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 - kstack_depth_to_print [ X86 only ]
39 - modprobe ==> Documentation/debugging-modules.txt
41 - msg_next_id [ sysv ipc ]
52 - panic_on_unrecovered_nmi
53 - panic_on_stackoverflow
55 - powersave-nap [ PPC only ]
59 - printk_ratelimit_burst
61 - real-root-dev ==> Documentation/initrd.txt
62 - reboot-cmd [ SPARC only ]
66 - sem_next_id [ sysv ipc ]
67 - sg-big-buff [ generic SCSI device (sg) ]
68 - shm_next_id [ sysv ipc ]
73 - stop-a [ SPARC only ]
74 - sysrq ==> Documentation/sysrq.txt
81 ==============================================================
85 highwater lowwater frequency
87 If BSD-style process accounting is enabled these values control
88 its behaviour. If free space on filesystem where the log lives
89 goes below <lowwater>% accounting suspends. If free space gets
90 above <highwater>% accounting resumes. <Frequency> determines
91 how often do we check the amount of free space (value is in
94 That is, suspend accounting if there left <= 2% free; resume it
95 if we got >=4%; consider information about amount of free space
98 ==============================================================
104 See Doc*/kernel/power/video.txt, it allows mode of video boot to be
107 ==============================================================
111 Enables/Disables automatic recomputing of msgmni upon memory add/remove
112 or upon ipc namespace creation/removal (see the msgmni description
113 above). Echoing "1" into this file enables msgmni automatic recomputing.
114 Echoing "0" turns it off. auto_msgmni default value is 1.
117 ==============================================================
121 x86 bootloader identification
123 This gives the bootloader type number as indicated by the bootloader,
124 shifted left by 4, and OR'd with the low four bits of the bootloader
125 version. The reason for this encoding is that this used to match the
126 type_of_loader field in the kernel header; the encoding is kept for
127 backwards compatibility. That is, if the full bootloader type number
128 is 0x15 and the full version number is 0x234, this file will contain
129 the value 340 = 0x154.
131 See the type_of_loader and ext_loader_type fields in
132 Documentation/x86/boot.txt for additional information.
134 ==============================================================
138 x86 bootloader version
140 The complete bootloader version number. In the example above, this
141 file will contain the value 564 = 0x234.
143 See the type_of_loader and ext_loader_ver fields in
144 Documentation/x86/boot.txt for additional information.
146 ==============================================================
150 Controls the kernel's callhome behavior in case of a kernel panic.
152 The s390 hardware allows an operating system to send a notification
153 to a service organization (callhome) in case of an operating system panic.
155 When the value in this file is 0 (which is the default behavior)
156 nothing happens in case of a kernel panic. If this value is set to "1"
157 the complete kernel oops message is send to the IBM customer service
158 organization in case the mainframe the Linux operating system is running
159 on has a service contract with IBM.
161 ==============================================================
165 Highest valid capability of the running kernel. Exports
166 CAP_LAST_CAP from the kernel.
168 ==============================================================
172 core_pattern is used to specify a core dumpfile pattern name.
173 . max length 128 characters; default value is "core"
174 . core_pattern is used as a pattern template for the output filename;
175 certain string patterns (beginning with '%') are substituted with
177 . backward compatibility with core_uses_pid:
178 If core_pattern does not include "%p" (default does not)
179 and core_uses_pid is set, then .PID will be appended to
181 . corename format specifiers:
182 %<NUL> '%' is dropped
185 %P global pid (init PID namespace)
188 %d dump mode, matches PR_SET_DUMPABLE and
189 /proc/sys/fs/suid_dumpable
193 %e executable filename (may be shortened)
195 %<OTHER> both are dropped
196 . If the first character of the pattern is a '|', the kernel will treat
197 the rest of the pattern as a command to run. The core dump will be
198 written to the standard input of that program instead of to a file.
200 ==============================================================
204 This sysctl is only applicable when core_pattern is configured to pipe
205 core files to a user space helper (when the first character of
206 core_pattern is a '|', see above). When collecting cores via a pipe
207 to an application, it is occasionally useful for the collecting
208 application to gather data about the crashing process from its
209 /proc/pid directory. In order to do this safely, the kernel must wait
210 for the collecting process to exit, so as not to remove the crashing
211 processes proc files prematurely. This in turn creates the
212 possibility that a misbehaving userspace collecting process can block
213 the reaping of a crashed process simply by never exiting. This sysctl
214 defends against that. It defines how many concurrent crashing
215 processes may be piped to user space applications in parallel. If
216 this value is exceeded, then those crashing processes above that value
217 are noted via the kernel log and their cores are skipped. 0 is a
218 special value, indicating that unlimited processes may be captured in
219 parallel, but that no waiting will take place (i.e. the collecting
220 process is not guaranteed access to /proc/<crashing pid>/). This
223 ==============================================================
227 The default coredump filename is "core". By setting
228 core_uses_pid to 1, the coredump filename becomes core.PID.
229 If core_pattern does not include "%p" (default does not)
230 and core_uses_pid is set, then .PID will be appended to
233 ==============================================================
237 When the value in this file is 0, ctrl-alt-del is trapped and
238 sent to the init(1) program to handle a graceful restart.
239 When, however, the value is > 0, Linux's reaction to a Vulcan
240 Nerve Pinch (tm) will be an immediate reboot, without even
241 syncing its dirty buffers.
243 Note: when a program (like dosemu) has the keyboard in 'raw'
244 mode, the ctrl-alt-del is intercepted by the program before it
245 ever reaches the kernel tty layer, and it's up to the program
246 to decide what to do with it.
248 ==============================================================
252 This toggle indicates whether unprivileged users are prevented
253 from using dmesg(8) to view messages from the kernel's log buffer.
254 When dmesg_restrict is set to (0) there are no restrictions. When
255 dmesg_restrict is set set to (1), users must have CAP_SYSLOG to use
258 The kernel config option CONFIG_SECURITY_DMESG_RESTRICT sets the
259 default value of dmesg_restrict.
261 ==============================================================
263 domainname & hostname:
265 These files can be used to set the NIS/YP domainname and the
266 hostname of your box in exactly the same way as the commands
267 domainname and hostname, i.e.:
268 # echo "darkstar" > /proc/sys/kernel/hostname
269 # echo "mydomain" > /proc/sys/kernel/domainname
270 has the same effect as
271 # hostname "darkstar"
272 # domainname "mydomain"
274 Note, however, that the classic darkstar.frop.org has the
275 hostname "darkstar" and DNS (Internet Domain Name Server)
276 domainname "frop.org", not to be confused with the NIS (Network
277 Information Service) or YP (Yellow Pages) domainname. These two
278 domain names are in general different. For a detailed discussion
279 see the hostname(1) man page.
281 ==============================================================
285 Path for the hotplug policy agent.
286 Default value is "/sbin/hotplug".
288 ==============================================================
292 This toggle indicates whether restrictions are placed on
293 exposing kernel addresses via /proc and other interfaces.
295 When kptr_restrict is set to (0), the default, there are no restrictions.
297 When kptr_restrict is set to (1), kernel pointers printed using the %pK
298 format specifier will be replaced with 0's unless the user has CAP_SYSLOG
299 and effective user and group ids are equal to the real ids. This is
300 because %pK checks are done at read() time rather than open() time, so
301 if permissions are elevated between the open() and the read() (e.g via
302 a setuid binary) then %pK will not leak kernel pointers to unprivileged
303 users. Note, this is a temporary solution only. The correct long-term
304 solution is to do the permission checks at open() time. Consider removing
305 world read permissions from files that use %pK, and using dmesg_restrict
306 to protect against uses of %pK in dmesg(8) if leaking kernel pointer
307 values to unprivileged users is a concern.
309 When kptr_restrict is set to (2), kernel pointers printed using
310 %pK will be replaced with 0's regardless of privileges.
312 ==============================================================
314 kstack_depth_to_print: (X86 only)
316 Controls the number of words to print when dumping the raw
319 ==============================================================
323 This flag controls the L2 cache of G3 processor boards. If
324 0, the cache is disabled. Enabled if nonzero.
326 ==============================================================
330 A toggle value indicating if modules are allowed to be loaded
331 in an otherwise modular kernel. This toggle defaults to off
332 (0), but can be set true (1). Once true, modules can be
333 neither loaded nor unloaded, and the toggle cannot be set back
336 ==============================================================
338 msg_next_id, sem_next_id, and shm_next_id:
340 These three toggles allows to specify desired id for next allocated IPC
341 object: message, semaphore or shared memory respectively.
343 By default they are equal to -1, which means generic allocation logic.
344 Possible values to set are in range {0..INT_MAX}.
347 1) kernel doesn't guarantee, that new object will have desired id. So,
348 it's up to userspace, how to handle an object with "wrong" id.
349 2) Toggle with non-default value will be set back to -1 by kernel after
350 successful IPC object allocation.
352 ==============================================================
356 Enables/Disables the NMI watchdog on x86 systems. When the value is
357 non-zero the NMI watchdog is enabled and will continuously test all
358 online cpus to determine whether or not they are still functioning
359 properly. Currently, passing "nmi_watchdog=" parameter at boot time is
360 required for this function to work.
362 If LAPIC NMI watchdog method is in use (nmi_watchdog=2 kernel
363 parameter), the NMI watchdog shares registers with oprofile. By
364 disabling the NMI watchdog, oprofile may have more registers to
367 ==============================================================
371 Enables/disables automatic page fault based NUMA memory
372 balancing. Memory is moved automatically to nodes
373 that access it often.
375 Enables/disables automatic NUMA memory balancing. On NUMA machines, there
376 is a performance penalty if remote memory is accessed by a CPU. When this
377 feature is enabled the kernel samples what task thread is accessing memory
378 by periodically unmapping pages and later trapping a page fault. At the
379 time of the page fault, it is determined if the data being accessed should
380 be migrated to a local memory node.
382 The unmapping of pages and trapping faults incur additional overhead that
383 ideally is offset by improved memory locality but there is no universal
384 guarantee. If the target workload is already bound to NUMA nodes then this
385 feature should be disabled. Otherwise, if the system overhead from the
386 feature is too high then the rate the kernel samples for NUMA hinting
387 faults may be controlled by the numa_balancing_scan_period_min_ms,
388 numa_balancing_scan_delay_ms, numa_balancing_scan_period_max_ms,
389 numa_balancing_scan_size_mb, numa_balancing_settle_count sysctls and
390 numa_balancing_migrate_deferred.
392 ==============================================================
394 numa_balancing_scan_period_min_ms, numa_balancing_scan_delay_ms,
395 numa_balancing_scan_period_max_ms, numa_balancing_scan_size_mb
397 Automatic NUMA balancing scans tasks address space and unmaps pages to
398 detect if pages are properly placed or if the data should be migrated to a
399 memory node local to where the task is running. Every "scan delay" the task
400 scans the next "scan size" number of pages in its address space. When the
401 end of the address space is reached the scanner restarts from the beginning.
403 In combination, the "scan delay" and "scan size" determine the scan rate.
404 When "scan delay" decreases, the scan rate increases. The scan delay and
405 hence the scan rate of every task is adaptive and depends on historical
406 behaviour. If pages are properly placed then the scan delay increases,
407 otherwise the scan delay decreases. The "scan size" is not adaptive but
408 the higher the "scan size", the higher the scan rate.
410 Higher scan rates incur higher system overhead as page faults must be
411 trapped and potentially data must be migrated. However, the higher the scan
412 rate, the more quickly a tasks memory is migrated to a local node if the
413 workload pattern changes and minimises performance impact due to remote
414 memory accesses. These sysctls control the thresholds for scan delays and
415 the number of pages scanned.
417 numa_balancing_scan_period_min_ms is the minimum time in milliseconds to
418 scan a tasks virtual memory. It effectively controls the maximum scanning
421 numa_balancing_scan_delay_ms is the starting "scan delay" used for a task
422 when it initially forks.
424 numa_balancing_scan_period_max_ms is the maximum time in milliseconds to
425 scan a tasks virtual memory. It effectively controls the minimum scanning
428 numa_balancing_scan_size_mb is how many megabytes worth of pages are
429 scanned for a given scan.
431 numa_balancing_settle_count is how many scan periods must complete before
432 the schedule balancer stops pushing the task towards a preferred node. This
433 gives the scheduler a chance to place the task on an alternative node if the
434 preferred node is overloaded.
436 numa_balancing_migrate_deferred is how many page migrations get skipped
437 unconditionally, after a page migration is skipped because a page is shared
438 with other tasks. This reduces page migration overhead, and determines
439 how much stronger the "move task near its memory" policy scheduler becomes,
440 versus the "move memory near its task" memory management policy, for workloads
443 ==============================================================
445 osrelease, ostype & version:
452 #5 Wed Feb 25 21:49:24 MET 1998
454 The files osrelease and ostype should be clear enough. Version
455 needs a little more clarification however. The '#5' means that
456 this is the fifth kernel built from this source base and the
457 date behind it indicates the time the kernel was built.
458 The only way to tune these values is to rebuild the kernel :-)
460 ==============================================================
462 overflowgid & overflowuid:
464 if your architecture did not always support 32-bit UIDs (i.e. arm,
465 i386, m68k, sh, and sparc32), a fixed UID and GID will be returned to
466 applications that use the old 16-bit UID/GID system calls, if the
467 actual UID or GID would exceed 65535.
469 These sysctls allow you to change the value of the fixed UID and GID.
470 The default is 65534.
472 ==============================================================
476 The value in this file represents the number of seconds the kernel
477 waits before rebooting on a panic. When you use the software watchdog,
478 the recommended setting is 60.
480 ==============================================================
482 panic_on_unrecovered_nmi:
484 The default Linux behaviour on an NMI of either memory or unknown is
485 to continue operation. For many environments such as scientific
486 computing it is preferable that the box is taken out and the error
487 dealt with than an uncorrected parity/ECC error get propagated.
489 A small number of systems do generate NMI's for bizarre random reasons
490 such as power management so the default is off. That sysctl works like
491 the existing panic controls already in that directory.
493 ==============================================================
497 Controls the kernel's behaviour when an oops or BUG is encountered.
499 0: try to continue operation
501 1: panic immediately. If the `panic' sysctl is also non-zero then the
502 machine will be rebooted.
504 ==============================================================
506 panic_on_stackoverflow:
508 Controls the kernel's behavior when detecting the overflows of
509 kernel, IRQ and exception stacks except a user stack.
510 This file shows up if CONFIG_DEBUG_STACKOVERFLOW is enabled.
512 0: try to continue operation.
514 1: panic immediately.
516 ==============================================================
518 perf_cpu_time_max_percent:
520 Hints to the kernel how much CPU time it should be allowed to
521 use to handle perf sampling events. If the perf subsystem
522 is informed that its samples are exceeding this limit, it
523 will drop its sampling frequency to attempt to reduce its CPU
526 Some perf sampling happens in NMIs. If these samples
527 unexpectedly take too long to execute, the NMIs can become
528 stacked up next to each other so much that nothing else is
531 0: disable the mechanism. Do not monitor or correct perf's
532 sampling rate no matter how CPU time it takes.
534 1-100: attempt to throttle perf's sample rate to this
535 percentage of CPU. Note: the kernel calculates an
536 "expected" length of each sample event. 100 here means
537 100% of that expected length. Even if this is set to
538 100, you may still see sample throttling if this
539 length is exceeded. Set to 0 if you truly do not care
540 how much CPU is consumed.
542 ==============================================================
547 PID allocation wrap value. When the kernel's next PID value
548 reaches this value, it wraps back to a minimum PID value.
549 PIDs of value pid_max or larger are not allocated.
551 ==============================================================
555 The last pid allocated in the current (the one task using this sysctl
556 lives in) pid namespace. When selecting a pid for a next task on fork
557 kernel tries to allocate a number starting from this one.
559 ==============================================================
561 powersave-nap: (PPC only)
563 If set, Linux-PPC will use the 'nap' mode of powersaving,
564 otherwise the 'doze' mode will be used.
566 ==============================================================
570 The four values in printk denote: console_loglevel,
571 default_message_loglevel, minimum_console_loglevel and
572 default_console_loglevel respectively.
574 These values influence printk() behavior when printing or
575 logging error messages. See 'man 2 syslog' for more info on
576 the different loglevels.
578 - console_loglevel: messages with a higher priority than
579 this will be printed to the console
580 - default_message_loglevel: messages without an explicit priority
581 will be printed with this priority
582 - minimum_console_loglevel: minimum (highest) value to which
583 console_loglevel can be set
584 - default_console_loglevel: default value for console_loglevel
586 ==============================================================
590 Delay each printk message in printk_delay milliseconds
592 Value from 0 - 10000 is allowed.
594 ==============================================================
598 Some warning messages are rate limited. printk_ratelimit specifies
599 the minimum length of time between these messages (in jiffies), by
600 default we allow one every 5 seconds.
602 A value of 0 will disable rate limiting.
604 ==============================================================
606 printk_ratelimit_burst:
608 While long term we enforce one message per printk_ratelimit
609 seconds, we do allow a burst of messages to pass through.
610 printk_ratelimit_burst specifies the number of messages we can
611 send before ratelimiting kicks in.
613 ==============================================================
617 This option can be used to select the type of process address
618 space randomization that is used in the system, for architectures
619 that support this feature.
621 0 - Turn the process address space randomization off. This is the
622 default for architectures that do not support this feature anyways,
623 and kernels that are booted with the "norandmaps" parameter.
625 1 - Make the addresses of mmap base, stack and VDSO page randomized.
626 This, among other things, implies that shared libraries will be
627 loaded to random addresses. Also for PIE-linked binaries, the
628 location of code start is randomized. This is the default if the
629 CONFIG_COMPAT_BRK option is enabled.
631 2 - Additionally enable heap randomization. This is the default if
632 CONFIG_COMPAT_BRK is disabled.
634 There are a few legacy applications out there (such as some ancient
635 versions of libc.so.5 from 1996) that assume that brk area starts
636 just after the end of the code+bss. These applications break when
637 start of the brk area is randomized. There are however no known
638 non-legacy applications that would be broken this way, so for most
639 systems it is safe to choose full randomization.
641 Systems with ancient and/or broken binaries should be configured
642 with CONFIG_COMPAT_BRK enabled, which excludes the heap from process
643 address space randomization.
645 ==============================================================
647 reboot-cmd: (Sparc only)
649 ??? This seems to be a way to give an argument to the Sparc
650 ROM/Flash boot loader. Maybe to tell it what to do after
653 ==============================================================
655 rtsig-max & rtsig-nr:
657 The file rtsig-max can be used to tune the maximum number
658 of POSIX realtime (queued) signals that can be outstanding
661 rtsig-nr shows the number of RT signals currently queued.
663 ==============================================================
667 This file shows the size of the generic SCSI (sg) buffer.
668 You can't tune it just yet, but you could change it on
669 compile time by editing include/scsi/sg.h and changing
670 the value of SG_BIG_BUFF.
672 There shouldn't be any reason to change this value. If
673 you can come up with one, you probably know what you
676 ==============================================================
680 This parameter sets the total amount of shared memory pages that
681 can be used system wide. Hence, SHMALL should always be at least
682 ceil(shmmax/PAGE_SIZE).
684 If you are not sure what the default PAGE_SIZE is on your Linux
685 system, you can run the following command:
689 ==============================================================
693 This value can be used to query and set the run time limit
694 on the maximum shared memory segment size that can be created.
695 Shared memory segments up to 1Gb are now supported in the
696 kernel. This value defaults to SHMMAX.
698 ==============================================================
702 Linux lets you set resource limits, including how much memory one
703 process can consume, via setrlimit(2). Unfortunately, shared memory
704 segments are allowed to exist without association with any process, and
705 thus might not be counted against any resource limits. If enabled,
706 shared memory segments are automatically destroyed when their attach
707 count becomes zero after a detach or a process termination. It will
708 also destroy segments that were created, but never attached to, on exit
709 from the process. The only use left for IPC_RMID is to immediately
710 destroy an unattached segment. Of course, this breaks the way things are
711 defined, so some applications might stop working. Note that this
712 feature will do you no good unless you also configure your resource
713 limits (in particular, RLIMIT_AS and RLIMIT_NPROC). Most systems don't
716 Note that if you change this from 0 to 1, already created segments
717 without users and with a dead originative process will be destroyed.
719 ==============================================================
723 Non-zero if the kernel has been tainted. Numeric values, which
724 can be ORed together:
726 1 - A module with a non-GPL license has been loaded, this
727 includes modules with no license.
728 Set by modutils >= 2.4.9 and module-init-tools.
729 2 - A module was force loaded by insmod -f.
730 Set by modutils >= 2.4.9 and module-init-tools.
731 4 - Unsafe SMP processors: SMP with CPUs not designed for SMP.
732 8 - A module was forcibly unloaded from the system by rmmod -f.
733 16 - A hardware machine check error occurred on the system.
734 32 - A bad page was discovered on the system.
735 64 - The user has asked that the system be marked "tainted". This
736 could be because they are running software that directly modifies
737 the hardware, or for other reasons.
738 128 - The system has died.
739 256 - The ACPI DSDT has been overridden with one supplied by the user
740 instead of using the one provided by the hardware.
741 512 - A kernel warning has occurred.
742 1024 - A module from drivers/staging was loaded.
743 2048 - The system is working around a severe firmware bug.
744 4096 - An out-of-tree module has been loaded.
746 ==============================================================
750 The value in this file affects behavior of handling NMI. When the
751 value is non-zero, unknown NMI is trapped and then panic occurs. At
752 that time, kernel debugging information is displayed on console.
754 NMI switch that most IA32 servers have fires unknown NMI up, for
755 example. If a system hangs up, try pressing the NMI switch.
757 ==============================================================
761 This value can be used to control the frequency of hrtimer and NMI
762 events and the soft and hard lockup thresholds. The default threshold
765 The softlockup threshold is (2 * watchdog_thresh). Setting this
766 tunable to zero will disable lockup detection altogether.
768 ==============================================================