1 Documentation for /proc/sys/kernel/* kernel version 2.2.10
2 (c) 1998, 1999, Rik van Riel <riel@nl.linux.org>
3 (c) 2009, Shen Feng<shen@cn.fujitsu.com>
5 For general info and legal blurb, please look in README.
7 ==============================================================
9 This file contains documentation for the sysctl files in
10 /proc/sys/kernel/ and is valid for Linux kernel version 2.2.
12 The files in this directory can be used to tune and monitor
13 miscellaneous and general things in the operation of the Linux
14 kernel. Since some of the files _can_ be used to screw up your
15 system, it is advisable to read both documentation and source
16 before actually making adjustments.
18 Currently, these files might (depending on your configuration)
19 show up in /proc/sys/kernel:
24 - bootloader_type [ X86 only ]
25 - bootloader_version [ X86 only ]
26 - callhome [ S390 only ]
37 - hung_task_check_count
38 - hung_task_timeout_secs
42 - kstack_depth_to_print [ X86 only ]
44 - modprobe ==> Documentation/debugging-modules.txt
46 - msg_next_id [ sysv ipc ]
57 - panic_on_unrecovered_nmi
58 - panic_on_stackoverflow
60 - powersave-nap [ PPC only ]
64 - printk_ratelimit_burst
66 - real-root-dev ==> Documentation/initrd.txt
67 - reboot-cmd [ SPARC only ]
71 - sem_next_id [ sysv ipc ]
72 - sg-big-buff [ generic SCSI device (sg) ]
73 - shm_next_id [ sysv ipc ]
78 - stop-a [ SPARC only ]
79 - sysrq ==> Documentation/sysrq.txt
86 ==============================================================
90 highwater lowwater frequency
92 If BSD-style process accounting is enabled these values control
93 its behaviour. If free space on filesystem where the log lives
94 goes below <lowwater>% accounting suspends. If free space gets
95 above <highwater>% accounting resumes. <Frequency> determines
96 how often do we check the amount of free space (value is in
99 That is, suspend accounting if there left <= 2% free; resume it
100 if we got >=4%; consider information about amount of free space
101 valid for 30 seconds.
103 ==============================================================
109 See Doc*/kernel/power/video.txt, it allows mode of video boot to be
112 ==============================================================
116 Enables/Disables automatic recomputing of msgmni upon memory add/remove
117 or upon ipc namespace creation/removal (see the msgmni description
118 above). Echoing "1" into this file enables msgmni automatic recomputing.
119 Echoing "0" turns it off. auto_msgmni default value is 1.
122 ==============================================================
126 x86 bootloader identification
128 This gives the bootloader type number as indicated by the bootloader,
129 shifted left by 4, and OR'd with the low four bits of the bootloader
130 version. The reason for this encoding is that this used to match the
131 type_of_loader field in the kernel header; the encoding is kept for
132 backwards compatibility. That is, if the full bootloader type number
133 is 0x15 and the full version number is 0x234, this file will contain
134 the value 340 = 0x154.
136 See the type_of_loader and ext_loader_type fields in
137 Documentation/x86/boot.txt for additional information.
139 ==============================================================
143 x86 bootloader version
145 The complete bootloader version number. In the example above, this
146 file will contain the value 564 = 0x234.
148 See the type_of_loader and ext_loader_ver fields in
149 Documentation/x86/boot.txt for additional information.
151 ==============================================================
155 Controls the kernel's callhome behavior in case of a kernel panic.
157 The s390 hardware allows an operating system to send a notification
158 to a service organization (callhome) in case of an operating system panic.
160 When the value in this file is 0 (which is the default behavior)
161 nothing happens in case of a kernel panic. If this value is set to "1"
162 the complete kernel oops message is send to the IBM customer service
163 organization in case the mainframe the Linux operating system is running
164 on has a service contract with IBM.
166 ==============================================================
170 Highest valid capability of the running kernel. Exports
171 CAP_LAST_CAP from the kernel.
173 ==============================================================
177 core_pattern is used to specify a core dumpfile pattern name.
178 . max length 128 characters; default value is "core"
179 . core_pattern is used as a pattern template for the output filename;
180 certain string patterns (beginning with '%') are substituted with
182 . backward compatibility with core_uses_pid:
183 If core_pattern does not include "%p" (default does not)
184 and core_uses_pid is set, then .PID will be appended to
186 . corename format specifiers:
187 %<NUL> '%' is dropped
190 %P global pid (init PID namespace)
193 %d dump mode, matches PR_SET_DUMPABLE and
194 /proc/sys/fs/suid_dumpable
198 %e executable filename (may be shortened)
200 %<OTHER> both are dropped
201 . If the first character of the pattern is a '|', the kernel will treat
202 the rest of the pattern as a command to run. The core dump will be
203 written to the standard input of that program instead of to a file.
205 ==============================================================
209 This sysctl is only applicable when core_pattern is configured to pipe
210 core files to a user space helper (when the first character of
211 core_pattern is a '|', see above). When collecting cores via a pipe
212 to an application, it is occasionally useful for the collecting
213 application to gather data about the crashing process from its
214 /proc/pid directory. In order to do this safely, the kernel must wait
215 for the collecting process to exit, so as not to remove the crashing
216 processes proc files prematurely. This in turn creates the
217 possibility that a misbehaving userspace collecting process can block
218 the reaping of a crashed process simply by never exiting. This sysctl
219 defends against that. It defines how many concurrent crashing
220 processes may be piped to user space applications in parallel. If
221 this value is exceeded, then those crashing processes above that value
222 are noted via the kernel log and their cores are skipped. 0 is a
223 special value, indicating that unlimited processes may be captured in
224 parallel, but that no waiting will take place (i.e. the collecting
225 process is not guaranteed access to /proc/<crashing pid>/). This
228 ==============================================================
232 The default coredump filename is "core". By setting
233 core_uses_pid to 1, the coredump filename becomes core.PID.
234 If core_pattern does not include "%p" (default does not)
235 and core_uses_pid is set, then .PID will be appended to
238 ==============================================================
242 When the value in this file is 0, ctrl-alt-del is trapped and
243 sent to the init(1) program to handle a graceful restart.
244 When, however, the value is > 0, Linux's reaction to a Vulcan
245 Nerve Pinch (tm) will be an immediate reboot, without even
246 syncing its dirty buffers.
248 Note: when a program (like dosemu) has the keyboard in 'raw'
249 mode, the ctrl-alt-del is intercepted by the program before it
250 ever reaches the kernel tty layer, and it's up to the program
251 to decide what to do with it.
253 ==============================================================
257 This toggle indicates whether unprivileged users are prevented
258 from using dmesg(8) to view messages from the kernel's log buffer.
259 When dmesg_restrict is set to (0) there are no restrictions. When
260 dmesg_restrict is set set to (1), users must have CAP_SYSLOG to use
263 The kernel config option CONFIG_SECURITY_DMESG_RESTRICT sets the
264 default value of dmesg_restrict.
266 ==============================================================
268 domainname & hostname:
270 These files can be used to set the NIS/YP domainname and the
271 hostname of your box in exactly the same way as the commands
272 domainname and hostname, i.e.:
273 # echo "darkstar" > /proc/sys/kernel/hostname
274 # echo "mydomain" > /proc/sys/kernel/domainname
275 has the same effect as
276 # hostname "darkstar"
277 # domainname "mydomain"
279 Note, however, that the classic darkstar.frop.org has the
280 hostname "darkstar" and DNS (Internet Domain Name Server)
281 domainname "frop.org", not to be confused with the NIS (Network
282 Information Service) or YP (Yellow Pages) domainname. These two
283 domain names are in general different. For a detailed discussion
284 see the hostname(1) man page.
286 ==============================================================
290 Path for the hotplug policy agent.
291 Default value is "/sbin/hotplug".
293 ==============================================================
297 Controls the kernel's behavior when a hung task is detected.
298 This file shows up if CONFIG_DETECT_HUNG_TASK is enabled.
300 0: continue operation. This is the default behavior.
302 1: panic immediately.
304 ==============================================================
306 hung_task_check_count:
308 The upper bound on the number of tasks that are checked.
309 This file shows up if CONFIG_DETECT_HUNG_TASK is enabled.
311 ==============================================================
313 hung_task_timeout_secs:
315 Check interval. When a task in D state did not get scheduled
316 for more than this value report a warning.
317 This file shows up if CONFIG_DETECT_HUNG_TASK is enabled.
319 0: means infinite timeout - no checking done.
320 Possible values to set are in range {0..LONG_MAX/HZ}.
322 ==============================================================
326 The maximum number of warnings to report. During a check interval
327 When this value is reached, no more the warnings will be reported.
328 This file shows up if CONFIG_DETECT_HUNG_TASK is enabled.
330 -1: report an infinite number of warnings.
332 ==============================================================
336 A toggle indicating if the kexec_load syscall has been disabled. This
337 value defaults to 0 (false: kexec_load enabled), but can be set to 1
338 (true: kexec_load disabled). Once true, kexec can no longer be used, and
339 the toggle cannot be set back to false. This allows a kexec image to be
340 loaded before disabling the syscall, allowing a system to set up (and
341 later use) an image without it being altered. Generally used together
342 with the "modules_disabled" sysctl.
344 ==============================================================
348 This toggle indicates whether restrictions are placed on
349 exposing kernel addresses via /proc and other interfaces.
351 When kptr_restrict is set to (0), the default, there are no restrictions.
353 When kptr_restrict is set to (1), kernel pointers printed using the %pK
354 format specifier will be replaced with 0's unless the user has CAP_SYSLOG
355 and effective user and group ids are equal to the real ids. This is
356 because %pK checks are done at read() time rather than open() time, so
357 if permissions are elevated between the open() and the read() (e.g via
358 a setuid binary) then %pK will not leak kernel pointers to unprivileged
359 users. Note, this is a temporary solution only. The correct long-term
360 solution is to do the permission checks at open() time. Consider removing
361 world read permissions from files that use %pK, and using dmesg_restrict
362 to protect against uses of %pK in dmesg(8) if leaking kernel pointer
363 values to unprivileged users is a concern.
365 When kptr_restrict is set to (2), kernel pointers printed using
366 %pK will be replaced with 0's regardless of privileges.
368 ==============================================================
370 kstack_depth_to_print: (X86 only)
372 Controls the number of words to print when dumping the raw
375 ==============================================================
379 This flag controls the L2 cache of G3 processor boards. If
380 0, the cache is disabled. Enabled if nonzero.
382 ==============================================================
386 A toggle value indicating if modules are allowed to be loaded
387 in an otherwise modular kernel. This toggle defaults to off
388 (0), but can be set true (1). Once true, modules can be
389 neither loaded nor unloaded, and the toggle cannot be set back
390 to false. Generally used with the "kexec_load_disabled" toggle.
392 ==============================================================
394 msg_next_id, sem_next_id, and shm_next_id:
396 These three toggles allows to specify desired id for next allocated IPC
397 object: message, semaphore or shared memory respectively.
399 By default they are equal to -1, which means generic allocation logic.
400 Possible values to set are in range {0..INT_MAX}.
403 1) kernel doesn't guarantee, that new object will have desired id. So,
404 it's up to userspace, how to handle an object with "wrong" id.
405 2) Toggle with non-default value will be set back to -1 by kernel after
406 successful IPC object allocation.
408 ==============================================================
412 Enables/Disables the NMI watchdog on x86 systems. When the value is
413 non-zero the NMI watchdog is enabled and will continuously test all
414 online cpus to determine whether or not they are still functioning
415 properly. Currently, passing "nmi_watchdog=" parameter at boot time is
416 required for this function to work.
418 If LAPIC NMI watchdog method is in use (nmi_watchdog=2 kernel
419 parameter), the NMI watchdog shares registers with oprofile. By
420 disabling the NMI watchdog, oprofile may have more registers to
423 ==============================================================
427 Enables/disables automatic page fault based NUMA memory
428 balancing. Memory is moved automatically to nodes
429 that access it often.
431 Enables/disables automatic NUMA memory balancing. On NUMA machines, there
432 is a performance penalty if remote memory is accessed by a CPU. When this
433 feature is enabled the kernel samples what task thread is accessing memory
434 by periodically unmapping pages and later trapping a page fault. At the
435 time of the page fault, it is determined if the data being accessed should
436 be migrated to a local memory node.
438 The unmapping of pages and trapping faults incur additional overhead that
439 ideally is offset by improved memory locality but there is no universal
440 guarantee. If the target workload is already bound to NUMA nodes then this
441 feature should be disabled. Otherwise, if the system overhead from the
442 feature is too high then the rate the kernel samples for NUMA hinting
443 faults may be controlled by the numa_balancing_scan_period_min_ms,
444 numa_balancing_scan_delay_ms, numa_balancing_scan_period_max_ms,
445 numa_balancing_scan_size_mb, numa_balancing_settle_count sysctls and
446 numa_balancing_migrate_deferred.
448 ==============================================================
450 numa_balancing_scan_period_min_ms, numa_balancing_scan_delay_ms,
451 numa_balancing_scan_period_max_ms, numa_balancing_scan_size_mb
453 Automatic NUMA balancing scans tasks address space and unmaps pages to
454 detect if pages are properly placed or if the data should be migrated to a
455 memory node local to where the task is running. Every "scan delay" the task
456 scans the next "scan size" number of pages in its address space. When the
457 end of the address space is reached the scanner restarts from the beginning.
459 In combination, the "scan delay" and "scan size" determine the scan rate.
460 When "scan delay" decreases, the scan rate increases. The scan delay and
461 hence the scan rate of every task is adaptive and depends on historical
462 behaviour. If pages are properly placed then the scan delay increases,
463 otherwise the scan delay decreases. The "scan size" is not adaptive but
464 the higher the "scan size", the higher the scan rate.
466 Higher scan rates incur higher system overhead as page faults must be
467 trapped and potentially data must be migrated. However, the higher the scan
468 rate, the more quickly a tasks memory is migrated to a local node if the
469 workload pattern changes and minimises performance impact due to remote
470 memory accesses. These sysctls control the thresholds for scan delays and
471 the number of pages scanned.
473 numa_balancing_scan_period_min_ms is the minimum time in milliseconds to
474 scan a tasks virtual memory. It effectively controls the maximum scanning
477 numa_balancing_scan_delay_ms is the starting "scan delay" used for a task
478 when it initially forks.
480 numa_balancing_scan_period_max_ms is the maximum time in milliseconds to
481 scan a tasks virtual memory. It effectively controls the minimum scanning
484 numa_balancing_scan_size_mb is how many megabytes worth of pages are
485 scanned for a given scan.
487 numa_balancing_migrate_deferred is how many page migrations get skipped
488 unconditionally, after a page migration is skipped because a page is shared
489 with other tasks. This reduces page migration overhead, and determines
490 how much stronger the "move task near its memory" policy scheduler becomes,
491 versus the "move memory near its task" memory management policy, for workloads
494 ==============================================================
496 osrelease, ostype & version:
503 #5 Wed Feb 25 21:49:24 MET 1998
505 The files osrelease and ostype should be clear enough. Version
506 needs a little more clarification however. The '#5' means that
507 this is the fifth kernel built from this source base and the
508 date behind it indicates the time the kernel was built.
509 The only way to tune these values is to rebuild the kernel :-)
511 ==============================================================
513 overflowgid & overflowuid:
515 if your architecture did not always support 32-bit UIDs (i.e. arm,
516 i386, m68k, sh, and sparc32), a fixed UID and GID will be returned to
517 applications that use the old 16-bit UID/GID system calls, if the
518 actual UID or GID would exceed 65535.
520 These sysctls allow you to change the value of the fixed UID and GID.
521 The default is 65534.
523 ==============================================================
527 The value in this file represents the number of seconds the kernel
528 waits before rebooting on a panic. When you use the software watchdog,
529 the recommended setting is 60.
531 ==============================================================
533 panic_on_unrecovered_nmi:
535 The default Linux behaviour on an NMI of either memory or unknown is
536 to continue operation. For many environments such as scientific
537 computing it is preferable that the box is taken out and the error
538 dealt with than an uncorrected parity/ECC error get propagated.
540 A small number of systems do generate NMI's for bizarre random reasons
541 such as power management so the default is off. That sysctl works like
542 the existing panic controls already in that directory.
544 ==============================================================
548 Controls the kernel's behaviour when an oops or BUG is encountered.
550 0: try to continue operation
552 1: panic immediately. If the `panic' sysctl is also non-zero then the
553 machine will be rebooted.
555 ==============================================================
557 panic_on_stackoverflow:
559 Controls the kernel's behavior when detecting the overflows of
560 kernel, IRQ and exception stacks except a user stack.
561 This file shows up if CONFIG_DEBUG_STACKOVERFLOW is enabled.
563 0: try to continue operation.
565 1: panic immediately.
567 ==============================================================
569 perf_cpu_time_max_percent:
571 Hints to the kernel how much CPU time it should be allowed to
572 use to handle perf sampling events. If the perf subsystem
573 is informed that its samples are exceeding this limit, it
574 will drop its sampling frequency to attempt to reduce its CPU
577 Some perf sampling happens in NMIs. If these samples
578 unexpectedly take too long to execute, the NMIs can become
579 stacked up next to each other so much that nothing else is
582 0: disable the mechanism. Do not monitor or correct perf's
583 sampling rate no matter how CPU time it takes.
585 1-100: attempt to throttle perf's sample rate to this
586 percentage of CPU. Note: the kernel calculates an
587 "expected" length of each sample event. 100 here means
588 100% of that expected length. Even if this is set to
589 100, you may still see sample throttling if this
590 length is exceeded. Set to 0 if you truly do not care
591 how much CPU is consumed.
593 ==============================================================
598 PID allocation wrap value. When the kernel's next PID value
599 reaches this value, it wraps back to a minimum PID value.
600 PIDs of value pid_max or larger are not allocated.
602 ==============================================================
606 The last pid allocated in the current (the one task using this sysctl
607 lives in) pid namespace. When selecting a pid for a next task on fork
608 kernel tries to allocate a number starting from this one.
610 ==============================================================
612 powersave-nap: (PPC only)
614 If set, Linux-PPC will use the 'nap' mode of powersaving,
615 otherwise the 'doze' mode will be used.
617 ==============================================================
621 The four values in printk denote: console_loglevel,
622 default_message_loglevel, minimum_console_loglevel and
623 default_console_loglevel respectively.
625 These values influence printk() behavior when printing or
626 logging error messages. See 'man 2 syslog' for more info on
627 the different loglevels.
629 - console_loglevel: messages with a higher priority than
630 this will be printed to the console
631 - default_message_loglevel: messages without an explicit priority
632 will be printed with this priority
633 - minimum_console_loglevel: minimum (highest) value to which
634 console_loglevel can be set
635 - default_console_loglevel: default value for console_loglevel
637 ==============================================================
641 Delay each printk message in printk_delay milliseconds
643 Value from 0 - 10000 is allowed.
645 ==============================================================
649 Some warning messages are rate limited. printk_ratelimit specifies
650 the minimum length of time between these messages (in jiffies), by
651 default we allow one every 5 seconds.
653 A value of 0 will disable rate limiting.
655 ==============================================================
657 printk_ratelimit_burst:
659 While long term we enforce one message per printk_ratelimit
660 seconds, we do allow a burst of messages to pass through.
661 printk_ratelimit_burst specifies the number of messages we can
662 send before ratelimiting kicks in.
664 ==============================================================
668 This option can be used to select the type of process address
669 space randomization that is used in the system, for architectures
670 that support this feature.
672 0 - Turn the process address space randomization off. This is the
673 default for architectures that do not support this feature anyways,
674 and kernels that are booted with the "norandmaps" parameter.
676 1 - Make the addresses of mmap base, stack and VDSO page randomized.
677 This, among other things, implies that shared libraries will be
678 loaded to random addresses. Also for PIE-linked binaries, the
679 location of code start is randomized. This is the default if the
680 CONFIG_COMPAT_BRK option is enabled.
682 2 - Additionally enable heap randomization. This is the default if
683 CONFIG_COMPAT_BRK is disabled.
685 There are a few legacy applications out there (such as some ancient
686 versions of libc.so.5 from 1996) that assume that brk area starts
687 just after the end of the code+bss. These applications break when
688 start of the brk area is randomized. There are however no known
689 non-legacy applications that would be broken this way, so for most
690 systems it is safe to choose full randomization.
692 Systems with ancient and/or broken binaries should be configured
693 with CONFIG_COMPAT_BRK enabled, which excludes the heap from process
694 address space randomization.
696 ==============================================================
698 reboot-cmd: (Sparc only)
700 ??? This seems to be a way to give an argument to the Sparc
701 ROM/Flash boot loader. Maybe to tell it what to do after
704 ==============================================================
706 rtsig-max & rtsig-nr:
708 The file rtsig-max can be used to tune the maximum number
709 of POSIX realtime (queued) signals that can be outstanding
712 rtsig-nr shows the number of RT signals currently queued.
714 ==============================================================
718 This file shows the size of the generic SCSI (sg) buffer.
719 You can't tune it just yet, but you could change it on
720 compile time by editing include/scsi/sg.h and changing
721 the value of SG_BIG_BUFF.
723 There shouldn't be any reason to change this value. If
724 you can come up with one, you probably know what you
727 ==============================================================
731 This parameter sets the total amount of shared memory pages that
732 can be used system wide. Hence, SHMALL should always be at least
733 ceil(shmmax/PAGE_SIZE).
735 If you are not sure what the default PAGE_SIZE is on your Linux
736 system, you can run the following command:
740 ==============================================================
744 This value can be used to query and set the run time limit
745 on the maximum shared memory segment size that can be created.
746 Shared memory segments up to 1Gb are now supported in the
747 kernel. This value defaults to SHMMAX.
749 ==============================================================
753 Linux lets you set resource limits, including how much memory one
754 process can consume, via setrlimit(2). Unfortunately, shared memory
755 segments are allowed to exist without association with any process, and
756 thus might not be counted against any resource limits. If enabled,
757 shared memory segments are automatically destroyed when their attach
758 count becomes zero after a detach or a process termination. It will
759 also destroy segments that were created, but never attached to, on exit
760 from the process. The only use left for IPC_RMID is to immediately
761 destroy an unattached segment. Of course, this breaks the way things are
762 defined, so some applications might stop working. Note that this
763 feature will do you no good unless you also configure your resource
764 limits (in particular, RLIMIT_AS and RLIMIT_NPROC). Most systems don't
767 Note that if you change this from 0 to 1, already created segments
768 without users and with a dead originative process will be destroyed.
770 ==============================================================
774 Non-zero if the kernel has been tainted. Numeric values, which
775 can be ORed together:
777 1 - A module with a non-GPL license has been loaded, this
778 includes modules with no license.
779 Set by modutils >= 2.4.9 and module-init-tools.
780 2 - A module was force loaded by insmod -f.
781 Set by modutils >= 2.4.9 and module-init-tools.
782 4 - Unsafe SMP processors: SMP with CPUs not designed for SMP.
783 8 - A module was forcibly unloaded from the system by rmmod -f.
784 16 - A hardware machine check error occurred on the system.
785 32 - A bad page was discovered on the system.
786 64 - The user has asked that the system be marked "tainted". This
787 could be because they are running software that directly modifies
788 the hardware, or for other reasons.
789 128 - The system has died.
790 256 - The ACPI DSDT has been overridden with one supplied by the user
791 instead of using the one provided by the hardware.
792 512 - A kernel warning has occurred.
793 1024 - A module from drivers/staging was loaded.
794 2048 - The system is working around a severe firmware bug.
795 4096 - An out-of-tree module has been loaded.
797 ==============================================================
801 The value in this file affects behavior of handling NMI. When the
802 value is non-zero, unknown NMI is trapped and then panic occurs. At
803 that time, kernel debugging information is displayed on console.
805 NMI switch that most IA32 servers have fires unknown NMI up, for
806 example. If a system hangs up, try pressing the NMI switch.
808 ==============================================================
812 This value can be used to control the frequency of hrtimer and NMI
813 events and the soft and hard lockup thresholds. The default threshold
816 The softlockup threshold is (2 * watchdog_thresh). Setting this
817 tunable to zero will disable lockup detection altogether.
819 ==============================================================