• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1Documentation 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>
4
5For general info and legal blurb, please look in README.
6
7==============================================================
8
9This file contains documentation for the sysctl files in
10/proc/sys/kernel/ and is valid for Linux kernel version 2.2.
11
12The files in this directory can be used to tune and monitor
13miscellaneous and general things in the operation of the Linux
14kernel. Since some of the files _can_ be used to screw up your
15system, it is advisable to read both documentation and source
16before actually making adjustments.
17
18Currently, these files might (depending on your configuration)
19show up in /proc/sys/kernel:
20
21- acct
22- acpi_video_flags
23- auto_msgmni
24- bootloader_type	     [ X86 only ]
25- bootloader_version	     [ X86 only ]
26- callhome		     [ S390 only ]
27- cap_last_cap
28- core_pattern
29- core_pipe_limit
30- core_uses_pid
31- ctrl-alt-del
32- dmesg_restrict
33- domainname
34- hostname
35- hotplug
36- hardlockup_all_cpu_backtrace
37- hung_task_panic
38- hung_task_check_count
39- hung_task_timeout_secs
40- hung_task_warnings
41- kexec_load_disabled
42- kptr_restrict
43- kstack_depth_to_print       [ X86 only ]
44- l2cr                        [ PPC only ]
45- modprobe                    ==> Documentation/debugging-modules.txt
46- modules_disabled
47- msg_next_id		      [ sysv ipc ]
48- msgmax
49- msgmnb
50- msgmni
51- nmi_watchdog
52- osrelease
53- ostype
54- overflowgid
55- overflowuid
56- panic
57- panic_on_oops
58- panic_on_stackoverflow
59- panic_on_unrecovered_nmi
60- panic_on_warn
61- perf_cpu_time_max_percent
62- perf_event_paranoid
63- pid_max
64- powersave-nap               [ PPC only ]
65- printk
66- printk_delay
67- printk_ratelimit
68- printk_ratelimit_burst
69- randomize_va_space
70- real-root-dev               ==> Documentation/initrd.txt
71- reboot-cmd                  [ SPARC only ]
72- rtsig-max
73- rtsig-nr
74- sem
75- sem_next_id		      [ sysv ipc ]
76- sg-big-buff                 [ generic SCSI device (sg) ]
77- shm_next_id		      [ sysv ipc ]
78- shm_rmid_forced
79- shmall
80- shmmax                      [ sysv ipc ]
81- shmmni
82- softlockup_all_cpu_backtrace
83- soft_watchdog
84- stop-a                      [ SPARC only ]
85- sysrq                       ==> Documentation/sysrq.txt
86- sysctl_writes_strict
87- tainted
88- threads-max
89- unknown_nmi_panic
90- watchdog
91- watchdog_thresh
92- version
93
94==============================================================
95
96acct:
97
98highwater lowwater frequency
99
100If BSD-style process accounting is enabled these values control
101its behaviour. If free space on filesystem where the log lives
102goes below <lowwater>% accounting suspends. If free space gets
103above <highwater>% accounting resumes. <Frequency> determines
104how often do we check the amount of free space (value is in
105seconds). Default:
1064 2 30
107That is, suspend accounting if there left <= 2% free; resume it
108if we got >=4%; consider information about amount of free space
109valid for 30 seconds.
110
111==============================================================
112
113acpi_video_flags:
114
115flags
116
117See Doc*/kernel/power/video.txt, it allows mode of video boot to be
118set during run time.
119
120==============================================================
121
122auto_msgmni:
123
124This variable has no effect and may be removed in future kernel
125releases. Reading it always returns 0.
126Up to Linux 3.17, it enabled/disabled automatic recomputing of msgmni
127upon memory add/remove or upon ipc namespace creation/removal.
128Echoing "1" into this file enabled msgmni automatic recomputing.
129Echoing "0" turned it off. auto_msgmni default value was 1.
130
131
132==============================================================
133
134bootloader_type:
135
136x86 bootloader identification
137
138This gives the bootloader type number as indicated by the bootloader,
139shifted left by 4, and OR'd with the low four bits of the bootloader
140version.  The reason for this encoding is that this used to match the
141type_of_loader field in the kernel header; the encoding is kept for
142backwards compatibility.  That is, if the full bootloader type number
143is 0x15 and the full version number is 0x234, this file will contain
144the value 340 = 0x154.
145
146See the type_of_loader and ext_loader_type fields in
147Documentation/x86/boot.txt for additional information.
148
149==============================================================
150
151bootloader_version:
152
153x86 bootloader version
154
155The complete bootloader version number.  In the example above, this
156file will contain the value 564 = 0x234.
157
158See the type_of_loader and ext_loader_ver fields in
159Documentation/x86/boot.txt for additional information.
160
161==============================================================
162
163callhome:
164
165Controls the kernel's callhome behavior in case of a kernel panic.
166
167The s390 hardware allows an operating system to send a notification
168to a service organization (callhome) in case of an operating system panic.
169
170When the value in this file is 0 (which is the default behavior)
171nothing happens in case of a kernel panic. If this value is set to "1"
172the complete kernel oops message is send to the IBM customer service
173organization in case the mainframe the Linux operating system is running
174on has a service contract with IBM.
175
176==============================================================
177
178cap_last_cap
179
180Highest valid capability of the running kernel.  Exports
181CAP_LAST_CAP from the kernel.
182
183==============================================================
184
185core_pattern:
186
187core_pattern is used to specify a core dumpfile pattern name.
188. max length 128 characters; default value is "core"
189. core_pattern is used as a pattern template for the output filename;
190  certain string patterns (beginning with '%') are substituted with
191  their actual values.
192. backward compatibility with core_uses_pid:
193	If core_pattern does not include "%p" (default does not)
194	and core_uses_pid is set, then .PID will be appended to
195	the filename.
196. corename format specifiers:
197	%<NUL>	'%' is dropped
198	%%	output one '%'
199	%p	pid
200	%P	global pid (init PID namespace)
201	%i	tid
202	%I	global tid (init PID namespace)
203	%u	uid (in initial user namespace)
204	%g	gid (in initial user namespace)
205	%d	dump mode, matches PR_SET_DUMPABLE and
206		/proc/sys/fs/suid_dumpable
207	%s	signal number
208	%t	UNIX time of dump
209	%h	hostname
210	%e	executable filename (may be shortened)
211	%E	executable path
212	%<OTHER> both are dropped
213. If the first character of the pattern is a '|', the kernel will treat
214  the rest of the pattern as a command to run.  The core dump will be
215  written to the standard input of that program instead of to a file.
216
217==============================================================
218
219core_pipe_limit:
220
221This sysctl is only applicable when core_pattern is configured to pipe
222core files to a user space helper (when the first character of
223core_pattern is a '|', see above).  When collecting cores via a pipe
224to an application, it is occasionally useful for the collecting
225application to gather data about the crashing process from its
226/proc/pid directory.  In order to do this safely, the kernel must wait
227for the collecting process to exit, so as not to remove the crashing
228processes proc files prematurely.  This in turn creates the
229possibility that a misbehaving userspace collecting process can block
230the reaping of a crashed process simply by never exiting.  This sysctl
231defends against that.  It defines how many concurrent crashing
232processes may be piped to user space applications in parallel.  If
233this value is exceeded, then those crashing processes above that value
234are noted via the kernel log and their cores are skipped.  0 is a
235special value, indicating that unlimited processes may be captured in
236parallel, but that no waiting will take place (i.e. the collecting
237process is not guaranteed access to /proc/<crashing pid>/).  This
238value defaults to 0.
239
240==============================================================
241
242core_uses_pid:
243
244The default coredump filename is "core".  By setting
245core_uses_pid to 1, the coredump filename becomes core.PID.
246If core_pattern does not include "%p" (default does not)
247and core_uses_pid is set, then .PID will be appended to
248the filename.
249
250==============================================================
251
252ctrl-alt-del:
253
254When the value in this file is 0, ctrl-alt-del is trapped and
255sent to the init(1) program to handle a graceful restart.
256When, however, the value is > 0, Linux's reaction to a Vulcan
257Nerve Pinch (tm) will be an immediate reboot, without even
258syncing its dirty buffers.
259
260Note: when a program (like dosemu) has the keyboard in 'raw'
261mode, the ctrl-alt-del is intercepted by the program before it
262ever reaches the kernel tty layer, and it's up to the program
263to decide what to do with it.
264
265==============================================================
266
267dmesg_restrict:
268
269This toggle indicates whether unprivileged users are prevented
270from using dmesg(8) to view messages from the kernel's log buffer.
271When dmesg_restrict is set to (0) there are no restrictions. When
272dmesg_restrict is set set to (1), users must have CAP_SYSLOG to use
273dmesg(8).
274
275The kernel config option CONFIG_SECURITY_DMESG_RESTRICT sets the
276default value of dmesg_restrict.
277
278==============================================================
279
280domainname & hostname:
281
282These files can be used to set the NIS/YP domainname and the
283hostname of your box in exactly the same way as the commands
284domainname and hostname, i.e.:
285# echo "darkstar" > /proc/sys/kernel/hostname
286# echo "mydomain" > /proc/sys/kernel/domainname
287has the same effect as
288# hostname "darkstar"
289# domainname "mydomain"
290
291Note, however, that the classic darkstar.frop.org has the
292hostname "darkstar" and DNS (Internet Domain Name Server)
293domainname "frop.org", not to be confused with the NIS (Network
294Information Service) or YP (Yellow Pages) domainname. These two
295domain names are in general different. For a detailed discussion
296see the hostname(1) man page.
297
298==============================================================
299hardlockup_all_cpu_backtrace:
300
301This value controls the hard lockup detector behavior when a hard
302lockup condition is detected as to whether or not to gather further
303debug information. If enabled, arch-specific all-CPU stack dumping
304will be initiated.
305
3060: do nothing. This is the default behavior.
307
3081: on detection capture more debug information.
309==============================================================
310
311hotplug:
312
313Path for the hotplug policy agent.
314Default value is "/sbin/hotplug".
315
316==============================================================
317
318hung_task_panic:
319
320Controls the kernel's behavior when a hung task is detected.
321This file shows up if CONFIG_DETECT_HUNG_TASK is enabled.
322
3230: continue operation. This is the default behavior.
324
3251: panic immediately.
326
327==============================================================
328
329hung_task_check_count:
330
331The upper bound on the number of tasks that are checked.
332This file shows up if CONFIG_DETECT_HUNG_TASK is enabled.
333
334==============================================================
335
336hung_task_timeout_secs:
337
338Check interval. When a task in D state did not get scheduled
339for more than this value report a warning.
340This file shows up if CONFIG_DETECT_HUNG_TASK is enabled.
341
3420: means infinite timeout - no checking done.
343Possible values to set are in range {0..LONG_MAX/HZ}.
344
345==============================================================
346
347hung_task_warnings:
348
349The maximum number of warnings to report. During a check interval
350if a hung task is detected, this value is decreased by 1.
351When this value reaches 0, no more warnings will be reported.
352This file shows up if CONFIG_DETECT_HUNG_TASK is enabled.
353
354-1: report an infinite number of warnings.
355
356==============================================================
357
358kexec_load_disabled:
359
360A toggle indicating if the kexec_load syscall has been disabled. This
361value defaults to 0 (false: kexec_load enabled), but can be set to 1
362(true: kexec_load disabled). Once true, kexec can no longer be used, and
363the toggle cannot be set back to false. This allows a kexec image to be
364loaded before disabling the syscall, allowing a system to set up (and
365later use) an image without it being altered. Generally used together
366with the "modules_disabled" sysctl.
367
368==============================================================
369
370kptr_restrict:
371
372This toggle indicates whether restrictions are placed on
373exposing kernel addresses via /proc and other interfaces.
374
375When kptr_restrict is set to (0), the default, there are no restrictions.
376
377When kptr_restrict is set to (1), kernel pointers printed using the %pK
378format specifier will be replaced with 0's unless the user has CAP_SYSLOG
379and effective user and group ids are equal to the real ids. This is
380because %pK checks are done at read() time rather than open() time, so
381if permissions are elevated between the open() and the read() (e.g via
382a setuid binary) then %pK will not leak kernel pointers to unprivileged
383users. Note, this is a temporary solution only. The correct long-term
384solution is to do the permission checks at open() time. Consider removing
385world read permissions from files that use %pK, and using dmesg_restrict
386to protect against uses of %pK in dmesg(8) if leaking kernel pointer
387values to unprivileged users is a concern.
388
389When kptr_restrict is set to (2), kernel pointers printed using
390%pK will be replaced with 0's regardless of privileges.
391
392==============================================================
393
394kstack_depth_to_print: (X86 only)
395
396Controls the number of words to print when dumping the raw
397kernel stack.
398
399==============================================================
400
401l2cr: (PPC only)
402
403This flag controls the L2 cache of G3 processor boards. If
4040, the cache is disabled. Enabled if nonzero.
405
406==============================================================
407
408modules_disabled:
409
410A toggle value indicating if modules are allowed to be loaded
411in an otherwise modular kernel.  This toggle defaults to off
412(0), but can be set true (1).  Once true, modules can be
413neither loaded nor unloaded, and the toggle cannot be set back
414to false.  Generally used with the "kexec_load_disabled" toggle.
415
416==============================================================
417
418msg_next_id, sem_next_id, and shm_next_id:
419
420These three toggles allows to specify desired id for next allocated IPC
421object: message, semaphore or shared memory respectively.
422
423By default they are equal to -1, which means generic allocation logic.
424Possible values to set are in range {0..INT_MAX}.
425
426Notes:
4271) kernel doesn't guarantee, that new object will have desired id. So,
428it's up to userspace, how to handle an object with "wrong" id.
4292) Toggle with non-default value will be set back to -1 by kernel after
430successful IPC object allocation.
431
432==============================================================
433
434nmi_watchdog:
435
436This parameter can be used to control the NMI watchdog
437(i.e. the hard lockup detector) on x86 systems.
438
439   0 - disable the hard lockup detector
440   1 - enable the hard lockup detector
441
442The hard lockup detector monitors each CPU for its ability to respond to
443timer interrupts. The mechanism utilizes CPU performance counter registers
444that are programmed to generate Non-Maskable Interrupts (NMIs) periodically
445while a CPU is busy. Hence, the alternative name 'NMI watchdog'.
446
447The NMI watchdog is disabled by default if the kernel is running as a guest
448in a KVM virtual machine. This default can be overridden by adding
449
450   nmi_watchdog=1
451
452to the guest kernel command line (see Documentation/kernel-parameters.txt).
453
454==============================================================
455
456numa_balancing
457
458Enables/disables automatic page fault based NUMA memory
459balancing. Memory is moved automatically to nodes
460that access it often.
461
462Enables/disables automatic NUMA memory balancing. On NUMA machines, there
463is a performance penalty if remote memory is accessed by a CPU. When this
464feature is enabled the kernel samples what task thread is accessing memory
465by periodically unmapping pages and later trapping a page fault. At the
466time of the page fault, it is determined if the data being accessed should
467be migrated to a local memory node.
468
469The unmapping of pages and trapping faults incur additional overhead that
470ideally is offset by improved memory locality but there is no universal
471guarantee. If the target workload is already bound to NUMA nodes then this
472feature should be disabled. Otherwise, if the system overhead from the
473feature is too high then the rate the kernel samples for NUMA hinting
474faults may be controlled by the numa_balancing_scan_period_min_ms,
475numa_balancing_scan_delay_ms, numa_balancing_scan_period_max_ms,
476numa_balancing_scan_size_mb, and numa_balancing_settle_count sysctls.
477
478==============================================================
479
480numa_balancing_scan_period_min_ms, numa_balancing_scan_delay_ms,
481numa_balancing_scan_period_max_ms, numa_balancing_scan_size_mb
482
483Automatic NUMA balancing scans tasks address space and unmaps pages to
484detect if pages are properly placed or if the data should be migrated to a
485memory node local to where the task is running.  Every "scan delay" the task
486scans the next "scan size" number of pages in its address space. When the
487end of the address space is reached the scanner restarts from the beginning.
488
489In combination, the "scan delay" and "scan size" determine the scan rate.
490When "scan delay" decreases, the scan rate increases.  The scan delay and
491hence the scan rate of every task is adaptive and depends on historical
492behaviour. If pages are properly placed then the scan delay increases,
493otherwise the scan delay decreases.  The "scan size" is not adaptive but
494the higher the "scan size", the higher the scan rate.
495
496Higher scan rates incur higher system overhead as page faults must be
497trapped and potentially data must be migrated. However, the higher the scan
498rate, the more quickly a tasks memory is migrated to a local node if the
499workload pattern changes and minimises performance impact due to remote
500memory accesses. These sysctls control the thresholds for scan delays and
501the number of pages scanned.
502
503numa_balancing_scan_period_min_ms is the minimum time in milliseconds to
504scan a tasks virtual memory. It effectively controls the maximum scanning
505rate for each task.
506
507numa_balancing_scan_delay_ms is the starting "scan delay" used for a task
508when it initially forks.
509
510numa_balancing_scan_period_max_ms is the maximum time in milliseconds to
511scan a tasks virtual memory. It effectively controls the minimum scanning
512rate for each task.
513
514numa_balancing_scan_size_mb is how many megabytes worth of pages are
515scanned for a given scan.
516
517==============================================================
518
519osrelease, ostype & version:
520
521# cat osrelease
5222.1.88
523# cat ostype
524Linux
525# cat version
526#5 Wed Feb 25 21:49:24 MET 1998
527
528The files osrelease and ostype should be clear enough. Version
529needs a little more clarification however. The '#5' means that
530this is the fifth kernel built from this source base and the
531date behind it indicates the time the kernel was built.
532The only way to tune these values is to rebuild the kernel :-)
533
534==============================================================
535
536overflowgid & overflowuid:
537
538if your architecture did not always support 32-bit UIDs (i.e. arm,
539i386, m68k, sh, and sparc32), a fixed UID and GID will be returned to
540applications that use the old 16-bit UID/GID system calls, if the
541actual UID or GID would exceed 65535.
542
543These sysctls allow you to change the value of the fixed UID and GID.
544The default is 65534.
545
546==============================================================
547
548panic:
549
550The value in this file represents the number of seconds the kernel
551waits before rebooting on a panic. When you use the software watchdog,
552the recommended setting is 60.
553
554==============================================================
555
556panic_on_oops:
557
558Controls the kernel's behaviour when an oops or BUG is encountered.
559
5600: try to continue operation
561
5621: panic immediately.  If the `panic' sysctl is also non-zero then the
563   machine will be rebooted.
564
565==============================================================
566
567panic_on_stackoverflow:
568
569Controls the kernel's behavior when detecting the overflows of
570kernel, IRQ and exception stacks except a user stack.
571This file shows up if CONFIG_DEBUG_STACKOVERFLOW is enabled.
572
5730: try to continue operation.
574
5751: panic immediately.
576
577==============================================================
578
579panic_on_unrecovered_nmi:
580
581The default Linux behaviour on an NMI of either memory or unknown is
582to continue operation. For many environments such as scientific
583computing it is preferable that the box is taken out and the error
584dealt with than an uncorrected parity/ECC error get propagated.
585
586A small number of systems do generate NMI's for bizarre random reasons
587such as power management so the default is off. That sysctl works like
588the existing panic controls already in that directory.
589
590==============================================================
591
592panic_on_warn:
593
594Calls panic() in the WARN() path when set to 1.  This is useful to avoid
595a kernel rebuild when attempting to kdump at the location of a WARN().
596
5970: only WARN(), default behaviour.
598
5991: call panic() after printing out WARN() location.
600
601==============================================================
602
603perf_cpu_time_max_percent:
604
605Hints to the kernel how much CPU time it should be allowed to
606use to handle perf sampling events.  If the perf subsystem
607is informed that its samples are exceeding this limit, it
608will drop its sampling frequency to attempt to reduce its CPU
609usage.
610
611Some perf sampling happens in NMIs.  If these samples
612unexpectedly take too long to execute, the NMIs can become
613stacked up next to each other so much that nothing else is
614allowed to execute.
615
6160: disable the mechanism.  Do not monitor or correct perf's
617   sampling rate no matter how CPU time it takes.
618
6191-100: attempt to throttle perf's sample rate to this
620   percentage of CPU.  Note: the kernel calculates an
621   "expected" length of each sample event.  100 here means
622   100% of that expected length.  Even if this is set to
623   100, you may still see sample throttling if this
624   length is exceeded.  Set to 0 if you truly do not care
625   how much CPU is consumed.
626
627==============================================================
628
629perf_event_paranoid:
630
631Controls use of the performance events system by unprivileged
632users (without CAP_SYS_ADMIN).  The default value is 3 if
633CONFIG_SECURITY_PERF_EVENTS_RESTRICT is set, or 1 otherwise.
634
635 -1: Allow use of (almost) all events by all users
636>=0: Disallow raw tracepoint access by users without CAP_IOC_LOCK
637>=1: Disallow CPU event access by users without CAP_SYS_ADMIN
638>=2: Disallow kernel profiling by users without CAP_SYS_ADMIN
639>=3: Disallow all event access by users without CAP_SYS_ADMIN
640
641==============================================================
642
643pid_max:
644
645PID allocation wrap value.  When the kernel's next PID value
646reaches this value, it wraps back to a minimum PID value.
647PIDs of value pid_max or larger are not allocated.
648
649==============================================================
650
651ns_last_pid:
652
653The last pid allocated in the current (the one task using this sysctl
654lives in) pid namespace. When selecting a pid for a next task on fork
655kernel tries to allocate a number starting from this one.
656
657==============================================================
658
659powersave-nap: (PPC only)
660
661If set, Linux-PPC will use the 'nap' mode of powersaving,
662otherwise the 'doze' mode will be used.
663
664==============================================================
665
666printk:
667
668The four values in printk denote: console_loglevel,
669default_message_loglevel, minimum_console_loglevel and
670default_console_loglevel respectively.
671
672These values influence printk() behavior when printing or
673logging error messages. See 'man 2 syslog' for more info on
674the different loglevels.
675
676- console_loglevel: messages with a higher priority than
677  this will be printed to the console
678- default_message_loglevel: messages without an explicit priority
679  will be printed with this priority
680- minimum_console_loglevel: minimum (highest) value to which
681  console_loglevel can be set
682- default_console_loglevel: default value for console_loglevel
683
684==============================================================
685
686printk_delay:
687
688Delay each printk message in printk_delay milliseconds
689
690Value from 0 - 10000 is allowed.
691
692==============================================================
693
694printk_ratelimit:
695
696Some warning messages are rate limited. printk_ratelimit specifies
697the minimum length of time between these messages (in jiffies), by
698default we allow one every 5 seconds.
699
700A value of 0 will disable rate limiting.
701
702==============================================================
703
704printk_ratelimit_burst:
705
706While long term we enforce one message per printk_ratelimit
707seconds, we do allow a burst of messages to pass through.
708printk_ratelimit_burst specifies the number of messages we can
709send before ratelimiting kicks in.
710
711==============================================================
712
713randomize_va_space:
714
715This option can be used to select the type of process address
716space randomization that is used in the system, for architectures
717that support this feature.
718
7190 - Turn the process address space randomization off.  This is the
720    default for architectures that do not support this feature anyways,
721    and kernels that are booted with the "norandmaps" parameter.
722
7231 - Make the addresses of mmap base, stack and VDSO page randomized.
724    This, among other things, implies that shared libraries will be
725    loaded to random addresses.  Also for PIE-linked binaries, the
726    location of code start is randomized.  This is the default if the
727    CONFIG_COMPAT_BRK option is enabled.
728
7292 - Additionally enable heap randomization.  This is the default if
730    CONFIG_COMPAT_BRK is disabled.
731
732    There are a few legacy applications out there (such as some ancient
733    versions of libc.so.5 from 1996) that assume that brk area starts
734    just after the end of the code+bss.  These applications break when
735    start of the brk area is randomized.  There are however no known
736    non-legacy applications that would be broken this way, so for most
737    systems it is safe to choose full randomization.
738
739    Systems with ancient and/or broken binaries should be configured
740    with CONFIG_COMPAT_BRK enabled, which excludes the heap from process
741    address space randomization.
742
743==============================================================
744
745reboot-cmd: (Sparc only)
746
747??? This seems to be a way to give an argument to the Sparc
748ROM/Flash boot loader. Maybe to tell it what to do after
749rebooting. ???
750
751==============================================================
752
753rtsig-max & rtsig-nr:
754
755The file rtsig-max can be used to tune the maximum number
756of POSIX realtime (queued) signals that can be outstanding
757in the system.
758
759rtsig-nr shows the number of RT signals currently queued.
760
761==============================================================
762
763sg-big-buff:
764
765This file shows the size of the generic SCSI (sg) buffer.
766You can't tune it just yet, but you could change it on
767compile time by editing include/scsi/sg.h and changing
768the value of SG_BIG_BUFF.
769
770There shouldn't be any reason to change this value. If
771you can come up with one, you probably know what you
772are doing anyway :)
773
774==============================================================
775
776shmall:
777
778This parameter sets the total amount of shared memory pages that
779can be used system wide. Hence, SHMALL should always be at least
780ceil(shmmax/PAGE_SIZE).
781
782If you are not sure what the default PAGE_SIZE is on your Linux
783system, you can run the following command:
784
785# getconf PAGE_SIZE
786
787==============================================================
788
789shmmax:
790
791This value can be used to query and set the run time limit
792on the maximum shared memory segment size that can be created.
793Shared memory segments up to 1Gb are now supported in the
794kernel.  This value defaults to SHMMAX.
795
796==============================================================
797
798shm_rmid_forced:
799
800Linux lets you set resource limits, including how much memory one
801process can consume, via setrlimit(2).  Unfortunately, shared memory
802segments are allowed to exist without association with any process, and
803thus might not be counted against any resource limits.  If enabled,
804shared memory segments are automatically destroyed when their attach
805count becomes zero after a detach or a process termination.  It will
806also destroy segments that were created, but never attached to, on exit
807from the process.  The only use left for IPC_RMID is to immediately
808destroy an unattached segment.  Of course, this breaks the way things are
809defined, so some applications might stop working.  Note that this
810feature will do you no good unless you also configure your resource
811limits (in particular, RLIMIT_AS and RLIMIT_NPROC).  Most systems don't
812need this.
813
814Note that if you change this from 0 to 1, already created segments
815without users and with a dead originative process will be destroyed.
816
817==============================================================
818
819sysctl_writes_strict:
820
821Control how file position affects the behavior of updating sysctl values
822via the /proc/sys interface:
823
824  -1 - Legacy per-write sysctl value handling, with no printk warnings.
825       Each write syscall must fully contain the sysctl value to be
826       written, and multiple writes on the same sysctl file descriptor
827       will rewrite the sysctl value, regardless of file position.
828   0 - Same behavior as above, but warn about processes that perform writes
829       to a sysctl file descriptor when the file position is not 0.
830   1 - (default) Respect file position when writing sysctl strings. Multiple
831       writes will append to the sysctl value buffer. Anything past the max
832       length of the sysctl value buffer will be ignored. Writes to numeric
833       sysctl entries must always be at file position 0 and the value must
834       be fully contained in the buffer sent in the write syscall.
835
836==============================================================
837
838softlockup_all_cpu_backtrace:
839
840This value controls the soft lockup detector thread's behavior
841when a soft lockup condition is detected as to whether or not
842to gather further debug information. If enabled, each cpu will
843be issued an NMI and instructed to capture stack trace.
844
845This feature is only applicable for architectures which support
846NMI.
847
8480: do nothing. This is the default behavior.
849
8501: on detection capture more debug information.
851
852==============================================================
853
854soft_watchdog
855
856This parameter can be used to control the soft lockup detector.
857
858   0 - disable the soft lockup detector
859   1 - enable the soft lockup detector
860
861The soft lockup detector monitors CPUs for threads that are hogging the CPUs
862without rescheduling voluntarily, and thus prevent the 'watchdog/N' threads
863from running. The mechanism depends on the CPUs ability to respond to timer
864interrupts which are needed for the 'watchdog/N' threads to be woken up by
865the watchdog timer function, otherwise the NMI watchdog - if enabled - can
866detect a hard lockup condition.
867
868==============================================================
869
870tainted:
871
872Non-zero if the kernel has been tainted.  Numeric values, which
873can be ORed together:
874
875   1 - A module with a non-GPL license has been loaded, this
876       includes modules with no license.
877       Set by modutils >= 2.4.9 and module-init-tools.
878   2 - A module was force loaded by insmod -f.
879       Set by modutils >= 2.4.9 and module-init-tools.
880   4 - Unsafe SMP processors: SMP with CPUs not designed for SMP.
881   8 - A module was forcibly unloaded from the system by rmmod -f.
882  16 - A hardware machine check error occurred on the system.
883  32 - A bad page was discovered on the system.
884  64 - The user has asked that the system be marked "tainted".  This
885       could be because they are running software that directly modifies
886       the hardware, or for other reasons.
887 128 - The system has died.
888 256 - The ACPI DSDT has been overridden with one supplied by the user
889        instead of using the one provided by the hardware.
890 512 - A kernel warning has occurred.
8911024 - A module from drivers/staging was loaded.
8922048 - The system is working around a severe firmware bug.
8934096 - An out-of-tree module has been loaded.
8948192 - An unsigned module has been loaded in a kernel supporting module
895       signature.
89616384 - A soft lockup has previously occurred on the system.
89732768 - The kernel has been live patched.
898
899==============================================================
900
901threads-max
902
903This value controls the maximum number of threads that can be created
904using fork().
905
906During initialization the kernel sets this value such that even if the
907maximum number of threads is created, the thread structures occupy only
908a part (1/8th) of the available RAM pages.
909
910The minimum value that can be written to threads-max is 20.
911The maximum value that can be written to threads-max is given by the
912constant FUTEX_TID_MASK (0x3fffffff).
913If a value outside of this range is written to threads-max an error
914EINVAL occurs.
915
916The value written is checked against the available RAM pages. If the
917thread structures would occupy too much (more than 1/8th) of the
918available RAM pages threads-max is reduced accordingly.
919
920==============================================================
921
922unknown_nmi_panic:
923
924The value in this file affects behavior of handling NMI. When the
925value is non-zero, unknown NMI is trapped and then panic occurs. At
926that time, kernel debugging information is displayed on console.
927
928NMI switch that most IA32 servers have fires unknown NMI up, for
929example.  If a system hangs up, try pressing the NMI switch.
930
931==============================================================
932
933watchdog:
934
935This parameter can be used to disable or enable the soft lockup detector
936_and_ the NMI watchdog (i.e. the hard lockup detector) at the same time.
937
938   0 - disable both lockup detectors
939   1 - enable both lockup detectors
940
941The soft lockup detector and the NMI watchdog can also be disabled or
942enabled individually, using the soft_watchdog and nmi_watchdog parameters.
943If the watchdog parameter is read, for example by executing
944
945   cat /proc/sys/kernel/watchdog
946
947the output of this command (0 or 1) shows the logical OR of soft_watchdog
948and nmi_watchdog.
949
950==============================================================
951
952watchdog_cpumask:
953
954This value can be used to control on which cpus the watchdog may run.
955The default cpumask is all possible cores, but if NO_HZ_FULL is
956enabled in the kernel config, and cores are specified with the
957nohz_full= boot argument, those cores are excluded by default.
958Offline cores can be included in this mask, and if the core is later
959brought online, the watchdog will be started based on the mask value.
960
961Typically this value would only be touched in the nohz_full case
962to re-enable cores that by default were not running the watchdog,
963if a kernel lockup was suspected on those cores.
964
965The argument value is the standard cpulist format for cpumasks,
966so for example to enable the watchdog on cores 0, 2, 3, and 4 you
967might say:
968
969  echo 0,2-4 > /proc/sys/kernel/watchdog_cpumask
970
971==============================================================
972
973watchdog_thresh:
974
975This value can be used to control the frequency of hrtimer and NMI
976events and the soft and hard lockup thresholds. The default threshold
977is 10 seconds.
978
979The softlockup threshold is (2 * watchdog_thresh). Setting this
980tunable to zero will disable lockup detection altogether.
981
982==============================================================
983