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 ]
38 - kstack_depth_to_print [ X86 only ]
40 - modprobe ==> Documentation/debugging-modules.txt
42 - msg_next_id [ sysv ipc ]
53 - panic_on_unrecovered_nmi
54 - panic_on_stackoverflow
56 - powersave-nap [ PPC only ]
60 - printk_ratelimit_burst
62 - real-root-dev ==> Documentation/initrd.txt
63 - reboot-cmd [ SPARC only ]
67 - sem_next_id [ sysv ipc ]
68 - sg-big-buff [ generic SCSI device (sg) ]
69 - shm_next_id [ sysv ipc ]
74 - stop-a [ SPARC only ]
75 - sysrq ==> Documentation/sysrq.txt
82 ==============================================================
86 highwater lowwater frequency
88 If BSD-style process accounting is enabled these values control
89 its behaviour. If free space on filesystem where the log lives
90 goes below <lowwater>% accounting suspends. If free space gets
91 above <highwater>% accounting resumes. <Frequency> determines
92 how often do we check the amount of free space (value is in
95 That is, suspend accounting if there left <= 2% free; resume it
96 if we got >=4%; consider information about amount of free space
99 ==============================================================
105 See Doc*/kernel/power/video.txt, it allows mode of video boot to be
108 ==============================================================
112 Enables/Disables automatic recomputing of msgmni upon memory add/remove
113 or upon ipc namespace creation/removal (see the msgmni description
114 above). Echoing "1" into this file enables msgmni automatic recomputing.
115 Echoing "0" turns it off. auto_msgmni default value is 1.
118 ==============================================================
122 x86 bootloader identification
124 This gives the bootloader type number as indicated by the bootloader,
125 shifted left by 4, and OR'd with the low four bits of the bootloader
126 version. The reason for this encoding is that this used to match the
127 type_of_loader field in the kernel header; the encoding is kept for
128 backwards compatibility. That is, if the full bootloader type number
129 is 0x15 and the full version number is 0x234, this file will contain
130 the value 340 = 0x154.
132 See the type_of_loader and ext_loader_type fields in
133 Documentation/x86/boot.txt for additional information.
135 ==============================================================
139 x86 bootloader version
141 The complete bootloader version number. In the example above, this
142 file will contain the value 564 = 0x234.
144 See the type_of_loader and ext_loader_ver fields in
145 Documentation/x86/boot.txt for additional information.
147 ==============================================================
151 Controls the kernel's callhome behavior in case of a kernel panic.
153 The s390 hardware allows an operating system to send a notification
154 to a service organization (callhome) in case of an operating system panic.
156 When the value in this file is 0 (which is the default behavior)
157 nothing happens in case of a kernel panic. If this value is set to "1"
158 the complete kernel oops message is send to the IBM customer service
159 organization in case the mainframe the Linux operating system is running
160 on has a service contract with IBM.
162 ==============================================================
166 Highest valid capability of the running kernel. Exports
167 CAP_LAST_CAP from the kernel.
169 ==============================================================
173 core_pattern is used to specify a core dumpfile pattern name.
174 . max length 128 characters; default value is "core"
175 . core_pattern is used as a pattern template for the output filename;
176 certain string patterns (beginning with '%') are substituted with
178 . backward compatibility with core_uses_pid:
179 If core_pattern does not include "%p" (default does not)
180 and core_uses_pid is set, then .PID will be appended to
182 . corename format specifiers:
183 %<NUL> '%' is dropped
186 %P global pid (init PID namespace)
189 %d dump mode, matches PR_SET_DUMPABLE and
190 /proc/sys/fs/suid_dumpable
194 %e executable filename (may be shortened)
196 %<OTHER> both are dropped
197 . If the first character of the pattern is a '|', the kernel will treat
198 the rest of the pattern as a command to run. The core dump will be
199 written to the standard input of that program instead of to a file.
201 ==============================================================
205 This sysctl is only applicable when core_pattern is configured to pipe
206 core files to a user space helper (when the first character of
207 core_pattern is a '|', see above). When collecting cores via a pipe
208 to an application, it is occasionally useful for the collecting
209 application to gather data about the crashing process from its
210 /proc/pid directory. In order to do this safely, the kernel must wait
211 for the collecting process to exit, so as not to remove the crashing
212 processes proc files prematurely. This in turn creates the
213 possibility that a misbehaving userspace collecting process can block
214 the reaping of a crashed process simply by never exiting. This sysctl
215 defends against that. It defines how many concurrent crashing
216 processes may be piped to user space applications in parallel. If
217 this value is exceeded, then those crashing processes above that value
218 are noted via the kernel log and their cores are skipped. 0 is a
219 special value, indicating that unlimited processes may be captured in
220 parallel, but that no waiting will take place (i.e. the collecting
221 process is not guaranteed access to /proc/<crashing pid>/). This
224 ==============================================================
228 The default coredump filename is "core". By setting
229 core_uses_pid to 1, the coredump filename becomes core.PID.
230 If core_pattern does not include "%p" (default does not)
231 and core_uses_pid is set, then .PID will be appended to
234 ==============================================================
238 When the value in this file is 0, ctrl-alt-del is trapped and
239 sent to the init(1) program to handle a graceful restart.
240 When, however, the value is > 0, Linux's reaction to a Vulcan
241 Nerve Pinch (tm) will be an immediate reboot, without even
242 syncing its dirty buffers.
244 Note: when a program (like dosemu) has the keyboard in 'raw'
245 mode, the ctrl-alt-del is intercepted by the program before it
246 ever reaches the kernel tty layer, and it's up to the program
247 to decide what to do with it.
249 ==============================================================
253 This toggle indicates whether unprivileged users are prevented
254 from using dmesg(8) to view messages from the kernel's log buffer.
255 When dmesg_restrict is set to (0) there are no restrictions. When
256 dmesg_restrict is set set to (1), users must have CAP_SYSLOG to use
259 The kernel config option CONFIG_SECURITY_DMESG_RESTRICT sets the
260 default value of dmesg_restrict.
262 ==============================================================
264 domainname & hostname:
266 These files can be used to set the NIS/YP domainname and the
267 hostname of your box in exactly the same way as the commands
268 domainname and hostname, i.e.:
269 # echo "darkstar" > /proc/sys/kernel/hostname
270 # echo "mydomain" > /proc/sys/kernel/domainname
271 has the same effect as
272 # hostname "darkstar"
273 # domainname "mydomain"
275 Note, however, that the classic darkstar.frop.org has the
276 hostname "darkstar" and DNS (Internet Domain Name Server)
277 domainname "frop.org", not to be confused with the NIS (Network
278 Information Service) or YP (Yellow Pages) domainname. These two
279 domain names are in general different. For a detailed discussion
280 see the hostname(1) man page.
282 ==============================================================
286 Path for the hotplug policy agent.
287 Default value is "/sbin/hotplug".
289 ==============================================================
293 A toggle indicating if the kexec_load syscall has been disabled. This
294 value defaults to 0 (false: kexec_load enabled), but can be set to 1
295 (true: kexec_load disabled). Once true, kexec can no longer be used, and
296 the toggle cannot be set back to false. This allows a kexec image to be
297 loaded before disabling the syscall, allowing a system to set up (and
298 later use) an image without it being altered. Generally used together
299 with the "modules_disabled" sysctl.
301 ==============================================================
305 This toggle indicates whether restrictions are placed on
306 exposing kernel addresses via /proc and other interfaces.
308 When kptr_restrict is set to (0), the default, there are no restrictions.
310 When kptr_restrict is set to (1), kernel pointers printed using the %pK
311 format specifier will be replaced with 0's unless the user has CAP_SYSLOG
312 and effective user and group ids are equal to the real ids. This is
313 because %pK checks are done at read() time rather than open() time, so
314 if permissions are elevated between the open() and the read() (e.g via
315 a setuid binary) then %pK will not leak kernel pointers to unprivileged
316 users. Note, this is a temporary solution only. The correct long-term
317 solution is to do the permission checks at open() time. Consider removing
318 world read permissions from files that use %pK, and using dmesg_restrict
319 to protect against uses of %pK in dmesg(8) if leaking kernel pointer
320 values to unprivileged users is a concern.
322 When kptr_restrict is set to (2), kernel pointers printed using
323 %pK will be replaced with 0's regardless of privileges.
325 ==============================================================
327 kstack_depth_to_print: (X86 only)
329 Controls the number of words to print when dumping the raw
332 ==============================================================
336 This flag controls the L2 cache of G3 processor boards. If
337 0, the cache is disabled. Enabled if nonzero.
339 ==============================================================
343 A toggle value indicating if modules are allowed to be loaded
344 in an otherwise modular kernel. This toggle defaults to off
345 (0), but can be set true (1). Once true, modules can be
346 neither loaded nor unloaded, and the toggle cannot be set back
347 to false. Generally used with the "kexec_load_disabled" toggle.
349 ==============================================================
351 msg_next_id, sem_next_id, and shm_next_id:
353 These three toggles allows to specify desired id for next allocated IPC
354 object: message, semaphore or shared memory respectively.
356 By default they are equal to -1, which means generic allocation logic.
357 Possible values to set are in range {0..INT_MAX}.
360 1) kernel doesn't guarantee, that new object will have desired id. So,
361 it's up to userspace, how to handle an object with "wrong" id.
362 2) Toggle with non-default value will be set back to -1 by kernel after
363 successful IPC object allocation.
365 ==============================================================
369 Enables/Disables the NMI watchdog on x86 systems. When the value is
370 non-zero the NMI watchdog is enabled and will continuously test all
371 online cpus to determine whether or not they are still functioning
372 properly. Currently, passing "nmi_watchdog=" parameter at boot time is
373 required for this function to work.
375 If LAPIC NMI watchdog method is in use (nmi_watchdog=2 kernel
376 parameter), the NMI watchdog shares registers with oprofile. By
377 disabling the NMI watchdog, oprofile may have more registers to
380 ==============================================================
384 Enables/disables automatic page fault based NUMA memory
385 balancing. Memory is moved automatically to nodes
386 that access it often.
388 Enables/disables automatic NUMA memory balancing. On NUMA machines, there
389 is a performance penalty if remote memory is accessed by a CPU. When this
390 feature is enabled the kernel samples what task thread is accessing memory
391 by periodically unmapping pages and later trapping a page fault. At the
392 time of the page fault, it is determined if the data being accessed should
393 be migrated to a local memory node.
395 The unmapping of pages and trapping faults incur additional overhead that
396 ideally is offset by improved memory locality but there is no universal
397 guarantee. If the target workload is already bound to NUMA nodes then this
398 feature should be disabled. Otherwise, if the system overhead from the
399 feature is too high then the rate the kernel samples for NUMA hinting
400 faults may be controlled by the numa_balancing_scan_period_min_ms,
401 numa_balancing_scan_delay_ms, numa_balancing_scan_period_max_ms,
402 numa_balancing_scan_size_mb, numa_balancing_settle_count sysctls and
403 numa_balancing_migrate_deferred.
405 ==============================================================
407 numa_balancing_scan_period_min_ms, numa_balancing_scan_delay_ms,
408 numa_balancing_scan_period_max_ms, numa_balancing_scan_size_mb
410 Automatic NUMA balancing scans tasks address space and unmaps pages to
411 detect if pages are properly placed or if the data should be migrated to a
412 memory node local to where the task is running. Every "scan delay" the task
413 scans the next "scan size" number of pages in its address space. When the
414 end of the address space is reached the scanner restarts from the beginning.
416 In combination, the "scan delay" and "scan size" determine the scan rate.
417 When "scan delay" decreases, the scan rate increases. The scan delay and
418 hence the scan rate of every task is adaptive and depends on historical
419 behaviour. If pages are properly placed then the scan delay increases,
420 otherwise the scan delay decreases. The "scan size" is not adaptive but
421 the higher the "scan size", the higher the scan rate.
423 Higher scan rates incur higher system overhead as page faults must be
424 trapped and potentially data must be migrated. However, the higher the scan
425 rate, the more quickly a tasks memory is migrated to a local node if the
426 workload pattern changes and minimises performance impact due to remote
427 memory accesses. These sysctls control the thresholds for scan delays and
428 the number of pages scanned.
430 numa_balancing_scan_period_min_ms is the minimum time in milliseconds to
431 scan a tasks virtual memory. It effectively controls the maximum scanning
434 numa_balancing_scan_delay_ms is the starting "scan delay" used for a task
435 when it initially forks.
437 numa_balancing_scan_period_max_ms is the maximum time in milliseconds to
438 scan a tasks virtual memory. It effectively controls the minimum scanning
441 numa_balancing_scan_size_mb is how many megabytes worth of pages are
442 scanned for a given scan.
444 numa_balancing_migrate_deferred is how many page migrations get skipped
445 unconditionally, after a page migration is skipped because a page is shared
446 with other tasks. This reduces page migration overhead, and determines
447 how much stronger the "move task near its memory" policy scheduler becomes,
448 versus the "move memory near its task" memory management policy, for workloads
451 ==============================================================
453 osrelease, ostype & version:
460 #5 Wed Feb 25 21:49:24 MET 1998
462 The files osrelease and ostype should be clear enough. Version
463 needs a little more clarification however. The '#5' means that
464 this is the fifth kernel built from this source base and the
465 date behind it indicates the time the kernel was built.
466 The only way to tune these values is to rebuild the kernel :-)
468 ==============================================================
470 overflowgid & overflowuid:
472 if your architecture did not always support 32-bit UIDs (i.e. arm,
473 i386, m68k, sh, and sparc32), a fixed UID and GID will be returned to
474 applications that use the old 16-bit UID/GID system calls, if the
475 actual UID or GID would exceed 65535.
477 These sysctls allow you to change the value of the fixed UID and GID.
478 The default is 65534.
480 ==============================================================
484 The value in this file represents the number of seconds the kernel
485 waits before rebooting on a panic. When you use the software watchdog,
486 the recommended setting is 60.
488 ==============================================================
490 panic_on_unrecovered_nmi:
492 The default Linux behaviour on an NMI of either memory or unknown is
493 to continue operation. For many environments such as scientific
494 computing it is preferable that the box is taken out and the error
495 dealt with than an uncorrected parity/ECC error get propagated.
497 A small number of systems do generate NMI's for bizarre random reasons
498 such as power management so the default is off. That sysctl works like
499 the existing panic controls already in that directory.
501 ==============================================================
505 Controls the kernel's behaviour when an oops or BUG is encountered.
507 0: try to continue operation
509 1: panic immediately. If the `panic' sysctl is also non-zero then the
510 machine will be rebooted.
512 ==============================================================
514 panic_on_stackoverflow:
516 Controls the kernel's behavior when detecting the overflows of
517 kernel, IRQ and exception stacks except a user stack.
518 This file shows up if CONFIG_DEBUG_STACKOVERFLOW is enabled.
520 0: try to continue operation.
522 1: panic immediately.
524 ==============================================================
526 perf_cpu_time_max_percent:
528 Hints to the kernel how much CPU time it should be allowed to
529 use to handle perf sampling events. If the perf subsystem
530 is informed that its samples are exceeding this limit, it
531 will drop its sampling frequency to attempt to reduce its CPU
534 Some perf sampling happens in NMIs. If these samples
535 unexpectedly take too long to execute, the NMIs can become
536 stacked up next to each other so much that nothing else is
539 0: disable the mechanism. Do not monitor or correct perf's
540 sampling rate no matter how CPU time it takes.
542 1-100: attempt to throttle perf's sample rate to this
543 percentage of CPU. Note: the kernel calculates an
544 "expected" length of each sample event. 100 here means
545 100% of that expected length. Even if this is set to
546 100, you may still see sample throttling if this
547 length is exceeded. Set to 0 if you truly do not care
548 how much CPU is consumed.
550 ==============================================================
555 PID allocation wrap value. When the kernel's next PID value
556 reaches this value, it wraps back to a minimum PID value.
557 PIDs of value pid_max or larger are not allocated.
559 ==============================================================
563 The last pid allocated in the current (the one task using this sysctl
564 lives in) pid namespace. When selecting a pid for a next task on fork
565 kernel tries to allocate a number starting from this one.
567 ==============================================================
569 powersave-nap: (PPC only)
571 If set, Linux-PPC will use the 'nap' mode of powersaving,
572 otherwise the 'doze' mode will be used.
574 ==============================================================
578 The four values in printk denote: console_loglevel,
579 default_message_loglevel, minimum_console_loglevel and
580 default_console_loglevel respectively.
582 These values influence printk() behavior when printing or
583 logging error messages. See 'man 2 syslog' for more info on
584 the different loglevels.
586 - console_loglevel: messages with a higher priority than
587 this will be printed to the console
588 - default_message_loglevel: messages without an explicit priority
589 will be printed with this priority
590 - minimum_console_loglevel: minimum (highest) value to which
591 console_loglevel can be set
592 - default_console_loglevel: default value for console_loglevel
594 ==============================================================
598 Delay each printk message in printk_delay milliseconds
600 Value from 0 - 10000 is allowed.
602 ==============================================================
606 Some warning messages are rate limited. printk_ratelimit specifies
607 the minimum length of time between these messages (in jiffies), by
608 default we allow one every 5 seconds.
610 A value of 0 will disable rate limiting.
612 ==============================================================
614 printk_ratelimit_burst:
616 While long term we enforce one message per printk_ratelimit
617 seconds, we do allow a burst of messages to pass through.
618 printk_ratelimit_burst specifies the number of messages we can
619 send before ratelimiting kicks in.
621 ==============================================================
625 This option can be used to select the type of process address
626 space randomization that is used in the system, for architectures
627 that support this feature.
629 0 - Turn the process address space randomization off. This is the
630 default for architectures that do not support this feature anyways,
631 and kernels that are booted with the "norandmaps" parameter.
633 1 - Make the addresses of mmap base, stack and VDSO page randomized.
634 This, among other things, implies that shared libraries will be
635 loaded to random addresses. Also for PIE-linked binaries, the
636 location of code start is randomized. This is the default if the
637 CONFIG_COMPAT_BRK option is enabled.
639 2 - Additionally enable heap randomization. This is the default if
640 CONFIG_COMPAT_BRK is disabled.
642 There are a few legacy applications out there (such as some ancient
643 versions of libc.so.5 from 1996) that assume that brk area starts
644 just after the end of the code+bss. These applications break when
645 start of the brk area is randomized. There are however no known
646 non-legacy applications that would be broken this way, so for most
647 systems it is safe to choose full randomization.
649 Systems with ancient and/or broken binaries should be configured
650 with CONFIG_COMPAT_BRK enabled, which excludes the heap from process
651 address space randomization.
653 ==============================================================
655 reboot-cmd: (Sparc only)
657 ??? This seems to be a way to give an argument to the Sparc
658 ROM/Flash boot loader. Maybe to tell it what to do after
661 ==============================================================
663 rtsig-max & rtsig-nr:
665 The file rtsig-max can be used to tune the maximum number
666 of POSIX realtime (queued) signals that can be outstanding
669 rtsig-nr shows the number of RT signals currently queued.
671 ==============================================================
675 This file shows the size of the generic SCSI (sg) buffer.
676 You can't tune it just yet, but you could change it on
677 compile time by editing include/scsi/sg.h and changing
678 the value of SG_BIG_BUFF.
680 There shouldn't be any reason to change this value. If
681 you can come up with one, you probably know what you
684 ==============================================================
688 This parameter sets the total amount of shared memory pages that
689 can be used system wide. Hence, SHMALL should always be at least
690 ceil(shmmax/PAGE_SIZE).
692 If you are not sure what the default PAGE_SIZE is on your Linux
693 system, you can run the following command:
697 ==============================================================
701 This value can be used to query and set the run time limit
702 on the maximum shared memory segment size that can be created.
703 Shared memory segments up to 1Gb are now supported in the
704 kernel. This value defaults to SHMMAX.
706 ==============================================================
710 Linux lets you set resource limits, including how much memory one
711 process can consume, via setrlimit(2). Unfortunately, shared memory
712 segments are allowed to exist without association with any process, and
713 thus might not be counted against any resource limits. If enabled,
714 shared memory segments are automatically destroyed when their attach
715 count becomes zero after a detach or a process termination. It will
716 also destroy segments that were created, but never attached to, on exit
717 from the process. The only use left for IPC_RMID is to immediately
718 destroy an unattached segment. Of course, this breaks the way things are
719 defined, so some applications might stop working. Note that this
720 feature will do you no good unless you also configure your resource
721 limits (in particular, RLIMIT_AS and RLIMIT_NPROC). Most systems don't
724 Note that if you change this from 0 to 1, already created segments
725 without users and with a dead originative process will be destroyed.
727 ==============================================================
731 Non-zero if the kernel has been tainted. Numeric values, which
732 can be ORed together:
734 1 - A module with a non-GPL license has been loaded, this
735 includes modules with no license.
736 Set by modutils >= 2.4.9 and module-init-tools.
737 2 - A module was force loaded by insmod -f.
738 Set by modutils >= 2.4.9 and module-init-tools.
739 4 - Unsafe SMP processors: SMP with CPUs not designed for SMP.
740 8 - A module was forcibly unloaded from the system by rmmod -f.
741 16 - A hardware machine check error occurred on the system.
742 32 - A bad page was discovered on the system.
743 64 - The user has asked that the system be marked "tainted". This
744 could be because they are running software that directly modifies
745 the hardware, or for other reasons.
746 128 - The system has died.
747 256 - The ACPI DSDT has been overridden with one supplied by the user
748 instead of using the one provided by the hardware.
749 512 - A kernel warning has occurred.
750 1024 - A module from drivers/staging was loaded.
751 2048 - The system is working around a severe firmware bug.
752 4096 - An out-of-tree module has been loaded.
754 ==============================================================
758 The value in this file affects behavior of handling NMI. When the
759 value is non-zero, unknown NMI is trapped and then panic occurs. At
760 that time, kernel debugging information is displayed on console.
762 NMI switch that most IA32 servers have fires unknown NMI up, for
763 example. If a system hangs up, try pressing the NMI switch.
765 ==============================================================
769 This value can be used to control the frequency of hrtimer and NMI
770 events and the soft and hard lockup thresholds. The default threshold
773 The softlockup threshold is (2 * watchdog_thresh). Setting this
774 tunable to zero will disable lockup detection altogether.
776 ==============================================================