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- panic_on_rcu_stall 62- perf_cpu_time_max_percent 63- perf_event_paranoid 64- perf_event_max_stack 65- perf_event_max_contexts_per_stack 66- pid_max 67- powersave-nap [ PPC only ] 68- printk 69- printk_delay 70- printk_ratelimit 71- printk_ratelimit_burst 72- pty ==> Documentation/filesystems/devpts.txt 73- randomize_va_space 74- real-root-dev ==> Documentation/initrd.txt 75- reboot-cmd [ SPARC only ] 76- rtsig-max 77- rtsig-nr 78- sem 79- sem_next_id [ sysv ipc ] 80- sg-big-buff [ generic SCSI device (sg) ] 81- shm_next_id [ sysv ipc ] 82- shm_rmid_forced 83- shmall 84- shmmax [ sysv ipc ] 85- shmmni 86- softlockup_all_cpu_backtrace 87- soft_watchdog 88- stop-a [ SPARC only ] 89- sysrq ==> Documentation/sysrq.txt 90- sysctl_writes_strict 91- tainted 92- threads-max 93- unknown_nmi_panic 94- watchdog 95- watchdog_thresh 96- version 97 98============================================================== 99 100acct: 101 102highwater lowwater frequency 103 104If BSD-style process accounting is enabled these values control 105its behaviour. If free space on filesystem where the log lives 106goes below <lowwater>% accounting suspends. If free space gets 107above <highwater>% accounting resumes. <Frequency> determines 108how often do we check the amount of free space (value is in 109seconds). Default: 1104 2 30 111That is, suspend accounting if there left <= 2% free; resume it 112if we got >=4%; consider information about amount of free space 113valid for 30 seconds. 114 115============================================================== 116 117acpi_video_flags: 118 119flags 120 121See Doc*/kernel/power/video.txt, it allows mode of video boot to be 122set during run time. 123 124============================================================== 125 126auto_msgmni: 127 128This variable has no effect and may be removed in future kernel 129releases. Reading it always returns 0. 130Up to Linux 3.17, it enabled/disabled automatic recomputing of msgmni 131upon memory add/remove or upon ipc namespace creation/removal. 132Echoing "1" into this file enabled msgmni automatic recomputing. 133Echoing "0" turned it off. auto_msgmni default value was 1. 134 135 136============================================================== 137 138bootloader_type: 139 140x86 bootloader identification 141 142This gives the bootloader type number as indicated by the bootloader, 143shifted left by 4, and OR'd with the low four bits of the bootloader 144version. The reason for this encoding is that this used to match the 145type_of_loader field in the kernel header; the encoding is kept for 146backwards compatibility. That is, if the full bootloader type number 147is 0x15 and the full version number is 0x234, this file will contain 148the value 340 = 0x154. 149 150See the type_of_loader and ext_loader_type fields in 151Documentation/x86/boot.txt for additional information. 152 153============================================================== 154 155bootloader_version: 156 157x86 bootloader version 158 159The complete bootloader version number. In the example above, this 160file will contain the value 564 = 0x234. 161 162See the type_of_loader and ext_loader_ver fields in 163Documentation/x86/boot.txt for additional information. 164 165============================================================== 166 167callhome: 168 169Controls the kernel's callhome behavior in case of a kernel panic. 170 171The s390 hardware allows an operating system to send a notification 172to a service organization (callhome) in case of an operating system panic. 173 174When the value in this file is 0 (which is the default behavior) 175nothing happens in case of a kernel panic. If this value is set to "1" 176the complete kernel oops message is send to the IBM customer service 177organization in case the mainframe the Linux operating system is running 178on has a service contract with IBM. 179 180============================================================== 181 182cap_last_cap 183 184Highest valid capability of the running kernel. Exports 185CAP_LAST_CAP from the kernel. 186 187============================================================== 188 189core_pattern: 190 191core_pattern is used to specify a core dumpfile pattern name. 192. max length 128 characters; default value is "core" 193. core_pattern is used as a pattern template for the output filename; 194 certain string patterns (beginning with '%') are substituted with 195 their actual values. 196. backward compatibility with core_uses_pid: 197 If core_pattern does not include "%p" (default does not) 198 and core_uses_pid is set, then .PID will be appended to 199 the filename. 200. corename format specifiers: 201 %<NUL> '%' is dropped 202 %% output one '%' 203 %p pid 204 %P global pid (init PID namespace) 205 %i tid 206 %I global tid (init PID namespace) 207 %u uid (in initial user namespace) 208 %g gid (in initial user namespace) 209 %d dump mode, matches PR_SET_DUMPABLE and 210 /proc/sys/fs/suid_dumpable 211 %s signal number 212 %t UNIX time of dump 213 %h hostname 214 %e executable filename (may be shortened) 215 %E executable path 216 %<OTHER> both are dropped 217. If the first character of the pattern is a '|', the kernel will treat 218 the rest of the pattern as a command to run. The core dump will be 219 written to the standard input of that program instead of to a file. 220 221============================================================== 222 223core_pipe_limit: 224 225This sysctl is only applicable when core_pattern is configured to pipe 226core files to a user space helper (when the first character of 227core_pattern is a '|', see above). When collecting cores via a pipe 228to an application, it is occasionally useful for the collecting 229application to gather data about the crashing process from its 230/proc/pid directory. In order to do this safely, the kernel must wait 231for the collecting process to exit, so as not to remove the crashing 232processes proc files prematurely. This in turn creates the 233possibility that a misbehaving userspace collecting process can block 234the reaping of a crashed process simply by never exiting. This sysctl 235defends against that. It defines how many concurrent crashing 236processes may be piped to user space applications in parallel. If 237this value is exceeded, then those crashing processes above that value 238are noted via the kernel log and their cores are skipped. 0 is a 239special value, indicating that unlimited processes may be captured in 240parallel, but that no waiting will take place (i.e. the collecting 241process is not guaranteed access to /proc/<crashing pid>/). This 242value defaults to 0. 243 244============================================================== 245 246core_uses_pid: 247 248The default coredump filename is "core". By setting 249core_uses_pid to 1, the coredump filename becomes core.PID. 250If core_pattern does not include "%p" (default does not) 251and core_uses_pid is set, then .PID will be appended to 252the filename. 253 254============================================================== 255 256ctrl-alt-del: 257 258When the value in this file is 0, ctrl-alt-del is trapped and 259sent to the init(1) program to handle a graceful restart. 260When, however, the value is > 0, Linux's reaction to a Vulcan 261Nerve Pinch (tm) will be an immediate reboot, without even 262syncing its dirty buffers. 263 264Note: when a program (like dosemu) has the keyboard in 'raw' 265mode, the ctrl-alt-del is intercepted by the program before it 266ever reaches the kernel tty layer, and it's up to the program 267to decide what to do with it. 268 269============================================================== 270 271dmesg_restrict: 272 273This toggle indicates whether unprivileged users are prevented 274from using dmesg(8) to view messages from the kernel's log buffer. 275When dmesg_restrict is set to (0) there are no restrictions. When 276dmesg_restrict is set set to (1), users must have CAP_SYSLOG to use 277dmesg(8). 278 279The kernel config option CONFIG_SECURITY_DMESG_RESTRICT sets the 280default value of dmesg_restrict. 281 282============================================================== 283 284domainname & hostname: 285 286These files can be used to set the NIS/YP domainname and the 287hostname of your box in exactly the same way as the commands 288domainname and hostname, i.e.: 289# echo "darkstar" > /proc/sys/kernel/hostname 290# echo "mydomain" > /proc/sys/kernel/domainname 291has the same effect as 292# hostname "darkstar" 293# domainname "mydomain" 294 295Note, however, that the classic darkstar.frop.org has the 296hostname "darkstar" and DNS (Internet Domain Name Server) 297domainname "frop.org", not to be confused with the NIS (Network 298Information Service) or YP (Yellow Pages) domainname. These two 299domain names are in general different. For a detailed discussion 300see the hostname(1) man page. 301 302============================================================== 303hardlockup_all_cpu_backtrace: 304 305This value controls the hard lockup detector behavior when a hard 306lockup condition is detected as to whether or not to gather further 307debug information. If enabled, arch-specific all-CPU stack dumping 308will be initiated. 309 3100: do nothing. This is the default behavior. 311 3121: on detection capture more debug information. 313============================================================== 314 315hotplug: 316 317Path for the hotplug policy agent. 318Default value is "/sbin/hotplug". 319 320============================================================== 321 322hung_task_panic: 323 324Controls the kernel's behavior when a hung task is detected. 325This file shows up if CONFIG_DETECT_HUNG_TASK is enabled. 326 3270: continue operation. This is the default behavior. 328 3291: panic immediately. 330 331============================================================== 332 333hung_task_check_count: 334 335The upper bound on the number of tasks that are checked. 336This file shows up if CONFIG_DETECT_HUNG_TASK is enabled. 337 338============================================================== 339 340hung_task_timeout_secs: 341 342Check interval. When a task in D state did not get scheduled 343for more than this value report a warning. 344This file shows up if CONFIG_DETECT_HUNG_TASK is enabled. 345 3460: means infinite timeout - no checking done. 347Possible values to set are in range {0..LONG_MAX/HZ}. 348 349============================================================== 350 351hung_task_warnings: 352 353The maximum number of warnings to report. During a check interval 354if a hung task is detected, this value is decreased by 1. 355When this value reaches 0, no more warnings will be reported. 356This file shows up if CONFIG_DETECT_HUNG_TASK is enabled. 357 358-1: report an infinite number of warnings. 359 360============================================================== 361 362kexec_load_disabled: 363 364A toggle indicating if the kexec_load syscall has been disabled. This 365value defaults to 0 (false: kexec_load enabled), but can be set to 1 366(true: kexec_load disabled). Once true, kexec can no longer be used, and 367the toggle cannot be set back to false. This allows a kexec image to be 368loaded before disabling the syscall, allowing a system to set up (and 369later use) an image without it being altered. Generally used together 370with the "modules_disabled" sysctl. 371 372============================================================== 373 374kptr_restrict: 375 376This toggle indicates whether restrictions are placed on 377exposing kernel addresses via /proc and other interfaces. 378 379When kptr_restrict is set to (0), the default, there are no restrictions. 380 381When kptr_restrict is set to (1), kernel pointers printed using the %pK 382format specifier will be replaced with 0's unless the user has CAP_SYSLOG 383and effective user and group ids are equal to the real ids. This is 384because %pK checks are done at read() time rather than open() time, so 385if permissions are elevated between the open() and the read() (e.g via 386a setuid binary) then %pK will not leak kernel pointers to unprivileged 387users. Note, this is a temporary solution only. The correct long-term 388solution is to do the permission checks at open() time. Consider removing 389world read permissions from files that use %pK, and using dmesg_restrict 390to protect against uses of %pK in dmesg(8) if leaking kernel pointer 391values to unprivileged users is a concern. 392 393When kptr_restrict is set to (2), kernel pointers printed using 394%pK will be replaced with 0's regardless of privileges. 395 396============================================================== 397 398kstack_depth_to_print: (X86 only) 399 400Controls the number of words to print when dumping the raw 401kernel stack. 402 403============================================================== 404 405l2cr: (PPC only) 406 407This flag controls the L2 cache of G3 processor boards. If 4080, the cache is disabled. Enabled if nonzero. 409 410============================================================== 411 412modules_disabled: 413 414A toggle value indicating if modules are allowed to be loaded 415in an otherwise modular kernel. This toggle defaults to off 416(0), but can be set true (1). Once true, modules can be 417neither loaded nor unloaded, and the toggle cannot be set back 418to false. Generally used with the "kexec_load_disabled" toggle. 419 420============================================================== 421 422msg_next_id, sem_next_id, and shm_next_id: 423 424These three toggles allows to specify desired id for next allocated IPC 425object: message, semaphore or shared memory respectively. 426 427By default they are equal to -1, which means generic allocation logic. 428Possible values to set are in range {0..INT_MAX}. 429 430Notes: 4311) kernel doesn't guarantee, that new object will have desired id. So, 432it's up to userspace, how to handle an object with "wrong" id. 4332) Toggle with non-default value will be set back to -1 by kernel after 434successful IPC object allocation. 435 436============================================================== 437 438nmi_watchdog: 439 440This parameter can be used to control the NMI watchdog 441(i.e. the hard lockup detector) on x86 systems. 442 443 0 - disable the hard lockup detector 444 1 - enable the hard lockup detector 445 446The hard lockup detector monitors each CPU for its ability to respond to 447timer interrupts. The mechanism utilizes CPU performance counter registers 448that are programmed to generate Non-Maskable Interrupts (NMIs) periodically 449while a CPU is busy. Hence, the alternative name 'NMI watchdog'. 450 451The NMI watchdog is disabled by default if the kernel is running as a guest 452in a KVM virtual machine. This default can be overridden by adding 453 454 nmi_watchdog=1 455 456to the guest kernel command line (see Documentation/kernel-parameters.txt). 457 458============================================================== 459 460numa_balancing 461 462Enables/disables automatic page fault based NUMA memory 463balancing. Memory is moved automatically to nodes 464that access it often. 465 466Enables/disables automatic NUMA memory balancing. On NUMA machines, there 467is a performance penalty if remote memory is accessed by a CPU. When this 468feature is enabled the kernel samples what task thread is accessing memory 469by periodically unmapping pages and later trapping a page fault. At the 470time of the page fault, it is determined if the data being accessed should 471be migrated to a local memory node. 472 473The unmapping of pages and trapping faults incur additional overhead that 474ideally is offset by improved memory locality but there is no universal 475guarantee. If the target workload is already bound to NUMA nodes then this 476feature should be disabled. Otherwise, if the system overhead from the 477feature is too high then the rate the kernel samples for NUMA hinting 478faults may be controlled by the numa_balancing_scan_period_min_ms, 479numa_balancing_scan_delay_ms, numa_balancing_scan_period_max_ms, 480numa_balancing_scan_size_mb, and numa_balancing_settle_count sysctls. 481 482============================================================== 483 484numa_balancing_scan_period_min_ms, numa_balancing_scan_delay_ms, 485numa_balancing_scan_period_max_ms, numa_balancing_scan_size_mb 486 487Automatic NUMA balancing scans tasks address space and unmaps pages to 488detect if pages are properly placed or if the data should be migrated to a 489memory node local to where the task is running. Every "scan delay" the task 490scans the next "scan size" number of pages in its address space. When the 491end of the address space is reached the scanner restarts from the beginning. 492 493In combination, the "scan delay" and "scan size" determine the scan rate. 494When "scan delay" decreases, the scan rate increases. The scan delay and 495hence the scan rate of every task is adaptive and depends on historical 496behaviour. If pages are properly placed then the scan delay increases, 497otherwise the scan delay decreases. The "scan size" is not adaptive but 498the higher the "scan size", the higher the scan rate. 499 500Higher scan rates incur higher system overhead as page faults must be 501trapped and potentially data must be migrated. However, the higher the scan 502rate, the more quickly a tasks memory is migrated to a local node if the 503workload pattern changes and minimises performance impact due to remote 504memory accesses. These sysctls control the thresholds for scan delays and 505the number of pages scanned. 506 507numa_balancing_scan_period_min_ms is the minimum time in milliseconds to 508scan a tasks virtual memory. It effectively controls the maximum scanning 509rate for each task. 510 511numa_balancing_scan_delay_ms is the starting "scan delay" used for a task 512when it initially forks. 513 514numa_balancing_scan_period_max_ms is the maximum time in milliseconds to 515scan a tasks virtual memory. It effectively controls the minimum scanning 516rate for each task. 517 518numa_balancing_scan_size_mb is how many megabytes worth of pages are 519scanned for a given scan. 520 521============================================================== 522 523osrelease, ostype & version: 524 525# cat osrelease 5262.1.88 527# cat ostype 528Linux 529# cat version 530#5 Wed Feb 25 21:49:24 MET 1998 531 532The files osrelease and ostype should be clear enough. Version 533needs a little more clarification however. The '#5' means that 534this is the fifth kernel built from this source base and the 535date behind it indicates the time the kernel was built. 536The only way to tune these values is to rebuild the kernel :-) 537 538============================================================== 539 540overflowgid & overflowuid: 541 542if your architecture did not always support 32-bit UIDs (i.e. arm, 543i386, m68k, sh, and sparc32), a fixed UID and GID will be returned to 544applications that use the old 16-bit UID/GID system calls, if the 545actual UID or GID would exceed 65535. 546 547These sysctls allow you to change the value of the fixed UID and GID. 548The default is 65534. 549 550============================================================== 551 552panic: 553 554The value in this file represents the number of seconds the kernel 555waits before rebooting on a panic. When you use the software watchdog, 556the recommended setting is 60. 557 558============================================================== 559 560panic_on_io_nmi: 561 562Controls the kernel's behavior when a CPU receives an NMI caused by 563an IO error. 564 5650: try to continue operation (default) 566 5671: panic immediately. The IO error triggered an NMI. This indicates a 568 serious system condition which could result in IO data corruption. 569 Rather than continuing, panicking might be a better choice. Some 570 servers issue this sort of NMI when the dump button is pushed, 571 and you can use this option to take a crash dump. 572 573============================================================== 574 575panic_on_oops: 576 577Controls the kernel's behaviour when an oops or BUG is encountered. 578 5790: try to continue operation 580 5811: panic immediately. If the `panic' sysctl is also non-zero then the 582 machine will be rebooted. 583 584============================================================== 585 586panic_on_stackoverflow: 587 588Controls the kernel's behavior when detecting the overflows of 589kernel, IRQ and exception stacks except a user stack. 590This file shows up if CONFIG_DEBUG_STACKOVERFLOW is enabled. 591 5920: try to continue operation. 593 5941: panic immediately. 595 596============================================================== 597 598panic_on_unrecovered_nmi: 599 600The default Linux behaviour on an NMI of either memory or unknown is 601to continue operation. For many environments such as scientific 602computing it is preferable that the box is taken out and the error 603dealt with than an uncorrected parity/ECC error get propagated. 604 605A small number of systems do generate NMI's for bizarre random reasons 606such as power management so the default is off. That sysctl works like 607the existing panic controls already in that directory. 608 609============================================================== 610 611panic_on_warn: 612 613Calls panic() in the WARN() path when set to 1. This is useful to avoid 614a kernel rebuild when attempting to kdump at the location of a WARN(). 615 6160: only WARN(), default behaviour. 617 6181: call panic() after printing out WARN() location. 619 620============================================================== 621 622panic_on_rcu_stall: 623 624When set to 1, calls panic() after RCU stall detection messages. This 625is useful to define the root cause of RCU stalls using a vmcore. 626 6270: do not panic() when RCU stall takes place, default behavior. 628 6291: panic() after printing RCU stall messages. 630 631============================================================== 632 633perf_cpu_time_max_percent: 634 635Hints to the kernel how much CPU time it should be allowed to 636use to handle perf sampling events. If the perf subsystem 637is informed that its samples are exceeding this limit, it 638will drop its sampling frequency to attempt to reduce its CPU 639usage. 640 641Some perf sampling happens in NMIs. If these samples 642unexpectedly take too long to execute, the NMIs can become 643stacked up next to each other so much that nothing else is 644allowed to execute. 645 6460: disable the mechanism. Do not monitor or correct perf's 647 sampling rate no matter how CPU time it takes. 648 6491-100: attempt to throttle perf's sample rate to this 650 percentage of CPU. Note: the kernel calculates an 651 "expected" length of each sample event. 100 here means 652 100% of that expected length. Even if this is set to 653 100, you may still see sample throttling if this 654 length is exceeded. Set to 0 if you truly do not care 655 how much CPU is consumed. 656 657============================================================== 658 659perf_event_paranoid: 660 661Controls use of the performance events system by unprivileged 662users (without CAP_SYS_ADMIN). The default value is 3 if 663CONFIG_SECURITY_PERF_EVENTS_RESTRICT is set, or 2 otherwise. 664 665 -1: Allow use of (almost) all events by all users 666>=0: Disallow raw tracepoint access by users without CAP_IOC_LOCK 667>=1: Disallow CPU event access by users without CAP_SYS_ADMIN 668>=2: Disallow kernel profiling by users without CAP_SYS_ADMIN 669>=3: Disallow all event access by users without CAP_SYS_ADMIN 670 671============================================================== 672 673perf_event_max_stack: 674 675Controls maximum number of stack frames to copy for (attr.sample_type & 676PERF_SAMPLE_CALLCHAIN) configured events, for instance, when using 677'perf record -g' or 'perf trace --call-graph fp'. 678 679This can only be done when no events are in use that have callchains 680enabled, otherwise writing to this file will return -EBUSY. 681 682The default value is 127. 683 684============================================================== 685 686perf_event_max_contexts_per_stack: 687 688Controls maximum number of stack frame context entries for 689(attr.sample_type & PERF_SAMPLE_CALLCHAIN) configured events, for 690instance, when using 'perf record -g' or 'perf trace --call-graph fp'. 691 692This can only be done when no events are in use that have callchains 693enabled, otherwise writing to this file will return -EBUSY. 694 695The default value is 8. 696 697============================================================== 698 699pid_max: 700 701PID allocation wrap value. When the kernel's next PID value 702reaches this value, it wraps back to a minimum PID value. 703PIDs of value pid_max or larger are not allocated. 704 705============================================================== 706 707ns_last_pid: 708 709The last pid allocated in the current (the one task using this sysctl 710lives in) pid namespace. When selecting a pid for a next task on fork 711kernel tries to allocate a number starting from this one. 712 713============================================================== 714 715powersave-nap: (PPC only) 716 717If set, Linux-PPC will use the 'nap' mode of powersaving, 718otherwise the 'doze' mode will be used. 719 720============================================================== 721 722printk: 723 724The four values in printk denote: console_loglevel, 725default_message_loglevel, minimum_console_loglevel and 726default_console_loglevel respectively. 727 728These values influence printk() behavior when printing or 729logging error messages. See 'man 2 syslog' for more info on 730the different loglevels. 731 732- console_loglevel: messages with a higher priority than 733 this will be printed to the console 734- default_message_loglevel: messages without an explicit priority 735 will be printed with this priority 736- minimum_console_loglevel: minimum (highest) value to which 737 console_loglevel can be set 738- default_console_loglevel: default value for console_loglevel 739 740============================================================== 741 742printk_delay: 743 744Delay each printk message in printk_delay milliseconds 745 746Value from 0 - 10000 is allowed. 747 748============================================================== 749 750printk_ratelimit: 751 752Some warning messages are rate limited. printk_ratelimit specifies 753the minimum length of time between these messages (in jiffies), by 754default we allow one every 5 seconds. 755 756A value of 0 will disable rate limiting. 757 758============================================================== 759 760printk_ratelimit_burst: 761 762While long term we enforce one message per printk_ratelimit 763seconds, we do allow a burst of messages to pass through. 764printk_ratelimit_burst specifies the number of messages we can 765send before ratelimiting kicks in. 766 767============================================================== 768 769printk_devkmsg: 770 771Control the logging to /dev/kmsg from userspace: 772 773ratelimit: default, ratelimited 774on: unlimited logging to /dev/kmsg from userspace 775off: logging to /dev/kmsg disabled 776 777The kernel command line parameter printk.devkmsg= overrides this and is 778a one-time setting until next reboot: once set, it cannot be changed by 779this sysctl interface anymore. 780 781============================================================== 782 783randomize_va_space: 784 785This option can be used to select the type of process address 786space randomization that is used in the system, for architectures 787that support this feature. 788 7890 - Turn the process address space randomization off. This is the 790 default for architectures that do not support this feature anyways, 791 and kernels that are booted with the "norandmaps" parameter. 792 7931 - Make the addresses of mmap base, stack and VDSO page randomized. 794 This, among other things, implies that shared libraries will be 795 loaded to random addresses. Also for PIE-linked binaries, the 796 location of code start is randomized. This is the default if the 797 CONFIG_COMPAT_BRK option is enabled. 798 7992 - Additionally enable heap randomization. This is the default if 800 CONFIG_COMPAT_BRK is disabled. 801 802 There are a few legacy applications out there (such as some ancient 803 versions of libc.so.5 from 1996) that assume that brk area starts 804 just after the end of the code+bss. These applications break when 805 start of the brk area is randomized. There are however no known 806 non-legacy applications that would be broken this way, so for most 807 systems it is safe to choose full randomization. 808 809 Systems with ancient and/or broken binaries should be configured 810 with CONFIG_COMPAT_BRK enabled, which excludes the heap from process 811 address space randomization. 812 813============================================================== 814 815reboot-cmd: (Sparc only) 816 817??? This seems to be a way to give an argument to the Sparc 818ROM/Flash boot loader. Maybe to tell it what to do after 819rebooting. ??? 820 821============================================================== 822 823rtsig-max & rtsig-nr: 824 825The file rtsig-max can be used to tune the maximum number 826of POSIX realtime (queued) signals that can be outstanding 827in the system. 828 829rtsig-nr shows the number of RT signals currently queued. 830 831============================================================== 832 833sched_schedstats: 834 835Enables/disables scheduler statistics. Enabling this feature 836incurs a small amount of overhead in the scheduler but is 837useful for debugging and performance tuning. 838 839============================================================== 840 841sg-big-buff: 842 843This file shows the size of the generic SCSI (sg) buffer. 844You can't tune it just yet, but you could change it on 845compile time by editing include/scsi/sg.h and changing 846the value of SG_BIG_BUFF. 847 848There shouldn't be any reason to change this value. If 849you can come up with one, you probably know what you 850are doing anyway :) 851 852============================================================== 853 854shmall: 855 856This parameter sets the total amount of shared memory pages that 857can be used system wide. Hence, SHMALL should always be at least 858ceil(shmmax/PAGE_SIZE). 859 860If you are not sure what the default PAGE_SIZE is on your Linux 861system, you can run the following command: 862 863# getconf PAGE_SIZE 864 865============================================================== 866 867shmmax: 868 869This value can be used to query and set the run time limit 870on the maximum shared memory segment size that can be created. 871Shared memory segments up to 1Gb are now supported in the 872kernel. This value defaults to SHMMAX. 873 874============================================================== 875 876shm_rmid_forced: 877 878Linux lets you set resource limits, including how much memory one 879process can consume, via setrlimit(2). Unfortunately, shared memory 880segments are allowed to exist without association with any process, and 881thus might not be counted against any resource limits. If enabled, 882shared memory segments are automatically destroyed when their attach 883count becomes zero after a detach or a process termination. It will 884also destroy segments that were created, but never attached to, on exit 885from the process. The only use left for IPC_RMID is to immediately 886destroy an unattached segment. Of course, this breaks the way things are 887defined, so some applications might stop working. Note that this 888feature will do you no good unless you also configure your resource 889limits (in particular, RLIMIT_AS and RLIMIT_NPROC). Most systems don't 890need this. 891 892Note that if you change this from 0 to 1, already created segments 893without users and with a dead originative process will be destroyed. 894 895============================================================== 896 897sysctl_writes_strict: 898 899Control how file position affects the behavior of updating sysctl values 900via the /proc/sys interface: 901 902 -1 - Legacy per-write sysctl value handling, with no printk warnings. 903 Each write syscall must fully contain the sysctl value to be 904 written, and multiple writes on the same sysctl file descriptor 905 will rewrite the sysctl value, regardless of file position. 906 0 - Same behavior as above, but warn about processes that perform writes 907 to a sysctl file descriptor when the file position is not 0. 908 1 - (default) Respect file position when writing sysctl strings. Multiple 909 writes will append to the sysctl value buffer. Anything past the max 910 length of the sysctl value buffer will be ignored. Writes to numeric 911 sysctl entries must always be at file position 0 and the value must 912 be fully contained in the buffer sent in the write syscall. 913 914============================================================== 915 916softlockup_all_cpu_backtrace: 917 918This value controls the soft lockup detector thread's behavior 919when a soft lockup condition is detected as to whether or not 920to gather further debug information. If enabled, each cpu will 921be issued an NMI and instructed to capture stack trace. 922 923This feature is only applicable for architectures which support 924NMI. 925 9260: do nothing. This is the default behavior. 927 9281: on detection capture more debug information. 929 930============================================================== 931 932soft_watchdog 933 934This parameter can be used to control the soft lockup detector. 935 936 0 - disable the soft lockup detector 937 1 - enable the soft lockup detector 938 939The soft lockup detector monitors CPUs for threads that are hogging the CPUs 940without rescheduling voluntarily, and thus prevent the 'watchdog/N' threads 941from running. The mechanism depends on the CPUs ability to respond to timer 942interrupts which are needed for the 'watchdog/N' threads to be woken up by 943the watchdog timer function, otherwise the NMI watchdog - if enabled - can 944detect a hard lockup condition. 945 946============================================================== 947 948tainted: 949 950Non-zero if the kernel has been tainted. Numeric values, which 951can be ORed together: 952 953 1 - A module with a non-GPL license has been loaded, this 954 includes modules with no license. 955 Set by modutils >= 2.4.9 and module-init-tools. 956 2 - A module was force loaded by insmod -f. 957 Set by modutils >= 2.4.9 and module-init-tools. 958 4 - Unsafe SMP processors: SMP with CPUs not designed for SMP. 959 8 - A module was forcibly unloaded from the system by rmmod -f. 960 16 - A hardware machine check error occurred on the system. 961 32 - A bad page was discovered on the system. 962 64 - The user has asked that the system be marked "tainted". This 963 could be because they are running software that directly modifies 964 the hardware, or for other reasons. 965 128 - The system has died. 966 256 - The ACPI DSDT has been overridden with one supplied by the user 967 instead of using the one provided by the hardware. 968 512 - A kernel warning has occurred. 9691024 - A module from drivers/staging was loaded. 9702048 - The system is working around a severe firmware bug. 9714096 - An out-of-tree module has been loaded. 9728192 - An unsigned module has been loaded in a kernel supporting module 973 signature. 97416384 - A soft lockup has previously occurred on the system. 97532768 - The kernel has been live patched. 976 977============================================================== 978 979threads-max 980 981This value controls the maximum number of threads that can be created 982using fork(). 983 984During initialization the kernel sets this value such that even if the 985maximum number of threads is created, the thread structures occupy only 986a part (1/8th) of the available RAM pages. 987 988The minimum value that can be written to threads-max is 20. 989The maximum value that can be written to threads-max is given by the 990constant FUTEX_TID_MASK (0x3fffffff). 991If a value outside of this range is written to threads-max an error 992EINVAL occurs. 993 994The value written is checked against the available RAM pages. If the 995thread structures would occupy too much (more than 1/8th) of the 996available RAM pages threads-max is reduced accordingly. 997 998============================================================== 999 1000unknown_nmi_panic: 1001 1002The value in this file affects behavior of handling NMI. When the 1003value is non-zero, unknown NMI is trapped and then panic occurs. At 1004that time, kernel debugging information is displayed on console. 1005 1006NMI switch that most IA32 servers have fires unknown NMI up, for 1007example. If a system hangs up, try pressing the NMI switch. 1008 1009============================================================== 1010 1011watchdog: 1012 1013This parameter can be used to disable or enable the soft lockup detector 1014_and_ the NMI watchdog (i.e. the hard lockup detector) at the same time. 1015 1016 0 - disable both lockup detectors 1017 1 - enable both lockup detectors 1018 1019The soft lockup detector and the NMI watchdog can also be disabled or 1020enabled individually, using the soft_watchdog and nmi_watchdog parameters. 1021If the watchdog parameter is read, for example by executing 1022 1023 cat /proc/sys/kernel/watchdog 1024 1025the output of this command (0 or 1) shows the logical OR of soft_watchdog 1026and nmi_watchdog. 1027 1028============================================================== 1029 1030watchdog_cpumask: 1031 1032This value can be used to control on which cpus the watchdog may run. 1033The default cpumask is all possible cores, but if NO_HZ_FULL is 1034enabled in the kernel config, and cores are specified with the 1035nohz_full= boot argument, those cores are excluded by default. 1036Offline cores can be included in this mask, and if the core is later 1037brought online, the watchdog will be started based on the mask value. 1038 1039Typically this value would only be touched in the nohz_full case 1040to re-enable cores that by default were not running the watchdog, 1041if a kernel lockup was suspected on those cores. 1042 1043The argument value is the standard cpulist format for cpumasks, 1044so for example to enable the watchdog on cores 0, 2, 3, and 4 you 1045might say: 1046 1047 echo 0,2-4 > /proc/sys/kernel/watchdog_cpumask 1048 1049============================================================== 1050 1051watchdog_thresh: 1052 1053This value can be used to control the frequency of hrtimer and NMI 1054events and the soft and hard lockup thresholds. The default threshold 1055is 10 seconds. 1056 1057The softlockup threshold is (2 * watchdog_thresh). Setting this 1058tunable to zero will disable lockup detection altogether. 1059 1060============================================================== 1061