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- pid_max 62- powersave-nap [ PPC only ] 63- printk 64- printk_delay 65- printk_ratelimit 66- printk_ratelimit_burst 67- randomize_va_space 68- real-root-dev ==> Documentation/initrd.txt 69- reboot-cmd [ SPARC only ] 70- rtsig-max 71- rtsig-nr 72- sem 73- sem_next_id [ sysv ipc ] 74- sg-big-buff [ generic SCSI device (sg) ] 75- shm_next_id [ sysv ipc ] 76- shm_rmid_forced 77- shmall 78- shmmax [ sysv ipc ] 79- shmmni 80- softlockup_all_cpu_backtrace 81- soft_watchdog 82- stop-a [ SPARC only ] 83- sysrq ==> Documentation/sysrq.txt 84- sysctl_writes_strict 85- tainted 86- threads-max 87- unknown_nmi_panic 88- watchdog 89- watchdog_thresh 90- version 91 92============================================================== 93 94acct: 95 96highwater lowwater frequency 97 98If BSD-style process accounting is enabled these values control 99its behaviour. If free space on filesystem where the log lives 100goes below <lowwater>% accounting suspends. If free space gets 101above <highwater>% accounting resumes. <Frequency> determines 102how often do we check the amount of free space (value is in 103seconds). Default: 1044 2 30 105That is, suspend accounting if there left <= 2% free; resume it 106if we got >=4%; consider information about amount of free space 107valid for 30 seconds. 108 109============================================================== 110 111acpi_video_flags: 112 113flags 114 115See Doc*/kernel/power/video.txt, it allows mode of video boot to be 116set during run time. 117 118============================================================== 119 120auto_msgmni: 121 122This variable has no effect and may be removed in future kernel 123releases. Reading it always returns 0. 124Up to Linux 3.17, it enabled/disabled automatic recomputing of msgmni 125upon memory add/remove or upon ipc namespace creation/removal. 126Echoing "1" into this file enabled msgmni automatic recomputing. 127Echoing "0" turned it off. auto_msgmni default value was 1. 128 129 130============================================================== 131 132bootloader_type: 133 134x86 bootloader identification 135 136This gives the bootloader type number as indicated by the bootloader, 137shifted left by 4, and OR'd with the low four bits of the bootloader 138version. The reason for this encoding is that this used to match the 139type_of_loader field in the kernel header; the encoding is kept for 140backwards compatibility. That is, if the full bootloader type number 141is 0x15 and the full version number is 0x234, this file will contain 142the value 340 = 0x154. 143 144See the type_of_loader and ext_loader_type fields in 145Documentation/x86/boot.txt for additional information. 146 147============================================================== 148 149bootloader_version: 150 151x86 bootloader version 152 153The complete bootloader version number. In the example above, this 154file will contain the value 564 = 0x234. 155 156See the type_of_loader and ext_loader_ver fields in 157Documentation/x86/boot.txt for additional information. 158 159============================================================== 160 161callhome: 162 163Controls the kernel's callhome behavior in case of a kernel panic. 164 165The s390 hardware allows an operating system to send a notification 166to a service organization (callhome) in case of an operating system panic. 167 168When the value in this file is 0 (which is the default behavior) 169nothing happens in case of a kernel panic. If this value is set to "1" 170the complete kernel oops message is send to the IBM customer service 171organization in case the mainframe the Linux operating system is running 172on has a service contract with IBM. 173 174============================================================== 175 176cap_last_cap 177 178Highest valid capability of the running kernel. Exports 179CAP_LAST_CAP from the kernel. 180 181============================================================== 182 183core_pattern: 184 185core_pattern is used to specify a core dumpfile pattern name. 186. max length 128 characters; default value is "core" 187. core_pattern is used as a pattern template for the output filename; 188 certain string patterns (beginning with '%') are substituted with 189 their actual values. 190. backward compatibility with core_uses_pid: 191 If core_pattern does not include "%p" (default does not) 192 and core_uses_pid is set, then .PID will be appended to 193 the filename. 194. corename format specifiers: 195 %<NUL> '%' is dropped 196 %% output one '%' 197 %p pid 198 %P global pid (init PID namespace) 199 %i tid 200 %I global tid (init PID namespace) 201 %u uid (in initial user namespace) 202 %g gid (in initial user namespace) 203 %d dump mode, matches PR_SET_DUMPABLE and 204 /proc/sys/fs/suid_dumpable 205 %s signal number 206 %t UNIX time of dump 207 %h hostname 208 %e executable filename (may be shortened) 209 %E executable path 210 %<OTHER> both are dropped 211. If the first character of the pattern is a '|', the kernel will treat 212 the rest of the pattern as a command to run. The core dump will be 213 written to the standard input of that program instead of to a file. 214 215============================================================== 216 217core_pipe_limit: 218 219This sysctl is only applicable when core_pattern is configured to pipe 220core files to a user space helper (when the first character of 221core_pattern is a '|', see above). When collecting cores via a pipe 222to an application, it is occasionally useful for the collecting 223application to gather data about the crashing process from its 224/proc/pid directory. In order to do this safely, the kernel must wait 225for the collecting process to exit, so as not to remove the crashing 226processes proc files prematurely. This in turn creates the 227possibility that a misbehaving userspace collecting process can block 228the reaping of a crashed process simply by never exiting. This sysctl 229defends against that. It defines how many concurrent crashing 230processes may be piped to user space applications in parallel. If 231this value is exceeded, then those crashing processes above that value 232are noted via the kernel log and their cores are skipped. 0 is a 233special value, indicating that unlimited processes may be captured in 234parallel, but that no waiting will take place (i.e. the collecting 235process is not guaranteed access to /proc/<crashing pid>/). This 236value defaults to 0. 237 238============================================================== 239 240core_uses_pid: 241 242The default coredump filename is "core". By setting 243core_uses_pid to 1, the coredump filename becomes core.PID. 244If core_pattern does not include "%p" (default does not) 245and core_uses_pid is set, then .PID will be appended to 246the filename. 247 248============================================================== 249 250ctrl-alt-del: 251 252When the value in this file is 0, ctrl-alt-del is trapped and 253sent to the init(1) program to handle a graceful restart. 254When, however, the value is > 0, Linux's reaction to a Vulcan 255Nerve Pinch (tm) will be an immediate reboot, without even 256syncing its dirty buffers. 257 258Note: when a program (like dosemu) has the keyboard in 'raw' 259mode, the ctrl-alt-del is intercepted by the program before it 260ever reaches the kernel tty layer, and it's up to the program 261to decide what to do with it. 262 263============================================================== 264 265dmesg_restrict: 266 267This toggle indicates whether unprivileged users are prevented 268from using dmesg(8) to view messages from the kernel's log buffer. 269When dmesg_restrict is set to (0) there are no restrictions. When 270dmesg_restrict is set set to (1), users must have CAP_SYSLOG to use 271dmesg(8). 272 273The kernel config option CONFIG_SECURITY_DMESG_RESTRICT sets the 274default value of dmesg_restrict. 275 276============================================================== 277 278domainname & hostname: 279 280These files can be used to set the NIS/YP domainname and the 281hostname of your box in exactly the same way as the commands 282domainname and hostname, i.e.: 283# echo "darkstar" > /proc/sys/kernel/hostname 284# echo "mydomain" > /proc/sys/kernel/domainname 285has the same effect as 286# hostname "darkstar" 287# domainname "mydomain" 288 289Note, however, that the classic darkstar.frop.org has the 290hostname "darkstar" and DNS (Internet Domain Name Server) 291domainname "frop.org", not to be confused with the NIS (Network 292Information Service) or YP (Yellow Pages) domainname. These two 293domain names are in general different. For a detailed discussion 294see the hostname(1) man page. 295 296============================================================== 297hardlockup_all_cpu_backtrace: 298 299This value controls the hard lockup detector behavior when a hard 300lockup condition is detected as to whether or not to gather further 301debug information. If enabled, arch-specific all-CPU stack dumping 302will be initiated. 303 3040: do nothing. This is the default behavior. 305 3061: on detection capture more debug information. 307============================================================== 308 309hotplug: 310 311Path for the hotplug policy agent. 312Default value is "/sbin/hotplug". 313 314============================================================== 315 316hung_task_panic: 317 318Controls the kernel's behavior when a hung task is detected. 319This file shows up if CONFIG_DETECT_HUNG_TASK is enabled. 320 3210: continue operation. This is the default behavior. 322 3231: panic immediately. 324 325============================================================== 326 327hung_task_check_count: 328 329The upper bound on the number of tasks that are checked. 330This file shows up if CONFIG_DETECT_HUNG_TASK is enabled. 331 332============================================================== 333 334hung_task_timeout_secs: 335 336Check interval. When a task in D state did not get scheduled 337for more than this value report a warning. 338This file shows up if CONFIG_DETECT_HUNG_TASK is enabled. 339 3400: means infinite timeout - no checking done. 341Possible values to set are in range {0..LONG_MAX/HZ}. 342 343============================================================== 344 345hung_task_warnings: 346 347The maximum number of warnings to report. During a check interval 348if a hung task is detected, this value is decreased by 1. 349When this value reaches 0, no more warnings will be reported. 350This file shows up if CONFIG_DETECT_HUNG_TASK is enabled. 351 352-1: report an infinite number of warnings. 353 354============================================================== 355 356kexec_load_disabled: 357 358A toggle indicating if the kexec_load syscall has been disabled. This 359value defaults to 0 (false: kexec_load enabled), but can be set to 1 360(true: kexec_load disabled). Once true, kexec can no longer be used, and 361the toggle cannot be set back to false. This allows a kexec image to be 362loaded before disabling the syscall, allowing a system to set up (and 363later use) an image without it being altered. Generally used together 364with the "modules_disabled" sysctl. 365 366============================================================== 367 368kptr_restrict: 369 370This toggle indicates whether restrictions are placed on 371exposing kernel addresses via /proc and other interfaces. 372 373When kptr_restrict is set to (0), the default, there are no restrictions. 374 375When kptr_restrict is set to (1), kernel pointers printed using the %pK 376format specifier will be replaced with 0's unless the user has CAP_SYSLOG 377and effective user and group ids are equal to the real ids. This is 378because %pK checks are done at read() time rather than open() time, so 379if permissions are elevated between the open() and the read() (e.g via 380a setuid binary) then %pK will not leak kernel pointers to unprivileged 381users. Note, this is a temporary solution only. The correct long-term 382solution is to do the permission checks at open() time. Consider removing 383world read permissions from files that use %pK, and using dmesg_restrict 384to protect against uses of %pK in dmesg(8) if leaking kernel pointer 385values to unprivileged users is a concern. 386 387When kptr_restrict is set to (2), kernel pointers printed using 388%pK will be replaced with 0's regardless of privileges. 389 390============================================================== 391 392kstack_depth_to_print: (X86 only) 393 394Controls the number of words to print when dumping the raw 395kernel stack. 396 397============================================================== 398 399l2cr: (PPC only) 400 401This flag controls the L2 cache of G3 processor boards. If 4020, the cache is disabled. Enabled if nonzero. 403 404============================================================== 405 406modules_disabled: 407 408A toggle value indicating if modules are allowed to be loaded 409in an otherwise modular kernel. This toggle defaults to off 410(0), but can be set true (1). Once true, modules can be 411neither loaded nor unloaded, and the toggle cannot be set back 412to false. Generally used with the "kexec_load_disabled" toggle. 413 414============================================================== 415 416msg_next_id, sem_next_id, and shm_next_id: 417 418These three toggles allows to specify desired id for next allocated IPC 419object: message, semaphore or shared memory respectively. 420 421By default they are equal to -1, which means generic allocation logic. 422Possible values to set are in range {0..INT_MAX}. 423 424Notes: 4251) kernel doesn't guarantee, that new object will have desired id. So, 426it's up to userspace, how to handle an object with "wrong" id. 4272) Toggle with non-default value will be set back to -1 by kernel after 428successful IPC object allocation. 429 430============================================================== 431 432nmi_watchdog: 433 434This parameter can be used to control the NMI watchdog 435(i.e. the hard lockup detector) on x86 systems. 436 437 0 - disable the hard lockup detector 438 1 - enable the hard lockup detector 439 440The hard lockup detector monitors each CPU for its ability to respond to 441timer interrupts. The mechanism utilizes CPU performance counter registers 442that are programmed to generate Non-Maskable Interrupts (NMIs) periodically 443while a CPU is busy. Hence, the alternative name 'NMI watchdog'. 444 445The NMI watchdog is disabled by default if the kernel is running as a guest 446in a KVM virtual machine. This default can be overridden by adding 447 448 nmi_watchdog=1 449 450to the guest kernel command line (see Documentation/kernel-parameters.txt). 451 452============================================================== 453 454numa_balancing 455 456Enables/disables automatic page fault based NUMA memory 457balancing. Memory is moved automatically to nodes 458that access it often. 459 460Enables/disables automatic NUMA memory balancing. On NUMA machines, there 461is a performance penalty if remote memory is accessed by a CPU. When this 462feature is enabled the kernel samples what task thread is accessing memory 463by periodically unmapping pages and later trapping a page fault. At the 464time of the page fault, it is determined if the data being accessed should 465be migrated to a local memory node. 466 467The unmapping of pages and trapping faults incur additional overhead that 468ideally is offset by improved memory locality but there is no universal 469guarantee. If the target workload is already bound to NUMA nodes then this 470feature should be disabled. Otherwise, if the system overhead from the 471feature is too high then the rate the kernel samples for NUMA hinting 472faults may be controlled by the numa_balancing_scan_period_min_ms, 473numa_balancing_scan_delay_ms, numa_balancing_scan_period_max_ms, 474numa_balancing_scan_size_mb, and numa_balancing_settle_count sysctls. 475 476============================================================== 477 478numa_balancing_scan_period_min_ms, numa_balancing_scan_delay_ms, 479numa_balancing_scan_period_max_ms, numa_balancing_scan_size_mb 480 481Automatic NUMA balancing scans tasks address space and unmaps pages to 482detect if pages are properly placed or if the data should be migrated to a 483memory node local to where the task is running. Every "scan delay" the task 484scans the next "scan size" number of pages in its address space. When the 485end of the address space is reached the scanner restarts from the beginning. 486 487In combination, the "scan delay" and "scan size" determine the scan rate. 488When "scan delay" decreases, the scan rate increases. The scan delay and 489hence the scan rate of every task is adaptive and depends on historical 490behaviour. If pages are properly placed then the scan delay increases, 491otherwise the scan delay decreases. The "scan size" is not adaptive but 492the higher the "scan size", the higher the scan rate. 493 494Higher scan rates incur higher system overhead as page faults must be 495trapped and potentially data must be migrated. However, the higher the scan 496rate, the more quickly a tasks memory is migrated to a local node if the 497workload pattern changes and minimises performance impact due to remote 498memory accesses. These sysctls control the thresholds for scan delays and 499the number of pages scanned. 500 501numa_balancing_scan_period_min_ms is the minimum time in milliseconds to 502scan a tasks virtual memory. It effectively controls the maximum scanning 503rate for each task. 504 505numa_balancing_scan_delay_ms is the starting "scan delay" used for a task 506when it initially forks. 507 508numa_balancing_scan_period_max_ms is the maximum time in milliseconds to 509scan a tasks virtual memory. It effectively controls the minimum scanning 510rate for each task. 511 512numa_balancing_scan_size_mb is how many megabytes worth of pages are 513scanned for a given scan. 514 515============================================================== 516 517osrelease, ostype & version: 518 519# cat osrelease 5202.1.88 521# cat ostype 522Linux 523# cat version 524#5 Wed Feb 25 21:49:24 MET 1998 525 526The files osrelease and ostype should be clear enough. Version 527needs a little more clarification however. The '#5' means that 528this is the fifth kernel built from this source base and the 529date behind it indicates the time the kernel was built. 530The only way to tune these values is to rebuild the kernel :-) 531 532============================================================== 533 534overflowgid & overflowuid: 535 536if your architecture did not always support 32-bit UIDs (i.e. arm, 537i386, m68k, sh, and sparc32), a fixed UID and GID will be returned to 538applications that use the old 16-bit UID/GID system calls, if the 539actual UID or GID would exceed 65535. 540 541These sysctls allow you to change the value of the fixed UID and GID. 542The default is 65534. 543 544============================================================== 545 546panic: 547 548The value in this file represents the number of seconds the kernel 549waits before rebooting on a panic. When you use the software watchdog, 550the recommended setting is 60. 551 552============================================================== 553 554panic_on_oops: 555 556Controls the kernel's behaviour when an oops or BUG is encountered. 557 5580: try to continue operation 559 5601: panic immediately. If the `panic' sysctl is also non-zero then the 561 machine will be rebooted. 562 563============================================================== 564 565panic_on_stackoverflow: 566 567Controls the kernel's behavior when detecting the overflows of 568kernel, IRQ and exception stacks except a user stack. 569This file shows up if CONFIG_DEBUG_STACKOVERFLOW is enabled. 570 5710: try to continue operation. 572 5731: panic immediately. 574 575============================================================== 576 577panic_on_unrecovered_nmi: 578 579The default Linux behaviour on an NMI of either memory or unknown is 580to continue operation. For many environments such as scientific 581computing it is preferable that the box is taken out and the error 582dealt with than an uncorrected parity/ECC error get propagated. 583 584A small number of systems do generate NMI's for bizarre random reasons 585such as power management so the default is off. That sysctl works like 586the existing panic controls already in that directory. 587 588============================================================== 589 590panic_on_warn: 591 592Calls panic() in the WARN() path when set to 1. This is useful to avoid 593a kernel rebuild when attempting to kdump at the location of a WARN(). 594 5950: only WARN(), default behaviour. 596 5971: call panic() after printing out WARN() location. 598 599============================================================== 600 601perf_cpu_time_max_percent: 602 603Hints to the kernel how much CPU time it should be allowed to 604use to handle perf sampling events. If the perf subsystem 605is informed that its samples are exceeding this limit, it 606will drop its sampling frequency to attempt to reduce its CPU 607usage. 608 609Some perf sampling happens in NMIs. If these samples 610unexpectedly take too long to execute, the NMIs can become 611stacked up next to each other so much that nothing else is 612allowed to execute. 613 6140: disable the mechanism. Do not monitor or correct perf's 615 sampling rate no matter how CPU time it takes. 616 6171-100: attempt to throttle perf's sample rate to this 618 percentage of CPU. Note: the kernel calculates an 619 "expected" length of each sample event. 100 here means 620 100% of that expected length. Even if this is set to 621 100, you may still see sample throttling if this 622 length is exceeded. Set to 0 if you truly do not care 623 how much CPU is consumed. 624 625============================================================== 626 627 628pid_max: 629 630PID allocation wrap value. When the kernel's next PID value 631reaches this value, it wraps back to a minimum PID value. 632PIDs of value pid_max or larger are not allocated. 633 634============================================================== 635 636ns_last_pid: 637 638The last pid allocated in the current (the one task using this sysctl 639lives in) pid namespace. When selecting a pid for a next task on fork 640kernel tries to allocate a number starting from this one. 641 642============================================================== 643 644powersave-nap: (PPC only) 645 646If set, Linux-PPC will use the 'nap' mode of powersaving, 647otherwise the 'doze' mode will be used. 648 649============================================================== 650 651printk: 652 653The four values in printk denote: console_loglevel, 654default_message_loglevel, minimum_console_loglevel and 655default_console_loglevel respectively. 656 657These values influence printk() behavior when printing or 658logging error messages. See 'man 2 syslog' for more info on 659the different loglevels. 660 661- console_loglevel: messages with a higher priority than 662 this will be printed to the console 663- default_message_loglevel: messages without an explicit priority 664 will be printed with this priority 665- minimum_console_loglevel: minimum (highest) value to which 666 console_loglevel can be set 667- default_console_loglevel: default value for console_loglevel 668 669============================================================== 670 671printk_delay: 672 673Delay each printk message in printk_delay milliseconds 674 675Value from 0 - 10000 is allowed. 676 677============================================================== 678 679printk_ratelimit: 680 681Some warning messages are rate limited. printk_ratelimit specifies 682the minimum length of time between these messages (in jiffies), by 683default we allow one every 5 seconds. 684 685A value of 0 will disable rate limiting. 686 687============================================================== 688 689printk_ratelimit_burst: 690 691While long term we enforce one message per printk_ratelimit 692seconds, we do allow a burst of messages to pass through. 693printk_ratelimit_burst specifies the number of messages we can 694send before ratelimiting kicks in. 695 696============================================================== 697 698randomize_va_space: 699 700This option can be used to select the type of process address 701space randomization that is used in the system, for architectures 702that support this feature. 703 7040 - Turn the process address space randomization off. This is the 705 default for architectures that do not support this feature anyways, 706 and kernels that are booted with the "norandmaps" parameter. 707 7081 - Make the addresses of mmap base, stack and VDSO page randomized. 709 This, among other things, implies that shared libraries will be 710 loaded to random addresses. Also for PIE-linked binaries, the 711 location of code start is randomized. This is the default if the 712 CONFIG_COMPAT_BRK option is enabled. 713 7142 - Additionally enable heap randomization. This is the default if 715 CONFIG_COMPAT_BRK is disabled. 716 717 There are a few legacy applications out there (such as some ancient 718 versions of libc.so.5 from 1996) that assume that brk area starts 719 just after the end of the code+bss. These applications break when 720 start of the brk area is randomized. There are however no known 721 non-legacy applications that would be broken this way, so for most 722 systems it is safe to choose full randomization. 723 724 Systems with ancient and/or broken binaries should be configured 725 with CONFIG_COMPAT_BRK enabled, which excludes the heap from process 726 address space randomization. 727 728============================================================== 729 730reboot-cmd: (Sparc only) 731 732??? This seems to be a way to give an argument to the Sparc 733ROM/Flash boot loader. Maybe to tell it what to do after 734rebooting. ??? 735 736============================================================== 737 738rtsig-max & rtsig-nr: 739 740The file rtsig-max can be used to tune the maximum number 741of POSIX realtime (queued) signals that can be outstanding 742in the system. 743 744rtsig-nr shows the number of RT signals currently queued. 745 746============================================================== 747 748sg-big-buff: 749 750This file shows the size of the generic SCSI (sg) buffer. 751You can't tune it just yet, but you could change it on 752compile time by editing include/scsi/sg.h and changing 753the value of SG_BIG_BUFF. 754 755There shouldn't be any reason to change this value. If 756you can come up with one, you probably know what you 757are doing anyway :) 758 759============================================================== 760 761shmall: 762 763This parameter sets the total amount of shared memory pages that 764can be used system wide. Hence, SHMALL should always be at least 765ceil(shmmax/PAGE_SIZE). 766 767If you are not sure what the default PAGE_SIZE is on your Linux 768system, you can run the following command: 769 770# getconf PAGE_SIZE 771 772============================================================== 773 774shmmax: 775 776This value can be used to query and set the run time limit 777on the maximum shared memory segment size that can be created. 778Shared memory segments up to 1Gb are now supported in the 779kernel. This value defaults to SHMMAX. 780 781============================================================== 782 783shm_rmid_forced: 784 785Linux lets you set resource limits, including how much memory one 786process can consume, via setrlimit(2). Unfortunately, shared memory 787segments are allowed to exist without association with any process, and 788thus might not be counted against any resource limits. If enabled, 789shared memory segments are automatically destroyed when their attach 790count becomes zero after a detach or a process termination. It will 791also destroy segments that were created, but never attached to, on exit 792from the process. The only use left for IPC_RMID is to immediately 793destroy an unattached segment. Of course, this breaks the way things are 794defined, so some applications might stop working. Note that this 795feature will do you no good unless you also configure your resource 796limits (in particular, RLIMIT_AS and RLIMIT_NPROC). Most systems don't 797need this. 798 799Note that if you change this from 0 to 1, already created segments 800without users and with a dead originative process will be destroyed. 801 802============================================================== 803 804sysctl_writes_strict: 805 806Control how file position affects the behavior of updating sysctl values 807via the /proc/sys interface: 808 809 -1 - Legacy per-write sysctl value handling, with no printk warnings. 810 Each write syscall must fully contain the sysctl value to be 811 written, and multiple writes on the same sysctl file descriptor 812 will rewrite the sysctl value, regardless of file position. 813 0 - (default) Same behavior as above, but warn about processes that 814 perform writes to a sysctl file descriptor when the file position 815 is not 0. 816 1 - Respect file position when writing sysctl strings. Multiple writes 817 will append to the sysctl value buffer. Anything past the max length 818 of the sysctl value buffer will be ignored. Writes to numeric sysctl 819 entries must always be at file position 0 and the value must be 820 fully contained in the buffer sent in the write syscall. 821 822============================================================== 823 824softlockup_all_cpu_backtrace: 825 826This value controls the soft lockup detector thread's behavior 827when a soft lockup condition is detected as to whether or not 828to gather further debug information. If enabled, each cpu will 829be issued an NMI and instructed to capture stack trace. 830 831This feature is only applicable for architectures which support 832NMI. 833 8340: do nothing. This is the default behavior. 835 8361: on detection capture more debug information. 837 838============================================================== 839 840soft_watchdog 841 842This parameter can be used to control the soft lockup detector. 843 844 0 - disable the soft lockup detector 845 1 - enable the soft lockup detector 846 847The soft lockup detector monitors CPUs for threads that are hogging the CPUs 848without rescheduling voluntarily, and thus prevent the 'watchdog/N' threads 849from running. The mechanism depends on the CPUs ability to respond to timer 850interrupts which are needed for the 'watchdog/N' threads to be woken up by 851the watchdog timer function, otherwise the NMI watchdog - if enabled - can 852detect a hard lockup condition. 853 854============================================================== 855 856tainted: 857 858Non-zero if the kernel has been tainted. Numeric values, which 859can be ORed together: 860 861 1 - A module with a non-GPL license has been loaded, this 862 includes modules with no license. 863 Set by modutils >= 2.4.9 and module-init-tools. 864 2 - A module was force loaded by insmod -f. 865 Set by modutils >= 2.4.9 and module-init-tools. 866 4 - Unsafe SMP processors: SMP with CPUs not designed for SMP. 867 8 - A module was forcibly unloaded from the system by rmmod -f. 868 16 - A hardware machine check error occurred on the system. 869 32 - A bad page was discovered on the system. 870 64 - The user has asked that the system be marked "tainted". This 871 could be because they are running software that directly modifies 872 the hardware, or for other reasons. 873 128 - The system has died. 874 256 - The ACPI DSDT has been overridden with one supplied by the user 875 instead of using the one provided by the hardware. 876 512 - A kernel warning has occurred. 8771024 - A module from drivers/staging was loaded. 8782048 - The system is working around a severe firmware bug. 8794096 - An out-of-tree module has been loaded. 8808192 - An unsigned module has been loaded in a kernel supporting module 881 signature. 88216384 - A soft lockup has previously occurred on the system. 88332768 - The kernel has been live patched. 884 885============================================================== 886 887threads-max 888 889This value controls the maximum number of threads that can be created 890using fork(). 891 892During initialization the kernel sets this value such that even if the 893maximum number of threads is created, the thread structures occupy only 894a part (1/8th) of the available RAM pages. 895 896The minimum value that can be written to threads-max is 20. 897The maximum value that can be written to threads-max is given by the 898constant FUTEX_TID_MASK (0x3fffffff). 899If a value outside of this range is written to threads-max an error 900EINVAL occurs. 901 902The value written is checked against the available RAM pages. If the 903thread structures would occupy too much (more than 1/8th) of the 904available RAM pages threads-max is reduced accordingly. 905 906============================================================== 907 908unknown_nmi_panic: 909 910The value in this file affects behavior of handling NMI. When the 911value is non-zero, unknown NMI is trapped and then panic occurs. At 912that time, kernel debugging information is displayed on console. 913 914NMI switch that most IA32 servers have fires unknown NMI up, for 915example. If a system hangs up, try pressing the NMI switch. 916 917============================================================== 918 919watchdog: 920 921This parameter can be used to disable or enable the soft lockup detector 922_and_ the NMI watchdog (i.e. the hard lockup detector) at the same time. 923 924 0 - disable both lockup detectors 925 1 - enable both lockup detectors 926 927The soft lockup detector and the NMI watchdog can also be disabled or 928enabled individually, using the soft_watchdog and nmi_watchdog parameters. 929If the watchdog parameter is read, for example by executing 930 931 cat /proc/sys/kernel/watchdog 932 933the output of this command (0 or 1) shows the logical OR of soft_watchdog 934and nmi_watchdog. 935 936============================================================== 937 938watchdog_cpumask: 939 940This value can be used to control on which cpus the watchdog may run. 941The default cpumask is all possible cores, but if NO_HZ_FULL is 942enabled in the kernel config, and cores are specified with the 943nohz_full= boot argument, those cores are excluded by default. 944Offline cores can be included in this mask, and if the core is later 945brought online, the watchdog will be started based on the mask value. 946 947Typically this value would only be touched in the nohz_full case 948to re-enable cores that by default were not running the watchdog, 949if a kernel lockup was suspected on those cores. 950 951The argument value is the standard cpulist format for cpumasks, 952so for example to enable the watchdog on cores 0, 2, 3, and 4 you 953might say: 954 955 echo 0,2-4 > /proc/sys/kernel/watchdog_cpumask 956 957============================================================== 958 959watchdog_thresh: 960 961This value can be used to control the frequency of hrtimer and NMI 962events and the soft and hard lockup thresholds. The default threshold 963is 10 seconds. 964 965The softlockup threshold is (2 * watchdog_thresh). Setting this 966tunable to zero will disable lockup detection altogether. 967 968============================================================== 969