``INT_MAX``}.
548 1) kernel doesn't guarantee, that new object will have desired id. So,
549 it's up to userspace, how to handle an object with "wrong" id.
550 2) Toggle with non-default value will be set back to -1 by kernel after
551 successful IPC object allocation. If an IPC object allocation syscall
552 fails, it is undefined if the value remains unmodified or is reset to -1.
558 Maximum number of supplementary groups, _i.e._ the maximum size which
559 ``setgroups`` will accept. Exports ``NGROUPS_MAX`` from the kernel.
566 This parameter can be used to control the NMI watchdog
567 (i.e. the hard lockup detector) on x86 systems.
569 = =================================
570 0 Disable the hard lockup detector.
571 1 Enable the hard lockup detector.
572 = =================================
574 The hard lockup detector monitors each CPU for its ability to respond to
575 timer interrupts. The mechanism utilizes CPU performance counter registers
576 that are programmed to generate Non-Maskable Interrupts (NMIs) periodically
577 while a CPU is busy. Hence, the alternative name 'NMI watchdog'.
579 The NMI watchdog is disabled by default if the kernel is running as a guest
580 in a KVM virtual machine. This default can be overridden by adding::
584 to the guest kernel command line (see :doc:`/admin-guide/kernel-parameters`).
590 Enables/disables automatic page fault based NUMA memory
591 balancing. Memory is moved automatically to nodes
592 that access it often.
594 Enables/disables automatic NUMA memory balancing. On NUMA machines, there
595 is a performance penalty if remote memory is accessed by a CPU. When this
596 feature is enabled the kernel samples what task thread is accessing memory
597 by periodically unmapping pages and later trapping a page fault. At the
598 time of the page fault, it is determined if the data being accessed should
599 be migrated to a local memory node.
601 The unmapping of pages and trapping faults incur additional overhead that
602 ideally is offset by improved memory locality but there is no universal
603 guarantee. If the target workload is already bound to NUMA nodes then this
604 feature should be disabled. Otherwise, if the system overhead from the
605 feature is too high then the rate the kernel samples for NUMA hinting
606 faults may be controlled by the `numa_balancing_scan_period_min_ms,
607 numa_balancing_scan_delay_ms, numa_balancing_scan_period_max_ms,
608 numa_balancing_scan_size_mb`_, and numa_balancing_settle_count sysctls.
611 numa_balancing_scan_period_min_ms, numa_balancing_scan_delay_ms, numa_balancing_scan_period_max_ms, numa_balancing_scan_size_mb
612 ===============================================================================================================================
615 Automatic NUMA balancing scans tasks address space and unmaps pages to
616 detect if pages are properly placed or if the data should be migrated to a
617 memory node local to where the task is running. Every "scan delay" the task
618 scans the next "scan size" number of pages in its address space. When the
619 end of the address space is reached the scanner restarts from the beginning.
621 In combination, the "scan delay" and "scan size" determine the scan rate.
622 When "scan delay" decreases, the scan rate increases. The scan delay and
623 hence the scan rate of every task is adaptive and depends on historical
624 behaviour. If pages are properly placed then the scan delay increases,
625 otherwise the scan delay decreases. The "scan size" is not adaptive but
626 the higher the "scan size", the higher the scan rate.
628 Higher scan rates incur higher system overhead as page faults must be
629 trapped and potentially data must be migrated. However, the higher the scan
630 rate, the more quickly a tasks memory is migrated to a local node if the
631 workload pattern changes and minimises performance impact due to remote
632 memory accesses. These sysctls control the thresholds for scan delays and
633 the number of pages scanned.
635 ``numa_balancing_scan_period_min_ms`` is the minimum time in milliseconds to
636 scan a tasks virtual memory. It effectively controls the maximum scanning
639 ``numa_balancing_scan_delay_ms`` is the starting "scan delay" used for a task
640 when it initially forks.
642 ``numa_balancing_scan_period_max_ms`` is the maximum time in milliseconds to
643 scan a tasks virtual memory. It effectively controls the minimum scanning
646 ``numa_balancing_scan_size_mb`` is how many megabytes worth of pages are
647 scanned for a given scan.
650 oops_all_cpu_backtrace
651 ======================
653 If this option is set, the kernel will send an NMI to all CPUs to dump
654 their backtraces when an oops event occurs. It should be used as a last
655 resort in case a panic cannot be triggered (to protect VMs running, for
656 example) or kdump can't be collected. This file shows up if CONFIG_SMP
659 0: Won't show all CPUs backtraces when an oops is detected.
660 This is the default behavior.
662 1: Will non-maskably interrupt all CPUs and dump their backtraces when
663 an oops event is detected.
666 osrelease, ostype & version
667 ===========================
676 #5 Wed Feb 25 21:49:24 MET 1998
678 The files ``osrelease`` and ``ostype`` should be clear enough.
680 needs a little more clarification however. The '#5' means that
681 this is the fifth kernel built from this source base and the
682 date behind it indicates the time the kernel was built.
683 The only way to tune these values is to rebuild the kernel :-)
686 overflowgid & overflowuid
687 =========================
689 if your architecture did not always support 32-bit UIDs (i.e. arm,
690 i386, m68k, sh, and sparc32), a fixed UID and GID will be returned to
691 applications that use the old 16-bit UID/GID system calls, if the
692 actual UID or GID would exceed 65535.
694 These sysctls allow you to change the value of the fixed UID and GID.
695 The default is 65534.
701 The value in this file determines the behaviour of the kernel on a
704 * if zero, the kernel will loop forever;
705 * if negative, the kernel will reboot immediately;
706 * if positive, the kernel will reboot after the corresponding number
709 When you use the software watchdog, the recommended setting is 60.
715 Controls the kernel's behavior when a CPU receives an NMI caused by
718 = ==================================================================
719 0 Try to continue operation (default).
720 1 Panic immediately. The IO error triggered an NMI. This indicates a
721 serious system condition which could result in IO data corruption.
722 Rather than continuing, panicking might be a better choice. Some
723 servers issue this sort of NMI when the dump button is pushed,
724 and you can use this option to take a crash dump.
725 = ==================================================================
731 Controls the kernel's behaviour when an oops or BUG is encountered.
733 = ===================================================================
734 0 Try to continue operation.
735 1 Panic immediately. If the `panic` sysctl is also non-zero then the
736 machine will be rebooted.
737 = ===================================================================
740 panic_on_stackoverflow
741 ======================
743 Controls the kernel's behavior when detecting the overflows of
744 kernel, IRQ and exception stacks except a user stack.
745 This file shows up if ``CONFIG_DEBUG_STACKOVERFLOW`` is enabled.
747 = ==========================
748 0 Try to continue operation.
750 = ==========================
753 panic_on_unrecovered_nmi
754 ========================
756 The default Linux behaviour on an NMI of either memory or unknown is
757 to continue operation. For many environments such as scientific
758 computing it is preferable that the box is taken out and the error
759 dealt with than an uncorrected parity/ECC error get propagated.
761 A small number of systems do generate NMIs for bizarre random reasons
762 such as power management so the default is off. That sysctl works like
763 the existing panic controls already in that directory.
769 Calls panic() in the WARN() path when set to 1. This is useful to avoid
770 a kernel rebuild when attempting to kdump at the location of a WARN().
772 = ================================================
773 0 Only WARN(), default behaviour.
774 1 Call panic() after printing out WARN() location.
775 = ================================================
781 Bitmask for printing system info when panic happens. User can chose
782 combination of the following bits:
784 ===== ============================================
785 bit 0 print all tasks info
786 bit 1 print system memory info
787 bit 2 print timer info
788 bit 3 print locks info if ``CONFIG_LOCKDEP`` is on
789 bit 4 print ftrace buffer
790 ===== ============================================
792 So for example to print tasks and memory info on panic, user can::
794 echo 3 > /proc/sys/kernel/panic_print
800 When set to 1, calls panic() after RCU stall detection messages. This
801 is useful to define the root cause of RCU stalls using a vmcore.
803 = ============================================================
804 0 Do not panic() when RCU stall takes place, default behavior.
805 1 panic() after printing RCU stall messages.
806 = ============================================================
809 perf_cpu_time_max_percent
810 =========================
812 Hints to the kernel how much CPU time it should be allowed to
813 use to handle perf sampling events. If the perf subsystem
814 is informed that its samples are exceeding this limit, it
815 will drop its sampling frequency to attempt to reduce its CPU
818 Some perf sampling happens in NMIs. If these samples
819 unexpectedly take too long to execute, the NMIs can become
820 stacked up next to each other so much that nothing else is
823 ===== ========================================================
824 0 Disable the mechanism. Do not monitor or correct perf's
825 sampling rate no matter how CPU time it takes.
827 1-100 Attempt to throttle perf's sample rate to this
828 percentage of CPU. Note: the kernel calculates an
829 "expected" length of each sample event. 100 here means
830 100% of that expected length. Even if this is set to
831 100, you may still see sample throttling if this
832 length is exceeded. Set to 0 if you truly do not care
833 how much CPU is consumed.
834 ===== ========================================================
840 Controls use of the performance events system by unprivileged
841 users (without CAP_PERFMON). The default value is 2.
843 For backward compatibility reasons access to system performance
844 monitoring and observability remains open for CAP_SYS_ADMIN
845 privileged processes but CAP_SYS_ADMIN usage for secure system
846 performance monitoring and observability operations is discouraged
847 with respect to CAP_PERFMON use cases.
849 === ==================================================================
850 -1 Allow use of (almost) all events by all users.
852 Ignore mlock limit after perf_event_mlock_kb without
855 >=0 Disallow ftrace function tracepoint by users without
858 Disallow raw tracepoint access by users without ``CAP_PERFMON``.
860 >=1 Disallow CPU event access by users without ``CAP_PERFMON``.
862 >=2 Disallow kernel profiling by users without ``CAP_PERFMON``.
863 === ==================================================================
869 Controls maximum number of stack frames to copy for (``attr.sample_type &
870 PERF_SAMPLE_CALLCHAIN``) configured events, for instance, when using
871 '``perf record -g``' or '``perf trace --call-graph fp``'.
873 This can only be done when no events are in use that have callchains
874 enabled, otherwise writing to this file will return ``-EBUSY``.
876 The default value is 127.
882 Control size of per-cpu ring buffer not counted against mlock limit.
884 The default value is 512 + 1 page
887 perf_event_max_contexts_per_stack
888 =================================
890 Controls maximum number of stack frame context entries for
891 (``attr.sample_type & PERF_SAMPLE_CALLCHAIN``) configured events, for
892 instance, when using '``perf record -g``' or '``perf trace --call-graph fp``'.
894 This can only be done when no events are in use that have callchains
895 enabled, otherwise writing to this file will return ``-EBUSY``.
897 The default value is 8.
903 PID allocation wrap value. When the kernel's next PID value
904 reaches this value, it wraps back to a minimum PID value.
905 PIDs of value ``pid_max`` or larger are not allocated.
911 The last pid allocated in the current (the one task using this sysctl
912 lives in) pid namespace. When selecting a pid for a next task on fork
913 kernel tries to allocate a number starting from this one.
916 powersave-nap (PPC only)
917 ========================
919 If set, Linux-PPC will use the 'nap' mode of powersaving,
920 otherwise the 'doze' mode will be used.
923 ==============================================================
928 The four values in printk denote: ``console_loglevel``,
929 ``default_message_loglevel``, ``minimum_console_loglevel`` and
930 ``default_console_loglevel`` respectively.
932 These values influence printk() behavior when printing or
933 logging error messages. See '``man 2 syslog``' for more info on
934 the different loglevels.
936 ======================== =====================================
937 console_loglevel messages with a higher priority than
938 this will be printed to the console
939 default_message_loglevel messages without an explicit priority
940 will be printed with this priority
941 minimum_console_loglevel minimum (highest) value to which
942 console_loglevel can be set
943 default_console_loglevel default value for console_loglevel
944 ======================== =====================================
950 Delay each printk message in ``printk_delay`` milliseconds
952 Value from 0 - 10000 is allowed.
958 Some warning messages are rate limited. ``printk_ratelimit`` specifies
959 the minimum length of time between these messages (in seconds).
960 The default value is 5 seconds.
962 A value of 0 will disable rate limiting.
965 printk_ratelimit_burst
966 ======================
968 While long term we enforce one message per `printk_ratelimit`_
969 seconds, we do allow a burst of messages to pass through.
970 ``printk_ratelimit_burst`` specifies the number of messages we can
971 send before ratelimiting kicks in.
973 The default value is 10 messages.
979 Control the logging to ``/dev/kmsg`` from userspace:
981 ========= =============================================
982 ratelimit default, ratelimited
983 on unlimited logging to /dev/kmsg from userspace
984 off logging to /dev/kmsg disabled
985 ========= =============================================
987 The kernel command line parameter ``printk.devkmsg=`` overrides this and is
988 a one-time setting until next reboot: once set, it cannot be changed by
989 this sysctl interface anymore.
991 ==============================================================
997 See Documentation/filesystems/devpts.rst.
1003 This is a directory, with the following entries:
1005 * ``boot_id``: a UUID generated the first time this is retrieved, and
1006 unvarying after that;
1008 * ``entropy_avail``: the pool's entropy count, in bits;
1010 * ``poolsize``: the entropy pool size, in bits;
1012 * ``urandom_min_reseed_secs``: obsolete (used to determine the minimum
1013 number of seconds between urandom pool reseeding).
1015 * ``uuid``: a UUID generated every time this is retrieved (this can
1016 thus be used to generate UUIDs at will);
1018 * ``write_wakeup_threshold``: when the entropy count drops below this
1019 (as a number of bits), processes waiting to write to ``/dev/random``
1022 If ``drivers/char/random.c`` is built with ``ADD_INTERRUPT_BENCH``
1023 defined, these additional entries are present:
1025 * ``add_interrupt_avg_cycles``: the average number of cycles between
1026 interrupts used to feed the pool;
1028 * ``add_interrupt_avg_deviation``: the standard deviation seen on the
1029 number of cycles between interrupts used to feed the pool.
1035 This option can be used to select the type of process address
1036 space randomization that is used in the system, for architectures
1037 that support this feature.
1039 == ===========================================================================
1040 0 Turn the process address space randomization off. This is the
1041 default for architectures that do not support this feature anyways,
1042 and kernels that are booted with the "norandmaps" parameter.
1044 1 Make the addresses of mmap base, stack and VDSO page randomized.
1045 This, among other things, implies that shared libraries will be
1046 loaded to random addresses. Also for PIE-linked binaries, the
1047 location of code start is randomized. This is the default if the
1048 ``CONFIG_COMPAT_BRK`` option is enabled.
1050 2 Additionally enable heap randomization. This is the default if
1051 ``CONFIG_COMPAT_BRK`` is disabled.
1053 There are a few legacy applications out there (such as some ancient
1054 versions of libc.so.5 from 1996) that assume that brk area starts
1055 just after the end of the code+bss. These applications break when
1056 start of the brk area is randomized. There are however no known
1057 non-legacy applications that would be broken this way, so for most
1058 systems it is safe to choose full randomization.
1060 Systems with ancient and/or broken binaries should be configured
1061 with ``CONFIG_COMPAT_BRK`` enabled, which excludes the heap from process
1062 address space randomization.
1063 == ===========================================================================
1069 See :doc:`/admin-guide/initrd`.
1072 reboot-cmd (SPARC only)
1073 =======================
1075 ??? This seems to be a way to give an argument to the Sparc
1076 ROM/Flash boot loader. Maybe to tell it what to do after
1083 Enables/disables Energy Aware Scheduling (EAS). EAS starts
1084 automatically on platforms where it can run (that is,
1085 platforms with asymmetric CPU topologies and having an Energy
1086 Model available). If your platform happens to meet the
1087 requirements for EAS but you do not want to use it, change
1094 Enables/disables scheduler statistics. Enabling this feature
1095 incurs a small amount of overhead in the scheduler but is
1096 useful for debugging and performance tuning.
1098 sched_util_clamp_min
1099 ====================
1101 Max allowed *minimum* utilization.
1103 Default value is 1024, which is the maximum possible value.
1105 It means that any requested uclamp.min value cannot be greater than
1106 sched_util_clamp_min, i.e., it is restricted to the range
1107 [0:sched_util_clamp_min].
1109 sched_util_clamp_max
1110 ====================
1112 Max allowed *maximum* utilization.
1114 Default value is 1024, which is the maximum possible value.
1116 It means that any requested uclamp.max value cannot be greater than
1117 sched_util_clamp_max, i.e., it is restricted to the range
1118 [0:sched_util_clamp_max].
1120 sched_util_clamp_min_rt_default
1121 ===============================
1123 By default Linux is tuned for performance. Which means that RT tasks always run
1124 at the highest frequency and most capable (highest capacity) CPU (in
1125 heterogeneous systems).
1127 Uclamp achieves this by setting the requested uclamp.min of all RT tasks to
1128 1024 by default, which effectively boosts the tasks to run at the highest
1129 frequency and biases them to run on the biggest CPU.
1131 This knob allows admins to change the default behavior when uclamp is being
1132 used. In battery powered devices particularly, running at the maximum
1133 capacity and frequency will increase energy consumption and shorten the battery
1136 This knob is only effective for RT tasks which the user hasn't modified their
1137 requested uclamp.min value via sched_setattr() syscall.
1139 This knob will not escape the range constraint imposed by sched_util_clamp_min
1144 sched_util_clamp_min_rt_default = 800
1145 sched_util_clamp_min = 600
1147 Then the boost will be clamped to 600 because 800 is outside of the permissible
1148 range of [0:600]. This could happen for instance if a powersave mode will
1149 restrict all boosts temporarily by modifying sched_util_clamp_min. As soon as
1150 this restriction is lifted, the requested sched_util_clamp_min_rt_default
1156 See :doc:`/userspace-api/seccomp_filter`.
1162 This file shows the size of the generic SCSI (sg) buffer.
1163 You can't tune it just yet, but you could change it on
1164 compile time by editing ``include/scsi/sg.h`` and changing
1165 the value of ``SG_BIG_BUFF``.
1167 There shouldn't be any reason to change this value. If
1168 you can come up with one, you probably know what you
1175 This parameter sets the total amount of shared memory pages that
1176 can be used system wide. Hence, ``shmall`` should always be at least
1177 ``ceil(shmmax/PAGE_SIZE)``.
1179 If you are not sure what the default ``PAGE_SIZE`` is on your Linux
1180 system, you can run the following command::
1188 This value can be used to query and set the run time limit
1189 on the maximum shared memory segment size that can be created.
1190 Shared memory segments up to 1Gb are now supported in the
1191 kernel. This value defaults to ``SHMMAX``.
1197 This value determines the maximum number of shared memory segments.
1198 4096 by default (``SHMMNI``).
1204 Linux lets you set resource limits, including how much memory one
1205 process can consume, via ``setrlimit(2)``. Unfortunately, shared memory
1206 segments are allowed to exist without association with any process, and
1207 thus might not be counted against any resource limits. If enabled,
1208 shared memory segments are automatically destroyed when their attach
1209 count becomes zero after a detach or a process termination. It will
1210 also destroy segments that were created, but never attached to, on exit
1211 from the process. The only use left for ``IPC_RMID`` is to immediately
1212 destroy an unattached segment. Of course, this breaks the way things are
1213 defined, so some applications might stop working. Note that this
1214 feature will do you no good unless you also configure your resource
1215 limits (in particular, ``RLIMIT_AS`` and ``RLIMIT_NPROC``). Most systems don't
1218 Note that if you change this from 0 to 1, already created segments
1219 without users and with a dead originative process will be destroyed.
1222 sysctl_writes_strict
1223 ====================
1225 Control how file position affects the behavior of updating sysctl values
1226 via the ``/proc/sys`` interface:
1228 == ======================================================================
1229 -1 Legacy per-write sysctl value handling, with no printk warnings.
1230 Each write syscall must fully contain the sysctl value to be
1231 written, and multiple writes on the same sysctl file descriptor
1232 will rewrite the sysctl value, regardless of file position.
1233 0 Same behavior as above, but warn about processes that perform writes
1234 to a sysctl file descriptor when the file position is not 0.
1235 1 (default) Respect file position when writing sysctl strings. Multiple
1236 writes will append to the sysctl value buffer. Anything past the max
1237 length of the sysctl value buffer will be ignored. Writes to numeric
1238 sysctl entries must always be at file position 0 and the value must
1239 be fully contained in the buffer sent in the write syscall.
1240 == ======================================================================
1243 softlockup_all_cpu_backtrace
1244 ============================
1246 This value controls the soft lockup detector thread's behavior
1247 when a soft lockup condition is detected as to whether or not
1248 to gather further debug information. If enabled, each cpu will
1249 be issued an NMI and instructed to capture stack trace.
1251 This feature is only applicable for architectures which support
1254 = ============================================
1255 0 Do nothing. This is the default behavior.
1256 1 On detection capture more debug information.
1257 = ============================================
1263 This parameter can be used to control whether the kernel panics
1264 when a soft lockup is detected.
1266 = ============================================
1267 0 Don't panic on soft lockup.
1268 1 Panic on soft lockup.
1269 = ============================================
1271 This can also be set using the softlockup_panic kernel parameter.
1277 This parameter can be used to control the soft lockup detector.
1279 = =================================
1280 0 Disable the soft lockup detector.
1281 1 Enable the soft lockup detector.
1282 = =================================
1284 The soft lockup detector monitors CPUs for threads that are hogging the CPUs
1285 without rescheduling voluntarily, and thus prevent the 'watchdog/N' threads
1286 from running. The mechanism depends on the CPUs ability to respond to timer
1287 interrupts which are needed for the 'watchdog/N' threads to be woken up by
1288 the watchdog timer function, otherwise the NMI watchdog — if enabled — can
1289 detect a hard lockup condition.
1295 This parameter can be used to control kernel stack erasing at the end
1296 of syscalls for kernels built with ``CONFIG_GCC_PLUGIN_STACKLEAK``.
1298 That erasing reduces the information which kernel stack leak bugs
1299 can reveal and blocks some uninitialized stack variable attacks.
1300 The tradeoff is the performance impact: on a single CPU system kernel
1301 compilation sees a 1% slowdown, other systems and workloads may vary.
1303 = ====================================================================
1304 0 Kernel stack erasing is disabled, STACKLEAK_METRICS are not updated.
1305 1 Kernel stack erasing is enabled (default), it is performed before
1306 returning to the userspace at the end of syscalls.
1307 = ====================================================================
1315 = ====================================
1316 0 Stop-A has no effect.
1317 1 Stop-A breaks to the PROM (default).
1318 = ====================================
1320 Stop-A is always enabled on a panic, so that the user can return to
1327 See :doc:`/admin-guide/sysrq`.
1333 Non-zero if the kernel has been tainted. Numeric values, which can be
1334 ORed together. The letters are seen in "Tainted" line of Oops reports.
1336 ====== ===== ==============================================================
1337 1 `(P)` proprietary module was loaded
1338 2 `(F)` module was force loaded
1339 4 `(S)` kernel running on an out of specification system
1340 8 `(R)` module was force unloaded
1341 16 `(M)` processor reported a Machine Check Exception (MCE)
1342 32 `(B)` bad page referenced or some unexpected page flags
1343 64 `(U)` taint requested by userspace application
1344 128 `(D)` kernel died recently, i.e. there was an OOPS or BUG
1345 256 `(A)` an ACPI table was overridden by user
1346 512 `(W)` kernel issued warning
1347 1024 `(C)` staging driver was loaded
1348 2048 `(I)` workaround for bug in platform firmware applied
1349 4096 `(O)` externally-built ("out-of-tree") module was loaded
1350 8192 `(E)` unsigned module was loaded
1351 16384 `(L)` soft lockup occurred
1352 32768 `(K)` kernel has been live patched
1353 65536 `(X)` Auxiliary taint, defined and used by for distros
1354 131072 `(T)` The kernel was built with the struct randomization plugin
1355 ====== ===== ==============================================================
1357 See :doc:`/admin-guide/tainted-kernels` for more information.
1360 writes to this sysctl interface will fail with ``EINVAL`` if the kernel is
1361 booted with the command line option ``panic_on_taint=<bitmask>,nousertaint``
1362 and any of the ORed together values being written to ``tainted`` match with
1363 the bitmask declared on panic_on_taint.
1364 See :doc:`/admin-guide/kernel-parameters` for more details on that particular
1365 kernel command line option and its optional ``nousertaint`` switch.
1370 This value controls the maximum number of threads that can be created
1373 During initialization the kernel sets this value such that even if the
1374 maximum number of threads is created, the thread structures occupy only
1375 a part (1/8th) of the available RAM pages.
1377 The minimum value that can be written to ``threads-max`` is 1.
1379 The maximum value that can be written to ``threads-max`` is given by the
1380 constant ``FUTEX_TID_MASK`` (0x3fffffff).
1382 If a value outside of this range is written to ``threads-max`` an
1383 ``EINVAL`` error occurs.
1389 When set, disables tracing (see :doc:`/trace/ftrace`) when a
1396 When tracepoints are sent to printk() (enabled by the ``tp_printk``
1397 boot parameter), this entry provides runtime control::
1399 echo 0 > /proc/sys/kernel/tracepoint_printk
1401 will stop tracepoints from being sent to printk(), and::
1403 echo 1 > /proc/sys/kernel/tracepoint_printk
1405 will send them to printk() again.
1407 This only works if the kernel was booted with ``tp_printk`` enabled.
1409 See :doc:`/admin-guide/kernel-parameters` and
1410 :doc:`/trace/boottime-trace`.
1413 .. _unaligned-dump-stack:
1415 unaligned-dump-stack (ia64)
1416 ===========================
1418 When logging unaligned accesses, controls whether the stack is
1421 = ===================================================
1422 0 Do not dump the stack. This is the default setting.
1424 = ===================================================
1426 See also `ignore-unaligned-usertrap`_.
1432 On architectures where unaligned accesses cause traps, and where this
1433 feature is supported (``CONFIG_SYSCTL_ARCH_UNALIGN_ALLOW``; currently,
1434 ``arc`` and ``parisc``), controls whether unaligned traps are caught
1435 and emulated (instead of failing).
1437 = ========================================================
1438 0 Do not emulate unaligned accesses.
1439 1 Emulate unaligned accesses. This is the default setting.
1440 = ========================================================
1442 See also `ignore-unaligned-usertrap`_.
1448 The value in this file affects behavior of handling NMI. When the
1449 value is non-zero, unknown NMI is trapped and then panic occurs. At
1450 that time, kernel debugging information is displayed on console.
1452 NMI switch that most IA32 servers have fires unknown NMI up, for
1453 example. If a system hangs up, try pressing the NMI switch.
1456 unprivileged_bpf_disabled
1457 =========================
1459 Writing 1 to this entry will disable unprivileged calls to ``bpf()``;
1460 once disabled, calling ``bpf()`` without ``CAP_SYS_ADMIN`` will return
1463 Once set, this can't be cleared.
1469 This parameter can be used to disable or enable the soft lockup detector
1470 *and* the NMI watchdog (i.e. the hard lockup detector) at the same time.
1472 = ==============================
1473 0 Disable both lockup detectors.
1474 1 Enable both lockup detectors.
1475 = ==============================
1477 The soft lockup detector and the NMI watchdog can also be disabled or
1478 enabled individually, using the ``soft_watchdog`` and ``nmi_watchdog``
1480 If the ``watchdog`` parameter is read, for example by executing::
1482 cat /proc/sys/kernel/watchdog
1484 the output of this command (0 or 1) shows the logical OR of
1485 ``soft_watchdog`` and ``nmi_watchdog``.
1491 This value can be used to control on which cpus the watchdog may run.
1492 The default cpumask is all possible cores, but if ``NO_HZ_FULL`` is
1493 enabled in the kernel config, and cores are specified with the
1494 ``nohz_full=`` boot argument, those cores are excluded by default.
1495 Offline cores can be included in this mask, and if the core is later
1496 brought online, the watchdog will be started based on the mask value.
1498 Typically this value would only be touched in the ``nohz_full`` case
1499 to re-enable cores that by default were not running the watchdog,
1500 if a kernel lockup was suspected on those cores.
1502 The argument value is the standard cpulist format for cpumasks,
1503 so for example to enable the watchdog on cores 0, 2, 3, and 4 you
1506 echo 0,2-4 > /proc/sys/kernel/watchdog_cpumask
1512 This value can be used to control the frequency of hrtimer and NMI
1513 events and the soft and hard lockup thresholds. The default threshold
1516 The softlockup threshold is (``2 * watchdog_thresh``). Setting this
1517 tunable to zero will disable lockup detection altogether.