1 ===================================
2 Documentation for /proc/sys/kernel/
3 ===================================
7 Copyright (c) 1998, 1999, Rik van Riel <riel@nl.linux.org>
9 Copyright (c) 2009, Shen Feng<shen@cn.fujitsu.com>
11 For general info and legal blurb, please look in index.rst.
13 ------------------------------------------------------------------------------
15 This file contains documentation for the sysctl files in
16 /proc/sys/kernel/ and is valid for Linux kernel version 2.2.
18 The files in this directory can be used to tune and monitor
19 miscellaneous and general things in the operation of the Linux
20 kernel. Since some of the files _can_ be used to screw up your
21 system, it is advisable to read both documentation and source
22 before actually making adjustments.
24 Currently, these files might (depending on your configuration)
25 show up in /proc/sys/kernel:
30 - bootloader_type [ X86 only ]
31 - bootloader_version [ X86 only ]
41 - hardlockup_all_cpu_backtrace
44 - hung_task_check_count
45 - hung_task_timeout_secs
46 - hung_task_check_interval_secs
48 - hyperv_record_panic_msg
52 - modprobe ==> Documentation/debugging-modules.txt
54 - msg_next_id [ sysv ipc ]
65 - panic_on_stackoverflow
66 - panic_on_unrecovered_nmi
70 - perf_cpu_time_max_percent
72 - perf_event_max_stack
74 - perf_event_max_contexts_per_stack
76 - powersave-nap [ PPC only ]
80 - printk_ratelimit_burst
81 - pty ==> Documentation/filesystems/devpts.txt
83 - real-root-dev ==> Documentation/admin-guide/initrd.rst
84 - reboot-cmd [ SPARC only ]
88 - seccomp/ ==> Documentation/userspace-api/seccomp_filter.rst
90 - sem_next_id [ sysv ipc ]
91 - sg-big-buff [ generic SCSI device (sg) ]
92 - shm_next_id [ sysv ipc ]
97 - softlockup_all_cpu_backtrace
100 - stop-a [ SPARC only ]
101 - sysrq ==> Documentation/admin-guide/sysrq.rst
102 - sysctl_writes_strict
103 - tainted ==> Documentation/admin-guide/tainted-kernels.rst
114 highwater lowwater frequency
116 If BSD-style process accounting is enabled these values control
117 its behaviour. If free space on filesystem where the log lives
118 goes below <lowwater>% accounting suspends. If free space gets
119 above <highwater>% accounting resumes. <Frequency> determines
120 how often do we check the amount of free space (value is in
123 That is, suspend accounting if there left <= 2% free; resume it
124 if we got >=4%; consider information about amount of free space
125 valid for 30 seconds.
133 See Doc*/kernel/power/video.txt, it allows mode of video boot to be
140 This variable has no effect and may be removed in future kernel
141 releases. Reading it always returns 0.
142 Up to Linux 3.17, it enabled/disabled automatic recomputing of msgmni
143 upon memory add/remove or upon ipc namespace creation/removal.
144 Echoing "1" into this file enabled msgmni automatic recomputing.
145 Echoing "0" turned it off. auto_msgmni default value was 1.
151 x86 bootloader identification
153 This gives the bootloader type number as indicated by the bootloader,
154 shifted left by 4, and OR'd with the low four bits of the bootloader
155 version. The reason for this encoding is that this used to match the
156 type_of_loader field in the kernel header; the encoding is kept for
157 backwards compatibility. That is, if the full bootloader type number
158 is 0x15 and the full version number is 0x234, this file will contain
159 the value 340 = 0x154.
161 See the type_of_loader and ext_loader_type fields in
162 Documentation/x86/boot.rst for additional information.
168 x86 bootloader version
170 The complete bootloader version number. In the example above, this
171 file will contain the value 564 = 0x234.
173 See the type_of_loader and ext_loader_ver fields in
174 Documentation/x86/boot.rst for additional information.
180 Highest valid capability of the running kernel. Exports
181 CAP_LAST_CAP from the kernel.
187 core_pattern is used to specify a core dumpfile pattern name.
189 * max length 127 characters; default value is "core"
190 * core_pattern is used as a pattern template for the output filename;
191 certain string patterns (beginning with '%') are substituted with
193 * backward compatibility with core_uses_pid:
195 If core_pattern does not include "%p" (default does not)
196 and core_uses_pid is set, then .PID will be appended to
199 * corename format specifiers::
201 %<NUL> '%' is dropped
204 %P global pid (init PID namespace)
206 %I global tid (init PID namespace)
207 %u uid (in initial user namespace)
208 %g gid (in initial user namespace)
209 %d dump mode, matches PR_SET_DUMPABLE and
210 /proc/sys/fs/suid_dumpable
214 %e executable filename (may be shortened)
216 %<OTHER> both are dropped
218 * If the first character of the pattern is a '|', the kernel will treat
219 the rest of the pattern as a command to run. The core dump will be
220 written to the standard input of that program instead of to a file.
226 This sysctl is only applicable when core_pattern is configured to pipe
227 core files to a user space helper (when the first character of
228 core_pattern is a '|', see above). When collecting cores via a pipe
229 to an application, it is occasionally useful for the collecting
230 application to gather data about the crashing process from its
231 /proc/pid directory. In order to do this safely, the kernel must wait
232 for the collecting process to exit, so as not to remove the crashing
233 processes proc files prematurely. This in turn creates the
234 possibility that a misbehaving userspace collecting process can block
235 the reaping of a crashed process simply by never exiting. This sysctl
236 defends against that. It defines how many concurrent crashing
237 processes may be piped to user space applications in parallel. If
238 this value is exceeded, then those crashing processes above that value
239 are noted via the kernel log and their cores are skipped. 0 is a
240 special value, indicating that unlimited processes may be captured in
241 parallel, but that no waiting will take place (i.e. the collecting
242 process is not guaranteed access to /proc/<crashing pid>/). This
249 The default coredump filename is "core". By setting
250 core_uses_pid to 1, the coredump filename becomes core.PID.
251 If core_pattern does not include "%p" (default does not)
252 and core_uses_pid is set, then .PID will be appended to
259 When the value in this file is 0, ctrl-alt-del is trapped and
260 sent to the init(1) program to handle a graceful restart.
261 When, however, the value is > 0, Linux's reaction to a Vulcan
262 Nerve Pinch (tm) will be an immediate reboot, without even
263 syncing its dirty buffers.
266 when a program (like dosemu) has the keyboard in 'raw'
267 mode, the ctrl-alt-del is intercepted by the program before it
268 ever reaches the kernel tty layer, and it's up to the program
269 to decide what to do with it.
275 This toggle indicates whether unprivileged users are prevented
276 from using dmesg(8) to view messages from the kernel's log buffer.
277 When dmesg_restrict is set to (0) there are no restrictions. When
278 dmesg_restrict is set set to (1), users must have CAP_SYSLOG to use
281 The kernel config option CONFIG_SECURITY_DMESG_RESTRICT sets the
282 default value of dmesg_restrict.
285 domainname & hostname:
286 ======================
288 These files can be used to set the NIS/YP domainname and the
289 hostname of your box in exactly the same way as the commands
290 domainname and hostname, i.e.::
292 # echo "darkstar" > /proc/sys/kernel/hostname
293 # echo "mydomain" > /proc/sys/kernel/domainname
295 has the same effect as::
297 # hostname "darkstar"
298 # domainname "mydomain"
300 Note, however, that the classic darkstar.frop.org has the
301 hostname "darkstar" and DNS (Internet Domain Name Server)
302 domainname "frop.org", not to be confused with the NIS (Network
303 Information Service) or YP (Yellow Pages) domainname. These two
304 domain names are in general different. For a detailed discussion
305 see the hostname(1) man page.
308 hardlockup_all_cpu_backtrace:
309 =============================
311 This value controls the hard lockup detector behavior when a hard
312 lockup condition is detected as to whether or not to gather further
313 debug information. If enabled, arch-specific all-CPU stack dumping
316 0: do nothing. This is the default behavior.
318 1: on detection capture more debug information.
324 This parameter can be used to control whether the kernel panics
325 when a hard lockup is detected.
327 0 - don't panic on hard lockup
328 1 - panic on hard lockup
330 See Documentation/lockup-watchdogs.txt for more information. This can
331 also be set using the nmi_watchdog kernel parameter.
337 Path for the hotplug policy agent.
338 Default value is "/sbin/hotplug".
344 Controls the kernel's behavior when a hung task is detected.
345 This file shows up if CONFIG_DETECT_HUNG_TASK is enabled.
347 0: continue operation. This is the default behavior.
349 1: panic immediately.
352 hung_task_check_count:
353 ======================
355 The upper bound on the number of tasks that are checked.
356 This file shows up if CONFIG_DETECT_HUNG_TASK is enabled.
359 hung_task_timeout_secs:
360 =======================
362 When a task in D state did not get scheduled
363 for more than this value report a warning.
364 This file shows up if CONFIG_DETECT_HUNG_TASK is enabled.
366 0: means infinite timeout - no checking done.
368 Possible values to set are in range {0..LONG_MAX/HZ}.
371 hung_task_check_interval_secs:
372 ==============================
374 Hung task check interval. If hung task checking is enabled
375 (see hung_task_timeout_secs), the check is done every
376 hung_task_check_interval_secs seconds.
377 This file shows up if CONFIG_DETECT_HUNG_TASK is enabled.
379 0 (default): means use hung_task_timeout_secs as checking interval.
380 Possible values to set are in range {0..LONG_MAX/HZ}.
386 The maximum number of warnings to report. During a check interval
387 if a hung task is detected, this value is decreased by 1.
388 When this value reaches 0, no more warnings will be reported.
389 This file shows up if CONFIG_DETECT_HUNG_TASK is enabled.
391 -1: report an infinite number of warnings.
394 hyperv_record_panic_msg:
395 ========================
397 Controls whether the panic kmsg data should be reported to Hyper-V.
399 0: do not report panic kmsg data.
401 1: report the panic kmsg data. This is the default behavior.
407 A toggle indicating if the kexec_load syscall has been disabled. This
408 value defaults to 0 (false: kexec_load enabled), but can be set to 1
409 (true: kexec_load disabled). Once true, kexec can no longer be used, and
410 the toggle cannot be set back to false. This allows a kexec image to be
411 loaded before disabling the syscall, allowing a system to set up (and
412 later use) an image without it being altered. Generally used together
413 with the "modules_disabled" sysctl.
419 This toggle indicates whether restrictions are placed on
420 exposing kernel addresses via /proc and other interfaces.
422 When kptr_restrict is set to 0 (the default) the address is hashed before
423 printing. (This is the equivalent to %p.)
425 When kptr_restrict is set to (1), kernel pointers printed using the %pK
426 format specifier will be replaced with 0's unless the user has CAP_SYSLOG
427 and effective user and group ids are equal to the real ids. This is
428 because %pK checks are done at read() time rather than open() time, so
429 if permissions are elevated between the open() and the read() (e.g via
430 a setuid binary) then %pK will not leak kernel pointers to unprivileged
431 users. Note, this is a temporary solution only. The correct long-term
432 solution is to do the permission checks at open() time. Consider removing
433 world read permissions from files that use %pK, and using dmesg_restrict
434 to protect against uses of %pK in dmesg(8) if leaking kernel pointer
435 values to unprivileged users is a concern.
437 When kptr_restrict is set to (2), kernel pointers printed using
438 %pK will be replaced with 0's regardless of privileges.
444 This flag controls the L2 cache of G3 processor boards. If
445 0, the cache is disabled. Enabled if nonzero.
451 A toggle value indicating if modules are allowed to be loaded
452 in an otherwise modular kernel. This toggle defaults to off
453 (0), but can be set true (1). Once true, modules can be
454 neither loaded nor unloaded, and the toggle cannot be set back
455 to false. Generally used with the "kexec_load_disabled" toggle.
458 msg_next_id, sem_next_id, and shm_next_id:
459 ==========================================
461 These three toggles allows to specify desired id for next allocated IPC
462 object: message, semaphore or shared memory respectively.
464 By default they are equal to -1, which means generic allocation logic.
465 Possible values to set are in range {0..INT_MAX}.
468 1) kernel doesn't guarantee, that new object will have desired id. So,
469 it's up to userspace, how to handle an object with "wrong" id.
470 2) Toggle with non-default value will be set back to -1 by kernel after
471 successful IPC object allocation. If an IPC object allocation syscall
472 fails, it is undefined if the value remains unmodified or is reset to -1.
478 This parameter can be used to control the NMI watchdog
479 (i.e. the hard lockup detector) on x86 systems.
481 0 - disable the hard lockup detector
483 1 - enable the hard lockup detector
485 The hard lockup detector monitors each CPU for its ability to respond to
486 timer interrupts. The mechanism utilizes CPU performance counter registers
487 that are programmed to generate Non-Maskable Interrupts (NMIs) periodically
488 while a CPU is busy. Hence, the alternative name 'NMI watchdog'.
490 The NMI watchdog is disabled by default if the kernel is running as a guest
491 in a KVM virtual machine. This default can be overridden by adding::
495 to the guest kernel command line (see Documentation/admin-guide/kernel-parameters.rst).
501 Enables/disables automatic page fault based NUMA memory
502 balancing. Memory is moved automatically to nodes
503 that access it often.
505 Enables/disables automatic NUMA memory balancing. On NUMA machines, there
506 is a performance penalty if remote memory is accessed by a CPU. When this
507 feature is enabled the kernel samples what task thread is accessing memory
508 by periodically unmapping pages and later trapping a page fault. At the
509 time of the page fault, it is determined if the data being accessed should
510 be migrated to a local memory node.
512 The unmapping of pages and trapping faults incur additional overhead that
513 ideally is offset by improved memory locality but there is no universal
514 guarantee. If the target workload is already bound to NUMA nodes then this
515 feature should be disabled. Otherwise, if the system overhead from the
516 feature is too high then the rate the kernel samples for NUMA hinting
517 faults may be controlled by the numa_balancing_scan_period_min_ms,
518 numa_balancing_scan_delay_ms, numa_balancing_scan_period_max_ms,
519 numa_balancing_scan_size_mb, and numa_balancing_settle_count sysctls.
521 numa_balancing_scan_period_min_ms, numa_balancing_scan_delay_ms, numa_balancing_scan_period_max_ms, numa_balancing_scan_size_mb
522 ===============================================================================================================================
525 Automatic NUMA balancing scans tasks address space and unmaps pages to
526 detect if pages are properly placed or if the data should be migrated to a
527 memory node local to where the task is running. Every "scan delay" the task
528 scans the next "scan size" number of pages in its address space. When the
529 end of the address space is reached the scanner restarts from the beginning.
531 In combination, the "scan delay" and "scan size" determine the scan rate.
532 When "scan delay" decreases, the scan rate increases. The scan delay and
533 hence the scan rate of every task is adaptive and depends on historical
534 behaviour. If pages are properly placed then the scan delay increases,
535 otherwise the scan delay decreases. The "scan size" is not adaptive but
536 the higher the "scan size", the higher the scan rate.
538 Higher scan rates incur higher system overhead as page faults must be
539 trapped and potentially data must be migrated. However, the higher the scan
540 rate, the more quickly a tasks memory is migrated to a local node if the
541 workload pattern changes and minimises performance impact due to remote
542 memory accesses. These sysctls control the thresholds for scan delays and
543 the number of pages scanned.
545 numa_balancing_scan_period_min_ms is the minimum time in milliseconds to
546 scan a tasks virtual memory. It effectively controls the maximum scanning
549 numa_balancing_scan_delay_ms is the starting "scan delay" used for a task
550 when it initially forks.
552 numa_balancing_scan_period_max_ms is the maximum time in milliseconds to
553 scan a tasks virtual memory. It effectively controls the minimum scanning
556 numa_balancing_scan_size_mb is how many megabytes worth of pages are
557 scanned for a given scan.
560 osrelease, ostype & version:
561 ============================
570 #5 Wed Feb 25 21:49:24 MET 1998
572 The files osrelease and ostype should be clear enough. Version
573 needs a little more clarification however. The '#5' means that
574 this is the fifth kernel built from this source base and the
575 date behind it indicates the time the kernel was built.
576 The only way to tune these values is to rebuild the kernel :-)
579 overflowgid & overflowuid:
580 ==========================
582 if your architecture did not always support 32-bit UIDs (i.e. arm,
583 i386, m68k, sh, and sparc32), a fixed UID and GID will be returned to
584 applications that use the old 16-bit UID/GID system calls, if the
585 actual UID or GID would exceed 65535.
587 These sysctls allow you to change the value of the fixed UID and GID.
588 The default is 65534.
594 The value in this file represents the number of seconds the kernel
595 waits before rebooting on a panic. When you use the software watchdog,
596 the recommended setting is 60.
602 Controls the kernel's behavior when a CPU receives an NMI caused by
605 0: try to continue operation (default)
607 1: panic immediately. The IO error triggered an NMI. This indicates a
608 serious system condition which could result in IO data corruption.
609 Rather than continuing, panicking might be a better choice. Some
610 servers issue this sort of NMI when the dump button is pushed,
611 and you can use this option to take a crash dump.
617 Controls the kernel's behaviour when an oops or BUG is encountered.
619 0: try to continue operation
621 1: panic immediately. If the `panic` sysctl is also non-zero then the
622 machine will be rebooted.
625 panic_on_stackoverflow:
626 =======================
628 Controls the kernel's behavior when detecting the overflows of
629 kernel, IRQ and exception stacks except a user stack.
630 This file shows up if CONFIG_DEBUG_STACKOVERFLOW is enabled.
632 0: try to continue operation.
634 1: panic immediately.
637 panic_on_unrecovered_nmi:
638 =========================
640 The default Linux behaviour on an NMI of either memory or unknown is
641 to continue operation. For many environments such as scientific
642 computing it is preferable that the box is taken out and the error
643 dealt with than an uncorrected parity/ECC error get propagated.
645 A small number of systems do generate NMI's for bizarre random reasons
646 such as power management so the default is off. That sysctl works like
647 the existing panic controls already in that directory.
653 Calls panic() in the WARN() path when set to 1. This is useful to avoid
654 a kernel rebuild when attempting to kdump at the location of a WARN().
656 0: only WARN(), default behaviour.
658 1: call panic() after printing out WARN() location.
664 Bitmask for printing system info when panic happens. User can chose
665 combination of the following bits:
667 ===== ========================================
668 bit 0 print all tasks info
669 bit 1 print system memory info
670 bit 2 print timer info
671 bit 3 print locks info if CONFIG_LOCKDEP is on
672 bit 4 print ftrace buffer
673 ===== ========================================
675 So for example to print tasks and memory info on panic, user can::
677 echo 3 > /proc/sys/kernel/panic_print
683 When set to 1, calls panic() after RCU stall detection messages. This
684 is useful to define the root cause of RCU stalls using a vmcore.
686 0: do not panic() when RCU stall takes place, default behavior.
688 1: panic() after printing RCU stall messages.
691 perf_cpu_time_max_percent:
692 ==========================
694 Hints to the kernel how much CPU time it should be allowed to
695 use to handle perf sampling events. If the perf subsystem
696 is informed that its samples are exceeding this limit, it
697 will drop its sampling frequency to attempt to reduce its CPU
700 Some perf sampling happens in NMIs. If these samples
701 unexpectedly take too long to execute, the NMIs can become
702 stacked up next to each other so much that nothing else is
706 disable the mechanism. Do not monitor or correct perf's
707 sampling rate no matter how CPU time it takes.
710 attempt to throttle perf's sample rate to this
711 percentage of CPU. Note: the kernel calculates an
712 "expected" length of each sample event. 100 here means
713 100% of that expected length. Even if this is set to
714 100, you may still see sample throttling if this
715 length is exceeded. Set to 0 if you truly do not care
716 how much CPU is consumed.
722 Controls use of the performance events system by unprivileged
723 users (without CAP_SYS_ADMIN). The default value is 2.
725 === ==================================================================
726 -1 Allow use of (almost) all events by all users
728 Ignore mlock limit after perf_event_mlock_kb without CAP_IPC_LOCK
730 >=0 Disallow ftrace function tracepoint by users without CAP_SYS_ADMIN
732 Disallow raw tracepoint access by users without CAP_SYS_ADMIN
734 >=1 Disallow CPU event access by users without CAP_SYS_ADMIN
736 >=2 Disallow kernel profiling by users without CAP_SYS_ADMIN
737 === ==================================================================
740 perf_event_max_stack:
741 =====================
743 Controls maximum number of stack frames to copy for (attr.sample_type &
744 PERF_SAMPLE_CALLCHAIN) configured events, for instance, when using
745 'perf record -g' or 'perf trace --call-graph fp'.
747 This can only be done when no events are in use that have callchains
748 enabled, otherwise writing to this file will return -EBUSY.
750 The default value is 127.
756 Control size of per-cpu ring buffer not counted agains mlock limit.
758 The default value is 512 + 1 page
761 perf_event_max_contexts_per_stack:
762 ==================================
764 Controls maximum number of stack frame context entries for
765 (attr.sample_type & PERF_SAMPLE_CALLCHAIN) configured events, for
766 instance, when using 'perf record -g' or 'perf trace --call-graph fp'.
768 This can only be done when no events are in use that have callchains
769 enabled, otherwise writing to this file will return -EBUSY.
771 The default value is 8.
777 PID allocation wrap value. When the kernel's next PID value
778 reaches this value, it wraps back to a minimum PID value.
779 PIDs of value pid_max or larger are not allocated.
785 The last pid allocated in the current (the one task using this sysctl
786 lives in) pid namespace. When selecting a pid for a next task on fork
787 kernel tries to allocate a number starting from this one.
790 powersave-nap: (PPC only)
791 =========================
793 If set, Linux-PPC will use the 'nap' mode of powersaving,
794 otherwise the 'doze' mode will be used.
796 ==============================================================
801 The four values in printk denote: console_loglevel,
802 default_message_loglevel, minimum_console_loglevel and
803 default_console_loglevel respectively.
805 These values influence printk() behavior when printing or
806 logging error messages. See 'man 2 syslog' for more info on
807 the different loglevels.
810 messages with a higher priority than
811 this will be printed to the console
812 - default_message_loglevel:
813 messages without an explicit priority
814 will be printed with this priority
815 - minimum_console_loglevel:
816 minimum (highest) value to which
817 console_loglevel can be set
818 - default_console_loglevel:
819 default value for console_loglevel
825 Delay each printk message in printk_delay milliseconds
827 Value from 0 - 10000 is allowed.
833 Some warning messages are rate limited. printk_ratelimit specifies
834 the minimum length of time between these messages (in jiffies), by
835 default we allow one every 5 seconds.
837 A value of 0 will disable rate limiting.
840 printk_ratelimit_burst:
841 =======================
843 While long term we enforce one message per printk_ratelimit
844 seconds, we do allow a burst of messages to pass through.
845 printk_ratelimit_burst specifies the number of messages we can
846 send before ratelimiting kicks in.
852 Control the logging to /dev/kmsg from userspace:
857 on: unlimited logging to /dev/kmsg from userspace
859 off: logging to /dev/kmsg disabled
861 The kernel command line parameter printk.devkmsg= overrides this and is
862 a one-time setting until next reboot: once set, it cannot be changed by
863 this sysctl interface anymore.
869 This option can be used to select the type of process address
870 space randomization that is used in the system, for architectures
871 that support this feature.
873 == ===========================================================================
874 0 Turn the process address space randomization off. This is the
875 default for architectures that do not support this feature anyways,
876 and kernels that are booted with the "norandmaps" parameter.
878 1 Make the addresses of mmap base, stack and VDSO page randomized.
879 This, among other things, implies that shared libraries will be
880 loaded to random addresses. Also for PIE-linked binaries, the
881 location of code start is randomized. This is the default if the
882 CONFIG_COMPAT_BRK option is enabled.
884 2 Additionally enable heap randomization. This is the default if
885 CONFIG_COMPAT_BRK is disabled.
887 There are a few legacy applications out there (such as some ancient
888 versions of libc.so.5 from 1996) that assume that brk area starts
889 just after the end of the code+bss. These applications break when
890 start of the brk area is randomized. There are however no known
891 non-legacy applications that would be broken this way, so for most
892 systems it is safe to choose full randomization.
894 Systems with ancient and/or broken binaries should be configured
895 with CONFIG_COMPAT_BRK enabled, which excludes the heap from process
896 address space randomization.
897 == ===========================================================================
900 reboot-cmd: (Sparc only)
901 ========================
903 ??? This seems to be a way to give an argument to the Sparc
904 ROM/Flash boot loader. Maybe to tell it what to do after
908 rtsig-max & rtsig-nr:
909 =====================
911 The file rtsig-max can be used to tune the maximum number
912 of POSIX realtime (queued) signals that can be outstanding
915 rtsig-nr shows the number of RT signals currently queued.
921 Enables/disables Energy Aware Scheduling (EAS). EAS starts
922 automatically on platforms where it can run (that is,
923 platforms with asymmetric CPU topologies and having an Energy
924 Model available). If your platform happens to meet the
925 requirements for EAS but you do not want to use it, change
932 Enables/disables scheduler statistics. Enabling this feature
933 incurs a small amount of overhead in the scheduler but is
934 useful for debugging and performance tuning.
940 This file shows the size of the generic SCSI (sg) buffer.
941 You can't tune it just yet, but you could change it on
942 compile time by editing include/scsi/sg.h and changing
943 the value of SG_BIG_BUFF.
945 There shouldn't be any reason to change this value. If
946 you can come up with one, you probably know what you
953 This parameter sets the total amount of shared memory pages that
954 can be used system wide. Hence, SHMALL should always be at least
955 ceil(shmmax/PAGE_SIZE).
957 If you are not sure what the default PAGE_SIZE is on your Linux
958 system, you can run the following command:
966 This value can be used to query and set the run time limit
967 on the maximum shared memory segment size that can be created.
968 Shared memory segments up to 1Gb are now supported in the
969 kernel. This value defaults to SHMMAX.
975 Linux lets you set resource limits, including how much memory one
976 process can consume, via setrlimit(2). Unfortunately, shared memory
977 segments are allowed to exist without association with any process, and
978 thus might not be counted against any resource limits. If enabled,
979 shared memory segments are automatically destroyed when their attach
980 count becomes zero after a detach or a process termination. It will
981 also destroy segments that were created, but never attached to, on exit
982 from the process. The only use left for IPC_RMID is to immediately
983 destroy an unattached segment. Of course, this breaks the way things are
984 defined, so some applications might stop working. Note that this
985 feature will do you no good unless you also configure your resource
986 limits (in particular, RLIMIT_AS and RLIMIT_NPROC). Most systems don't
989 Note that if you change this from 0 to 1, already created segments
990 without users and with a dead originative process will be destroyed.
993 sysctl_writes_strict:
994 =====================
996 Control how file position affects the behavior of updating sysctl values
997 via the /proc/sys interface:
999 == ======================================================================
1000 -1 Legacy per-write sysctl value handling, with no printk warnings.
1001 Each write syscall must fully contain the sysctl value to be
1002 written, and multiple writes on the same sysctl file descriptor
1003 will rewrite the sysctl value, regardless of file position.
1004 0 Same behavior as above, but warn about processes that perform writes
1005 to a sysctl file descriptor when the file position is not 0.
1006 1 (default) Respect file position when writing sysctl strings. Multiple
1007 writes will append to the sysctl value buffer. Anything past the max
1008 length of the sysctl value buffer will be ignored. Writes to numeric
1009 sysctl entries must always be at file position 0 and the value must
1010 be fully contained in the buffer sent in the write syscall.
1011 == ======================================================================
1014 softlockup_all_cpu_backtrace:
1015 =============================
1017 This value controls the soft lockup detector thread's behavior
1018 when a soft lockup condition is detected as to whether or not
1019 to gather further debug information. If enabled, each cpu will
1020 be issued an NMI and instructed to capture stack trace.
1022 This feature is only applicable for architectures which support
1025 0: do nothing. This is the default behavior.
1027 1: on detection capture more debug information.
1033 This parameter can be used to control the soft lockup detector.
1035 0 - disable the soft lockup detector
1037 1 - enable the soft lockup detector
1039 The soft lockup detector monitors CPUs for threads that are hogging the CPUs
1040 without rescheduling voluntarily, and thus prevent the 'watchdog/N' threads
1041 from running. The mechanism depends on the CPUs ability to respond to timer
1042 interrupts which are needed for the 'watchdog/N' threads to be woken up by
1043 the watchdog timer function, otherwise the NMI watchdog - if enabled - can
1044 detect a hard lockup condition.
1050 This parameter can be used to control kernel stack erasing at the end
1051 of syscalls for kernels built with CONFIG_GCC_PLUGIN_STACKLEAK.
1053 That erasing reduces the information which kernel stack leak bugs
1054 can reveal and blocks some uninitialized stack variable attacks.
1055 The tradeoff is the performance impact: on a single CPU system kernel
1056 compilation sees a 1% slowdown, other systems and workloads may vary.
1058 0: kernel stack erasing is disabled, STACKLEAK_METRICS are not updated.
1060 1: kernel stack erasing is enabled (default), it is performed before
1061 returning to the userspace at the end of syscalls.
1067 Non-zero if the kernel has been tainted. Numeric values, which can be
1068 ORed together. The letters are seen in "Tainted" line of Oops reports.
1070 ====== ===== ==============================================================
1071 1 `(P)` proprietary module was loaded
1072 2 `(F)` module was force loaded
1073 4 `(S)` SMP kernel oops on an officially SMP incapable processor
1074 8 `(R)` module was force unloaded
1075 16 `(M)` processor reported a Machine Check Exception (MCE)
1076 32 `(B)` bad page referenced or some unexpected page flags
1077 64 `(U)` taint requested by userspace application
1078 128 `(D)` kernel died recently, i.e. there was an OOPS or BUG
1079 256 `(A)` an ACPI table was overridden by user
1080 512 `(W)` kernel issued warning
1081 1024 `(C)` staging driver was loaded
1082 2048 `(I)` workaround for bug in platform firmware applied
1083 4096 `(O)` externally-built ("out-of-tree") module was loaded
1084 8192 `(E)` unsigned module was loaded
1085 16384 `(L)` soft lockup occurred
1086 32768 `(K)` kernel has been live patched
1087 65536 `(X)` Auxiliary taint, defined and used by for distros
1088 131072 `(T)` The kernel was built with the struct randomization plugin
1089 ====== ===== ==============================================================
1091 See Documentation/admin-guide/tainted-kernels.rst for more information.
1097 This value controls the maximum number of threads that can be created
1100 During initialization the kernel sets this value such that even if the
1101 maximum number of threads is created, the thread structures occupy only
1102 a part (1/8th) of the available RAM pages.
1104 The minimum value that can be written to threads-max is 20.
1106 The maximum value that can be written to threads-max is given by the
1107 constant FUTEX_TID_MASK (0x3fffffff).
1109 If a value outside of this range is written to threads-max an error
1112 The value written is checked against the available RAM pages. If the
1113 thread structures would occupy too much (more than 1/8th) of the
1114 available RAM pages threads-max is reduced accordingly.
1120 The value in this file affects behavior of handling NMI. When the
1121 value is non-zero, unknown NMI is trapped and then panic occurs. At
1122 that time, kernel debugging information is displayed on console.
1124 NMI switch that most IA32 servers have fires unknown NMI up, for
1125 example. If a system hangs up, try pressing the NMI switch.
1131 This parameter can be used to disable or enable the soft lockup detector
1132 _and_ the NMI watchdog (i.e. the hard lockup detector) at the same time.
1134 0 - disable both lockup detectors
1136 1 - enable both lockup detectors
1138 The soft lockup detector and the NMI watchdog can also be disabled or
1139 enabled individually, using the soft_watchdog and nmi_watchdog parameters.
1140 If the watchdog parameter is read, for example by executing::
1142 cat /proc/sys/kernel/watchdog
1144 the output of this command (0 or 1) shows the logical OR of soft_watchdog
1151 This value can be used to control on which cpus the watchdog may run.
1152 The default cpumask is all possible cores, but if NO_HZ_FULL is
1153 enabled in the kernel config, and cores are specified with the
1154 nohz_full= boot argument, those cores are excluded by default.
1155 Offline cores can be included in this mask, and if the core is later
1156 brought online, the watchdog will be started based on the mask value.
1158 Typically this value would only be touched in the nohz_full case
1159 to re-enable cores that by default were not running the watchdog,
1160 if a kernel lockup was suspected on those cores.
1162 The argument value is the standard cpulist format for cpumasks,
1163 so for example to enable the watchdog on cores 0, 2, 3, and 4 you
1166 echo 0,2-4 > /proc/sys/kernel/watchdog_cpumask
1172 This value can be used to control the frequency of hrtimer and NMI
1173 events and the soft and hard lockup thresholds. The default threshold
1176 The softlockup threshold is (2 * watchdog_thresh). Setting this
1177 tunable to zero will disable lockup detection altogether.