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 DRM Direct Rendering Management support is enabled. 12 DYNAMIC_DEBUG Build in debug messages and enable them at runtime 13 EARLY Parameter processed too early to be embedded in initrd. 14 EDD BIOS Enhanced Disk Drive Services (EDD) is enabled 15 EFI EFI Partitioning (GPT) is enabled 16 EVM Extended Verification Module 17 FB The frame buffer device is enabled. 18 FTRACE Function tracing enabled. 19 GCOV GCOV profiling is enabled. 20 HIBERNATION HIBERNATION is enabled. 21 HW Appropriate hardware is enabled. 22 HYPER_V HYPERV support is enabled. 23 IMA Integrity measurement architecture is enabled. 24 IP_PNP IP DHCP, BOOTP, or RARP is enabled. 25 IPV6 IPv6 support is enabled. 26 ISAPNP ISA PnP code is enabled. 27 ISOL CPU Isolation is enabled. 28 JOY Appropriate joystick support is enabled. 29 KGDB Kernel debugger support is enabled. 30 KVM Kernel Virtual Machine support is enabled. 31 LIBATA Libata driver is enabled 32 LOONGARCH LoongArch architecture is enabled. 33 LOOP Loopback device support is enabled. 34 LP Printer support is enabled. 35 M68k M68k architecture is enabled. 36 These options have more detailed description inside of 37 Documentation/arch/m68k/kernel-options.rst. 38 MIPS MIPS architecture is enabled. 39 MOUSE Appropriate mouse support is enabled. 40 MSI Message Signaled Interrupts (PCI). 41 MTD MTD (Memory Technology Device) support is enabled. 42 NET Appropriate network support is enabled. 43 NFS Appropriate NFS support is enabled. 44 NUMA NUMA support is enabled. 45 OF Devicetree is enabled. 46 PARISC The PA-RISC architecture is enabled. 47 PCI PCI bus support is enabled. 48 PCIE PCI Express support is enabled. 49 PCMCIA The PCMCIA subsystem is enabled. 50 PM Power Management support is enabled. 51 PNP Plug & Play support is enabled. 52 PPC PowerPC architecture is enabled. 53 PPT Parallel port support is enabled. 54 PS2 Appropriate PS/2 support is enabled. 55 PV_OPS A paravirtualized kernel is enabled. 56 RAM RAM disk support is enabled. 57 RDT Intel Resource Director Technology. 58 RISCV RISCV architecture is enabled. 59 S390 S390 architecture is enabled. 60 SCSI Appropriate SCSI support is enabled. 61 A lot of drivers have their options described inside 62 the Documentation/scsi/ sub-directory. 63 SDW SoundWire support is enabled. 64 SECURITY Different security models are enabled. 65 SELINUX SELinux support is enabled. 66 SERIAL Serial support is enabled. 67 SH SuperH architecture is enabled. 68 SMP The kernel is an SMP kernel. 69 SPARC Sparc architecture is enabled. 70 SUSPEND System suspend states are enabled. 71 SWSUSP Software suspend (hibernation) is enabled. 72 TPM TPM drivers are enabled. 73 UMS USB Mass Storage support is enabled. 74 USB USB support is enabled. 75 NVME NVMe support is enabled 76 USBHID USB Human Interface Device support is enabled. 77 V4L Video For Linux support is enabled. 78 VGA The VGA console has been enabled. 79 VMMIO Driver for memory mapped virtio devices is enabled. 80 VT Virtual terminal support is enabled. 81 WDT Watchdog support is enabled. 82 X86-32 X86-32, aka i386 architecture is enabled. 83 X86-64 X86-64 architecture is enabled. 84 X86 Either 32-bit or 64-bit x86 (same as X86-32+X86-64) 85 X86_UV SGI UV support is enabled. 86 XEN Xen support is enabled 87 XTENSA xtensa architecture is enabled. 88 89In addition, the following text indicates that the option 90 91 BOOT Is a boot loader parameter. 92 BUGS= Relates to possible processor bugs on the said processor. 93 KNL Is a kernel start-up parameter. 94 95 96Kernel parameters 97 98 accept_memory= [MM] 99 Format: { eager | lazy } 100 default: lazy 101 By default, unaccepted memory is accepted lazily to 102 avoid prolonged boot times. The lazy option will add 103 some runtime overhead until all memory is eventually 104 accepted. In most cases the overhead is negligible. 105 For some workloads or for debugging purposes 106 accept_memory=eager can be used to accept all memory 107 at once during boot. 108 109 acpi= [HW,ACPI,X86,ARM64,RISCV64,EARLY] 110 Advanced Configuration and Power Interface 111 Format: { force | on | off | strict | noirq | rsdt | 112 copy_dsdt | nospcr } 113 force -- enable ACPI if default was off 114 on -- enable ACPI but allow fallback to DT [arm64,riscv64] 115 off -- disable ACPI if default was on 116 noirq -- do not use ACPI for IRQ routing 117 strict -- Be less tolerant of platforms that are not 118 strictly ACPI specification compliant. 119 rsdt -- prefer RSDT over (default) XSDT 120 copy_dsdt -- copy DSDT to memory 121 nocmcff -- Disable firmware first mode for corrected 122 errors. This disables parsing the HEST CMC error 123 source to check if firmware has set the FF flag. This 124 may result in duplicate corrected error reports. 125 nospcr -- disable console in ACPI SPCR table as 126 default _serial_ console on ARM64 127 spcr -- enable console in ACPI SPCR table as 128 default _serial_ console on x86 129 For ARM64, ONLY "acpi=off", "acpi=on", "acpi=force" or 130 "acpi=nospcr" are available 131 For RISCV64, ONLY "acpi=off", "acpi=on" or "acpi=force" 132 are available 133 134 See also Documentation/power/runtime_pm.rst, pci=noacpi 135 136 acpi_apic_instance= [ACPI,IOAPIC,EARLY] 137 Format: <int> 138 2: use 2nd APIC table, if available 139 1,0: use 1st APIC table 140 default: 0 141 142 acpi_backlight= [HW,ACPI] 143 { vendor | video | native | none } 144 If set to vendor, prefer vendor-specific driver 145 (e.g. thinkpad_acpi, sony_acpi, etc.) instead 146 of the ACPI video.ko driver. 147 If set to video, use the ACPI video.ko driver. 148 If set to native, use the device's native backlight mode. 149 If set to none, disable the ACPI backlight interface. 150 151 acpi_force_32bit_fadt_addr [ACPI,EARLY] 152 force FADT to use 32 bit addresses rather than the 153 64 bit X_* addresses. Some firmware have broken 64 154 bit addresses for force ACPI ignore these and use 155 the older legacy 32 bit addresses. 156 157 acpica_no_return_repair [HW, ACPI] 158 Disable AML predefined validation mechanism 159 This mechanism can repair the evaluation result to make 160 the return objects more ACPI specification compliant. 161 This option is useful for developers to identify the 162 root cause of an AML interpreter issue when the issue 163 has something to do with the repair mechanism. 164 165 acpi.debug_layer= [HW,ACPI,ACPI_DEBUG] 166 acpi.debug_level= [HW,ACPI,ACPI_DEBUG] 167 Format: <int> 168 CONFIG_ACPI_DEBUG must be enabled to produce any ACPI 169 debug output. Bits in debug_layer correspond to a 170 _COMPONENT in an ACPI source file, e.g., 171 #define _COMPONENT ACPI_EVENTS 172 Bits in debug_level correspond to a level in 173 ACPI_DEBUG_PRINT statements, e.g., 174 ACPI_DEBUG_PRINT((ACPI_DB_INFO, ... 175 The debug_level mask defaults to "info". See 176 Documentation/firmware-guide/acpi/debug.rst for more information about 177 debug layers and levels. 178 179 Enable processor driver info messages: 180 acpi.debug_layer=0x20000000 181 Enable AML "Debug" output, i.e., stores to the Debug 182 object while interpreting AML: 183 acpi.debug_layer=0xffffffff acpi.debug_level=0x2 184 Enable all messages related to ACPI hardware: 185 acpi.debug_layer=0x2 acpi.debug_level=0xffffffff 186 187 Some values produce so much output that the system is 188 unusable. The "log_buf_len" parameter may be useful 189 if you need to capture more output. 190 191 acpi.poweroff_on_fatal= [ACPI] 192 {0 | 1} 193 Causes the system to poweroff when the ACPI bytecode signals 194 a fatal error. The default value of this setting is 1. 195 Overriding this value should only be done for diagnosing 196 ACPI firmware problems, as the system might behave erratically 197 after having encountered a fatal ACPI error. 198 199 acpi_enforce_resources= [ACPI] 200 { strict | lax | no } 201 Check for resource conflicts between native drivers 202 and ACPI OperationRegions (SystemIO and SystemMemory 203 only). IO ports and memory declared in ACPI might be 204 used by the ACPI subsystem in arbitrary AML code and 205 can interfere with legacy drivers. 206 strict (default): access to resources claimed by ACPI 207 is denied; legacy drivers trying to access reserved 208 resources will fail to bind to device using them. 209 lax: access to resources claimed by ACPI is allowed; 210 legacy drivers trying to access reserved resources 211 will bind successfully but a warning message is logged. 212 no: ACPI OperationRegions are not marked as reserved, 213 no further checks are performed. 214 215 acpi_force_table_verification [HW,ACPI,EARLY] 216 Enable table checksum verification during early stage. 217 By default, this is disabled due to x86 early mapping 218 size limitation. 219 220 acpi_irq_balance [HW,ACPI] 221 ACPI will balance active IRQs 222 default in APIC mode 223 224 acpi_irq_nobalance [HW,ACPI] 225 ACPI will not move active IRQs (default) 226 default in PIC mode 227 228 acpi_irq_isa= [HW,ACPI] If irq_balance, mark listed IRQs used by ISA 229 Format: <irq>,<irq>... 230 231 acpi_irq_pci= [HW,ACPI] If irq_balance, clear listed IRQs for 232 use by PCI 233 Format: <irq>,<irq>... 234 235 acpi_mask_gpe= [HW,ACPI] 236 Due to the existence of _Lxx/_Exx, some GPEs triggered 237 by unsupported hardware/firmware features can result in 238 GPE floodings that cannot be automatically disabled by 239 the GPE dispatcher. 240 This facility can be used to prevent such uncontrolled 241 GPE floodings. 242 Format: <byte> or <bitmap-list> 243 244 acpi_no_auto_serialize [HW,ACPI] 245 Disable auto-serialization of AML methods 246 AML control methods that contain the opcodes to create 247 named objects will be marked as "Serialized" by the 248 auto-serialization feature. 249 This feature is enabled by default. 250 This option allows to turn off the feature. 251 252 acpi_no_memhotplug [ACPI] Disable memory hotplug. Useful for kdump 253 kernels. 254 255 acpi_no_static_ssdt [HW,ACPI,EARLY] 256 Disable installation of static SSDTs at early boot time 257 By default, SSDTs contained in the RSDT/XSDT will be 258 installed automatically and they will appear under 259 /sys/firmware/acpi/tables. 260 This option turns off this feature. 261 Note that specifying this option does not affect 262 dynamic table installation which will install SSDT 263 tables to /sys/firmware/acpi/tables/dynamic. 264 265 acpi_no_watchdog [HW,ACPI,WDT] 266 Ignore the ACPI-based watchdog interface (WDAT) and let 267 a native driver control the watchdog device instead. 268 269 acpi_rsdp= [ACPI,EFI,KEXEC,EARLY] 270 Pass the RSDP address to the kernel, mostly used 271 on machines running EFI runtime service to boot the 272 second kernel for kdump. 273 274 acpi_os_name= [HW,ACPI] Tell ACPI BIOS the name of the OS 275 Format: To spoof as Windows 98: ="Microsoft Windows" 276 277 acpi_rev_override [ACPI] Override the _REV object to return 5 (instead 278 of 2 which is mandated by ACPI 6) as the supported ACPI 279 specification revision (when using this switch, it may 280 be necessary to carry out a cold reboot _twice_ in a 281 row to make it take effect on the platform firmware). 282 283 acpi_osi= [HW,ACPI] Modify list of supported OS interface strings 284 acpi_osi="string1" # add string1 285 acpi_osi="!string2" # remove string2 286 acpi_osi=!* # remove all strings 287 acpi_osi=! # disable all built-in OS vendor 288 strings 289 acpi_osi=!! # enable all built-in OS vendor 290 strings 291 acpi_osi= # disable all strings 292 293 'acpi_osi=!' can be used in combination with single or 294 multiple 'acpi_osi="string1"' to support specific OS 295 vendor string(s). Note that such command can only 296 affect the default state of the OS vendor strings, thus 297 it cannot affect the default state of the feature group 298 strings and the current state of the OS vendor strings, 299 specifying it multiple times through kernel command line 300 is meaningless. This command is useful when one do not 301 care about the state of the feature group strings which 302 should be controlled by the OSPM. 303 Examples: 304 1. 'acpi_osi=! acpi_osi="Windows 2000"' is equivalent 305 to 'acpi_osi="Windows 2000" acpi_osi=!', they all 306 can make '_OSI("Windows 2000")' TRUE. 307 308 'acpi_osi=' cannot be used in combination with other 309 'acpi_osi=' command lines, the _OSI method will not 310 exist in the ACPI namespace. NOTE that such command can 311 only affect the _OSI support state, thus specifying it 312 multiple times through kernel command line is also 313 meaningless. 314 Examples: 315 1. 'acpi_osi=' can make 'CondRefOf(_OSI, Local1)' 316 FALSE. 317 318 'acpi_osi=!*' can be used in combination with single or 319 multiple 'acpi_osi="string1"' to support specific 320 string(s). Note that such command can affect the 321 current state of both the OS vendor strings and the 322 feature group strings, thus specifying it multiple times 323 through kernel command line is meaningful. But it may 324 still not able to affect the final state of a string if 325 there are quirks related to this string. This command 326 is useful when one want to control the state of the 327 feature group strings to debug BIOS issues related to 328 the OSPM features. 329 Examples: 330 1. 'acpi_osi="Module Device" acpi_osi=!*' can make 331 '_OSI("Module Device")' FALSE. 332 2. 'acpi_osi=!* acpi_osi="Module Device"' can make 333 '_OSI("Module Device")' TRUE. 334 3. 'acpi_osi=! acpi_osi=!* acpi_osi="Windows 2000"' is 335 equivalent to 336 'acpi_osi=!* acpi_osi=! acpi_osi="Windows 2000"' 337 and 338 'acpi_osi=!* acpi_osi="Windows 2000" acpi_osi=!', 339 they all will make '_OSI("Windows 2000")' TRUE. 340 341 acpi_pm_good [X86] 342 Override the pmtimer bug detection: force the kernel 343 to assume that this machine's pmtimer latches its value 344 and always returns good values. 345 346 acpi_sci= [HW,ACPI,EARLY] ACPI System Control Interrupt trigger mode 347 Format: { level | edge | high | low } 348 349 acpi_skip_timer_override [HW,ACPI,EARLY] 350 Recognize and ignore IRQ0/pin2 Interrupt Override. 351 For broken nForce2 BIOS resulting in XT-PIC timer. 352 353 acpi_sleep= [HW,ACPI] Sleep options 354 Format: { s3_bios, s3_mode, s3_beep, s4_hwsig, 355 s4_nohwsig, old_ordering, nonvs, 356 sci_force_enable, nobl } 357 See Documentation/power/video.rst for information on 358 s3_bios and s3_mode. 359 s3_beep is for debugging; it makes the PC's speaker beep 360 as soon as the kernel's real-mode entry point is called. 361 s4_hwsig causes the kernel to check the ACPI hardware 362 signature during resume from hibernation, and gracefully 363 refuse to resume if it has changed. This complies with 364 the ACPI specification but not with reality, since 365 Windows does not do this and many laptops do change it 366 on docking. So the default behaviour is to allow resume 367 and simply warn when the signature changes, unless the 368 s4_hwsig option is enabled. 369 s4_nohwsig prevents ACPI hardware signature from being 370 used (or even warned about) during resume. 371 old_ordering causes the ACPI 1.0 ordering of the _PTS 372 control method, with respect to putting devices into 373 low power states, to be enforced (the ACPI 2.0 ordering 374 of _PTS is used by default). 375 nonvs prevents the kernel from saving/restoring the 376 ACPI NVS memory during suspend/hibernation and resume. 377 sci_force_enable causes the kernel to set SCI_EN directly 378 on resume from S1/S3 (which is against the ACPI spec, 379 but some broken systems don't work without it). 380 nobl causes the internal blacklist of systems known to 381 behave incorrectly in some ways with respect to system 382 suspend and resume to be ignored (use wisely). 383 384 acpi_use_timer_override [HW,ACPI,EARLY] 385 Use timer override. For some broken Nvidia NF5 boards 386 that require a timer override, but don't have HPET 387 388 add_efi_memmap [EFI,X86,EARLY] Include EFI memory map in 389 kernel's map of available physical RAM. 390 391 agp= [AGP] 392 { off | try_unsupported } 393 off: disable AGP support 394 try_unsupported: try to drive unsupported chipsets 395 (may crash computer or cause data corruption) 396 397 ALSA [HW,ALSA] 398 See Documentation/sound/alsa-configuration.rst 399 400 alignment= [KNL,ARM] 401 Allow the default userspace alignment fault handler 402 behaviour to be specified. Bit 0 enables warnings, 403 bit 1 enables fixups, and bit 2 sends a segfault. 404 405 align_va_addr= [X86-64] 406 Align virtual addresses by clearing slice [14:12] when 407 allocating a VMA at process creation time. This option 408 gives you up to 3% performance improvement on AMD F15h 409 machines (where it is enabled by default) for a 410 CPU-intensive style benchmark, and it can vary highly in 411 a microbenchmark depending on workload and compiler. 412 413 32: only for 32-bit processes 414 64: only for 64-bit processes 415 on: enable for both 32- and 64-bit processes 416 off: disable for both 32- and 64-bit processes 417 418 alloc_snapshot [FTRACE] 419 Allocate the ftrace snapshot buffer on boot up when the 420 main buffer is allocated. This is handy if debugging 421 and you need to use tracing_snapshot() on boot up, and 422 do not want to use tracing_snapshot_alloc() as it needs 423 to be done where GFP_KERNEL allocations are allowed. 424 425 allow_mismatched_32bit_el0 [ARM64,EARLY] 426 Allow execve() of 32-bit applications and setting of the 427 PER_LINUX32 personality on systems where only a strict 428 subset of the CPUs support 32-bit EL0. When this 429 parameter is present, the set of CPUs supporting 32-bit 430 EL0 is indicated by /sys/devices/system/cpu/aarch32_el0 431 and hot-unplug operations may be restricted. 432 433 See Documentation/arch/arm64/asymmetric-32bit.rst for more 434 information. 435 436 amd_iommu= [HW,X86-64] 437 Pass parameters to the AMD IOMMU driver in the system. 438 Possible values are: 439 fullflush - Deprecated, equivalent to iommu.strict=1 440 off - do not initialize any AMD IOMMU found in 441 the system 442 force_isolation - Force device isolation for all 443 devices. The IOMMU driver is not 444 allowed anymore to lift isolation 445 requirements as needed. This option 446 does not override iommu=pt 447 force_enable - Force enable the IOMMU on platforms known 448 to be buggy with IOMMU enabled. Use this 449 option with care. 450 pgtbl_v1 - Use v1 page table for DMA-API (Default). 451 pgtbl_v2 - Use v2 page table for DMA-API. 452 irtcachedis - Disable Interrupt Remapping Table (IRT) caching. 453 nohugepages - Limit page-sizes used for v1 page-tables 454 to 4 KiB. 455 v2_pgsizes_only - Limit page-sizes used for v1 page-tables 456 to 4KiB/2Mib/1GiB. 457 458 459 amd_iommu_dump= [HW,X86-64] 460 Enable AMD IOMMU driver option to dump the ACPI table 461 for AMD IOMMU. With this option enabled, AMD IOMMU 462 driver will print ACPI tables for AMD IOMMU during 463 IOMMU initialization. 464 465 amd_iommu_intr= [HW,X86-64] 466 Specifies one of the following AMD IOMMU interrupt 467 remapping modes: 468 legacy - Use legacy interrupt remapping mode. 469 vapic - Use virtual APIC mode, which allows IOMMU 470 to inject interrupts directly into guest. 471 This mode requires kvm-amd.avic=1. 472 (Default when IOMMU HW support is present.) 473 474 amd_pstate= [X86,EARLY] 475 disable 476 Do not enable amd_pstate as the default 477 scaling driver for the supported processors 478 passive 479 Use amd_pstate with passive mode as a scaling driver. 480 In this mode autonomous selection is disabled. 481 Driver requests a desired performance level and platform 482 tries to match the same performance level if it is 483 satisfied by guaranteed performance level. 484 active 485 Use amd_pstate_epp driver instance as the scaling driver, 486 driver provides a hint to the hardware if software wants 487 to bias toward performance (0x0) or energy efficiency (0xff) 488 to the CPPC firmware. then CPPC power algorithm will 489 calculate the runtime workload and adjust the realtime cores 490 frequency. 491 guided 492 Activate guided autonomous mode. Driver requests minimum and 493 maximum performance level and the platform autonomously 494 selects a performance level in this range and appropriate 495 to the current workload. 496 497 amd_prefcore= 498 [X86] 499 disable 500 Disable amd-pstate preferred core. 501 502 amd_dynamic_epp= 503 [X86] 504 disable 505 Disable amd-pstate dynamic EPP. 506 enable 507 Enable amd-pstate dynamic EPP. 508 509 amijoy.map= [HW,JOY] Amiga joystick support 510 Map of devices attached to JOY0DAT and JOY1DAT 511 Format: <a>,<b> 512 See also Documentation/input/joydev/joystick.rst 513 514 analog.map= [HW,JOY] Analog joystick and gamepad support 515 Specifies type or capabilities of an analog joystick 516 connected to one of 16 gameports 517 Format: <type1>,<type2>,..<type16> 518 519 apc= [HW,SPARC] 520 Power management functions (SPARCstation-4/5 + deriv.) 521 Format: noidle 522 Disable APC CPU standby support. SPARCstation-Fox does 523 not play well with APC CPU idle - disable it if you have 524 APC and your system crashes randomly. 525 526 apic [APIC,X86-64] Use IO-APIC. Default. 527 528 apic= [APIC,X86,EARLY] Advanced Programmable Interrupt Controller 529 Change the output verbosity while booting 530 Format: { quiet (default) | verbose | debug } 531 Change the amount of debugging information output 532 when initialising the APIC and IO-APIC components. 533 534 apic_extnmi= [APIC,X86,EARLY] External NMI delivery setting 535 Format: { bsp (default) | all | none } 536 bsp: External NMI is delivered only to CPU 0 537 all: External NMIs are broadcast to all CPUs as a 538 backup of CPU 0 539 none: External NMI is masked for all CPUs. This is 540 useful so that a dump capture kernel won't be 541 shot down by NMI 542 543 apicpmtimer Do APIC timer calibration using the pmtimer. Implies 544 apicmaintimer. Useful when your PIT timer is totally 545 broken. 546 547 autoconf= [IPV6] 548 See Documentation/networking/ipv6.rst. 549 550 apm= [APM] Advanced Power Management 551 See header of arch/x86/kernel/apm_32.c. 552 553 apparmor= [APPARMOR] Disable or enable AppArmor at boot time 554 Format: { "0" | "1" } 555 See security/apparmor/Kconfig help text 556 0 -- disable. 557 1 -- enable. 558 Default value is set via kernel config option. 559 560 arm64.no32bit_el0 [ARM64] Unconditionally disable the execution of 561 32 bit applications. 562 563 arm64.nobti [ARM64] Unconditionally disable Branch Target 564 Identification support 565 566 arm64.nogcs [ARM64] Unconditionally disable Guarded Control Stack 567 support 568 569 arm64.nohaft [ARM64] Unconditionally disable Hardware managed Access 570 Flag for Table descriptors support 571 572 arm64.nomops [ARM64] Unconditionally disable Memory Copy and Memory 573 Set instructions support 574 575 arm64.nompam [ARM64] Unconditionally disable Memory Partitioning And 576 Monitoring support 577 578 arm64.nomte [ARM64] Unconditionally disable Memory Tagging Extension 579 support 580 581 arm64.nopauth [ARM64] Unconditionally disable Pointer Authentication 582 support 583 584 arm64.nosme [ARM64] Unconditionally disable Scalable Matrix 585 Extension support 586 587 arm64.nosve [ARM64] Unconditionally disable Scalable Vector 588 Extension support 589 590 ataflop= [HW,M68k] 591 592 atarimouse= [HW,MOUSE] Atari Mouse 593 594 atkbd.extra= [HW] Enable extra LEDs and keys on IBM RapidAccess, 595 EzKey and similar keyboards 596 597 atkbd.reset= [HW] Reset keyboard during initialization 598 599 atkbd.set= [HW] Select keyboard code set 600 Format: <int> (2 = AT (default), 3 = PS/2) 601 602 atkbd.scroll= [HW] Enable scroll wheel on MS Office and similar 603 keyboards 604 605 atkbd.softraw= [HW] Choose between synthetic and real raw mode 606 Format: <bool> (0 = real, 1 = synthetic (default)) 607 608 atkbd.softrepeat= [HW] 609 Use software keyboard repeat 610 611 audit= [KNL] Enable the audit sub-system 612 Format: { "0" | "1" | "off" | "on" } 613 0 | off - kernel audit is disabled and can not be 614 enabled until the next reboot 615 unset - kernel audit is initialized but disabled and 616 will be fully enabled by the userspace auditd. 617 1 | on - kernel audit is initialized and partially 618 enabled, storing at most audit_backlog_limit 619 messages in RAM until it is fully enabled by the 620 userspace auditd. 621 Default: unset 622 623 audit_backlog_limit= [KNL] Set the audit queue size limit. 624 Format: <int> (must be >=0) 625 Default: 64 626 627 bau= [X86_UV] Enable the BAU on SGI UV. The default 628 behavior is to disable the BAU (i.e. bau=0). 629 Format: { "0" | "1" } 630 0 - Disable the BAU. 631 1 - Enable the BAU. 632 unset - Disable the BAU. 633 634 bdev_allow_write_mounted= 635 Format: <bool> 636 Control the ability to open a mounted block device 637 for writing, i.e., allow / disallow writes that bypass 638 the FS. This was implemented as a means to prevent 639 fuzzers from crashing the kernel by overwriting the 640 metadata underneath a mounted FS without its awareness. 641 This also prevents destructive formatting of mounted 642 filesystems by naive storage tooling that don't use 643 O_EXCL. Default is Y and can be changed through the 644 Kconfig option CONFIG_BLK_DEV_WRITE_MOUNTED. 645 646 bert_disable [ACPI] 647 Disable BERT OS support on buggy BIOSes. 648 649 bgrt_disable [ACPI,X86,EARLY] 650 Disable BGRT to avoid flickering OEM logo. 651 652 blkdevparts= Manual partition parsing of block device(s) for 653 embedded devices based on command line input. 654 See Documentation/block/cmdline-partition.rst 655 656 boot_delay= [KNL,EARLY] 657 Milliseconds to delay each printk during boot. 658 Only works if CONFIG_BOOT_PRINTK_DELAY is enabled, 659 and you may also have to specify "lpj=". Boot_delay 660 values larger than 10 seconds (10000) are assumed 661 erroneous and ignored. 662 Format: integer 663 664 bootconfig [KNL,EARLY] 665 Extended command line options can be added to an initrd 666 and this will cause the kernel to look for it. 667 668 See Documentation/admin-guide/bootconfig.rst 669 670 bttv.card= [HW,V4L] bttv (bt848 + bt878 based grabber cards) 671 bttv.radio= Most important insmod options are available as 672 kernel args too. 673 bttv.pll= See Documentation/admin-guide/media/bttv.rst 674 bttv.tuner= 675 676 bulk_remove=off [PPC] This parameter disables the use of the pSeries 677 firmware feature for flushing multiple hpte entries 678 at a time. 679 680 c101= [NET] Moxa C101 synchronous serial card 681 682 cachesize= [BUGS=X86-32] Override level 2 CPU cache size detection. 683 Sometimes CPU hardware bugs make them report the cache 684 size incorrectly. The kernel will attempt work arounds 685 to fix known problems, but for some CPUs it is not 686 possible to determine what the correct size should be. 687 This option provides an override for these situations. 688 689 carrier_timeout= 690 [NET] Specifies amount of time (in seconds) that 691 the kernel should wait for a network carrier. By default 692 it waits 120 seconds. 693 694 ca_keys= [KEYS] This parameter identifies a specific key(s) on 695 the system trusted keyring to be used for certificate 696 trust validation. 697 format: { id:<keyid> | builtin } 698 699 cca= [MIPS,EARLY] Override the kernel pages' cache coherency 700 algorithm. Accepted values range from 0 to 7 701 inclusive. See arch/mips/include/asm/pgtable-bits.h 702 for platform specific values (SB1, Loongson3 and 703 others). 704 705 ccw_timeout_log [S390] 706 See Documentation/arch/s390/common_io.rst for details. 707 708 cfi= [X86-64] Set Control Flow Integrity checking features 709 when CONFIG_FINEIBT is enabled. 710 Format: feature[,feature...] 711 Default: auto 712 713 auto: Use FineIBT if IBT available, otherwise kCFI. 714 Under FineIBT, enable "paranoid" mode when 715 FRED is not available. 716 off: Turn off CFI checking. 717 kcfi: Use kCFI (disable FineIBT). 718 fineibt: Use FineIBT (even if IBT not available). 719 norand: Do not re-randomize CFI hashes. 720 paranoid: Add caller hash checking under FineIBT. 721 bhi: Enable register poisoning to stop speculation 722 across FineIBT. (Disabled by default.) 723 warn: Do not enforce CFI checking: warn only. 724 debug: Report CFI initialization details. 725 726 cgroup_disable= [KNL] Disable a particular controller or optional feature 727 Format: {name of the controller(s) or feature(s) to disable} 728 The effects of cgroup_disable=foo are: 729 - foo isn't auto-mounted if you mount all cgroups in 730 a single hierarchy 731 - foo isn't visible as an individually mountable 732 subsystem 733 - if foo is an optional feature then the feature is 734 disabled and corresponding cgroup files are not 735 created 736 {Currently only "memory" controller deal with this and 737 cut the overhead, others just disable the usage. So 738 only cgroup_disable=memory is actually worthy} 739 Specifying "pressure" disables per-cgroup pressure 740 stall information accounting feature 741 742 cgroup_no_v1= [KNL] Disable cgroup controllers and named hierarchies in v1 743 Format: { { controller | "all" | "named" } 744 [,{ controller | "all" | "named" }...] } 745 Like cgroup_disable, but only applies to cgroup v1; 746 the blacklisted controllers remain available in cgroup2. 747 "all" blacklists all controllers and "named" disables 748 named mounts. Specifying both "all" and "named" disables 749 all v1 hierarchies. 750 751 cgroup_v1_proc= [KNL] Show also missing controllers in /proc/cgroups 752 Format: { "true" | "false" } 753 /proc/cgroups lists only v1 controllers by default. 754 This compatibility option enables listing also v2 755 controllers (whose v1 code is not compiled!), so that 756 semi-legacy software can check this file to decide 757 about usage of v2 (sic) controllers. 758 759 cgroup_favordynmods= [KNL] Enable or Disable favordynmods. 760 Format: { "true" | "false" } 761 Defaults to the value of CONFIG_CGROUP_FAVOR_DYNMODS. 762 763 cgroup.memory= [KNL] Pass options to the cgroup memory controller. 764 Format: <string> 765 nosocket -- Disable socket memory accounting. 766 nokmem -- Disable kernel memory accounting. 767 nobpf -- Disable BPF memory accounting. 768 769 check_pages= [MM,EARLY] Enable sanity checking of pages after 770 allocations / before freeing. This adds checks to catch 771 double-frees, use-after-frees, and other sources of 772 page corruption by inspecting page internals (flags, 773 mapcount/refcount, memcg_data, etc.). 774 Format: { "0" | "1" } 775 Default: 0 (1 if CONFIG_DEBUG_VM is set) 776 777 checkreqprot= [SELINUX] Set initial checkreqprot flag value. 778 Format: { "0" | "1" } 779 See security/selinux/Kconfig help text. 780 0 -- check protection applied by kernel (includes 781 any implied execute protection). 782 1 -- check protection requested by application. 783 Default value is set via a kernel config option. 784 Value can be changed at runtime via 785 /sys/fs/selinux/checkreqprot. 786 Setting checkreqprot to 1 is deprecated. 787 788 cio_ignore= [S390] 789 See Documentation/arch/s390/common_io.rst for details. 790 791 clk_ignore_unused 792 [CLK] 793 Prevents the clock framework from automatically gating 794 clocks that have not been explicitly enabled by a Linux 795 device driver but are enabled in hardware at reset or 796 by the bootloader/firmware. Note that this does not 797 force such clocks to be always-on nor does it reserve 798 those clocks in any way. This parameter is useful for 799 debug and development, but should not be needed on a 800 platform with proper driver support. For more 801 information, see Documentation/driver-api/clk.rst. 802 803 clock= [BUGS=X86-32, HW] gettimeofday clocksource override. 804 [Deprecated] 805 Forces specified clocksource (if available) to be used 806 when calculating gettimeofday(). If specified 807 clocksource is not available, it defaults to PIT. 808 Format: { pit | tsc | cyclone | pmtmr } 809 810 clocksource= Override the default clocksource 811 Format: <string> 812 Override the default clocksource and use the clocksource 813 with the name specified. 814 Some clocksource names to choose from, depending on 815 the platform: 816 [all] jiffies (this is the base, fallback clocksource) 817 [ACPI] acpi_pm 818 [ARM] imx_timer1,OSTS,netx_timer,mpu_timer2, 819 pxa_timer,timer3,32k_counter,timer0_1 820 [X86-32] pit,hpet,tsc; 821 scx200_hrt on Geode; cyclone on IBM x440 822 [MIPS] MIPS 823 [PARISC] cr16 824 [S390] tod 825 [SH] SuperH 826 [SPARC64] tick 827 [X86-64] hpet,tsc 828 829 clocksource.arm_arch_timer.evtstrm= 830 [ARM,ARM64,EARLY] 831 Format: <bool> 832 Enable/disable the eventstream feature of the ARM 833 architected timer so that code using WFE-based polling 834 loops can be debugged more effectively on production 835 systems. 836 837 clocksource.verify_n_cpus= [KNL] 838 Limit the number of CPUs checked for clocksources 839 marked with CLOCK_SOURCE_VERIFY_PERCPU that 840 are marked unstable due to excessive skew. 841 A negative value says to check all CPUs, while 842 zero says not to check any. Values larger than 843 nr_cpu_ids are silently truncated to nr_cpu_ids. 844 The actual CPUs are chosen randomly, with 845 no replacement if the same CPU is chosen twice. 846 847 clocksource-wdtest.holdoff= [KNL] 848 Set the time in seconds that the clocksource 849 watchdog test waits before commencing its tests. 850 Defaults to zero when built as a module and to 851 10 seconds when built into the kernel. 852 853 cma=nn[MG]@[start[MG][-end[MG]]] 854 [KNL,CMA,EARLY] 855 Sets the size of kernel global memory area for 856 contiguous memory allocations and optionally the 857 placement constraint by the physical address range of 858 memory allocations. A value of 0 disables CMA 859 altogether. For more information, see 860 kernel/dma/contiguous.c 861 862 cma_pernuma=nn[MG] 863 [KNL,CMA,EARLY] 864 Sets the size of kernel per-numa memory area for 865 contiguous memory allocations. A value of 0 disables 866 per-numa CMA altogether. And If this option is not 867 specified, the default value is 0. 868 With per-numa CMA enabled, DMA users on node nid will 869 first try to allocate buffer from the pernuma area 870 which is located in node nid, if the allocation fails, 871 they will fallback to the global default memory area. 872 873 numa_cma=<node>:nn[MG][,<node>:nn[MG]] 874 [KNL,CMA,EARLY] 875 Sets the size of kernel numa memory area for 876 contiguous memory allocations. It will reserve CMA 877 area for the specified node. 878 879 If it is setup together with upper 'cmd_pernuma=' 880 (unlikely), its size setting takes priority for the 881 specified numa nodes. 882 883 With numa CMA enabled, DMA users on node nid will 884 first try to allocate buffer from the numa area 885 which is located in node nid, if the allocation fails, 886 they will fallback to the global default memory area. 887 888 cmo_free_hint= [PPC] Format: { yes | no } 889 Specify whether pages are marked as being inactive 890 when they are freed. This is used in CMO environments 891 to determine OS memory pressure for page stealing by 892 a hypervisor. 893 Default: yes 894 895 coherent_pool=nn[KMG] [ARM,KNL,EARLY] 896 Sets the size of memory pool for coherent, atomic dma 897 allocations, by default set to 256K. 898 899 condev= [HW,S390] console device 900 conmode= 901 902 con3215_drop= [S390,EARLY] 3215 console drop mode. 903 Format: y|n|Y|N|1|0 904 When set to true, drop data on the 3215 console when 905 the console buffer is full. In this case the 906 operator using a 3270 terminal emulator (for example 907 x3270) does not have to enter the clear key for the 908 console output to advance and the kernel to continue. 909 This leads to a much faster boot time when a 3270 910 terminal emulator is active. If no 3270 terminal 911 emulator is used, this parameter has no effect. 912 913 console= [KNL] Output console device and options. 914 915 tty<n> Use the virtual console device <n>. 916 917 ttyS<n>[,options] 918 ttyUSB0[,options] 919 Use the specified serial port. The options are of 920 the form "bbbbpnf", where "bbbb" is the baud rate, 921 "p" is parity ("n", "o", or "e"), "n" is number of 922 bits, and "f" is flow control ("r" for RTS or 923 omit it). Default is "9600n8". 924 925 See Documentation/admin-guide/serial-console.rst for more 926 information. See 927 Documentation/networking/netconsole.rst for an 928 alternative. 929 930 <DEVNAME>:<n>.<n>[,options] 931 Use the specified serial port on the serial core bus. 932 The addressing uses DEVNAME of the physical serial port 933 device, followed by the serial core controller instance, 934 and the serial port instance. The options are the same 935 as documented for the ttyS addressing above. 936 937 The mapping of the serial ports to the tty instances 938 can be viewed with: 939 940 $ ls -d /sys/bus/serial-base/devices/*:*.*/tty/* 941 /sys/bus/serial-base/devices/00:04:0.0/tty/ttyS0 942 943 In the above example, the console can be addressed with 944 console=00:04:0.0. Note that a console addressed this 945 way will only get added when the related device driver 946 is ready. The use of an earlycon parameter in addition to 947 the console may be desired for console output early on. 948 949 uart[8250],io,<addr>[,options] 950 uart[8250],mmio,<addr>[,options] 951 uart[8250],mmio16,<addr>[,options] 952 uart[8250],mmio32,<addr>[,options] 953 uart[8250],0x<addr>[,options] 954 Start an early, polled-mode console on the 8250/16550 955 UART at the specified I/O port or MMIO address, 956 switching to the matching ttyS device later. 957 MMIO inter-register address stride is either 8-bit 958 (mmio), 16-bit (mmio16), or 32-bit (mmio32). 959 If none of [io|mmio|mmio16|mmio32], <addr> is assumed 960 to be equivalent to 'mmio'. 'options' are specified in 961 the same format described for ttyS above; if unspecified, 962 the h/w is not re-initialized. 963 964 hvc<n> Use the hypervisor console device <n>. This is for 965 both Xen and PowerPC hypervisors. 966 967 { null | "" } 968 Use to disable console output, i.e., to have kernel 969 console messages discarded. 970 This must be the only console= parameter used on the 971 kernel command line. 972 973 If the device connected to the port is not a TTY but a braille 974 device, prepend "brl," before the device type, for instance 975 console=brl,ttyS0 976 For now, only VisioBraille is supported. 977 978 console_msg_format= 979 [KNL] Change console messages format 980 default 981 By default we print messages on consoles in 982 "[time stamp] text\n" format (time stamp may not be 983 printed, depending on CONFIG_PRINTK_TIME or 984 `printk_time' param). 985 syslog 986 Switch to syslog format: "<%u>[time stamp] text\n" 987 IOW, each message will have a facility and loglevel 988 prefix. The format is similar to one used by syslog() 989 syscall, or to executing "dmesg -S --raw" or to reading 990 from /proc/kmsg. 991 992 consoleblank= [KNL] The console blank (screen saver) timeout in 993 seconds. A value of 0 disables the blank timer. 994 Defaults to 0. 995 996 coredump_filter= 997 [KNL] Change the default value for 998 /proc/<pid>/coredump_filter. 999 See also Documentation/filesystems/proc.rst. 1000 1001 coresight_cpu_debug.enable 1002 [ARM,ARM64] 1003 Format: <bool> 1004 Enable/disable the CPU sampling based debugging. 1005 0: default value, disable debugging 1006 1: enable debugging at boot time 1007 1008 cpcihp_generic= [HW,PCI] Generic port I/O CompactPCI driver 1009 Format: 1010 <first_slot>,<last_slot>,<port>,<enum_bit>[,<debug>] 1011 1012 cpuidle.off=1 [CPU_IDLE] 1013 disable the cpuidle sub-system 1014 1015 cpuidle.governor= 1016 [CPU_IDLE] Name of the cpuidle governor to use. 1017 1018 cpufreq.off=1 [CPU_FREQ] 1019 disable the cpufreq sub-system 1020 1021 cpufreq.default_governor= 1022 [CPU_FREQ] Name of the default cpufreq governor or 1023 policy to use. This governor must be registered in the 1024 kernel before the cpufreq driver probes. 1025 1026 cpu_init_udelay=N 1027 [X86,EARLY] Delay for N microsec between assert and de-assert 1028 of APIC INIT to start processors. This delay occurs 1029 on every CPU online, such as boot, and resume from suspend. 1030 Default: 10000 1031 1032 cpuhp.parallel= 1033 [SMP] Enable/disable parallel bringup of secondary CPUs 1034 Format: <bool> 1035 Default is enabled if CONFIG_HOTPLUG_PARALLEL=y. Otherwise 1036 the parameter has no effect. 1037 1038 crash_kexec_post_notifiers 1039 Only jump to kdump kernel after running the panic 1040 notifiers and dumping kmsg. This option increases 1041 the risks of a kdump failure, since some panic 1042 notifiers can make the crashed kernel more unstable. 1043 In configurations where kdump may not be reliable, 1044 running the panic notifiers could allow collecting 1045 more data on dmesg, like stack traces from other CPUS 1046 or extra data dumped by panic_print. Note that some 1047 configurations enable this option unconditionally, 1048 like Hyper-V, PowerPC (fadump) and AMD SEV-SNP. 1049 1050 crashkernel=size[KMG][@offset[KMG]] 1051 [KNL,EARLY] Using kexec, Linux can switch to a 'crash kernel' 1052 upon panic. This parameter reserves the physical 1053 memory region [offset, offset + size] for that kernel 1054 image. If '@offset' is omitted, then a suitable offset 1055 is selected automatically. 1056 [KNL, X86-64, ARM64, RISCV, LoongArch] Select a region 1057 under 4G first, and fall back to reserve region above 1058 4G when '@offset' hasn't been specified. 1059 See Documentation/admin-guide/kdump/kdump.rst for further details. 1060 1061 crashkernel=range1:size1[,range2:size2,...][@offset] 1062 [KNL] Same as above, but depends on the memory 1063 in the running system. The syntax of range is 1064 start-[end] where start and end are both 1065 a memory unit (amount[KMG]). See also 1066 Documentation/admin-guide/kdump/kdump.rst for an example. 1067 1068 crashkernel=size[KMG],high 1069 [KNL, X86-64, ARM64, RISCV, LoongArch] range could be 1070 above 4G. 1071 Allow kernel to allocate physical memory region from top, 1072 so could be above 4G if system have more than 4G ram 1073 installed. Otherwise memory region will be allocated 1074 below 4G, if available. 1075 It will be ignored if crashkernel=X is specified. 1076 crashkernel=size[KMG],low 1077 [KNL, X86-64, ARM64, RISCV, LoongArch] range under 4G. 1078 When crashkernel=X,high is passed, kernel could allocate 1079 physical memory region above 4G, that cause second kernel 1080 crash on system that require some amount of low memory, 1081 e.g. swiotlb requires at least 64M+32K low memory, also 1082 enough extra low memory is needed to make sure DMA buffers 1083 for 32-bit devices won't run out. Kernel would try to allocate 1084 default size of memory below 4G automatically. The default 1085 size is platform dependent. 1086 --> x86: max(swiotlb_size_or_default() + 8MiB, 256MiB) 1087 --> arm64: 128MiB 1088 --> riscv: 128MiB 1089 --> loongarch: 128MiB 1090 This one lets the user specify own low range under 4G 1091 for second kernel instead. 1092 0: to disable low allocation. 1093 It will be ignored when crashkernel=X,high is not used 1094 or memory reserved is below 4G. 1095 crashkernel=size[KMG],cma 1096 [KNL, X86, ARM64, RISCV, PPC] Reserve additional crash 1097 kernel memory from CMA. This reservation is usable by 1098 the first system's userspace memory and kernel movable 1099 allocations (memory balloon, zswap). Pages allocated 1100 from this memory range will not be included in the vmcore 1101 so this should not be used if dumping of userspace memory 1102 is intended and it has to be expected that some movable 1103 kernel pages may be missing from the dump. 1104 1105 A standard crashkernel reservation, as described above, 1106 is still needed to hold the crash kernel and initrd. 1107 1108 This option increases the risk of a kdump failure: DMA 1109 transfers configured by the first kernel may end up 1110 corrupting the second kernel's memory. 1111 1112 This reservation method is intended for systems that 1113 can't afford to sacrifice enough memory for standard 1114 crashkernel reservation and where less reliable and 1115 possibly incomplete kdump is preferable to no kdump at 1116 all. 1117 1118 cryptomgr.notests 1119 [KNL] Disable crypto self-tests 1120 1121 cs89x0_dma= [HW,NET] 1122 Format: <dma> 1123 1124 cs89x0_media= [HW,NET] 1125 Format: { rj45 | aui | bnc } 1126 1127 csdlock_debug= [KNL] Enable or disable debug add-ons of cross-CPU 1128 function call handling. When switched on, 1129 additional debug data is printed to the console 1130 in case a hanging CPU is detected, and that 1131 CPU is pinged again in order to try to resolve 1132 the hang situation. The default value of this 1133 option depends on the CSD_LOCK_WAIT_DEBUG_DEFAULT 1134 Kconfig option. 1135 1136 dasd= [HW,NET] 1137 See header of drivers/s390/block/dasd_devmap.c. 1138 1139 db9.dev[2|3]= [HW,JOY] Multisystem joystick support via parallel port 1140 (one device per port) 1141 Format: <port#>,<type> 1142 See also Documentation/input/devices/joystick-parport.rst 1143 1144 debug [KNL,EARLY] Enable kernel debugging (events log level). 1145 1146 debug_boot_weak_hash 1147 [KNL,EARLY] Enable printing [hashed] pointers early in the 1148 boot sequence. If enabled, we use a weak hash instead 1149 of siphash to hash pointers. Use this option if you are 1150 seeing instances of '(___ptrval___)') and need to see a 1151 value (hashed pointer) instead. Cryptographically 1152 insecure, please do not use on production kernels. 1153 1154 debug_locks_verbose= 1155 [KNL] verbose locking self-tests 1156 Format: <int> 1157 Print debugging info while doing the locking API 1158 self-tests. 1159 Bitmask for the various LOCKTYPE_ tests. Defaults to 0 1160 (no extra messages), setting it to -1 (all bits set) 1161 will print _a_lot_ more information - normally only 1162 useful to lockdep developers. 1163 1164 debug_objects [KNL,EARLY] Enable object debugging 1165 1166 debug_guardpage_minorder= 1167 [KNL,EARLY] When CONFIG_DEBUG_PAGEALLOC is set, this 1168 parameter allows control of the order of pages that will 1169 be intentionally kept free (and hence protected) by the 1170 buddy allocator. Bigger value increase the probability 1171 of catching random memory corruption, but reduce the 1172 amount of memory for normal system use. The maximum 1173 possible value is MAX_PAGE_ORDER/2. Setting this 1174 parameter to 1 or 2 should be enough to identify most 1175 random memory corruption problems caused by bugs in 1176 kernel or driver code when a CPU writes to (or reads 1177 from) a random memory location. Note that there exists 1178 a class of memory corruptions problems caused by buggy 1179 H/W or F/W or by drivers badly programming DMA 1180 (basically when memory is written at bus level and the 1181 CPU MMU is bypassed) which are not detectable by 1182 CONFIG_DEBUG_PAGEALLOC, hence this option will not 1183 help tracking down these problems. 1184 1185 debug_pagealloc= 1186 [KNL,EARLY] When CONFIG_DEBUG_PAGEALLOC is set, this parameter 1187 enables the feature at boot time. By default, it is 1188 disabled and the system will work mostly the same as a 1189 kernel built without CONFIG_DEBUG_PAGEALLOC. 1190 Note: to get most of debug_pagealloc error reports, it's 1191 useful to also enable the page_owner functionality. 1192 on: enable the feature 1193 1194 debugfs= [KNL,EARLY] This parameter enables what is exposed to 1195 userspace and debugfs internal clients. 1196 Format: { on, off } 1197 on: All functions are enabled. 1198 off: Filesystem is not registered and clients 1199 get a -EPERM as result when trying to register files 1200 or directories within debugfs. 1201 This is equivalent of the runtime functionality if 1202 debugfs was not enabled in the kernel at all. 1203 Default value is set in build-time with a kernel configuration. 1204 1205 debugpat [X86] Enable PAT debugging 1206 1207 default_hugepagesz= 1208 [HW] The size of the default HugeTLB page. This is 1209 the size represented by the legacy /proc/ hugepages 1210 APIs. In addition, this is the default hugetlb size 1211 used for shmget(), mmap() and mounting hugetlbfs 1212 filesystems. If not specified, defaults to the 1213 architecture's default huge page size. Huge page 1214 sizes are architecture dependent. See also 1215 Documentation/admin-guide/mm/hugetlbpage.rst. 1216 Format: size[KMG] 1217 1218 deferred_probe_timeout= 1219 [KNL] Debugging option to set a timeout in seconds for 1220 deferred probe to give up waiting on dependencies to 1221 probe. Only specific dependencies (subsystems or 1222 drivers) that have opted in will be ignored. A timeout 1223 of 0 will timeout at the end of initcalls; a negative 1224 value is treated as an infinite timeout value. If the 1225 timeout hasn't expired, it'll be restarted by each 1226 successful driver registration. This option will also 1227 dump out devices still on the deferred probe list after 1228 retrying. 1229 1230 delayacct [KNL] Enable per-task delay accounting 1231 1232 dell_smm_hwmon.ignore_dmi= 1233 [HW] Continue probing hardware even if DMI data 1234 indicates that the driver is running on unsupported 1235 hardware. 1236 1237 dell_smm_hwmon.force= 1238 [HW] Activate driver even if SMM BIOS signature does 1239 not match list of supported models and enable otherwise 1240 blacklisted features. 1241 1242 dell_smm_hwmon.power_status= 1243 [HW] Report power status in /proc/i8k 1244 (disabled by default). 1245 1246 dell_smm_hwmon.restricted= 1247 [HW] Allow controlling fans only if SYS_ADMIN 1248 capability is set. 1249 1250 dell_smm_hwmon.fan_mult= 1251 [HW] Factor to multiply fan speed with. 1252 1253 dell_smm_hwmon.fan_max= 1254 [HW] Maximum configurable fan speed. 1255 1256 dfltcc= [HW,S390] 1257 Format: { on | off | def_only | inf_only | always } 1258 on: s390 zlib hardware support for compression on 1259 level 1 and decompression (default) 1260 off: No s390 zlib hardware support 1261 def_only: s390 zlib hardware support for deflate 1262 only (compression on level 1) 1263 inf_only: s390 zlib hardware support for inflate 1264 only (decompression) 1265 always: Same as 'on' but ignores the selected compression 1266 level always using hardware support (used for debugging) 1267 1268 dhash_entries= [KNL] 1269 Set number of hash buckets for dentry cache. 1270 1271 disable_1tb_segments [PPC,EARLY] 1272 Disables the use of 1TB hash page table segments. This 1273 causes the kernel to fall back to 256MB segments which 1274 can be useful when debugging issues that require an SLB 1275 miss to occur. 1276 1277 disable= [IPV6] 1278 See Documentation/networking/ipv6.rst. 1279 1280 disable_radix [PPC,EARLY] 1281 Disable RADIX MMU mode on POWER9 1282 1283 disable_tlbie [PPC] 1284 Disable TLBIE instruction. Currently does not work 1285 with KVM, with HASH MMU, or with coherent accelerators. 1286 1287 disable_ddw [PPC/PSERIES,EARLY] 1288 Disable Dynamic DMA Window support. Use this 1289 to workaround buggy firmware. 1290 1291 disable_ipv6= [IPV6] 1292 See Documentation/networking/ipv6.rst. 1293 1294 disable_mtrr_cleanup [X86,EARLY] 1295 The kernel tries to adjust MTRR layout from continuous 1296 to discrete, to make X server driver able to add WB 1297 entry later. This parameter disables that. 1298 1299 disable_mtrr_trim [X86, Intel and AMD only,EARLY] 1300 By default the kernel will trim any uncacheable 1301 memory out of your available memory pool based on 1302 MTRR settings. This parameter disables that behavior, 1303 possibly causing your machine to run very slowly. 1304 1305 disable_timer_pin_1 [X86,EARLY] 1306 Disable PIN 1 of APIC timer 1307 Can be useful to work around chipset bugs. 1308 1309 dis_ucode_ldr [X86] Disable the microcode loader. 1310 1311 dma_debug=off If the kernel is compiled with DMA_API_DEBUG support, 1312 this option disables the debugging code at boot. 1313 1314 dma_debug_entries=<number> 1315 This option allows to tune the number of preallocated 1316 entries for DMA-API debugging code. One entry is 1317 required per DMA-API allocation. Use this if the 1318 DMA-API debugging code disables itself because the 1319 architectural default is too low. 1320 1321 dma_debug_driver=<driver_name> 1322 With this option the DMA-API debugging driver 1323 filter feature can be enabled at boot time. Just 1324 pass the driver to filter for as the parameter. 1325 The filter can be disabled or changed to another 1326 driver later using sysfs. 1327 1328 reg_file_data_sampling= 1329 [X86] Controls mitigation for Register File Data 1330 Sampling (RFDS) vulnerability. RFDS is a CPU 1331 vulnerability which may allow userspace to infer 1332 kernel data values previously stored in floating point 1333 registers, vector registers, or integer registers. 1334 RFDS only affects Intel Atom processors. 1335 1336 on: Turns ON the mitigation. 1337 off: Turns OFF the mitigation. 1338 1339 This parameter overrides the compile time default set 1340 by CONFIG_MITIGATION_RFDS. Mitigation cannot be 1341 disabled when other VERW based mitigations (like MDS) 1342 are enabled. In order to disable RFDS mitigation all 1343 VERW based mitigations need to be disabled. 1344 1345 For details see: 1346 Documentation/admin-guide/hw-vuln/reg-file-data-sampling.rst 1347 1348 dm_verity.keyring_unsealed= 1349 [KNL] When set to 1, leave the dm-verity keyring 1350 unsealed after initialization so userspace can 1351 provision keys. Once the keyring is restricted 1352 it becomes active and is searched during signature 1353 verification. 1354 1355 driver_async_probe= [KNL] 1356 List of driver names to be probed asynchronously. * 1357 matches with all driver names. If * is specified, the 1358 rest of the listed driver names are those that will NOT 1359 match the *. 1360 Format: <driver_name1>,<driver_name2>... 1361 1362 drm.edid_firmware=[<connector>:]<file>[,[<connector>:]<file>] 1363 Broken monitors, graphic adapters, KVMs and EDIDless 1364 panels may send no or incorrect EDID data sets. 1365 This parameter allows to specify an EDID data sets 1366 in the /lib/firmware directory that are used instead. 1367 An EDID data set will only be used for a particular 1368 connector, if its name and a colon are prepended to 1369 the EDID name. Each connector may use a unique EDID 1370 data set by separating the files with a comma. An EDID 1371 data set with no connector name will be used for 1372 any connectors not explicitly specified. 1373 1374 dscc4.setup= [NET] 1375 1376 dt_cpu_ftrs= [PPC,EARLY] 1377 Format: {"off" | "known"} 1378 Control how the dt_cpu_ftrs device-tree binding is 1379 used for CPU feature discovery and setup (if it 1380 exists). 1381 off: Do not use it, fall back to legacy cpu table. 1382 known: Do not pass through unknown features to guests 1383 or userspace, only those that the kernel is aware of. 1384 1385 dump_apple_properties [X86] 1386 Dump name and content of EFI device properties on 1387 x86 Macs. Useful for driver authors to determine 1388 what data is available or for reverse-engineering. 1389 1390 dyndbg[="val"] [KNL,DYNAMIC_DEBUG] 1391 <module>.dyndbg[="val"] 1392 Enable debug messages at boot time. See 1393 Documentation/admin-guide/dynamic-debug-howto.rst 1394 for details. 1395 1396 early_ioremap_debug [KNL,EARLY] 1397 Enable debug messages in early_ioremap support. This 1398 is useful for tracking down temporary early mappings 1399 which are not unmapped. 1400 1401 earlycon= [KNL,EARLY] Output early console device and options. 1402 1403 When used with no options, the early console is 1404 determined by stdout-path property in device tree's 1405 chosen node or the ACPI SPCR table if supported by 1406 the platform. 1407 1408 cdns,<addr>[,options] 1409 Start an early, polled-mode console on a Cadence 1410 (xuartps) serial port at the specified address. Only 1411 supported option is baud rate. If baud rate is not 1412 specified, the serial port must already be setup and 1413 configured. 1414 1415 uart[8250],io,<addr>[,options[,uartclk]] 1416 uart[8250],mmio,<addr>[,options[,uartclk]] 1417 uart[8250],mmio32,<addr>[,options[,uartclk]] 1418 uart[8250],mmio32be,<addr>[,options[,uartclk]] 1419 uart[8250],0x<addr>[,options] 1420 Start an early, polled-mode console on the 8250/16550 1421 UART at the specified I/O port or MMIO address. 1422 MMIO inter-register address stride is either 8-bit 1423 (mmio) or 32-bit (mmio32 or mmio32be). 1424 If none of [io|mmio|mmio32|mmio32be], <addr> is assumed 1425 to be equivalent to 'mmio'. 'options' are specified 1426 in the same format described for "console=ttyS<n>"; if 1427 unspecified, the h/w is not initialized. 'uartclk' is 1428 the uart clock frequency; if unspecified, it is set 1429 to 'BASE_BAUD' * 16. 1430 1431 pl011,<addr> 1432 pl011,mmio32,<addr> 1433 Start an early, polled-mode console on a pl011 serial 1434 port at the specified address. The pl011 serial port 1435 must already be setup and configured. Options are not 1436 yet supported. If 'mmio32' is specified, then only 1437 the driver will use only 32-bit accessors to read/write 1438 the device registers. 1439 1440 liteuart,<addr> 1441 Start an early console on a litex serial port at the 1442 specified address. The serial port must already be 1443 setup and configured. Options are not yet supported. 1444 1445 meson,<addr> 1446 Start an early, polled-mode console on a meson serial 1447 port at the specified address. The serial port must 1448 already be setup and configured. Options are not yet 1449 supported. 1450 1451 msm_serial,<addr> 1452 Start an early, polled-mode console on an msm serial 1453 port at the specified address. The serial port 1454 must already be setup and configured. Options are not 1455 yet supported. 1456 1457 msm_serial_dm,<addr> 1458 Start an early, polled-mode console on an msm serial 1459 dm port at the specified address. The serial port 1460 must already be setup and configured. Options are not 1461 yet supported. 1462 1463 owl,<addr> 1464 Start an early, polled-mode console on a serial port 1465 of an Actions Semi SoC, such as S500 or S900, at the 1466 specified address. The serial port must already be 1467 setup and configured. Options are not yet supported. 1468 1469 rda,<addr> 1470 Start an early, polled-mode console on a serial port 1471 of an RDA Micro SoC, such as RDA8810PL, at the 1472 specified address. The serial port must already be 1473 setup and configured. Options are not yet supported. 1474 1475 sbi 1476 Use RISC-V SBI (Supervisor Binary Interface) for early 1477 console. 1478 1479 smh Use ARM semihosting calls for early console. 1480 1481 s3c2410,<addr> 1482 s3c2412,<addr> 1483 s3c2440,<addr> 1484 s3c6400,<addr> 1485 s5pv210,<addr> 1486 exynos4210,<addr> 1487 Use early console provided by serial driver available 1488 on Samsung SoCs, requires selecting proper type and 1489 a correct base address of the selected UART port. The 1490 serial port must already be setup and configured. 1491 Options are not yet supported. 1492 1493 lantiq,<addr> 1494 Start an early, polled-mode console on a lantiq serial 1495 (lqasc) port at the specified address. The serial port 1496 must already be setup and configured. Options are not 1497 yet supported. 1498 1499 lpuart,<addr> 1500 lpuart32,<addr> 1501 Use early console provided by Freescale LP UART driver 1502 found on Freescale Vybrid and QorIQ LS1021A processors. 1503 A valid base address must be provided, and the serial 1504 port must already be setup and configured. 1505 1506 ec_imx21,<addr> 1507 ec_imx6q,<addr> 1508 Start an early, polled-mode, output-only console on the 1509 Freescale i.MX UART at the specified address. The UART 1510 must already be setup and configured. 1511 1512 ar3700_uart,<addr> 1513 Start an early, polled-mode console on the 1514 Armada 3700 serial port at the specified 1515 address. The serial port must already be setup 1516 and configured. Options are not yet supported. 1517 1518 qcom_geni,<addr> 1519 Start an early, polled-mode console on a Qualcomm 1520 Generic Interface (GENI) based serial port at the 1521 specified address. The serial port must already be 1522 setup and configured. Options are not yet supported. 1523 1524 efifb,[options] 1525 Start an early, unaccelerated console on the EFI 1526 memory mapped framebuffer (if available). On cache 1527 coherent non-x86 systems that use system memory for 1528 the framebuffer, pass the 'ram' option so that it is 1529 mapped with the correct attributes. 1530 1531 linflex,<addr> 1532 Use early console provided by Freescale LINFlexD UART 1533 serial driver for NXP S32V234 SoCs. A valid base 1534 address must be provided, and the serial port must 1535 already be setup and configured. 1536 1537 earlyprintk= [X86,SH,ARM,M68k,S390,UM,EARLY] 1538 earlyprintk=vga 1539 earlyprintk=sclp 1540 earlyprintk=xen 1541 earlyprintk=serial[,ttySn[,baudrate]] 1542 earlyprintk=serial[,0x...[,baudrate]] 1543 earlyprintk=ttySn[,baudrate] 1544 earlyprintk=dbgp[debugController#] 1545 earlyprintk=mmio32,membase[,{nocfg|baudrate}] 1546 earlyprintk=pciserial[,force],bus:device.function[,{nocfg|baudrate}] 1547 earlyprintk=xdbc[xhciController#] 1548 earlyprintk=bios 1549 1550 earlyprintk is useful when the kernel crashes before 1551 the normal console is initialized. It is not enabled by 1552 default because it has some cosmetic problems. 1553 1554 Use "nocfg" to skip UART configuration, assume 1555 BIOS/firmware has configured UART correctly. 1556 1557 Append ",keep" to not disable it when the real console 1558 takes over. 1559 1560 Only one of vga, serial, or usb debug port can 1561 be used at a time. 1562 1563 Currently only ttyS0 and ttyS1 may be specified by 1564 name. Other I/O ports may be explicitly specified 1565 on some architectures (x86 and arm at least) by 1566 replacing ttySn with an I/O port address, like this: 1567 earlyprintk=serial,0x1008,115200 1568 You can find the port for a given device in 1569 /proc/tty/driver/serial: 1570 2: uart:ST16650V2 port:00001008 irq:18 ... 1571 1572 Interaction with the standard serial driver is not 1573 very good. 1574 1575 The VGA output is eventually overwritten by 1576 the real console. 1577 1578 The xen option can only be used in Xen domains. 1579 1580 The sclp output can only be used on s390. 1581 1582 The bios output can only be used on SuperH. 1583 1584 The optional "force" to "pciserial" enables use of a 1585 PCI device even when its classcode is not of the 1586 UART class. 1587 1588 edac_report= [HW,EDAC] Control how to report EDAC event 1589 Format: {"on" | "off" | "force"} 1590 on: enable EDAC to report H/W event. May be overridden 1591 by other higher priority error reporting module. 1592 off: disable H/W event reporting through EDAC. 1593 force: enforce the use of EDAC to report H/W event. 1594 default: on. 1595 1596 edd= [EDD] 1597 Format: {"off" | "on" | "skip[mbr]"} 1598 1599 efi= [EFI,EARLY] 1600 Format: { "debug", "disable_early_pci_dma", 1601 "nochunk", "noruntime", "nosoftreserve", 1602 "novamap", "no_disable_early_pci_dma" } 1603 debug: enable misc debug output. 1604 disable_early_pci_dma: disable the busmaster bit on all 1605 PCI bridges while in the EFI boot stub. 1606 nochunk: disable reading files in "chunks" in the EFI 1607 boot stub, as chunking can cause problems with some 1608 firmware implementations. 1609 noruntime : disable EFI runtime services support 1610 nosoftreserve: The EFI_MEMORY_SP (Specific Purpose) 1611 attribute may cause the kernel to reserve the 1612 memory range for a memory mapping driver to 1613 claim. Specify efi=nosoftreserve to disable this 1614 reservation and treat the memory by its base type 1615 (i.e. EFI_CONVENTIONAL_MEMORY / "System RAM"). 1616 novamap: do not call SetVirtualAddressMap(). 1617 no_disable_early_pci_dma: Leave the busmaster bit set 1618 on all PCI bridges while in the EFI boot stub 1619 1620 efi_no_storage_paranoia [EFI,X86,EARLY] 1621 The kernel reserves 5KB of EFI variable storage for 1622 safety, because some UEFI implementation may fail to 1623 boot if there's insufficient space in the EFI variable 1624 storage. 1625 1626 Using this parameter, you can use the 5KB reservation 1627 in the EFI variable storage. 1628 1629 However, Use this parameter only if you are really 1630 sure that your UEFI does sane gc and fulfills the spec 1631 otherwise your board may brick. 1632 1633 efivar_ssdt= [EFI; X86] Name of an EFI variable that contains an SSDT 1634 that is to be dynamically loaded by Linux. If there are 1635 multiple variables with the same name but with different 1636 vendor GUIDs, all of them will be loaded. See 1637 Documentation/admin-guide/acpi/ssdt-overlays.rst for details. 1638 1639 1640 eisa_irq_edge= [PARISC,HW] 1641 See header of drivers/parisc/eisa.c. 1642 1643 ekgdboc= [X86,KGDB,EARLY] Allow early kernel console debugging 1644 Format: ekgdboc=kbd 1645 1646 This is designed to be used in conjunction with 1647 the boot argument: earlyprintk=vga 1648 1649 This parameter works in place of the kgdboc parameter 1650 but can only be used if the backing tty is available 1651 very early in the boot process. For early debugging 1652 via a serial port see kgdboc_earlycon instead. 1653 1654 elfcorehdr=[size[KMG]@]offset[KMG] [PPC,SH,X86,S390,EARLY] 1655 Specifies physical address of start of kernel core 1656 image elf header and optionally the size. Generally 1657 kexec loader will pass this option to capture kernel. 1658 See Documentation/admin-guide/kdump/kdump.rst for details. 1659 1660 enable_mtrr_cleanup [X86,EARLY] 1661 The kernel tries to adjust MTRR layout from continuous 1662 to discrete, to make X server driver able to add WB 1663 entry later. This parameter enables that. 1664 1665 enable_timer_pin_1 [X86] 1666 Enable PIN 1 of APIC timer 1667 Can be useful to work around chipset bugs 1668 (in particular on some ATI chipsets). 1669 The kernel tries to set a reasonable default. 1670 1671 enforcing= [SELINUX] Set initial enforcing status. 1672 Format: {"0" | "1"} 1673 See security/selinux/Kconfig help text. 1674 0 -- permissive (log only, no denials). 1675 1 -- enforcing (deny and log). 1676 Default value is 0. 1677 Value can be changed at runtime via 1678 /sys/fs/selinux/enforce. 1679 1680 erst_disable [ACPI] 1681 Disable Error Record Serialization Table (ERST) 1682 support. 1683 1684 ether= [HW,NET] Ethernet cards parameters 1685 This option is obsoleted by the "netdev=" option, which 1686 has equivalent usage. See its documentation for details. 1687 1688 evm= [EVM] 1689 Format: { "fix" } 1690 Permit 'security.evm' to be updated regardless of 1691 current integrity status. 1692 1693 early_page_ext [KNL,EARLY] Enforces page_ext initialization to earlier 1694 stages so cover more early boot allocations. 1695 Please note that as side effect some optimizations 1696 might be disabled to achieve that (e.g. parallelized 1697 memory initialization is disabled) so the boot process 1698 might take longer, especially on systems with a lot of 1699 memory. Available with CONFIG_PAGE_EXTENSION=y. 1700 1701 failslab= 1702 fail_usercopy= 1703 fail_page_alloc= 1704 fail_skb_realloc= 1705 fail_make_request=[KNL] 1706 General fault injection mechanism. 1707 Format: <interval>,<probability>,<space>,<times> 1708 See also Documentation/fault-injection/. 1709 1710 fb_tunnels= [NET] 1711 Format: { initns | none } 1712 See Documentation/admin-guide/sysctl/net.rst for 1713 fb_tunnels_only_for_init_ns 1714 1715 floppy= [HW] 1716 See Documentation/admin-guide/blockdev/floppy.rst. 1717 1718 forcepae [X86-32] 1719 Forcefully enable Physical Address Extension (PAE). 1720 Many Pentium M systems disable PAE but may have a 1721 functionally usable PAE implementation. 1722 Warning: use of this parameter will taint the kernel 1723 and may cause unknown problems. 1724 1725 fred= [X86-64] 1726 Enable/disable Flexible Return and Event Delivery. 1727 Format: { on | off } 1728 on: enable FRED when it's present, the default setting. 1729 off: disable FRED. 1730 1731 ftrace=[tracer] 1732 [FTRACE] will set and start the specified tracer 1733 as early as possible in order to facilitate early 1734 boot debugging. 1735 1736 ftrace_boot_snapshot 1737 [FTRACE] On boot up, a snapshot will be taken of the 1738 ftrace ring buffer that can be read at: 1739 /sys/kernel/tracing/snapshot. 1740 This is useful if you need tracing information from kernel 1741 boot up that is likely to be overridden by user space 1742 start up functionality. 1743 1744 Optionally, the snapshot can also be defined for a tracing 1745 instance that was created by the trace_instance= command 1746 line parameter. 1747 1748 trace_instance=foo,sched_switch ftrace_boot_snapshot=foo 1749 1750 The above will cause the "foo" tracing instance to trigger 1751 a snapshot at the end of boot up. 1752 1753 ftrace_dump_on_oops[=2(orig_cpu) | =<instance>][,<instance> | 1754 ,<instance>=2(orig_cpu)] 1755 [FTRACE] will dump the trace buffers on oops. 1756 If no parameter is passed, ftrace will dump global 1757 buffers of all CPUs, if you pass 2 or orig_cpu, it 1758 will dump only the buffer of the CPU that triggered 1759 the oops, or the specific instance will be dumped if 1760 its name is passed. Multiple instance dump is also 1761 supported, and instances are separated by commas. Each 1762 instance supports only dump on CPU that triggered the 1763 oops by passing 2 or orig_cpu to it. 1764 1765 ftrace_dump_on_oops=foo=orig_cpu 1766 1767 The above will dump only the buffer of "foo" instance 1768 on CPU that triggered the oops. 1769 1770 ftrace_dump_on_oops,foo,bar=orig_cpu 1771 1772 The above will dump global buffer on all CPUs, the 1773 buffer of "foo" instance on all CPUs and the buffer 1774 of "bar" instance on CPU that triggered the oops. 1775 1776 ftrace_filter=[function-list] 1777 [FTRACE] Limit the functions traced by the function 1778 tracer at boot up. function-list is a comma-separated 1779 list of functions. This list can be changed at run 1780 time by the set_ftrace_filter file in the debugfs 1781 tracing directory. 1782 1783 ftrace_notrace=[function-list] 1784 [FTRACE] Do not trace the functions specified in 1785 function-list. This list can be changed at run time 1786 by the set_ftrace_notrace file in the debugfs 1787 tracing directory. 1788 1789 ftrace_graph_filter=[function-list] 1790 [FTRACE] Limit the top level callers functions traced 1791 by the function graph tracer at boot up. 1792 function-list is a comma-separated list of functions 1793 that can be changed at run time by the 1794 set_graph_function file in the debugfs tracing directory. 1795 1796 ftrace_graph_notrace=[function-list] 1797 [FTRACE] Do not trace from the functions specified in 1798 function-list. This list is a comma-separated list of 1799 functions that can be changed at run time by the 1800 set_graph_notrace file in the debugfs tracing directory. 1801 1802 ftrace_graph_max_depth=<uint> 1803 [FTRACE] Used with the function graph tracer. This is 1804 the max depth it will trace into a function. This value 1805 can be changed at run time by the max_graph_depth file 1806 in the tracefs tracing directory. default: 0 (no limit) 1807 1808 fw_devlink= [KNL,EARLY] Create device links between consumer and supplier 1809 devices by scanning the firmware to infer the 1810 consumer/supplier relationships. This feature is 1811 especially useful when drivers are loaded as modules as 1812 it ensures proper ordering of tasks like device probing 1813 (suppliers first, then consumers), supplier boot state 1814 clean up (only after all consumers have probed), 1815 suspend/resume & runtime PM (consumers first, then 1816 suppliers). 1817 Format: { off | permissive | on | rpm } 1818 off -- Don't create device links from firmware info. 1819 permissive -- Create device links from firmware info 1820 but use it only for ordering boot state clean 1821 up (sync_state() calls). 1822 on -- Create device links from firmware info and use it 1823 to enforce probe and suspend/resume ordering. 1824 rpm -- Like "on", but also use to order runtime PM. 1825 1826 fw_devlink.strict=<bool> 1827 [KNL,EARLY] Treat all inferred dependencies as mandatory 1828 dependencies. This only applies for fw_devlink=on|rpm. 1829 Format: <bool> 1830 1831 fw_devlink.sync_state = 1832 [KNL,EARLY] When all devices that could probe have finished 1833 probing, this parameter controls what to do with 1834 devices that haven't yet received their sync_state() 1835 calls. 1836 Format: { strict | timeout } 1837 strict -- Default. Continue waiting on consumers to 1838 probe successfully. 1839 timeout -- Give up waiting on consumers and call 1840 sync_state() on any devices that haven't yet 1841 received their sync_state() calls after 1842 deferred_probe_timeout has expired or by 1843 late_initcall() if !CONFIG_MODULES. 1844 1845 gamecon.map[2|3]= 1846 [HW,JOY] Multisystem joystick and NES/SNES/PSX pad 1847 support via parallel port (up to 5 devices per port) 1848 Format: <port#>,<pad1>,<pad2>,<pad3>,<pad4>,<pad5> 1849 See also Documentation/input/devices/joystick-parport.rst 1850 1851 gamma= [HW,DRM] 1852 1853 gart_fix_e820= [X86-64,EARLY] disable the fix e820 for K8 GART 1854 Format: off | on 1855 default: on 1856 1857 gather_data_sampling= 1858 [X86,INTEL,EARLY] Control the Gather Data Sampling (GDS) 1859 mitigation. 1860 1861 Gather Data Sampling is a hardware vulnerability which 1862 allows unprivileged speculative access to data which was 1863 previously stored in vector registers. 1864 1865 This issue is mitigated by default in updated microcode. 1866 The mitigation may have a performance impact but can be 1867 disabled. On systems without the microcode mitigation 1868 disabling AVX serves as a mitigation. 1869 1870 force: Disable AVX to mitigate systems without 1871 microcode mitigation. No effect if the microcode 1872 mitigation is present. Known to cause crashes in 1873 userspace with buggy AVX enumeration. 1874 1875 off: Disable GDS mitigation. 1876 1877 gbpages [X86] Use GB pages for kernel direct mappings. 1878 1879 gcov_persist= [GCOV] When non-zero (default), profiling data for 1880 kernel modules is saved and remains accessible via 1881 debugfs, even when the module is unloaded/reloaded. 1882 When zero, profiling data is discarded and associated 1883 debugfs files are removed at module unload time. 1884 1885 goldfish [X86] Enable the goldfish android emulator platform. 1886 Don't use this when you are not running on the 1887 android emulator 1888 1889 gpio-mockup.gpio_mockup_ranges 1890 [HW] Sets the ranges of gpiochip of for this device. 1891 Format: <start1>,<end1>,<start2>,<end2>... 1892 gpio-mockup.gpio_mockup_named_lines 1893 [HW] Let the driver know GPIO lines should be named. 1894 1895 gpt [EFI] Forces disk with valid GPT signature but 1896 invalid Protective MBR to be treated as GPT. If the 1897 primary GPT is corrupted, it enables the backup/alternate 1898 GPT to be used instead. 1899 1900 grcan.enable0= [HW] Configuration of physical interface 0. Determines 1901 the "Enable 0" bit of the configuration register. 1902 Format: 0 | 1 1903 Default: 0 1904 grcan.enable1= [HW] Configuration of physical interface 1. Determines 1905 the "Enable 0" bit of the configuration register. 1906 Format: 0 | 1 1907 Default: 0 1908 grcan.select= [HW] Select which physical interface to use. 1909 Format: 0 | 1 1910 Default: 0 1911 grcan.txsize= [HW] Sets the size of the tx buffer. 1912 Format: <unsigned int> such that (txsize & ~0x1fffc0) == 0. 1913 Default: 1024 1914 grcan.rxsize= [HW] Sets the size of the rx buffer. 1915 Format: <unsigned int> such that (rxsize & ~0x1fffc0) == 0. 1916 Default: 1024 1917 1918 hardened_usercopy= 1919 [KNL] Under CONFIG_HARDENED_USERCOPY, whether 1920 hardening is enabled for this boot. Hardened 1921 usercopy checking is used to protect the kernel 1922 from reading or writing beyond known memory 1923 allocation boundaries as a proactive defense 1924 against bounds-checking flaws in the kernel's 1925 copy_to_user()/copy_from_user() interface. 1926 The default is determined by 1927 CONFIG_HARDENED_USERCOPY_DEFAULT_ON. 1928 on Perform hardened usercopy checks. 1929 off Disable hardened usercopy checks. 1930 1931 hardlockup_all_cpu_backtrace= 1932 [KNL] Should the hard-lockup detector generate 1933 backtraces on all cpus. 1934 Format: 0 | 1 1935 1936 hash_pointers= 1937 [KNL,EARLY] 1938 By default, when pointers are printed to the console 1939 or buffers via the %p format string, that pointer is 1940 "hashed", i.e. obscured by hashing the pointer value. 1941 This is a security feature that hides actual kernel 1942 addresses from unprivileged users, but it also makes 1943 debugging the kernel more difficult since unequal 1944 pointers can no longer be compared. The choices are: 1945 Format: { auto | always | never } 1946 Default: auto 1947 1948 auto - Hash pointers unless slab_debug is enabled. 1949 always - Always hash pointers (even if slab_debug is 1950 enabled). 1951 never - Never hash pointers. This option should only 1952 be specified when debugging the kernel. Do 1953 not use on production kernels. The boot 1954 param "no_hash_pointers" is an alias for 1955 this mode. 1956 1957 For controlling hashing dynamically at runtime, 1958 use the "kernel.kptr_restrict" sysctl instead. 1959 1960 hashdist= [KNL,NUMA] Large hashes allocated during boot 1961 are distributed across NUMA nodes. Defaults on 1962 for 64-bit NUMA, off otherwise. 1963 Format: 0 | 1 (for off | on) 1964 1965 hd= [EIDE] (E)IDE hard drive subsystem geometry 1966 Format: <cyl>,<head>,<sect> 1967 1968 hest_disable [ACPI] 1969 Disable Hardware Error Source Table (HEST) support; 1970 corresponding firmware-first mode error processing 1971 logic will be disabled. 1972 1973 hibernate= [HIBERNATION] 1974 noresume Don't check if there's a hibernation image 1975 present during boot. 1976 nocompress Don't compress/decompress hibernation images. 1977 no Disable hibernation and resume. 1978 protect_image Turn on image protection during restoration 1979 (that will set all pages holding image data 1980 during restoration read-only). 1981 1982 hibernate.compressor= [HIBERNATION] Compression algorithm to be 1983 used with hibernation. 1984 Format: { lzo | lz4 } 1985 Default: lzo 1986 1987 lzo: Select LZO compression algorithm to 1988 compress/decompress hibernation image. 1989 1990 lz4: Select LZ4 compression algorithm to 1991 compress/decompress hibernation image. 1992 1993 hibernate.pm_test_delay= 1994 [HIBERNATION] 1995 Sets the number of seconds to remain in a hibernation test 1996 mode before resuming the system (see 1997 /sys/power/pm_test). Only available when CONFIG_PM_DEBUG 1998 is set. Default value is 5. 1999 2000 hibernate_compression_threads= 2001 [HIBERNATION] 2002 Set the number of threads used for compressing or decompressing 2003 hibernation images. 2004 2005 Format: <integer> 2006 Default: 3 2007 Minimum: 1 2008 Example: hibernate_compression_threads=4 2009 2010 highmem=nn[KMG] [KNL,BOOT,EARLY] forces the highmem zone to have an exact 2011 size of <nn>. This works even on boxes that have no 2012 highmem otherwise. This also works to reduce highmem 2013 size on bigger boxes. 2014 2015 highres= [KNL] Enable/disable high resolution timer mode. 2016 Valid parameters: "on", "off" 2017 Default: "on" 2018 2019 hlt [BUGS=ARM,SH] 2020 2021 hostname= [KNL,EARLY] Set the hostname (aka UTS nodename). 2022 Format: <string> 2023 This allows setting the system's hostname during early 2024 startup. This sets the name returned by gethostname. 2025 Using this parameter to set the hostname makes it 2026 possible to ensure the hostname is correctly set before 2027 any userspace processes run, avoiding the possibility 2028 that a process may call gethostname before the hostname 2029 has been explicitly set, resulting in the calling 2030 process getting an incorrect result. The string must 2031 not exceed the maximum allowed hostname length (usually 2032 64 characters) and will be truncated otherwise. 2033 2034 hpet= [X86-32,HPET] option to control HPET usage 2035 Format: { enable (default) | disable | force | 2036 verbose } 2037 disable: disable HPET and use PIT instead 2038 force: allow force enabled of undocumented chips (ICH4, 2039 VIA, nVidia) 2040 verbose: show contents of HPET registers during setup 2041 2042 hpet_mmap= [X86, HPET_MMAP] Allow userspace to mmap HPET 2043 registers. Default set by CONFIG_HPET_MMAP_DEFAULT. 2044 2045 hugepages= [HW,EARLY] Number of HugeTLB pages to allocate at boot. 2046 If this follows hugepagesz (below), it specifies 2047 the number of pages of hugepagesz to be allocated. 2048 If this is the first HugeTLB parameter on the command 2049 line, it specifies the number of pages to allocate for 2050 the default huge page size. If using node format, the 2051 number of pages to allocate per-node can be specified. 2052 See also Documentation/admin-guide/mm/hugetlbpage.rst. 2053 Format: <integer> or (node format) 2054 <node>:<integer>[,<node>:<integer>] 2055 2056 hugepagesz= 2057 [HW,EARLY] The size of the HugeTLB pages. This is 2058 used in conjunction with hugepages (above) to 2059 allocate huge pages of a specific size at boot. The 2060 pair hugepagesz=X hugepages=Y can be specified once 2061 for each supported huge page size. Huge page sizes 2062 are architecture dependent. See also 2063 Documentation/admin-guide/mm/hugetlbpage.rst. 2064 Format: size[KMG] 2065 2066 hugepage_alloc_threads= 2067 [HW] The number of threads that should be used to 2068 allocate hugepages during boot. This option can be 2069 used to improve system bootup time when allocating 2070 a large amount of huge pages. 2071 The default value is 25% of the available hardware threads. 2072 2073 Note that this parameter only applies to non-gigantic huge pages. 2074 2075 hugetlb_cma= [HW,CMA,EARLY] The size of a CMA area used for allocation 2076 of gigantic hugepages. Or using node format, the size 2077 of a CMA area per node can be specified. 2078 Format: nn[KMGTPE] or (node format) 2079 <node>:nn[KMGTPE][,<node>:nn[KMGTPE]] 2080 2081 The size must be a multiple of the gigantic page size. 2082 When using node format, this applies to each per-node size. 2083 Missaligned values are dropped with a warning. 2084 2085 Reserve a CMA area of given size and allocate gigantic 2086 hugepages using the CMA allocator. If enabled, the 2087 boot-time allocation of gigantic hugepages is skipped. 2088 2089 hugetlb_cma_only= 2090 [HW,CMA,EARLY] When allocating new HugeTLB pages, only 2091 try to allocate from the CMA areas. 2092 2093 This option does nothing if hugetlb_cma= is not also 2094 specified. 2095 2096 hugetlb_free_vmemmap= 2097 [KNL] Requires CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP 2098 enabled. 2099 Control if HugeTLB Vmemmap Optimization (HVO) is enabled. 2100 Allows heavy hugetlb users to free up some more 2101 memory (7 * PAGE_SIZE for each 2MB hugetlb page). 2102 Format: { on | off (default) } 2103 2104 on: enable HVO 2105 off: disable HVO 2106 2107 Built with CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP_DEFAULT_ON=y, 2108 the default is on. 2109 2110 Note that the vmemmap pages may be allocated from the added 2111 memory block itself when memory_hotplug.memmap_on_memory is 2112 enabled, those vmemmap pages cannot be optimized even if this 2113 feature is enabled. Other vmemmap pages not allocated from 2114 the added memory block itself do not be affected. 2115 2116 hung_task_panic= 2117 [KNL] Number of hung tasks to trigger kernel panic. 2118 Format: <int> 2119 2120 When set to a non-zero value, a kernel panic will be triggered if 2121 the number of detected hung tasks reaches this value. 2122 2123 0: don't panic 2124 1: panic immediately on first hung task 2125 N: panic after N hung tasks are detected in a single scan 2126 2127 The default value is controlled by the 2128 CONFIG_BOOTPARAM_HUNG_TASK_PANIC build-time option. The value 2129 selected by this boot parameter can be changed later by the 2130 kernel.hung_task_panic sysctl. 2131 2132 hvc_iucv= [S390] Number of z/VM IUCV hypervisor console (HVC) 2133 terminal devices. Valid values: 0..8 2134 hvc_iucv_allow= [S390] Comma-separated list of z/VM user IDs. 2135 If specified, z/VM IUCV HVC accepts connections 2136 from listed z/VM user IDs only. 2137 2138 hv_nopvspin [X86,HYPER_V,EARLY] 2139 Disables the paravirt spinlock optimizations 2140 which allow the hypervisor to 'idle' the guest 2141 on lock contention. 2142 2143 hw_protection= [HW] 2144 Format: reboot | shutdown 2145 2146 Hardware protection action taken on critical events like 2147 overtemperature or imminent voltage loss. 2148 2149 i2c_bus= [HW] Override the default board specific I2C bus speed 2150 or register an additional I2C bus that is not 2151 registered from board initialization code. 2152 Format: 2153 <bus_id>,<clkrate> 2154 2155 i2c_touchscreen_props= [HW,ACPI,X86] 2156 Set device-properties for ACPI-enumerated I2C-attached 2157 touchscreen, to e.g. fix coordinates of upside-down 2158 mounted touchscreens. If you need this option please 2159 submit a drivers/platform/x86/touchscreen_dmi.c patch 2160 adding a DMI quirk for this. 2161 2162 Format: 2163 <ACPI_HW_ID>:<prop_name>=<val>[:prop_name=val][:...] 2164 Where <val> is one of: 2165 Omit "=<val>" entirely Set a boolean device-property 2166 Unsigned number Set a u32 device-property 2167 Anything else Set a string device-property 2168 2169 Examples (split over multiple lines): 2170 i2c_touchscreen_props=GDIX1001:touchscreen-inverted-x: 2171 touchscreen-inverted-y 2172 2173 i2c_touchscreen_props=MSSL1680:touchscreen-size-x=1920: 2174 touchscreen-size-y=1080:touchscreen-inverted-y: 2175 firmware-name=gsl1680-vendor-model.fw:silead,home-button 2176 2177 i8042.debug [HW] Toggle i8042 debug mode 2178 i8042.unmask_kbd_data 2179 [HW] Enable printing of interrupt data from the KBD port 2180 (disabled by default, and as a pre-condition 2181 requires that i8042.debug=1 be enabled) 2182 i8042.direct [HW] Put keyboard port into non-translated mode 2183 i8042.dumbkbd [HW] Pretend that controller can only read data from 2184 keyboard and cannot control its state 2185 (Don't attempt to blink the leds) 2186 i8042.noaux [HW] Don't check for auxiliary (== mouse) port 2187 i8042.nokbd [HW] Don't check/create keyboard port 2188 i8042.noloop [HW] Disable the AUX Loopback command while probing 2189 for the AUX port 2190 i8042.nomux [HW] Don't check presence of an active multiplexing 2191 controller 2192 i8042.nopnp [HW] Don't use ACPIPnP / PnPBIOS to discover KBD/AUX 2193 controllers 2194 i8042.notimeout [HW] Ignore timeout condition signalled by controller 2195 i8042.reset [HW] Reset the controller during init, cleanup and 2196 suspend-to-ram transitions, only during s2r 2197 transitions, or never reset 2198 Format: { 1 | Y | y | 0 | N | n } 2199 1, Y, y: always reset controller 2200 0, N, n: don't ever reset controller 2201 Default: only on s2r transitions on x86; most other 2202 architectures force reset to be always executed 2203 i8042.unlock [HW] Unlock (ignore) the keylock 2204 i8042.kbdreset [HW] Reset device connected to KBD port 2205 i8042.probe_defer 2206 [HW] Allow deferred probing upon i8042 probe errors 2207 2208 i810= [HW,DRM] 2209 2210 i915.invert_brightness= 2211 [DRM] Invert the sense of the variable that is used to 2212 set the brightness of the panel backlight. Normally a 2213 brightness value of 0 indicates backlight switched off, 2214 and the maximum of the brightness value sets the backlight 2215 to maximum brightness. If this parameter is set to 0 2216 (default) and the machine requires it, or this parameter 2217 is set to 1, a brightness value of 0 sets the backlight 2218 to maximum brightness, and the maximum of the brightness 2219 value switches the backlight off. 2220 -1 -- never invert brightness 2221 0 -- machine default 2222 1 -- force brightness inversion 2223 2224 ia32_emulation= [X86-64] 2225 Format: <bool> 2226 When true, allows loading 32-bit programs and executing 32-bit 2227 syscalls, essentially overriding IA32_EMULATION_DEFAULT_DISABLED at 2228 boot time. When false, unconditionally disables IA32 emulation. 2229 2230 2231 idle= [X86,EARLY] 2232 Format: idle=poll, idle=halt, idle=nomwait 2233 2234 idle=poll: Don't do power saving in the idle loop 2235 using HLT, but poll for rescheduling event. This will 2236 make the CPUs eat a lot more power, but may be useful 2237 to get slightly better performance in multiprocessor 2238 benchmarks. It also makes some profiling using 2239 performance counters more accurate. Please note that 2240 on systems with MONITOR/MWAIT support (like Intel 2241 EM64T CPUs) this option has no performance advantage 2242 over the normal idle loop. It may also interact badly 2243 with hyperthreading. 2244 2245 idle=halt: Halt is forced to be used for CPU idle. 2246 In such case C2/C3 won't be used again. 2247 2248 idle=nomwait: Disable mwait for CPU C-states 2249 2250 [ARM64,EARLY] 2251 Format: idle=wfi, idle=yield, idle=nop 2252 2253 idle=wfi: Use the WFI (Wait For Interrupt) hint 2254 instruction in the idle loop. This is the default and 2255 allows the CPU to enter a low-power state until an 2256 interrupt arrives. 2257 2258 idle=yield: Use the YIELD hint instruction instead of 2259 WFI. CPUs supporting simultaneous multi-threading (SMT), 2260 can continue executing another thread when the current 2261 thread reaches the idle loop. This will make the CPUs 2262 eat more power, but may be useful to get slightly better 2263 performance in some applications, since the CPUs will 2264 not enter a low-power state. 2265 2266 idle=nop: Do not execute any idle instruction in the 2267 idle loop. This is useful on platforms where WFI 2268 misbehaves, leading to system instability or loss of CPU 2269 state. This will make the CPUs eat more power, but may 2270 give slightly better performance in some applications, 2271 since the CPUs will not enter a low-power state. 2272 2273 idxd.sva= [HW] 2274 Format: <bool> 2275 Allow force disabling of Shared Virtual Memory (SVA) 2276 support for the idxd driver. By default it is set to 2277 true (1). 2278 2279 idxd.tc_override= [HW] 2280 Format: <bool> 2281 Allow override of default traffic class configuration 2282 for the device. By default it is set to false (0). 2283 2284 ieee754= [MIPS] Select IEEE Std 754 conformance mode 2285 Format: { strict | legacy | 2008 | relaxed | emulated } 2286 Default: strict 2287 2288 Choose which programs will be accepted for execution 2289 based on the IEEE 754 NaN encoding(s) supported by 2290 the FPU and the NaN encoding requested with the value 2291 of an ELF file header flag individually set by each 2292 binary. Hardware implementations are permitted to 2293 support either or both of the legacy and the 2008 NaN 2294 encoding mode. 2295 2296 Available settings are as follows: 2297 strict accept binaries that request a NaN encoding 2298 supported by the FPU 2299 legacy only accept legacy-NaN binaries, if supported 2300 by the FPU 2301 2008 only accept 2008-NaN binaries, if supported 2302 by the FPU 2303 relaxed accept any binaries regardless of whether 2304 supported by the FPU 2305 emulated accept any binaries but enable FPU emulator 2306 if binary mode is unsupported by the FPU. 2307 2308 The FPU emulator is always able to support both NaN 2309 encodings, so if no FPU hardware is present or it has 2310 been disabled with 'nofpu', then the settings of 2311 'legacy' and '2008' strap the emulator accordingly, 2312 'relaxed' straps the emulator for both legacy-NaN and 2313 2008-NaN, whereas 'strict' enables legacy-NaN only on 2314 legacy processors and both NaN encodings on MIPS32 or 2315 MIPS64 CPUs. 2316 2317 The setting for ABS.fmt/NEG.fmt instruction execution 2318 mode generally follows that for the NaN encoding, 2319 except where unsupported by hardware. 2320 2321 ignore_loglevel [KNL,EARLY] 2322 Ignore loglevel setting - this will print /all/ 2323 kernel messages to the console. Useful for debugging. 2324 We also add it as printk module parameter, so users 2325 could change it dynamically, usually by 2326 /sys/module/printk/parameters/ignore_loglevel. 2327 2328 ignore_rlimit_data 2329 Ignore RLIMIT_DATA setting for data mappings, 2330 print warning at first misuse. Can be changed via 2331 /sys/module/kernel/parameters/ignore_rlimit_data. 2332 2333 ihash_entries= [KNL] 2334 Set number of hash buckets for inode cache. 2335 2336 ima_appraise= [IMA] appraise integrity measurements 2337 Format: { "off" | "enforce" | "fix" | "log" } 2338 default: "enforce" 2339 2340 ima_appraise_tcb [IMA] Deprecated. Use ima_policy= instead. 2341 The builtin appraise policy appraises all files 2342 owned by uid=0. 2343 2344 ima_canonical_fmt [IMA] 2345 Use the canonical format for the binary runtime 2346 measurements, instead of host native format. 2347 2348 ima_flush_htable [IMA] 2349 Flush the IMA hash table when deleting all the 2350 staged measurement records, to achieve maximum 2351 memory saving at the cost of having duplicate 2352 records across the staged measurement lists. 2353 2354 ima_hash= [IMA] 2355 Format: { md5 | sha1 | rmd160 | sha256 | sha384 2356 | sha512 | ... } 2357 default: "sha1" 2358 2359 The list of supported hash algorithms is defined 2360 in crypto/hash_info.h. 2361 2362 ima_policy= [IMA] 2363 The builtin policies to load during IMA setup. 2364 Format: "tcb | appraise_tcb | secure_boot | 2365 fail_securely | critical_data" 2366 2367 The "tcb" policy measures all programs exec'd, files 2368 mmap'd for exec, and all files opened with the read 2369 mode bit set by either the effective uid (euid=0) or 2370 uid=0. 2371 2372 The "appraise_tcb" policy appraises the integrity of 2373 all files owned by root. 2374 2375 The "secure_boot" policy appraises the integrity 2376 of files (eg. kexec kernel image, kernel modules, 2377 firmware, policy, etc) based on file signatures. 2378 2379 The "fail_securely" policy forces file signature 2380 verification failure also on privileged mounted 2381 filesystems with the SB_I_UNVERIFIABLE_SIGNATURE 2382 flag. 2383 2384 The "critical_data" policy measures kernel integrity 2385 critical data. 2386 2387 ima_tcb [IMA] Deprecated. Use ima_policy= instead. 2388 Load a policy which meets the needs of the Trusted 2389 Computing Base. This means IMA will measure all 2390 programs exec'd, files mmap'd for exec, and all files 2391 opened for read by uid=0. 2392 2393 ima_template= [IMA] 2394 Select one of defined IMA measurements template formats. 2395 Formats: { "ima" | "ima-ng" | "ima-ngv2" | "ima-sig" | 2396 "ima-sigv2" } 2397 Default: "ima-ng" 2398 2399 ima_template_fmt= 2400 [IMA] Define a custom template format. 2401 Format: { "field1|...|fieldN" } 2402 2403 ima= [IMA] Enable or disable IMA 2404 Format: { "off" | "on" } 2405 Default: "on" 2406 Note that disabling IMA is limited to kdump kernel. 2407 2408 indirect_target_selection= [X86,Intel] Mitigation control for Indirect 2409 Target Selection(ITS) bug in Intel CPUs. Updated 2410 microcode is also required for a fix in IBPB. 2411 2412 on: Enable mitigation (default). 2413 off: Disable mitigation. 2414 force: Force the ITS bug and deploy default 2415 mitigation. 2416 vmexit: Only deploy mitigation if CPU is affected by 2417 guest/host isolation part of ITS. 2418 stuff: Deploy RSB-fill mitigation when retpoline is 2419 also deployed. Otherwise, deploy the default 2420 mitigation. 2421 2422 For details see: 2423 Documentation/admin-guide/hw-vuln/indirect-target-selection.rst 2424 2425 init= [KNL] 2426 Format: <full_path> 2427 Run specified binary instead of /sbin/init as init 2428 process. 2429 2430 initcall_debug [KNL] Trace initcalls as they are executed. Useful 2431 for working out where the kernel is dying during 2432 startup. 2433 2434 initcall_blacklist= [KNL] Do not execute a comma-separated list of 2435 initcall functions. Useful for debugging built-in 2436 modules and initcalls. 2437 2438 initramfs_async= [KNL] 2439 Format: <bool> 2440 Default: 1 2441 This parameter controls whether the initramfs 2442 image is unpacked asynchronously, concurrently 2443 with devices being probed and 2444 initialized. This should normally just work, 2445 but as a debugging aid, one can get the 2446 historical behaviour of the initramfs 2447 unpacking being completed before device_ and 2448 late_ initcalls. 2449 2450 initrd= [BOOT,EARLY] Specify the location of the initial ramdisk 2451 2452 initrdmem= [KNL,EARLY] Specify a physical address and size from which to 2453 load the initrd. If an initrd is compiled in or 2454 specified in the bootparams, it takes priority over this 2455 setting. 2456 Format: ss[KMG],nn[KMG] 2457 Default is 0, 0 2458 2459 init_on_alloc= [MM,EARLY] Fill newly allocated pages and heap objects with 2460 zeroes. 2461 Format: 0 | 1 2462 Default set by CONFIG_INIT_ON_ALLOC_DEFAULT_ON. 2463 2464 init_on_free= [MM,EARLY] Fill freed pages and heap objects with zeroes. 2465 Format: 0 | 1 2466 Default set by CONFIG_INIT_ON_FREE_DEFAULT_ON. 2467 2468 init_pkru= [X86] Specify the default memory protection keys rights 2469 register contents for all processes. 0x55555554 by 2470 default (disallow access to all but pkey 0). Can 2471 override in debugfs after boot. 2472 2473 inport.irq= [HW] Inport (ATI XL and Microsoft) busmouse driver 2474 Format: <irq> 2475 2476 int_pln_enable [X86] Enable power limit notification interrupt 2477 2478 integrity_audit=[IMA] 2479 Format: { "0" | "1" } 2480 0 -- basic integrity auditing messages. (Default) 2481 1 -- additional integrity auditing messages. 2482 2483 intel_iommu= [DMAR] Intel IOMMU driver (DMAR) option 2484 on 2485 Enable intel iommu driver. 2486 off 2487 Disable intel iommu driver. 2488 igfx_off [Default Off] 2489 By default, gfx is mapped as normal device. If a gfx 2490 device has a dedicated DMAR unit, the DMAR unit is 2491 bypassed by not enabling DMAR with this option. In 2492 this case, gfx device will use physical address for 2493 DMA. 2494 strict [Default Off] 2495 Deprecated, equivalent to iommu.strict=1. 2496 sp_off [Default Off] 2497 By default, super page will be supported if Intel IOMMU 2498 has the capability. With this option, super page will 2499 not be supported. 2500 sm_on 2501 Enable the Intel IOMMU scalable mode if the hardware 2502 advertises that it has support for the scalable mode 2503 translation. 2504 sm_off 2505 Disallow use of the Intel IOMMU scalable mode. 2506 tboot_noforce [Default Off] 2507 Do not force the Intel IOMMU enabled under tboot. 2508 By default, tboot will force Intel IOMMU on, which 2509 could harm performance of some high-throughput 2510 devices like 40GBit network cards, even if identity 2511 mapping is enabled. 2512 Note that using this option lowers the security 2513 provided by tboot because it makes the system 2514 vulnerable to DMA attacks. 2515 2516 intel_idle.max_cstate= [KNL,HW,ACPI,X86] 2517 0 disables intel_idle and fall back on acpi_idle. 2518 1 to 9 specify maximum depth of C-state. 2519 2520 intel_pstate= [X86,EARLY] 2521 disable 2522 Do not enable intel_pstate as the default 2523 scaling driver for the supported processors 2524 active 2525 Use intel_pstate driver to bypass the scaling 2526 governors layer of cpufreq and provides it own 2527 algorithms for p-state selection. There are two 2528 P-state selection algorithms provided by 2529 intel_pstate in the active mode: powersave and 2530 performance. The way they both operate depends 2531 on whether or not the hardware managed P-states 2532 (HWP) feature has been enabled in the processor 2533 and possibly on the processor model. 2534 passive 2535 Use intel_pstate as a scaling driver, but configure it 2536 to work with generic cpufreq governors (instead of 2537 enabling its internal governor). This mode cannot be 2538 used along with the hardware-managed P-states (HWP) 2539 feature. 2540 force 2541 Enable intel_pstate on systems that prohibit it by default 2542 in favor of acpi-cpufreq. Forcing the intel_pstate driver 2543 instead of acpi-cpufreq may disable platform features, such 2544 as thermal controls and power capping, that rely on ACPI 2545 P-States information being indicated to OSPM and therefore 2546 should be used with caution. This option does not work with 2547 processors that aren't supported by the intel_pstate driver 2548 or on platforms that use pcc-cpufreq instead of acpi-cpufreq. 2549 no_hwp 2550 Do not enable hardware P state control (HWP) 2551 if available. 2552 hwp_only 2553 Only load intel_pstate on systems which support 2554 hardware P state control (HWP) if available. 2555 support_acpi_ppc 2556 Enforce ACPI _PPC performance limits. If the Fixed ACPI 2557 Description Table, specifies preferred power management 2558 profile as "Enterprise Server" or "Performance Server", 2559 then this feature is turned on by default. 2560 per_cpu_perf_limits 2561 Allow per-logical-CPU P-State performance control limits using 2562 cpufreq sysfs interface 2563 no_cas 2564 Do not enable capacity-aware scheduling (CAS) on 2565 hybrid systems 2566 2567 intremap= [X86-64,Intel-IOMMU,EARLY] 2568 on enable Interrupt Remapping (default) 2569 off disable Interrupt Remapping 2570 nosid disable Source ID checking 2571 no_x2apic_optout 2572 BIOS x2APIC opt-out request will be ignored 2573 nopost disable Interrupt Posting 2574 posted_msi 2575 enable MSIs delivered as posted interrupts 2576 2577 iomem= Disable strict checking of access to MMIO memory 2578 strict regions from userspace. 2579 relaxed 2580 2581 iommu= [X86,EARLY] 2582 2583 off 2584 Don't initialize and use any kind of IOMMU. 2585 2586 force 2587 Force the use of the hardware IOMMU even when 2588 it is not actually needed (e.g. because < 3 GB 2589 memory). 2590 2591 noforce 2592 Don't force hardware IOMMU usage when it is not 2593 needed. (default). 2594 2595 merge 2596 Do scatter-gather (SG) merging. Implies "force" 2597 (experimental). 2598 2599 nomerge 2600 Don't do scatter-gather (SG) merging. 2601 2602 biomerge 2603 Do scatter-gather (SG) merging. Implies "force" 2604 (experimental). [same as "merge"] 2605 2606 panic 2607 Always panic when IOMMU overflows. 2608 2609 nopanic 2610 Don't panic on IOMMU overflows. 2611 2612 pt 2613 Use passththrough mode by default 2614 (Equivalent to iommu.passthrough=1) 2615 2616 nopt 2617 Use translated mode for DMA by default 2618 (Equivalent to iommu.passthrough=0) 2619 2620 soft 2621 Use software bounce buffering (SWIOTLB) (default for 2622 Intel machines). This can be used to prevent the usage 2623 of an available hardware IOMMU. 2624 2625 usedac 2626 Use the DAC on VIA PCI bridge 2627 (default: disable the VIA PCI bridge DAC) 2628 2629 AMD Gart HW IOMMU-specific options: (CONFIG_GART_IOMMU) 2630 2631 <size> 2632 Set the size of the remapping area in bytes. 2633 2634 allowed 2635 Overwrite iommu off workarounds for specific chipsets 2636 2637 force 2638 Overwrite iommu off workarounds for specific chipsets 2639 2640 fullflush 2641 Flush IOMMU on each allocation (default). 2642 2643 nofullflush 2644 Don't use IOMMU fullflush. 2645 2646 memaper[=<order>] 2647 Allocate an own aperture over RAM with size 2648 32MB<<order. (default: order=1, i.e. 64MB) 2649 2650 noaperture 2651 Ask the IOMMU not to touch the aperture for AGP. 2652 2653 noagp 2654 Don't initialize the AGP driver and use full aperture. 2655 2656 iommu= [PPC/POWERNV] 2657 2658 nobypass 2659 Disable IOMMU bypass, using IOMMU for PCI devices. 2660 2661 iommu.forcedac= [ARM64,X86,EARLY] Control IOVA allocation for PCI devices. 2662 Format: { "0" | "1" } 2663 0 - Try to allocate a 32-bit DMA address first, before 2664 falling back to the full range if needed. 2665 1 - Allocate directly from the full usable range, 2666 forcing Dual Address Cycle for PCI cards supporting 2667 greater than 32-bit addressing. 2668 2669 iommu.strict= [ARM64,X86,S390,EARLY] Configure TLB invalidation behaviour 2670 Format: { "0" | "1" } 2671 0 - Lazy mode. 2672 Request that DMA unmap operations use deferred 2673 invalidation of hardware TLBs, for increased 2674 throughput at the cost of reduced device isolation. 2675 Will fall back to strict mode if not supported by 2676 the relevant IOMMU driver. 2677 1 - Strict mode. 2678 DMA unmap operations invalidate IOMMU hardware TLBs 2679 synchronously. 2680 unset - Use value of CONFIG_IOMMU_DEFAULT_DMA_{LAZY,STRICT}. 2681 Note: on x86, strict mode specified via one of the 2682 legacy driver-specific options takes precedence. 2683 2684 iommu.passthrough= 2685 [ARM64,X86,EARLY] Configure DMA to bypass the IOMMU by default. 2686 Format: { "0" | "1" } 2687 0 - Use IOMMU translation for DMA. 2688 1 - Bypass the IOMMU for DMA. 2689 unset - Use value of CONFIG_IOMMU_DEFAULT_PASSTHROUGH. 2690 2691 iommu.debug_pagealloc= 2692 [KNL,EARLY] When CONFIG_IOMMU_DEBUG_PAGEALLOC is set, this 2693 parameter enables the feature at boot time. By default, it 2694 is disabled and the system behaves the same way as a kernel 2695 built without CONFIG_IOMMU_DEBUG_PAGEALLOC. 2696 Format: { "0" | "1" } 2697 0 - Sanitizer disabled. 2698 1 - Sanitizer enabled, expect runtime overhead. 2699 2700 io7= [HW] IO7 for Marvel-based Alpha systems 2701 See comment before marvel_specify_io7 in 2702 arch/alpha/kernel/core_marvel.c. 2703 2704 io_delay= [X86,EARLY] I/O delay method 2705 0x80 2706 Standard port 0x80 based delay 2707 0xed 2708 Alternate port 0xed based delay (needed on some systems) 2709 udelay 2710 Simple two microseconds delay 2711 none 2712 No delay 2713 2714 ip= [IP_PNP] 2715 See Documentation/admin-guide/nfs/nfsroot.rst. 2716 2717 ipcmni_extend [KNL,EARLY] Extend the maximum number of unique System V 2718 IPC identifiers from 32,768 to 16,777,216. 2719 2720 ipe.enforce= [IPE] 2721 Format: <bool> 2722 Determine whether IPE starts in permissive (0) or 2723 enforce (1) mode. The default is enforce. 2724 2725 ipe.success_audit= 2726 [IPE] 2727 Format: <bool> 2728 Start IPE with success auditing enabled, emitting 2729 an audit event when a binary is allowed. The default 2730 is 0. 2731 2732 irqaffinity= [SMP] Set the default irq affinity mask 2733 The argument is a cpu list, as described above. 2734 2735 irqchip.gicv2_force_probe= 2736 [ARM,ARM64,EARLY] 2737 Format: <bool> 2738 Force the kernel to look for the second 4kB page 2739 of a GICv2 controller even if the memory range 2740 exposed by the device tree is too small. 2741 2742 irqchip.gicv3_nolpi= 2743 [ARM,ARM64,EARLY] 2744 Force the kernel to ignore the availability of 2745 LPIs (and by consequence ITSs). Intended for system 2746 that use the kernel as a bootloader, and thus want 2747 to let secondary kernels in charge of setting up 2748 LPIs. 2749 2750 irqchip.gicv3_pseudo_nmi= [ARM64,EARLY] 2751 Enables support for pseudo-NMIs in the kernel. This 2752 requires the kernel to be built with 2753 CONFIG_ARM64_PSEUDO_NMI. 2754 2755 irqchip.riscv_imsic_noipi 2756 [RISC-V,EARLY] 2757 Force the kernel to not use IMSIC software injected MSIs 2758 as IPIs. Intended for system where IMSIC is trap-n-emulated, 2759 and thus want to reduce MMIO traps when triggering IPIs 2760 to multiple harts. 2761 2762 irqfixup [HW] 2763 When an interrupt is not handled search all handlers 2764 for it. Intended to get systems with badly broken 2765 firmware running. 2766 2767 irqhandler.duration_warn_us= [KNL] 2768 Warn if an IRQ handler exceeds the specified duration 2769 threshold in microseconds. Useful for identifying 2770 long-running IRQs in the system. 2771 2772 irqpoll [HW] 2773 When an interrupt is not handled search all handlers 2774 for it. Also check all handlers each timer 2775 interrupt. Intended to get systems with badly broken 2776 firmware running. 2777 2778 isapnp= [ISAPNP] 2779 Format: <RDP>,<reset>,<pci_scan>,<verbosity> 2780 2781 isolcpus= [KNL,SMP,ISOL] Isolate a given set of CPUs from disturbance. 2782 Format: [flag-list,]<cpu-list> 2783 2784 Specify one or more CPUs to isolate from disturbances 2785 specified in the flag list (default: domain): 2786 2787 nohz 2788 Disable the tick when a single task runs as well as 2789 disabling other kernel noises like having RCU callbacks 2790 offloaded. This is equivalent to the nohz_full parameter. 2791 2792 A residual 1Hz tick is offloaded to workqueues, which you 2793 need to affine to housekeeping through the global 2794 workqueue's affinity configured via the 2795 /sys/devices/virtual/workqueue/cpumask sysfs file, or 2796 by using the 'domain' flag described below. 2797 2798 NOTE: by default the global workqueue runs on all CPUs, 2799 so to protect individual CPUs the 'cpumask' file has to 2800 be configured manually after bootup. 2801 2802 domain 2803 Isolate from the general SMP balancing and scheduling 2804 algorithms. Note that performing domain isolation this way 2805 is irreversible: it's not possible to bring back a CPU to 2806 the domains once isolated through this boot time 2807 configuration. Use cpusets for a dynamic configuration 2808 which can be altered at runtime. For details see 2809 Documentation/admin-guide/cpu-isolation.rst. 2810 2811 You can move a process onto or off an "isolated" CPU via 2812 the CPU affinity syscalls or cpuset. 2813 <cpu number> begins at 0 and the maximum value is 2814 "number of CPUs in system - 1". 2815 2816 managed_irq 2817 2818 Isolate from being targeted by managed interrupts 2819 which have an interrupt mask containing isolated 2820 CPUs. The affinity of managed interrupts is 2821 handled by the kernel and cannot be changed via 2822 the /proc/irq/* interfaces. 2823 2824 This isolation is best effort and only effective 2825 if the automatically assigned interrupt mask of a 2826 device queue contains isolated and housekeeping 2827 CPUs. If housekeeping CPUs are online then such 2828 interrupts are directed to the housekeeping CPU 2829 so that IO submitted on the housekeeping CPU 2830 cannot disturb the isolated CPU. 2831 2832 If a queue's affinity mask contains only isolated 2833 CPUs then this parameter has no effect on the 2834 interrupt routing decision, though interrupts are 2835 only delivered when tasks running on those 2836 isolated CPUs submit IO. IO submitted on 2837 housekeeping CPUs has no influence on those 2838 queues. 2839 2840 The format of <cpu-list> is described above. 2841 2842 iucv= [HW,NET] 2843 2844 ivrs_ioapic [HW,X86-64] 2845 Provide an override to the IOAPIC-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 IOAPIC-ID decimal 10 to 2850 PCI segment 0x1 and PCI device 00:14.0, 2851 write the parameter as: 2852 ivrs_ioapic=10@0001:00:14.0 2853 2854 Deprecated formats: 2855 * To map IOAPIC-ID decimal 10 to PCI device 00:14.0 2856 write the parameter as: 2857 ivrs_ioapic[10]=00:14.0 2858 * To map IOAPIC-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_hpet [HW,X86-64] 2863 Provide an override to the HPET-ID<->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 HPET-ID decimal 10 to 2868 PCI segment 0x1 and PCI device 00:14.0, 2869 write the parameter as: 2870 ivrs_hpet=10@0001:00:14.0 2871 2872 Deprecated formats: 2873 * To map HPET-ID decimal 0 to PCI device 00:14.0 2874 write the parameter as: 2875 ivrs_hpet[0]=00:14.0 2876 * To map HPET-ID decimal 10 to PCI segment 0x1 and 2877 PCI device 00:14.0 write the parameter as: 2878 ivrs_ioapic[10]=0001:00:14.0 2879 2880 ivrs_acpihid [HW,X86-64] 2881 Provide an override to the ACPI-HID:UID<->DEVICE-ID 2882 mapping provided in the IVRS ACPI table. 2883 By default, PCI segment is 0, and can be omitted. 2884 2885 For example, to map UART-HID:UID AMD0020:0 to 2886 PCI segment 0x1 and PCI device ID 00:14.5, 2887 write the parameter as: 2888 ivrs_acpihid=AMD0020:0@0001:00:14.5 2889 2890 Deprecated formats: 2891 * To map UART-HID:UID AMD0020:0 to PCI segment is 0, 2892 PCI device ID 00:14.5, write the parameter as: 2893 ivrs_acpihid[00:14.5]=AMD0020:0 2894 * To map UART-HID:UID AMD0020:0 to PCI segment 0x1 and 2895 PCI device ID 00:14.5, write the parameter as: 2896 ivrs_acpihid[0001:00:14.5]=AMD0020:0 2897 2898 js= [HW,JOY] Analog joystick 2899 See Documentation/input/joydev/joystick.rst. 2900 2901 kasan_multi_shot 2902 [KNL] Enforce KASAN (Kernel Address Sanitizer) to print 2903 report on every invalid memory access. Without this 2904 parameter KASAN will print report only for the first 2905 invalid access. 2906 2907 keep_bootcon [KNL,EARLY] 2908 Do not unregister boot console at start. This is only 2909 useful for debugging when something happens in the window 2910 between unregistering the boot console and initializing 2911 the real console. 2912 2913 keepinitrd [HW,ARM] See retain_initrd. 2914 2915 kernelcore= [KNL,X86,PPC,EARLY] 2916 Format: nn[KMGTPE] | nn% | "mirror" 2917 This parameter specifies the amount of memory usable by 2918 the kernel for non-movable allocations. The requested 2919 amount is spread evenly throughout all nodes in the 2920 system as ZONE_NORMAL. The remaining memory is used for 2921 movable memory in its own zone, ZONE_MOVABLE. In the 2922 event, a node is too small to have both ZONE_NORMAL and 2923 ZONE_MOVABLE, kernelcore memory will take priority and 2924 other nodes will have a larger ZONE_MOVABLE. 2925 2926 ZONE_MOVABLE is used for the allocation of pages that 2927 may be reclaimed or moved by the page migration 2928 subsystem. Note that allocations like PTEs-from-HighMem 2929 still use the HighMem zone if it exists, and the Normal 2930 zone if it does not. 2931 2932 It is possible to specify the exact amount of memory in 2933 the form of "nn[KMGTPE]", a percentage of total system 2934 memory in the form of "nn%", or "mirror". If "mirror" 2935 option is specified, mirrored (reliable) memory is used 2936 for non-movable allocations and remaining memory is used 2937 for Movable pages. "nn[KMGTPE]", "nn%", and "mirror" 2938 are exclusive, so you cannot specify multiple forms. 2939 2940 kfence.burst= [MM,KFENCE] The number of additional successive 2941 allocations to be attempted through KFENCE for each 2942 sample interval. 2943 Format: <unsigned integer> 2944 Default: 0 2945 2946 kfence.check_on_panic= 2947 [MM,KFENCE] Whether to check all KFENCE-managed objects' 2948 canaries on panic. 2949 Format: <bool> 2950 Default: false 2951 2952 kfence.deferrable= 2953 [MM,KFENCE] Whether to use a deferrable timer to trigger 2954 allocations. This avoids forcing CPU wake-ups if the 2955 system is idle, at the risk of a less predictable 2956 sample interval. 2957 Format: <bool> 2958 Default: CONFIG_KFENCE_DEFERRABLE 2959 2960 kfence.fault= [MM,KFENCE] Controls the behavior when a KFENCE 2961 error is detected. 2962 report - print the error report and continue (default). 2963 oops - print the error report and oops. 2964 panic - print the error report and panic. 2965 2966 kfence.sample_interval= 2967 [MM,KFENCE] KFENCE's sample interval in milliseconds. 2968 Format: <unsigned integer> 2969 0 - Disable KFENCE. 2970 >0 - Enabled KFENCE with given sample interval. 2971 Default: CONFIG_KFENCE_SAMPLE_INTERVAL 2972 2973 kfence.skip_covered_thresh= 2974 [MM,KFENCE] If pool utilization reaches this threshold 2975 (pool usage%), KFENCE limits currently covered 2976 allocations of the same source from further filling 2977 up the pool. 2978 Format: <unsigned integer> 2979 Default: 75 2980 2981 kgdbdbgp= [KGDB,HW,EARLY] kgdb over EHCI usb debug port. 2982 Format: <Controller#>[,poll interval] 2983 The controller # is the number of the ehci usb debug 2984 port as it is probed via PCI. The poll interval is 2985 optional and is the number seconds in between 2986 each poll cycle to the debug port in case you need 2987 the functionality for interrupting the kernel with 2988 gdb or control-c on the dbgp connection. When 2989 not using this parameter you use sysrq-g to break into 2990 the kernel debugger. 2991 2992 kgdboc= [KGDB,HW] kgdb over consoles. 2993 Requires a tty driver that supports console polling, 2994 or a supported polling keyboard driver (non-usb). 2995 Serial only format: <serial_device>[,baud] 2996 keyboard only format: kbd 2997 keyboard and serial format: kbd,<serial_device>[,baud] 2998 Optional Kernel mode setting: 2999 kms, kbd format: kms,kbd 3000 kms, kbd and serial format: kms,kbd,<ser_dev>[,baud] 3001 3002 kgdboc_earlycon= [KGDB,HW,EARLY] 3003 If the boot console provides the ability to read 3004 characters and can work in polling mode, you can use 3005 this parameter to tell kgdb to use it as a backend 3006 until the normal console is registered. Intended to 3007 be used together with the kgdboc parameter which 3008 specifies the normal console to transition to. 3009 3010 The name of the early console should be specified 3011 as the value of this parameter. Note that the name of 3012 the early console might be different than the tty 3013 name passed to kgdboc. It's OK to leave the value 3014 blank and the first boot console that implements 3015 read() will be picked. 3016 3017 kgdbwait [KGDB,EARLY] Stop kernel execution and enter the 3018 kernel debugger at the earliest opportunity. 3019 3020 kho= [KEXEC,EARLY] 3021 Format: { "0" | "1" | "off" | "on" | "y" | "n" } 3022 Enables or disables Kexec HandOver. 3023 "0" | "off" | "n" - kexec handover is disabled 3024 "1" | "on" | "y" - kexec handover is enabled 3025 3026 kho_scratch= [KEXEC,EARLY] 3027 Format: ll[KMG],mm[KMG],nn[KMG] | nn% 3028 Defines the size of the KHO scratch region. The KHO 3029 scratch regions are physically contiguous memory 3030 ranges that can only be used for non-kernel 3031 allocations. That way, even when memory is heavily 3032 fragmented with handed over memory, the kexeced 3033 kernel will always have enough contiguous ranges to 3034 bootstrap itself. 3035 3036 It is possible to specify the exact amount of 3037 memory in the form of "ll[KMG],mm[KMG],nn[KMG]" 3038 where the first parameter defines the size of a low 3039 memory scratch area, the second parameter defines 3040 the size of a global scratch area and the third 3041 parameter defines the size of additional per-node 3042 scratch areas. The form "nn%" defines scale factor 3043 (in percents) of memory that was used during boot. 3044 3045 kmac= [MIPS] Korina ethernet MAC address. 3046 Configure the RouterBoard 532 series on-chip 3047 Ethernet adapter MAC address. 3048 3049 kmemleak= [KNL,EARLY] Boot-time kmemleak enable/disable 3050 Valid arguments: on, off 3051 Default: on 3052 Built with CONFIG_DEBUG_KMEMLEAK_DEFAULT_OFF=y, 3053 the default is off. 3054 3055 kprobe_event=[probe-list] 3056 [FTRACE] Add kprobe events and enable at boot time. 3057 The probe-list is a semicolon delimited list of probe 3058 definitions. Each definition is same as kprobe_events 3059 interface, but the parameters are comma delimited. 3060 For example, to add a kprobe event on vfs_read with 3061 arg1 and arg2, add to the command line; 3062 3063 kprobe_event=p,vfs_read,$arg1,$arg2 3064 3065 See also Documentation/trace/kprobetrace.rst "Kernel 3066 Boot Parameter" section. 3067 3068 kpti= [ARM64,EARLY] Control page table isolation of 3069 user and kernel address spaces. 3070 Default: enabled on cores which need mitigation. 3071 0: force disabled 3072 1: force enabled 3073 3074 kunit.enable= [KUNIT] Enable executing KUnit tests. Requires 3075 CONFIG_KUNIT to be set to be fully enabled. The 3076 default value can be overridden via 3077 KUNIT_DEFAULT_ENABLED. 3078 Default is 1 (enabled) 3079 3080 kvm.ignore_msrs=[KVM] Ignore guest accesses to unhandled MSRs. 3081 Default is 0 (don't ignore, but inject #GP) 3082 3083 kvm.eager_page_split= 3084 [KVM,X86] Controls whether or not KVM will try to 3085 proactively split all huge pages during dirty logging. 3086 Eager page splitting reduces interruptions to vCPU 3087 execution by eliminating the write-protection faults 3088 and MMU lock contention that would otherwise be 3089 required to split huge pages lazily. 3090 3091 VM workloads that rarely perform writes or that write 3092 only to a small region of VM memory may benefit from 3093 disabling eager page splitting to allow huge pages to 3094 still be used for reads. 3095 3096 The behavior of eager page splitting depends on whether 3097 KVM_DIRTY_LOG_INITIALLY_SET is enabled or disabled. If 3098 disabled, all huge pages in a memslot will be eagerly 3099 split when dirty logging is enabled on that memslot. If 3100 enabled, eager page splitting will be performed during 3101 the KVM_CLEAR_DIRTY ioctl, and only for the pages being 3102 cleared. 3103 3104 Eager page splitting is only supported when kvm.tdp_mmu=Y. 3105 3106 Default is Y (on). 3107 3108 kvm.enable_pmu=[KVM,X86] 3109 If enabled, KVM will virtualize PMU functionality based 3110 on the virtual CPU model defined by userspace. This 3111 can be overridden on a per-VM basis via 3112 KVM_CAP_PMU_CAPABILITY. 3113 3114 If disabled, KVM will not virtualize PMU functionality, 3115 e.g. MSRs, PMCs, PMIs, etc., even if userspace defines 3116 a virtual CPU model that contains PMU assets. 3117 3118 Note, KVM's vPMU support implicitly requires running 3119 with an in-kernel local APIC, e.g. to deliver PMIs to 3120 the guest. Running without an in-kernel local APIC is 3121 not supported, though KVM will allow such a combination 3122 (with severely degraded functionality). 3123 3124 See also enable_mediated_pmu. 3125 3126 Default is Y (on). 3127 3128 kvm.enable_virt_at_load=[KVM,ARM64,LOONGARCH,MIPS,RISCV,X86] 3129 If enabled, KVM will enable virtualization in hardware 3130 when KVM is loaded, and disable virtualization when KVM 3131 is unloaded (if KVM is built as a module). 3132 3133 If disabled, KVM will dynamically enable and disable 3134 virtualization on-demand when creating and destroying 3135 VMs, i.e. on the 0=>1 and 1=>0 transitions of the 3136 number of VMs. 3137 3138 Enabling virtualization at module load avoids potential 3139 latency for creation of the 0=>1 VM, as KVM serializes 3140 virtualization enabling across all online CPUs. The 3141 "cost" of enabling virtualization when KVM is loaded, 3142 is that doing so may interfere with using out-of-tree 3143 hypervisors that want to "own" virtualization hardware. 3144 3145 kvm.enable_vmware_backdoor=[KVM] Support VMware backdoor PV interface. 3146 Default is false (don't support). 3147 3148 kvm.nx_huge_pages= 3149 [KVM] Controls the software workaround for the 3150 X86_BUG_ITLB_MULTIHIT bug. 3151 force : Always deploy workaround. 3152 off : Never deploy workaround. 3153 auto : Deploy workaround based on the presence of 3154 X86_BUG_ITLB_MULTIHIT. 3155 3156 Default is 'auto'. 3157 3158 If the software workaround is enabled for the host, 3159 guests do need not to enable it for nested guests. 3160 3161 kvm.nx_huge_pages_recovery_ratio= 3162 [KVM] Controls how many 4KiB pages are periodically zapped 3163 back to huge pages. 0 disables the recovery, otherwise if 3164 the value is N KVM will zap 1/Nth of the 4KiB pages every 3165 period (see below). The default is 60. 3166 3167 kvm.nx_huge_pages_recovery_period_ms= 3168 [KVM] Controls the time period at which KVM zaps 4KiB pages 3169 back to huge pages. If the value is a non-zero N, KVM will 3170 zap a portion (see ratio above) of the pages every N msecs. 3171 If the value is 0 (the default), KVM will pick a period based 3172 on the ratio, such that a page is zapped after 1 hour on average. 3173 3174 kvm-{amd,intel}.enable_mediated_pmu=[KVM,AMD,INTEL] 3175 If enabled, KVM will provide a mediated virtual PMU, 3176 instead of the default perf-based virtual PMU (if 3177 kvm.enable_pmu is true and PMU is enumerated via the 3178 virtual CPU model). 3179 3180 With a perf-based vPMU, KVM operates as a user of perf, 3181 i.e. emulates guest PMU counters using perf events. 3182 KVM-created perf events are managed by perf as regular 3183 (guest-only) events, e.g. are scheduled in/out, contend 3184 for hardware resources, etc. Using a perf-based vPMU 3185 allows guest and host usage of the PMU to co-exist, but 3186 incurs non-trivial overhead and can result in silently 3187 dropped guest events (due to resource contention). 3188 3189 With a mediated vPMU, hardware PMU state is context 3190 switched around the world switch to/from the guest. 3191 KVM mediates which events the guest can utilize, but 3192 gives the guest direct access to all other PMU assets 3193 when possible (KVM may intercept some accesses if the 3194 virtual CPU model provides a subset of hardware PMU 3195 functionality). Using a mediated vPMU significantly 3196 reduces PMU virtualization overhead and eliminates lost 3197 guest events, but is mutually exclusive with using perf 3198 to profile KVM guests and adds latency to most VM-Exits 3199 (to context switch PMU state). 3200 3201 Default is N (off). 3202 3203 kvm-amd.nested= [KVM,AMD] Control nested virtualization feature in 3204 KVM/SVM. Default is 1 (enabled). 3205 3206 kvm-amd.npt= [KVM,AMD] Control KVM's use of Nested Page Tables, 3207 a.k.a. Two-Dimensional Page Tables. Default is 1 3208 (enabled). Disable by KVM if hardware lacks support 3209 for NPT. 3210 3211 kvm-amd.ciphertext_hiding_asids= 3212 [KVM,AMD] Ciphertext hiding prevents disallowed accesses 3213 to SNP private memory from reading ciphertext. Instead, 3214 reads will see constant default values (0xff). 3215 3216 If ciphertext hiding is enabled, the joint SEV-ES and 3217 SEV-SNP ASID space is partitioned into separate SEV-ES 3218 and SEV-SNP ASID ranges, with the SEV-SNP range being 3219 [1..max_snp_asid] and the SEV-ES range being 3220 (max_snp_asid..min_sev_asid), where min_sev_asid is 3221 enumerated by CPUID.0x.8000_001F[EDX]. 3222 3223 A non-zero value enables SEV-SNP ciphertext hiding and 3224 adjusts the ASID ranges for SEV-ES and SEV-SNP guests. 3225 KVM caps the number of SEV-SNP ASIDs at the maximum 3226 possible value, e.g. specifying -1u will assign all 3227 joint SEV-ES and SEV-SNP ASIDs to SEV-SNP. Note, 3228 assigning all joint ASIDs to SEV-SNP, i.e. configuring 3229 max_snp_asid == min_sev_asid-1, will effectively make 3230 SEV-ES unusable. 3231 3232 kvm-arm.mode= 3233 [KVM,ARM,EARLY] Select one of KVM/arm64's modes of 3234 operation. 3235 3236 none: Forcefully disable KVM. 3237 3238 nvhe: Standard nVHE-based mode, without support for 3239 protected guests. 3240 3241 protected: Mode with support for guests whose state is 3242 kept private from the host, using VHE or 3243 nVHE depending on HW support. 3244 3245 nested: VHE-based mode with support for nested 3246 virtualization. Requires at least ARMv8.4 3247 hardware (with FEAT_NV2). 3248 3249 Defaults to VHE/nVHE based on hardware support. Setting 3250 mode to "protected" will disable kexec and hibernation 3251 for the host. To force nVHE on VHE hardware, add 3252 "arm64_sw.hvhe=0 id_aa64mmfr1.vh=0" to the 3253 command-line. 3254 "nested" and "protected" are experimental and should be 3255 used with extreme caution. 3256 3257 kvm-arm.vgic_v3_group0_trap= 3258 [KVM,ARM,EARLY] Trap guest accesses to GICv3 group-0 3259 system registers 3260 3261 kvm-arm.vgic_v3_group1_trap= 3262 [KVM,ARM,EARLY] Trap guest accesses to GICv3 group-1 3263 system registers 3264 3265 kvm-arm.vgic_v3_common_trap= 3266 [KVM,ARM,EARLY] Trap guest accesses to GICv3 common 3267 system registers 3268 3269 kvm-arm.vgic_v4_enable= 3270 [KVM,ARM,EARLY] Allow use of GICv4 for direct 3271 injection of LPIs. 3272 3273 kvm-arm.wfe_trap_policy= 3274 [KVM,ARM] Control when to set WFE instruction trap for 3275 KVM VMs. Traps are allowed but not guaranteed by the 3276 CPU architecture. 3277 3278 trap: set WFE instruction trap 3279 3280 notrap: clear WFE instruction trap 3281 3282 kvm-arm.wfi_trap_policy= 3283 [KVM,ARM] Control when to set WFI instruction trap for 3284 KVM VMs. Traps are allowed but not guaranteed by the 3285 CPU architecture. 3286 3287 trap: set WFI instruction trap 3288 3289 notrap: clear WFI instruction trap 3290 3291 kvm_cma_resv_ratio=n [PPC,EARLY] 3292 Reserves given percentage from system memory area for 3293 contiguous memory allocation for KVM hash pagetable 3294 allocation. 3295 By default it reserves 5% of total system memory. 3296 Format: <integer> 3297 Default: 5 3298 3299 kvm-intel.ept= [KVM,Intel] Control KVM's use of Extended Page Tables, 3300 a.k.a. Two-Dimensional Page Tables. Default is 1 3301 (enabled). Disable by KVM if hardware lacks support 3302 for EPT. 3303 3304 kvm-intel.emulate_invalid_guest_state= 3305 [KVM,Intel] Control whether to emulate invalid guest 3306 state. Ignored if kvm-intel.enable_unrestricted_guest=1, 3307 as guest state is never invalid for unrestricted 3308 guests. This param doesn't apply to nested guests (L2), 3309 as KVM never emulates invalid L2 guest state. 3310 Default is 1 (enabled). 3311 3312 kvm-intel.flexpriority= 3313 [KVM,Intel] Control KVM's use of FlexPriority feature 3314 (TPR shadow). Default is 1 (enabled). Disable by KVM if 3315 hardware lacks support for it. 3316 3317 kvm-intel.nested= 3318 [KVM,Intel] Control nested virtualization feature in 3319 KVM/VMX. Default is 1 (enabled). 3320 3321 kvm-intel.unrestricted_guest= 3322 [KVM,Intel] Control KVM's use of unrestricted guest 3323 feature (virtualized real and unpaged mode). Default 3324 is 1 (enabled). Disable by KVM if EPT is disabled or 3325 hardware lacks support for it. 3326 3327 kvm-intel.vmentry_l1d_flush=[KVM,Intel] Mitigation for L1 Terminal Fault 3328 CVE-2018-3620. 3329 3330 Valid arguments: never, cond, always 3331 3332 always: L1D cache flush on every VMENTER. 3333 cond: Flush L1D on VMENTER only when the code between 3334 VMEXIT and VMENTER can leak host memory. 3335 never: Disables the mitigation 3336 3337 Default is cond (do L1 cache flush in specific instances) 3338 3339 kvm-intel.vpid= [KVM,Intel] Control KVM's use of Virtual Processor 3340 Identification feature (tagged TLBs). Default is 1 3341 (enabled). Disable by KVM if hardware lacks support 3342 for it. 3343 3344 l1d_flush= [X86,INTEL,EARLY] 3345 Control mitigation for L1D based snooping vulnerability. 3346 3347 Certain CPUs are vulnerable to an exploit against CPU 3348 internal buffers which can forward information to a 3349 disclosure gadget under certain conditions. 3350 3351 In vulnerable processors, the speculatively 3352 forwarded data can be used in a cache side channel 3353 attack, to access data to which the attacker does 3354 not have direct access. 3355 3356 This parameter controls the mitigation. The 3357 options are: 3358 3359 on - enable the interface for the mitigation 3360 3361 l1tf= [X86,EARLY] Control mitigation of the L1TF vulnerability on 3362 affected CPUs 3363 3364 The kernel PTE inversion protection is unconditionally 3365 enabled and cannot be disabled. 3366 3367 full 3368 Provides all available mitigations for the 3369 L1TF vulnerability. Disables SMT and 3370 enables all mitigations in the 3371 hypervisors, i.e. unconditional L1D flush. 3372 3373 SMT control and L1D flush control via the 3374 sysfs interface is still possible after 3375 boot. Hypervisors will issue a warning 3376 when the first VM is started in a 3377 potentially insecure configuration, 3378 i.e. SMT enabled or L1D flush disabled. 3379 3380 full,force 3381 Same as 'full', but disables SMT and L1D 3382 flush runtime control. Implies the 3383 'nosmt=force' command line option. 3384 (i.e. sysfs control of SMT is disabled.) 3385 3386 flush 3387 Leaves SMT enabled and enables the default 3388 hypervisor mitigation, i.e. conditional 3389 L1D flush. 3390 3391 SMT control and L1D flush control via the 3392 sysfs interface is still possible after 3393 boot. Hypervisors will issue a warning 3394 when the first VM is started in a 3395 potentially insecure configuration, 3396 i.e. SMT enabled or L1D flush disabled. 3397 3398 flush,nosmt 3399 3400 Disables SMT and enables the default 3401 hypervisor mitigation. 3402 3403 SMT control and L1D flush control via the 3404 sysfs interface is still possible after 3405 boot. Hypervisors will issue a warning 3406 when the first VM is started in a 3407 potentially insecure configuration, 3408 i.e. SMT enabled or L1D flush disabled. 3409 3410 flush,nowarn 3411 Same as 'flush', but hypervisors will not 3412 warn when a VM is started in a potentially 3413 insecure configuration. 3414 3415 off 3416 Disables hypervisor mitigations and doesn't 3417 emit any warnings. 3418 It also drops the swap size and available 3419 RAM limit restriction on both hypervisor and 3420 bare metal. 3421 3422 Default is 'flush'. 3423 3424 For details see: Documentation/admin-guide/hw-vuln/l1tf.rst 3425 3426 l2cr= [PPC] 3427 3428 l3cr= [PPC] 3429 3430 lapic [X86-32,APIC,EARLY] Enable the local APIC even if BIOS 3431 disabled it. 3432 3433 lapic= [X86,APIC] Do not use TSC deadline 3434 value for LAPIC timer one-shot implementation. Default 3435 back to the programmable timer unit in the LAPIC. 3436 Format: notscdeadline 3437 3438 lapic_timer_c2_ok [X86,APIC,EARLY] trust the local apic timer 3439 in C2 power state. 3440 3441 libata.dma= [LIBATA] DMA control 3442 libata.dma=0 Disable all PATA and SATA DMA 3443 libata.dma=1 PATA and SATA Disk DMA only 3444 libata.dma=2 ATAPI (CDROM) DMA only 3445 libata.dma=4 Compact Flash DMA only 3446 Combinations also work, so libata.dma=3 enables DMA 3447 for disks and CDROMs, but not CFs. 3448 3449 libata.ignore_hpa= [LIBATA] Ignore HPA limit 3450 libata.ignore_hpa=0 keep BIOS limits (default) 3451 libata.ignore_hpa=1 ignore limits, using full disk 3452 3453 libata.noacpi [LIBATA] Disables use of ACPI in libata suspend/resume 3454 when set. 3455 Format: <int> 3456 3457 libata.force= [LIBATA] Force configurations. The format is a comma- 3458 separated list of "[ID:]VAL" where ID is PORT[.DEVICE]. 3459 PORT and DEVICE are decimal numbers matching port, link 3460 or device. Basically, it matches the ATA ID string 3461 printed on console by libata. If the whole ID part is 3462 omitted, the last PORT and DEVICE values are used. If 3463 ID hasn't been specified yet, the configuration applies 3464 to all ports, links and devices. 3465 3466 If only DEVICE is omitted, the parameter applies to 3467 the port and all links and devices behind it. DEVICE 3468 number of 0 either selects the first device or the 3469 first fan-out link behind PMP device. It does not 3470 select the host link. DEVICE number of 15 selects the 3471 host link and device attached to it. 3472 3473 The VAL specifies the configuration to force. As long 3474 as there is no ambiguity, shortcut notation is allowed. 3475 For example, both 1.5 and 1.5G would work for 1.5Gbps. 3476 The following configurations can be forced. 3477 3478 * Cable type: 40c, 80c, short40c, unk, ign or sata. 3479 Any ID with matching PORT is used. 3480 3481 * SATA link speed limit: 1.5Gbps or 3.0Gbps. 3482 3483 * Transfer mode: pio[0-7], mwdma[0-4] and udma[0-7]. 3484 udma[/][16,25,33,44,66,100,133] notation is also 3485 allowed. 3486 3487 * nohrst, nosrst, norst: suppress hard, soft and both 3488 resets. 3489 3490 * rstonce: only attempt one reset during hot-unplug 3491 link recovery. 3492 3493 * [no]dbdelay: Enable or disable the extra 200ms delay 3494 before debouncing a link PHY and device presence 3495 detection. 3496 3497 * [no]ncq: Turn on or off NCQ. 3498 3499 * [no]ncqtrim: Enable or disable queued DSM TRIM. 3500 3501 * [no]ncqati: Enable or disable NCQ trim on ATI chipset. 3502 3503 * [no]trim: Enable or disable (unqueued) TRIM. 3504 3505 * trim_zero: Indicate that TRIM command zeroes data. 3506 3507 * max_trim_128m: Set 128M maximum trim size limit. 3508 3509 * [no]dma: Turn on or off DMA transfers. 3510 3511 * atapi_dmadir: Enable ATAPI DMADIR bridge support. 3512 3513 * atapi_mod16_dma: Enable the use of ATAPI DMA for 3514 commands that are not a multiple of 16 bytes. 3515 3516 * [no]dmalog: Enable or disable the use of the 3517 READ LOG DMA EXT command to access logs. 3518 3519 * [no]iddevlog: Enable or disable access to the 3520 identify device data log. 3521 3522 * [no]logdir: Enable or disable access to the general 3523 purpose log directory. 3524 3525 * max_sec=<sectors>: Set the transfer size limit, in 3526 number of 512-byte sectors, to the value specified in 3527 <sectors>. The value specified in <sectors> has to be 3528 a non-zero positive integer. 3529 3530 * max_sec_128: Set transfer size limit to 128 sectors. 3531 3532 * max_sec_1024: Set or clear transfer size limit to 3533 1024 sectors. 3534 3535 * max_sec_lba48: Set or clear transfer size limit to 3536 65535 sectors. 3537 3538 * external: Mark port as external (hotplug-capable). 3539 3540 * [no]lpm: Enable or disable link power management. 3541 3542 * [no]setxfer: Indicate if transfer speed mode setting 3543 should be skipped. 3544 3545 * [no]fua: Disable or enable FUA (Force Unit Access) 3546 support for devices supporting this feature. 3547 3548 * dump_id: Dump IDENTIFY data. 3549 3550 * disable: Disable this device. 3551 3552 If there are multiple matching configurations changing 3553 the same attribute, the last one is used. 3554 3555 liveupdate= [KNL,EARLY] 3556 Format: <bool> 3557 Enable Live Update Orchestrator (LUO). 3558 Default: off. 3559 3560 lockd.nlm_grace_period=P [NFS] Assign grace period. 3561 Format: <integer> 3562 3563 lockd.nlm_tcpport=N [NFS] Assign TCP port. 3564 Format: <integer> 3565 3566 lockd.nlm_timeout=T [NFS] Assign timeout value. 3567 Format: <integer> 3568 3569 lockd.nlm_udpport=M [NFS] Assign UDP port. 3570 Format: <integer> 3571 3572 lockdown= [SECURITY,EARLY] 3573 { integrity | confidentiality } 3574 Enable the kernel lockdown feature. If set to 3575 integrity, kernel features that allow userland to 3576 modify the running kernel are disabled. If set to 3577 confidentiality, kernel features that allow userland 3578 to extract confidential information from the kernel 3579 are also disabled. 3580 3581 locktorture.acq_writer_lim= [KNL] 3582 Set the time limit in jiffies for a lock 3583 acquisition. Acquisitions exceeding this limit 3584 will result in a splat once they do complete. 3585 3586 locktorture.bind_readers= [KNL] 3587 Specify the list of CPUs to which the readers are 3588 to be bound. 3589 3590 locktorture.bind_writers= [KNL] 3591 Specify the list of CPUs to which the writers are 3592 to be bound. 3593 3594 locktorture.call_rcu_chains= [KNL] 3595 Specify the number of self-propagating call_rcu() 3596 chains to set up. These are used to ensure that 3597 there is a high probability of an RCU grace period 3598 in progress at any given time. Defaults to 0, 3599 which disables these call_rcu() chains. 3600 3601 locktorture.long_hold= [KNL] 3602 Specify the duration in milliseconds for the 3603 occasional long-duration lock hold time. Defaults 3604 to 100 milliseconds. Select 0 to disable. 3605 3606 locktorture.nested_locks= [KNL] 3607 Specify the maximum lock nesting depth that 3608 locktorture is to exercise, up to a limit of 8 3609 (MAX_NESTED_LOCKS). Specify zero to disable. 3610 Note that this parameter is ineffective on types 3611 of locks that do not support nested acquisition. 3612 3613 locktorture.nreaders_stress= [KNL] 3614 Set the number of locking read-acquisition kthreads. 3615 Defaults to being automatically set based on the 3616 number of online CPUs. 3617 3618 locktorture.nwriters_stress= [KNL] 3619 Set the number of locking write-acquisition kthreads. 3620 3621 locktorture.onoff_holdoff= [KNL] 3622 Set time (s) after boot for CPU-hotplug testing. 3623 3624 locktorture.onoff_interval= [KNL] 3625 Set time (s) between CPU-hotplug operations, or 3626 zero to disable CPU-hotplug testing. 3627 3628 locktorture.rt_boost= [KNL] 3629 Do periodic testing of real-time lock priority 3630 boosting. Select 0 to disable, 1 to boost 3631 only rt_mutex, and 2 to boost unconditionally. 3632 Defaults to 2, which might seem to be an 3633 odd choice, but which should be harmless for 3634 non-real-time spinlocks, due to their disabling 3635 of preemption. Note that non-realtime mutexes 3636 disable boosting. 3637 3638 locktorture.rt_boost_factor= [KNL] 3639 Number that determines how often and for how 3640 long priority boosting is exercised. This is 3641 scaled down by the number of writers, so that the 3642 number of boosts per unit time remains roughly 3643 constant as the number of writers increases. 3644 On the other hand, the duration of each boost 3645 increases with the number of writers. 3646 3647 locktorture.shuffle_interval= [KNL] 3648 Set task-shuffle interval (jiffies). Shuffling 3649 tasks allows some CPUs to go into dyntick-idle 3650 mode during the locktorture test. 3651 3652 locktorture.shutdown_secs= [KNL] 3653 Set time (s) after boot system shutdown. This 3654 is useful for hands-off automated testing. 3655 3656 locktorture.stat_interval= [KNL] 3657 Time (s) between statistics printk()s. 3658 3659 locktorture.stutter= [KNL] 3660 Time (s) to stutter testing, for example, 3661 specifying five seconds causes the test to run for 3662 five seconds, wait for five seconds, and so on. 3663 This tests the locking primitive's ability to 3664 transition abruptly to and from idle. 3665 3666 locktorture.torture_type= [KNL] 3667 Specify the locking implementation to test. 3668 3669 locktorture.verbose= [KNL] 3670 Enable additional printk() statements. 3671 3672 locktorture.writer_fifo= [KNL] 3673 Run the write-side locktorture kthreads at 3674 sched_set_fifo() real-time priority. 3675 3676 logibm.irq= [HW,MOUSE] Logitech Bus Mouse Driver 3677 Format: <irq> 3678 3679 loglevel= [KNL,EARLY] 3680 All Kernel Messages with a loglevel smaller than the 3681 console loglevel will be printed to the console. It can 3682 also be changed with klogd or other programs. The 3683 loglevels are defined as follows: 3684 3685 0 (KERN_EMERG) system is unusable 3686 1 (KERN_ALERT) action must be taken immediately 3687 2 (KERN_CRIT) critical conditions 3688 3 (KERN_ERR) error conditions 3689 4 (KERN_WARNING) warning conditions 3690 5 (KERN_NOTICE) normal but significant condition 3691 6 (KERN_INFO) informational 3692 7 (KERN_DEBUG) debug-level messages 3693 3694 log_buf_len=n[KMG] [KNL,EARLY] 3695 Sets the size of the printk ring buffer, in bytes. 3696 n must be a power of two and greater than the 3697 minimal size. The minimal size is defined by 3698 LOG_BUF_SHIFT kernel config parameter. There 3699 is also CONFIG_LOG_CPU_MAX_BUF_SHIFT config 3700 parameter that allows to increase the default size 3701 depending on the number of CPUs. See init/Kconfig 3702 for more details. 3703 3704 logo.nologo [FB] Disables display of the built-in Linux logo. 3705 This may be used to provide more screen space for 3706 kernel log messages and is useful when debugging 3707 kernel boot problems. 3708 3709 lp=0 [LP] Specify parallel ports to use, e.g, 3710 lp=port[,port...] lp=none,parport0 (lp0 not configured, lp1 uses 3711 lp=reset first parallel port). 'lp=0' disables the 3712 lp=auto printer driver. 'lp=reset' (which can be 3713 specified in addition to the ports) causes 3714 attached printers to be reset. Using 3715 lp=port1,port2,... specifies the parallel ports 3716 to associate lp devices with, starting with 3717 lp0. A port specification may be 'none' to skip 3718 that lp device, or a parport name such as 3719 'parport0'. Specifying 'lp=auto' instead of a 3720 port specification list means that device IDs 3721 from each port should be examined, to see if 3722 an IEEE 1284-compliant printer is attached; if 3723 so, the driver will manage that printer. 3724 See also header of drivers/char/lp.c. 3725 3726 lpj=n [KNL] 3727 Sets loops_per_jiffy to given constant, thus avoiding 3728 time-consuming boot-time autodetection (up to 250 ms per 3729 CPU). 0 enables autodetection (default). To determine 3730 the correct value for your kernel, boot with normal 3731 autodetection and see what value is printed. Note that 3732 on SMP systems the preset will be applied to all CPUs, 3733 which is likely to cause problems if your CPUs need 3734 significantly divergent settings. An incorrect value 3735 will cause delays in the kernel to be wrong, leading to 3736 unpredictable I/O errors and other breakage. Although 3737 unlikely, in the extreme case this might damage your 3738 hardware. 3739 3740 lsm.debug [SECURITY] Enable LSM initialization debugging output. 3741 3742 lsm=lsm1,...,lsmN 3743 [SECURITY] Choose order of LSM initialization. This 3744 overrides CONFIG_LSM, and the "security=" parameter. 3745 3746 machtype= [Loongson] Share the same kernel image file between 3747 different yeeloong laptops. 3748 Example: machtype=lemote-yeeloong-2f-7inch 3749 3750 maxcpus= [SMP,EARLY] Maximum number of processors that an SMP kernel 3751 will bring up during bootup. maxcpus=n : n >= 0 limits 3752 the kernel to bring up 'n' processors. Surely after 3753 bootup you can bring up the other plugged cpu by executing 3754 "echo 1 > /sys/devices/system/cpu/cpuX/online". So maxcpus 3755 only takes effect during system bootup. 3756 While n=0 is a special case, it is equivalent to "nosmp", 3757 which also disables the IO APIC. 3758 3759 max_loop= [LOOP] The number of loop block devices that get 3760 (loop.max_loop) unconditionally pre-created at init time. The default 3761 number is configured by BLK_DEV_LOOP_MIN_COUNT. Instead 3762 of statically allocating a predefined number, loop 3763 devices can be requested on-demand with the 3764 /dev/loop-control interface. 3765 3766 mce= [X86-{32,64}] 3767 3768 Please see Documentation/arch/x86/x86_64/machinecheck.rst for sysfs runtime tunables. 3769 3770 off 3771 disable machine check 3772 3773 no_cmci 3774 disable CMCI(Corrected Machine Check Interrupt) that 3775 Intel processor supports. Usually this disablement is 3776 not recommended, but it might be handy if your 3777 hardware is misbehaving. 3778 3779 Note that you'll get more problems without CMCI than 3780 with due to the shared banks, i.e. you might get 3781 duplicated error logs. 3782 3783 dont_log_ce 3784 don't make logs for corrected errors. All events 3785 reported as corrected are silently cleared by OS. This 3786 option will be useful if you have no interest in any 3787 of corrected errors. 3788 3789 ignore_ce 3790 disable features for corrected errors, e.g. 3791 polling timer and CMCI. All events reported as 3792 corrected are not cleared by OS and remained in its 3793 error banks. 3794 3795 Usually this disablement is not recommended, however 3796 if there is an agent checking/clearing corrected 3797 errors (e.g. BIOS or hardware monitoring 3798 applications), conflicting with OS's error handling, 3799 and you cannot deactivate the agent, then this option 3800 will be a help. 3801 3802 no_lmce 3803 do not opt-in to Local MCE delivery. Use legacy method 3804 to broadcast MCEs. 3805 3806 bootlog 3807 enable logging of machine checks left over from 3808 booting. Disabled by default on AMD Fam10h and older 3809 because some BIOS leave bogus ones. 3810 3811 If your BIOS doesn't do that it's a good idea to 3812 enable though to make sure you log even machine check 3813 events that result in a reboot. On Intel systems it is 3814 enabled by default. 3815 3816 nobootlog 3817 disable boot machine check logging. 3818 3819 print_all 3820 print all machine check logs to the console. 3821 3822 monarchtimeout (number) 3823 sets the time in us to wait for other CPUs on machine 3824 checks. 0 to disable. 3825 3826 bios_cmci_threshold 3827 don't overwrite the bios-set CMCI threshold. This boot 3828 option prevents Linux from overwriting the CMCI 3829 threshold set by the bios. Without this option, Linux 3830 always sets the CMCI threshold to 1. Enabling this may 3831 make memory predictive failure analysis less effective 3832 if the bios sets thresholds for memory errors since we 3833 will not see details for all errors. 3834 3835 recovery 3836 force-enable recoverable machine check code paths 3837 3838 Everything else is in sysfs now. 3839 3840 3841 md= [HW] RAID subsystems devices and level 3842 See Documentation/admin-guide/md.rst. 3843 3844 mds= [X86,INTEL,EARLY] 3845 Control mitigation for the Micro-architectural Data 3846 Sampling (MDS) vulnerability. 3847 3848 Certain CPUs are vulnerable to an exploit against CPU 3849 internal buffers which can forward information to a 3850 disclosure gadget under certain conditions. 3851 3852 In vulnerable processors, the speculatively 3853 forwarded data can be used in a cache side channel 3854 attack, to access data to which the attacker does 3855 not have direct access. 3856 3857 This parameter controls the MDS mitigation. The 3858 options are: 3859 3860 full - Enable MDS mitigation on vulnerable CPUs 3861 full,nosmt - Enable MDS mitigation and disable 3862 SMT on vulnerable CPUs 3863 off - Unconditionally disable MDS mitigation 3864 3865 On TAA-affected machines, mds=off can be prevented by 3866 an active TAA mitigation as both vulnerabilities are 3867 mitigated with the same mechanism so in order to disable 3868 this mitigation, you need to specify tsx_async_abort=off 3869 too. 3870 3871 Not specifying this option is equivalent to 3872 mds=full. 3873 3874 For details see: Documentation/admin-guide/hw-vuln/mds.rst 3875 3876 mem=nn[KMG] [HEXAGON,EARLY] Set the memory size. 3877 Must be specified, otherwise memory size will be 0. 3878 3879 mem=nn[KMG] [KNL,BOOT,EARLY] Force usage of a specific amount 3880 of memory Amount of memory to be used in cases 3881 as follows: 3882 3883 1 for test; 3884 2 when the kernel is not able to see the whole system memory; 3885 3 memory that lies after 'mem=' boundary is excluded from 3886 the hypervisor, then assigned to KVM guests. 3887 4 to limit the memory available for kdump kernel. 3888 3889 [ARC,MICROBLAZE] - the limit applies only to low memory, 3890 high memory is not affected. 3891 3892 [ARM64] - only limits memory covered by the linear 3893 mapping. The NOMAP regions are not affected. 3894 3895 [X86] Work as limiting max address. Use together 3896 with memmap= to avoid physical address space collisions. 3897 Without memmap= PCI devices could be placed at addresses 3898 belonging to unused RAM. 3899 3900 Note that this only takes effects during boot time since 3901 in above case 3, memory may need be hot added after boot 3902 if system memory of hypervisor is not sufficient. 3903 3904 mem=nn[KMG]@ss[KMG] 3905 [ARM,MIPS,EARLY] - override the memory layout 3906 reported by firmware. 3907 Define a memory region of size nn[KMG] starting at 3908 ss[KMG]. 3909 Multiple different regions can be specified with 3910 multiple mem= parameters on the command line. 3911 3912 mem=nopentium [BUGS=X86-32] Disable usage of 4MB pages for kernel 3913 memory. 3914 3915 memblock=debug [KNL,EARLY] Enable memblock debug messages. 3916 3917 memchunk=nn[KMG] 3918 [KNL,SH] Allow user to override the default size for 3919 per-device physically contiguous DMA buffers. 3920 3921 memhp_default_state=online/offline/online_kernel/online_movable 3922 [KNL] Set the initial state for the memory hotplug 3923 onlining policy. If not specified, the default value is 3924 set according to the 3925 CONFIG_MHP_DEFAULT_ONLINE_TYPE kernel config 3926 options. 3927 See Documentation/admin-guide/mm/memory-hotplug.rst. 3928 3929 memmap=exactmap [KNL,X86,EARLY] Enable setting of an exact 3930 E820 memory map, as specified by the user. 3931 Such memmap=exactmap lines can be constructed based on 3932 BIOS output or other requirements. See the memmap=nn@ss 3933 option description. 3934 3935 memmap=nn[KMG]@ss[KMG] 3936 [KNL, X86,MIPS,XTENSA,EARLY] Force usage of a specific region of memory. 3937 Region of memory to be used is from ss to ss+nn. 3938 If @ss[KMG] is omitted, it is equivalent to mem=nn[KMG], 3939 which limits max address to nn[KMG]. 3940 Multiple different regions can be specified, 3941 comma delimited. 3942 Example: 3943 memmap=100M@2G,100M#3G,1G!1024G 3944 3945 memmap=nn[KMG]#ss[KMG] 3946 [KNL,ACPI,EARLY] Mark specific memory as ACPI data. 3947 Region of memory to be marked is from ss to ss+nn. 3948 3949 memmap=nn[KMG]$ss[KMG] 3950 [KNL,ACPI,EARLY] Mark specific memory as reserved. 3951 Region of memory to be reserved is from ss to ss+nn. 3952 Example: Exclude memory from 0x18690000-0x1869ffff 3953 memmap=64K$0x18690000 3954 or 3955 memmap=0x10000$0x18690000 3956 Some bootloaders may need an escape character before '$', 3957 like Grub2, otherwise '$' and the following number 3958 will be eaten. 3959 3960 memmap=nn[KMG]!ss[KMG,EARLY] 3961 [KNL,X86] Mark specific memory as protected. 3962 Region of memory to be used, from ss to ss+nn. 3963 The memory region may be marked as e820 type 12 (0xc) 3964 and is NVDIMM or ADR memory. 3965 3966 memmap=<size>%<offset>-<oldtype>+<newtype> 3967 [KNL,ACPI,EARLY] Convert memory within the specified region 3968 from <oldtype> to <newtype>. If "-<oldtype>" is left 3969 out, the whole region will be marked as <newtype>, 3970 even if previously unavailable. If "+<newtype>" is left 3971 out, matching memory will be removed. Types are 3972 specified as e820 types, e.g., 1 = RAM, 2 = reserved, 3973 3 = ACPI, 12 = PRAM. 3974 3975 memory_corruption_check=0/1 [X86,EARLY] 3976 Some BIOSes seem to corrupt the first 64k of 3977 memory when doing things like suspend/resume. 3978 Setting this option will scan the memory 3979 looking for corruption. Enabling this will 3980 both detect corruption and prevent the kernel 3981 from using the memory being corrupted. 3982 However, it's intended as a diagnostic tool; if 3983 repeatable BIOS-originated corruption always 3984 affects the same memory, you can use memmap= 3985 to prevent the kernel from using that memory. 3986 3987 memory_corruption_check_size=size [X86,EARLY] 3988 By default it checks for corruption in the low 3989 64k, making this memory unavailable for normal 3990 use. Use this parameter to scan for 3991 corruption in more or less memory. 3992 3993 memory_corruption_check_period=seconds [X86,EARLY] 3994 By default it checks for corruption every 60 3995 seconds. Use this parameter to check at some 3996 other rate. 0 disables periodic checking. 3997 3998 memory_hotplug.memmap_on_memory 3999 [KNL,X86,ARM] Boolean flag to enable this feature. 4000 Format: {on | off (default)} 4001 When enabled, runtime hotplugged memory will 4002 allocate its internal metadata (struct pages, 4003 those vmemmap pages cannot be optimized even 4004 if hugetlb_free_vmemmap is enabled) from the 4005 hotadded memory which will allow to hotadd a 4006 lot of memory without requiring additional 4007 memory to do so. 4008 This feature is disabled by default because it 4009 has some implication on large (e.g. GB) 4010 allocations in some configurations (e.g. small 4011 memory blocks). 4012 The state of the flag can be read in 4013 /sys/module/memory_hotplug/parameters/memmap_on_memory. 4014 Note that even when enabled, there are a few cases where 4015 the feature is not effective. 4016 4017 memtest= [KNL,X86,ARM,M68K,PPC,RISCV,EARLY] Enable memtest 4018 Format: <integer> 4019 default : 0 <disable> 4020 Specifies the number of memtest passes to be 4021 performed. Each pass selects another test 4022 pattern from a given set of patterns. Memtest 4023 fills the memory with this pattern, validates 4024 memory contents and reserves bad memory 4025 regions that are detected. 4026 4027 mem_encrypt= [X86-64] AMD Secure Memory Encryption (SME) control 4028 Valid arguments: on, off 4029 Default: off 4030 mem_encrypt=on: Activate SME 4031 mem_encrypt=off: Do not activate SME 4032 4033 Refer to Documentation/virt/kvm/x86/amd-memory-encryption.rst 4034 for details on when memory encryption can be activated. 4035 4036 mem_sleep_default= [SUSPEND] Default system suspend mode: 4037 s2idle - Suspend-To-Idle 4038 shallow - Power-On Suspend or equivalent (if supported) 4039 deep - Suspend-To-RAM or equivalent (if supported) 4040 See Documentation/admin-guide/pm/sleep-states.rst. 4041 4042 mfgptfix [X86-32] Fix MFGPT timers on AMD Geode platforms when 4043 the BIOS has incorrectly applied a workaround. TinyBIOS 4044 version 0.98 is known to be affected, 0.99 fixes the 4045 problem by letting the user disable the workaround. 4046 4047 mga= [HW,DRM] 4048 4049 microcode= [X86] Control the behavior of the microcode loader. 4050 Available options, comma separated: 4051 4052 base_rev=X - with <X> with format: <u32> 4053 Set the base microcode revision of each thread when in 4054 debug mode. 4055 4056 dis_ucode_ldr: disable the microcode loader 4057 4058 force_minrev: 4059 Enable or disable the microcode minimal revision 4060 enforcement for the runtime microcode loader. 4061 4062 mini2440= [ARM,HW,KNL] 4063 Format:[0..2][b][c][t] 4064 Default: "0tb" 4065 MINI2440 configuration specification: 4066 0 - The attached screen is the 3.5" TFT 4067 1 - The attached screen is the 7" TFT 4068 2 - The VGA Shield is attached (1024x768) 4069 Leaving out the screen size parameter will not load 4070 the TFT driver, and the framebuffer will be left 4071 unconfigured. 4072 b - Enable backlight. The TFT backlight pin will be 4073 linked to the kernel VESA blanking code and a GPIO 4074 LED. This parameter is not necessary when using the 4075 VGA shield. 4076 c - Enable the s3c camera interface. 4077 t - Reserved for enabling touchscreen support. The 4078 touchscreen support is not enabled in the mainstream 4079 kernel as of 2.6.30, a preliminary port can be found 4080 in the "bleeding edge" mini2440 support kernel at 4081 https://repo.or.cz/w/linux-2.6/mini2440.git 4082 4083 mitigations= 4084 [X86,PPC,S390,ARM64,EARLY] Control optional mitigations for 4085 CPU vulnerabilities. This is a set of curated, 4086 arch-independent options, each of which is an 4087 aggregation of existing arch-specific options. 4088 4089 Note, "mitigations" is supported if and only if the 4090 kernel was built with CPU_MITIGATIONS=y. 4091 4092 off 4093 Disable all optional CPU mitigations. This 4094 improves system performance, but it may also 4095 expose users to several CPU vulnerabilities. 4096 Equivalent to: if nokaslr then kpti=0 [ARM64] 4097 gather_data_sampling=off [X86] 4098 indirect_target_selection=off [X86] 4099 kvm.nx_huge_pages=off [X86] 4100 l1tf=off [X86] 4101 mds=off [X86] 4102 mmio_stale_data=off [X86] 4103 no_entry_flush [PPC] 4104 no_uaccess_flush [PPC] 4105 nobp=0 [S390] 4106 nopti [X86,PPC] 4107 nospectre_bhb [ARM64] 4108 nospectre_v1 [X86,PPC] 4109 nospectre_v2 [X86,PPC,S390,ARM64] 4110 reg_file_data_sampling=off [X86] 4111 retbleed=off [X86] 4112 spec_rstack_overflow=off [X86] 4113 spec_store_bypass_disable=off [X86,PPC] 4114 spectre_bhi=off [X86] 4115 spectre_v2_user=off [X86] 4116 srbds=off [X86,INTEL] 4117 ssbd=force-off [ARM64] 4118 tsa=off [X86,AMD] 4119 tsx_async_abort=off [X86] 4120 vmscape=off [X86] 4121 4122 Exceptions: 4123 This does not have any effect on 4124 kvm.nx_huge_pages when 4125 kvm.nx_huge_pages=force. 4126 4127 auto (default) 4128 Mitigate all CPU vulnerabilities, but leave SMT 4129 enabled, even if it's vulnerable. This is for 4130 users who don't want to be surprised by SMT 4131 getting disabled across kernel upgrades, or who 4132 have other ways of avoiding SMT-based attacks. 4133 Equivalent to: (default behavior) 4134 4135 auto,nosmt 4136 Mitigate all CPU vulnerabilities, disabling SMT 4137 if needed. This is for users who always want to 4138 be fully mitigated, even if it means losing SMT. 4139 Equivalent to: l1tf=flush,nosmt [X86] 4140 mds=full,nosmt [X86] 4141 tsx_async_abort=full,nosmt [X86] 4142 mmio_stale_data=full,nosmt [X86] 4143 retbleed=auto,nosmt [X86] 4144 4145 [X86] After one of the above options, additionally 4146 supports attack-vector based controls as documented in 4147 Documentation/admin-guide/hw-vuln/attack_vector_controls.rst 4148 4149 mminit_loglevel= 4150 [KNL,EARLY] When CONFIG_DEBUG_MEMORY_INIT is set, this 4151 parameter allows control of the logging verbosity for 4152 the additional memory initialisation checks. A value 4153 of 0 disables mminit logging and a level of 4 will 4154 log everything. Information is printed at KERN_DEBUG 4155 so loglevel=8 may also need to be specified. 4156 4157 mmio_stale_data= 4158 [X86,INTEL,EARLY] Control mitigation for the Processor 4159 MMIO Stale Data vulnerabilities. 4160 4161 Processor MMIO Stale Data is a class of 4162 vulnerabilities that may expose data after an MMIO 4163 operation. Exposed data could originate or end in 4164 the same CPU buffers as affected by MDS and TAA. 4165 Therefore, similar to MDS and TAA, the mitigation 4166 is to clear the affected CPU buffers. 4167 4168 This parameter controls the mitigation. The 4169 options are: 4170 4171 full - Enable mitigation on vulnerable CPUs 4172 4173 full,nosmt - Enable mitigation and disable SMT on 4174 vulnerable CPUs. 4175 4176 off - Unconditionally disable mitigation 4177 4178 On MDS or TAA affected machines, 4179 mmio_stale_data=off can be prevented by an active 4180 MDS or TAA mitigation as these vulnerabilities are 4181 mitigated with the same mechanism so in order to 4182 disable this mitigation, you need to specify 4183 mds=off and tsx_async_abort=off too. 4184 4185 Not specifying this option is equivalent to 4186 mmio_stale_data=full. 4187 4188 For details see: 4189 Documentation/admin-guide/hw-vuln/processor_mmio_stale_data.rst 4190 4191 <module>.async_probe[=<bool>] [KNL] 4192 If no <bool> value is specified or if the value 4193 specified is not a valid <bool>, enable asynchronous 4194 probe on this module. Otherwise, enable/disable 4195 asynchronous probe on this module as indicated by the 4196 <bool> value. See also: module.async_probe 4197 4198 module.async_probe=<bool> 4199 [KNL] When set to true, modules will use async probing 4200 by default. To enable/disable async probing for a 4201 specific module, use the module specific control that 4202 is documented under <module>.async_probe. When both 4203 module.async_probe and <module>.async_probe are 4204 specified, <module>.async_probe takes precedence for 4205 the specific module. 4206 4207 module.enable_dups_trace 4208 [KNL] When CONFIG_MODULE_DEBUG_AUTOLOAD_DUPS is set, 4209 this means that duplicate request_module() calls will 4210 trigger a WARN_ON() instead of a pr_warn(). Note that 4211 if MODULE_DEBUG_AUTOLOAD_DUPS_TRACE is set, WARN_ON()s 4212 will always be issued and this option does nothing. 4213 module.sig_enforce 4214 [KNL] When CONFIG_MODULE_SIG is set, this means that 4215 modules without (valid) signatures will fail to load. 4216 Note that if CONFIG_MODULE_SIG_FORCE is set, that 4217 is always true, so this option does nothing. 4218 4219 module_blacklist= [KNL] Do not load a comma-separated list of 4220 modules. Useful for debugging problem modules. 4221 4222 mousedev.tap_time= 4223 [MOUSE] Maximum time between finger touching and 4224 leaving touchpad surface for touch to be considered 4225 a tap and be reported as a left button click (for 4226 touchpads working in absolute mode only). 4227 Format: <msecs> 4228 mousedev.xres= [MOUSE] Horizontal screen resolution, used for devices 4229 reporting absolute coordinates, such as tablets 4230 mousedev.yres= [MOUSE] Vertical screen resolution, used for devices 4231 reporting absolute coordinates, such as tablets 4232 4233 movablecore= [KNL,X86,PPC,EARLY] 4234 Format: nn[KMGTPE] | nn% 4235 This parameter is the complement to kernelcore=, it 4236 specifies the amount of memory used for migratable 4237 allocations. If both kernelcore and movablecore is 4238 specified, then kernelcore will be at *least* the 4239 specified value but may be more. If movablecore on its 4240 own is specified, the administrator must be careful 4241 that the amount of memory usable for all allocations 4242 is not too small. 4243 4244 movable_node [KNL,EARLY] Boot-time switch to make hotplugable memory 4245 NUMA nodes to be movable. This means that the memory 4246 of such nodes will be usable only for movable 4247 allocations which rules out almost all kernel 4248 allocations. Use with caution! 4249 4250 MTD_Partition= [MTD] 4251 Format: <name>,<region-number>,<size>,<offset> 4252 4253 MTD_Region= [MTD] Format: 4254 <name>,<region-number>[,<base>,<size>,<buswidth>,<altbuswidth>] 4255 4256 mtdparts= [MTD] 4257 See drivers/mtd/parsers/cmdlinepart.c 4258 4259 mtouchusb.raw_coordinates= 4260 [HW] Make the MicroTouch USB driver use raw coordinates 4261 ('y', default) or cooked coordinates ('n') 4262 4263 mtrr=debug [X86,EARLY] 4264 Enable printing debug information related to MTRR 4265 registers at boot time. 4266 4267 mtrr_chunk_size=nn[KMG,X86,EARLY] 4268 used for mtrr cleanup. It is largest continuous chunk 4269 that could hold holes aka. UC entries. 4270 4271 mtrr_gran_size=nn[KMG,X86,EARLY] 4272 Used for mtrr cleanup. It is granularity of mtrr block. 4273 Default is 1. 4274 Large value could prevent small alignment from 4275 using up MTRRs. 4276 4277 mtrr_spare_reg_nr=n [X86,EARLY] 4278 Format: <integer> 4279 Range: 0,7 : spare reg number 4280 Default : 1 4281 Used for mtrr cleanup. It is spare mtrr entries number. 4282 Set to 2 or more if your graphical card needs more. 4283 4284 multitce=off [PPC] This parameter disables the use of the pSeries 4285 firmware feature for updating multiple TCE entries 4286 at a time. 4287 4288 n2= [NET] SDL Inc. RISCom/N2 synchronous serial card 4289 4290 netdev= [NET] NE2000 ISA network devices parameters 4291 Format: <irq>,<io>,<mem_start>,<mem_end>,<name> 4292 4293 netpoll.carrier_timeout= 4294 [NET] Specifies amount of time (in seconds) that 4295 netpoll should wait for a carrier. By default netpoll 4296 waits 4 seconds. 4297 4298 nf_conntrack.acct= 4299 [NETFILTER] Enable connection tracking flow accounting 4300 0 to disable accounting 4301 1 to enable accounting 4302 Default value is 0. 4303 4304 nfs.cache_getent= 4305 [NFS] sets the pathname to the program which is used 4306 to update the NFS client cache entries. 4307 4308 nfs.cache_getent_timeout= 4309 [NFS] sets the timeout after which an attempt to 4310 update a cache entry is deemed to have failed. 4311 4312 nfs.callback_nr_threads= 4313 [NFSv4] set the total number of threads that the 4314 NFS client will assign to service NFSv4 callback 4315 requests. 4316 4317 nfs.callback_tcpport= 4318 [NFS] set the TCP port on which the NFSv4 callback 4319 channel should listen. 4320 4321 nfs.delay_retrans= 4322 [NFS] specifies the number of times the NFSv4 client 4323 retries the request before returning an EAGAIN error, 4324 after a reply of NFS4ERR_DELAY from the server. 4325 Only applies if the softerr mount option is enabled, 4326 and the specified value is >= 0. 4327 4328 nfs.idmap_cache_timeout= 4329 [NFS] set the maximum lifetime for idmapper cache 4330 entries. 4331 4332 nfs.max_session_cb_slots= 4333 [NFSv4.1] Sets the maximum number of session 4334 slots the client will assign to the callback 4335 channel. This determines the maximum number of 4336 callbacks the client will process in parallel for 4337 a particular server. 4338 4339 nfs.max_session_slots= 4340 [NFSv4.1] Sets the maximum number of session slots 4341 the client will attempt to negotiate with the server. 4342 This limits the number of simultaneous RPC requests 4343 that the client can send to the NFSv4.1 server. 4344 Note that there is little point in setting this 4345 value higher than the max_tcp_slot_table_limit. 4346 4347 nfs.nfs4_disable_idmapping= 4348 [NFSv4] When set to the default of '1', this option 4349 ensures that both the RPC level authentication 4350 scheme and the NFS level operations agree to use 4351 numeric uids/gids if the mount is using the 4352 'sec=sys' security flavour. In effect it is 4353 disabling idmapping, which can make migration from 4354 legacy NFSv2/v3 systems to NFSv4 easier. 4355 Servers that do not support this mode of operation 4356 will be autodetected by the client, and it will fall 4357 back to using the idmapper. 4358 To turn off this behaviour, set the value to '0'. 4359 4360 nfs.nfs4_unique_id= 4361 [NFS4] Specify an additional fixed unique ident- 4362 ification string that NFSv4 clients can insert into 4363 their nfs_client_id4 string. This is typically a 4364 UUID that is generated at system install time. 4365 4366 nfs.recover_lost_locks= 4367 [NFSv4] Attempt to recover locks that were lost due 4368 to a lease timeout on the server. Please note that 4369 doing this risks data corruption, since there are 4370 no guarantees that the file will remain unchanged 4371 after the locks are lost. 4372 If you want to enable the kernel legacy behaviour of 4373 attempting to recover these locks, then set this 4374 parameter to '1'. 4375 The default parameter value of '0' causes the kernel 4376 not to attempt recovery of lost locks. 4377 4378 nfs.send_implementation_id= 4379 [NFSv4.1] Send client implementation identification 4380 information in exchange_id requests. 4381 If zero, no implementation identification information 4382 will be sent. 4383 The default is to send the implementation identification 4384 information. 4385 4386 nfs4.layoutstats_timer= 4387 [NFSv4.2] Change the rate at which the kernel sends 4388 layoutstats to the pNFS metadata server. 4389 4390 Setting this to value to 0 causes the kernel to use 4391 whatever value is the default set by the layout 4392 driver. A non-zero value sets the minimum interval 4393 in seconds between layoutstats transmissions. 4394 4395 nfsd.inter_copy_offload_enable= 4396 [NFSv4.2] When set to 1, the server will support 4397 server-to-server copies for which this server is 4398 the destination of the copy. 4399 4400 nfsd.nfs4_disable_idmapping= 4401 [NFSv4] When set to the default of '1', the NFSv4 4402 server will return only numeric uids and gids to 4403 clients using auth_sys, and will accept numeric uids 4404 and gids from such clients. This is intended to ease 4405 migration from NFSv2/v3. 4406 4407 nfsd.nfsd4_ssc_umount_timeout= 4408 [NFSv4.2] When used as the destination of a 4409 server-to-server copy, knfsd temporarily mounts 4410 the source server. It caches the mount in case 4411 it will be needed again, and discards it if not 4412 used for the number of milliseconds specified by 4413 this parameter. 4414 4415 nfsaddrs= [NFS] Deprecated. Use ip= instead. 4416 See Documentation/admin-guide/nfs/nfsroot.rst. 4417 4418 nfsroot= [NFS] nfs root filesystem for disk-less boxes. 4419 See Documentation/admin-guide/nfs/nfsroot.rst. 4420 4421 nfsrootdebug [NFS] enable nfsroot debugging messages. 4422 See Documentation/admin-guide/nfs/nfsroot.rst. 4423 4424 nmi_backtrace.backtrace_idle [KNL] 4425 Dump stacks even of idle CPUs in response to an 4426 NMI stack-backtrace request. 4427 4428 nmi_debug= [KNL,SH] Specify one or more actions to take 4429 when a NMI is triggered. 4430 Format: [state][,regs][,debounce][,die] 4431 4432 nmi_watchdog= [KNL,BUGS=X86] Debugging features for SMP kernels 4433 Format: [panic,][nopanic,][rNNN,][num] 4434 Valid num: 0 or 1 4435 0 - turn hardlockup detector in nmi_watchdog off 4436 1 - turn hardlockup detector in nmi_watchdog on 4437 rNNN - configure the watchdog with raw perf event 0xNNN 4438 4439 When panic is specified, panic when an NMI watchdog 4440 timeout occurs (or 'nopanic' to not panic on an NMI 4441 watchdog, if CONFIG_BOOTPARAM_HARDLOCKUP_PANIC is set) 4442 To disable both hard and soft lockup detectors, 4443 please see 'nowatchdog'. 4444 This is useful when you use a panic=... timeout and 4445 need the box quickly up again. 4446 4447 These settings can be accessed at runtime via 4448 the nmi_watchdog and hardlockup_panic sysctls. 4449 4450 no4lvl [RISCV,EARLY] Disable 4-level and 5-level paging modes. 4451 Forces kernel to use 3-level paging instead. 4452 4453 no5lvl [X86-64,RISCV,EARLY] Disable 5-level paging mode. Forces 4454 kernel to use 4-level paging instead. 4455 4456 noalign [KNL,ARM] 4457 4458 noapic [SMP,APIC,EARLY] Tells the kernel to not make use of any 4459 IOAPICs that may be present in the system. 4460 4461 noapictimer [APIC,X86] Don't set up the APIC timer 4462 4463 noautogroup Disable scheduler automatic task group creation. 4464 4465 nocache [ARM,EARLY] 4466 4467 no_console_suspend 4468 [HW] Never suspend the console 4469 Disable suspending of consoles during suspend and 4470 hibernate operations. Once disabled, debugging 4471 messages can reach various consoles while the rest 4472 of the system is being put to sleep (ie, while 4473 debugging driver suspend/resume hooks). This may 4474 not work reliably with all consoles, but is known 4475 to work with serial and VGA consoles. 4476 To facilitate more flexible debugging, we also add 4477 console_suspend, a printk module parameter to control 4478 it. Users could use console_suspend (usually 4479 /sys/module/printk/parameters/console_suspend) to 4480 turn on/off it dynamically. 4481 4482 no_debug_objects 4483 [KNL,EARLY] Disable object debugging 4484 4485 nodsp [SH] Disable hardware DSP at boot time. 4486 4487 noefi [EFI,EARLY] Disable EFI runtime services support. 4488 4489 no_entry_flush [PPC,EARLY] Don't flush the L1-D cache when entering the kernel. 4490 4491 noexec32 [X86-64] 4492 This affects only 32-bit executables. 4493 noexec32=on: enable non-executable mappings (default) 4494 read doesn't imply executable mappings 4495 noexec32=off: disable non-executable mappings 4496 read implies executable mappings 4497 4498 no_file_caps Tells the kernel not to honor file capabilities. The 4499 only way then for a file to be executed with privilege 4500 is to be setuid root or executed by root. 4501 4502 nofpu [MIPS,SH] Disable hardware FPU at boot time. 4503 4504 nofsgsbase [X86] Disables FSGSBASE instructions. 4505 4506 nofxsr [BUGS=X86-32] Disables x86 floating point extended 4507 register save and restore. The kernel will only save 4508 legacy floating-point registers on task switch. 4509 4510 nogbpages [X86] Do not use GB pages for kernel direct mappings. 4511 4512 no_hash_pointers 4513 [KNL,EARLY] 4514 Alias for "hash_pointers=never". 4515 4516 nohibernate [HIBERNATION] Disable hibernation and resume. 4517 4518 nohlt [ARM,ARM64,MICROBLAZE,MIPS,PPC,RISCV,SH] Forces the kernel to 4519 busy wait in do_idle() and not use the arch_cpu_idle() 4520 implementation; requires CONFIG_GENERIC_IDLE_POLL_SETUP 4521 to be effective. This is useful on platforms where the 4522 sleep(SH) or wfi(ARM,ARM64) instructions do not work 4523 correctly or when doing power measurements to evaluate 4524 the impact of the sleep instructions. This is also 4525 useful when using JTAG debugger. 4526 4527 nohpet [X86] Don't use the HPET timer. 4528 4529 nohugeiomap [KNL,X86,PPC,ARM64,EARLY] Disable kernel huge I/O mappings. 4530 4531 nohugevmalloc [KNL,X86,PPC,ARM64,EARLY] Disable kernel huge vmalloc mappings. 4532 4533 nohz= [KNL] Boottime enable/disable dynamic ticks 4534 Valid arguments: on, off 4535 Default: on 4536 4537 nohz_full= [KNL,BOOT,SMP,ISOL] 4538 The argument is a cpu list, as described above. 4539 In kernels built with CONFIG_NO_HZ_FULL=y, set 4540 the specified list of CPUs whose tick will be stopped 4541 whenever possible. The boot CPU will be forced outside 4542 the range to maintain the timekeeping. Any CPUs 4543 in this list will have their RCU callbacks offloaded, 4544 just as if they had also been called out in the 4545 rcu_nocbs= boot parameter. 4546 4547 Note that this argument takes precedence over 4548 the CONFIG_RCU_NOCB_CPU_DEFAULT_ALL option. 4549 4550 noinitrd [Deprecated,RAM] Tells the kernel not to load any configured 4551 initial RAM disk. Currently this parameter applies to 4552 initrd only, not to initramfs. But it applies to both 4553 in EFI mode. 4554 4555 nointremap [X86-64,Intel-IOMMU,EARLY] Do not enable interrupt 4556 remapping. 4557 [Deprecated - use intremap=off] 4558 4559 noinvpcid [X86,EARLY] Disable the INVPCID cpu feature. 4560 4561 noiotrap [SH] Disables trapped I/O port accesses. 4562 4563 noirqdebug [X86-32] Disables the code which attempts to detect and 4564 disable unhandled interrupt sources. 4565 4566 noisapnp [ISAPNP] Disables ISA PnP code. 4567 4568 nokaslr [KNL,EARLY] 4569 When CONFIG_RANDOMIZE_BASE is set, this disables 4570 kernel and module base offset ASLR (Address Space 4571 Layout Randomization). 4572 4573 no-kvmapf [X86,KVM,EARLY] Disable paravirtualized asynchronous page 4574 fault handling. 4575 4576 no-kvmclock [X86,KVM,EARLY] Disable paravirtualized KVM clock driver 4577 4578 nolapic [X86-32,APIC,EARLY] Do not enable or use the local APIC. 4579 4580 nolapic_timer [X86-32,APIC,EARLY] Do not use the local APIC timer. 4581 4582 nomce [X86-32] Disable Machine Check Exception 4583 4584 nomfgpt [X86-32] Disable Multi-Function General Purpose 4585 Timer usage (for AMD Geode machines). 4586 4587 nomodeset Disable kernel modesetting. Most systems' firmware 4588 sets up a display mode and provides framebuffer memory 4589 for output. With nomodeset, DRM and fbdev drivers will 4590 not load if they could possibly displace the pre- 4591 initialized output. Only the system framebuffer will 4592 be available for use. The respective drivers will not 4593 perform display-mode changes or accelerated rendering. 4594 4595 Useful as error fallback, or for testing and debugging. 4596 4597 nomodule Disable module load 4598 4599 nonmi_ipi [X86] Disable using NMI IPIs during panic/reboot to 4600 shutdown the other cpus. Instead use the REBOOT_VECTOR 4601 irq. 4602 4603 nopat [X86,EARLY] Disable PAT (page attribute table extension of 4604 pagetables) support. 4605 4606 nopcid [X86-64,EARLY] Disable the PCID cpu feature. 4607 4608 nopku [X86] Disable Memory Protection Keys CPU feature found 4609 in some Intel CPUs. 4610 4611 nopti [X86-64,EARLY] 4612 Equivalent to pti=off 4613 4614 nopv= [X86,XEN,KVM,HYPER_V,VMWARE,EARLY] 4615 Disables the PV optimizations forcing the guest to run 4616 as generic guest with no PV drivers. Currently support 4617 XEN HVM, KVM, HYPER_V and VMWARE guest. 4618 4619 nopvspin [X86,XEN,KVM,EARLY] 4620 Disables the qspinlock slow path using PV optimizations 4621 which allow the hypervisor to 'idle' the guest on lock 4622 contention. 4623 4624 norandmaps Don't use address space randomization. Equivalent to 4625 echo 0 > /proc/sys/kernel/randomize_va_space 4626 4627 noresume [SWSUSP] Disables resume and restores original swap 4628 space. 4629 4630 no-scroll [VGA] Disables scrollback. 4631 This is required for the Braillex ib80-piezo Braille 4632 reader made by F.H. Papenmeier (Germany). 4633 4634 nosgx [X86-64,SGX,EARLY] Disables Intel SGX kernel support. 4635 4636 nosmap [PPC,EARLY] 4637 Disable SMAP (Supervisor Mode Access Prevention) 4638 even if it is supported by processor. 4639 4640 nosmep [PPC64s,EARLY] 4641 Disable SMEP (Supervisor Mode Execution Prevention) 4642 even if it is supported by processor. 4643 4644 nosmp [SMP,EARLY] Tells an SMP kernel to act as a UP kernel, 4645 and disable the IO APIC. legacy for "maxcpus=0". 4646 4647 nosmt [KNL,MIPS,PPC,EARLY] Disable symmetric multithreading (SMT). 4648 Equivalent to smt=1. 4649 4650 [KNL,LOONGARCH,X86,ARM64,PPC,S390] Disable symmetric multithreading (SMT). 4651 nosmt=force: Force disable SMT, cannot be undone 4652 via the sysfs control file. 4653 4654 nosoftlockup [KNL] Disable the soft-lockup detector. 4655 4656 nospec_store_bypass_disable 4657 [HW,EARLY] Disable all mitigations for the Speculative 4658 Store Bypass vulnerability 4659 4660 nospectre_bhb [ARM64,EARLY] Disable all mitigations for Spectre-BHB (branch 4661 history injection) vulnerability. System may allow data leaks 4662 with this option. 4663 4664 nospectre_v1 [X86,PPC,EARLY] Disable mitigations for Spectre Variant 1 4665 (bounds check bypass). With this option data leaks are 4666 possible in the system. 4667 4668 nospectre_v2 [X86,PPC_E500,ARM64,EARLY] Disable all mitigations 4669 for the Spectre variant 2 (indirect branch 4670 prediction) vulnerability. System may allow data 4671 leaks with this option. 4672 4673 no-steal-acc [X86,PV_OPS,ARM64,PPC/PSERIES,RISCV,LOONGARCH,EARLY] 4674 Disable paravirtualized steal time accounting. steal time 4675 is computed, but won't influence scheduler behaviour 4676 4677 nosync [HW,M68K] Disables sync negotiation for all devices. 4678 4679 no_timer_check [X86,APIC] Disables the code which tests for broken 4680 timer IRQ sources, i.e., the IO-APIC timer. This can 4681 work around problems with incorrect timer 4682 initialization on some boards. 4683 4684 no_uaccess_flush 4685 [PPC,EARLY] Don't flush the L1-D cache after accessing user data. 4686 4687 novmcoredd [KNL,KDUMP] 4688 Disable device dump. Device dump allows drivers to 4689 append dump data to vmcore so you can collect driver 4690 specified debug info. Drivers can append the data 4691 without any limit and this data is stored in memory, 4692 so this may cause significant memory stress. Disabling 4693 device dump can help save memory but the driver debug 4694 data will be no longer available. This parameter 4695 is only available when CONFIG_PROC_VMCORE_DEVICE_DUMP 4696 is set. 4697 4698 no-vmw-sched-clock 4699 [X86,PV_OPS,EARLY] Disable paravirtualized VMware 4700 scheduler clock and use the default one. 4701 4702 nowatchdog [KNL] Disable both lockup detectors, i.e. 4703 soft-lockup and NMI watchdog (hard-lockup). 4704 4705 nowb [ARM,EARLY] 4706 4707 nox2apic [X86-64,APIC,EARLY] Do not enable x2APIC mode. 4708 4709 NOTE: this parameter will be ignored on systems with the 4710 LEGACY_XAPIC_DISABLED bit set in the 4711 IA32_XAPIC_DISABLE_STATUS MSR. 4712 4713 noxsave [BUGS=X86] Disables x86 extended register state save 4714 and restore using xsave. The kernel will fallback to 4715 enabling legacy floating-point and sse state. 4716 4717 noxsaveopt [X86] Disables xsaveopt used in saving x86 extended 4718 register states. The kernel will fall back to use 4719 xsave to save the states. By using this parameter, 4720 performance of saving the states is degraded because 4721 xsave doesn't support modified optimization while 4722 xsaveopt supports it on xsaveopt enabled systems. 4723 4724 noxsaves [X86] Disables xsaves and xrstors used in saving and 4725 restoring x86 extended register state in compacted 4726 form of xsave area. The kernel will fall back to use 4727 xsaveopt and xrstor to save and restore the states 4728 in standard form of xsave area. By using this 4729 parameter, xsave area per process might occupy more 4730 memory on xsaves enabled systems. 4731 4732 nr_cpus= [SMP,EARLY] Maximum number of processors that an SMP kernel 4733 could support. nr_cpus=n : n >= 1 limits the kernel to 4734 support 'n' processors. It could be larger than the 4735 number of already plugged CPU during bootup, later in 4736 runtime you can physically add extra cpu until it reaches 4737 n. So during boot up some boot time memory for per-cpu 4738 variables need be pre-allocated for later physical cpu 4739 hot plugging. 4740 4741 nr_uarts= [SERIAL] maximum number of UARTs to be registered. 4742 4743 numa=off [KNL, ARM64, PPC, RISCV, SPARC, X86, EARLY] 4744 Disable NUMA, Only set up a single NUMA node 4745 spanning all memory. 4746 4747 numa=fake=<size>[MG] 4748 [KNL, ARM64, RISCV, X86, EARLY] 4749 If given as a memory unit, fills all system RAM with 4750 nodes of size interleaved over physical nodes. 4751 4752 numa=fake=<N> 4753 [KNL, ARM64, RISCV, X86, EARLY] 4754 If given as an integer, fills all system RAM with N 4755 fake nodes interleaved over physical nodes. 4756 4757 numa=fake=<N>U 4758 [KNL, ARM64, RISCV, X86, EARLY] 4759 If given as an integer followed by 'U', it will 4760 divide each physical node into N emulated nodes. 4761 4762 numa=noacpi [X86] Don't parse the SRAT table for NUMA setup 4763 4764 numa=nohmat [X86] Don't parse the HMAT table for NUMA setup, or 4765 soft-reserved memory partitioning. 4766 4767 numa_balancing= [KNL,ARM64,PPC,RISCV,S390,X86] Enable or disable automatic 4768 NUMA balancing. 4769 Allowed values are enable and disable 4770 4771 numa_zonelist_order= [KNL, BOOT] Select zonelist order for NUMA. 4772 'node', 'default' can be specified 4773 This can be set from sysctl after boot. 4774 See Documentation/admin-guide/sysctl/vm.rst for details. 4775 4776 nvme.quirks= [NVME] A list of quirk entries to augment the built-in 4777 nvme quirk list. List entries are separated by a 4778 '-' character. 4779 Each entry has the form VendorID:ProductID:quirk_names. 4780 The IDs are 4-digits hex numbers and quirk_names is a 4781 list of quirk names separated by commas. A quirk name 4782 can be prefixed by '^', meaning that the specified 4783 quirk must be disabled. 4784 4785 Example: 4786 nvme.quirks=7710:2267:bogus_nid,^identify_cns-9900:7711:broken_msi 4787 4788 ohci1394_dma=early [HW,EARLY] enable debugging via the ohci1394 driver. 4789 See Documentation/core-api/debugging-via-ohci1394.rst for more 4790 info. 4791 4792 olpc_ec_timeout= [OLPC] ms delay when issuing EC commands 4793 Rather than timing out after 20 ms if an EC 4794 command is not properly ACKed, override the length 4795 of the timeout. We have interrupts disabled while 4796 waiting for the ACK, so if this is set too high 4797 interrupts *may* be lost! 4798 4799 omap_mux= [OMAP] Override bootloader pin multiplexing. 4800 Format: <mux_mode0.mode_name=value>... 4801 For example, to override I2C bus2: 4802 omap_mux=i2c2_scl.i2c2_scl=0x100,i2c2_sda.i2c2_sda=0x100 4803 4804 onenand.bdry= [HW,MTD] Flex-OneNAND Boundary Configuration 4805 4806 Format: [die0_boundary][,die0_lock][,die1_boundary][,die1_lock] 4807 4808 boundary - index of last SLC block on Flex-OneNAND. 4809 The remaining blocks are configured as MLC blocks. 4810 lock - Configure if Flex-OneNAND boundary should be locked. 4811 Once locked, the boundary cannot be changed. 4812 1 indicates lock status, 0 indicates unlock status. 4813 4814 oops=panic [KNL,EARLY] 4815 Always panic on oopses. Default is to just kill the 4816 process, but there is a small probability of 4817 deadlocking the machine. 4818 This will also cause panics on machine check exceptions. 4819 Useful together with panic=30 to trigger a reboot. 4820 4821 page_alloc.shuffle= 4822 [KNL] Boolean flag to control whether the page allocator 4823 should randomize its free lists. This parameter can be 4824 used to enable/disable page randomization. The state of 4825 the flag can be read from sysfs at: 4826 /sys/module/page_alloc/parameters/shuffle. 4827 This parameter is only available if CONFIG_SHUFFLE_PAGE_ALLOCATOR=y. 4828 4829 page_owner= [KNL,EARLY] Boot-time page_owner enabling option. 4830 Storage of the information about who allocated 4831 each page is disabled in default. With this switch, 4832 we can turn it on. 4833 on: enable the feature 4834 4835 page_poison= [KNL,EARLY] Boot-time parameter changing the state of 4836 poisoning on the buddy allocator, available with 4837 CONFIG_PAGE_POISONING=y. 4838 off: turn off poisoning (default) 4839 on: turn on poisoning 4840 4841 page_reporting.page_reporting_order= 4842 [KNL] Minimal page reporting order 4843 Format: <integer> 4844 Adjust the minimal page reporting order. The page 4845 reporting is disabled when it exceeds MAX_PAGE_ORDER. 4846 4847 panic= [KNL] Kernel behaviour on panic: delay <timeout> 4848 timeout > 0: seconds before rebooting 4849 timeout = 0: wait forever 4850 timeout < 0: reboot immediately 4851 Format: <timeout> 4852 4853 panic_on_taint= [KNL,EARLY] 4854 Bitmask for conditionally calling panic() in add_taint() 4855 Format: <hex>[,nousertaint] 4856 Hexadecimal bitmask representing the set of TAINT flags 4857 that will cause the kernel to panic when add_taint() is 4858 called with any of the flags in this set. 4859 The optional switch "nousertaint" can be utilized to 4860 prevent userspace forced crashes by writing to sysctl 4861 /proc/sys/kernel/tainted any flagset matching with the 4862 bitmask set on panic_on_taint. 4863 See Documentation/admin-guide/tainted-kernels.rst for 4864 extra details on the taint flags that users can pick 4865 to compose the bitmask to assign to panic_on_taint. 4866 4867 panic_on_warn=1 panic() instead of WARN(). Useful to cause kdump 4868 on a WARN(). 4869 4870 panic_force_cpu= 4871 [KNL,SMP] Force panic handling to execute on a specific CPU. 4872 Format: <cpu number> 4873 Some platforms require panic handling to occur on a 4874 specific CPU for the crash kernel to function correctly. 4875 This can be due to firmware limitations, interrupt routing 4876 constraints, or platform-specific requirements where only 4877 a particular CPU can safely enter the crash kernel. 4878 When set, panic() will redirect execution to the specified 4879 CPU before proceeding with the normal panic and kexec flow. 4880 If the target CPU is offline or unavailable, panic proceeds 4881 on the current CPU. 4882 This option should only be used for systems with the above 4883 constraints as it might cause the panic operation to be less reliable. 4884 4885 panic_print= Bitmask for printing system info when panic happens. 4886 User can chose combination of the following bits: 4887 bit 0: print all tasks info 4888 bit 1: print system memory info 4889 bit 2: print timer info 4890 bit 3: print locks info if CONFIG_LOCKDEP is on 4891 bit 4: print ftrace buffer 4892 bit 5: replay all kernel messages on consoles at the end of panic 4893 bit 6: print all CPUs backtrace (if available in the arch) 4894 bit 7: print only tasks in uninterruptible (blocked) state 4895 *Be aware* that this option may print a _lot_ of lines, 4896 so there are risks of losing older messages in the log. 4897 Use this option carefully, maybe worth to setup a 4898 bigger log buffer with "log_buf_len" along with this. 4899 4900 panic_sys_info= A comma separated list of extra information to be dumped 4901 on panic. 4902 Format: val[,val...] 4903 Where @val can be any of the following: 4904 4905 tasks: print all tasks info 4906 mem: print system memory info 4907 timers: print timers info 4908 locks: print locks info if CONFIG_LOCKDEP is on 4909 ftrace: print ftrace buffer 4910 all_bt: print all CPUs backtrace (if available in the arch) 4911 blocked_tasks: print only tasks in uninterruptible (blocked) state 4912 4913 This is a human readable alternative to the 'panic_print' option. 4914 4915 panic_console_replay 4916 When panic happens, replay all kernel messages on 4917 consoles at the end of panic. 4918 4919 parkbd.port= [HW] Parallel port number the keyboard adapter is 4920 connected to, default is 0. 4921 Format: <parport#> 4922 parkbd.mode= [HW] Parallel port keyboard adapter mode of operation, 4923 0 for XT, 1 for AT (default is AT). 4924 Format: <mode> 4925 4926 parport= [HW,PPT] Specify parallel ports. 0 disables. 4927 Format: { 0 | auto | 0xBBB[,IRQ[,DMA]] } 4928 Use 'auto' to force the driver to use any 4929 IRQ/DMA settings detected (the default is to 4930 ignore detected IRQ/DMA settings because of 4931 possible conflicts). You can specify the base 4932 address, IRQ, and DMA settings; IRQ and DMA 4933 should be numbers, or 'auto' (for using detected 4934 settings on that particular port), or 'nofifo' 4935 (to avoid using a FIFO even if it is detected). 4936 Parallel ports are assigned in the order they 4937 are specified on the command line, starting 4938 with parport0. 4939 4940 parport_init_mode= [HW,PPT] 4941 Configure VIA parallel port to operate in 4942 a specific mode. This is necessary on Pegasos 4943 computer where firmware has no options for setting 4944 up parallel port mode and sets it to spp. 4945 Currently this function knows 686a and 8231 chips. 4946 Format: [spp|ps2|epp|ecp|ecpepp] 4947 4948 pata_legacy.all= [HW,LIBATA] 4949 Format: <int> 4950 Set to non-zero to probe primary and secondary ISA 4951 port ranges on PCI systems where no PCI PATA device 4952 has been found at either range. Disabled by default. 4953 4954 pata_legacy.autospeed= [HW,LIBATA] 4955 Format: <int> 4956 Set to non-zero if a chip is present that snoops speed 4957 changes. Disabled by default. 4958 4959 pata_legacy.ht6560a= [HW,LIBATA] 4960 Format: <int> 4961 Set to 1, 2, or 3 for HT 6560A on the primary channel, 4962 the secondary channel, or both channels respectively. 4963 Disabled by default. 4964 4965 pata_legacy.ht6560b= [HW,LIBATA] 4966 Format: <int> 4967 Set to 1, 2, or 3 for HT 6560B on the primary channel, 4968 the secondary channel, or both channels respectively. 4969 Disabled by default. 4970 4971 pata_legacy.iordy_mask= [HW,LIBATA] 4972 Format: <int> 4973 IORDY enable mask. Set individual bits to allow IORDY 4974 for the respective channel. Bit 0 is for the first 4975 legacy channel handled by this driver, bit 1 is for 4976 the second channel, and so on. The sequence will often 4977 correspond to the primary legacy channel, the secondary 4978 legacy channel, and so on, but the handling of a PCI 4979 bus and the use of other driver options may interfere 4980 with the sequence. By default IORDY is allowed across 4981 all channels. 4982 4983 pata_legacy.opti82c46x= [HW,LIBATA] 4984 Format: <int> 4985 Set to 1, 2, or 3 for Opti 82c611A on the primary 4986 channel, the secondary channel, or both channels 4987 respectively. Disabled by default. 4988 4989 pata_legacy.opti82c611a= [HW,LIBATA] 4990 Format: <int> 4991 Set to 1, 2, or 3 for Opti 82c465MV on the primary 4992 channel, the secondary channel, or both channels 4993 respectively. Disabled by default. 4994 4995 pata_legacy.pio_mask= [HW,LIBATA] 4996 Format: <int> 4997 PIO mode mask for autospeed devices. Set individual 4998 bits to allow the use of the respective PIO modes. 4999 Bit 0 is for mode 0, bit 1 is for mode 1, and so on. 5000 All modes allowed by default. 5001 5002 pata_legacy.probe_all= [HW,LIBATA] 5003 Format: <int> 5004 Set to non-zero to probe tertiary and further ISA 5005 port ranges on PCI systems. Disabled by default. 5006 5007 pata_legacy.probe_mask= [HW,LIBATA] 5008 Format: <int> 5009 Probe mask for legacy ISA PATA ports. Depending on 5010 platform configuration and the use of other driver 5011 options up to 6 legacy ports are supported: 0x1f0, 5012 0x170, 0x1e8, 0x168, 0x1e0, 0x160, however probing 5013 of individual ports can be disabled by setting the 5014 corresponding bits in the mask to 1. Bit 0 is for 5015 the first port in the list above (0x1f0), and so on. 5016 By default all supported ports are probed. 5017 5018 pata_legacy.qdi= [HW,LIBATA] 5019 Format: <int> 5020 Set to non-zero to probe QDI controllers. By default 5021 set to 1 if CONFIG_PATA_QDI_MODULE, 0 otherwise. 5022 5023 pata_legacy.winbond= [HW,LIBATA] 5024 Format: <int> 5025 Set to non-zero to probe Winbond controllers. Use 5026 the standard I/O port (0x130) if 1, otherwise the 5027 value given is the I/O port to use (typically 0x1b0). 5028 By default set to 1 if CONFIG_PATA_WINBOND_VLB_MODULE, 5029 0 otherwise. 5030 5031 pata_platform.pio_mask= [HW,LIBATA] 5032 Format: <int> 5033 Supported PIO mode mask. Set individual bits to allow 5034 the use of the respective PIO modes. Bit 0 is for 5035 mode 0, bit 1 is for mode 1, and so on. Mode 0 only 5036 allowed by default. 5037 5038 pause_on_oops=<int> 5039 Halt all CPUs after the first oops has been printed for 5040 the specified number of seconds. This is to be used if 5041 your oopses keep scrolling off the screen. 5042 5043 pci=option[,option...] [PCI,EARLY] various PCI subsystem options. 5044 5045 Some options herein operate on a specific device 5046 or a set of devices (<pci_dev>). These are 5047 specified in one of the following formats: 5048 5049 [<domain>:]<bus>:<dev>.<func>[/<dev>.<func>]* 5050 pci:<vendor>:<device>[:<subvendor>:<subdevice>] 5051 5052 Note: the first format specifies a PCI 5053 bus/device/function address which may change 5054 if new hardware is inserted, if motherboard 5055 firmware changes, or due to changes caused 5056 by other kernel parameters. If the 5057 domain is left unspecified, it is 5058 taken to be zero. Optionally, a path 5059 to a device through multiple device/function 5060 addresses can be specified after the base 5061 address (this is more robust against 5062 renumbering issues). The second format 5063 selects devices using IDs from the 5064 configuration space which may match multiple 5065 devices in the system. 5066 5067 earlydump dump PCI config space before the kernel 5068 changes anything 5069 off [X86] don't probe for the PCI bus 5070 bios [X86-32] force use of PCI BIOS, don't access 5071 the hardware directly. Use this if your machine 5072 has a non-standard PCI host bridge. 5073 nobios [X86-32] disallow use of PCI BIOS, only direct 5074 hardware access methods are allowed. Use this 5075 if you experience crashes upon bootup and you 5076 suspect they are caused by the BIOS. 5077 conf1 [X86] Force use of PCI Configuration Access 5078 Mechanism 1 (config address in IO port 0xCF8, 5079 data in IO port 0xCFC, both 32-bit). 5080 conf2 [X86] Force use of PCI Configuration Access 5081 Mechanism 2 (IO port 0xCF8 is an 8-bit port for 5082 the function, IO port 0xCFA, also 8-bit, sets 5083 bus number. The config space is then accessed 5084 through ports 0xC000-0xCFFF). 5085 See http://wiki.osdev.org/PCI for more info 5086 on the configuration access mechanisms. 5087 noaer [PCIE] If the PCIEAER kernel config parameter is 5088 enabled, this kernel boot option can be used to 5089 disable the use of PCIE advanced error reporting. 5090 nodomains [PCI] Disable support for multiple PCI 5091 root domains (aka PCI segments, in ACPI-speak). 5092 nommconf [X86] Disable use of MMCONFIG for PCI 5093 Configuration 5094 check_enable_amd_mmconf [X86] check for and enable 5095 properly configured MMIO access to PCI 5096 config space on AMD family 10h CPU 5097 nomsi [MSI] If the PCI_MSI kernel config parameter is 5098 enabled, this kernel boot option can be used to 5099 disable the use of MSI interrupts system-wide. 5100 noioapicquirk [APIC] Disable all boot interrupt quirks. 5101 Safety option to keep boot IRQs enabled. This 5102 should never be necessary. 5103 ioapicreroute [APIC] Enable rerouting of boot IRQs to the 5104 primary IO-APIC for bridges that cannot disable 5105 boot IRQs. This fixes a source of spurious IRQs 5106 when the system masks IRQs. 5107 noioapicreroute [APIC] Disable workaround that uses the 5108 boot IRQ equivalent of an IRQ that connects to 5109 a chipset where boot IRQs cannot be disabled. 5110 The opposite of ioapicreroute. 5111 biosirq [X86-32] Use PCI BIOS calls to get the interrupt 5112 routing table. These calls are known to be buggy 5113 on several machines and they hang the machine 5114 when used, but on other computers it's the only 5115 way to get the interrupt routing table. Try 5116 this option if the kernel is unable to allocate 5117 IRQs or discover secondary PCI buses on your 5118 motherboard. 5119 rom [X86] Assign address space to expansion ROMs. 5120 Use with caution as certain devices share 5121 address decoders between ROMs and other 5122 resources. 5123 norom [X86] Do not assign address space to 5124 expansion ROMs that do not already have 5125 BIOS assigned address ranges. 5126 nobar [X86] Do not assign address space to the 5127 BARs that weren't assigned by the BIOS. 5128 irqmask=0xMMMM [X86] Set a bit mask of IRQs allowed to be 5129 assigned automatically to PCI devices. You can 5130 make the kernel exclude IRQs of your ISA cards 5131 this way. 5132 pirqaddr=0xAAAAA [X86] Specify the physical address 5133 of the PIRQ table (normally generated 5134 by the BIOS) if it is outside the 5135 F0000h-100000h range. 5136 lastbus=N [X86] Scan all buses thru bus #N. Can be 5137 useful if the kernel is unable to find your 5138 secondary buses and you want to tell it 5139 explicitly which ones they are. 5140 assign-busses [X86] Always assign all PCI bus 5141 numbers ourselves, overriding 5142 whatever the firmware may have done. 5143 usepirqmask [X86] Honor the possible IRQ mask stored 5144 in the BIOS $PIR table. This is needed on 5145 some systems with broken BIOSes, notably 5146 some HP Pavilion N5400 and Omnibook XE3 5147 notebooks. This will have no effect if ACPI 5148 IRQ routing is enabled. 5149 noacpi [X86] Do not use ACPI for IRQ routing 5150 or for PCI scanning. 5151 use_crs [X86] Use PCI host bridge window information 5152 from ACPI. On BIOSes from 2008 or later, this 5153 is enabled by default. If you need to use this, 5154 please report a bug. 5155 nocrs [X86] Ignore PCI host bridge windows from ACPI. 5156 If you need to use this, please report a bug. 5157 use_e820 [X86] Use E820 reservations to exclude parts of 5158 PCI host bridge windows. This is a workaround 5159 for BIOS defects in host bridge _CRS methods. 5160 If you need to use this, please report a bug to 5161 <linux-pci@vger.kernel.org>. 5162 no_e820 [X86] Ignore E820 reservations for PCI host 5163 bridge windows. This is the default on modern 5164 hardware. If you need to use this, please report 5165 a bug to <linux-pci@vger.kernel.org>. 5166 routeirq Do IRQ routing for all PCI devices. 5167 This is normally done in pci_enable_device(), 5168 so this option is a temporary workaround 5169 for broken drivers that don't call it. 5170 skip_isa_align [X86] do not align io start addr, so can 5171 handle more pci cards 5172 noearly [X86] Don't do any early type 1 scanning. 5173 This might help on some broken boards which 5174 machine check when some devices' config space 5175 is read. But various workarounds are disabled 5176 and some IOMMU drivers will not work. 5177 bfsort Sort PCI devices into breadth-first order. 5178 This sorting is done to get a device 5179 order compatible with older (<= 2.4) kernels. 5180 nobfsort Don't sort PCI devices into breadth-first order. 5181 pcie_bus_tune_off Disable PCIe MPS (Max Payload Size) 5182 tuning and use the BIOS-configured MPS defaults. 5183 pcie_bus_safe Set every device's MPS to the largest value 5184 supported by all devices below the root complex. 5185 pcie_bus_perf Set device MPS to the largest allowable MPS 5186 based on its parent bus. Also set MRRS (Max 5187 Read Request Size) to the largest supported 5188 value (no larger than the MPS that the device 5189 or bus can support) for best performance. 5190 pcie_bus_peer2peer Set every device's MPS to 128B, which 5191 every device is guaranteed to support. This 5192 configuration allows peer-to-peer DMA between 5193 any pair of devices, possibly at the cost of 5194 reduced performance. This also guarantees 5195 that hot-added devices will work. 5196 cbiosize=nn[KMG] The fixed amount of bus space which is 5197 reserved for the CardBus bridge's IO window. 5198 The default value is 256 bytes. 5199 cbmemsize=nn[KMG] The fixed amount of bus space which is 5200 reserved for the CardBus bridge's memory 5201 window. The default value is 64 megabytes. 5202 resource_alignment= 5203 Format: 5204 [<order of align>@]<pci_dev>[; ...] 5205 Specifies alignment and device to reassign 5206 aligned memory resources. How to 5207 specify the device is described above. 5208 If <order of align> is not specified, 5209 PAGE_SIZE is used as alignment. 5210 A PCI-PCI bridge can be specified if resource 5211 windows need to be expanded. 5212 To specify the alignment for several 5213 instances of a device, the PCI vendor, 5214 device, subvendor, and subdevice may be 5215 specified, e.g., 12@pci:8086:9c22:103c:198f 5216 for 4096-byte alignment. 5217 ecrc= Enable/disable PCIe ECRC (transaction layer 5218 end-to-end CRC checking). Only effective if 5219 OS has native AER control (either granted by 5220 ACPI _OSC or forced via "pcie_ports=native") 5221 bios: Use BIOS/firmware settings. This is the 5222 the default. 5223 off: Turn ECRC off 5224 on: Turn ECRC on. 5225 hpiosize=nn[KMG] The fixed amount of bus space which is 5226 reserved for hotplug bridge's IO window. 5227 Default size is 256 bytes. 5228 hpmmiosize=nn[KMG] The fixed amount of bus space which is 5229 reserved for hotplug bridge's MMIO window. 5230 Default size is 2 megabytes. 5231 hpmmioprefsize=nn[KMG] The fixed amount of bus space which is 5232 reserved for hotplug bridge's MMIO_PREF window. 5233 Default size is 2 megabytes. 5234 hpmemsize=nn[KMG] The fixed amount of bus space which is 5235 reserved for hotplug bridge's MMIO and 5236 MMIO_PREF window. 5237 Default size is 2 megabytes. 5238 hpbussize=nn The minimum amount of additional bus numbers 5239 reserved for buses below a hotplug bridge. 5240 Default is 1. 5241 realloc= Enable/disable reallocating PCI bridge resources 5242 if allocations done by BIOS are too small to 5243 accommodate resources required by all child 5244 devices. 5245 off: Turn realloc off 5246 on: Turn realloc on 5247 realloc same as realloc=on 5248 noari do not use PCIe ARI. 5249 noats [PCIE, Intel-IOMMU, AMD-IOMMU] 5250 do not use PCIe ATS (and IOMMU device IOTLB). 5251 pcie_scan_all Scan all possible PCIe devices. Otherwise we 5252 only look for one device below a PCIe downstream 5253 port. 5254 big_root_window Try to add a big 64bit memory window to the PCIe 5255 root complex on AMD CPUs. Some GFX hardware 5256 can resize a BAR to allow access to all VRAM. 5257 Adding the window is slightly risky (it may 5258 conflict with unreported devices), so this 5259 taints the kernel. 5260 disable_acs_redir=<pci_dev>[; ...] 5261 Specify one or more PCI devices (in the format 5262 specified above) separated by semicolons. 5263 Each device specified will have the PCI ACS 5264 redirect capabilities forced off which will 5265 allow P2P traffic between devices through 5266 bridges without forcing it upstream. Note: 5267 this removes isolation between devices and 5268 may put more devices in an IOMMU group. 5269 config_acs= 5270 Format: 5271 <ACS flags>@<pci_dev>[; ...] 5272 Specify one or more PCI devices (in the format 5273 specified above) optionally prepended with flags 5274 and separated by semicolons. The respective 5275 capabilities will be enabled, disabled or 5276 unchanged based on what is specified in 5277 flags. 5278 5279 ACS Flags is defined as follows: 5280 bit-0 : ACS Source Validation 5281 bit-1 : ACS Translation Blocking 5282 bit-2 : ACS P2P Request Redirect 5283 bit-3 : ACS P2P Completion Redirect 5284 bit-4 : ACS Upstream Forwarding 5285 bit-5 : ACS P2P Egress Control 5286 bit-6 : ACS Direct Translated P2P 5287 Each bit can be marked as: 5288 '0' – force disabled 5289 '1' – force enabled 5290 'x' – unchanged 5291 For example, 5292 pci=config_acs=10x@pci:0:0 5293 would configure all devices that support 5294 ACS to enable P2P Request Redirect, disable 5295 Translation Blocking, and leave Source 5296 Validation unchanged from whatever power-up 5297 or firmware set it to. 5298 5299 Note: this may remove isolation between devices 5300 and may put more devices in an IOMMU group. 5301 force_floating [S390] Force usage of floating interrupts. 5302 nomio [S390] Do not use MIO instructions. 5303 norid [S390] ignore the RID field and force use of 5304 one PCI domain per PCI function 5305 notph [PCIE] If the PCIE_TPH kernel config parameter 5306 is enabled, this kernel boot option can be used 5307 to disable PCIe TLP Processing Hints support 5308 system-wide. 5309 5310 pcie_aspm= [PCIE] Forcibly enable or ignore PCIe Active State Power 5311 Management. 5312 off Don't touch ASPM configuration at all. Leave any 5313 configuration done by firmware unchanged. 5314 force Enable ASPM even on devices that claim not to support it. 5315 WARNING: Forcing ASPM on may cause system lockups. 5316 5317 pcie_ports= [PCIE] PCIe port services handling: 5318 native Use native PCIe services (PME, AER, DPC, PCIe hotplug) 5319 even if the platform doesn't give the OS permission to 5320 use them. This may cause conflicts if the platform 5321 also tries to use these services. 5322 dpc-native Use native PCIe service for DPC only. May 5323 cause conflicts if firmware uses AER or DPC. 5324 compat Disable native PCIe services (PME, AER, DPC, PCIe 5325 hotplug). 5326 5327 pcie_port_pm= [PCIE] PCIe port power management handling: 5328 off Disable power management of all PCIe ports 5329 force Forcibly enable power management of all PCIe ports 5330 5331 pcie_pme= [PCIE,PM] Native PCIe PME signaling options: 5332 nomsi Do not use MSI for native PCIe PME signaling (this makes 5333 all PCIe root ports use INTx for all services). 5334 5335 pcmv= [HW,PCMCIA] BadgePAD 4 5336 5337 pd_ignore_unused 5338 [PM] 5339 Keep all power-domains already enabled by bootloader on, 5340 even if no driver has claimed them. This is useful 5341 for debug and development, but should not be 5342 needed on a platform with proper driver support. 5343 5344 pdcchassis= [PARISC,HW] Disable/Enable PDC Chassis Status codes at 5345 boot time. 5346 Format: { 0 | 1 } 5347 See arch/parisc/kernel/pdc_chassis.c 5348 5349 percpu_alloc= [MM,EARLY] 5350 Select which percpu first chunk allocator to use. 5351 Currently supported values are "embed" and "page". 5352 Archs may support subset or none of the selections. 5353 See comments in mm/percpu.c for details on each 5354 allocator. This parameter is primarily for debugging 5355 and performance comparison. 5356 5357 pirq= [SMP,APIC] Manual mp-table setup 5358 See Documentation/arch/x86/i386/IO-APIC.rst. 5359 5360 plip= [PPT,NET] Parallel port network link 5361 Format: { parport<nr> | timid | 0 } 5362 See also Documentation/admin-guide/parport.rst. 5363 5364 pmtmr= [X86] Manual setup of pmtmr I/O Port. 5365 Override pmtimer IOPort with a hex value. 5366 e.g. pmtmr=0x508 5367 5368 pmu_override= [PPC] Override the PMU. 5369 This option takes over the PMU facility, so it is no 5370 longer usable by perf. Setting this option starts the 5371 PMU counters by setting MMCR0 to 0 (the FC bit is 5372 cleared). If a number is given, then MMCR1 is set to 5373 that number, otherwise (e.g., 'pmu_override=on'), MMCR1 5374 remains 0. 5375 5376 pm_async= [PM] 5377 Format: off 5378 This parameter sets the initial value of the 5379 /sys/power/pm_async sysfs knob at boot time. 5380 If set to "off", disables asynchronous suspend and 5381 resume of devices during system-wide power transitions. 5382 This can be useful on platforms where device 5383 dependencies are not well-defined, or for debugging 5384 power management issues. Asynchronous operations are 5385 enabled by default. 5386 5387 5388 pm_debug_messages [SUSPEND,KNL] 5389 Enable suspend/resume debug messages during boot up. 5390 5391 pm.dpm_watchdog_enabled= 5392 [PM] Enable or disable the DPM watchdog. Requires 5393 CONFIG_PM_SLEEP and CONFIG_DPM_WATCHDOG enabled. 5394 Format: <bool> 5395 Default value is set by CONFIG_DPM_WATCHDOG_ENABLED. 5396 5397 pnp.debug=1 [PNP] 5398 Enable PNP debug messages (depends on the 5399 CONFIG_PNP_DEBUG_MESSAGES option). Change at run-time 5400 via /sys/module/pnp/parameters/debug. We always show 5401 current resource usage; turning this on also shows 5402 possible settings and some assignment information. 5403 5404 pnpacpi= [ACPI] 5405 { off } 5406 5407 pnpbios= [ISAPNP] 5408 { on | off | curr | res | no-curr | no-res } 5409 5410 pnp_reserve_irq= 5411 [ISAPNP] Exclude IRQs for the autoconfiguration 5412 5413 pnp_reserve_dma= 5414 [ISAPNP] Exclude DMAs for the autoconfiguration 5415 5416 pnp_reserve_io= [ISAPNP] Exclude I/O ports for the autoconfiguration 5417 Ranges are in pairs (I/O port base and size). 5418 5419 pnp_reserve_mem= 5420 [ISAPNP] Exclude memory regions for the 5421 autoconfiguration. 5422 Ranges are in pairs (memory base and size). 5423 5424 ports= [IP_VS_FTP] IPVS ftp helper module 5425 Default is 21. 5426 Up to 8 (IP_VS_APP_MAX_PORTS) ports 5427 may be specified. 5428 Format: <port>,<port>.... 5429 5430 possible_cpus= [SMP,S390,X86] 5431 Format: <unsigned int> 5432 Set the number of possible CPUs, overriding the 5433 regular discovery mechanisms (such as ACPI/FW, etc). 5434 5435 powersave=off [PPC] This option disables power saving features. 5436 It specifically disables cpuidle and sets the 5437 platform machine description specific power_save 5438 function to NULL. On Idle the CPU just reduces 5439 execution priority. 5440 5441 ppc_strict_facility_enable 5442 [PPC,ENABLE] This option catches any kernel floating point, 5443 Altivec, VSX and SPE outside of regions specifically 5444 allowed (eg kernel_enable_fpu()/kernel_disable_fpu()). 5445 There is some performance impact when enabling this. 5446 5447 ppc_tm= [PPC,EARLY] 5448 Format: {"off"} 5449 Disable Hardware Transactional Memory 5450 5451 preempt= [KNL] 5452 Select preemption mode if you have CONFIG_PREEMPT_DYNAMIC 5453 none - Limited to cond_resched() calls 5454 voluntary - Limited to cond_resched() and might_sleep() calls 5455 full - Any section that isn't explicitly preempt disabled 5456 can be preempted anytime. Tasks will also yield 5457 contended spinlocks (if the critical section isn't 5458 explicitly preempt disabled beyond the lock itself). 5459 lazy - Scheduler controlled. Similar to full but instead 5460 of preempting the task immediately, the task gets 5461 one HZ tick time to yield itself before the 5462 preemption will be forced. One preemption is when the 5463 task returns to user space. 5464 5465 print-fatal-signals= 5466 [KNL] debug: print fatal signals 5467 5468 If enabled, warn about various signal handling 5469 related application anomalies: too many signals, 5470 too many POSIX.1 timers, fatal signals causing a 5471 coredump - etc. 5472 5473 If you hit the warning due to signal overflow, 5474 you might want to try "ulimit -i unlimited". 5475 5476 default: off. 5477 5478 printk.always_kmsg_dump= 5479 Trigger kmsg_dump for cases other than kernel oops or 5480 panics 5481 Format: <bool> (1/Y/y=enable, 0/N/n=disable) 5482 default: disabled 5483 5484 printk.console_no_auto_verbose= 5485 Disable console loglevel raise on oops, panic 5486 or lockdep-detected issues (only if lock debug is on). 5487 With an exception to setups with low baudrate on 5488 serial console, keeping this 0 is a good choice 5489 in order to provide more debug information. 5490 Format: <bool> 5491 default: 0 (auto_verbose is enabled) 5492 5493 printk.debug_non_panic_cpus= 5494 Allows storing messages from non-panic CPUs into 5495 the printk log buffer during panic(). They are 5496 flushed to consoles by the panic-CPU on 5497 a best-effort basis. 5498 Format: <bool> (1/Y/y=enable, 0/N/n=disable) 5499 Default: disabled 5500 5501 printk.devkmsg={on,off,ratelimit} 5502 Control writing to /dev/kmsg. 5503 on - unlimited logging to /dev/kmsg from userspace 5504 off - logging to /dev/kmsg disabled 5505 ratelimit - ratelimit the logging 5506 Default: ratelimit 5507 5508 printk.time= Show timing data prefixed to each printk message line 5509 Format: <bool> (1/Y/y=enable, 0/N/n=disable) 5510 5511 proc_mem.force_override= [KNL] 5512 Format: {always | ptrace | never} 5513 Traditionally /proc/pid/mem allows memory permissions to be 5514 overridden without restrictions. This option may be set to 5515 restrict that. Can be one of: 5516 - 'always': traditional behavior always allows mem overrides. 5517 - 'ptrace': only allow mem overrides for active ptracers. 5518 - 'never': never allow mem overrides. 5519 If not specified, default is the CONFIG_PROC_MEM_* choice. 5520 5521 processor.max_cstate= [HW,ACPI] 5522 Limit processor to maximum C-state 5523 max_cstate=9 overrides any DMI blacklist limit. 5524 5525 processor.nocst [HW,ACPI] 5526 Ignore the _CST method to determine C-states, 5527 instead using the legacy FADT method 5528 5529 profile= [KNL] Enable kernel profiling via /proc/profile 5530 Format: [<profiletype>,]<number> 5531 Param: <profiletype>: "schedule" or "kvm" 5532 [defaults to kernel profiling] 5533 Param: "schedule" - profile schedule points. 5534 Param: "kvm" - profile VM exits. 5535 Param: <number> - step/bucket size as a power of 2 for 5536 statistical time based profiling. 5537 5538 prot_virt= [S390] enable hosting protected virtual machines 5539 isolated from the hypervisor (if hardware supports 5540 that). If enabled, the default kernel base address 5541 might be overridden even when Kernel Address Space 5542 Layout Randomization is disabled. 5543 Format: <bool> 5544 5545 psi= [KNL] Enable or disable pressure stall information 5546 tracking. 5547 Format: <bool> 5548 5549 psmouse.proto= [HW,MOUSE] Highest PS2 mouse protocol extension to 5550 probe for; one of (bare|imps|exps|lifebook|any). 5551 psmouse.rate= [HW,MOUSE] Set desired mouse report rate, in reports 5552 per second. 5553 psmouse.resetafter= [HW,MOUSE] 5554 Try to reset the device after so many bad packets 5555 (0 = never). 5556 psmouse.resolution= 5557 [HW,MOUSE] Set desired mouse resolution, in dpi. 5558 psmouse.smartscroll= 5559 [HW,MOUSE] Controls Logitech smartscroll autorepeat. 5560 0 = disabled, 1 = enabled (default). 5561 5562 pstore.backend= Specify the name of the pstore backend to use 5563 5564 pti= [X86-64] Control Page Table Isolation of user and 5565 kernel address spaces. Disabling this feature 5566 removes hardening, but improves performance of 5567 system calls and interrupts. 5568 5569 on - unconditionally enable 5570 off - unconditionally disable 5571 auto - kernel detects whether your CPU model is 5572 vulnerable to issues that PTI mitigates 5573 5574 Not specifying this option is equivalent to pti=auto. 5575 5576 pty.legacy_count= 5577 [KNL] Number of legacy pty's. Overwrites compiled-in 5578 default number. 5579 5580 quiet [KNL,EARLY] Disable most log messages 5581 5582 r128= [HW,DRM] 5583 5584 radix_hcall_invalidate=on [PPC/PSERIES] 5585 Disable RADIX GTSE feature and use hcall for TLB 5586 invalidate. 5587 5588 raid= [HW,RAID] 5589 See Documentation/admin-guide/md.rst. 5590 5591 ramdisk_size= [RAM] Sizes of RAM disks in kilobytes 5592 See Documentation/admin-guide/blockdev/ramdisk.rst. 5593 5594 ramdisk_start= [Deprecated,RAM] RAM disk image start address 5595 5596 random.trust_cpu=off 5597 [KNL,EARLY] Disable trusting the use of the CPU's 5598 random number generator (if available) to 5599 initialize the kernel's RNG. 5600 5601 random.trust_bootloader=off 5602 [KNL,EARLY] Disable trusting the use of the a seed 5603 passed by the bootloader (if available) to 5604 initialize the kernel's RNG. 5605 5606 randomize_kstack_offset= 5607 [KNL,EARLY] Enable or disable kernel stack offset 5608 randomization, which provides roughly 5 bits of 5609 entropy, frustrating memory corruption attacks 5610 that depend on stack address determinism or 5611 cross-syscall address exposures. This is only 5612 available on architectures that have defined 5613 CONFIG_HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET. 5614 Format: <bool> (1/Y/y=enable, 0/N/n=disable) 5615 Default is CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT. 5616 5617 ras=option[,option,...] [KNL] RAS-specific options 5618 5619 cec_disable [X86] 5620 Disable the Correctable Errors Collector, 5621 see CONFIG_RAS_CEC help text. 5622 5623 rcu_nocbs[=cpu-list] 5624 [KNL] The optional argument is a cpu list, 5625 as described above. 5626 5627 In kernels built with CONFIG_RCU_NOCB_CPU=y, 5628 enable the no-callback CPU mode, which prevents 5629 such CPUs' callbacks from being invoked in 5630 softirq context. Invocation of such CPUs' RCU 5631 callbacks will instead be offloaded to "rcuox/N" 5632 kthreads created for that purpose, where "x" is 5633 "p" for RCU-preempt, "s" for RCU-sched, and "g" 5634 for the kthreads that mediate grace periods; and 5635 "N" is the CPU number. This reduces OS jitter on 5636 the offloaded CPUs, which can be useful for HPC 5637 and real-time workloads. It can also improve 5638 energy efficiency for asymmetric multiprocessors. 5639 5640 If a cpulist is passed as an argument, the specified 5641 list of CPUs is set to no-callback mode from boot. 5642 5643 Otherwise, if the '=' sign and the cpulist 5644 arguments are omitted, no CPU will be set to 5645 no-callback mode from boot but the mode may be 5646 toggled at runtime via cpusets. 5647 5648 Note that this argument takes precedence over 5649 the CONFIG_RCU_NOCB_CPU_DEFAULT_ALL option. 5650 5651 rcu_nocb_poll [KNL] 5652 Rather than requiring that offloaded CPUs 5653 (specified by rcu_nocbs= above) explicitly 5654 awaken the corresponding "rcuoN" kthreads, 5655 make these kthreads poll for callbacks. 5656 This improves the real-time response for the 5657 offloaded CPUs by relieving them of the need to 5658 wake up the corresponding kthread, but degrades 5659 energy efficiency by requiring that the kthreads 5660 periodically wake up to do the polling. 5661 5662 rcutree.blimit= [KNL] 5663 Set maximum number of finished RCU callbacks to 5664 process in one batch. 5665 5666 rcutree.csd_lock_suppress_rcu_stall= [KNL] 5667 Do only a one-line RCU CPU stall warning when 5668 there is an ongoing too-long CSD-lock wait. 5669 5670 rcutree.do_rcu_barrier= [KNL] 5671 Request a call to rcu_barrier(). This is 5672 throttled so that userspace tests can safely 5673 hammer on the sysfs variable if they so choose. 5674 If triggered before the RCU grace-period machinery 5675 is fully active, this will error out with EAGAIN. 5676 5677 rcutree.dump_tree= [KNL] 5678 Dump the structure of the rcu_node combining tree 5679 out at early boot. This is used for diagnostic 5680 purposes, to verify correct tree setup. 5681 5682 rcutree.gp_cleanup_delay= [KNL] 5683 Set the number of jiffies to delay each step of 5684 RCU grace-period cleanup. 5685 5686 rcutree.gp_init_delay= [KNL] 5687 Set the number of jiffies to delay each step of 5688 RCU grace-period initialization. 5689 5690 rcutree.gp_preinit_delay= [KNL] 5691 Set the number of jiffies to delay each step of 5692 RCU grace-period pre-initialization, that is, 5693 the propagation of recent CPU-hotplug changes up 5694 the rcu_node combining tree. 5695 5696 rcutree.jiffies_till_first_fqs= [KNL] 5697 Set delay from grace-period initialization to 5698 first attempt to force quiescent states. 5699 Units are jiffies, minimum value is zero, 5700 and maximum value is HZ. 5701 5702 rcutree.jiffies_till_next_fqs= [KNL] 5703 Set delay between subsequent attempts to force 5704 quiescent states. Units are jiffies, minimum 5705 value is one, and maximum value is HZ. 5706 5707 rcutree.jiffies_till_sched_qs= [KNL] 5708 Set required age in jiffies for a 5709 given grace period before RCU starts 5710 soliciting quiescent-state help from 5711 rcu_note_context_switch() and cond_resched(). 5712 If not specified, the kernel will calculate 5713 a value based on the most recent settings 5714 of rcutree.jiffies_till_first_fqs 5715 and rcutree.jiffies_till_next_fqs. 5716 This calculated value may be viewed in 5717 rcutree.jiffies_to_sched_qs. Any attempt to set 5718 rcutree.jiffies_to_sched_qs will be cheerfully 5719 overwritten. 5720 5721 rcutree.kthread_prio= [KNL,BOOT] 5722 Set the SCHED_FIFO priority of the RCU per-CPU 5723 kthreads (rcuc/N). This value is also used for 5724 the priority of the RCU boost threads (rcub/N) 5725 and for the RCU grace-period kthreads (rcu_bh, 5726 rcu_preempt, and rcu_sched). If RCU_BOOST is 5727 set, valid values are 1-99 and the default is 1 5728 (the least-favored priority). Otherwise, when 5729 RCU_BOOST is not set, valid values are 0-99 and 5730 the default is zero (non-realtime operation). 5731 When RCU_NOCB_CPU is set, also adjust the 5732 priority of NOCB callback kthreads. 5733 5734 rcutree.nocb_nobypass_lim_per_jiffy= [KNL] 5735 On callback-offloaded (rcu_nocbs) CPUs, 5736 RCU reduces the lock contention that would 5737 otherwise be caused by callback floods through 5738 use of the ->nocb_bypass list. However, in the 5739 common non-flooded case, RCU queues directly to 5740 the main ->cblist in order to avoid the extra 5741 overhead of the ->nocb_bypass list and its lock. 5742 But if there are too many callbacks queued during 5743 a single jiffy, RCU pre-queues the callbacks into 5744 the ->nocb_bypass queue. The definition of "too 5745 many" is supplied by this kernel boot parameter. 5746 5747 rcutree.nohz_full_patience_delay= [KNL] 5748 On callback-offloaded (rcu_nocbs) CPUs, avoid 5749 disturbing RCU unless the grace period has 5750 reached the specified age in milliseconds. 5751 Defaults to zero. Large values will be capped 5752 at five seconds. All values will be rounded down 5753 to the nearest value representable by jiffies. 5754 5755 rcutree.qhimark= [KNL] 5756 Set threshold of queued RCU callbacks beyond which 5757 batch limiting is disabled. 5758 5759 rcutree.qlowmark= [KNL] 5760 Set threshold of queued RCU callbacks below which 5761 batch limiting is re-enabled. 5762 5763 rcutree.qovld= [KNL] 5764 Set threshold of queued RCU callbacks beyond which 5765 RCU's force-quiescent-state scan will aggressively 5766 enlist help from cond_resched() and sched IPIs to 5767 help CPUs more quickly reach quiescent states. 5768 Set to less than zero to make this be set based 5769 on rcutree.qhimark at boot time and to zero to 5770 disable more aggressive help enlistment. 5771 5772 rcutree.rcu_delay_page_cache_fill_msec= [KNL] 5773 Set the page-cache refill delay (in milliseconds) 5774 in response to low-memory conditions. The range 5775 of permitted values is in the range 0:100000. 5776 5777 rcutree.rcu_divisor= [KNL] 5778 Set the shift-right count to use to compute 5779 the callback-invocation batch limit bl from 5780 the number of callbacks queued on this CPU. 5781 The result will be bounded below by the value of 5782 the rcutree.blimit kernel parameter. Every bl 5783 callbacks, the softirq handler will exit in 5784 order to allow the CPU to do other work. 5785 5786 Please note that this callback-invocation batch 5787 limit applies only to non-offloaded callback 5788 invocation. Offloaded callbacks are instead 5789 invoked in the context of an rcuoc kthread, which 5790 scheduler will preempt as it does any other task. 5791 5792 rcutree.rcu_fanout_exact= [KNL] 5793 Disable autobalancing of the rcu_node combining 5794 tree. This is used by rcutorture, and might 5795 possibly be useful for architectures having high 5796 cache-to-cache transfer latencies. 5797 5798 rcutree.rcu_fanout_leaf= [KNL] 5799 Change the number of CPUs assigned to each 5800 leaf rcu_node structure. Useful for very 5801 large systems, which will choose the value 64, 5802 and for NUMA systems with large remote-access 5803 latencies, which will choose a value aligned 5804 with the appropriate hardware boundaries. 5805 5806 rcutree.rcu_min_cached_objs= [KNL] 5807 Minimum number of objects which are cached and 5808 maintained per one CPU. Object size is equal 5809 to PAGE_SIZE. The cache allows to reduce the 5810 pressure to page allocator, also it makes the 5811 whole algorithm to behave better in low memory 5812 condition. 5813 5814 rcutree.rcu_nocb_gp_stride= [KNL] 5815 Set the number of NOCB callback kthreads in 5816 each group, which defaults to the square root 5817 of the number of CPUs. Larger numbers reduce 5818 the wakeup overhead on the global grace-period 5819 kthread, but increases that same overhead on 5820 each group's NOCB grace-period kthread. 5821 5822 rcutree.rcu_kick_kthreads= [KNL] 5823 Cause the grace-period kthread to get an extra 5824 wake_up() if it sleeps three times longer than 5825 it should at force-quiescent-state time. 5826 This wake_up() will be accompanied by a 5827 WARN_ONCE() splat and an ftrace_dump(). 5828 5829 rcutree.rcu_resched_ns= [KNL] 5830 Limit the time spend invoking a batch of RCU 5831 callbacks to the specified number of nanoseconds. 5832 By default, this limit is checked only once 5833 every 32 callbacks in order to limit the pain 5834 inflicted by local_clock() overhead. 5835 5836 rcutree.rcu_unlock_delay= [KNL] 5837 In CONFIG_RCU_STRICT_GRACE_PERIOD=y kernels, 5838 this specifies an rcu_read_unlock()-time delay 5839 in microseconds. This defaults to zero. 5840 Larger delays increase the probability of 5841 catching RCU pointer leaks, that is, buggy use 5842 of RCU-protected pointers after the relevant 5843 rcu_read_unlock() has completed. 5844 5845 rcutree.sysrq_rcu= [KNL] 5846 Commandeer a sysrq key to dump out Tree RCU's 5847 rcu_node tree with an eye towards determining 5848 why a new grace period has not yet started. 5849 5850 rcutree.use_softirq= [KNL] 5851 If set to zero, move all RCU_SOFTIRQ processing to 5852 per-CPU rcuc kthreads. Defaults to a non-zero 5853 value, meaning that RCU_SOFTIRQ is used by default. 5854 Specify rcutree.use_softirq=0 to use rcuc kthreads. 5855 5856 But note that CONFIG_PREEMPT_RT=y kernels disable 5857 this kernel boot parameter, forcibly setting it 5858 to zero. 5859 5860 rcutree.enable_rcu_lazy= [KNL] 5861 To save power, batch RCU callbacks and flush after 5862 delay, memory pressure or callback list growing too 5863 big. 5864 5865 rcutree.rcu_normal_wake_from_gp= [KNL] 5866 Reduces a latency of synchronize_rcu() call. This approach 5867 maintains its own track of synchronize_rcu() callers, so it 5868 does not interact with regular callbacks because it does not 5869 use a call_rcu[_hurry]() path. Please note, this is for a 5870 normal grace period. 5871 5872 How to disable it: 5873 5874 echo 0 > /sys/module/rcutree/parameters/rcu_normal_wake_from_gp 5875 or pass a boot parameter "rcutree.rcu_normal_wake_from_gp=0" 5876 5877 Default is 1 if it is not explicitly disabled by the boot parameter 5878 passing 0. 5879 5880 rcuscale.gp_async= [KNL] 5881 Measure performance of asynchronous 5882 grace-period primitives such as call_rcu(). 5883 5884 rcuscale.gp_async_max= [KNL] 5885 Specify the maximum number of outstanding 5886 callbacks per writer thread. When a writer 5887 thread exceeds this limit, it invokes the 5888 corresponding flavor of rcu_barrier() to allow 5889 previously posted callbacks to drain. 5890 5891 rcuscale.gp_exp= [KNL] 5892 Measure performance of expedited synchronous 5893 grace-period primitives. 5894 5895 rcuscale.holdoff= [KNL] 5896 Set test-start holdoff period. The purpose of 5897 this parameter is to delay the start of the 5898 test until boot completes in order to avoid 5899 interference. 5900 5901 rcuscale.kfree_by_call_rcu= [KNL] 5902 In kernels built with CONFIG_RCU_LAZY=y, test 5903 call_rcu() instead of kfree_rcu(). 5904 5905 rcuscale.kfree_mult= [KNL] 5906 Instead of allocating an object of size kfree_obj, 5907 allocate one of kfree_mult * sizeof(kfree_obj). 5908 Defaults to 1. 5909 5910 rcuscale.kfree_rcu_test= [KNL] 5911 Set to measure performance of kfree_rcu() flooding. 5912 5913 rcuscale.kfree_rcu_test_double= [KNL] 5914 Test the double-argument variant of kfree_rcu(). 5915 If this parameter has the same value as 5916 rcuscale.kfree_rcu_test_single, both the single- 5917 and double-argument variants are tested. 5918 5919 rcuscale.kfree_rcu_test_single= [KNL] 5920 Test the single-argument variant of kfree_rcu(). 5921 If this parameter has the same value as 5922 rcuscale.kfree_rcu_test_double, both the single- 5923 and double-argument variants are tested. 5924 5925 rcuscale.kfree_nthreads= [KNL] 5926 The number of threads running loops of kfree_rcu(). 5927 5928 rcuscale.kfree_alloc_num= [KNL] 5929 Number of allocations and frees done in an iteration. 5930 5931 rcuscale.kfree_loops= [KNL] 5932 Number of loops doing rcuscale.kfree_alloc_num number 5933 of allocations and frees. 5934 5935 rcuscale.minruntime= [KNL] 5936 Set the minimum test run time in seconds. This 5937 does not affect the data-collection interval, 5938 but instead allows better measurement of things 5939 like CPU consumption. 5940 5941 rcuscale.nreaders= [KNL] 5942 Set number of RCU readers. The value -1 selects 5943 N, where N is the number of CPUs. A value 5944 "n" less than -1 selects N-n+1, where N is again 5945 the number of CPUs. For example, -2 selects N 5946 (the number of CPUs), -3 selects N+1, and so on. 5947 A value of "n" less than or equal to -N selects 5948 a single reader. 5949 5950 rcuscale.nwriters= [KNL] 5951 Set number of RCU writers. The values operate 5952 the same as for rcuscale.nreaders. 5953 N, where N is the number of CPUs 5954 5955 rcuscale.scale_type= [KNL] 5956 Specify the RCU implementation to test. 5957 5958 rcuscale.shutdown= [KNL] 5959 Shut the system down after performance tests 5960 complete. This is useful for hands-off automated 5961 testing. 5962 5963 rcuscale.verbose= [KNL] 5964 Enable additional printk() statements. 5965 5966 rcuscale.writer_holdoff= [KNL] 5967 Write-side holdoff between grace periods, 5968 in microseconds. The default of zero says 5969 no holdoff. 5970 5971 rcuscale.writer_holdoff_jiffies= [KNL] 5972 Additional write-side holdoff between grace 5973 periods, but in jiffies. The default of zero 5974 says no holdoff. 5975 5976 rcutorture.fqs_duration= [KNL] 5977 Set duration of force_quiescent_state bursts 5978 in microseconds. 5979 5980 rcutorture.fqs_holdoff= [KNL] 5981 Set holdoff time within force_quiescent_state bursts 5982 in microseconds. 5983 5984 rcutorture.fqs_stutter= [KNL] 5985 Set wait time between force_quiescent_state bursts 5986 in seconds. 5987 5988 rcutorture.fwd_progress= [KNL] 5989 Specifies the number of kthreads to be used 5990 for RCU grace-period forward-progress testing 5991 for the types of RCU supporting this notion. 5992 Defaults to 1 kthread, values less than zero or 5993 greater than the number of CPUs cause the number 5994 of CPUs to be used. 5995 5996 rcutorture.fwd_progress_div= [KNL] 5997 Specify the fraction of a CPU-stall-warning 5998 period to do tight-loop forward-progress testing. 5999 6000 rcutorture.fwd_progress_holdoff= [KNL] 6001 Number of seconds to wait between successive 6002 forward-progress tests. 6003 6004 rcutorture.fwd_progress_need_resched= [KNL] 6005 Enclose cond_resched() calls within checks for 6006 need_resched() during tight-loop forward-progress 6007 testing. 6008 6009 rcutorture.gp_cond= [KNL] 6010 Use conditional/asynchronous update-side 6011 normal-grace-period primitives, if available. 6012 6013 rcutorture.gp_cond_exp= [KNL] 6014 Use conditional/asynchronous update-side 6015 expedited-grace-period primitives, if available. 6016 6017 rcutorture.gp_cond_full= [KNL] 6018 Use conditional/asynchronous update-side 6019 normal-grace-period primitives that also take 6020 concurrent expedited grace periods into account, 6021 if available. 6022 6023 rcutorture.gp_cond_exp_full= [KNL] 6024 Use conditional/asynchronous update-side 6025 expedited-grace-period primitives that also take 6026 concurrent normal grace periods into account, 6027 if available. 6028 6029 rcutorture.gp_cond_wi= [KNL] 6030 Nominal wait interval for normal conditional 6031 grace periods (specified by rcutorture's 6032 gp_cond and gp_cond_full module parameters), 6033 in microseconds. The actual wait interval will 6034 be randomly selected to nanosecond granularity up 6035 to this wait interval. Defaults to 16 jiffies, 6036 for example, 16,000 microseconds on a system 6037 with HZ=1000. 6038 6039 rcutorture.gp_cond_wi_exp= [KNL] 6040 Nominal wait interval for expedited conditional 6041 grace periods (specified by rcutorture's 6042 gp_cond_exp and gp_cond_exp_full module 6043 parameters), in microseconds. The actual wait 6044 interval will be randomly selected to nanosecond 6045 granularity up to this wait interval. Defaults to 6046 128 microseconds. 6047 6048 rcutorture.gp_exp= [KNL] 6049 Use expedited update-side primitives, if available. 6050 6051 rcutorture.gp_normal= [KNL] 6052 Use normal (non-expedited) asynchronous 6053 update-side primitives, if available. 6054 6055 rcutorture.gp_poll= [KNL] 6056 Use polled update-side normal-grace-period 6057 primitives, if available. 6058 6059 rcutorture.gp_poll_exp= [KNL] 6060 Use polled update-side expedited-grace-period 6061 primitives, if available. 6062 6063 rcutorture.gp_poll_full= [KNL] 6064 Use polled update-side normal-grace-period 6065 primitives that also take concurrent expedited 6066 grace periods into account, if available. 6067 6068 rcutorture.gp_poll_exp_full= [KNL] 6069 Use polled update-side expedited-grace-period 6070 primitives that also take concurrent normal 6071 grace periods into account, if available. 6072 6073 rcutorture.gp_poll_wi= [KNL] 6074 Nominal wait interval for normal conditional 6075 grace periods (specified by rcutorture's 6076 gp_poll and gp_poll_full module parameters), 6077 in microseconds. The actual wait interval will 6078 be randomly selected to nanosecond granularity up 6079 to this wait interval. Defaults to 16 jiffies, 6080 for example, 16,000 microseconds on a system 6081 with HZ=1000. 6082 6083 rcutorture.gp_poll_wi_exp= [KNL] 6084 Nominal wait interval for expedited conditional 6085 grace periods (specified by rcutorture's 6086 gp_poll_exp and gp_poll_exp_full module 6087 parameters), in microseconds. The actual wait 6088 interval will be randomly selected to nanosecond 6089 granularity up to this wait interval. Defaults to 6090 128 microseconds. 6091 6092 rcutorture.gp_sync= [KNL] 6093 Use normal (non-expedited) synchronous 6094 update-side primitives, if available. If all 6095 of rcutorture.gp_cond=, rcutorture.gp_exp=, 6096 rcutorture.gp_normal=, and rcutorture.gp_sync= 6097 are zero, rcutorture acts as if is interpreted 6098 they are all non-zero. 6099 6100 rcutorture.gpwrap_lag= [KNL] 6101 Enable grace-period wrap lag testing. Setting 6102 to false prevents the gpwrap lag test from 6103 running. Default is true. 6104 6105 rcutorture.gpwrap_lag_gps= [KNL] 6106 Set the value for grace-period wrap lag during 6107 active lag testing periods. This controls how many 6108 grace periods differences we tolerate between 6109 rdp and rnp's gp_seq before setting overflow flag. 6110 The default is always set to 8. 6111 6112 rcutorture.gpwrap_lag_cycle_mins= [KNL] 6113 Set the total cycle duration for gpwrap lag 6114 testing in minutes. This is the total time for 6115 one complete cycle of active and inactive 6116 testing periods. Default is 30 minutes. 6117 6118 rcutorture.gpwrap_lag_active_mins= [KNL] 6119 Set the duration for which gpwrap lag is active 6120 within each cycle, in minutes. During this time, 6121 the grace-period wrap lag will be set to the 6122 value specified by gpwrap_lag_gps. Default is 6123 5 minutes. 6124 6125 rcutorture.irqreader= [KNL] 6126 Run RCU readers from irq handlers, or, more 6127 accurately, from a timer handler. Not all RCU 6128 flavors take kindly to this sort of thing. 6129 6130 rcutorture.leakpointer= [KNL] 6131 Leak an RCU-protected pointer out of the reader. 6132 This can of course result in splats, and is 6133 intended to test the ability of things like 6134 CONFIG_RCU_STRICT_GRACE_PERIOD=y to detect 6135 such leaks. 6136 6137 rcutorture.n_barrier_cbs= [KNL] 6138 Set callbacks/threads for rcu_barrier() testing. 6139 6140 rcutorture.nfakewriters= [KNL] 6141 Set number of concurrent RCU writers. These just 6142 stress RCU, they don't participate in the actual 6143 test, hence the "fake". 6144 6145 rcutorture.nocbs_nthreads= [KNL] 6146 Set number of RCU callback-offload togglers. 6147 Zero (the default) disables toggling. 6148 6149 rcutorture.nocbs_toggle= [KNL] 6150 Set the delay in milliseconds between successive 6151 callback-offload toggling attempts. 6152 6153 rcutorture.nreaders= [KNL] 6154 Set number of RCU readers. The value -1 selects 6155 N-1, where N is the number of CPUs. A value 6156 "n" less than -1 selects N-n-2, where N is again 6157 the number of CPUs. For example, -2 selects N 6158 (the number of CPUs), -3 selects N+1, and so on. 6159 6160 rcutorture.object_debug= [KNL] 6161 Enable debug-object double-call_rcu() testing. 6162 6163 rcutorture.onoff_holdoff= [KNL] 6164 Set time (s) after boot for CPU-hotplug testing. 6165 6166 rcutorture.onoff_interval= [KNL] 6167 Set time (jiffies) between CPU-hotplug operations, 6168 or zero to disable CPU-hotplug testing. 6169 6170 rcutorture.preempt_duration= [KNL] 6171 Set duration (in milliseconds) of preemptions 6172 by a high-priority FIFO real-time task. Set to 6173 zero (the default) to disable. The CPUs to 6174 preempt are selected randomly from the set that 6175 are online at a given point in time. Races with 6176 CPUs going offline are ignored, with that attempt 6177 at preemption skipped. 6178 6179 rcutorture.preempt_interval= [KNL] 6180 Set interval (in milliseconds, defaulting to one 6181 second) between preemptions by a high-priority 6182 FIFO real-time task. This delay is mediated 6183 by an hrtimer and is further fuzzed to avoid 6184 inadvertent synchronizations. 6185 6186 rcutorture.read_exit_burst= [KNL] 6187 The number of times in a given read-then-exit 6188 episode that a set of read-then-exit kthreads 6189 is spawned. 6190 6191 rcutorture.read_exit_delay= [KNL] 6192 The delay, in seconds, between successive 6193 read-then-exit testing episodes. 6194 6195 rcutorture.reader_flavor= [KNL] 6196 A bit mask indicating which readers to use. 6197 If there is more than one bit set, the readers 6198 are entered from low-order bit up, and are 6199 exited in the opposite order. For SRCU, the 6200 0x1 bit is normal readers, 0x2 NMI-safe readers, 6201 and 0x4 light-weight readers. 6202 6203 rcutorture.shuffle_interval= [KNL] 6204 Set task-shuffle interval (s). Shuffling tasks 6205 allows some CPUs to go into dyntick-idle mode 6206 during the rcutorture test. 6207 6208 rcutorture.shutdown_secs= [KNL] 6209 Set time (s) after boot system shutdown. This 6210 is useful for hands-off automated testing. 6211 6212 rcutorture.stall_cpu= [KNL] 6213 Duration of CPU stall (s) to test RCU CPU stall 6214 warnings, zero to disable. 6215 6216 rcutorture.stall_cpu_block= [KNL] 6217 Sleep while stalling if set. This will result 6218 in warnings from preemptible RCU in addition to 6219 any other stall-related activity. Note that 6220 in kernels built with CONFIG_PREEMPTION=n and 6221 CONFIG_PREEMPT_COUNT=y, this parameter will 6222 cause the CPU to pass through a quiescent state. 6223 Given CONFIG_PREEMPTION=n, this will suppress 6224 RCU CPU stall warnings, but will instead result 6225 in scheduling-while-atomic splats. 6226 6227 Use of this module parameter results in splats. 6228 6229 6230 rcutorture.stall_cpu_holdoff= [KNL] 6231 Time to wait (s) after boot before inducing stall. 6232 6233 rcutorture.stall_cpu_irqsoff= [KNL] 6234 Disable interrupts while stalling if set, but only 6235 on the first stall in the set. 6236 6237 rcutorture.stall_cpu_repeat= [KNL] 6238 Number of times to repeat the stall sequence, 6239 so that rcutorture.stall_cpu_repeat=3 will result 6240 in four stall sequences. 6241 6242 rcutorture.stall_gp_kthread= [KNL] 6243 Duration (s) of forced sleep within RCU 6244 grace-period kthread to test RCU CPU stall 6245 warnings, zero to disable. If both stall_cpu 6246 and stall_gp_kthread are specified, the 6247 kthread is starved first, then the CPU. 6248 6249 rcutorture.stat_interval= [KNL] 6250 Time (s) between statistics printk()s. 6251 6252 rcutorture.stutter= [KNL] 6253 Time (s) to stutter testing, for example, specifying 6254 five seconds causes the test to run for five seconds, 6255 wait for five seconds, and so on. This tests RCU's 6256 ability to transition abruptly to and from idle. 6257 6258 rcutorture.test_boost= [KNL] 6259 Test RCU priority boosting? 0=no, 1=maybe, 2=yes. 6260 "Maybe" means test if the RCU implementation 6261 under test support RCU priority boosting. 6262 6263 rcutorture.test_boost_duration= [KNL] 6264 Duration (s) of each individual boost test. 6265 6266 rcutorture.test_boost_holdoff= [KNL] 6267 Holdoff time (s) from start of test to the start 6268 of RCU priority-boost testing. Defaults to zero, 6269 that is, no holdoff. 6270 6271 rcutorture.test_boost_interval= [KNL] 6272 Interval (s) between each boost test. 6273 6274 rcutorture.test_no_idle_hz= [KNL] 6275 Test RCU's dyntick-idle handling. See also the 6276 rcutorture.shuffle_interval parameter. 6277 6278 rcutorture.torture_type= [KNL] 6279 Specify the RCU implementation to test. 6280 6281 rcutorture.verbose= [KNL] 6282 Enable additional printk() statements. 6283 6284 rcupdate.rcu_cpu_stall_ftrace_dump= [KNL] 6285 Dump ftrace buffer after reporting RCU CPU 6286 stall warning. 6287 6288 rcupdate.rcu_cpu_stall_notifiers= [KNL] 6289 Provide RCU CPU stall notifiers, but see the 6290 warnings in the RCU_CPU_STALL_NOTIFIER Kconfig 6291 option's help text. TL;DR: You almost certainly 6292 do not want rcupdate.rcu_cpu_stall_notifiers. 6293 6294 rcupdate.rcu_cpu_stall_suppress= [KNL] 6295 Suppress RCU CPU stall warning messages. 6296 6297 rcupdate.rcu_cpu_stall_suppress_at_boot= [KNL] 6298 Suppress RCU CPU stall warning messages and 6299 rcutorture writer stall warnings that occur 6300 during early boot, that is, during the time 6301 before the init task is spawned. 6302 6303 rcupdate.rcu_cpu_stall_timeout= [KNL] 6304 Set timeout for RCU CPU stall warning messages. 6305 The value is in seconds and the maximum allowed 6306 value is 300 seconds. 6307 6308 rcupdate.rcu_exp_cpu_stall_timeout= [KNL] 6309 Set timeout for expedited RCU CPU stall warning 6310 messages. The value is in milliseconds 6311 and the maximum allowed value is 21000 6312 milliseconds. Please note that this value is 6313 adjusted to an arch timer tick resolution. 6314 Setting this to zero causes the value from 6315 rcupdate.rcu_cpu_stall_timeout to be used (after 6316 conversion from seconds to milliseconds). 6317 6318 rcupdate.rcu_cpu_stall_cputime= [KNL] 6319 Provide statistics on the cputime and count of 6320 interrupts and tasks during the sampling period. For 6321 multiple continuous RCU stalls, all sampling periods 6322 begin at half of the first RCU stall timeout. 6323 6324 rcupdate.rcu_exp_stall_task_details= [KNL] 6325 Print stack dumps of any tasks blocking the 6326 current expedited RCU grace period during an 6327 expedited RCU CPU stall warning. 6328 6329 rcupdate.rcu_expedited= [KNL] 6330 Use expedited grace-period primitives, for 6331 example, synchronize_rcu_expedited() instead 6332 of synchronize_rcu(). This reduces latency, 6333 but can increase CPU utilization, degrade 6334 real-time latency, and degrade energy efficiency. 6335 No effect on CONFIG_TINY_RCU kernels. 6336 6337 rcupdate.rcu_normal= [KNL] 6338 Use only normal grace-period primitives, 6339 for example, synchronize_rcu() instead of 6340 synchronize_rcu_expedited(). This improves 6341 real-time latency, CPU utilization, and 6342 energy efficiency, but can expose users to 6343 increased grace-period latency. This parameter 6344 overrides rcupdate.rcu_expedited. No effect on 6345 CONFIG_TINY_RCU kernels. 6346 6347 rcupdate.rcu_normal_after_boot= [KNL] 6348 Once boot has completed (that is, after 6349 rcu_end_inkernel_boot() has been invoked), use 6350 only normal grace-period primitives. No effect 6351 on CONFIG_TINY_RCU kernels. 6352 6353 But note that CONFIG_PREEMPT_RT=y kernels enables 6354 this kernel boot parameter, forcibly setting 6355 it to the value one, that is, converting any 6356 post-boot attempt at an expedited RCU grace 6357 period to instead use normal non-expedited 6358 grace-period processing. 6359 6360 rcupdate.rcu_task_collapse_lim= [KNL] 6361 Set the maximum number of callbacks present 6362 at the beginning of a grace period that allows 6363 the RCU Tasks flavors to collapse back to using 6364 a single callback queue. This switching only 6365 occurs when rcupdate.rcu_task_enqueue_lim is 6366 set to the default value of -1. 6367 6368 rcupdate.rcu_task_contend_lim= [KNL] 6369 Set the minimum number of callback-queuing-time 6370 lock-contention events per jiffy required to 6371 cause the RCU Tasks flavors to switch to per-CPU 6372 callback queuing. This switching only occurs 6373 when rcupdate.rcu_task_enqueue_lim is set to 6374 the default value of -1. 6375 6376 rcupdate.rcu_task_enqueue_lim= [KNL] 6377 Set the number of callback queues to use for the 6378 RCU Tasks family of RCU flavors. The default 6379 of -1 allows this to be automatically (and 6380 dynamically) adjusted. This parameter is intended 6381 for use in testing. 6382 6383 rcupdate.rcu_task_lazy_lim= [KNL] 6384 Number of callbacks on a given CPU that will 6385 cancel laziness on that CPU. Use -1 to disable 6386 cancellation of laziness, but be advised that 6387 doing so increases the danger of OOM due to 6388 callback flooding. 6389 6390 rcupdate.rcu_task_stall_info= [KNL] 6391 Set initial timeout in jiffies for RCU task stall 6392 informational messages, which give some indication 6393 of the problem for those not patient enough to 6394 wait for ten minutes. Informational messages are 6395 only printed prior to the stall-warning message 6396 for a given grace period. Disable with a value 6397 less than or equal to zero. Defaults to ten 6398 seconds. A change in value does not take effect 6399 until the beginning of the next grace period. 6400 6401 rcupdate.rcu_task_stall_info_mult= [KNL] 6402 Multiplier for time interval between successive 6403 RCU task stall informational messages for a given 6404 RCU tasks grace period. This value is clamped 6405 to one through ten, inclusive. It defaults to 6406 the value three, so that the first informational 6407 message is printed 10 seconds into the grace 6408 period, the second at 40 seconds, the third at 6409 160 seconds, and then the stall warning at 600 6410 seconds would prevent a fourth at 640 seconds. 6411 6412 rcupdate.rcu_task_stall_timeout= [KNL] 6413 Set timeout in jiffies for RCU task stall 6414 warning messages. Disable with a value less 6415 than or equal to zero. Defaults to ten minutes. 6416 A change in value does not take effect until 6417 the beginning of the next grace period. 6418 6419 rcupdate.rcu_tasks_lazy_ms= [KNL] 6420 Set timeout in milliseconds RCU Tasks asynchronous 6421 callback batching for call_rcu_tasks(). 6422 A negative value will take the default. A value 6423 of zero will disable batching. Batching is 6424 always disabled for synchronize_rcu_tasks(). 6425 6426 rcupdate.rcu_self_test= [KNL] 6427 Run the RCU early boot self tests 6428 6429 rdinit= [KNL] 6430 Format: <full_path> 6431 Run specified binary instead of /init from the ramdisk, 6432 used for early userspace startup. See initrd. 6433 6434 rdrand= [X86,EARLY] 6435 force - Override the decision by the kernel to hide the 6436 advertisement of RDRAND support (this affects 6437 certain AMD processors because of buggy BIOS 6438 support, specifically around the suspend/resume 6439 path). 6440 6441 rdt= [HW,X86,RDT] 6442 Turn on/off individual RDT features. List is: 6443 cmt, mbmtotal, mbmlocal, l3cat, l3cdp, l2cat, l2cdp, 6444 mba, smba, bmec, abmc, sdciae, energy[:guid], 6445 perf[:guid]. 6446 E.g. to turn on cmt and turn off mba use: 6447 rdt=cmt,!mba 6448 To turn off all energy telemetry monitoring and ensure that 6449 perf telemetry monitoring associated with guid 0x12345 6450 is enabled use: 6451 rdt=!energy,perf:0x12345 6452 6453 reboot= [KNL] 6454 Format (x86 or x86_64): 6455 [w[arm] | c[old] | h[ard] | s[oft] | g[pio]] | d[efault] \ 6456 [[,]s[mp]####] \ 6457 [[,]b[ios] | a[cpi] | k[bd] | t[riple] | e[fi] | p[ci]] \ 6458 [[,]f[orce]] 6459 Where reboot_mode is one of warm (soft) or cold (hard) or gpio 6460 (prefix with 'panic_' to set mode for panic 6461 reboot only), 6462 reboot_type is one of bios, acpi, kbd, triple, efi, or pci, 6463 reboot_force is either force or not specified, 6464 reboot_cpu is s[mp]#### with #### being the processor 6465 to be used for rebooting. 6466 6467 acpi 6468 Use the ACPI RESET_REG in the FADT. If ACPI is not 6469 configured or the ACPI reset does not work, the reboot 6470 path attempts the reset using the keyboard controller. 6471 6472 bios 6473 Use the CPU reboot vector for warm reset 6474 6475 cold 6476 Set the cold reboot flag 6477 6478 default 6479 There are some built-in platform specific "quirks" 6480 - you may see: "reboot: <name> series board detected. 6481 Selecting <type> for reboots." In the case where you 6482 think the quirk is in error (e.g. you have newer BIOS, 6483 or newer board) using this option will ignore the 6484 built-in quirk table, and use the generic default 6485 reboot actions. 6486 6487 efi 6488 Use efi reset_system runtime service. If EFI is not 6489 configured or the EFI reset does not work, the reboot 6490 path attempts the reset using the keyboard controller. 6491 6492 force 6493 Don't stop other CPUs on reboot. This can make reboot 6494 more reliable in some cases. 6495 6496 kbd 6497 Use the keyboard controller. cold reset (default) 6498 6499 pci 6500 Use a write to the PCI config space register 0xcf9 to 6501 trigger reboot. 6502 6503 triple 6504 Force a triple fault (init) 6505 6506 warm 6507 Don't set the cold reboot flag 6508 6509 Using warm reset will be much faster especially on big 6510 memory systems because the BIOS will not go through 6511 the memory check. Disadvantage is that not all 6512 hardware will be completely reinitialized on reboot so 6513 there may be boot problems on some systems. 6514 6515 6516 refscale.holdoff= [KNL] 6517 Set test-start holdoff period. The purpose of 6518 this parameter is to delay the start of the 6519 test until boot completes in order to avoid 6520 interference. 6521 6522 refscale.lookup_instances= [KNL] 6523 Number of data elements to use for the forms of 6524 SLAB_TYPESAFE_BY_RCU testing. A negative number 6525 is negated and multiplied by nr_cpu_ids, while 6526 zero specifies nr_cpu_ids. 6527 6528 refscale.loops= [KNL] 6529 Set the number of loops over the synchronization 6530 primitive under test. Increasing this number 6531 reduces noise due to loop start/end overhead, 6532 but the default has already reduced the per-pass 6533 noise to a handful of picoseconds on ca. 2020 6534 x86 laptops. 6535 6536 refscale.nreaders= [KNL] 6537 Set number of readers. The default value of -1 6538 selects N, where N is roughly 75% of the number 6539 of CPUs. A value of zero is an interesting choice. 6540 6541 refscale.nruns= [KNL] 6542 Set number of runs, each of which is dumped onto 6543 the console log. 6544 6545 refscale.readdelay= [KNL] 6546 Set the read-side critical-section duration, 6547 measured in microseconds. 6548 6549 refscale.scale_type= [KNL] 6550 Specify the read-protection implementation to test. 6551 6552 refscale.shutdown= [KNL] 6553 Shut down the system at the end of the performance 6554 test. This defaults to 1 (shut it down) when 6555 refscale is built into the kernel and to 0 (leave 6556 it running) when refscale is built as a module. 6557 6558 refscale.verbose= [KNL] 6559 Enable additional printk() statements. 6560 6561 refscale.verbose_batched= [KNL] 6562 Batch the additional printk() statements. If zero 6563 (the default) or negative, print everything. Otherwise, 6564 print every Nth verbose statement, where N is the value 6565 specified. 6566 6567 regulator_ignore_unused 6568 [REGULATOR] 6569 Prevents regulator framework from disabling regulators 6570 that are unused, due no driver claiming them. This may 6571 be useful for debug and development, but should not be 6572 needed on a platform with proper driver support. 6573 6574 relax_domain_level= 6575 [KNL, SMP] Set scheduler's default relax_domain_level. 6576 See Documentation/admin-guide/cgroup-v1/cpusets.rst. 6577 6578 reserve= [KNL,BUGS] Force kernel to ignore I/O ports or memory 6579 Format: <base1>,<size1>[,<base2>,<size2>,...] 6580 Reserve I/O ports or memory so the kernel won't use 6581 them. If <base> is less than 0x10000, the region 6582 is assumed to be I/O ports; otherwise it is memory. 6583 6584 reserve_mem= [RAM] 6585 Format: nn[KMG]:<align>:<label> 6586 Reserve physical memory and label it with a name that 6587 other subsystems can use to access it. This is typically 6588 used for systems that do not wipe the RAM, and this command 6589 line will try to reserve the same physical memory on 6590 soft reboots. Note, it is not guaranteed to be the same 6591 location. For example, if anything about the system changes 6592 or if booting a different kernel. It can also fail if KASLR 6593 places the kernel at the location of where the RAM reservation 6594 was from a previous boot, the new reservation will be at a 6595 different location. 6596 Any subsystem using this feature must add a way to verify 6597 that the contents of the physical memory is from a previous 6598 boot, as there may be cases where the memory will not be 6599 located at the same location. 6600 6601 The format is size:align:label for example, to request 6602 12 megabytes of 4096 alignment for ramoops: 6603 6604 reserve_mem=12M:4096:oops ramoops.mem_name=oops 6605 6606 reservetop= [X86-32,EARLY] 6607 Format: nn[KMG] 6608 Reserves a hole at the top of the kernel virtual 6609 address space. 6610 6611 reset_devices [KNL] Force drivers to reset the underlying device 6612 during initialization. 6613 6614 resume= [SWSUSP] 6615 Specify the partition device for software suspend 6616 Format: 6617 {/dev/<dev> | PARTUUID=<uuid> | <int>:<int> | <hex>} 6618 6619 resume_offset= [SWSUSP] 6620 Specify the offset from the beginning of the partition 6621 given by "resume=" at which the swap header is located, 6622 in <PAGE_SIZE> units (needed only for swap files). 6623 See Documentation/power/swsusp-and-swap-files.rst 6624 6625 resumedelay= [HIBERNATION] Delay (in seconds) to pause before attempting to 6626 read the resume files 6627 6628 resumewait [HIBERNATION] Wait (indefinitely) for resume device to show up. 6629 Useful for devices that are detected asynchronously 6630 (e.g. USB and MMC devices). 6631 6632 retain_initrd [RAM] Keep initrd memory after extraction. After boot, it will 6633 be accessible via /sys/firmware/initrd. 6634 6635 retbleed= [X86] Control mitigation of RETBleed (Arbitrary 6636 Speculative Code Execution with Return Instructions) 6637 vulnerability. 6638 6639 AMD-based UNRET and IBPB mitigations alone do not stop 6640 sibling threads from influencing the predictions of other 6641 sibling threads. For that reason, STIBP is used on pro- 6642 cessors that support it, and mitigate SMT on processors 6643 that don't. 6644 6645 off - no mitigation 6646 auto - automatically select a mitigation 6647 auto,nosmt - automatically select a mitigation, 6648 disabling SMT if necessary for 6649 the full mitigation (only on Zen1 6650 and older without STIBP). 6651 ibpb - On AMD, mitigate short speculation 6652 windows on basic block boundaries too. 6653 Safe, highest perf impact. It also 6654 enables STIBP if present. Not suitable 6655 on Intel. 6656 ibpb,nosmt - Like "ibpb" above but will disable SMT 6657 when STIBP is not available. This is 6658 the alternative for systems which do not 6659 have STIBP. 6660 unret - Force enable untrained return thunks, 6661 only effective on AMD f15h-f17h based 6662 systems. 6663 unret,nosmt - Like unret, but will disable SMT when STIBP 6664 is not available. This is the alternative for 6665 systems which do not have STIBP. 6666 6667 Selecting 'auto' will choose a mitigation method at run 6668 time according to the CPU. 6669 6670 Not specifying this option is equivalent to retbleed=auto. 6671 6672 rfkill.default_state= 6673 0 "airplane mode". All wifi, bluetooth, wimax, gps, fm, 6674 etc. communication is blocked by default. 6675 1 Unblocked. 6676 6677 rfkill.master_switch_mode= 6678 0 The "airplane mode" button does nothing. 6679 1 The "airplane mode" button toggles between everything 6680 blocked and the previous configuration. 6681 2 The "airplane mode" button toggles between everything 6682 blocked and everything unblocked. 6683 6684 ring3mwait=disable 6685 [KNL] Disable ring 3 MONITOR/MWAIT feature on supported 6686 CPUs. 6687 6688 riscv_isa_fallback [RISCV,EARLY] 6689 When CONFIG_RISCV_ISA_FALLBACK is not enabled, permit 6690 falling back to detecting extension support by parsing 6691 "riscv,isa" property on devicetree systems when the 6692 replacement properties are not found. See the Kconfig 6693 entry for RISCV_ISA_FALLBACK. 6694 6695 riscv_nousercfi= 6696 all Disable user CFI ABI to userspace even if cpu extension 6697 are available. 6698 bcfi Disable user backward CFI ABI to userspace even if 6699 the shadow stack extension is available. 6700 fcfi Disable user forward CFI ABI to userspace even if the 6701 landing pad extension is available. 6702 6703 ro [KNL] Mount root device read-only on boot 6704 6705 rodata= [KNL,EARLY] 6706 on Mark read-only kernel memory as read-only (default). 6707 off Leave read-only kernel memory writable for debugging. 6708 noalias Mark read-only kernel memory as read-only but retain 6709 writable aliases in the direct map for regions outside 6710 of the kernel image. [arm64] 6711 6712 rockchip.usb_uart 6713 [EARLY] 6714 Enable the uart passthrough on the designated usb port 6715 on Rockchip SoCs. When active, the signals of the 6716 debug-uart get routed to the D+ and D- pins of the usb 6717 port and the regular usb controller gets disabled. 6718 6719 root= [KNL] Root filesystem 6720 Usually this is a block device specifier of some kind, 6721 see the early_lookup_bdev comment in 6722 block/early-lookup.c for details. 6723 Alternatively this can be "ram" for the legacy initial 6724 ramdisk, "nfs" and "cifs" for root on a network file 6725 system, or "mtd" and "ubi" for mounting from raw flash. 6726 6727 rootdelay= [KNL] Delay (in seconds) to pause before attempting to 6728 mount the root filesystem 6729 6730 rootflags= [KNL] Set root filesystem mount option string 6731 6732 rseq_slice_ext= [KNL] RSEQ based time slice extension 6733 Format: boolean 6734 Control enablement of RSEQ based time slice extension. 6735 Default is 'on'. 6736 6737 initramfs_options= [KNL] 6738 Specify mount options for the initramfs mount. 6739 6740 rootfstype= [KNL] Set root filesystem type 6741 6742 rootwait [KNL] Wait (indefinitely) for root device to show up. 6743 Useful for devices that are detected asynchronously 6744 (e.g. USB and MMC devices). 6745 6746 rootwait= [KNL] Maximum time (in seconds) to wait for root device 6747 to show up before attempting to mount the root 6748 filesystem. 6749 6750 rproc_mem=nn[KMG][@address] 6751 [KNL,ARM,CMA] Remoteproc physical memory block. 6752 Memory area to be used by remote processor image, 6753 managed by CMA. 6754 6755 rseq_debug= [KNL] Enable or disable restartable sequence 6756 debug mode. Defaults to CONFIG_RSEQ_DEBUG_DEFAULT_ENABLE. 6757 Format: <bool> 6758 6759 rt_group_sched= [KNL] Enable or disable SCHED_RR/FIFO group scheduling 6760 when CONFIG_RT_GROUP_SCHED=y. Defaults to 6761 !CONFIG_RT_GROUP_SCHED_DEFAULT_DISABLED. 6762 Format: <bool> 6763 6764 rw [KNL] Mount root device read-write on boot 6765 6766 S [KNL] Run init in single mode 6767 6768 s390_iommu= [HW,S390] 6769 Set s390 IOTLB flushing mode 6770 strict 6771 With strict flushing every unmap operation will result 6772 in an IOTLB flush. Default is lazy flushing before 6773 reuse, which is faster. Deprecated, equivalent to 6774 iommu.strict=1. 6775 6776 s390_iommu_aperture= [KNL,S390] 6777 Specifies the size of the per device DMA address space 6778 accessible through the DMA and IOMMU APIs as a decimal 6779 factor of the size of main memory. 6780 The default is 1 meaning that one can concurrently use 6781 as many DMA addresses as physical memory is installed, 6782 if supported by hardware, and thus map all of memory 6783 once. With a value of 2 one can map all of memory twice 6784 and so on. As a special case a factor of 0 imposes no 6785 restrictions other than those given by hardware at the 6786 cost of significant additional memory use for tables. 6787 6788 6789 sched_proxy_exec= [KNL] 6790 Enables or disables "proxy execution" style 6791 solution to mutex-based priority inversion. 6792 Format: <bool> 6793 6794 sched_verbose [KNL,EARLY] Enables verbose scheduler debug messages. 6795 6796 schedstats= [KNL,X86] Enable or disable scheduled statistics. 6797 Allowed values are enable and disable. This feature 6798 incurs a small amount of overhead in the scheduler 6799 but is useful for debugging and performance tuning. 6800 6801 sched_thermal_decay_shift= 6802 [Deprecated] 6803 [KNL, SMP] Set a decay shift for scheduler thermal 6804 pressure signal. Thermal pressure signal follows the 6805 default decay period of other scheduler pelt 6806 signals(usually 32 ms but configurable). Setting 6807 sched_thermal_decay_shift will left shift the decay 6808 period for the thermal pressure signal by the shift 6809 value. 6810 i.e. with the default pelt decay period of 32 ms 6811 sched_thermal_decay_shift thermal pressure decay pr 6812 1 64 ms 6813 2 128 ms 6814 and so on. 6815 Format: integer between 0 and 10 6816 Default is 0. 6817 6818 scftorture.holdoff= [KNL] 6819 Number of seconds to hold off before starting 6820 test. Defaults to zero for module insertion and 6821 to 10 seconds for built-in smp_call_function() 6822 tests. 6823 6824 scftorture.longwait= [KNL] 6825 Request ridiculously long waits randomly selected 6826 up to the chosen limit in seconds. Zero (the 6827 default) disables this feature. Please note 6828 that requesting even small non-zero numbers of 6829 seconds can result in RCU CPU stall warnings, 6830 softlockup complaints, and so on. 6831 6832 scftorture.nthreads= [KNL] 6833 Number of kthreads to spawn to invoke the 6834 smp_call_function() family of functions. 6835 The default of -1 specifies a number of kthreads 6836 equal to the number of CPUs. 6837 6838 scftorture.onoff_holdoff= [KNL] 6839 Number seconds to wait after the start of the 6840 test before initiating CPU-hotplug operations. 6841 6842 scftorture.onoff_interval= [KNL] 6843 Number seconds to wait between successive 6844 CPU-hotplug operations. Specifying zero (which 6845 is the default) disables CPU-hotplug operations. 6846 6847 scftorture.shutdown_secs= [KNL] 6848 The number of seconds following the start of the 6849 test after which to shut down the system. The 6850 default of zero avoids shutting down the system. 6851 Non-zero values are useful for automated tests. 6852 6853 scftorture.stat_interval= [KNL] 6854 The number of seconds between outputting the 6855 current test statistics to the console. A value 6856 of zero disables statistics output. 6857 6858 scftorture.stutter_cpus= [KNL] 6859 The number of jiffies to wait between each change 6860 to the set of CPUs under test. 6861 6862 scftorture.use_cpus_read_lock= [KNL] 6863 Use use_cpus_read_lock() instead of the default 6864 preempt_disable() to disable CPU hotplug 6865 while invoking one of the smp_call_function*() 6866 functions. 6867 6868 scftorture.verbose= [KNL] 6869 Enable additional printk() statements. 6870 6871 scftorture.weight_single= [KNL] 6872 The probability weighting to use for the 6873 smp_call_function_single() function with a zero 6874 "wait" parameter. A value of -1 selects the 6875 default if all other weights are -1. However, 6876 if at least one weight has some other value, a 6877 value of -1 will instead select a weight of zero. 6878 6879 scftorture.weight_single_wait= [KNL] 6880 The probability weighting to use for the 6881 smp_call_function_single() function with a 6882 non-zero "wait" parameter. See weight_single. 6883 6884 scftorture.weight_many= [KNL] 6885 The probability weighting to use for the 6886 smp_call_function_many() function with a zero 6887 "wait" parameter. See weight_single. 6888 Note well that setting a high probability for 6889 this weighting can place serious IPI load 6890 on the system. 6891 6892 scftorture.weight_many_wait= [KNL] 6893 The probability weighting to use for the 6894 smp_call_function_many() function with a 6895 non-zero "wait" parameter. See weight_single 6896 and weight_many. 6897 6898 scftorture.weight_all= [KNL] 6899 The probability weighting to use for the 6900 smp_call_function_all() function with a zero 6901 "wait" parameter. See weight_single and 6902 weight_many. 6903 6904 scftorture.weight_all_wait= [KNL] 6905 The probability weighting to use for the 6906 smp_call_function_all() function with a 6907 non-zero "wait" parameter. See weight_single 6908 and weight_many. 6909 6910 sdw_mclk_divider=[SDW] 6911 Specify the MCLK divider for Intel SoundWire buses in 6912 case the BIOS does not provide the clock rate properly. 6913 6914 skew_tick= [KNL,EARLY] Offset the periodic timer tick per cpu to mitigate 6915 xtime_lock contention on larger systems, and/or RCU lock 6916 contention on all systems with CONFIG_MAXSMP set. 6917 Format: { "0" | "1" } 6918 0 -- disable. (may be 1 via CONFIG_CMDLINE="skew_tick=1") 6919 1 -- enable. 6920 Note: increases power consumption, thus should only be 6921 enabled if running jitter sensitive (HPC/RT) workloads. 6922 6923 security= [SECURITY] Choose a legacy "major" security module to 6924 enable at boot. This has been deprecated by the 6925 "lsm=" parameter. 6926 6927 selinux= [SELINUX] Disable or enable SELinux at boot time. 6928 Format: { "0" | "1" } 6929 See security/selinux/Kconfig help text. 6930 0 -- disable. 6931 1 -- enable. 6932 Default value is 1. 6933 6934 serialnumber [BUGS=X86-32] 6935 6936 sev=option[,option...] [X86-64] 6937 6938 debug 6939 Enable debug messages. 6940 6941 nosnp 6942 Do not enable SEV-SNP (applies to host/hypervisor 6943 only). Setting 'nosnp' avoids the RMP check overhead 6944 in memory accesses when users do not want to run 6945 SEV-SNP guests. 6946 6947 shapers= [NET] 6948 Maximal number of shapers. 6949 6950 show_lapic= [APIC,X86] Advanced Programmable Interrupt Controller 6951 Limit apic dumping. The parameter defines the maximal 6952 number of local apics being dumped. Also it is possible 6953 to set it to "all" by meaning -- no limit here. 6954 Format: { 1 (default) | 2 | ... | all }. 6955 The parameter valid if only apic=debug or 6956 apic=verbose is specified. 6957 Example: apic=debug show_lapic=all 6958 6959 slab_debug[=options[,slabs][;[options[,slabs]]...]] [MM] 6960 Enabling slab_debug allows one to determine the 6961 culprit if slab objects become corrupted. Enabling 6962 slab_debug can create guard zones around objects and 6963 may poison objects when not in use. Also tracks the 6964 last alloc / free. For more information see 6965 Documentation/admin-guide/mm/slab.rst. 6966 (slub_debug legacy name also accepted for now) 6967 6968 Using this option implies the "no_hash_pointers" 6969 option which can be undone by adding the 6970 "hash_pointers=always" option. 6971 6972 slab_max_order= [MM] 6973 Determines the maximum allowed order for slabs. 6974 A high setting may cause OOMs due to memory 6975 fragmentation. For more information see 6976 Documentation/admin-guide/mm/slab.rst. 6977 (slub_max_order legacy name also accepted for now) 6978 6979 slab_merge [MM] 6980 Enable merging of slabs with similar size when the 6981 kernel is built without CONFIG_SLAB_MERGE_DEFAULT. 6982 (slub_merge legacy name also accepted for now) 6983 6984 slab_min_objects= [MM] 6985 The minimum number of objects per slab. SLUB will 6986 increase the slab order up to slab_max_order to 6987 generate a sufficiently large slab able to contain 6988 the number of objects indicated. The higher the number 6989 of objects the smaller the overhead of tracking slabs 6990 and the less frequently locks need to be acquired. 6991 For more information see 6992 Documentation/admin-guide/mm/slab.rst. 6993 (slub_min_objects legacy name also accepted for now) 6994 6995 slab_min_order= [MM] 6996 Determines the minimum page order for slabs. Must be 6997 lower or equal to slab_max_order. For more information see 6998 Documentation/admin-guide/mm/slab.rst. 6999 (slub_min_order legacy name also accepted for now) 7000 7001 slab_nomerge [MM] 7002 Disable merging of slabs with similar size. May be 7003 necessary if there is some reason to distinguish 7004 allocs to different slabs, especially in hardened 7005 environments where the risk of heap overflows and 7006 layout control by attackers can usually be 7007 frustrated by disabling merging. This will reduce 7008 most of the exposure of a heap attack to a single 7009 cache (risks via metadata attacks are mostly 7010 unchanged). Debug options disable merging on their 7011 own. 7012 For more information see 7013 Documentation/admin-guide/mm/slab.rst. 7014 (slub_nomerge legacy name also accepted for now) 7015 7016 slab_strict_numa [MM] 7017 Support memory policies on a per object level 7018 in the slab allocator. The default is for memory 7019 policies to be applied at the folio level when 7020 a new folio is needed or a partial folio is 7021 retrieved from the lists. Increases overhead 7022 in the slab fastpaths but gains more accurate 7023 NUMA kernel object placement which helps with slow 7024 interconnects in NUMA systems. 7025 7026 slram= [HW,MTD] 7027 7028 smart2= [HW] 7029 Format: <io1>[,<io2>[,...,<io8>]] 7030 7031 smp.csd_lock_timeout= [KNL] 7032 Specify the period of time in milliseconds 7033 that smp_call_function() and friends will wait 7034 for a CPU to release the CSD lock. This is 7035 useful when diagnosing bugs involving CPUs 7036 disabling interrupts for extended periods 7037 of time. Defaults to 5,000 milliseconds, and 7038 setting a value of zero disables this feature. 7039 This feature may be more efficiently disabled 7040 using the csdlock_debug- kernel parameter. 7041 7042 smp.panic_on_ipistall= [KNL] 7043 If a csd_lock_timeout extends for more than 7044 the specified number of milliseconds, panic the 7045 system. By default, let CSD-lock acquisition 7046 take as long as they take. Specifying 300,000 7047 for this value provides a 5-minute timeout. 7048 7049 smt= [KNL,MIPS,S390,EARLY] Set the maximum number of threads 7050 (logical CPUs) to use per physical CPU on systems 7051 capable of symmetric multithreading (SMT). Will 7052 be capped to the actual hardware limit. 7053 Format: <integer> 7054 Default: -1 (no limit) 7055 7056 softlockup_panic= 7057 [KNL] Should the soft-lockup detector generate panics. 7058 Format: <int> 7059 7060 A value of non-zero instructs the soft-lockup detector 7061 to panic the machine when a soft-lockup duration exceeds 7062 N thresholds. It is also controlled by the kernel.softlockup_panic 7063 sysctl and CONFIG_BOOTPARAM_SOFTLOCKUP_PANIC, which is the 7064 respective build-time switch to that functionality. 7065 7066 softlockup_all_cpu_backtrace= 7067 [KNL] Should the soft-lockup detector generate 7068 backtraces on all cpus. 7069 Format: 0 | 1 7070 7071 sonypi.*= [HW] Sony Programmable I/O Control Device driver 7072 See Documentation/admin-guide/laptops/sonypi.rst 7073 7074 spectre_bhi= [X86] Control mitigation of Branch History Injection 7075 (BHI) vulnerability. This setting affects the 7076 deployment of the HW BHI control and the SW BHB 7077 clearing sequence. 7078 7079 on - (default) Enable the HW or SW mitigation as 7080 needed. This protects the kernel from 7081 both syscalls and VMs. 7082 vmexit - On systems which don't have the HW mitigation 7083 available, enable the SW mitigation on vmexit 7084 ONLY. On such systems, the host kernel is 7085 protected from VM-originated BHI attacks, but 7086 may still be vulnerable to syscall attacks. 7087 off - Disable the mitigation. 7088 7089 spectre_v2= [X86,EARLY] Control mitigation of Spectre variant 2 7090 (indirect branch speculation) vulnerability. 7091 The default operation protects the kernel from 7092 user space attacks. 7093 7094 on - unconditionally enable, implies 7095 spectre_v2_user=on 7096 off - unconditionally disable, implies 7097 spectre_v2_user=off 7098 auto - kernel detects whether your CPU model is 7099 vulnerable 7100 7101 Selecting 'on' will, and 'auto' may, choose a 7102 mitigation method at run time according to the 7103 CPU, the available microcode, the setting of the 7104 CONFIG_MITIGATION_RETPOLINE configuration option, 7105 and the compiler with which the kernel was built. 7106 7107 Selecting 'on' will also enable the mitigation 7108 against user space to user space task attacks. 7109 Selecting specific mitigation does not force enable 7110 user mitigations. 7111 7112 Selecting 'off' will disable both the kernel and 7113 the user space protections. 7114 7115 Specific mitigations can also be selected manually: 7116 7117 retpoline - replace indirect branches 7118 retpoline,generic - Retpolines 7119 retpoline,lfence - LFENCE; indirect branch 7120 retpoline,amd - alias for retpoline,lfence 7121 eibrs - Enhanced/Auto IBRS 7122 eibrs,retpoline - Enhanced/Auto IBRS + Retpolines 7123 eibrs,lfence - Enhanced/Auto IBRS + LFENCE 7124 ibrs - use IBRS to protect kernel 7125 7126 Not specifying this option is equivalent to 7127 spectre_v2=auto. 7128 7129 spectre_v2_user= 7130 [X86] Control mitigation of Spectre variant 2 7131 (indirect branch speculation) vulnerability between 7132 user space tasks 7133 7134 on - Unconditionally enable mitigations. Is 7135 enforced by spectre_v2=on 7136 7137 off - Unconditionally disable mitigations. Is 7138 enforced by spectre_v2=off 7139 7140 prctl - Indirect branch speculation is enabled, 7141 but mitigation can be enabled via prctl 7142 per thread. The mitigation control state 7143 is inherited on fork. 7144 7145 prctl,ibpb 7146 - Like "prctl" above, but only STIBP is 7147 controlled per thread. IBPB is issued 7148 always when switching between different user 7149 space processes. 7150 7151 seccomp 7152 - Same as "prctl" above, but all seccomp 7153 threads will enable the mitigation unless 7154 they explicitly opt out. 7155 7156 seccomp,ibpb 7157 - Like "seccomp" above, but only STIBP is 7158 controlled per thread. IBPB is issued 7159 always when switching between different 7160 user space processes. 7161 7162 auto - Kernel selects the mitigation depending on 7163 the available CPU features and vulnerability. 7164 7165 Default mitigation: "prctl" 7166 7167 Not specifying this option is equivalent to 7168 spectre_v2_user=auto. 7169 7170 spec_rstack_overflow= 7171 [X86,EARLY] Control RAS overflow mitigation on AMD Zen CPUs 7172 7173 off - Disable mitigation 7174 microcode - Enable microcode mitigation only 7175 safe-ret - Enable sw-only safe RET mitigation (default) 7176 ibpb - Enable mitigation by issuing IBPB on 7177 kernel entry 7178 ibpb-vmexit - Issue IBPB only on VMEXIT 7179 (cloud-specific mitigation) 7180 7181 spec_store_bypass_disable= 7182 [HW,EARLY] Control Speculative Store Bypass (SSB) Disable mitigation 7183 (Speculative Store Bypass vulnerability) 7184 7185 Certain CPUs are vulnerable to an exploit against a 7186 a common industry wide performance optimization known 7187 as "Speculative Store Bypass" in which recent stores 7188 to the same memory location may not be observed by 7189 later loads during speculative execution. The idea 7190 is that such stores are unlikely and that they can 7191 be detected prior to instruction retirement at the 7192 end of a particular speculation execution window. 7193 7194 In vulnerable processors, the speculatively forwarded 7195 store can be used in a cache side channel attack, for 7196 example to read memory to which the attacker does not 7197 directly have access (e.g. inside sandboxed code). 7198 7199 This parameter controls whether the Speculative Store 7200 Bypass optimization is used. 7201 7202 On x86 the options are: 7203 7204 on - Unconditionally disable Speculative Store Bypass 7205 off - Unconditionally enable Speculative Store Bypass 7206 auto - Kernel detects whether the CPU model contains an 7207 implementation of Speculative Store Bypass and 7208 picks the most appropriate mitigation. If the 7209 CPU is not vulnerable, "off" is selected. If the 7210 CPU is vulnerable the default mitigation is 7211 architecture and Kconfig dependent. See below. 7212 prctl - Control Speculative Store Bypass per thread 7213 via prctl. Speculative Store Bypass is enabled 7214 for a process by default. The state of the control 7215 is inherited on fork. 7216 seccomp - Same as "prctl" above, but all seccomp threads 7217 will disable SSB unless they explicitly opt out. 7218 7219 Default mitigations: 7220 X86: "prctl" 7221 7222 On powerpc the options are: 7223 7224 on,auto - On Power8 and Power9 insert a store-forwarding 7225 barrier on kernel entry and exit. On Power7 7226 perform a software flush on kernel entry and 7227 exit. 7228 off - No action. 7229 7230 Not specifying this option is equivalent to 7231 spec_store_bypass_disable=auto. 7232 7233 split_llc= 7234 [X86,EARLY] Split the LLC N-ways 7235 7236 When set, the LLC is split this many ways by matching 7237 'core_id % n'. This is setup before SMP bringup and 7238 used during SMP bringup before it knows the full 7239 topology. If your core count doesn't nicely divide by 7240 the number given, you get to keep the pieces. 7241 7242 This is mostly a debug feature to emulate multiple LLCs 7243 on hardware that only have a single LLC. 7244 7245 split_lock_detect= 7246 [X86] Enable split lock detection or bus lock detection 7247 7248 When enabled (and if hardware support is present), atomic 7249 instructions that access data across cache line 7250 boundaries will result in an alignment check exception 7251 for split lock detection or a debug exception for 7252 bus lock detection. 7253 7254 off - not enabled 7255 7256 warn - the kernel will emit rate-limited warnings 7257 about applications triggering the #AC 7258 exception or the #DB exception. This mode is 7259 the default on CPUs that support split lock 7260 detection or bus lock detection. Default 7261 behavior is by #AC if both features are 7262 enabled in hardware. 7263 7264 fatal - the kernel will send SIGBUS to applications 7265 that trigger the #AC exception or the #DB 7266 exception. Default behavior is by #AC if 7267 both features are enabled in hardware. 7268 7269 ratelimit:N - 7270 Set system wide rate limit to N bus locks 7271 per second for bus lock detection. 7272 0 < N <= 1000. 7273 7274 N/A for split lock detection. 7275 7276 7277 If an #AC exception is hit in the kernel or in 7278 firmware (i.e. not while executing in user mode) 7279 the kernel will oops in either "warn" or "fatal" 7280 mode. 7281 7282 #DB exception for bus lock is triggered only when 7283 CPL > 0. 7284 7285 srbds= [X86,INTEL,EARLY] 7286 Control the Special Register Buffer Data Sampling 7287 (SRBDS) mitigation. 7288 7289 Certain CPUs are vulnerable to an MDS-like 7290 exploit which can leak bits from the random 7291 number generator. 7292 7293 By default, this issue is mitigated by 7294 microcode. However, the microcode fix can cause 7295 the RDRAND and RDSEED instructions to become 7296 much slower. Among other effects, this will 7297 result in reduced throughput from /dev/urandom. 7298 7299 The microcode mitigation can be disabled with 7300 the following option: 7301 7302 off: Disable mitigation and remove 7303 performance impact to RDRAND and RDSEED 7304 7305 srcutree.big_cpu_lim [KNL] 7306 Specifies the number of CPUs constituting a 7307 large system, such that srcu_struct structures 7308 should immediately allocate an srcu_node array. 7309 This kernel-boot parameter defaults to 128, 7310 but takes effect only when the low-order four 7311 bits of srcutree.convert_to_big is equal to 3 7312 (decide at boot). 7313 7314 srcutree.convert_to_big [KNL] 7315 Specifies under what conditions an SRCU tree 7316 srcu_struct structure will be converted to big 7317 form, that is, with an rcu_node tree: 7318 7319 0: Never. 7320 1: At init_srcu_struct() time. 7321 2: When rcutorture decides to. 7322 3: Decide at boot time (default). 7323 0x1X: Above plus if high contention. 7324 7325 Either way, the srcu_node tree will be sized based 7326 on the actual runtime number of CPUs (nr_cpu_ids) 7327 instead of the compile-time CONFIG_NR_CPUS. 7328 7329 srcutree.counter_wrap_check [KNL] 7330 Specifies how frequently to check for 7331 grace-period sequence counter wrap for the 7332 srcu_data structure's ->srcu_gp_seq_needed field. 7333 The greater the number of bits set in this kernel 7334 parameter, the less frequently counter wrap will 7335 be checked for. Note that the bottom two bits 7336 are ignored. 7337 7338 srcutree.exp_holdoff [KNL] 7339 Specifies how many nanoseconds must elapse 7340 since the end of the last SRCU grace period for 7341 a given srcu_struct until the next normal SRCU 7342 grace period will be considered for automatic 7343 expediting. Set to zero to disable automatic 7344 expediting. 7345 7346 srcutree.srcu_max_nodelay [KNL] 7347 Specifies the number of no-delay instances 7348 per jiffy for which the SRCU grace period 7349 worker thread will be rescheduled with zero 7350 delay. Beyond this limit, worker thread will 7351 be rescheduled with a sleep delay of one jiffy. 7352 7353 srcutree.srcu_max_nodelay_phase [KNL] 7354 Specifies the per-grace-period phase, number of 7355 non-sleeping polls of readers. Beyond this limit, 7356 grace period worker thread will be rescheduled 7357 with a sleep delay of one jiffy, between each 7358 rescan of the readers, for a grace period phase. 7359 7360 srcutree.srcu_retry_check_delay [KNL] 7361 Specifies number of microseconds of non-sleeping 7362 delay between each non-sleeping poll of readers. 7363 7364 srcutree.small_contention_lim [KNL] 7365 Specifies the number of update-side contention 7366 events per jiffy will be tolerated before 7367 initiating a conversion of an srcu_struct 7368 structure to big form. Note that the value of 7369 srcutree.convert_to_big must have the 0x10 bit 7370 set for contention-based conversions to occur. 7371 7372 ssbd= [ARM64,HW,EARLY] 7373 Speculative Store Bypass Disable control 7374 7375 On CPUs that are vulnerable to the Speculative 7376 Store Bypass vulnerability and offer a 7377 firmware based mitigation, this parameter 7378 indicates how the mitigation should be used: 7379 7380 force-on: Unconditionally enable mitigation for 7381 for both kernel and userspace 7382 force-off: Unconditionally disable mitigation for 7383 for both kernel and userspace 7384 kernel: Always enable mitigation in the 7385 kernel, and offer a prctl interface 7386 to allow userspace to register its 7387 interest in being mitigated too. 7388 7389 stack_guard_gap= [MM] 7390 override the default stack gap protection. The value 7391 is in page units and it defines how many pages prior 7392 to (for stacks growing down) resp. after (for stacks 7393 growing up) the main stack are reserved for no other 7394 mapping. Default value is 256 pages. 7395 7396 stack_depot_disable= [KNL,EARLY] 7397 Setting this to true through kernel command line will 7398 disable the stack depot thereby saving the static memory 7399 consumed by the stack hash table. By default this is set 7400 to false. 7401 7402 stack_depot_max_pools= [KNL,EARLY] 7403 Specify the maximum number of pools to use for storing 7404 stack traces. Pools are allocated on-demand up to this 7405 limit. Default value is 8191 pools. 7406 7407 stacktrace [FTRACE] 7408 Enable the stack tracer on boot up. 7409 7410 stacktrace_filter=[function-list] 7411 [FTRACE] Limit the functions that the stack tracer 7412 will trace at boot up. function-list is a comma-separated 7413 list of functions. This list can be changed at run 7414 time by the stack_trace_filter file in the debugfs 7415 tracing directory. Note, this enables stack tracing 7416 and the stacktrace above is not needed. 7417 7418 sti= [PARISC,HW] 7419 Format: <num> 7420 Set the STI (builtin display/keyboard on the HP-PARISC 7421 machines) console (graphic card) which should be used 7422 as the initial boot-console. 7423 See also comment in drivers/video/sticore.c. 7424 7425 sti_font= [HW] 7426 See comment in drivers/video/sticore.c. 7427 7428 stifb= [HW] 7429 Format: bpp:<bpp1>[:<bpp2>[:<bpp3>...]] 7430 7431 strict_sas_size= 7432 [X86] 7433 Format: <bool> 7434 Enable or disable strict sigaltstack size checks 7435 against the required signal frame size which 7436 depends on the supported FPU features. This can 7437 be used to filter out binaries which have 7438 not yet been made aware of AT_MINSIGSTKSZ. 7439 7440 stress_hpt [PPC,EARLY] 7441 Limits the number of kernel HPT entries in the hash 7442 page table to increase the rate of hash page table 7443 faults on kernel addresses. 7444 7445 stress_slb [PPC,EARLY] 7446 Limits the number of kernel SLB entries, and flushes 7447 them frequently to increase the rate of SLB faults 7448 on kernel addresses. 7449 7450 no_slb_preload [PPC,EARLY] 7451 Disables slb preloading for userspace. 7452 7453 sunrpc.min_resvport= 7454 sunrpc.max_resvport= 7455 [NFS,SUNRPC] 7456 SunRPC servers often require that client requests 7457 originate from a privileged port (i.e. a port in the 7458 range 0 < portnr < 1024). 7459 An administrator who wishes to reserve some of these 7460 ports for other uses may adjust the range that the 7461 kernel's sunrpc client considers to be privileged 7462 using these two parameters to set the minimum and 7463 maximum port values. 7464 7465 sunrpc.svc_rpc_per_connection_limit= 7466 [NFS,SUNRPC] 7467 Limit the number of requests that the server will 7468 process in parallel from a single connection. 7469 The default value is 0 (no limit). 7470 7471 sunrpc.pool_mode= 7472 [NFS] 7473 Deprecated. The NFS server now always uses one 7474 service thread pool per NUMA node (equivalent to a 7475 single global pool on non-NUMA machines). All of 7476 the previously accepted values (auto, global, 7477 percpu, pernode) are still accepted for backward 7478 compatibility but are ignored: the mode is always 7479 pernode, and reads always return "pernode". 7480 7481 sunrpc.tcp_slot_table_entries= 7482 sunrpc.udp_slot_table_entries= 7483 [NFS,SUNRPC] 7484 Sets the upper limit on the number of simultaneous 7485 RPC calls that can be sent from the client to a 7486 server. Increasing these values may allow you to 7487 improve throughput, but will also increase the 7488 amount of memory reserved for use by the client. 7489 7490 suspend.pm_test_delay= 7491 [SUSPEND] 7492 Sets the number of seconds to remain in a suspend test 7493 mode before resuming the system (see 7494 /sys/power/pm_test). Only available when CONFIG_PM_DEBUG 7495 is set. Default value is 5. 7496 7497 svm= [PPC] 7498 Format: { on | off | y | n | 1 | 0 } 7499 This parameter controls use of the Protected 7500 Execution Facility on pSeries. 7501 7502 swiotlb= [ARM,PPC,MIPS,X86,S390,EARLY] 7503 Format: { <int> [,<int>] | force | noforce } 7504 <int> -- Number of I/O TLB slabs 7505 <int> -- Second integer after comma. Number of swiotlb 7506 areas with their own lock. Will be rounded up 7507 to a power of 2. 7508 force -- force using of bounce buffers even if they 7509 wouldn't be automatically used by the kernel 7510 noforce -- Never use bounce buffers (for debugging) 7511 7512 switches= [HW,M68k,EARLY] 7513 7514 sysctl.*= [KNL] 7515 Set a sysctl parameter, right before loading the init 7516 process, as if the value was written to the respective 7517 /proc/sys/... file. Both '.' and '/' are recognized as 7518 separators. Unrecognized parameters and invalid values 7519 are reported in the kernel log. Sysctls registered 7520 later by a loaded module cannot be set this way. 7521 Example: sysctl.vm.swappiness=40 7522 7523 sysrq_always_enabled 7524 [KNL] 7525 Ignore sysrq setting - this boot parameter will 7526 neutralize any effect of /proc/sys/kernel/sysrq. 7527 Useful for debugging. 7528 7529 tcpmhash_entries= [KNL,NET] 7530 Set the number of tcp_metrics_hash slots. 7531 Default value is 8192 or 16384 depending on total 7532 ram pages. This is used to specify the TCP metrics 7533 cache size. See Documentation/networking/ip-sysctl.rst 7534 "tcp_no_metrics_save" section for more details. 7535 7536 tdfx= [HW,DRM] 7537 7538 test_suspend= [SUSPEND] 7539 Format: { "mem" | "standby" | "freeze" }[,N] 7540 Specify "mem" (for Suspend-to-RAM) or "standby" (for 7541 standby suspend) or "freeze" (for suspend type freeze) 7542 as the system sleep state during system startup with 7543 the optional capability to repeat N number of times. 7544 The system is woken from this state using a 7545 wakeup-capable RTC alarm. 7546 7547 thash_entries= [KNL,NET] 7548 Set number of hash buckets for TCP connection 7549 7550 thermal.act= [HW,ACPI] 7551 -1: disable all active trip points in all thermal zones 7552 <degrees C>: override all lowest active trip points 7553 7554 thermal.crt= [HW,ACPI] 7555 -1: disable all critical trip points in all thermal zones 7556 <degrees C>: override all critical trip points 7557 7558 thermal.off= [HW,ACPI] 7559 1: disable ACPI thermal control 7560 7561 thermal.psv= [HW,ACPI] 7562 -1: disable all passive trip points 7563 <degrees C>: override all passive trip points to this 7564 value 7565 7566 thermal.tzp= [HW,ACPI] 7567 Specify global default ACPI thermal zone polling rate 7568 <deci-seconds>: poll all this frequency 7569 0: no polling (default) 7570 7571 thp_anon= [KNL] 7572 Format: <size>[KMG],<size>[KMG]:<state>;<size>[KMG]-<size>[KMG]:<state> 7573 state is one of "always", "madvise", "never" or "inherit". 7574 Control the default behavior of the system with respect 7575 to anonymous transparent hugepages. 7576 Can be used multiple times for multiple anon THP sizes. 7577 See Documentation/admin-guide/mm/transhuge.rst for more 7578 details. 7579 7580 threadirqs [KNL,EARLY] 7581 Force threading of all interrupt handlers except those 7582 marked explicitly IRQF_NO_THREAD. 7583 7584 thp_shmem= [KNL] 7585 Format: <size>[KMG],<size>[KMG]:<policy>;<size>[KMG]-<size>[KMG]:<policy> 7586 Control the default policy of each hugepage size for the 7587 internal shmem mount. <policy> is one of policies available 7588 for the shmem mount ("always", "inherit", "never", "within_size", 7589 and "advise"). 7590 It can be used multiple times for multiple shmem THP sizes. 7591 See Documentation/admin-guide/mm/transhuge.rst for more 7592 details. 7593 7594 tlbi= [X86-64] 7595 Format: {ipi} 7596 ipi: switch to IPI-based TLB flushing 7597 7598 topology= [S390,EARLY] 7599 Format: {off | on} 7600 Specify if the kernel should make use of the cpu 7601 topology information if the hardware supports this. 7602 The scheduler will make use of this information and 7603 e.g. base its process migration decisions on it. 7604 Default is on. 7605 7606 torture.disable_onoff_at_boot= [KNL] 7607 Prevent the CPU-hotplug component of torturing 7608 until after init has spawned. 7609 7610 torture.ftrace_dump_at_shutdown= [KNL] 7611 Dump the ftrace buffer at torture-test shutdown, 7612 even if there were no errors. This can be a 7613 very costly operation when many torture tests 7614 are running concurrently, especially on systems 7615 with rotating-rust storage. 7616 7617 torture.verbose_sleep_frequency= [KNL] 7618 Specifies how many verbose printk()s should be 7619 emitted between each sleep. The default of zero 7620 disables verbose-printk() sleeping. 7621 7622 torture.verbose_sleep_duration= [KNL] 7623 Duration of each verbose-printk() sleep in jiffies. 7624 7625 tpm.disable_pcr_integrity= [HW,TPM] 7626 Do not protect PCR registers from unintended physical 7627 access, or interposers in the bus by the means of 7628 having an integrity protected session wrapped around 7629 TPM2_PCR_Extend command. Consider this in a situation 7630 where TPM is heavily utilized by IMA, thus protection 7631 causing a major performance hit, and the space where 7632 machines are deployed is by other means guarded. 7633 7634 tpm_crb_ffa.busy_timeout_ms= [ARM64,TPM] 7635 Maximum time in milliseconds to retry sending a message 7636 to the TPM service before giving up. This parameter controls 7637 how long the system will continue retrying when the TPM 7638 service is busy. 7639 Format: <unsigned int> 7640 Default: 2000 (2 seconds) 7641 7642 tpm_suspend_pcr=[HW,TPM] 7643 Format: integer pcr id 7644 Specify that at suspend time, the tpm driver 7645 should extend the specified pcr with zeros, 7646 as a workaround for some chips which fail to 7647 flush the last written pcr on TPM_SaveState. 7648 This will guarantee that all the other pcrs 7649 are saved. 7650 7651 tpm_tis.interrupts= [HW,TPM] 7652 Enable interrupts for the MMIO based physical layer 7653 for the FIFO interface. By default it is set to false 7654 (0). For more information about TPM hardware interfaces 7655 defined by Trusted Computing Group (TCG) see 7656 https://trustedcomputinggroup.org/resource/pc-client-platform-tpm-profile-ptp-specification/ 7657 7658 tp_printk [FTRACE] 7659 Have the tracepoints sent to printk as well as the 7660 tracing ring buffer. This is useful for early boot up 7661 where the system hangs or reboots and does not give the 7662 option for reading the tracing buffer or performing a 7663 ftrace_dump_on_oops. 7664 7665 To turn off having tracepoints sent to printk, 7666 echo 0 > /proc/sys/kernel/tracepoint_printk 7667 Note, echoing 1 into this file without the 7668 tp_printk kernel cmdline option has no effect. 7669 7670 The tp_printk_stop_on_boot (see below) can also be used 7671 to stop the printing of events to console at 7672 late_initcall_sync. 7673 7674 ** CAUTION ** 7675 7676 Having tracepoints sent to printk() and activating high 7677 frequency tracepoints such as irq or sched, can cause 7678 the system to live lock. 7679 7680 tp_printk_stop_on_boot [FTRACE] 7681 When tp_printk (above) is set, it can cause a lot of noise 7682 on the console. It may be useful to only include the 7683 printing of events during boot up, as user space may 7684 make the system inoperable. 7685 7686 This command line option will stop the printing of events 7687 to console at the late_initcall_sync() time frame. 7688 7689 trace_buf_size=nn[KMG] 7690 [FTRACE] will set tracing buffer size on each cpu. 7691 7692 trace_clock= [FTRACE] Set the clock used for tracing events 7693 at boot up. 7694 local - Use the per CPU time stamp counter 7695 (converted into nanoseconds). Fast, but 7696 depending on the architecture, may not be 7697 in sync between CPUs. 7698 global - Event time stamps are synchronized across 7699 CPUs. May be slower than the local clock, 7700 but better for some race conditions. 7701 counter - Simple counting of events (1, 2, ..) 7702 note, some counts may be skipped due to the 7703 infrastructure grabbing the clock more than 7704 once per event. 7705 uptime - Use jiffies as the time stamp. 7706 perf - Use the same clock that perf uses. 7707 mono - Use ktime_get_mono_fast_ns() for time stamps. 7708 mono_raw - Use ktime_get_raw_fast_ns() for time 7709 stamps. 7710 boot - Use ktime_get_boot_fast_ns() for time stamps. 7711 Architectures may add more clocks. See 7712 Documentation/trace/ftrace.rst for more details. 7713 7714 trace_event=[event-list] 7715 [FTRACE] Set and start specified trace events in order 7716 to facilitate early boot debugging. The event-list is a 7717 comma-separated list of trace events to enable. See 7718 also Documentation/trace/events.rst 7719 7720 To enable modules, use :mod: keyword: 7721 7722 trace_event=:mod:<module> 7723 7724 The value before :mod: will only enable specific events 7725 that are part of the module. See the above mentioned 7726 document for more information. 7727 7728 trace_instance=[instance-info] 7729 [FTRACE] Create a ring buffer instance early in boot up. 7730 This will be listed in: 7731 7732 /sys/kernel/tracing/instances 7733 7734 Events can be enabled at the time the instance is created 7735 via: 7736 7737 trace_instance=<name>,<system1>:<event1>,<system2>:<event2> 7738 7739 Note, the "<system*>:" portion is optional if the event is 7740 unique. 7741 7742 trace_instance=foo,sched:sched_switch,irq_handler_entry,initcall 7743 7744 will enable the "sched_switch" event (note, the "sched:" is optional, and 7745 the same thing would happen if it was left off). The irq_handler_entry 7746 event, and all events under the "initcall" system. 7747 7748 Flags can be added to the instance to modify its behavior when it is 7749 created. The flags are separated by '^'. 7750 7751 The available flags are: 7752 7753 traceoff - Have the tracing instance tracing disabled after it is created. 7754 traceprintk - Have trace_printk() write into this trace instance 7755 (note, "printk" and "trace_printk" can also be used) 7756 7757 trace_instance=foo^traceoff^traceprintk,sched,irq 7758 7759 The flags must come before the defined events. 7760 7761 If memory has been reserved (see memmap for x86), the instance 7762 can use that memory: 7763 7764 memmap=12M$0x284500000 trace_instance=boot_map@0x284500000:12M 7765 7766 The above will create a "boot_map" instance that uses the physical 7767 memory at 0x284500000 that is 12Megs. The per CPU buffers of that 7768 instance will be split up accordingly. 7769 7770 Alternatively, the memory can be reserved by the reserve_mem option: 7771 7772 reserve_mem=12M:4096:trace trace_instance=boot_map@trace 7773 7774 This will reserve 12 megabytes at boot up with a 4096 byte alignment 7775 and place the ring buffer in this memory. Note that due to KASLR, the 7776 memory may not be the same location each time, which will not preserve 7777 the buffer content. 7778 7779 Also note that the layout of the ring buffer data may change between 7780 kernel versions where the validator will fail and reset the ring buffer 7781 if the layout is not the same as the previous kernel. 7782 7783 If the ring buffer is used for persistent bootups and has events enabled, 7784 it is recommend to disable tracing so that events from a previous boot do not 7785 mix with events of the current boot (unless you are debugging a random crash 7786 at boot up). 7787 7788 reserve_mem=12M:4096:trace trace_instance=boot_map^traceoff^traceprintk@trace,sched,irq 7789 7790 Note, saving the trace buffer across reboots does require that the system 7791 is set up to not wipe memory. For instance, CONFIG_RESET_ATTACK_MITIGATION 7792 can force a memory reset on boot which will clear any trace that was stored. 7793 This is just one of many ways that can clear memory. Make sure your system 7794 keeps the content of memory across reboots before relying on this option. 7795 7796 NB: Both the mapped address and size must be page aligned for the architecture. 7797 7798 See also Documentation/trace/debugging.rst 7799 7800 7801 trace_options=[option-list] 7802 [FTRACE] Enable or disable tracer options at boot. 7803 The option-list is a comma delimited list of options 7804 that can be enabled or disabled just as if you were 7805 to echo the option name into 7806 7807 /sys/kernel/tracing/trace_options 7808 7809 For example, to enable stacktrace option (to dump the 7810 stack trace of each event), add to the command line: 7811 7812 trace_options=stacktrace 7813 7814 See also Documentation/trace/ftrace.rst "trace options" 7815 section. 7816 7817 trace_trigger=[trigger-list] 7818 [FTRACE] Add an event trigger on specific events. 7819 Set a trigger on top of a specific event, with an optional 7820 filter. 7821 7822 The format is "trace_trigger=<event>.<trigger>[ if <filter>],..." 7823 Where more than one trigger may be specified that are comma delimited. 7824 7825 For example: 7826 7827 trace_trigger="sched_switch.stacktrace if prev_state == 2" 7828 7829 The above will enable the "stacktrace" trigger on the "sched_switch" 7830 event but only trigger it if the "prev_state" of the "sched_switch" 7831 event is "2" (TASK_UNINTERRUPTIBLE). 7832 7833 See also "Event triggers" in Documentation/trace/events.rst 7834 7835 7836 traceoff_after_boot 7837 [FTRACE] Sometimes tracing is used to debug issues 7838 during the boot process. Since the trace buffer has a 7839 limited amount of storage, it may be prudent to 7840 disable tracing after the boot is finished, otherwise 7841 the critical information may be overwritten. With this 7842 option, the main tracing buffer will be turned off at 7843 the end of the boot process. 7844 7845 traceoff_on_warning 7846 [FTRACE] enable this option to disable tracing when a 7847 warning is hit. This turns off "tracing_on". Tracing can 7848 be enabled again by echoing '1' into the "tracing_on" 7849 file located in /sys/kernel/tracing/ 7850 7851 This option is useful, as it disables the trace before 7852 the WARNING dump is called, which prevents the trace to 7853 be filled with content caused by the warning output. 7854 7855 This option can also be set at run time via the sysctl 7856 option: kernel/traceoff_on_warning 7857 7858 transparent_hugepage= 7859 [KNL] 7860 Format: [always|madvise|never] 7861 Can be used to control the default behavior of the system 7862 with respect to transparent hugepages. 7863 See Documentation/admin-guide/mm/transhuge.rst 7864 for more details. 7865 7866 transparent_hugepage_shmem= [KNL] 7867 Format: [always|within_size|advise|never|deny|force] 7868 Can be used to control the hugepage allocation policy for 7869 the internal shmem mount. 7870 See Documentation/admin-guide/mm/transhuge.rst 7871 for more details. 7872 7873 transparent_hugepage_tmpfs= [KNL] 7874 Format: [always|within_size|advise|never] 7875 Can be used to control the default hugepage allocation policy 7876 for the tmpfs mount. 7877 See Documentation/admin-guide/mm/transhuge.rst 7878 for more details. 7879 7880 trusted.source= [KEYS] 7881 Format: <string> 7882 This parameter identifies the trust source as a backend 7883 for trusted keys implementation. Supported trust 7884 sources: 7885 - "tpm" 7886 - "tee" 7887 - "caam" 7888 - "dcp" 7889 - "pkwm" 7890 If not specified then it defaults to iterating through 7891 the trust source list starting with TPM and assigns the 7892 first trust source as a backend which is initialized 7893 successfully during iteration. 7894 7895 trusted.debug= [KEYS] 7896 Format: <bool> 7897 Enable trusted keys debug traces at runtime when 7898 CONFIG_TRUSTED_KEYS_DEBUG=y. 7899 7900 To make the traces visible after enabling the option, 7901 use trusted.dyndbg='+p' as needed. By convention, 7902 the subsystem uses pr_debug() for these traces. 7903 7904 SAFETY: The traces can leak sensitive data, so be 7905 cautious before enabling this. They remain inactive 7906 unless this parameter is set this option to a true 7907 value. 7908 7909 Default: false 7910 7911 trusted.rng= [KEYS] 7912 Format: <string> 7913 The RNG used to generate key material for trusted keys. 7914 Can be one of: 7915 - "kernel" 7916 - the same value as trusted.source: "tpm" or "tee" 7917 - "default" 7918 If not specified, "default" is used. In this case, 7919 the RNG's choice is left to each individual trust source. 7920 7921 trusted.dcp_use_otp_key 7922 This is intended to be used in combination with 7923 trusted.source=dcp and will select the DCP OTP key 7924 instead of the DCP UNIQUE key blob encryption. 7925 7926 trusted.dcp_skip_zk_test 7927 This is intended to be used in combination with 7928 trusted.source=dcp and will disable the check if the 7929 blob key is all zeros. This is helpful for situations where 7930 having this key zero'ed is acceptable. E.g. in testing 7931 scenarios. 7932 7933 tsa= [X86] Control mitigation for Transient Scheduler 7934 Attacks on AMD CPUs. Search the following in your 7935 favourite search engine for more details: 7936 7937 "Technical guidance for mitigating transient scheduler 7938 attacks". 7939 7940 off - disable the mitigation 7941 on - enable the mitigation (default) 7942 user - mitigate only user/kernel transitions 7943 vm - mitigate only guest/host transitions 7944 7945 7946 tsc= Disable clocksource stability checks for TSC. 7947 Format: <string> 7948 [x86] reliable: mark tsc clocksource as reliable, this 7949 disables clocksource verification at runtime, as well 7950 as the stability checks done at bootup. Used to enable 7951 high-resolution timer mode on older hardware, and in 7952 virtualized environment. 7953 [x86] noirqtime: Do not use TSC to do irq accounting. 7954 Used to run time disable IRQ_TIME_ACCOUNTING on any 7955 platforms where RDTSC is slow and this accounting 7956 can add overhead. 7957 [x86] unstable: mark the TSC clocksource as unstable, this 7958 marks the TSC unconditionally unstable at bootup and 7959 avoids any further wobbles once the TSC watchdog notices. 7960 [x86] nowatchdog: disable clocksource watchdog. Used 7961 in situations with strict latency requirements (where 7962 interruptions from clocksource watchdog are not 7963 acceptable). 7964 [x86] recalibrate: force recalibration against a HW timer 7965 (HPET or PM timer) on systems whose TSC frequency was 7966 obtained from HW or FW using either an MSR or CPUID(0x15). 7967 Warn if the difference is more than 500 ppm. 7968 [x86] watchdog: Enforce the clocksource watchdog on TSC 7969 7970 tsc_early_khz= [X86,EARLY] Skip early TSC calibration and use the given 7971 value instead. Useful when the early TSC frequency discovery 7972 procedure is not reliable, such as on overclocked systems 7973 with CPUID.16h support and partial CPUID.15h support. 7974 Format: <unsigned int> 7975 7976 tsx= [X86] Control Transactional Synchronization 7977 Extensions (TSX) feature in Intel processors that 7978 support TSX control. 7979 7980 This parameter controls the TSX feature. The options are: 7981 7982 on - Enable TSX on the system. Although there are 7983 mitigations for all known security vulnerabilities, 7984 TSX has been known to be an accelerator for 7985 several previous speculation-related CVEs, and 7986 so there may be unknown security risks associated 7987 with leaving it enabled. 7988 7989 off - Disable TSX on the system. (Note that this 7990 option takes effect only on newer CPUs which are 7991 not vulnerable to MDS, i.e., have 7992 MSR_IA32_ARCH_CAPABILITIES.MDS_NO=1 and which get 7993 the new IA32_TSX_CTRL MSR through a microcode 7994 update. This new MSR allows for the reliable 7995 deactivation of the TSX functionality.) 7996 7997 auto - Disable TSX if X86_BUG_TAA is present, 7998 otherwise enable TSX on the system. 7999 8000 Not specifying this option is equivalent to tsx=off. 8001 8002 See Documentation/admin-guide/hw-vuln/tsx_async_abort.rst 8003 for more details. 8004 8005 tsx_async_abort= [X86,INTEL,EARLY] Control mitigation for the TSX Async 8006 Abort (TAA) vulnerability. 8007 8008 Similar to Micro-architectural Data Sampling (MDS) 8009 certain CPUs that support Transactional 8010 Synchronization Extensions (TSX) are vulnerable to an 8011 exploit against CPU internal buffers which can forward 8012 information to a disclosure gadget under certain 8013 conditions. 8014 8015 In vulnerable processors, the speculatively forwarded 8016 data can be used in a cache side channel attack, to 8017 access data to which the attacker does not have direct 8018 access. 8019 8020 This parameter controls the TAA mitigation. The 8021 options are: 8022 8023 full - Enable TAA mitigation on vulnerable CPUs 8024 if TSX is enabled. 8025 8026 full,nosmt - Enable TAA mitigation and disable SMT on 8027 vulnerable CPUs. If TSX is disabled, SMT 8028 is not disabled because CPU is not 8029 vulnerable to cross-thread TAA attacks. 8030 off - Unconditionally disable TAA mitigation 8031 8032 On MDS-affected machines, tsx_async_abort=off can be 8033 prevented by an active MDS mitigation as both vulnerabilities 8034 are mitigated with the same mechanism so in order to disable 8035 this mitigation, you need to specify mds=off too. 8036 8037 Not specifying this option is equivalent to 8038 tsx_async_abort=full. On CPUs which are MDS affected 8039 and deploy MDS mitigation, TAA mitigation is not 8040 required and doesn't provide any additional 8041 mitigation. 8042 8043 For details see: 8044 Documentation/admin-guide/hw-vuln/tsx_async_abort.rst 8045 8046 turbografx.map[2|3]= [HW,JOY] 8047 TurboGraFX parallel port interface 8048 Format: 8049 <port#>,<js1>,<js2>,<js3>,<js4>,<js5>,<js6>,<js7> 8050 See also Documentation/input/devices/joystick-parport.rst 8051 8052 udbg-immortal [PPC] When debugging early kernel crashes that 8053 happen after console_init() and before a proper 8054 console driver takes over, this boot options might 8055 help "seeing" what's going on. 8056 8057 uhash_entries= [KNL,NET] 8058 Set number of hash buckets for UDP/UDP-Lite connections 8059 8060 uhci-hcd.ignore_oc= 8061 [USB] Ignore overcurrent events (default N). 8062 Some badly-designed motherboards generate lots of 8063 bogus events, for ports that aren't wired to 8064 anything. Set this parameter to avoid log spamming. 8065 Note that genuine overcurrent events won't be 8066 reported either. 8067 8068 unaligned_scalar_speed= 8069 [RISCV] 8070 Format: {slow | fast | unsupported} 8071 Allow skipping scalar unaligned access speed tests. This 8072 is useful for testing alternative code paths and to skip 8073 the tests in environments where they run too slowly. All 8074 CPUs must have the same scalar unaligned access speed. 8075 8076 unaligned_vector_speed= 8077 [RISCV] 8078 Format: {slow | fast | unsupported} 8079 Allow skipping vector unaligned access speed tests. This 8080 is useful for testing alternative code paths and to skip 8081 the tests in environments where they run too slowly. All 8082 CPUs must have the same vector unaligned access speed. 8083 8084 unknown_nmi_panic 8085 [X86] Cause panic on unknown NMI. 8086 8087 unwind_debug [X86-64,EARLY] 8088 Enable unwinder debug output. This can be 8089 useful for debugging certain unwinder error 8090 conditions, including corrupt stacks and 8091 bad/missing unwinder metadata. 8092 8093 usbcore.authorized_default= 8094 [USB] Default USB device authorization: 8095 (default -1 = authorized (same as 1), 8096 0 = not authorized, 1 = authorized, 2 = authorized 8097 if device connected to internal port) 8098 8099 usbcore.autosuspend= 8100 [USB] The autosuspend time delay (in seconds) used 8101 for newly-detected USB devices (default 2). This 8102 is the time required before an idle device will be 8103 autosuspended. Devices for which the delay is set 8104 to a negative value won't be autosuspended at all. 8105 8106 usbcore.usbfs_snoop= 8107 [USB] Set to log all usbfs traffic (default 0 = off). 8108 8109 usbcore.usbfs_snoop_max= 8110 [USB] Maximum number of bytes to snoop in each URB 8111 (default = 65536). 8112 8113 usbcore.blinkenlights= 8114 [USB] Set to cycle leds on hubs (default 0 = off). 8115 8116 usbcore.old_scheme_first= 8117 [USB] Start with the old device initialization 8118 scheme (default 0 = off). 8119 8120 usbcore.usbfs_memory_mb= 8121 [USB] Memory limit (in MB) for buffers allocated by 8122 usbfs (default = 16, 0 = max = 2047). 8123 8124 usbcore.use_both_schemes= 8125 [USB] Try the other device initialization scheme 8126 if the first one fails (default 1 = enabled). 8127 8128 usbcore.initial_descriptor_timeout= 8129 [USB] Specifies timeout for the initial 64-byte 8130 USB_REQ_GET_DESCRIPTOR request in milliseconds 8131 (default 5000 = 5.0 seconds). 8132 8133 usbcore.nousb [USB] Disable the USB subsystem 8134 8135 usbcore.quirks= 8136 [USB] A list of quirk entries to augment the built-in 8137 usb core quirk list. List entries are separated by 8138 commas. Each entry has the form 8139 VendorID:ProductID:Flags. The IDs are 4-digit hex 8140 numbers and Flags is a set of letters. Each letter 8141 will change the built-in quirk; setting it if it is 8142 clear and clearing it if it is set. The letters have 8143 the following meanings: 8144 a = USB_QUIRK_STRING_FETCH_255 (string 8145 descriptors must not be fetched using 8146 a 255-byte read); 8147 b = USB_QUIRK_RESET_RESUME (device can't resume 8148 correctly so reset it instead); 8149 c = USB_QUIRK_NO_SET_INTF (device can't handle 8150 Set-Interface requests); 8151 d = USB_QUIRK_CONFIG_INTF_STRINGS (device can't 8152 handle its Configuration or Interface 8153 strings); 8154 e = USB_QUIRK_RESET (device can't be reset 8155 (e.g morph devices), don't use reset); 8156 f = USB_QUIRK_HONOR_BNUMINTERFACES (device has 8157 more interface descriptions than the 8158 bNumInterfaces count, and can't handle 8159 talking to these interfaces); 8160 g = USB_QUIRK_DELAY_INIT (device needs a pause 8161 during initialization, after we read 8162 the device descriptor); 8163 h = USB_QUIRK_LINEAR_UFRAME_INTR_BINTERVAL (For 8164 high speed and super speed interrupt 8165 endpoints, the USB 2.0 and USB 3.0 spec 8166 require the interval in microframes (1 8167 microframe = 125 microseconds) to be 8168 calculated as interval = 2 ^ 8169 (bInterval-1). 8170 Devices with this quirk report their 8171 bInterval as the result of this 8172 calculation instead of the exponent 8173 variable used in the calculation); 8174 i = USB_QUIRK_DEVICE_QUALIFIER (device can't 8175 handle device_qualifier descriptor 8176 requests); 8177 j = USB_QUIRK_IGNORE_REMOTE_WAKEUP (device 8178 generates spurious wakeup, ignore 8179 remote wakeup capability); 8180 k = USB_QUIRK_NO_LPM (device can't handle Link 8181 Power Management); 8182 l = USB_QUIRK_LINEAR_FRAME_INTR_BINTERVAL 8183 (Device reports its bInterval as linear 8184 frames instead of the USB 2.0 8185 calculation); 8186 m = USB_QUIRK_DISCONNECT_SUSPEND (Device needs 8187 to be disconnected before suspend to 8188 prevent spurious wakeup); 8189 n = USB_QUIRK_DELAY_CTRL_MSG (Device needs a 8190 pause after every control message); 8191 o = USB_QUIRK_HUB_SLOW_RESET (Hub needs extra 8192 delay after resetting its port); 8193 p = USB_QUIRK_SHORT_SET_ADDRESS_REQ_TIMEOUT 8194 (Reduce timeout of the SET_ADDRESS 8195 request from 5000 ms to 500 ms); 8196 q = USB_QUIRK_FORCE_ONE_CONFIG (Device 8197 claims zero configurations, 8198 forcing to 1); 8199 r = USB_QUIRK_WINDOWS_CONFIG_REQ_SIZE (Device 8200 fails during initialization when asked for 8201 9-bytes configuration descriptor request. 8202 Ask for 255-bytes request instead to mirror 8203 Windows' behavior); 8204 Example: quirks=0781:5580:bk,0a5c:5834:gij 8205 8206 usbhid.mousepoll= 8207 [USBHID] The interval which mice are to be polled at. 8208 8209 usbhid.jspoll= 8210 [USBHID] The interval which joysticks are to be polled at. 8211 8212 usbhid.kbpoll= 8213 [USBHID] The interval which keyboards are to be polled at. 8214 8215 usb-storage.delay_use= 8216 [UMS] The delay in seconds before a new device is 8217 scanned for Logical Units (default 1). 8218 Optionally the delay in milliseconds if the value has 8219 suffix with "ms". 8220 Example: delay_use=2567ms 8221 8222 usb-storage.quirks= 8223 [UMS] A list of quirks entries to supplement or 8224 override the built-in unusual_devs list. List 8225 entries are separated by commas. Each entry has 8226 the form VID:PID:Flags where VID and PID are Vendor 8227 and Product ID values (4-digit hex numbers) and 8228 Flags is a set of characters, each corresponding 8229 to a common usb-storage quirk flag as follows: 8230 a = SANE_SENSE (collect more than 18 bytes 8231 of sense data, not on uas); 8232 b = BAD_SENSE (don't collect more than 18 8233 bytes of sense data, not on uas); 8234 c = FIX_CAPACITY (decrease the reported 8235 device capacity by one sector); 8236 d = NO_READ_DISC_INFO (don't use 8237 READ_DISC_INFO command, not on uas); 8238 e = NO_READ_CAPACITY_16 (don't use 8239 READ_CAPACITY_16 command); 8240 f = NO_REPORT_OPCODES (don't use report opcodes 8241 command, uas only); 8242 g = MAX_SECTORS_240 (don't transfer more than 8243 240 sectors at a time, uas only); 8244 h = CAPACITY_HEURISTICS (decrease the 8245 reported device capacity by one 8246 sector if the number is odd); 8247 i = IGNORE_DEVICE (don't bind to this 8248 device); 8249 j = NO_REPORT_LUNS (don't use report luns 8250 command, uas only); 8251 k = NO_SAME (do not use WRITE_SAME, uas only) 8252 l = NOT_LOCKABLE (don't try to lock and 8253 unlock ejectable media, not on uas); 8254 m = MAX_SECTORS_64 (don't transfer more 8255 than 64 sectors = 32 KB at a time, 8256 not on uas); 8257 n = INITIAL_READ10 (force a retry of the 8258 initial READ(10) command, not on uas); 8259 o = CAPACITY_OK (accept the capacity 8260 reported by the device, not on uas); 8261 p = WRITE_CACHE (the device cache is ON 8262 by default, not on uas); 8263 r = IGNORE_RESIDUE (the device reports 8264 bogus residue values, not on uas); 8265 s = SINGLE_LUN (the device has only one 8266 Logical Unit); 8267 t = NO_ATA_1X (don't allow ATA(12) and ATA(16) 8268 commands, uas only); 8269 u = IGNORE_UAS (don't bind to the uas driver); 8270 w = NO_WP_DETECT (don't test whether the 8271 medium is write-protected). 8272 y = ALWAYS_SYNC (issue a SYNCHRONIZE_CACHE 8273 even if the device claims no cache, 8274 not on uas) 8275 Example: quirks=0419:aaf5:rl,0421:0433:rc 8276 8277 user_debug= [KNL,ARM] 8278 Format: <int> 8279 See arch/arm/Kconfig.debug help text. 8280 1 - undefined instruction events 8281 2 - system calls 8282 4 - invalid data aborts 8283 8 - SIGSEGV faults 8284 16 - SIGBUS faults 8285 Example: user_debug=31 8286 8287 vdso= [X86,SH,SPARC] 8288 On X86_32, this is an alias for vdso32=. Otherwise: 8289 8290 vdso=1: enable VDSO (the default) 8291 vdso=0: disable VDSO mapping 8292 8293 vdso32= [X86] Control the 32-bit vDSO 8294 vdso32=1: enable 32-bit VDSO 8295 vdso32=0 or vdso32=2: disable 32-bit VDSO 8296 8297 See the help text for CONFIG_COMPAT_VDSO for more 8298 details. If CONFIG_COMPAT_VDSO is set, the default is 8299 vdso32=0; otherwise, the default is vdso32=1. 8300 8301 For compatibility with older kernels, vdso32=2 is an 8302 alias for vdso32=0. 8303 8304 Try vdso32=0 if you encounter an error that says: 8305 dl_main: Assertion `(void *) ph->p_vaddr == _rtld_local._dl_sysinfo_dso' failed! 8306 8307 video= [FB,EARLY] Frame buffer configuration 8308 See Documentation/fb/modedb.rst. 8309 8310 video.brightness_switch_enabled= [ACPI] 8311 Format: [0|1] 8312 If set to 1, on receiving an ACPI notify event 8313 generated by hotkey, video driver will adjust brightness 8314 level and then send out the event to user space through 8315 the allocated input device. If set to 0, video driver 8316 will only send out the event without touching backlight 8317 brightness level. 8318 default: 1 8319 8320 virtio_mmio.device= 8321 [VMMIO] Memory mapped virtio (platform) device. 8322 8323 <size>@<baseaddr>:<irq>[:<id>] 8324 where: 8325 <size> := size (can use standard suffixes 8326 like K, M and G) 8327 <baseaddr> := physical base address 8328 <irq> := interrupt number (as passed to 8329 request_irq()) 8330 <id> := (optional) platform device id 8331 example: 8332 virtio_mmio.device=1K@0x100b0000:48:7 8333 8334 Can be used multiple times for multiple devices. 8335 8336 vga= [BOOT,X86-32] Select a particular video mode 8337 See Documentation/arch/x86/boot.rst and 8338 Documentation/admin-guide/svga.rst. 8339 Use vga=ask for menu. 8340 This is actually a boot loader parameter; the value is 8341 passed to the kernel using a special protocol. 8342 8343 vm_debug[=options] [KNL] Available with CONFIG_DEBUG_VM=y. 8344 May slow down system boot speed, especially when 8345 enabled on systems with a large amount of memory. 8346 All options are enabled by default, and this 8347 interface is meant to allow for selectively 8348 enabling or disabling specific virtual memory 8349 debugging features. 8350 8351 Available options are: 8352 P Enable page structure init time poisoning 8353 - Disable all of the above options 8354 8355 vmalloc=nn[KMG] [KNL,BOOT,EARLY] Forces the vmalloc area to have an 8356 exact size of <nn>. This can be used to increase 8357 the minimum size (128MB on x86, arm32 platforms). 8358 It can also be used to decrease the size and leave more room 8359 for directly mapped kernel RAM. Note that this parameter does 8360 not exist on many other platforms (including arm64, alpha, 8361 loongarch, arc, csky, hexagon, microblaze, mips, nios2, openrisc, 8362 parisc, m64k, powerpc, riscv, sh, um, xtensa, s390, sparc). 8363 8364 vmcp_cma=nn[MG] [KNL,S390,EARLY] 8365 Sets the memory size reserved for contiguous memory 8366 allocations for the vmcp device driver. 8367 8368 vmhalt= [KNL,S390] Perform z/VM CP command after system halt. 8369 Format: <command> 8370 8371 vmpanic= [KNL,S390] Perform z/VM CP command after kernel panic. 8372 Format: <command> 8373 8374 vmpoff= [KNL,S390] Perform z/VM CP command after power off. 8375 Format: <command> 8376 8377 vmscape= [X86] Controls mitigation for VMscape attacks. 8378 VMscape attacks can leak information from a userspace 8379 hypervisor to a guest via speculative side-channels. 8380 8381 off - disable the mitigation 8382 ibpb - use Indirect Branch Prediction Barrier 8383 (IBPB) mitigation (default) 8384 force - force vulnerability detection even on 8385 unaffected processors 8386 8387 vsyscall= [X86-64,EARLY] 8388 Controls the behavior of vsyscalls (i.e. calls to 8389 fixed addresses of 0xffffffffff600x00 from legacy 8390 code). Most statically-linked binaries and older 8391 versions of glibc use these calls. Because these 8392 functions are at fixed addresses, they make nice 8393 targets for exploits that can control RIP. 8394 8395 emulate Vsyscalls turn into traps and are emulated 8396 reasonably safely. The vsyscall page is 8397 readable. This disables the Linear 8398 Address Space Separation (LASS) security 8399 feature and makes the system less secure. 8400 8401 xonly [default] Vsyscalls turn into traps and are 8402 emulated reasonably safely. The vsyscall 8403 page is not readable. 8404 8405 none Vsyscalls don't work at all. This makes 8406 them quite hard to use for exploits but 8407 might break your system. 8408 8409 vt.color= [VT] Default text color. 8410 Format: 0xYX, X = foreground, Y = background. 8411 Default: 0x07 = light gray on black. 8412 8413 vt.cur_default= [VT] Default cursor shape. 8414 Format: 0xCCBBAA, where AA, BB, and CC are the same as 8415 the parameters of the <Esc>[?A;B;Cc escape sequence; 8416 see vga-softcursor.rst. Default: 2 = underline. 8417 8418 vt.default_blu= [VT] 8419 Format: <blue0>,<blue1>,<blue2>,...,<blue15> 8420 Change the default blue palette of the console. 8421 This is a 16-member array composed of values 8422 ranging from 0-255. 8423 8424 vt.default_grn= [VT] 8425 Format: <green0>,<green1>,<green2>,...,<green15> 8426 Change the default green palette of the console. 8427 This is a 16-member array composed of values 8428 ranging from 0-255. 8429 8430 vt.default_red= [VT] 8431 Format: <red0>,<red1>,<red2>,...,<red15> 8432 Change the default red palette of the console. 8433 This is a 16-member array composed of values 8434 ranging from 0-255. 8435 8436 vt.default_utf8= 8437 [VT] 8438 Format=<0|1> 8439 Set system-wide default UTF-8 mode for all tty's. 8440 Default is 1, i.e. UTF-8 mode is enabled for all 8441 newly opened terminals. 8442 8443 vt.global_cursor_default= 8444 [VT] 8445 Format=<-1|0|1> 8446 Set system-wide default for whether a cursor 8447 is shown on new VTs. Default is -1, 8448 i.e. cursors will be created by default unless 8449 overridden by individual drivers. 0 will hide 8450 cursors, 1 will display them. 8451 8452 vt.italic= [VT] Default color for italic text; 0-15. 8453 Default: 2 = green. 8454 8455 vt.underline= [VT] Default color for underlined text; 0-15. 8456 Default: 3 = cyan. 8457 8458 watchdog timers [HW,WDT] For information on watchdog timers, 8459 see Documentation/watchdog/watchdog-parameters.rst 8460 or other driver-specific files in the 8461 Documentation/watchdog/ directory. 8462 8463 watchdog_thresh= 8464 [KNL] 8465 Set the hard lockup detector stall duration 8466 threshold in seconds. The soft lockup detector 8467 threshold is set to twice the value. A value of 0 8468 disables both lockup detectors. Default is 10 8469 seconds. 8470 8471 workqueue.unbound_cpus= 8472 [KNL,SMP] Specify to constrain one or some CPUs 8473 to use in unbound workqueues. 8474 Format: <cpu-list> 8475 By default, all online CPUs are available for 8476 unbound workqueues. 8477 8478 workqueue.watchdog_thresh= 8479 If CONFIG_WQ_WATCHDOG is configured, workqueue can 8480 warn stall conditions and dump internal state to 8481 help debugging. 0 disables workqueue stall 8482 detection; otherwise, it's the stall threshold 8483 duration in seconds. The default value is 30 and 8484 it can be updated at runtime by writing to the 8485 corresponding sysfs file. 8486 8487 workqueue.panic_on_stall=<uint> 8488 Panic when workqueue stall is detected by 8489 CONFIG_WQ_WATCHDOG. It sets the number times of the 8490 stall to trigger panic. 8491 8492 The default is set by CONFIG_BOOTPARAM_WQ_STALL_PANIC, 8493 which is 0 (disabled) if not configured. 8494 8495 workqueue.panic_on_stall_time=<uint> 8496 Panic when a workqueue stall has been continuous for 8497 the specified number of seconds. Unlike panic_on_stall 8498 which counts accumulated stall events, this triggers 8499 based on the duration of a single continuous stall. 8500 8501 The default is 0, which disables the time-based panic. 8502 8503 workqueue.cpu_intensive_thresh_us= 8504 Per-cpu work items which run for longer than this 8505 threshold are automatically considered CPU intensive 8506 and excluded from concurrency management to prevent 8507 them from noticeably delaying other per-cpu work 8508 items. Default is 10000 (10ms). 8509 8510 If CONFIG_WQ_CPU_INTENSIVE_REPORT is set, the kernel 8511 will report the work functions which violate this 8512 threshold repeatedly. They are likely good 8513 candidates for using WQ_UNBOUND workqueues instead. 8514 8515 workqueue.cpu_intensive_warning_thresh=<uint> 8516 If CONFIG_WQ_CPU_INTENSIVE_REPORT is set, the kernel 8517 will report the work functions which violate the 8518 intensive_threshold_us repeatedly. In order to prevent 8519 spurious warnings, start printing only after a work 8520 function has violated this threshold number of times. 8521 8522 The default is 4 times. 0 disables the warning. 8523 8524 workqueue.power_efficient 8525 Per-cpu workqueues are generally preferred because 8526 they show better performance thanks to cache 8527 locality; unfortunately, per-cpu workqueues tend to 8528 be more power hungry than unbound workqueues. 8529 8530 Enabling this makes the per-cpu workqueues which 8531 were observed to contribute significantly to power 8532 consumption unbound, leading to measurably lower 8533 power usage at the cost of small performance 8534 overhead. 8535 8536 The default value of this parameter is determined by 8537 the config option CONFIG_WQ_POWER_EFFICIENT_DEFAULT. 8538 8539 workqueue.default_affinity_scope= 8540 Select the default affinity scope to use for unbound 8541 workqueues. Can be one of "cpu", "smt", "cache", 8542 "cache_shard", "numa" and "system". Default is 8543 "cache_shard". For more 8544 information, see the Affinity Scopes section in 8545 Documentation/core-api/workqueue.rst. 8546 8547 This can be changed after boot by writing to the 8548 matching /sys/module/workqueue/parameters file. All 8549 workqueues with the "default" affinity scope will be 8550 updated accordingly. 8551 8552 workqueue.debug_force_rr_cpu 8553 Workqueue used to implicitly guarantee that work 8554 items queued without explicit CPU specified are put 8555 on the local CPU. This guarantee is no longer true 8556 and while local CPU is still preferred work items 8557 may be put on foreign CPUs. This debug option 8558 forces round-robin CPU selection to flush out 8559 usages which depend on the now broken guarantee. 8560 When enabled, memory and cache locality will be 8561 impacted. 8562 8563 writecombine= [LOONGARCH,EARLY] Control the MAT (Memory Access 8564 Type) of ioremap_wc(). 8565 8566 on - Enable writecombine, use WUC for ioremap_wc() 8567 off - Disable writecombine, use SUC for ioremap_wc() 8568 8569 x2apic_phys [X86-64,APIC,EARLY] Use x2apic physical mode instead of 8570 default x2apic cluster mode on platforms 8571 supporting x2apic. 8572 8573 xen_512gb_limit [KNL,X86-64,XEN] 8574 Restricts the kernel running paravirtualized under Xen 8575 to use only up to 512 GB of RAM. The reason to do so is 8576 crash analysis tools and Xen tools for doing domain 8577 save/restore/migration must be enabled to handle larger 8578 domains. 8579 8580 xen_console_io [XEN,EARLY] 8581 Boolean option to enable/disable the usage of the Xen 8582 console_io hypercalls to read and write to the console. 8583 Mostly useful for debugging and development. 8584 8585 xen_emul_unplug= [HW,X86,XEN,EARLY] 8586 Unplug Xen emulated devices 8587 Format: [unplug0,][unplug1] 8588 ide-disks -- unplug primary master IDE devices 8589 aux-ide-disks -- unplug non-primary-master IDE devices 8590 nics -- unplug network devices 8591 all -- unplug all emulated devices (NICs and IDE disks) 8592 unnecessary -- unplugging emulated devices is 8593 unnecessary even if the host did not respond to 8594 the unplug protocol 8595 never -- do not unplug even if version check succeeds 8596 8597 xen_legacy_crash [X86,XEN,EARLY] 8598 Crash from Xen panic notifier, without executing late 8599 panic() code such as dumping handler. 8600 8601 xen_mc_debug [X86,XEN,EARLY] 8602 Enable multicall debugging when running as a Xen PV guest. 8603 Enabling this feature will reduce performance a little 8604 bit, so it should only be enabled for obtaining extended 8605 debug data in case of multicall errors. 8606 8607 xen_msr_safe= [X86,XEN,EARLY] 8608 Format: <bool> 8609 Select whether to always use non-faulting (safe) MSR 8610 access functions when running as Xen PV guest. The 8611 default value is controlled by CONFIG_XEN_PV_MSR_SAFE. 8612 8613 xen_nopv [X86] 8614 Disables the PV optimizations forcing the HVM guest to 8615 run as generic HVM guest with no PV drivers. 8616 This option is obsoleted by the "nopv" option, which 8617 has equivalent effect for XEN platform. 8618 8619 xen_no_vector_callback 8620 [KNL,X86,XEN,EARLY] Disable the vector callback for Xen 8621 event channel interrupts. 8622 8623 xen_scrub_pages= [XEN] 8624 Boolean option to control scrubbing pages before giving them back 8625 to Xen, for use by other domains. Can be also changed at runtime 8626 with /sys/devices/system/xen_memory/xen_memory0/scrub_pages. 8627 Default value controlled with CONFIG_XEN_SCRUB_PAGES_DEFAULT. 8628 8629 xen_timer_slop= [X86-64,XEN,EARLY] 8630 Set the timer slop (in nanoseconds) for the virtual Xen 8631 timers (default is 100000). This adjusts the minimum 8632 delta of virtualized Xen timers, where lower values 8633 improve timer resolution at the expense of processing 8634 more timer interrupts. 8635 8636 xen.balloon_boot_timeout= [XEN] 8637 The time (in seconds) to wait before giving up to boot 8638 in case initial ballooning fails to free enough memory. 8639 Applies only when running as HVM or PVH guest and 8640 started with less memory configured than allowed at 8641 max. Default is 180. 8642 8643 xen.event_eoi_delay= [XEN] 8644 How long to delay EOI handling in case of event 8645 storms (jiffies). Default is 10. 8646 8647 xen.event_loop_timeout= [XEN] 8648 After which time (jiffies) the event handling loop 8649 should start to delay EOI handling. Default is 2. 8650 8651 xen.fifo_events= [XEN] 8652 Boolean parameter to disable using fifo event handling 8653 even if available. Normally fifo event handling is 8654 preferred over the 2-level event handling, as it is 8655 fairer and the number of possible event channels is 8656 much higher. Default is on (use fifo events). 8657 8658 xirc2ps_cs= [NET,PCMCIA] 8659 Format: 8660 <irq>,<irq_mask>,<io>,<full_duplex>,<do_sound>,<lockup_hack>[,<irq2>[,<irq3>[,<irq4>]]] 8661 8662 xive= [PPC] 8663 By default on POWER9 and above, the kernel will 8664 natively use the XIVE interrupt controller. This option 8665 allows the fallback firmware mode to be used: 8666 8667 off Fallback to firmware control of XIVE interrupt 8668 controller on both pseries and powernv 8669 platforms. Only useful on POWER9 and above. 8670 8671 xive.store-eoi=off [PPC] 8672 By default on POWER10 and above, the kernel will use 8673 stores for EOI handling when the XIVE interrupt mode 8674 is active. This option allows the XIVE driver to use 8675 loads instead, as on POWER9. 8676 8677 xhci-hcd.quirks [USB,KNL] 8678 A hex value specifying bitmask with supplemental xhci 8679 host controller quirks. Meaning of each bit can be 8680 consulted in header drivers/usb/host/xhci.h. 8681 8682 xmon [PPC,EARLY] 8683 Format: { early | on | rw | ro | off } 8684 Controls if xmon debugger is enabled. Default is off. 8685 Passing only "xmon" is equivalent to "xmon=early". 8686 early Call xmon as early as possible on boot; xmon 8687 debugger is called from setup_arch(). 8688 on xmon debugger hooks will be installed so xmon 8689 is only called on a kernel crash. Default mode, 8690 i.e. either "ro" or "rw" mode, is controlled 8691 with CONFIG_XMON_DEFAULT_RO_MODE. 8692 rw xmon debugger hooks will be installed so xmon 8693 is called only on a kernel crash, mode is write, 8694 meaning SPR registers, memory and, other data 8695 can be written using xmon commands. 8696 ro same as "rw" option above but SPR registers, 8697 memory, and other data can't be written using 8698 xmon commands. 8699 off xmon is disabled. 8700