7689 lines · plain
1 accept_memory= [MM]2 Format: { eager | lazy }3 default: lazy4 By default, unaccepted memory is accepted lazily to5 avoid prolonged boot times. The lazy option will add6 some runtime overhead until all memory is eventually7 accepted. In most cases the overhead is negligible.8 For some workloads or for debugging purposes9 accept_memory=eager can be used to accept all memory10 at once during boot.11 12 acpi= [HW,ACPI,X86,ARM64,RISCV64,EARLY]13 Advanced Configuration and Power Interface14 Format: { force | on | off | strict | noirq | rsdt |15 copy_dsdt | nospcr }16 force -- enable ACPI if default was off17 on -- enable ACPI but allow fallback to DT [arm64,riscv64]18 off -- disable ACPI if default was on19 noirq -- do not use ACPI for IRQ routing20 strict -- Be less tolerant of platforms that are not21 strictly ACPI specification compliant.22 rsdt -- prefer RSDT over (default) XSDT23 copy_dsdt -- copy DSDT to memory24 nospcr -- disable console in ACPI SPCR table as25 default _serial_ console on ARM6426 For ARM64, ONLY "acpi=off", "acpi=on", "acpi=force" or27 "acpi=nospcr" are available28 For RISCV64, ONLY "acpi=off", "acpi=on" or "acpi=force"29 are available30 31 See also Documentation/power/runtime_pm.rst, pci=noacpi32 33 acpi_apic_instance= [ACPI,IOAPIC,EARLY]34 Format: <int>35 2: use 2nd APIC table, if available36 1,0: use 1st APIC table37 default: 038 39 acpi_backlight= [HW,ACPI]40 { vendor | video | native | none }41 If set to vendor, prefer vendor-specific driver42 (e.g. thinkpad_acpi, sony_acpi, etc.) instead43 of the ACPI video.ko driver.44 If set to video, use the ACPI video.ko driver.45 If set to native, use the device's native backlight mode.46 If set to none, disable the ACPI backlight interface.47 48 acpi_force_32bit_fadt_addr [ACPI,EARLY]49 force FADT to use 32 bit addresses rather than the50 64 bit X_* addresses. Some firmware have broken 6451 bit addresses for force ACPI ignore these and use52 the older legacy 32 bit addresses.53 54 acpica_no_return_repair [HW, ACPI]55 Disable AML predefined validation mechanism56 This mechanism can repair the evaluation result to make57 the return objects more ACPI specification compliant.58 This option is useful for developers to identify the59 root cause of an AML interpreter issue when the issue60 has something to do with the repair mechanism.61 62 acpi.debug_layer= [HW,ACPI,ACPI_DEBUG]63 acpi.debug_level= [HW,ACPI,ACPI_DEBUG]64 Format: <int>65 CONFIG_ACPI_DEBUG must be enabled to produce any ACPI66 debug output. Bits in debug_layer correspond to a67 _COMPONENT in an ACPI source file, e.g.,68 #define _COMPONENT ACPI_EVENTS69 Bits in debug_level correspond to a level in70 ACPI_DEBUG_PRINT statements, e.g.,71 ACPI_DEBUG_PRINT((ACPI_DB_INFO, ...72 The debug_level mask defaults to "info". See73 Documentation/firmware-guide/acpi/debug.rst for more information about74 debug layers and levels.75 76 Enable processor driver info messages:77 acpi.debug_layer=0x2000000078 Enable AML "Debug" output, i.e., stores to the Debug79 object while interpreting AML:80 acpi.debug_layer=0xffffffff acpi.debug_level=0x281 Enable all messages related to ACPI hardware:82 acpi.debug_layer=0x2 acpi.debug_level=0xffffffff83 84 Some values produce so much output that the system is85 unusable. The "log_buf_len" parameter may be useful86 if you need to capture more output.87 88 acpi_enforce_resources= [ACPI]89 { strict | lax | no }90 Check for resource conflicts between native drivers91 and ACPI OperationRegions (SystemIO and SystemMemory92 only). IO ports and memory declared in ACPI might be93 used by the ACPI subsystem in arbitrary AML code and94 can interfere with legacy drivers.95 strict (default): access to resources claimed by ACPI96 is denied; legacy drivers trying to access reserved97 resources will fail to bind to device using them.98 lax: access to resources claimed by ACPI is allowed;99 legacy drivers trying to access reserved resources100 will bind successfully but a warning message is logged.101 no: ACPI OperationRegions are not marked as reserved,102 no further checks are performed.103 104 acpi_force_table_verification [HW,ACPI,EARLY]105 Enable table checksum verification during early stage.106 By default, this is disabled due to x86 early mapping107 size limitation.108 109 acpi_irq_balance [HW,ACPI]110 ACPI will balance active IRQs111 default in APIC mode112 113 acpi_irq_nobalance [HW,ACPI]114 ACPI will not move active IRQs (default)115 default in PIC mode116 117 acpi_irq_isa= [HW,ACPI] If irq_balance, mark listed IRQs used by ISA118 Format: <irq>,<irq>...119 120 acpi_irq_pci= [HW,ACPI] If irq_balance, clear listed IRQs for121 use by PCI122 Format: <irq>,<irq>...123 124 acpi_mask_gpe= [HW,ACPI]125 Due to the existence of _Lxx/_Exx, some GPEs triggered126 by unsupported hardware/firmware features can result in127 GPE floodings that cannot be automatically disabled by128 the GPE dispatcher.129 This facility can be used to prevent such uncontrolled130 GPE floodings.131 Format: <byte> or <bitmap-list>132 133 acpi_no_auto_serialize [HW,ACPI]134 Disable auto-serialization of AML methods135 AML control methods that contain the opcodes to create136 named objects will be marked as "Serialized" by the137 auto-serialization feature.138 This feature is enabled by default.139 This option allows to turn off the feature.140 141 acpi_no_memhotplug [ACPI] Disable memory hotplug. Useful for kdump142 kernels.143 144 acpi_no_static_ssdt [HW,ACPI,EARLY]145 Disable installation of static SSDTs at early boot time146 By default, SSDTs contained in the RSDT/XSDT will be147 installed automatically and they will appear under148 /sys/firmware/acpi/tables.149 This option turns off this feature.150 Note that specifying this option does not affect151 dynamic table installation which will install SSDT152 tables to /sys/firmware/acpi/tables/dynamic.153 154 acpi_no_watchdog [HW,ACPI,WDT]155 Ignore the ACPI-based watchdog interface (WDAT) and let156 a native driver control the watchdog device instead.157 158 acpi_rsdp= [ACPI,EFI,KEXEC,EARLY]159 Pass the RSDP address to the kernel, mostly used160 on machines running EFI runtime service to boot the161 second kernel for kdump.162 163 acpi_os_name= [HW,ACPI] Tell ACPI BIOS the name of the OS164 Format: To spoof as Windows 98: ="Microsoft Windows"165 166 acpi_rev_override [ACPI] Override the _REV object to return 5 (instead167 of 2 which is mandated by ACPI 6) as the supported ACPI168 specification revision (when using this switch, it may169 be necessary to carry out a cold reboot _twice_ in a170 row to make it take effect on the platform firmware).171 172 acpi_osi= [HW,ACPI] Modify list of supported OS interface strings173 acpi_osi="string1" # add string1174 acpi_osi="!string2" # remove string2175 acpi_osi=!* # remove all strings176 acpi_osi=! # disable all built-in OS vendor177 strings178 acpi_osi=!! # enable all built-in OS vendor179 strings180 acpi_osi= # disable all strings181 182 'acpi_osi=!' can be used in combination with single or183 multiple 'acpi_osi="string1"' to support specific OS184 vendor string(s). Note that such command can only185 affect the default state of the OS vendor strings, thus186 it cannot affect the default state of the feature group187 strings and the current state of the OS vendor strings,188 specifying it multiple times through kernel command line189 is meaningless. This command is useful when one do not190 care about the state of the feature group strings which191 should be controlled by the OSPM.192 Examples:193 1. 'acpi_osi=! acpi_osi="Windows 2000"' is equivalent194 to 'acpi_osi="Windows 2000" acpi_osi=!', they all195 can make '_OSI("Windows 2000")' TRUE.196 197 'acpi_osi=' cannot be used in combination with other198 'acpi_osi=' command lines, the _OSI method will not199 exist in the ACPI namespace. NOTE that such command can200 only affect the _OSI support state, thus specifying it201 multiple times through kernel command line is also202 meaningless.203 Examples:204 1. 'acpi_osi=' can make 'CondRefOf(_OSI, Local1)'205 FALSE.206 207 'acpi_osi=!*' can be used in combination with single or208 multiple 'acpi_osi="string1"' to support specific209 string(s). Note that such command can affect the210 current state of both the OS vendor strings and the211 feature group strings, thus specifying it multiple times212 through kernel command line is meaningful. But it may213 still not able to affect the final state of a string if214 there are quirks related to this string. This command215 is useful when one want to control the state of the216 feature group strings to debug BIOS issues related to217 the OSPM features.218 Examples:219 1. 'acpi_osi="Module Device" acpi_osi=!*' can make220 '_OSI("Module Device")' FALSE.221 2. 'acpi_osi=!* acpi_osi="Module Device"' can make222 '_OSI("Module Device")' TRUE.223 3. 'acpi_osi=! acpi_osi=!* acpi_osi="Windows 2000"' is224 equivalent to225 'acpi_osi=!* acpi_osi=! acpi_osi="Windows 2000"'226 and227 'acpi_osi=!* acpi_osi="Windows 2000" acpi_osi=!',228 they all will make '_OSI("Windows 2000")' TRUE.229 230 acpi_pm_good [X86]231 Override the pmtimer bug detection: force the kernel232 to assume that this machine's pmtimer latches its value233 and always returns good values.234 235 acpi_sci= [HW,ACPI,EARLY] ACPI System Control Interrupt trigger mode236 Format: { level | edge | high | low }237 238 acpi_skip_timer_override [HW,ACPI,EARLY]239 Recognize and ignore IRQ0/pin2 Interrupt Override.240 For broken nForce2 BIOS resulting in XT-PIC timer.241 242 acpi_sleep= [HW,ACPI] Sleep options243 Format: { s3_bios, s3_mode, s3_beep, s4_hwsig,244 s4_nohwsig, old_ordering, nonvs,245 sci_force_enable, nobl }246 See Documentation/power/video.rst for information on247 s3_bios and s3_mode.248 s3_beep is for debugging; it makes the PC's speaker beep249 as soon as the kernel's real-mode entry point is called.250 s4_hwsig causes the kernel to check the ACPI hardware251 signature during resume from hibernation, and gracefully252 refuse to resume if it has changed. This complies with253 the ACPI specification but not with reality, since254 Windows does not do this and many laptops do change it255 on docking. So the default behaviour is to allow resume256 and simply warn when the signature changes, unless the257 s4_hwsig option is enabled.258 s4_nohwsig prevents ACPI hardware signature from being259 used (or even warned about) during resume.260 old_ordering causes the ACPI 1.0 ordering of the _PTS261 control method, with respect to putting devices into262 low power states, to be enforced (the ACPI 2.0 ordering263 of _PTS is used by default).264 nonvs prevents the kernel from saving/restoring the265 ACPI NVS memory during suspend/hibernation and resume.266 sci_force_enable causes the kernel to set SCI_EN directly267 on resume from S1/S3 (which is against the ACPI spec,268 but some broken systems don't work without it).269 nobl causes the internal blacklist of systems known to270 behave incorrectly in some ways with respect to system271 suspend and resume to be ignored (use wisely).272 273 acpi_use_timer_override [HW,ACPI,EARLY]274 Use timer override. For some broken Nvidia NF5 boards275 that require a timer override, but don't have HPET276 277 add_efi_memmap [EFI,X86,EARLY] Include EFI memory map in278 kernel's map of available physical RAM.279 280 agp= [AGP]281 { off | try_unsupported }282 off: disable AGP support283 try_unsupported: try to drive unsupported chipsets284 (may crash computer or cause data corruption)285 286 ALSA [HW,ALSA]287 See Documentation/sound/alsa-configuration.rst288 289 alignment= [KNL,ARM]290 Allow the default userspace alignment fault handler291 behaviour to be specified. Bit 0 enables warnings,292 bit 1 enables fixups, and bit 2 sends a segfault.293 294 align_va_addr= [X86-64]295 Align virtual addresses by clearing slice [14:12] when296 allocating a VMA at process creation time. This option297 gives you up to 3% performance improvement on AMD F15h298 machines (where it is enabled by default) for a299 CPU-intensive style benchmark, and it can vary highly in300 a microbenchmark depending on workload and compiler.301 302 32: only for 32-bit processes303 64: only for 64-bit processes304 on: enable for both 32- and 64-bit processes305 off: disable for both 32- and 64-bit processes306 307 alloc_snapshot [FTRACE]308 Allocate the ftrace snapshot buffer on boot up when the309 main buffer is allocated. This is handy if debugging310 and you need to use tracing_snapshot() on boot up, and311 do not want to use tracing_snapshot_alloc() as it needs312 to be done where GFP_KERNEL allocations are allowed.313 314 allow_mismatched_32bit_el0 [ARM64,EARLY]315 Allow execve() of 32-bit applications and setting of the316 PER_LINUX32 personality on systems where only a strict317 subset of the CPUs support 32-bit EL0. When this318 parameter is present, the set of CPUs supporting 32-bit319 EL0 is indicated by /sys/devices/system/cpu/aarch32_el0320 and hot-unplug operations may be restricted.321 322 See Documentation/arch/arm64/asymmetric-32bit.rst for more323 information.324 325 amd_iommu= [HW,X86-64]326 Pass parameters to the AMD IOMMU driver in the system.327 Possible values are:328 fullflush - Deprecated, equivalent to iommu.strict=1329 off - do not initialize any AMD IOMMU found in330 the system331 force_isolation - Force device isolation for all332 devices. The IOMMU driver is not333 allowed anymore to lift isolation334 requirements as needed. This option335 does not override iommu=pt336 force_enable - Force enable the IOMMU on platforms known337 to be buggy with IOMMU enabled. Use this338 option with care.339 pgtbl_v1 - Use v1 page table for DMA-API (Default).340 pgtbl_v2 - Use v2 page table for DMA-API.341 irtcachedis - Disable Interrupt Remapping Table (IRT) caching.342 nohugepages - Limit page-sizes used for v1 page-tables343 to 4 KiB.344 v2_pgsizes_only - Limit page-sizes used for v1 page-tables345 to 4KiB/2Mib/1GiB.346 347 348 amd_iommu_dump= [HW,X86-64]349 Enable AMD IOMMU driver option to dump the ACPI table350 for AMD IOMMU. With this option enabled, AMD IOMMU351 driver will print ACPI tables for AMD IOMMU during352 IOMMU initialization.353 354 amd_iommu_intr= [HW,X86-64]355 Specifies one of the following AMD IOMMU interrupt356 remapping modes:357 legacy - Use legacy interrupt remapping mode.358 vapic - Use virtual APIC mode, which allows IOMMU359 to inject interrupts directly into guest.360 This mode requires kvm-amd.avic=1.361 (Default when IOMMU HW support is present.)362 363 amd_pstate= [X86,EARLY]364 disable365 Do not enable amd_pstate as the default366 scaling driver for the supported processors367 passive368 Use amd_pstate with passive mode as a scaling driver.369 In this mode autonomous selection is disabled.370 Driver requests a desired performance level and platform371 tries to match the same performance level if it is372 satisfied by guaranteed performance level.373 active374 Use amd_pstate_epp driver instance as the scaling driver,375 driver provides a hint to the hardware if software wants376 to bias toward performance (0x0) or energy efficiency (0xff)377 to the CPPC firmware. then CPPC power algorithm will378 calculate the runtime workload and adjust the realtime cores379 frequency.380 guided381 Activate guided autonomous mode. Driver requests minimum and382 maximum performance level and the platform autonomously383 selects a performance level in this range and appropriate384 to the current workload.385 386 amd_prefcore=387 [X86]388 disable389 Disable amd-pstate preferred core.390 391 amijoy.map= [HW,JOY] Amiga joystick support392 Map of devices attached to JOY0DAT and JOY1DAT393 Format: <a>,<b>394 See also Documentation/input/joydev/joystick.rst395 396 analog.map= [HW,JOY] Analog joystick and gamepad support397 Specifies type or capabilities of an analog joystick398 connected to one of 16 gameports399 Format: <type1>,<type2>,..<type16>400 401 apc= [HW,SPARC]402 Power management functions (SPARCstation-4/5 + deriv.)403 Format: noidle404 Disable APC CPU standby support. SPARCstation-Fox does405 not play well with APC CPU idle - disable it if you have406 APC and your system crashes randomly.407 408 apic= [APIC,X86,EARLY] Advanced Programmable Interrupt Controller409 Change the output verbosity while booting410 Format: { quiet (default) | verbose | debug }411 Change the amount of debugging information output412 when initialising the APIC and IO-APIC components.413 For X86-32, this can also be used to specify an APIC414 driver name.415 Format: apic=driver_name416 Examples: apic=bigsmp417 418 apic_extnmi= [APIC,X86,EARLY] External NMI delivery setting419 Format: { bsp (default) | all | none }420 bsp: External NMI is delivered only to CPU 0421 all: External NMIs are broadcast to all CPUs as a422 backup of CPU 0423 none: External NMI is masked for all CPUs. This is424 useful so that a dump capture kernel won't be425 shot down by NMI426 427 autoconf= [IPV6]428 See Documentation/networking/ipv6.rst.429 430 apm= [APM] Advanced Power Management431 See header of arch/x86/kernel/apm_32.c.432 433 apparmor= [APPARMOR] Disable or enable AppArmor at boot time434 Format: { "0" | "1" }435 See security/apparmor/Kconfig help text436 0 -- disable.437 1 -- enable.438 Default value is set via kernel config option.439 440 arcrimi= [HW,NET] ARCnet - "RIM I" (entirely mem-mapped) cards441 Format: <io>,<irq>,<nodeID>442 443 arm64.no32bit_el0 [ARM64] Unconditionally disable the execution of444 32 bit applications.445 446 arm64.nobti [ARM64] Unconditionally disable Branch Target447 Identification support448 449 arm64.nomops [ARM64] Unconditionally disable Memory Copy and Memory450 Set instructions support451 452 arm64.nomte [ARM64] Unconditionally disable Memory Tagging Extension453 support454 455 arm64.nopauth [ARM64] Unconditionally disable Pointer Authentication456 support457 458 arm64.nosme [ARM64] Unconditionally disable Scalable Matrix459 Extension support460 461 arm64.nosve [ARM64] Unconditionally disable Scalable Vector462 Extension support463 464 ataflop= [HW,M68k]465 466 atarimouse= [HW,MOUSE] Atari Mouse467 468 atkbd.extra= [HW] Enable extra LEDs and keys on IBM RapidAccess,469 EzKey and similar keyboards470 471 atkbd.reset= [HW] Reset keyboard during initialization472 473 atkbd.set= [HW] Select keyboard code set474 Format: <int> (2 = AT (default), 3 = PS/2)475 476 atkbd.scroll= [HW] Enable scroll wheel on MS Office and similar477 keyboards478 479 atkbd.softraw= [HW] Choose between synthetic and real raw mode480 Format: <bool> (0 = real, 1 = synthetic (default))481 482 atkbd.softrepeat= [HW]483 Use software keyboard repeat484 485 audit= [KNL] Enable the audit sub-system486 Format: { "0" | "1" | "off" | "on" }487 0 | off - kernel audit is disabled and can not be488 enabled until the next reboot489 unset - kernel audit is initialized but disabled and490 will be fully enabled by the userspace auditd.491 1 | on - kernel audit is initialized and partially492 enabled, storing at most audit_backlog_limit493 messages in RAM until it is fully enabled by the494 userspace auditd.495 Default: unset496 497 audit_backlog_limit= [KNL] Set the audit queue size limit.498 Format: <int> (must be >=0)499 Default: 64500 501 bau= [X86_UV] Enable the BAU on SGI UV. The default502 behavior is to disable the BAU (i.e. bau=0).503 Format: { "0" | "1" }504 0 - Disable the BAU.505 1 - Enable the BAU.506 unset - Disable the BAU.507 508 baycom_epp= [HW,AX25]509 Format: <io>,<mode>510 511 baycom_par= [HW,AX25] BayCom Parallel Port AX.25 Modem512 Format: <io>,<mode>513 See header of drivers/net/hamradio/baycom_par.c.514 515 baycom_ser_fdx= [HW,AX25]516 BayCom Serial Port AX.25 Modem (Full Duplex Mode)517 Format: <io>,<irq>,<mode>[,<baud>]518 See header of drivers/net/hamradio/baycom_ser_fdx.c.519 520 baycom_ser_hdx= [HW,AX25]521 BayCom Serial Port AX.25 Modem (Half Duplex Mode)522 Format: <io>,<irq>,<mode>523 See header of drivers/net/hamradio/baycom_ser_hdx.c.524 525 bdev_allow_write_mounted=526 Format: <bool>527 Control the ability to open a mounted block device528 for writing, i.e., allow / disallow writes that bypass529 the FS. This was implemented as a means to prevent530 fuzzers from crashing the kernel by overwriting the531 metadata underneath a mounted FS without its awareness.532 This also prevents destructive formatting of mounted533 filesystems by naive storage tooling that don't use534 O_EXCL. Default is Y and can be changed through the535 Kconfig option CONFIG_BLK_DEV_WRITE_MOUNTED.536 537 bert_disable [ACPI]538 Disable BERT OS support on buggy BIOSes.539 540 bgrt_disable [ACPI,X86,EARLY]541 Disable BGRT to avoid flickering OEM logo.542 543 blkdevparts= Manual partition parsing of block device(s) for544 embedded devices based on command line input.545 See Documentation/block/cmdline-partition.rst546 547 boot_delay= [KNL,EARLY]548 Milliseconds to delay each printk during boot.549 Only works if CONFIG_BOOT_PRINTK_DELAY is enabled,550 and you may also have to specify "lpj=". Boot_delay551 values larger than 10 seconds (10000) are assumed552 erroneous and ignored.553 Format: integer554 555 bootconfig [KNL,EARLY]556 Extended command line options can be added to an initrd557 and this will cause the kernel to look for it.558 559 See Documentation/admin-guide/bootconfig.rst560 561 bttv.card= [HW,V4L] bttv (bt848 + bt878 based grabber cards)562 bttv.radio= Most important insmod options are available as563 kernel args too.564 bttv.pll= See Documentation/admin-guide/media/bttv.rst565 bttv.tuner=566 567 bulk_remove=off [PPC] This parameter disables the use of the pSeries568 firmware feature for flushing multiple hpte entries569 at a time.570 571 c101= [NET] Moxa C101 synchronous serial card572 573 cachesize= [BUGS=X86-32] Override level 2 CPU cache size detection.574 Sometimes CPU hardware bugs make them report the cache575 size incorrectly. The kernel will attempt work arounds576 to fix known problems, but for some CPUs it is not577 possible to determine what the correct size should be.578 This option provides an override for these situations.579 580 carrier_timeout=581 [NET] Specifies amount of time (in seconds) that582 the kernel should wait for a network carrier. By default583 it waits 120 seconds.584 585 ca_keys= [KEYS] This parameter identifies a specific key(s) on586 the system trusted keyring to be used for certificate587 trust validation.588 format: { id:<keyid> | builtin }589 590 cca= [MIPS,EARLY] Override the kernel pages' cache coherency591 algorithm. Accepted values range from 0 to 7592 inclusive. See arch/mips/include/asm/pgtable-bits.h593 for platform specific values (SB1, Loongson3 and594 others).595 596 ccw_timeout_log [S390]597 See Documentation/arch/s390/common_io.rst for details.598 599 cgroup_disable= [KNL] Disable a particular controller or optional feature600 Format: {name of the controller(s) or feature(s) to disable}601 The effects of cgroup_disable=foo are:602 - foo isn't auto-mounted if you mount all cgroups in603 a single hierarchy604 - foo isn't visible as an individually mountable605 subsystem606 - if foo is an optional feature then the feature is607 disabled and corresponding cgroup files are not608 created609 {Currently only "memory" controller deal with this and610 cut the overhead, others just disable the usage. So611 only cgroup_disable=memory is actually worthy}612 Specifying "pressure" disables per-cgroup pressure613 stall information accounting feature614 615 cgroup_no_v1= [KNL] Disable cgroup controllers and named hierarchies in v1616 Format: { { controller | "all" | "named" }617 [,{ controller | "all" | "named" }...] }618 Like cgroup_disable, but only applies to cgroup v1;619 the blacklisted controllers remain available in cgroup2.620 "all" blacklists all controllers and "named" disables621 named mounts. Specifying both "all" and "named" disables622 all v1 hierarchies.623 624 cgroup_favordynmods= [KNL] Enable or Disable favordynmods.625 Format: { "true" | "false" }626 Defaults to the value of CONFIG_CGROUP_FAVOR_DYNMODS.627 628 cgroup.memory= [KNL] Pass options to the cgroup memory controller.629 Format: <string>630 nosocket -- Disable socket memory accounting.631 nokmem -- Disable kernel memory accounting.632 nobpf -- Disable BPF memory accounting.633 634 checkreqprot= [SELINUX] Set initial checkreqprot flag value.635 Format: { "0" | "1" }636 See security/selinux/Kconfig help text.637 0 -- check protection applied by kernel (includes638 any implied execute protection).639 1 -- check protection requested by application.640 Default value is set via a kernel config option.641 Value can be changed at runtime via642 /sys/fs/selinux/checkreqprot.643 Setting checkreqprot to 1 is deprecated.644 645 cio_ignore= [S390]646 See Documentation/arch/s390/common_io.rst for details.647 648 clearcpuid=X[,X...] [X86]649 Disable CPUID feature X for the kernel. See650 arch/x86/include/asm/cpufeatures.h for the valid bit651 numbers X. Note the Linux-specific bits are not necessarily652 stable over kernel options, but the vendor-specific653 ones should be.654 X can also be a string as appearing in the flags: line655 in /proc/cpuinfo which does not have the above656 instability issue. However, not all features have names657 in /proc/cpuinfo.658 Note that using this option will taint your kernel.659 Also note that user programs calling CPUID directly660 or using the feature without checking anything661 will still see it. This just prevents it from662 being used by the kernel or shown in /proc/cpuinfo.663 Also note the kernel might malfunction if you disable664 some critical bits.665 666 clk_ignore_unused667 [CLK]668 Prevents the clock framework from automatically gating669 clocks that have not been explicitly enabled by a Linux670 device driver but are enabled in hardware at reset or671 by the bootloader/firmware. Note that this does not672 force such clocks to be always-on nor does it reserve673 those clocks in any way. This parameter is useful for674 debug and development, but should not be needed on a675 platform with proper driver support. For more676 information, see Documentation/driver-api/clk.rst.677 678 clock= [BUGS=X86-32, HW] gettimeofday clocksource override.679 [Deprecated]680 Forces specified clocksource (if available) to be used681 when calculating gettimeofday(). If specified682 clocksource is not available, it defaults to PIT.683 Format: { pit | tsc | cyclone | pmtmr }684 685 clocksource= Override the default clocksource686 Format: <string>687 Override the default clocksource and use the clocksource688 with the name specified.689 Some clocksource names to choose from, depending on690 the platform:691 [all] jiffies (this is the base, fallback clocksource)692 [ACPI] acpi_pm693 [ARM] imx_timer1,OSTS,netx_timer,mpu_timer2,694 pxa_timer,timer3,32k_counter,timer0_1695 [X86-32] pit,hpet,tsc;696 scx200_hrt on Geode; cyclone on IBM x440697 [MIPS] MIPS698 [PARISC] cr16699 [S390] tod700 [SH] SuperH701 [SPARC64] tick702 [X86-64] hpet,tsc703 704 clocksource.arm_arch_timer.evtstrm=705 [ARM,ARM64,EARLY]706 Format: <bool>707 Enable/disable the eventstream feature of the ARM708 architected timer so that code using WFE-based polling709 loops can be debugged more effectively on production710 systems.711 712 clocksource.verify_n_cpus= [KNL]713 Limit the number of CPUs checked for clocksources714 marked with CLOCK_SOURCE_VERIFY_PERCPU that715 are marked unstable due to excessive skew.716 A negative value says to check all CPUs, while717 zero says not to check any. Values larger than718 nr_cpu_ids are silently truncated to nr_cpu_ids.719 The actual CPUs are chosen randomly, with720 no replacement if the same CPU is chosen twice.721 722 clocksource-wdtest.holdoff= [KNL]723 Set the time in seconds that the clocksource724 watchdog test waits before commencing its tests.725 Defaults to zero when built as a module and to726 10 seconds when built into the kernel.727 728 cma=nn[MG]@[start[MG][-end[MG]]]729 [KNL,CMA,EARLY]730 Sets the size of kernel global memory area for731 contiguous memory allocations and optionally the732 placement constraint by the physical address range of733 memory allocations. A value of 0 disables CMA734 altogether. For more information, see735 kernel/dma/contiguous.c736 737 cma_pernuma=nn[MG]738 [KNL,CMA,EARLY]739 Sets the size of kernel per-numa memory area for740 contiguous memory allocations. A value of 0 disables741 per-numa CMA altogether. And If this option is not742 specified, the default value is 0.743 With per-numa CMA enabled, DMA users on node nid will744 first try to allocate buffer from the pernuma area745 which is located in node nid, if the allocation fails,746 they will fallback to the global default memory area.747 748 numa_cma=<node>:nn[MG][,<node>:nn[MG]]749 [KNL,CMA,EARLY]750 Sets the size of kernel numa memory area for751 contiguous memory allocations. It will reserve CMA752 area for the specified node.753 754 With numa CMA enabled, DMA users on node nid will755 first try to allocate buffer from the numa area756 which is located in node nid, if the allocation fails,757 they will fallback to the global default memory area.758 759 cmo_free_hint= [PPC] Format: { yes | no }760 Specify whether pages are marked as being inactive761 when they are freed. This is used in CMO environments762 to determine OS memory pressure for page stealing by763 a hypervisor.764 Default: yes765 766 coherent_pool=nn[KMG] [ARM,KNL,EARLY]767 Sets the size of memory pool for coherent, atomic dma768 allocations, by default set to 256K.769 770 com20020= [HW,NET] ARCnet - COM20020 chipset771 Format:772 <io>[,<irq>[,<nodeID>[,<backplane>[,<ckp>[,<timeout>]]]]]773 774 com90io= [HW,NET] ARCnet - COM90xx chipset (IO-mapped buffers)775 Format: <io>[,<irq>]776 777 com90xx= [HW,NET]778 ARCnet - COM90xx chipset (memory-mapped buffers)779 Format: <io>[,<irq>[,<memstart>]]780 781 condev= [HW,S390] console device782 conmode=783 784 con3215_drop= [S390,EARLY] 3215 console drop mode.785 Format: y|n|Y|N|1|0786 When set to true, drop data on the 3215 console when787 the console buffer is full. In this case the788 operator using a 3270 terminal emulator (for example789 x3270) does not have to enter the clear key for the790 console output to advance and the kernel to continue.791 This leads to a much faster boot time when a 3270792 terminal emulator is active. If no 3270 terminal793 emulator is used, this parameter has no effect.794 795 console= [KNL] Output console device and options.796 797 tty<n> Use the virtual console device <n>.798 799 ttyS<n>[,options]800 ttyUSB0[,options]801 Use the specified serial port. The options are of802 the form "bbbbpnf", where "bbbb" is the baud rate,803 "p" is parity ("n", "o", or "e"), "n" is number of804 bits, and "f" is flow control ("r" for RTS or805 omit it). Default is "9600n8".806 807 See Documentation/admin-guide/serial-console.rst for more808 information. See809 Documentation/networking/netconsole.rst for an810 alternative.811 812 <DEVNAME>:<n>.<n>[,options]813 Use the specified serial port on the serial core bus.814 The addressing uses DEVNAME of the physical serial port815 device, followed by the serial core controller instance,816 and the serial port instance. The options are the same817 as documented for the ttyS addressing above.818 819 The mapping of the serial ports to the tty instances820 can be viewed with:821 822 $ ls -d /sys/bus/serial-base/devices/*:*.*/tty/*823 /sys/bus/serial-base/devices/00:04:0.0/tty/ttyS0824 825 In the above example, the console can be addressed with826 console=00:04:0.0. Note that a console addressed this827 way will only get added when the related device driver828 is ready. The use of an earlycon parameter in addition to829 the console may be desired for console output early on.830 831 uart[8250],io,<addr>[,options]832 uart[8250],mmio,<addr>[,options]833 uart[8250],mmio16,<addr>[,options]834 uart[8250],mmio32,<addr>[,options]835 uart[8250],0x<addr>[,options]836 Start an early, polled-mode console on the 8250/16550837 UART at the specified I/O port or MMIO address,838 switching to the matching ttyS device later.839 MMIO inter-register address stride is either 8-bit840 (mmio), 16-bit (mmio16), or 32-bit (mmio32).841 If none of [io|mmio|mmio16|mmio32], <addr> is assumed842 to be equivalent to 'mmio'. 'options' are specified in843 the same format described for ttyS above; if unspecified,844 the h/w is not re-initialized.845 846 hvc<n> Use the hypervisor console device <n>. This is for847 both Xen and PowerPC hypervisors.848 849 { null | "" }850 Use to disable console output, i.e., to have kernel851 console messages discarded.852 This must be the only console= parameter used on the853 kernel command line.854 855 If the device connected to the port is not a TTY but a braille856 device, prepend "brl," before the device type, for instance857 console=brl,ttyS0858 For now, only VisioBraille is supported.859 860 console_msg_format=861 [KNL] Change console messages format862 default863 By default we print messages on consoles in864 "[time stamp] text\n" format (time stamp may not be865 printed, depending on CONFIG_PRINTK_TIME or866 `printk_time' param).867 syslog868 Switch to syslog format: "<%u>[time stamp] text\n"869 IOW, each message will have a facility and loglevel870 prefix. The format is similar to one used by syslog()871 syscall, or to executing "dmesg -S --raw" or to reading872 from /proc/kmsg.873 874 consoleblank= [KNL] The console blank (screen saver) timeout in875 seconds. A value of 0 disables the blank timer.876 Defaults to 0.877 878 coredump_filter=879 [KNL] Change the default value for880 /proc/<pid>/coredump_filter.881 See also Documentation/filesystems/proc.rst.882 883 coresight_cpu_debug.enable884 [ARM,ARM64]885 Format: <bool>886 Enable/disable the CPU sampling based debugging.887 0: default value, disable debugging888 1: enable debugging at boot time889 890 cpcihp_generic= [HW,PCI] Generic port I/O CompactPCI driver891 Format:892 <first_slot>,<last_slot>,<port>,<enum_bit>[,<debug>]893 894 cpuidle.off=1 [CPU_IDLE]895 disable the cpuidle sub-system896 897 cpuidle.governor=898 [CPU_IDLE] Name of the cpuidle governor to use.899 900 cpufreq.off=1 [CPU_FREQ]901 disable the cpufreq sub-system902 903 cpufreq.default_governor=904 [CPU_FREQ] Name of the default cpufreq governor or905 policy to use. This governor must be registered in the906 kernel before the cpufreq driver probes.907 908 cpu_init_udelay=N909 [X86,EARLY] Delay for N microsec between assert and de-assert910 of APIC INIT to start processors. This delay occurs911 on every CPU online, such as boot, and resume from suspend.912 Default: 10000913 914 cpuhp.parallel=915 [SMP] Enable/disable parallel bringup of secondary CPUs916 Format: <bool>917 Default is enabled if CONFIG_HOTPLUG_PARALLEL=y. Otherwise918 the parameter has no effect.919 920 crash_kexec_post_notifiers921 Run kdump after running panic-notifiers and dumping922 kmsg. This only for the users who doubt kdump always923 succeeds in any situation.924 Note that this also increases risks of kdump failure,925 because some panic notifiers can make the crashed926 kernel more unstable.927 928 crashkernel=size[KMG][@offset[KMG]]929 [KNL,EARLY] Using kexec, Linux can switch to a 'crash kernel'930 upon panic. This parameter reserves the physical931 memory region [offset, offset + size] for that kernel932 image. If '@offset' is omitted, then a suitable offset933 is selected automatically.934 [KNL, X86-64, ARM64, RISCV, LoongArch] Select a region935 under 4G first, and fall back to reserve region above936 4G when '@offset' hasn't been specified.937 See Documentation/admin-guide/kdump/kdump.rst for further details.938 939 crashkernel=range1:size1[,range2:size2,...][@offset]940 [KNL] Same as above, but depends on the memory941 in the running system. The syntax of range is942 start-[end] where start and end are both943 a memory unit (amount[KMG]). See also944 Documentation/admin-guide/kdump/kdump.rst for an example.945 946 crashkernel=size[KMG],high947 [KNL, X86-64, ARM64, RISCV, LoongArch] range could be948 above 4G.949 Allow kernel to allocate physical memory region from top,950 so could be above 4G if system have more than 4G ram951 installed. Otherwise memory region will be allocated952 below 4G, if available.953 It will be ignored if crashkernel=X is specified.954 crashkernel=size[KMG],low955 [KNL, X86-64, ARM64, RISCV, LoongArch] range under 4G.956 When crashkernel=X,high is passed, kernel could allocate957 physical memory region above 4G, that cause second kernel958 crash on system that require some amount of low memory,959 e.g. swiotlb requires at least 64M+32K low memory, also960 enough extra low memory is needed to make sure DMA buffers961 for 32-bit devices won't run out. Kernel would try to allocate962 default size of memory below 4G automatically. The default963 size is platform dependent.964 --> x86: max(swiotlb_size_or_default() + 8MiB, 256MiB)965 --> arm64: 128MiB966 --> riscv: 128MiB967 --> loongarch: 128MiB968 This one lets the user specify own low range under 4G969 for second kernel instead.970 0: to disable low allocation.971 It will be ignored when crashkernel=X,high is not used972 or memory reserved is below 4G.973 974 cryptomgr.notests975 [KNL] Disable crypto self-tests976 977 cs89x0_dma= [HW,NET]978 Format: <dma>979 980 cs89x0_media= [HW,NET]981 Format: { rj45 | aui | bnc }982 983 csdlock_debug= [KNL] Enable or disable debug add-ons of cross-CPU984 function call handling. When switched on,985 additional debug data is printed to the console986 in case a hanging CPU is detected, and that987 CPU is pinged again in order to try to resolve988 the hang situation. The default value of this989 option depends on the CSD_LOCK_WAIT_DEBUG_DEFAULT990 Kconfig option.991 992 dasd= [HW,NET]993 See header of drivers/s390/block/dasd_devmap.c.994 995 db9.dev[2|3]= [HW,JOY] Multisystem joystick support via parallel port996 (one device per port)997 Format: <port#>,<type>998 See also Documentation/input/devices/joystick-parport.rst999 1000 debug [KNL,EARLY] Enable kernel debugging (events log level).1001 1002 debug_boot_weak_hash1003 [KNL,EARLY] Enable printing [hashed] pointers early in the1004 boot sequence. If enabled, we use a weak hash instead1005 of siphash to hash pointers. Use this option if you are1006 seeing instances of '(___ptrval___)') and need to see a1007 value (hashed pointer) instead. Cryptographically1008 insecure, please do not use on production kernels.1009 1010 debug_locks_verbose=1011 [KNL] verbose locking self-tests1012 Format: <int>1013 Print debugging info while doing the locking API1014 self-tests.1015 Bitmask for the various LOCKTYPE_ tests. Defaults to 01016 (no extra messages), setting it to -1 (all bits set)1017 will print _a_lot_ more information - normally only1018 useful to lockdep developers.1019 1020 debug_objects [KNL,EARLY] Enable object debugging1021 1022 debug_guardpage_minorder=1023 [KNL,EARLY] When CONFIG_DEBUG_PAGEALLOC is set, this1024 parameter allows control of the order of pages that will1025 be intentionally kept free (and hence protected) by the1026 buddy allocator. Bigger value increase the probability1027 of catching random memory corruption, but reduce the1028 amount of memory for normal system use. The maximum1029 possible value is MAX_PAGE_ORDER/2. Setting this1030 parameter to 1 or 2 should be enough to identify most1031 random memory corruption problems caused by bugs in1032 kernel or driver code when a CPU writes to (or reads1033 from) a random memory location. Note that there exists1034 a class of memory corruptions problems caused by buggy1035 H/W or F/W or by drivers badly programming DMA1036 (basically when memory is written at bus level and the1037 CPU MMU is bypassed) which are not detectable by1038 CONFIG_DEBUG_PAGEALLOC, hence this option will not1039 help tracking down these problems.1040 1041 debug_pagealloc=1042 [KNL,EARLY] When CONFIG_DEBUG_PAGEALLOC is set, this parameter1043 enables the feature at boot time. By default, it is1044 disabled and the system will work mostly the same as a1045 kernel built without CONFIG_DEBUG_PAGEALLOC.1046 Note: to get most of debug_pagealloc error reports, it's1047 useful to also enable the page_owner functionality.1048 on: enable the feature1049 1050 debugfs= [KNL,EARLY] This parameter enables what is exposed to1051 userspace and debugfs internal clients.1052 Format: { on, no-mount, off }1053 on: All functions are enabled.1054 no-mount:1055 Filesystem is not registered but kernel clients can1056 access APIs and a crashkernel can be used to read1057 its content. There is nothing to mount.1058 off: Filesystem is not registered and clients1059 get a -EPERM as result when trying to register files1060 or directories within debugfs.1061 This is equivalent of the runtime functionality if1062 debugfs was not enabled in the kernel at all.1063 Default value is set in build-time with a kernel configuration.1064 1065 debugpat [X86] Enable PAT debugging1066 1067 default_hugepagesz=1068 [HW] The size of the default HugeTLB page. This is1069 the size represented by the legacy /proc/ hugepages1070 APIs. In addition, this is the default hugetlb size1071 used for shmget(), mmap() and mounting hugetlbfs1072 filesystems. If not specified, defaults to the1073 architecture's default huge page size. Huge page1074 sizes are architecture dependent. See also1075 Documentation/admin-guide/mm/hugetlbpage.rst.1076 Format: size[KMG]1077 1078 deferred_probe_timeout=1079 [KNL] Debugging option to set a timeout in seconds for1080 deferred probe to give up waiting on dependencies to1081 probe. Only specific dependencies (subsystems or1082 drivers) that have opted in will be ignored. A timeout1083 of 0 will timeout at the end of initcalls. If the time1084 out hasn't expired, it'll be restarted by each1085 successful driver registration. This option will also1086 dump out devices still on the deferred probe list after1087 retrying.1088 1089 delayacct [KNL] Enable per-task delay accounting1090 1091 dell_smm_hwmon.ignore_dmi=1092 [HW] Continue probing hardware even if DMI data1093 indicates that the driver is running on unsupported1094 hardware.1095 1096 dell_smm_hwmon.force=1097 [HW] Activate driver even if SMM BIOS signature does1098 not match list of supported models and enable otherwise1099 blacklisted features.1100 1101 dell_smm_hwmon.power_status=1102 [HW] Report power status in /proc/i8k1103 (disabled by default).1104 1105 dell_smm_hwmon.restricted=1106 [HW] Allow controlling fans only if SYS_ADMIN1107 capability is set.1108 1109 dell_smm_hwmon.fan_mult=1110 [HW] Factor to multiply fan speed with.1111 1112 dell_smm_hwmon.fan_max=1113 [HW] Maximum configurable fan speed.1114 1115 dfltcc= [HW,S390]1116 Format: { on | off | def_only | inf_only | always }1117 on: s390 zlib hardware support for compression on1118 level 1 and decompression (default)1119 off: No s390 zlib hardware support1120 def_only: s390 zlib hardware support for deflate1121 only (compression on level 1)1122 inf_only: s390 zlib hardware support for inflate1123 only (decompression)1124 always: Same as 'on' but ignores the selected compression1125 level always using hardware support (used for debugging)1126 1127 dhash_entries= [KNL]1128 Set number of hash buckets for dentry cache.1129 1130 disable_1tb_segments [PPC,EARLY]1131 Disables the use of 1TB hash page table segments. This1132 causes the kernel to fall back to 256MB segments which1133 can be useful when debugging issues that require an SLB1134 miss to occur.1135 1136 disable= [IPV6]1137 See Documentation/networking/ipv6.rst.1138 1139 disable_radix [PPC,EARLY]1140 Disable RADIX MMU mode on POWER91141 1142 disable_tlbie [PPC]1143 Disable TLBIE instruction. Currently does not work1144 with KVM, with HASH MMU, or with coherent accelerators.1145 1146 disable_ddw [PPC/PSERIES,EARLY]1147 Disable Dynamic DMA Window support. Use this1148 to workaround buggy firmware.1149 1150 disable_ipv6= [IPV6]1151 See Documentation/networking/ipv6.rst.1152 1153 disable_mtrr_cleanup [X86,EARLY]1154 The kernel tries to adjust MTRR layout from continuous1155 to discrete, to make X server driver able to add WB1156 entry later. This parameter disables that.1157 1158 disable_mtrr_trim [X86, Intel and AMD only,EARLY]1159 By default the kernel will trim any uncacheable1160 memory out of your available memory pool based on1161 MTRR settings. This parameter disables that behavior,1162 possibly causing your machine to run very slowly.1163 1164 disable_timer_pin_1 [X86,EARLY]1165 Disable PIN 1 of APIC timer1166 Can be useful to work around chipset bugs.1167 1168 dis_ucode_ldr [X86] Disable the microcode loader.1169 1170 dma_debug=off If the kernel is compiled with DMA_API_DEBUG support,1171 this option disables the debugging code at boot.1172 1173 dma_debug_entries=<number>1174 This option allows to tune the number of preallocated1175 entries for DMA-API debugging code. One entry is1176 required per DMA-API allocation. Use this if the1177 DMA-API debugging code disables itself because the1178 architectural default is too low.1179 1180 dma_debug_driver=<driver_name>1181 With this option the DMA-API debugging driver1182 filter feature can be enabled at boot time. Just1183 pass the driver to filter for as the parameter.1184 The filter can be disabled or changed to another1185 driver later using sysfs.1186 1187 reg_file_data_sampling=1188 [X86] Controls mitigation for Register File Data1189 Sampling (RFDS) vulnerability. RFDS is a CPU1190 vulnerability which may allow userspace to infer1191 kernel data values previously stored in floating point1192 registers, vector registers, or integer registers.1193 RFDS only affects Intel Atom processors.1194 1195 on: Turns ON the mitigation.1196 off: Turns OFF the mitigation.1197 1198 This parameter overrides the compile time default set1199 by CONFIG_MITIGATION_RFDS. Mitigation cannot be1200 disabled when other VERW based mitigations (like MDS)1201 are enabled. In order to disable RFDS mitigation all1202 VERW based mitigations need to be disabled.1203 1204 For details see:1205 Documentation/admin-guide/hw-vuln/reg-file-data-sampling.rst1206 1207 driver_async_probe= [KNL]1208 List of driver names to be probed asynchronously. *1209 matches with all driver names. If * is specified, the1210 rest of the listed driver names are those that will NOT1211 match the *.1212 Format: <driver_name1>,<driver_name2>...1213 1214 drm.edid_firmware=[<connector>:]<file>[,[<connector>:]<file>]1215 Broken monitors, graphic adapters, KVMs and EDIDless1216 panels may send no or incorrect EDID data sets.1217 This parameter allows to specify an EDID data sets1218 in the /lib/firmware directory that are used instead.1219 An EDID data set will only be used for a particular1220 connector, if its name and a colon are prepended to1221 the EDID name. Each connector may use a unique EDID1222 data set by separating the files with a comma. An EDID1223 data set with no connector name will be used for1224 any connectors not explicitly specified.1225 1226 dscc4.setup= [NET]1227 1228 dt_cpu_ftrs= [PPC,EARLY]1229 Format: {"off" | "known"}1230 Control how the dt_cpu_ftrs device-tree binding is1231 used for CPU feature discovery and setup (if it1232 exists).1233 off: Do not use it, fall back to legacy cpu table.1234 known: Do not pass through unknown features to guests1235 or userspace, only those that the kernel is aware of.1236 1237 dump_apple_properties [X86]1238 Dump name and content of EFI device properties on1239 x86 Macs. Useful for driver authors to determine1240 what data is available or for reverse-engineering.1241 1242 dyndbg[="val"] [KNL,DYNAMIC_DEBUG]1243 <module>.dyndbg[="val"]1244 Enable debug messages at boot time. See1245 Documentation/admin-guide/dynamic-debug-howto.rst1246 for details.1247 1248 early_ioremap_debug [KNL,EARLY]1249 Enable debug messages in early_ioremap support. This1250 is useful for tracking down temporary early mappings1251 which are not unmapped.1252 1253 earlycon= [KNL,EARLY] Output early console device and options.1254 1255 When used with no options, the early console is1256 determined by stdout-path property in device tree's1257 chosen node or the ACPI SPCR table if supported by1258 the platform.1259 1260 cdns,<addr>[,options]1261 Start an early, polled-mode console on a Cadence1262 (xuartps) serial port at the specified address. Only1263 supported option is baud rate. If baud rate is not1264 specified, the serial port must already be setup and1265 configured.1266 1267 uart[8250],io,<addr>[,options[,uartclk]]1268 uart[8250],mmio,<addr>[,options[,uartclk]]1269 uart[8250],mmio32,<addr>[,options[,uartclk]]1270 uart[8250],mmio32be,<addr>[,options[,uartclk]]1271 uart[8250],0x<addr>[,options]1272 Start an early, polled-mode console on the 8250/165501273 UART at the specified I/O port or MMIO address.1274 MMIO inter-register address stride is either 8-bit1275 (mmio) or 32-bit (mmio32 or mmio32be).1276 If none of [io|mmio|mmio32|mmio32be], <addr> is assumed1277 to be equivalent to 'mmio'. 'options' are specified1278 in the same format described for "console=ttyS<n>"; if1279 unspecified, the h/w is not initialized. 'uartclk' is1280 the uart clock frequency; if unspecified, it is set1281 to 'BASE_BAUD' * 16.1282 1283 pl011,<addr>1284 pl011,mmio32,<addr>1285 Start an early, polled-mode console on a pl011 serial1286 port at the specified address. The pl011 serial port1287 must already be setup and configured. Options are not1288 yet supported. If 'mmio32' is specified, then only1289 the driver will use only 32-bit accessors to read/write1290 the device registers.1291 1292 liteuart,<addr>1293 Start an early console on a litex serial port at the1294 specified address. The serial port must already be1295 setup and configured. Options are not yet supported.1296 1297 meson,<addr>1298 Start an early, polled-mode console on a meson serial1299 port at the specified address. The serial port must1300 already be setup and configured. Options are not yet1301 supported.1302 1303 msm_serial,<addr>1304 Start an early, polled-mode console on an msm serial1305 port at the specified address. The serial port1306 must already be setup and configured. Options are not1307 yet supported.1308 1309 msm_serial_dm,<addr>1310 Start an early, polled-mode console on an msm serial1311 dm port at the specified address. The serial port1312 must already be setup and configured. Options are not1313 yet supported.1314 1315 owl,<addr>1316 Start an early, polled-mode console on a serial port1317 of an Actions Semi SoC, such as S500 or S900, at the1318 specified address. The serial port must already be1319 setup and configured. Options are not yet supported.1320 1321 rda,<addr>1322 Start an early, polled-mode console on a serial port1323 of an RDA Micro SoC, such as RDA8810PL, at the1324 specified address. The serial port must already be1325 setup and configured. Options are not yet supported.1326 1327 sbi1328 Use RISC-V SBI (Supervisor Binary Interface) for early1329 console.1330 1331 smh Use ARM semihosting calls for early console.1332 1333 s3c2410,<addr>1334 s3c2412,<addr>1335 s3c2440,<addr>1336 s3c6400,<addr>1337 s5pv210,<addr>1338 exynos4210,<addr>1339 Use early console provided by serial driver available1340 on Samsung SoCs, requires selecting proper type and1341 a correct base address of the selected UART port. The1342 serial port must already be setup and configured.1343 Options are not yet supported.1344 1345 lantiq,<addr>1346 Start an early, polled-mode console on a lantiq serial1347 (lqasc) port at the specified address. The serial port1348 must already be setup and configured. Options are not1349 yet supported.1350 1351 lpuart,<addr>1352 lpuart32,<addr>1353 Use early console provided by Freescale LP UART driver1354 found on Freescale Vybrid and QorIQ LS1021A processors.1355 A valid base address must be provided, and the serial1356 port must already be setup and configured.1357 1358 ec_imx21,<addr>1359 ec_imx6q,<addr>1360 Start an early, polled-mode, output-only console on the1361 Freescale i.MX UART at the specified address. The UART1362 must already be setup and configured.1363 1364 ar3700_uart,<addr>1365 Start an early, polled-mode console on the1366 Armada 3700 serial port at the specified1367 address. The serial port must already be setup1368 and configured. Options are not yet supported.1369 1370 qcom_geni,<addr>1371 Start an early, polled-mode console on a Qualcomm1372 Generic Interface (GENI) based serial port at the1373 specified address. The serial port must already be1374 setup and configured. Options are not yet supported.1375 1376 efifb,[options]1377 Start an early, unaccelerated console on the EFI1378 memory mapped framebuffer (if available). On cache1379 coherent non-x86 systems that use system memory for1380 the framebuffer, pass the 'ram' option so that it is1381 mapped with the correct attributes.1382 1383 linflex,<addr>1384 Use early console provided by Freescale LINFlexD UART1385 serial driver for NXP S32V234 SoCs. A valid base1386 address must be provided, and the serial port must1387 already be setup and configured.1388 1389 earlyprintk= [X86,SH,ARM,M68k,S390,UM,EARLY]1390 earlyprintk=vga1391 earlyprintk=sclp1392 earlyprintk=xen1393 earlyprintk=serial[,ttySn[,baudrate]]1394 earlyprintk=serial[,0x...[,baudrate]]1395 earlyprintk=ttySn[,baudrate]1396 earlyprintk=dbgp[debugController#]1397 earlyprintk=pciserial[,force],bus:device.function[,baudrate]1398 earlyprintk=xdbc[xhciController#]1399 earlyprintk=bios1400 1401 earlyprintk is useful when the kernel crashes before1402 the normal console is initialized. It is not enabled by1403 default because it has some cosmetic problems.1404 1405 Append ",keep" to not disable it when the real console1406 takes over.1407 1408 Only one of vga, serial, or usb debug port can1409 be used at a time.1410 1411 Currently only ttyS0 and ttyS1 may be specified by1412 name. Other I/O ports may be explicitly specified1413 on some architectures (x86 and arm at least) by1414 replacing ttySn with an I/O port address, like this:1415 earlyprintk=serial,0x1008,1152001416 You can find the port for a given device in1417 /proc/tty/driver/serial:1418 2: uart:ST16650V2 port:00001008 irq:18 ...1419 1420 Interaction with the standard serial driver is not1421 very good.1422 1423 The VGA output is eventually overwritten by1424 the real console.1425 1426 The xen option can only be used in Xen domains.1427 1428 The sclp output can only be used on s390.1429 1430 The bios output can only be used on SuperH.1431 1432 The optional "force" to "pciserial" enables use of a1433 PCI device even when its classcode is not of the1434 UART class.1435 1436 edac_report= [HW,EDAC] Control how to report EDAC event1437 Format: {"on" | "off" | "force"}1438 on: enable EDAC to report H/W event. May be overridden1439 by other higher priority error reporting module.1440 off: disable H/W event reporting through EDAC.1441 force: enforce the use of EDAC to report H/W event.1442 default: on.1443 1444 edd= [EDD]1445 Format: {"off" | "on" | "skip[mbr]"}1446 1447 efi= [EFI,EARLY]1448 Format: { "debug", "disable_early_pci_dma",1449 "nochunk", "noruntime", "nosoftreserve",1450 "novamap", "no_disable_early_pci_dma" }1451 debug: enable misc debug output.1452 disable_early_pci_dma: disable the busmaster bit on all1453 PCI bridges while in the EFI boot stub.1454 nochunk: disable reading files in "chunks" in the EFI1455 boot stub, as chunking can cause problems with some1456 firmware implementations.1457 noruntime : disable EFI runtime services support1458 nosoftreserve: The EFI_MEMORY_SP (Specific Purpose)1459 attribute may cause the kernel to reserve the1460 memory range for a memory mapping driver to1461 claim. Specify efi=nosoftreserve to disable this1462 reservation and treat the memory by its base type1463 (i.e. EFI_CONVENTIONAL_MEMORY / "System RAM").1464 novamap: do not call SetVirtualAddressMap().1465 no_disable_early_pci_dma: Leave the busmaster bit set1466 on all PCI bridges while in the EFI boot stub1467 1468 efi_no_storage_paranoia [EFI,X86,EARLY]1469 Using this parameter you can use more than 50% of1470 your efi variable storage. Use this parameter only if1471 you are really sure that your UEFI does sane gc and1472 fulfills the spec otherwise your board may brick.1473 1474 efivar_ssdt= [EFI; X86] Name of an EFI variable that contains an SSDT1475 that is to be dynamically loaded by Linux. If there are1476 multiple variables with the same name but with different1477 vendor GUIDs, all of them will be loaded. See1478 Documentation/admin-guide/acpi/ssdt-overlays.rst for details.1479 1480 1481 eisa_irq_edge= [PARISC,HW]1482 See header of drivers/parisc/eisa.c.1483 1484 ekgdboc= [X86,KGDB,EARLY] Allow early kernel console debugging1485 Format: ekgdboc=kbd1486 1487 This is designed to be used in conjunction with1488 the boot argument: earlyprintk=vga1489 1490 This parameter works in place of the kgdboc parameter1491 but can only be used if the backing tty is available1492 very early in the boot process. For early debugging1493 via a serial port see kgdboc_earlycon instead.1494 1495 elanfreq= [X86-32]1496 See comment before function elanfreq_setup() in1497 arch/x86/kernel/cpu/cpufreq/elanfreq.c.1498 1499 elfcorehdr=[size[KMG]@]offset[KMG] [PPC,SH,X86,S390,EARLY]1500 Specifies physical address of start of kernel core1501 image elf header and optionally the size. Generally1502 kexec loader will pass this option to capture kernel.1503 See Documentation/admin-guide/kdump/kdump.rst for details.1504 1505 enable_mtrr_cleanup [X86,EARLY]1506 The kernel tries to adjust MTRR layout from continuous1507 to discrete, to make X server driver able to add WB1508 entry later. This parameter enables that.1509 1510 enable_timer_pin_1 [X86]1511 Enable PIN 1 of APIC timer1512 Can be useful to work around chipset bugs1513 (in particular on some ATI chipsets).1514 The kernel tries to set a reasonable default.1515 1516 enforcing= [SELINUX] Set initial enforcing status.1517 Format: {"0" | "1"}1518 See security/selinux/Kconfig help text.1519 0 -- permissive (log only, no denials).1520 1 -- enforcing (deny and log).1521 Default value is 0.1522 Value can be changed at runtime via1523 /sys/fs/selinux/enforce.1524 1525 erst_disable [ACPI]1526 Disable Error Record Serialization Table (ERST)1527 support.1528 1529 ether= [HW,NET] Ethernet cards parameters1530 This option is obsoleted by the "netdev=" option, which1531 has equivalent usage. See its documentation for details.1532 1533 evm= [EVM]1534 Format: { "fix" }1535 Permit 'security.evm' to be updated regardless of1536 current integrity status.1537 1538 early_page_ext [KNL,EARLY] Enforces page_ext initialization to earlier1539 stages so cover more early boot allocations.1540 Please note that as side effect some optimizations1541 might be disabled to achieve that (e.g. parallelized1542 memory initialization is disabled) so the boot process1543 might take longer, especially on systems with a lot of1544 memory. Available with CONFIG_PAGE_EXTENSION=y.1545 1546 failslab=1547 fail_usercopy=1548 fail_page_alloc=1549 fail_make_request=[KNL]1550 General fault injection mechanism.1551 Format: <interval>,<probability>,<space>,<times>1552 See also Documentation/fault-injection/.1553 1554 fb_tunnels= [NET]1555 Format: { initns | none }1556 See Documentation/admin-guide/sysctl/net.rst for1557 fb_tunnels_only_for_init_ns1558 1559 floppy= [HW]1560 See Documentation/admin-guide/blockdev/floppy.rst.1561 1562 forcepae [X86-32]1563 Forcefully enable Physical Address Extension (PAE).1564 Many Pentium M systems disable PAE but may have a1565 functionally usable PAE implementation.1566 Warning: use of this parameter will taint the kernel1567 and may cause unknown problems.1568 1569 fred= [X86-64]1570 Enable/disable Flexible Return and Event Delivery.1571 Format: { on | off }1572 on: enable FRED when it's present.1573 off: disable FRED, the default setting.1574 1575 ftrace=[tracer]1576 [FTRACE] will set and start the specified tracer1577 as early as possible in order to facilitate early1578 boot debugging.1579 1580 ftrace_boot_snapshot1581 [FTRACE] On boot up, a snapshot will be taken of the1582 ftrace ring buffer that can be read at:1583 /sys/kernel/tracing/snapshot.1584 This is useful if you need tracing information from kernel1585 boot up that is likely to be overridden by user space1586 start up functionality.1587 1588 Optionally, the snapshot can also be defined for a tracing1589 instance that was created by the trace_instance= command1590 line parameter.1591 1592 trace_instance=foo,sched_switch ftrace_boot_snapshot=foo1593 1594 The above will cause the "foo" tracing instance to trigger1595 a snapshot at the end of boot up.1596 1597 ftrace_dump_on_oops[=2(orig_cpu) | =<instance>][,<instance> |1598 ,<instance>=2(orig_cpu)]1599 [FTRACE] will dump the trace buffers on oops.1600 If no parameter is passed, ftrace will dump global1601 buffers of all CPUs, if you pass 2 or orig_cpu, it1602 will dump only the buffer of the CPU that triggered1603 the oops, or the specific instance will be dumped if1604 its name is passed. Multiple instance dump is also1605 supported, and instances are separated by commas. Each1606 instance supports only dump on CPU that triggered the1607 oops by passing 2 or orig_cpu to it.1608 1609 ftrace_dump_on_oops=foo=orig_cpu1610 1611 The above will dump only the buffer of "foo" instance1612 on CPU that triggered the oops.1613 1614 ftrace_dump_on_oops,foo,bar=orig_cpu1615 1616 The above will dump global buffer on all CPUs, the1617 buffer of "foo" instance on all CPUs and the buffer1618 of "bar" instance on CPU that triggered the oops.1619 1620 ftrace_filter=[function-list]1621 [FTRACE] Limit the functions traced by the function1622 tracer at boot up. function-list is a comma-separated1623 list of functions. This list can be changed at run1624 time by the set_ftrace_filter file in the debugfs1625 tracing directory.1626 1627 ftrace_notrace=[function-list]1628 [FTRACE] Do not trace the functions specified in1629 function-list. This list can be changed at run time1630 by the set_ftrace_notrace file in the debugfs1631 tracing directory.1632 1633 ftrace_graph_filter=[function-list]1634 [FTRACE] Limit the top level callers functions traced1635 by the function graph tracer at boot up.1636 function-list is a comma-separated list of functions1637 that can be changed at run time by the1638 set_graph_function file in the debugfs tracing directory.1639 1640 ftrace_graph_notrace=[function-list]1641 [FTRACE] Do not trace from the functions specified in1642 function-list. This list is a comma-separated list of1643 functions that can be changed at run time by the1644 set_graph_notrace file in the debugfs tracing directory.1645 1646 ftrace_graph_max_depth=<uint>1647 [FTRACE] Used with the function graph tracer. This is1648 the max depth it will trace into a function. This value1649 can be changed at run time by the max_graph_depth file1650 in the tracefs tracing directory. default: 0 (no limit)1651 1652 fw_devlink= [KNL,EARLY] Create device links between consumer and supplier1653 devices by scanning the firmware to infer the1654 consumer/supplier relationships. This feature is1655 especially useful when drivers are loaded as modules as1656 it ensures proper ordering of tasks like device probing1657 (suppliers first, then consumers), supplier boot state1658 clean up (only after all consumers have probed),1659 suspend/resume & runtime PM (consumers first, then1660 suppliers).1661 Format: { off | permissive | on | rpm }1662 off -- Don't create device links from firmware info.1663 permissive -- Create device links from firmware info1664 but use it only for ordering boot state clean1665 up (sync_state() calls).1666 on -- Create device links from firmware info and use it1667 to enforce probe and suspend/resume ordering.1668 rpm -- Like "on", but also use to order runtime PM.1669 1670 fw_devlink.strict=<bool>1671 [KNL,EARLY] Treat all inferred dependencies as mandatory1672 dependencies. This only applies for fw_devlink=on|rpm.1673 Format: <bool>1674 1675 fw_devlink.sync_state =1676 [KNL,EARLY] When all devices that could probe have finished1677 probing, this parameter controls what to do with1678 devices that haven't yet received their sync_state()1679 calls.1680 Format: { strict | timeout }1681 strict -- Default. Continue waiting on consumers to1682 probe successfully.1683 timeout -- Give up waiting on consumers and call1684 sync_state() on any devices that haven't yet1685 received their sync_state() calls after1686 deferred_probe_timeout has expired or by1687 late_initcall() if !CONFIG_MODULES.1688 1689 gamecon.map[2|3]=1690 [HW,JOY] Multisystem joystick and NES/SNES/PSX pad1691 support via parallel port (up to 5 devices per port)1692 Format: <port#>,<pad1>,<pad2>,<pad3>,<pad4>,<pad5>1693 See also Documentation/input/devices/joystick-parport.rst1694 1695 gamma= [HW,DRM]1696 1697 gart_fix_e820= [X86-64,EARLY] disable the fix e820 for K8 GART1698 Format: off | on1699 default: on1700 1701 gather_data_sampling=1702 [X86,INTEL,EARLY] Control the Gather Data Sampling (GDS)1703 mitigation.1704 1705 Gather Data Sampling is a hardware vulnerability which1706 allows unprivileged speculative access to data which was1707 previously stored in vector registers.1708 1709 This issue is mitigated by default in updated microcode.1710 The mitigation may have a performance impact but can be1711 disabled. On systems without the microcode mitigation1712 disabling AVX serves as a mitigation.1713 1714 force: Disable AVX to mitigate systems without1715 microcode mitigation. No effect if the microcode1716 mitigation is present. Known to cause crashes in1717 userspace with buggy AVX enumeration.1718 1719 off: Disable GDS mitigation.1720 1721 gcov_persist= [GCOV] When non-zero (default), profiling data for1722 kernel modules is saved and remains accessible via1723 debugfs, even when the module is unloaded/reloaded.1724 When zero, profiling data is discarded and associated1725 debugfs files are removed at module unload time.1726 1727 goldfish [X86] Enable the goldfish android emulator platform.1728 Don't use this when you are not running on the1729 android emulator1730 1731 gpio-mockup.gpio_mockup_ranges1732 [HW] Sets the ranges of gpiochip of for this device.1733 Format: <start1>,<end1>,<start2>,<end2>...1734 gpio-mockup.gpio_mockup_named_lines1735 [HW] Let the driver know GPIO lines should be named.1736 1737 gpt [EFI] Forces disk with valid GPT signature but1738 invalid Protective MBR to be treated as GPT. If the1739 primary GPT is corrupted, it enables the backup/alternate1740 GPT to be used instead.1741 1742 grcan.enable0= [HW] Configuration of physical interface 0. Determines1743 the "Enable 0" bit of the configuration register.1744 Format: 0 | 11745 Default: 01746 grcan.enable1= [HW] Configuration of physical interface 1. Determines1747 the "Enable 0" bit of the configuration register.1748 Format: 0 | 11749 Default: 01750 grcan.select= [HW] Select which physical interface to use.1751 Format: 0 | 11752 Default: 01753 grcan.txsize= [HW] Sets the size of the tx buffer.1754 Format: <unsigned int> such that (txsize & ~0x1fffc0) == 0.1755 Default: 10241756 grcan.rxsize= [HW] Sets the size of the rx buffer.1757 Format: <unsigned int> such that (rxsize & ~0x1fffc0) == 0.1758 Default: 10241759 1760 hardened_usercopy=1761 [KNL] Under CONFIG_HARDENED_USERCOPY, whether1762 hardening is enabled for this boot. Hardened1763 usercopy checking is used to protect the kernel1764 from reading or writing beyond known memory1765 allocation boundaries as a proactive defense1766 against bounds-checking flaws in the kernel's1767 copy_to_user()/copy_from_user() interface.1768 on Perform hardened usercopy checks (default).1769 off Disable hardened usercopy checks.1770 1771 hardlockup_all_cpu_backtrace=1772 [KNL] Should the hard-lockup detector generate1773 backtraces on all cpus.1774 Format: 0 | 11775 1776 hashdist= [KNL,NUMA] Large hashes allocated during boot1777 are distributed across NUMA nodes. Defaults on1778 for 64-bit NUMA, off otherwise.1779 Format: 0 | 1 (for off | on)1780 1781 hd= [EIDE] (E)IDE hard drive subsystem geometry1782 Format: <cyl>,<head>,<sect>1783 1784 hest_disable [ACPI]1785 Disable Hardware Error Source Table (HEST) support;1786 corresponding firmware-first mode error processing1787 logic will be disabled.1788 1789 hibernate= [HIBERNATION]1790 noresume Don't check if there's a hibernation image1791 present during boot.1792 nocompress Don't compress/decompress hibernation images.1793 no Disable hibernation and resume.1794 protect_image Turn on image protection during restoration1795 (that will set all pages holding image data1796 during restoration read-only).1797 1798 hibernate.compressor= [HIBERNATION] Compression algorithm to be1799 used with hibernation.1800 Format: { lzo | lz4 }1801 Default: lzo1802 1803 lzo: Select LZO compression algorithm to1804 compress/decompress hibernation image.1805 1806 lz4: Select LZ4 compression algorithm to1807 compress/decompress hibernation image.1808 1809 highmem=nn[KMG] [KNL,BOOT,EARLY] forces the highmem zone to have an exact1810 size of <nn>. This works even on boxes that have no1811 highmem otherwise. This also works to reduce highmem1812 size on bigger boxes.1813 1814 highres= [KNL] Enable/disable high resolution timer mode.1815 Valid parameters: "on", "off"1816 Default: "on"1817 1818 hlt [BUGS=ARM,SH]1819 1820 hostname= [KNL,EARLY] Set the hostname (aka UTS nodename).1821 Format: <string>1822 This allows setting the system's hostname during early1823 startup. This sets the name returned by gethostname.1824 Using this parameter to set the hostname makes it1825 possible to ensure the hostname is correctly set before1826 any userspace processes run, avoiding the possibility1827 that a process may call gethostname before the hostname1828 has been explicitly set, resulting in the calling1829 process getting an incorrect result. The string must1830 not exceed the maximum allowed hostname length (usually1831 64 characters) and will be truncated otherwise.1832 1833 hpet= [X86-32,HPET] option to control HPET usage1834 Format: { enable (default) | disable | force |1835 verbose }1836 disable: disable HPET and use PIT instead1837 force: allow force enabled of undocumented chips (ICH4,1838 VIA, nVidia)1839 verbose: show contents of HPET registers during setup1840 1841 hpet_mmap= [X86, HPET_MMAP] Allow userspace to mmap HPET1842 registers. Default set by CONFIG_HPET_MMAP_DEFAULT.1843 1844 hugepages= [HW] Number of HugeTLB pages to allocate at boot.1845 If this follows hugepagesz (below), it specifies1846 the number of pages of hugepagesz to be allocated.1847 If this is the first HugeTLB parameter on the command1848 line, it specifies the number of pages to allocate for1849 the default huge page size. If using node format, the1850 number of pages to allocate per-node can be specified.1851 See also Documentation/admin-guide/mm/hugetlbpage.rst.1852 Format: <integer> or (node format)1853 <node>:<integer>[,<node>:<integer>]1854 1855 hugepagesz=1856 [HW] The size of the HugeTLB pages. This is used in1857 conjunction with hugepages (above) to allocate huge1858 pages of a specific size at boot. The pair1859 hugepagesz=X hugepages=Y can be specified once for1860 each supported huge page size. Huge page sizes are1861 architecture dependent. See also1862 Documentation/admin-guide/mm/hugetlbpage.rst.1863 Format: size[KMG]1864 1865 hugetlb_cma= [HW,CMA,EARLY] The size of a CMA area used for allocation1866 of gigantic hugepages. Or using node format, the size1867 of a CMA area per node can be specified.1868 Format: nn[KMGTPE] or (node format)1869 <node>:nn[KMGTPE][,<node>:nn[KMGTPE]]1870 1871 Reserve a CMA area of given size and allocate gigantic1872 hugepages using the CMA allocator. If enabled, the1873 boot-time allocation of gigantic hugepages is skipped.1874 1875 hugetlb_free_vmemmap=1876 [KNL] Requires CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP1877 enabled.1878 Control if HugeTLB Vmemmap Optimization (HVO) is enabled.1879 Allows heavy hugetlb users to free up some more1880 memory (7 * PAGE_SIZE for each 2MB hugetlb page).1881 Format: { on | off (default) }1882 1883 on: enable HVO1884 off: disable HVO1885 1886 Built with CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP_DEFAULT_ON=y,1887 the default is on.1888 1889 Note that the vmemmap pages may be allocated from the added1890 memory block itself when memory_hotplug.memmap_on_memory is1891 enabled, those vmemmap pages cannot be optimized even if this1892 feature is enabled. Other vmemmap pages not allocated from1893 the added memory block itself do not be affected.1894 1895 hung_task_panic=1896 [KNL] Should the hung task detector generate panics.1897 Format: 0 | 11898 1899 A value of 1 instructs the kernel to panic when a1900 hung task is detected. The default value is controlled1901 by the CONFIG_BOOTPARAM_HUNG_TASK_PANIC build-time1902 option. The value selected by this boot parameter can1903 be changed later by the kernel.hung_task_panic sysctl.1904 1905 hvc_iucv= [S390] Number of z/VM IUCV hypervisor console (HVC)1906 terminal devices. Valid values: 0..81907 hvc_iucv_allow= [S390] Comma-separated list of z/VM user IDs.1908 If specified, z/VM IUCV HVC accepts connections1909 from listed z/VM user IDs only.1910 1911 hv_nopvspin [X86,HYPER_V,EARLY]1912 Disables the paravirt spinlock optimizations1913 which allow the hypervisor to 'idle' the guest1914 on lock contention.1915 1916 i2c_bus= [HW] Override the default board specific I2C bus speed1917 or register an additional I2C bus that is not1918 registered from board initialization code.1919 Format:1920 <bus_id>,<clkrate>1921 1922 i2c_touchscreen_props= [HW,ACPI,X86]1923 Set device-properties for ACPI-enumerated I2C-attached1924 touchscreen, to e.g. fix coordinates of upside-down1925 mounted touchscreens. If you need this option please1926 submit a drivers/platform/x86/touchscreen_dmi.c patch1927 adding a DMI quirk for this.1928 1929 Format:1930 <ACPI_HW_ID>:<prop_name>=<val>[:prop_name=val][:...]1931 Where <val> is one of:1932 Omit "=<val>" entirely Set a boolean device-property1933 Unsigned number Set a u32 device-property1934 Anything else Set a string device-property1935 1936 Examples (split over multiple lines):1937 i2c_touchscreen_props=GDIX1001:touchscreen-inverted-x:1938 touchscreen-inverted-y1939 1940 i2c_touchscreen_props=MSSL1680:touchscreen-size-x=1920:1941 touchscreen-size-y=1080:touchscreen-inverted-y:1942 firmware-name=gsl1680-vendor-model.fw:silead,home-button1943 1944 i8042.debug [HW] Toggle i8042 debug mode1945 i8042.unmask_kbd_data1946 [HW] Enable printing of interrupt data from the KBD port1947 (disabled by default, and as a pre-condition1948 requires that i8042.debug=1 be enabled)1949 i8042.direct [HW] Put keyboard port into non-translated mode1950 i8042.dumbkbd [HW] Pretend that controller can only read data from1951 keyboard and cannot control its state1952 (Don't attempt to blink the leds)1953 i8042.noaux [HW] Don't check for auxiliary (== mouse) port1954 i8042.nokbd [HW] Don't check/create keyboard port1955 i8042.noloop [HW] Disable the AUX Loopback command while probing1956 for the AUX port1957 i8042.nomux [HW] Don't check presence of an active multiplexing1958 controller1959 i8042.nopnp [HW] Don't use ACPIPnP / PnPBIOS to discover KBD/AUX1960 controllers1961 i8042.notimeout [HW] Ignore timeout condition signalled by controller1962 i8042.reset [HW] Reset the controller during init, cleanup and1963 suspend-to-ram transitions, only during s2r1964 transitions, or never reset1965 Format: { 1 | Y | y | 0 | N | n }1966 1, Y, y: always reset controller1967 0, N, n: don't ever reset controller1968 Default: only on s2r transitions on x86; most other1969 architectures force reset to be always executed1970 i8042.unlock [HW] Unlock (ignore) the keylock1971 i8042.kbdreset [HW] Reset device connected to KBD port1972 i8042.probe_defer1973 [HW] Allow deferred probing upon i8042 probe errors1974 1975 i810= [HW,DRM]1976 1977 i915.invert_brightness=1978 [DRM] Invert the sense of the variable that is used to1979 set the brightness of the panel backlight. Normally a1980 brightness value of 0 indicates backlight switched off,1981 and the maximum of the brightness value sets the backlight1982 to maximum brightness. If this parameter is set to 01983 (default) and the machine requires it, or this parameter1984 is set to 1, a brightness value of 0 sets the backlight1985 to maximum brightness, and the maximum of the brightness1986 value switches the backlight off.1987 -1 -- never invert brightness1988 0 -- machine default1989 1 -- force brightness inversion1990 1991 ia32_emulation= [X86-64]1992 Format: <bool>1993 When true, allows loading 32-bit programs and executing 32-bit1994 syscalls, essentially overriding IA32_EMULATION_DEFAULT_DISABLED at1995 boot time. When false, unconditionally disables IA32 emulation.1996 1997 icn= [HW,ISDN]1998 Format: <io>[,<membase>[,<icn_id>[,<icn_id2>]]]1999 2000 2001 idle= [X86,EARLY]2002 Format: idle=poll, idle=halt, idle=nomwait2003 Poll forces a polling idle loop that can slightly2004 improve the performance of waking up a idle CPU, but2005 will use a lot of power and make the system run hot.2006 Not recommended.2007 idle=halt: Halt is forced to be used for CPU idle.2008 In such case C2/C3 won't be used again.2009 idle=nomwait: Disable mwait for CPU C-states2010 2011 idxd.sva= [HW]2012 Format: <bool>2013 Allow force disabling of Shared Virtual Memory (SVA)2014 support for the idxd driver. By default it is set to2015 true (1).2016 2017 idxd.tc_override= [HW]2018 Format: <bool>2019 Allow override of default traffic class configuration2020 for the device. By default it is set to false (0).2021 2022 ieee754= [MIPS] Select IEEE Std 754 conformance mode2023 Format: { strict | legacy | 2008 | relaxed | emulated }2024 Default: strict2025 2026 Choose which programs will be accepted for execution2027 based on the IEEE 754 NaN encoding(s) supported by2028 the FPU and the NaN encoding requested with the value2029 of an ELF file header flag individually set by each2030 binary. Hardware implementations are permitted to2031 support either or both of the legacy and the 2008 NaN2032 encoding mode.2033 2034 Available settings are as follows:2035 strict accept binaries that request a NaN encoding2036 supported by the FPU2037 legacy only accept legacy-NaN binaries, if supported2038 by the FPU2039 2008 only accept 2008-NaN binaries, if supported2040 by the FPU2041 relaxed accept any binaries regardless of whether2042 supported by the FPU2043 emulated accept any binaries but enable FPU emulator2044 if binary mode is unsupported by the FPU.2045 2046 The FPU emulator is always able to support both NaN2047 encodings, so if no FPU hardware is present or it has2048 been disabled with 'nofpu', then the settings of2049 'legacy' and '2008' strap the emulator accordingly,2050 'relaxed' straps the emulator for both legacy-NaN and2051 2008-NaN, whereas 'strict' enables legacy-NaN only on2052 legacy processors and both NaN encodings on MIPS32 or2053 MIPS64 CPUs.2054 2055 The setting for ABS.fmt/NEG.fmt instruction execution2056 mode generally follows that for the NaN encoding,2057 except where unsupported by hardware.2058 2059 ignore_loglevel [KNL,EARLY]2060 Ignore loglevel setting - this will print /all/2061 kernel messages to the console. Useful for debugging.2062 We also add it as printk module parameter, so users2063 could change it dynamically, usually by2064 /sys/module/printk/parameters/ignore_loglevel.2065 2066 ignore_rlimit_data2067 Ignore RLIMIT_DATA setting for data mappings,2068 print warning at first misuse. Can be changed via2069 /sys/module/kernel/parameters/ignore_rlimit_data.2070 2071 ihash_entries= [KNL]2072 Set number of hash buckets for inode cache.2073 2074 ima_appraise= [IMA] appraise integrity measurements2075 Format: { "off" | "enforce" | "fix" | "log" }2076 default: "enforce"2077 2078 ima_appraise_tcb [IMA] Deprecated. Use ima_policy= instead.2079 The builtin appraise policy appraises all files2080 owned by uid=0.2081 2082 ima_canonical_fmt [IMA]2083 Use the canonical format for the binary runtime2084 measurements, instead of host native format.2085 2086 ima_hash= [IMA]2087 Format: { md5 | sha1 | rmd160 | sha256 | sha3842088 | sha512 | ... }2089 default: "sha1"2090 2091 The list of supported hash algorithms is defined2092 in crypto/hash_info.h.2093 2094 ima_policy= [IMA]2095 The builtin policies to load during IMA setup.2096 Format: "tcb | appraise_tcb | secure_boot |2097 fail_securely | critical_data"2098 2099 The "tcb" policy measures all programs exec'd, files2100 mmap'd for exec, and all files opened with the read2101 mode bit set by either the effective uid (euid=0) or2102 uid=0.2103 2104 The "appraise_tcb" policy appraises the integrity of2105 all files owned by root.2106 2107 The "secure_boot" policy appraises the integrity2108 of files (eg. kexec kernel image, kernel modules,2109 firmware, policy, etc) based on file signatures.2110 2111 The "fail_securely" policy forces file signature2112 verification failure also on privileged mounted2113 filesystems with the SB_I_UNVERIFIABLE_SIGNATURE2114 flag.2115 2116 The "critical_data" policy measures kernel integrity2117 critical data.2118 2119 ima_tcb [IMA] Deprecated. Use ima_policy= instead.2120 Load a policy which meets the needs of the Trusted2121 Computing Base. This means IMA will measure all2122 programs exec'd, files mmap'd for exec, and all files2123 opened for read by uid=0.2124 2125 ima_template= [IMA]2126 Select one of defined IMA measurements template formats.2127 Formats: { "ima" | "ima-ng" | "ima-ngv2" | "ima-sig" |2128 "ima-sigv2" }2129 Default: "ima-ng"2130 2131 ima_template_fmt=2132 [IMA] Define a custom template format.2133 Format: { "field1|...|fieldN" }2134 2135 ima.ahash_minsize= [IMA] Minimum file size for asynchronous hash usage2136 Format: <min_file_size>2137 Set the minimal file size for using asynchronous hash.2138 If left unspecified, ahash usage is disabled.2139 2140 ahash performance varies for different data sizes on2141 different crypto accelerators. This option can be used2142 to achieve the best performance for a particular HW.2143 2144 ima.ahash_bufsize= [IMA] Asynchronous hash buffer size2145 Format: <bufsize>2146 Set hashing buffer size. Default: 4k.2147 2148 ahash performance varies for different chunk sizes on2149 different crypto accelerators. This option can be used2150 to achieve best performance for particular HW.2151 2152 init= [KNL]2153 Format: <full_path>2154 Run specified binary instead of /sbin/init as init2155 process.2156 2157 initcall_debug [KNL] Trace initcalls as they are executed. Useful2158 for working out where the kernel is dying during2159 startup.2160 2161 initcall_blacklist= [KNL] Do not execute a comma-separated list of2162 initcall functions. Useful for debugging built-in2163 modules and initcalls.2164 2165 initramfs_async= [KNL]2166 Format: <bool>2167 Default: 12168 This parameter controls whether the initramfs2169 image is unpacked asynchronously, concurrently2170 with devices being probed and2171 initialized. This should normally just work,2172 but as a debugging aid, one can get the2173 historical behaviour of the initramfs2174 unpacking being completed before device_ and2175 late_ initcalls.2176 2177 initrd= [BOOT,EARLY] Specify the location of the initial ramdisk2178 2179 initrdmem= [KNL,EARLY] Specify a physical address and size from which to2180 load the initrd. If an initrd is compiled in or2181 specified in the bootparams, it takes priority over this2182 setting.2183 Format: ss[KMG],nn[KMG]2184 Default is 0, 02185 2186 init_on_alloc= [MM,EARLY] Fill newly allocated pages and heap objects with2187 zeroes.2188 Format: 0 | 12189 Default set by CONFIG_INIT_ON_ALLOC_DEFAULT_ON.2190 2191 init_on_free= [MM,EARLY] Fill freed pages and heap objects with zeroes.2192 Format: 0 | 12193 Default set by CONFIG_INIT_ON_FREE_DEFAULT_ON.2194 2195 init_pkru= [X86] Specify the default memory protection keys rights2196 register contents for all processes. 0x55555554 by2197 default (disallow access to all but pkey 0). Can2198 override in debugfs after boot.2199 2200 inport.irq= [HW] Inport (ATI XL and Microsoft) busmouse driver2201 Format: <irq>2202 2203 int_pln_enable [X86] Enable power limit notification interrupt2204 2205 integrity_audit=[IMA]2206 Format: { "0" | "1" }2207 0 -- basic integrity auditing messages. (Default)2208 1 -- additional integrity auditing messages.2209 2210 intel_iommu= [DMAR] Intel IOMMU driver (DMAR) option2211 on2212 Enable intel iommu driver.2213 off2214 Disable intel iommu driver.2215 igfx_off [Default Off]2216 By default, gfx is mapped as normal device. If a gfx2217 device has a dedicated DMAR unit, the DMAR unit is2218 bypassed by not enabling DMAR with this option. In2219 this case, gfx device will use physical address for2220 DMA.2221 strict [Default Off]2222 Deprecated, equivalent to iommu.strict=1.2223 sp_off [Default Off]2224 By default, super page will be supported if Intel IOMMU2225 has the capability. With this option, super page will2226 not be supported.2227 sm_on2228 Enable the Intel IOMMU scalable mode if the hardware2229 advertises that it has support for the scalable mode2230 translation.2231 sm_off2232 Disallow use of the Intel IOMMU scalable mode.2233 tboot_noforce [Default Off]2234 Do not force the Intel IOMMU enabled under tboot.2235 By default, tboot will force Intel IOMMU on, which2236 could harm performance of some high-throughput2237 devices like 40GBit network cards, even if identity2238 mapping is enabled.2239 Note that using this option lowers the security2240 provided by tboot because it makes the system2241 vulnerable to DMA attacks.2242 2243 intel_idle.max_cstate= [KNL,HW,ACPI,X86]2244 0 disables intel_idle and fall back on acpi_idle.2245 1 to 9 specify maximum depth of C-state.2246 2247 intel_pstate= [X86,EARLY]2248 disable2249 Do not enable intel_pstate as the default2250 scaling driver for the supported processors2251 active2252 Use intel_pstate driver to bypass the scaling2253 governors layer of cpufreq and provides it own2254 algorithms for p-state selection. There are two2255 P-state selection algorithms provided by2256 intel_pstate in the active mode: powersave and2257 performance. The way they both operate depends2258 on whether or not the hardware managed P-states2259 (HWP) feature has been enabled in the processor2260 and possibly on the processor model.2261 passive2262 Use intel_pstate as a scaling driver, but configure it2263 to work with generic cpufreq governors (instead of2264 enabling its internal governor). This mode cannot be2265 used along with the hardware-managed P-states (HWP)2266 feature.2267 force2268 Enable intel_pstate on systems that prohibit it by default2269 in favor of acpi-cpufreq. Forcing the intel_pstate driver2270 instead of acpi-cpufreq may disable platform features, such2271 as thermal controls and power capping, that rely on ACPI2272 P-States information being indicated to OSPM and therefore2273 should be used with caution. This option does not work with2274 processors that aren't supported by the intel_pstate driver2275 or on platforms that use pcc-cpufreq instead of acpi-cpufreq.2276 no_hwp2277 Do not enable hardware P state control (HWP)2278 if available.2279 hwp_only2280 Only load intel_pstate on systems which support2281 hardware P state control (HWP) if available.2282 support_acpi_ppc2283 Enforce ACPI _PPC performance limits. If the Fixed ACPI2284 Description Table, specifies preferred power management2285 profile as "Enterprise Server" or "Performance Server",2286 then this feature is turned on by default.2287 per_cpu_perf_limits2288 Allow per-logical-CPU P-State performance control limits using2289 cpufreq sysfs interface2290 2291 intremap= [X86-64,Intel-IOMMU,EARLY]2292 on enable Interrupt Remapping (default)2293 off disable Interrupt Remapping2294 nosid disable Source ID checking2295 no_x2apic_optout2296 BIOS x2APIC opt-out request will be ignored2297 nopost disable Interrupt Posting2298 posted_msi2299 enable MSIs delivered as posted interrupts2300 2301 iomem= Disable strict checking of access to MMIO memory2302 strict regions from userspace.2303 relaxed2304 2305 iommu= [X86,EARLY]2306 off2307 force2308 noforce2309 biomerge2310 panic2311 nopanic2312 merge2313 nomerge2314 soft2315 pt [X86]2316 nopt [X86]2317 nobypass [PPC/POWERNV]2318 Disable IOMMU bypass, using IOMMU for PCI devices.2319 2320 iommu.forcedac= [ARM64,X86,EARLY] Control IOVA allocation for PCI devices.2321 Format: { "0" | "1" }2322 0 - Try to allocate a 32-bit DMA address first, before2323 falling back to the full range if needed.2324 1 - Allocate directly from the full usable range,2325 forcing Dual Address Cycle for PCI cards supporting2326 greater than 32-bit addressing.2327 2328 iommu.strict= [ARM64,X86,S390,EARLY] Configure TLB invalidation behaviour2329 Format: { "0" | "1" }2330 0 - Lazy mode.2331 Request that DMA unmap operations use deferred2332 invalidation of hardware TLBs, for increased2333 throughput at the cost of reduced device isolation.2334 Will fall back to strict mode if not supported by2335 the relevant IOMMU driver.2336 1 - Strict mode.2337 DMA unmap operations invalidate IOMMU hardware TLBs2338 synchronously.2339 unset - Use value of CONFIG_IOMMU_DEFAULT_DMA_{LAZY,STRICT}.2340 Note: on x86, strict mode specified via one of the2341 legacy driver-specific options takes precedence.2342 2343 iommu.passthrough=2344 [ARM64,X86,EARLY] Configure DMA to bypass the IOMMU by default.2345 Format: { "0" | "1" }2346 0 - Use IOMMU translation for DMA.2347 1 - Bypass the IOMMU for DMA.2348 unset - Use value of CONFIG_IOMMU_DEFAULT_PASSTHROUGH.2349 2350 io7= [HW] IO7 for Marvel-based Alpha systems2351 See comment before marvel_specify_io7 in2352 arch/alpha/kernel/core_marvel.c.2353 2354 io_delay= [X86,EARLY] I/O delay method2355 0x802356 Standard port 0x80 based delay2357 0xed2358 Alternate port 0xed based delay (needed on some systems)2359 udelay2360 Simple two microseconds delay2361 none2362 No delay2363 2364 ip= [IP_PNP]2365 See Documentation/admin-guide/nfs/nfsroot.rst.2366 2367 ipcmni_extend [KNL,EARLY] Extend the maximum number of unique System V2368 IPC identifiers from 32,768 to 16,777,216.2369 2370 ipe.enforce= [IPE]2371 Format: <bool>2372 Determine whether IPE starts in permissive (0) or2373 enforce (1) mode. The default is enforce.2374 2375 ipe.success_audit=2376 [IPE]2377 Format: <bool>2378 Start IPE with success auditing enabled, emitting2379 an audit event when a binary is allowed. The default2380 is 0.2381 2382 irqaffinity= [SMP] Set the default irq affinity mask2383 The argument is a cpu list, as described above.2384 2385 irqchip.gicv2_force_probe=2386 [ARM,ARM64,EARLY]2387 Format: <bool>2388 Force the kernel to look for the second 4kB page2389 of a GICv2 controller even if the memory range2390 exposed by the device tree is too small.2391 2392 irqchip.gicv3_nolpi=2393 [ARM,ARM64,EARLY]2394 Force the kernel to ignore the availability of2395 LPIs (and by consequence ITSs). Intended for system2396 that use the kernel as a bootloader, and thus want2397 to let secondary kernels in charge of setting up2398 LPIs.2399 2400 irqchip.gicv3_pseudo_nmi= [ARM64,EARLY]2401 Enables support for pseudo-NMIs in the kernel. This2402 requires the kernel to be built with2403 CONFIG_ARM64_PSEUDO_NMI.2404 2405 irqfixup [HW]2406 When an interrupt is not handled search all handlers2407 for it. Intended to get systems with badly broken2408 firmware running.2409 2410 irqpoll [HW]2411 When an interrupt is not handled search all handlers2412 for it. Also check all handlers each timer2413 interrupt. Intended to get systems with badly broken2414 firmware running.2415 2416 isapnp= [ISAPNP]2417 Format: <RDP>,<reset>,<pci_scan>,<verbosity>2418 2419 isolcpus= [KNL,SMP,ISOL] Isolate a given set of CPUs from disturbance.2420 [Deprecated - use cpusets instead]2421 Format: [flag-list,]<cpu-list>2422 2423 Specify one or more CPUs to isolate from disturbances2424 specified in the flag list (default: domain):2425 2426 nohz2427 Disable the tick when a single task runs.2428 2429 A residual 1Hz tick is offloaded to workqueues, which you2430 need to affine to housekeeping through the global2431 workqueue's affinity configured via the2432 /sys/devices/virtual/workqueue/cpumask sysfs file, or2433 by using the 'domain' flag described below.2434 2435 NOTE: by default the global workqueue runs on all CPUs,2436 so to protect individual CPUs the 'cpumask' file has to2437 be configured manually after bootup.2438 2439 domain2440 Isolate from the general SMP balancing and scheduling2441 algorithms. Note that performing domain isolation this way2442 is irreversible: it's not possible to bring back a CPU to2443 the domains once isolated through isolcpus. It's strongly2444 advised to use cpusets instead to disable scheduler load2445 balancing through the "cpuset.sched_load_balance" file.2446 It offers a much more flexible interface where CPUs can2447 move in and out of an isolated set anytime.2448 2449 You can move a process onto or off an "isolated" CPU via2450 the CPU affinity syscalls or cpuset.2451 <cpu number> begins at 0 and the maximum value is2452 "number of CPUs in system - 1".2453 2454 managed_irq2455 2456 Isolate from being targeted by managed interrupts2457 which have an interrupt mask containing isolated2458 CPUs. The affinity of managed interrupts is2459 handled by the kernel and cannot be changed via2460 the /proc/irq/* interfaces.2461 2462 This isolation is best effort and only effective2463 if the automatically assigned interrupt mask of a2464 device queue contains isolated and housekeeping2465 CPUs. If housekeeping CPUs are online then such2466 interrupts are directed to the housekeeping CPU2467 so that IO submitted on the housekeeping CPU2468 cannot disturb the isolated CPU.2469 2470 If a queue's affinity mask contains only isolated2471 CPUs then this parameter has no effect on the2472 interrupt routing decision, though interrupts are2473 only delivered when tasks running on those2474 isolated CPUs submit IO. IO submitted on2475 housekeeping CPUs has no influence on those2476 queues.2477 2478 The format of <cpu-list> is described above.2479 2480 iucv= [HW,NET]2481 2482 ivrs_ioapic [HW,X86-64]2483 Provide an override to the IOAPIC-ID<->DEVICE-ID2484 mapping provided in the IVRS ACPI table.2485 By default, PCI segment is 0, and can be omitted.2486 2487 For example, to map IOAPIC-ID decimal 10 to2488 PCI segment 0x1 and PCI device 00:14.0,2489 write the parameter as:2490 ivrs_ioapic=10@0001:00:14.02491 2492 Deprecated formats:2493 * To map IOAPIC-ID decimal 10 to PCI device 00:14.02494 write the parameter as:2495 ivrs_ioapic[10]=00:14.02496 * To map IOAPIC-ID decimal 10 to PCI segment 0x1 and2497 PCI device 00:14.0 write the parameter as:2498 ivrs_ioapic[10]=0001:00:14.02499 2500 ivrs_hpet [HW,X86-64]2501 Provide an override to the HPET-ID<->DEVICE-ID2502 mapping provided in the IVRS ACPI table.2503 By default, PCI segment is 0, and can be omitted.2504 2505 For example, to map HPET-ID decimal 10 to2506 PCI segment 0x1 and PCI device 00:14.0,2507 write the parameter as:2508 ivrs_hpet=10@0001:00:14.02509 2510 Deprecated formats:2511 * To map HPET-ID decimal 0 to PCI device 00:14.02512 write the parameter as:2513 ivrs_hpet[0]=00:14.02514 * To map HPET-ID decimal 10 to PCI segment 0x1 and2515 PCI device 00:14.0 write the parameter as:2516 ivrs_ioapic[10]=0001:00:14.02517 2518 ivrs_acpihid [HW,X86-64]2519 Provide an override to the ACPI-HID:UID<->DEVICE-ID2520 mapping provided in the IVRS ACPI table.2521 By default, PCI segment is 0, and can be omitted.2522 2523 For example, to map UART-HID:UID AMD0020:0 to2524 PCI segment 0x1 and PCI device ID 00:14.5,2525 write the parameter as:2526 ivrs_acpihid=AMD0020:0@0001:00:14.52527 2528 Deprecated formats:2529 * To map UART-HID:UID AMD0020:0 to PCI segment is 0,2530 PCI device ID 00:14.5, write the parameter as:2531 ivrs_acpihid[00:14.5]=AMD0020:02532 * To map UART-HID:UID AMD0020:0 to PCI segment 0x1 and2533 PCI device ID 00:14.5, write the parameter as:2534 ivrs_acpihid[0001:00:14.5]=AMD0020:02535 2536 js= [HW,JOY] Analog joystick2537 See Documentation/input/joydev/joystick.rst.2538 2539 kasan_multi_shot2540 [KNL] Enforce KASAN (Kernel Address Sanitizer) to print2541 report on every invalid memory access. Without this2542 parameter KASAN will print report only for the first2543 invalid access.2544 2545 keep_bootcon [KNL,EARLY]2546 Do not unregister boot console at start. This is only2547 useful for debugging when something happens in the window2548 between unregistering the boot console and initializing2549 the real console.2550 2551 keepinitrd [HW,ARM] See retain_initrd.2552 2553 kernelcore= [KNL,X86,PPC,EARLY]2554 Format: nn[KMGTPE] | nn% | "mirror"2555 This parameter specifies the amount of memory usable by2556 the kernel for non-movable allocations. The requested2557 amount is spread evenly throughout all nodes in the2558 system as ZONE_NORMAL. The remaining memory is used for2559 movable memory in its own zone, ZONE_MOVABLE. In the2560 event, a node is too small to have both ZONE_NORMAL and2561 ZONE_MOVABLE, kernelcore memory will take priority and2562 other nodes will have a larger ZONE_MOVABLE.2563 2564 ZONE_MOVABLE is used for the allocation of pages that2565 may be reclaimed or moved by the page migration2566 subsystem. Note that allocations like PTEs-from-HighMem2567 still use the HighMem zone if it exists, and the Normal2568 zone if it does not.2569 2570 It is possible to specify the exact amount of memory in2571 the form of "nn[KMGTPE]", a percentage of total system2572 memory in the form of "nn%", or "mirror". If "mirror"2573 option is specified, mirrored (reliable) memory is used2574 for non-movable allocations and remaining memory is used2575 for Movable pages. "nn[KMGTPE]", "nn%", and "mirror"2576 are exclusive, so you cannot specify multiple forms.2577 2578 kgdbdbgp= [KGDB,HW,EARLY] kgdb over EHCI usb debug port.2579 Format: <Controller#>[,poll interval]2580 The controller # is the number of the ehci usb debug2581 port as it is probed via PCI. The poll interval is2582 optional and is the number seconds in between2583 each poll cycle to the debug port in case you need2584 the functionality for interrupting the kernel with2585 gdb or control-c on the dbgp connection. When2586 not using this parameter you use sysrq-g to break into2587 the kernel debugger.2588 2589 kgdboc= [KGDB,HW] kgdb over consoles.2590 Requires a tty driver that supports console polling,2591 or a supported polling keyboard driver (non-usb).2592 Serial only format: <serial_device>[,baud]2593 keyboard only format: kbd2594 keyboard and serial format: kbd,<serial_device>[,baud]2595 Optional Kernel mode setting:2596 kms, kbd format: kms,kbd2597 kms, kbd and serial format: kms,kbd,<ser_dev>[,baud]2598 2599 kgdboc_earlycon= [KGDB,HW,EARLY]2600 If the boot console provides the ability to read2601 characters and can work in polling mode, you can use2602 this parameter to tell kgdb to use it as a backend2603 until the normal console is registered. Intended to2604 be used together with the kgdboc parameter which2605 specifies the normal console to transition to.2606 2607 The name of the early console should be specified2608 as the value of this parameter. Note that the name of2609 the early console might be different than the tty2610 name passed to kgdboc. It's OK to leave the value2611 blank and the first boot console that implements2612 read() will be picked.2613 2614 kgdbwait [KGDB,EARLY] Stop kernel execution and enter the2615 kernel debugger at the earliest opportunity.2616 2617 kmac= [MIPS] Korina ethernet MAC address.2618 Configure the RouterBoard 532 series on-chip2619 Ethernet adapter MAC address.2620 2621 kmemleak= [KNL,EARLY] Boot-time kmemleak enable/disable2622 Valid arguments: on, off2623 Default: on2624 Built with CONFIG_DEBUG_KMEMLEAK_DEFAULT_OFF=y,2625 the default is off.2626 2627 kprobe_event=[probe-list]2628 [FTRACE] Add kprobe events and enable at boot time.2629 The probe-list is a semicolon delimited list of probe2630 definitions. Each definition is same as kprobe_events2631 interface, but the parameters are comma delimited.2632 For example, to add a kprobe event on vfs_read with2633 arg1 and arg2, add to the command line;2634 2635 kprobe_event=p,vfs_read,$arg1,$arg22636 2637 See also Documentation/trace/kprobetrace.rst "Kernel2638 Boot Parameter" section.2639 2640 kpti= [ARM64,EARLY] Control page table isolation of2641 user and kernel address spaces.2642 Default: enabled on cores which need mitigation.2643 0: force disabled2644 1: force enabled2645 2646 kunit.enable= [KUNIT] Enable executing KUnit tests. Requires2647 CONFIG_KUNIT to be set to be fully enabled. The2648 default value can be overridden via2649 KUNIT_DEFAULT_ENABLED.2650 Default is 1 (enabled)2651 2652 kvm.ignore_msrs=[KVM] Ignore guest accesses to unhandled MSRs.2653 Default is 0 (don't ignore, but inject #GP)2654 2655 kvm.eager_page_split=2656 [KVM,X86] Controls whether or not KVM will try to2657 proactively split all huge pages during dirty logging.2658 Eager page splitting reduces interruptions to vCPU2659 execution by eliminating the write-protection faults2660 and MMU lock contention that would otherwise be2661 required to split huge pages lazily.2662 2663 VM workloads that rarely perform writes or that write2664 only to a small region of VM memory may benefit from2665 disabling eager page splitting to allow huge pages to2666 still be used for reads.2667 2668 The behavior of eager page splitting depends on whether2669 KVM_DIRTY_LOG_INITIALLY_SET is enabled or disabled. If2670 disabled, all huge pages in a memslot will be eagerly2671 split when dirty logging is enabled on that memslot. If2672 enabled, eager page splitting will be performed during2673 the KVM_CLEAR_DIRTY ioctl, and only for the pages being2674 cleared.2675 2676 Eager page splitting is only supported when kvm.tdp_mmu=Y.2677 2678 Default is Y (on).2679 2680 kvm.enable_virt_at_load=[KVM,ARM64,LOONGARCH,MIPS,RISCV,X86]2681 If enabled, KVM will enable virtualization in hardware2682 when KVM is loaded, and disable virtualization when KVM2683 is unloaded (if KVM is built as a module).2684 2685 If disabled, KVM will dynamically enable and disable2686 virtualization on-demand when creating and destroying2687 VMs, i.e. on the 0=>1 and 1=>0 transitions of the2688 number of VMs.2689 2690 Enabling virtualization at module lode avoids potential2691 latency for creation of the 0=>1 VM, as KVM serializes2692 virtualization enabling across all online CPUs. The2693 "cost" of enabling virtualization when KVM is loaded,2694 is that doing so may interfere with using out-of-tree2695 hypervisors that want to "own" virtualization hardware.2696 2697 kvm.enable_vmware_backdoor=[KVM] Support VMware backdoor PV interface.2698 Default is false (don't support).2699 2700 kvm.nx_huge_pages=2701 [KVM] Controls the software workaround for the2702 X86_BUG_ITLB_MULTIHIT bug.2703 force : Always deploy workaround.2704 off : Never deploy workaround.2705 auto : Deploy workaround based on the presence of2706 X86_BUG_ITLB_MULTIHIT.2707 2708 Default is 'auto'.2709 2710 If the software workaround is enabled for the host,2711 guests do need not to enable it for nested guests.2712 2713 kvm.nx_huge_pages_recovery_ratio=2714 [KVM] Controls how many 4KiB pages are periodically zapped2715 back to huge pages. 0 disables the recovery, otherwise if2716 the value is N KVM will zap 1/Nth of the 4KiB pages every2717 period (see below). The default is 60.2718 2719 kvm.nx_huge_pages_recovery_period_ms=2720 [KVM] Controls the time period at which KVM zaps 4KiB pages2721 back to huge pages. If the value is a non-zero N, KVM will2722 zap a portion (see ratio above) of the pages every N msecs.2723 If the value is 0 (the default), KVM will pick a period based2724 on the ratio, such that a page is zapped after 1 hour on average.2725 2726 kvm-amd.nested= [KVM,AMD] Control nested virtualization feature in2727 KVM/SVM. Default is 1 (enabled).2728 2729 kvm-amd.npt= [KVM,AMD] Control KVM's use of Nested Page Tables,2730 a.k.a. Two-Dimensional Page Tables. Default is 12731 (enabled). Disable by KVM if hardware lacks support2732 for NPT.2733 2734 kvm-arm.mode=2735 [KVM,ARM,EARLY] Select one of KVM/arm64's modes of2736 operation.2737 2738 none: Forcefully disable KVM.2739 2740 nvhe: Standard nVHE-based mode, without support for2741 protected guests.2742 2743 protected: nVHE-based mode with support for guests whose2744 state is kept private from the host.2745 2746 nested: VHE-based mode with support for nested2747 virtualization. Requires at least ARMv8.32748 hardware.2749 2750 Defaults to VHE/nVHE based on hardware support. Setting2751 mode to "protected" will disable kexec and hibernation2752 for the host. "nested" is experimental and should be2753 used with extreme caution.2754 2755 kvm-arm.vgic_v3_group0_trap=2756 [KVM,ARM,EARLY] Trap guest accesses to GICv3 group-02757 system registers2758 2759 kvm-arm.vgic_v3_group1_trap=2760 [KVM,ARM,EARLY] Trap guest accesses to GICv3 group-12761 system registers2762 2763 kvm-arm.vgic_v3_common_trap=2764 [KVM,ARM,EARLY] Trap guest accesses to GICv3 common2765 system registers2766 2767 kvm-arm.vgic_v4_enable=2768 [KVM,ARM,EARLY] Allow use of GICv4 for direct2769 injection of LPIs.2770 2771 kvm-arm.wfe_trap_policy=2772 [KVM,ARM] Control when to set WFE instruction trap for2773 KVM VMs. Traps are allowed but not guaranteed by the2774 CPU architecture.2775 2776 trap: set WFE instruction trap2777 2778 notrap: clear WFE instruction trap2779 2780 kvm-arm.wfi_trap_policy=2781 [KVM,ARM] Control when to set WFI instruction trap for2782 KVM VMs. Traps are allowed but not guaranteed by the2783 CPU architecture.2784 2785 trap: set WFI instruction trap2786 2787 notrap: clear WFI instruction trap2788 2789 kvm_cma_resv_ratio=n [PPC,EARLY]2790 Reserves given percentage from system memory area for2791 contiguous memory allocation for KVM hash pagetable2792 allocation.2793 By default it reserves 5% of total system memory.2794 Format: <integer>2795 Default: 52796 2797 kvm-intel.ept= [KVM,Intel] Control KVM's use of Extended Page Tables,2798 a.k.a. Two-Dimensional Page Tables. Default is 12799 (enabled). Disable by KVM if hardware lacks support2800 for EPT.2801 2802 kvm-intel.emulate_invalid_guest_state=2803 [KVM,Intel] Control whether to emulate invalid guest2804 state. Ignored if kvm-intel.enable_unrestricted_guest=1,2805 as guest state is never invalid for unrestricted2806 guests. This param doesn't apply to nested guests (L2),2807 as KVM never emulates invalid L2 guest state.2808 Default is 1 (enabled).2809 2810 kvm-intel.flexpriority=2811 [KVM,Intel] Control KVM's use of FlexPriority feature2812 (TPR shadow). Default is 1 (enabled). Disable by KVM if2813 hardware lacks support for it.2814 2815 kvm-intel.nested=2816 [KVM,Intel] Control nested virtualization feature in2817 KVM/VMX. Default is 1 (enabled).2818 2819 kvm-intel.unrestricted_guest=2820 [KVM,Intel] Control KVM's use of unrestricted guest2821 feature (virtualized real and unpaged mode). Default2822 is 1 (enabled). Disable by KVM if EPT is disabled or2823 hardware lacks support for it.2824 2825 kvm-intel.vmentry_l1d_flush=[KVM,Intel] Mitigation for L1 Terminal Fault2826 CVE-2018-3620.2827 2828 Valid arguments: never, cond, always2829 2830 always: L1D cache flush on every VMENTER.2831 cond: Flush L1D on VMENTER only when the code between2832 VMEXIT and VMENTER can leak host memory.2833 never: Disables the mitigation2834 2835 Default is cond (do L1 cache flush in specific instances)2836 2837 kvm-intel.vpid= [KVM,Intel] Control KVM's use of Virtual Processor2838 Identification feature (tagged TLBs). Default is 12839 (enabled). Disable by KVM if hardware lacks support2840 for it.2841 2842 l1d_flush= [X86,INTEL,EARLY]2843 Control mitigation for L1D based snooping vulnerability.2844 2845 Certain CPUs are vulnerable to an exploit against CPU2846 internal buffers which can forward information to a2847 disclosure gadget under certain conditions.2848 2849 In vulnerable processors, the speculatively2850 forwarded data can be used in a cache side channel2851 attack, to access data to which the attacker does2852 not have direct access.2853 2854 This parameter controls the mitigation. The2855 options are:2856 2857 on - enable the interface for the mitigation2858 2859 l1tf= [X86,EARLY] Control mitigation of the L1TF vulnerability on2860 affected CPUs2861 2862 The kernel PTE inversion protection is unconditionally2863 enabled and cannot be disabled.2864 2865 full2866 Provides all available mitigations for the2867 L1TF vulnerability. Disables SMT and2868 enables all mitigations in the2869 hypervisors, i.e. unconditional L1D flush.2870 2871 SMT control and L1D flush control via the2872 sysfs interface is still possible after2873 boot. Hypervisors will issue a warning2874 when the first VM is started in a2875 potentially insecure configuration,2876 i.e. SMT enabled or L1D flush disabled.2877 2878 full,force2879 Same as 'full', but disables SMT and L1D2880 flush runtime control. Implies the2881 'nosmt=force' command line option.2882 (i.e. sysfs control of SMT is disabled.)2883 2884 flush2885 Leaves SMT enabled and enables the default2886 hypervisor mitigation, i.e. conditional2887 L1D flush.2888 2889 SMT control and L1D flush control via the2890 sysfs interface is still possible after2891 boot. Hypervisors will issue a warning2892 when the first VM is started in a2893 potentially insecure configuration,2894 i.e. SMT enabled or L1D flush disabled.2895 2896 flush,nosmt2897 2898 Disables SMT and enables the default2899 hypervisor mitigation.2900 2901 SMT control and L1D flush control via the2902 sysfs interface is still possible after2903 boot. Hypervisors will issue a warning2904 when the first VM is started in a2905 potentially insecure configuration,2906 i.e. SMT enabled or L1D flush disabled.2907 2908 flush,nowarn2909 Same as 'flush', but hypervisors will not2910 warn when a VM is started in a potentially2911 insecure configuration.2912 2913 off2914 Disables hypervisor mitigations and doesn't2915 emit any warnings.2916 It also drops the swap size and available2917 RAM limit restriction on both hypervisor and2918 bare metal.2919 2920 Default is 'flush'.2921 2922 For details see: Documentation/admin-guide/hw-vuln/l1tf.rst2923 2924 l2cr= [PPC]2925 2926 l3cr= [PPC]2927 2928 lapic [X86-32,APIC,EARLY] Enable the local APIC even if BIOS2929 disabled it.2930 2931 lapic= [X86,APIC] Do not use TSC deadline2932 value for LAPIC timer one-shot implementation. Default2933 back to the programmable timer unit in the LAPIC.2934 Format: notscdeadline2935 2936 lapic_timer_c2_ok [X86,APIC,EARLY] trust the local apic timer2937 in C2 power state.2938 2939 libata.dma= [LIBATA] DMA control2940 libata.dma=0 Disable all PATA and SATA DMA2941 libata.dma=1 PATA and SATA Disk DMA only2942 libata.dma=2 ATAPI (CDROM) DMA only2943 libata.dma=4 Compact Flash DMA only2944 Combinations also work, so libata.dma=3 enables DMA2945 for disks and CDROMs, but not CFs.2946 2947 libata.ignore_hpa= [LIBATA] Ignore HPA limit2948 libata.ignore_hpa=0 keep BIOS limits (default)2949 libata.ignore_hpa=1 ignore limits, using full disk2950 2951 libata.noacpi [LIBATA] Disables use of ACPI in libata suspend/resume2952 when set.2953 Format: <int>2954 2955 libata.force= [LIBATA] Force configurations. The format is a comma-2956 separated list of "[ID:]VAL" where ID is PORT[.DEVICE].2957 PORT and DEVICE are decimal numbers matching port, link2958 or device. Basically, it matches the ATA ID string2959 printed on console by libata. If the whole ID part is2960 omitted, the last PORT and DEVICE values are used. If2961 ID hasn't been specified yet, the configuration applies2962 to all ports, links and devices.2963 2964 If only DEVICE is omitted, the parameter applies to2965 the port and all links and devices behind it. DEVICE2966 number of 0 either selects the first device or the2967 first fan-out link behind PMP device. It does not2968 select the host link. DEVICE number of 15 selects the2969 host link and device attached to it.2970 2971 The VAL specifies the configuration to force. As long2972 as there is no ambiguity, shortcut notation is allowed.2973 For example, both 1.5 and 1.5G would work for 1.5Gbps.2974 The following configurations can be forced.2975 2976 * Cable type: 40c, 80c, short40c, unk, ign or sata.2977 Any ID with matching PORT is used.2978 2979 * SATA link speed limit: 1.5Gbps or 3.0Gbps.2980 2981 * Transfer mode: pio[0-7], mwdma[0-4] and udma[0-7].2982 udma[/][16,25,33,44,66,100,133] notation is also2983 allowed.2984 2985 * nohrst, nosrst, norst: suppress hard, soft and both2986 resets.2987 2988 * rstonce: only attempt one reset during hot-unplug2989 link recovery.2990 2991 * [no]dbdelay: Enable or disable the extra 200ms delay2992 before debouncing a link PHY and device presence2993 detection.2994 2995 * [no]ncq: Turn on or off NCQ.2996 2997 * [no]ncqtrim: Enable or disable queued DSM TRIM.2998 2999 * [no]ncqati: Enable or disable NCQ trim on ATI chipset.3000 3001 * [no]trim: Enable or disable (unqueued) TRIM.3002 3003 * trim_zero: Indicate that TRIM command zeroes data.3004 3005 * max_trim_128m: Set 128M maximum trim size limit.3006 3007 * [no]dma: Turn on or off DMA transfers.3008 3009 * atapi_dmadir: Enable ATAPI DMADIR bridge support.3010 3011 * atapi_mod16_dma: Enable the use of ATAPI DMA for3012 commands that are not a multiple of 16 bytes.3013 3014 * [no]dmalog: Enable or disable the use of the3015 READ LOG DMA EXT command to access logs.3016 3017 * [no]iddevlog: Enable or disable access to the3018 identify device data log.3019 3020 * [no]logdir: Enable or disable access to the general3021 purpose log directory.3022 3023 * max_sec_128: Set transfer size limit to 128 sectors.3024 3025 * max_sec_1024: Set or clear transfer size limit to3026 1024 sectors.3027 3028 * max_sec_lba48: Set or clear transfer size limit to3029 65535 sectors.3030 3031 * [no]lpm: Enable or disable link power management.3032 3033 * [no]setxfer: Indicate if transfer speed mode setting3034 should be skipped.3035 3036 * [no]fua: Disable or enable FUA (Force Unit Access)3037 support for devices supporting this feature.3038 3039 * dump_id: Dump IDENTIFY data.3040 3041 * disable: Disable this device.3042 3043 If there are multiple matching configurations changing3044 the same attribute, the last one is used.3045 3046 load_ramdisk= [RAM] [Deprecated]3047 3048 lockd.nlm_grace_period=P [NFS] Assign grace period.3049 Format: <integer>3050 3051 lockd.nlm_tcpport=N [NFS] Assign TCP port.3052 Format: <integer>3053 3054 lockd.nlm_timeout=T [NFS] Assign timeout value.3055 Format: <integer>3056 3057 lockd.nlm_udpport=M [NFS] Assign UDP port.3058 Format: <integer>3059 3060 lockdown= [SECURITY,EARLY]3061 { integrity | confidentiality }3062 Enable the kernel lockdown feature. If set to3063 integrity, kernel features that allow userland to3064 modify the running kernel are disabled. If set to3065 confidentiality, kernel features that allow userland3066 to extract confidential information from the kernel3067 are also disabled.3068 3069 locktorture.acq_writer_lim= [KNL]3070 Set the time limit in jiffies for a lock3071 acquisition. Acquisitions exceeding this limit3072 will result in a splat once they do complete.3073 3074 locktorture.bind_readers= [KNL]3075 Specify the list of CPUs to which the readers are3076 to be bound.3077 3078 locktorture.bind_writers= [KNL]3079 Specify the list of CPUs to which the writers are3080 to be bound.3081 3082 locktorture.call_rcu_chains= [KNL]3083 Specify the number of self-propagating call_rcu()3084 chains to set up. These are used to ensure that3085 there is a high probability of an RCU grace period3086 in progress at any given time. Defaults to 0,3087 which disables these call_rcu() chains.3088 3089 locktorture.long_hold= [KNL]3090 Specify the duration in milliseconds for the3091 occasional long-duration lock hold time. Defaults3092 to 100 milliseconds. Select 0 to disable.3093 3094 locktorture.nested_locks= [KNL]3095 Specify the maximum lock nesting depth that3096 locktorture is to exercise, up to a limit of 83097 (MAX_NESTED_LOCKS). Specify zero to disable.3098 Note that this parameter is ineffective on types3099 of locks that do not support nested acquisition.3100 3101 locktorture.nreaders_stress= [KNL]3102 Set the number of locking read-acquisition kthreads.3103 Defaults to being automatically set based on the3104 number of online CPUs.3105 3106 locktorture.nwriters_stress= [KNL]3107 Set the number of locking write-acquisition kthreads.3108 3109 locktorture.onoff_holdoff= [KNL]3110 Set time (s) after boot for CPU-hotplug testing.3111 3112 locktorture.onoff_interval= [KNL]3113 Set time (s) between CPU-hotplug operations, or3114 zero to disable CPU-hotplug testing.3115 3116 locktorture.rt_boost= [KNL]3117 Do periodic testing of real-time lock priority3118 boosting. Select 0 to disable, 1 to boost3119 only rt_mutex, and 2 to boost unconditionally.3120 Defaults to 2, which might seem to be an3121 odd choice, but which should be harmless for3122 non-real-time spinlocks, due to their disabling3123 of preemption. Note that non-realtime mutexes3124 disable boosting.3125 3126 locktorture.rt_boost_factor= [KNL]3127 Number that determines how often and for how3128 long priority boosting is exercised. This is3129 scaled down by the number of writers, so that the3130 number of boosts per unit time remains roughly3131 constant as the number of writers increases.3132 On the other hand, the duration of each boost3133 increases with the number of writers.3134 3135 locktorture.shuffle_interval= [KNL]3136 Set task-shuffle interval (jiffies). Shuffling3137 tasks allows some CPUs to go into dyntick-idle3138 mode during the locktorture test.3139 3140 locktorture.shutdown_secs= [KNL]3141 Set time (s) after boot system shutdown. This3142 is useful for hands-off automated testing.3143 3144 locktorture.stat_interval= [KNL]3145 Time (s) between statistics printk()s.3146 3147 locktorture.stutter= [KNL]3148 Time (s) to stutter testing, for example,3149 specifying five seconds causes the test to run for3150 five seconds, wait for five seconds, and so on.3151 This tests the locking primitive's ability to3152 transition abruptly to and from idle.3153 3154 locktorture.torture_type= [KNL]3155 Specify the locking implementation to test.3156 3157 locktorture.verbose= [KNL]3158 Enable additional printk() statements.3159 3160 locktorture.writer_fifo= [KNL]3161 Run the write-side locktorture kthreads at3162 sched_set_fifo() real-time priority.3163 3164 logibm.irq= [HW,MOUSE] Logitech Bus Mouse Driver3165 Format: <irq>3166 3167 loglevel= [KNL,EARLY]3168 All Kernel Messages with a loglevel smaller than the3169 console loglevel will be printed to the console. It can3170 also be changed with klogd or other programs. The3171 loglevels are defined as follows:3172 3173 0 (KERN_EMERG) system is unusable3174 1 (KERN_ALERT) action must be taken immediately3175 2 (KERN_CRIT) critical conditions3176 3 (KERN_ERR) error conditions3177 4 (KERN_WARNING) warning conditions3178 5 (KERN_NOTICE) normal but significant condition3179 6 (KERN_INFO) informational3180 7 (KERN_DEBUG) debug-level messages3181 3182 log_buf_len=n[KMG] [KNL,EARLY]3183 Sets the size of the printk ring buffer, in bytes.3184 n must be a power of two and greater than the3185 minimal size. The minimal size is defined by3186 LOG_BUF_SHIFT kernel config parameter. There3187 is also CONFIG_LOG_CPU_MAX_BUF_SHIFT config3188 parameter that allows to increase the default size3189 depending on the number of CPUs. See init/Kconfig3190 for more details.3191 3192 logo.nologo [FB] Disables display of the built-in Linux logo.3193 This may be used to provide more screen space for3194 kernel log messages and is useful when debugging3195 kernel boot problems.3196 3197 lp=0 [LP] Specify parallel ports to use, e.g,3198 lp=port[,port...] lp=none,parport0 (lp0 not configured, lp1 uses3199 lp=reset first parallel port). 'lp=0' disables the3200 lp=auto printer driver. 'lp=reset' (which can be3201 specified in addition to the ports) causes3202 attached printers to be reset. Using3203 lp=port1,port2,... specifies the parallel ports3204 to associate lp devices with, starting with3205 lp0. A port specification may be 'none' to skip3206 that lp device, or a parport name such as3207 'parport0'. Specifying 'lp=auto' instead of a3208 port specification list means that device IDs3209 from each port should be examined, to see if3210 an IEEE 1284-compliant printer is attached; if3211 so, the driver will manage that printer.3212 See also header of drivers/char/lp.c.3213 3214 lpj=n [KNL]3215 Sets loops_per_jiffy to given constant, thus avoiding3216 time-consuming boot-time autodetection (up to 250 ms per3217 CPU). 0 enables autodetection (default). To determine3218 the correct value for your kernel, boot with normal3219 autodetection and see what value is printed. Note that3220 on SMP systems the preset will be applied to all CPUs,3221 which is likely to cause problems if your CPUs need3222 significantly divergent settings. An incorrect value3223 will cause delays in the kernel to be wrong, leading to3224 unpredictable I/O errors and other breakage. Although3225 unlikely, in the extreme case this might damage your3226 hardware.3227 3228 lsm.debug [SECURITY] Enable LSM initialization debugging output.3229 3230 lsm=lsm1,...,lsmN3231 [SECURITY] Choose order of LSM initialization. This3232 overrides CONFIG_LSM, and the "security=" parameter.3233 3234 machtype= [Loongson] Share the same kernel image file between3235 different yeeloong laptops.3236 Example: machtype=lemote-yeeloong-2f-7inch3237 3238 maxcpus= [SMP,EARLY] Maximum number of processors that an SMP kernel3239 will bring up during bootup. maxcpus=n : n >= 0 limits3240 the kernel to bring up 'n' processors. Surely after3241 bootup you can bring up the other plugged cpu by executing3242 "echo 1 > /sys/devices/system/cpu/cpuX/online". So maxcpus3243 only takes effect during system bootup.3244 While n=0 is a special case, it is equivalent to "nosmp",3245 which also disables the IO APIC.3246 3247 max_loop= [LOOP] The number of loop block devices that get3248 (loop.max_loop) unconditionally pre-created at init time. The default3249 number is configured by BLK_DEV_LOOP_MIN_COUNT. Instead3250 of statically allocating a predefined number, loop3251 devices can be requested on-demand with the3252 /dev/loop-control interface.3253 3254 mce [X86-32] Machine Check Exception3255 3256 mce=option [X86-64] See Documentation/arch/x86/x86_64/boot-options.rst3257 3258 md= [HW] RAID subsystems devices and level3259 See Documentation/admin-guide/md.rst.3260 3261 mdacon= [MDA]3262 Format: <first>,<last>3263 Specifies range of consoles to be captured by the MDA.3264 3265 mds= [X86,INTEL,EARLY]3266 Control mitigation for the Micro-architectural Data3267 Sampling (MDS) vulnerability.3268 3269 Certain CPUs are vulnerable to an exploit against CPU3270 internal buffers which can forward information to a3271 disclosure gadget under certain conditions.3272 3273 In vulnerable processors, the speculatively3274 forwarded data can be used in a cache side channel3275 attack, to access data to which the attacker does3276 not have direct access.3277 3278 This parameter controls the MDS mitigation. The3279 options are:3280 3281 full - Enable MDS mitigation on vulnerable CPUs3282 full,nosmt - Enable MDS mitigation and disable3283 SMT on vulnerable CPUs3284 off - Unconditionally disable MDS mitigation3285 3286 On TAA-affected machines, mds=off can be prevented by3287 an active TAA mitigation as both vulnerabilities are3288 mitigated with the same mechanism so in order to disable3289 this mitigation, you need to specify tsx_async_abort=off3290 too.3291 3292 Not specifying this option is equivalent to3293 mds=full.3294 3295 For details see: Documentation/admin-guide/hw-vuln/mds.rst3296 3297 mem=nn[KMG] [HEXAGON,EARLY] Set the memory size.3298 Must be specified, otherwise memory size will be 0.3299 3300 mem=nn[KMG] [KNL,BOOT,EARLY] Force usage of a specific amount3301 of memory Amount of memory to be used in cases3302 as follows:3303 3304 1 for test;3305 2 when the kernel is not able to see the whole system memory;3306 3 memory that lies after 'mem=' boundary is excluded from3307 the hypervisor, then assigned to KVM guests.3308 4 to limit the memory available for kdump kernel.3309 3310 [ARC,MICROBLAZE] - the limit applies only to low memory,3311 high memory is not affected.3312 3313 [ARM64] - only limits memory covered by the linear3314 mapping. The NOMAP regions are not affected.3315 3316 [X86] Work as limiting max address. Use together3317 with memmap= to avoid physical address space collisions.3318 Without memmap= PCI devices could be placed at addresses3319 belonging to unused RAM.3320 3321 Note that this only takes effects during boot time since3322 in above case 3, memory may need be hot added after boot3323 if system memory of hypervisor is not sufficient.3324 3325 mem=nn[KMG]@ss[KMG]3326 [ARM,MIPS,EARLY] - override the memory layout3327 reported by firmware.3328 Define a memory region of size nn[KMG] starting at3329 ss[KMG].3330 Multiple different regions can be specified with3331 multiple mem= parameters on the command line.3332 3333 mem=nopentium [BUGS=X86-32] Disable usage of 4MB pages for kernel3334 memory.3335 3336 memblock=debug [KNL,EARLY] Enable memblock debug messages.3337 3338 memchunk=nn[KMG]3339 [KNL,SH] Allow user to override the default size for3340 per-device physically contiguous DMA buffers.3341 3342 memhp_default_state=online/offline/online_kernel/online_movable3343 [KNL] Set the initial state for the memory hotplug3344 onlining policy. If not specified, the default value is3345 set according to the3346 CONFIG_MEMORY_HOTPLUG_DEFAULT_ONLINE kernel config3347 option.3348 See Documentation/admin-guide/mm/memory-hotplug.rst.3349 3350 memmap=exactmap [KNL,X86,EARLY] Enable setting of an exact3351 E820 memory map, as specified by the user.3352 Such memmap=exactmap lines can be constructed based on3353 BIOS output or other requirements. See the memmap=nn@ss3354 option description.3355 3356 memmap=nn[KMG]@ss[KMG]3357 [KNL, X86,MIPS,XTENSA,EARLY] Force usage of a specific region of memory.3358 Region of memory to be used is from ss to ss+nn.3359 If @ss[KMG] is omitted, it is equivalent to mem=nn[KMG],3360 which limits max address to nn[KMG].3361 Multiple different regions can be specified,3362 comma delimited.3363 Example:3364 memmap=100M@2G,100M#3G,1G!1024G3365 3366 memmap=nn[KMG]#ss[KMG]3367 [KNL,ACPI,EARLY] Mark specific memory as ACPI data.3368 Region of memory to be marked is from ss to ss+nn.3369 3370 memmap=nn[KMG]$ss[KMG]3371 [KNL,ACPI,EARLY] Mark specific memory as reserved.3372 Region of memory to be reserved is from ss to ss+nn.3373 Example: Exclude memory from 0x18690000-0x1869ffff3374 memmap=64K$0x186900003375 or3376 memmap=0x10000$0x186900003377 Some bootloaders may need an escape character before '$',3378 like Grub2, otherwise '$' and the following number3379 will be eaten.3380 3381 memmap=nn[KMG]!ss[KMG,EARLY]3382 [KNL,X86] Mark specific memory as protected.3383 Region of memory to be used, from ss to ss+nn.3384 The memory region may be marked as e820 type 12 (0xc)3385 and is NVDIMM or ADR memory.3386 3387 memmap=<size>%<offset>-<oldtype>+<newtype>3388 [KNL,ACPI,EARLY] Convert memory within the specified region3389 from <oldtype> to <newtype>. If "-<oldtype>" is left3390 out, the whole region will be marked as <newtype>,3391 even if previously unavailable. If "+<newtype>" is left3392 out, matching memory will be removed. Types are3393 specified as e820 types, e.g., 1 = RAM, 2 = reserved,3394 3 = ACPI, 12 = PRAM.3395 3396 memory_corruption_check=0/1 [X86,EARLY]3397 Some BIOSes seem to corrupt the first 64k of3398 memory when doing things like suspend/resume.3399 Setting this option will scan the memory3400 looking for corruption. Enabling this will3401 both detect corruption and prevent the kernel3402 from using the memory being corrupted.3403 However, its intended as a diagnostic tool; if3404 repeatable BIOS-originated corruption always3405 affects the same memory, you can use memmap=3406 to prevent the kernel from using that memory.3407 3408 memory_corruption_check_size=size [X86,EARLY]3409 By default it checks for corruption in the low3410 64k, making this memory unavailable for normal3411 use. Use this parameter to scan for3412 corruption in more or less memory.3413 3414 memory_corruption_check_period=seconds [X86,EARLY]3415 By default it checks for corruption every 603416 seconds. Use this parameter to check at some3417 other rate. 0 disables periodic checking.3418 3419 memory_hotplug.memmap_on_memory3420 [KNL,X86,ARM] Boolean flag to enable this feature.3421 Format: {on | off (default)}3422 When enabled, runtime hotplugged memory will3423 allocate its internal metadata (struct pages,3424 those vmemmap pages cannot be optimized even3425 if hugetlb_free_vmemmap is enabled) from the3426 hotadded memory which will allow to hotadd a3427 lot of memory without requiring additional3428 memory to do so.3429 This feature is disabled by default because it3430 has some implication on large (e.g. GB)3431 allocations in some configurations (e.g. small3432 memory blocks).3433 The state of the flag can be read in3434 /sys/module/memory_hotplug/parameters/memmap_on_memory.3435 Note that even when enabled, there are a few cases where3436 the feature is not effective.3437 3438 memtest= [KNL,X86,ARM,M68K,PPC,RISCV,EARLY] Enable memtest3439 Format: <integer>3440 default : 0 <disable>3441 Specifies the number of memtest passes to be3442 performed. Each pass selects another test3443 pattern from a given set of patterns. Memtest3444 fills the memory with this pattern, validates3445 memory contents and reserves bad memory3446 regions that are detected.3447 3448 mem_encrypt= [X86-64] AMD Secure Memory Encryption (SME) control3449 Valid arguments: on, off3450 Default: off3451 mem_encrypt=on: Activate SME3452 mem_encrypt=off: Do not activate SME3453 3454 Refer to Documentation/virt/kvm/x86/amd-memory-encryption.rst3455 for details on when memory encryption can be activated.3456 3457 mem_sleep_default= [SUSPEND] Default system suspend mode:3458 s2idle - Suspend-To-Idle3459 shallow - Power-On Suspend or equivalent (if supported)3460 deep - Suspend-To-RAM or equivalent (if supported)3461 See Documentation/admin-guide/pm/sleep-states.rst.3462 3463 mfgptfix [X86-32] Fix MFGPT timers on AMD Geode platforms when3464 the BIOS has incorrectly applied a workaround. TinyBIOS3465 version 0.98 is known to be affected, 0.99 fixes the3466 problem by letting the user disable the workaround.3467 3468 mga= [HW,DRM]3469 3470 microcode.force_minrev= [X86]3471 Format: <bool>3472 Enable or disable the microcode minimal revision3473 enforcement for the runtime microcode loader.3474 3475 mini2440= [ARM,HW,KNL]3476 Format:[0..2][b][c][t]3477 Default: "0tb"3478 MINI2440 configuration specification:3479 0 - The attached screen is the 3.5" TFT3480 1 - The attached screen is the 7" TFT3481 2 - The VGA Shield is attached (1024x768)3482 Leaving out the screen size parameter will not load3483 the TFT driver, and the framebuffer will be left3484 unconfigured.3485 b - Enable backlight. The TFT backlight pin will be3486 linked to the kernel VESA blanking code and a GPIO3487 LED. This parameter is not necessary when using the3488 VGA shield.3489 c - Enable the s3c camera interface.3490 t - Reserved for enabling touchscreen support. The3491 touchscreen support is not enabled in the mainstream3492 kernel as of 2.6.30, a preliminary port can be found3493 in the "bleeding edge" mini2440 support kernel at3494 https://repo.or.cz/w/linux-2.6/mini2440.git3495 3496 mitigations=3497 [X86,PPC,S390,ARM64,EARLY] Control optional mitigations for3498 CPU vulnerabilities. This is a set of curated,3499 arch-independent options, each of which is an3500 aggregation of existing arch-specific options.3501 3502 Note, "mitigations" is supported if and only if the3503 kernel was built with CPU_MITIGATIONS=y.3504 3505 off3506 Disable all optional CPU mitigations. This3507 improves system performance, but it may also3508 expose users to several CPU vulnerabilities.3509 Equivalent to: if nokaslr then kpti=0 [ARM64]3510 gather_data_sampling=off [X86]3511 kvm.nx_huge_pages=off [X86]3512 l1tf=off [X86]3513 mds=off [X86]3514 mmio_stale_data=off [X86]3515 no_entry_flush [PPC]3516 no_uaccess_flush [PPC]3517 nobp=0 [S390]3518 nopti [X86,PPC]3519 nospectre_bhb [ARM64]3520 nospectre_v1 [X86,PPC]3521 nospectre_v2 [X86,PPC,S390,ARM64]3522 reg_file_data_sampling=off [X86]3523 retbleed=off [X86]3524 spec_rstack_overflow=off [X86]3525 spec_store_bypass_disable=off [X86,PPC]3526 spectre_bhi=off [X86]3527 spectre_v2_user=off [X86]3528 srbds=off [X86,INTEL]3529 ssbd=force-off [ARM64]3530 tsx_async_abort=off [X86]3531 3532 Exceptions:3533 This does not have any effect on3534 kvm.nx_huge_pages when3535 kvm.nx_huge_pages=force.3536 3537 auto (default)3538 Mitigate all CPU vulnerabilities, but leave SMT3539 enabled, even if it's vulnerable. This is for3540 users who don't want to be surprised by SMT3541 getting disabled across kernel upgrades, or who3542 have other ways of avoiding SMT-based attacks.3543 Equivalent to: (default behavior)3544 3545 auto,nosmt3546 Mitigate all CPU vulnerabilities, disabling SMT3547 if needed. This is for users who always want to3548 be fully mitigated, even if it means losing SMT.3549 Equivalent to: l1tf=flush,nosmt [X86]3550 mds=full,nosmt [X86]3551 tsx_async_abort=full,nosmt [X86]3552 mmio_stale_data=full,nosmt [X86]3553 retbleed=auto,nosmt [X86]3554 3555 mminit_loglevel=3556 [KNL,EARLY] When CONFIG_DEBUG_MEMORY_INIT is set, this3557 parameter allows control of the logging verbosity for3558 the additional memory initialisation checks. A value3559 of 0 disables mminit logging and a level of 4 will3560 log everything. Information is printed at KERN_DEBUG3561 so loglevel=8 may also need to be specified.3562 3563 mmio_stale_data=3564 [X86,INTEL,EARLY] Control mitigation for the Processor3565 MMIO Stale Data vulnerabilities.3566 3567 Processor MMIO Stale Data is a class of3568 vulnerabilities that may expose data after an MMIO3569 operation. Exposed data could originate or end in3570 the same CPU buffers as affected by MDS and TAA.3571 Therefore, similar to MDS and TAA, the mitigation3572 is to clear the affected CPU buffers.3573 3574 This parameter controls the mitigation. The3575 options are:3576 3577 full - Enable mitigation on vulnerable CPUs3578 3579 full,nosmt - Enable mitigation and disable SMT on3580 vulnerable CPUs.3581 3582 off - Unconditionally disable mitigation3583 3584 On MDS or TAA affected machines,3585 mmio_stale_data=off can be prevented by an active3586 MDS or TAA mitigation as these vulnerabilities are3587 mitigated with the same mechanism so in order to3588 disable this mitigation, you need to specify3589 mds=off and tsx_async_abort=off too.3590 3591 Not specifying this option is equivalent to3592 mmio_stale_data=full.3593 3594 For details see:3595 Documentation/admin-guide/hw-vuln/processor_mmio_stale_data.rst3596 3597 <module>.async_probe[=<bool>] [KNL]3598 If no <bool> value is specified or if the value3599 specified is not a valid <bool>, enable asynchronous3600 probe on this module. Otherwise, enable/disable3601 asynchronous probe on this module as indicated by the3602 <bool> value. See also: module.async_probe3603 3604 module.async_probe=<bool>3605 [KNL] When set to true, modules will use async probing3606 by default. To enable/disable async probing for a3607 specific module, use the module specific control that3608 is documented under <module>.async_probe. When both3609 module.async_probe and <module>.async_probe are3610 specified, <module>.async_probe takes precedence for3611 the specific module.3612 3613 module.enable_dups_trace3614 [KNL] When CONFIG_MODULE_DEBUG_AUTOLOAD_DUPS is set,3615 this means that duplicate request_module() calls will3616 trigger a WARN_ON() instead of a pr_warn(). Note that3617 if MODULE_DEBUG_AUTOLOAD_DUPS_TRACE is set, WARN_ON()s3618 will always be issued and this option does nothing.3619 module.sig_enforce3620 [KNL] When CONFIG_MODULE_SIG is set, this means that3621 modules without (valid) signatures will fail to load.3622 Note that if CONFIG_MODULE_SIG_FORCE is set, that3623 is always true, so this option does nothing.3624 3625 module_blacklist= [KNL] Do not load a comma-separated list of3626 modules. Useful for debugging problem modules.3627 3628 mousedev.tap_time=3629 [MOUSE] Maximum time between finger touching and3630 leaving touchpad surface for touch to be considered3631 a tap and be reported as a left button click (for3632 touchpads working in absolute mode only).3633 Format: <msecs>3634 mousedev.xres= [MOUSE] Horizontal screen resolution, used for devices3635 reporting absolute coordinates, such as tablets3636 mousedev.yres= [MOUSE] Vertical screen resolution, used for devices3637 reporting absolute coordinates, such as tablets3638 3639 movablecore= [KNL,X86,PPC,EARLY]3640 Format: nn[KMGTPE] | nn%3641 This parameter is the complement to kernelcore=, it3642 specifies the amount of memory used for migratable3643 allocations. If both kernelcore and movablecore is3644 specified, then kernelcore will be at *least* the3645 specified value but may be more. If movablecore on its3646 own is specified, the administrator must be careful3647 that the amount of memory usable for all allocations3648 is not too small.3649 3650 movable_node [KNL,EARLY] Boot-time switch to make hotplugable memory3651 NUMA nodes to be movable. This means that the memory3652 of such nodes will be usable only for movable3653 allocations which rules out almost all kernel3654 allocations. Use with caution!3655 3656 MTD_Partition= [MTD]3657 Format: <name>,<region-number>,<size>,<offset>3658 3659 MTD_Region= [MTD] Format:3660 <name>,<region-number>[,<base>,<size>,<buswidth>,<altbuswidth>]3661 3662 mtdparts= [MTD]3663 See drivers/mtd/parsers/cmdlinepart.c3664 3665 mtouchusb.raw_coordinates=3666 [HW] Make the MicroTouch USB driver use raw coordinates3667 ('y', default) or cooked coordinates ('n')3668 3669 mtrr=debug [X86,EARLY]3670 Enable printing debug information related to MTRR3671 registers at boot time.3672 3673 mtrr_chunk_size=nn[KMG,X86,EARLY]3674 used for mtrr cleanup. It is largest continuous chunk3675 that could hold holes aka. UC entries.3676 3677 mtrr_gran_size=nn[KMG,X86,EARLY]3678 Used for mtrr cleanup. It is granularity of mtrr block.3679 Default is 1.3680 Large value could prevent small alignment from3681 using up MTRRs.3682 3683 mtrr_spare_reg_nr=n [X86,EARLY]3684 Format: <integer>3685 Range: 0,7 : spare reg number3686 Default : 13687 Used for mtrr cleanup. It is spare mtrr entries number.3688 Set to 2 or more if your graphical card needs more.3689 3690 multitce=off [PPC] This parameter disables the use of the pSeries3691 firmware feature for updating multiple TCE entries3692 at a time.3693 3694 n2= [NET] SDL Inc. RISCom/N2 synchronous serial card3695 3696 netdev= [NET] Network devices parameters3697 Format: <irq>,<io>,<mem_start>,<mem_end>,<name>3698 Note that mem_start is often overloaded to mean3699 something different and driver-specific.3700 This usage is only documented in each driver source3701 file if at all.3702 3703 netpoll.carrier_timeout=3704 [NET] Specifies amount of time (in seconds) that3705 netpoll should wait for a carrier. By default netpoll3706 waits 4 seconds.3707 3708 nf_conntrack.acct=3709 [NETFILTER] Enable connection tracking flow accounting3710 0 to disable accounting3711 1 to enable accounting3712 Default value is 0.3713 3714 nfs.cache_getent=3715 [NFS] sets the pathname to the program which is used3716 to update the NFS client cache entries.3717 3718 nfs.cache_getent_timeout=3719 [NFS] sets the timeout after which an attempt to3720 update a cache entry is deemed to have failed.3721 3722 nfs.callback_nr_threads=3723 [NFSv4] set the total number of threads that the3724 NFS client will assign to service NFSv4 callback3725 requests.3726 3727 nfs.callback_tcpport=3728 [NFS] set the TCP port on which the NFSv4 callback3729 channel should listen.3730 3731 nfs.delay_retrans=3732 [NFS] specifies the number of times the NFSv4 client3733 retries the request before returning an EAGAIN error,3734 after a reply of NFS4ERR_DELAY from the server.3735 Only applies if the softerr mount option is enabled,3736 and the specified value is >= 0.3737 3738 nfs.enable_ino64=3739 [NFS] enable 64-bit inode numbers.3740 If zero, the NFS client will fake up a 32-bit inode3741 number for the readdir() and stat() syscalls instead3742 of returning the full 64-bit number.3743 The default is to return 64-bit inode numbers.3744 3745 nfs.idmap_cache_timeout=3746 [NFS] set the maximum lifetime for idmapper cache3747 entries.3748 3749 nfs.max_session_cb_slots=3750 [NFSv4.1] Sets the maximum number of session3751 slots the client will assign to the callback3752 channel. This determines the maximum number of3753 callbacks the client will process in parallel for3754 a particular server.3755 3756 nfs.max_session_slots=3757 [NFSv4.1] Sets the maximum number of session slots3758 the client will attempt to negotiate with the server.3759 This limits the number of simultaneous RPC requests3760 that the client can send to the NFSv4.1 server.3761 Note that there is little point in setting this3762 value higher than the max_tcp_slot_table_limit.3763 3764 nfs.nfs4_disable_idmapping=3765 [NFSv4] When set to the default of '1', this option3766 ensures that both the RPC level authentication3767 scheme and the NFS level operations agree to use3768 numeric uids/gids if the mount is using the3769 'sec=sys' security flavour. In effect it is3770 disabling idmapping, which can make migration from3771 legacy NFSv2/v3 systems to NFSv4 easier.3772 Servers that do not support this mode of operation3773 will be autodetected by the client, and it will fall3774 back to using the idmapper.3775 To turn off this behaviour, set the value to '0'.3776 3777 nfs.nfs4_unique_id=3778 [NFS4] Specify an additional fixed unique ident-3779 ification string that NFSv4 clients can insert into3780 their nfs_client_id4 string. This is typically a3781 UUID that is generated at system install time.3782 3783 nfs.recover_lost_locks=3784 [NFSv4] Attempt to recover locks that were lost due3785 to a lease timeout on the server. Please note that3786 doing this risks data corruption, since there are3787 no guarantees that the file will remain unchanged3788 after the locks are lost.3789 If you want to enable the kernel legacy behaviour of3790 attempting to recover these locks, then set this3791 parameter to '1'.3792 The default parameter value of '0' causes the kernel3793 not to attempt recovery of lost locks.3794 3795 nfs.send_implementation_id=3796 [NFSv4.1] Send client implementation identification3797 information in exchange_id requests.3798 If zero, no implementation identification information3799 will be sent.3800 The default is to send the implementation identification3801 information.3802 3803 nfs4.layoutstats_timer=3804 [NFSv4.2] Change the rate at which the kernel sends3805 layoutstats to the pNFS metadata server.3806 3807 Setting this to value to 0 causes the kernel to use3808 whatever value is the default set by the layout3809 driver. A non-zero value sets the minimum interval3810 in seconds between layoutstats transmissions.3811 3812 nfsd.inter_copy_offload_enable=3813 [NFSv4.2] When set to 1, the server will support3814 server-to-server copies for which this server is3815 the destination of the copy.3816 3817 nfsd.nfs4_disable_idmapping=3818 [NFSv4] When set to the default of '1', the NFSv43819 server will return only numeric uids and gids to3820 clients using auth_sys, and will accept numeric uids3821 and gids from such clients. This is intended to ease3822 migration from NFSv2/v3.3823 3824 nfsd.nfsd4_ssc_umount_timeout=3825 [NFSv4.2] When used as the destination of a3826 server-to-server copy, knfsd temporarily mounts3827 the source server. It caches the mount in case3828 it will be needed again, and discards it if not3829 used for the number of milliseconds specified by3830 this parameter.3831 3832 nfsaddrs= [NFS] Deprecated. Use ip= instead.3833 See Documentation/admin-guide/nfs/nfsroot.rst.3834 3835 nfsroot= [NFS] nfs root filesystem for disk-less boxes.3836 See Documentation/admin-guide/nfs/nfsroot.rst.3837 3838 nfsrootdebug [NFS] enable nfsroot debugging messages.3839 See Documentation/admin-guide/nfs/nfsroot.rst.3840 3841 nmi_backtrace.backtrace_idle [KNL]3842 Dump stacks even of idle CPUs in response to an3843 NMI stack-backtrace request.3844 3845 nmi_debug= [KNL,SH] Specify one or more actions to take3846 when a NMI is triggered.3847 Format: [state][,regs][,debounce][,die]3848 3849 nmi_watchdog= [KNL,BUGS=X86] Debugging features for SMP kernels3850 Format: [panic,][nopanic,][rNNN,][num]3851 Valid num: 0 or 13852 0 - turn hardlockup detector in nmi_watchdog off3853 1 - turn hardlockup detector in nmi_watchdog on3854 rNNN - configure the watchdog with raw perf event 0xNNN3855 3856 When panic is specified, panic when an NMI watchdog3857 timeout occurs (or 'nopanic' to not panic on an NMI3858 watchdog, if CONFIG_BOOTPARAM_HARDLOCKUP_PANIC is set)3859 To disable both hard and soft lockup detectors,3860 please see 'nowatchdog'.3861 This is useful when you use a panic=... timeout and3862 need the box quickly up again.3863 3864 These settings can be accessed at runtime via3865 the nmi_watchdog and hardlockup_panic sysctls.3866 3867 no387 [BUGS=X86-32] Tells the kernel to use the 387 maths3868 emulation library even if a 387 maths coprocessor3869 is present.3870 3871 no4lvl [RISCV,EARLY] Disable 4-level and 5-level paging modes.3872 Forces kernel to use 3-level paging instead.3873 3874 no5lvl [X86-64,RISCV,EARLY] Disable 5-level paging mode. Forces3875 kernel to use 4-level paging instead.3876 3877 noalign [KNL,ARM]3878 3879 noapic [SMP,APIC,EARLY] Tells the kernel to not make use of any3880 IOAPICs that may be present in the system.3881 3882 noautogroup Disable scheduler automatic task group creation.3883 3884 nocache [ARM,EARLY]3885 3886 no_console_suspend3887 [HW] Never suspend the console3888 Disable suspending of consoles during suspend and3889 hibernate operations. Once disabled, debugging3890 messages can reach various consoles while the rest3891 of the system is being put to sleep (ie, while3892 debugging driver suspend/resume hooks). This may3893 not work reliably with all consoles, but is known3894 to work with serial and VGA consoles.3895 To facilitate more flexible debugging, we also add3896 console_suspend, a printk module parameter to control3897 it. Users could use console_suspend (usually3898 /sys/module/printk/parameters/console_suspend) to3899 turn on/off it dynamically.3900 3901 no_debug_objects3902 [KNL,EARLY] Disable object debugging3903 3904 nodsp [SH] Disable hardware DSP at boot time.3905 3906 noefi [EFI,EARLY] Disable EFI runtime services support.3907 3908 no_entry_flush [PPC,EARLY] Don't flush the L1-D cache when entering the kernel.3909 3910 noexec32 [X86-64]3911 This affects only 32-bit executables.3912 noexec32=on: enable non-executable mappings (default)3913 read doesn't imply executable mappings3914 noexec32=off: disable non-executable mappings3915 read implies executable mappings3916 3917 no_file_caps Tells the kernel not to honor file capabilities. The3918 only way then for a file to be executed with privilege3919 is to be setuid root or executed by root.3920 3921 nofpu [MIPS,SH] Disable hardware FPU at boot time.3922 3923 nofsgsbase [X86] Disables FSGSBASE instructions.3924 3925 nofxsr [BUGS=X86-32] Disables x86 floating point extended3926 register save and restore. The kernel will only save3927 legacy floating-point registers on task switch.3928 3929 no_hash_pointers3930 [KNL,EARLY]3931 Force pointers printed to the console or buffers to be3932 unhashed. By default, when a pointer is printed via %p3933 format string, that pointer is "hashed", i.e. obscured3934 by hashing the pointer value. This is a security feature3935 that hides actual kernel addresses from unprivileged3936 users, but it also makes debugging the kernel more3937 difficult since unequal pointers can no longer be3938 compared. However, if this command-line option is3939 specified, then all normal pointers will have their true3940 value printed. This option should only be specified when3941 debugging the kernel. Please do not use on production3942 kernels.3943 3944 nohibernate [HIBERNATION] Disable hibernation and resume.3945 3946 nohlt [ARM,ARM64,MICROBLAZE,MIPS,PPC,RISCV,SH] Forces the kernel to3947 busy wait in do_idle() and not use the arch_cpu_idle()3948 implementation; requires CONFIG_GENERIC_IDLE_POLL_SETUP3949 to be effective. This is useful on platforms where the3950 sleep(SH) or wfi(ARM,ARM64) instructions do not work3951 correctly or when doing power measurements to evaluate3952 the impact of the sleep instructions. This is also3953 useful when using JTAG debugger.3954 3955 nohugeiomap [KNL,X86,PPC,ARM64,EARLY] Disable kernel huge I/O mappings.3956 3957 nohugevmalloc [KNL,X86,PPC,ARM64,EARLY] Disable kernel huge vmalloc mappings.3958 3959 nohz= [KNL] Boottime enable/disable dynamic ticks3960 Valid arguments: on, off3961 Default: on3962 3963 nohz_full= [KNL,BOOT,SMP,ISOL]3964 The argument is a cpu list, as described above.3965 In kernels built with CONFIG_NO_HZ_FULL=y, set3966 the specified list of CPUs whose tick will be stopped3967 whenever possible. The boot CPU will be forced outside3968 the range to maintain the timekeeping. Any CPUs3969 in this list will have their RCU callbacks offloaded,3970 just as if they had also been called out in the3971 rcu_nocbs= boot parameter.3972 3973 Note that this argument takes precedence over3974 the CONFIG_RCU_NOCB_CPU_DEFAULT_ALL option.3975 3976 noinitrd [RAM] Tells the kernel not to load any configured3977 initial RAM disk.3978 3979 nointremap [X86-64,Intel-IOMMU,EARLY] Do not enable interrupt3980 remapping.3981 [Deprecated - use intremap=off]3982 3983 noinvpcid [X86,EARLY] Disable the INVPCID cpu feature.3984 3985 noiotrap [SH] Disables trapped I/O port accesses.3986 3987 noirqdebug [X86-32] Disables the code which attempts to detect and3988 disable unhandled interrupt sources.3989 3990 noisapnp [ISAPNP] Disables ISA PnP code.3991 3992 nokaslr [KNL,EARLY]3993 When CONFIG_RANDOMIZE_BASE is set, this disables3994 kernel and module base offset ASLR (Address Space3995 Layout Randomization).3996 3997 no-kvmapf [X86,KVM,EARLY] Disable paravirtualized asynchronous page3998 fault handling.3999 4000 no-kvmclock [X86,KVM,EARLY] Disable paravirtualized KVM clock driver4001 4002 nolapic [X86-32,APIC,EARLY] Do not enable or use the local APIC.4003 4004 nolapic_timer [X86-32,APIC,EARLY] Do not use the local APIC timer.4005 4006 nomce [X86-32] Disable Machine Check Exception4007 4008 nomfgpt [X86-32] Disable Multi-Function General Purpose4009 Timer usage (for AMD Geode machines).4010 4011 nomodeset Disable kernel modesetting. Most systems' firmware4012 sets up a display mode and provides framebuffer memory4013 for output. With nomodeset, DRM and fbdev drivers will4014 not load if they could possibly displace the pre-4015 initialized output. Only the system framebuffer will4016 be available for use. The respective drivers will not4017 perform display-mode changes or accelerated rendering.4018 4019 Useful as error fallback, or for testing and debugging.4020 4021 nomodule Disable module load4022 4023 nonmi_ipi [X86] Disable using NMI IPIs during panic/reboot to4024 shutdown the other cpus. Instead use the REBOOT_VECTOR4025 irq.4026 4027 nopat [X86,EARLY] Disable PAT (page attribute table extension of4028 pagetables) support.4029 4030 nopcid [X86-64,EARLY] Disable the PCID cpu feature.4031 4032 nopku [X86] Disable Memory Protection Keys CPU feature found4033 in some Intel CPUs.4034 4035 nopti [X86-64,EARLY]4036 Equivalent to pti=off4037 4038 nopv= [X86,XEN,KVM,HYPER_V,VMWARE,EARLY]4039 Disables the PV optimizations forcing the guest to run4040 as generic guest with no PV drivers. Currently support4041 XEN HVM, KVM, HYPER_V and VMWARE guest.4042 4043 nopvspin [X86,XEN,KVM,EARLY]4044 Disables the qspinlock slow path using PV optimizations4045 which allow the hypervisor to 'idle' the guest on lock4046 contention.4047 4048 norandmaps Don't use address space randomization. Equivalent to4049 echo 0 > /proc/sys/kernel/randomize_va_space4050 4051 noreplace-smp [X86-32,SMP] Don't replace SMP instructions4052 with UP alternatives4053 4054 noresume [SWSUSP] Disables resume and restores original swap4055 space.4056 4057 no-scroll [VGA] Disables scrollback.4058 This is required for the Braillex ib80-piezo Braille4059 reader made by F.H. Papenmeier (Germany).4060 4061 nosgx [X86-64,SGX,EARLY] Disables Intel SGX kernel support.4062 4063 nosmap [PPC,EARLY]4064 Disable SMAP (Supervisor Mode Access Prevention)4065 even if it is supported by processor.4066 4067 nosmep [PPC64s,EARLY]4068 Disable SMEP (Supervisor Mode Execution Prevention)4069 even if it is supported by processor.4070 4071 nosmp [SMP,EARLY] Tells an SMP kernel to act as a UP kernel,4072 and disable the IO APIC. legacy for "maxcpus=0".4073 4074 nosmt [KNL,MIPS,PPC,S390,EARLY] Disable symmetric multithreading (SMT).4075 Equivalent to smt=1.4076 4077 [KNL,X86,PPC] Disable symmetric multithreading (SMT).4078 nosmt=force: Force disable SMT, cannot be undone4079 via the sysfs control file.4080 4081 nosoftlockup [KNL] Disable the soft-lockup detector.4082 4083 nospec_store_bypass_disable4084 [HW,EARLY] Disable all mitigations for the Speculative4085 Store Bypass vulnerability4086 4087 nospectre_bhb [ARM64,EARLY] Disable all mitigations for Spectre-BHB (branch4088 history injection) vulnerability. System may allow data leaks4089 with this option.4090 4091 nospectre_v1 [X86,PPC,EARLY] Disable mitigations for Spectre Variant 14092 (bounds check bypass). With this option data leaks are4093 possible in the system.4094 4095 nospectre_v2 [X86,PPC_E500,ARM64,EARLY] Disable all mitigations4096 for the Spectre variant 2 (indirect branch4097 prediction) vulnerability. System may allow data4098 leaks with this option.4099 4100 no-steal-acc [X86,PV_OPS,ARM64,PPC/PSERIES,RISCV,LOONGARCH,EARLY]4101 Disable paravirtualized steal time accounting. steal time4102 is computed, but won't influence scheduler behaviour4103 4104 nosync [HW,M68K] Disables sync negotiation for all devices.4105 4106 no_timer_check [X86,APIC] Disables the code which tests for4107 broken timer IRQ sources.4108 4109 no_uaccess_flush4110 [PPC,EARLY] Don't flush the L1-D cache after accessing user data.4111 4112 novmcoredd [KNL,KDUMP]4113 Disable device dump. Device dump allows drivers to4114 append dump data to vmcore so you can collect driver4115 specified debug info. Drivers can append the data4116 without any limit and this data is stored in memory,4117 so this may cause significant memory stress. Disabling4118 device dump can help save memory but the driver debug4119 data will be no longer available. This parameter4120 is only available when CONFIG_PROC_VMCORE_DEVICE_DUMP4121 is set.4122 4123 no-vmw-sched-clock4124 [X86,PV_OPS,EARLY] Disable paravirtualized VMware4125 scheduler clock and use the default one.4126 4127 nowatchdog [KNL] Disable both lockup detectors, i.e.4128 soft-lockup and NMI watchdog (hard-lockup).4129 4130 nowb [ARM,EARLY]4131 4132 nox2apic [X86-64,APIC,EARLY] Do not enable x2APIC mode.4133 4134 NOTE: this parameter will be ignored on systems with the4135 LEGACY_XAPIC_DISABLED bit set in the4136 IA32_XAPIC_DISABLE_STATUS MSR.4137 4138 noxsave [BUGS=X86] Disables x86 extended register state save4139 and restore using xsave. The kernel will fallback to4140 enabling legacy floating-point and sse state.4141 4142 noxsaveopt [X86] Disables xsaveopt used in saving x86 extended4143 register states. The kernel will fall back to use4144 xsave to save the states. By using this parameter,4145 performance of saving the states is degraded because4146 xsave doesn't support modified optimization while4147 xsaveopt supports it on xsaveopt enabled systems.4148 4149 noxsaves [X86] Disables xsaves and xrstors used in saving and4150 restoring x86 extended register state in compacted4151 form of xsave area. The kernel will fall back to use4152 xsaveopt and xrstor to save and restore the states4153 in standard form of xsave area. By using this4154 parameter, xsave area per process might occupy more4155 memory on xsaves enabled systems.4156 4157 nr_cpus= [SMP,EARLY] Maximum number of processors that an SMP kernel4158 could support. nr_cpus=n : n >= 1 limits the kernel to4159 support 'n' processors. It could be larger than the4160 number of already plugged CPU during bootup, later in4161 runtime you can physically add extra cpu until it reaches4162 n. So during boot up some boot time memory for per-cpu4163 variables need be pre-allocated for later physical cpu4164 hot plugging.4165 4166 nr_uarts= [SERIAL] maximum number of UARTs to be registered.4167 4168 numa=off [KNL, ARM64, PPC, RISCV, SPARC, X86, EARLY]4169 Disable NUMA, Only set up a single NUMA node4170 spanning all memory.4171 4172 numa=fake=<size>[MG]4173 [KNL, ARM64, RISCV, X86, EARLY]4174 If given as a memory unit, fills all system RAM with4175 nodes of size interleaved over physical nodes.4176 4177 numa=fake=<N>4178 [KNL, ARM64, RISCV, X86, EARLY]4179 If given as an integer, fills all system RAM with N4180 fake nodes interleaved over physical nodes.4181 4182 numa=fake=<N>U4183 [KNL, ARM64, RISCV, X86, EARLY]4184 If given as an integer followed by 'U', it will4185 divide each physical node into N emulated nodes.4186 4187 numa_balancing= [KNL,ARM64,PPC,RISCV,S390,X86] Enable or disable automatic4188 NUMA balancing.4189 Allowed values are enable and disable4190 4191 numa_zonelist_order= [KNL, BOOT] Select zonelist order for NUMA.4192 'node', 'default' can be specified4193 This can be set from sysctl after boot.4194 See Documentation/admin-guide/sysctl/vm.rst for details.4195 4196 ohci1394_dma=early [HW,EARLY] enable debugging via the ohci1394 driver.4197 See Documentation/core-api/debugging-via-ohci1394.rst for more4198 info.4199 4200 olpc_ec_timeout= [OLPC] ms delay when issuing EC commands4201 Rather than timing out after 20 ms if an EC4202 command is not properly ACKed, override the length4203 of the timeout. We have interrupts disabled while4204 waiting for the ACK, so if this is set too high4205 interrupts *may* be lost!4206 4207 omap_mux= [OMAP] Override bootloader pin multiplexing.4208 Format: <mux_mode0.mode_name=value>...4209 For example, to override I2C bus2:4210 omap_mux=i2c2_scl.i2c2_scl=0x100,i2c2_sda.i2c2_sda=0x1004211 4212 onenand.bdry= [HW,MTD] Flex-OneNAND Boundary Configuration4213 4214 Format: [die0_boundary][,die0_lock][,die1_boundary][,die1_lock]4215 4216 boundary - index of last SLC block on Flex-OneNAND.4217 The remaining blocks are configured as MLC blocks.4218 lock - Configure if Flex-OneNAND boundary should be locked.4219 Once locked, the boundary cannot be changed.4220 1 indicates lock status, 0 indicates unlock status.4221 4222 oops=panic [KNL,EARLY]4223 Always panic on oopses. Default is to just kill the4224 process, but there is a small probability of4225 deadlocking the machine.4226 This will also cause panics on machine check exceptions.4227 Useful together with panic=30 to trigger a reboot.4228 4229 page_alloc.shuffle=4230 [KNL] Boolean flag to control whether the page allocator4231 should randomize its free lists. This parameter can be4232 used to enable/disable page randomization. The state of4233 the flag can be read from sysfs at:4234 /sys/module/page_alloc/parameters/shuffle.4235 This parameter is only available if CONFIG_SHUFFLE_PAGE_ALLOCATOR=y.4236 4237 page_owner= [KNL,EARLY] Boot-time page_owner enabling option.4238 Storage of the information about who allocated4239 each page is disabled in default. With this switch,4240 we can turn it on.4241 on: enable the feature4242 4243 page_poison= [KNL,EARLY] Boot-time parameter changing the state of4244 poisoning on the buddy allocator, available with4245 CONFIG_PAGE_POISONING=y.4246 off: turn off poisoning (default)4247 on: turn on poisoning4248 4249 page_reporting.page_reporting_order=4250 [KNL] Minimal page reporting order4251 Format: <integer>4252 Adjust the minimal page reporting order. The page4253 reporting is disabled when it exceeds MAX_PAGE_ORDER.4254 4255 panic= [KNL] Kernel behaviour on panic: delay <timeout>4256 timeout > 0: seconds before rebooting4257 timeout = 0: wait forever4258 timeout < 0: reboot immediately4259 Format: <timeout>4260 4261 panic_on_taint= [KNL,EARLY]4262 Bitmask for conditionally calling panic() in add_taint()4263 Format: <hex>[,nousertaint]4264 Hexadecimal bitmask representing the set of TAINT flags4265 that will cause the kernel to panic when add_taint() is4266 called with any of the flags in this set.4267 The optional switch "nousertaint" can be utilized to4268 prevent userspace forced crashes by writing to sysctl4269 /proc/sys/kernel/tainted any flagset matching with the4270 bitmask set on panic_on_taint.4271 See Documentation/admin-guide/tainted-kernels.rst for4272 extra details on the taint flags that users can pick4273 to compose the bitmask to assign to panic_on_taint.4274 4275 panic_on_warn=1 panic() instead of WARN(). Useful to cause kdump4276 on a WARN().4277 4278 panic_print= Bitmask for printing system info when panic happens.4279 User can chose combination of the following bits:4280 bit 0: print all tasks info4281 bit 1: print system memory info4282 bit 2: print timer info4283 bit 3: print locks info if CONFIG_LOCKDEP is on4284 bit 4: print ftrace buffer4285 bit 5: print all printk messages in buffer4286 bit 6: print all CPUs backtrace (if available in the arch)4287 bit 7: print only tasks in uninterruptible (blocked) state4288 *Be aware* that this option may print a _lot_ of lines,4289 so there are risks of losing older messages in the log.4290 Use this option carefully, maybe worth to setup a4291 bigger log buffer with "log_buf_len" along with this.4292 4293 parkbd.port= [HW] Parallel port number the keyboard adapter is4294 connected to, default is 0.4295 Format: <parport#>4296 parkbd.mode= [HW] Parallel port keyboard adapter mode of operation,4297 0 for XT, 1 for AT (default is AT).4298 Format: <mode>4299 4300 parport= [HW,PPT] Specify parallel ports. 0 disables.4301 Format: { 0 | auto | 0xBBB[,IRQ[,DMA]] }4302 Use 'auto' to force the driver to use any4303 IRQ/DMA settings detected (the default is to4304 ignore detected IRQ/DMA settings because of4305 possible conflicts). You can specify the base4306 address, IRQ, and DMA settings; IRQ and DMA4307 should be numbers, or 'auto' (for using detected4308 settings on that particular port), or 'nofifo'4309 (to avoid using a FIFO even if it is detected).4310 Parallel ports are assigned in the order they4311 are specified on the command line, starting4312 with parport0.4313 4314 parport_init_mode= [HW,PPT]4315 Configure VIA parallel port to operate in4316 a specific mode. This is necessary on Pegasos4317 computer where firmware has no options for setting4318 up parallel port mode and sets it to spp.4319 Currently this function knows 686a and 8231 chips.4320 Format: [spp|ps2|epp|ecp|ecpepp]4321 4322 pata_legacy.all= [HW,LIBATA]4323 Format: <int>4324 Set to non-zero to probe primary and secondary ISA4325 port ranges on PCI systems where no PCI PATA device4326 has been found at either range. Disabled by default.4327 4328 pata_legacy.autospeed= [HW,LIBATA]4329 Format: <int>4330 Set to non-zero if a chip is present that snoops speed4331 changes. Disabled by default.4332 4333 pata_legacy.ht6560a= [HW,LIBATA]4334 Format: <int>4335 Set to 1, 2, or 3 for HT 6560A on the primary channel,4336 the secondary channel, or both channels respectively.4337 Disabled by default.4338 4339 pata_legacy.ht6560b= [HW,LIBATA]4340 Format: <int>4341 Set to 1, 2, or 3 for HT 6560B on the primary channel,4342 the secondary channel, or both channels respectively.4343 Disabled by default.4344 4345 pata_legacy.iordy_mask= [HW,LIBATA]4346 Format: <int>4347 IORDY enable mask. Set individual bits to allow IORDY4348 for the respective channel. Bit 0 is for the first4349 legacy channel handled by this driver, bit 1 is for4350 the second channel, and so on. The sequence will often4351 correspond to the primary legacy channel, the secondary4352 legacy channel, and so on, but the handling of a PCI4353 bus and the use of other driver options may interfere4354 with the sequence. By default IORDY is allowed across4355 all channels.4356 4357 pata_legacy.opti82c46x= [HW,LIBATA]4358 Format: <int>4359 Set to 1, 2, or 3 for Opti 82c611A on the primary4360 channel, the secondary channel, or both channels4361 respectively. Disabled by default.4362 4363 pata_legacy.opti82c611a= [HW,LIBATA]4364 Format: <int>4365 Set to 1, 2, or 3 for Opti 82c465MV on the primary4366 channel, the secondary channel, or both channels4367 respectively. Disabled by default.4368 4369 pata_legacy.pio_mask= [HW,LIBATA]4370 Format: <int>4371 PIO mode mask for autospeed devices. Set individual4372 bits to allow the use of the respective PIO modes.4373 Bit 0 is for mode 0, bit 1 is for mode 1, and so on.4374 All modes allowed by default.4375 4376 pata_legacy.probe_all= [HW,LIBATA]4377 Format: <int>4378 Set to non-zero to probe tertiary and further ISA4379 port ranges on PCI systems. Disabled by default.4380 4381 pata_legacy.probe_mask= [HW,LIBATA]4382 Format: <int>4383 Probe mask for legacy ISA PATA ports. Depending on4384 platform configuration and the use of other driver4385 options up to 6 legacy ports are supported: 0x1f0,4386 0x170, 0x1e8, 0x168, 0x1e0, 0x160, however probing4387 of individual ports can be disabled by setting the4388 corresponding bits in the mask to 1. Bit 0 is for4389 the first port in the list above (0x1f0), and so on.4390 By default all supported ports are probed.4391 4392 pata_legacy.qdi= [HW,LIBATA]4393 Format: <int>4394 Set to non-zero to probe QDI controllers. By default4395 set to 1 if CONFIG_PATA_QDI_MODULE, 0 otherwise.4396 4397 pata_legacy.winbond= [HW,LIBATA]4398 Format: <int>4399 Set to non-zero to probe Winbond controllers. Use4400 the standard I/O port (0x130) if 1, otherwise the4401 value given is the I/O port to use (typically 0x1b0).4402 By default set to 1 if CONFIG_PATA_WINBOND_VLB_MODULE,4403 0 otherwise.4404 4405 pata_platform.pio_mask= [HW,LIBATA]4406 Format: <int>4407 Supported PIO mode mask. Set individual bits to allow4408 the use of the respective PIO modes. Bit 0 is for4409 mode 0, bit 1 is for mode 1, and so on. Mode 0 only4410 allowed by default.4411 4412 pause_on_oops=<int>4413 Halt all CPUs after the first oops has been printed for4414 the specified number of seconds. This is to be used if4415 your oopses keep scrolling off the screen.4416 4417 pcbit= [HW,ISDN]4418 4419 pci=option[,option...] [PCI,EARLY] various PCI subsystem options.4420 4421 Some options herein operate on a specific device4422 or a set of devices (<pci_dev>). These are4423 specified in one of the following formats:4424 4425 [<domain>:]<bus>:<dev>.<func>[/<dev>.<func>]*4426 pci:<vendor>:<device>[:<subvendor>:<subdevice>]4427 4428 Note: the first format specifies a PCI4429 bus/device/function address which may change4430 if new hardware is inserted, if motherboard4431 firmware changes, or due to changes caused4432 by other kernel parameters. If the4433 domain is left unspecified, it is4434 taken to be zero. Optionally, a path4435 to a device through multiple device/function4436 addresses can be specified after the base4437 address (this is more robust against4438 renumbering issues). The second format4439 selects devices using IDs from the4440 configuration space which may match multiple4441 devices in the system.4442 4443 earlydump dump PCI config space before the kernel4444 changes anything4445 off [X86] don't probe for the PCI bus4446 bios [X86-32] force use of PCI BIOS, don't access4447 the hardware directly. Use this if your machine4448 has a non-standard PCI host bridge.4449 nobios [X86-32] disallow use of PCI BIOS, only direct4450 hardware access methods are allowed. Use this4451 if you experience crashes upon bootup and you4452 suspect they are caused by the BIOS.4453 conf1 [X86] Force use of PCI Configuration Access4454 Mechanism 1 (config address in IO port 0xCF8,4455 data in IO port 0xCFC, both 32-bit).4456 conf2 [X86] Force use of PCI Configuration Access4457 Mechanism 2 (IO port 0xCF8 is an 8-bit port for4458 the function, IO port 0xCFA, also 8-bit, sets4459 bus number. The config space is then accessed4460 through ports 0xC000-0xCFFF).4461 See http://wiki.osdev.org/PCI for more info4462 on the configuration access mechanisms.4463 noaer [PCIE] If the PCIEAER kernel config parameter is4464 enabled, this kernel boot option can be used to4465 disable the use of PCIE advanced error reporting.4466 nodomains [PCI] Disable support for multiple PCI4467 root domains (aka PCI segments, in ACPI-speak).4468 nommconf [X86] Disable use of MMCONFIG for PCI4469 Configuration4470 check_enable_amd_mmconf [X86] check for and enable4471 properly configured MMIO access to PCI4472 config space on AMD family 10h CPU4473 nomsi [MSI] If the PCI_MSI kernel config parameter is4474 enabled, this kernel boot option can be used to4475 disable the use of MSI interrupts system-wide.4476 noioapicquirk [APIC] Disable all boot interrupt quirks.4477 Safety option to keep boot IRQs enabled. This4478 should never be necessary.4479 ioapicreroute [APIC] Enable rerouting of boot IRQs to the4480 primary IO-APIC for bridges that cannot disable4481 boot IRQs. This fixes a source of spurious IRQs4482 when the system masks IRQs.4483 noioapicreroute [APIC] Disable workaround that uses the4484 boot IRQ equivalent of an IRQ that connects to4485 a chipset where boot IRQs cannot be disabled.4486 The opposite of ioapicreroute.4487 biosirq [X86-32] Use PCI BIOS calls to get the interrupt4488 routing table. These calls are known to be buggy4489 on several machines and they hang the machine4490 when used, but on other computers it's the only4491 way to get the interrupt routing table. Try4492 this option if the kernel is unable to allocate4493 IRQs or discover secondary PCI buses on your4494 motherboard.4495 rom [X86] Assign address space to expansion ROMs.4496 Use with caution as certain devices share4497 address decoders between ROMs and other4498 resources.4499 norom [X86] Do not assign address space to4500 expansion ROMs that do not already have4501 BIOS assigned address ranges.4502 nobar [X86] Do not assign address space to the4503 BARs that weren't assigned by the BIOS.4504 irqmask=0xMMMM [X86] Set a bit mask of IRQs allowed to be4505 assigned automatically to PCI devices. You can4506 make the kernel exclude IRQs of your ISA cards4507 this way.4508 pirqaddr=0xAAAAA [X86] Specify the physical address4509 of the PIRQ table (normally generated4510 by the BIOS) if it is outside the4511 F0000h-100000h range.4512 lastbus=N [X86] Scan all buses thru bus #N. Can be4513 useful if the kernel is unable to find your4514 secondary buses and you want to tell it4515 explicitly which ones they are.4516 assign-busses [X86] Always assign all PCI bus4517 numbers ourselves, overriding4518 whatever the firmware may have done.4519 usepirqmask [X86] Honor the possible IRQ mask stored4520 in the BIOS $PIR table. This is needed on4521 some systems with broken BIOSes, notably4522 some HP Pavilion N5400 and Omnibook XE34523 notebooks. This will have no effect if ACPI4524 IRQ routing is enabled.4525 noacpi [X86] Do not use ACPI for IRQ routing4526 or for PCI scanning.4527 use_crs [X86] Use PCI host bridge window information4528 from ACPI. On BIOSes from 2008 or later, this4529 is enabled by default. If you need to use this,4530 please report a bug.4531 nocrs [X86] Ignore PCI host bridge windows from ACPI.4532 If you need to use this, please report a bug.4533 use_e820 [X86] Use E820 reservations to exclude parts of4534 PCI host bridge windows. This is a workaround4535 for BIOS defects in host bridge _CRS methods.4536 If you need to use this, please report a bug to4537 <linux-pci@vger.kernel.org>.4538 no_e820 [X86] Ignore E820 reservations for PCI host4539 bridge windows. This is the default on modern4540 hardware. If you need to use this, please report4541 a bug to <linux-pci@vger.kernel.org>.4542 routeirq Do IRQ routing for all PCI devices.4543 This is normally done in pci_enable_device(),4544 so this option is a temporary workaround4545 for broken drivers that don't call it.4546 skip_isa_align [X86] do not align io start addr, so can4547 handle more pci cards4548 noearly [X86] Don't do any early type 1 scanning.4549 This might help on some broken boards which4550 machine check when some devices' config space4551 is read. But various workarounds are disabled4552 and some IOMMU drivers will not work.4553 bfsort Sort PCI devices into breadth-first order.4554 This sorting is done to get a device4555 order compatible with older (<= 2.4) kernels.4556 nobfsort Don't sort PCI devices into breadth-first order.4557 pcie_bus_tune_off Disable PCIe MPS (Max Payload Size)4558 tuning and use the BIOS-configured MPS defaults.4559 pcie_bus_safe Set every device's MPS to the largest value4560 supported by all devices below the root complex.4561 pcie_bus_perf Set device MPS to the largest allowable MPS4562 based on its parent bus. Also set MRRS (Max4563 Read Request Size) to the largest supported4564 value (no larger than the MPS that the device4565 or bus can support) for best performance.4566 pcie_bus_peer2peer Set every device's MPS to 128B, which4567 every device is guaranteed to support. This4568 configuration allows peer-to-peer DMA between4569 any pair of devices, possibly at the cost of4570 reduced performance. This also guarantees4571 that hot-added devices will work.4572 cbiosize=nn[KMG] The fixed amount of bus space which is4573 reserved for the CardBus bridge's IO window.4574 The default value is 256 bytes.4575 cbmemsize=nn[KMG] The fixed amount of bus space which is4576 reserved for the CardBus bridge's memory4577 window. The default value is 64 megabytes.4578 resource_alignment=4579 Format:4580 [<order of align>@]<pci_dev>[; ...]4581 Specifies alignment and device to reassign4582 aligned memory resources. How to4583 specify the device is described above.4584 If <order of align> is not specified,4585 PAGE_SIZE is used as alignment.4586 A PCI-PCI bridge can be specified if resource4587 windows need to be expanded.4588 To specify the alignment for several4589 instances of a device, the PCI vendor,4590 device, subvendor, and subdevice may be4591 specified, e.g., 12@pci:8086:9c22:103c:198f4592 for 4096-byte alignment.4593 ecrc= Enable/disable PCIe ECRC (transaction layer4594 end-to-end CRC checking). Only effective if4595 OS has native AER control (either granted by4596 ACPI _OSC or forced via "pcie_ports=native")4597 bios: Use BIOS/firmware settings. This is the4598 the default.4599 off: Turn ECRC off4600 on: Turn ECRC on.4601 hpiosize=nn[KMG] The fixed amount of bus space which is4602 reserved for hotplug bridge's IO window.4603 Default size is 256 bytes.4604 hpmmiosize=nn[KMG] The fixed amount of bus space which is4605 reserved for hotplug bridge's MMIO window.4606 Default size is 2 megabytes.4607 hpmmioprefsize=nn[KMG] The fixed amount of bus space which is4608 reserved for hotplug bridge's MMIO_PREF window.4609 Default size is 2 megabytes.4610 hpmemsize=nn[KMG] The fixed amount of bus space which is4611 reserved for hotplug bridge's MMIO and4612 MMIO_PREF window.4613 Default size is 2 megabytes.4614 hpbussize=nn The minimum amount of additional bus numbers4615 reserved for buses below a hotplug bridge.4616 Default is 1.4617 realloc= Enable/disable reallocating PCI bridge resources4618 if allocations done by BIOS are too small to4619 accommodate resources required by all child4620 devices.4621 off: Turn realloc off4622 on: Turn realloc on4623 realloc same as realloc=on4624 noari do not use PCIe ARI.4625 noats [PCIE, Intel-IOMMU, AMD-IOMMU]4626 do not use PCIe ATS (and IOMMU device IOTLB).4627 pcie_scan_all Scan all possible PCIe devices. Otherwise we4628 only look for one device below a PCIe downstream4629 port.4630 big_root_window Try to add a big 64bit memory window to the PCIe4631 root complex on AMD CPUs. Some GFX hardware4632 can resize a BAR to allow access to all VRAM.4633 Adding the window is slightly risky (it may4634 conflict with unreported devices), so this4635 taints the kernel.4636 disable_acs_redir=<pci_dev>[; ...]4637 Specify one or more PCI devices (in the format4638 specified above) separated by semicolons.4639 Each device specified will have the PCI ACS4640 redirect capabilities forced off which will4641 allow P2P traffic between devices through4642 bridges without forcing it upstream. Note:4643 this removes isolation between devices and4644 may put more devices in an IOMMU group.4645 config_acs=4646 Format:4647 <ACS flags>@<pci_dev>[; ...]4648 Specify one or more PCI devices (in the format4649 specified above) optionally prepended with flags4650 and separated by semicolons. The respective4651 capabilities will be enabled, disabled or4652 unchanged based on what is specified in4653 flags.4654 4655 ACS Flags is defined as follows:4656 bit-0 : ACS Source Validation4657 bit-1 : ACS Translation Blocking4658 bit-2 : ACS P2P Request Redirect4659 bit-3 : ACS P2P Completion Redirect4660 bit-4 : ACS Upstream Forwarding4661 bit-5 : ACS P2P Egress Control4662 bit-6 : ACS Direct Translated P2P4663 Each bit can be marked as:4664 '0' – force disabled4665 '1' – force enabled4666 'x' – unchanged4667 For example,4668 pci=config_acs=10x4669 would configure all devices that support4670 ACS to enable P2P Request Redirect, disable4671 Translation Blocking, and leave Source4672 Validation unchanged from whatever power-up4673 or firmware set it to.4674 4675 Note: this may remove isolation between devices4676 and may put more devices in an IOMMU group.4677 force_floating [S390] Force usage of floating interrupts.4678 nomio [S390] Do not use MIO instructions.4679 norid [S390] ignore the RID field and force use of4680 one PCI domain per PCI function4681 4682 pcie_aspm= [PCIE] Forcibly enable or ignore PCIe Active State Power4683 Management.4684 off Don't touch ASPM configuration at all. Leave any4685 configuration done by firmware unchanged.4686 force Enable ASPM even on devices that claim not to support it.4687 WARNING: Forcing ASPM on may cause system lockups.4688 4689 pcie_ports= [PCIE] PCIe port services handling:4690 native Use native PCIe services (PME, AER, DPC, PCIe hotplug)4691 even if the platform doesn't give the OS permission to4692 use them. This may cause conflicts if the platform4693 also tries to use these services.4694 dpc-native Use native PCIe service for DPC only. May4695 cause conflicts if firmware uses AER or DPC.4696 compat Disable native PCIe services (PME, AER, DPC, PCIe4697 hotplug).4698 4699 pcie_port_pm= [PCIE] PCIe port power management handling:4700 off Disable power management of all PCIe ports4701 force Forcibly enable power management of all PCIe ports4702 4703 pcie_pme= [PCIE,PM] Native PCIe PME signaling options:4704 nomsi Do not use MSI for native PCIe PME signaling (this makes4705 all PCIe root ports use INTx for all services).4706 4707 pcmv= [HW,PCMCIA] BadgePAD 44708 4709 pd_ignore_unused4710 [PM]4711 Keep all power-domains already enabled by bootloader on,4712 even if no driver has claimed them. This is useful4713 for debug and development, but should not be4714 needed on a platform with proper driver support.4715 4716 pdcchassis= [PARISC,HW] Disable/Enable PDC Chassis Status codes at4717 boot time.4718 Format: { 0 | 1 }4719 See arch/parisc/kernel/pdc_chassis.c4720 4721 percpu_alloc= [MM,EARLY]4722 Select which percpu first chunk allocator to use.4723 Currently supported values are "embed" and "page".4724 Archs may support subset or none of the selections.4725 See comments in mm/percpu.c for details on each4726 allocator. This parameter is primarily for debugging4727 and performance comparison.4728 4729 pirq= [SMP,APIC] Manual mp-table setup4730 See Documentation/arch/x86/i386/IO-APIC.rst.4731 4732 plip= [PPT,NET] Parallel port network link4733 Format: { parport<nr> | timid | 0 }4734 See also Documentation/admin-guide/parport.rst.4735 4736 pmtmr= [X86] Manual setup of pmtmr I/O Port.4737 Override pmtimer IOPort with a hex value.4738 e.g. pmtmr=0x5084739 4740 pmu_override= [PPC] Override the PMU.4741 This option takes over the PMU facility, so it is no4742 longer usable by perf. Setting this option starts the4743 PMU counters by setting MMCR0 to 0 (the FC bit is4744 cleared). If a number is given, then MMCR1 is set to4745 that number, otherwise (e.g., 'pmu_override=on'), MMCR14746 remains 0.4747 4748 pm_debug_messages [SUSPEND,KNL]4749 Enable suspend/resume debug messages during boot up.4750 4751 pnp.debug=1 [PNP]4752 Enable PNP debug messages (depends on the4753 CONFIG_PNP_DEBUG_MESSAGES option). Change at run-time4754 via /sys/module/pnp/parameters/debug. We always show4755 current resource usage; turning this on also shows4756 possible settings and some assignment information.4757 4758 pnpacpi= [ACPI]4759 { off }4760 4761 pnpbios= [ISAPNP]4762 { on | off | curr | res | no-curr | no-res }4763 4764 pnp_reserve_irq=4765 [ISAPNP] Exclude IRQs for the autoconfiguration4766 4767 pnp_reserve_dma=4768 [ISAPNP] Exclude DMAs for the autoconfiguration4769 4770 pnp_reserve_io= [ISAPNP] Exclude I/O ports for the autoconfiguration4771 Ranges are in pairs (I/O port base and size).4772 4773 pnp_reserve_mem=4774 [ISAPNP] Exclude memory regions for the4775 autoconfiguration.4776 Ranges are in pairs (memory base and size).4777 4778 ports= [IP_VS_FTP] IPVS ftp helper module4779 Default is 21.4780 Up to 8 (IP_VS_APP_MAX_PORTS) ports4781 may be specified.4782 Format: <port>,<port>....4783 4784 possible_cpus= [SMP,S390,X86]4785 Format: <unsigned int>4786 Set the number of possible CPUs, overriding the4787 regular discovery mechanisms (such as ACPI/FW, etc).4788 4789 powersave=off [PPC] This option disables power saving features.4790 It specifically disables cpuidle and sets the4791 platform machine description specific power_save4792 function to NULL. On Idle the CPU just reduces4793 execution priority.4794 4795 ppc_strict_facility_enable4796 [PPC,ENABLE] This option catches any kernel floating point,4797 Altivec, VSX and SPE outside of regions specifically4798 allowed (eg kernel_enable_fpu()/kernel_disable_fpu()).4799 There is some performance impact when enabling this.4800 4801 ppc_tm= [PPC,EARLY]4802 Format: {"off"}4803 Disable Hardware Transactional Memory4804 4805 preempt= [KNL]4806 Select preemption mode if you have CONFIG_PREEMPT_DYNAMIC4807 none - Limited to cond_resched() calls4808 voluntary - Limited to cond_resched() and might_sleep() calls4809 full - Any section that isn't explicitly preempt disabled4810 can be preempted anytime. Tasks will also yield4811 contended spinlocks (if the critical section isn't4812 explicitly preempt disabled beyond the lock itself).4813 4814 print-fatal-signals=4815 [KNL] debug: print fatal signals4816 4817 If enabled, warn about various signal handling4818 related application anomalies: too many signals,4819 too many POSIX.1 timers, fatal signals causing a4820 coredump - etc.4821 4822 If you hit the warning due to signal overflow,4823 you might want to try "ulimit -i unlimited".4824 4825 default: off.4826 4827 printk.always_kmsg_dump=4828 Trigger kmsg_dump for cases other than kernel oops or4829 panics4830 Format: <bool> (1/Y/y=enable, 0/N/n=disable)4831 default: disabled4832 4833 printk.console_no_auto_verbose=4834 Disable console loglevel raise on oops, panic4835 or lockdep-detected issues (only if lock debug is on).4836 With an exception to setups with low baudrate on4837 serial console, keeping this 0 is a good choice4838 in order to provide more debug information.4839 Format: <bool>4840 default: 0 (auto_verbose is enabled)4841 4842 printk.devkmsg={on,off,ratelimit}4843 Control writing to /dev/kmsg.4844 on - unlimited logging to /dev/kmsg from userspace4845 off - logging to /dev/kmsg disabled4846 ratelimit - ratelimit the logging4847 Default: ratelimit4848 4849 printk.time= Show timing data prefixed to each printk message line4850 Format: <bool> (1/Y/y=enable, 0/N/n=disable)4851 4852 proc_mem.force_override= [KNL]4853 Format: {always | ptrace | never}4854 Traditionally /proc/pid/mem allows memory permissions to be4855 overridden without restrictions. This option may be set to4856 restrict that. Can be one of:4857 - 'always': traditional behavior always allows mem overrides.4858 - 'ptrace': only allow mem overrides for active ptracers.4859 - 'never': never allow mem overrides.4860 If not specified, default is the CONFIG_PROC_MEM_* choice.4861 4862 processor.max_cstate= [HW,ACPI]4863 Limit processor to maximum C-state4864 max_cstate=9 overrides any DMI blacklist limit.4865 4866 processor.nocst [HW,ACPI]4867 Ignore the _CST method to determine C-states,4868 instead using the legacy FADT method4869 4870 profile= [KNL] Enable kernel profiling via /proc/profile4871 Format: [<profiletype>,]<number>4872 Param: <profiletype>: "schedule" or "kvm"4873 [defaults to kernel profiling]4874 Param: "schedule" - profile schedule points.4875 Param: "kvm" - profile VM exits.4876 Param: <number> - step/bucket size as a power of 2 for4877 statistical time based profiling.4878 4879 prompt_ramdisk= [RAM] [Deprecated]4880 4881 prot_virt= [S390] enable hosting protected virtual machines4882 isolated from the hypervisor (if hardware supports4883 that). If enabled, the default kernel base address4884 might be overridden even when Kernel Address Space4885 Layout Randomization is disabled.4886 Format: <bool>4887 4888 psi= [KNL] Enable or disable pressure stall information4889 tracking.4890 Format: <bool>4891 4892 psmouse.proto= [HW,MOUSE] Highest PS2 mouse protocol extension to4893 probe for; one of (bare|imps|exps|lifebook|any).4894 psmouse.rate= [HW,MOUSE] Set desired mouse report rate, in reports4895 per second.4896 psmouse.resetafter= [HW,MOUSE]4897 Try to reset the device after so many bad packets4898 (0 = never).4899 psmouse.resolution=4900 [HW,MOUSE] Set desired mouse resolution, in dpi.4901 psmouse.smartscroll=4902 [HW,MOUSE] Controls Logitech smartscroll autorepeat.4903 0 = disabled, 1 = enabled (default).4904 4905 pstore.backend= Specify the name of the pstore backend to use4906 4907 pti= [X86-64] Control Page Table Isolation of user and4908 kernel address spaces. Disabling this feature4909 removes hardening, but improves performance of4910 system calls and interrupts.4911 4912 on - unconditionally enable4913 off - unconditionally disable4914 auto - kernel detects whether your CPU model is4915 vulnerable to issues that PTI mitigates4916 4917 Not specifying this option is equivalent to pti=auto.4918 4919 pty.legacy_count=4920 [KNL] Number of legacy pty's. Overwrites compiled-in4921 default number.4922 4923 quiet [KNL,EARLY] Disable most log messages4924 4925 r128= [HW,DRM]4926 4927 radix_hcall_invalidate=on [PPC/PSERIES]4928 Disable RADIX GTSE feature and use hcall for TLB4929 invalidate.4930 4931 raid= [HW,RAID]4932 See Documentation/admin-guide/md.rst.4933 4934 ramdisk_size= [RAM] Sizes of RAM disks in kilobytes4935 See Documentation/admin-guide/blockdev/ramdisk.rst.4936 4937 ramdisk_start= [RAM] RAM disk image start address4938 4939 random.trust_cpu=off4940 [KNL,EARLY] Disable trusting the use of the CPU's4941 random number generator (if available) to4942 initialize the kernel's RNG.4943 4944 random.trust_bootloader=off4945 [KNL,EARLY] Disable trusting the use of the a seed4946 passed by the bootloader (if available) to4947 initialize the kernel's RNG.4948 4949 randomize_kstack_offset=4950 [KNL,EARLY] Enable or disable kernel stack offset4951 randomization, which provides roughly 5 bits of4952 entropy, frustrating memory corruption attacks4953 that depend on stack address determinism or4954 cross-syscall address exposures. This is only4955 available on architectures that have defined4956 CONFIG_HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET.4957 Format: <bool> (1/Y/y=enable, 0/N/n=disable)4958 Default is CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT.4959 4960 ras=option[,option,...] [KNL] RAS-specific options4961 4962 cec_disable [X86]4963 Disable the Correctable Errors Collector,4964 see CONFIG_RAS_CEC help text.4965 4966 rcu_nocbs[=cpu-list]4967 [KNL] The optional argument is a cpu list,4968 as described above.4969 4970 In kernels built with CONFIG_RCU_NOCB_CPU=y,4971 enable the no-callback CPU mode, which prevents4972 such CPUs' callbacks from being invoked in4973 softirq context. Invocation of such CPUs' RCU4974 callbacks will instead be offloaded to "rcuox/N"4975 kthreads created for that purpose, where "x" is4976 "p" for RCU-preempt, "s" for RCU-sched, and "g"4977 for the kthreads that mediate grace periods; and4978 "N" is the CPU number. This reduces OS jitter on4979 the offloaded CPUs, which can be useful for HPC4980 and real-time workloads. It can also improve4981 energy efficiency for asymmetric multiprocessors.4982 4983 If a cpulist is passed as an argument, the specified4984 list of CPUs is set to no-callback mode from boot.4985 4986 Otherwise, if the '=' sign and the cpulist4987 arguments are omitted, no CPU will be set to4988 no-callback mode from boot but the mode may be4989 toggled at runtime via cpusets.4990 4991 Note that this argument takes precedence over4992 the CONFIG_RCU_NOCB_CPU_DEFAULT_ALL option.4993 4994 rcu_nocb_poll [KNL]4995 Rather than requiring that offloaded CPUs4996 (specified by rcu_nocbs= above) explicitly4997 awaken the corresponding "rcuoN" kthreads,4998 make these kthreads poll for callbacks.4999 This improves the real-time response for the5000 offloaded CPUs by relieving them of the need to5001 wake up the corresponding kthread, but degrades5002 energy efficiency by requiring that the kthreads5003 periodically wake up to do the polling.5004 5005 rcutree.blimit= [KNL]5006 Set maximum number of finished RCU callbacks to5007 process in one batch.5008 5009 rcutree.csd_lock_suppress_rcu_stall= [KNL]5010 Do only a one-line RCU CPU stall warning when5011 there is an ongoing too-long CSD-lock wait.5012 5013 rcutree.do_rcu_barrier= [KNL]5014 Request a call to rcu_barrier(). This is5015 throttled so that userspace tests can safely5016 hammer on the sysfs variable if they so choose.5017 If triggered before the RCU grace-period machinery5018 is fully active, this will error out with EAGAIN.5019 5020 rcutree.dump_tree= [KNL]5021 Dump the structure of the rcu_node combining tree5022 out at early boot. This is used for diagnostic5023 purposes, to verify correct tree setup.5024 5025 rcutree.gp_cleanup_delay= [KNL]5026 Set the number of jiffies to delay each step of5027 RCU grace-period cleanup.5028 5029 rcutree.gp_init_delay= [KNL]5030 Set the number of jiffies to delay each step of5031 RCU grace-period initialization.5032 5033 rcutree.gp_preinit_delay= [KNL]5034 Set the number of jiffies to delay each step of5035 RCU grace-period pre-initialization, that is,5036 the propagation of recent CPU-hotplug changes up5037 the rcu_node combining tree.5038 5039 rcutree.jiffies_till_first_fqs= [KNL]5040 Set delay from grace-period initialization to5041 first attempt to force quiescent states.5042 Units are jiffies, minimum value is zero,5043 and maximum value is HZ.5044 5045 rcutree.jiffies_till_next_fqs= [KNL]5046 Set delay between subsequent attempts to force5047 quiescent states. Units are jiffies, minimum5048 value is one, and maximum value is HZ.5049 5050 rcutree.jiffies_till_sched_qs= [KNL]5051 Set required age in jiffies for a5052 given grace period before RCU starts5053 soliciting quiescent-state help from5054 rcu_note_context_switch() and cond_resched().5055 If not specified, the kernel will calculate5056 a value based on the most recent settings5057 of rcutree.jiffies_till_first_fqs5058 and rcutree.jiffies_till_next_fqs.5059 This calculated value may be viewed in5060 rcutree.jiffies_to_sched_qs. Any attempt to set5061 rcutree.jiffies_to_sched_qs will be cheerfully5062 overwritten.5063 5064 rcutree.kthread_prio= [KNL,BOOT]5065 Set the SCHED_FIFO priority of the RCU per-CPU5066 kthreads (rcuc/N). This value is also used for5067 the priority of the RCU boost threads (rcub/N)5068 and for the RCU grace-period kthreads (rcu_bh,5069 rcu_preempt, and rcu_sched). If RCU_BOOST is5070 set, valid values are 1-99 and the default is 15071 (the least-favored priority). Otherwise, when5072 RCU_BOOST is not set, valid values are 0-99 and5073 the default is zero (non-realtime operation).5074 When RCU_NOCB_CPU is set, also adjust the5075 priority of NOCB callback kthreads.5076 5077 rcutree.nocb_nobypass_lim_per_jiffy= [KNL]5078 On callback-offloaded (rcu_nocbs) CPUs,5079 RCU reduces the lock contention that would5080 otherwise be caused by callback floods through5081 use of the ->nocb_bypass list. However, in the5082 common non-flooded case, RCU queues directly to5083 the main ->cblist in order to avoid the extra5084 overhead of the ->nocb_bypass list and its lock.5085 But if there are too many callbacks queued during5086 a single jiffy, RCU pre-queues the callbacks into5087 the ->nocb_bypass queue. The definition of "too5088 many" is supplied by this kernel boot parameter.5089 5090 rcutree.nohz_full_patience_delay= [KNL]5091 On callback-offloaded (rcu_nocbs) CPUs, avoid5092 disturbing RCU unless the grace period has5093 reached the specified age in milliseconds.5094 Defaults to zero. Large values will be capped5095 at five seconds. All values will be rounded down5096 to the nearest value representable by jiffies.5097 5098 rcutree.qhimark= [KNL]5099 Set threshold of queued RCU callbacks beyond which5100 batch limiting is disabled.5101 5102 rcutree.qlowmark= [KNL]5103 Set threshold of queued RCU callbacks below which5104 batch limiting is re-enabled.5105 5106 rcutree.qovld= [KNL]5107 Set threshold of queued RCU callbacks beyond which5108 RCU's force-quiescent-state scan will aggressively5109 enlist help from cond_resched() and sched IPIs to5110 help CPUs more quickly reach quiescent states.5111 Set to less than zero to make this be set based5112 on rcutree.qhimark at boot time and to zero to5113 disable more aggressive help enlistment.5114 5115 rcutree.rcu_delay_page_cache_fill_msec= [KNL]5116 Set the page-cache refill delay (in milliseconds)5117 in response to low-memory conditions. The range5118 of permitted values is in the range 0:100000.5119 5120 rcutree.rcu_divisor= [KNL]5121 Set the shift-right count to use to compute5122 the callback-invocation batch limit bl from5123 the number of callbacks queued on this CPU.5124 The result will be bounded below by the value of5125 the rcutree.blimit kernel parameter. Every bl5126 callbacks, the softirq handler will exit in5127 order to allow the CPU to do other work.5128 5129 Please note that this callback-invocation batch5130 limit applies only to non-offloaded callback5131 invocation. Offloaded callbacks are instead5132 invoked in the context of an rcuoc kthread, which5133 scheduler will preempt as it does any other task.5134 5135 rcutree.rcu_fanout_exact= [KNL]5136 Disable autobalancing of the rcu_node combining5137 tree. This is used by rcutorture, and might5138 possibly be useful for architectures having high5139 cache-to-cache transfer latencies.5140 5141 rcutree.rcu_fanout_leaf= [KNL]5142 Change the number of CPUs assigned to each5143 leaf rcu_node structure. Useful for very5144 large systems, which will choose the value 64,5145 and for NUMA systems with large remote-access5146 latencies, which will choose a value aligned5147 with the appropriate hardware boundaries.5148 5149 rcutree.rcu_min_cached_objs= [KNL]5150 Minimum number of objects which are cached and5151 maintained per one CPU. Object size is equal5152 to PAGE_SIZE. The cache allows to reduce the5153 pressure to page allocator, also it makes the5154 whole algorithm to behave better in low memory5155 condition.5156 5157 rcutree.rcu_nocb_gp_stride= [KNL]5158 Set the number of NOCB callback kthreads in5159 each group, which defaults to the square root5160 of the number of CPUs. Larger numbers reduce5161 the wakeup overhead on the global grace-period5162 kthread, but increases that same overhead on5163 each group's NOCB grace-period kthread.5164 5165 rcutree.rcu_kick_kthreads= [KNL]5166 Cause the grace-period kthread to get an extra5167 wake_up() if it sleeps three times longer than5168 it should at force-quiescent-state time.5169 This wake_up() will be accompanied by a5170 WARN_ONCE() splat and an ftrace_dump().5171 5172 rcutree.rcu_resched_ns= [KNL]5173 Limit the time spend invoking a batch of RCU5174 callbacks to the specified number of nanoseconds.5175 By default, this limit is checked only once5176 every 32 callbacks in order to limit the pain5177 inflicted by local_clock() overhead.5178 5179 rcutree.rcu_unlock_delay= [KNL]5180 In CONFIG_RCU_STRICT_GRACE_PERIOD=y kernels,5181 this specifies an rcu_read_unlock()-time delay5182 in microseconds. This defaults to zero.5183 Larger delays increase the probability of5184 catching RCU pointer leaks, that is, buggy use5185 of RCU-protected pointers after the relevant5186 rcu_read_unlock() has completed.5187 5188 rcutree.sysrq_rcu= [KNL]5189 Commandeer a sysrq key to dump out Tree RCU's5190 rcu_node tree with an eye towards determining5191 why a new grace period has not yet started.5192 5193 rcutree.use_softirq= [KNL]5194 If set to zero, move all RCU_SOFTIRQ processing to5195 per-CPU rcuc kthreads. Defaults to a non-zero5196 value, meaning that RCU_SOFTIRQ is used by default.5197 Specify rcutree.use_softirq=0 to use rcuc kthreads.5198 5199 But note that CONFIG_PREEMPT_RT=y kernels disable5200 this kernel boot parameter, forcibly setting it5201 to zero.5202 5203 rcutree.enable_rcu_lazy= [KNL]5204 To save power, batch RCU callbacks and flush after5205 delay, memory pressure or callback list growing too5206 big.5207 5208 rcutree.rcu_normal_wake_from_gp= [KNL]5209 Reduces a latency of synchronize_rcu() call. This approach5210 maintains its own track of synchronize_rcu() callers, so it5211 does not interact with regular callbacks because it does not5212 use a call_rcu[_hurry]() path. Please note, this is for a5213 normal grace period.5214 5215 How to enable it:5216 5217 echo 1 > /sys/module/rcutree/parameters/rcu_normal_wake_from_gp5218 or pass a boot parameter "rcutree.rcu_normal_wake_from_gp=1"5219 5220 Default is 0.5221 5222 rcuscale.gp_async= [KNL]5223 Measure performance of asynchronous5224 grace-period primitives such as call_rcu().5225 5226 rcuscale.gp_async_max= [KNL]5227 Specify the maximum number of outstanding5228 callbacks per writer thread. When a writer5229 thread exceeds this limit, it invokes the5230 corresponding flavor of rcu_barrier() to allow5231 previously posted callbacks to drain.5232 5233 rcuscale.gp_exp= [KNL]5234 Measure performance of expedited synchronous5235 grace-period primitives.5236 5237 rcuscale.holdoff= [KNL]5238 Set test-start holdoff period. The purpose of5239 this parameter is to delay the start of the5240 test until boot completes in order to avoid5241 interference.5242 5243 rcuscale.kfree_by_call_rcu= [KNL]5244 In kernels built with CONFIG_RCU_LAZY=y, test5245 call_rcu() instead of kfree_rcu().5246 5247 rcuscale.kfree_mult= [KNL]5248 Instead of allocating an object of size kfree_obj,5249 allocate one of kfree_mult * sizeof(kfree_obj).5250 Defaults to 1.5251 5252 rcuscale.kfree_rcu_test= [KNL]5253 Set to measure performance of kfree_rcu() flooding.5254 5255 rcuscale.kfree_rcu_test_double= [KNL]5256 Test the double-argument variant of kfree_rcu().5257 If this parameter has the same value as5258 rcuscale.kfree_rcu_test_single, both the single-5259 and double-argument variants are tested.5260 5261 rcuscale.kfree_rcu_test_single= [KNL]5262 Test the single-argument variant of kfree_rcu().5263 If this parameter has the same value as5264 rcuscale.kfree_rcu_test_double, both the single-5265 and double-argument variants are tested.5266 5267 rcuscale.kfree_nthreads= [KNL]5268 The number of threads running loops of kfree_rcu().5269 5270 rcuscale.kfree_alloc_num= [KNL]5271 Number of allocations and frees done in an iteration.5272 5273 rcuscale.kfree_loops= [KNL]5274 Number of loops doing rcuscale.kfree_alloc_num number5275 of allocations and frees.5276 5277 rcuscale.minruntime= [KNL]5278 Set the minimum test run time in seconds. This5279 does not affect the data-collection interval,5280 but instead allows better measurement of things5281 like CPU consumption.5282 5283 rcuscale.nreaders= [KNL]5284 Set number of RCU readers. The value -1 selects5285 N, where N is the number of CPUs. A value5286 "n" less than -1 selects N-n+1, where N is again5287 the number of CPUs. For example, -2 selects N5288 (the number of CPUs), -3 selects N+1, and so on.5289 A value of "n" less than or equal to -N selects5290 a single reader.5291 5292 rcuscale.nwriters= [KNL]5293 Set number of RCU writers. The values operate5294 the same as for rcuscale.nreaders.5295 N, where N is the number of CPUs5296 5297 rcuscale.scale_type= [KNL]5298 Specify the RCU implementation to test.5299 5300 rcuscale.shutdown= [KNL]5301 Shut the system down after performance tests5302 complete. This is useful for hands-off automated5303 testing.5304 5305 rcuscale.verbose= [KNL]5306 Enable additional printk() statements.5307 5308 rcuscale.writer_holdoff= [KNL]5309 Write-side holdoff between grace periods,5310 in microseconds. The default of zero says5311 no holdoff.5312 5313 rcuscale.writer_holdoff_jiffies= [KNL]5314 Additional write-side holdoff between grace5315 periods, but in jiffies. The default of zero5316 says no holdoff.5317 5318 rcutorture.fqs_duration= [KNL]5319 Set duration of force_quiescent_state bursts5320 in microseconds.5321 5322 rcutorture.fqs_holdoff= [KNL]5323 Set holdoff time within force_quiescent_state bursts5324 in microseconds.5325 5326 rcutorture.fqs_stutter= [KNL]5327 Set wait time between force_quiescent_state bursts5328 in seconds.5329 5330 rcutorture.fwd_progress= [KNL]5331 Specifies the number of kthreads to be used5332 for RCU grace-period forward-progress testing5333 for the types of RCU supporting this notion.5334 Defaults to 1 kthread, values less than zero or5335 greater than the number of CPUs cause the number5336 of CPUs to be used.5337 5338 rcutorture.fwd_progress_div= [KNL]5339 Specify the fraction of a CPU-stall-warning5340 period to do tight-loop forward-progress testing.5341 5342 rcutorture.fwd_progress_holdoff= [KNL]5343 Number of seconds to wait between successive5344 forward-progress tests.5345 5346 rcutorture.fwd_progress_need_resched= [KNL]5347 Enclose cond_resched() calls within checks for5348 need_resched() during tight-loop forward-progress5349 testing.5350 5351 rcutorture.gp_cond= [KNL]5352 Use conditional/asynchronous update-side5353 primitives, if available.5354 5355 rcutorture.gp_exp= [KNL]5356 Use expedited update-side primitives, if available.5357 5358 rcutorture.gp_normal= [KNL]5359 Use normal (non-expedited) asynchronous5360 update-side primitives, if available.5361 5362 rcutorture.gp_sync= [KNL]5363 Use normal (non-expedited) synchronous5364 update-side primitives, if available. If all5365 of rcutorture.gp_cond=, rcutorture.gp_exp=,5366 rcutorture.gp_normal=, and rcutorture.gp_sync=5367 are zero, rcutorture acts as if is interpreted5368 they are all non-zero.5369 5370 rcutorture.irqreader= [KNL]5371 Run RCU readers from irq handlers, or, more5372 accurately, from a timer handler. Not all RCU5373 flavors take kindly to this sort of thing.5374 5375 rcutorture.leakpointer= [KNL]5376 Leak an RCU-protected pointer out of the reader.5377 This can of course result in splats, and is5378 intended to test the ability of things like5379 CONFIG_RCU_STRICT_GRACE_PERIOD=y to detect5380 such leaks.5381 5382 rcutorture.n_barrier_cbs= [KNL]5383 Set callbacks/threads for rcu_barrier() testing.5384 5385 rcutorture.nfakewriters= [KNL]5386 Set number of concurrent RCU writers. These just5387 stress RCU, they don't participate in the actual5388 test, hence the "fake".5389 5390 rcutorture.nocbs_nthreads= [KNL]5391 Set number of RCU callback-offload togglers.5392 Zero (the default) disables toggling.5393 5394 rcutorture.nocbs_toggle= [KNL]5395 Set the delay in milliseconds between successive5396 callback-offload toggling attempts.5397 5398 rcutorture.nreaders= [KNL]5399 Set number of RCU readers. The value -1 selects5400 N-1, where N is the number of CPUs. A value5401 "n" less than -1 selects N-n-2, where N is again5402 the number of CPUs. For example, -2 selects N5403 (the number of CPUs), -3 selects N+1, and so on.5404 5405 rcutorture.object_debug= [KNL]5406 Enable debug-object double-call_rcu() testing.5407 5408 rcutorture.onoff_holdoff= [KNL]5409 Set time (s) after boot for CPU-hotplug testing.5410 5411 rcutorture.onoff_interval= [KNL]5412 Set time (jiffies) between CPU-hotplug operations,5413 or zero to disable CPU-hotplug testing.5414 5415 rcutorture.read_exit= [KNL]5416 Set the number of read-then-exit kthreads used5417 to test the interaction of RCU updaters and5418 task-exit processing.5419 5420 rcutorture.read_exit_burst= [KNL]5421 The number of times in a given read-then-exit5422 episode that a set of read-then-exit kthreads5423 is spawned.5424 5425 rcutorture.read_exit_delay= [KNL]5426 The delay, in seconds, between successive5427 read-then-exit testing episodes.5428 5429 rcutorture.shuffle_interval= [KNL]5430 Set task-shuffle interval (s). Shuffling tasks5431 allows some CPUs to go into dyntick-idle mode5432 during the rcutorture test.5433 5434 rcutorture.shutdown_secs= [KNL]5435 Set time (s) after boot system shutdown. This5436 is useful for hands-off automated testing.5437 5438 rcutorture.stall_cpu= [KNL]5439 Duration of CPU stall (s) to test RCU CPU stall5440 warnings, zero to disable.5441 5442 rcutorture.stall_cpu_block= [KNL]5443 Sleep while stalling if set. This will result5444 in warnings from preemptible RCU in addition to5445 any other stall-related activity. Note that5446 in kernels built with CONFIG_PREEMPTION=n and5447 CONFIG_PREEMPT_COUNT=y, this parameter will5448 cause the CPU to pass through a quiescent state.5449 Given CONFIG_PREEMPTION=n, this will suppress5450 RCU CPU stall warnings, but will instead result5451 in scheduling-while-atomic splats.5452 5453 Use of this module parameter results in splats.5454 5455 5456 rcutorture.stall_cpu_holdoff= [KNL]5457 Time to wait (s) after boot before inducing stall.5458 5459 rcutorture.stall_cpu_irqsoff= [KNL]5460 Disable interrupts while stalling if set, but only5461 on the first stall in the set.5462 5463 rcutorture.stall_cpu_repeat= [KNL]5464 Number of times to repeat the stall sequence,5465 so that rcutorture.stall_cpu_repeat=3 will result5466 in four stall sequences.5467 5468 rcutorture.stall_gp_kthread= [KNL]5469 Duration (s) of forced sleep within RCU5470 grace-period kthread to test RCU CPU stall5471 warnings, zero to disable. If both stall_cpu5472 and stall_gp_kthread are specified, the5473 kthread is starved first, then the CPU.5474 5475 rcutorture.stat_interval= [KNL]5476 Time (s) between statistics printk()s.5477 5478 rcutorture.stutter= [KNL]5479 Time (s) to stutter testing, for example, specifying5480 five seconds causes the test to run for five seconds,5481 wait for five seconds, and so on. This tests RCU's5482 ability to transition abruptly to and from idle.5483 5484 rcutorture.test_boost= [KNL]5485 Test RCU priority boosting? 0=no, 1=maybe, 2=yes.5486 "Maybe" means test if the RCU implementation5487 under test support RCU priority boosting.5488 5489 rcutorture.test_boost_duration= [KNL]5490 Duration (s) of each individual boost test.5491 5492 rcutorture.test_boost_interval= [KNL]5493 Interval (s) between each boost test.5494 5495 rcutorture.test_no_idle_hz= [KNL]5496 Test RCU's dyntick-idle handling. See also the5497 rcutorture.shuffle_interval parameter.5498 5499 rcutorture.torture_type= [KNL]5500 Specify the RCU implementation to test.5501 5502 rcutorture.verbose= [KNL]5503 Enable additional printk() statements.5504 5505 rcupdate.rcu_cpu_stall_ftrace_dump= [KNL]5506 Dump ftrace buffer after reporting RCU CPU5507 stall warning.5508 5509 rcupdate.rcu_cpu_stall_notifiers= [KNL]5510 Provide RCU CPU stall notifiers, but see the5511 warnings in the RCU_CPU_STALL_NOTIFIER Kconfig5512 option's help text. TL;DR: You almost certainly5513 do not want rcupdate.rcu_cpu_stall_notifiers.5514 5515 rcupdate.rcu_cpu_stall_suppress= [KNL]5516 Suppress RCU CPU stall warning messages.5517 5518 rcupdate.rcu_cpu_stall_suppress_at_boot= [KNL]5519 Suppress RCU CPU stall warning messages and5520 rcutorture writer stall warnings that occur5521 during early boot, that is, during the time5522 before the init task is spawned.5523 5524 rcupdate.rcu_cpu_stall_timeout= [KNL]5525 Set timeout for RCU CPU stall warning messages.5526 The value is in seconds and the maximum allowed5527 value is 300 seconds.5528 5529 rcupdate.rcu_exp_cpu_stall_timeout= [KNL]5530 Set timeout for expedited RCU CPU stall warning5531 messages. The value is in milliseconds5532 and the maximum allowed value is 210005533 milliseconds. Please note that this value is5534 adjusted to an arch timer tick resolution.5535 Setting this to zero causes the value from5536 rcupdate.rcu_cpu_stall_timeout to be used (after5537 conversion from seconds to milliseconds).5538 5539 rcupdate.rcu_cpu_stall_cputime= [KNL]5540 Provide statistics on the cputime and count of5541 interrupts and tasks during the sampling period. For5542 multiple continuous RCU stalls, all sampling periods5543 begin at half of the first RCU stall timeout.5544 5545 rcupdate.rcu_exp_stall_task_details= [KNL]5546 Print stack dumps of any tasks blocking the5547 current expedited RCU grace period during an5548 expedited RCU CPU stall warning.5549 5550 rcupdate.rcu_expedited= [KNL]5551 Use expedited grace-period primitives, for5552 example, synchronize_rcu_expedited() instead5553 of synchronize_rcu(). This reduces latency,5554 but can increase CPU utilization, degrade5555 real-time latency, and degrade energy efficiency.5556 No effect on CONFIG_TINY_RCU kernels.5557 5558 rcupdate.rcu_normal= [KNL]5559 Use only normal grace-period primitives,5560 for example, synchronize_rcu() instead of5561 synchronize_rcu_expedited(). This improves5562 real-time latency, CPU utilization, and5563 energy efficiency, but can expose users to5564 increased grace-period latency. This parameter5565 overrides rcupdate.rcu_expedited. No effect on5566 CONFIG_TINY_RCU kernels.5567 5568 rcupdate.rcu_normal_after_boot= [KNL]5569 Once boot has completed (that is, after5570 rcu_end_inkernel_boot() has been invoked), use5571 only normal grace-period primitives. No effect5572 on CONFIG_TINY_RCU kernels.5573 5574 But note that CONFIG_PREEMPT_RT=y kernels enables5575 this kernel boot parameter, forcibly setting5576 it to the value one, that is, converting any5577 post-boot attempt at an expedited RCU grace5578 period to instead use normal non-expedited5579 grace-period processing.5580 5581 rcupdate.rcu_task_collapse_lim= [KNL]5582 Set the maximum number of callbacks present5583 at the beginning of a grace period that allows5584 the RCU Tasks flavors to collapse back to using5585 a single callback queue. This switching only5586 occurs when rcupdate.rcu_task_enqueue_lim is5587 set to the default value of -1.5588 5589 rcupdate.rcu_task_contend_lim= [KNL]5590 Set the minimum number of callback-queuing-time5591 lock-contention events per jiffy required to5592 cause the RCU Tasks flavors to switch to per-CPU5593 callback queuing. This switching only occurs5594 when rcupdate.rcu_task_enqueue_lim is set to5595 the default value of -1.5596 5597 rcupdate.rcu_task_enqueue_lim= [KNL]5598 Set the number of callback queues to use for the5599 RCU Tasks family of RCU flavors. The default5600 of -1 allows this to be automatically (and5601 dynamically) adjusted. This parameter is intended5602 for use in testing.5603 5604 rcupdate.rcu_task_ipi_delay= [KNL]5605 Set time in jiffies during which RCU tasks will5606 avoid sending IPIs, starting with the beginning5607 of a given grace period. Setting a large5608 number avoids disturbing real-time workloads,5609 but lengthens grace periods.5610 5611 rcupdate.rcu_task_lazy_lim= [KNL]5612 Number of callbacks on a given CPU that will5613 cancel laziness on that CPU. Use -1 to disable5614 cancellation of laziness, but be advised that5615 doing so increases the danger of OOM due to5616 callback flooding.5617 5618 rcupdate.rcu_task_stall_info= [KNL]5619 Set initial timeout in jiffies for RCU task stall5620 informational messages, which give some indication5621 of the problem for those not patient enough to5622 wait for ten minutes. Informational messages are5623 only printed prior to the stall-warning message5624 for a given grace period. Disable with a value5625 less than or equal to zero. Defaults to ten5626 seconds. A change in value does not take effect5627 until the beginning of the next grace period.5628 5629 rcupdate.rcu_task_stall_info_mult= [KNL]5630 Multiplier for time interval between successive5631 RCU task stall informational messages for a given5632 RCU tasks grace period. This value is clamped5633 to one through ten, inclusive. It defaults to5634 the value three, so that the first informational5635 message is printed 10 seconds into the grace5636 period, the second at 40 seconds, the third at5637 160 seconds, and then the stall warning at 6005638 seconds would prevent a fourth at 640 seconds.5639 5640 rcupdate.rcu_task_stall_timeout= [KNL]5641 Set timeout in jiffies for RCU task stall5642 warning messages. Disable with a value less5643 than or equal to zero. Defaults to ten minutes.5644 A change in value does not take effect until5645 the beginning of the next grace period.5646 5647 rcupdate.rcu_tasks_lazy_ms= [KNL]5648 Set timeout in milliseconds RCU Tasks asynchronous5649 callback batching for call_rcu_tasks().5650 A negative value will take the default. A value5651 of zero will disable batching. Batching is5652 always disabled for synchronize_rcu_tasks().5653 5654 rcupdate.rcu_tasks_trace_lazy_ms= [KNL]5655 Set timeout in milliseconds RCU Tasks5656 Trace asynchronous callback batching for5657 call_rcu_tasks_trace(). A negative value5658 will take the default. A value of zero will5659 disable batching. Batching is always disabled5660 for synchronize_rcu_tasks_trace().5661 5662 rcupdate.rcu_self_test= [KNL]5663 Run the RCU early boot self tests5664 5665 rdinit= [KNL]5666 Format: <full_path>5667 Run specified binary instead of /init from the ramdisk,5668 used for early userspace startup. See initrd.5669 5670 rdrand= [X86,EARLY]5671 force - Override the decision by the kernel to hide the5672 advertisement of RDRAND support (this affects5673 certain AMD processors because of buggy BIOS5674 support, specifically around the suspend/resume5675 path).5676 5677 rdt= [HW,X86,RDT]5678 Turn on/off individual RDT features. List is:5679 cmt, mbmtotal, mbmlocal, l3cat, l3cdp, l2cat, l2cdp,5680 mba, smba, bmec.5681 E.g. to turn on cmt and turn off mba use:5682 rdt=cmt,!mba5683 5684 reboot= [KNL]5685 Format (x86 or x86_64):5686 [w[arm] | c[old] | h[ard] | s[oft] | g[pio]] | d[efault] \5687 [[,]s[mp]#### \5688 [[,]b[ios] | a[cpi] | k[bd] | t[riple] | e[fi] | p[ci]] \5689 [[,]f[orce]5690 Where reboot_mode is one of warm (soft) or cold (hard) or gpio5691 (prefix with 'panic_' to set mode for panic5692 reboot only),5693 reboot_type is one of bios, acpi, kbd, triple, efi, or pci,5694 reboot_force is either force or not specified,5695 reboot_cpu is s[mp]#### with #### being the processor5696 to be used for rebooting.5697 5698 refscale.holdoff= [KNL]5699 Set test-start holdoff period. The purpose of5700 this parameter is to delay the start of the5701 test until boot completes in order to avoid5702 interference.5703 5704 refscale.lookup_instances= [KNL]5705 Number of data elements to use for the forms of5706 SLAB_TYPESAFE_BY_RCU testing. A negative number5707 is negated and multiplied by nr_cpu_ids, while5708 zero specifies nr_cpu_ids.5709 5710 refscale.loops= [KNL]5711 Set the number of loops over the synchronization5712 primitive under test. Increasing this number5713 reduces noise due to loop start/end overhead,5714 but the default has already reduced the per-pass5715 noise to a handful of picoseconds on ca. 20205716 x86 laptops.5717 5718 refscale.nreaders= [KNL]5719 Set number of readers. The default value of -15720 selects N, where N is roughly 75% of the number5721 of CPUs. A value of zero is an interesting choice.5722 5723 refscale.nruns= [KNL]5724 Set number of runs, each of which is dumped onto5725 the console log.5726 5727 refscale.readdelay= [KNL]5728 Set the read-side critical-section duration,5729 measured in microseconds.5730 5731 refscale.scale_type= [KNL]5732 Specify the read-protection implementation to test.5733 5734 refscale.shutdown= [KNL]5735 Shut down the system at the end of the performance5736 test. This defaults to 1 (shut it down) when5737 refscale is built into the kernel and to 0 (leave5738 it running) when refscale is built as a module.5739 5740 refscale.verbose= [KNL]5741 Enable additional printk() statements.5742 5743 refscale.verbose_batched= [KNL]5744 Batch the additional printk() statements. If zero5745 (the default) or negative, print everything. Otherwise,5746 print every Nth verbose statement, where N is the value5747 specified.5748 5749 regulator_ignore_unused5750 [REGULATOR]5751 Prevents regulator framework from disabling regulators5752 that are unused, due no driver claiming them. This may5753 be useful for debug and development, but should not be5754 needed on a platform with proper driver support.5755 5756 relax_domain_level=5757 [KNL, SMP] Set scheduler's default relax_domain_level.5758 See Documentation/admin-guide/cgroup-v1/cpusets.rst.5759 5760 reserve= [KNL,BUGS] Force kernel to ignore I/O ports or memory5761 Format: <base1>,<size1>[,<base2>,<size2>,...]5762 Reserve I/O ports or memory so the kernel won't use5763 them. If <base> is less than 0x10000, the region5764 is assumed to be I/O ports; otherwise it is memory.5765 5766 reserve_mem= [RAM]5767 Format: nn[KNG]:<align>:<label>5768 Reserve physical memory and label it with a name that5769 other subsystems can use to access it. This is typically5770 used for systems that do not wipe the RAM, and this command5771 line will try to reserve the same physical memory on5772 soft reboots. Note, it is not guaranteed to be the same5773 location. For example, if anything about the system changes5774 or if booting a different kernel. It can also fail if KASLR5775 places the kernel at the location of where the RAM reservation5776 was from a previous boot, the new reservation will be at a5777 different location.5778 Any subsystem using this feature must add a way to verify5779 that the contents of the physical memory is from a previous5780 boot, as there may be cases where the memory will not be5781 located at the same location.5782 5783 The format is size:align:label for example, to request5784 12 megabytes of 4096 alignment for ramoops:5785 5786 reserve_mem=12M:4096:oops ramoops.mem_name=oops5787 5788 reservetop= [X86-32,EARLY]5789 Format: nn[KMG]5790 Reserves a hole at the top of the kernel virtual5791 address space.5792 5793 reset_devices [KNL] Force drivers to reset the underlying device5794 during initialization.5795 5796 resume= [SWSUSP]5797 Specify the partition device for software suspend5798 Format:5799 {/dev/<dev> | PARTUUID=<uuid> | <int>:<int> | <hex>}5800 5801 resume_offset= [SWSUSP]5802 Specify the offset from the beginning of the partition5803 given by "resume=" at which the swap header is located,5804 in <PAGE_SIZE> units (needed only for swap files).5805 See Documentation/power/swsusp-and-swap-files.rst5806 5807 resumedelay= [HIBERNATION] Delay (in seconds) to pause before attempting to5808 read the resume files5809 5810 resumewait [HIBERNATION] Wait (indefinitely) for resume device to show up.5811 Useful for devices that are detected asynchronously5812 (e.g. USB and MMC devices).5813 5814 retain_initrd [RAM] Keep initrd memory after extraction. After boot, it will5815 be accessible via /sys/firmware/initrd.5816 5817 retbleed= [X86] Control mitigation of RETBleed (Arbitrary5818 Speculative Code Execution with Return Instructions)5819 vulnerability.5820 5821 AMD-based UNRET and IBPB mitigations alone do not stop5822 sibling threads from influencing the predictions of other5823 sibling threads. For that reason, STIBP is used on pro-5824 cessors that support it, and mitigate SMT on processors5825 that don't.5826 5827 off - no mitigation5828 auto - automatically select a migitation5829 auto,nosmt - automatically select a mitigation,5830 disabling SMT if necessary for5831 the full mitigation (only on Zen15832 and older without STIBP).5833 ibpb - On AMD, mitigate short speculation5834 windows on basic block boundaries too.5835 Safe, highest perf impact. It also5836 enables STIBP if present. Not suitable5837 on Intel.5838 ibpb,nosmt - Like "ibpb" above but will disable SMT5839 when STIBP is not available. This is5840 the alternative for systems which do not5841 have STIBP.5842 unret - Force enable untrained return thunks,5843 only effective on AMD f15h-f17h based5844 systems.5845 unret,nosmt - Like unret, but will disable SMT when STIBP5846 is not available. This is the alternative for5847 systems which do not have STIBP.5848 5849 Selecting 'auto' will choose a mitigation method at run5850 time according to the CPU.5851 5852 Not specifying this option is equivalent to retbleed=auto.5853 5854 rfkill.default_state=5855 0 "airplane mode". All wifi, bluetooth, wimax, gps, fm,5856 etc. communication is blocked by default.5857 1 Unblocked.5858 5859 rfkill.master_switch_mode=5860 0 The "airplane mode" button does nothing.5861 1 The "airplane mode" button toggles between everything5862 blocked and the previous configuration.5863 2 The "airplane mode" button toggles between everything5864 blocked and everything unblocked.5865 5866 ring3mwait=disable5867 [KNL] Disable ring 3 MONITOR/MWAIT feature on supported5868 CPUs.5869 5870 riscv_isa_fallback [RISCV,EARLY]5871 When CONFIG_RISCV_ISA_FALLBACK is not enabled, permit5872 falling back to detecting extension support by parsing5873 "riscv,isa" property on devicetree systems when the5874 replacement properties are not found. See the Kconfig5875 entry for RISCV_ISA_FALLBACK.5876 5877 ro [KNL] Mount root device read-only on boot5878 5879 rodata= [KNL,EARLY]5880 on Mark read-only kernel memory as read-only (default).5881 off Leave read-only kernel memory writable for debugging.5882 full Mark read-only kernel memory and aliases as read-only5883 [arm64]5884 5885 rockchip.usb_uart5886 [EARLY]5887 Enable the uart passthrough on the designated usb port5888 on Rockchip SoCs. When active, the signals of the5889 debug-uart get routed to the D+ and D- pins of the usb5890 port and the regular usb controller gets disabled.5891 5892 root= [KNL] Root filesystem5893 Usually this a a block device specifier of some kind,5894 see the early_lookup_bdev comment in5895 block/early-lookup.c for details.5896 Alternatively this can be "ram" for the legacy initial5897 ramdisk, "nfs" and "cifs" for root on a network file5898 system, or "mtd" and "ubi" for mounting from raw flash.5899 5900 rootdelay= [KNL] Delay (in seconds) to pause before attempting to5901 mount the root filesystem5902 5903 rootflags= [KNL] Set root filesystem mount option string5904 5905 rootfstype= [KNL] Set root filesystem type5906 5907 rootwait [KNL] Wait (indefinitely) for root device to show up.5908 Useful for devices that are detected asynchronously5909 (e.g. USB and MMC devices).5910 5911 rootwait= [KNL] Maximum time (in seconds) to wait for root device5912 to show up before attempting to mount the root5913 filesystem.5914 5915 rproc_mem=nn[KMG][@address]5916 [KNL,ARM,CMA] Remoteproc physical memory block.5917 Memory area to be used by remote processor image,5918 managed by CMA.5919 5920 rw [KNL] Mount root device read-write on boot5921 5922 S [KNL] Run init in single mode5923 5924 s390_iommu= [HW,S390]5925 Set s390 IOTLB flushing mode5926 strict5927 With strict flushing every unmap operation will result5928 in an IOTLB flush. Default is lazy flushing before5929 reuse, which is faster. Deprecated, equivalent to5930 iommu.strict=1.5931 5932 s390_iommu_aperture= [KNL,S390]5933 Specifies the size of the per device DMA address space5934 accessible through the DMA and IOMMU APIs as a decimal5935 factor of the size of main memory.5936 The default is 1 meaning that one can concurrently use5937 as many DMA addresses as physical memory is installed,5938 if supported by hardware, and thus map all of memory5939 once. With a value of 2 one can map all of memory twice5940 and so on. As a special case a factor of 0 imposes no5941 restrictions other than those given by hardware at the5942 cost of significant additional memory use for tables.5943 5944 sa1100ir [NET]5945 See drivers/net/irda/sa1100_ir.c.5946 5947 sched_verbose [KNL,EARLY] Enables verbose scheduler debug messages.5948 5949 schedstats= [KNL,X86] Enable or disable scheduled statistics.5950 Allowed values are enable and disable. This feature5951 incurs a small amount of overhead in the scheduler5952 but is useful for debugging and performance tuning.5953 5954 sched_thermal_decay_shift=5955 [Deprecated]5956 [KNL, SMP] Set a decay shift for scheduler thermal5957 pressure signal. Thermal pressure signal follows the5958 default decay period of other scheduler pelt5959 signals(usually 32 ms but configurable). Setting5960 sched_thermal_decay_shift will left shift the decay5961 period for the thermal pressure signal by the shift5962 value.5963 i.e. with the default pelt decay period of 32 ms5964 sched_thermal_decay_shift thermal pressure decay pr5965 1 64 ms5966 2 128 ms5967 and so on.5968 Format: integer between 0 and 105969 Default is 0.5970 5971 scftorture.holdoff= [KNL]5972 Number of seconds to hold off before starting5973 test. Defaults to zero for module insertion and5974 to 10 seconds for built-in smp_call_function()5975 tests.5976 5977 scftorture.longwait= [KNL]5978 Request ridiculously long waits randomly selected5979 up to the chosen limit in seconds. Zero (the5980 default) disables this feature. Please note5981 that requesting even small non-zero numbers of5982 seconds can result in RCU CPU stall warnings,5983 softlockup complaints, and so on.5984 5985 scftorture.nthreads= [KNL]5986 Number of kthreads to spawn to invoke the5987 smp_call_function() family of functions.5988 The default of -1 specifies a number of kthreads5989 equal to the number of CPUs.5990 5991 scftorture.onoff_holdoff= [KNL]5992 Number seconds to wait after the start of the5993 test before initiating CPU-hotplug operations.5994 5995 scftorture.onoff_interval= [KNL]5996 Number seconds to wait between successive5997 CPU-hotplug operations. Specifying zero (which5998 is the default) disables CPU-hotplug operations.5999 6000 scftorture.shutdown_secs= [KNL]6001 The number of seconds following the start of the6002 test after which to shut down the system. The6003 default of zero avoids shutting down the system.6004 Non-zero values are useful for automated tests.6005 6006 scftorture.stat_interval= [KNL]6007 The number of seconds between outputting the6008 current test statistics to the console. A value6009 of zero disables statistics output.6010 6011 scftorture.stutter_cpus= [KNL]6012 The number of jiffies to wait between each change6013 to the set of CPUs under test.6014 6015 scftorture.use_cpus_read_lock= [KNL]6016 Use use_cpus_read_lock() instead of the default6017 preempt_disable() to disable CPU hotplug6018 while invoking one of the smp_call_function*()6019 functions.6020 6021 scftorture.verbose= [KNL]6022 Enable additional printk() statements.6023 6024 scftorture.weight_single= [KNL]6025 The probability weighting to use for the6026 smp_call_function_single() function with a zero6027 "wait" parameter. A value of -1 selects the6028 default if all other weights are -1. However,6029 if at least one weight has some other value, a6030 value of -1 will instead select a weight of zero.6031 6032 scftorture.weight_single_wait= [KNL]6033 The probability weighting to use for the6034 smp_call_function_single() function with a6035 non-zero "wait" parameter. See weight_single.6036 6037 scftorture.weight_many= [KNL]6038 The probability weighting to use for the6039 smp_call_function_many() function with a zero6040 "wait" parameter. See weight_single.6041 Note well that setting a high probability for6042 this weighting can place serious IPI load6043 on the system.6044 6045 scftorture.weight_many_wait= [KNL]6046 The probability weighting to use for the6047 smp_call_function_many() function with a6048 non-zero "wait" parameter. See weight_single6049 and weight_many.6050 6051 scftorture.weight_all= [KNL]6052 The probability weighting to use for the6053 smp_call_function_all() function with a zero6054 "wait" parameter. See weight_single and6055 weight_many.6056 6057 scftorture.weight_all_wait= [KNL]6058 The probability weighting to use for the6059 smp_call_function_all() function with a6060 non-zero "wait" parameter. See weight_single6061 and weight_many.6062 6063 skew_tick= [KNL,EARLY] Offset the periodic timer tick per cpu to mitigate6064 xtime_lock contention on larger systems, and/or RCU lock6065 contention on all systems with CONFIG_MAXSMP set.6066 Format: { "0" | "1" }6067 0 -- disable. (may be 1 via CONFIG_CMDLINE="skew_tick=1"6068 1 -- enable.6069 Note: increases power consumption, thus should only be6070 enabled if running jitter sensitive (HPC/RT) workloads.6071 6072 security= [SECURITY] Choose a legacy "major" security module to6073 enable at boot. This has been deprecated by the6074 "lsm=" parameter.6075 6076 selinux= [SELINUX] Disable or enable SELinux at boot time.6077 Format: { "0" | "1" }6078 See security/selinux/Kconfig help text.6079 0 -- disable.6080 1 -- enable.6081 Default value is 1.6082 6083 serialnumber [BUGS=X86-32]6084 6085 sev=option[,option...] [X86-64] See Documentation/arch/x86/x86_64/boot-options.rst6086 6087 shapers= [NET]6088 Maximal number of shapers.6089 6090 show_lapic= [APIC,X86] Advanced Programmable Interrupt Controller6091 Limit apic dumping. The parameter defines the maximal6092 number of local apics being dumped. Also it is possible6093 to set it to "all" by meaning -- no limit here.6094 Format: { 1 (default) | 2 | ... | all }.6095 The parameter valid if only apic=debug or6096 apic=verbose is specified.6097 Example: apic=debug show_lapic=all6098 6099 slab_debug[=options[,slabs][;[options[,slabs]]...] [MM]6100 Enabling slab_debug allows one to determine the6101 culprit if slab objects become corrupted. Enabling6102 slab_debug can create guard zones around objects and6103 may poison objects when not in use. Also tracks the6104 last alloc / free. For more information see6105 Documentation/mm/slub.rst.6106 (slub_debug legacy name also accepted for now)6107 6108 slab_max_order= [MM]6109 Determines the maximum allowed order for slabs.6110 A high setting may cause OOMs due to memory6111 fragmentation. For more information see6112 Documentation/mm/slub.rst.6113 (slub_max_order legacy name also accepted for now)6114 6115 slab_merge [MM]6116 Enable merging of slabs with similar size when the6117 kernel is built without CONFIG_SLAB_MERGE_DEFAULT.6118 (slub_merge legacy name also accepted for now)6119 6120 slab_min_objects= [MM]6121 The minimum number of objects per slab. SLUB will6122 increase the slab order up to slab_max_order to6123 generate a sufficiently large slab able to contain6124 the number of objects indicated. The higher the number6125 of objects the smaller the overhead of tracking slabs6126 and the less frequently locks need to be acquired.6127 For more information see Documentation/mm/slub.rst.6128 (slub_min_objects legacy name also accepted for now)6129 6130 slab_min_order= [MM]6131 Determines the minimum page order for slabs. Must be6132 lower or equal to slab_max_order. For more information see6133 Documentation/mm/slub.rst.6134 (slub_min_order legacy name also accepted for now)6135 6136 slab_nomerge [MM]6137 Disable merging of slabs with similar size. May be6138 necessary if there is some reason to distinguish6139 allocs to different slabs, especially in hardened6140 environments where the risk of heap overflows and6141 layout control by attackers can usually be6142 frustrated by disabling merging. This will reduce6143 most of the exposure of a heap attack to a single6144 cache (risks via metadata attacks are mostly6145 unchanged). Debug options disable merging on their6146 own.6147 For more information see Documentation/mm/slub.rst.6148 (slub_nomerge legacy name also accepted for now)6149 6150 slram= [HW,MTD]6151 6152 smart2= [HW]6153 Format: <io1>[,<io2>[,...,<io8>]]6154 6155 smp.csd_lock_timeout= [KNL]6156 Specify the period of time in milliseconds6157 that smp_call_function() and friends will wait6158 for a CPU to release the CSD lock. This is6159 useful when diagnosing bugs involving CPUs6160 disabling interrupts for extended periods6161 of time. Defaults to 5,000 milliseconds, and6162 setting a value of zero disables this feature.6163 This feature may be more efficiently disabled6164 using the csdlock_debug- kernel parameter.6165 6166 smp.panic_on_ipistall= [KNL]6167 If a csd_lock_timeout extends for more than6168 the specified number of milliseconds, panic the6169 system. By default, let CSD-lock acquisition6170 take as long as they take. Specifying 300,0006171 for this value provides a 5-minute timeout.6172 6173 smsc-ircc2.nopnp [HW] Don't use PNP to discover SMC devices6174 smsc-ircc2.ircc_cfg= [HW] Device configuration I/O port6175 smsc-ircc2.ircc_sir= [HW] SIR base I/O port6176 smsc-ircc2.ircc_fir= [HW] FIR base I/O port6177 smsc-ircc2.ircc_irq= [HW] IRQ line6178 smsc-ircc2.ircc_dma= [HW] DMA channel6179 smsc-ircc2.ircc_transceiver= [HW] Transceiver type:6180 0: Toshiba Satellite 1800 (GP data pin select)6181 1: Fast pin select (default)6182 2: ATC IRMode6183 6184 smt= [KNL,MIPS,S390,EARLY] Set the maximum number of threads6185 (logical CPUs) to use per physical CPU on systems6186 capable of symmetric multithreading (SMT). Will6187 be capped to the actual hardware limit.6188 Format: <integer>6189 Default: -1 (no limit)6190 6191 softlockup_panic=6192 [KNL] Should the soft-lockup detector generate panics.6193 Format: 0 | 16194 6195 A value of 1 instructs the soft-lockup detector6196 to panic the machine when a soft-lockup occurs. It is6197 also controlled by the kernel.softlockup_panic sysctl6198 and CONFIG_BOOTPARAM_SOFTLOCKUP_PANIC, which is the6199 respective build-time switch to that functionality.6200 6201 softlockup_all_cpu_backtrace=6202 [KNL] Should the soft-lockup detector generate6203 backtraces on all cpus.6204 Format: 0 | 16205 6206 sonypi.*= [HW] Sony Programmable I/O Control Device driver6207 See Documentation/admin-guide/laptops/sonypi.rst6208 6209 spectre_bhi= [X86] Control mitigation of Branch History Injection6210 (BHI) vulnerability. This setting affects the6211 deployment of the HW BHI control and the SW BHB6212 clearing sequence.6213 6214 on - (default) Enable the HW or SW mitigation as6215 needed. This protects the kernel from6216 both syscalls and VMs.6217 vmexit - On systems which don't have the HW mitigation6218 available, enable the SW mitigation on vmexit6219 ONLY. On such systems, the host kernel is6220 protected from VM-originated BHI attacks, but6221 may still be vulnerable to syscall attacks.6222 off - Disable the mitigation.6223 6224 spectre_v2= [X86,EARLY] Control mitigation of Spectre variant 26225 (indirect branch speculation) vulnerability.6226 The default operation protects the kernel from6227 user space attacks.6228 6229 on - unconditionally enable, implies6230 spectre_v2_user=on6231 off - unconditionally disable, implies6232 spectre_v2_user=off6233 auto - kernel detects whether your CPU model is6234 vulnerable6235 6236 Selecting 'on' will, and 'auto' may, choose a6237 mitigation method at run time according to the6238 CPU, the available microcode, the setting of the6239 CONFIG_MITIGATION_RETPOLINE configuration option,6240 and the compiler with which the kernel was built.6241 6242 Selecting 'on' will also enable the mitigation6243 against user space to user space task attacks.6244 6245 Selecting 'off' will disable both the kernel and6246 the user space protections.6247 6248 Specific mitigations can also be selected manually:6249 6250 retpoline - replace indirect branches6251 retpoline,generic - Retpolines6252 retpoline,lfence - LFENCE; indirect branch6253 retpoline,amd - alias for retpoline,lfence6254 eibrs - Enhanced/Auto IBRS6255 eibrs,retpoline - Enhanced/Auto IBRS + Retpolines6256 eibrs,lfence - Enhanced/Auto IBRS + LFENCE6257 ibrs - use IBRS to protect kernel6258 6259 Not specifying this option is equivalent to6260 spectre_v2=auto.6261 6262 spectre_v2_user=6263 [X86] Control mitigation of Spectre variant 26264 (indirect branch speculation) vulnerability between6265 user space tasks6266 6267 on - Unconditionally enable mitigations. Is6268 enforced by spectre_v2=on6269 6270 off - Unconditionally disable mitigations. Is6271 enforced by spectre_v2=off6272 6273 prctl - Indirect branch speculation is enabled,6274 but mitigation can be enabled via prctl6275 per thread. The mitigation control state6276 is inherited on fork.6277 6278 prctl,ibpb6279 - Like "prctl" above, but only STIBP is6280 controlled per thread. IBPB is issued6281 always when switching between different user6282 space processes.6283 6284 seccomp6285 - Same as "prctl" above, but all seccomp6286 threads will enable the mitigation unless6287 they explicitly opt out.6288 6289 seccomp,ibpb6290 - Like "seccomp" above, but only STIBP is6291 controlled per thread. IBPB is issued6292 always when switching between different6293 user space processes.6294 6295 auto - Kernel selects the mitigation depending on6296 the available CPU features and vulnerability.6297 6298 Default mitigation: "prctl"6299 6300 Not specifying this option is equivalent to6301 spectre_v2_user=auto.6302 6303 spec_rstack_overflow=6304 [X86,EARLY] Control RAS overflow mitigation on AMD Zen CPUs6305 6306 off - Disable mitigation6307 microcode - Enable microcode mitigation only6308 safe-ret - Enable sw-only safe RET mitigation (default)6309 ibpb - Enable mitigation by issuing IBPB on6310 kernel entry6311 ibpb-vmexit - Issue IBPB only on VMEXIT6312 (cloud-specific mitigation)6313 6314 spec_store_bypass_disable=6315 [HW,EARLY] Control Speculative Store Bypass (SSB) Disable mitigation6316 (Speculative Store Bypass vulnerability)6317 6318 Certain CPUs are vulnerable to an exploit against a6319 a common industry wide performance optimization known6320 as "Speculative Store Bypass" in which recent stores6321 to the same memory location may not be observed by6322 later loads during speculative execution. The idea6323 is that such stores are unlikely and that they can6324 be detected prior to instruction retirement at the6325 end of a particular speculation execution window.6326 6327 In vulnerable processors, the speculatively forwarded6328 store can be used in a cache side channel attack, for6329 example to read memory to which the attacker does not6330 directly have access (e.g. inside sandboxed code).6331 6332 This parameter controls whether the Speculative Store6333 Bypass optimization is used.6334 6335 On x86 the options are:6336 6337 on - Unconditionally disable Speculative Store Bypass6338 off - Unconditionally enable Speculative Store Bypass6339 auto - Kernel detects whether the CPU model contains an6340 implementation of Speculative Store Bypass and6341 picks the most appropriate mitigation. If the6342 CPU is not vulnerable, "off" is selected. If the6343 CPU is vulnerable the default mitigation is6344 architecture and Kconfig dependent. See below.6345 prctl - Control Speculative Store Bypass per thread6346 via prctl. Speculative Store Bypass is enabled6347 for a process by default. The state of the control6348 is inherited on fork.6349 seccomp - Same as "prctl" above, but all seccomp threads6350 will disable SSB unless they explicitly opt out.6351 6352 Default mitigations:6353 X86: "prctl"6354 6355 On powerpc the options are:6356 6357 on,auto - On Power8 and Power9 insert a store-forwarding6358 barrier on kernel entry and exit. On Power76359 perform a software flush on kernel entry and6360 exit.6361 off - No action.6362 6363 Not specifying this option is equivalent to6364 spec_store_bypass_disable=auto.6365 6366 split_lock_detect=6367 [X86] Enable split lock detection or bus lock detection6368 6369 When enabled (and if hardware support is present), atomic6370 instructions that access data across cache line6371 boundaries will result in an alignment check exception6372 for split lock detection or a debug exception for6373 bus lock detection.6374 6375 off - not enabled6376 6377 warn - the kernel will emit rate-limited warnings6378 about applications triggering the #AC6379 exception or the #DB exception. This mode is6380 the default on CPUs that support split lock6381 detection or bus lock detection. Default6382 behavior is by #AC if both features are6383 enabled in hardware.6384 6385 fatal - the kernel will send SIGBUS to applications6386 that trigger the #AC exception or the #DB6387 exception. Default behavior is by #AC if6388 both features are enabled in hardware.6389 6390 ratelimit:N -6391 Set system wide rate limit to N bus locks6392 per second for bus lock detection.6393 0 < N <= 1000.6394 6395 N/A for split lock detection.6396 6397 6398 If an #AC exception is hit in the kernel or in6399 firmware (i.e. not while executing in user mode)6400 the kernel will oops in either "warn" or "fatal"6401 mode.6402 6403 #DB exception for bus lock is triggered only when6404 CPL > 0.6405 6406 srbds= [X86,INTEL,EARLY]6407 Control the Special Register Buffer Data Sampling6408 (SRBDS) mitigation.6409 6410 Certain CPUs are vulnerable to an MDS-like6411 exploit which can leak bits from the random6412 number generator.6413 6414 By default, this issue is mitigated by6415 microcode. However, the microcode fix can cause6416 the RDRAND and RDSEED instructions to become6417 much slower. Among other effects, this will6418 result in reduced throughput from /dev/urandom.6419 6420 The microcode mitigation can be disabled with6421 the following option:6422 6423 off: Disable mitigation and remove6424 performance impact to RDRAND and RDSEED6425 6426 srcutree.big_cpu_lim [KNL]6427 Specifies the number of CPUs constituting a6428 large system, such that srcu_struct structures6429 should immediately allocate an srcu_node array.6430 This kernel-boot parameter defaults to 128,6431 but takes effect only when the low-order four6432 bits of srcutree.convert_to_big is equal to 36433 (decide at boot).6434 6435 srcutree.convert_to_big [KNL]6436 Specifies under what conditions an SRCU tree6437 srcu_struct structure will be converted to big6438 form, that is, with an rcu_node tree:6439 6440 0: Never.6441 1: At init_srcu_struct() time.6442 2: When rcutorture decides to.6443 3: Decide at boot time (default).6444 0x1X: Above plus if high contention.6445 6446 Either way, the srcu_node tree will be sized based6447 on the actual runtime number of CPUs (nr_cpu_ids)6448 instead of the compile-time CONFIG_NR_CPUS.6449 6450 srcutree.counter_wrap_check [KNL]6451 Specifies how frequently to check for6452 grace-period sequence counter wrap for the6453 srcu_data structure's ->srcu_gp_seq_needed field.6454 The greater the number of bits set in this kernel6455 parameter, the less frequently counter wrap will6456 be checked for. Note that the bottom two bits6457 are ignored.6458 6459 srcutree.exp_holdoff [KNL]6460 Specifies how many nanoseconds must elapse6461 since the end of the last SRCU grace period for6462 a given srcu_struct until the next normal SRCU6463 grace period will be considered for automatic6464 expediting. Set to zero to disable automatic6465 expediting.6466 6467 srcutree.srcu_max_nodelay [KNL]6468 Specifies the number of no-delay instances6469 per jiffy for which the SRCU grace period6470 worker thread will be rescheduled with zero6471 delay. Beyond this limit, worker thread will6472 be rescheduled with a sleep delay of one jiffy.6473 6474 srcutree.srcu_max_nodelay_phase [KNL]6475 Specifies the per-grace-period phase, number of6476 non-sleeping polls of readers. Beyond this limit,6477 grace period worker thread will be rescheduled6478 with a sleep delay of one jiffy, between each6479 rescan of the readers, for a grace period phase.6480 6481 srcutree.srcu_retry_check_delay [KNL]6482 Specifies number of microseconds of non-sleeping6483 delay between each non-sleeping poll of readers.6484 6485 srcutree.small_contention_lim [KNL]6486 Specifies the number of update-side contention6487 events per jiffy will be tolerated before6488 initiating a conversion of an srcu_struct6489 structure to big form. Note that the value of6490 srcutree.convert_to_big must have the 0x10 bit6491 set for contention-based conversions to occur.6492 6493 ssbd= [ARM64,HW,EARLY]6494 Speculative Store Bypass Disable control6495 6496 On CPUs that are vulnerable to the Speculative6497 Store Bypass vulnerability and offer a6498 firmware based mitigation, this parameter6499 indicates how the mitigation should be used:6500 6501 force-on: Unconditionally enable mitigation for6502 for both kernel and userspace6503 force-off: Unconditionally disable mitigation for6504 for both kernel and userspace6505 kernel: Always enable mitigation in the6506 kernel, and offer a prctl interface6507 to allow userspace to register its6508 interest in being mitigated too.6509 6510 stack_guard_gap= [MM]6511 override the default stack gap protection. The value6512 is in page units and it defines how many pages prior6513 to (for stacks growing down) resp. after (for stacks6514 growing up) the main stack are reserved for no other6515 mapping. Default value is 256 pages.6516 6517 stack_depot_disable= [KNL,EARLY]6518 Setting this to true through kernel command line will6519 disable the stack depot thereby saving the static memory6520 consumed by the stack hash table. By default this is set6521 to false.6522 6523 stacktrace [FTRACE]6524 Enabled the stack tracer on boot up.6525 6526 stacktrace_filter=[function-list]6527 [FTRACE] Limit the functions that the stack tracer6528 will trace at boot up. function-list is a comma-separated6529 list of functions. This list can be changed at run6530 time by the stack_trace_filter file in the debugfs6531 tracing directory. Note, this enables stack tracing6532 and the stacktrace above is not needed.6533 6534 sti= [PARISC,HW]6535 Format: <num>6536 Set the STI (builtin display/keyboard on the HP-PARISC6537 machines) console (graphic card) which should be used6538 as the initial boot-console.6539 See also comment in drivers/video/console/sticore.c.6540 6541 sti_font= [HW]6542 See comment in drivers/video/console/sticore.c.6543 6544 stifb= [HW]6545 Format: bpp:<bpp1>[:<bpp2>[:<bpp3>...]]6546 6547 strict_sas_size=6548 [X86]6549 Format: <bool>6550 Enable or disable strict sigaltstack size checks6551 against the required signal frame size which6552 depends on the supported FPU features. This can6553 be used to filter out binaries which have6554 not yet been made aware of AT_MINSIGSTKSZ.6555 6556 stress_hpt [PPC,EARLY]6557 Limits the number of kernel HPT entries in the hash6558 page table to increase the rate of hash page table6559 faults on kernel addresses.6560 6561 stress_slb [PPC,EARLY]6562 Limits the number of kernel SLB entries, and flushes6563 them frequently to increase the rate of SLB faults6564 on kernel addresses.6565 6566 sunrpc.min_resvport=6567 sunrpc.max_resvport=6568 [NFS,SUNRPC]6569 SunRPC servers often require that client requests6570 originate from a privileged port (i.e. a port in the6571 range 0 < portnr < 1024).6572 An administrator who wishes to reserve some of these6573 ports for other uses may adjust the range that the6574 kernel's sunrpc client considers to be privileged6575 using these two parameters to set the minimum and6576 maximum port values.6577 6578 sunrpc.svc_rpc_per_connection_limit=6579 [NFS,SUNRPC]6580 Limit the number of requests that the server will6581 process in parallel from a single connection.6582 The default value is 0 (no limit).6583 6584 sunrpc.pool_mode=6585 [NFS]6586 Control how the NFS server code allocates CPUs to6587 service thread pools. Depending on how many NICs6588 you have and where their interrupts are bound, this6589 option will affect which CPUs will do NFS serving.6590 Note: this parameter cannot be changed while the6591 NFS server is running.6592 6593 auto the server chooses an appropriate mode6594 automatically using heuristics6595 global a single global pool contains all CPUs6596 percpu one pool for each CPU6597 pernode one pool for each NUMA node (equivalent6598 to global on non-NUMA machines)6599 6600 sunrpc.tcp_slot_table_entries=6601 sunrpc.udp_slot_table_entries=6602 [NFS,SUNRPC]6603 Sets the upper limit on the number of simultaneous6604 RPC calls that can be sent from the client to a6605 server. Increasing these values may allow you to6606 improve throughput, but will also increase the6607 amount of memory reserved for use by the client.6608 6609 suspend.pm_test_delay=6610 [SUSPEND]6611 Sets the number of seconds to remain in a suspend test6612 mode before resuming the system (see6613 /sys/power/pm_test). Only available when CONFIG_PM_DEBUG6614 is set. Default value is 5.6615 6616 svm= [PPC]6617 Format: { on | off | y | n | 1 | 0 }6618 This parameter controls use of the Protected6619 Execution Facility on pSeries.6620 6621 swiotlb= [ARM,PPC,MIPS,X86,S390,EARLY]6622 Format: { <int> [,<int>] | force | noforce }6623 <int> -- Number of I/O TLB slabs6624 <int> -- Second integer after comma. Number of swiotlb6625 areas with their own lock. Will be rounded up6626 to a power of 2.6627 force -- force using of bounce buffers even if they6628 wouldn't be automatically used by the kernel6629 noforce -- Never use bounce buffers (for debugging)6630 6631 switches= [HW,M68k,EARLY]6632 6633 sysctl.*= [KNL]6634 Set a sysctl parameter, right before loading the init6635 process, as if the value was written to the respective6636 /proc/sys/... file. Both '.' and '/' are recognized as6637 separators. Unrecognized parameters and invalid values6638 are reported in the kernel log. Sysctls registered6639 later by a loaded module cannot be set this way.6640 Example: sysctl.vm.swappiness=406641 6642 sysrq_always_enabled6643 [KNL]6644 Ignore sysrq setting - this boot parameter will6645 neutralize any effect of /proc/sys/kernel/sysrq.6646 Useful for debugging.6647 6648 tcpmhash_entries= [KNL,NET]6649 Set the number of tcp_metrics_hash slots.6650 Default value is 8192 or 16384 depending on total6651 ram pages. This is used to specify the TCP metrics6652 cache size. See Documentation/networking/ip-sysctl.rst6653 "tcp_no_metrics_save" section for more details.6654 6655 tdfx= [HW,DRM]6656 6657 test_suspend= [SUSPEND]6658 Format: { "mem" | "standby" | "freeze" }[,N]6659 Specify "mem" (for Suspend-to-RAM) or "standby" (for6660 standby suspend) or "freeze" (for suspend type freeze)6661 as the system sleep state during system startup with6662 the optional capability to repeat N number of times.6663 The system is woken from this state using a6664 wakeup-capable RTC alarm.6665 6666 thash_entries= [KNL,NET]6667 Set number of hash buckets for TCP connection6668 6669 thermal.act= [HW,ACPI]6670 -1: disable all active trip points in all thermal zones6671 <degrees C>: override all lowest active trip points6672 6673 thermal.crt= [HW,ACPI]6674 -1: disable all critical trip points in all thermal zones6675 <degrees C>: override all critical trip points6676 6677 thermal.off= [HW,ACPI]6678 1: disable ACPI thermal control6679 6680 thermal.psv= [HW,ACPI]6681 -1: disable all passive trip points6682 <degrees C>: override all passive trip points to this6683 value6684 6685 thermal.tzp= [HW,ACPI]6686 Specify global default ACPI thermal zone polling rate6687 <deci-seconds>: poll all this frequency6688 0: no polling (default)6689 6690 thp_anon= [KNL]6691 Format: <size>[KMG],<size>[KMG]:<state>;<size>[KMG]-<size>[KMG]:<state>6692 state is one of "always", "madvise", "never" or "inherit".6693 Control the default behavior of the system with respect6694 to anonymous transparent hugepages.6695 Can be used multiple times for multiple anon THP sizes.6696 See Documentation/admin-guide/mm/transhuge.rst for more6697 details.6698 6699 threadirqs [KNL,EARLY]6700 Force threading of all interrupt handlers except those6701 marked explicitly IRQF_NO_THREAD.6702 6703 topology= [S390,EARLY]6704 Format: {off | on}6705 Specify if the kernel should make use of the cpu6706 topology information if the hardware supports this.6707 The scheduler will make use of this information and6708 e.g. base its process migration decisions on it.6709 Default is on.6710 6711 torture.disable_onoff_at_boot= [KNL]6712 Prevent the CPU-hotplug component of torturing6713 until after init has spawned.6714 6715 torture.ftrace_dump_at_shutdown= [KNL]6716 Dump the ftrace buffer at torture-test shutdown,6717 even if there were no errors. This can be a6718 very costly operation when many torture tests6719 are running concurrently, especially on systems6720 with rotating-rust storage.6721 6722 torture.verbose_sleep_frequency= [KNL]6723 Specifies how many verbose printk()s should be6724 emitted between each sleep. The default of zero6725 disables verbose-printk() sleeping.6726 6727 torture.verbose_sleep_duration= [KNL]6728 Duration of each verbose-printk() sleep in jiffies.6729 6730 tpm.disable_pcr_integrity= [HW,TPM]6731 Do not protect PCR registers from unintended physical6732 access, or interposers in the bus by the means of6733 having an integrity protected session wrapped around6734 TPM2_PCR_Extend command. Consider this in a situation6735 where TPM is heavily utilized by IMA, thus protection6736 causing a major performance hit, and the space where6737 machines are deployed is by other means guarded.6738 6739 tpm_suspend_pcr=[HW,TPM]6740 Format: integer pcr id6741 Specify that at suspend time, the tpm driver6742 should extend the specified pcr with zeros,6743 as a workaround for some chips which fail to6744 flush the last written pcr on TPM_SaveState.6745 This will guarantee that all the other pcrs6746 are saved.6747 6748 tpm_tis.interrupts= [HW,TPM]6749 Enable interrupts for the MMIO based physical layer6750 for the FIFO interface. By default it is set to false6751 (0). For more information about TPM hardware interfaces6752 defined by Trusted Computing Group (TCG) see6753 https://trustedcomputinggroup.org/resource/pc-client-platform-tpm-profile-ptp-specification/6754 6755 tp_printk [FTRACE]6756 Have the tracepoints sent to printk as well as the6757 tracing ring buffer. This is useful for early boot up6758 where the system hangs or reboots and does not give the6759 option for reading the tracing buffer or performing a6760 ftrace_dump_on_oops.6761 6762 To turn off having tracepoints sent to printk,6763 echo 0 > /proc/sys/kernel/tracepoint_printk6764 Note, echoing 1 into this file without the6765 tp_printk kernel cmdline option has no effect.6766 6767 The tp_printk_stop_on_boot (see below) can also be used6768 to stop the printing of events to console at6769 late_initcall_sync.6770 6771 ** CAUTION **6772 6773 Having tracepoints sent to printk() and activating high6774 frequency tracepoints such as irq or sched, can cause6775 the system to live lock.6776 6777 tp_printk_stop_on_boot [FTRACE]6778 When tp_printk (above) is set, it can cause a lot of noise6779 on the console. It may be useful to only include the6780 printing of events during boot up, as user space may6781 make the system inoperable.6782 6783 This command line option will stop the printing of events6784 to console at the late_initcall_sync() time frame.6785 6786 trace_buf_size=nn[KMG]6787 [FTRACE] will set tracing buffer size on each cpu.6788 6789 trace_clock= [FTRACE] Set the clock used for tracing events6790 at boot up.6791 local - Use the per CPU time stamp counter6792 (converted into nanoseconds). Fast, but6793 depending on the architecture, may not be6794 in sync between CPUs.6795 global - Event time stamps are synchronize across6796 CPUs. May be slower than the local clock,6797 but better for some race conditions.6798 counter - Simple counting of events (1, 2, ..)6799 note, some counts may be skipped due to the6800 infrastructure grabbing the clock more than6801 once per event.6802 uptime - Use jiffies as the time stamp.6803 perf - Use the same clock that perf uses.6804 mono - Use ktime_get_mono_fast_ns() for time stamps.6805 mono_raw - Use ktime_get_raw_fast_ns() for time6806 stamps.6807 boot - Use ktime_get_boot_fast_ns() for time stamps.6808 Architectures may add more clocks. See6809 Documentation/trace/ftrace.rst for more details.6810 6811 trace_event=[event-list]6812 [FTRACE] Set and start specified trace events in order6813 to facilitate early boot debugging. The event-list is a6814 comma-separated list of trace events to enable. See6815 also Documentation/trace/events.rst6816 6817 trace_instance=[instance-info]6818 [FTRACE] Create a ring buffer instance early in boot up.6819 This will be listed in:6820 6821 /sys/kernel/tracing/instances6822 6823 Events can be enabled at the time the instance is created6824 via:6825 6826 trace_instance=<name>,<system1>:<event1>,<system2>:<event2>6827 6828 Note, the "<system*>:" portion is optional if the event is6829 unique.6830 6831 trace_instance=foo,sched:sched_switch,irq_handler_entry,initcall6832 6833 will enable the "sched_switch" event (note, the "sched:" is optional, and6834 the same thing would happen if it was left off). The irq_handler_entry6835 event, and all events under the "initcall" system.6836 6837 Flags can be added to the instance to modify its behavior when it is6838 created. The flags are separated by '^'.6839 6840 The available flags are:6841 6842 traceoff - Have the tracing instance tracing disabled after it is created.6843 traceprintk - Have trace_printk() write into this trace instance6844 (note, "printk" and "trace_printk" can also be used)6845 6846 trace_instance=foo^traceoff^traceprintk,sched,irq6847 6848 The flags must come before the defined events.6849 6850 If memory has been reserved (see memmap for x86), the instance6851 can use that memory:6852 6853 memmap=12M$0x284500000 trace_instance=boot_map@0x284500000:12M6854 6855 The above will create a "boot_map" instance that uses the physical6856 memory at 0x284500000 that is 12Megs. The per CPU buffers of that6857 instance will be split up accordingly.6858 6859 Alternatively, the memory can be reserved by the reserve_mem option:6860 6861 reserve_mem=12M:4096:trace trace_instance=boot_map@trace6862 6863 This will reserve 12 megabytes at boot up with a 4096 byte alignment6864 and place the ring buffer in this memory. Note that due to KASLR, the6865 memory may not be the same location each time, which will not preserve6866 the buffer content.6867 6868 Also note that the layout of the ring buffer data may change between6869 kernel versions where the validator will fail and reset the ring buffer6870 if the layout is not the same as the previous kernel.6871 6872 If the ring buffer is used for persistent bootups and has events enabled,6873 it is recommend to disable tracing so that events from a previous boot do not6874 mix with events of the current boot (unless you are debugging a random crash6875 at boot up).6876 6877 reserve_mem=12M:4096:trace trace_instance=boot_map^traceoff^traceprintk@trace,sched,irq6878 6879 See also Documentation/trace/debugging.rst6880 6881 6882 trace_options=[option-list]6883 [FTRACE] Enable or disable tracer options at boot.6884 The option-list is a comma delimited list of options6885 that can be enabled or disabled just as if you were6886 to echo the option name into6887 6888 /sys/kernel/tracing/trace_options6889 6890 For example, to enable stacktrace option (to dump the6891 stack trace of each event), add to the command line:6892 6893 trace_options=stacktrace6894 6895 See also Documentation/trace/ftrace.rst "trace options"6896 section.6897 6898 trace_trigger=[trigger-list]6899 [FTRACE] Add a event trigger on specific events.6900 Set a trigger on top of a specific event, with an optional6901 filter.6902 6903 The format is is "trace_trigger=<event>.<trigger>[ if <filter>],..."6904 Where more than one trigger may be specified that are comma deliminated.6905 6906 For example:6907 6908 trace_trigger="sched_switch.stacktrace if prev_state == 2"6909 6910 The above will enable the "stacktrace" trigger on the "sched_switch"6911 event but only trigger it if the "prev_state" of the "sched_switch"6912 event is "2" (TASK_UNINTERUPTIBLE).6913 6914 See also "Event triggers" in Documentation/trace/events.rst6915 6916 6917 traceoff_on_warning6918 [FTRACE] enable this option to disable tracing when a6919 warning is hit. This turns off "tracing_on". Tracing can6920 be enabled again by echoing '1' into the "tracing_on"6921 file located in /sys/kernel/tracing/6922 6923 This option is useful, as it disables the trace before6924 the WARNING dump is called, which prevents the trace to6925 be filled with content caused by the warning output.6926 6927 This option can also be set at run time via the sysctl6928 option: kernel/traceoff_on_warning6929 6930 transparent_hugepage=6931 [KNL]6932 Format: [always|madvise|never]6933 Can be used to control the default behavior of the system6934 with respect to transparent hugepages.6935 See Documentation/admin-guide/mm/transhuge.rst6936 for more details.6937 6938 trusted.source= [KEYS]6939 Format: <string>6940 This parameter identifies the trust source as a backend6941 for trusted keys implementation. Supported trust6942 sources:6943 - "tpm"6944 - "tee"6945 - "caam"6946 - "dcp"6947 If not specified then it defaults to iterating through6948 the trust source list starting with TPM and assigns the6949 first trust source as a backend which is initialized6950 successfully during iteration.6951 6952 trusted.rng= [KEYS]6953 Format: <string>6954 The RNG used to generate key material for trusted keys.6955 Can be one of:6956 - "kernel"6957 - the same value as trusted.source: "tpm" or "tee"6958 - "default"6959 If not specified, "default" is used. In this case,6960 the RNG's choice is left to each individual trust source.6961 6962 trusted.dcp_use_otp_key6963 This is intended to be used in combination with6964 trusted.source=dcp and will select the DCP OTP key6965 instead of the DCP UNIQUE key blob encryption.6966 6967 trusted.dcp_skip_zk_test6968 This is intended to be used in combination with6969 trusted.source=dcp and will disable the check if the6970 blob key is all zeros. This is helpful for situations where6971 having this key zero'ed is acceptable. E.g. in testing6972 scenarios.6973 6974 tsc= Disable clocksource stability checks for TSC.6975 Format: <string>6976 [x86] reliable: mark tsc clocksource as reliable, this6977 disables clocksource verification at runtime, as well6978 as the stability checks done at bootup. Used to enable6979 high-resolution timer mode on older hardware, and in6980 virtualized environment.6981 [x86] noirqtime: Do not use TSC to do irq accounting.6982 Used to run time disable IRQ_TIME_ACCOUNTING on any6983 platforms where RDTSC is slow and this accounting6984 can add overhead.6985 [x86] unstable: mark the TSC clocksource as unstable, this6986 marks the TSC unconditionally unstable at bootup and6987 avoids any further wobbles once the TSC watchdog notices.6988 [x86] nowatchdog: disable clocksource watchdog. Used6989 in situations with strict latency requirements (where6990 interruptions from clocksource watchdog are not6991 acceptable).6992 [x86] recalibrate: force recalibration against a HW timer6993 (HPET or PM timer) on systems whose TSC frequency was6994 obtained from HW or FW using either an MSR or CPUID(0x15).6995 Warn if the difference is more than 500 ppm.6996 [x86] watchdog: Use TSC as the watchdog clocksource with6997 which to check other HW timers (HPET or PM timer), but6998 only on systems where TSC has been deemed trustworthy.6999 This will be suppressed by an earlier tsc=nowatchdog and7000 can be overridden by a later tsc=nowatchdog. A console7001 message will flag any such suppression or overriding.7002 7003 tsc_early_khz= [X86,EARLY] Skip early TSC calibration and use the given7004 value instead. Useful when the early TSC frequency discovery7005 procedure is not reliable, such as on overclocked systems7006 with CPUID.16h support and partial CPUID.15h support.7007 Format: <unsigned int>7008 7009 tsx= [X86] Control Transactional Synchronization7010 Extensions (TSX) feature in Intel processors that7011 support TSX control.7012 7013 This parameter controls the TSX feature. The options are:7014 7015 on - Enable TSX on the system. Although there are7016 mitigations for all known security vulnerabilities,7017 TSX has been known to be an accelerator for7018 several previous speculation-related CVEs, and7019 so there may be unknown security risks associated7020 with leaving it enabled.7021 7022 off - Disable TSX on the system. (Note that this7023 option takes effect only on newer CPUs which are7024 not vulnerable to MDS, i.e., have7025 MSR_IA32_ARCH_CAPABILITIES.MDS_NO=1 and which get7026 the new IA32_TSX_CTRL MSR through a microcode7027 update. This new MSR allows for the reliable7028 deactivation of the TSX functionality.)7029 7030 auto - Disable TSX if X86_BUG_TAA is present,7031 otherwise enable TSX on the system.7032 7033 Not specifying this option is equivalent to tsx=off.7034 7035 See Documentation/admin-guide/hw-vuln/tsx_async_abort.rst7036 for more details.7037 7038 tsx_async_abort= [X86,INTEL,EARLY] Control mitigation for the TSX Async7039 Abort (TAA) vulnerability.7040 7041 Similar to Micro-architectural Data Sampling (MDS)7042 certain CPUs that support Transactional7043 Synchronization Extensions (TSX) are vulnerable to an7044 exploit against CPU internal buffers which can forward7045 information to a disclosure gadget under certain7046 conditions.7047 7048 In vulnerable processors, the speculatively forwarded7049 data can be used in a cache side channel attack, to7050 access data to which the attacker does not have direct7051 access.7052 7053 This parameter controls the TAA mitigation. The7054 options are:7055 7056 full - Enable TAA mitigation on vulnerable CPUs7057 if TSX is enabled.7058 7059 full,nosmt - Enable TAA mitigation and disable SMT on7060 vulnerable CPUs. If TSX is disabled, SMT7061 is not disabled because CPU is not7062 vulnerable to cross-thread TAA attacks.7063 off - Unconditionally disable TAA mitigation7064 7065 On MDS-affected machines, tsx_async_abort=off can be7066 prevented by an active MDS mitigation as both vulnerabilities7067 are mitigated with the same mechanism so in order to disable7068 this mitigation, you need to specify mds=off too.7069 7070 Not specifying this option is equivalent to7071 tsx_async_abort=full. On CPUs which are MDS affected7072 and deploy MDS mitigation, TAA mitigation is not7073 required and doesn't provide any additional7074 mitigation.7075 7076 For details see:7077 Documentation/admin-guide/hw-vuln/tsx_async_abort.rst7078 7079 turbografx.map[2|3]= [HW,JOY]7080 TurboGraFX parallel port interface7081 Format:7082 <port#>,<js1>,<js2>,<js3>,<js4>,<js5>,<js6>,<js7>7083 See also Documentation/input/devices/joystick-parport.rst7084 7085 udbg-immortal [PPC] When debugging early kernel crashes that7086 happen after console_init() and before a proper7087 console driver takes over, this boot options might7088 help "seeing" what's going on.7089 7090 uhash_entries= [KNL,NET]7091 Set number of hash buckets for UDP/UDP-Lite connections7092 7093 uhci-hcd.ignore_oc=7094 [USB] Ignore overcurrent events (default N).7095 Some badly-designed motherboards generate lots of7096 bogus events, for ports that aren't wired to7097 anything. Set this parameter to avoid log spamming.7098 Note that genuine overcurrent events won't be7099 reported either.7100 7101 unknown_nmi_panic7102 [X86] Cause panic on unknown NMI.7103 7104 unwind_debug [X86-64,EARLY]7105 Enable unwinder debug output. This can be7106 useful for debugging certain unwinder error7107 conditions, including corrupt stacks and7108 bad/missing unwinder metadata.7109 7110 usbcore.authorized_default=7111 [USB] Default USB device authorization:7112 (default -1 = authorized (same as 1),7113 0 = not authorized, 1 = authorized, 2 = authorized7114 if device connected to internal port)7115 7116 usbcore.autosuspend=7117 [USB] The autosuspend time delay (in seconds) used7118 for newly-detected USB devices (default 2). This7119 is the time required before an idle device will be7120 autosuspended. Devices for which the delay is set7121 to a negative value won't be autosuspended at all.7122 7123 usbcore.usbfs_snoop=7124 [USB] Set to log all usbfs traffic (default 0 = off).7125 7126 usbcore.usbfs_snoop_max=7127 [USB] Maximum number of bytes to snoop in each URB7128 (default = 65536).7129 7130 usbcore.blinkenlights=7131 [USB] Set to cycle leds on hubs (default 0 = off).7132 7133 usbcore.old_scheme_first=7134 [USB] Start with the old device initialization7135 scheme (default 0 = off).7136 7137 usbcore.usbfs_memory_mb=7138 [USB] Memory limit (in MB) for buffers allocated by7139 usbfs (default = 16, 0 = max = 2047).7140 7141 usbcore.use_both_schemes=7142 [USB] Try the other device initialization scheme7143 if the first one fails (default 1 = enabled).7144 7145 usbcore.initial_descriptor_timeout=7146 [USB] Specifies timeout for the initial 64-byte7147 USB_REQ_GET_DESCRIPTOR request in milliseconds7148 (default 5000 = 5.0 seconds).7149 7150 usbcore.nousb [USB] Disable the USB subsystem7151 7152 usbcore.quirks=7153 [USB] A list of quirk entries to augment the built-in7154 usb core quirk list. List entries are separated by7155 commas. Each entry has the form7156 VendorID:ProductID:Flags. The IDs are 4-digit hex7157 numbers and Flags is a set of letters. Each letter7158 will change the built-in quirk; setting it if it is7159 clear and clearing it if it is set. The letters have7160 the following meanings:7161 a = USB_QUIRK_STRING_FETCH_255 (string7162 descriptors must not be fetched using7163 a 255-byte read);7164 b = USB_QUIRK_RESET_RESUME (device can't resume7165 correctly so reset it instead);7166 c = USB_QUIRK_NO_SET_INTF (device can't handle7167 Set-Interface requests);7168 d = USB_QUIRK_CONFIG_INTF_STRINGS (device can't7169 handle its Configuration or Interface7170 strings);7171 e = USB_QUIRK_RESET (device can't be reset7172 (e.g morph devices), don't use reset);7173 f = USB_QUIRK_HONOR_BNUMINTERFACES (device has7174 more interface descriptions than the7175 bNumInterfaces count, and can't handle7176 talking to these interfaces);7177 g = USB_QUIRK_DELAY_INIT (device needs a pause7178 during initialization, after we read7179 the device descriptor);7180 h = USB_QUIRK_LINEAR_UFRAME_INTR_BINTERVAL (For7181 high speed and super speed interrupt7182 endpoints, the USB 2.0 and USB 3.0 spec7183 require the interval in microframes (17184 microframe = 125 microseconds) to be7185 calculated as interval = 2 ^7186 (bInterval-1).7187 Devices with this quirk report their7188 bInterval as the result of this7189 calculation instead of the exponent7190 variable used in the calculation);7191 i = USB_QUIRK_DEVICE_QUALIFIER (device can't7192 handle device_qualifier descriptor7193 requests);7194 j = USB_QUIRK_IGNORE_REMOTE_WAKEUP (device7195 generates spurious wakeup, ignore7196 remote wakeup capability);7197 k = USB_QUIRK_NO_LPM (device can't handle Link7198 Power Management);7199 l = USB_QUIRK_LINEAR_FRAME_INTR_BINTERVAL7200 (Device reports its bInterval as linear7201 frames instead of the USB 2.07202 calculation);7203 m = USB_QUIRK_DISCONNECT_SUSPEND (Device needs7204 to be disconnected before suspend to7205 prevent spurious wakeup);7206 n = USB_QUIRK_DELAY_CTRL_MSG (Device needs a7207 pause after every control message);7208 o = USB_QUIRK_HUB_SLOW_RESET (Hub needs extra7209 delay after resetting its port);7210 p = USB_QUIRK_SHORT_SET_ADDRESS_REQ_TIMEOUT7211 (Reduce timeout of the SET_ADDRESS7212 request from 5000 ms to 500 ms);7213 Example: quirks=0781:5580:bk,0a5c:5834:gij7214 7215 usbhid.mousepoll=7216 [USBHID] The interval which mice are to be polled at.7217 7218 usbhid.jspoll=7219 [USBHID] The interval which joysticks are to be polled at.7220 7221 usbhid.kbpoll=7222 [USBHID] The interval which keyboards are to be polled at.7223 7224 usb-storage.delay_use=7225 [UMS] The delay in seconds before a new device is7226 scanned for Logical Units (default 1).7227 Optionally the delay in milliseconds if the value has7228 suffix with "ms".7229 Example: delay_use=2567ms7230 7231 usb-storage.quirks=7232 [UMS] A list of quirks entries to supplement or7233 override the built-in unusual_devs list. List7234 entries are separated by commas. Each entry has7235 the form VID:PID:Flags where VID and PID are Vendor7236 and Product ID values (4-digit hex numbers) and7237 Flags is a set of characters, each corresponding7238 to a common usb-storage quirk flag as follows:7239 a = SANE_SENSE (collect more than 18 bytes7240 of sense data, not on uas);7241 b = BAD_SENSE (don't collect more than 187242 bytes of sense data, not on uas);7243 c = FIX_CAPACITY (decrease the reported7244 device capacity by one sector);7245 d = NO_READ_DISC_INFO (don't use7246 READ_DISC_INFO command, not on uas);7247 e = NO_READ_CAPACITY_16 (don't use7248 READ_CAPACITY_16 command);7249 f = NO_REPORT_OPCODES (don't use report opcodes7250 command, uas only);7251 g = MAX_SECTORS_240 (don't transfer more than7252 240 sectors at a time, uas only);7253 h = CAPACITY_HEURISTICS (decrease the7254 reported device capacity by one7255 sector if the number is odd);7256 i = IGNORE_DEVICE (don't bind to this7257 device);7258 j = NO_REPORT_LUNS (don't use report luns7259 command, uas only);7260 k = NO_SAME (do not use WRITE_SAME, uas only)7261 l = NOT_LOCKABLE (don't try to lock and7262 unlock ejectable media, not on uas);7263 m = MAX_SECTORS_64 (don't transfer more7264 than 64 sectors = 32 KB at a time,7265 not on uas);7266 n = INITIAL_READ10 (force a retry of the7267 initial READ(10) command, not on uas);7268 o = CAPACITY_OK (accept the capacity7269 reported by the device, not on uas);7270 p = WRITE_CACHE (the device cache is ON7271 by default, not on uas);7272 r = IGNORE_RESIDUE (the device reports7273 bogus residue values, not on uas);7274 s = SINGLE_LUN (the device has only one7275 Logical Unit);7276 t = NO_ATA_1X (don't allow ATA(12) and ATA(16)7277 commands, uas only);7278 u = IGNORE_UAS (don't bind to the uas driver);7279 w = NO_WP_DETECT (don't test whether the7280 medium is write-protected).7281 y = ALWAYS_SYNC (issue a SYNCHRONIZE_CACHE7282 even if the device claims no cache,7283 not on uas)7284 Example: quirks=0419:aaf5:rl,0421:0433:rc7285 7286 user_debug= [KNL,ARM]7287 Format: <int>7288 See arch/arm/Kconfig.debug help text.7289 1 - undefined instruction events7290 2 - system calls7291 4 - invalid data aborts7292 8 - SIGSEGV faults7293 16 - SIGBUS faults7294 Example: user_debug=317295 7296 userpte=7297 [X86,EARLY] Flags controlling user PTE allocations.7298 7299 nohigh = do not allocate PTE pages in7300 HIGHMEM regardless of setting7301 of CONFIG_HIGHPTE.7302 7303 vdso= [X86,SH,SPARC]7304 On X86_32, this is an alias for vdso32=. Otherwise:7305 7306 vdso=1: enable VDSO (the default)7307 vdso=0: disable VDSO mapping7308 7309 vdso32= [X86] Control the 32-bit vDSO7310 vdso32=1: enable 32-bit VDSO7311 vdso32=0 or vdso32=2: disable 32-bit VDSO7312 7313 See the help text for CONFIG_COMPAT_VDSO for more7314 details. If CONFIG_COMPAT_VDSO is set, the default is7315 vdso32=0; otherwise, the default is vdso32=1.7316 7317 For compatibility with older kernels, vdso32=2 is an7318 alias for vdso32=0.7319 7320 Try vdso32=0 if you encounter an error that says:7321 dl_main: Assertion `(void *) ph->p_vaddr == _rtld_local._dl_sysinfo_dso' failed!7322 7323 video= [FB,EARLY] Frame buffer configuration7324 See Documentation/fb/modedb.rst.7325 7326 video.brightness_switch_enabled= [ACPI]7327 Format: [0|1]7328 If set to 1, on receiving an ACPI notify event7329 generated by hotkey, video driver will adjust brightness7330 level and then send out the event to user space through7331 the allocated input device. If set to 0, video driver7332 will only send out the event without touching backlight7333 brightness level.7334 default: 17335 7336 virtio_mmio.device=7337 [VMMIO] Memory mapped virtio (platform) device.7338 7339 <size>@<baseaddr>:<irq>[:<id>]7340 where:7341 <size> := size (can use standard suffixes7342 like K, M and G)7343 <baseaddr> := physical base address7344 <irq> := interrupt number (as passed to7345 request_irq())7346 <id> := (optional) platform device id7347 example:7348 virtio_mmio.device=1K@0x100b0000:48:77349 7350 Can be used multiple times for multiple devices.7351 7352 vga= [BOOT,X86-32] Select a particular video mode7353 See Documentation/arch/x86/boot.rst and7354 Documentation/admin-guide/svga.rst.7355 Use vga=ask for menu.7356 This is actually a boot loader parameter; the value is7357 passed to the kernel using a special protocol.7358 7359 vm_debug[=options] [KNL] Available with CONFIG_DEBUG_VM=y.7360 May slow down system boot speed, especially when7361 enabled on systems with a large amount of memory.7362 All options are enabled by default, and this7363 interface is meant to allow for selectively7364 enabling or disabling specific virtual memory7365 debugging features.7366 7367 Available options are:7368 P Enable page structure init time poisoning7369 - Disable all of the above options7370 7371 vmalloc=nn[KMG] [KNL,BOOT,EARLY] Forces the vmalloc area to have an7372 exact size of <nn>. This can be used to increase7373 the minimum size (128MB on x86, arm32 platforms).7374 It can also be used to decrease the size and leave more room7375 for directly mapped kernel RAM. Note that this parameter does7376 not exist on many other platforms (including arm64, alpha,7377 loongarch, arc, csky, hexagon, microblaze, mips, nios2, openrisc,7378 parisc, m64k, powerpc, riscv, sh, um, xtensa, s390, sparc).7379 7380 vmcp_cma=nn[MG] [KNL,S390,EARLY]7381 Sets the memory size reserved for contiguous memory7382 allocations for the vmcp device driver.7383 7384 vmhalt= [KNL,S390] Perform z/VM CP command after system halt.7385 Format: <command>7386 7387 vmpanic= [KNL,S390] Perform z/VM CP command after kernel panic.7388 Format: <command>7389 7390 vmpoff= [KNL,S390] Perform z/VM CP command after power off.7391 Format: <command>7392 7393 vsyscall= [X86-64,EARLY]7394 Controls the behavior of vsyscalls (i.e. calls to7395 fixed addresses of 0xffffffffff600x00 from legacy7396 code). Most statically-linked binaries and older7397 versions of glibc use these calls. Because these7398 functions are at fixed addresses, they make nice7399 targets for exploits that can control RIP.7400 7401 emulate Vsyscalls turn into traps and are emulated7402 reasonably safely. The vsyscall page is7403 readable.7404 7405 xonly [default] Vsyscalls turn into traps and are7406 emulated reasonably safely. The vsyscall7407 page is not readable.7408 7409 none Vsyscalls don't work at all. This makes7410 them quite hard to use for exploits but7411 might break your system.7412 7413 vt.color= [VT] Default text color.7414 Format: 0xYX, X = foreground, Y = background.7415 Default: 0x07 = light gray on black.7416 7417 vt.cur_default= [VT] Default cursor shape.7418 Format: 0xCCBBAA, where AA, BB, and CC are the same as7419 the parameters of the <Esc>[?A;B;Cc escape sequence;7420 see VGA-softcursor.txt. Default: 2 = underline.7421 7422 vt.default_blu= [VT]7423 Format: <blue0>,<blue1>,<blue2>,...,<blue15>7424 Change the default blue palette of the console.7425 This is a 16-member array composed of values7426 ranging from 0-255.7427 7428 vt.default_grn= [VT]7429 Format: <green0>,<green1>,<green2>,...,<green15>7430 Change the default green palette of the console.7431 This is a 16-member array composed of values7432 ranging from 0-255.7433 7434 vt.default_red= [VT]7435 Format: <red0>,<red1>,<red2>,...,<red15>7436 Change the default red palette of the console.7437 This is a 16-member array composed of values7438 ranging from 0-255.7439 7440 vt.default_utf8=7441 [VT]7442 Format=<0|1>7443 Set system-wide default UTF-8 mode for all tty's.7444 Default is 1, i.e. UTF-8 mode is enabled for all7445 newly opened terminals.7446 7447 vt.global_cursor_default=7448 [VT]7449 Format=<-1|0|1>7450 Set system-wide default for whether a cursor7451 is shown on new VTs. Default is -1,7452 i.e. cursors will be created by default unless7453 overridden by individual drivers. 0 will hide7454 cursors, 1 will display them.7455 7456 vt.italic= [VT] Default color for italic text; 0-15.7457 Default: 2 = green.7458 7459 vt.underline= [VT] Default color for underlined text; 0-15.7460 Default: 3 = cyan.7461 7462 watchdog timers [HW,WDT] For information on watchdog timers,7463 see Documentation/watchdog/watchdog-parameters.rst7464 or other driver-specific files in the7465 Documentation/watchdog/ directory.7466 7467 watchdog_thresh=7468 [KNL]7469 Set the hard lockup detector stall duration7470 threshold in seconds. The soft lockup detector7471 threshold is set to twice the value. A value of 07472 disables both lockup detectors. Default is 107473 seconds.7474 7475 workqueue.unbound_cpus=7476 [KNL,SMP] Specify to constrain one or some CPUs7477 to use in unbound workqueues.7478 Format: <cpu-list>7479 By default, all online CPUs are available for7480 unbound workqueues.7481 7482 workqueue.watchdog_thresh=7483 If CONFIG_WQ_WATCHDOG is configured, workqueue can7484 warn stall conditions and dump internal state to7485 help debugging. 0 disables workqueue stall7486 detection; otherwise, it's the stall threshold7487 duration in seconds. The default value is 30 and7488 it can be updated at runtime by writing to the7489 corresponding sysfs file.7490 7491 workqueue.panic_on_stall=<uint>7492 Panic when workqueue stall is detected by7493 CONFIG_WQ_WATCHDOG. It sets the number times of the7494 stall to trigger panic.7495 7496 The default is 0, which disables the panic on stall.7497 7498 workqueue.cpu_intensive_thresh_us=7499 Per-cpu work items which run for longer than this7500 threshold are automatically considered CPU intensive7501 and excluded from concurrency management to prevent7502 them from noticeably delaying other per-cpu work7503 items. Default is 10000 (10ms).7504 7505 If CONFIG_WQ_CPU_INTENSIVE_REPORT is set, the kernel7506 will report the work functions which violate this7507 threshold repeatedly. They are likely good7508 candidates for using WQ_UNBOUND workqueues instead.7509 7510 workqueue.cpu_intensive_warning_thresh=<uint>7511 If CONFIG_WQ_CPU_INTENSIVE_REPORT is set, the kernel7512 will report the work functions which violate the7513 intensive_threshold_us repeatedly. In order to prevent7514 spurious warnings, start printing only after a work7515 function has violated this threshold number of times.7516 7517 The default is 4 times. 0 disables the warning.7518 7519 workqueue.power_efficient7520 Per-cpu workqueues are generally preferred because7521 they show better performance thanks to cache7522 locality; unfortunately, per-cpu workqueues tend to7523 be more power hungry than unbound workqueues.7524 7525 Enabling this makes the per-cpu workqueues which7526 were observed to contribute significantly to power7527 consumption unbound, leading to measurably lower7528 power usage at the cost of small performance7529 overhead.7530 7531 The default value of this parameter is determined by7532 the config option CONFIG_WQ_POWER_EFFICIENT_DEFAULT.7533 7534 workqueue.default_affinity_scope=7535 Select the default affinity scope to use for unbound7536 workqueues. Can be one of "cpu", "smt", "cache",7537 "numa" and "system". Default is "cache". For more7538 information, see the Affinity Scopes section in7539 Documentation/core-api/workqueue.rst.7540 7541 This can be changed after boot by writing to the7542 matching /sys/module/workqueue/parameters file. All7543 workqueues with the "default" affinity scope will be7544 updated accordingly.7545 7546 workqueue.debug_force_rr_cpu7547 Workqueue used to implicitly guarantee that work7548 items queued without explicit CPU specified are put7549 on the local CPU. This guarantee is no longer true7550 and while local CPU is still preferred work items7551 may be put on foreign CPUs. This debug option7552 forces round-robin CPU selection to flush out7553 usages which depend on the now broken guarantee.7554 When enabled, memory and cache locality will be7555 impacted.7556 7557 writecombine= [LOONGARCH,EARLY] Control the MAT (Memory Access7558 Type) of ioremap_wc().7559 7560 on - Enable writecombine, use WUC for ioremap_wc()7561 off - Disable writecombine, use SUC for ioremap_wc()7562 7563 x2apic_phys [X86-64,APIC,EARLY] Use x2apic physical mode instead of7564 default x2apic cluster mode on platforms7565 supporting x2apic.7566 7567 xen_512gb_limit [KNL,X86-64,XEN]7568 Restricts the kernel running paravirtualized under Xen7569 to use only up to 512 GB of RAM. The reason to do so is7570 crash analysis tools and Xen tools for doing domain7571 save/restore/migration must be enabled to handle larger7572 domains.7573 7574 xen_emul_unplug= [HW,X86,XEN,EARLY]7575 Unplug Xen emulated devices7576 Format: [unplug0,][unplug1]7577 ide-disks -- unplug primary master IDE devices7578 aux-ide-disks -- unplug non-primary-master IDE devices7579 nics -- unplug network devices7580 all -- unplug all emulated devices (NICs and IDE disks)7581 unnecessary -- unplugging emulated devices is7582 unnecessary even if the host did not respond to7583 the unplug protocol7584 never -- do not unplug even if version check succeeds7585 7586 xen_legacy_crash [X86,XEN,EARLY]7587 Crash from Xen panic notifier, without executing late7588 panic() code such as dumping handler.7589 7590 xen_mc_debug [X86,XEN,EARLY]7591 Enable multicall debugging when running as a Xen PV guest.7592 Enabling this feature will reduce performance a little7593 bit, so it should only be enabled for obtaining extended7594 debug data in case of multicall errors.7595 7596 xen_msr_safe= [X86,XEN,EARLY]7597 Format: <bool>7598 Select whether to always use non-faulting (safe) MSR7599 access functions when running as Xen PV guest. The7600 default value is controlled by CONFIG_XEN_PV_MSR_SAFE.7601 7602 xen_nopv [X86]7603 Disables the PV optimizations forcing the HVM guest to7604 run as generic HVM guest with no PV drivers.7605 This option is obsoleted by the "nopv" option, which7606 has equivalent effect for XEN platform.7607 7608 xen_no_vector_callback7609 [KNL,X86,XEN,EARLY] Disable the vector callback for Xen7610 event channel interrupts.7611 7612 xen_scrub_pages= [XEN]7613 Boolean option to control scrubbing pages before giving them back7614 to Xen, for use by other domains. Can be also changed at runtime7615 with /sys/devices/system/xen_memory/xen_memory0/scrub_pages.7616 Default value controlled with CONFIG_XEN_SCRUB_PAGES_DEFAULT.7617 7618 xen_timer_slop= [X86-64,XEN,EARLY]7619 Set the timer slop (in nanoseconds) for the virtual Xen7620 timers (default is 100000). This adjusts the minimum7621 delta of virtualized Xen timers, where lower values7622 improve timer resolution at the expense of processing7623 more timer interrupts.7624 7625 xen.balloon_boot_timeout= [XEN]7626 The time (in seconds) to wait before giving up to boot7627 in case initial ballooning fails to free enough memory.7628 Applies only when running as HVM or PVH guest and7629 started with less memory configured than allowed at7630 max. Default is 180.7631 7632 xen.event_eoi_delay= [XEN]7633 How long to delay EOI handling in case of event7634 storms (jiffies). Default is 10.7635 7636 xen.event_loop_timeout= [XEN]7637 After which time (jiffies) the event handling loop7638 should start to delay EOI handling. Default is 2.7639 7640 xen.fifo_events= [XEN]7641 Boolean parameter to disable using fifo event handling7642 even if available. Normally fifo event handling is7643 preferred over the 2-level event handling, as it is7644 fairer and the number of possible event channels is7645 much higher. Default is on (use fifo events).7646 7647 xirc2ps_cs= [NET,PCMCIA]7648 Format:7649 <irq>,<irq_mask>,<io>,<full_duplex>,<do_sound>,<lockup_hack>[,<irq2>[,<irq3>[,<irq4>]]]7650 7651 xive= [PPC]7652 By default on POWER9 and above, the kernel will7653 natively use the XIVE interrupt controller. This option7654 allows the fallback firmware mode to be used:7655 7656 off Fallback to firmware control of XIVE interrupt7657 controller on both pseries and powernv7658 platforms. Only useful on POWER9 and above.7659 7660 xive.store-eoi=off [PPC]7661 By default on POWER10 and above, the kernel will use7662 stores for EOI handling when the XIVE interrupt mode7663 is active. This option allows the XIVE driver to use7664 loads instead, as on POWER9.7665 7666 xhci-hcd.quirks [USB,KNL]7667 A hex value specifying bitmask with supplemental xhci7668 host controller quirks. Meaning of each bit can be7669 consulted in header drivers/usb/host/xhci.h.7670 7671 xmon [PPC,EARLY]7672 Format: { early | on | rw | ro | off }7673 Controls if xmon debugger is enabled. Default is off.7674 Passing only "xmon" is equivalent to "xmon=early".7675 early Call xmon as early as possible on boot; xmon7676 debugger is called from setup_arch().7677 on xmon debugger hooks will be installed so xmon7678 is only called on a kernel crash. Default mode,7679 i.e. either "ro" or "rw" mode, is controlled7680 with CONFIG_XMON_DEFAULT_RO_MODE.7681 rw xmon debugger hooks will be installed so xmon7682 is called only on a kernel crash, mode is write,7683 meaning SPR registers, memory and, other data7684 can be written using xmon commands.7685 ro same as "rw" option above but SPR registers,7686 memory, and other data can't be written using7687 xmon commands.7688 off xmon is disabled.7689