1 ===================================
2 Documentation for /proc/sys/kernel/
3 ===================================
5 .. See scripts/check-sysctl-docs to keep this up to date
8 Copyright (c) 1998, 1999, Rik van Riel <riel@nl.linux.org>
10 Copyright (c) 2009, Shen Feng<shen@cn.fujitsu.com>
12 For general info and legal blurb, please look in
13 Documentation/admin-guide/sysctl/index.rst.
15 ------------------------------------------------------------------------------
17 This file contains documentation for the sysctl files in
18 ``/proc/sys/kernel/``.
20 The files in this directory can be used to tune and monitor
21 miscellaneous and general things in the operation of the Linux
22 kernel. Since some of the files *can* be used to screw up your
23 system, it is advisable to read both documentation and source
24 before actually making adjustments.
26 Currently, these files might (depending on your configuration)
27 show up in ``/proc/sys/kernel``:
37 highwater lowwater frequency
39 If BSD-style process accounting is enabled these values control
40 its behaviour. If free space on filesystem where the log lives
41 goes below ``lowwater``\ % accounting suspends. If free space gets
42 above ``highwater``\ % accounting resumes. ``frequency`` determines
43 how often do we check the amount of free space (value is in
50 That is, suspend accounting if free space drops below 2%; resume it
51 if it increases to at least 4%; consider information about amount of
52 free space valid for 30 seconds.
58 See Documentation/power/video.rst. This allows the video resume mode to be set,
59 in a similar fashion to the ``acpi_sleep`` kernel parameter, by
60 combining the following values:
71 The machine hardware name, the same output as ``uname -m``
72 (e.g. ``x86_64`` or ``aarch64``).
77 This variable has no effect and may be removed in future kernel
78 releases. Reading it always returns 0.
79 Up to Linux 3.17, it enabled/disabled automatic recomputing of
81 upon memory add/remove or upon IPC namespace creation/removal.
82 Echoing "1" into this file enabled msgmni automatic recomputing.
83 Echoing "0" turned it off. The default value was 1.
86 bootloader_type (x86 only)
87 ==========================
89 This gives the bootloader type number as indicated by the bootloader,
90 shifted left by 4, and OR'd with the low four bits of the bootloader
91 version. The reason for this encoding is that this used to match the
92 ``type_of_loader`` field in the kernel header; the encoding is kept for
93 backwards compatibility. That is, if the full bootloader type number
94 is 0x15 and the full version number is 0x234, this file will contain
95 the value 340 = 0x154.
97 See the ``type_of_loader`` and ``ext_loader_type`` fields in
98 Documentation/arch/x86/boot.rst for additional information.
101 bootloader_version (x86 only)
102 =============================
104 The complete bootloader version number. In the example above, this
105 file will contain the value 564 = 0x234.
107 See the ``type_of_loader`` and ``ext_loader_ver`` fields in
108 Documentation/arch/x86/boot.rst for additional information.
114 Controls whether the kernel should collect statistics on BPF programs
115 (total time spent running, number of times run...). Enabling
116 statistics causes a slight reduction in performance on each program
117 run. The statistics can be seen using ``bpftool``.
119 = ===================================
120 0 Don't collect statistics (default).
121 1 Collect statistics.
122 = ===================================
128 This is the pid which will be signalled on reboot (notably, by
129 Ctrl-Alt-Delete). Writing a value to this file which doesn't
130 correspond to a running process will result in ``-ESRCH``.
132 See also `ctrl-alt-del`_.
138 Highest valid capability of the running kernel. Exports
139 ``CAP_LAST_CAP`` from the kernel.
147 ``core_pattern`` is used to specify a core dumpfile pattern name.
149 * max length 127 characters; default value is "core"
150 * ``core_pattern`` is used as a pattern template for the output
151 filename; certain string patterns (beginning with '%') are
152 substituted with their actual values.
153 * backward compatibility with ``core_uses_pid``:
155 If ``core_pattern`` does not include "%p" (default does not)
156 and ``core_uses_pid`` is set, then .PID will be appended to
159 * corename format specifiers
161 ======== ==========================================
162 %<NUL> '%' is dropped
165 %P global pid (init PID namespace)
167 %I global tid (init PID namespace)
168 %u uid (in initial user namespace)
169 %g gid (in initial user namespace)
170 %d dump mode, matches ``PR_SET_DUMPABLE`` and
171 ``/proc/sys/fs/suid_dumpable``
175 %e executable filename (may be shortened, could be changed by prctl etc)
176 %f executable filename
178 %c maximum size of core file by resource limit RLIMIT_CORE
179 %C CPU the task ran on
180 %<OTHER> both are dropped
181 ======== ==========================================
183 * If the first character of the pattern is a '|', the kernel will treat
184 the rest of the pattern as a command to run. The core dump will be
185 written to the standard input of that program instead of to a file.
191 This sysctl is only applicable when `core_pattern`_ is configured to
192 pipe core files to a user space helper (when the first character of
193 ``core_pattern`` is a '|', see above).
194 When collecting cores via a pipe to an application, it is occasionally
195 useful for the collecting application to gather data about the
196 crashing process from its ``/proc/pid`` directory.
197 In order to do this safely, the kernel must wait for the collecting
198 process to exit, so as not to remove the crashing processes proc files
200 This in turn creates the possibility that a misbehaving userspace
201 collecting process can block the reaping of a crashed process simply
203 This sysctl defends against that.
204 It defines how many concurrent crashing processes may be piped to user
205 space applications in parallel.
206 If this value is exceeded, then those crashing processes above that
207 value are noted via the kernel log and their cores are skipped.
208 0 is a special value, indicating that unlimited processes may be
209 captured in parallel, but that no waiting will take place (i.e. the
210 collecting process is not guaranteed access to ``/proc/<crashing
212 This value defaults to 0.
218 The default coredump filename is "core". By setting
219 ``core_uses_pid`` to 1, the coredump filename becomes core.PID.
220 If `core_pattern`_ does not include "%p" (default does not)
221 and ``core_uses_pid`` is set, then .PID will be appended to
228 When the value in this file is 0, ctrl-alt-del is trapped and
229 sent to the ``init(1)`` program to handle a graceful restart.
230 When, however, the value is > 0, Linux's reaction to a Vulcan
231 Nerve Pinch (tm) will be an immediate reboot, without even
232 syncing its dirty buffers.
235 when a program (like dosemu) has the keyboard in 'raw'
236 mode, the ctrl-alt-del is intercepted by the program before it
237 ever reaches the kernel tty layer, and it's up to the program
238 to decide what to do with it.
244 This toggle indicates whether unprivileged users are prevented
245 from using ``dmesg(8)`` to view messages from the kernel's log
247 When ``dmesg_restrict`` is set to 0 there are no restrictions.
248 When ``dmesg_restrict`` is set to 1, users must have
249 ``CAP_SYSLOG`` to use ``dmesg(8)``.
251 The kernel config option ``CONFIG_SECURITY_DMESG_RESTRICT`` sets the
252 default value of ``dmesg_restrict``.
255 domainname & hostname
256 =====================
258 These files can be used to set the NIS/YP domainname and the
259 hostname of your box in exactly the same way as the commands
260 domainname and hostname, i.e.::
262 # echo "darkstar" > /proc/sys/kernel/hostname
263 # echo "mydomain" > /proc/sys/kernel/domainname
265 has the same effect as::
267 # hostname "darkstar"
268 # domainname "mydomain"
270 Note, however, that the classic darkstar.frop.org has the
271 hostname "darkstar" and DNS (Internet Domain Name Server)
272 domainname "frop.org", not to be confused with the NIS (Network
273 Information Service) or YP (Yellow Pages) domainname. These two
274 domain names are in general different. For a detailed discussion
275 see the ``hostname(1)`` man page.
281 See Documentation/driver-api/firmware/fallback-mechanisms.rst.
283 The entries in this directory allow the firmware loader helper
284 fallback to be controlled:
286 * ``force_sysfs_fallback``, when set to 1, forces the use of the
288 * ``ignore_sysfs_fallback``, when set to 1, ignores any fallback.
294 Determines whether ``ftrace_dump()`` should be called on an oops (or
295 kernel panic). This will output the contents of the ftrace buffers to
296 the console. This is very useful for capturing traces that lead to
297 crashes and outputting them to a serial console.
299 = ===================================================
300 0 Disabled (default).
301 1 Dump buffers of all CPUs.
302 2 Dump the buffer of the CPU that triggered the oops.
303 = ===================================================
306 ftrace_enabled, stack_tracer_enabled
307 ====================================
309 See Documentation/trace/ftrace.rst.
312 hardlockup_all_cpu_backtrace
313 ============================
315 This value controls the hard lockup detector behavior when a hard
316 lockup condition is detected as to whether or not to gather further
317 debug information. If enabled, arch-specific all-CPU stack dumping
320 = ============================================
321 0 Do nothing. This is the default behavior.
322 1 On detection capture more debug information.
323 = ============================================
329 This parameter can be used to control whether the kernel panics
330 when a hard lockup is detected.
332 = ===========================
333 0 Don't panic on hard lockup.
334 1 Panic on hard lockup.
335 = ===========================
337 See Documentation/admin-guide/lockup-watchdogs.rst for more information.
338 This can also be set using the nmi_watchdog kernel parameter.
344 Path for the hotplug policy agent.
345 Default value is ``CONFIG_UEVENT_HELPER_PATH``, which in turn defaults
348 This file only exists when ``CONFIG_UEVENT_HELPER`` is enabled. Most
349 modern systems rely exclusively on the netlink-based uevent source and
353 hung_task_all_cpu_backtrace
354 ===========================
356 If this option is set, the kernel will send an NMI to all CPUs to dump
357 their backtraces when a hung task is detected. This file shows up if
358 CONFIG_DETECT_HUNG_TASK and CONFIG_SMP are enabled.
360 0: Won't show all CPUs backtraces when a hung task is detected.
361 This is the default behavior.
363 1: Will non-maskably interrupt all CPUs and dump their backtraces when
364 a hung task is detected.
370 Controls the kernel's behavior when a hung task is detected.
371 This file shows up if ``CONFIG_DETECT_HUNG_TASK`` is enabled.
373 = =================================================
374 0 Continue operation. This is the default behavior.
376 = =================================================
379 hung_task_check_count
380 =====================
382 The upper bound on the number of tasks that are checked.
383 This file shows up if ``CONFIG_DETECT_HUNG_TASK`` is enabled.
386 hung_task_timeout_secs
387 ======================
389 When a task in D state did not get scheduled
390 for more than this value report a warning.
391 This file shows up if ``CONFIG_DETECT_HUNG_TASK`` is enabled.
393 0 means infinite timeout, no checking is done.
395 Possible values to set are in range {0:``LONG_MAX``/``HZ``}.
398 hung_task_check_interval_secs
399 =============================
401 Hung task check interval. If hung task checking is enabled
402 (see `hung_task_timeout_secs`_), the check is done every
403 ``hung_task_check_interval_secs`` seconds.
404 This file shows up if ``CONFIG_DETECT_HUNG_TASK`` is enabled.
406 0 (default) means use ``hung_task_timeout_secs`` as checking
409 Possible values to set are in range {0:``LONG_MAX``/``HZ``}.
415 The maximum number of warnings to report. During a check interval
416 if a hung task is detected, this value is decreased by 1.
417 When this value reaches 0, no more warnings will be reported.
418 This file shows up if ``CONFIG_DETECT_HUNG_TASK`` is enabled.
420 -1: report an infinite number of warnings.
423 hyperv_record_panic_msg
424 =======================
426 Controls whether the panic kmsg data should be reported to Hyper-V.
428 = =========================================================
429 0 Do not report panic kmsg data.
430 1 Report the panic kmsg data. This is the default behavior.
431 = =========================================================
434 ignore-unaligned-usertrap
435 =========================
437 On architectures where unaligned accesses cause traps, and where this
438 feature is supported (``CONFIG_SYSCTL_ARCH_UNALIGN_NO_WARN``;
439 currently, ``arc``, ``ia64`` and ``loongarch``), controls whether all
440 unaligned traps are logged.
442 = =============================================================
443 0 Log all unaligned accesses.
444 1 Only warn the first time a process traps. This is the default
446 = =============================================================
448 See also `unaligned-trap`_ and `unaligned-dump-stack`_. On ``ia64``,
449 this allows system administrators to override the
450 ``IA64_THREAD_UAC_NOPRINT`` ``prctl`` and avoid logs being flooded.
456 Prevents all processes from creating new io_uring instances. Enabling this
457 shrinks the kernel's attack surface.
459 = ======================================================================
460 0 All processes can create io_uring instances as normal. This is the
462 1 io_uring creation is disabled (io_uring_setup() will fail with
463 -EPERM) for unprivileged processes not in the io_uring_group group.
464 Existing io_uring instances can still be used. See the
465 documentation for io_uring_group for more information.
466 2 io_uring creation is disabled for all processes. io_uring_setup()
467 always fails with -EPERM. Existing io_uring instances can still be
469 = ======================================================================
475 When io_uring_disabled is set to 1, a process must either be
476 privileged (CAP_SYS_ADMIN) or be in the io_uring_group group in order
477 to create an io_uring instance. If io_uring_group is set to -1 (the
478 default), only processes with the CAP_SYS_ADMIN capability may create
485 A toggle indicating if the syscalls ``kexec_load`` and
486 ``kexec_file_load`` have been disabled.
487 This value defaults to 0 (false: ``kexec_*load`` enabled), but can be
488 set to 1 (true: ``kexec_*load`` disabled).
489 Once true, kexec can no longer be used, and the toggle cannot be set
491 This allows a kexec image to be loaded before disabling the syscall,
492 allowing a system to set up (and later use) an image without it being
494 Generally used together with the `modules_disabled`_ sysctl.
496 kexec_load_limit_panic
497 ======================
499 This parameter specifies a limit to the number of times the syscalls
500 ``kexec_load`` and ``kexec_file_load`` can be called with a crash
501 image. It can only be set with a more restrictive value than the
504 == ======================================================
505 -1 Unlimited calls to kexec. This is the default setting.
506 N Number of calls left.
507 == ======================================================
509 kexec_load_limit_reboot
510 =======================
512 Similar functionality as ``kexec_load_limit_panic``, but for a normal
518 This toggle indicates whether restrictions are placed on
519 exposing kernel addresses via ``/proc`` and other interfaces.
521 When ``kptr_restrict`` is set to 0 (the default) the address is hashed
523 (This is the equivalent to %p.)
525 When ``kptr_restrict`` is set to 1, kernel pointers printed using the
526 %pK format specifier will be replaced with 0s unless the user has
527 ``CAP_SYSLOG`` and effective user and group ids are equal to the real
529 This is because %pK checks are done at read() time rather than open()
530 time, so if permissions are elevated between the open() and the read()
531 (e.g via a setuid binary) then %pK will not leak kernel pointers to
533 Note, this is a temporary solution only.
534 The correct long-term solution is to do the permission checks at
536 Consider removing world read permissions from files that use %pK, and
537 using `dmesg_restrict`_ to protect against uses of %pK in ``dmesg(8)``
538 if leaking kernel pointer values to unprivileged users is a concern.
540 When ``kptr_restrict`` is set to 2, kernel pointers printed using
541 %pK will be replaced with 0s regardless of privileges.
547 The full path to the usermode helper for autoloading kernel modules,
548 by default ``CONFIG_MODPROBE_PATH``, which in turn defaults to
549 "/sbin/modprobe". This binary is executed when the kernel requests a
550 module. For example, if userspace passes an unknown filesystem type
551 to mount(), then the kernel will automatically request the
552 corresponding filesystem module by executing this usermode helper.
553 This usermode helper should insert the needed module into the kernel.
555 This sysctl only affects module autoloading. It has no effect on the
556 ability to explicitly insert modules.
558 This sysctl can be used to debug module loading requests::
560 echo '#! /bin/sh' > /tmp/modprobe
561 echo 'echo "$@" >> /tmp/modprobe.log' >> /tmp/modprobe
562 echo 'exec /sbin/modprobe "$@"' >> /tmp/modprobe
563 chmod a+x /tmp/modprobe
564 echo /tmp/modprobe > /proc/sys/kernel/modprobe
566 Alternatively, if this sysctl is set to the empty string, then module
567 autoloading is completely disabled. The kernel will not try to
568 execute a usermode helper at all, nor will it call the
569 kernel_module_request LSM hook.
571 If CONFIG_STATIC_USERMODEHELPER=y is set in the kernel configuration,
572 then the configured static usermode helper overrides this sysctl,
573 except that the empty string is still accepted to completely disable
574 module autoloading as described above.
579 A toggle value indicating if modules are allowed to be loaded
580 in an otherwise modular kernel. This toggle defaults to off
581 (0), but can be set true (1). Once true, modules can be
582 neither loaded nor unloaded, and the toggle cannot be set back
583 to false. Generally used with the `kexec_load_disabled`_ toggle.
588 msgmax, msgmnb, and msgmni
589 ==========================
591 ``msgmax`` is the maximum size of an IPC message, in bytes. 8192 by
592 default (``MSGMAX``).
594 ``msgmnb`` is the maximum size of an IPC queue, in bytes. 16384 by
595 default (``MSGMNB``).
597 ``msgmni`` is the maximum number of IPC queues. 32000 by default
601 msg_next_id, sem_next_id, and shm_next_id (System V IPC)
602 ========================================================
604 These three toggles allows to specify desired id for next allocated IPC
605 object: message, semaphore or shared memory respectively.
607 By default they are equal to -1, which means generic allocation logic.
608 Possible values to set are in range {0:``INT_MAX``}.
611 1) kernel doesn't guarantee, that new object will have desired id. So,
612 it's up to userspace, how to handle an object with "wrong" id.
613 2) Toggle with non-default value will be set back to -1 by kernel after
614 successful IPC object allocation. If an IPC object allocation syscall
615 fails, it is undefined if the value remains unmodified or is reset to -1.
621 Maximum number of supplementary groups, _i.e._ the maximum size which
622 ``setgroups`` will accept. Exports ``NGROUPS_MAX`` from the kernel.
629 This parameter can be used to control the NMI watchdog
630 (i.e. the hard lockup detector) on x86 systems.
632 = =================================
633 0 Disable the hard lockup detector.
634 1 Enable the hard lockup detector.
635 = =================================
637 The hard lockup detector monitors each CPU for its ability to respond to
638 timer interrupts. The mechanism utilizes CPU performance counter registers
639 that are programmed to generate Non-Maskable Interrupts (NMIs) periodically
640 while a CPU is busy. Hence, the alternative name 'NMI watchdog'.
642 The NMI watchdog is disabled by default if the kernel is running as a guest
643 in a KVM virtual machine. This default can be overridden by adding::
647 to the guest kernel command line (see
648 Documentation/admin-guide/kernel-parameters.rst).
651 nmi_wd_lpm_factor (PPC only)
652 ============================
654 Factor to apply to the NMI watchdog timeout (only when ``nmi_watchdog`` is
655 set to 1). This factor represents the percentage added to
656 ``watchdog_thresh`` when calculating the NMI watchdog timeout during an
657 LPM. The soft lockup timeout is not impacted.
659 A value of 0 means no change. The default value is 200 meaning the NMI
660 watchdog is set to 30s (based on ``watchdog_thresh`` equal to 10).
666 Enables/disables and configures automatic page fault based NUMA memory
667 balancing. Memory is moved automatically to nodes that access it often.
668 The value to set can be the result of ORing the following:
670 = =================================
671 0 NUMA_BALANCING_DISABLED
672 1 NUMA_BALANCING_NORMAL
673 2 NUMA_BALANCING_MEMORY_TIERING
674 = =================================
676 Or NUMA_BALANCING_NORMAL to optimize page placement among different
677 NUMA nodes to reduce remote accessing. On NUMA machines, there is a
678 performance penalty if remote memory is accessed by a CPU. When this
679 feature is enabled the kernel samples what task thread is accessing
680 memory by periodically unmapping pages and later trapping a page
681 fault. At the time of the page fault, it is determined if the data
682 being accessed should be migrated to a local memory node.
684 The unmapping of pages and trapping faults incur additional overhead that
685 ideally is offset by improved memory locality but there is no universal
686 guarantee. If the target workload is already bound to NUMA nodes then this
687 feature should be disabled.
689 Or NUMA_BALANCING_MEMORY_TIERING to optimize page placement among
690 different types of memory (represented as different NUMA nodes) to
691 place the hot pages in the fast memory. This is implemented based on
692 unmapping and page fault too.
694 numa_balancing_promote_rate_limit_MBps
695 ======================================
697 Too high promotion/demotion throughput between different memory types
698 may hurt application latency. This can be used to rate limit the
699 promotion throughput. The per-node max promotion throughput in MB/s
700 will be limited to be no more than the set value.
702 A rule of thumb is to set this to less than 1/10 of the PMEM node
705 oops_all_cpu_backtrace
706 ======================
708 If this option is set, the kernel will send an NMI to all CPUs to dump
709 their backtraces when an oops event occurs. It should be used as a last
710 resort in case a panic cannot be triggered (to protect VMs running, for
711 example) or kdump can't be collected. This file shows up if CONFIG_SMP
714 0: Won't show all CPUs backtraces when an oops is detected.
715 This is the default behavior.
717 1: Will non-maskably interrupt all CPUs and dump their backtraces when
718 an oops event is detected.
724 Number of kernel oopses after which the kernel should panic when
725 ``panic_on_oops`` is not set. Setting this to 0 disables checking
726 the count. Setting this to 1 has the same effect as setting
727 ``panic_on_oops=1``. The default value is 10000.
730 osrelease, ostype & version
731 ===========================
740 #5 Wed Feb 25 21:49:24 MET 1998
742 The files ``osrelease`` and ``ostype`` should be clear enough.
744 needs a little more clarification however. The '#5' means that
745 this is the fifth kernel built from this source base and the
746 date behind it indicates the time the kernel was built.
747 The only way to tune these values is to rebuild the kernel :-)
750 overflowgid & overflowuid
751 =========================
753 if your architecture did not always support 32-bit UIDs (i.e. arm,
754 i386, m68k, sh, and sparc32), a fixed UID and GID will be returned to
755 applications that use the old 16-bit UID/GID system calls, if the
756 actual UID or GID would exceed 65535.
758 These sysctls allow you to change the value of the fixed UID and GID.
759 The default is 65534.
765 The value in this file determines the behaviour of the kernel on a
768 * if zero, the kernel will loop forever;
769 * if negative, the kernel will reboot immediately;
770 * if positive, the kernel will reboot after the corresponding number
773 When you use the software watchdog, the recommended setting is 60.
779 Controls the kernel's behavior when a CPU receives an NMI caused by
782 = ==================================================================
783 0 Try to continue operation (default).
784 1 Panic immediately. The IO error triggered an NMI. This indicates a
785 serious system condition which could result in IO data corruption.
786 Rather than continuing, panicking might be a better choice. Some
787 servers issue this sort of NMI when the dump button is pushed,
788 and you can use this option to take a crash dump.
789 = ==================================================================
795 Controls the kernel's behaviour when an oops or BUG is encountered.
797 = ===================================================================
798 0 Try to continue operation.
799 1 Panic immediately. If the `panic` sysctl is also non-zero then the
800 machine will be rebooted.
801 = ===================================================================
804 panic_on_stackoverflow
805 ======================
807 Controls the kernel's behavior when detecting the overflows of
808 kernel, IRQ and exception stacks except a user stack.
809 This file shows up if ``CONFIG_DEBUG_STACKOVERFLOW`` is enabled.
811 = ==========================
812 0 Try to continue operation.
814 = ==========================
817 panic_on_unrecovered_nmi
818 ========================
820 The default Linux behaviour on an NMI of either memory or unknown is
821 to continue operation. For many environments such as scientific
822 computing it is preferable that the box is taken out and the error
823 dealt with than an uncorrected parity/ECC error get propagated.
825 A small number of systems do generate NMIs for bizarre random reasons
826 such as power management so the default is off. That sysctl works like
827 the existing panic controls already in that directory.
833 Calls panic() in the WARN() path when set to 1. This is useful to avoid
834 a kernel rebuild when attempting to kdump at the location of a WARN().
836 = ================================================
837 0 Only WARN(), default behaviour.
838 1 Call panic() after printing out WARN() location.
839 = ================================================
845 Bitmask for printing system info when panic happens. User can chose
846 combination of the following bits:
848 ===== ============================================
849 bit 0 print all tasks info
850 bit 1 print system memory info
851 bit 2 print timer info
852 bit 3 print locks info if ``CONFIG_LOCKDEP`` is on
853 bit 4 print ftrace buffer
854 bit 5 print all printk messages in buffer
855 bit 6 print all CPUs backtrace (if available in the arch)
856 ===== ============================================
858 So for example to print tasks and memory info on panic, user can::
860 echo 3 > /proc/sys/kernel/panic_print
866 When set to 1, calls panic() after RCU stall detection messages. This
867 is useful to define the root cause of RCU stalls using a vmcore.
869 = ============================================================
870 0 Do not panic() when RCU stall takes place, default behavior.
871 1 panic() after printing RCU stall messages.
872 = ============================================================
874 max_rcu_stall_to_panic
875 ======================
877 When ``panic_on_rcu_stall`` is set to 1, this value determines the
878 number of times that RCU can stall before panic() is called.
880 When ``panic_on_rcu_stall`` is set to 0, this value is has no effect.
882 perf_cpu_time_max_percent
883 =========================
885 Hints to the kernel how much CPU time it should be allowed to
886 use to handle perf sampling events. If the perf subsystem
887 is informed that its samples are exceeding this limit, it
888 will drop its sampling frequency to attempt to reduce its CPU
891 Some perf sampling happens in NMIs. If these samples
892 unexpectedly take too long to execute, the NMIs can become
893 stacked up next to each other so much that nothing else is
896 ===== ========================================================
897 0 Disable the mechanism. Do not monitor or correct perf's
898 sampling rate no matter how CPU time it takes.
900 1-100 Attempt to throttle perf's sample rate to this
901 percentage of CPU. Note: the kernel calculates an
902 "expected" length of each sample event. 100 here means
903 100% of that expected length. Even if this is set to
904 100, you may still see sample throttling if this
905 length is exceeded. Set to 0 if you truly do not care
906 how much CPU is consumed.
907 ===== ========================================================
913 Controls use of the performance events system by unprivileged
914 users (without CAP_PERFMON). The default value is 2.
916 For backward compatibility reasons access to system performance
917 monitoring and observability remains open for CAP_SYS_ADMIN
918 privileged processes but CAP_SYS_ADMIN usage for secure system
919 performance monitoring and observability operations is discouraged
920 with respect to CAP_PERFMON use cases.
922 === ==================================================================
923 -1 Allow use of (almost) all events by all users.
925 Ignore mlock limit after perf_event_mlock_kb without
928 >=0 Disallow ftrace function tracepoint by users without
931 Disallow raw tracepoint access by users without ``CAP_PERFMON``.
933 >=1 Disallow CPU event access by users without ``CAP_PERFMON``.
935 >=2 Disallow kernel profiling by users without ``CAP_PERFMON``.
936 === ==================================================================
942 Controls maximum number of stack frames to copy for (``attr.sample_type &
943 PERF_SAMPLE_CALLCHAIN``) configured events, for instance, when using
944 '``perf record -g``' or '``perf trace --call-graph fp``'.
946 This can only be done when no events are in use that have callchains
947 enabled, otherwise writing to this file will return ``-EBUSY``.
949 The default value is 127.
955 Control size of per-cpu ring buffer not counted against mlock limit.
957 The default value is 512 + 1 page
960 perf_event_max_contexts_per_stack
961 =================================
963 Controls maximum number of stack frame context entries for
964 (``attr.sample_type & PERF_SAMPLE_CALLCHAIN``) configured events, for
965 instance, when using '``perf record -g``' or '``perf trace --call-graph fp``'.
967 This can only be done when no events are in use that have callchains
968 enabled, otherwise writing to this file will return ``-EBUSY``.
970 The default value is 8.
973 perf_user_access (arm64 and riscv only)
974 =======================================
976 Controls user space access for reading perf event counters.
981 The default value is 0 (access disabled).
983 When set to 1, user space can read performance monitor counter registers
986 See Documentation/arch/arm64/perf.rst for more information.
991 When set to 0, user space access is disabled.
993 The default value is 1, user space can read performance monitor counter
994 registers through perf, any direct access without perf intervention will trigger
995 an illegal instruction.
997 When set to 2, which enables legacy mode (user space has direct access to cycle
998 and insret CSRs only). Note that this legacy value is deprecated and will be
999 removed once all user space applications are fixed.
1001 Note that the time CSR is always directly accessible to all modes.
1006 PID allocation wrap value. When the kernel's next PID value
1007 reaches this value, it wraps back to a minimum PID value.
1008 PIDs of value ``pid_max`` or larger are not allocated.
1014 The last pid allocated in the current (the one task using this sysctl
1015 lives in) pid namespace. When selecting a pid for a next task on fork
1016 kernel tries to allocate a number starting from this one.
1019 powersave-nap (PPC only)
1020 ========================
1022 If set, Linux-PPC will use the 'nap' mode of powersaving,
1023 otherwise the 'doze' mode will be used.
1026 ==============================================================
1031 The four values in printk denote: ``console_loglevel``,
1032 ``default_message_loglevel``, ``minimum_console_loglevel`` and
1033 ``default_console_loglevel`` respectively.
1035 These values influence printk() behavior when printing or
1036 logging error messages. See '``man 2 syslog``' for more info on
1037 the different loglevels.
1039 ======================== =====================================
1040 console_loglevel messages with a higher priority than
1041 this will be printed to the console
1042 default_message_loglevel messages without an explicit priority
1043 will be printed with this priority
1044 minimum_console_loglevel minimum (highest) value to which
1045 console_loglevel can be set
1046 default_console_loglevel default value for console_loglevel
1047 ======================== =====================================
1053 Delay each printk message in ``printk_delay`` milliseconds
1055 Value from 0 - 10000 is allowed.
1061 Some warning messages are rate limited. ``printk_ratelimit`` specifies
1062 the minimum length of time between these messages (in seconds).
1063 The default value is 5 seconds.
1065 A value of 0 will disable rate limiting.
1068 printk_ratelimit_burst
1069 ======================
1071 While long term we enforce one message per `printk_ratelimit`_
1072 seconds, we do allow a burst of messages to pass through.
1073 ``printk_ratelimit_burst`` specifies the number of messages we can
1074 send before ratelimiting kicks in.
1076 The default value is 10 messages.
1082 Control the logging to ``/dev/kmsg`` from userspace:
1084 ========= =============================================
1085 ratelimit default, ratelimited
1086 on unlimited logging to /dev/kmsg from userspace
1087 off logging to /dev/kmsg disabled
1088 ========= =============================================
1090 The kernel command line parameter ``printk.devkmsg=`` overrides this and is
1091 a one-time setting until next reboot: once set, it cannot be changed by
1092 this sysctl interface anymore.
1094 ==============================================================
1100 See Documentation/filesystems/devpts.rst.
1106 This is a directory, with the following entries:
1108 * ``boot_id``: a UUID generated the first time this is retrieved, and
1109 unvarying after that;
1111 * ``uuid``: a UUID generated every time this is retrieved (this can
1112 thus be used to generate UUIDs at will);
1114 * ``entropy_avail``: the pool's entropy count, in bits;
1116 * ``poolsize``: the entropy pool size, in bits;
1118 * ``urandom_min_reseed_secs``: obsolete (used to determine the minimum
1119 number of seconds between urandom pool reseeding). This file is
1120 writable for compatibility purposes, but writing to it has no effect
1121 on any RNG behavior;
1123 * ``write_wakeup_threshold``: when the entropy count drops below this
1124 (as a number of bits), processes waiting to write to ``/dev/random``
1125 are woken up. This file is writable for compatibility purposes, but
1126 writing to it has no effect on any RNG behavior.
1132 This option can be used to select the type of process address
1133 space randomization that is used in the system, for architectures
1134 that support this feature.
1136 == ===========================================================================
1137 0 Turn the process address space randomization off. This is the
1138 default for architectures that do not support this feature anyways,
1139 and kernels that are booted with the "norandmaps" parameter.
1141 1 Make the addresses of mmap base, stack and VDSO page randomized.
1142 This, among other things, implies that shared libraries will be
1143 loaded to random addresses. Also for PIE-linked binaries, the
1144 location of code start is randomized. This is the default if the
1145 ``CONFIG_COMPAT_BRK`` option is enabled.
1147 2 Additionally enable heap randomization. This is the default if
1148 ``CONFIG_COMPAT_BRK`` is disabled.
1150 There are a few legacy applications out there (such as some ancient
1151 versions of libc.so.5 from 1996) that assume that brk area starts
1152 just after the end of the code+bss. These applications break when
1153 start of the brk area is randomized. There are however no known
1154 non-legacy applications that would be broken this way, so for most
1155 systems it is safe to choose full randomization.
1157 Systems with ancient and/or broken binaries should be configured
1158 with ``CONFIG_COMPAT_BRK`` enabled, which excludes the heap from process
1159 address space randomization.
1160 == ===========================================================================
1166 See Documentation/admin-guide/initrd.rst.
1169 reboot-cmd (SPARC only)
1170 =======================
1172 ??? This seems to be a way to give an argument to the Sparc
1173 ROM/Flash boot loader. Maybe to tell it what to do after
1180 Enables/disables Energy Aware Scheduling (EAS). EAS starts
1181 automatically on platforms where it can run (that is,
1182 platforms with asymmetric CPU topologies and having an Energy
1183 Model available). If your platform happens to meet the
1184 requirements for EAS but you do not want to use it, change
1190 Enables/disables task delay accounting (see
1191 Documentation/accounting/delay-accounting.rst. Enabling this feature incurs
1192 a small amount of overhead in the scheduler but is useful for debugging
1193 and performance tuning. It is required by some tools such as iotop.
1198 Enables/disables scheduler statistics. Enabling this feature
1199 incurs a small amount of overhead in the scheduler but is
1200 useful for debugging and performance tuning.
1202 sched_util_clamp_min
1203 ====================
1205 Max allowed *minimum* utilization.
1207 Default value is 1024, which is the maximum possible value.
1209 It means that any requested uclamp.min value cannot be greater than
1210 sched_util_clamp_min, i.e., it is restricted to the range
1211 [0:sched_util_clamp_min].
1213 sched_util_clamp_max
1214 ====================
1216 Max allowed *maximum* utilization.
1218 Default value is 1024, which is the maximum possible value.
1220 It means that any requested uclamp.max value cannot be greater than
1221 sched_util_clamp_max, i.e., it is restricted to the range
1222 [0:sched_util_clamp_max].
1224 sched_util_clamp_min_rt_default
1225 ===============================
1227 By default Linux is tuned for performance. Which means that RT tasks always run
1228 at the highest frequency and most capable (highest capacity) CPU (in
1229 heterogeneous systems).
1231 Uclamp achieves this by setting the requested uclamp.min of all RT tasks to
1232 1024 by default, which effectively boosts the tasks to run at the highest
1233 frequency and biases them to run on the biggest CPU.
1235 This knob allows admins to change the default behavior when uclamp is being
1236 used. In battery powered devices particularly, running at the maximum
1237 capacity and frequency will increase energy consumption and shorten the battery
1240 This knob is only effective for RT tasks which the user hasn't modified their
1241 requested uclamp.min value via sched_setattr() syscall.
1243 This knob will not escape the range constraint imposed by sched_util_clamp_min
1248 sched_util_clamp_min_rt_default = 800
1249 sched_util_clamp_min = 600
1251 Then the boost will be clamped to 600 because 800 is outside of the permissible
1252 range of [0:600]. This could happen for instance if a powersave mode will
1253 restrict all boosts temporarily by modifying sched_util_clamp_min. As soon as
1254 this restriction is lifted, the requested sched_util_clamp_min_rt_default
1260 See Documentation/userspace-api/seccomp_filter.rst.
1266 This file shows the size of the generic SCSI (sg) buffer.
1267 You can't tune it just yet, but you could change it on
1268 compile time by editing ``include/scsi/sg.h`` and changing
1269 the value of ``SG_BIG_BUFF``.
1271 There shouldn't be any reason to change this value. If
1272 you can come up with one, you probably know what you
1279 This parameter sets the total amount of shared memory pages that
1280 can be used system wide. Hence, ``shmall`` should always be at least
1281 ``ceil(shmmax/PAGE_SIZE)``.
1283 If you are not sure what the default ``PAGE_SIZE`` is on your Linux
1284 system, you can run the following command::
1292 This value can be used to query and set the run time limit
1293 on the maximum shared memory segment size that can be created.
1294 Shared memory segments up to 1Gb are now supported in the
1295 kernel. This value defaults to ``SHMMAX``.
1301 This value determines the maximum number of shared memory segments.
1302 4096 by default (``SHMMNI``).
1308 Linux lets you set resource limits, including how much memory one
1309 process can consume, via ``setrlimit(2)``. Unfortunately, shared memory
1310 segments are allowed to exist without association with any process, and
1311 thus might not be counted against any resource limits. If enabled,
1312 shared memory segments are automatically destroyed when their attach
1313 count becomes zero after a detach or a process termination. It will
1314 also destroy segments that were created, but never attached to, on exit
1315 from the process. The only use left for ``IPC_RMID`` is to immediately
1316 destroy an unattached segment. Of course, this breaks the way things are
1317 defined, so some applications might stop working. Note that this
1318 feature will do you no good unless you also configure your resource
1319 limits (in particular, ``RLIMIT_AS`` and ``RLIMIT_NPROC``). Most systems don't
1322 Note that if you change this from 0 to 1, already created segments
1323 without users and with a dead originative process will be destroyed.
1326 sysctl_writes_strict
1327 ====================
1329 Control how file position affects the behavior of updating sysctl values
1330 via the ``/proc/sys`` interface:
1332 == ======================================================================
1333 -1 Legacy per-write sysctl value handling, with no printk warnings.
1334 Each write syscall must fully contain the sysctl value to be
1335 written, and multiple writes on the same sysctl file descriptor
1336 will rewrite the sysctl value, regardless of file position.
1337 0 Same behavior as above, but warn about processes that perform writes
1338 to a sysctl file descriptor when the file position is not 0.
1339 1 (default) Respect file position when writing sysctl strings. Multiple
1340 writes will append to the sysctl value buffer. Anything past the max
1341 length of the sysctl value buffer will be ignored. Writes to numeric
1342 sysctl entries must always be at file position 0 and the value must
1343 be fully contained in the buffer sent in the write syscall.
1344 == ======================================================================
1347 softlockup_all_cpu_backtrace
1348 ============================
1350 This value controls the soft lockup detector thread's behavior
1351 when a soft lockup condition is detected as to whether or not
1352 to gather further debug information. If enabled, each cpu will
1353 be issued an NMI and instructed to capture stack trace.
1355 This feature is only applicable for architectures which support
1358 = ============================================
1359 0 Do nothing. This is the default behavior.
1360 1 On detection capture more debug information.
1361 = ============================================
1367 This parameter can be used to control whether the kernel panics
1368 when a soft lockup is detected.
1370 = ============================================
1371 0 Don't panic on soft lockup.
1372 1 Panic on soft lockup.
1373 = ============================================
1375 This can also be set using the softlockup_panic kernel parameter.
1381 This parameter can be used to control the soft lockup detector.
1383 = =================================
1384 0 Disable the soft lockup detector.
1385 1 Enable the soft lockup detector.
1386 = =================================
1388 The soft lockup detector monitors CPUs for threads that are hogging the CPUs
1389 without rescheduling voluntarily, and thus prevent the 'migration/N' threads
1390 from running, causing the watchdog work fail to execute. The mechanism depends
1391 on the CPUs ability to respond to timer interrupts which are needed for the
1392 watchdog work to be queued by the watchdog timer function, otherwise the NMI
1393 watchdog — if enabled — can detect a hard lockup condition.
1396 split_lock_mitigate (x86 only)
1397 ==============================
1399 On x86, each "split lock" imposes a system-wide performance penalty. On larger
1400 systems, large numbers of split locks from unprivileged users can result in
1401 denials of service to well-behaved and potentially more important users.
1403 The kernel mitigates these bad users by detecting split locks and imposing
1404 penalties: forcing them to wait and only allowing one core to execute split
1407 These mitigations can make those bad applications unbearably slow. Setting
1408 split_lock_mitigate=0 may restore some application performance, but will also
1409 increase system exposure to denial of service attacks from split lock users.
1411 = ===================================================================
1412 0 Disable the mitigation mode - just warns the split lock on kernel log
1413 and exposes the system to denials of service from the split lockers.
1414 1 Enable the mitigation mode (this is the default) - penalizes the split
1415 lockers with intentional performance degradation.
1416 = ===================================================================
1422 This parameter can be used to control kernel stack erasing at the end
1423 of syscalls for kernels built with ``CONFIG_GCC_PLUGIN_STACKLEAK``.
1425 That erasing reduces the information which kernel stack leak bugs
1426 can reveal and blocks some uninitialized stack variable attacks.
1427 The tradeoff is the performance impact: on a single CPU system kernel
1428 compilation sees a 1% slowdown, other systems and workloads may vary.
1430 = ====================================================================
1431 0 Kernel stack erasing is disabled, STACKLEAK_METRICS are not updated.
1432 1 Kernel stack erasing is enabled (default), it is performed before
1433 returning to the userspace at the end of syscalls.
1434 = ====================================================================
1442 = ====================================
1443 0 Stop-A has no effect.
1444 1 Stop-A breaks to the PROM (default).
1445 = ====================================
1447 Stop-A is always enabled on a panic, so that the user can return to
1454 See Documentation/admin-guide/sysrq.rst.
1460 Non-zero if the kernel has been tainted. Numeric values, which can be
1461 ORed together. The letters are seen in "Tainted" line of Oops reports.
1463 ====== ===== ==============================================================
1464 1 `(P)` proprietary module was loaded
1465 2 `(F)` module was force loaded
1466 4 `(S)` kernel running on an out of specification system
1467 8 `(R)` module was force unloaded
1468 16 `(M)` processor reported a Machine Check Exception (MCE)
1469 32 `(B)` bad page referenced or some unexpected page flags
1470 64 `(U)` taint requested by userspace application
1471 128 `(D)` kernel died recently, i.e. there was an OOPS or BUG
1472 256 `(A)` an ACPI table was overridden by user
1473 512 `(W)` kernel issued warning
1474 1024 `(C)` staging driver was loaded
1475 2048 `(I)` workaround for bug in platform firmware applied
1476 4096 `(O)` externally-built ("out-of-tree") module was loaded
1477 8192 `(E)` unsigned module was loaded
1478 16384 `(L)` soft lockup occurred
1479 32768 `(K)` kernel has been live patched
1480 65536 `(X)` Auxiliary taint, defined and used by for distros
1481 131072 `(T)` The kernel was built with the struct randomization plugin
1482 ====== ===== ==============================================================
1484 See Documentation/admin-guide/tainted-kernels.rst for more information.
1487 writes to this sysctl interface will fail with ``EINVAL`` if the kernel is
1488 booted with the command line option ``panic_on_taint=<bitmask>,nousertaint``
1489 and any of the ORed together values being written to ``tainted`` match with
1490 the bitmask declared on panic_on_taint.
1491 See Documentation/admin-guide/kernel-parameters.rst for more details on
1492 that particular kernel command line option and its optional
1493 ``nousertaint`` switch.
1498 This value controls the maximum number of threads that can be created
1501 During initialization the kernel sets this value such that even if the
1502 maximum number of threads is created, the thread structures occupy only
1503 a part (1/8th) of the available RAM pages.
1505 The minimum value that can be written to ``threads-max`` is 1.
1507 The maximum value that can be written to ``threads-max`` is given by the
1508 constant ``FUTEX_TID_MASK`` (0x3fffffff).
1510 If a value outside of this range is written to ``threads-max`` an
1511 ``EINVAL`` error occurs.
1517 When set, disables tracing (see Documentation/trace/ftrace.rst) when a
1524 When tracepoints are sent to printk() (enabled by the ``tp_printk``
1525 boot parameter), this entry provides runtime control::
1527 echo 0 > /proc/sys/kernel/tracepoint_printk
1529 will stop tracepoints from being sent to printk(), and::
1531 echo 1 > /proc/sys/kernel/tracepoint_printk
1533 will send them to printk() again.
1535 This only works if the kernel was booted with ``tp_printk`` enabled.
1537 See Documentation/admin-guide/kernel-parameters.rst and
1538 Documentation/trace/boottime-trace.rst.
1541 .. _unaligned-dump-stack:
1543 unaligned-dump-stack (ia64)
1544 ===========================
1546 When logging unaligned accesses, controls whether the stack is
1549 = ===================================================
1550 0 Do not dump the stack. This is the default setting.
1552 = ===================================================
1554 See also `ignore-unaligned-usertrap`_.
1560 On architectures where unaligned accesses cause traps, and where this
1561 feature is supported (``CONFIG_SYSCTL_ARCH_UNALIGN_ALLOW``; currently,
1562 ``arc``, ``parisc`` and ``loongarch``), controls whether unaligned traps
1563 are caught and emulated (instead of failing).
1565 = ========================================================
1566 0 Do not emulate unaligned accesses.
1567 1 Emulate unaligned accesses. This is the default setting.
1568 = ========================================================
1570 See also `ignore-unaligned-usertrap`_.
1576 The value in this file affects behavior of handling NMI. When the
1577 value is non-zero, unknown NMI is trapped and then panic occurs. At
1578 that time, kernel debugging information is displayed on console.
1580 NMI switch that most IA32 servers have fires unknown NMI up, for
1581 example. If a system hangs up, try pressing the NMI switch.
1584 unprivileged_bpf_disabled
1585 =========================
1587 Writing 1 to this entry will disable unprivileged calls to ``bpf()``;
1588 once disabled, calling ``bpf()`` without ``CAP_SYS_ADMIN`` or ``CAP_BPF``
1589 will return ``-EPERM``. Once set to 1, this can't be cleared from the
1590 running kernel anymore.
1592 Writing 2 to this entry will also disable unprivileged calls to ``bpf()``,
1593 however, an admin can still change this setting later on, if needed, by
1594 writing 0 or 1 to this entry.
1596 If ``BPF_UNPRIV_DEFAULT_OFF`` is enabled in the kernel config, then this
1597 entry will default to 2 instead of 0.
1599 = =============================================================
1600 0 Unprivileged calls to ``bpf()`` are enabled
1601 1 Unprivileged calls to ``bpf()`` are disabled without recovery
1602 2 Unprivileged calls to ``bpf()`` are disabled
1603 = =============================================================
1609 Number of kernel warnings after which the kernel should panic when
1610 ``panic_on_warn`` is not set. Setting this to 0 disables checking
1611 the warning count. Setting this to 1 has the same effect as setting
1612 ``panic_on_warn=1``. The default value is 0.
1618 This parameter can be used to disable or enable the soft lockup detector
1619 *and* the NMI watchdog (i.e. the hard lockup detector) at the same time.
1621 = ==============================
1622 0 Disable both lockup detectors.
1623 1 Enable both lockup detectors.
1624 = ==============================
1626 The soft lockup detector and the NMI watchdog can also be disabled or
1627 enabled individually, using the ``soft_watchdog`` and ``nmi_watchdog``
1629 If the ``watchdog`` parameter is read, for example by executing::
1631 cat /proc/sys/kernel/watchdog
1633 the output of this command (0 or 1) shows the logical OR of
1634 ``soft_watchdog`` and ``nmi_watchdog``.
1640 This value can be used to control on which cpus the watchdog may run.
1641 The default cpumask is all possible cores, but if ``NO_HZ_FULL`` is
1642 enabled in the kernel config, and cores are specified with the
1643 ``nohz_full=`` boot argument, those cores are excluded by default.
1644 Offline cores can be included in this mask, and if the core is later
1645 brought online, the watchdog will be started based on the mask value.
1647 Typically this value would only be touched in the ``nohz_full`` case
1648 to re-enable cores that by default were not running the watchdog,
1649 if a kernel lockup was suspected on those cores.
1651 The argument value is the standard cpulist format for cpumasks,
1652 so for example to enable the watchdog on cores 0, 2, 3, and 4 you
1655 echo 0,2-4 > /proc/sys/kernel/watchdog_cpumask
1661 This value can be used to control the frequency of hrtimer and NMI
1662 events and the soft and hard lockup thresholds. The default threshold
1665 The softlockup threshold is (``2 * watchdog_thresh``). Setting this
1666 tunable to zero will disable lockup detection altogether.