1676 lines · plain
1===================================2Documentation for /proc/sys/kernel/3===================================4 5.. See scripts/check-sysctl-docs to keep this up to date6 7 8Copyright (c) 1998, 1999, Rik van Riel <riel@nl.linux.org>9 10Copyright (c) 2009, Shen Feng<shen@cn.fujitsu.com>11 12For general info and legal blurb, please look in13Documentation/admin-guide/sysctl/index.rst.14 15------------------------------------------------------------------------------16 17This file contains documentation for the sysctl files in18``/proc/sys/kernel/``.19 20The files in this directory can be used to tune and monitor21miscellaneous and general things in the operation of the Linux22kernel. Since some of the files *can* be used to screw up your23system, it is advisable to read both documentation and source24before actually making adjustments.25 26Currently, these files might (depending on your configuration)27show up in ``/proc/sys/kernel``:28 29.. contents:: :local:30 31 32acct33====34 35::36 37 highwater lowwater frequency38 39If BSD-style process accounting is enabled these values control40its behaviour. If free space on filesystem where the log lives41goes below ``lowwater``\ % accounting suspends. If free space gets42above ``highwater``\ % accounting resumes. ``frequency`` determines43how often do we check the amount of free space (value is in44seconds). Default:45 46::47 48 4 2 3049 50That is, suspend accounting if free space drops below 2%; resume it51if it increases to at least 4%; consider information about amount of52free space valid for 30 seconds.53 54 55acpi_video_flags56================57 58See Documentation/power/video.rst. This allows the video resume mode to be set,59in a similar fashion to the ``acpi_sleep`` kernel parameter, by60combining the following values:61 62= =======631 s3_bios642 s3_mode654 s3_beep66= =======67 68arch69====70 71The machine hardware name, the same output as ``uname -m``72(e.g. ``x86_64`` or ``aarch64``).73 74auto_msgmni75===========76 77This variable has no effect and may be removed in future kernel78releases. Reading it always returns 0.79Up to Linux 3.17, it enabled/disabled automatic recomputing of80`msgmni`_81upon memory add/remove or upon IPC namespace creation/removal.82Echoing "1" into this file enabled msgmni automatic recomputing.83Echoing "0" turned it off. The default value was 1.84 85 86bootloader_type (x86 only)87==========================88 89This gives the bootloader type number as indicated by the bootloader,90shifted left by 4, and OR'd with the low four bits of the bootloader91version. The reason for this encoding is that this used to match the92``type_of_loader`` field in the kernel header; the encoding is kept for93backwards compatibility. That is, if the full bootloader type number94is 0x15 and the full version number is 0x234, this file will contain95the value 340 = 0x154.96 97See the ``type_of_loader`` and ``ext_loader_type`` fields in98Documentation/arch/x86/boot.rst for additional information.99 100 101bootloader_version (x86 only)102=============================103 104The complete bootloader version number. In the example above, this105file will contain the value 564 = 0x234.106 107See the ``type_of_loader`` and ``ext_loader_ver`` fields in108Documentation/arch/x86/boot.rst for additional information.109 110 111bpf_stats_enabled112=================113 114Controls whether the kernel should collect statistics on BPF programs115(total time spent running, number of times run...). Enabling116statistics causes a slight reduction in performance on each program117run. The statistics can be seen using ``bpftool``.118 119= ===================================1200 Don't collect statistics (default).1211 Collect statistics.122= ===================================123 124 125cad_pid126=======127 128This is the pid which will be signalled on reboot (notably, by129Ctrl-Alt-Delete). Writing a value to this file which doesn't130correspond to a running process will result in ``-ESRCH``.131 132See also `ctrl-alt-del`_.133 134 135cap_last_cap136============137 138Highest valid capability of the running kernel. Exports139``CAP_LAST_CAP`` from the kernel.140 141 142.. _core_pattern:143 144core_pattern145============146 147``core_pattern`` is used to specify a core dumpfile pattern name.148 149* max length 127 characters; default value is "core"150* ``core_pattern`` is used as a pattern template for the output151 filename; certain string patterns (beginning with '%') are152 substituted with their actual values.153* backward compatibility with ``core_uses_pid``:154 155 If ``core_pattern`` does not include "%p" (default does not)156 and ``core_uses_pid`` is set, then .PID will be appended to157 the filename.158 159* corename format specifiers160 161 ======== ==========================================162 %<NUL> '%' is dropped163 %% output one '%'164 %p pid165 %P global pid (init PID namespace)166 %i tid167 %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`` and171 ``/proc/sys/fs/suid_dumpable``172 %s signal number173 %t UNIX time of dump174 %h hostname175 %e executable filename (may be shortened, could be changed by prctl etc)176 %f executable filename177 %E executable path178 %c maximum size of core file by resource limit RLIMIT_CORE179 %C CPU the task ran on180 %<OTHER> both are dropped181 ======== ==========================================182 183* If the first character of the pattern is a '|', the kernel will treat184 the rest of the pattern as a command to run. The core dump will be185 written to the standard input of that program instead of to a file.186 187 188core_pipe_limit189===============190 191This sysctl is only applicable when `core_pattern`_ is configured to192pipe core files to a user space helper (when the first character of193``core_pattern`` is a '|', see above).194When collecting cores via a pipe to an application, it is occasionally195useful for the collecting application to gather data about the196crashing process from its ``/proc/pid`` directory.197In order to do this safely, the kernel must wait for the collecting198process to exit, so as not to remove the crashing processes proc files199prematurely.200This in turn creates the possibility that a misbehaving userspace201collecting process can block the reaping of a crashed process simply202by never exiting.203This sysctl defends against that.204It defines how many concurrent crashing processes may be piped to user205space applications in parallel.206If this value is exceeded, then those crashing processes above that207value are noted via the kernel log and their cores are skipped.2080 is a special value, indicating that unlimited processes may be209captured in parallel, but that no waiting will take place (i.e. the210collecting process is not guaranteed access to ``/proc/<crashing211pid>/``).212This value defaults to 0.213 214 215core_uses_pid216=============217 218The default coredump filename is "core". By setting219``core_uses_pid`` to 1, the coredump filename becomes core.PID.220If `core_pattern`_ does not include "%p" (default does not)221and ``core_uses_pid`` is set, then .PID will be appended to222the filename.223 224 225ctrl-alt-del226============227 228When the value in this file is 0, ctrl-alt-del is trapped and229sent to the ``init(1)`` program to handle a graceful restart.230When, however, the value is > 0, Linux's reaction to a Vulcan231Nerve Pinch (tm) will be an immediate reboot, without even232syncing its dirty buffers.233 234Note:235 when a program (like dosemu) has the keyboard in 'raw'236 mode, the ctrl-alt-del is intercepted by the program before it237 ever reaches the kernel tty layer, and it's up to the program238 to decide what to do with it.239 240 241dmesg_restrict242==============243 244This toggle indicates whether unprivileged users are prevented245from using ``dmesg(8)`` to view messages from the kernel's log246buffer.247When ``dmesg_restrict`` is set to 0 there are no restrictions.248When ``dmesg_restrict`` is set to 1, users must have249``CAP_SYSLOG`` to use ``dmesg(8)``.250 251The kernel config option ``CONFIG_SECURITY_DMESG_RESTRICT`` sets the252default value of ``dmesg_restrict``.253 254 255domainname & hostname256=====================257 258These files can be used to set the NIS/YP domainname and the259hostname of your box in exactly the same way as the commands260domainname and hostname, i.e.::261 262 # echo "darkstar" > /proc/sys/kernel/hostname263 # echo "mydomain" > /proc/sys/kernel/domainname264 265has the same effect as::266 267 # hostname "darkstar"268 # domainname "mydomain"269 270Note, however, that the classic darkstar.frop.org has the271hostname "darkstar" and DNS (Internet Domain Name Server)272domainname "frop.org", not to be confused with the NIS (Network273Information Service) or YP (Yellow Pages) domainname. These two274domain names are in general different. For a detailed discussion275see the ``hostname(1)`` man page.276 277 278firmware_config279===============280 281See Documentation/driver-api/firmware/fallback-mechanisms.rst.282 283The entries in this directory allow the firmware loader helper284fallback to be controlled:285 286* ``force_sysfs_fallback``, when set to 1, forces the use of the287 fallback;288* ``ignore_sysfs_fallback``, when set to 1, ignores any fallback.289 290 291ftrace_dump_on_oops292===================293 294Determines whether ``ftrace_dump()`` should be called on an oops (or295kernel panic). This will output the contents of the ftrace buffers to296the console. This is very useful for capturing traces that lead to297crashes and outputting them to a serial console.298 299======================= ===========================================3000 Disabled (default).3011 Dump buffers of all CPUs.3022(orig_cpu) Dump the buffer of the CPU that triggered the303 oops.304<instance> Dump the specific instance buffer on all CPUs.305<instance>=2(orig_cpu) Dump the specific instance buffer on the CPU306 that triggered the oops.307======================= ===========================================308 309Multiple instance dump is also supported, and instances are separated310by commas. If global buffer also needs to be dumped, please specify311the dump mode (1/2/orig_cpu) first for global buffer.312 313So for example to dump "foo" and "bar" instance buffer on all CPUs,314user can::315 316 echo "foo,bar" > /proc/sys/kernel/ftrace_dump_on_oops317 318To dump global buffer and "foo" instance buffer on all319CPUs along with the "bar" instance buffer on CPU that triggered the320oops, user can::321 322 echo "1,foo,bar=2" > /proc/sys/kernel/ftrace_dump_on_oops323 324ftrace_enabled, stack_tracer_enabled325====================================326 327See Documentation/trace/ftrace.rst.328 329 330hardlockup_all_cpu_backtrace331============================332 333This value controls the hard lockup detector behavior when a hard334lockup condition is detected as to whether or not to gather further335debug information. If enabled, arch-specific all-CPU stack dumping336will be initiated.337 338= ============================================3390 Do nothing. This is the default behavior.3401 On detection capture more debug information.341= ============================================342 343 344hardlockup_panic345================346 347This parameter can be used to control whether the kernel panics348when a hard lockup is detected.349 350= ===========================3510 Don't panic on hard lockup.3521 Panic on hard lockup.353= ===========================354 355See Documentation/admin-guide/lockup-watchdogs.rst for more information.356This can also be set using the nmi_watchdog kernel parameter.357 358 359hotplug360=======361 362Path for the hotplug policy agent.363Default value is ``CONFIG_UEVENT_HELPER_PATH``, which in turn defaults364to the empty string.365 366This file only exists when ``CONFIG_UEVENT_HELPER`` is enabled. Most367modern systems rely exclusively on the netlink-based uevent source and368don't need this.369 370 371hung_task_all_cpu_backtrace372===========================373 374If this option is set, the kernel will send an NMI to all CPUs to dump375their backtraces when a hung task is detected. This file shows up if376CONFIG_DETECT_HUNG_TASK and CONFIG_SMP are enabled.377 3780: Won't show all CPUs backtraces when a hung task is detected.379This is the default behavior.380 3811: Will non-maskably interrupt all CPUs and dump their backtraces when382a hung task is detected.383 384 385hung_task_panic386===============387 388Controls the kernel's behavior when a hung task is detected.389This file shows up if ``CONFIG_DETECT_HUNG_TASK`` is enabled.390 391= =================================================3920 Continue operation. This is the default behavior.3931 Panic immediately.394= =================================================395 396 397hung_task_check_count398=====================399 400The upper bound on the number of tasks that are checked.401This file shows up if ``CONFIG_DETECT_HUNG_TASK`` is enabled.402 403 404hung_task_timeout_secs405======================406 407When a task in D state did not get scheduled408for more than this value report a warning.409This file shows up if ``CONFIG_DETECT_HUNG_TASK`` is enabled.410 4110 means infinite timeout, no checking is done.412 413Possible values to set are in range {0:``LONG_MAX``/``HZ``}.414 415 416hung_task_check_interval_secs417=============================418 419Hung task check interval. If hung task checking is enabled420(see `hung_task_timeout_secs`_), the check is done every421``hung_task_check_interval_secs`` seconds.422This file shows up if ``CONFIG_DETECT_HUNG_TASK`` is enabled.423 4240 (default) means use ``hung_task_timeout_secs`` as checking425interval.426 427Possible values to set are in range {0:``LONG_MAX``/``HZ``}.428 429 430hung_task_warnings431==================432 433The maximum number of warnings to report. During a check interval434if a hung task is detected, this value is decreased by 1.435When this value reaches 0, no more warnings will be reported.436This file shows up if ``CONFIG_DETECT_HUNG_TASK`` is enabled.437 438-1: report an infinite number of warnings.439 440 441hyperv_record_panic_msg442=======================443 444Controls whether the panic kmsg data should be reported to Hyper-V.445 446= =========================================================4470 Do not report panic kmsg data.4481 Report the panic kmsg data. This is the default behavior.449= =========================================================450 451 452ignore-unaligned-usertrap453=========================454 455On architectures where unaligned accesses cause traps, and where this456feature is supported (``CONFIG_SYSCTL_ARCH_UNALIGN_NO_WARN``;457currently, ``arc``, ``parisc`` and ``loongarch``), controls whether all458unaligned traps are logged.459 460= =============================================================4610 Log all unaligned accesses.4621 Only warn the first time a process traps. This is the default463 setting.464= =============================================================465 466See also `unaligned-trap`_.467 468io_uring_disabled469=================470 471Prevents all processes from creating new io_uring instances. Enabling this472shrinks the kernel's attack surface.473 474= ======================================================================4750 All processes can create io_uring instances as normal. This is the476 default setting.4771 io_uring creation is disabled (io_uring_setup() will fail with478 -EPERM) for unprivileged processes not in the io_uring_group group.479 Existing io_uring instances can still be used. See the480 documentation for io_uring_group for more information.4812 io_uring creation is disabled for all processes. io_uring_setup()482 always fails with -EPERM. Existing io_uring instances can still be483 used.484= ======================================================================485 486 487io_uring_group488==============489 490When io_uring_disabled is set to 1, a process must either be491privileged (CAP_SYS_ADMIN) or be in the io_uring_group group in order492to create an io_uring instance. If io_uring_group is set to -1 (the493default), only processes with the CAP_SYS_ADMIN capability may create494io_uring instances.495 496 497kexec_load_disabled498===================499 500A toggle indicating if the syscalls ``kexec_load`` and501``kexec_file_load`` have been disabled.502This value defaults to 0 (false: ``kexec_*load`` enabled), but can be503set to 1 (true: ``kexec_*load`` disabled).504Once true, kexec can no longer be used, and the toggle cannot be set505back to false.506This allows a kexec image to be loaded before disabling the syscall,507allowing a system to set up (and later use) an image without it being508altered.509Generally used together with the `modules_disabled`_ sysctl.510 511kexec_load_limit_panic512======================513 514This parameter specifies a limit to the number of times the syscalls515``kexec_load`` and ``kexec_file_load`` can be called with a crash516image. It can only be set with a more restrictive value than the517current one.518 519== ======================================================520-1 Unlimited calls to kexec. This is the default setting.521N Number of calls left.522== ======================================================523 524kexec_load_limit_reboot525=======================526 527Similar functionality as ``kexec_load_limit_panic``, but for a normal528image.529 530kptr_restrict531=============532 533This toggle indicates whether restrictions are placed on534exposing kernel addresses via ``/proc`` and other interfaces.535 536When ``kptr_restrict`` is set to 0 (the default) the address is hashed537before printing.538(This is the equivalent to %p.)539 540When ``kptr_restrict`` is set to 1, kernel pointers printed using the541%pK format specifier will be replaced with 0s unless the user has542``CAP_SYSLOG`` and effective user and group ids are equal to the real543ids.544This is because %pK checks are done at read() time rather than open()545time, so if permissions are elevated between the open() and the read()546(e.g via a setuid binary) then %pK will not leak kernel pointers to547unprivileged users.548Note, this is a temporary solution only.549The correct long-term solution is to do the permission checks at550open() time.551Consider removing world read permissions from files that use %pK, and552using `dmesg_restrict`_ to protect against uses of %pK in ``dmesg(8)``553if leaking kernel pointer values to unprivileged users is a concern.554 555When ``kptr_restrict`` is set to 2, kernel pointers printed using556%pK will be replaced with 0s regardless of privileges.557 558 559modprobe560========561 562The full path to the usermode helper for autoloading kernel modules,563by default ``CONFIG_MODPROBE_PATH``, which in turn defaults to564"/sbin/modprobe". This binary is executed when the kernel requests a565module. For example, if userspace passes an unknown filesystem type566to mount(), then the kernel will automatically request the567corresponding filesystem module by executing this usermode helper.568This usermode helper should insert the needed module into the kernel.569 570This sysctl only affects module autoloading. It has no effect on the571ability to explicitly insert modules.572 573This sysctl can be used to debug module loading requests::574 575 echo '#! /bin/sh' > /tmp/modprobe576 echo 'echo "$@" >> /tmp/modprobe.log' >> /tmp/modprobe577 echo 'exec /sbin/modprobe "$@"' >> /tmp/modprobe578 chmod a+x /tmp/modprobe579 echo /tmp/modprobe > /proc/sys/kernel/modprobe580 581Alternatively, if this sysctl is set to the empty string, then module582autoloading is completely disabled. The kernel will not try to583execute a usermode helper at all, nor will it call the584kernel_module_request LSM hook.585 586If CONFIG_STATIC_USERMODEHELPER=y is set in the kernel configuration,587then the configured static usermode helper overrides this sysctl,588except that the empty string is still accepted to completely disable589module autoloading as described above.590 591modules_disabled592================593 594A toggle value indicating if modules are allowed to be loaded595in an otherwise modular kernel. This toggle defaults to off596(0), but can be set true (1). Once true, modules can be597neither loaded nor unloaded, and the toggle cannot be set back598to false. Generally used with the `kexec_load_disabled`_ toggle.599 600 601.. _msgmni:602 603msgmax, msgmnb, and msgmni604==========================605 606``msgmax`` is the maximum size of an IPC message, in bytes. 8192 by607default (``MSGMAX``).608 609``msgmnb`` is the maximum size of an IPC queue, in bytes. 16384 by610default (``MSGMNB``).611 612``msgmni`` is the maximum number of IPC queues. 32000 by default613(``MSGMNI``).614 615All of these parameters are set per ipc namespace. The maximum number of bytes616in POSIX message queues is limited by ``RLIMIT_MSGQUEUE``. This limit is617respected hierarchically in the each user namespace.618 619msg_next_id, sem_next_id, and shm_next_id (System V IPC)620========================================================621 622These three toggles allows to specify desired id for next allocated IPC623object: message, semaphore or shared memory respectively.624 625By default they are equal to -1, which means generic allocation logic.626Possible values to set are in range {0:``INT_MAX``}.627 628Notes:629 1) kernel doesn't guarantee, that new object will have desired id. So,630 it's up to userspace, how to handle an object with "wrong" id.631 2) Toggle with non-default value will be set back to -1 by kernel after632 successful IPC object allocation. If an IPC object allocation syscall633 fails, it is undefined if the value remains unmodified or is reset to -1.634 635 636ngroups_max637===========638 639Maximum number of supplementary groups, _i.e._ the maximum size which640``setgroups`` will accept. Exports ``NGROUPS_MAX`` from the kernel.641 642 643 644nmi_watchdog645============646 647This parameter can be used to control the NMI watchdog648(i.e. the hard lockup detector) on x86 systems.649 650= =================================6510 Disable the hard lockup detector.6521 Enable the hard lockup detector.653= =================================654 655The hard lockup detector monitors each CPU for its ability to respond to656timer interrupts. The mechanism utilizes CPU performance counter registers657that are programmed to generate Non-Maskable Interrupts (NMIs) periodically658while a CPU is busy. Hence, the alternative name 'NMI watchdog'.659 660The NMI watchdog is disabled by default if the kernel is running as a guest661in a KVM virtual machine. This default can be overridden by adding::662 663 nmi_watchdog=1664 665to the guest kernel command line (see666Documentation/admin-guide/kernel-parameters.rst).667 668 669nmi_wd_lpm_factor (PPC only)670============================671 672Factor to apply to the NMI watchdog timeout (only when ``nmi_watchdog`` is673set to 1). This factor represents the percentage added to674``watchdog_thresh`` when calculating the NMI watchdog timeout during an675LPM. The soft lockup timeout is not impacted.676 677A value of 0 means no change. The default value is 200 meaning the NMI678watchdog is set to 30s (based on ``watchdog_thresh`` equal to 10).679 680 681numa_balancing682==============683 684Enables/disables and configures automatic page fault based NUMA memory685balancing. Memory is moved automatically to nodes that access it often.686The value to set can be the result of ORing the following:687 688= =================================6890 NUMA_BALANCING_DISABLED6901 NUMA_BALANCING_NORMAL6912 NUMA_BALANCING_MEMORY_TIERING692= =================================693 694Or NUMA_BALANCING_NORMAL to optimize page placement among different695NUMA nodes to reduce remote accessing. On NUMA machines, there is a696performance penalty if remote memory is accessed by a CPU. When this697feature is enabled the kernel samples what task thread is accessing698memory by periodically unmapping pages and later trapping a page699fault. At the time of the page fault, it is determined if the data700being accessed should be migrated to a local memory node.701 702The unmapping of pages and trapping faults incur additional overhead that703ideally is offset by improved memory locality but there is no universal704guarantee. If the target workload is already bound to NUMA nodes then this705feature should be disabled.706 707Or NUMA_BALANCING_MEMORY_TIERING to optimize page placement among708different types of memory (represented as different NUMA nodes) to709place the hot pages in the fast memory. This is implemented based on710unmapping and page fault too.711 712numa_balancing_promote_rate_limit_MBps713======================================714 715Too high promotion/demotion throughput between different memory types716may hurt application latency. This can be used to rate limit the717promotion throughput. The per-node max promotion throughput in MB/s718will be limited to be no more than the set value.719 720A rule of thumb is to set this to less than 1/10 of the PMEM node721write bandwidth.722 723oops_all_cpu_backtrace724======================725 726If this option is set, the kernel will send an NMI to all CPUs to dump727their backtraces when an oops event occurs. It should be used as a last728resort in case a panic cannot be triggered (to protect VMs running, for729example) or kdump can't be collected. This file shows up if CONFIG_SMP730is enabled.731 7320: Won't show all CPUs backtraces when an oops is detected.733This is the default behavior.734 7351: Will non-maskably interrupt all CPUs and dump their backtraces when736an oops event is detected.737 738 739oops_limit740==========741 742Number of kernel oopses after which the kernel should panic when743``panic_on_oops`` is not set. Setting this to 0 disables checking744the count. Setting this to 1 has the same effect as setting745``panic_on_oops=1``. The default value is 10000.746 747 748osrelease, ostype & version749===========================750 751::752 753 # cat osrelease754 2.1.88755 # cat ostype756 Linux757 # cat version758 #5 Wed Feb 25 21:49:24 MET 1998759 760The files ``osrelease`` and ``ostype`` should be clear enough.761``version``762needs a little more clarification however. The '#5' means that763this is the fifth kernel built from this source base and the764date behind it indicates the time the kernel was built.765The only way to tune these values is to rebuild the kernel :-)766 767 768overflowgid & overflowuid769=========================770 771if your architecture did not always support 32-bit UIDs (i.e. arm,772i386, m68k, sh, and sparc32), a fixed UID and GID will be returned to773applications that use the old 16-bit UID/GID system calls, if the774actual UID or GID would exceed 65535.775 776These sysctls allow you to change the value of the fixed UID and GID.777The default is 65534.778 779 780panic781=====782 783The value in this file determines the behaviour of the kernel on a784panic:785 786* if zero, the kernel will loop forever;787* if negative, the kernel will reboot immediately;788* if positive, the kernel will reboot after the corresponding number789 of seconds.790 791When you use the software watchdog, the recommended setting is 60.792 793 794panic_on_io_nmi795===============796 797Controls the kernel's behavior when a CPU receives an NMI caused by798an IO error.799 800= ==================================================================8010 Try to continue operation (default).8021 Panic immediately. The IO error triggered an NMI. This indicates a803 serious system condition which could result in IO data corruption.804 Rather than continuing, panicking might be a better choice. Some805 servers issue this sort of NMI when the dump button is pushed,806 and you can use this option to take a crash dump.807= ==================================================================808 809 810panic_on_oops811=============812 813Controls the kernel's behaviour when an oops or BUG is encountered.814 815= ===================================================================8160 Try to continue operation.8171 Panic immediately. If the `panic` sysctl is also non-zero then the818 machine will be rebooted.819= ===================================================================820 821 822panic_on_stackoverflow823======================824 825Controls the kernel's behavior when detecting the overflows of826kernel, IRQ and exception stacks except a user stack.827This file shows up if ``CONFIG_DEBUG_STACKOVERFLOW`` is enabled.828 829= ==========================8300 Try to continue operation.8311 Panic immediately.832= ==========================833 834 835panic_on_unrecovered_nmi836========================837 838The default Linux behaviour on an NMI of either memory or unknown is839to continue operation. For many environments such as scientific840computing it is preferable that the box is taken out and the error841dealt with than an uncorrected parity/ECC error get propagated.842 843A small number of systems do generate NMIs for bizarre random reasons844such as power management so the default is off. That sysctl works like845the existing panic controls already in that directory.846 847 848panic_on_warn849=============850 851Calls panic() in the WARN() path when set to 1. This is useful to avoid852a kernel rebuild when attempting to kdump at the location of a WARN().853 854= ================================================8550 Only WARN(), default behaviour.8561 Call panic() after printing out WARN() location.857= ================================================858 859 860panic_print861===========862 863Bitmask for printing system info when panic happens. User can chose864combination of the following bits:865 866===== ============================================867bit 0 print all tasks info868bit 1 print system memory info869bit 2 print timer info870bit 3 print locks info if ``CONFIG_LOCKDEP`` is on871bit 4 print ftrace buffer872bit 5 print all printk messages in buffer873bit 6 print all CPUs backtrace (if available in the arch)874bit 7 print only tasks in uninterruptible (blocked) state875===== ============================================876 877So for example to print tasks and memory info on panic, user can::878 879 echo 3 > /proc/sys/kernel/panic_print880 881 882panic_on_rcu_stall883==================884 885When set to 1, calls panic() after RCU stall detection messages. This886is useful to define the root cause of RCU stalls using a vmcore.887 888= ============================================================8890 Do not panic() when RCU stall takes place, default behavior.8901 panic() after printing RCU stall messages.891= ============================================================892 893max_rcu_stall_to_panic894======================895 896When ``panic_on_rcu_stall`` is set to 1, this value determines the897number of times that RCU can stall before panic() is called.898 899When ``panic_on_rcu_stall`` is set to 0, this value is has no effect.900 901perf_cpu_time_max_percent902=========================903 904Hints to the kernel how much CPU time it should be allowed to905use to handle perf sampling events. If the perf subsystem906is informed that its samples are exceeding this limit, it907will drop its sampling frequency to attempt to reduce its CPU908usage.909 910Some perf sampling happens in NMIs. If these samples911unexpectedly take too long to execute, the NMIs can become912stacked up next to each other so much that nothing else is913allowed to execute.914 915===== ========================================================9160 Disable the mechanism. Do not monitor or correct perf's917 sampling rate no matter how CPU time it takes.918 9191-100 Attempt to throttle perf's sample rate to this920 percentage of CPU. Note: the kernel calculates an921 "expected" length of each sample event. 100 here means922 100% of that expected length. Even if this is set to923 100, you may still see sample throttling if this924 length is exceeded. Set to 0 if you truly do not care925 how much CPU is consumed.926===== ========================================================927 928 929perf_event_paranoid930===================931 932Controls use of the performance events system by unprivileged933users (without CAP_PERFMON). The default value is 2.934 935For backward compatibility reasons access to system performance936monitoring and observability remains open for CAP_SYS_ADMIN937privileged processes but CAP_SYS_ADMIN usage for secure system938performance monitoring and observability operations is discouraged939with respect to CAP_PERFMON use cases.940 941=== ==================================================================942 -1 Allow use of (almost) all events by all users.943 944 Ignore mlock limit after perf_event_mlock_kb without945 ``CAP_IPC_LOCK``.946 947>=0 Disallow ftrace function tracepoint by users without948 ``CAP_PERFMON``.949 950 Disallow raw tracepoint access by users without ``CAP_PERFMON``.951 952>=1 Disallow CPU event access by users without ``CAP_PERFMON``.953 954>=2 Disallow kernel profiling by users without ``CAP_PERFMON``.955=== ==================================================================956 957 958perf_event_max_stack959====================960 961Controls maximum number of stack frames to copy for (``attr.sample_type &962PERF_SAMPLE_CALLCHAIN``) configured events, for instance, when using963'``perf record -g``' or '``perf trace --call-graph fp``'.964 965This can only be done when no events are in use that have callchains966enabled, otherwise writing to this file will return ``-EBUSY``.967 968The default value is 127.969 970 971perf_event_mlock_kb972===================973 974Control size of per-cpu ring buffer not counted against mlock limit.975 976The default value is 512 + 1 page977 978 979perf_event_max_contexts_per_stack980=================================981 982Controls maximum number of stack frame context entries for983(``attr.sample_type & PERF_SAMPLE_CALLCHAIN``) configured events, for984instance, when using '``perf record -g``' or '``perf trace --call-graph fp``'.985 986This can only be done when no events are in use that have callchains987enabled, otherwise writing to this file will return ``-EBUSY``.988 989The default value is 8.990 991 992perf_user_access (arm64 and riscv only)993=======================================994 995Controls user space access for reading perf event counters.996 997arm64998=====999 1000The default value is 0 (access disabled).1001 1002When set to 1, user space can read performance monitor counter registers1003directly.1004 1005See Documentation/arch/arm64/perf.rst for more information.1006 1007riscv1008=====1009 1010When set to 0, user space access is disabled.1011 1012The default value is 1, user space can read performance monitor counter1013registers through perf, any direct access without perf intervention will trigger1014an illegal instruction.1015 1016When set to 2, which enables legacy mode (user space has direct access to cycle1017and insret CSRs only). Note that this legacy value is deprecated and will be1018removed once all user space applications are fixed.1019 1020Note that the time CSR is always directly accessible to all modes.1021 1022pid_max1023=======1024 1025PID allocation wrap value. When the kernel's next PID value1026reaches this value, it wraps back to a minimum PID value.1027PIDs of value ``pid_max`` or larger are not allocated.1028 1029 1030ns_last_pid1031===========1032 1033The last pid allocated in the current (the one task using this sysctl1034lives in) pid namespace. When selecting a pid for a next task on fork1035kernel tries to allocate a number starting from this one.1036 1037 1038powersave-nap (PPC only)1039========================1040 1041If set, Linux-PPC will use the 'nap' mode of powersaving,1042otherwise the 'doze' mode will be used.1043 1044 1045==============================================================1046 1047printk1048======1049 1050The four values in printk denote: ``console_loglevel``,1051``default_message_loglevel``, ``minimum_console_loglevel`` and1052``default_console_loglevel`` respectively.1053 1054These values influence printk() behavior when printing or1055logging error messages. See '``man 2 syslog``' for more info on1056the different loglevels.1057 1058======================== =====================================1059console_loglevel messages with a higher priority than1060 this will be printed to the console1061default_message_loglevel messages without an explicit priority1062 will be printed with this priority1063minimum_console_loglevel minimum (highest) value to which1064 console_loglevel can be set1065default_console_loglevel default value for console_loglevel1066======================== =====================================1067 1068 1069printk_delay1070============1071 1072Delay each printk message in ``printk_delay`` milliseconds1073 1074Value from 0 - 10000 is allowed.1075 1076 1077printk_ratelimit1078================1079 1080Some warning messages are rate limited. ``printk_ratelimit`` specifies1081the minimum length of time between these messages (in seconds).1082The default value is 5 seconds.1083 1084A value of 0 will disable rate limiting.1085 1086 1087printk_ratelimit_burst1088======================1089 1090While long term we enforce one message per `printk_ratelimit`_1091seconds, we do allow a burst of messages to pass through.1092``printk_ratelimit_burst`` specifies the number of messages we can1093send before ratelimiting kicks in.1094 1095The default value is 10 messages.1096 1097 1098printk_devkmsg1099==============1100 1101Control the logging to ``/dev/kmsg`` from userspace:1102 1103========= =============================================1104ratelimit default, ratelimited1105on unlimited logging to /dev/kmsg from userspace1106off logging to /dev/kmsg disabled1107========= =============================================1108 1109The kernel command line parameter ``printk.devkmsg=`` overrides this and is1110a one-time setting until next reboot: once set, it cannot be changed by1111this sysctl interface anymore.1112 1113==============================================================1114 1115 1116pty1117===1118 1119See Documentation/filesystems/devpts.rst.1120 1121 1122random1123======1124 1125This is a directory, with the following entries:1126 1127* ``boot_id``: a UUID generated the first time this is retrieved, and1128 unvarying after that;1129 1130* ``uuid``: a UUID generated every time this is retrieved (this can1131 thus be used to generate UUIDs at will);1132 1133* ``entropy_avail``: the pool's entropy count, in bits;1134 1135* ``poolsize``: the entropy pool size, in bits;1136 1137* ``urandom_min_reseed_secs``: obsolete (used to determine the minimum1138 number of seconds between urandom pool reseeding). This file is1139 writable for compatibility purposes, but writing to it has no effect1140 on any RNG behavior;1141 1142* ``write_wakeup_threshold``: when the entropy count drops below this1143 (as a number of bits), processes waiting to write to ``/dev/random``1144 are woken up. This file is writable for compatibility purposes, but1145 writing to it has no effect on any RNG behavior.1146 1147 1148randomize_va_space1149==================1150 1151This option can be used to select the type of process address1152space randomization that is used in the system, for architectures1153that support this feature.1154 1155== ===========================================================================11560 Turn the process address space randomization off. This is the1157 default for architectures that do not support this feature anyways,1158 and kernels that are booted with the "norandmaps" parameter.1159 11601 Make the addresses of mmap base, stack and VDSO page randomized.1161 This, among other things, implies that shared libraries will be1162 loaded to random addresses. Also for PIE-linked binaries, the1163 location of code start is randomized. This is the default if the1164 ``CONFIG_COMPAT_BRK`` option is enabled.1165 11662 Additionally enable heap randomization. This is the default if1167 ``CONFIG_COMPAT_BRK`` is disabled.1168 1169 There are a few legacy applications out there (such as some ancient1170 versions of libc.so.5 from 1996) that assume that brk area starts1171 just after the end of the code+bss. These applications break when1172 start of the brk area is randomized. There are however no known1173 non-legacy applications that would be broken this way, so for most1174 systems it is safe to choose full randomization.1175 1176 Systems with ancient and/or broken binaries should be configured1177 with ``CONFIG_COMPAT_BRK`` enabled, which excludes the heap from process1178 address space randomization.1179== ===========================================================================1180 1181 1182real-root-dev1183=============1184 1185See Documentation/admin-guide/initrd.rst.1186 1187 1188reboot-cmd (SPARC only)1189=======================1190 1191??? This seems to be a way to give an argument to the Sparc1192ROM/Flash boot loader. Maybe to tell it what to do after1193rebooting. ???1194 1195 1196sched_energy_aware1197==================1198 1199Enables/disables Energy Aware Scheduling (EAS). EAS starts1200automatically on platforms where it can run (that is,1201platforms with asymmetric CPU topologies and having an Energy1202Model available). If your platform happens to meet the1203requirements for EAS but you do not want to use it, change1204this value to 0. On Non-EAS platforms, write operation fails and1205read doesn't return anything.1206 1207task_delayacct1208===============1209 1210Enables/disables task delay accounting (see1211Documentation/accounting/delay-accounting.rst. Enabling this feature incurs1212a small amount of overhead in the scheduler but is useful for debugging1213and performance tuning. It is required by some tools such as iotop.1214 1215sched_schedstats1216================1217 1218Enables/disables scheduler statistics. Enabling this feature1219incurs a small amount of overhead in the scheduler but is1220useful for debugging and performance tuning.1221 1222sched_util_clamp_min1223====================1224 1225Max allowed *minimum* utilization.1226 1227Default value is 1024, which is the maximum possible value.1228 1229It means that any requested uclamp.min value cannot be greater than1230sched_util_clamp_min, i.e., it is restricted to the range1231[0:sched_util_clamp_min].1232 1233sched_util_clamp_max1234====================1235 1236Max allowed *maximum* utilization.1237 1238Default value is 1024, which is the maximum possible value.1239 1240It means that any requested uclamp.max value cannot be greater than1241sched_util_clamp_max, i.e., it is restricted to the range1242[0:sched_util_clamp_max].1243 1244sched_util_clamp_min_rt_default1245===============================1246 1247By default Linux is tuned for performance. Which means that RT tasks always run1248at the highest frequency and most capable (highest capacity) CPU (in1249heterogeneous systems).1250 1251Uclamp achieves this by setting the requested uclamp.min of all RT tasks to12521024 by default, which effectively boosts the tasks to run at the highest1253frequency and biases them to run on the biggest CPU.1254 1255This knob allows admins to change the default behavior when uclamp is being1256used. In battery powered devices particularly, running at the maximum1257capacity and frequency will increase energy consumption and shorten the battery1258life.1259 1260This knob is only effective for RT tasks which the user hasn't modified their1261requested uclamp.min value via sched_setattr() syscall.1262 1263This knob will not escape the range constraint imposed by sched_util_clamp_min1264defined above.1265 1266For example if1267 1268 sched_util_clamp_min_rt_default = 8001269 sched_util_clamp_min = 6001270 1271Then the boost will be clamped to 600 because 800 is outside of the permissible1272range of [0:600]. This could happen for instance if a powersave mode will1273restrict all boosts temporarily by modifying sched_util_clamp_min. As soon as1274this restriction is lifted, the requested sched_util_clamp_min_rt_default1275will take effect.1276 1277seccomp1278=======1279 1280See Documentation/userspace-api/seccomp_filter.rst.1281 1282 1283sg-big-buff1284===========1285 1286This file shows the size of the generic SCSI (sg) buffer.1287You can't tune it just yet, but you could change it on1288compile time by editing ``include/scsi/sg.h`` and changing1289the value of ``SG_BIG_BUFF``.1290 1291There shouldn't be any reason to change this value. If1292you can come up with one, you probably know what you1293are doing anyway :)1294 1295 1296shmall1297======1298 1299This parameter sets the total amount of shared memory pages that can be used1300inside ipc namespace. The shared memory pages counting occurs for each ipc1301namespace separately and is not inherited. Hence, ``shmall`` should always be at1302least ``ceil(shmmax/PAGE_SIZE)``.1303 1304If you are not sure what the default ``PAGE_SIZE`` is on your Linux1305system, you can run the following command::1306 1307 # getconf PAGE_SIZE1308 1309To reduce or disable the ability to allocate shared memory, you must create a1310new ipc namespace, set this parameter to the required value and prohibit the1311creation of a new ipc namespace in the current user namespace or cgroups can1312be used.1313 1314shmmax1315======1316 1317This value can be used to query and set the run time limit1318on the maximum shared memory segment size that can be created.1319Shared memory segments up to 1Gb are now supported in the1320kernel. This value defaults to ``SHMMAX``.1321 1322 1323shmmni1324======1325 1326This value determines the maximum number of shared memory segments.13274096 by default (``SHMMNI``).1328 1329 1330shm_rmid_forced1331===============1332 1333Linux lets you set resource limits, including how much memory one1334process can consume, via ``setrlimit(2)``. Unfortunately, shared memory1335segments are allowed to exist without association with any process, and1336thus might not be counted against any resource limits. If enabled,1337shared memory segments are automatically destroyed when their attach1338count becomes zero after a detach or a process termination. It will1339also destroy segments that were created, but never attached to, on exit1340from the process. The only use left for ``IPC_RMID`` is to immediately1341destroy an unattached segment. Of course, this breaks the way things are1342defined, so some applications might stop working. Note that this1343feature will do you no good unless you also configure your resource1344limits (in particular, ``RLIMIT_AS`` and ``RLIMIT_NPROC``). Most systems don't1345need this.1346 1347Note that if you change this from 0 to 1, already created segments1348without users and with a dead originative process will be destroyed.1349 1350 1351sysctl_writes_strict1352====================1353 1354Control how file position affects the behavior of updating sysctl values1355via the ``/proc/sys`` interface:1356 1357 == ======================================================================1358 -1 Legacy per-write sysctl value handling, with no printk warnings.1359 Each write syscall must fully contain the sysctl value to be1360 written, and multiple writes on the same sysctl file descriptor1361 will rewrite the sysctl value, regardless of file position.1362 0 Same behavior as above, but warn about processes that perform writes1363 to a sysctl file descriptor when the file position is not 0.1364 1 (default) Respect file position when writing sysctl strings. Multiple1365 writes will append to the sysctl value buffer. Anything past the max1366 length of the sysctl value buffer will be ignored. Writes to numeric1367 sysctl entries must always be at file position 0 and the value must1368 be fully contained in the buffer sent in the write syscall.1369 == ======================================================================1370 1371 1372softlockup_all_cpu_backtrace1373============================1374 1375This value controls the soft lockup detector thread's behavior1376when a soft lockup condition is detected as to whether or not1377to gather further debug information. If enabled, each cpu will1378be issued an NMI and instructed to capture stack trace.1379 1380This feature is only applicable for architectures which support1381NMI.1382 1383= ============================================13840 Do nothing. This is the default behavior.13851 On detection capture more debug information.1386= ============================================1387 1388 1389softlockup_panic1390=================1391 1392This parameter can be used to control whether the kernel panics1393when a soft lockup is detected.1394 1395= ============================================13960 Don't panic on soft lockup.13971 Panic on soft lockup.1398= ============================================1399 1400This can also be set using the softlockup_panic kernel parameter.1401 1402 1403soft_watchdog1404=============1405 1406This parameter can be used to control the soft lockup detector.1407 1408= =================================14090 Disable the soft lockup detector.14101 Enable the soft lockup detector.1411= =================================1412 1413The soft lockup detector monitors CPUs for threads that are hogging the CPUs1414without rescheduling voluntarily, and thus prevent the 'migration/N' threads1415from running, causing the watchdog work fail to execute. The mechanism depends1416on the CPUs ability to respond to timer interrupts which are needed for the1417watchdog work to be queued by the watchdog timer function, otherwise the NMI1418watchdog — if enabled — can detect a hard lockup condition.1419 1420 1421split_lock_mitigate (x86 only)1422==============================1423 1424On x86, each "split lock" imposes a system-wide performance penalty. On larger1425systems, large numbers of split locks from unprivileged users can result in1426denials of service to well-behaved and potentially more important users.1427 1428The kernel mitigates these bad users by detecting split locks and imposing1429penalties: forcing them to wait and only allowing one core to execute split1430locks at a time.1431 1432These mitigations can make those bad applications unbearably slow. Setting1433split_lock_mitigate=0 may restore some application performance, but will also1434increase system exposure to denial of service attacks from split lock users.1435 1436= ===================================================================14370 Disable the mitigation mode - just warns the split lock on kernel log1438 and exposes the system to denials of service from the split lockers.14391 Enable the mitigation mode (this is the default) - penalizes the split1440 lockers with intentional performance degradation.1441= ===================================================================1442 1443 1444stack_erasing1445=============1446 1447This parameter can be used to control kernel stack erasing at the end1448of syscalls for kernels built with ``CONFIG_GCC_PLUGIN_STACKLEAK``.1449 1450That erasing reduces the information which kernel stack leak bugs1451can reveal and blocks some uninitialized stack variable attacks.1452The tradeoff is the performance impact: on a single CPU system kernel1453compilation sees a 1% slowdown, other systems and workloads may vary.1454 1455= ====================================================================14560 Kernel stack erasing is disabled, STACKLEAK_METRICS are not updated.14571 Kernel stack erasing is enabled (default), it is performed before1458 returning to the userspace at the end of syscalls.1459= ====================================================================1460 1461 1462stop-a (SPARC only)1463===================1464 1465Controls Stop-A:1466 1467= ====================================14680 Stop-A has no effect.14691 Stop-A breaks to the PROM (default).1470= ====================================1471 1472Stop-A is always enabled on a panic, so that the user can return to1473the boot PROM.1474 1475 1476sysrq1477=====1478 1479See Documentation/admin-guide/sysrq.rst.1480 1481 1482tainted1483=======1484 1485Non-zero if the kernel has been tainted. Numeric values, which can be1486ORed together. The letters are seen in "Tainted" line of Oops reports.1487 1488====== ===== ==============================================================1489 1 `(P)` proprietary module was loaded1490 2 `(F)` module was force loaded1491 4 `(S)` kernel running on an out of specification system1492 8 `(R)` module was force unloaded1493 16 `(M)` processor reported a Machine Check Exception (MCE)1494 32 `(B)` bad page referenced or some unexpected page flags1495 64 `(U)` taint requested by userspace application1496 128 `(D)` kernel died recently, i.e. there was an OOPS or BUG1497 256 `(A)` an ACPI table was overridden by user1498 512 `(W)` kernel issued warning1499 1024 `(C)` staging driver was loaded1500 2048 `(I)` workaround for bug in platform firmware applied1501 4096 `(O)` externally-built ("out-of-tree") module was loaded1502 8192 `(E)` unsigned module was loaded1503 16384 `(L)` soft lockup occurred1504 32768 `(K)` kernel has been live patched1505 65536 `(X)` Auxiliary taint, defined and used by for distros1506131072 `(T)` The kernel was built with the struct randomization plugin1507====== ===== ==============================================================1508 1509See Documentation/admin-guide/tainted-kernels.rst for more information.1510 1511Note:1512 writes to this sysctl interface will fail with ``EINVAL`` if the kernel is1513 booted with the command line option ``panic_on_taint=<bitmask>,nousertaint``1514 and any of the ORed together values being written to ``tainted`` match with1515 the bitmask declared on panic_on_taint.1516 See Documentation/admin-guide/kernel-parameters.rst for more details on1517 that particular kernel command line option and its optional1518 ``nousertaint`` switch.1519 1520threads-max1521===========1522 1523This value controls the maximum number of threads that can be created1524using ``fork()``.1525 1526During initialization the kernel sets this value such that even if the1527maximum number of threads is created, the thread structures occupy only1528a part (1/8th) of the available RAM pages.1529 1530The minimum value that can be written to ``threads-max`` is 1.1531 1532The maximum value that can be written to ``threads-max`` is given by the1533constant ``FUTEX_TID_MASK`` (0x3fffffff).1534 1535If a value outside of this range is written to ``threads-max`` an1536``EINVAL`` error occurs.1537 1538 1539traceoff_on_warning1540===================1541 1542When set, disables tracing (see Documentation/trace/ftrace.rst) when a1543``WARN()`` is hit.1544 1545 1546tracepoint_printk1547=================1548 1549When tracepoints are sent to printk() (enabled by the ``tp_printk``1550boot parameter), this entry provides runtime control::1551 1552 echo 0 > /proc/sys/kernel/tracepoint_printk1553 1554will stop tracepoints from being sent to printk(), and::1555 1556 echo 1 > /proc/sys/kernel/tracepoint_printk1557 1558will send them to printk() again.1559 1560This only works if the kernel was booted with ``tp_printk`` enabled.1561 1562See Documentation/admin-guide/kernel-parameters.rst and1563Documentation/trace/boottime-trace.rst.1564 1565 1566unaligned-trap1567==============1568 1569On architectures where unaligned accesses cause traps, and where this1570feature is supported (``CONFIG_SYSCTL_ARCH_UNALIGN_ALLOW``; currently,1571``arc``, ``parisc`` and ``loongarch``), controls whether unaligned traps1572are caught and emulated (instead of failing).1573 1574= ========================================================15750 Do not emulate unaligned accesses.15761 Emulate unaligned accesses. This is the default setting.1577= ========================================================1578 1579See also `ignore-unaligned-usertrap`_.1580 1581 1582unknown_nmi_panic1583=================1584 1585The value in this file affects behavior of handling NMI. When the1586value is non-zero, unknown NMI is trapped and then panic occurs. At1587that time, kernel debugging information is displayed on console.1588 1589NMI switch that most IA32 servers have fires unknown NMI up, for1590example. If a system hangs up, try pressing the NMI switch.1591 1592 1593unprivileged_bpf_disabled1594=========================1595 1596Writing 1 to this entry will disable unprivileged calls to ``bpf()``;1597once disabled, calling ``bpf()`` without ``CAP_SYS_ADMIN`` or ``CAP_BPF``1598will return ``-EPERM``. Once set to 1, this can't be cleared from the1599running kernel anymore.1600 1601Writing 2 to this entry will also disable unprivileged calls to ``bpf()``,1602however, an admin can still change this setting later on, if needed, by1603writing 0 or 1 to this entry.1604 1605If ``BPF_UNPRIV_DEFAULT_OFF`` is enabled in the kernel config, then this1606entry will default to 2 instead of 0.1607 1608= =============================================================16090 Unprivileged calls to ``bpf()`` are enabled16101 Unprivileged calls to ``bpf()`` are disabled without recovery16112 Unprivileged calls to ``bpf()`` are disabled1612= =============================================================1613 1614 1615warn_limit1616==========1617 1618Number of kernel warnings after which the kernel should panic when1619``panic_on_warn`` is not set. Setting this to 0 disables checking1620the warning count. Setting this to 1 has the same effect as setting1621``panic_on_warn=1``. The default value is 0.1622 1623 1624watchdog1625========1626 1627This parameter can be used to disable or enable the soft lockup detector1628*and* the NMI watchdog (i.e. the hard lockup detector) at the same time.1629 1630= ==============================16310 Disable both lockup detectors.16321 Enable both lockup detectors.1633= ==============================1634 1635The soft lockup detector and the NMI watchdog can also be disabled or1636enabled individually, using the ``soft_watchdog`` and ``nmi_watchdog``1637parameters.1638If the ``watchdog`` parameter is read, for example by executing::1639 1640 cat /proc/sys/kernel/watchdog1641 1642the output of this command (0 or 1) shows the logical OR of1643``soft_watchdog`` and ``nmi_watchdog``.1644 1645 1646watchdog_cpumask1647================1648 1649This value can be used to control on which cpus the watchdog may run.1650The default cpumask is all possible cores, but if ``NO_HZ_FULL`` is1651enabled in the kernel config, and cores are specified with the1652``nohz_full=`` boot argument, those cores are excluded by default.1653Offline cores can be included in this mask, and if the core is later1654brought online, the watchdog will be started based on the mask value.1655 1656Typically this value would only be touched in the ``nohz_full`` case1657to re-enable cores that by default were not running the watchdog,1658if a kernel lockup was suspected on those cores.1659 1660The argument value is the standard cpulist format for cpumasks,1661so for example to enable the watchdog on cores 0, 2, 3, and 4 you1662might say::1663 1664 echo 0,2-4 > /proc/sys/kernel/watchdog_cpumask1665 1666 1667watchdog_thresh1668===============1669 1670This value can be used to control the frequency of hrtimer and NMI1671events and the soft and hard lockup thresholds. The default threshold1672is 10 seconds.1673 1674The softlockup threshold is (``2 * watchdog_thresh``). Setting this1675tunable to zero will disable lockup detection altogether.1676