1 ACPI ACPI support is enabled. 2 AGP AGP (Accelerated Graphics Port) is enabled. 3 ALSA ALSA sound support is enabled. 4 APIC APIC support is enabled. 5 APM Advanced Power Management support is enabled. 6 APPARMOR AppArmor support is enabled. 7 ARM ARM architecture is enabled. 8 ARM64 ARM64 architecture is enabled. 9 CLK Common clock infrastructure is enabled. 10 CMA Contiguous Memory Area support is enabled. 11 CPU_FREQ CPU frequency scaling is enabled. 12 CPU_IDLE CPU idle time management is enabled. 13 DRM Direct Rendering Management support is enabled. 14 DYNAMIC_DEBUG Build in debug messages and enable them at runtime 15 EARLY Parameter processed too early to be embedded in initrd. 16 EDD BIOS Enhanced Disk Drive Services (EDD) is enabled 17 EFI EFI Partitioning (GPT) is enabled 18 EVM Extended Verification Module 19 FB The frame buffer device is enabled. 20 FTRACE Function tracing enabled. 21 GCOV GCOV profiling is enabled. 22 HIBERNATION HIBERNATION is enabled. 23 HW Appropriate hardware is enabled. 24 HYPER_V HYPERV support is enabled. 25 IMA Integrity measurement architecture is enabled. 26 IP_PNP IP DHCP, BOOTP, or RARP is enabled. 27 IPV6 IPv6 support is enabled. 28 ISAPNP ISA PnP code is enabled. 29 ISOL CPU Isolation is enabled. 30 JOY Appropriate joystick support is enabled. 31 KGDB Kernel debugger support is enabled. 32 KVM Kernel Virtual Machine support is enabled. 33 LIBATA Libata driver is enabled 34 LOONGARCH LoongArch architecture is enabled. 35 LOOP Loopback device support is enabled. 36 LP Printer support is enabled. 37 M68k M68k architecture is enabled. 38 These options have more detailed description inside of 39 Documentation/arch/m68k/kernel-options.rst. 40 MIPS MIPS architecture is enabled. 41 MOUSE Appropriate mouse support is enabled. 42 MSI Message Signaled Interrupts (PCI). 43 MTD MTD (Memory Technology Device) support is enabled. 44 NET Appropriate network support is enabled. 45 NFS Appropriate NFS support is enabled. 46 NUMA NUMA support is enabled. 47 OF Devicetree is enabled. 48 PARISC The PA-RISC architecture is enabled. 49 PCI PCI bus support is enabled. 50 PCIE PCI Express support is enabled. 51 PCMCIA The PCMCIA subsystem is enabled. 52 PM Power Management support is enabled. 53 PNP Plug & Play support is enabled. 54 PPC PowerPC architecture is enabled. 55 PPT Parallel port support is enabled. 56 PS2 Appropriate PS/2 support is enabled. 57 PV_OPS A paravirtualized kernel is enabled. 58 RAM RAM disk support is enabled. 59 RDT Intel Resource Director Technology. 60 RISCV RISCV architecture is enabled. 61 S390 S390 architecture is enabled. 62 SCSI Appropriate SCSI support is enabled. 63 A lot of drivers have their options described inside 64 the Documentation/scsi/ sub-directory. 65 SDW SoundWire support is enabled. 66 SECURITY Different security models are enabled. 67 SELINUX SELinux support is enabled. 68 SERIAL Serial support is enabled. 69 SH SuperH architecture is enabled. 70 SMP The kernel is an SMP kernel. 71 SPARC Sparc architecture is enabled. 72 SUSPEND System suspend states are enabled. 73 SWSUSP Software suspend (hibernation) is enabled. 74 TPM TPM drivers are enabled. 75 UMS USB Mass Storage support is enabled. 76 USB USB support is enabled. 77 NVME NVMe support is enabled 78 USBHID USB Human Interface Device support is enabled. 79 V4L Video For Linux support is enabled. 80 VGA The VGA console has been enabled. 81 VMMIO Driver for memory mapped virtio devices is enabled. 82 VT Virtual terminal support is enabled. 83 WDT Watchdog support is enabled. 84 X86-32 X86-32, aka i386 architecture is enabled. 85 X86-64 X86-64 architecture is enabled. 86 X86 Either 32-bit or 64-bit x86 (same as X86-32+X86-64) 87 X86_UV SGI UV support is enabled. 88 XEN Xen support is enabled 89 XTENSA xtensa architecture is enabled. 90 91In addition, the following text indicates that the option 92 93 BOOT Is a boot loader parameter. 94 BUGS= Relates to possible processor bugs on the said processor. 95 KNL Is a kernel start-up parameter. 96 97 98Kernel parameters 99 100 accept_memory= [MM] 101 Format: { eager | lazy } 102 default: lazy 103 By default, unaccepted memory is accepted lazily to 104 avoid prolonged boot times. The lazy option will add 105 some runtime overhead until all memory is eventually 106 accepted. In most cases the overhead is negligible. 107 For some workloads or for debugging purposes 108 accept_memory=eager can be used to accept all memory 109 at once during boot. 110 111 acpi= [HW,ACPI,X86,ARM64,RISCV64,EARLY] 112 Advanced Configuration and Power Interface 113 Format: { force | on | off | strict | noirq | rsdt | 114 copy_dsdt | nospcr } 115 force -- enable ACPI if default was off 116 on -- enable ACPI but allow fallback to DT [arm64,riscv64] 117 off -- disable ACPI if default was on 118 noirq -- do not use ACPI for IRQ routing 119 strict -- Be less tolerant of platforms that are not 120 strictly ACPI specification compliant. 121 rsdt -- prefer RSDT over (default) XSDT 122 copy_dsdt -- copy DSDT to memory 123 nocmcff -- Disable firmware first mode for corrected 124 errors. This disables parsing the HEST CMC error 125 source to check if firmware has set the FF flag. This 126 may result in duplicate corrected error reports. 127 nospcr -- disable console in ACPI SPCR table as 128 default _serial_ console on ARM64 129 spcr -- enable console in ACPI SPCR table as 130 default _serial_ console on x86 131 For ARM64, ONLY "acpi=off", "acpi=on", "acpi=force" or 132 "acpi=nospcr" are available 133 For RISCV64, ONLY "acpi=off", "acpi=on" or "acpi=force" 134 are available 135 136 See also Documentation/power/runtime_pm.rst, pci=noacpi 137 138 acpi_apic_instance= [ACPI,IOAPIC,EARLY] 139 Format: <int> 140 2: use 2nd APIC table, if available 141 1,0: use 1st APIC table 142 default: 0 143 144 acpi_backlight= [HW,ACPI] 145 { vendor | video | native | none } 146 If set to vendor, prefer vendor-specific driver 147 (e.g. thinkpad_acpi, sony_acpi, etc.) instead 148 of the ACPI video.ko driver. 149 If set to video, use the ACPI video.ko driver. 150 If set to native, use the device's native backlight mode. 151 If set to none, disable the ACPI backlight interface. 152 153 acpi_force_32bit_fadt_addr [ACPI,EARLY] 154 force FADT to use 32 bit addresses rather than the 155 64 bit X_* addresses. Some firmware have broken 64 156 bit addresses for force ACPI ignore these and use 157 the older legacy 32 bit addresses. 158 159 acpica_no_return_repair [HW, ACPI] 160 Disable AML predefined validation mechanism 161 This mechanism can repair the evaluation result to make 162 the return objects more ACPI specification compliant. 163 This option is useful for developers to identify the 164 root cause of an AML interpreter issue when the issue 165 has something to do with the repair mechanism. 166 167 acpi.debug_layer= [HW,ACPI,ACPI_DEBUG] 168 acpi.debug_level= [HW,ACPI,ACPI_DEBUG] 169 Format: <int> 170 CONFIG_ACPI_DEBUG must be enabled to produce any ACPI 171 debug output. Bits in debug_layer correspond to a 172 _COMPONENT in an ACPI source file, e.g., 173 #define _COMPONENT ACPI_EVENTS 174 Bits in debug_level correspond to a level in 175 ACPI_DEBUG_PRINT statements, e.g., 176 ACPI_DEBUG_PRINT((ACPI_DB_INFO, ... 177 The debug_level mask defaults to "info". See 178 Documentation/firmware-guide/acpi/debug.rst for more information about 179 debug layers and levels. 180 181 Enable processor driver info messages: 182 acpi.debug_layer=0x20000000 183 Enable AML "Debug" output, i.e., stores to the Debug 184 object while interpreting AML: 185 acpi.debug_layer=0xffffffff acpi.debug_level=0x2 186 Enable all messages related to ACPI hardware: 187 acpi.debug_layer=0x2 acpi.debug_level=0xffffffff 188 189 Some values produce so much output that the system is 190 unusable. The "log_buf_len" parameter may be useful 191 if you need to capture more output. 192 193 acpi.poweroff_on_fatal= [ACPI] 194 {0 | 1} 195 Causes the system to poweroff when the ACPI bytecode signals 196 a fatal error. The default value of this setting is 1. 197 Overriding this value should only be done for diagnosing 198 ACPI firmware problems, as the system might behave erratically 199 after having encountered a fatal ACPI error. 200 201 acpi_enforce_resources= [ACPI] 202 { strict | lax | no } 203 Check for resource conflicts between native drivers 204 and ACPI OperationRegions (SystemIO and SystemMemory 205 only). IO ports and memory declared in ACPI might be 206 used by the ACPI subsystem in arbitrary AML code and 207 can interfere with legacy drivers. 208 strict (default): access to resources claimed by ACPI 209 is denied; legacy drivers trying to access reserved 210 resources will fail to bind to device using them. 211 lax: access to resources claimed by ACPI is allowed; 212 legacy drivers trying to access reserved resources 213 will bind successfully but a warning message is logged. 214 no: ACPI OperationRegions are not marked as reserved, 215 no further checks are performed. 216 217 acpi_force_table_verification [HW,ACPI,EARLY] 218 Enable table checksum verification during early stage. 219 By default, this is disabled due to x86 early mapping 220 size limitation. 221 222 acpi_irq_balance [HW,ACPI] 223 ACPI will balance active IRQs 224 default in APIC mode 225 226 acpi_irq_nobalance [HW,ACPI] 227 ACPI will not move active IRQs (default) 228 default in PIC mode 229 230 acpi_irq_isa= [HW,ACPI] If irq_balance, mark listed IRQs used by ISA 231 Format: <irq>,<irq>... 232 233 acpi_irq_pci= [HW,ACPI] If irq_balance, clear listed IRQs for 234 use by PCI 235 Format: <irq>,<irq>... 236 237 acpi_mask_gpe= [HW,ACPI] 238 Due to the existence of _Lxx/_Exx, some GPEs triggered 239 by unsupported hardware/firmware features can result in 240 GPE floodings that cannot be automatically disabled by 241 the GPE dispatcher. 242 This facility can be used to prevent such uncontrolled 243 GPE floodings. 244 Format: <byte> or <bitmap-list> 245 246 acpi_no_auto_serialize [HW,ACPI] 247 Disable auto-serialization of AML methods 248 AML control methods that contain the opcodes to create 249 named objects will be marked as "Serialized" by the 250 auto-serialization feature. 251 This feature is enabled by default. 252 This option allows to turn off the feature. 253 254 acpi_no_memhotplug [ACPI] Disable memory hotplug. Useful for kdump 255 kernels. 256 257 acpi_no_static_ssdt [HW,ACPI,EARLY] 258 Disable installation of static SSDTs at early boot time 259 By default, SSDTs contained in the RSDT/XSDT will be 260 installed automatically and they will appear under 261 /sys/firmware/acpi/tables. 262 This option turns off this feature. 263 Note that specifying this option does not affect 264 dynamic table installation which will install SSDT 265 tables to /sys/firmware/acpi/tables/dynamic. 266 267 acpi_no_watchdog [HW,ACPI,WDT] 268 Ignore the ACPI-based watchdog interface (WDAT) and let 269 a native driver control the watchdog device instead. 270 271 acpi_rsdp= [ACPI,EFI,KEXEC,EARLY] 272 Pass the RSDP address to the kernel, mostly used 273 on machines running EFI runtime service to boot the 274 second kernel for kdump. 275 276 acpi_os_name= [HW,ACPI] Tell ACPI BIOS the name of the OS 277 Format: To spoof as Windows 98: ="Microsoft Windows" 278 279 acpi_rev_override [ACPI] Override the _REV object to return 5 (instead 280 of 2 which is mandated by ACPI 6) as the supported ACPI 281 specification revision (when using this switch, it may 282 be necessary to carry out a cold reboot _twice_ in a 283 row to make it take effect on the platform firmware). 284 285 acpi_osi= [HW,ACPI] Modify list of supported OS interface strings 286 acpi_osi="string1" # add string1 287 acpi_osi="!string2" # remove string2 288 acpi_osi=!* # remove all strings 289 acpi_osi=! # disable all built-in OS vendor 290 strings 291 acpi_osi=!! # enable all built-in OS vendor 292 strings 293 acpi_osi= # disable all strings 294 295 'acpi_osi=!' can be used in combination with single or 296 multiple 'acpi_osi="string1"' to support specific OS 297 vendor string(s). Note that such command can only 298 affect the default state of the OS vendor strings, thus 299 it cannot affect the default state of the feature group 300 strings and the current state of the OS vendor strings, 301 specifying it multiple times through kernel command line 302 is meaningless. This command is useful when one do not 303 care about the state of the feature group strings which 304 should be controlled by the OSPM. 305 Examples: 306 1. 'acpi_osi=! acpi_osi="Windows 2000"' is equivalent 307 to 'acpi_osi="Windows 2000" acpi_osi=!', they all 308 can make '_OSI("Windows 2000")' TRUE. 309 310 'acpi_osi=' cannot be used in combination with other 311 'acpi_osi=' command lines, the _OSI method will not 312 exist in the ACPI namespace. NOTE that such command can 313 only affect the _OSI support state, thus specifying it 314 multiple times through kernel command line is also 315 meaningless. 316 Examples: 317 1. 'acpi_osi=' can make 'CondRefOf(_OSI, Local1)' 318 FALSE. 319 320 'acpi_osi=!*' can be used in combination with single or 321 multiple 'acpi_osi="string1"' to support specific 322 string(s). Note that such command can affect the 323 current state of both the OS vendor strings and the 324 feature group strings, thus specifying it multiple times 325 through kernel command line is meaningful. But it may 326 still not able to affect the final state of a string if 327 there are quirks related to this string. This command 328 is useful when one want to control the state of the 329 feature group strings to debug BIOS issues related to 330 the OSPM features. 331 Examples: 332 1. 'acpi_osi="Module Device" acpi_osi=!*' can make 333 '_OSI("Module Device")' FALSE. 334 2. 'acpi_osi=!* acpi_osi="Module Device"' can make 335 '_OSI("Module Device")' TRUE. 336 3. 'acpi_osi=! acpi_osi=!* acpi_osi="Windows 2000"' is 337 equivalent to 338 'acpi_osi=!* acpi_osi=! acpi_osi="Windows 2000"' 339 and 340 'acpi_osi=!* acpi_osi="Windows 2000" acpi_osi=!', 341 they all will make '_OSI("Windows 2000")' TRUE. 342 343 acpi_pm_good [X86] 344 Override the pmtimer bug detection: force the kernel 345 to assume that this machine's pmtimer latches its value 346 and always returns good values. 347 348 acpi_sci= [HW,ACPI,EARLY] ACPI System Control Interrupt trigger mode 349 Format: { level | edge | high | low } 350 351 acpi_skip_timer_override [HW,ACPI,EARLY] 352 Recognize and ignore IRQ0/pin2 Interrupt Override. 353 For broken nForce2 BIOS resulting in XT-PIC timer. 354 355 acpi_sleep= [HW,ACPI] Sleep options 356 Format: { s3_bios, s3_mode, s3_beep, s4_hwsig, 357 s4_nohwsig, old_ordering, nonvs, 358 sci_force_enable, nobl } 359 See Documentation/power/video.rst for information on 360 s3_bios and s3_mode. 361 s3_beep is for debugging; it makes the PC's speaker beep 362 as soon as the kernel's real-mode entry point is called. 363 s4_hwsig causes the kernel to check the ACPI hardware 364 signature during resume from hibernation, and gracefully 365 refuse to resume if it has changed. This complies with 366 the ACPI specification but not with reality, since 367 Windows does not do this and many laptops do change it 368 on docking. So the default behaviour is to allow resume 369 and simply warn when the signature changes, unless the 370 s4_hwsig option is enabled. 371 s4_nohwsig prevents ACPI hardware signature from being 372 used (or even warned about) during resume. 373 old_ordering causes the ACPI 1.0 ordering of the _PTS 374 control method, with respect to putting devices into 375 low power states, to be enforced (the ACPI 2.0 ordering 376 of _PTS is used by default). 377 nonvs prevents the kernel from saving/restoring the 378 ACPI NVS memory during suspend/hibernation and resume. 379 sci_force_enable causes the kernel to set SCI_EN directly 380 on resume from S1/S3 (which is against the ACPI spec, 381 but some broken systems don't work without it). 382 nobl causes the internal blacklist of systems known to 383 behave incorrectly in some ways with respect to system 384 suspend and resume to be ignored (use wisely). 385 386 acpi_use_timer_override [HW,ACPI,EARLY] 387 Use timer override. For some broken Nvidia NF5 boards 388 that require a timer override, but don't have HPET 389 390 add_efi_memmap [EFI,X86,EARLY] Include EFI memory map in 391 kernel's map of available physical RAM. 392 393 agp= [AGP] 394 { off | try_unsupported } 395 off: disable AGP support 396 try_unsupported: try to drive unsupported chipsets 397 (may crash computer or cause data corruption) 398 399 ALSA [HW,ALSA] 400 See Documentation/sound/alsa-configuration.rst 401 402 alignment= [KNL,ARM] 403 Allow the default userspace alignment fault handler 404 behaviour to be specified. Bit 0 enables warnings, 405 bit 1 enables fixups, and bit 2 sends a segfault. 406 407 align_va_addr= [X86-64] 408 Align virtual addresses by clearing slice [14:12] when 409 allocating a VMA at process creation time. This option 410 gives you up to 3% performance improvement on AMD F15h 411 machines (where it is enabled by default) for a 412 CPU-intensive style benchmark, and it can vary highly in 413 a microbenchmark depending on workload and compiler. 414 415 32: only for 32-bit processes 416 64: only for 64-bit processes 417 on: enable for both 32- and 64-bit processes 418 off: disable for both 32- and 64-bit processes 419 420 alloc_snapshot [FTRACE] 421 Allocate the ftrace snapshot buffer on boot up when the 422 main buffer is allocated. This is handy if debugging 423 and you need to use tracing_snapshot() on boot up, and 424 do not want to use tracing_snapshot_alloc() as it needs 425 to be done where GFP_KERNEL allocations are allowed. 426 427 allow_mismatched_32bit_el0 [ARM64,EARLY] 428 Allow execve() of 32-bit applications and setting of the 429 PER_LINUX32 personality on systems where only a strict 430 subset of the CPUs support 32-bit EL0. When this 431 parameter is present, the set of CPUs supporting 32-bit 432 EL0 is indicated by /sys/devices/system/cpu/aarch32_el0 433 and hot-unplug operations may be restricted. 434 435 See Documentation/arch/arm64/asymmetric-32bit.rst for more 436 information. 437 438 amd_iommu= [HW,X86-64] 439 Pass parameters to the AMD IOMMU driver in the system. 440 Possible values are: 441 fullflush - Deprecated, equivalent to iommu.strict=1 442 off - do not initialize any AMD IOMMU found in 443 the system 444 force_isolation - Force device isolation for all 445 devices. The IOMMU driver is not 446 allowed anymore to lift isolation 447 requirements as needed. This option 448 does not override iommu=pt 449 force_enable - Force enable the IOMMU on platforms known 450 to be buggy with IOMMU enabled. Use this 451 option with care. 452 pgtbl_v1 - Use v1 page table for DMA-API (Default). 453 pgtbl_v2 - Use v2 page table for DMA-API. 454 irtcachedis - Disable Interrupt Remapping Table (IRT) caching. 455 nohugepages - Limit page-sizes used for v1 page-tables 456 to 4 KiB. 457 v2_pgsizes_only - Limit page-sizes used for v1 page-tables 458 to 4KiB/2Mib/1GiB. 459 460 461 amd_iommu_dump= [HW,X86-64] 462 Enable AMD IOMMU driver option to dump the ACPI table 463 for AMD IOMMU. With this option enabled, AMD IOMMU 464 driver will print ACPI tables for AMD IOMMU during 465 IOMMU initialization. 466 467 amd_iommu_intr= [HW,X86-64] 468 Specifies one of the following AMD IOMMU interrupt 469 remapping modes: 470 legacy - Use legacy interrupt remapping mode. 471 vapic - Use virtual APIC mode, which allows IOMMU 472 to inject interrupts directly into guest. 473 This mode requires kvm-amd.avic=1. 474 (Default when IOMMU HW support is present.) 475 476 amd_pstate= [X86,EARLY] 477 disable 478 Do not enable amd_pstate as the default 479 scaling driver for the supported processors 480 passive 481 Use amd_pstate with passive mode as a scaling driver. 482 In this mode autonomous selection is disabled. 483 Driver requests a desired performance level and platform 484 tries to match the same performance level if it is 485 satisfied by guaranteed performance level. 486 active 487 Use amd_pstate_epp driver instance as the scaling driver, 488 driver provides a hint to the hardware if software wants 489 to bias toward performance (0x0) or energy efficiency (0xff) 490 to the CPPC firmware. then CPPC power algorithm will 491 calculate the runtime workload and adjust the realtime cores 492 frequency. 493 guided 494 Activate guided autonomous mode. Driver requests minimum and 495 maximum performance level and the platform autonomously 496 selects a performance level in this range and appropriate 497 to the current workload. 498 499 amd_prefcore= 500 [X86] 501 disable 502 Disable amd-pstate preferred core. 503 504 amijoy.map= [HW,JOY] Amiga joystick support 505 Map of devices attached to JOY0DAT and JOY1DAT 506 Format: <a>,<b> 507 See also Documentation/input/joydev/joystick.rst 508 509 analog.map= [HW,JOY] Analog joystick and gamepad support 510 Specifies type or capabilities of an analog joystick 511 connected to one of 16 gameports 512 Format: <type1>,<type2>,..<type16> 513 514 apc= [HW,SPARC] 515 Power management functions (SPARCstation-4/5 + deriv.) 516 Format: noidle 517 Disable APC CPU standby support. SPARCstation-Fox does 518 not play well with APC CPU idle - disable it if you have 519 APC and your system crashes randomly. 520 521 apic [APIC,X86-64] Use IO-APIC. Default. 522 523 apic= [APIC,X86,EARLY] Advanced Programmable Interrupt Controller 524 Change the output verbosity while booting 525 Format: { quiet (default) | verbose | debug } 526 Change the amount of debugging information output 527 when initialising the APIC and IO-APIC components. 528 529 apic_extnmi= [APIC,X86,EARLY] External NMI delivery setting 530 Format: { bsp (default) | all | none } 531 bsp: External NMI is delivered only to CPU 0 532 all: External NMIs are broadcast to all CPUs as a 533 backup of CPU 0 534 none: External NMI is masked for all CPUs. This is 535 useful so that a dump capture kernel won't be 536 shot down by NMI 537 538 apicpmtimer Do APIC timer calibration using the pmtimer. Implies 539 apicmaintimer. Useful when your PIT timer is totally 540 broken. 541 542 autoconf= [IPV6] 543 See Documentation/networking/ipv6.rst. 544 545 apm= [APM] Advanced Power Management 546 See header of arch/x86/kernel/apm_32.c. 547 548 apparmor= [APPARMOR] Disable or enable AppArmor at boot time 549 Format: { "0" | "1" } 550 See security/apparmor/Kconfig help text 551 0 -- disable. 552 1 -- enable. 553 Default value is set via kernel config option. 554 555 arm64.no32bit_el0 [ARM64] Unconditionally disable the execution of 556 32 bit applications. 557 558 arm64.nobti [ARM64] Unconditionally disable Branch Target 559 Identification support 560 561 arm64.nogcs [ARM64] Unconditionally disable Guarded Control Stack 562 support 563 564 arm64.nohaft [ARM64] Unconditionally disable Hardware managed Access 565 Flag for Table descriptors support 566 567 arm64.nomops [ARM64] Unconditionally disable Memory Copy and Memory 568 Set instructions support 569 570 arm64.nompam [ARM64] Unconditionally disable Memory Partitioning And 571 Monitoring support. Only for a machine that does not 572 boot without it. See Documentation/arch/arm64/mpam.rst 573 574 arm64.nomte [ARM64] Unconditionally disable Memory Tagging Extension 575 support 576 577 arm64.nopauth [ARM64] Unconditionally disable Pointer Authentication 578 support 579 580 arm64.nosme [ARM64] Unconditionally disable Scalable Matrix 581 Extension support 582 583 arm64.nosve [ARM64] Unconditionally disable Scalable Vector 584 Extension support 585 586 arm64.nowfxt [ARM64] Unconditionally disable Wait For Event with 587 Timeout and Wait For Interrupt with Timeout support 588 589 ataflop= [HW,M68k] 590 591 atkbd.extra= [HW] Enable extra LEDs and keys on IBM RapidAccess, 592 EzKey and similar keyboards 593 594 atkbd.reset= [HW] Reset keyboard during initialization 595 596 atkbd.set= [HW] Select keyboard code set 597 Format: <int> (2 = AT (default), 3 = PS/2) 598 599 atkbd.scroll= [HW] Enable scroll wheel on MS Office and similar 600 keyboards 601 602 atkbd.softraw= [HW] Choose between synthetic and real raw mode 603 Format: <bool> (0 = real, 1 = synthetic (default)) 604 605 atkbd.softrepeat= [HW] 606 Use software keyboard repeat 607 608 audit= [KNL] Enable the audit sub-system 609 Format: { "0" | "1" | "off" | "on" } 610 0 | off - kernel audit is disabled and can not be 611 enabled until the next reboot 612 unset - kernel audit is initialized but disabled and 613 will be fully enabled by the userspace auditd. 614 1 | on - kernel audit is initialized and partially 615 enabled, storing at most audit_backlog_limit 616 messages in RAM until it is fully enabled by the 617 userspace auditd. 618 Default: unset 619 620 audit_backlog_limit= [KNL] Set the audit queue size limit. 621 Format: <int> (must be >=0) 622 Default: 64 623 624 bdev_allow_write_mounted= 625 Format: <bool> 626 Control the ability to open a mounted block device 627 for writing, i.e., allow / disallow writes that bypass 628 the FS. This was implemented as a means to prevent 629 fuzzers from crashing the kernel by overwriting the 630 metadata underneath a mounted FS without its awareness. 631 This also prevents destructive formatting of mounted 632 filesystems by naive storage tooling that don't use 633 O_EXCL. Default is Y and can be changed through the 634 Kconfig option CONFIG_BLK_DEV_WRITE_MOUNTED. 635 636 bert_disable [ACPI] 637 Disable BERT OS support on buggy BIOSes. 638 639 bgrt_disable [ACPI,X86,EARLY] 640 Disable BGRT to avoid flickering OEM logo. 641 642 blkdevparts= Manual partition parsing of block device(s) for 643 embedded devices based on command line input. 644 See Documentation/block/cmdline-partition.rst 645 646 boot_delay= [KNL,EARLY] 647 Milliseconds to delay each printk during boot. 648 Only works if CONFIG_BOOT_PRINTK_DELAY is enabled, 649 and you may also have to specify "lpj=". Boot_delay 650 values larger than 10 seconds (10000) are assumed 651 erroneous and ignored. 652 Format: integer 653 654 bootconfig [KNL,EARLY] 655 Extended command line options can be added to an initrd 656 and this will cause the kernel to look for it. 657 658 See Documentation/admin-guide/bootconfig.rst 659 660 bttv.card= [HW,V4L] bttv (bt848 + bt878 based grabber cards) 661 bttv.radio= Most important insmod options are available as 662 kernel args too. 663 bttv.pll= See Documentation/admin-guide/media/bttv.rst 664 bttv.tuner= 665 666 bulk_remove=off [PPC] This parameter disables the use of the pSeries 667 firmware feature for flushing multiple hpte entries 668 at a time. 669 670 c101= [NET] Moxa C101 synchronous serial card 671 672 cachesize= [BUGS=X86-32] Override level 2 CPU cache size detection. 673 Sometimes CPU hardware bugs make them report the cache 674 size incorrectly. The kernel will attempt work arounds 675 to fix known problems, but for some CPUs it is not 676 possible to determine what the correct size should be. 677 This option provides an override for these situations. 678 679 carrier_timeout= 680 [NET] Specifies amount of time (in seconds) that 681 the kernel should wait for a network carrier. By default 682 it waits 120 seconds. 683 684 ca_keys= [KEYS] This parameter identifies a specific key(s) on 685 the system trusted keyring to be used for certificate 686 trust validation. 687 format: { id:<keyid> | builtin } 688 689 cca= [MIPS,EARLY] Override the kernel pages' cache coherency 690 algorithm. Accepted values range from 0 to 7 691 inclusive. See arch/mips/include/asm/pgtable-bits.h 692 for platform specific values (SB1, Loongson3 and 693 others). 694 695 ccw_timeout_log [S390] 696 See Documentation/arch/s390/common_io.rst for details. 697 698 cfi= [X86-64] Set Control Flow Integrity checking features 699 when CONFIG_FINEIBT is enabled. 700 Format: feature[,feature...] 701 Default: auto 702 703 auto: Use FineIBT if IBT available, otherwise kCFI. 704 Under FineIBT, enable "paranoid" mode when 705 FRED is not available. 706 off: Turn off CFI checking. 707 kcfi: Use kCFI (disable FineIBT). 708 fineibt: Use FineIBT (even if IBT not available). 709 norand: Do not re-randomize CFI hashes. 710 paranoid: Add caller hash checking under FineIBT. 711 bhi: Enable register poisoning to stop speculation 712 across FineIBT. (Disabled by default.) 713 warn: Do not enforce CFI checking: warn only. 714 debug: Report CFI initialization details. 715 716 cgroup_disable= [KNL] Disable a particular controller or optional feature 717 Format: {name of the controller(s) or feature(s) to disable} 718 The effects of cgroup_disable=foo are: 719 - foo isn't auto-mounted if you mount all cgroups in 720 a single hierarchy 721 - foo isn't visible as an individually mountable 722 subsystem 723 - if foo is an optional feature then the feature is 724 disabled and corresponding cgroup files are not 725 created 726 {Currently only "memory" controller deal with this and 727 cut the overhead, others just disable the usage. So 728 only cgroup_disable=memory is actually worthy} 729 Specifying "pressure" disables per-cgroup pressure 730 stall information accounting feature 731 732 cgroup_no_v1= [KNL] Disable cgroup controllers and named hierarchies in v1 733 Format: { { controller | "all" | "named" } 734 [,{ controller | "all" | "named" }...] } 735 Like cgroup_disable, but only applies to cgroup v1; 736 the blacklisted controllers remain available in cgroup2. 737 "all" blacklists all controllers and "named" disables 738 named mounts. Specifying both "all" and "named" disables 739 all v1 hierarchies. 740 741 cgroup_v1_proc= [KNL] Show also missing controllers in /proc/cgroups 742 Format: { "true" | "false" } 743 /proc/cgroups lists only v1 controllers by default. 744 This compatibility option enables listing also v2 745 controllers (whose v1 code is not compiled!), so that 746 semi-legacy software can check this file to decide 747 about usage of v2 (sic) controllers. 748 749 cgroup_favordynmods= [KNL] Enable or Disable favordynmods. 750 Format: { "true" | "false" } 751 Defaults to the value of CONFIG_CGROUP_FAVOR_DYNMODS. 752 753 cgroup.memory= [KNL] Pass options to the cgroup memory controller. 754 Format: <string> 755 nosocket -- Disable socket memory accounting. 756 nokmem -- Disable kernel memory accounting. 757 nobpf -- Disable BPF memory accounting. 758 759 check_pages= [MM,EARLY] Enable sanity checking of pages after 760 allocations / before freeing. This adds checks to catch 761 double-frees, use-after-frees, and other sources of 762 page corruption by inspecting page internals (flags, 763 mapcount/refcount, memcg_data, etc.). 764 Format: { "0" | "1" } 765 Default: 0 (1 if CONFIG_DEBUG_VM is set) 766 767 checkreqprot= [SELINUX] Set initial checkreqprot flag value. 768 Format: { "0" | "1" } 769 See security/selinux/Kconfig help text. 770 0 -- check protection applied by kernel (includes 771 any implied execute protection). 772 1 -- check protection requested by application. 773 Default value is set via a kernel config option. 774 Value can be changed at runtime via 775 /sys/fs/selinux/checkreqprot. 776 Setting checkreqprot to 1 is deprecated. 777 778 cio_ignore= [S390] 779 See Documentation/arch/s390/common_io.rst for details. 780 781 clk_ignore_unused 782 [CLK] 783 Prevents the clock framework from automatically gating 784 clocks that have not been explicitly enabled by a Linux 785 device driver but are enabled in hardware at reset or 786 by the bootloader/firmware. Note that this does not 787 force such clocks to be always-on nor does it reserve 788 those clocks in any way. This parameter is useful for 789 debug and development, but should not be needed on a 790 platform with proper driver support. For more 791 information, see Documentation/driver-api/clk.rst. 792 793 clock= [BUGS=X86-32, HW] gettimeofday clocksource override. 794 [Deprecated] 795 Forces specified clocksource (if available) to be used 796 when calculating gettimeofday(). If specified 797 clocksource is not available, it defaults to PIT. 798 Format: { pit | tsc | cyclone | pmtmr } 799 800 clocksource= Override the default clocksource 801 Format: <string> 802 Override the default clocksource and use the clocksource 803 with the name specified. 804 Some clocksource names to choose from, depending on 805 the platform: 806 [all] jiffies (this is the base, fallback clocksource) 807 [ACPI] acpi_pm 808 [ARM] imx_timer1,OSTS,netx_timer,mpu_timer2, 809 pxa_timer,timer3,32k_counter,timer0_1 810 [X86-32] pit,hpet,tsc; 811 scx200_hrt on Geode; cyclone on IBM x440 812 [MIPS] MIPS 813 [PARISC] cr16 814 [S390] tod 815 [SH] SuperH 816 [SPARC64] tick 817 [X86-64] hpet,tsc 818 819 clocksource.arm_arch_timer.evtstrm= 820 [ARM,ARM64,EARLY] 821 Format: <bool> 822 Enable/disable the eventstream feature of the ARM 823 architected timer so that code using WFE-based polling 824 loops can be debugged more effectively on production 825 systems. 826 827 clocksource.verify_n_cpus= [KNL] 828 Limit the number of CPUs checked for clocksources 829 marked with CLOCK_SOURCE_VERIFY_PERCPU that 830 are marked unstable due to excessive skew. 831 A negative value says to check all CPUs, while 832 zero says not to check any. Values larger than 833 nr_cpu_ids are silently truncated to nr_cpu_ids. 834 The actual CPUs are chosen randomly, with 835 no replacement if the same CPU is chosen twice. 836 837 clocksource-wdtest.holdoff= [KNL] 838 Set the time in seconds that the clocksource 839 watchdog test waits before commencing its tests. 840 Defaults to zero when built as a module and to 841 10 seconds when built into the kernel. 842 843 cma=nn[MG]@[start[MG][-end[MG]]] 844 [KNL,CMA,EARLY] 845 Sets the size of kernel global memory area for 846 contiguous memory allocations and optionally the 847 placement constraint by the physical address range of 848 memory allocations. A value of 0 disables CMA 849 altogether. For more information, see 850 kernel/dma/contiguous.c 851 852 cma_pernuma=nn[MG] 853 [KNL,CMA,EARLY] 854 Sets the size of kernel per-numa memory area for 855 contiguous memory allocations. A value of 0 disables 856 per-numa CMA altogether. And If this option is not 857 specified, the default value is 0. 858 With per-numa CMA enabled, DMA users on node nid will 859 first try to allocate buffer from the pernuma area 860 which is located in node nid, if the allocation fails, 861 they will fallback to the global default memory area. 862 863 numa_cma=<node>:nn[MG][,<node>:nn[MG]] 864 [KNL,CMA,EARLY] 865 Sets the size of kernel numa memory area for 866 contiguous memory allocations. It will reserve CMA 867 area for the specified node. 868 869 If it is setup together with upper 'cmd_pernuma=' 870 (unlikely), its size setting takes priority for the 871 specified numa nodes. 872 873 With numa CMA enabled, DMA users on node nid will 874 first try to allocate buffer from the numa area 875 which is located in node nid, if the allocation fails, 876 they will fallback to the global default memory area. 877 878 cmo_free_hint= [PPC] Format: { yes | no } 879 Specify whether pages are marked as being inactive 880 when they are freed. This is used in CMO environments 881 to determine OS memory pressure for page stealing by 882 a hypervisor. 883 Default: yes 884 885 coherent_pool=nn[KMG] [ARM,KNL,EARLY] 886 Sets the size of memory pool for coherent, atomic dma 887 allocations, by default set to 256K. 888 889 condev= [HW,S390] console device 890 conmode= 891 892 con3215_drop= [S390,EARLY] 3215 console drop mode. 893 Format: y|n|Y|N|1|0 894 When set to true, drop data on the 3215 console when 895 the console buffer is full. In this case the 896 operator using a 3270 terminal emulator (for example 897 x3270) does not have to enter the clear key for the 898 console output to advance and the kernel to continue. 899 This leads to a much faster boot time when a 3270 900 terminal emulator is active. If no 3270 terminal 901 emulator is used, this parameter has no effect. 902 903 console= [KNL] Output console device and options. 904 905 tty<n> Use the virtual console device <n>. 906 907 ttyS<n>[,options] 908 ttyUSB0[,options] 909 Use the specified serial port. The options are of 910 the form "bbbbpnf", where "bbbb" is the baud rate, 911 "p" is parity ("n", "o", or "e"), "n" is number of 912 bits, and "f" is flow control ("r" for RTS or 913 omit it). Default is "9600n8". 914 915 See Documentation/admin-guide/serial-console.rst for more 916 information. See 917 Documentation/networking/netconsole.rst for an 918 alternative. 919 920 <DEVNAME>:<n>.<n>[,options] 921 Use the specified serial port on the serial core bus. 922 The addressing uses DEVNAME of the physical serial port 923 device, followed by the serial core controller instance, 924 and the serial port instance. The options are the same 925 as documented for the ttyS addressing above. 926 927 The mapping of the serial ports to the tty instances 928 can be viewed with: 929 930 $ ls -d /sys/bus/serial-base/devices/*:*.*/tty/* 931 /sys/bus/serial-base/devices/00:04:0.0/tty/ttyS0 932 933 In the above example, the console can be addressed with 934 console=00:04:0.0. Note that a console addressed this 935 way will only get added when the related device driver 936 is ready. The use of an earlycon parameter in addition to 937 the console may be desired for console output early on. 938 939 uart[8250],io,<addr>[,options] 940 uart[8250],mmio,<addr>[,options] 941 uart[8250],mmio16,<addr>[,options] 942 uart[8250],mmio32,<addr>[,options] 943 uart[8250],0x<addr>[,options] 944 Start an early, polled-mode console on the 8250/16550 945 UART at the specified I/O port or MMIO address, 946 switching to the matching ttyS device later. 947 MMIO inter-register address stride is either 8-bit 948 (mmio), 16-bit (mmio16), or 32-bit (mmio32). 949 If none of [io|mmio|mmio16|mmio32], <addr> is assumed 950 to be equivalent to 'mmio'. 'options' are specified in 951 the same format described for ttyS above; if unspecified, 952 the h/w is not re-initialized. 953 954 hvc<n> Use the hypervisor console device <n>. This is for 955 both Xen and PowerPC hypervisors. 956 957 { null | "" } 958 Use to disable console output, i.e., to have kernel 959 console messages discarded. 960 This must be the only console= parameter used on the 961 kernel command line. 962 963 If the device connected to the port is not a TTY but a braille 964 device, prepend "brl," before the device type, for instance 965 console=brl,ttyS0 966 For now, only VisioBraille is supported. 967 968 console_msg_format= 969 [KNL] Change console messages format 970 default 971 By default we print messages on consoles in 972 "[time stamp] text\n" format (time stamp may not be 973 printed, depending on CONFIG_PRINTK_TIME or 974 `printk_time' param). 975 syslog 976 Switch to syslog format: "<%u>[time stamp] text\n" 977 IOW, each message will have a facility and loglevel 978 prefix. The format is similar to one used by syslog() 979 syscall, or to executing "dmesg -S --raw" or to reading 980 from /proc/kmsg. 981 982 consoleblank= [KNL] The console blank (screen saver) timeout in 983 seconds. A value of 0 disables the blank timer. 984 Defaults to 0. 985 986 coredump_filter= 987 [KNL] Change the default value for 988 /proc/<pid>/coredump_filter. 989 See also Documentation/filesystems/proc.rst. 990 991 coresight_cpu_debug.enable 992 [ARM,ARM64] 993 Format: <bool> 994 Enable/disable the CPU sampling based debugging. 995 0: default value, disable debugging 996 1: enable debugging at boot time 997 998 cpcihp_generic= [HW,PCI] Generic port I/O CompactPCI driver 999 Format: 1000 <first_slot>,<last_slot>,<port>,<enum_bit>[,<debug>] 1001 1002 cpuidle.off=1 [CPU_IDLE] 1003 disable the cpuidle sub-system 1004 1005 cpuidle.governor= 1006 [CPU_IDLE] Name of the cpuidle governor to use. 1007 1008 cpufreq.off=1 [CPU_FREQ] 1009 disable the cpufreq sub-system 1010 1011 cpufreq.default_governor= 1012 [CPU_FREQ] Name of the default cpufreq governor or 1013 policy to use. This governor must be registered in the 1014 kernel before the cpufreq driver probes. 1015 1016 cpu_init_udelay=N 1017 [X86,EARLY] Delay for N microsec between assert and de-assert 1018 of APIC INIT to start processors. This delay occurs 1019 on every CPU online, such as boot, and resume from suspend. 1020 Default: 10000 1021 1022 cpuhp.parallel= 1023 [SMP] Enable/disable parallel bringup of secondary CPUs 1024 Format: <bool> 1025 Default is enabled if CONFIG_HOTPLUG_PARALLEL=y. Otherwise 1026 the parameter has no effect. 1027 1028 crash_kexec_post_notifiers 1029 Only jump to kdump kernel after running the panic 1030 notifiers and dumping kmsg. This option increases 1031 the risks of a kdump failure, since some panic 1032 notifiers can make the crashed kernel more unstable. 1033 In configurations where kdump may not be reliable, 1034 running the panic notifiers could allow collecting 1035 more data on dmesg, like stack traces from other CPUS 1036 or extra data dumped by panic_print. Note that some 1037 configurations enable this option unconditionally, 1038 like Hyper-V, PowerPC (fadump) and AMD SEV-SNP. 1039 1040 crashkernel=size[KMG][@offset[KMG]] 1041 [KNL,EARLY] Using kexec, Linux can switch to a 'crash kernel' 1042 upon panic. This parameter reserves the physical 1043 memory region [offset, offset + size] for that kernel 1044 image. If '@offset' is omitted, then a suitable offset 1045 is selected automatically. 1046 [KNL, X86-64, ARM64, RISCV, LoongArch] Select a region 1047 under 4G first, and fall back to reserve region above 1048 4G when '@offset' hasn't been specified. 1049 See Documentation/admin-guide/kdump/kdump.rst for further details. 1050 1051 crashkernel=range1:size1[,range2:size2,...][@offset] 1052 [KNL] Same as above, but depends on the memory 1053 in the running system. The syntax of range is 1054 start-[end] where start and end are both 1055 a memory unit (amount[KMG]). See also 1056 Documentation/admin-guide/kdump/kdump.rst for an example. 1057 1058 crashkernel=size[KMG],high 1059 [KNL, X86-64, ARM64, RISCV, LoongArch] range could be 1060 above 4G. 1061 Allow kernel to allocate physical memory region from top, 1062 so could be above 4G if system have more than 4G ram 1063 installed. Otherwise memory region will be allocated 1064 below 4G, if available. 1065 It will be ignored if crashkernel=X is specified. 1066 crashkernel=size[KMG],low 1067 [KNL, X86-64, ARM64, RISCV, LoongArch] range under 4G. 1068 When crashkernel=X,high is passed, kernel could allocate 1069 physical memory region above 4G, that cause second kernel 1070 crash on system that require some amount of low memory, 1071 e.g. swiotlb requires at least 64M+32K low memory, also 1072 enough extra low memory is needed to make sure DMA buffers 1073 for 32-bit devices won't run out. Kernel would try to allocate 1074 default size of memory below 4G automatically. The default 1075 size is platform dependent. 1076 --> x86: max(swiotlb_size_or_default() + 8MiB, 256MiB) 1077 --> arm64: 128MiB 1078 --> riscv: 128MiB 1079 --> loongarch: 128MiB 1080 This one lets the user specify own low range under 4G 1081 for second kernel instead. 1082 0: to disable low allocation. 1083 It will be ignored when crashkernel=X,high is not used 1084 or memory reserved is below 4G. 1085 crashkernel=size[KMG],cma 1086 [KNL, X86, ARM64, RISCV, PPC] Reserve additional crash 1087 kernel memory from CMA. This reservation is usable by 1088 the first system's userspace memory and kernel movable 1089 allocations (memory balloon, zswap). Pages allocated 1090 from this memory range will not be included in the vmcore 1091 so this should not be used if dumping of userspace memory 1092 is intended and it has to be expected that some movable 1093 kernel pages may be missing from the dump. 1094 1095 A standard crashkernel reservation, as described above, 1096 is still needed to hold the crash kernel and initrd. 1097 1098 This option increases the risk of a kdump failure: DMA 1099 transfers configured by the first kernel may end up 1100 corrupting the second kernel's memory. 1101 1102 This reservation method is intended for systems that 1103 can't afford to sacrifice enough memory for standard 1104 crashkernel reservation and where less reliable and 1105 possibly incomplete kdump is preferable to no kdump at 1106 all. 1107 1108 cryptomgr.notests 1109 [KNL] Disable crypto self-tests 1110 1111 cs89x0_dma= [HW,NET] 1112 Format: <dma> 1113 1114 cs89x0_media= [HW,NET] 1115 Format: { rj45 | aui | bnc } 1116 1117 csdlock_debug= [KNL] Enable or disable debug add-ons of cross-CPU 1118 function call handling. When switched on, 1119 additional debug data is printed to the console 1120 in case a hanging CPU is detected, and that 1121 CPU is pinged again in order to try to resolve 1122 the hang situation. The default value of this 1123 option depends on the CSD_LOCK_WAIT_DEBUG_DEFAULT 1124 Kconfig option. 1125 1126 dasd= [HW,NET] 1127 See header of drivers/s390/block/dasd_devmap.c. 1128 1129 db9.dev[2|3]= [HW,JOY] Multisystem joystick support via parallel port 1130 (one device per port) 1131 Format: <port#>,<type> 1132 See also Documentation/input/devices/joystick-parport.rst 1133 1134 debug [KNL,EARLY] Enable kernel debugging (events log level). 1135 1136 debug_boot_weak_hash 1137 [KNL,EARLY] Enable printing [hashed] pointers early in the 1138 boot sequence. If enabled, we use a weak hash instead 1139 of siphash to hash pointers. Use this option if you are 1140 seeing instances of '(___ptrval___)') and need to see a 1141 value (hashed pointer) instead. Cryptographically 1142 insecure, please do not use on production kernels. 1143 1144 debug_locks_verbose= 1145 [KNL] verbose locking self-tests 1146 Format: <int> 1147 Print debugging info while doing the locking API 1148 self-tests. 1149 Bitmask for the various LOCKTYPE_ tests. Defaults to 0 1150 (no extra messages), setting it to -1 (all bits set) 1151 will print _a_lot_ more information - normally only 1152 useful to lockdep developers. 1153 1154 debug_objects [KNL,EARLY] Enable object debugging 1155 1156 debug_guardpage_minorder= 1157 [KNL,EARLY] When CONFIG_DEBUG_PAGEALLOC is set, this 1158 parameter allows control of the order of pages that will 1159 be intentionally kept free (and hence protected) by the 1160 buddy allocator. Bigger value increase the probability 1161 of catching random memory corruption, but reduce the 1162 amount of memory for normal system use. The maximum 1163 possible value is MAX_PAGE_ORDER/2. Setting this 1164 parameter to 1 or 2 should be enough to identify most 1165 random memory corruption problems caused by bugs in 1166 kernel or driver code when a CPU writes to (or reads 1167 from) a random memory location. Note that there exists 1168 a class of memory corruptions problems caused by buggy 1169 H/W or F/W or by drivers badly programming DMA 1170 (basically when memory is written at bus level and the 1171 CPU MMU is bypassed) which are not detectable by 1172 CONFIG_DEBUG_PAGEALLOC, hence this option will not 1173 help tracking down these problems. 1174 1175 debug_pagealloc= 1176 [KNL,EARLY] When CONFIG_DEBUG_PAGEALLOC is set, this parameter 1177 enables the feature at boot time. By default, it is 1178 disabled and the system will work mostly the same as a 1179 kernel built without CONFIG_DEBUG_PAGEALLOC. 1180 Note: to get most of debug_pagealloc error reports, it's 1181 useful to also enable the page_owner functionality. 1182 on: enable the feature 1183 1184 debugfs= [KNL,EARLY] This parameter enables what is exposed to 1185 userspace and debugfs internal clients. 1186 Format: { on, off } 1187 on: All functions are enabled. 1188 off: Filesystem is not registered and clients 1189 get a -EPERM as result when trying to register files 1190 or directories within debugfs. 1191 This is equivalent of the runtime functionality if 1192 debugfs was not enabled in the kernel at all. 1193 Default value is set in build-time with a kernel configuration. 1194 1195 debugpat [X86] Enable PAT debugging 1196 1197 default_hugepagesz= 1198 [HW] The size of the default HugeTLB page. This is 1199 the size represented by the legacy /proc/ hugepages 1200 APIs. In addition, this is the default hugetlb size 1201 used for shmget(), mmap() and mounting hugetlbfs 1202 filesystems. If not specified, defaults to the 1203 architecture's default huge page size. Huge page 1204 sizes are architecture dependent. See also 1205 Documentation/admin-guide/mm/hugetlbpage.rst. 1206 Format: size[KMG] 1207 1208 deferred_probe_timeout= 1209 [KNL] Debugging option to set a timeout in seconds for 1210 deferred probe to give up waiting on dependencies to 1211 probe. Only specific dependencies (subsystems or 1212 drivers) that have opted in will be ignored. A timeout 1213 of 0 will timeout at the end of initcalls; a negative 1214 value is treated as an infinite timeout value. If the 1215 timeout hasn't expired, it'll be restarted by each 1216 successful driver registration. This option will also 1217 dump out devices still on the deferred probe list after 1218 retrying. 1219 1220 delayacct [KNL] Enable per-task delay accounting 1221 1222 dell_smm_hwmon.ignore_dmi= 1223 [HW] Continue probing hardware even if DMI data 1224 indicates that the driver is running on unsupported 1225 hardware. 1226 1227 dell_smm_hwmon.force= 1228 [HW] Activate driver even if SMM BIOS signature does 1229 not match list of supported models and enable otherwise 1230 blacklisted features. 1231 1232 dell_smm_hwmon.power_status= 1233 [HW] Report power status in /proc/i8k 1234 (disabled by default). 1235 1236 dell_smm_hwmon.restricted= 1237 [HW] Allow controlling fans only if SYS_ADMIN 1238 capability is set. 1239 1240 dell_smm_hwmon.fan_mult= 1241 [HW] Factor to multiply fan speed with. 1242 1243 dell_smm_hwmon.fan_max= 1244 [HW] Maximum configurable fan speed. 1245 1246 dfltcc= [HW,S390] 1247 Format: { on | off | def_only | inf_only | always } 1248 on: s390 zlib hardware support for compression on 1249 level 1 and decompression (default) 1250 off: No s390 zlib hardware support 1251 def_only: s390 zlib hardware support for deflate 1252 only (compression on level 1) 1253 inf_only: s390 zlib hardware support for inflate 1254 only (decompression) 1255 always: Same as 'on' but ignores the selected compression 1256 level always using hardware support (used for debugging) 1257 1258 dhash_entries= [KNL] 1259 Set number of hash buckets for dentry cache. 1260 1261 disable_1tb_segments [PPC,EARLY] 1262 Disables the use of 1TB hash page table segments. This 1263 causes the kernel to fall back to 256MB segments which 1264 can be useful when debugging issues that require an SLB 1265 miss to occur. 1266 1267 disable= [IPV6] 1268 See Documentation/networking/ipv6.rst. 1269 1270 disable_radix [PPC,EARLY] 1271 Disable RADIX MMU mode on POWER9 1272 1273 disable_tlbie [PPC] 1274 Disable TLBIE instruction. Currently does not work 1275 with KVM, with HASH MMU, or with coherent accelerators. 1276 1277 disable_ddw [PPC/PSERIES,EARLY] 1278 Disable Dynamic DMA Window support. Use this 1279 to workaround buggy firmware. 1280 1281 disable_ipv6= [IPV6] 1282 See Documentation/networking/ipv6.rst. 1283 1284 disable_mtrr_cleanup [X86,EARLY] 1285 The kernel tries to adjust MTRR layout from continuous 1286 to discrete, to make X server driver able to add WB 1287 entry later. This parameter disables that. 1288 1289 disable_mtrr_trim [X86, Intel and AMD only,EARLY] 1290 By default the kernel will trim any uncacheable 1291 memory out of your available memory pool based on 1292 MTRR settings. This parameter disables that behavior, 1293 possibly causing your machine to run very slowly. 1294 1295 disable_timer_pin_1 [X86,EARLY] 1296 Disable PIN 1 of APIC timer 1297 Can be useful to work around chipset bugs. 1298 1299 dis_ucode_ldr [X86] Disable the microcode loader. 1300 1301 dma_debug=off If the kernel is compiled with DMA_API_DEBUG support, 1302 this option disables the debugging code at boot. 1303 1304 dma_debug_entries=<number> 1305 This option allows to tune the number of preallocated 1306 entries for DMA-API debugging code. One entry is 1307 required per DMA-API allocation. Use this if the 1308 DMA-API debugging code disables itself because the 1309 architectural default is too low. 1310 1311 dma_debug_driver=<driver_name> 1312 With this option the DMA-API debugging driver 1313 filter feature can be enabled at boot time. Just 1314 pass the driver to filter for as the parameter. 1315 The filter can be disabled or changed to another 1316 driver later using sysfs. 1317 1318 reg_file_data_sampling= 1319 [X86] Controls mitigation for Register File Data 1320 Sampling (RFDS) vulnerability. RFDS is a CPU 1321 vulnerability which may allow userspace to infer 1322 kernel data values previously stored in floating point 1323 registers, vector registers, or integer registers. 1324 RFDS only affects Intel Atom processors. 1325 1326 on: Turns ON the mitigation. 1327 off: Turns OFF the mitigation. 1328 1329 This parameter overrides the compile time default set 1330 by CONFIG_MITIGATION_RFDS. Mitigation cannot be 1331 disabled when other VERW based mitigations (like MDS) 1332 are enabled. In order to disable RFDS mitigation all 1333 VERW based mitigations need to be disabled. 1334 1335 For details see: 1336 Documentation/admin-guide/hw-vuln/reg-file-data-sampling.rst 1337 1338 dm_verity.keyring_unsealed= 1339 [KNL] When set to 1, leave the dm-verity keyring 1340 unsealed after initialization so userspace can 1341 provision keys. Once the keyring is restricted 1342 it becomes active and is searched during signature 1343 verification. 1344 1345 driver_async_probe= [KNL] 1346 List of driver names to be probed asynchronously. * 1347 matches with all driver names. If * is specified, the 1348 rest of the listed driver names are those that will NOT 1349 match the *. 1350 Format: <driver_name1>,<driver_name2>... 1351 1352 drm.edid_firmware=[<connector>:]<file>[,[<connector>:]<file>] 1353 Broken monitors, graphic adapters, KVMs and EDIDless 1354 panels may send no or incorrect EDID data sets. 1355 This parameter allows to specify an EDID data sets 1356 in the /lib/firmware directory that are used instead. 1357 An EDID data set will only be used for a particular 1358 connector, if its name and a colon are prepended to 1359 the EDID name. Each connector may use a unique EDID 1360 data set by separating the files with a comma. An EDID 1361 data set with no connector name will be used for 1362 any connectors not explicitly specified. 1363 1364 dscc4.setup= [NET] 1365 1366 dt_cpu_ftrs= [PPC,EARLY] 1367 Format: {"off" | "known"} 1368 Control how the dt_cpu_ftrs device-tree binding is 1369 used for CPU feature discovery and setup (if it 1370 exists). 1371 off: Do not use it, fall back to legacy cpu table. 1372 known: Do not pass through unknown features to guests 1373 or userspace, only those that the kernel is aware of. 1374 1375 dump_apple_properties [X86] 1376 Dump name and content of EFI device properties on 1377 x86 Macs. Useful for driver authors to determine 1378 what data is available or for reverse-engineering. 1379 1380 dyndbg[="val"] [KNL,DYNAMIC_DEBUG] 1381 <module>.dyndbg[="val"] 1382 Enable debug messages at boot time. See 1383 Documentation/admin-guide/dynamic-debug-howto.rst 1384 for details. 1385 1386 early_ioremap_debug [KNL,EARLY] 1387 Enable debug messages in early_ioremap support. This 1388 is useful for tracking down temporary early mappings 1389 which are not unmapped. 1390 1391 earlycon= [KNL,EARLY] Output early console device and options. 1392 1393 When used with no options, the early console is 1394 determined by stdout-path property in device tree's 1395 chosen node or the ACPI SPCR table if supported by 1396 the platform. 1397 1398 cdns,<addr>[,options] 1399 Start an early, polled-mode console on a Cadence 1400 (xuartps) serial port at the specified address. Only 1401 supported option is baud rate. If baud rate is not 1402 specified, the serial port must already be setup and 1403 configured. 1404 1405 uart[8250],io,<addr>[,options[,uartclk]] 1406 uart[8250],mmio,<addr>[,options[,uartclk]] 1407 uart[8250],mmio32,<addr>[,options[,uartclk]] 1408 uart[8250],mmio32be,<addr>[,options[,uartclk]] 1409 uart[8250],0x<addr>[,options] 1410 Start an early, polled-mode console on the 8250/16550 1411 UART at the specified I/O port or MMIO address. 1412 MMIO inter-register address stride is either 8-bit 1413 (mmio) or 32-bit (mmio32 or mmio32be). 1414 If none of [io|mmio|mmio32|mmio32be], <addr> is assumed 1415 to be equivalent to 'mmio'. 'options' are specified 1416 in the same format described for "console=ttyS<n>"; if 1417 unspecified, the h/w is not initialized. 'uartclk' is 1418 the uart clock frequency; if unspecified, it is set 1419 to 'BASE_BAUD' * 16. 1420 1421 pl011,<addr> 1422 pl011,mmio32,<addr> 1423 Start an early, polled-mode console on a pl011 serial 1424 port at the specified address. The pl011 serial port 1425 must already be setup and configured. Options are not 1426 yet supported. If 'mmio32' is specified, then only 1427 the driver will use only 32-bit accessors to read/write 1428 the device registers. 1429 1430 liteuart,<addr> 1431 Start an early console on a litex serial port at the 1432 specified address. The serial port must already be 1433 setup and configured. Options are not yet supported. 1434 1435 meson,<addr> 1436 Start an early, polled-mode console on a meson serial 1437 port at the specified address. The serial port must 1438 already be setup and configured. Options are not yet 1439 supported. 1440 1441 msm_serial,<addr> 1442 Start an early, polled-mode console on an msm serial 1443 port at the specified address. The serial port 1444 must already be setup and configured. Options are not 1445 yet supported. 1446 1447 msm_serial_dm,<addr> 1448 Start an early, polled-mode console on an msm serial 1449 dm port at the specified address. The serial port 1450 must already be setup and configured. Options are not 1451 yet supported. 1452 1453 owl,<addr> 1454 Start an early, polled-mode console on a serial port 1455 of an Actions Semi SoC, such as S500 or S900, at the 1456 specified address. The serial port must already be 1457 setup and configured. Options are not yet supported. 1458 1459 rda,<addr> 1460 Start an early, polled-mode console on a serial port 1461 of an RDA Micro SoC, such as RDA8810PL, at the 1462 specified address. The serial port must already be 1463 setup and configured. Options are not yet supported. 1464 1465 sbi 1466 Use RISC-V SBI (Supervisor Binary Interface) for early 1467 console. 1468 1469 smh Use ARM semihosting calls for early console. 1470 1471 s3c2410,<addr> 1472 s3c2412,<addr> 1473 s3c2440,<addr> 1474 s3c6400,<addr> 1475 s5pv210,<addr> 1476 exynos4210,<addr> 1477 Use early console provided by serial driver available 1478 on Samsung SoCs, requires selecting proper type and 1479 a correct base address of the selected UART port. The 1480 serial port must already be setup and configured. 1481 Options are not yet supported. 1482 1483 lantiq,<addr> 1484 Start an early, polled-mode console on a lantiq serial 1485 (lqasc) port at the specified address. The serial port 1486 must already be setup and configured. Options are not 1487 yet supported. 1488 1489 lpuart,<addr> 1490 lpuart32,<addr> 1491 Use early console provided by Freescale LP UART driver 1492 found on Freescale Vybrid and QorIQ LS1021A processors. 1493 A valid base address must be provided, and the serial 1494 port must already be setup and configured. 1495 1496 ec_imx21,<addr> 1497 ec_imx6q,<addr> 1498 Start an early, polled-mode, output-only console on the 1499 Freescale i.MX UART at the specified address. The UART 1500 must already be setup and configured. 1501 1502 ar3700_uart,<addr> 1503 Start an early, polled-mode console on the 1504 Armada 3700 serial port at the specified 1505 address. The serial port must already be setup 1506 and configured. Options are not yet supported. 1507 1508 qcom_geni,<addr> 1509 Start an early, polled-mode console on a Qualcomm 1510 Generic Interface (GENI) based serial port at the 1511 specified address. The serial port must already be 1512 setup and configured. Options are not yet supported. 1513 1514 efifb,[options] 1515 Start an early, unaccelerated console on the EFI 1516 memory mapped framebuffer (if available). On cache 1517 coherent non-x86 systems that use system memory for 1518 the framebuffer, pass the 'ram' option so that it is 1519 mapped with the correct attributes. 1520 1521 linflex,<addr> 1522 Use early console provided by Freescale LINFlexD UART 1523 serial driver for NXP S32V234 SoCs. A valid base 1524 address must be provided, and the serial port must 1525 already be setup and configured. 1526 1527 earlyprintk= [X86,SH,ARM,M68k,S390,UM,EARLY] 1528 earlyprintk=vga 1529 earlyprintk=sclp 1530 earlyprintk=xen 1531 earlyprintk=serial[,ttySn[,baudrate]] 1532 earlyprintk=serial[,0x...[,baudrate]] 1533 earlyprintk=ttySn[,baudrate] 1534 earlyprintk=dbgp[debugController#] 1535 earlyprintk=mmio32,membase[,{nocfg|baudrate}] 1536 earlyprintk=pciserial[,force],bus:device.function[,{nocfg|baudrate}] 1537 earlyprintk=xdbc[xhciController#] 1538 earlyprintk=bios 1539 1540 earlyprintk is useful when the kernel crashes before 1541 the normal console is initialized. It is not enabled by 1542 default because it has some cosmetic problems. 1543 1544 Use "nocfg" to skip UART configuration, assume 1545 BIOS/firmware has configured UART correctly. 1546 1547 Append ",keep" to not disable it when the real console 1548 takes over. 1549 1550 Only one of vga, serial, or usb debug port can 1551 be used at a time. 1552 1553 Currently only ttyS0 and ttyS1 may be specified by 1554 name. Other I/O ports may be explicitly specified 1555 on some architectures (x86 and arm at least) by 1556 replacing ttySn with an I/O port address, like this: 1557 earlyprintk=serial,0x1008,115200 1558 You can find the port for a given device in 1559 /proc/tty/driver/serial: 1560 2: uart:ST16650V2 port:00001008 irq:18 ... 1561 1562 Interaction with the standard serial driver is not 1563 very good. 1564 1565 The VGA output is eventually overwritten by 1566 the real console. 1567 1568 The xen option can only be used in Xen domains. 1569 1570 The sclp output can only be used on s390. 1571 1572 The bios output can only be used on SuperH. 1573 1574 The optional "force" to "pciserial" enables use of a 1575 PCI device even when its classcode is not of the 1576 UART class. 1577 1578 edd= [EDD] 1579 Format: {"off" | "on" | "skip[mbr]"} 1580 1581 efi= [EFI,EARLY] 1582 Format: { "debug", "disable_early_pci_dma", 1583 "nochunk", "noruntime", "nosoftreserve", 1584 "novamap", "no_disable_early_pci_dma" } 1585 debug: enable misc debug output. 1586 disable_early_pci_dma: disable the busmaster bit on all 1587 PCI bridges while in the EFI boot stub. 1588 nochunk: disable reading files in "chunks" in the EFI 1589 boot stub, as chunking can cause problems with some 1590 firmware implementations. 1591 noruntime : disable EFI runtime services support 1592 nosoftreserve: The EFI_MEMORY_SP (Specific Purpose) 1593 attribute may cause the kernel to reserve the 1594 memory range for a memory mapping driver to 1595 claim. Specify efi=nosoftreserve to disable this 1596 reservation and treat the memory by its base type 1597 (i.e. EFI_CONVENTIONAL_MEMORY / "System RAM"). 1598 novamap: do not call SetVirtualAddressMap(). 1599 no_disable_early_pci_dma: Leave the busmaster bit set 1600 on all PCI bridges while in the EFI boot stub 1601 1602 efi_no_storage_paranoia [EFI,X86,EARLY] 1603 The kernel reserves 5KB of EFI variable storage for 1604 safety, because some UEFI implementation may fail to 1605 boot if there's insufficient space in the EFI variable 1606 storage. 1607 1608 Using this parameter, you can use the 5KB reservation 1609 in the EFI variable storage. 1610 1611 However, Use this parameter only if you are really 1612 sure that your UEFI does sane gc and fulfills the spec 1613 otherwise your board may brick. 1614 1615 efivar_ssdt= [EFI; X86] Name of an EFI variable that contains an SSDT 1616 that is to be dynamically loaded by Linux. If there are 1617 multiple variables with the same name but with different 1618 vendor GUIDs, all of them will be loaded. See 1619 Documentation/admin-guide/acpi/ssdt-overlays.rst for details. 1620 1621 1622 eisa_irq_edge= [PARISC,HW] 1623 See header of drivers/parisc/eisa.c. 1624 1625 ekgdboc= [X86,KGDB,EARLY] Allow early kernel console debugging 1626 Format: ekgdboc=kbd 1627 1628 This is designed to be used in conjunction with 1629 the boot argument: earlyprintk=vga 1630 1631 This parameter works in place of the kgdboc parameter 1632 but can only be used if the backing tty is available 1633 very early in the boot process. For early debugging 1634 via a serial port see kgdboc_earlycon instead. 1635 1636 elfcorehdr=[size[KMG]@]offset[KMG] [PPC,SH,X86,S390,EARLY] 1637 Specifies physical address of start of kernel core 1638 image elf header and optionally the size. Generally 1639 kexec loader will pass this option to capture kernel. 1640 See Documentation/admin-guide/kdump/kdump.rst for details. 1641 1642 enable_mtrr_cleanup [X86,EARLY] 1643 The kernel tries to adjust MTRR layout from continuous 1644 to discrete, to make X server driver able to add WB 1645 entry later. This parameter enables that. 1646 1647 enforcing= [SELINUX] Set initial enforcing status. 1648 Format: {"0" | "1"} 1649 See security/selinux/Kconfig help text. 1650 0 -- permissive (log only, no denials). 1651 1 -- enforcing (deny and log). 1652 Default value is 0. 1653 Value can be changed at runtime via 1654 /sys/fs/selinux/enforce. 1655 1656 erst_disable [ACPI] 1657 Disable Error Record Serialization Table (ERST) 1658 support. 1659 1660 ether= [HW,NET] Ethernet cards parameters 1661 This option is obsoleted by the "netdev=" option, which 1662 has equivalent usage. See its documentation for details. 1663 1664 evm= [EVM] 1665 Format: { "fix" } 1666 Permit 'security.evm' to be updated regardless of 1667 current integrity status. 1668 1669 early_page_ext [KNL,EARLY] Enforces page_ext initialization to earlier 1670 stages so cover more early boot allocations. 1671 Please note that as side effect some optimizations 1672 might be disabled to achieve that (e.g. parallelized 1673 memory initialization is disabled) so the boot process 1674 might take longer, especially on systems with a lot of 1675 memory. Available with CONFIG_PAGE_EXTENSION=y. 1676 1677 failslab= 1678 fail_usercopy= 1679 fail_page_alloc= 1680 fail_skb_realloc= 1681 fail_make_request=[KNL] 1682 General fault injection mechanism. 1683 Format: <interval>,<probability>,<space>,<times> 1684 See also Documentation/fault-injection/. 1685 1686 fb_tunnels= [NET] 1687 Format: { initns | none } 1688 See Documentation/admin-guide/sysctl/net.rst for 1689 fb_tunnels_only_for_init_ns 1690 1691 floppy= [HW] 1692 See Documentation/admin-guide/blockdev/floppy.rst. 1693 1694 forcepae [X86-32] 1695 Forcefully enable Physical Address Extension (PAE). 1696 Many Pentium M systems disable PAE but may have a 1697 functionally usable PAE implementation. 1698 Warning: use of this parameter will taint the kernel 1699 and may cause unknown problems. 1700 1701 fred= [X86-64] 1702 Enable/disable Flexible Return and Event Delivery. 1703 Format: { on | off } 1704 on: enable FRED when it's present, the default setting. 1705 off: disable FRED. 1706 1707 ftrace=[tracer] 1708 [FTRACE] will set and start the specified tracer 1709 as early as possible in order to facilitate early 1710 boot debugging. 1711 1712 ftrace_boot_snapshot 1713 [FTRACE] On boot up, a snapshot will be taken of the 1714 ftrace ring buffer that can be read at: 1715 /sys/kernel/tracing/snapshot. 1716 This is useful if you need tracing information from kernel 1717 boot up that is likely to be overridden by user space 1718 start up functionality. 1719 1720 Optionally, the snapshot can also be defined for a tracing 1721 instance that was created by the trace_instance= command 1722 line parameter. 1723 1724 trace_instance=foo,sched_switch ftrace_boot_snapshot=foo 1725 1726 The above will cause the "foo" tracing instance to trigger 1727 a snapshot at the end of boot up. 1728 1729 ftrace_dump_on_oops[=2(orig_cpu) | =<instance>][,<instance> | 1730 ,<instance>=2(orig_cpu)] 1731 [FTRACE] will dump the trace buffers on oops. 1732 If no parameter is passed, ftrace will dump global 1733 buffers of all CPUs, if you pass 2 or orig_cpu, it 1734 will dump only the buffer of the CPU that triggered 1735 the oops, or the specific instance will be dumped if 1736 its name is passed. Multiple instance dump is also 1737 supported, and instances are separated by commas. Each 1738 instance supports only dump on CPU that triggered the 1739 oops by passing 2 or orig_cpu to it. 1740 1741 ftrace_dump_on_oops=foo=orig_cpu 1742 1743 The above will dump only the buffer of "foo" instance 1744 on CPU that triggered the oops. 1745 1746 ftrace_dump_on_oops,foo,bar=orig_cpu 1747 1748 The above will dump global buffer on all CPUs, the 1749 buffer of "foo" instance on all CPUs and the buffer 1750 of "bar" instance on CPU that triggered the oops. 1751 1752 ftrace_filter=[function-list] 1753 [FTRACE] Limit the functions traced by the function 1754 tracer at boot up. function-list is a comma-separated 1755 list of functions. This list can be changed at run 1756 time by the set_ftrace_filter file in the debugfs 1757 tracing directory. 1758 1759 ftrace_notrace=[function-list] 1760 [FTRACE] Do not trace the functions specified in 1761 function-list. This list can be changed at run time 1762 by the set_ftrace_notrace file in the debugfs 1763 tracing directory. 1764 1765 ftrace_graph_filter=[function-list] 1766 [FTRACE] Limit the top level callers functions traced 1767 by the function graph tracer at boot up. 1768 function-list is a comma-separated list of functions 1769 that can be changed at run time by the 1770 set_graph_function file in the debugfs tracing directory. 1771 1772 ftrace_graph_notrace=[function-list] 1773 [FTRACE] Do not trace from the functions specified in 1774 function-list. This list is a comma-separated list of 1775 functions that can be changed at run time by the 1776 set_graph_notrace file in the debugfs tracing directory. 1777 1778 ftrace_graph_max_depth=<uint> 1779 [FTRACE] Used with the function graph tracer. This is 1780 the max depth it will trace into a function. This value 1781 can be changed at run time by the max_graph_depth file 1782 in the tracefs tracing directory. default: 0 (no limit) 1783 1784 fw_devlink= [KNL,EARLY] Create device links between consumer and supplier 1785 devices by scanning the firmware to infer the 1786 consumer/supplier relationships. This feature is 1787 especially useful when drivers are loaded as modules as 1788 it ensures proper ordering of tasks like device probing 1789 (suppliers first, then consumers), supplier boot state 1790 clean up (only after all consumers have probed), 1791 suspend/resume & runtime PM (consumers first, then 1792 suppliers). 1793 Format: { off | permissive | on | rpm } 1794 off -- Don't create device links from firmware info. 1795 permissive -- Create device links from firmware info 1796 but use it only for ordering boot state clean 1797 up (sync_state() calls). 1798 on -- Create device links from firmware info and use it 1799 to enforce probe and suspend/resume ordering. 1800 rpm -- Like "on", but also use to order runtime PM. 1801 1802 fw_devlink.strict=<bool> 1803 [KNL,EARLY] Treat all inferred dependencies as mandatory 1804 dependencies. This only applies for fw_devlink=on|rpm. 1805 Format: <bool> 1806 1807 fw_devlink.sync_state = 1808 [KNL,EARLY] When all devices that could probe have finished 1809 probing, this parameter controls what to do with 1810 devices that haven't yet received their sync_state() 1811 calls. 1812 Format: { strict | timeout } 1813 strict -- Default. Continue waiting on consumers to 1814 probe successfully. 1815 timeout -- Give up waiting on consumers and call 1816 sync_state() on any devices that haven't yet 1817 received their sync_state() calls after 1818 deferred_probe_timeout has expired or by 1819 late_initcall() if !CONFIG_MODULES. 1820 1821 gamecon.map[2|3]= 1822 [HW,JOY] Multisystem joystick and NES/SNES/PSX pad 1823 support via parallel port (up to 5 devices per port) 1824 Format: <port#>,<pad1>,<pad2>,<pad3>,<pad4>,<pad5> 1825 See also Documentation/input/devices/joystick-parport.rst 1826 1827 gamma= [HW,DRM] 1828 1829 gart_fix_e820= [X86-64,EARLY] disable the fix e820 for K8 GART 1830 Format: off | on 1831 default: on 1832 1833 gather_data_sampling= 1834 [X86,INTEL,EARLY] Control the Gather Data Sampling (GDS) 1835 mitigation. 1836 1837 Gather Data Sampling is a hardware vulnerability which 1838 allows unprivileged speculative access to data which was 1839 previously stored in vector registers. 1840 1841 This issue is mitigated by default in updated microcode. 1842 The mitigation may have a performance impact but can be 1843 disabled. On systems without the microcode mitigation 1844 disabling AVX serves as a mitigation. 1845 1846 force: Disable AVX to mitigate systems without 1847 microcode mitigation. No effect if the microcode 1848 mitigation is present. Known to cause crashes in 1849 userspace with buggy AVX enumeration. 1850 1851 off: Disable GDS mitigation. 1852 1853 gbpages [X86] Use GB pages for kernel direct mappings. 1854 1855 gcov_persist= [GCOV] When non-zero (default), profiling data for 1856 kernel modules is saved and remains accessible via 1857 debugfs, even when the module is unloaded/reloaded. 1858 When zero, profiling data is discarded and associated 1859 debugfs files are removed at module unload time. 1860 1861 goldfish [X86] Enable the goldfish android emulator platform. 1862 Don't use this when you are not running on the 1863 android emulator 1864 1865 gpio-mockup.gpio_mockup_ranges 1866 [HW] Sets the ranges of gpiochip of for this device. 1867 Format: <start1>,<end1>,<start2>,<end2>... 1868 gpio-mockup.gpio_mockup_named_lines 1869 [HW] Let the driver know GPIO lines should be named. 1870 1871 gpt [EFI] Forces disk with valid GPT signature but 1872 invalid Protective MBR to be treated as GPT. If the 1873 primary GPT is corrupted, it enables the backup/alternate 1874 GPT to be used instead. 1875 1876 grcan.enable0= [HW] Configuration of physical interface 0. Determines 1877 the "Enable 0" bit of the configuration register. 1878 Format: 0 | 1 1879 Default: 0 1880 grcan.enable1= [HW] Configuration of physical interface 1. Determines 1881 the "Enable 0" bit of the configuration register. 1882 Format: 0 | 1 1883 Default: 0 1884 grcan.select= [HW] Select which physical interface to use. 1885 Format: 0 | 1 1886 Default: 0 1887 grcan.txsize= [HW] Sets the size of the tx buffer. 1888 Format: <unsigned int> such that (txsize & ~0x1fffc0) == 0. 1889 Default: 1024 1890 grcan.rxsize= [HW] Sets the size of the rx buffer. 1891 Format: <unsigned int> such that (rxsize & ~0x1fffc0) == 0. 1892 Default: 1024 1893 1894 hardened_usercopy= 1895 [KNL] Under CONFIG_HARDENED_USERCOPY, whether 1896 hardening is enabled for this boot. Hardened 1897 usercopy checking is used to protect the kernel 1898 from reading or writing beyond known memory 1899 allocation boundaries as a proactive defense 1900 against bounds-checking flaws in the kernel's 1901 copy_to_user()/copy_from_user() interface. 1902 The default is determined by 1903 CONFIG_HARDENED_USERCOPY_DEFAULT_ON. 1904 on Perform hardened usercopy checks. 1905 off Disable hardened usercopy checks. 1906 1907 hardlockup_all_cpu_backtrace= 1908 [KNL] Should the hard-lockup detector generate 1909 backtraces on all cpus. 1910 Format: 0 | 1 1911 1912 hash_pointers= 1913 [KNL,EARLY] 1914 By default, when pointers are printed to the console 1915 or buffers via the %p format string, that pointer is 1916 "hashed", i.e. obscured by hashing the pointer value. 1917 This is a security feature that hides actual kernel 1918 addresses from unprivileged users, but it also makes 1919 debugging the kernel more difficult since unequal 1920 pointers can no longer be compared. The choices are: 1921 Format: { auto | always | never } 1922 Default: auto 1923 1924 auto - Hash pointers unless slab_debug is enabled. 1925 always - Always hash pointers (even if slab_debug is 1926 enabled). 1927 never - Never hash pointers. This option should only 1928 be specified when debugging the kernel. Do 1929 not use on production kernels. The boot 1930 param "no_hash_pointers" is an alias for 1931 this mode. 1932 1933 For controlling hashing dynamically at runtime, 1934 use the "kernel.kptr_restrict" sysctl instead. 1935 1936 hashdist= [KNL,NUMA] Large hashes allocated during boot 1937 are distributed across NUMA nodes. Defaults on 1938 for 64-bit NUMA, off otherwise. 1939 Format: 0 | 1 (for off | on) 1940 1941 hd= [EIDE] (E)IDE hard drive subsystem geometry 1942 Format: <cyl>,<head>,<sect> 1943 1944 hest_disable [ACPI] 1945 Disable Hardware Error Source Table (HEST) support; 1946 corresponding firmware-first mode error processing 1947 logic will be disabled. 1948 1949 hibernate= [HIBERNATION] 1950 noresume Don't check if there's a hibernation image 1951 present during boot. 1952 nocompress Don't compress/decompress hibernation images. 1953 no Disable hibernation and resume. 1954 protect_image Turn on image protection during restoration 1955 (that will set all pages holding image data 1956 during restoration read-only). 1957 1958 hibernate.compressor= [HIBERNATION] Compression algorithm to be 1959 used with hibernation. 1960 Format: { lzo | lz4 } 1961 Default: lzo 1962 1963 lzo: Select LZO compression algorithm to 1964 compress/decompress hibernation image. 1965 1966 lz4: Select LZ4 compression algorithm to 1967 compress/decompress hibernation image. 1968 1969 hibernate.pm_test_delay= 1970 [HIBERNATION] 1971 Sets the number of seconds to remain in a hibernation test 1972 mode before resuming the system (see 1973 /sys/power/pm_test). Only available when CONFIG_PM_DEBUG 1974 is set. Default value is 5. 1975 1976 hibernate_compression_threads= 1977 [HIBERNATION] 1978 Set the number of threads used for compressing or decompressing 1979 hibernation images. 1980 1981 Format: <integer> 1982 Default: 3 1983 Minimum: 1 1984 Example: hibernate_compression_threads=4 1985 1986 highmem=nn[KMG] [KNL,BOOT,EARLY] forces the highmem zone to have an exact 1987 size of <nn>. This works even on boxes that have no 1988 highmem otherwise. This also works to reduce highmem 1989 size on bigger boxes. 1990 1991 highres= [KNL] Enable/disable high resolution timer mode. 1992 Valid parameters: "on", "off" 1993 Default: "on" 1994 1995 hlt [BUGS=ARM,SH] 1996 1997 hostname= [KNL,EARLY] Set the hostname (aka UTS nodename). 1998 Format: <string> 1999 This allows setting the system's hostname during early 2000 startup. This sets the name returned by gethostname. 2001 Using this parameter to set the hostname makes it 2002 possible to ensure the hostname is correctly set before 2003 any userspace processes run, avoiding the possibility 2004 that a process may call gethostname before the hostname 2005 has been explicitly set, resulting in the calling 2006 process getting an incorrect result. The string must 2007 not exceed the maximum allowed hostname length (usually 2008 64 characters) and will be truncated otherwise. 2009 2010 hpet= [X86-32,HPET] option to control HPET usage 2011 Format: { enable (default) | disable | force | 2012 verbose } 2013 disable: disable HPET and use PIT instead 2014 force: allow force enabled of undocumented chips (ICH4, 2015 VIA, nVidia) 2016 verbose: show contents of HPET registers during setup 2017 2018 hpet_mmap= [X86, HPET_MMAP] Allow userspace to mmap HPET 2019 registers. Default set by CONFIG_HPET_MMAP_DEFAULT. 2020 2021 hugepages= [HW,EARLY] Number of HugeTLB pages to allocate at boot. 2022 If this follows hugepagesz (below), it specifies 2023 the number of pages of hugepagesz to be allocated. 2024 If this is the first HugeTLB parameter on the command 2025 line, it specifies the number of pages to allocate for 2026 the default huge page size. If using node format, the 2027 number of pages to allocate per-node can be specified. 2028 See also Documentation/admin-guide/mm/hugetlbpage.rst. 2029 Format: <integer> or (node format) 2030 <node>:<integer>[,<node>:<integer>] 2031 2032 hugepagesz= 2033 [HW,EARLY] The size of the HugeTLB pages. This is 2034 used in conjunction with hugepages (above) to 2035 allocate huge pages of a specific size at boot. The 2036 pair hugepagesz=X hugepages=Y can be specified once 2037 for each supported huge page size. Huge page sizes 2038 are architecture dependent. See also 2039 Documentation/admin-guide/mm/hugetlbpage.rst. 2040 Format: size[KMG] 2041 2042 hugepage_alloc_threads= 2043 [HW] The number of threads that should be used to 2044 allocate hugepages during boot. This option can be 2045 used to improve system bootup time when allocating 2046 a large amount of huge pages. 2047 The default value is 25% of the available hardware threads. 2048 2049 Note that this parameter only applies to non-gigantic huge pages. 2050 2051 hugetlb_cma= [HW,CMA,EARLY] The size of a CMA area used for allocation 2052 of gigantic hugepages. Or using node format, the size 2053 of a CMA area per node can be specified. 2054 The size can be an absolute value (e.g., 2G) or a 2055 percentage of the total memory or node memory (e.g., 20%). 2056 Percentage-derived sizes are rounded down to a multiple of 2057 the architecture's gigantic hugepage size and may become 2058 zero. 2059 Format: nn[KMGTPE] or nn% or (node format) 2060 <node>:nn[KMGTPE][,<node>:nn[KMGTPE]] or 2061 <node>:nn%[,<node>:nn%] 2062 2063 The size must be a multiple of the gigantic page size. 2064 When using node format, this applies to each per-node size. 2065 Missaligned values are dropped with a warning. 2066 2067 Reserve a CMA area of given size and allocate gigantic 2068 hugepages using the CMA allocator. If enabled, the 2069 boot-time allocation of gigantic hugepages is skipped. 2070 2071 hugetlb_cma_only= 2072 [HW,CMA,EARLY] When allocating new HugeTLB pages, only 2073 try to allocate from the CMA areas. 2074 2075 This option does nothing if hugetlb_cma= is not also 2076 specified. 2077 2078 hugetlb_free_vmemmap= 2079 [KNL] Requires CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP 2080 enabled. 2081 Control if HugeTLB Vmemmap Optimization (HVO) is enabled. 2082 Allows heavy hugetlb users to free up some more 2083 memory (7 * PAGE_SIZE for each 2MB hugetlb page). 2084 Format: { on | off (default) } 2085 2086 on: enable HVO 2087 off: disable HVO 2088 2089 Built with CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP_DEFAULT_ON=y, 2090 the default is on. 2091 2092 Note that the vmemmap pages may be allocated from the added 2093 memory block itself when memory_hotplug.memmap_on_memory is 2094 enabled, those vmemmap pages cannot be optimized even if this 2095 feature is enabled. Other vmemmap pages not allocated from 2096 the added memory block itself do not be affected. 2097 2098 hung_task_panic= 2099 [KNL] Number of hung tasks to trigger kernel panic. 2100 Format: <int> 2101 2102 When set to a non-zero value, a kernel panic will be triggered if 2103 the number of detected hung tasks reaches this value. 2104 2105 0: don't panic 2106 1: panic immediately on first hung task 2107 N: panic after N hung tasks are detected in a single scan 2108 2109 The default value is controlled by the 2110 CONFIG_BOOTPARAM_HUNG_TASK_PANIC build-time option. The value 2111 selected by this boot parameter can be changed later by the 2112 kernel.hung_task_panic sysctl. 2113 2114 hvc_iucv= [S390] Number of z/VM IUCV hypervisor console (HVC) 2115 terminal devices. Valid values: 0..8 2116 hvc_iucv_allow= [S390] Comma-separated list of z/VM user IDs. 2117 If specified, z/VM IUCV HVC accepts connections 2118 from listed z/VM user IDs only. 2119 2120 hv_nopvspin [X86,HYPER_V,EARLY] 2121 Disables the paravirt spinlock optimizations 2122 which allow the hypervisor to 'idle' the guest 2123 on lock contention. 2124 2125 hw_protection= [HW] 2126 Format: reboot | shutdown 2127 2128 Hardware protection action taken on critical events like 2129 overtemperature or imminent voltage loss. 2130 2131 i2c_bus= [HW] Override the default board specific I2C bus speed 2132 or register an additional I2C bus that is not 2133 registered from board initialization code. 2134 Format: 2135 <bus_id>,<clkrate> 2136 2137 i2c_touchscreen_props= [HW,ACPI,X86] 2138 Set device-properties for ACPI-enumerated I2C-attached 2139 touchscreen, to e.g. fix coordinates of upside-down 2140 mounted touchscreens. If you need this option please 2141 submit a drivers/platform/x86/touchscreen_dmi.c patch 2142 adding a DMI quirk for this. 2143 2144 Format: 2145 <ACPI_HW_ID>:<prop_name>=<val>[:prop_name=val][:...] 2146 Where <val> is one of: 2147 Omit "=<val>" entirely Set a boolean device-property 2148 Unsigned number Set a u32 device-property 2149 Anything else Set a string device-property 2150 2151 Examples (split over multiple lines): 2152 i2c_touchscreen_props=GDIX1001:touchscreen-inverted-x: 2153 touchscreen-inverted-y 2154 2155 i2c_touchscreen_props=MSSL1680:touchscreen-size-x=1920: 2156 touchscreen-size-y=1080:touchscreen-inverted-y: 2157 firmware-name=gsl1680-vendor-model.fw:silead,home-button 2158 2159 i8042.debug [HW] Toggle i8042 debug mode 2160 i8042.unmask_kbd_data 2161 [HW] Enable printing of interrupt data from the KBD port 2162 (disabled by default, and as a pre-condition 2163 requires that i8042.debug=1 be enabled) 2164 i8042.direct [HW] Put keyboard port into non-translated mode 2165 i8042.dumbkbd [HW] Pretend that controller can only read data from 2166 keyboard and cannot control its state 2167 (Don't attempt to blink the leds) 2168 i8042.noaux [HW] Don't check for auxiliary (== mouse) port 2169 i8042.nokbd [HW] Don't check/create keyboard port 2170 i8042.noloop [HW] Disable the AUX Loopback command while probing 2171 for the AUX port 2172 i8042.nomux [HW] Don't check presence of an active multiplexing 2173 controller 2174 i8042.nopnp [HW] Don't use ACPIPnP / PnPBIOS to discover KBD/AUX 2175 controllers 2176 i8042.notimeout [HW] Ignore timeout condition signalled by controller 2177 i8042.reset [HW] Reset the controller during init, cleanup and 2178 suspend-to-ram transitions, only during s2r 2179 transitions, or never reset 2180 Format: { 1 | Y | y | 0 | N | n } 2181 1, Y, y: always reset controller 2182 0, N, n: don't ever reset controller 2183 Default: only on s2r transitions on x86; most other 2184 architectures force reset to be always executed 2185 i8042.unlock [HW] Unlock (ignore) the keylock 2186 i8042.kbdreset [HW] Reset device connected to KBD port 2187 i8042.probe_defer 2188 [HW] Allow deferred probing upon i8042 probe errors 2189 2190 i810= [HW,DRM] 2191 2192 i915.invert_brightness= 2193 [DRM] Invert the sense of the variable that is used to 2194 set the brightness of the panel backlight. Normally a 2195 brightness value of 0 indicates backlight switched off, 2196 and the maximum of the brightness value sets the backlight 2197 to maximum brightness. If this parameter is set to 0 2198 (default) and the machine requires it, or this parameter 2199 is set to 1, a brightness value of 0 sets the backlight 2200 to maximum brightness, and the maximum of the brightness 2201 value switches the backlight off. 2202 -1 -- never invert brightness 2203 0 -- machine default 2204 1 -- force brightness inversion 2205 2206 ia32_emulation= [X86-64] 2207 Format: <bool> 2208 When true, allows loading 32-bit programs and executing 32-bit 2209 syscalls, essentially overriding IA32_EMULATION_DEFAULT_DISABLED at 2210 boot time. When false, unconditionally disables IA32 emulation. 2211 2212 2213 idle= [X86,EARLY] 2214 Format: idle=poll, idle=halt, idle=nomwait 2215 2216 idle=poll: Don't do power saving in the idle loop 2217 using HLT, but poll for rescheduling event. This will 2218 make the CPUs eat a lot more power, but may be useful 2219 to get slightly better performance in multiprocessor 2220 benchmarks. It also makes some profiling using 2221 performance counters more accurate. Please note that 2222 on systems with MONITOR/MWAIT support (like Intel 2223 EM64T CPUs) this option has no performance advantage 2224 over the normal idle loop. It may also interact badly 2225 with hyperthreading. 2226 2227 idle=halt: Halt is forced to be used for CPU idle. 2228 In such case C2/C3 won't be used again. 2229 2230 idle=nomwait: Disable mwait for CPU C-states 2231 2232 [ARM64,EARLY] 2233 Format: idle=wfi, idle=yield, idle=nop 2234 2235 idle=wfi: Use the WFI (Wait For Interrupt) hint 2236 instruction in the idle loop. This is the default and 2237 allows the CPU to enter a low-power state until an 2238 interrupt arrives. 2239 2240 idle=yield: Use the YIELD hint instruction instead of 2241 WFI. CPUs supporting simultaneous multi-threading (SMT), 2242 can continue executing another thread when the current 2243 thread reaches the idle loop. This will make the CPUs 2244 eat more power, but may be useful to get slightly better 2245 performance in some applications, since the CPUs will 2246 not enter a low-power state. 2247 2248 idle=nop: Do not execute any idle instruction in the 2249 idle loop. This is useful on platforms where WFI 2250 misbehaves, leading to system instability or loss of CPU 2251 state. This will make the CPUs eat more power, but may 2252 give slightly better performance in some applications, 2253 since the CPUs will not enter a low-power state. 2254 2255 idxd.sva= [HW] 2256 Format: <bool> 2257 Allow force disabling of Shared Virtual Memory (SVA) 2258 support for the idxd driver. By default it is set to 2259 true (1). 2260 2261 idxd.tc_override= [HW] 2262 Format: <bool> 2263 Allow override of default traffic class configuration 2264 for the device. By default it is set to false (0). 2265 2266 ieee754= [MIPS] Select IEEE Std 754 conformance mode 2267 Format: { strict | legacy | 2008 | relaxed | emulated } 2268 Default: strict 2269 2270 Choose which programs will be accepted for execution 2271 based on the IEEE 754 NaN encoding(s) supported by 2272 the FPU and the NaN encoding requested with the value 2273 of an ELF file header flag individually set by each 2274 binary. Hardware implementations are permitted to 2275 support either or both of the legacy and the 2008 NaN 2276 encoding mode. 2277 2278 Available settings are as follows: 2279 strict accept binaries that request a NaN encoding 2280 supported by the FPU 2281 legacy only accept legacy-NaN binaries, if supported 2282 by the FPU 2283 2008 only accept 2008-NaN binaries, if supported 2284 by the FPU 2285 relaxed accept any binaries regardless of whether 2286 supported by the FPU 2287 emulated accept any binaries but enable FPU emulator 2288 if binary mode is unsupported by the FPU. 2289 2290 The FPU emulator is always able to support both NaN 2291 encodings, so if no FPU hardware is present or it has 2292 been disabled with 'nofpu', then the settings of 2293 'legacy' and '2008' strap the emulator accordingly, 2294 'relaxed' straps the emulator for both legacy-NaN and 2295 2008-NaN, whereas 'strict' enables legacy-NaN only on 2296 legacy processors and both NaN encodings on MIPS32 or 2297 MIPS64 CPUs. 2298 2299 The setting for ABS.fmt/NEG.fmt instruction execution 2300 mode generally follows that for the NaN encoding, 2301 except where unsupported by hardware. 2302 2303 ignore_loglevel [KNL,EARLY] 2304 Ignore loglevel setting - this will print /all/ 2305 kernel messages to the console. Useful for debugging. 2306 We also add it as printk module parameter, so users 2307 could change it dynamically, usually by 2308 /sys/module/printk/parameters/ignore_loglevel. 2309 2310 ignore_rlimit_data 2311 Ignore RLIMIT_DATA setting for data mappings, 2312 print warning at first misuse. Can be changed via 2313 /sys/module/kernel/parameters/ignore_rlimit_data. 2314 2315 ihash_entries= [KNL] 2316 Set number of hash buckets for inode cache. 2317 2318 ima_appraise= [IMA] appraise integrity measurements 2319 Format: { "off" | "enforce" | "fix" | "log" } 2320 default: "enforce" 2321 2322 ima_appraise_tcb [IMA] Deprecated. Use ima_policy= instead. 2323 The builtin appraise policy appraises all files 2324 owned by uid=0. 2325 2326 ima_canonical_fmt [IMA] 2327 Use the canonical format for the binary runtime 2328 measurements, instead of host native format. 2329 2330 ima_flush_htable [IMA] 2331 Flush the IMA hash table when deleting all the 2332 staged measurement records, to achieve maximum 2333 memory saving at the cost of having duplicate 2334 records across the staged measurement lists. 2335 2336 ima_hash= [IMA] 2337 Format: { md5 | sha1 | rmd160 | sha256 | sha384 2338 | sha512 | ... } 2339 default: "sha1" 2340 2341 The list of supported hash algorithms is defined 2342 in crypto/hash_info.h. 2343 2344 ima_policy= [IMA] 2345 The builtin policies to load during IMA setup. 2346 Format: "tcb | appraise_tcb | secure_boot | 2347 fail_securely | critical_data" 2348 2349 The "tcb" policy measures all programs exec'd, files 2350 mmap'd for exec, and all files opened with the read 2351 mode bit set by either the effective uid (euid=0) or 2352 uid=0. 2353 2354 The "appraise_tcb" policy appraises the integrity of 2355 all files owned by root. 2356 2357 The "secure_boot" policy appraises the integrity 2358 of files (eg. kexec kernel image, kernel modules, 2359 firmware, policy, etc) based on file signatures. 2360 2361 The "fail_securely" policy forces file signature 2362 verification failure also on privileged mounted 2363 filesystems with the SB_I_UNVERIFIABLE_SIGNATURE 2364 flag. 2365 2366 The "critical_data" policy measures kernel integrity 2367 critical data. 2368 2369 ima_tcb [IMA] Deprecated. Use ima_policy= instead. 2370 Load a policy which meets the needs of the Trusted 2371 Computing Base. This means IMA will measure all 2372 programs exec'd, files mmap'd for exec, and all files 2373 opened for read by uid=0. 2374 2375 ima_template= [IMA] 2376 Select one of defined IMA measurements template formats. 2377 Formats: { "ima" | "ima-ng" | "ima-ngv2" | "ima-sig" | 2378 "ima-sigv2" } 2379 Default: "ima-ng" 2380 2381 ima_template_fmt= 2382 [IMA] Define a custom template format. 2383 Format: { "field1|...|fieldN" } 2384 2385 ima= [IMA] Enable or disable IMA 2386 Format: { "off" | "on" } 2387 Default: "on" 2388 Note that disabling IMA is limited to kdump kernel. 2389 2390 indirect_target_selection= [X86,Intel] Mitigation control for Indirect 2391 Target Selection(ITS) bug in Intel CPUs. Updated 2392 microcode is also required for a fix in IBPB. 2393 2394 on: Enable mitigation (default). 2395 off: Disable mitigation. 2396 force: Force the ITS bug and deploy default 2397 mitigation. 2398 vmexit: Only deploy mitigation if CPU is affected by 2399 guest/host isolation part of ITS. 2400 stuff: Deploy RSB-fill mitigation when retpoline is 2401 also deployed. Otherwise, deploy the default 2402 mitigation. 2403 2404 For details see: 2405 Documentation/admin-guide/hw-vuln/indirect-target-selection.rst 2406 2407 init= [KNL] 2408 Format: <full_path> 2409 Run specified binary instead of /sbin/init as init 2410 process. 2411 2412 initcall_debug [KNL] Trace initcalls as they are executed. Useful 2413 for working out where the kernel is dying during 2414 startup. 2415 2416 initcall_blacklist= [KNL] Do not execute a comma-separated list of 2417 initcall functions. Useful for debugging built-in 2418 modules and initcalls. 2419 2420 initramfs_async= [KNL] 2421 Format: <bool> 2422 Default: 1 2423 This parameter controls whether the initramfs 2424 image is unpacked asynchronously, concurrently 2425 with devices being probed and 2426 initialized. This should normally just work, 2427 but as a debugging aid, one can get the 2428 historical behaviour of the initramfs 2429 unpacking being completed before device_ and 2430 late_ initcalls. 2431 2432 initrd= [BOOT,EARLY] Specify the location of the initial ramdisk 2433 2434 initrdmem= [KNL,EARLY] Specify a physical address and size from which to 2435 load the initrd. If an initrd is compiled in or 2436 specified in the bootparams, it takes priority over this 2437 setting. 2438 Format: ss[KMG],nn[KMG] 2439 Default is 0, 0 2440 2441 init_on_alloc= [MM,EARLY] Fill newly allocated pages and heap objects with 2442 zeroes. 2443 Format: 0 | 1 2444 Default set by CONFIG_INIT_ON_ALLOC_DEFAULT_ON. 2445 2446 init_on_free= [MM,EARLY] Fill freed pages and heap objects with zeroes. 2447 Format: 0 | 1 2448 Default set by CONFIG_INIT_ON_FREE_DEFAULT_ON. 2449 2450 init_pkru= [X86] Specify the default memory protection keys rights 2451 register contents for all processes. 0x55555554 by 2452 default (disallow access to all but pkey 0). Can 2453 override in debugfs after boot. 2454 2455 inport.irq= [HW] Inport (ATI XL and Microsoft) busmouse driver 2456 Format: <irq> 2457 2458 int_pln_enable [X86] Enable power limit notification interrupt 2459 2460 integrity_audit=[IMA] 2461 Format: { "0" | "1" } 2462 0 -- basic integrity auditing messages. (Default) 2463 1 -- additional integrity auditing messages. 2464 2465 intel_iommu= [DMAR] Intel IOMMU driver (DMAR) option 2466 on 2467 Enable intel iommu driver. 2468 off 2469 Disable intel iommu driver. 2470 igfx_off [Default Off] 2471 By default, gfx is mapped as normal device. If a gfx 2472 device has a dedicated DMAR unit, the DMAR unit is 2473 bypassed by not enabling DMAR with this option. In 2474 this case, gfx device will use physical address for 2475 DMA. 2476 strict [Default Off] 2477 Deprecated, equivalent to iommu.strict=1. 2478 sp_off [Default Off] 2479 By default, super page will be supported if Intel IOMMU 2480 has the capability. With this option, super page will 2481 not be supported. 2482 sm_on 2483 Enable the Intel IOMMU scalable mode if the hardware 2484 advertises that it has support for the scalable mode 2485 translation. 2486 sm_off 2487 Disallow use of the Intel IOMMU scalable mode. 2488 tboot_noforce [Default Off] 2489 Do not force the Intel IOMMU enabled under tboot. 2490 By default, tboot will force Intel IOMMU on, which 2491 could harm performance of some high-throughput 2492 devices like 40GBit network cards, even if identity 2493 mapping is enabled. 2494 Note that using this option lowers the security 2495 provided by tboot because it makes the system 2496 vulnerable to DMA attacks. 2497 2498 intel_idle.max_cstate= [KNL,HW,ACPI,X86] 2499 0 disables intel_idle and fall back on acpi_idle. 2500 1 to 9 specify maximum depth of C-state. 2501 2502 intel_pstate= [X86,EARLY] 2503 disable 2504 Do not enable intel_pstate as the default 2505 scaling driver for the supported processors 2506 active 2507 Use intel_pstate driver to bypass the scaling 2508 governors layer of cpufreq and provides it own 2509 algorithms for p-state selection. There are two 2510 P-state selection algorithms provided by 2511 intel_pstate in the active mode: powersave and 2512 performance. The way they both operate depends 2513 on whether or not the hardware managed P-states 2514 (HWP) feature has been enabled in the processor 2515 and possibly on the processor model. 2516 passive 2517 Use intel_pstate as a scaling driver, but configure it 2518 to work with generic cpufreq governors (instead of 2519 enabling its internal governor). This mode cannot be 2520 used along with the hardware-managed P-states (HWP) 2521 feature. 2522 force 2523 Enable intel_pstate on systems that prohibit it by default 2524 in favor of acpi-cpufreq. Forcing the intel_pstate driver 2525 instead of acpi-cpufreq may disable platform features, such 2526 as thermal controls and power capping, that rely on ACPI 2527 P-States information being indicated to OSPM and therefore 2528 should be used with caution. This option does not work with 2529 processors that aren't supported by the intel_pstate driver 2530 or on platforms that use pcc-cpufreq instead of acpi-cpufreq. 2531 no_hwp 2532 Do not enable hardware P state control (HWP) 2533 if available. 2534 hwp_only 2535 Only load intel_pstate on systems which support 2536 hardware P state control (HWP) if available. 2537 support_acpi_ppc 2538 Enforce ACPI _PPC performance limits. If the Fixed ACPI 2539 Description Table, specifies preferred power management 2540 profile as "Enterprise Server" or "Performance Server", 2541 then this feature is turned on by default. 2542 per_cpu_perf_limits 2543 Allow per-logical-CPU P-State performance control limits using 2544 cpufreq sysfs interface 2545 no_cas 2546 Do not enable capacity-aware scheduling (CAS) on 2547 hybrid systems 2548 2549 intremap= [X86-64,Intel-IOMMU,EARLY] 2550 on enable Interrupt Remapping (default) 2551 off disable Interrupt Remapping 2552 nosid disable Source ID checking 2553 no_x2apic_optout 2554 BIOS x2APIC opt-out request will be ignored 2555 nopost disable Interrupt Posting 2556 posted_msi 2557 enable MSIs delivered as posted interrupts 2558 2559 iomem= Disable strict checking of access to MMIO memory 2560 strict regions from userspace. 2561 relaxed 2562 2563 iommu= [X86,EARLY] 2564 2565 off 2566 Don't initialize and use any kind of IOMMU. 2567 2568 force 2569 Force the use of the hardware IOMMU even when 2570 it is not actually needed (e.g. because < 3 GB 2571 memory). 2572 2573 noforce 2574 Don't force hardware IOMMU usage when it is not 2575 needed. (default). 2576 2577 merge 2578 Do scatter-gather (SG) merging. Implies "force" 2579 (experimental). 2580 2581 nomerge 2582 Don't do scatter-gather (SG) merging. 2583 2584 biomerge 2585 Do scatter-gather (SG) merging. Implies "force" 2586 (experimental). [same as "merge"] 2587 2588 panic 2589 Always panic when IOMMU overflows. 2590 2591 nopanic 2592 Don't panic on IOMMU overflows. 2593 2594 pt 2595 Use passththrough mode by default 2596 (Equivalent to iommu.passthrough=1) 2597 2598 nopt 2599 Use translated mode for DMA by default 2600 (Equivalent to iommu.passthrough=0) 2601 2602 soft 2603 Use software bounce buffering (SWIOTLB) (default for 2604 Intel machines). This can be used to prevent the usage 2605 of an available hardware IOMMU. 2606 2607 usedac 2608 Use the DAC on VIA PCI bridge 2609 (default: disable the VIA PCI bridge DAC) 2610 2611 AMD Gart HW IOMMU-specific options: (CONFIG_GART_IOMMU) 2612 2613 <size> 2614 Set the size of the remapping area in bytes. 2615 2616 allowed 2617 Overwrite iommu off workarounds for specific chipsets 2618 2619 force 2620 Overwrite iommu off workarounds for specific chipsets 2621 2622 fullflush 2623 Flush IOMMU on each allocation (default). 2624 2625 nofullflush 2626 Don't use IOMMU fullflush. 2627 2628 memaper[=<order>] 2629 Allocate an own aperture over RAM with size 2630 32MB<<order. (default: order=1, i.e. 64MB) 2631 2632 noaperture 2633 Ask the IOMMU not to touch the aperture for AGP. 2634 2635 noagp 2636 Don't initialize the AGP driver and use full aperture. 2637 2638 iommu= [PPC/POWERNV] 2639 2640 nobypass 2641 Disable IOMMU bypass, using IOMMU for PCI devices. 2642 2643 iommu.forcedac= [ARM64,X86,EARLY] Control IOVA allocation for PCI devices. 2644 Format: { "0" | "1" } 2645 0 - Try to allocate a 32-bit DMA address first, before 2646 falling back to the full range if needed. 2647 1 - Allocate directly from the full usable range, 2648 forcing Dual Address Cycle for PCI cards supporting 2649 greater than 32-bit addressing. 2650 2651 iommu.strict= [ARM64,X86,S390,EARLY] Configure TLB invalidation behaviour 2652 Format: { "0" | "1" } 2653 0 - Lazy mode. 2654 Request that DMA unmap operations use deferred 2655 invalidation of hardware TLBs, for increased 2656 throughput at the cost of reduced device isolation. 2657 Will fall back to strict mode if not supported by 2658 the relevant IOMMU driver. 2659 1 - Strict mode. 2660 DMA unmap operations invalidate IOMMU hardware TLBs 2661 synchronously. 2662 unset - Use value of CONFIG_IOMMU_DEFAULT_DMA_{LAZY,STRICT}. 2663 Note: on x86, strict mode specified via one of the 2664 legacy driver-specific options takes precedence. 2665 2666 iommu.passthrough= 2667 [ARM64,X86,EARLY] Configure DMA to bypass the IOMMU by default. 2668 Format: { "0" | "1" } 2669 0 - Use IOMMU translation for DMA. 2670 1 - Bypass the IOMMU for DMA. 2671 unset - Use value of CONFIG_IOMMU_DEFAULT_PASSTHROUGH. 2672 2673 iommu.debug_pagealloc= 2674 [KNL,EARLY] When CONFIG_IOMMU_DEBUG_PAGEALLOC is set, this 2675 parameter enables the feature at boot time. By default, it 2676 is disabled and the system behaves the same way as a kernel 2677 built without CONFIG_IOMMU_DEBUG_PAGEALLOC. 2678 Format: { "0" | "1" } 2679 0 - Sanitizer disabled. 2680 1 - Sanitizer enabled, expect runtime overhead. 2681 2682 io7= [HW] IO7 for Marvel-based Alpha systems 2683 See comment before marvel_specify_io7 in 2684 arch/alpha/kernel/core_marvel.c. 2685 2686 io_delay= [X86,EARLY] I/O delay method 2687 0x80 2688 Standard port 0x80 based delay 2689 0xed 2690 Alternate port 0xed based delay (needed on some systems) 2691 udelay 2692 Simple two microseconds delay 2693 none 2694 No delay 2695 2696 ip= [IP_PNP] 2697 See Documentation/admin-guide/nfs/nfsroot.rst. 2698 2699 ipcmni_extend [KNL,EARLY] Extend the maximum number of unique System V 2700 IPC identifiers from 32,768 to 16,777,216. 2701 2702 ipe.enforce= [IPE] 2703 Format: <bool> 2704 Determine whether IPE starts in permissive (0) or 2705 enforce (1) mode. The default is enforce. 2706 2707 ipe.success_audit= 2708 [IPE] 2709 Format: <bool> 2710 Start IPE with success auditing enabled, emitting 2711 an audit event when a binary is allowed. The default 2712 is 0. 2713 2714 irqaffinity= [SMP] Set the default irq affinity mask 2715 The argument is a cpu list, as described above. 2716 2717 irqchip.gicv2_force_probe= 2718 [ARM,ARM64,EARLY] 2719 Format: <bool> 2720 Force the kernel to look for the second 4kB page 2721 of a GICv2 controller even if the memory range 2722 exposed by the device tree is too small. 2723 2724 irqchip.gicv3_nolpi= 2725 [ARM,ARM64,EARLY] 2726 Force the kernel to ignore the availability of 2727 LPIs (and by consequence ITSs). Intended for system 2728 that use the kernel as a bootloader, and thus want 2729 to let secondary kernels in charge of setting up 2730 LPIs. 2731 2732 irqchip.gicv3_pseudo_nmi= [ARM64,EARLY] 2733 Enables support for pseudo-NMIs in the kernel. This 2734 requires the kernel to be built with 2735 CONFIG_ARM64_PSEUDO_NMI. 2736 2737 irqchip.riscv_imsic_noipi 2738 [RISC-V,EARLY] 2739 Force the kernel to not use IMSIC software injected MSIs 2740 as IPIs. Intended for system where IMSIC is trap-n-emulated, 2741 and thus want to reduce MMIO traps when triggering IPIs 2742 to multiple harts. 2743 2744 irqfixup [HW] 2745 When an interrupt is not handled search all handlers 2746 for it. Intended to get systems with badly broken 2747 firmware running. 2748 2749 irqhandler.duration_warn_us= [KNL] 2750 Warn if an IRQ handler exceeds the specified duration 2751 threshold in microseconds. Useful for identifying 2752 long-running IRQs in the system. 2753 2754 irqpoll [HW] 2755 When an interrupt is not handled search all handlers 2756 for it. Also check all handlers each timer 2757 interrupt. Intended to get systems with badly broken 2758 firmware running. 2759 2760 isapnp= [ISAPNP] 2761 Format: <RDP>,<reset>,<pci_scan>,<verbosity> 2762 2763 isolcpus= [KNL,SMP,ISOL] Isolate a given set of CPUs from disturbance. 2764 Format: [flag-list,]<cpu-list> 2765 2766 Specify one or more CPUs to isolate from disturbances 2767 specified in the flag list (default: domain): 2768 2769 nohz 2770 Disable the tick when a single task runs as well as 2771 disabling other kernel noises like having RCU callbacks 2772 offloaded. This is equivalent to the nohz_full parameter. 2773 2774 A residual 1Hz tick is offloaded to workqueues, which you 2775 need to affine to housekeeping through the global 2776 workqueue's affinity configured via the 2777 /sys/devices/virtual/workqueue/cpumask sysfs file, or 2778 by using the 'domain' flag described below. 2779 2780 NOTE: by default the global workqueue runs on all CPUs, 2781 so to protect individual CPUs the 'cpumask' file has to 2782 be configured manually after bootup. 2783 2784 domain 2785 Isolate from the general SMP balancing and scheduling 2786 algorithms. Note that performing domain isolation this way 2787 is irreversible: it's not possible to bring back a CPU to 2788 the domains once isolated through this boot time 2789 configuration. Use cpusets for a dynamic configuration 2790 which can be altered at runtime. For details see 2791 Documentation/admin-guide/cpu-isolation.rst. 2792 2793 You can move a process onto or off an "isolated" CPU via 2794 the CPU affinity syscalls or cpuset. 2795 <cpu number> begins at 0 and the maximum value is 2796 "number of CPUs in system - 1". 2797 2798 managed_irq 2799 2800 Isolate from being targeted by managed interrupts 2801 which have an interrupt mask containing isolated 2802 CPUs. The affinity of managed interrupts is 2803 handled by the kernel and cannot be changed via 2804 the /proc/irq/* interfaces. 2805 2806 This isolation is best effort and only effective 2807 if the automatically assigned interrupt mask of a 2808 device queue contains isolated and housekeeping 2809 CPUs. If housekeeping CPUs are online then such 2810 interrupts are directed to the housekeeping CPU 2811 so that IO submitted on the housekeeping CPU 2812 cannot disturb the isolated CPU. 2813 2814 If a queue's affinity mask contains only isolated 2815 CPUs then this parameter has no effect on the 2816 interrupt routing decision, though interrupts are 2817 only delivered when tasks running on those 2818 isolated CPUs submit IO. IO submitted on 2819 housekeeping CPUs has no influence on those 2820 queues. 2821 2822 The format of <cpu-list> is described above. 2823 2824 iucv= [HW,NET] 2825 2826 ivrs_ioapic [HW,X86-64] 2827 Provide an override to the IOAPIC-ID<->DEVICE-ID 2828 mapping provided in the IVRS ACPI table. 2829 By default, PCI segment is 0, and can be omitted. 2830 2831 For example, to map IOAPIC-ID decimal 10 to 2832 PCI segment 0x1 and PCI device 00:14.0, 2833 write the parameter as: 2834 ivrs_ioapic=10@0001:00:14.0 2835 2836 Deprecated formats: 2837 * To map IOAPIC-ID decimal 10 to PCI device 00:14.0 2838 write the parameter as: 2839 ivrs_ioapic[10]=00:14.0 2840 * To map IOAPIC-ID decimal 10 to PCI segment 0x1 and 2841 PCI device 00:14.0 write the parameter as: 2842 ivrs_ioapic[10]=0001:00:14.0 2843 2844 ivrs_hpet [HW,X86-64] 2845 Provide an override to the HPET-ID<->DEVICE-ID 2846 mapping provided in the IVRS ACPI table. 2847 By default, PCI segment is 0, and can be omitted. 2848 2849 For example, to map HPET-ID decimal 10 to 2850 PCI segment 0x1 and PCI device 00:14.0, 2851 write the parameter as: 2852 ivrs_hpet=10@0001:00:14.0 2853 2854 Deprecated formats: 2855 * To map HPET-ID decimal 0 to PCI device 00:14.0 2856 write the parameter as: 2857 ivrs_hpet[0]=00:14.0 2858 * To map HPET-ID decimal 10 to PCI segment 0x1 and 2859 PCI device 00:14.0 write the parameter as: 2860 ivrs_ioapic[10]=0001:00:14.0 2861 2862 ivrs_acpihid [HW,X86-64] 2863 Provide an override to the ACPI-HID:UID<->DEVICE-ID 2864 mapping provided in the IVRS ACPI table. 2865 By default, PCI segment is 0, and can be omitted. 2866 2867 For example, to map UART-HID:UID AMD0020:0 to 2868 PCI segment 0x1 and PCI device ID 00:14.5, 2869 write the parameter as: 2870 ivrs_acpihid=AMD0020:0@0001:00:14.5 2871 2872 Deprecated formats: 2873 * To map UART-HID:UID AMD0020:0 to PCI segment is 0, 2874 PCI device ID 00:14.5, write the parameter as: 2875 ivrs_acpihid[00:14.5]=AMD0020:0 2876 * To map UART-HID:UID AMD0020:0 to PCI segment 0x1 and 2877 PCI device ID 00:14.5, write the parameter as: 2878 ivrs_acpihid[0001:00:14.5]=AMD0020:0 2879 2880 js= [HW,JOY] Analog joystick 2881 See Documentation/input/joydev/joystick.rst. 2882 2883 kasan_multi_shot 2884 [KNL] Enforce KASAN (Kernel Address Sanitizer) to print 2885 report on every invalid memory access. Without this 2886 parameter KASAN will print report only for the first 2887 invalid access. 2888 2889 keep_bootcon [KNL,EARLY] 2890 Do not unregister boot console at start. This is only 2891 useful for debugging when something happens in the window 2892 between unregistering the boot console and initializing 2893 the real console. 2894 2895 keepinitrd [HW,ARM] See retain_initrd. 2896 2897 kernelcore= [KNL,X86,PPC,EARLY] 2898 Format: nn[KMGTPE] | nn% | "mirror" 2899 This parameter specifies the amount of memory usable by 2900 the kernel for non-movable allocations. The requested 2901 amount is spread evenly throughout all nodes in the 2902 system as ZONE_NORMAL. The remaining memory is used for 2903 movable memory in its own zone, ZONE_MOVABLE. In the 2904 event, a node is too small to have both ZONE_NORMAL and 2905 ZONE_MOVABLE, kernelcore memory will take priority and 2906 other nodes will have a larger ZONE_MOVABLE. 2907 2908 ZONE_MOVABLE is used for the allocation of pages that 2909 may be reclaimed or moved by the page migration 2910 subsystem. Note that allocations like PTEs-from-HighMem 2911 still use the HighMem zone if it exists, and the Normal 2912 zone if it does not. 2913 2914 It is possible to specify the exact amount of memory in 2915 the form of "nn[KMGTPE]", a percentage of total system 2916 memory in the form of "nn%", or "mirror". If "mirror" 2917 option is specified, mirrored (reliable) memory is used 2918 for non-movable allocations and remaining memory is used 2919 for Movable pages. "nn[KMGTPE]", "nn%", and "mirror" 2920 are exclusive, so you cannot specify multiple forms. 2921 2922 kfence.burst= [MM,KFENCE] The number of additional successive 2923 allocations to be attempted through KFENCE for each 2924 sample interval. 2925 Format: <unsigned integer> 2926 Default: 0 2927 2928 kfence.check_on_panic= 2929 [MM,KFENCE] Whether to check all KFENCE-managed objects' 2930 canaries on panic. 2931 Format: <bool> 2932 Default: false 2933 2934 kfence.deferrable= 2935 [MM,KFENCE] Whether to use a deferrable timer to trigger 2936 allocations. This avoids forcing CPU wake-ups if the 2937 system is idle, at the risk of a less predictable 2938 sample interval. 2939 Format: <bool> 2940 Default: CONFIG_KFENCE_DEFERRABLE 2941 2942 kfence.fault= [MM,KFENCE] Controls the behavior when a KFENCE 2943 error is detected. 2944 report - print the error report and continue (default). 2945 oops - print the error report and oops. 2946 panic - print the error report and panic. 2947 2948 kfence.sample_interval= 2949 [MM,KFENCE] KFENCE's sample interval in milliseconds. 2950 Format: <unsigned integer> 2951 0 - Disable KFENCE. 2952 >0 - Enabled KFENCE with given sample interval. 2953 Default: CONFIG_KFENCE_SAMPLE_INTERVAL 2954 2955 kfence.skip_covered_thresh= 2956 [MM,KFENCE] If pool utilization reaches this threshold 2957 (pool usage%), KFENCE limits currently covered 2958 allocations of the same source from further filling 2959 up the pool. 2960 Format: <unsigned integer> 2961 Default: 75 2962 2963 kgdbdbgp= [KGDB,HW,EARLY] kgdb over EHCI usb debug port. 2964 Format: <Controller#>[,poll interval] 2965 The controller # is the number of the ehci usb debug 2966 port as it is probed via PCI. The poll interval is 2967 optional and is the number seconds in between 2968 each poll cycle to the debug port in case you need 2969 the functionality for interrupting the kernel with 2970 gdb or control-c on the dbgp connection. When 2971 not using this parameter you use sysrq-g to break into 2972 the kernel debugger. 2973 2974 kgdboc= [KGDB,HW] kgdb over consoles. 2975 Requires a tty driver that supports console polling, 2976 or a supported polling keyboard driver (non-usb). 2977 Serial only format: <serial_device>[,baud] 2978 keyboard only format: kbd 2979 keyboard and serial format: kbd,<serial_device>[,baud] 2980 Optional Kernel mode setting: 2981 kms, kbd format: kms,kbd 2982 kms, kbd and serial format: kms,kbd,<ser_dev>[,baud] 2983 2984 kgdboc_earlycon= [KGDB,HW,EARLY] 2985 If the boot console provides the ability to read 2986 characters and can work in polling mode, you can use 2987 this parameter to tell kgdb to use it as a backend 2988 until the normal console is registered. Intended to 2989 be used together with the kgdboc parameter which 2990 specifies the normal console to transition to. 2991 2992 The name of the early console should be specified 2993 as the value of this parameter. Note that the name of 2994 the early console might be different than the tty 2995 name passed to kgdboc. It's OK to leave the value 2996 blank and the first boot console that implements 2997 read() will be picked. 2998 2999 kgdbwait [KGDB,EARLY] Stop kernel execution and enter the 3000 kernel debugger at the earliest opportunity. 3001 3002 kho= [KEXEC,EARLY] 3003 Format: { "0" | "1" | "off" | "on" | "y" | "n" } 3004 Enables or disables Kexec HandOver. 3005 "0" | "off" | "n" - kexec handover is disabled 3006 "1" | "on" | "y" - kexec handover is enabled 3007 3008 kho_scratch= [KEXEC,EARLY] 3009 Format: ll[KMG],mm[KMG],nn[KMG] | nn% 3010 Defines the size of the KHO scratch region. The KHO 3011 scratch regions are physically contiguous memory 3012 ranges that can only be used for non-kernel 3013 allocations. That way, even when memory is heavily 3014 fragmented with handed over memory, the kexeced 3015 kernel will always have enough contiguous ranges to 3016 bootstrap itself. 3017 3018 It is possible to specify the exact amount of 3019 memory in the form of "ll[KMG],mm[KMG],nn[KMG]" 3020 where the first parameter defines the size of a low 3021 memory scratch area, the second parameter defines 3022 the size of a global scratch area and the third 3023 parameter defines the size of additional per-node 3024 scratch areas. The form "nn%" defines scale factor 3025 (in percents) of memory that was used during boot. 3026 3027 kmac= [MIPS] Korina ethernet MAC address. 3028 Configure the RouterBoard 532 series on-chip 3029 Ethernet adapter MAC address. 3030 3031 kmemleak= [KNL,EARLY] Boot-time kmemleak enable/disable 3032 Valid arguments: on, off 3033 Default: on 3034 Built with CONFIG_DEBUG_KMEMLEAK_DEFAULT_OFF=y, 3035 the default is off. 3036 3037 kprobe_event=[probe-list] 3038 [FTRACE] Add kprobe events and enable at boot time. 3039 The probe-list is a semicolon delimited list of probe 3040 definitions. Each definition is same as kprobe_events 3041 interface, but the parameters are comma delimited. 3042 For example, to add a kprobe event on vfs_read with 3043 arg1 and arg2, add to the command line; 3044 3045 kprobe_event=p,vfs_read,$arg1,$arg2 3046 3047 See also Documentation/trace/kprobetrace.rst "Kernel 3048 Boot Parameter" section. 3049 3050 kpti= [ARM64,EARLY] Control page table isolation of 3051 user and kernel address spaces. 3052 Default: enabled on cores which need mitigation. 3053 0: force disabled 3054 1: force enabled 3055 3056 kunit.enable= [KUNIT] Enable executing KUnit tests. Requires 3057 CONFIG_KUNIT to be set to be fully enabled. The 3058 default value can be overridden via 3059 KUNIT_DEFAULT_ENABLED. 3060 Default is 1 (enabled) 3061 3062 kvm.ignore_msrs=[KVM] Ignore guest accesses to unhandled MSRs. 3063 Default is 0 (don't ignore, but inject #GP) 3064 3065 kvm.eager_page_split= 3066 [KVM,X86,RISCV] Controls whether or not KVM will try to 3067 proactively split all huge pages during dirty logging. 3068 Eager page splitting reduces interruptions to vCPU 3069 execution by eliminating the write-protection faults 3070 and MMU lock contention that would otherwise be 3071 required to split huge pages lazily. 3072 3073 VM workloads that rarely perform writes or that write 3074 only to a small region of VM memory may benefit from 3075 disabling eager page splitting to allow huge pages to 3076 still be used for reads. 3077 3078 The behavior of eager page splitting depends on whether 3079 KVM_DIRTY_LOG_INITIALLY_SET is enabled or disabled. If 3080 disabled, all huge pages in a memslot will be eagerly 3081 split when dirty logging is enabled on that memslot. If 3082 enabled, eager page splitting will be performed during 3083 the KVM_CLEAR_DIRTY ioctl, and only for the pages being 3084 cleared. 3085 3086 On x86, eager page splitting is only supported when 3087 kvm.tdp_mmu=Y. 3088 3089 On RISCV, eager page splitting is supported by default. 3090 3091 Default is Y (on). 3092 3093 kvm.enable_pmu=[KVM,X86] 3094 If enabled, KVM will virtualize PMU functionality based 3095 on the virtual CPU model defined by userspace. This 3096 can be overridden on a per-VM basis via 3097 KVM_CAP_PMU_CAPABILITY. 3098 3099 If disabled, KVM will not virtualize PMU functionality, 3100 e.g. MSRs, PMCs, PMIs, etc., even if userspace defines 3101 a virtual CPU model that contains PMU assets. 3102 3103 Note, KVM's vPMU support implicitly requires running 3104 with an in-kernel local APIC, e.g. to deliver PMIs to 3105 the guest. Running without an in-kernel local APIC is 3106 not supported, though KVM will allow such a combination 3107 (with severely degraded functionality). 3108 3109 See also enable_mediated_pmu. 3110 3111 Default is Y (on). 3112 3113 kvm.enable_virt_at_load=[KVM,ARM64,LOONGARCH,MIPS,RISCV,X86] 3114 If enabled, KVM will enable virtualization in hardware 3115 when KVM is loaded, and disable virtualization when KVM 3116 is unloaded (if KVM is built as a module). 3117 3118 If disabled, KVM will dynamically enable and disable 3119 virtualization on-demand when creating and destroying 3120 VMs, i.e. on the 0=>1 and 1=>0 transitions of the 3121 number of VMs. 3122 3123 Enabling virtualization at module load avoids potential 3124 latency for creation of the 0=>1 VM, as KVM serializes 3125 virtualization enabling across all online CPUs. The 3126 "cost" of enabling virtualization when KVM is loaded, 3127 is that doing so may interfere with using out-of-tree 3128 hypervisors that want to "own" virtualization hardware. 3129 3130 kvm.enable_vmware_backdoor=[KVM] Support VMware backdoor PV interface. 3131 Default is false (don't support). 3132 3133 kvm.nx_huge_pages= 3134 [KVM] Controls the software workaround for the 3135 X86_BUG_ITLB_MULTIHIT bug. 3136 force : Always deploy workaround. 3137 off : Never deploy workaround. 3138 auto : Deploy workaround based on the presence of 3139 X86_BUG_ITLB_MULTIHIT. 3140 3141 Default is 'auto'. 3142 3143 If the software workaround is enabled for the host, 3144 guests do need not to enable it for nested guests. 3145 3146 kvm.nx_huge_pages_recovery_ratio= 3147 [KVM] Controls how many 4KiB pages are periodically zapped 3148 back to huge pages. 0 disables the recovery, otherwise if 3149 the value is N KVM will zap 1/Nth of the 4KiB pages every 3150 period (see below). The default is 60. 3151 3152 kvm.nx_huge_pages_recovery_period_ms= 3153 [KVM] Controls the time period at which KVM zaps 4KiB pages 3154 back to huge pages. If the value is a non-zero N, KVM will 3155 zap a portion (see ratio above) of the pages every N msecs. 3156 If the value is 0 (the default), KVM will pick a period based 3157 on the ratio, such that a page is zapped after 1 hour on average. 3158 3159 kvm-{amd,intel}.enable_mediated_pmu=[KVM,AMD,INTEL] 3160 If enabled, KVM will provide a mediated virtual PMU, 3161 instead of the default perf-based virtual PMU (if 3162 kvm.enable_pmu is true and PMU is enumerated via the 3163 virtual CPU model). 3164 3165 With a perf-based vPMU, KVM operates as a user of perf, 3166 i.e. emulates guest PMU counters using perf events. 3167 KVM-created perf events are managed by perf as regular 3168 (guest-only) events, e.g. are scheduled in/out, contend 3169 for hardware resources, etc. Using a perf-based vPMU 3170 allows guest and host usage of the PMU to co-exist, but 3171 incurs non-trivial overhead and can result in silently 3172 dropped guest events (due to resource contention). 3173 3174 With a mediated vPMU, hardware PMU state is context 3175 switched around the world switch to/from the guest. 3176 KVM mediates which events the guest can utilize, but 3177 gives the guest direct access to all other PMU assets 3178 when possible (KVM may intercept some accesses if the 3179 virtual CPU model provides a subset of hardware PMU 3180 functionality). Using a mediated vPMU significantly 3181 reduces PMU virtualization overhead and eliminates lost 3182 guest events, but is mutually exclusive with using perf 3183 to profile KVM guests and adds latency to most VM-Exits 3184 (to context switch PMU state). 3185 3186 Default is N (off). 3187 3188 kvm-amd.nested= [KVM,AMD] Control nested virtualization feature in 3189 KVM/SVM. Default is 1 (enabled). 3190 3191 kvm-amd.npt= [KVM,AMD] Control KVM's use of Nested Page Tables, 3192 a.k.a. Two-Dimensional Page Tables. Default is 1 3193 (enabled). Disable by KVM if hardware lacks support 3194 for NPT. 3195 3196 kvm-amd.ciphertext_hiding_asids= 3197 [KVM,AMD] Ciphertext hiding prevents disallowed accesses 3198 to SNP private memory from reading ciphertext. Instead, 3199 reads will see constant default values (0xff). 3200 3201 If ciphertext hiding is enabled, the joint SEV-ES and 3202 SEV-SNP ASID space is partitioned into separate SEV-ES 3203 and SEV-SNP ASID ranges, with the SEV-SNP range being 3204 [1..max_snp_asid] and the SEV-ES range being 3205 (max_snp_asid..min_sev_asid), where min_sev_asid is 3206 enumerated by CPUID.0x.8000_001F[EDX]. 3207 3208 A non-zero value enables SEV-SNP ciphertext hiding and 3209 adjusts the ASID ranges for SEV-ES and SEV-SNP guests. 3210 KVM caps the number of SEV-SNP ASIDs at the maximum 3211 possible value, e.g. specifying -1u will assign all 3212 joint SEV-ES and SEV-SNP ASIDs to SEV-SNP. Note, 3213 assigning all joint ASIDs to SEV-SNP, i.e. configuring 3214 max_snp_asid == min_sev_asid-1, will effectively make 3215 SEV-ES unusable. 3216 3217 kvm-arm.mode= 3218 [KVM,ARM,EARLY] Select one of KVM/arm64's modes of 3219 operation. 3220 3221 none: Forcefully disable KVM. 3222 3223 nvhe: Standard nVHE-based mode, without support for 3224 protected guests. 3225 3226 protected: Mode with support for guests whose state is 3227 kept private from the host, using VHE or 3228 nVHE depending on HW support. 3229 3230 nested: VHE-based mode with support for nested 3231 virtualization. Requires at least ARMv8.4 3232 hardware (with FEAT_NV2). 3233 3234 Defaults to VHE/nVHE based on hardware support. Setting 3235 mode to "protected" will disable kexec and hibernation 3236 for the host. To force nVHE on VHE hardware, add 3237 "arm64_sw.hvhe=0 id_aa64mmfr1.vh=0" to the 3238 command-line. 3239 "nested" and "protected" are experimental and should be 3240 used with extreme caution. 3241 3242 kvm-arm.vgic_v3_group0_trap= 3243 [KVM,ARM,EARLY] Trap guest accesses to GICv3 group-0 3244 system registers 3245 3246 kvm-arm.vgic_v3_group1_trap= 3247 [KVM,ARM,EARLY] Trap guest accesses to GICv3 group-1 3248 system registers 3249 3250 kvm-arm.vgic_v3_common_trap= 3251 [KVM,ARM,EARLY] Trap guest accesses to GICv3 common 3252 system registers 3253 3254 kvm-arm.vgic_v4_enable= 3255 [KVM,ARM,EARLY] Allow use of GICv4 for direct 3256 injection of LPIs. 3257 3258 kvm-arm.wfe_trap_policy= 3259 [KVM,ARM] Control when to set WFE instruction trap for 3260 KVM VMs. Traps are allowed but not guaranteed by the 3261 CPU architecture. 3262 3263 trap: set WFE instruction trap 3264 3265 notrap: clear WFE instruction trap 3266 3267 kvm-arm.wfi_trap_policy= 3268 [KVM,ARM] Control when to set WFI instruction trap for 3269 KVM VMs. Traps are allowed but not guaranteed by the 3270 CPU architecture. 3271 3272 trap: set WFI instruction trap 3273 3274 notrap: clear WFI instruction trap 3275 3276 kvm_cma_resv_ratio=n [PPC,EARLY] 3277 Reserves given percentage from system memory area for 3278 contiguous memory allocation for KVM hash pagetable 3279 allocation. 3280 By default it reserves 5% of total system memory. 3281 Format: <integer> 3282 Default: 5 3283 3284 kvm-intel.ept= [KVM,Intel] Control KVM's use of Extended Page Tables, 3285 a.k.a. Two-Dimensional Page Tables. Default is 1 3286 (enabled). Disable by KVM if hardware lacks support 3287 for EPT. 3288 3289 kvm-intel.emulate_invalid_guest_state= 3290 [KVM,Intel] Control whether to emulate invalid guest 3291 state. Ignored if kvm-intel.enable_unrestricted_guest=1, 3292 as guest state is never invalid for unrestricted 3293 guests. This param doesn't apply to nested guests (L2), 3294 as KVM never emulates invalid L2 guest state. 3295 Default is 1 (enabled). 3296 3297 kvm-intel.flexpriority= 3298 [KVM,Intel] Control KVM's use of FlexPriority feature 3299 (TPR shadow). Default is 1 (enabled). Disable by KVM if 3300 hardware lacks support for it. 3301 3302 kvm-intel.nested= 3303 [KVM,Intel] Control nested virtualization feature in 3304 KVM/VMX. Default is 1 (enabled). 3305 3306 kvm-intel.unrestricted_guest= 3307 [KVM,Intel] Control KVM's use of unrestricted guest 3308 feature (virtualized real and unpaged mode). Default 3309 is 1 (enabled). Disable by KVM if EPT is disabled or 3310 hardware lacks support for it. 3311 3312 kvm-intel.vmentry_l1d_flush=[KVM,Intel] Mitigation for L1 Terminal Fault 3313 CVE-2018-3620. 3314 3315 Valid arguments: never, cond, always 3316 3317 always: L1D cache flush on every VMENTER. 3318 cond: Flush L1D on VMENTER only when the code between 3319 VMEXIT and VMENTER can leak host memory. 3320 never: Disables the mitigation 3321 3322 Default is cond (do L1 cache flush in specific instances) 3323 3324 kvm-intel.vpid= [KVM,Intel] Control KVM's use of Virtual Processor 3325 Identification feature (tagged TLBs). Default is 1 3326 (enabled). Disable by KVM if hardware lacks support 3327 for it. 3328 3329 l1d_flush= [X86,INTEL,EARLY] 3330 Control mitigation for L1D based snooping vulnerability. 3331 3332 Certain CPUs are vulnerable to an exploit against CPU 3333 internal buffers which can forward information to a 3334 disclosure gadget under certain conditions. 3335 3336 In vulnerable processors, the speculatively 3337 forwarded data can be used in a cache side channel 3338 attack, to access data to which the attacker does 3339 not have direct access. 3340 3341 This parameter controls the mitigation. The 3342 options are: 3343 3344 on - enable the interface for the mitigation 3345 3346 l1tf= [X86,EARLY] Control mitigation of the L1TF vulnerability on 3347 affected CPUs 3348 3349 The kernel PTE inversion protection is unconditionally 3350 enabled and cannot be disabled. 3351 3352 full 3353 Provides all available mitigations for the 3354 L1TF vulnerability. Disables SMT and 3355 enables all mitigations in the 3356 hypervisors, i.e. unconditional L1D flush. 3357 3358 SMT control and L1D flush control via the 3359 sysfs interface is still possible after 3360 boot. Hypervisors will issue a warning 3361 when the first VM is started in a 3362 potentially insecure configuration, 3363 i.e. SMT enabled or L1D flush disabled. 3364 3365 full,force 3366 Same as 'full', but disables SMT and L1D 3367 flush runtime control. Implies the 3368 'nosmt=force' command line option. 3369 (i.e. sysfs control of SMT is disabled.) 3370 3371 flush 3372 Leaves SMT enabled and enables the default 3373 hypervisor mitigation, i.e. conditional 3374 L1D flush. 3375 3376 SMT control and L1D flush control via the 3377 sysfs interface is still possible after 3378 boot. Hypervisors will issue a warning 3379 when the first VM is started in a 3380 potentially insecure configuration, 3381 i.e. SMT enabled or L1D flush disabled. 3382 3383 flush,nosmt 3384 3385 Disables SMT and enables the default 3386 hypervisor mitigation. 3387 3388 SMT control and L1D flush control via the 3389 sysfs interface is still possible after 3390 boot. Hypervisors will issue a warning 3391 when the first VM is started in a 3392 potentially insecure configuration, 3393 i.e. SMT enabled or L1D flush disabled. 3394 3395 flush,nowarn 3396 Same as 'flush', but hypervisors will not 3397 warn when a VM is started in a potentially 3398 insecure configuration. 3399 3400 off 3401 Disables hypervisor mitigations and doesn't 3402 emit any warnings. 3403 It also drops the swap size and available 3404 RAM limit restriction on both hypervisor and 3405 bare metal. 3406 3407 Default is 'flush'. 3408 3409 For details see: Documentation/admin-guide/hw-vuln/l1tf.rst 3410 3411 l2cr= [PPC] 3412 3413 l3cr= [PPC] 3414 3415 lapic [X86-32,APIC,EARLY] Enable the local APIC even if BIOS 3416 disabled it. 3417 3418 lapic= [X86,APIC] Do not use TSC deadline 3419 value for LAPIC timer one-shot implementation. Default 3420 back to the programmable timer unit in the LAPIC. 3421 Format: notscdeadline 3422 3423 lapic_timer_c2_ok [X86,APIC,EARLY] trust the local apic timer 3424 in C2 power state. 3425 3426 libata.dma= [LIBATA] DMA control 3427 libata.dma=0 Disable all PATA and SATA DMA 3428 libata.dma=1 PATA and SATA Disk DMA only 3429 libata.dma=2 ATAPI (CDROM) DMA only 3430 libata.dma=4 Compact Flash DMA only 3431 Combinations also work, so libata.dma=3 enables DMA 3432 for disks and CDROMs, but not CFs. 3433 3434 libata.ignore_hpa= [LIBATA] Ignore HPA limit 3435 libata.ignore_hpa=0 keep BIOS limits (default) 3436 libata.ignore_hpa=1 ignore limits, using full disk 3437 3438 libata.noacpi [LIBATA] Disables use of ACPI in libata suspend/resume 3439 when set. 3440 Format: <int> 3441 3442 libata.force= [LIBATA] Force configurations. The format is a comma- 3443 separated list of "[ID:]VAL" where ID is PORT[.DEVICE]. 3444 PORT and DEVICE are decimal numbers matching port, link 3445 or device. Basically, it matches the ATA ID string 3446 printed on console by libata. If the whole ID part is 3447 omitted, the last PORT and DEVICE values are used. If 3448 ID hasn't been specified yet, the configuration applies 3449 to all ports, links and devices. 3450 3451 If only DEVICE is omitted, the parameter applies to 3452 the port and all links and devices behind it. DEVICE 3453 number of 0 either selects the first device or the 3454 first fan-out link behind PMP device. It does not 3455 select the host link. DEVICE number of 15 selects the 3456 host link and device attached to it. 3457 3458 The VAL specifies the configuration to force. As long 3459 as there is no ambiguity, shortcut notation is allowed. 3460 For example, both 1.5 and 1.5G would work for 1.5Gbps. 3461 The following configurations can be forced. 3462 3463 * Cable type: 40c, 80c, short40c, unk, ign or sata. 3464 Any ID with matching PORT is used. 3465 3466 * SATA link speed limit: 1.5Gbps or 3.0Gbps. 3467 3468 * Transfer mode: pio[0-6], mwdma[0-4] and udma[0-7]. 3469 udma[/][16,25,33,44,66,100,133] notation is also 3470 allowed. 3471 3472 * nohrst, nosrst, norst: suppress hard, soft and both 3473 resets. 3474 3475 * rstonce: only attempt one reset during hot-unplug 3476 link recovery. 3477 3478 * [no]dbdelay: Enable or disable the extra 200ms delay 3479 before debouncing a link PHY and device presence 3480 detection. 3481 3482 * [no]ncq: Turn on or off NCQ. 3483 3484 * [no]ncqtrim: Enable or disable queued DSM TRIM. 3485 3486 * [no]ncqati: Enable or disable NCQ on ATI chipsets. 3487 3488 * [no]trim: Enable or disable (unqueued) TRIM. 3489 3490 * trim_zero: Indicate that TRIM command zeroes data. 3491 3492 * max_trim_128m: Set 128M maximum trim size limit. 3493 3494 * [no]dma: Turn on or off DMA transfers. 3495 3496 * atapi_dmadir: Enable ATAPI DMADIR bridge support. 3497 3498 * atapi_mod16_dma: Enable the use of ATAPI DMA for 3499 commands that are not a multiple of 16 bytes. 3500 3501 * [no]dmalog: Enable or disable the use of the 3502 READ LOG DMA EXT command to access logs. 3503 3504 * [no]iddevlog: Enable or disable access to the 3505 identify device data log. 3506 3507 * [no]logdir: Enable or disable access to the general 3508 purpose log directory. 3509 3510 * max_sec=<sectors>: Set the transfer size limit, in 3511 number of 512-byte sectors, to the value specified in 3512 <sectors>. The value specified in <sectors> has to be 3513 a non-zero positive integer. 3514 3515 * max_sec_128: Set transfer size limit to 128 sectors. 3516 3517 * max_sec_1024: Set transfer size limit to 1024 sectors. 3518 3519 * max_sec_lba48: Set transfer size limit to 65535 sectors. 3520 3521 * external: Mark port as external (hotplug-capable). 3522 3523 * [no]lpm: Enable or disable link power management. 3524 3525 * [no]setxfer: Indicate if transfer speed mode setting 3526 should be skipped. 3527 3528 * [no]fua: Disable or enable FUA (Force Unit Access) 3529 support for devices supporting this feature. 3530 3531 * dump_id: Dump IDENTIFY data. 3532 3533 * disable: Disable this device. 3534 3535 If there are multiple matching configurations changing 3536 the same attribute, the last one is used. 3537 3538 liveupdate= [KNL,EARLY] 3539 Format: <bool> 3540 Enable Live Update Orchestrator (LUO). 3541 Default: off. 3542 3543 lockd.nlm_grace_period=P [NFS] Assign grace period. 3544 Format: <integer> 3545 3546 lockd.nlm_tcpport=N [NFS] Assign TCP port. 3547 Format: <integer> 3548 3549 lockd.nlm_timeout=T [NFS] Assign timeout value. 3550 Format: <integer> 3551 3552 lockd.nlm_udpport=M [NFS] Assign UDP port. 3553 Format: <integer> 3554 3555 lockdown= [SECURITY,EARLY] 3556 { integrity | confidentiality } 3557 Enable the kernel lockdown feature. If set to 3558 integrity, kernel features that allow userland to 3559 modify the running kernel are disabled. If set to 3560 confidentiality, kernel features that allow userland 3561 to extract confidential information from the kernel 3562 are also disabled. 3563 3564 locktorture.acq_writer_lim= [KNL] 3565 Set the time limit in jiffies for a lock 3566 acquisition. Acquisitions exceeding this limit 3567 will result in a splat once they do complete. 3568 3569 locktorture.bind_readers= [KNL] 3570 Specify the list of CPUs to which the readers are 3571 to be bound. 3572 3573 locktorture.bind_writers= [KNL] 3574 Specify the list of CPUs to which the writers are 3575 to be bound. 3576 3577 locktorture.call_rcu_chains= [KNL] 3578 Specify the number of self-propagating call_rcu() 3579 chains to set up. These are used to ensure that 3580 there is a high probability of an RCU grace period 3581 in progress at any given time. Defaults to 0, 3582 which disables these call_rcu() chains. 3583 3584 locktorture.long_hold= [KNL] 3585 Specify the duration in milliseconds for the 3586 occasional long-duration lock hold time. Defaults 3587 to 100 milliseconds. Select 0 to disable. 3588 3589 locktorture.nested_locks= [KNL] 3590 Specify the maximum lock nesting depth that 3591 locktorture is to exercise, up to a limit of 8 3592 (MAX_NESTED_LOCKS). Specify zero to disable. 3593 Note that this parameter is ineffective on types 3594 of locks that do not support nested acquisition. 3595 3596 locktorture.nreaders_stress= [KNL] 3597 Set the number of locking read-acquisition kthreads. 3598 Defaults to being automatically set based on the 3599 number of online CPUs. 3600 3601 locktorture.nwriters_stress= [KNL] 3602 Set the number of locking write-acquisition kthreads. 3603 3604 locktorture.onoff_holdoff= [KNL] 3605 Set time (s) after boot for CPU-hotplug testing. 3606 3607 locktorture.onoff_interval= [KNL] 3608 Set time (s) between CPU-hotplug operations, or 3609 zero to disable CPU-hotplug testing. 3610 3611 locktorture.rt_boost= [KNL] 3612 Do periodic testing of real-time lock priority 3613 boosting. Select 0 to disable, 1 to boost 3614 only rt_mutex, and 2 to boost unconditionally. 3615 Defaults to 2, which might seem to be an 3616 odd choice, but which should be harmless for 3617 non-real-time spinlocks, due to their disabling 3618 of preemption. Note that non-realtime mutexes 3619 disable boosting. 3620 3621 locktorture.rt_boost_factor= [KNL] 3622 Number that determines how often and for how 3623 long priority boosting is exercised. This is 3624 scaled down by the number of writers, so that the 3625 number of boosts per unit time remains roughly 3626 constant as the number of writers increases. 3627 On the other hand, the duration of each boost 3628 increases with the number of writers. 3629 3630 locktorture.shuffle_interval= [KNL] 3631 Set task-shuffle interval (jiffies). Shuffling 3632 tasks allows some CPUs to go into dyntick-idle 3633 mode during the locktorture test. 3634 3635 locktorture.shutdown_secs= [KNL] 3636 Set time (s) after boot system shutdown. This 3637 is useful for hands-off automated testing. 3638 3639 locktorture.stat_interval= [KNL] 3640 Time (s) between statistics printk()s. 3641 3642 locktorture.stutter= [KNL] 3643 Time (s) to stutter testing, for example, 3644 specifying five seconds causes the test to run for 3645 five seconds, wait for five seconds, and so on. 3646 This tests the locking primitive's ability to 3647 transition abruptly to and from idle. 3648 3649 locktorture.torture_type= [KNL] 3650 Specify the locking implementation to test. 3651 3652 locktorture.verbose= [KNL] 3653 Enable additional printk() statements. 3654 3655 locktorture.writer_fifo= [KNL] 3656 Run the write-side locktorture kthreads at 3657 sched_set_fifo() real-time priority. 3658 3659 logibm.irq= [HW,MOUSE] Logitech Bus Mouse Driver 3660 Format: <irq> 3661 3662 loglevel= [KNL,EARLY] 3663 All Kernel Messages with a loglevel smaller than the 3664 console loglevel will be printed to the console. It can 3665 also be changed with klogd or other programs. The 3666 loglevels are defined as follows: 3667 3668 0 (KERN_EMERG) system is unusable 3669 1 (KERN_ALERT) action must be taken immediately 3670 2 (KERN_CRIT) critical conditions 3671 3 (KERN_ERR) error conditions 3672 4 (KERN_WARNING) warning conditions 3673 5 (KERN_NOTICE) normal but significant condition 3674 6 (KERN_INFO) informational 3675 7 (KERN_DEBUG) debug-level messages 3676 3677 log_buf_len=n[KMG] [KNL,EARLY] 3678 Sets the size of the printk ring buffer, in bytes. 3679 n must be a power of two and greater than the 3680 minimal size. The minimal size is defined by 3681 LOG_BUF_SHIFT kernel config parameter. There 3682 is also CONFIG_LOG_CPU_MAX_BUF_SHIFT config 3683 parameter that allows to increase the default size 3684 depending on the number of CPUs. See init/Kconfig 3685 for more details. 3686 3687 logo.nologo [FB] Disables display of the built-in Linux logo. 3688 This may be used to provide more screen space for 3689 kernel log messages and is useful when debugging 3690 kernel boot problems. 3691 3692 lp=0 [LP] Specify parallel ports to use, e.g, 3693 lp=port[,port...] lp=none,parport0 (lp0 not configured, lp1 uses 3694 lp=reset first parallel port). 'lp=0' disables the 3695 lp=auto printer driver. 'lp=reset' (which can be 3696 specified in addition to the ports) causes 3697 attached printers to be reset. Using 3698 lp=port1,port2,... specifies the parallel ports 3699 to associate lp devices with, starting with 3700 lp0. A port specification may be 'none' to skip 3701 that lp device, or a parport name such as 3702 'parport0'. Specifying 'lp=auto' instead of a 3703 port specification list means that device IDs 3704 from each port should be examined, to see if 3705 an IEEE 1284-compliant printer is attached; if 3706 so, the driver will manage that printer. 3707 See also header of drivers/char/lp.c. 3708 3709 lpj=n [KNL] 3710 Sets loops_per_jiffy to given constant, thus avoiding 3711 time-consuming boot-time autodetection (up to 250 ms per 3712 CPU). 0 enables autodetection (default). To determine 3713 the correct value for your kernel, boot with normal 3714 autodetection and see what value is printed. Note that 3715 on SMP systems the preset will be applied to all CPUs, 3716 which is likely to cause problems if your CPUs need 3717 significantly divergent settings. An incorrect value 3718 will cause delays in the kernel to be wrong, leading to 3719 unpredictable I/O errors and other breakage. Although 3720 unlikely, in the extreme case this might damage your 3721 hardware. 3722 3723 lsm.debug [SECURITY] Enable LSM initialization debugging output. 3724 3725 lsm=lsm1,...,lsmN 3726 [SECURITY] Choose order of LSM initialization. This 3727 overrides CONFIG_LSM, and the "security=" parameter. 3728 3729 machtype= [Loongson] Share the same kernel image file between 3730 different yeeloong laptops. 3731 Example: machtype=lemote-yeeloong-2f-7inch 3732 3733 maxcpus= [SMP,EARLY] Maximum number of processors that an SMP kernel 3734 will bring up during bootup. maxcpus=n : n >= 0 limits 3735 the kernel to bring up 'n' processors. Surely after 3736 bootup you can bring up the other plugged cpu by executing 3737 "echo 1 > /sys/devices/system/cpu/cpuX/online". So maxcpus 3738 only takes effect during system bootup. 3739 While n=0 is a special case, it is equivalent to "nosmp", 3740 which also disables the IO APIC. 3741 3742 max_loop= [LOOP] The number of loop block devices that get 3743 (loop.max_loop) unconditionally pre-created at init time. The default 3744 number is configured by BLK_DEV_LOOP_MIN_COUNT. Instead 3745 of statically allocating a predefined number, loop 3746 devices can be requested on-demand with the 3747 /dev/loop-control interface. 3748 3749 mce= [X86-{32,64}] 3750 3751 Please see Documentation/arch/x86/x86_64/machinecheck.rst for sysfs runtime tunables. 3752 3753 off 3754 disable machine check 3755 3756 no_cmci 3757 disable CMCI(Corrected Machine Check Interrupt) that 3758 Intel processor supports. Usually this disablement is 3759 not recommended, but it might be handy if your 3760 hardware is misbehaving. 3761 3762 Note that you'll get more problems without CMCI than 3763 with due to the shared banks, i.e. you might get 3764 duplicated error logs. 3765 3766 dont_log_ce 3767 don't make logs for corrected errors. All events 3768 reported as corrected are silently cleared by OS. This 3769 option will be useful if you have no interest in any 3770 of corrected errors. 3771 3772 ignore_ce 3773 disable features for corrected errors, e.g. 3774 polling timer and CMCI. All events reported as 3775 corrected are not cleared by OS and remained in its 3776 error banks. 3777 3778 Usually this disablement is not recommended, however 3779 if there is an agent checking/clearing corrected 3780 errors (e.g. BIOS or hardware monitoring 3781 applications), conflicting with OS's error handling, 3782 and you cannot deactivate the agent, then this option 3783 will be a help. 3784 3785 no_lmce 3786 do not opt-in to Local MCE delivery. Use legacy method 3787 to broadcast MCEs. 3788 3789 bootlog 3790 enable logging of machine checks left over from 3791 booting. Disabled by default on AMD Fam10h and older 3792 because some BIOS leave bogus ones. 3793 3794 If your BIOS doesn't do that it's a good idea to 3795 enable though to make sure you log even machine check 3796 events that result in a reboot. On Intel systems it is 3797 enabled by default. 3798 3799 nobootlog 3800 disable boot machine check logging. 3801 3802 print_all 3803 print all machine check logs to the console. 3804 3805 monarchtimeout (number) 3806 sets the time in us to wait for other CPUs on machine 3807 checks. 0 to disable. 3808 3809 bios_cmci_threshold 3810 don't overwrite the bios-set CMCI threshold. This boot 3811 option prevents Linux from overwriting the CMCI 3812 threshold set by the bios. Without this option, Linux 3813 always sets the CMCI threshold to 1. Enabling this may 3814 make memory predictive failure analysis less effective 3815 if the bios sets thresholds for memory errors since we 3816 will not see details for all errors. 3817 3818 recovery 3819 force-enable recoverable machine check code paths 3820 3821 Everything else is in sysfs now. 3822 3823 3824 md= [HW] RAID subsystems devices and level 3825 See Documentation/admin-guide/md.rst. 3826 3827 mds= [X86,INTEL,EARLY] 3828 Control mitigation for the Micro-architectural Data 3829 Sampling (MDS) vulnerability. 3830 3831 Certain CPUs are vulnerable to an exploit against CPU 3832 internal buffers which can forward information to a 3833 disclosure gadget under certain conditions. 3834 3835 In vulnerable processors, the speculatively 3836 forwarded data can be used in a cache side channel 3837 attack, to access data to which the attacker does 3838 not have direct access. 3839 3840 This parameter controls the MDS mitigation. The 3841 options are: 3842 3843 full - Enable MDS mitigation on vulnerable CPUs 3844 full,nosmt - Enable MDS mitigation and disable 3845 SMT on vulnerable CPUs 3846 off - Unconditionally disable MDS mitigation 3847 3848 On TAA-affected machines, mds=off can be prevented by 3849 an active TAA mitigation as both vulnerabilities are 3850 mitigated with the same mechanism so in order to disable 3851 this mitigation, you need to specify tsx_async_abort=off 3852 too. 3853 3854 Not specifying this option is equivalent to 3855 mds=full. 3856 3857 For details see: Documentation/admin-guide/hw-vuln/mds.rst 3858 3859 mem=nn[KMG] [HEXAGON,EARLY] Set the memory size. 3860 Must be specified, otherwise memory size will be 0. 3861 3862 mem=nn[KMG] [KNL,BOOT,EARLY] Force usage of a specific amount 3863 of memory Amount of memory to be used in cases 3864 as follows: 3865 3866 1 for test; 3867 2 when the kernel is not able to see the whole system memory; 3868 3 memory that lies after 'mem=' boundary is excluded from 3869 the hypervisor, then assigned to KVM guests. 3870 4 to limit the memory available for kdump kernel. 3871 3872 [ARC,MICROBLAZE] - the limit applies only to low memory, 3873 high memory is not affected. 3874 3875 [ARM64] - only limits memory covered by the linear 3876 mapping. The NOMAP regions are not affected. 3877 3878 [X86] Work as limiting max address. Use together 3879 with memmap= to avoid physical address space collisions. 3880 Without memmap= PCI devices could be placed at addresses 3881 belonging to unused RAM. 3882 3883 Note that this only takes effects during boot time since 3884 in above case 3, memory may need be hot added after boot 3885 if system memory of hypervisor is not sufficient. 3886 3887 mem=nn[KMG]@ss[KMG] 3888 [ARM,MIPS,EARLY] - override the memory layout 3889 reported by firmware. 3890 Define a memory region of size nn[KMG] starting at 3891 ss[KMG]. 3892 Multiple different regions can be specified with 3893 multiple mem= parameters on the command line. 3894 3895 mem=nopentium [BUGS=X86-32] Disable usage of 4MB pages for kernel 3896 memory. 3897 3898 memblock=debug [KNL,EARLY] Enable memblock debug messages. 3899 3900 memchunk=nn[KMG] 3901 [KNL,SH] Allow user to override the default size for 3902 per-device physically contiguous DMA buffers. 3903 3904 memhp_default_state=online/offline/online_kernel/online_movable 3905 [KNL] Set the initial state for the memory hotplug 3906 onlining policy. If not specified, the default value is 3907 set according to the 3908 CONFIG_MHP_DEFAULT_ONLINE_TYPE kernel config 3909 options. 3910 See Documentation/admin-guide/mm/memory-hotplug.rst. 3911 3912 memmap=exactmap [KNL,X86,EARLY] Enable setting of an exact 3913 E820 memory map, as specified by the user. 3914 Such memmap=exactmap lines can be constructed based on 3915 BIOS output or other requirements. See the memmap=nn@ss 3916 option description. 3917 3918 memmap=nn[KMG]@ss[KMG] 3919 [KNL, X86,MIPS,XTENSA,EARLY] Force usage of a specific region of memory. 3920 Region of memory to be used is from ss to ss+nn. 3921 If @ss[KMG] is omitted, it is equivalent to mem=nn[KMG], 3922 which limits max address to nn[KMG]. 3923 Multiple different regions can be specified, 3924 comma delimited. 3925 Example: 3926 memmap=100M@2G,100M#3G,1G!1024G 3927 3928 memmap=nn[KMG]#ss[KMG] 3929 [KNL,ACPI,EARLY] Mark specific memory as ACPI data. 3930 Region of memory to be marked is from ss to ss+nn. 3931 3932 memmap=nn[KMG]$ss[KMG] 3933 [KNL,ACPI,EARLY] Mark specific memory as reserved. 3934 Region of memory to be reserved is from ss to ss+nn. 3935 Example: Exclude memory from 0x18690000-0x1869ffff 3936 memmap=64K$0x18690000 3937 or 3938 memmap=0x10000$0x18690000 3939 Some bootloaders may need an escape character before '$', 3940 like Grub2, otherwise '$' and the following number 3941 will be eaten. 3942 3943 memmap=nn[KMG]!ss[KMG] 3944 [KNL,X86,EARLY] Mark specific memory as protected. 3945 Region of memory to be used, from ss to ss+nn. 3946 The memory region may be marked as e820 type 12 (0xc) 3947 and is NVDIMM or ADR memory. 3948 3949 memmap=<size>%<offset>-<oldtype>+<newtype> 3950 [KNL,ACPI,EARLY] Convert memory within the specified region 3951 from <oldtype> to <newtype>. If "-<oldtype>" is left 3952 out, the whole region will be marked as <newtype>, 3953 even if previously unavailable. If "+<newtype>" is left 3954 out, matching memory will be removed. Types are 3955 specified as e820 types, e.g., 1 = RAM, 2 = reserved, 3956 3 = ACPI, 12 = PRAM. 3957 3958 memory_corruption_check=0/1 [X86,EARLY] 3959 Some BIOSes seem to corrupt the first 64k of 3960 memory when doing things like suspend/resume. 3961 Setting this option will scan the memory 3962 looking for corruption. Enabling this will 3963 both detect corruption and prevent the kernel 3964 from using the memory being corrupted. 3965 However, it's intended as a diagnostic tool; if 3966 repeatable BIOS-originated corruption always 3967 affects the same memory, you can use memmap= 3968 to prevent the kernel from using that memory. 3969 3970 memory_corruption_check_size=size [X86,EARLY] 3971 By default it checks for corruption in the low 3972 64k, making this memory unavailable for normal 3973 use. Use this parameter to scan for 3974 corruption in more or less memory. 3975 3976 memory_corruption_check_period=seconds [X86,EARLY] 3977 By default it checks for corruption every 60 3978 seconds. Use this parameter to check at some 3979 other rate. 0 disables periodic checking. 3980 3981 memory_hotplug.memmap_on_memory 3982 [KNL,X86,ARM] Boolean flag to enable this feature. 3983 Format: {on | off (default)} 3984 When enabled, runtime hotplugged memory will 3985 allocate its internal metadata (struct pages, 3986 those vmemmap pages cannot be optimized even 3987 if hugetlb_free_vmemmap is enabled) from the 3988 hotadded memory which will allow to hotadd a 3989 lot of memory without requiring additional 3990 memory to do so. 3991 This feature is disabled by default because it 3992 has some implication on large (e.g. GB) 3993 allocations in some configurations (e.g. small 3994 memory blocks). 3995 The state of the flag can be read in 3996 /sys/module/memory_hotplug/parameters/memmap_on_memory. 3997 Note that even when enabled, there are a few cases where 3998 the feature is not effective. 3999 4000 mempool_debug [MM] 4001 Enable mempool debugging. This enables element 4002 poison checking when freeing elements back to the 4003 pool. Useful for debugging mempool corruption. 4004 4005 memtest= [KNL,ALPHA,X86,ARM,M68K,PPC,RISCV,EARLY] Enable memtest 4006 Format: <integer> 4007 default : 0 <disable> 4008 Specifies the number of memtest passes to be 4009 performed. Each pass selects another test 4010 pattern from a given set of patterns. Memtest 4011 fills the memory with this pattern, validates 4012 memory contents and reserves bad memory 4013 regions that are detected. 4014 4015 mem_encrypt= [X86-64] AMD Secure Memory Encryption (SME) control 4016 Valid arguments: on, off 4017 Default: off 4018 mem_encrypt=on: Activate SME 4019 mem_encrypt=off: Do not activate SME 4020 4021 Refer to Documentation/virt/kvm/x86/amd-memory-encryption.rst 4022 for details on when memory encryption can be activated. 4023 4024 mem_sleep_default= [SUSPEND] Default system suspend mode: 4025 s2idle - Suspend-To-Idle 4026 shallow - Power-On Suspend or equivalent (if supported) 4027 deep - Suspend-To-RAM or equivalent (if supported) 4028 See Documentation/admin-guide/pm/sleep-states.rst. 4029 4030 mfgptfix [X86-32] Fix MFGPT timers on AMD Geode platforms when 4031 the BIOS has incorrectly applied a workaround. TinyBIOS 4032 version 0.98 is known to be affected, 0.99 fixes the 4033 problem by letting the user disable the workaround. 4034 4035 mga= [HW,DRM] 4036 4037 microcode= [X86] Control the behavior of the microcode loader. 4038 Available options, comma separated: 4039 4040 base_rev=X - with <X> with format: <u32> 4041 Set the base microcode revision of each thread when in 4042 debug mode. 4043 4044 dis_ucode_ldr: disable the microcode loader 4045 4046 force_minrev: 4047 Enable or disable the microcode minimal revision 4048 enforcement for the runtime microcode loader. 4049 4050 mitigations= 4051 [X86,PPC,S390,ARM64,EARLY] Control optional mitigations for 4052 CPU vulnerabilities. This is a set of curated, 4053 arch-independent options, each of which is an 4054 aggregation of existing arch-specific options. 4055 4056 Note, "mitigations" is supported if and only if the 4057 kernel was built with CPU_MITIGATIONS=y. 4058 4059 off 4060 Disable all optional CPU mitigations. This 4061 improves system performance, but it may also 4062 expose users to several CPU vulnerabilities. 4063 Equivalent to: if nokaslr then kpti=0 [ARM64] 4064 gather_data_sampling=off [X86] 4065 indirect_target_selection=off [X86] 4066 kvm.nx_huge_pages=off [X86] 4067 l1tf=off [X86] 4068 mds=off [X86] 4069 mmio_stale_data=off [X86] 4070 no_entry_flush [PPC] 4071 no_uaccess_flush [PPC] 4072 nobp=0 [S390] 4073 nopti [X86,PPC] 4074 nospectre_bhb [ARM64] 4075 nospectre_v1 [X86,PPC] 4076 nospectre_v2 [X86,PPC,S390,ARM64] 4077 reg_file_data_sampling=off [X86] 4078 retbleed=off [X86] 4079 spec_rstack_overflow=off [X86] 4080 spec_store_bypass_disable=off [X86,PPC] 4081 spectre_bhi=off [X86] 4082 spectre_v2_user=off [X86] 4083 srbds=off [X86,INTEL] 4084 ssbd=force-off [ARM64] 4085 tsa=off [X86,AMD] 4086 tsx_async_abort=off [X86] 4087 vmscape=off [X86] 4088 4089 Exceptions: 4090 This does not have any effect on 4091 kvm.nx_huge_pages when 4092 kvm.nx_huge_pages=force. 4093 4094 auto (default) 4095 Mitigate all CPU vulnerabilities, but leave SMT 4096 enabled, even if it's vulnerable. This is for 4097 users who don't want to be surprised by SMT 4098 getting disabled across kernel upgrades, or who 4099 have other ways of avoiding SMT-based attacks. 4100 Equivalent to: (default behavior) 4101 4102 auto,nosmt 4103 Mitigate all CPU vulnerabilities, disabling SMT 4104 if needed. This is for users who always want to 4105 be fully mitigated, even if it means losing SMT. 4106 Equivalent to: l1tf=flush,nosmt [X86] 4107 mds=full,nosmt [X86] 4108 tsx_async_abort=full,nosmt [X86] 4109 mmio_stale_data=full,nosmt [X86] 4110 retbleed=auto,nosmt [X86] 4111 4112 [X86] After one of the above options, additionally 4113 supports attack-vector based controls as documented in 4114 Documentation/admin-guide/hw-vuln/attack_vector_controls.rst 4115 4116 mminit_loglevel= 4117 [KNL,EARLY] When CONFIG_DEBUG_MEMORY_INIT is set, this 4118 parameter allows control of the logging verbosity for 4119 the additional memory initialisation checks. A value 4120 of 0 disables mminit logging and a level of 4 will 4121 log everything. Information is printed at KERN_DEBUG 4122 so loglevel=8 may also need to be specified. 4123 4124 mmio_stale_data= 4125 [X86,INTEL,EARLY] Control mitigation for the Processor 4126 MMIO Stale Data vulnerabilities. 4127 4128 Processor MMIO Stale Data is a class of 4129 vulnerabilities that may expose data after an MMIO 4130 operation. Exposed data could originate or end in 4131 the same CPU buffers as affected by MDS and TAA. 4132 Therefore, similar to MDS and TAA, the mitigation 4133 is to clear the affected CPU buffers. 4134 4135 This parameter controls the mitigation. The 4136 options are: 4137 4138 full - Enable mitigation on vulnerable CPUs 4139 4140 full,nosmt - Enable mitigation and disable SMT on 4141 vulnerable CPUs. 4142 4143 off - Unconditionally disable mitigation 4144 4145 On MDS or TAA affected machines, 4146 mmio_stale_data=off can be prevented by an active 4147 MDS or TAA mitigation as these vulnerabilities are 4148 mitigated with the same mechanism so in order to 4149 disable this mitigation, you need to specify 4150 mds=off and tsx_async_abort=off too. 4151 4152 Not specifying this option is equivalent to 4153 mmio_stale_data=full. 4154 4155 For details see: 4156 Documentation/admin-guide/hw-vuln/processor_mmio_stale_data.rst 4157 4158 <module>.async_probe[=<bool>] [KNL] 4159 If no <bool> value is specified or if the value 4160 specified is not a valid <bool>, enable asynchronous 4161 probe on this module. Otherwise, enable/disable 4162 asynchronous probe on this module as indicated by the 4163 <bool> value. See also: module.async_probe 4164 4165 module.async_probe=<bool> 4166 [KNL] When set to true, modules will use async probing 4167 by default. To enable/disable async probing for a 4168 specific module, use the module specific control that 4169 is documented under <module>.async_probe. When both 4170 module.async_probe and <module>.async_probe are 4171 specified, <module>.async_probe takes precedence for 4172 the specific module. 4173 4174 module.enable_dups_trace 4175 [KNL] When CONFIG_MODULE_DEBUG_AUTOLOAD_DUPS is set, 4176 this means that duplicate request_module() calls will 4177 trigger a WARN_ON() instead of a pr_warn(). Note that 4178 if MODULE_DEBUG_AUTOLOAD_DUPS_TRACE is set, WARN_ON()s 4179 will always be issued and this option does nothing. 4180 module.sig_enforce 4181 [KNL] When CONFIG_MODULE_SIG is set, this means that 4182 modules without (valid) signatures will fail to load. 4183 Note that if CONFIG_MODULE_SIG_FORCE is set, that 4184 is always true, so this option does nothing. 4185 4186 module_blacklist= [KNL] Do not load a comma-separated list of 4187 modules. Useful for debugging problem modules. 4188 4189 mousedev.tap_time= 4190 [MOUSE] Maximum time between finger touching and 4191 leaving touchpad surface for touch to be considered 4192 a tap and be reported as a left button click (for 4193 touchpads working in absolute mode only). 4194 Format: <msecs> 4195 mousedev.xres= [MOUSE] Horizontal screen resolution, used for devices 4196 reporting absolute coordinates, such as tablets 4197 mousedev.yres= [MOUSE] Vertical screen resolution, used for devices 4198 reporting absolute coordinates, such as tablets 4199 4200 movablecore= [KNL,X86,PPC,EARLY] 4201 Format: nn[KMGTPE] | nn% 4202 This parameter is the complement to kernelcore=, it 4203 specifies the amount of memory used for migratable 4204 allocations. If both kernelcore and movablecore is 4205 specified, then kernelcore will be at *least* the 4206 specified value but may be more. If movablecore on its 4207 own is specified, the administrator must be careful 4208 that the amount of memory usable for all allocations 4209 is not too small. 4210 4211 movable_node [KNL,EARLY] Boot-time switch to make hotplugable memory 4212 NUMA nodes to be movable. This means that the memory 4213 of such nodes will be usable only for movable 4214 allocations which rules out almost all kernel 4215 allocations. Use with caution! 4216 4217 MTD_Partition= [MTD] 4218 Format: <name>,<region-number>,<size>,<offset> 4219 4220 MTD_Region= [MTD] Format: 4221 <name>,<region-number>[,<base>,<size>,<buswidth>,<altbuswidth>] 4222 4223 mtdparts= [MTD] 4224 See drivers/mtd/parsers/cmdlinepart.c 4225 4226 mtouchusb.raw_coordinates= 4227 [HW] Make the MicroTouch USB driver use raw coordinates 4228 ('y', default) or cooked coordinates ('n') 4229 4230 mtrr=debug [X86,EARLY] 4231 Enable printing debug information related to MTRR 4232 registers at boot time. 4233 4234 mtrr_chunk_size=nn[KMG] [X86,EARLY] 4235 used for mtrr cleanup. It is largest continuous chunk 4236 that could hold holes aka. UC entries. 4237 4238 mtrr_gran_size=nn[KMG] [X86,EARLY] 4239 Used for mtrr cleanup. It is granularity of mtrr block. 4240 Default is 1. 4241 Large value could prevent small alignment from 4242 using up MTRRs. 4243 4244 mtrr_spare_reg_nr=n [X86,EARLY] 4245 Format: <integer> 4246 Range: 0,7 : spare reg number 4247 Default : 1 4248 Used for mtrr cleanup. It is spare mtrr entries number. 4249 Set to 2 or more if your graphical card needs more. 4250 4251 multitce=off [PPC] This parameter disables the use of the pSeries 4252 firmware feature for updating multiple TCE entries 4253 at a time. 4254 4255 n2= [NET] SDL Inc. RISCom/N2 synchronous serial card 4256 4257 netdev= [NET] NE2000 ISA network devices parameters 4258 Format: <irq>,<io>,<mem_start>,<mem_end>,<name> 4259 4260 netpoll.carrier_timeout= 4261 [NET] Specifies amount of time (in seconds) that 4262 netpoll should wait for a carrier. By default netpoll 4263 waits 4 seconds. 4264 4265 nf_conntrack.acct= 4266 [NETFILTER] Enable connection tracking flow accounting 4267 0 to disable accounting 4268 1 to enable accounting 4269 Default value is 0. 4270 4271 nfs.cache_getent= 4272 [NFS] sets the pathname to the program which is used 4273 to update the NFS client cache entries. 4274 4275 nfs.cache_getent_timeout= 4276 [NFS] sets the timeout after which an attempt to 4277 update a cache entry is deemed to have failed. 4278 4279 nfs.callback_nr_threads= 4280 [NFSv4] set the total number of threads that the 4281 NFS client will assign to service NFSv4 callback 4282 requests. 4283 4284 nfs.callback_tcpport= 4285 [NFS] set the TCP port on which the NFSv4 callback 4286 channel should listen. 4287 4288 nfs.delay_retrans= 4289 [NFS] specifies the number of times the NFSv4 client 4290 retries the request before returning an EAGAIN error, 4291 after a reply of NFS4ERR_DELAY from the server. 4292 Only applies if the softerr mount option is enabled, 4293 and the specified value is >= 0. 4294 4295 nfs.idmap_cache_timeout= 4296 [NFS] set the maximum lifetime for idmapper cache 4297 entries. 4298 4299 nfs.max_session_cb_slots= 4300 [NFSv4.1] Sets the maximum number of session 4301 slots the client will assign to the callback 4302 channel. This determines the maximum number of 4303 callbacks the client will process in parallel for 4304 a particular server. 4305 4306 nfs.max_session_slots= 4307 [NFSv4.1] Sets the maximum number of session slots 4308 the client will attempt to negotiate with the server. 4309 This limits the number of simultaneous RPC requests 4310 that the client can send to the NFSv4.1 server. 4311 Note that there is little point in setting this 4312 value higher than the max_tcp_slot_table_limit. 4313 4314 nfs.nfs4_disable_idmapping= 4315 [NFSv4] When set to the default of '1', this option 4316 ensures that both the RPC level authentication 4317 scheme and the NFS level operations agree to use 4318 numeric uids/gids if the mount is using the 4319 'sec=sys' security flavour. In effect it is 4320 disabling idmapping, which can make migration from 4321 legacy NFSv2/v3 systems to NFSv4 easier. 4322 Servers that do not support this mode of operation 4323 will be autodetected by the client, and it will fall 4324 back to using the idmapper. 4325 To turn off this behaviour, set the value to '0'. 4326 4327 nfs.nfs4_unique_id= 4328 [NFS4] Specify an additional fixed unique ident- 4329 ification string that NFSv4 clients can insert into 4330 their nfs_client_id4 string. This is typically a 4331 UUID that is generated at system install time. 4332 4333 nfs.recover_lost_locks= 4334 [NFSv4] Attempt to recover locks that were lost due 4335 to a lease timeout on the server. Please note that 4336 doing this risks data corruption, since there are 4337 no guarantees that the file will remain unchanged 4338 after the locks are lost. 4339 If you want to enable the kernel legacy behaviour of 4340 attempting to recover these locks, then set this 4341 parameter to '1'. 4342 The default parameter value of '0' causes the kernel 4343 not to attempt recovery of lost locks. 4344 4345 nfs.send_implementation_id= 4346 [NFSv4.1] Send client implementation identification 4347 information in exchange_id requests. 4348 If zero, no implementation identification information 4349 will be sent. 4350 The default is to send the implementation identification 4351 information. 4352 4353 nfs4.layoutstats_timer= 4354 [NFSv4.2] Change the rate at which the kernel sends 4355 layoutstats to the pNFS metadata server. 4356 4357 Setting this to value to 0 causes the kernel to use 4358 whatever value is the default set by the layout 4359 driver. A non-zero value sets the minimum interval 4360 in seconds between layoutstats transmissions. 4361 4362 nfsd.inter_copy_offload_enable= 4363 [NFSv4.2] When set to 1, the server will support 4364 server-to-server copies for which this server is 4365 the destination of the copy. 4366 4367 nfsd.nfs4_disable_idmapping= 4368 [NFSv4] When set to the default of '1', the NFSv4 4369 server will return only numeric uids and gids to 4370 clients using auth_sys, and will accept numeric uids 4371 and gids from such clients. This is intended to ease 4372 migration from NFSv2/v3. 4373 4374 nfsd.nfsd4_ssc_umount_timeout= 4375 [NFSv4.2] When used as the destination of a 4376 server-to-server copy, knfsd temporarily mounts 4377 the source server. It caches the mount in case 4378 it will be needed again, and discards it if not 4379 used for the number of milliseconds specified by 4380 this parameter. 4381 4382 nfsaddrs= [NFS] Deprecated. Use ip= instead. 4383 See Documentation/admin-guide/nfs/nfsroot.rst. 4384 4385 nfsroot= [NFS] nfs root filesystem for disk-less boxes. 4386 See Documentation/admin-guide/nfs/nfsroot.rst. 4387 4388 nfsrootdebug [NFS] enable nfsroot debugging messages. 4389 See Documentation/admin-guide/nfs/nfsroot.rst. 4390 4391 nmi_backtrace.backtrace_idle [KNL] 4392 Dump stacks even of idle CPUs in response to an 4393 NMI stack-backtrace request. 4394 4395 nmi_debug= [KNL,SH] Specify one or more actions to take 4396 when a NMI is triggered. 4397 Format: [state][,regs][,debounce][,die] 4398 4399 nmi_watchdog= [KNL,BUGS=X86] Debugging features for SMP kernels 4400 Format: [panic,][nopanic,][rNNN,][num] 4401 Valid num: 0 or 1 4402 0 - turn hardlockup detector in nmi_watchdog off 4403 1 - turn hardlockup detector in nmi_watchdog on 4404 rNNN - configure the watchdog with raw perf event 0xNNN 4405 4406 When panic is specified, panic when an NMI watchdog 4407 timeout occurs (or 'nopanic' to not panic on an NMI 4408 watchdog, if CONFIG_BOOTPARAM_HARDLOCKUP_PANIC is set) 4409 To disable both hard and soft lockup detectors, 4410 please see 'nowatchdog'. 4411 This is useful when you use a panic=... timeout and 4412 need the box quickly up again. 4413 4414 These settings can be accessed at runtime via 4415 the nmi_watchdog and hardlockup_panic sysctls. 4416 4417 no4lvl [RISCV,EARLY] Disable 4-level and 5-level paging modes. 4418 Forces kernel to use 3-level paging instead. 4419 4420 no5lvl [X86-64,RISCV,EARLY] Disable 5-level paging mode. Forces 4421 kernel to use 4-level paging instead. 4422 4423 noalign [KNL,ARM] 4424 4425 noapic [SMP,APIC,EARLY] Tells the kernel to not make use of any 4426 IOAPICs that may be present in the system. 4427 4428 noapictimer [APIC,X86] Don't set up the APIC timer 4429 4430 noautogroup Disable scheduler automatic task group creation. 4431 4432 nocache [ARM,EARLY] 4433 4434 no_console_suspend 4435 [HW] Never suspend the console 4436 Disable suspending of consoles during suspend and 4437 hibernate operations. Once disabled, debugging 4438 messages can reach various consoles while the rest 4439 of the system is being put to sleep (ie, while 4440 debugging driver suspend/resume hooks). This may 4441 not work reliably with all consoles, but is known 4442 to work with serial and VGA consoles. 4443 To facilitate more flexible debugging, we also add 4444 console_suspend, a printk module parameter to control 4445 it. Users could use console_suspend (usually 4446 /sys/module/printk/parameters/console_suspend) to 4447 turn on/off it dynamically. 4448 4449 no_debug_objects 4450 [KNL,EARLY] Disable object debugging 4451 4452 nodsp [SH] Disable hardware DSP at boot time. 4453 4454 noefi [EFI,EARLY] Disable EFI runtime services support. 4455 4456 no_entry_flush [PPC,EARLY] Don't flush the L1-D cache when entering the kernel. 4457 4458 noexec32 [X86-64] 4459 This affects only 32-bit executables. 4460 noexec32=on: enable non-executable mappings (default) 4461 read doesn't imply executable mappings 4462 noexec32=off: disable non-executable mappings 4463 read implies executable mappings 4464 4465 no_file_caps Tells the kernel not to honor file capabilities. The 4466 only way then for a file to be executed with privilege 4467 is to be setuid root or executed by root. 4468 4469 nofpu [MIPS,SH] Disable hardware FPU at boot time. 4470 4471 nofsgsbase [X86] Disables FSGSBASE instructions. 4472 4473 nofxsr [BUGS=X86-32] Disables x86 floating point extended 4474 register save and restore. The kernel will only save 4475 legacy floating-point registers on task switch. 4476 4477 nogbpages [X86] Do not use GB pages for kernel direct mappings. 4478 4479 no_hash_pointers 4480 [KNL,EARLY] 4481 Alias for "hash_pointers=never". 4482 4483 nohibernate [HIBERNATION] Disable hibernation and resume. 4484 4485 nohlt [ARM,ARM64,MICROBLAZE,MIPS,PPC,RISCV,SH] Forces the kernel to 4486 busy wait in do_idle() and not use the arch_cpu_idle() 4487 implementation; requires CONFIG_GENERIC_IDLE_POLL_SETUP 4488 to be effective. This is useful on platforms where the 4489 sleep(SH) or wfi(ARM,ARM64) instructions do not work 4490 correctly or when doing power measurements to evaluate 4491 the impact of the sleep instructions. This is also 4492 useful when using JTAG debugger. 4493 4494 nohpet [X86] Don't use the HPET timer. 4495 4496 nohugeiomap [KNL,X86,PPC,ARM64,EARLY] Disable kernel huge I/O mappings. 4497 4498 nohugevmalloc [KNL,X86,PPC,ARM64,EARLY] Disable kernel huge vmalloc mappings. 4499 4500 nohz= [KNL] Boottime enable/disable dynamic ticks 4501 Valid arguments: on, off 4502 Default: on 4503 4504 nohz_full= [KNL,BOOT,SMP,ISOL] 4505 The argument is a cpu list, as described above. 4506 In kernels built with CONFIG_NO_HZ_FULL=y, set 4507 the specified list of CPUs whose tick will be stopped 4508 whenever possible. The boot CPU will be forced outside 4509 the range to maintain the timekeeping. Any CPUs 4510 in this list will have their RCU callbacks offloaded, 4511 just as if they had also been called out in the 4512 rcu_nocbs= boot parameter. 4513 4514 Note that this argument takes precedence over 4515 the CONFIG_RCU_NOCB_CPU_DEFAULT_ALL option. 4516 4517 noinitrd [Deprecated,RAM] Tells the kernel not to load any configured 4518 initial RAM disk. Currently this parameter applies to 4519 initrd only, not to initramfs. But it applies to both 4520 in EFI mode. 4521 4522 nointremap [X86-64,Intel-IOMMU,EARLY] Do not enable interrupt 4523 remapping. 4524 [Deprecated - use intremap=off] 4525 4526 noinvpcid [X86,EARLY] Disable the INVPCID cpu feature. 4527 4528 noiotrap [SH] Disables trapped I/O port accesses. 4529 4530 noirqdebug [X86-32] Disables the code which attempts to detect and 4531 disable unhandled interrupt sources. 4532 4533 noisapnp [ISAPNP] Disables ISA PnP code. 4534 4535 nokaslr [KNL,EARLY] 4536 When CONFIG_RANDOMIZE_BASE is set, this disables 4537 kernel and module base offset ASLR (Address Space 4538 Layout Randomization). 4539 4540 no-kvmapf [X86,KVM,EARLY] Disable paravirtualized asynchronous page 4541 fault handling. 4542 4543 no-kvmclock [X86,KVM,EARLY] Disable paravirtualized KVM clock driver 4544 4545 nolapic [X86-32,APIC,EARLY] Do not enable or use the local APIC. 4546 4547 nolapic_timer [X86-32,APIC,EARLY] Do not use the local APIC timer. 4548 4549 nomce [X86-32] Disable Machine Check Exception 4550 4551 nomodeset Disable kernel modesetting. Most systems' firmware 4552 sets up a display mode and provides framebuffer memory 4553 for output. With nomodeset, DRM and fbdev drivers will 4554 not load if they could possibly displace the pre- 4555 initialized output. Only the system framebuffer will 4556 be available for use. The respective drivers will not 4557 perform display-mode changes or accelerated rendering. 4558 4559 Useful as error fallback, or for testing and debugging. 4560 4561 nomodule Disable module load 4562 4563 nonmi_ipi [X86] Disable using NMI IPIs during panic/reboot to 4564 shutdown the other cpus. Instead use the REBOOT_VECTOR 4565 irq. 4566 4567 nopat [X86,EARLY] Disable PAT (page attribute table extension of 4568 pagetables) support. 4569 4570 nopcid [X86-64,EARLY] Disable the PCID cpu feature. 4571 4572 nopku [X86] Disable Memory Protection Keys CPU feature found 4573 in some Intel CPUs. 4574 4575 nopti [X86-64,EARLY] 4576 Equivalent to pti=off 4577 4578 nopv= [X86,XEN,KVM,HYPER_V,VMWARE,EARLY] 4579 Disables the PV optimizations forcing the guest to run 4580 as generic guest with no PV drivers. Currently support 4581 XEN HVM, KVM, HYPER_V and VMWARE guest. 4582 4583 nopvspin [X86,XEN,KVM,EARLY] 4584 Disables the qspinlock slow path using PV optimizations 4585 which allow the hypervisor to 'idle' the guest on lock 4586 contention. 4587 4588 norandmaps Don't use address space randomization. Equivalent to 4589 echo 0 > /proc/sys/kernel/randomize_va_space 4590 4591 noresume [SWSUSP] Disables resume and restores original swap 4592 space. 4593 4594 no-scroll [VGA] Disables scrollback. 4595 This is required for the Braillex ib80-piezo Braille 4596 reader made by F.H. Papenmeier (Germany). 4597 4598 nosgx [X86-64,SGX,EARLY] Disables Intel SGX kernel support. 4599 4600 nosmap [PPC,EARLY] 4601 Disable SMAP (Supervisor Mode Access Prevention) 4602 even if it is supported by processor. 4603 4604 nosmep [PPC64s,EARLY] 4605 Disable SMEP (Supervisor Mode Execution Prevention) 4606 even if it is supported by processor. 4607 4608 nosmp [SMP,EARLY] Tells an SMP kernel to act as a UP kernel, 4609 and disable the IO APIC. legacy for "maxcpus=0". 4610 4611 nosmt [KNL,MIPS,PPC,EARLY] Disable symmetric multithreading (SMT). 4612 Equivalent to smt=1. 4613 4614 [KNL,LOONGARCH,X86,ARM64,PPC,S390] Disable symmetric multithreading (SMT). 4615 nosmt=force: Force disable SMT, cannot be undone 4616 via the sysfs control file. 4617 4618 nosoftlockup [KNL] Disable the soft-lockup detector. 4619 4620 nospec_store_bypass_disable 4621 [HW,EARLY] Disable all mitigations for the Speculative 4622 Store Bypass vulnerability 4623 4624 nospectre_bhb [ARM64,EARLY] Disable all mitigations for Spectre-BHB (branch 4625 history injection) vulnerability. System may allow data leaks 4626 with this option. 4627 4628 nospectre_v1 [X86,PPC,EARLY] Disable mitigations for Spectre Variant 1 4629 (bounds check bypass). With this option data leaks are 4630 possible in the system. 4631 4632 nospectre_v2 [X86,PPC_E500,ARM64,EARLY] Disable all mitigations 4633 for the Spectre variant 2 (indirect branch 4634 prediction) vulnerability. System may allow data 4635 leaks with this option. 4636 4637 no-steal-acc [X86,PV_OPS,ARM64,PPC/PSERIES,RISCV,LOONGARCH,EARLY] 4638 Disable paravirtualized steal time accounting. steal time 4639 is computed, but won't influence scheduler behaviour 4640 4641 nosync [HW,M68K] Disables sync negotiation for all devices. 4642 4643 no_timer_check [X86,APIC] Disables the code which tests for broken 4644 timer IRQ sources, i.e., the IO-APIC timer. This can 4645 work around problems with incorrect timer 4646 initialization on some boards. 4647 4648 no_uaccess_flush 4649 [PPC,EARLY] Don't flush the L1-D cache after accessing user data. 4650 4651 novmcoredd [KNL,KDUMP] 4652 Disable device dump. Device dump allows drivers to 4653 append dump data to vmcore so you can collect driver 4654 specified debug info. Drivers can append the data 4655 without any limit and this data is stored in memory, 4656 so this may cause significant memory stress. Disabling 4657 device dump can help save memory but the driver debug 4658 data will be no longer available. This parameter 4659 is only available when CONFIG_PROC_VMCORE_DEVICE_DUMP 4660 is set. 4661 4662 no-vmw-sched-clock 4663 [X86,PV_OPS,EARLY] Disable paravirtualized VMware 4664 scheduler clock and use the default one. 4665 4666 nowatchdog [KNL] Disable both lockup detectors, i.e. 4667 soft-lockup and NMI watchdog (hard-lockup). 4668 4669 nowb [ARM,EARLY] 4670 4671 nox2apic [X86-64,APIC,EARLY] Do not enable x2APIC mode. 4672 4673 NOTE: this parameter will be ignored on systems with the 4674 LEGACY_XAPIC_DISABLED bit set in the 4675 IA32_XAPIC_DISABLE_STATUS MSR. 4676 4677 noxsave [BUGS=X86] Disables x86 extended register state save 4678 and restore using xsave. The kernel will fallback to 4679 enabling legacy floating-point and sse state. 4680 4681 noxsaveopt [X86] Disables xsaveopt used in saving x86 extended 4682 register states. The kernel will fall back to use 4683 xsave to save the states. By using this parameter, 4684 performance of saving the states is degraded because 4685 xsave doesn't support modified optimization while 4686 xsaveopt supports it on xsaveopt enabled systems. 4687 4688 noxsaves [X86] Disables xsaves and xrstors used in saving and 4689 restoring x86 extended register state in compacted 4690 form of xsave area. The kernel will fall back to use 4691 xsaveopt and xrstor to save and restore the states 4692 in standard form of xsave area. By using this 4693 parameter, xsave area per process might occupy more 4694 memory on xsaves enabled systems. 4695 4696 nr_cpus= [SMP,EARLY] Maximum number of processors that an SMP kernel 4697 could support. nr_cpus=n : n >= 1 limits the kernel to 4698 support 'n' processors. It could be larger than the 4699 number of already plugged CPU during bootup, later in 4700 runtime you can physically add extra cpu until it reaches 4701 n. So during boot up some boot time memory for per-cpu 4702 variables need be pre-allocated for later physical cpu 4703 hot plugging. 4704 4705 nr_uarts= [SERIAL] maximum number of UARTs to be registered. 4706 4707 numa=off [KNL, ARM64, PPC, RISCV, SPARC, X86, EARLY] 4708 Disable NUMA, Only set up a single NUMA node 4709 spanning all memory. 4710 4711 numa=fake=<size>[MG] 4712 [KNL, ARM64, RISCV, X86, EARLY] 4713 If given as a memory unit, fills all system RAM with 4714 nodes of size interleaved over physical nodes. 4715 4716 numa=fake=<N> 4717 [KNL, ARM64, RISCV, X86, EARLY] 4718 If given as an integer, fills all system RAM with N 4719 fake nodes interleaved over physical nodes. 4720 4721 numa=fake=<N>U 4722 [KNL, ARM64, RISCV, X86, EARLY] 4723 If given as an integer followed by 'U', it will 4724 divide each physical node into N emulated nodes. 4725 4726 numa=noacpi [X86] Don't parse the SRAT table for NUMA setup 4727 4728 numa=nohmat [X86] Don't parse the HMAT table for NUMA setup, or 4729 soft-reserved memory partitioning. 4730 4731 numa_balancing= [KNL,ARM64,PPC,RISCV,S390,X86] Enable or disable automatic 4732 NUMA balancing. 4733 Allowed values are enable and disable 4734 4735 numa_zonelist_order= [KNL, BOOT] Select zonelist order for NUMA. 4736 'node', 'default' can be specified 4737 This can be set from sysctl after boot. 4738 See Documentation/admin-guide/sysctl/vm.rst for details. 4739 4740 nvme.quirks= [NVME] A list of quirk entries to augment the built-in 4741 nvme quirk list. List entries are separated by a 4742 '-' character. 4743 Each entry has the form VendorID:ProductID:quirk_names. 4744 The IDs are 4-digits hex numbers and quirk_names is a 4745 list of quirk names separated by commas. A quirk name 4746 can be prefixed by '^', meaning that the specified 4747 quirk must be disabled. 4748 4749 Example: 4750 nvme.quirks=7710:2267:bogus_nid,^identify_cns-9900:7711:broken_msi 4751 4752 ohci1394_dma=early [HW,EARLY] enable debugging via the ohci1394 driver. 4753 See Documentation/core-api/debugging-via-ohci1394.rst for more 4754 info. 4755 4756 olpc_ec_timeout= [OLPC] ms delay when issuing EC commands 4757 Rather than timing out after 20 ms if an EC 4758 command is not properly ACKed, override the length 4759 of the timeout. We have interrupts disabled while 4760 waiting for the ACK, so if this is set too high 4761 interrupts *may* be lost! 4762 4763 onenand.bdry= [HW,MTD] Flex-OneNAND Boundary Configuration 4764 4765 Format: [die0_boundary][,die0_lock][,die1_boundary][,die1_lock] 4766 4767 boundary - index of last SLC block on Flex-OneNAND. 4768 The remaining blocks are configured as MLC blocks. 4769 lock - Configure if Flex-OneNAND boundary should be locked. 4770 Once locked, the boundary cannot be changed. 4771 1 indicates lock status, 0 indicates unlock status. 4772 4773 oops=panic [KNL,EARLY] 4774 Always panic on oopses. Default is to just kill the 4775 process, but there is a small probability of 4776 deadlocking the machine. 4777 This will also cause panics on machine check exceptions. 4778 Useful together with panic=30 to trigger a reboot. 4779 4780 page_alloc.shuffle= 4781 [KNL] Boolean flag to control whether the page allocator 4782 should randomize its free lists. This parameter can be 4783 used to enable/disable page randomization. The state of 4784 the flag can be read from sysfs at: 4785 /sys/module/page_alloc/parameters/shuffle. 4786 This parameter is only available if CONFIG_SHUFFLE_PAGE_ALLOCATOR=y. 4787 4788 page_owner= [KNL,EARLY] Boot-time page_owner enabling option. 4789 Storage of the information about who allocated 4790 each page is disabled in default. With this switch, 4791 we can turn it on. 4792 on: enable the feature 4793 4794 page_poison= [KNL,EARLY] Boot-time parameter changing the state of 4795 poisoning on the buddy allocator, available with 4796 CONFIG_PAGE_POISONING=y. 4797 off: turn off poisoning (default) 4798 on: turn on poisoning 4799 4800 page_reporting.page_reporting_order= 4801 [KNL] Minimal page reporting order 4802 Format: <integer> 4803 Adjust the minimal page reporting order. The page 4804 reporting is disabled when it exceeds MAX_PAGE_ORDER. 4805 4806 page_reporting.page_reporting_delay_ms= 4807 [KNL] Free page reporting delay in milliseconds 4808 Format: <unsigned integer> 4809 Adjust the delay in milliseconds between free page 4810 reporting intervals. Default is 2000 (2 seconds). 4811 4812 panic= [KNL] Kernel behaviour on panic: delay <timeout> 4813 timeout > 0: seconds before rebooting 4814 timeout = 0: wait forever 4815 timeout < 0: reboot immediately 4816 Format: <timeout> 4817 4818 panic_on_taint= [KNL,EARLY] 4819 Bitmask for conditionally calling panic() in add_taint() 4820 Format: <hex>[,nousertaint] 4821 Hexadecimal bitmask representing the set of TAINT flags 4822 that will cause the kernel to panic when add_taint() is 4823 called with any of the flags in this set. 4824 The optional switch "nousertaint" can be utilized to 4825 prevent userspace forced crashes by writing to sysctl 4826 /proc/sys/kernel/tainted any flagset matching with the 4827 bitmask set on panic_on_taint. 4828 See Documentation/admin-guide/tainted-kernels.rst for 4829 extra details on the taint flags that users can pick 4830 to compose the bitmask to assign to panic_on_taint. 4831 4832 panic_on_warn=1 panic() instead of WARN(). Useful to cause kdump 4833 on a WARN(). 4834 4835 panic_force_cpu= 4836 [KNL,SMP] Force panic handling to execute on a specific CPU. 4837 Format: <cpu number> 4838 Some platforms require panic handling to occur on a 4839 specific CPU for the crash kernel to function correctly. 4840 This can be due to firmware limitations, interrupt routing 4841 constraints, or platform-specific requirements where only 4842 a particular CPU can safely enter the crash kernel. 4843 When set, panic() will redirect execution to the specified 4844 CPU before proceeding with the normal panic and kexec flow. 4845 If the target CPU is offline or unavailable, panic proceeds 4846 on the current CPU. 4847 This option should only be used for systems with the above 4848 constraints as it might cause the panic operation to be less reliable. 4849 4850 panic_print= Bitmask for printing system info when panic happens. 4851 User can chose combination of the following bits: 4852 4853 bit 0: print all tasks info 4854 bit 1: print system memory info 4855 bit 2: print timer info 4856 bit 3: print locks info if CONFIG_LOCKDEP is on 4857 bit 4: print ftrace buffer 4858 bit 5: replay all kernel messages on consoles at the end of panic 4859 bit 6: print all CPUs backtrace (if available in the arch) 4860 bit 7: print only tasks in uninterruptible (blocked) state 4861 4862 *Be aware* that this option may print a _lot_ of lines. 4863 4864 There is an increased risk of losing older messages in 4865 the log. Maybe worth to setup a bigger log buffer with 4866 "log_buf_len" along with this. 4867 4868 Also consider using "printk.console_no_auto_verbose=Y" 4869 if using this along with pstore, to avoid extra delays 4870 due to increased console verbosity during panic. 4871 4872 panic_sys_info= A comma separated list of extra information to be dumped 4873 on panic. 4874 Format: val[,val...] 4875 Where @val can be any of the following: 4876 4877 tasks: print all tasks info 4878 mem: print system memory info 4879 timers: print timers info 4880 locks: print locks info if CONFIG_LOCKDEP is on 4881 ftrace: print ftrace buffer 4882 all_bt: print all CPUs backtrace (if available in the arch) 4883 blocked_tasks: print only tasks in uninterruptible (blocked) state 4884 4885 This is a human readable alternative to the 'panic_print' 4886 option above. Please be aware of the caveats using this, 4887 check the 'panic_print' parameter for more details. 4888 4889 panic_console_replay 4890 When panic happens, replay all kernel messages on 4891 consoles at the end of panic. 4892 4893 parkbd.port= [HW] Parallel port number the keyboard adapter is 4894 connected to, default is 0. 4895 Format: <parport#> 4896 parkbd.mode= [HW] Parallel port keyboard adapter mode of operation, 4897 0 for XT, 1 for AT (default is AT). 4898 Format: <mode> 4899 4900 parport= [HW,PPT] Specify parallel ports. 0 disables. 4901 Format: { 0 | auto | 0xBBB[,IRQ[,DMA]] } 4902 Use 'auto' to force the driver to use any 4903 IRQ/DMA settings detected (the default is to 4904 ignore detected IRQ/DMA settings because of 4905 possible conflicts). You can specify the base 4906 address, IRQ, and DMA settings; IRQ and DMA 4907 should be numbers, or 'auto' (for using detected 4908 settings on that particular port), or 'nofifo' 4909 (to avoid using a FIFO even if it is detected). 4910 Parallel ports are assigned in the order they 4911 are specified on the command line, starting 4912 with parport0. 4913 4914 parport_init_mode= [HW,PPT] 4915 Configure VIA parallel port to operate in 4916 a specific mode. This is necessary on Pegasos 4917 computer where firmware has no options for setting 4918 up parallel port mode and sets it to spp. 4919 Currently this function knows 686a and 8231 chips. 4920 Format: [spp|ps2|epp|ecp|ecpepp] 4921 4922 pata_legacy.all= [HW,LIBATA] 4923 Format: <int> 4924 Set to non-zero to probe primary and secondary ISA 4925 port ranges on PCI systems where no PCI PATA device 4926 has been found at either range. Disabled by default. 4927 4928 pata_legacy.autospeed= [HW,LIBATA] 4929 Format: <int> 4930 Set to non-zero if a chip is present that snoops speed 4931 changes. Disabled by default. 4932 4933 pata_legacy.iordy_mask= [HW,LIBATA] 4934 Format: <int> 4935 IORDY enable mask. Set individual bits to allow IORDY 4936 for the respective channel. Bit 0 is for the first 4937 legacy channel handled by this driver, bit 1 is for 4938 the second channel, and so on. The sequence will often 4939 correspond to the primary legacy channel, the secondary 4940 legacy channel, and so on, but the handling of a PCI 4941 bus and the use of other driver options may interfere 4942 with the sequence. By default IORDY is allowed across 4943 all channels. 4944 4945 pata_legacy.pio_mask= [HW,LIBATA] 4946 Format: <int> 4947 PIO mode mask for autospeed devices. Set individual 4948 bits to allow the use of the respective PIO modes. 4949 Bit 0 is for mode 0, bit 1 is for mode 1, and so on. 4950 All modes allowed by default. 4951 4952 pata_legacy.probe_all= [HW,LIBATA] 4953 Format: <int> 4954 Set to non-zero to probe tertiary and further ISA 4955 port ranges on PCI systems. Disabled by default. 4956 4957 pata_legacy.probe_mask= [HW,LIBATA] 4958 Format: <int> 4959 Probe mask for legacy ISA PATA ports. Depending on 4960 platform configuration and the use of other driver 4961 options up to 6 legacy ports are supported: 0x1f0, 4962 0x170, 0x1e8, 0x168, 0x1e0, 0x160, however probing 4963 of individual ports can be disabled by setting the 4964 corresponding bits in the mask to 1. Bit 0 is for 4965 the first port in the list above (0x1f0), and so on. 4966 By default all supported ports are probed. 4967 4968 pata_platform.pio_mask= [HW,LIBATA] 4969 Format: <int> 4970 Supported PIO mode mask. Set individual bits to allow 4971 the use of the respective PIO modes. Bit 0 is for 4972 mode 0, bit 1 is for mode 1, and so on. Mode 0 only 4973 allowed by default. 4974 4975 pause_on_oops=<int> 4976 Halt all CPUs after the first oops has been printed for 4977 the specified number of seconds. This is to be used if 4978 your oopses keep scrolling off the screen. 4979 4980 pci=option[,option...] [PCI,EARLY] various PCI subsystem options. 4981 4982 Some options herein operate on a specific device 4983 or a set of devices (<pci_dev>). These are 4984 specified in one of the following formats: 4985 4986 [<domain>:]<bus>:<dev>.<func>[/<dev>.<func>]* 4987 pci:<vendor>:<device>[:<subvendor>:<subdevice>] 4988 4989 Note: the first format specifies a PCI 4990 bus/device/function address which may change 4991 if new hardware is inserted, if motherboard 4992 firmware changes, or due to changes caused 4993 by other kernel parameters. If the 4994 domain is left unspecified, it is 4995 taken to be zero. Optionally, a path 4996 to a device through multiple device/function 4997 addresses can be specified after the base 4998 address (this is more robust against 4999 renumbering issues). The second format 5000 selects devices using IDs from the 5001 configuration space which may match multiple 5002 devices in the system. 5003 5004 earlydump dump PCI config space before the kernel 5005 changes anything 5006 off [X86] don't probe for the PCI bus 5007 bios [X86-32] force use of PCI BIOS, don't access 5008 the hardware directly. Use this if your machine 5009 has a non-standard PCI host bridge. 5010 nobios [X86-32] disallow use of PCI BIOS, only direct 5011 hardware access methods are allowed. Use this 5012 if you experience crashes upon bootup and you 5013 suspect they are caused by the BIOS. 5014 conf1 [X86] Force use of PCI Configuration Access 5015 Mechanism 1 (config address in IO port 0xCF8, 5016 data in IO port 0xCFC, both 32-bit). 5017 conf2 [X86] Force use of PCI Configuration Access 5018 Mechanism 2 (IO port 0xCF8 is an 8-bit port for 5019 the function, IO port 0xCFA, also 8-bit, sets 5020 bus number. The config space is then accessed 5021 through ports 0xC000-0xCFFF). 5022 See http://wiki.osdev.org/PCI for more info 5023 on the configuration access mechanisms. 5024 noaer [PCIE] If the PCIEAER kernel config parameter is 5025 enabled, this kernel boot option can be used to 5026 disable the use of PCIE advanced error reporting. 5027 nodomains [PCI] Disable support for multiple PCI 5028 root domains (aka PCI segments, in ACPI-speak). 5029 nommconf [X86] Disable use of MMCONFIG for PCI 5030 Configuration 5031 check_enable_amd_mmconf [X86] check for and enable 5032 properly configured MMIO access to PCI 5033 config space on AMD family 10h CPU 5034 nomsi [MSI] If the PCI_MSI kernel config parameter is 5035 enabled, this kernel boot option can be used to 5036 disable the use of MSI interrupts system-wide. 5037 noioapicquirk [APIC] Disable all boot interrupt quirks. 5038 Safety option to keep boot IRQs enabled. This 5039 should never be necessary. 5040 ioapicreroute [APIC] Enable rerouting of boot IRQs to the 5041 primary IO-APIC for bridges that cannot disable 5042 boot IRQs. This fixes a source of spurious IRQs 5043 when the system masks IRQs. 5044 noioapicreroute [APIC] Disable workaround that uses the 5045 boot IRQ equivalent of an IRQ that connects to 5046 a chipset where boot IRQs cannot be disabled. 5047 The opposite of ioapicreroute. 5048 biosirq [X86-32] Use PCI BIOS calls to get the interrupt 5049 routing table. These calls are known to be buggy 5050 on several machines and they hang the machine 5051 when used, but on other computers it's the only 5052 way to get the interrupt routing table. Try 5053 this option if the kernel is unable to allocate 5054 IRQs or discover secondary PCI buses on your 5055 motherboard. 5056 rom [X86] Assign address space to expansion ROMs. 5057 Use with caution as certain devices share 5058 address decoders between ROMs and other 5059 resources. 5060 norom [X86] Do not assign address space to 5061 expansion ROMs that do not already have 5062 BIOS assigned address ranges. 5063 nobar [X86] Do not assign address space to the 5064 BARs that weren't assigned by the BIOS. 5065 irqmask=0xMMMM [X86] Set a bit mask of IRQs allowed to be 5066 assigned automatically to PCI devices. You can 5067 make the kernel exclude IRQs of your ISA cards 5068 this way. 5069 pirqaddr=0xAAAAA [X86] Specify the physical address 5070 of the PIRQ table (normally generated 5071 by the BIOS) if it is outside the 5072 F0000h-100000h range. 5073 lastbus=N [X86] Scan all buses thru bus #N. Can be 5074 useful if the kernel is unable to find your 5075 secondary buses and you want to tell it 5076 explicitly which ones they are. 5077 assign-busses [X86] Always assign all PCI bus 5078 numbers ourselves, overriding 5079 whatever the firmware may have done. 5080 usepirqmask [X86] Honor the possible IRQ mask stored 5081 in the BIOS $PIR table. This is needed on 5082 some systems with broken BIOSes, notably 5083 some HP Pavilion N5400 and Omnibook XE3 5084 notebooks. This will have no effect if ACPI 5085 IRQ routing is enabled. 5086 noacpi [X86] Do not use ACPI for IRQ routing 5087 or for PCI scanning. 5088 use_crs [X86] Use PCI host bridge window information 5089 from ACPI. On BIOSes from 2008 or later, this 5090 is enabled by default. If you need to use this, 5091 please report a bug. 5092 nocrs [X86] Ignore PCI host bridge windows from ACPI. 5093 If you need to use this, please report a bug. 5094 use_e820 [X86] Use E820 reservations to exclude parts of 5095 PCI host bridge windows. This is a workaround 5096 for BIOS defects in host bridge _CRS methods. 5097 If you need to use this, please report a bug to 5098 <linux-pci@vger.kernel.org>. 5099 no_e820 [X86] Ignore E820 reservations for PCI host 5100 bridge windows. This is the default on modern 5101 hardware. If you need to use this, please report 5102 a bug to <linux-pci@vger.kernel.org>. 5103 routeirq Do IRQ routing for all PCI devices. 5104 This is normally done in pci_enable_device(), 5105 so this option is a temporary workaround 5106 for broken drivers that don't call it. 5107 skip_isa_align [X86] do not align io start addr, so can 5108 handle more pci cards 5109 noearly [X86] Don't do any early type 1 scanning. 5110 This might help on some broken boards which 5111 machine check when some devices' config space 5112 is read. But various workarounds are disabled 5113 and some IOMMU drivers will not work. 5114 bfsort Sort PCI devices into breadth-first order. 5115 This sorting is done to get a device 5116 order compatible with older (<= 2.4) kernels. 5117 nobfsort Don't sort PCI devices into breadth-first order. 5118 pcie_bus_tune_off Disable PCIe MPS (Max Payload Size) 5119 tuning and use the BIOS-configured MPS defaults. 5120 pcie_bus_safe Set every device's MPS to the largest value 5121 supported by all devices below the root complex. 5122 pcie_bus_perf Set device MPS to the largest allowable MPS 5123 based on its parent bus. Also set MRRS (Max 5124 Read Request Size) to the largest supported 5125 value (no larger than the MPS that the device 5126 or bus can support) for best performance. 5127 pcie_bus_peer2peer Set every device's MPS to 128B, which 5128 every device is guaranteed to support. This 5129 configuration allows peer-to-peer DMA between 5130 any pair of devices, possibly at the cost of 5131 reduced performance. This also guarantees 5132 that hot-added devices will work. 5133 cbiosize=nn[KMG] The fixed amount of bus space which is 5134 reserved for the CardBus bridge's IO window. 5135 The default value is 256 bytes. 5136 cbmemsize=nn[KMG] The fixed amount of bus space which is 5137 reserved for the CardBus bridge's memory 5138 window. The default value is 64 megabytes. 5139 resource_alignment= 5140 Format: 5141 [<order of align>@]<pci_dev>[; ...] 5142 Specifies alignment and device to reassign 5143 aligned memory resources. How to 5144 specify the device is described above. 5145 If <order of align> is not specified, 5146 PAGE_SIZE is used as alignment. 5147 A PCI-PCI bridge can be specified if resource 5148 windows need to be expanded. 5149 To specify the alignment for several 5150 instances of a device, the PCI vendor, 5151 device, subvendor, and subdevice may be 5152 specified, e.g., 12@pci:8086:9c22:103c:198f 5153 for 4096-byte alignment. 5154 ecrc= Enable/disable PCIe ECRC (transaction layer 5155 end-to-end CRC checking). Only effective if 5156 OS has native AER control (either granted by 5157 ACPI _OSC or forced via "pcie_ports=native") 5158 bios: Use BIOS/firmware settings. This is the 5159 the default. 5160 off: Turn ECRC off 5161 on: Turn ECRC on. 5162 hpiosize=nn[KMG] The fixed amount of bus space which is 5163 reserved for hotplug bridge's IO window. 5164 Default size is 256 bytes. 5165 hpmmiosize=nn[KMG] The fixed amount of bus space which is 5166 reserved for hotplug bridge's MMIO window. 5167 Default size is 2 megabytes. 5168 hpmmioprefsize=nn[KMG] The fixed amount of bus space which is 5169 reserved for hotplug bridge's MMIO_PREF window. 5170 Default size is 2 megabytes. 5171 hpmemsize=nn[KMG] The fixed amount of bus space which is 5172 reserved for hotplug bridge's MMIO and 5173 MMIO_PREF window. 5174 Default size is 2 megabytes. 5175 hpbussize=nn The minimum amount of additional bus numbers 5176 reserved for buses below a hotplug bridge. 5177 Default is 1. 5178 realloc= Enable/disable reallocating PCI bridge resources 5179 if allocations done by BIOS are too small to 5180 accommodate resources required by all child 5181 devices. 5182 off: Turn realloc off 5183 on: Turn realloc on 5184 realloc same as realloc=on 5185 noari do not use PCIe ARI. 5186 noats [PCIE, Intel-IOMMU, AMD-IOMMU] 5187 do not use PCIe ATS (and IOMMU device IOTLB). 5188 pcie_scan_all Scan all possible PCIe devices. Otherwise we 5189 only look for one device below a PCIe downstream 5190 port. 5191 big_root_window Try to add a big 64bit memory window to the PCIe 5192 root complex on AMD CPUs. Some GFX hardware 5193 can resize a BAR to allow access to all VRAM. 5194 Adding the window is slightly risky (it may 5195 conflict with unreported devices), so this 5196 taints the kernel. 5197 disable_acs_redir=<pci_dev>[; ...] 5198 Specify one or more PCI devices (in the format 5199 specified above) separated by semicolons. 5200 Each device specified will have the PCI ACS 5201 redirect capabilities forced off which will 5202 allow P2P traffic between devices through 5203 bridges without forcing it upstream. Note: 5204 this removes isolation between devices and 5205 may put more devices in an IOMMU group. 5206 config_acs= 5207 Format: 5208 <ACS flags>@<pci_dev>[; ...] 5209 Specify one or more PCI devices (in the format 5210 specified above) optionally prepended with flags 5211 and separated by semicolons. The respective 5212 capabilities will be enabled, disabled or 5213 unchanged based on what is specified in 5214 flags. 5215 5216 ACS Flags is defined as follows: 5217 bit-0 : ACS Source Validation 5218 bit-1 : ACS Translation Blocking 5219 bit-2 : ACS P2P Request Redirect 5220 bit-3 : ACS P2P Completion Redirect 5221 bit-4 : ACS Upstream Forwarding 5222 bit-5 : ACS P2P Egress Control 5223 bit-6 : ACS Direct Translated P2P 5224 Each bit can be marked as: 5225 '0' – force disabled 5226 '1' – force enabled 5227 'x' – unchanged 5228 For example, 5229 pci=config_acs=10x@pci:0:0 5230 would configure all devices that support 5231 ACS to enable P2P Request Redirect, disable 5232 Translation Blocking, and leave Source 5233 Validation unchanged from whatever power-up 5234 or firmware set it to. 5235 5236 Note: this may remove isolation between devices 5237 and may put more devices in an IOMMU group. 5238 force_floating [S390] Force usage of floating interrupts. 5239 nomio [S390] Do not use MIO instructions. 5240 norid [S390] ignore the RID field and force use of 5241 one PCI domain per PCI function 5242 notph [PCIE] If the PCIE_TPH kernel config parameter 5243 is enabled, this kernel boot option can be used 5244 to disable PCIe TLP Processing Hints support 5245 system-wide. 5246 5247 pcie_aspm= [PCIE] Forcibly enable or ignore PCIe Active State Power 5248 Management. 5249 off Don't touch ASPM configuration at all. Leave any 5250 configuration done by firmware unchanged. 5251 force Enable ASPM even on devices that claim not to support it. 5252 WARNING: Forcing ASPM on may cause system lockups. 5253 5254 pcie_ports= [PCIE] PCIe port services handling: 5255 native Use native PCIe services (PME, AER, DPC, PCIe hotplug) 5256 even if the platform doesn't give the OS permission to 5257 use them. This may cause conflicts if the platform 5258 also tries to use these services. 5259 dpc-native Use native PCIe service for DPC only. May 5260 cause conflicts if firmware uses AER or DPC. 5261 compat Disable native PCIe services (PME, AER, DPC, PCIe 5262 hotplug). 5263 5264 pcie_port_pm= [PCIE] PCIe port power management handling: 5265 off Disable power management of all PCIe ports 5266 force Forcibly enable power management of all PCIe ports 5267 5268 pcie_pme= [PCIE,PM] Native PCIe PME signaling options: 5269 nomsi Do not use MSI for native PCIe PME signaling (this makes 5270 all PCIe root ports use INTx for all services). 5271 5272 pd_ignore_unused 5273 [PM] 5274 Keep all power-domains already enabled by bootloader on, 5275 even if no driver has claimed them. This is useful 5276 for debug and development, but should not be 5277 needed on a platform with proper driver support. 5278 5279 pdcchassis= [PARISC,HW] Disable/Enable PDC Chassis Status codes at 5280 boot time. 5281 Format: { 0 | 1 } 5282 See arch/parisc/kernel/pdc_chassis.c 5283 5284 percpu_alloc= [MM,EARLY] 5285 Select which percpu first chunk allocator to use. 5286 Currently supported values are "embed" and "page". 5287 Archs may support subset or none of the selections. 5288 See comments in mm/percpu.c for details on each 5289 allocator. This parameter is primarily for debugging 5290 and performance comparison. 5291 5292 pirq= [SMP,APIC] Manual mp-table setup 5293 See Documentation/arch/x86/i386/IO-APIC.rst. 5294 5295 plip= [PPT,NET] Parallel port network link 5296 Format: { parport<nr> | timid | 0 } 5297 See also Documentation/admin-guide/parport.rst. 5298 5299 pmtmr= [X86] Manual setup of pmtmr I/O Port. 5300 Override pmtimer IOPort with a hex value. 5301 e.g. pmtmr=0x508 5302 5303 pmu_override= [PPC] Override the PMU. 5304 This option takes over the PMU facility, so it is no 5305 longer usable by perf. Setting this option starts the 5306 PMU counters by setting MMCR0 to 0 (the FC bit is 5307 cleared). If a number is given, then MMCR1 is set to 5308 that number, otherwise (e.g., 'pmu_override=on'), MMCR1 5309 remains 0. 5310 5311 pm_async= [PM] 5312 Format: off 5313 This parameter sets the initial value of the 5314 /sys/power/pm_async sysfs knob at boot time. 5315 If set to "off", disables asynchronous suspend and 5316 resume of devices during system-wide power transitions. 5317 This can be useful on platforms where device 5318 dependencies are not well-defined, or for debugging 5319 power management issues. Asynchronous operations are 5320 enabled by default. 5321 5322 5323 pm_debug_messages [SUSPEND,KNL] 5324 Enable suspend/resume debug messages during boot up. 5325 5326 pm.dpm_watchdog_enabled= 5327 [PM] Enable or disable the DPM watchdog. Requires 5328 CONFIG_PM_SLEEP and CONFIG_DPM_WATCHDOG enabled. 5329 Format: <bool> 5330 Default value is set by CONFIG_DPM_WATCHDOG_ENABLED. 5331 5332 pnp.debug=1 [PNP] 5333 Enable PNP debug messages (depends on the 5334 CONFIG_PNP_DEBUG_MESSAGES option). Change at run-time 5335 via /sys/module/pnp/parameters/debug. We always show 5336 current resource usage; turning this on also shows 5337 possible settings and some assignment information. 5338 5339 pnpacpi= [ACPI] 5340 { off } 5341 5342 pnpbios= [ISAPNP] 5343 { on | off | curr | res | no-curr | no-res } 5344 5345 pnp_reserve_irq= 5346 [ISAPNP] Exclude IRQs for the autoconfiguration 5347 5348 pnp_reserve_dma= 5349 [ISAPNP] Exclude DMAs for the autoconfiguration 5350 5351 pnp_reserve_io= [ISAPNP] Exclude I/O ports for the autoconfiguration 5352 Ranges are in pairs (I/O port base and size). 5353 5354 pnp_reserve_mem= 5355 [ISAPNP] Exclude memory regions for the 5356 autoconfiguration. 5357 Ranges are in pairs (memory base and size). 5358 5359 ports= [IP_VS_FTP] IPVS ftp helper module 5360 Default is 21. 5361 Up to 8 (IP_VS_APP_MAX_PORTS) ports 5362 may be specified. 5363 Format: <port>,<port>.... 5364 5365 possible_cpus= [SMP,S390,X86] 5366 Format: <unsigned int> 5367 Set the number of possible CPUs, overriding the 5368 regular discovery mechanisms (such as ACPI/FW, etc). 5369 5370 powersave=off [PPC] This option disables power saving features. 5371 It specifically disables cpuidle and sets the 5372 platform machine description specific power_save 5373 function to NULL. On Idle the CPU just reduces 5374 execution priority. 5375 5376 ppc_strict_facility_enable 5377 [PPC,ENABLE] This option catches any kernel floating point, 5378 Altivec, VSX and SPE outside of regions specifically 5379 allowed (eg kernel_enable_fpu()/kernel_disable_fpu()). 5380 There is some performance impact when enabling this. 5381 5382 ppc_tm= [PPC,EARLY] 5383 Format: {"off"} 5384 Disable Hardware Transactional Memory 5385 5386 preempt= [KNL] 5387 Select preemption mode if you have CONFIG_PREEMPT_DYNAMIC 5388 none - Limited to cond_resched() calls 5389 voluntary - Limited to cond_resched() and might_sleep() calls 5390 full - Any section that isn't explicitly preempt disabled 5391 can be preempted anytime. Tasks will also yield 5392 contended spinlocks (if the critical section isn't 5393 explicitly preempt disabled beyond the lock itself). 5394 lazy - Scheduler controlled. Similar to full but instead 5395 of preempting the task immediately, the task gets 5396 one HZ tick time to yield itself before the 5397 preemption will be forced. One such preemption 5398 point is when the task returns to user space. 5399 5400 print-fatal-signals= 5401 [KNL] debug: print fatal signals 5402 5403 If enabled, warn about various signal handling 5404 related application anomalies: too many signals, 5405 too many POSIX.1 timers, fatal signals causing a 5406 coredump - etc. 5407 5408 If you hit the warning due to signal overflow, 5409 you might want to try "ulimit -i unlimited". 5410 5411 default: off. 5412 5413 printk.always_kmsg_dump= 5414 Trigger kmsg_dump for cases other than kernel oops or 5415 panics 5416 Format: <bool> (1/Y/y=enable, 0/N/n=disable) 5417 default: disabled 5418 5419 printk.console_no_auto_verbose= 5420 Disable console loglevel raise on oops, panic 5421 or lockdep-detected issues (only if lock debug is on). 5422 With an exception to setups with low baudrate on 5423 serial console, keeping this 0 is a good choice 5424 in order to provide more debug information. 5425 Format: <bool> 5426 default: 0 (auto_verbose is enabled) 5427 5428 printk.debug_non_panic_cpus= 5429 Allows storing messages from non-panic CPUs into 5430 the printk log buffer during panic(). They are 5431 flushed to consoles by the panic-CPU on 5432 a best-effort basis. 5433 Format: <bool> (1/Y/y=enable, 0/N/n=disable) 5434 Default: disabled 5435 5436 printk.devkmsg={on,off,ratelimit} 5437 Control writing to /dev/kmsg. 5438 on - unlimited logging to /dev/kmsg from userspace 5439 off - logging to /dev/kmsg disabled 5440 ratelimit - ratelimit the logging 5441 Default: ratelimit 5442 5443 printk.time= Show timing data prefixed to each printk message line 5444 Format: <bool> (1/Y/y=enable, 0/N/n=disable) 5445 5446 proc_mem.force_override= [KNL] 5447 Format: {always | ptrace | never} 5448 Traditionally /proc/pid/mem allows memory permissions to be 5449 overridden without restrictions. This option may be set to 5450 restrict that. Can be one of: 5451 - 'always': traditional behavior always allows mem overrides. 5452 - 'ptrace': only allow mem overrides for active ptracers. 5453 - 'never': never allow mem overrides. 5454 If not specified, default is the CONFIG_PROC_MEM_* choice. 5455 5456 processor.max_cstate= [HW,ACPI] 5457 Limit processor to maximum C-state 5458 max_cstate=9 overrides any DMI blacklist limit. 5459 5460 processor.nocst [HW,ACPI] 5461 Ignore the _CST method to determine C-states, 5462 instead using the legacy FADT method 5463 5464 profile= [KNL] Enable kernel profiling via /proc/profile 5465 Format: [<profiletype>,]<number> 5466 Param: <profiletype>: "schedule" or "kvm" 5467 [defaults to kernel profiling] 5468 Param: "schedule" - profile schedule points. 5469 Param: "kvm" - profile VM exits. 5470 Param: <number> - step/bucket size as a power of 2 for 5471 statistical time based profiling. 5472 5473 prot_virt= [S390] enable hosting protected virtual machines 5474 isolated from the hypervisor (if hardware supports 5475 that). If enabled, the default kernel base address 5476 might be overridden even when Kernel Address Space 5477 Layout Randomization is disabled. 5478 Format: <bool> 5479 5480 psi= [KNL] Enable or disable pressure stall information 5481 tracking. 5482 Format: <bool> 5483 5484 psmouse.proto= [HW,MOUSE] Highest PS2 mouse protocol extension to 5485 probe for; one of (bare|imps|exps|lifebook|any). 5486 psmouse.rate= [HW,MOUSE] Set desired mouse report rate, in reports 5487 per second. 5488 psmouse.resetafter= [HW,MOUSE] 5489 Try to reset the device after so many bad packets 5490 (0 = never). 5491 psmouse.resolution= 5492 [HW,MOUSE] Set desired mouse resolution, in dpi. 5493 psmouse.smartscroll= 5494 [HW,MOUSE] Controls Logitech smartscroll autorepeat. 5495 0 = disabled, 1 = enabled (default). 5496 5497 pstore.backend= Specify the name of the pstore backend to use 5498 5499 pti= [X86-64] Control Page Table Isolation of user and 5500 kernel address spaces. Disabling this feature 5501 removes hardening, but improves performance of 5502 system calls and interrupts. 5503 5504 on - unconditionally enable 5505 off - unconditionally disable 5506 auto - kernel detects whether your CPU model is 5507 vulnerable to issues that PTI mitigates 5508 5509 Not specifying this option is equivalent to pti=auto. 5510 5511 pty.legacy_count= 5512 [KNL] Number of legacy pty's. Overwrites compiled-in 5513 default number. 5514 5515 quiet [KNL,EARLY] Disable most log messages 5516 5517 r128= [HW,DRM] 5518 5519 radix_hcall_invalidate=on [PPC/PSERIES] 5520 Disable RADIX GTSE feature and use hcall for TLB 5521 invalidate. 5522 5523 raid= [HW,RAID] 5524 See Documentation/admin-guide/md.rst. 5525 5526 ramdisk_size= [RAM] Sizes of RAM disks in kilobytes 5527 See Documentation/admin-guide/blockdev/ramdisk.rst. 5528 5529 ramdisk_start= [Deprecated,RAM] RAM disk image start address 5530 5531 random.trust_cpu=off 5532 [KNL,EARLY] Disable trusting the use of the CPU's 5533 random number generator (if available) to 5534 initialize the kernel's RNG. 5535 5536 random.trust_bootloader=off 5537 [KNL,EARLY] Disable trusting the use of the a seed 5538 passed by the bootloader (if available) to 5539 initialize the kernel's RNG. 5540 5541 randomize_kstack_offset= 5542 [KNL,EARLY] Enable or disable kernel stack offset 5543 randomization, which provides roughly 5 bits of 5544 entropy, frustrating memory corruption attacks 5545 that depend on stack address determinism or 5546 cross-syscall address exposures. This is only 5547 available on architectures that have defined 5548 CONFIG_HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET. 5549 Format: <bool> (1/Y/y=enable, 0/N/n=disable) 5550 Default is CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT. 5551 5552 ras=option[,option,...] [KNL] RAS-specific options 5553 5554 cec_disable [X86] 5555 Disable the Correctable Errors Collector, 5556 see CONFIG_RAS_CEC help text. 5557 5558 rcu_nocbs[=cpu-list] 5559 [KNL] The optional argument is a cpu list, 5560 as described above. 5561 5562 In kernels built with CONFIG_RCU_NOCB_CPU=y, 5563 enable the no-callback CPU mode, which prevents 5564 such CPUs' callbacks from being invoked in 5565 softirq context. Invocation of such CPUs' RCU 5566 callbacks will instead be offloaded to "rcuox/N" 5567 kthreads created for that purpose, where "x" is 5568 "p" for RCU-preempt, "s" for RCU-sched, and "g" 5569 for the kthreads that mediate grace periods; and 5570 "N" is the CPU number. This reduces OS jitter on 5571 the offloaded CPUs, which can be useful for HPC 5572 and real-time workloads. It can also improve 5573 energy efficiency for asymmetric multiprocessors. 5574 5575 If a cpulist is passed as an argument, the specified 5576 list of CPUs is set to no-callback mode from boot. 5577 5578 Otherwise, if the '=' sign and the cpulist 5579 arguments are omitted, no CPU will be set to 5580 no-callback mode from boot but the mode may be 5581 toggled at runtime via cpusets. 5582 5583 Note that this argument takes precedence over 5584 the CONFIG_RCU_NOCB_CPU_DEFAULT_ALL option. 5585 5586 rcu_nocb_poll [KNL] 5587 Rather than requiring that offloaded CPUs 5588 (specified by rcu_nocbs= above) explicitly 5589 awaken the corresponding "rcuoN" kthreads, 5590 make these kthreads poll for callbacks. 5591 This improves the real-time response for the 5592 offloaded CPUs by relieving them of the need to 5593 wake up the corresponding kthread, but degrades 5594 energy efficiency by requiring that the kthreads 5595 periodically wake up to do the polling. 5596 5597 rcutree.blimit= [KNL] 5598 Set maximum number of finished RCU callbacks to 5599 process in one batch. 5600 5601 rcutree.csd_lock_suppress_rcu_stall= [KNL] 5602 Do only a one-line RCU CPU stall warning when 5603 there is an ongoing too-long CSD-lock wait. 5604 5605 rcutree.do_rcu_barrier= [KNL] 5606 Wait for deferred kfree_rcu() frees and ordinary 5607 call_rcu() callbacks queued before this request to 5608 complete. This does not prevent new work from being 5609 queued concurrently. Requests are throttled so that 5610 userspace tests can safely hammer on the sysfs 5611 variable if they so choose. 5612 If triggered before the RCU grace-period machinery 5613 is fully active, this will error out with EAGAIN. 5614 5615 rcutree.dump_tree= [KNL] 5616 Dump the structure of the rcu_node combining tree 5617 out at early boot. This is used for diagnostic 5618 purposes, to verify correct tree setup. 5619 5620 rcutree.gp_cleanup_delay= [KNL] 5621 Set the number of jiffies to delay each step of 5622 RCU grace-period cleanup. 5623 5624 rcutree.gp_init_delay= [KNL] 5625 Set the number of jiffies to delay each step of 5626 RCU grace-period initialization. 5627 5628 rcutree.gp_preinit_delay= [KNL] 5629 Set the number of jiffies to delay each step of 5630 RCU grace-period pre-initialization, that is, 5631 the propagation of recent CPU-hotplug changes up 5632 the rcu_node combining tree. 5633 5634 rcutree.jiffies_till_first_fqs= [KNL] 5635 Set delay from grace-period initialization to 5636 first attempt to force quiescent states. 5637 Units are jiffies, minimum value is zero, 5638 and maximum value is HZ. 5639 5640 rcutree.jiffies_till_next_fqs= [KNL] 5641 Set delay between subsequent attempts to force 5642 quiescent states. Units are jiffies, minimum 5643 value is one, and maximum value is HZ. 5644 5645 rcutree.jiffies_till_sched_qs= [KNL] 5646 Set required age in jiffies for a 5647 given grace period before RCU starts 5648 soliciting quiescent-state help from 5649 rcu_note_context_switch() and cond_resched(). 5650 If not specified, the kernel will calculate 5651 a value based on the most recent settings 5652 of rcutree.jiffies_till_first_fqs 5653 and rcutree.jiffies_till_next_fqs. 5654 This calculated value may be viewed in 5655 rcutree.jiffies_to_sched_qs. Any attempt to set 5656 rcutree.jiffies_to_sched_qs will be cheerfully 5657 overwritten. 5658 5659 rcutree.kthread_prio= [KNL,BOOT] 5660 Set the SCHED_FIFO priority of the RCU per-CPU 5661 kthreads (rcuc/N). This value is also used for 5662 the priority of the RCU boost threads (rcub/N) 5663 and for the RCU grace-period kthreads (rcu_bh, 5664 rcu_preempt, and rcu_sched). If RCU_BOOST is 5665 set, valid values are 1-99 and the default is 1 5666 (the least-favored priority). Otherwise, when 5667 RCU_BOOST is not set, valid values are 0-99 and 5668 the default is zero (non-realtime operation). 5669 When RCU_NOCB_CPU is set, also adjust the 5670 priority of NOCB callback kthreads. 5671 5672 rcutree.nocb_nobypass_lim_per_jiffy= [KNL] 5673 On callback-offloaded (rcu_nocbs) CPUs, 5674 RCU reduces the lock contention that would 5675 otherwise be caused by callback floods through 5676 use of the ->nocb_bypass list. However, in the 5677 common non-flooded case, RCU queues directly to 5678 the main ->cblist in order to avoid the extra 5679 overhead of the ->nocb_bypass list and its lock. 5680 But if there are too many callbacks queued during 5681 a single jiffy, RCU pre-queues the callbacks into 5682 the ->nocb_bypass queue. The definition of "too 5683 many" is supplied by this kernel boot parameter. 5684 5685 rcutree.nohz_full_patience_delay= [KNL] 5686 On callback-offloaded (rcu_nocbs) CPUs, avoid 5687 disturbing RCU unless the grace period has 5688 reached the specified age in milliseconds. 5689 Defaults to zero. Large values will be capped 5690 at five seconds. All values will be rounded down 5691 to the nearest value representable by jiffies. 5692 5693 rcutree.qhimark= [KNL] 5694 Set threshold of queued RCU callbacks beyond which 5695 batch limiting is disabled. 5696 5697 rcutree.qlowmark= [KNL] 5698 Set threshold of queued RCU callbacks below which 5699 batch limiting is re-enabled. 5700 5701 rcutree.qovld= [KNL] 5702 Set threshold of queued RCU callbacks beyond which 5703 RCU's force-quiescent-state scan will aggressively 5704 enlist help from cond_resched() and sched IPIs to 5705 help CPUs more quickly reach quiescent states. 5706 Set to less than zero to make this be set based 5707 on rcutree.qhimark at boot time and to zero to 5708 disable more aggressive help enlistment. 5709 5710 rcutree.rcu_delay_page_cache_fill_msec= [KNL] 5711 Set the page-cache refill delay (in milliseconds) 5712 in response to low-memory conditions. The range 5713 of permitted values is in the range 0:100000. 5714 5715 rcutree.rcu_divisor= [KNL] 5716 Set the shift-right count to use to compute 5717 the callback-invocation batch limit bl from 5718 the number of callbacks queued on this CPU. 5719 The result will be bounded below by the value of 5720 the rcutree.blimit kernel parameter. Every bl 5721 callbacks, the softirq handler will exit in 5722 order to allow the CPU to do other work. 5723 5724 Please note that this callback-invocation batch 5725 limit applies only to non-offloaded callback 5726 invocation. Offloaded callbacks are instead 5727 invoked in the context of an rcuoc kthread, which 5728 scheduler will preempt as it does any other task. 5729 5730 rcutree.rcu_fanout_exact= [KNL] 5731 Disable autobalancing of the rcu_node combining 5732 tree. This is used by rcutorture, and might 5733 possibly be useful for architectures having high 5734 cache-to-cache transfer latencies. 5735 5736 rcutree.rcu_fanout_leaf= [KNL] 5737 Change the number of CPUs assigned to each 5738 leaf rcu_node structure. Useful for very 5739 large systems, which will choose the value 64, 5740 and for NUMA systems with large remote-access 5741 latencies, which will choose a value aligned 5742 with the appropriate hardware boundaries. 5743 5744 rcutree.rcu_min_cached_objs= [KNL] 5745 Minimum number of objects which are cached and 5746 maintained per one CPU. Object size is equal 5747 to PAGE_SIZE. The cache allows to reduce the 5748 pressure to page allocator, also it makes the 5749 whole algorithm to behave better in low memory 5750 condition. 5751 5752 rcutree.rcu_nocb_gp_stride= [KNL] 5753 Set the number of NOCB callback kthreads in 5754 each group, which defaults to the square root 5755 of the number of CPUs. Larger numbers reduce 5756 the wakeup overhead on the global grace-period 5757 kthread, but increases that same overhead on 5758 each group's NOCB grace-period kthread. 5759 5760 rcutree.rcu_kick_kthreads= [KNL] 5761 Cause the grace-period kthread to get an extra 5762 wake_up() if it sleeps three times longer than 5763 it should at force-quiescent-state time. 5764 This wake_up() will be accompanied by a 5765 WARN_ONCE() splat and an ftrace_dump(). 5766 5767 rcutree.rcu_resched_ns= [KNL] 5768 Limit the time spend invoking a batch of RCU 5769 callbacks to the specified number of nanoseconds. 5770 By default, this limit is checked only once 5771 every 32 callbacks in order to limit the pain 5772 inflicted by local_clock() overhead. 5773 5774 rcutree.rcu_unlock_delay= [KNL] 5775 In CONFIG_RCU_STRICT_GRACE_PERIOD=y kernels, 5776 this specifies an rcu_read_unlock()-time delay 5777 in microseconds. This defaults to zero. 5778 Larger delays increase the probability of 5779 catching RCU pointer leaks, that is, buggy use 5780 of RCU-protected pointers after the relevant 5781 rcu_read_unlock() has completed. 5782 5783 rcutree.sysrq_rcu= [KNL] 5784 Commandeer a sysrq key to dump out Tree RCU's 5785 rcu_node tree with an eye towards determining 5786 why a new grace period has not yet started. 5787 5788 rcutree.use_softirq= [KNL] 5789 If set to zero, move all RCU_SOFTIRQ processing to 5790 per-CPU rcuc kthreads. Defaults to a non-zero 5791 value, meaning that RCU_SOFTIRQ is used by default. 5792 Specify rcutree.use_softirq=0 to use rcuc kthreads. 5793 5794 But note that CONFIG_PREEMPT_RT=y kernels disable 5795 this kernel boot parameter, forcibly setting it 5796 to zero. 5797 5798 rcutree.enable_rcu_lazy= [KNL] 5799 To save power, batch RCU callbacks and flush after 5800 delay, memory pressure or callback list growing too 5801 big. 5802 5803 rcutree.rcu_normal_wake_from_gp= [KNL] 5804 Reduces a latency of synchronize_rcu() call. This approach 5805 maintains its own track of synchronize_rcu() callers, so it 5806 does not interact with regular callbacks because it does not 5807 use a call_rcu[_hurry]() path. Please note, this is for a 5808 normal grace period. 5809 5810 How to disable it: 5811 5812 echo 0 > /sys/module/rcutree/parameters/rcu_normal_wake_from_gp 5813 or pass a boot parameter "rcutree.rcu_normal_wake_from_gp=0" 5814 5815 Default is 1 if it is not explicitly disabled by the boot parameter 5816 passing 0. 5817 5818 rcuscale.gp_async= [KNL] 5819 Measure performance of asynchronous 5820 grace-period primitives such as call_rcu(). 5821 5822 rcuscale.gp_async_max= [KNL] 5823 Specify the maximum number of outstanding 5824 callbacks per writer thread. When a writer 5825 thread exceeds this limit, it invokes the 5826 corresponding flavor of rcu_barrier() to allow 5827 previously posted callbacks to drain. 5828 5829 rcuscale.gp_exp= [KNL] 5830 Measure performance of expedited synchronous 5831 grace-period primitives. 5832 5833 rcuscale.holdoff= [KNL] 5834 Set test-start holdoff period. The purpose of 5835 this parameter is to delay the start of the 5836 test until boot completes in order to avoid 5837 interference. 5838 5839 rcuscale.kfree_by_call_rcu= [KNL] 5840 In kernels built with CONFIG_RCU_LAZY=y, test 5841 call_rcu() instead of kfree_rcu(). 5842 5843 rcuscale.kfree_mult= [KNL] 5844 Instead of allocating an object of size kfree_obj, 5845 allocate one of kfree_mult * sizeof(kfree_obj). 5846 Defaults to 1. 5847 5848 rcuscale.kfree_rcu_test= [KNL] 5849 Set to measure performance of kfree_rcu() flooding. 5850 5851 rcuscale.kfree_rcu_test_double= [KNL] 5852 Test the double-argument variant of kfree_rcu(). 5853 If this parameter has the same value as 5854 rcuscale.kfree_rcu_test_single, both the single- 5855 and double-argument variants are tested. 5856 5857 rcuscale.kfree_rcu_test_single= [KNL] 5858 Test the single-argument variant of kfree_rcu(). 5859 If this parameter has the same value as 5860 rcuscale.kfree_rcu_test_double, both the single- 5861 and double-argument variants are tested. 5862 5863 rcuscale.kfree_nthreads= [KNL] 5864 The number of threads running loops of kfree_rcu(). 5865 5866 rcuscale.kfree_alloc_num= [KNL] 5867 Number of allocations and frees done in an iteration. 5868 5869 rcuscale.kfree_loops= [KNL] 5870 Number of loops doing rcuscale.kfree_alloc_num number 5871 of allocations and frees. 5872 5873 rcuscale.minruntime= [KNL] 5874 Set the minimum test run time in seconds. This 5875 does not affect the data-collection interval, 5876 but instead allows better measurement of things 5877 like CPU consumption. 5878 5879 rcuscale.nreaders= [KNL] 5880 Set number of RCU readers. The value -1 selects 5881 N, where N is the number of CPUs. A value 5882 "n" less than -1 selects N-n+1, where N is again 5883 the number of CPUs. For example, -2 selects N 5884 (the number of CPUs), -3 selects N+1, and so on. 5885 A value of "n" less than or equal to -N selects 5886 a single reader. 5887 5888 rcuscale.nwriters= [KNL] 5889 Set number of RCU writers. The values operate 5890 the same as for rcuscale.nreaders. 5891 N, where N is the number of CPUs 5892 5893 rcuscale.scale_type= [KNL] 5894 Specify the RCU implementation to test. 5895 5896 rcuscale.shutdown= [KNL] 5897 Shut the system down after performance tests 5898 complete. This is useful for hands-off automated 5899 testing. 5900 5901 rcuscale.verbose= [KNL] 5902 Enable additional printk() statements. 5903 5904 rcuscale.writer_holdoff= [KNL] 5905 Write-side holdoff between grace periods, 5906 in microseconds. The default of zero says 5907 no holdoff. 5908 5909 rcuscale.writer_holdoff_jiffies= [KNL] 5910 Additional write-side holdoff between grace 5911 periods, but in jiffies. The default of zero 5912 says no holdoff. 5913 5914 rcutorture.fqs_duration= [KNL] 5915 Set duration of force_quiescent_state bursts 5916 in microseconds. 5917 5918 rcutorture.fqs_holdoff= [KNL] 5919 Set holdoff time within force_quiescent_state bursts 5920 in microseconds. 5921 5922 rcutorture.fqs_stutter= [KNL] 5923 Set wait time between force_quiescent_state bursts 5924 in seconds. 5925 5926 rcutorture.fwd_progress= [KNL] 5927 Specifies the number of kthreads to be used 5928 for RCU grace-period forward-progress testing 5929 for the types of RCU supporting this notion. 5930 Defaults to 1 kthread, values less than zero or 5931 greater than the number of CPUs cause the number 5932 of CPUs to be used. 5933 5934 rcutorture.fwd_progress_div= [KNL] 5935 Specify the fraction of a CPU-stall-warning 5936 period to do tight-loop forward-progress testing. 5937 5938 rcutorture.fwd_progress_holdoff= [KNL] 5939 Number of seconds to wait between successive 5940 forward-progress tests. 5941 5942 rcutorture.fwd_progress_need_resched= [KNL] 5943 Enclose cond_resched() calls within checks for 5944 need_resched() during tight-loop forward-progress 5945 testing. 5946 5947 rcutorture.gp_cond= [KNL] 5948 Use conditional/asynchronous update-side 5949 normal-grace-period primitives, if available. 5950 5951 rcutorture.gp_cond_exp= [KNL] 5952 Use conditional/asynchronous update-side 5953 expedited-grace-period primitives, if available. 5954 5955 rcutorture.gp_cond_full= [KNL] 5956 Use conditional/asynchronous update-side 5957 normal-grace-period primitives that also take 5958 concurrent expedited grace periods into account, 5959 if available. 5960 5961 rcutorture.gp_cond_exp_full= [KNL] 5962 Use conditional/asynchronous update-side 5963 expedited-grace-period primitives that also take 5964 concurrent normal grace periods into account, 5965 if available. 5966 5967 rcutorture.gp_cond_wi= [KNL] 5968 Nominal wait interval for normal conditional 5969 grace periods (specified by rcutorture's 5970 gp_cond and gp_cond_full module parameters), 5971 in microseconds. The actual wait interval will 5972 be randomly selected to nanosecond granularity up 5973 to this wait interval. Defaults to 16 jiffies, 5974 for example, 16,000 microseconds on a system 5975 with HZ=1000. 5976 5977 rcutorture.gp_cond_wi_exp= [KNL] 5978 Nominal wait interval for expedited conditional 5979 grace periods (specified by rcutorture's 5980 gp_cond_exp and gp_cond_exp_full module 5981 parameters), in microseconds. The actual wait 5982 interval will be randomly selected to nanosecond 5983 granularity up to this wait interval. Defaults to 5984 128 microseconds. 5985 5986 rcutorture.gp_exp= [KNL] 5987 Use expedited update-side primitives, if available. 5988 5989 rcutorture.gp_normal= [KNL] 5990 Use normal (non-expedited) asynchronous 5991 update-side primitives, if available. 5992 5993 rcutorture.gp_poll= [KNL] 5994 Use polled update-side normal-grace-period 5995 primitives, if available. 5996 5997 rcutorture.gp_poll_exp= [KNL] 5998 Use polled update-side expedited-grace-period 5999 primitives, if available. 6000 6001 rcutorture.gp_poll_full= [KNL] 6002 Use polled update-side normal-grace-period 6003 primitives that also take concurrent expedited 6004 grace periods into account, if available. 6005 6006 rcutorture.gp_poll_exp_full= [KNL] 6007 Use polled update-side expedited-grace-period 6008 primitives that also take concurrent normal 6009 grace periods into account, if available. 6010 6011 rcutorture.gp_poll_wi= [KNL] 6012 Nominal wait interval for normal conditional 6013 grace periods (specified by rcutorture's 6014 gp_poll and gp_poll_full module parameters), 6015 in microseconds. The actual wait interval will 6016 be randomly selected to nanosecond granularity up 6017 to this wait interval. Defaults to 16 jiffies, 6018 for example, 16,000 microseconds on a system 6019 with HZ=1000. 6020 6021 rcutorture.gp_poll_wi_exp= [KNL] 6022 Nominal wait interval for expedited conditional 6023 grace periods (specified by rcutorture's 6024 gp_poll_exp and gp_poll_exp_full module 6025 parameters), in microseconds. The actual wait 6026 interval will be randomly selected to nanosecond 6027 granularity up to this wait interval. Defaults to 6028 128 microseconds. 6029 6030 rcutorture.gp_sync= [KNL] 6031 Use normal (non-expedited) synchronous 6032 update-side primitives, if available. If all 6033 of rcutorture.gp_cond=, rcutorture.gp_exp=, 6034 rcutorture.gp_normal=, and rcutorture.gp_sync= 6035 are zero, rcutorture acts as if is interpreted 6036 they are all non-zero. 6037 6038 rcutorture.gpwrap_lag= [KNL] 6039 Enable grace-period wrap lag testing. Setting 6040 to false prevents the gpwrap lag test from 6041 running. Default is true. 6042 6043 rcutorture.gpwrap_lag_gps= [KNL] 6044 Set the value for grace-period wrap lag during 6045 active lag testing periods. This controls how many 6046 grace periods differences we tolerate between 6047 rdp and rnp's gp_seq before setting overflow flag. 6048 The default is always set to 8. 6049 6050 rcutorture.gpwrap_lag_cycle_mins= [KNL] 6051 Set the total cycle duration for gpwrap lag 6052 testing in minutes. This is the total time for 6053 one complete cycle of active and inactive 6054 testing periods. Default is 30 minutes. 6055 6056 rcutorture.gpwrap_lag_active_mins= [KNL] 6057 Set the duration for which gpwrap lag is active 6058 within each cycle, in minutes. During this time, 6059 the grace-period wrap lag will be set to the 6060 value specified by gpwrap_lag_gps. Default is 6061 5 minutes. 6062 6063 rcutorture.irqreader= [KNL] 6064 Run RCU readers from irq handlers, or, more 6065 accurately, from a timer handler. Not all RCU 6066 flavors take kindly to this sort of thing. 6067 6068 rcutorture.leakpointer= [KNL] 6069 Leak an RCU-protected pointer out of the reader. 6070 This can of course result in splats, and is 6071 intended to test the ability of things like 6072 CONFIG_RCU_STRICT_GRACE_PERIOD=y to detect 6073 such leaks. 6074 6075 rcutorture.n_barrier_cbs= [KNL] 6076 Set callbacks/threads for rcu_barrier() testing. 6077 6078 rcutorture.nfakewriters= [KNL] 6079 Set number of concurrent RCU writers. These just 6080 stress RCU, they don't participate in the actual 6081 test, hence the "fake". 6082 6083 rcutorture.nmi_calls= [KNL] 6084 Enable issuing RCU callbacks from an NMI, on the 6085 RCU flavors that support it, to exercise the 6086 any-context callback path. Requires 6087 CONFIG_PERF_EVENTS and a hardware PMU; without 6088 both, nothing is issued. Defaults to enabled. 6089 6090 rcutorture.nocbs_nthreads= [KNL] 6091 Set number of RCU callback-offload togglers. 6092 Zero (the default) disables toggling. 6093 6094 rcutorture.nocbs_toggle= [KNL] 6095 Set the delay in milliseconds between successive 6096 callback-offload toggling attempts. 6097 6098 rcutorture.nreaders= [KNL] 6099 Set number of RCU readers. The value -1 selects 6100 N-1, where N is the number of CPUs. A value 6101 "n" less than -1 selects N-n-2, where N is again 6102 the number of CPUs. For example, -2 selects N 6103 (the number of CPUs), -3 selects N+1, and so on. 6104 6105 rcutorture.nwriters= [KNL] 6106 Set number of RCU writers, which must be either 6107 zero or one. For additional writers, use instead 6108 the rcutorture.nfakewriters parameter. 6109 6110 rcutorture.object_debug= [KNL] 6111 Enable debug-object double-call_rcu() testing. 6112 6113 rcutorture.onoff_holdoff= [KNL] 6114 Set time (s) after boot for CPU-hotplug testing. 6115 6116 rcutorture.onoff_interval= [KNL] 6117 Set time (jiffies) between CPU-hotplug operations, 6118 or zero to disable CPU-hotplug testing. 6119 6120 rcutorture.preempt_duration= [KNL] 6121 Set duration (in milliseconds) of preemptions 6122 by a high-priority FIFO real-time task. Set to 6123 zero (the default) to disable. The CPUs to 6124 preempt are selected randomly from the set that 6125 are online at a given point in time. Races with 6126 CPUs going offline are ignored, with that attempt 6127 at preemption skipped. 6128 6129 rcutorture.preempt_interval= [KNL] 6130 Set interval (in milliseconds, defaulting to one 6131 second) between preemptions by a high-priority 6132 FIFO real-time task. This delay is mediated 6133 by an hrtimer and is further fuzzed to avoid 6134 inadvertent synchronizations. 6135 6136 rcutorture.read_exit_burst= [KNL] 6137 The number of times in a given read-then-exit 6138 episode that a set of read-then-exit kthreads 6139 is spawned. 6140 6141 rcutorture.read_exit_delay= [KNL] 6142 The delay, in seconds, between successive 6143 read-then-exit testing episodes. 6144 6145 rcutorture.reader_flavor= [KNL] 6146 A bit mask indicating which readers to use. 6147 If there is more than one bit set, the readers 6148 are entered from low-order bit up, and are 6149 exited in the opposite order. For SRCU, the 6150 0x1 bit is normal readers, 0x2 NMI-safe readers, 6151 and 0x4 light-weight readers. 6152 6153 rcutorture.shuffle_interval= [KNL] 6154 Set task-shuffle interval (s). Shuffling tasks 6155 allows some CPUs to go into dyntick-idle mode 6156 during the rcutorture test. 6157 6158 rcutorture.shutdown_secs= [KNL] 6159 Set time (s) after boot system shutdown. This 6160 is useful for hands-off automated testing. 6161 6162 rcutorture.stall_cpu= [KNL] 6163 Duration of CPU stall (s) to test RCU CPU stall 6164 warnings, zero to disable. 6165 6166 rcutorture.stall_cpu_block= [KNL] 6167 Sleep while stalling if set. This will result 6168 in warnings from preemptible RCU in addition to 6169 any other stall-related activity. Note that 6170 in kernels built with CONFIG_PREEMPTION=n and 6171 CONFIG_PREEMPT_COUNT=y, this parameter will 6172 cause the CPU to pass through a quiescent state. 6173 Given CONFIG_PREEMPTION=n, this will suppress 6174 RCU CPU stall warnings, but will instead result 6175 in scheduling-while-atomic splats. 6176 6177 Use of this module parameter results in splats. 6178 6179 6180 rcutorture.stall_cpu_holdoff= [KNL] 6181 Time to wait (s) after boot before inducing stall. 6182 6183 rcutorture.stall_cpu_irqsoff= [KNL] 6184 Disable interrupts while stalling if set, but only 6185 on the first stall in the set. 6186 6187 rcutorture.stall_cpu_repeat= [KNL] 6188 Number of times to repeat the stall sequence, 6189 so that rcutorture.stall_cpu_repeat=3 will result 6190 in four stall sequences. 6191 6192 rcutorture.stall_gp_kthread= [KNL] 6193 Duration (s) of forced sleep within RCU 6194 grace-period kthread to test RCU CPU stall 6195 warnings, zero to disable. If both stall_cpu 6196 and stall_gp_kthread are specified, the 6197 kthread is starved first, then the CPU. 6198 6199 rcutorture.stall_only= [KNL] 6200 Shut off all rcutorture kthreads other than the 6201 RCU CPU stall-warning test kthreads. The purpose 6202 of this is to test production applictions' 6203 reactions to CPU stalls, and with minimal 6204 additional overhead. Or you can omit the 6205 stall-warning tests as well and get a heavy 6206 no-op, your choice! 6207 6208 rcutorture.stat_interval= [KNL] 6209 Time (s) between statistics printk()s. 6210 6211 rcutorture.stutter= [KNL] 6212 Time (s) to stutter testing, for example, specifying 6213 five seconds causes the test to run for five seconds, 6214 wait for five seconds, and so on. This tests RCU's 6215 ability to transition abruptly to and from idle. 6216 6217 rcutorture.test_boost= [KNL] 6218 Test RCU priority boosting? 0=no, 1=maybe, 2=yes. 6219 "Maybe" means test if the RCU implementation 6220 under test support RCU priority boosting. 6221 6222 rcutorture.test_boost_duration= [KNL] 6223 Duration (s) of each individual boost test. 6224 6225 rcutorture.test_boost_holdoff= [KNL] 6226 Holdoff time (s) from start of test to the start 6227 of RCU priority-boost testing. Defaults to zero, 6228 that is, no holdoff. 6229 6230 rcutorture.test_boost_interval= [KNL] 6231 Interval (s) between each boost test. 6232 6233 rcutorture.test_no_idle_hz= [KNL] 6234 Test RCU's dyntick-idle handling. See also the 6235 rcutorture.shuffle_interval parameter. 6236 6237 rcutorture.torture_type= [KNL] 6238 Specify the RCU implementation to test. 6239 6240 rcutorture.verbose= [KNL] 6241 Enable additional printk() statements. 6242 6243 rcupdate.rcu_cpu_stall_ftrace_dump= [KNL] 6244 Dump ftrace buffer after reporting RCU CPU 6245 stall warning. 6246 6247 rcupdate.rcu_cpu_stall_notifiers= [KNL] 6248 Provide RCU CPU stall notifiers, but see the 6249 warnings in the RCU_CPU_STALL_NOTIFIER Kconfig 6250 option's help text. TL;DR: You almost certainly 6251 do not want rcupdate.rcu_cpu_stall_notifiers. 6252 6253 rcupdate.rcu_cpu_stall_suppress= [KNL] 6254 Suppress RCU CPU stall warning messages. 6255 6256 rcupdate.rcu_cpu_stall_suppress_at_boot= [KNL] 6257 Suppress RCU CPU stall warning messages and 6258 rcutorture writer stall warnings that occur 6259 during early boot, that is, during the time 6260 before the init task is spawned. 6261 6262 rcupdate.rcu_cpu_stall_timeout= [KNL] 6263 Set timeout for RCU CPU stall warning messages. 6264 The value is in seconds and the maximum allowed 6265 value is 300 seconds. 6266 6267 rcupdate.rcu_exp_cpu_stall_timeout= [KNL] 6268 Set timeout for expedited RCU CPU stall warning 6269 messages. The value is in milliseconds 6270 and the maximum allowed value is 21000 6271 milliseconds. Please note that this value is 6272 adjusted to an arch timer tick resolution. 6273 Setting this to zero causes the value from 6274 rcupdate.rcu_cpu_stall_timeout to be used (after 6275 conversion from seconds to milliseconds). 6276 6277 rcupdate.rcu_cpu_stall_cputime= [KNL] 6278 Provide statistics on the cputime and count of 6279 interrupts and tasks during the sampling period. For 6280 multiple continuous RCU stalls, all sampling periods 6281 begin at half of the first RCU stall timeout. 6282 6283 rcupdate.rcu_exp_stall_task_details= [KNL] 6284 Print stack dumps of any tasks blocking the 6285 current expedited RCU grace period during an 6286 expedited RCU CPU stall warning. 6287 6288 rcupdate.rcu_expedited= [KNL] 6289 Use expedited grace-period primitives, for 6290 example, synchronize_rcu_expedited() instead 6291 of synchronize_rcu(). This reduces latency, 6292 but can increase CPU utilization, degrade 6293 real-time latency, and degrade energy efficiency. 6294 No effect on CONFIG_TINY_RCU kernels. 6295 6296 rcupdate.rcu_normal= [KNL] 6297 Use only normal grace-period primitives, 6298 for example, synchronize_rcu() instead of 6299 synchronize_rcu_expedited(). This improves 6300 real-time latency, CPU utilization, and 6301 energy efficiency, but can expose users to 6302 increased grace-period latency. This parameter 6303 overrides rcupdate.rcu_expedited. No effect on 6304 CONFIG_TINY_RCU kernels. 6305 6306 rcupdate.rcu_normal_after_boot= [KNL] 6307 Once boot has completed (that is, after 6308 rcu_end_inkernel_boot() has been invoked), use 6309 only normal grace-period primitives. No effect 6310 on CONFIG_TINY_RCU kernels. 6311 6312 But note that CONFIG_PREEMPT_RT=y kernels enables 6313 this kernel boot parameter, forcibly setting 6314 it to the value one, that is, converting any 6315 post-boot attempt at an expedited RCU grace 6316 period to instead use normal non-expedited 6317 grace-period processing. 6318 6319 rcupdate.rcu_task_collapse_debug= [KNL] 6320 Enable debugging prints that record when RCU Tasks 6321 and RCU Tasks Trace expand to per-CPU callback 6322 queuing and collapse back to CPU-0 queuing. 6323 This is default-disabled due to the fact that 6324 some workloads can make it quite noisy. 6325 6326 rcupdate.rcu_task_collapse_lim= [KNL] 6327 Set the maximum number of callbacks present 6328 at the beginning of a grace period that allows 6329 the RCU Tasks flavors to collapse back to using 6330 a single callback queue. This switching only 6331 occurs when rcupdate.rcu_task_enqueue_lim is 6332 set to the default value of -1. 6333 6334 rcupdate.rcu_task_contend_lim= [KNL] 6335 Set the minimum number of callback-queuing-time 6336 lock-contention events per jiffy required to 6337 cause the RCU Tasks flavors to switch to per-CPU 6338 callback queuing. This switching only occurs 6339 when rcupdate.rcu_task_enqueue_lim is set to 6340 the default value of -1. 6341 6342 rcupdate.rcu_task_enqueue_lim= [KNL] 6343 Set the number of callback queues to use for the 6344 RCU Tasks family of RCU flavors. The default 6345 of -1 allows this to be automatically (and 6346 dynamically) adjusted. This parameter is intended 6347 for use in testing. 6348 6349 rcupdate.rcu_task_lazy_lim= [KNL] 6350 Number of callbacks on a given CPU that will 6351 cancel laziness on that CPU. Use -1 to disable 6352 cancellation of laziness, but be advised that 6353 doing so increases the danger of OOM due to 6354 callback flooding. 6355 6356 rcupdate.rcu_task_stall_info= [KNL] 6357 Set initial timeout in jiffies for RCU task stall 6358 informational messages, which give some indication 6359 of the problem for those not patient enough to 6360 wait for ten minutes. Informational messages are 6361 only printed prior to the stall-warning message 6362 for a given grace period. Disable with a value 6363 less than or equal to zero. Defaults to ten 6364 seconds. A change in value does not take effect 6365 until the beginning of the next grace period. 6366 6367 rcupdate.rcu_task_stall_info_mult= [KNL] 6368 Multiplier for time interval between successive 6369 RCU task stall informational messages for a given 6370 RCU tasks grace period. This value is clamped 6371 to one through ten, inclusive. It defaults to 6372 the value three, so that the first informational 6373 message is printed 10 seconds into the grace 6374 period, the second at 40 seconds, the third at 6375 160 seconds, and then the stall warning at 600 6376 seconds would prevent a fourth at 640 seconds. 6377 6378 rcupdate.rcu_task_stall_timeout= [KNL] 6379 Set timeout in jiffies for RCU task stall 6380 warning messages. Disable with a value less 6381 than or equal to zero. Defaults to ten minutes. 6382 A change in value does not take effect until 6383 the beginning of the next grace period. 6384 6385 rcupdate.rcu_tasks_lazy_ms= [KNL] 6386 Set timeout in milliseconds RCU Tasks asynchronous 6387 callback batching for call_rcu_tasks(). 6388 A negative value will take the default. A value 6389 of zero will disable batching. Batching is 6390 always disabled for synchronize_rcu_tasks(). 6391 6392 rcupdate.rcu_self_test= [KNL] 6393 Run the RCU early boot self tests 6394 6395 rdinit= [KNL] 6396 Format: <full_path> 6397 Run specified binary instead of /init from the ramdisk, 6398 used for early userspace startup. See initrd. 6399 6400 rdrand= [X86,EARLY] 6401 force - Override the decision by the kernel to hide the 6402 advertisement of RDRAND support (this affects 6403 certain AMD processors because of buggy BIOS 6404 support, specifically around the suspend/resume 6405 path). 6406 6407 rdt= [HW,X86,RDT] 6408 Turn on/off individual RDT features. List is: 6409 cmt, mbmtotal, mbmlocal, l3cat, l3cdp, l2cat, l2cdp, 6410 mba, smba, bmec, abmc, sdciae, energy[:guid], 6411 perf[:guid]. 6412 E.g. to turn on cmt and turn off mba use: 6413 rdt=cmt,!mba 6414 To turn off all energy telemetry monitoring and ensure that 6415 perf telemetry monitoring associated with guid 0x12345 6416 is enabled use: 6417 rdt=!energy,perf:0x12345 6418 6419 reboot= [KNL] 6420 Format (x86 or x86_64): 6421 [w[arm] | c[old] | h[ard] | s[oft] | g[pio]] | d[efault] \ 6422 [[,]s[mp]####] \ 6423 [[,]b[ios] | a[cpi] | k[bd] | t[riple] | e[fi] | p[ci]] \ 6424 [[,]f[orce]] 6425 Where reboot_mode is one of warm (soft) or cold (hard) or gpio 6426 (prefix with 'panic_' to set mode for panic 6427 reboot only), 6428 reboot_type is one of bios, acpi, kbd, triple, efi, or pci, 6429 reboot_force is either force or not specified, 6430 reboot_cpu is s[mp]#### with #### being the processor 6431 to be used for rebooting. 6432 6433 acpi 6434 Use the ACPI RESET_REG in the FADT. If ACPI is not 6435 configured or the ACPI reset does not work, the reboot 6436 path attempts the reset using the keyboard controller. 6437 6438 bios 6439 Use the CPU reboot vector for warm reset 6440 6441 cold 6442 Set the cold reboot flag 6443 6444 default 6445 There are some built-in platform specific "quirks" 6446 - you may see: "reboot: <name> series board detected. 6447 Selecting <type> for reboots." In the case where you 6448 think the quirk is in error (e.g. you have newer BIOS, 6449 or newer board) using this option will ignore the 6450 built-in quirk table, and use the generic default 6451 reboot actions. 6452 6453 efi 6454 Use efi reset_system runtime service. If EFI is not 6455 configured or the EFI reset does not work, the reboot 6456 path attempts the reset using the keyboard controller. 6457 6458 force 6459 Don't stop other CPUs on reboot. This can make reboot 6460 more reliable in some cases. 6461 6462 kbd 6463 Use the keyboard controller. cold reset (default) 6464 6465 pci 6466 Use a write to the PCI config space register 0xcf9 to 6467 trigger reboot. 6468 6469 triple 6470 Force a triple fault (init) 6471 6472 warm 6473 Don't set the cold reboot flag 6474 6475 Using warm reset will be much faster especially on big 6476 memory systems because the BIOS will not go through 6477 the memory check. Disadvantage is that not all 6478 hardware will be completely reinitialized on reboot so 6479 there may be boot problems on some systems. 6480 6481 6482 refscale.holdoff= [KNL] 6483 Set test-start holdoff period. The purpose of 6484 this parameter is to delay the start of the 6485 test until boot completes in order to avoid 6486 interference. 6487 6488 refscale.lookup_instances= [KNL] 6489 Number of data elements to use for the forms of 6490 SLAB_TYPESAFE_BY_RCU testing. A negative number 6491 is negated and multiplied by nr_cpu_ids, while 6492 zero specifies nr_cpu_ids. 6493 6494 refscale.loops= [KNL] 6495 Set the number of loops over the synchronization 6496 primitive under test. Increasing this number 6497 reduces noise due to loop start/end overhead, 6498 but the default has already reduced the per-pass 6499 noise to a handful of picoseconds on ca. 2020 6500 x86 laptops. 6501 6502 refscale.nreaders= [KNL] 6503 Set number of readers. The default value of -1 6504 selects N, where N is roughly 75% of the number 6505 of CPUs. A value of zero is an interesting choice. 6506 6507 refscale.nruns= [KNL] 6508 Set number of runs, each of which is dumped onto 6509 the console log. 6510 6511 refscale.readdelay= [KNL] 6512 Set the read-side critical-section duration, 6513 measured in microseconds. 6514 6515 refscale.scale_type= [KNL] 6516 Specify the read-protection implementation to test. 6517 6518 refscale.shutdown= [KNL] 6519 Shut down the system at the end of the performance 6520 test. This defaults to 1 (shut it down) when 6521 refscale is built into the kernel and to 0 (leave 6522 it running) when refscale is built as a module. 6523 6524 refscale.verbose= [KNL] 6525 Enable additional printk() statements. 6526 6527 refscale.verbose_batched= [KNL] 6528 Batch the additional printk() statements. If zero 6529 (the default) or negative, print everything. Otherwise, 6530 print every Nth verbose statement, where N is the value 6531 specified. 6532 6533 regulator_ignore_unused 6534 [REGULATOR] 6535 Prevents regulator framework from disabling regulators 6536 that are unused, due no driver claiming them. This may 6537 be useful for debug and development, but should not be 6538 needed on a platform with proper driver support. 6539 6540 relax_domain_level= 6541 [KNL, SMP] Set scheduler's default relax_domain_level. 6542 See Documentation/admin-guide/cgroup-v1/cpusets.rst. 6543 6544 reserve= [KNL,BUGS] Force kernel to ignore I/O ports or memory 6545 Format: <base1>,<size1>[,<base2>,<size2>,...] 6546 Reserve I/O ports or memory so the kernel won't use 6547 them. If <base> is less than 0x10000, the region 6548 is assumed to be I/O ports; otherwise it is memory. 6549 6550 reserve_mem= [MM] 6551 Format: nn[KMG]:<align>:<label> 6552 Reserve physical memory and label it with a name that 6553 other subsystems can use to access it. This is typically 6554 used for systems that do not wipe the RAM, and this command 6555 line will try to reserve the same physical memory on 6556 soft reboots. Note, it is not guaranteed to be the same 6557 location. For example, if anything about the system changes 6558 or if booting a different kernel. It can also fail if KASLR 6559 places the kernel at the location of where the RAM reservation 6560 was from a previous boot, the new reservation will be at a 6561 different location. 6562 Any subsystem using this feature must add a way to verify 6563 that the contents of the physical memory is from a previous 6564 boot, as there may be cases where the memory will not be 6565 located at the same location. 6566 6567 The format is size:align:label for example, to request 6568 12 megabytes of 4096 alignment for ramoops: 6569 6570 reserve_mem=12M:4096:oops ramoops.mem_name=oops 6571 6572 reservetop= [X86-32,EARLY] 6573 Format: nn[KMG] 6574 Reserves a hole at the top of the kernel virtual 6575 address space. 6576 6577 reset_devices [KNL] Force drivers to reset the underlying device 6578 during initialization. 6579 6580 resume= [SWSUSP] 6581 Specify the partition device for software suspend 6582 Format: 6583 {/dev/<dev> | PARTUUID=<uuid> | <int>:<int> | <hex>} 6584 6585 resume_offset= [SWSUSP] 6586 Specify the offset from the beginning of the partition 6587 given by "resume=" at which the swap header is located, 6588 in <PAGE_SIZE> units (needed only for swap files). 6589 See Documentation/power/swsusp-and-swap-files.rst 6590 6591 resumedelay= [HIBERNATION] Delay (in seconds) to pause before attempting to 6592 read the resume files 6593 6594 resumewait [HIBERNATION] Wait (indefinitely) for resume device to show up. 6595 Useful for devices that are detected asynchronously 6596 (e.g. USB and MMC devices). 6597 6598 retain_initrd [RAM] Keep initrd memory after extraction. After boot, it will 6599 be accessible via /sys/firmware/initrd. 6600 6601 retbleed= [X86] Control mitigation of RETBleed (Arbitrary 6602 Speculative Code Execution with Return Instructions) 6603 vulnerability. 6604 6605 AMD-based UNRET and IBPB mitigations alone do not stop 6606 sibling threads from influencing the predictions of other 6607 sibling threads. For that reason, STIBP is used on pro- 6608 cessors that support it, and mitigate SMT on processors 6609 that don't. 6610 6611 off - no mitigation 6612 auto - automatically select a mitigation 6613 auto,nosmt - automatically select a mitigation, 6614 disabling SMT if necessary for 6615 the full mitigation (only on Zen1 6616 and older without STIBP). 6617 ibpb - On AMD, mitigate short speculation 6618 windows on basic block boundaries too. 6619 Safe, highest perf impact. It also 6620 enables STIBP if present. Not suitable 6621 on Intel. 6622 ibpb,nosmt - Like "ibpb" above but will disable SMT 6623 when STIBP is not available. This is 6624 the alternative for systems which do not 6625 have STIBP. 6626 unret - Force enable untrained return thunks, 6627 only effective on AMD f15h-f17h based 6628 systems. 6629 unret,nosmt - Like unret, but will disable SMT when STIBP 6630 is not available. This is the alternative for 6631 systems which do not have STIBP. 6632 6633 Selecting 'auto' will choose a mitigation method at run 6634 time according to the CPU. 6635 6636 Not specifying this option is equivalent to retbleed=auto. 6637 6638 rfkill.default_state= 6639 0 "airplane mode". All wifi, bluetooth, wimax, gps, fm, 6640 etc. communication is blocked by default. 6641 1 Unblocked. 6642 6643 rfkill.master_switch_mode= 6644 0 The "airplane mode" button does nothing. 6645 1 The "airplane mode" button toggles between everything 6646 blocked and the previous configuration. 6647 2 The "airplane mode" button toggles between everything 6648 blocked and everything unblocked. 6649 6650 ring3mwait=disable 6651 [KNL] Disable ring 3 MONITOR/MWAIT feature on supported 6652 CPUs. 6653 6654 riscv_isa_fallback [RISCV,EARLY] 6655 When CONFIG_RISCV_ISA_FALLBACK is not enabled, permit 6656 falling back to detecting extension support by parsing 6657 "riscv,isa" property on devicetree systems when the 6658 replacement properties are not found. See the Kconfig 6659 entry for RISCV_ISA_FALLBACK. 6660 6661 riscv_nousercfi= 6662 all Disable user CFI ABI to userspace even if cpu extension 6663 are available. 6664 bcfi Disable user backward CFI ABI to userspace even if 6665 the shadow stack extension is available. 6666 fcfi Disable user forward CFI ABI to userspace even if the 6667 landing pad extension is available. 6668 6669 ro [KNL] Mount root device read-only on boot 6670 6671 rodata= [KNL,EARLY] 6672 on Mark read-only kernel memory as read-only (default). 6673 off Leave read-only kernel memory writable for debugging. 6674 noalias Mark read-only kernel memory as read-only but retain 6675 writable aliases in the direct map for regions outside 6676 of the kernel image. [arm64] 6677 6678 rockchip.usb_uart 6679 [EARLY] 6680 Enable the uart passthrough on the designated usb port 6681 on Rockchip SoCs. When active, the signals of the 6682 debug-uart get routed to the D+ and D- pins of the usb 6683 port and the regular usb controller gets disabled. 6684 6685 root= [KNL] Root filesystem 6686 Usually this is a block device specifier of some kind, 6687 see the early_lookup_bdev comment in 6688 block/early-lookup.c for details. 6689 Alternatively this can be "ram" for the legacy initial 6690 ramdisk, "nfs" and "cifs" for root on a network file 6691 system, or "mtd" and "ubi" for mounting from raw flash. 6692 6693 rootdelay= [KNL] Delay (in seconds) to pause before attempting to 6694 mount the root filesystem 6695 6696 rootflags= [KNL] Set root filesystem mount option string 6697 6698 rseq_slice_ext= [KNL] RSEQ based time slice extension 6699 Format: boolean 6700 Control enablement of RSEQ based time slice extension. 6701 Default is 'on'. 6702 6703 initramfs_options= [KNL] 6704 Specify mount options for the initramfs mount. 6705 6706 rootfstype= [KNL] Set root filesystem type 6707 6708 rootwait [KNL] Wait (indefinitely) for root device to show up. 6709 Useful for devices that are detected asynchronously 6710 (e.g. USB and MMC devices). 6711 6712 rootwait= [KNL] Maximum time (in seconds) to wait for root device 6713 to show up before attempting to mount the root 6714 filesystem. 6715 6716 rproc_mem=nn[KMG][@address] 6717 [KNL,ARM,CMA] Remoteproc physical memory block. 6718 Memory area to be used by remote processor image, 6719 managed by CMA. 6720 6721 rseq_debug= [KNL] Enable or disable restartable sequence 6722 debug mode. Defaults to CONFIG_RSEQ_DEBUG_DEFAULT_ENABLE. 6723 Format: <bool> 6724 6725 rt_group_sched= [KNL] Enable or disable SCHED_RR/FIFO group scheduling 6726 when CONFIG_RT_GROUP_SCHED=y. Defaults to 6727 !CONFIG_RT_GROUP_SCHED_DEFAULT_DISABLED. 6728 Format: <bool> 6729 6730 rw [KNL] Mount root device read-write on boot 6731 6732 S [KNL] Run init in single mode 6733 6734 s390_iommu= [HW,S390] 6735 Set s390 IOTLB flushing mode 6736 strict 6737 With strict flushing every unmap operation will result 6738 in an IOTLB flush. Default is lazy flushing before 6739 reuse, which is faster. Deprecated, equivalent to 6740 iommu.strict=1. 6741 6742 s390_iommu_aperture= [KNL,S390] 6743 Specifies the size of the per device DMA address space 6744 accessible through the DMA and IOMMU APIs as a decimal 6745 factor of the size of main memory. 6746 The default is 1 meaning that one can concurrently use 6747 as many DMA addresses as physical memory is installed, 6748 if supported by hardware, and thus map all of memory 6749 once. With a value of 2 one can map all of memory twice 6750 and so on. As a special case a factor of 0 imposes no 6751 restrictions other than those given by hardware at the 6752 cost of significant additional memory use for tables. 6753 6754 6755 sched_proxy_exec= [KNL] 6756 Enables or disables "proxy execution" style 6757 solution to mutex-based priority inversion. 6758 Format: <bool> 6759 6760 sched_smt_asym_packing= [KNL,SMP] 6761 Format: on 6762 Force asymmetric packing at the SMT scheduling domain. 6763 Idle CPU selection prefers siblings with a higher 6764 arch_asym_cpu_priority(). The default implementation 6765 prefers lower-numbered logical CPUs, but architecture 6766 overrides remain authoritative. Equal priorities do 6767 not establish a sibling preference. For example, x86 6768 with CONFIG_SCHED_MC_PRIO normally assigns the same 6769 core priority to both SMT siblings, so this option 6770 does not favor either sibling. 6771 6772 sched_verbose [KNL,EARLY] Enables verbose scheduler debug messages. 6773 6774 schedstats= [KNL,X86] Enable or disable scheduled statistics. 6775 Allowed values are enable and disable. This feature 6776 incurs a small amount of overhead in the scheduler 6777 but is useful for debugging and performance tuning. 6778 6779 sched_thermal_decay_shift= 6780 [Deprecated] 6781 [KNL, SMP] Set a decay shift for scheduler thermal 6782 pressure signal. Thermal pressure signal follows the 6783 default decay period of other scheduler pelt 6784 signals(usually 32 ms but configurable). Setting 6785 sched_thermal_decay_shift will left shift the decay 6786 period for the thermal pressure signal by the shift 6787 value. 6788 i.e. with the default pelt decay period of 32 ms 6789 sched_thermal_decay_shift thermal pressure decay pr 6790 1 64 ms 6791 2 128 ms 6792 and so on. 6793 Format: integer between 0 and 10 6794 Default is 0. 6795 6796 scftorture.holdoff= [KNL] 6797 Number of seconds to hold off before starting 6798 test. Defaults to zero for module insertion and 6799 to 10 seconds for built-in smp_call_function() 6800 tests. 6801 6802 scftorture.longwait= [KNL] 6803 Request ridiculously long waits randomly selected 6804 up to the chosen limit in seconds. Zero (the 6805 default) disables this feature. Please note 6806 that requesting even small non-zero numbers of 6807 seconds can result in RCU CPU stall warnings, 6808 softlockup complaints, and so on. 6809 6810 scftorture.nthreads= [KNL] 6811 Number of kthreads to spawn to invoke the 6812 smp_call_function() family of functions. 6813 The default of -1 specifies a number of kthreads 6814 equal to the number of CPUs. 6815 6816 scftorture.onoff_holdoff= [KNL] 6817 Number seconds to wait after the start of the 6818 test before initiating CPU-hotplug operations. 6819 6820 scftorture.onoff_interval= [KNL] 6821 Number seconds to wait between successive 6822 CPU-hotplug operations. Specifying zero (which 6823 is the default) disables CPU-hotplug operations. 6824 6825 scftorture.shutdown_secs= [KNL] 6826 The number of seconds following the start of the 6827 test after which to shut down the system. The 6828 default of zero avoids shutting down the system. 6829 Non-zero values are useful for automated tests. 6830 6831 scftorture.stat_interval= [KNL] 6832 The number of seconds between outputting the 6833 current test statistics to the console. A value 6834 of zero disables statistics output. 6835 6836 scftorture.stutter_cpus= [KNL] 6837 The number of jiffies to wait between each change 6838 to the set of CPUs under test. 6839 6840 scftorture.use_cpus_read_lock= [KNL] 6841 Use use_cpus_read_lock() instead of the default 6842 preempt_disable() to disable CPU hotplug 6843 while invoking one of the smp_call_function*() 6844 functions. 6845 6846 scftorture.verbose= [KNL] 6847 Enable additional printk() statements. 6848 6849 scftorture.weight_single= [KNL] 6850 The probability weighting to use for the 6851 smp_call_function_single() function with a zero 6852 "wait" parameter. A value of -1 selects the 6853 default if all other weights are -1. However, 6854 if at least one weight has some other value, a 6855 value of -1 will instead select a weight of zero. 6856 6857 scftorture.weight_single_wait= [KNL] 6858 The probability weighting to use for the 6859 smp_call_function_single() function with a 6860 non-zero "wait" parameter. See weight_single. 6861 6862 scftorture.weight_many= [KNL] 6863 The probability weighting to use for the 6864 smp_call_function_many() function with a zero 6865 "wait" parameter. See weight_single. 6866 Note well that setting a high probability for 6867 this weighting can place serious IPI load 6868 on the system. 6869 6870 scftorture.weight_many_wait= [KNL] 6871 The probability weighting to use for the 6872 smp_call_function_many() function with a 6873 non-zero "wait" parameter. See weight_single 6874 and weight_many. 6875 6876 scftorture.weight_all= [KNL] 6877 The probability weighting to use for the 6878 smp_call_function_all() function with a zero 6879 "wait" parameter. See weight_single and 6880 weight_many. 6881 6882 scftorture.weight_all_wait= [KNL] 6883 The probability weighting to use for the 6884 smp_call_function_all() function with a 6885 non-zero "wait" parameter. See weight_single 6886 and weight_many. 6887 6888 sdw_mclk_divider=[SDW] 6889 Specify the MCLK divider for Intel SoundWire buses in 6890 case the BIOS does not provide the clock rate properly. 6891 6892 skew_tick= [KNL,EARLY] Offset the periodic timer tick per cpu to mitigate 6893 xtime_lock contention on larger systems, and/or RCU lock 6894 contention on all systems with CONFIG_MAXSMP set. 6895 Format: { "0" | "1" } 6896 0 -- disable. (may be 1 via CONFIG_CMDLINE="skew_tick=1") 6897 1 -- enable. 6898 Note: increases power consumption, thus should only be 6899 enabled if running jitter sensitive (HPC/RT) workloads. 6900 6901 security= [SECURITY] Choose a legacy "major" security module to 6902 enable at boot. This has been deprecated by the 6903 "lsm=" parameter. 6904 6905 selinux= [SELINUX] Disable or enable SELinux at boot time. 6906 Format: { "0" | "1" } 6907 See security/selinux/Kconfig help text. 6908 0 -- disable. 6909 1 -- enable. 6910 Default value is 1. 6911 6912 serialnumber [BUGS=X86-32] 6913 6914 sev=option[,option...] [X86-64] 6915 6916 debug 6917 Enable debug messages. 6918 6919 nosnp 6920 Do not enable SEV-SNP (applies to host/hypervisor 6921 only). Setting 'nosnp' avoids the RMP check overhead 6922 in memory accesses when users do not want to run 6923 SEV-SNP guests. 6924 6925 shapers= [NET] 6926 Maximal number of shapers. 6927 6928 show_lapic= [APIC,X86] Advanced Programmable Interrupt Controller 6929 Limit apic dumping. The parameter defines the maximal 6930 number of local apics being dumped. Also it is possible 6931 to set it to "all" by meaning -- no limit here. 6932 Format: { 1 (default) | 2 | ... | all }. 6933 The parameter valid if only apic=debug or 6934 apic=verbose is specified. 6935 Example: apic=debug show_lapic=all 6936 6937 slab_debug[=options[,slabs][;[options[,slabs]]...]] [MM] 6938 Enabling slab_debug allows one to determine the 6939 culprit if slab objects become corrupted. Enabling 6940 slab_debug can create guard zones around objects and 6941 may poison objects when not in use. Also tracks the 6942 last alloc / free. For more information see 6943 Documentation/admin-guide/mm/slab.rst. 6944 (slub_debug legacy name also accepted for now) 6945 6946 Using this option implies the "no_hash_pointers" 6947 option which can be undone by adding the 6948 "hash_pointers=always" option. 6949 6950 slab_max_order= [MM] 6951 Determines the maximum allowed order for slabs. 6952 A high setting may cause OOMs due to memory 6953 fragmentation. For more information see 6954 Documentation/admin-guide/mm/slab.rst. 6955 (slub_max_order legacy name also accepted for now) 6956 6957 slab_merge [MM] 6958 Enable merging of slabs with similar size when the 6959 kernel is built without CONFIG_SLAB_MERGE_DEFAULT. 6960 (slub_merge legacy name also accepted for now) 6961 6962 slab_min_objects= [MM] 6963 The minimum number of objects per slab. SLUB will 6964 increase the slab order up to slab_max_order to 6965 generate a sufficiently large slab able to contain 6966 the number of objects indicated. The higher the number 6967 of objects the smaller the overhead of tracking slabs 6968 and the less frequently locks need to be acquired. 6969 For more information see 6970 Documentation/admin-guide/mm/slab.rst. 6971 (slub_min_objects legacy name also accepted for now) 6972 6973 slab_min_order= [MM] 6974 Determines the minimum page order for slabs. Must be 6975 lower or equal to slab_max_order. For more information see 6976 Documentation/admin-guide/mm/slab.rst. 6977 (slub_min_order legacy name also accepted for now) 6978 6979 slab_nomerge [MM] 6980 Disable merging of slabs with similar size. May be 6981 necessary if there is some reason to distinguish 6982 allocs to different slabs, especially in hardened 6983 environments where the risk of heap overflows and 6984 layout control by attackers can usually be 6985 frustrated by disabling merging. This will reduce 6986 most of the exposure of a heap attack to a single 6987 cache (risks via metadata attacks are mostly 6988 unchanged). Debug options disable merging on their 6989 own. 6990 For more information see 6991 Documentation/admin-guide/mm/slab.rst. 6992 (slub_nomerge legacy name also accepted for now) 6993 6994 slab_strict_numa [MM] 6995 Support memory policies on a per object level 6996 in the slab allocator. The default is for memory 6997 policies to be applied at the folio level when 6998 a new folio is needed or a partial folio is 6999 retrieved from the lists. Increases overhead 7000 in the slab fastpaths but gains more accurate 7001 NUMA kernel object placement which helps with slow 7002 interconnects in NUMA systems. 7003 7004 slram= [HW,MTD] 7005 7006 smart2= [HW] 7007 Format: <io1>[,<io2>[,...,<io8>]] 7008 7009 smp.csd_lock_timeout= [KNL] 7010 Specify the period of time in milliseconds 7011 that smp_call_function() and friends will wait 7012 for a CPU to release the CSD lock. This is 7013 useful when diagnosing bugs involving CPUs 7014 disabling interrupts for extended periods 7015 of time. Defaults to 5,000 milliseconds, and 7016 setting a value of zero disables this feature. 7017 This feature may be more efficiently disabled 7018 using the csdlock_debug- kernel parameter. 7019 7020 smp.panic_on_ipistall= [KNL] 7021 If a csd_lock_timeout extends for more than 7022 the specified number of milliseconds, panic the 7023 system. By default, let CSD-lock acquisition 7024 take as long as they take. Specifying 300,000 7025 for this value provides a 5-minute timeout. 7026 7027 smt= [KNL,MIPS,S390,EARLY] Set the maximum number of threads 7028 (logical CPUs) to use per physical CPU on systems 7029 capable of symmetric multithreading (SMT). Will 7030 be capped to the actual hardware limit. 7031 Format: <integer> 7032 Default: -1 (no limit) 7033 7034 softlockup_panic= 7035 [KNL] Should the soft-lockup detector generate panics. 7036 Format: <int> 7037 7038 A value of non-zero instructs the soft-lockup detector 7039 to panic the machine when a soft-lockup duration exceeds 7040 N thresholds. It is also controlled by the kernel.softlockup_panic 7041 sysctl and CONFIG_BOOTPARAM_SOFTLOCKUP_PANIC, which is the 7042 respective build-time switch to that functionality. 7043 7044 softlockup_all_cpu_backtrace= 7045 [KNL] Should the soft-lockup detector generate 7046 backtraces on all cpus. 7047 Format: 0 | 1 7048 7049 sonypi.*= [HW] Sony Programmable I/O Control Device driver 7050 See Documentation/admin-guide/laptops/sonypi.rst 7051 7052 spectre_bhi= [X86] Control mitigation of Branch History Injection 7053 (BHI) vulnerability. This setting affects the 7054 deployment of the HW BHI control and the SW BHB 7055 clearing sequence. 7056 7057 on - (default) Enable the HW or SW mitigation as 7058 needed. This protects the kernel from 7059 both syscalls and VMs. 7060 vmexit - On systems which don't have the HW mitigation 7061 available, enable the SW mitigation on vmexit 7062 ONLY. On such systems, the host kernel is 7063 protected from VM-originated BHI attacks, but 7064 may still be vulnerable to syscall attacks. 7065 off - Disable the mitigation. 7066 7067 spectre_v2= [X86,EARLY] Control mitigation of Spectre variant 2 7068 (indirect branch speculation) vulnerability. 7069 The default operation protects the kernel from 7070 user space attacks. 7071 7072 on - unconditionally enable, implies 7073 spectre_v2_user=on 7074 off - unconditionally disable, implies 7075 spectre_v2_user=off 7076 auto - kernel detects whether your CPU model is 7077 vulnerable 7078 7079 Selecting 'on' will, and 'auto' may, choose a 7080 mitigation method at run time according to the 7081 CPU, the available microcode, the setting of the 7082 CONFIG_MITIGATION_RETPOLINE configuration option, 7083 and the compiler with which the kernel was built. 7084 7085 Selecting 'on' will also enable the mitigation 7086 against user space to user space task attacks. 7087 Selecting specific mitigation does not force enable 7088 user mitigations. 7089 7090 Selecting 'off' will disable both the kernel and 7091 the user space protections. 7092 7093 Specific mitigations can also be selected manually: 7094 7095 retpoline - replace indirect branches 7096 retpoline,generic - Retpolines 7097 retpoline,lfence - LFENCE; indirect branch 7098 retpoline,amd - alias for retpoline,lfence 7099 eibrs - Enhanced/Auto IBRS 7100 eibrs,retpoline - Enhanced/Auto IBRS + Retpolines 7101 eibrs,lfence - Enhanced/Auto IBRS + LFENCE 7102 ibrs - use IBRS to protect kernel 7103 7104 Not specifying this option is equivalent to 7105 spectre_v2=auto. 7106 7107 spectre_v2_user= 7108 [X86] Control mitigation of Spectre variant 2 7109 (indirect branch speculation) vulnerability between 7110 user space tasks 7111 7112 on - Unconditionally enable mitigations. Is 7113 enforced by spectre_v2=on 7114 7115 off - Unconditionally disable mitigations. Is 7116 enforced by spectre_v2=off 7117 7118 prctl - Indirect branch speculation is enabled, 7119 but mitigation can be enabled via prctl 7120 per thread. The mitigation control state 7121 is inherited on fork. 7122 7123 prctl,ibpb 7124 - Like "prctl" above, but only STIBP is 7125 controlled per thread. IBPB is issued 7126 always when switching between different user 7127 space processes. 7128 7129 seccomp 7130 - Same as "prctl" above, but all seccomp 7131 threads will enable the mitigation unless 7132 they explicitly opt out. 7133 7134 seccomp,ibpb 7135 - Like "seccomp" above, but only STIBP is 7136 controlled per thread. IBPB is issued 7137 always when switching between different 7138 user space processes. 7139 7140 auto - Kernel selects the mitigation depending on 7141 the available CPU features and vulnerability. 7142 7143 Default mitigation: "prctl" 7144 7145 Not specifying this option is equivalent to 7146 spectre_v2_user=auto. 7147 7148 spec_rstack_overflow= 7149 [X86,EARLY] Control RAS overflow mitigation on AMD Zen CPUs 7150 7151 off - Disable mitigation 7152 microcode - Enable microcode mitigation only 7153 safe-ret - Enable sw-only safe RET mitigation (default) 7154 ibpb - Enable mitigation by issuing IBPB on 7155 kernel entry 7156 ibpb-vmexit - Issue IBPB only on VMEXIT 7157 (cloud-specific mitigation) 7158 7159 spec_store_bypass_disable= 7160 [HW,EARLY] Control Speculative Store Bypass (SSB) Disable mitigation 7161 (Speculative Store Bypass vulnerability) 7162 7163 Certain CPUs are vulnerable to an exploit against a 7164 a common industry wide performance optimization known 7165 as "Speculative Store Bypass" in which recent stores 7166 to the same memory location may not be observed by 7167 later loads during speculative execution. The idea 7168 is that such stores are unlikely and that they can 7169 be detected prior to instruction retirement at the 7170 end of a particular speculation execution window. 7171 7172 In vulnerable processors, the speculatively forwarded 7173 store can be used in a cache side channel attack, for 7174 example to read memory to which the attacker does not 7175 directly have access (e.g. inside sandboxed code). 7176 7177 This parameter controls whether the Speculative Store 7178 Bypass optimization is used. 7179 7180 On x86 the options are: 7181 7182 on - Unconditionally disable Speculative Store Bypass 7183 off - Unconditionally enable Speculative Store Bypass 7184 auto - Kernel detects whether the CPU model contains an 7185 implementation of Speculative Store Bypass and 7186 picks the most appropriate mitigation. If the 7187 CPU is not vulnerable, "off" is selected. If the 7188 CPU is vulnerable the default mitigation is 7189 architecture and Kconfig dependent. See below. 7190 prctl - Control Speculative Store Bypass per thread 7191 via prctl. Speculative Store Bypass is enabled 7192 for a process by default. The state of the control 7193 is inherited on fork. 7194 seccomp - Same as "prctl" above, but all seccomp threads 7195 will disable SSB unless they explicitly opt out. 7196 7197 Default mitigations: 7198 X86: "prctl" 7199 7200 On powerpc the options are: 7201 7202 on,auto - On Power8 and Power9 insert a store-forwarding 7203 barrier on kernel entry and exit. On Power7 7204 perform a software flush on kernel entry and 7205 exit. 7206 off - No action. 7207 7208 Not specifying this option is equivalent to 7209 spec_store_bypass_disable=auto. 7210 7211 split_llc= 7212 [X86,EARLY] Split the LLC N-ways 7213 7214 When set, the LLC is split this many ways by matching 7215 'core_id % n'. This is setup before SMP bringup and 7216 used during SMP bringup before it knows the full 7217 topology. If your core count doesn't nicely divide by 7218 the number given, you get to keep the pieces. 7219 7220 This is mostly a debug feature to emulate multiple LLCs 7221 on hardware that only have a single LLC. 7222 7223 split_lock_detect= 7224 [X86] Enable split lock detection or bus lock detection 7225 7226 When enabled (and if hardware support is present), atomic 7227 instructions that access data across cache line 7228 boundaries will result in an alignment check exception 7229 for split lock detection or a debug exception for 7230 bus lock detection. 7231 7232 off - not enabled 7233 7234 warn - the kernel will emit rate-limited warnings 7235 about applications triggering the #AC 7236 exception or the #DB exception. This mode is 7237 the default on CPUs that support split lock 7238 detection or bus lock detection. Default 7239 behavior is by #AC if both features are 7240 enabled in hardware. 7241 7242 fatal - the kernel will send SIGBUS to applications 7243 that trigger the #AC exception or the #DB 7244 exception. Default behavior is by #AC if 7245 both features are enabled in hardware. 7246 7247 ratelimit:N - 7248 Set system wide rate limit to N bus locks 7249 per second for bus lock detection. 7250 0 < N <= 1000. 7251 7252 N/A for split lock detection. 7253 7254 7255 If an #AC exception is hit in the kernel or in 7256 firmware (i.e. not while executing in user mode) 7257 the kernel will oops in either "warn" or "fatal" 7258 mode. 7259 7260 #DB exception for bus lock is triggered only when 7261 CPL > 0. 7262 7263 srbds= [X86,INTEL,EARLY] 7264 Control the Special Register Buffer Data Sampling 7265 (SRBDS) mitigation. 7266 7267 Certain CPUs are vulnerable to an MDS-like 7268 exploit which can leak bits from the random 7269 number generator. 7270 7271 By default, this issue is mitigated by 7272 microcode. However, the microcode fix can cause 7273 the RDRAND and RDSEED instructions to become 7274 much slower. Among other effects, this will 7275 result in reduced throughput from /dev/urandom. 7276 7277 The microcode mitigation can be disabled with 7278 the following option: 7279 7280 off: Disable mitigation and remove 7281 performance impact to RDRAND and RDSEED 7282 7283 srcutree.big_cpu_lim [KNL] 7284 Specifies the number of CPUs constituting a 7285 large system, such that srcu_struct structures 7286 should immediately allocate an srcu_node array. 7287 This kernel-boot parameter defaults to 128, 7288 but takes effect only when the low-order four 7289 bits of srcutree.convert_to_big is equal to 3 7290 (decide at boot). 7291 7292 srcutree.convert_to_big [KNL] 7293 Specifies under what conditions an SRCU tree 7294 srcu_struct structure will be converted to big 7295 form, that is, with an rcu_node tree: 7296 7297 0: Never. 7298 1: At init_srcu_struct() time. 7299 2: When rcutorture decides to. 7300 3: Decide at boot time (default). 7301 0x1X: Above plus if high contention. 7302 7303 Either way, the srcu_node tree will be sized based 7304 on the actual runtime number of CPUs (nr_cpu_ids) 7305 instead of the compile-time CONFIG_NR_CPUS. 7306 7307 srcutree.counter_wrap_check [KNL] 7308 Specifies how frequently to check for 7309 grace-period sequence counter wrap for the 7310 srcu_data structure's ->srcu_gp_seq_needed field. 7311 The greater the number of bits set in this kernel 7312 parameter, the less frequently counter wrap will 7313 be checked for. Note that the bottom two bits 7314 are ignored. 7315 7316 srcutree.exp_holdoff [KNL] 7317 Specifies how many nanoseconds must elapse 7318 since the end of the last SRCU grace period for 7319 a given srcu_struct until the next normal SRCU 7320 grace period will be considered for automatic 7321 expediting. Set to zero to disable automatic 7322 expediting. 7323 7324 srcutree.srcu_max_nodelay [KNL] 7325 Specifies the number of no-delay instances 7326 per jiffy for which the SRCU grace period 7327 worker thread will be rescheduled with zero 7328 delay. Beyond this limit, worker thread will 7329 be rescheduled with a sleep delay of one jiffy. 7330 7331 srcutree.srcu_max_nodelay_phase [KNL] 7332 Specifies the per-grace-period phase, number of 7333 non-sleeping polls of readers. Beyond this limit, 7334 grace period worker thread will be rescheduled 7335 with a sleep delay of one jiffy, between each 7336 rescan of the readers, for a grace period phase. 7337 7338 srcutree.srcu_retry_check_delay [KNL] 7339 Specifies number of microseconds of non-sleeping 7340 delay between each non-sleeping poll of readers. 7341 7342 srcutree.small_contention_lim [KNL] 7343 Specifies the number of update-side contention 7344 events per jiffy will be tolerated before 7345 initiating a conversion of an srcu_struct 7346 structure to big form. Note that the value of 7347 srcutree.convert_to_big must have the 0x10 bit 7348 set for contention-based conversions to occur. 7349 7350 ssbd= [ARM64,HW,EARLY] 7351 Speculative Store Bypass Disable control 7352 7353 On CPUs that are vulnerable to the Speculative 7354 Store Bypass vulnerability and offer a 7355 firmware based mitigation, this parameter 7356 indicates how the mitigation should be used: 7357 7358 force-on: Unconditionally enable mitigation for 7359 for both kernel and userspace 7360 force-off: Unconditionally disable mitigation for 7361 for both kernel and userspace 7362 kernel: Always enable mitigation in the 7363 kernel, and offer a prctl interface 7364 to allow userspace to register its 7365 interest in being mitigated too. 7366 7367 stack_guard_gap= [MM] 7368 override the default stack gap protection. The value 7369 is in page units and it defines how many pages prior 7370 to (for stacks growing down) resp. after (for stacks 7371 growing up) the main stack are reserved for no other 7372 mapping. Default value is 256 pages. 7373 7374 stack_depot_disable= [KNL,EARLY] 7375 Setting this to true through kernel command line will 7376 disable the stack depot thereby saving the static memory 7377 consumed by the stack hash table. By default this is set 7378 to false. 7379 7380 stack_depot_max_pools= [KNL,EARLY] 7381 Specify the maximum number of pools to use for storing 7382 stack traces. Pools are allocated on-demand up to this 7383 limit. Default value is 8191 pools. 7384 7385 stacktrace [FTRACE] 7386 Enable the stack tracer on boot up. 7387 7388 stacktrace_filter=[function-list] 7389 [FTRACE] Limit the functions that the stack tracer 7390 will trace at boot up. function-list is a comma-separated 7391 list of functions. This list can be changed at run 7392 time by the stack_trace_filter file in the debugfs 7393 tracing directory. Note, this enables stack tracing 7394 and the stacktrace above is not needed. 7395 7396 sti= [PARISC,HW] 7397 Format: <num> 7398 Set the STI (builtin display/keyboard on the HP-PARISC 7399 machines) console (graphic card) which should be used 7400 as the initial boot-console. 7401 See also comment in drivers/video/sticore.c. 7402 7403 sti_font= [HW] 7404 See comment in drivers/video/sticore.c. 7405 7406 stifb= [HW] 7407 Format: bpp:<bpp1>[:<bpp2>[:<bpp3>...]] 7408 7409 strict_sas_size= 7410 [X86] 7411 Format: <bool> 7412 Enable or disable strict sigaltstack size checks 7413 against the required signal frame size which 7414 depends on the supported FPU features. This can 7415 be used to filter out binaries which have 7416 not yet been made aware of AT_MINSIGSTKSZ. 7417 7418 stress_hpt [PPC,EARLY] 7419 Limits the number of kernel HPT entries in the hash 7420 page table to increase the rate of hash page table 7421 faults on kernel addresses. 7422 7423 stress_slb [PPC,EARLY] 7424 Limits the number of kernel SLB entries, and flushes 7425 them frequently to increase the rate of SLB faults 7426 on kernel addresses. 7427 7428 no_slb_preload [PPC,EARLY] 7429 Disables slb preloading for userspace. 7430 7431 sunrpc.min_resvport= 7432 sunrpc.max_resvport= 7433 [NFS,SUNRPC] 7434 SunRPC servers often require that client requests 7435 originate from a privileged port (i.e. a port in the 7436 range 0 < portnr < 1024). 7437 An administrator who wishes to reserve some of these 7438 ports for other uses may adjust the range that the 7439 kernel's sunrpc client considers to be privileged 7440 using these two parameters to set the minimum and 7441 maximum port values. 7442 7443 sunrpc.svc_rpc_per_connection_limit= 7444 [NFS,SUNRPC] 7445 Limit the number of requests that the server will 7446 process in parallel from a single connection. 7447 The default value is 0 (no limit). 7448 7449 sunrpc.pool_mode= 7450 [NFS] 7451 Deprecated. The NFS server now always uses one 7452 service thread pool per NUMA node (equivalent to a 7453 single global pool on non-NUMA machines). All of 7454 the previously accepted values (auto, global, 7455 percpu, pernode) are still accepted for backward 7456 compatibility but are ignored: the mode is always 7457 pernode, and reads always return "pernode". 7458 7459 sunrpc.tcp_slot_table_entries= 7460 sunrpc.udp_slot_table_entries= 7461 [NFS,SUNRPC] 7462 Sets the upper limit on the number of simultaneous 7463 RPC calls that can be sent from the client to a 7464 server. Increasing these values may allow you to 7465 improve throughput, but will also increase the 7466 amount of memory reserved for use by the client. 7467 7468 suspend.pm_test_delay= 7469 [SUSPEND] 7470 Sets the number of seconds to remain in a suspend test 7471 mode before resuming the system (see 7472 /sys/power/pm_test). Only available when CONFIG_PM_DEBUG 7473 is set. Default value is 5. 7474 7475 svm= [PPC] 7476 Format: { on | off | y | n | 1 | 0 } 7477 This parameter controls use of the Protected 7478 Execution Facility on pSeries. 7479 7480 swiotlb= [ARM,PPC,MIPS,X86,S390,EARLY] 7481 Format: { <int> [,<int>] | force | noforce | track_hiwater} 7482 <int> -- Number of I/O TLB slabs 7483 <int> -- Second integer after comma. Number of swiotlb 7484 areas with their own lock. Will be rounded up 7485 to a power of 2. 7486 force -- force using of bounce buffers even if they 7487 wouldn't be automatically used by the kernel 7488 noforce -- Never use bounce buffers (for debugging) 7489 track_hiwater -- Track high watermark of swiotlb buffers. 7490 Only available when CONFIG_DEBUG_FS is set. 7491 7492 switches= [HW,M68k,EARLY] 7493 7494 sysctl.*= [KNL] 7495 Set a sysctl parameter, right before loading the init 7496 process, as if the value was written to the respective 7497 /proc/sys/... file. Both '.' and '/' are recognized as 7498 separators. Unrecognized parameters and invalid values 7499 are reported in the kernel log. Sysctls registered 7500 later by a loaded module cannot be set this way. 7501 Example: sysctl.vm.swappiness=40 7502 7503 sysrq_always_enabled 7504 [KNL] 7505 Ignore sysrq setting - this boot parameter will 7506 neutralize any effect of /proc/sys/kernel/sysrq. 7507 Useful for debugging. 7508 7509 tcpmhash_entries= [KNL,NET] 7510 Set the number of tcp_metrics_hash slots. 7511 Default value is 8192 or 16384 depending on total 7512 ram pages. This is used to specify the TCP metrics 7513 cache size. See Documentation/networking/ip-sysctl.rst 7514 "tcp_no_metrics_save" section for more details. 7515 7516 tdfx= [HW,DRM] 7517 7518 test_suspend= [SUSPEND] 7519 Format: { "mem" | "standby" | "freeze" }[,N] 7520 Specify "mem" (for Suspend-to-RAM) or "standby" (for 7521 standby suspend) or "freeze" (for suspend type freeze) 7522 as the system sleep state during system startup with 7523 the optional capability to repeat N number of times. 7524 The system is woken from this state using a 7525 wakeup-capable RTC alarm. 7526 7527 thash_entries= [KNL,NET] 7528 Set number of hash buckets for TCP connection 7529 7530 thermal.act= [HW,ACPI] 7531 -1: disable all active trip points in all thermal zones 7532 <degrees C>: override all lowest active trip points 7533 7534 thermal.crt= [HW,ACPI] 7535 -1: disable all critical trip points in all thermal zones 7536 <degrees C>: override all critical trip points 7537 7538 thermal.off= [HW,ACPI] 7539 1: disable ACPI thermal control 7540 7541 thermal.psv= [HW,ACPI] 7542 -1: disable all passive trip points 7543 <degrees C>: override all passive trip points to this 7544 value 7545 7546 thermal.tzp= [HW,ACPI] 7547 Specify global default ACPI thermal zone polling rate 7548 <deci-seconds>: poll all this frequency 7549 0: no polling (default) 7550 7551 thp_anon= [KNL] 7552 Format: <size>[KMG],<size>[KMG]:<state>;<size>[KMG]-<size>[KMG]:<state> 7553 state is one of "always", "madvise", "never" or "inherit". 7554 Control the default behavior of the system with respect 7555 to anonymous transparent hugepages. 7556 Can be used multiple times for multiple anon THP sizes. 7557 See Documentation/admin-guide/mm/transhuge.rst for more 7558 details. 7559 7560 threadirqs [KNL,EARLY] 7561 Force threading of all interrupt handlers except those 7562 marked explicitly IRQF_NO_THREAD. 7563 7564 thp_shmem= [KNL] 7565 Format: <size>[KMG],<size>[KMG]:<policy>;<size>[KMG]-<size>[KMG]:<policy> 7566 Control the default policy of each hugepage size for the 7567 internal shmem mount. <policy> is one of policies available 7568 for the shmem mount ("always", "inherit", "never", "within_size", 7569 and "advise"). 7570 It can be used multiple times for multiple shmem THP sizes. 7571 See Documentation/admin-guide/mm/transhuge.rst for more 7572 details. 7573 7574 tlbi= [X86-64] 7575 Format: {ipi} 7576 ipi: switch to IPI-based TLB flushing 7577 7578 topology= [S390,EARLY] 7579 Format: {off | on} 7580 Specify if the kernel should make use of the cpu 7581 topology information if the hardware supports this. 7582 The scheduler will make use of this information and 7583 e.g. base its process migration decisions on it. 7584 Default is on. 7585 7586 torture.disable_onoff_at_boot= [KNL] 7587 Prevent the CPU-hotplug component of torturing 7588 until after init has spawned. 7589 7590 torture.ftrace_dump_at_shutdown= [KNL] 7591 Dump the ftrace buffer at torture-test shutdown, 7592 even if there were no errors. This can be a 7593 very costly operation when many torture tests 7594 are running concurrently, especially on systems 7595 with rotating-rust storage. 7596 7597 torture.verbose_sleep_frequency= [KNL] 7598 Specifies how many verbose printk()s should be 7599 emitted between each sleep. The default of zero 7600 disables verbose-printk() sleeping. 7601 7602 torture.verbose_sleep_duration= [KNL] 7603 Duration of each verbose-printk() sleep in jiffies. 7604 7605 tpm.disable_pcr_integrity= [HW,TPM] 7606 Do not protect PCR registers from unintended physical 7607 access, or interposers in the bus by the means of 7608 having an integrity protected session wrapped around 7609 TPM2_PCR_Extend command. Consider this in a situation 7610 where TPM is heavily utilized by IMA, thus protection 7611 causing a major performance hit, and the space where 7612 machines are deployed is by other means guarded. 7613 7614 tpm_crb_ffa.busy_timeout_ms= [ARM64,TPM] 7615 Maximum time in milliseconds to retry sending a message 7616 to the TPM service before giving up. This parameter controls 7617 how long the system will continue retrying when the TPM 7618 service is busy. 7619 Format: <unsigned int> 7620 Default: 2000 (2 seconds) 7621 7622 tpm_suspend_pcr=[HW,TPM] 7623 Format: integer pcr id 7624 Specify that at suspend time, the tpm driver 7625 should extend the specified pcr with zeros, 7626 as a workaround for some chips which fail to 7627 flush the last written pcr on TPM_SaveState. 7628 This will guarantee that all the other pcrs 7629 are saved. 7630 7631 tpm_tis.interrupts= [HW,TPM] 7632 Enable interrupts for the MMIO based physical layer 7633 for the FIFO interface. By default it is set to false 7634 (0). For more information about TPM hardware interfaces 7635 defined by Trusted Computing Group (TCG) see 7636 https://trustedcomputinggroup.org/resource/pc-client-platform-tpm-profile-ptp-specification/ 7637 7638 tp_printk [FTRACE] 7639 Have the tracepoints sent to printk as well as the 7640 tracing ring buffer. This is useful for early boot up 7641 where the system hangs or reboots and does not give the 7642 option for reading the tracing buffer or performing a 7643 ftrace_dump_on_oops. 7644 7645 To turn off having tracepoints sent to printk, 7646 echo 0 > /proc/sys/kernel/tracepoint_printk 7647 Note, echoing 1 into this file without the 7648 tp_printk kernel cmdline option has no effect. 7649 7650 The tp_printk_stop_on_boot (see below) can also be used 7651 to stop the printing of events to console at 7652 late_initcall_sync. 7653 7654 ** CAUTION ** 7655 7656 Having tracepoints sent to printk() and activating high 7657 frequency tracepoints such as irq or sched, can cause 7658 the system to live lock. 7659 7660 tp_printk_stop_on_boot [FTRACE] 7661 When tp_printk (above) is set, it can cause a lot of noise 7662 on the console. It may be useful to only include the 7663 printing of events during boot up, as user space may 7664 make the system inoperable. 7665 7666 This command line option will stop the printing of events 7667 to console at the late_initcall_sync() time frame. 7668 7669 trace_buf_size=nn[KMG] 7670 [FTRACE] will set tracing buffer size on each cpu. 7671 7672 trace_clock= [FTRACE] Set the clock used for tracing events 7673 at boot up. 7674 local - Use the per CPU time stamp counter 7675 (converted into nanoseconds). Fast, but 7676 depending on the architecture, may not be 7677 in sync between CPUs. 7678 global - Event time stamps are synchronized across 7679 CPUs. May be slower than the local clock, 7680 but better for some race conditions. 7681 counter - Simple counting of events (1, 2, ..) 7682 note, some counts may be skipped due to the 7683 infrastructure grabbing the clock more than 7684 once per event. 7685 uptime - Use jiffies as the time stamp. 7686 perf - Use the same clock that perf uses. 7687 mono - Use ktime_get_mono_fast_ns() for time stamps. 7688 mono_raw - Use ktime_get_raw_fast_ns() for time 7689 stamps. 7690 boot - Use ktime_get_boot_fast_ns() for time stamps. 7691 Architectures may add more clocks. See 7692 Documentation/trace/ftrace.rst for more details. 7693 7694 trace_event=[event-list] 7695 [FTRACE] Set and start specified trace events in order 7696 to facilitate early boot debugging. The event-list is a 7697 comma-separated list of trace events to enable. See 7698 also Documentation/trace/events.rst 7699 7700 To enable modules, use :mod: keyword: 7701 7702 trace_event=:mod:<module> 7703 7704 The value before :mod: will only enable specific events 7705 that are part of the module. See the above mentioned 7706 document for more information. 7707 7708 trace_instance=[instance-info] 7709 [FTRACE] Create a ring buffer instance early in boot up. 7710 This will be listed in: 7711 7712 /sys/kernel/tracing/instances 7713 7714 Events can be enabled at the time the instance is created 7715 via: 7716 7717 trace_instance=<name>,<system1>:<event1>,<system2>:<event2> 7718 7719 Note, the "<system*>:" portion is optional if the event is 7720 unique. 7721 7722 trace_instance=foo,sched:sched_switch,irq_handler_entry,initcall 7723 7724 will enable the "sched_switch" event (note, the "sched:" is optional, and 7725 the same thing would happen if it was left off). The irq_handler_entry 7726 event, and all events under the "initcall" system. 7727 7728 Flags can be added to the instance to modify its behavior when it is 7729 created. The flags are separated by '^'. 7730 7731 The available flags are: 7732 7733 traceoff - Have the tracing instance tracing disabled after it is created. 7734 traceprintk - Have trace_printk() write into this trace instance 7735 (note, "printk" and "trace_printk" can also be used) 7736 7737 trace_instance=foo^traceoff^traceprintk,sched,irq 7738 7739 The flags must come before the defined events. 7740 7741 If memory has been reserved (see memmap for x86), the instance 7742 can use that memory: 7743 7744 memmap=12M$0x284500000 trace_instance=boot_map@0x284500000:12M 7745 7746 The above will create a "boot_map" instance that uses the physical 7747 memory at 0x284500000 that is 12Megs. The per CPU buffers of that 7748 instance will be split up accordingly. 7749 7750 Alternatively, the memory can be reserved by the reserve_mem option: 7751 7752 reserve_mem=12M:4096:trace trace_instance=boot_map@trace 7753 7754 This will reserve 12 megabytes at boot up with a 4096 byte alignment 7755 and place the ring buffer in this memory. Note that due to KASLR, the 7756 memory may not be the same location each time, which will not preserve 7757 the buffer content. 7758 7759 Also note that the layout of the ring buffer data may change between 7760 kernel versions where the validator will fail and reset the ring buffer 7761 if the layout is not the same as the previous kernel. 7762 7763 If the ring buffer is used for persistent bootups and has events enabled, 7764 it is recommend to disable tracing so that events from a previous boot do not 7765 mix with events of the current boot (unless you are debugging a random crash 7766 at boot up). 7767 7768 reserve_mem=12M:4096:trace trace_instance=boot_map^traceoff^traceprintk@trace,sched,irq 7769 7770 Note, saving the trace buffer across reboots does require that the system 7771 is set up to not wipe memory. For instance, CONFIG_RESET_ATTACK_MITIGATION 7772 can force a memory reset on boot which will clear any trace that was stored. 7773 This is just one of many ways that can clear memory. Make sure your system 7774 keeps the content of memory across reboots before relying on this option. 7775 7776 NB: Both the mapped address and size must be page aligned for the architecture. 7777 7778 See also Documentation/trace/debugging.rst 7779 7780 7781 trace_options=[option-list] 7782 [FTRACE] Enable or disable tracer options at boot. 7783 The option-list is a comma delimited list of options 7784 that can be enabled or disabled just as if you were 7785 to echo the option name into 7786 7787 /sys/kernel/tracing/trace_options 7788 7789 For example, to enable stacktrace option (to dump the 7790 stack trace of each event), add to the command line: 7791 7792 trace_options=stacktrace 7793 7794 See also Documentation/trace/ftrace.rst "trace options" 7795 section. 7796 7797 trace_trigger=[trigger-list] 7798 [FTRACE] Add an event trigger on specific events. 7799 Set a trigger on top of a specific event, with an optional 7800 filter. 7801 7802 The format is "trace_trigger=<event>.<trigger>[ if <filter>],..." 7803 Where more than one trigger may be specified that are comma delimited. 7804 7805 For example: 7806 7807 trace_trigger="sched_switch.stacktrace if prev_state == 2" 7808 7809 The above will enable the "stacktrace" trigger on the "sched_switch" 7810 event but only trigger it if the "prev_state" of the "sched_switch" 7811 event is "2" (TASK_UNINTERRUPTIBLE). 7812 7813 See also "Event triggers" in Documentation/trace/events.rst 7814 7815 7816 traceoff_after_boot 7817 [FTRACE] Sometimes tracing is used to debug issues 7818 during the boot process. Since the trace buffer has a 7819 limited amount of storage, it may be prudent to 7820 disable tracing after the boot is finished, otherwise 7821 the critical information may be overwritten. With this 7822 option, the main tracing buffer will be turned off at 7823 the end of the boot process. 7824 7825 traceoff_on_warning 7826 [FTRACE] enable this option to disable tracing when a 7827 warning is hit. This turns off "tracing_on". Tracing can 7828 be enabled again by echoing '1' into the "tracing_on" 7829 file located in /sys/kernel/tracing/ 7830 7831 This option is useful, as it disables the trace before 7832 the WARNING dump is called, which prevents the trace to 7833 be filled with content caused by the warning output. 7834 7835 This option can also be set at run time via the sysctl 7836 option: kernel/traceoff_on_warning 7837 7838 transparent_hugepage= 7839 [KNL] 7840 Format: [always|madvise|never] 7841 Can be used to control the default behavior of the system 7842 with respect to transparent hugepages. 7843 See Documentation/admin-guide/mm/transhuge.rst 7844 for more details. 7845 7846 transparent_hugepage_shmem= [KNL] 7847 Format: [always|within_size|advise|never|deny|force] 7848 Can be used to control the hugepage allocation policy for 7849 the internal shmem mount. 7850 See Documentation/admin-guide/mm/transhuge.rst 7851 for more details. 7852 7853 transparent_hugepage_tmpfs= [KNL] 7854 Format: [always|within_size|advise|never] 7855 Can be used to control the default hugepage allocation policy 7856 for the tmpfs mount. 7857 See Documentation/admin-guide/mm/transhuge.rst 7858 for more details. 7859 7860 trusted.source= [KEYS] 7861 Format: <string> 7862 This parameter identifies the trust source as a backend 7863 for trusted keys implementation. Supported trust 7864 sources: 7865 - "tpm" 7866 - "tee" 7867 - "caam" 7868 - "dcp" 7869 - "pkwm" 7870 If not specified then it defaults to iterating through 7871 the trust source list starting with TPM and assigns the 7872 first trust source as a backend which is initialized 7873 successfully during iteration. 7874 7875 trusted.debug= [KEYS] 7876 Format: <bool> 7877 Enable trusted keys debug traces at runtime when 7878 CONFIG_TRUSTED_KEYS_DEBUG=y. 7879 7880 To make the traces visible after enabling the option, 7881 use trusted.dyndbg='+p' as needed. By convention, 7882 the subsystem uses pr_debug() for these traces. 7883 7884 SAFETY: The traces can leak sensitive data, so be 7885 cautious before enabling this. They remain inactive 7886 unless this parameter is set this option to a true 7887 value. 7888 7889 Default: false 7890 7891 trusted.rng= [KEYS] 7892 Format: <string> 7893 The RNG used to generate key material for trusted keys. 7894 Can be one of: 7895 - "kernel" 7896 - the same value as trusted.source: "tpm" or "tee" 7897 - "default" 7898 If not specified, "default" is used. In this case, 7899 the RNG's choice is left to each individual trust source. 7900 7901 trusted.dcp_use_otp_key 7902 This is intended to be used in combination with 7903 trusted.source=dcp and will select the DCP OTP key 7904 instead of the DCP UNIQUE key blob encryption. 7905 7906 trusted.dcp_skip_zk_test 7907 This is intended to be used in combination with 7908 trusted.source=dcp and will disable the check if the 7909 blob key is all zeros. This is helpful for situations where 7910 having this key zero'ed is acceptable. E.g. in testing 7911 scenarios. 7912 7913 tsa= [X86] Control mitigation for Transient Scheduler 7914 Attacks on AMD CPUs. Search the following in your 7915 favourite search engine for more details: 7916 7917 "Technical guidance for mitigating transient scheduler 7918 attacks". 7919 7920 off - disable the mitigation 7921 on - enable the mitigation (default) 7922 user - mitigate only user/kernel transitions 7923 vm - mitigate only guest/host transitions 7924 7925 7926 tsc= Disable clocksource stability checks for TSC. 7927 Format: <string> 7928 [x86] reliable: mark tsc clocksource as reliable, this 7929 disables clocksource verification at runtime, as well 7930 as the stability checks done at bootup. Used to enable 7931 high-resolution timer mode on older hardware, and in 7932 virtualized environment. 7933 [x86] noirqtime: Do not use TSC to do irq accounting. 7934 Used to run time disable IRQ_TIME_ACCOUNTING on any 7935 platforms where RDTSC is slow and this accounting 7936 can add overhead. 7937 [x86] unstable: mark the TSC clocksource as unstable, this 7938 marks the TSC unconditionally unstable at bootup and 7939 avoids any further wobbles once the TSC watchdog notices. 7940 [x86] nowatchdog: disable clocksource watchdog. Used 7941 in situations with strict latency requirements (where 7942 interruptions from clocksource watchdog are not 7943 acceptable). 7944 [x86] recalibrate: force recalibration against a HW timer 7945 (HPET or PM timer) on systems whose TSC frequency was 7946 obtained from HW or FW using either an MSR or CPUID(0x15). 7947 Warn if the difference is more than 500 ppm. 7948 [x86] watchdog: Enforce the clocksource watchdog on TSC 7949 7950 tsc_early_khz= [X86,EARLY] Skip early TSC calibration and use the given 7951 value instead. Useful when the early TSC frequency discovery 7952 procedure is not reliable, such as on overclocked systems 7953 with CPUID.16h support and partial CPUID.15h support. 7954 Format: <unsigned int> 7955 7956 tsx= [X86] Control Transactional Synchronization 7957 Extensions (TSX) feature in Intel processors that 7958 support TSX control. 7959 7960 This parameter controls the TSX feature. The options are: 7961 7962 on - Enable TSX on the system. Although there are 7963 mitigations for all known security vulnerabilities, 7964 TSX has been known to be an accelerator for 7965 several previous speculation-related CVEs, and 7966 so there may be unknown security risks associated 7967 with leaving it enabled. 7968 7969 off - Disable TSX on the system. (Note that this 7970 option takes effect only on newer CPUs which are 7971 not vulnerable to MDS, i.e., have 7972 MSR_IA32_ARCH_CAPABILITIES.MDS_NO=1 and which get 7973 the new IA32_TSX_CTRL MSR through a microcode 7974 update. This new MSR allows for the reliable 7975 deactivation of the TSX functionality.) 7976 7977 auto - Disable TSX if X86_BUG_TAA is present, 7978 otherwise enable TSX on the system. 7979 7980 Not specifying this option is equivalent to tsx=off. 7981 7982 See Documentation/admin-guide/hw-vuln/tsx_async_abort.rst 7983 for more details. 7984 7985 tsx_async_abort= [X86,INTEL,EARLY] Control mitigation for the TSX Async 7986 Abort (TAA) vulnerability. 7987 7988 Similar to Micro-architectural Data Sampling (MDS) 7989 certain CPUs that support Transactional 7990 Synchronization Extensions (TSX) are vulnerable to an 7991 exploit against CPU internal buffers which can forward 7992 information to a disclosure gadget under certain 7993 conditions. 7994 7995 In vulnerable processors, the speculatively forwarded 7996 data can be used in a cache side channel attack, to 7997 access data to which the attacker does not have direct 7998 access. 7999 8000 This parameter controls the TAA mitigation. The 8001 options are: 8002 8003 full - Enable TAA mitigation on vulnerable CPUs 8004 if TSX is enabled. 8005 8006 full,nosmt - Enable TAA mitigation and disable SMT on 8007 vulnerable CPUs. If TSX is disabled, SMT 8008 is not disabled because CPU is not 8009 vulnerable to cross-thread TAA attacks. 8010 off - Unconditionally disable TAA mitigation 8011 8012 On MDS-affected machines, tsx_async_abort=off can be 8013 prevented by an active MDS mitigation as both vulnerabilities 8014 are mitigated with the same mechanism so in order to disable 8015 this mitigation, you need to specify mds=off too. 8016 8017 Not specifying this option is equivalent to 8018 tsx_async_abort=full. On CPUs which are MDS affected 8019 and deploy MDS mitigation, TAA mitigation is not 8020 required and doesn't provide any additional 8021 mitigation. 8022 8023 For details see: 8024 Documentation/admin-guide/hw-vuln/tsx_async_abort.rst 8025 8026 turbografx.map[2|3]= [HW,JOY] 8027 TurboGraFX parallel port interface 8028 Format: 8029 <port#>,<js1>,<js2>,<js3>,<js4>,<js5>,<js6>,<js7> 8030 See also Documentation/input/devices/joystick-parport.rst 8031 8032 udbg-immortal [PPC] When debugging early kernel crashes that 8033 happen after console_init() and before a proper 8034 console driver takes over, this boot options might 8035 help "seeing" what's going on. 8036 8037 uhash_entries= [KNL,NET] 8038 Set number of hash buckets for UDP/UDP-Lite connections 8039 8040 uhci-hcd.ignore_oc= 8041 [USB] Ignore overcurrent events (default N). 8042 Some badly-designed motherboards generate lots of 8043 bogus events, for ports that aren't wired to 8044 anything. Set this parameter to avoid log spamming. 8045 Note that genuine overcurrent events won't be 8046 reported either. 8047 8048 unaligned_scalar_speed= 8049 [RISCV] 8050 Format: {slow | fast | unsupported} 8051 Allow skipping scalar unaligned access speed tests. This 8052 is useful for testing alternative code paths and to skip 8053 the tests in environments where they run too slowly. All 8054 CPUs must have the same scalar unaligned access speed. 8055 8056 unaligned_vector_speed= 8057 [RISCV] 8058 Format: {slow | fast | unsupported} 8059 Allow skipping vector unaligned access speed tests. This 8060 is useful for testing alternative code paths and to skip 8061 the tests in environments where they run too slowly. All 8062 CPUs must have the same vector unaligned access speed. 8063 8064 unknown_nmi_panic 8065 [X86] Cause panic on unknown NMI. 8066 8067 unwind_debug [X86-64,EARLY] 8068 Enable unwinder debug output. This can be 8069 useful for debugging certain unwinder error 8070 conditions, including corrupt stacks and 8071 bad/missing unwinder metadata. 8072 8073 usbcore.authorized_default= 8074 [USB] Default USB device authorization: 8075 (default -1 = authorized (same as 1), 8076 0 = not authorized, 1 = authorized, 2 = authorized 8077 if device connected to internal port) 8078 8079 usbcore.autosuspend= 8080 [USB] The autosuspend time delay (in seconds) used 8081 for newly-detected USB devices (default 2). This 8082 is the time required before an idle device will be 8083 autosuspended. Devices for which the delay is set 8084 to a negative value won't be autosuspended at all. 8085 8086 usbcore.usbfs_snoop= 8087 [USB] Set to log all usbfs traffic (default 0 = off). 8088 8089 usbcore.usbfs_snoop_max= 8090 [USB] Maximum number of bytes to snoop in each URB 8091 (default = 65536). 8092 8093 usbcore.blinkenlights= 8094 [USB] Set to cycle leds on hubs (default 0 = off). 8095 8096 usbcore.old_scheme_first= 8097 [USB] Start with the old device initialization 8098 scheme (default 0 = off). 8099 8100 usbcore.usbfs_memory_mb= 8101 [USB] Memory limit (in MB) for buffers allocated by 8102 usbfs (default = 16, 0 = max = 2047). 8103 8104 usbcore.use_both_schemes= 8105 [USB] Try the other device initialization scheme 8106 if the first one fails (default 1 = enabled). 8107 8108 usbcore.initial_descriptor_timeout= 8109 [USB] Specifies timeout for the initial 64-byte 8110 USB_REQ_GET_DESCRIPTOR request in milliseconds 8111 (default 5000 = 5.0 seconds). 8112 8113 usbcore.nousb [USB] Disable the USB subsystem 8114 8115 usbcore.quirks= 8116 [USB] A list of quirk entries to augment the built-in 8117 usb core quirk list. List entries are separated by 8118 commas. Each entry has the form 8119 VendorID:ProductID:Flags. The IDs are 4-digit hex 8120 numbers and Flags is a set of letters. Each letter 8121 will change the built-in quirk; setting it if it is 8122 clear and clearing it if it is set. The letters have 8123 the following meanings: 8124 a = USB_QUIRK_STRING_FETCH_255 (string 8125 descriptors must not be fetched using 8126 a 255-byte read); 8127 b = USB_QUIRK_RESET_RESUME (device can't resume 8128 correctly so reset it instead); 8129 c = USB_QUIRK_NO_SET_INTF (device can't handle 8130 Set-Interface requests); 8131 d = USB_QUIRK_CONFIG_INTF_STRINGS (device can't 8132 handle its Configuration or Interface 8133 strings); 8134 e = USB_QUIRK_RESET (device can't be reset 8135 (e.g morph devices), don't use reset); 8136 f = USB_QUIRK_HONOR_BNUMINTERFACES (device has 8137 more interface descriptions than the 8138 bNumInterfaces count, and can't handle 8139 talking to these interfaces); 8140 g = USB_QUIRK_DELAY_INIT (device needs a pause 8141 during initialization, after we read 8142 the device descriptor); 8143 h = USB_QUIRK_LINEAR_UFRAME_INTR_BINTERVAL (For 8144 high speed and super speed interrupt 8145 endpoints, the USB 2.0 and USB 3.0 spec 8146 require the interval in microframes (1 8147 microframe = 125 microseconds) to be 8148 calculated as interval = 2 ^ 8149 (bInterval-1). 8150 Devices with this quirk report their 8151 bInterval as the result of this 8152 calculation instead of the exponent 8153 variable used in the calculation); 8154 i = USB_QUIRK_DEVICE_QUALIFIER (device can't 8155 handle device_qualifier descriptor 8156 requests); 8157 j = USB_QUIRK_IGNORE_REMOTE_WAKEUP (device 8158 generates spurious wakeup, ignore 8159 remote wakeup capability); 8160 k = USB_QUIRK_NO_LPM (device can't handle Link 8161 Power Management); 8162 l = USB_QUIRK_LINEAR_FRAME_INTR_BINTERVAL 8163 (Device reports its bInterval as linear 8164 frames instead of the USB 2.0 8165 calculation); 8166 m = USB_QUIRK_DISCONNECT_SUSPEND (Device needs 8167 to be disconnected before suspend to 8168 prevent spurious wakeup); 8169 n = USB_QUIRK_DELAY_CTRL_MSG (Device needs a 8170 pause after every control message); 8171 o = USB_QUIRK_HUB_SLOW_RESET (Hub needs extra 8172 delay after resetting its port); 8173 p = USB_QUIRK_SHORT_SET_ADDRESS_REQ_TIMEOUT 8174 (Reduce timeout of the SET_ADDRESS 8175 request from 5000 ms to 500 ms); 8176 q = USB_QUIRK_FORCE_ONE_CONFIG (Device 8177 claims zero configurations, 8178 forcing to 1); 8179 r = USB_QUIRK_WINDOWS_CONFIG_REQ_SIZE (Device 8180 fails during initialization when asked for 8181 9-bytes configuration descriptor request. 8182 Ask for 255-bytes request instead to mirror 8183 Windows' behavior); 8184 Example: quirks=0781:5580:bk,0a5c:5834:gij 8185 8186 usbhid.mousepoll= 8187 [USBHID] The interval which mice are to be polled at. 8188 8189 usbhid.jspoll= 8190 [USBHID] The interval which joysticks are to be polled at. 8191 8192 usbhid.kbpoll= 8193 [USBHID] The interval which keyboards are to be polled at. 8194 8195 usb-storage.delay_use= 8196 [UMS] The delay in seconds before a new device is 8197 scanned for Logical Units (default 1). 8198 Optionally the delay in milliseconds if the value has 8199 suffix with "ms". 8200 Example: delay_use=2567ms 8201 8202 usb-storage.quirks= 8203 [UMS] A list of quirks entries to supplement or 8204 override the built-in unusual_devs list. List 8205 entries are separated by commas. Each entry has 8206 the form VID:PID:Flags where VID and PID are Vendor 8207 and Product ID values (4-digit hex numbers) and 8208 Flags is a set of characters, each corresponding 8209 to a common usb-storage quirk flag as follows: 8210 a = SANE_SENSE (collect more than 18 bytes 8211 of sense data, not on uas); 8212 b = BAD_SENSE (don't collect more than 18 8213 bytes of sense data, not on uas); 8214 c = FIX_CAPACITY (decrease the reported 8215 device capacity by one sector); 8216 d = NO_READ_DISC_INFO (don't use 8217 READ_DISC_INFO command, not on uas); 8218 e = NO_READ_CAPACITY_16 (don't use 8219 READ_CAPACITY_16 command); 8220 f = NO_REPORT_OPCODES (don't use report opcodes 8221 command, uas only); 8222 g = MAX_SECTORS_240 (don't transfer more than 8223 240 sectors at a time, uas only); 8224 h = CAPACITY_HEURISTICS (decrease the 8225 reported device capacity by one 8226 sector if the number is odd); 8227 i = IGNORE_DEVICE (don't bind to this 8228 device); 8229 j = NO_REPORT_LUNS (don't use report luns 8230 command, uas only); 8231 k = NO_SAME (do not use WRITE_SAME, uas only) 8232 l = NOT_LOCKABLE (don't try to lock and 8233 unlock ejectable media, not on uas); 8234 m = MAX_SECTORS_64 (don't transfer more 8235 than 64 sectors = 32 KB at a time, 8236 not on uas); 8237 n = INITIAL_READ10 (force a retry of the 8238 initial READ(10) command, not on uas); 8239 o = CAPACITY_OK (accept the capacity 8240 reported by the device, not on uas); 8241 p = WRITE_CACHE (the device cache is ON 8242 by default, not on uas); 8243 r = IGNORE_RESIDUE (the device reports 8244 bogus residue values, not on uas); 8245 s = SINGLE_LUN (the device has only one 8246 Logical Unit); 8247 t = NO_ATA_1X (don't allow ATA(12) and ATA(16) 8248 commands, uas only); 8249 u = IGNORE_UAS (don't bind to the uas driver); 8250 w = NO_WP_DETECT (don't test whether the 8251 medium is write-protected). 8252 y = ALWAYS_SYNC (issue a SYNCHRONIZE_CACHE 8253 even if the device claims no cache, 8254 not on uas) 8255 Example: quirks=0419:aaf5:rl,0421:0433:rc 8256 8257 user_debug= [KNL,ARM] 8258 Format: <int> 8259 See arch/arm/Kconfig.debug help text. 8260 1 - undefined instruction events 8261 2 - system calls 8262 4 - invalid data aborts 8263 8 - SIGSEGV faults 8264 16 - SIGBUS faults 8265 Example: user_debug=31 8266 8267 vdso= [X86,SH,SPARC] 8268 On X86_32, this is an alias for vdso32=. Otherwise: 8269 8270 vdso=1: enable VDSO (the default) 8271 vdso=0: disable VDSO mapping 8272 8273 vdso32= [X86] Control the 32-bit vDSO 8274 vdso32=1: enable 32-bit VDSO 8275 vdso32=0 or vdso32=2: disable 32-bit VDSO 8276 8277 See the help text for CONFIG_COMPAT_VDSO for more 8278 details. If CONFIG_COMPAT_VDSO is set, the default is 8279 vdso32=0; otherwise, the default is vdso32=1. 8280 8281 For compatibility with older kernels, vdso32=2 is an 8282 alias for vdso32=0. 8283 8284 Try vdso32=0 if you encounter an error that says: 8285 dl_main: Assertion `(void *) ph->p_vaddr == _rtld_local._dl_sysinfo_dso' failed! 8286 8287 video= [FB,EARLY] Frame buffer configuration 8288 See Documentation/fb/modedb.rst. 8289 8290 video.brightness_switch_enabled= [ACPI] 8291 Format: [0|1] 8292 If set to 1, on receiving an ACPI notify event 8293 generated by hotkey, video driver will adjust brightness 8294 level and then send out the event to user space through 8295 the allocated input device. If set to 0, video driver 8296 will only send out the event without touching backlight 8297 brightness level. 8298 default: 1 8299 8300 virtio_mmio.device= 8301 [VMMIO] Memory mapped virtio (platform) device. 8302 8303 <size>@<baseaddr>:<irq>[:<id>] 8304 where: 8305 <size> := size (can use standard suffixes 8306 like K, M and G) 8307 <baseaddr> := physical base address 8308 <irq> := interrupt number (as passed to 8309 request_irq()) 8310 <id> := (optional) platform device id 8311 example: 8312 virtio_mmio.device=1K@0x100b0000:48:7 8313 8314 Can be used multiple times for multiple devices. 8315 8316 vga= [BOOT,X86-32] Select a particular video mode 8317 See Documentation/arch/x86/boot.rst and 8318 Documentation/admin-guide/svga.rst. 8319 Use vga=ask for menu. 8320 This is actually a boot loader parameter; the value is 8321 passed to the kernel using a special protocol. 8322 8323 vm_debug[=options] [KNL] Available with CONFIG_DEBUG_VM=y. 8324 May slow down system boot speed, especially when 8325 enabled on systems with a large amount of memory. 8326 All options are enabled by default, and this 8327 interface is meant to allow for selectively 8328 enabling or disabling specific virtual memory 8329 debugging features. 8330 8331 Available options are: 8332 P Enable page structure init time poisoning 8333 - Disable all of the above options 8334 8335 vmalloc=nn[KMG] [KNL,BOOT,EARLY] Forces the vmalloc area to have an 8336 exact size of <nn>. This can be used to increase 8337 the minimum size (128MB on x86, arm32 platforms). 8338 It can also be used to decrease the size and leave more room 8339 for directly mapped kernel RAM. Note that this parameter does 8340 not exist on many other platforms (including arm64, alpha, 8341 loongarch, arc, csky, hexagon, microblaze, mips, nios2, openrisc, 8342 parisc, m64k, powerpc, riscv, sh, um, xtensa, s390, sparc). 8343 8344 vmcp_cma=nn[MG] [KNL,S390,EARLY] 8345 Sets the memory size reserved for contiguous memory 8346 allocations for the vmcp device driver. 8347 8348 vmhalt= [KNL,S390] Perform z/VM CP command after system halt. 8349 Format: <command> 8350 8351 vmpanic= [KNL,S390] Perform z/VM CP command after kernel panic. 8352 Format: <command> 8353 8354 vmpoff= [KNL,S390] Perform z/VM CP command after power off. 8355 Format: <command> 8356 8357 vmscape= [X86] Controls mitigation for VMscape attacks. 8358 VMscape attacks can leak information from a userspace 8359 hypervisor to a guest via speculative side-channels. 8360 8361 off - disable the mitigation 8362 ibpb - use Indirect Branch Prediction Barrier 8363 (IBPB) mitigation (default) 8364 force - force vulnerability detection even on 8365 unaffected processors 8366 8367 vsyscall= [X86-64,EARLY] 8368 Controls the behavior of vsyscalls (i.e. calls to 8369 fixed addresses of 0xffffffffff600x00 from legacy 8370 code). Most statically-linked binaries and older 8371 versions of glibc use these calls. Because these 8372 functions are at fixed addresses, they make nice 8373 targets for exploits that can control RIP. 8374 8375 emulate Vsyscalls turn into traps and are emulated 8376 reasonably safely. The vsyscall page is 8377 readable. This disables the Linear 8378 Address Space Separation (LASS) security 8379 feature and makes the system less secure. 8380 8381 xonly [default] Vsyscalls turn into traps and are 8382 emulated reasonably safely. The vsyscall 8383 page is not readable. 8384 8385 none Vsyscalls don't work at all. This makes 8386 them quite hard to use for exploits but 8387 might break your system. 8388 8389 vt.color= [VT] Default text color. 8390 Format: 0xYX, X = foreground, Y = background. 8391 Default: 0x07 = light gray on black. 8392 8393 vt.cur_default= [VT] Default cursor shape. 8394 Format: 0xCCBBAA, where AA, BB, and CC are the same as 8395 the parameters of the <Esc>[?A;B;Cc escape sequence; 8396 see vga-softcursor.rst. Default: 2 = underline. 8397 8398 vt.default_blu= [VT] 8399 Format: <blue0>,<blue1>,<blue2>,...,<blue15> 8400 Change the default blue palette of the console. 8401 This is a 16-member array composed of values 8402 ranging from 0-255. 8403 8404 vt.default_grn= [VT] 8405 Format: <green0>,<green1>,<green2>,...,<green15> 8406 Change the default green palette of the console. 8407 This is a 16-member array composed of values 8408 ranging from 0-255. 8409 8410 vt.default_red= [VT] 8411 Format: <red0>,<red1>,<red2>,...,<red15> 8412 Change the default red palette of the console. 8413 This is a 16-member array composed of values 8414 ranging from 0-255. 8415 8416 vt.default_utf8= 8417 [VT] 8418 Format=<0|1> 8419 Set system-wide default UTF-8 mode for all tty's. 8420 Default is 1, i.e. UTF-8 mode is enabled for all 8421 newly opened terminals. 8422 8423 vt.global_cursor_default= 8424 [VT] 8425 Format=<-1|0|1> 8426 Set system-wide default for whether a cursor 8427 is shown on new VTs. Default is -1, 8428 i.e. cursors will be created by default unless 8429 overridden by individual drivers. 0 will hide 8430 cursors, 1 will display them. 8431 8432 vt.italic= [VT] Default color for italic text; 0-15. 8433 Default: 2 = green. 8434 8435 vt.underline= [VT] Default color for underlined text; 0-15. 8436 Default: 3 = cyan. 8437 8438 watchdog timers [HW,WDT] For information on watchdog timers, 8439 see Documentation/watchdog/watchdog-parameters.rst 8440 or other driver-specific files in the 8441 Documentation/watchdog/ directory. 8442 8443 watchdog_thresh= 8444 [KNL] 8445 Set the hard lockup detector stall duration 8446 threshold in seconds. The soft lockup detector 8447 threshold is set to twice the value. A value of 0 8448 disables both lockup detectors. Default is 10 8449 seconds. 8450 8451 workqueue.unbound_cpus= 8452 [KNL,SMP] Specify to constrain one or some CPUs 8453 to use in unbound workqueues. 8454 Format: <cpu-list> 8455 By default, all online CPUs are available for 8456 unbound workqueues. 8457 8458 workqueue.watchdog_thresh= 8459 If CONFIG_WQ_WATCHDOG is configured, workqueue can 8460 warn stall conditions and dump internal state to 8461 help debugging. 0 disables workqueue stall 8462 detection; otherwise, it's the stall threshold 8463 duration in seconds. The default value is 30 and 8464 it can be updated at runtime by writing to the 8465 corresponding sysfs file. 8466 8467 workqueue.panic_on_stall=<uint> 8468 Panic when workqueue stall is detected by 8469 CONFIG_WQ_WATCHDOG. It sets the number times of the 8470 stall to trigger panic. 8471 8472 The default is set by CONFIG_BOOTPARAM_WQ_STALL_PANIC, 8473 which is 0 (disabled) if not configured. 8474 8475 workqueue.panic_on_stall_time=<uint> 8476 Panic when a workqueue stall has been continuous for 8477 the specified number of seconds. Unlike panic_on_stall 8478 which counts accumulated stall events, this triggers 8479 based on the duration of a single continuous stall. 8480 8481 The default is 0, which disables the time-based panic. 8482 8483 workqueue.cpu_intensive_thresh_us= 8484 Per-cpu work items which run for longer than this 8485 threshold are automatically considered CPU intensive 8486 and excluded from concurrency management to prevent 8487 them from noticeably delaying other per-cpu work 8488 items. Default is 10000 (10ms). 8489 8490 If CONFIG_WQ_CPU_INTENSIVE_REPORT is set, the kernel 8491 will report the work functions which violate this 8492 threshold repeatedly. They are likely good 8493 candidates for using WQ_UNBOUND workqueues instead. 8494 8495 workqueue.cpu_intensive_warning_thresh=<uint> 8496 If CONFIG_WQ_CPU_INTENSIVE_REPORT is set, the kernel 8497 will report the work functions which violate the 8498 intensive_threshold_us repeatedly. In order to prevent 8499 spurious warnings, start printing only after a work 8500 function has violated this threshold number of times. 8501 8502 The default is 4 times. 0 disables the warning. 8503 8504 workqueue.power_efficient 8505 Per-cpu workqueues are generally preferred because 8506 they show better performance thanks to cache 8507 locality; unfortunately, per-cpu workqueues tend to 8508 be more power hungry than unbound workqueues. 8509 8510 Enabling this makes the per-cpu workqueues which 8511 were observed to contribute significantly to power 8512 consumption unbound, leading to measurably lower 8513 power usage at the cost of small performance 8514 overhead. 8515 8516 The default value of this parameter is determined by 8517 the config option CONFIG_WQ_POWER_EFFICIENT_DEFAULT. 8518 8519 workqueue.default_affinity_scope= 8520 Select the default affinity scope to use for unbound 8521 workqueues. Can be one of "cpu", "smt", "cache", 8522 "cache_shard", "numa" and "system". Default is 8523 "cache_shard". For more 8524 information, see the Affinity Scopes section in 8525 Documentation/core-api/workqueue.rst. 8526 8527 This can be changed after boot by writing to the 8528 matching /sys/module/workqueue/parameters file. All 8529 workqueues with the "default" affinity scope will be 8530 updated accordingly. 8531 8532 workqueue.debug_force_rr_cpu 8533 Workqueue used to implicitly guarantee that work 8534 items queued without explicit CPU specified are put 8535 on the local CPU. This guarantee is no longer true 8536 and while local CPU is still preferred work items 8537 may be put on foreign CPUs. This debug option 8538 forces round-robin CPU selection to flush out 8539 usages which depend on the now broken guarantee. 8540 When enabled, memory and cache locality will be 8541 impacted. 8542 8543 writecombine= [LOONGARCH,EARLY] Control the MAT (Memory Access 8544 Type) of ioremap_wc(). 8545 8546 on - Enable writecombine, use WUC for ioremap_wc() 8547 off - Disable writecombine, use SUC for ioremap_wc() 8548 8549 x2apic_phys [X86-64,APIC,EARLY] Use x2apic physical mode instead of 8550 default x2apic cluster mode on platforms 8551 supporting x2apic. 8552 8553 xen_512gb_limit [KNL,X86-64,XEN] 8554 Restricts the kernel running paravirtualized under Xen 8555 to use only up to 512 GB of RAM. The reason to do so is 8556 crash analysis tools and Xen tools for doing domain 8557 save/restore/migration must be enabled to handle larger 8558 domains. 8559 8560 xen_console_io [XEN,EARLY] 8561 Boolean option to enable/disable the usage of the Xen 8562 console_io hypercalls to read and write to the console. 8563 Mostly useful for debugging and development. 8564 8565 xen_emul_unplug= [HW,X86,XEN,EARLY] 8566 Unplug Xen emulated devices 8567 Format: [unplug0,][unplug1] 8568 ide-disks -- unplug primary master IDE devices 8569 aux-ide-disks -- unplug non-primary-master IDE devices 8570 nics -- unplug network devices 8571 all -- unplug all emulated devices (NICs and IDE disks) 8572 unnecessary -- unplugging emulated devices is 8573 unnecessary even if the host did not respond to 8574 the unplug protocol 8575 never -- do not unplug even if version check succeeds 8576 8577 xen_legacy_crash [X86,XEN,EARLY] 8578 Crash from Xen panic notifier, without executing late 8579 panic() code such as dumping handler. 8580 8581 xen_mc_debug [X86,XEN,EARLY] 8582 Enable multicall debugging when running as a Xen PV guest. 8583 Enabling this feature will reduce performance a little 8584 bit, so it should only be enabled for obtaining extended 8585 debug data in case of multicall errors. 8586 8587 xen_msr_safe= [X86,XEN,EARLY] 8588 Format: <bool> 8589 Select whether to always use non-faulting (safe) MSR 8590 access functions when running as Xen PV guest. The 8591 default value is controlled by CONFIG_XEN_PV_MSR_SAFE. 8592 8593 xen_nopv [X86] 8594 Disables the PV optimizations forcing the HVM guest to 8595 run as generic HVM guest with no PV drivers. 8596 This option is obsoleted by the "nopv" option, which 8597 has equivalent effect for XEN platform. 8598 8599 xen_no_vector_callback 8600 [KNL,X86,XEN,EARLY] Disable the vector callback for Xen 8601 event channel interrupts. 8602 8603 xen_scrub_pages= [XEN] 8604 Boolean option to control scrubbing pages before giving them back 8605 to Xen, for use by other domains. Can be also changed at runtime 8606 with /sys/devices/system/xen_memory/xen_memory0/scrub_pages. 8607 Default value controlled with CONFIG_XEN_SCRUB_PAGES_DEFAULT. 8608 8609 xen_timer_slop= [X86-64,XEN,EARLY] 8610 Set the timer slop (in nanoseconds) for the virtual Xen 8611 timers (default is 100000). This adjusts the minimum 8612 delta of virtualized Xen timers, where lower values 8613 improve timer resolution at the expense of processing 8614 more timer interrupts. 8615 8616 xen.balloon_boot_timeout= [XEN] 8617 The time (in seconds) to wait before giving up to boot 8618 in case initial ballooning fails to free enough memory. 8619 Applies only when running as HVM or PVH guest and 8620 started with less memory configured than allowed at 8621 max. Default is 180. 8622 8623 xen.event_eoi_delay= [XEN] 8624 How long to delay EOI handling in case of event 8625 storms (jiffies). Default is 10. 8626 8627 xen.event_loop_timeout= [XEN] 8628 After which time (jiffies) the event handling loop 8629 should start to delay EOI handling. Default is 2. 8630 8631 xen.fifo_events= [XEN] 8632 Boolean parameter to disable using fifo event handling 8633 even if available. Normally fifo event handling is 8634 preferred over the 2-level event handling, as it is 8635 fairer and the number of possible event channels is 8636 much higher. Default is on (use fifo events). 8637 8638 xirc2ps_cs= [NET,PCMCIA] 8639 Format: 8640 <irq>,<irq_mask>,<io>,<full_duplex>,<do_sound>,<lockup_hack>[,<irq2>[,<irq3>[,<irq4>]]] 8641 8642 xive= [PPC] 8643 By default on POWER9 and above, the kernel will 8644 natively use the XIVE interrupt controller. This option 8645 allows the fallback firmware mode to be used: 8646 8647 off Fallback to firmware control of XIVE interrupt 8648 controller on both pseries and powernv 8649 platforms. Only useful on POWER9 and above. 8650 8651 xive.store-eoi=off [PPC] 8652 By default on POWER10 and above, the kernel will use 8653 stores for EOI handling when the XIVE interrupt mode 8654 is active. This option allows the XIVE driver to use 8655 loads instead, as on POWER9. 8656 8657 xhci-hcd.quirks [USB,KNL] 8658 A hex value specifying bitmask with supplemental xhci 8659 host controller quirks. Meaning of each bit can be 8660 consulted in header drivers/usb/host/xhci.h. 8661 8662 xmon [PPC,EARLY] 8663 Format: { early | on | rw | ro | off } 8664 Controls if xmon debugger is enabled. Default is off. 8665 Passing only "xmon" is equivalent to "xmon=early". 8666 early Call xmon as early as possible on boot; xmon 8667 debugger is called from setup_arch(). 8668 on xmon debugger hooks will be installed so xmon 8669 is only called on a kernel crash. Default mode, 8670 i.e. either "ro" or "rw" mode, is controlled 8671 with CONFIG_XMON_DEFAULT_RO_MODE. 8672 rw xmon debugger hooks will be installed so xmon 8673 is called only on a kernel crash, mode is write, 8674 meaning SPR registers, memory and, other data 8675 can be written using xmon commands. 8676 ro same as "rw" option above but SPR registers, 8677 memory, and other data can't be written using 8678 xmon commands. 8679 off xmon is disabled. 8680