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 ISDN Appropriate ISDN support is enabled. 28 ISOL CPU Isolation is enabled. 29 JOY Appropriate joystick support is enabled. 30 KGDB Kernel debugger support is enabled. 31 KVM Kernel Virtual Machine support is enabled. 32 LIBATA Libata driver is enabled 33 LOONGARCH LoongArch architecture is enabled. 34 LOOP Loopback device support is enabled. 35 LP Printer support is enabled. 36 M68k M68k architecture is enabled. 37 These options have more detailed description inside of 38 Documentation/arch/m68k/kernel-options.rst. 39 MDA MDA console support is enabled. 40 MIPS MIPS architecture is enabled. 41 MOUSE Appropriate mouse support is enabled. 42 MSI Message Signaled Interrupts (PCI). 43 MTD MTD (Memory Technology Device) support is enabled. 44 NET Appropriate network support is enabled. 45 NFS Appropriate NFS support is enabled. 46 NUMA NUMA support is enabled. 47 OF Devicetree is enabled. 48 PARISC The PA-RISC architecture is enabled. 49 PCI PCI bus support is enabled. 50 PCIE PCI Express support is enabled. 51 PCMCIA The PCMCIA subsystem is enabled. 52 PNP Plug & Play support is enabled. 53 PPC PowerPC architecture is enabled. 54 PPT Parallel port support is enabled. 55 PS2 Appropriate PS/2 support is enabled. 56 PV_OPS A paravirtualized kernel is enabled. 57 RAM RAM disk support is enabled. 58 RDT Intel Resource Director Technology. 59 RISCV RISCV architecture is enabled. 60 S390 S390 architecture is enabled. 61 SCSI Appropriate SCSI support is enabled. 62 A lot of drivers have their options described inside 63 the Documentation/scsi/ sub-directory. 64 SDW SoundWire support is enabled. 65 SECURITY Different security models are enabled. 66 SELINUX SELinux support is enabled. 67 SERIAL Serial support is enabled. 68 SH SuperH architecture is enabled. 69 SMP The kernel is an SMP kernel. 70 SPARC Sparc architecture is enabled. 71 SUSPEND System suspend states are enabled. 72 SWSUSP Software suspend (hibernation) is enabled. 73 TPM TPM drivers are enabled. 74 UMS USB Mass Storage support is enabled. 75 USB USB support is enabled. 76 NVME NVMe support is enabled 77 USBHID USB Human Interface Device support is enabled. 78 V4L Video For Linux support is enabled. 79 VGA The VGA console has been enabled. 80 VMMIO Driver for memory mapped virtio devices is enabled. 81 VT Virtual terminal support is enabled. 82 WDT Watchdog support is enabled. 83 X86-32 X86-32, aka i386 architecture is enabled. 84 X86-64 X86-64 architecture is enabled. 85 X86 Either 32-bit or 64-bit x86 (same as X86-32+X86-64) 86 X86_UV SGI UV support is enabled. 87 XEN Xen support is enabled 88 XTENSA xtensa architecture is enabled. 89 90In addition, the following text indicates that the option 91 92 BOOT Is a boot loader parameter. 93 BUGS= Relates to possible processor bugs on the said processor. 94 KNL Is a kernel start-up parameter. 95 96 97Kernel parameters 98 99 accept_memory= [MM] 100 Format: { eager | lazy } 101 default: lazy 102 By default, unaccepted memory is accepted lazily to 103 avoid prolonged boot times. The lazy option will add 104 some runtime overhead until all memory is eventually 105 accepted. In most cases the overhead is negligible. 106 For some workloads or for debugging purposes 107 accept_memory=eager can be used to accept all memory 108 at once during boot. 109 110 acpi= [HW,ACPI,X86,ARM64,RISCV64,EARLY] 111 Advanced Configuration and Power Interface 112 Format: { force | on | off | strict | noirq | rsdt | 113 copy_dsdt | nospcr } 114 force -- enable ACPI if default was off 115 on -- enable ACPI but allow fallback to DT [arm64,riscv64] 116 off -- disable ACPI if default was on 117 noirq -- do not use ACPI for IRQ routing 118 strict -- Be less tolerant of platforms that are not 119 strictly ACPI specification compliant. 120 rsdt -- prefer RSDT over (default) XSDT 121 copy_dsdt -- copy DSDT to memory 122 nocmcff -- Disable firmware first mode for corrected 123 errors. This disables parsing the HEST CMC error 124 source to check if firmware has set the FF flag. This 125 may result in duplicate corrected error reports. 126 nospcr -- disable console in ACPI SPCR table as 127 default _serial_ console on ARM64 128 spcr -- enable console in ACPI SPCR table as 129 default _serial_ console on x86 130 For ARM64, ONLY "acpi=off", "acpi=on", "acpi=force" or 131 "acpi=nospcr" are available 132 For RISCV64, ONLY "acpi=off", "acpi=on" or "acpi=force" 133 are available 134 135 See also Documentation/power/runtime_pm.rst, pci=noacpi 136 137 acpi_apic_instance= [ACPI,IOAPIC,EARLY] 138 Format: <int> 139 2: use 2nd APIC table, if available 140 1,0: use 1st APIC table 141 default: 0 142 143 acpi_backlight= [HW,ACPI] 144 { vendor | video | native | none } 145 If set to vendor, prefer vendor-specific driver 146 (e.g. thinkpad_acpi, sony_acpi, etc.) instead 147 of the ACPI video.ko driver. 148 If set to video, use the ACPI video.ko driver. 149 If set to native, use the device's native backlight mode. 150 If set to none, disable the ACPI backlight interface. 151 152 acpi_force_32bit_fadt_addr [ACPI,EARLY] 153 force FADT to use 32 bit addresses rather than the 154 64 bit X_* addresses. Some firmware have broken 64 155 bit addresses for force ACPI ignore these and use 156 the older legacy 32 bit addresses. 157 158 acpica_no_return_repair [HW, ACPI] 159 Disable AML predefined validation mechanism 160 This mechanism can repair the evaluation result to make 161 the return objects more ACPI specification compliant. 162 This option is useful for developers to identify the 163 root cause of an AML interpreter issue when the issue 164 has something to do with the repair mechanism. 165 166 acpi.debug_layer= [HW,ACPI,ACPI_DEBUG] 167 acpi.debug_level= [HW,ACPI,ACPI_DEBUG] 168 Format: <int> 169 CONFIG_ACPI_DEBUG must be enabled to produce any ACPI 170 debug output. Bits in debug_layer correspond to a 171 _COMPONENT in an ACPI source file, e.g., 172 #define _COMPONENT ACPI_EVENTS 173 Bits in debug_level correspond to a level in 174 ACPI_DEBUG_PRINT statements, e.g., 175 ACPI_DEBUG_PRINT((ACPI_DB_INFO, ... 176 The debug_level mask defaults to "info". See 177 Documentation/firmware-guide/acpi/debug.rst for more information about 178 debug layers and levels. 179 180 Enable processor driver info messages: 181 acpi.debug_layer=0x20000000 182 Enable AML "Debug" output, i.e., stores to the Debug 183 object while interpreting AML: 184 acpi.debug_layer=0xffffffff acpi.debug_level=0x2 185 Enable all messages related to ACPI hardware: 186 acpi.debug_layer=0x2 acpi.debug_level=0xffffffff 187 188 Some values produce so much output that the system is 189 unusable. The "log_buf_len" parameter may be useful 190 if you need to capture more output. 191 192 acpi.poweroff_on_fatal= [ACPI] 193 {0 | 1} 194 Causes the system to poweroff when the ACPI bytecode signals 195 a fatal error. The default value of this setting is 1. 196 Overriding this value should only be done for diagnosing 197 ACPI firmware problems, as the system might behave erratically 198 after having encountered a fatal ACPI error. 199 200 acpi_enforce_resources= [ACPI] 201 { strict | lax | no } 202 Check for resource conflicts between native drivers 203 and ACPI OperationRegions (SystemIO and SystemMemory 204 only). IO ports and memory declared in ACPI might be 205 used by the ACPI subsystem in arbitrary AML code and 206 can interfere with legacy drivers. 207 strict (default): access to resources claimed by ACPI 208 is denied; legacy drivers trying to access reserved 209 resources will fail to bind to device using them. 210 lax: access to resources claimed by ACPI is allowed; 211 legacy drivers trying to access reserved resources 212 will bind successfully but a warning message is logged. 213 no: ACPI OperationRegions are not marked as reserved, 214 no further checks are performed. 215 216 acpi_force_table_verification [HW,ACPI,EARLY] 217 Enable table checksum verification during early stage. 218 By default, this is disabled due to x86 early mapping 219 size limitation. 220 221 acpi_irq_balance [HW,ACPI] 222 ACPI will balance active IRQs 223 default in APIC mode 224 225 acpi_irq_nobalance [HW,ACPI] 226 ACPI will not move active IRQs (default) 227 default in PIC mode 228 229 acpi_irq_isa= [HW,ACPI] If irq_balance, mark listed IRQs used by ISA 230 Format: <irq>,<irq>... 231 232 acpi_irq_pci= [HW,ACPI] If irq_balance, clear listed IRQs for 233 use by PCI 234 Format: <irq>,<irq>... 235 236 acpi_mask_gpe= [HW,ACPI] 237 Due to the existence of _Lxx/_Exx, some GPEs triggered 238 by unsupported hardware/firmware features can result in 239 GPE floodings that cannot be automatically disabled by 240 the GPE dispatcher. 241 This facility can be used to prevent such uncontrolled 242 GPE floodings. 243 Format: <byte> or <bitmap-list> 244 245 acpi_no_auto_serialize [HW,ACPI] 246 Disable auto-serialization of AML methods 247 AML control methods that contain the opcodes to create 248 named objects will be marked as "Serialized" by the 249 auto-serialization feature. 250 This feature is enabled by default. 251 This option allows to turn off the feature. 252 253 acpi_no_memhotplug [ACPI] Disable memory hotplug. Useful for kdump 254 kernels. 255 256 acpi_no_static_ssdt [HW,ACPI,EARLY] 257 Disable installation of static SSDTs at early boot time 258 By default, SSDTs contained in the RSDT/XSDT will be 259 installed automatically and they will appear under 260 /sys/firmware/acpi/tables. 261 This option turns off this feature. 262 Note that specifying this option does not affect 263 dynamic table installation which will install SSDT 264 tables to /sys/firmware/acpi/tables/dynamic. 265 266 acpi_no_watchdog [HW,ACPI,WDT] 267 Ignore the ACPI-based watchdog interface (WDAT) and let 268 a native driver control the watchdog device instead. 269 270 acpi_rsdp= [ACPI,EFI,KEXEC,EARLY] 271 Pass the RSDP address to the kernel, mostly used 272 on machines running EFI runtime service to boot the 273 second kernel for kdump. 274 275 acpi_os_name= [HW,ACPI] Tell ACPI BIOS the name of the OS 276 Format: To spoof as Windows 98: ="Microsoft Windows" 277 278 acpi_rev_override [ACPI] Override the _REV object to return 5 (instead 279 of 2 which is mandated by ACPI 6) as the supported ACPI 280 specification revision (when using this switch, it may 281 be necessary to carry out a cold reboot _twice_ in a 282 row to make it take effect on the platform firmware). 283 284 acpi_osi= [HW,ACPI] Modify list of supported OS interface strings 285 acpi_osi="string1" # add string1 286 acpi_osi="!string2" # remove string2 287 acpi_osi=!* # remove all strings 288 acpi_osi=! # disable all built-in OS vendor 289 strings 290 acpi_osi=!! # enable all built-in OS vendor 291 strings 292 acpi_osi= # disable all strings 293 294 'acpi_osi=!' can be used in combination with single or 295 multiple 'acpi_osi="string1"' to support specific OS 296 vendor string(s). Note that such command can only 297 affect the default state of the OS vendor strings, thus 298 it cannot affect the default state of the feature group 299 strings and the current state of the OS vendor strings, 300 specifying it multiple times through kernel command line 301 is meaningless. This command is useful when one do not 302 care about the state of the feature group strings which 303 should be controlled by the OSPM. 304 Examples: 305 1. 'acpi_osi=! acpi_osi="Windows 2000"' is equivalent 306 to 'acpi_osi="Windows 2000" acpi_osi=!', they all 307 can make '_OSI("Windows 2000")' TRUE. 308 309 'acpi_osi=' cannot be used in combination with other 310 'acpi_osi=' command lines, the _OSI method will not 311 exist in the ACPI namespace. NOTE that such command can 312 only affect the _OSI support state, thus specifying it 313 multiple times through kernel command line is also 314 meaningless. 315 Examples: 316 1. 'acpi_osi=' can make 'CondRefOf(_OSI, Local1)' 317 FALSE. 318 319 'acpi_osi=!*' can be used in combination with single or 320 multiple 'acpi_osi="string1"' to support specific 321 string(s). Note that such command can affect the 322 current state of both the OS vendor strings and the 323 feature group strings, thus specifying it multiple times 324 through kernel command line is meaningful. But it may 325 still not able to affect the final state of a string if 326 there are quirks related to this string. This command 327 is useful when one want to control the state of the 328 feature group strings to debug BIOS issues related to 329 the OSPM features. 330 Examples: 331 1. 'acpi_osi="Module Device" acpi_osi=!*' can make 332 '_OSI("Module Device")' FALSE. 333 2. 'acpi_osi=!* acpi_osi="Module Device"' can make 334 '_OSI("Module Device")' TRUE. 335 3. 'acpi_osi=! acpi_osi=!* acpi_osi="Windows 2000"' is 336 equivalent to 337 'acpi_osi=!* acpi_osi=! acpi_osi="Windows 2000"' 338 and 339 'acpi_osi=!* acpi_osi="Windows 2000" acpi_osi=!', 340 they all will make '_OSI("Windows 2000")' TRUE. 341 342 acpi_pm_good [X86] 343 Override the pmtimer bug detection: force the kernel 344 to assume that this machine's pmtimer latches its value 345 and always returns good values. 346 347 acpi_sci= [HW,ACPI,EARLY] ACPI System Control Interrupt trigger mode 348 Format: { level | edge | high | low } 349 350 acpi_skip_timer_override [HW,ACPI,EARLY] 351 Recognize and ignore IRQ0/pin2 Interrupt Override. 352 For broken nForce2 BIOS resulting in XT-PIC timer. 353 354 acpi_sleep= [HW,ACPI] Sleep options 355 Format: { s3_bios, s3_mode, s3_beep, s4_hwsig, 356 s4_nohwsig, old_ordering, nonvs, 357 sci_force_enable, nobl } 358 See Documentation/power/video.rst for information on 359 s3_bios and s3_mode. 360 s3_beep is for debugging; it makes the PC's speaker beep 361 as soon as the kernel's real-mode entry point is called. 362 s4_hwsig causes the kernel to check the ACPI hardware 363 signature during resume from hibernation, and gracefully 364 refuse to resume if it has changed. This complies with 365 the ACPI specification but not with reality, since 366 Windows does not do this and many laptops do change it 367 on docking. So the default behaviour is to allow resume 368 and simply warn when the signature changes, unless the 369 s4_hwsig option is enabled. 370 s4_nohwsig prevents ACPI hardware signature from being 371 used (or even warned about) during resume. 372 old_ordering causes the ACPI 1.0 ordering of the _PTS 373 control method, with respect to putting devices into 374 low power states, to be enforced (the ACPI 2.0 ordering 375 of _PTS is used by default). 376 nonvs prevents the kernel from saving/restoring the 377 ACPI NVS memory during suspend/hibernation and resume. 378 sci_force_enable causes the kernel to set SCI_EN directly 379 on resume from S1/S3 (which is against the ACPI spec, 380 but some broken systems don't work without it). 381 nobl causes the internal blacklist of systems known to 382 behave incorrectly in some ways with respect to system 383 suspend and resume to be ignored (use wisely). 384 385 acpi_use_timer_override [HW,ACPI,EARLY] 386 Use timer override. For some broken Nvidia NF5 boards 387 that require a timer override, but don't have HPET 388 389 add_efi_memmap [EFI,X86,EARLY] Include EFI memory map in 390 kernel's map of available physical RAM. 391 392 agp= [AGP] 393 { off | try_unsupported } 394 off: disable AGP support 395 try_unsupported: try to drive unsupported chipsets 396 (may crash computer or cause data corruption) 397 398 ALSA [HW,ALSA] 399 See Documentation/sound/alsa-configuration.rst 400 401 alignment= [KNL,ARM] 402 Allow the default userspace alignment fault handler 403 behaviour to be specified. Bit 0 enables warnings, 404 bit 1 enables fixups, and bit 2 sends a segfault. 405 406 align_va_addr= [X86-64] 407 Align virtual addresses by clearing slice [14:12] when 408 allocating a VMA at process creation time. This option 409 gives you up to 3% performance improvement on AMD F15h 410 machines (where it is enabled by default) for a 411 CPU-intensive style benchmark, and it can vary highly in 412 a microbenchmark depending on workload and compiler. 413 414 32: only for 32-bit processes 415 64: only for 64-bit processes 416 on: enable for both 32- and 64-bit processes 417 off: disable for both 32- and 64-bit processes 418 419 alloc_snapshot [FTRACE] 420 Allocate the ftrace snapshot buffer on boot up when the 421 main buffer is allocated. This is handy if debugging 422 and you need to use tracing_snapshot() on boot up, and 423 do not want to use tracing_snapshot_alloc() as it needs 424 to be done where GFP_KERNEL allocations are allowed. 425 426 allow_mismatched_32bit_el0 [ARM64,EARLY] 427 Allow execve() of 32-bit applications and setting of the 428 PER_LINUX32 personality on systems where only a strict 429 subset of the CPUs support 32-bit EL0. When this 430 parameter is present, the set of CPUs supporting 32-bit 431 EL0 is indicated by /sys/devices/system/cpu/aarch32_el0 432 and hot-unplug operations may be restricted. 433 434 See Documentation/arch/arm64/asymmetric-32bit.rst for more 435 information. 436 437 amd_iommu= [HW,X86-64] 438 Pass parameters to the AMD IOMMU driver in the system. 439 Possible values are: 440 fullflush - Deprecated, equivalent to iommu.strict=1 441 off - do not initialize any AMD IOMMU found in 442 the system 443 force_isolation - Force device isolation for all 444 devices. The IOMMU driver is not 445 allowed anymore to lift isolation 446 requirements as needed. This option 447 does not override iommu=pt 448 force_enable - Force enable the IOMMU on platforms known 449 to be buggy with IOMMU enabled. Use this 450 option with care. 451 pgtbl_v1 - Use v1 page table for DMA-API (Default). 452 pgtbl_v2 - Use v2 page table for DMA-API. 453 irtcachedis - Disable Interrupt Remapping Table (IRT) caching. 454 nohugepages - Limit page-sizes used for v1 page-tables 455 to 4 KiB. 456 v2_pgsizes_only - Limit page-sizes used for v1 page-tables 457 to 4KiB/2Mib/1GiB. 458 459 460 amd_iommu_dump= [HW,X86-64] 461 Enable AMD IOMMU driver option to dump the ACPI table 462 for AMD IOMMU. With this option enabled, AMD IOMMU 463 driver will print ACPI tables for AMD IOMMU during 464 IOMMU initialization. 465 466 amd_iommu_intr= [HW,X86-64] 467 Specifies one of the following AMD IOMMU interrupt 468 remapping modes: 469 legacy - Use legacy interrupt remapping mode. 470 vapic - Use virtual APIC mode, which allows IOMMU 471 to inject interrupts directly into guest. 472 This mode requires kvm-amd.avic=1. 473 (Default when IOMMU HW support is present.) 474 475 amd_pstate= [X86,EARLY] 476 disable 477 Do not enable amd_pstate as the default 478 scaling driver for the supported processors 479 passive 480 Use amd_pstate with passive mode as a scaling driver. 481 In this mode autonomous selection is disabled. 482 Driver requests a desired performance level and platform 483 tries to match the same performance level if it is 484 satisfied by guaranteed performance level. 485 active 486 Use amd_pstate_epp driver instance as the scaling driver, 487 driver provides a hint to the hardware if software wants 488 to bias toward performance (0x0) or energy efficiency (0xff) 489 to the CPPC firmware. then CPPC power algorithm will 490 calculate the runtime workload and adjust the realtime cores 491 frequency. 492 guided 493 Activate guided autonomous mode. Driver requests minimum and 494 maximum performance level and the platform autonomously 495 selects a performance level in this range and appropriate 496 to the current workload. 497 498 amd_prefcore= 499 [X86] 500 disable 501 Disable amd-pstate preferred core. 502 503 amd_dynamic_epp= 504 [X86] 505 disable 506 Disable amd-pstate dynamic EPP. 507 enable 508 Enable amd-pstate dynamic EPP. 509 510 amijoy.map= [HW,JOY] Amiga joystick support 511 Map of devices attached to JOY0DAT and JOY1DAT 512 Format: <a>,<b> 513 See also Documentation/input/joydev/joystick.rst 514 515 analog.map= [HW,JOY] Analog joystick and gamepad support 516 Specifies type or capabilities of an analog joystick 517 connected to one of 16 gameports 518 Format: <type1>,<type2>,..<type16> 519 520 apc= [HW,SPARC] 521 Power management functions (SPARCstation-4/5 + deriv.) 522 Format: noidle 523 Disable APC CPU standby support. SPARCstation-Fox does 524 not play well with APC CPU idle - disable it if you have 525 APC and your system crashes randomly. 526 527 apic [APIC,X86-64] Use IO-APIC. Default. 528 529 apic= [APIC,X86,EARLY] Advanced Programmable Interrupt Controller 530 Change the output verbosity while booting 531 Format: { quiet (default) | verbose | debug } 532 Change the amount of debugging information output 533 when initialising the APIC and IO-APIC components. 534 535 apic_extnmi= [APIC,X86,EARLY] External NMI delivery setting 536 Format: { bsp (default) | all | none } 537 bsp: External NMI is delivered only to CPU 0 538 all: External NMIs are broadcast to all CPUs as a 539 backup of CPU 0 540 none: External NMI is masked for all CPUs. This is 541 useful so that a dump capture kernel won't be 542 shot down by NMI 543 544 apicpmtimer Do APIC timer calibration using the pmtimer. Implies 545 apicmaintimer. Useful when your PIT timer is totally 546 broken. 547 548 autoconf= [IPV6] 549 See Documentation/networking/ipv6.rst. 550 551 apm= [APM] Advanced Power Management 552 See header of arch/x86/kernel/apm_32.c. 553 554 apparmor= [APPARMOR] Disable or enable AppArmor at boot time 555 Format: { "0" | "1" } 556 See security/apparmor/Kconfig help text 557 0 -- disable. 558 1 -- enable. 559 Default value is set via kernel config option. 560 561 arcrimi= [HW,NET] ARCnet - "RIM I" (entirely mem-mapped) cards 562 Format: <io>,<irq>,<nodeID> 563 564 arm64.no32bit_el0 [ARM64] Unconditionally disable the execution of 565 32 bit applications. 566 567 arm64.nobti [ARM64] Unconditionally disable Branch Target 568 Identification support 569 570 arm64.nogcs [ARM64] Unconditionally disable Guarded Control Stack 571 support 572 573 arm64.nomops [ARM64] Unconditionally disable Memory Copy and Memory 574 Set instructions support 575 576 arm64.nompam [ARM64] Unconditionally disable Memory Partitioning And 577 Monitoring support 578 579 arm64.nomte [ARM64] Unconditionally disable Memory Tagging Extension 580 support 581 582 arm64.nopauth [ARM64] Unconditionally disable Pointer Authentication 583 support 584 585 arm64.nosme [ARM64] Unconditionally disable Scalable Matrix 586 Extension support 587 588 arm64.nosve [ARM64] Unconditionally disable Scalable Vector 589 Extension support 590 591 ataflop= [HW,M68k] 592 593 atarimouse= [HW,MOUSE] Atari Mouse 594 595 atkbd.extra= [HW] Enable extra LEDs and keys on IBM RapidAccess, 596 EzKey and similar keyboards 597 598 atkbd.reset= [HW] Reset keyboard during initialization 599 600 atkbd.set= [HW] Select keyboard code set 601 Format: <int> (2 = AT (default), 3 = PS/2) 602 603 atkbd.scroll= [HW] Enable scroll wheel on MS Office and similar 604 keyboards 605 606 atkbd.softraw= [HW] Choose between synthetic and real raw mode 607 Format: <bool> (0 = real, 1 = synthetic (default)) 608 609 atkbd.softrepeat= [HW] 610 Use software keyboard repeat 611 612 audit= [KNL] Enable the audit sub-system 613 Format: { "0" | "1" | "off" | "on" } 614 0 | off - kernel audit is disabled and can not be 615 enabled until the next reboot 616 unset - kernel audit is initialized but disabled and 617 will be fully enabled by the userspace auditd. 618 1 | on - kernel audit is initialized and partially 619 enabled, storing at most audit_backlog_limit 620 messages in RAM until it is fully enabled by the 621 userspace auditd. 622 Default: unset 623 624 audit_backlog_limit= [KNL] Set the audit queue size limit. 625 Format: <int> (must be >=0) 626 Default: 64 627 628 bau= [X86_UV] Enable the BAU on SGI UV. The default 629 behavior is to disable the BAU (i.e. bau=0). 630 Format: { "0" | "1" } 631 0 - Disable the BAU. 632 1 - Enable the BAU. 633 unset - Disable the BAU. 634 635 bdev_allow_write_mounted= 636 Format: <bool> 637 Control the ability to open a mounted block device 638 for writing, i.e., allow / disallow writes that bypass 639 the FS. This was implemented as a means to prevent 640 fuzzers from crashing the kernel by overwriting the 641 metadata underneath a mounted FS without its awareness. 642 This also prevents destructive formatting of mounted 643 filesystems by naive storage tooling that don't use 644 O_EXCL. Default is Y and can be changed through the 645 Kconfig option CONFIG_BLK_DEV_WRITE_MOUNTED. 646 647 bert_disable [ACPI] 648 Disable BERT OS support on buggy BIOSes. 649 650 bgrt_disable [ACPI,X86,EARLY] 651 Disable BGRT to avoid flickering OEM logo. 652 653 blkdevparts= Manual partition parsing of block device(s) for 654 embedded devices based on command line input. 655 See Documentation/block/cmdline-partition.rst 656 657 boot_delay= [KNL,EARLY] 658 Milliseconds to delay each printk during boot. 659 Only works if CONFIG_BOOT_PRINTK_DELAY is enabled, 660 and you may also have to specify "lpj=". Boot_delay 661 values larger than 10 seconds (10000) are assumed 662 erroneous and ignored. 663 Format: integer 664 665 bootconfig [KNL,EARLY] 666 Extended command line options can be added to an initrd 667 and this will cause the kernel to look for it. 668 669 See Documentation/admin-guide/bootconfig.rst 670 671 bttv.card= [HW,V4L] bttv (bt848 + bt878 based grabber cards) 672 bttv.radio= Most important insmod options are available as 673 kernel args too. 674 bttv.pll= See Documentation/admin-guide/media/bttv.rst 675 bttv.tuner= 676 677 bulk_remove=off [PPC] This parameter disables the use of the pSeries 678 firmware feature for flushing multiple hpte entries 679 at a time. 680 681 c101= [NET] Moxa C101 synchronous serial card 682 683 cachesize= [BUGS=X86-32] Override level 2 CPU cache size detection. 684 Sometimes CPU hardware bugs make them report the cache 685 size incorrectly. The kernel will attempt work arounds 686 to fix known problems, but for some CPUs it is not 687 possible to determine what the correct size should be. 688 This option provides an override for these situations. 689 690 carrier_timeout= 691 [NET] Specifies amount of time (in seconds) that 692 the kernel should wait for a network carrier. By default 693 it waits 120 seconds. 694 695 ca_keys= [KEYS] This parameter identifies a specific key(s) on 696 the system trusted keyring to be used for certificate 697 trust validation. 698 format: { id:<keyid> | builtin } 699 700 cca= [MIPS,EARLY] Override the kernel pages' cache coherency 701 algorithm. Accepted values range from 0 to 7 702 inclusive. See arch/mips/include/asm/pgtable-bits.h 703 for platform specific values (SB1, Loongson3 and 704 others). 705 706 ccw_timeout_log [S390] 707 See Documentation/arch/s390/common_io.rst for details. 708 709 cfi= [X86-64] Set Control Flow Integrity checking features 710 when CONFIG_FINEIBT is enabled. 711 Format: feature[,feature...] 712 Default: auto 713 714 auto: Use FineIBT if IBT available, otherwise kCFI. 715 Under FineIBT, enable "paranoid" mode when 716 FRED is not available. 717 off: Turn off CFI checking. 718 kcfi: Use kCFI (disable FineIBT). 719 fineibt: Use FineIBT (even if IBT not available). 720 norand: Do not re-randomize CFI hashes. 721 paranoid: Add caller hash checking under FineIBT. 722 bhi: Enable register poisoning to stop speculation 723 across FineIBT. (Disabled by default.) 724 warn: Do not enforce CFI checking: warn only. 725 debug: Report CFI initialization details. 726 727 cgroup_disable= [KNL] Disable a particular controller or optional feature 728 Format: {name of the controller(s) or feature(s) to disable} 729 The effects of cgroup_disable=foo are: 730 - foo isn't auto-mounted if you mount all cgroups in 731 a single hierarchy 732 - foo isn't visible as an individually mountable 733 subsystem 734 - if foo is an optional feature then the feature is 735 disabled and corresponding cgroup files are not 736 created 737 {Currently only "memory" controller deal with this and 738 cut the overhead, others just disable the usage. So 739 only cgroup_disable=memory is actually worthy} 740 Specifying "pressure" disables per-cgroup pressure 741 stall information accounting feature 742 743 cgroup_no_v1= [KNL] Disable cgroup controllers and named hierarchies in v1 744 Format: { { controller | "all" | "named" } 745 [,{ controller | "all" | "named" }...] } 746 Like cgroup_disable, but only applies to cgroup v1; 747 the blacklisted controllers remain available in cgroup2. 748 "all" blacklists all controllers and "named" disables 749 named mounts. Specifying both "all" and "named" disables 750 all v1 hierarchies. 751 752 cgroup_v1_proc= [KNL] Show also missing controllers in /proc/cgroups 753 Format: { "true" | "false" } 754 /proc/cgroups lists only v1 controllers by default. 755 This compatibility option enables listing also v2 756 controllers (whose v1 code is not compiled!), so that 757 semi-legacy software can check this file to decide 758 about usage of v2 (sic) controllers. 759 760 cgroup_favordynmods= [KNL] Enable or Disable favordynmods. 761 Format: { "true" | "false" } 762 Defaults to the value of CONFIG_CGROUP_FAVOR_DYNMODS. 763 764 cgroup.memory= [KNL] Pass options to the cgroup memory controller. 765 Format: <string> 766 nosocket -- Disable socket memory accounting. 767 nokmem -- Disable kernel memory accounting. 768 nobpf -- Disable BPF memory accounting. 769 770 check_pages= [MM,EARLY] Enable sanity checking of pages after 771 allocations / before freeing. This adds checks to catch 772 double-frees, use-after-frees, and other sources of 773 page corruption by inspecting page internals (flags, 774 mapcount/refcount, memcg_data, etc.). 775 Format: { "0" | "1" } 776 Default: 0 (1 if CONFIG_DEBUG_VM is set) 777 778 checkreqprot= [SELINUX] Set initial checkreqprot flag value. 779 Format: { "0" | "1" } 780 See security/selinux/Kconfig help text. 781 0 -- check protection applied by kernel (includes 782 any implied execute protection). 783 1 -- check protection requested by application. 784 Default value is set via a kernel config option. 785 Value can be changed at runtime via 786 /sys/fs/selinux/checkreqprot. 787 Setting checkreqprot to 1 is deprecated. 788 789 cio_ignore= [S390] 790 See Documentation/arch/s390/common_io.rst for details. 791 792 clearcpuid=X[,X...] [X86] 793 Disable CPUID feature X for the kernel. See 794 arch/x86/include/asm/cpufeatures.h for the valid bit 795 numbers X. Note the Linux-specific bits are not necessarily 796 stable over kernel options, but the vendor-specific 797 ones should be. 798 X can also be a string as appearing in the flags: line 799 in /proc/cpuinfo which does not have the above 800 instability issue. However, not all features have names 801 in /proc/cpuinfo. 802 Note that using this option will taint your kernel. 803 Also note that user programs calling CPUID directly 804 or using the feature without checking anything 805 will still see it. This just prevents it from 806 being used by the kernel or shown in /proc/cpuinfo. 807 Also note the kernel might malfunction if you disable 808 some critical bits. 809 810 clk_ignore_unused 811 [CLK] 812 Prevents the clock framework from automatically gating 813 clocks that have not been explicitly enabled by a Linux 814 device driver but are enabled in hardware at reset or 815 by the bootloader/firmware. Note that this does not 816 force such clocks to be always-on nor does it reserve 817 those clocks in any way. This parameter is useful for 818 debug and development, but should not be needed on a 819 platform with proper driver support. For more 820 information, see Documentation/driver-api/clk.rst. 821 822 clock= [BUGS=X86-32, HW] gettimeofday clocksource override. 823 [Deprecated] 824 Forces specified clocksource (if available) to be used 825 when calculating gettimeofday(). If specified 826 clocksource is not available, it defaults to PIT. 827 Format: { pit | tsc | cyclone | pmtmr } 828 829 clocksource= Override the default clocksource 830 Format: <string> 831 Override the default clocksource and use the clocksource 832 with the name specified. 833 Some clocksource names to choose from, depending on 834 the platform: 835 [all] jiffies (this is the base, fallback clocksource) 836 [ACPI] acpi_pm 837 [ARM] imx_timer1,OSTS,netx_timer,mpu_timer2, 838 pxa_timer,timer3,32k_counter,timer0_1 839 [X86-32] pit,hpet,tsc; 840 scx200_hrt on Geode; cyclone on IBM x440 841 [MIPS] MIPS 842 [PARISC] cr16 843 [S390] tod 844 [SH] SuperH 845 [SPARC64] tick 846 [X86-64] hpet,tsc 847 848 clocksource.arm_arch_timer.evtstrm= 849 [ARM,ARM64,EARLY] 850 Format: <bool> 851 Enable/disable the eventstream feature of the ARM 852 architected timer so that code using WFE-based polling 853 loops can be debugged more effectively on production 854 systems. 855 856 clocksource.verify_n_cpus= [KNL] 857 Limit the number of CPUs checked for clocksources 858 marked with CLOCK_SOURCE_VERIFY_PERCPU that 859 are marked unstable due to excessive skew. 860 A negative value says to check all CPUs, while 861 zero says not to check any. Values larger than 862 nr_cpu_ids are silently truncated to nr_cpu_ids. 863 The actual CPUs are chosen randomly, with 864 no replacement if the same CPU is chosen twice. 865 866 clocksource-wdtest.holdoff= [KNL] 867 Set the time in seconds that the clocksource 868 watchdog test waits before commencing its tests. 869 Defaults to zero when built as a module and to 870 10 seconds when built into the kernel. 871 872 cma=nn[MG]@[start[MG][-end[MG]]] 873 [KNL,CMA,EARLY] 874 Sets the size of kernel global memory area for 875 contiguous memory allocations and optionally the 876 placement constraint by the physical address range of 877 memory allocations. A value of 0 disables CMA 878 altogether. For more information, see 879 kernel/dma/contiguous.c 880 881 cma_pernuma=nn[MG] 882 [KNL,CMA,EARLY] 883 Sets the size of kernel per-numa memory area for 884 contiguous memory allocations. A value of 0 disables 885 per-numa CMA altogether. And If this option is not 886 specified, the default value is 0. 887 With per-numa CMA enabled, DMA users on node nid will 888 first try to allocate buffer from the pernuma area 889 which is located in node nid, if the allocation fails, 890 they will fallback to the global default memory area. 891 892 numa_cma=<node>:nn[MG][,<node>:nn[MG]] 893 [KNL,CMA,EARLY] 894 Sets the size of kernel numa memory area for 895 contiguous memory allocations. It will reserve CMA 896 area for the specified node. 897 898 With numa CMA enabled, DMA users on node nid will 899 first try to allocate buffer from the numa area 900 which is located in node nid, if the allocation fails, 901 they will fallback to the global default memory area. 902 903 cmo_free_hint= [PPC] Format: { yes | no } 904 Specify whether pages are marked as being inactive 905 when they are freed. This is used in CMO environments 906 to determine OS memory pressure for page stealing by 907 a hypervisor. 908 Default: yes 909 910 coherent_pool=nn[KMG] [ARM,KNL,EARLY] 911 Sets the size of memory pool for coherent, atomic dma 912 allocations, by default set to 256K. 913 914 com20020= [HW,NET] ARCnet - COM20020 chipset 915 Format: 916 <io>[,<irq>[,<nodeID>[,<backplane>[,<ckp>[,<timeout>]]]]] 917 918 com90io= [HW,NET] ARCnet - COM90xx chipset (IO-mapped buffers) 919 Format: <io>[,<irq>] 920 921 com90xx= [HW,NET] 922 ARCnet - COM90xx chipset (memory-mapped buffers) 923 Format: <io>[,<irq>[,<memstart>]] 924 925 condev= [HW,S390] console device 926 conmode= 927 928 con3215_drop= [S390,EARLY] 3215 console drop mode. 929 Format: y|n|Y|N|1|0 930 When set to true, drop data on the 3215 console when 931 the console buffer is full. In this case the 932 operator using a 3270 terminal emulator (for example 933 x3270) does not have to enter the clear key for the 934 console output to advance and the kernel to continue. 935 This leads to a much faster boot time when a 3270 936 terminal emulator is active. If no 3270 terminal 937 emulator is used, this parameter has no effect. 938 939 console= [KNL] Output console device and options. 940 941 tty<n> Use the virtual console device <n>. 942 943 ttyS<n>[,options] 944 ttyUSB0[,options] 945 Use the specified serial port. The options are of 946 the form "bbbbpnf", where "bbbb" is the baud rate, 947 "p" is parity ("n", "o", or "e"), "n" is number of 948 bits, and "f" is flow control ("r" for RTS or 949 omit it). Default is "9600n8". 950 951 See Documentation/admin-guide/serial-console.rst for more 952 information. See 953 Documentation/networking/netconsole.rst for an 954 alternative. 955 956 <DEVNAME>:<n>.<n>[,options] 957 Use the specified serial port on the serial core bus. 958 The addressing uses DEVNAME of the physical serial port 959 device, followed by the serial core controller instance, 960 and the serial port instance. The options are the same 961 as documented for the ttyS addressing above. 962 963 The mapping of the serial ports to the tty instances 964 can be viewed with: 965 966 $ ls -d /sys/bus/serial-base/devices/*:*.*/tty/* 967 /sys/bus/serial-base/devices/00:04:0.0/tty/ttyS0 968 969 In the above example, the console can be addressed with 970 console=00:04:0.0. Note that a console addressed this 971 way will only get added when the related device driver 972 is ready. The use of an earlycon parameter in addition to 973 the console may be desired for console output early on. 974 975 uart[8250],io,<addr>[,options] 976 uart[8250],mmio,<addr>[,options] 977 uart[8250],mmio16,<addr>[,options] 978 uart[8250],mmio32,<addr>[,options] 979 uart[8250],0x<addr>[,options] 980 Start an early, polled-mode console on the 8250/16550 981 UART at the specified I/O port or MMIO address, 982 switching to the matching ttyS device later. 983 MMIO inter-register address stride is either 8-bit 984 (mmio), 16-bit (mmio16), or 32-bit (mmio32). 985 If none of [io|mmio|mmio16|mmio32], <addr> is assumed 986 to be equivalent to 'mmio'. 'options' are specified in 987 the same format described for ttyS above; if unspecified, 988 the h/w is not re-initialized. 989 990 hvc<n> Use the hypervisor console device <n>. This is for 991 both Xen and PowerPC hypervisors. 992 993 { null | "" } 994 Use to disable console output, i.e., to have kernel 995 console messages discarded. 996 This must be the only console= parameter used on the 997 kernel command line. 998 999 If the device connected to the port is not a TTY but a braille 1000 device, prepend "brl," before the device type, for instance 1001 console=brl,ttyS0 1002 For now, only VisioBraille is supported. 1003 1004 console_msg_format= 1005 [KNL] Change console messages format 1006 default 1007 By default we print messages on consoles in 1008 "[time stamp] text\n" format (time stamp may not be 1009 printed, depending on CONFIG_PRINTK_TIME or 1010 `printk_time' param). 1011 syslog 1012 Switch to syslog format: "<%u>[time stamp] text\n" 1013 IOW, each message will have a facility and loglevel 1014 prefix. The format is similar to one used by syslog() 1015 syscall, or to executing "dmesg -S --raw" or to reading 1016 from /proc/kmsg. 1017 1018 consoleblank= [KNL] The console blank (screen saver) timeout in 1019 seconds. A value of 0 disables the blank timer. 1020 Defaults to 0. 1021 1022 coredump_filter= 1023 [KNL] Change the default value for 1024 /proc/<pid>/coredump_filter. 1025 See also Documentation/filesystems/proc.rst. 1026 1027 coresight_cpu_debug.enable 1028 [ARM,ARM64] 1029 Format: <bool> 1030 Enable/disable the CPU sampling based debugging. 1031 0: default value, disable debugging 1032 1: enable debugging at boot time 1033 1034 cpcihp_generic= [HW,PCI] Generic port I/O CompactPCI driver 1035 Format: 1036 <first_slot>,<last_slot>,<port>,<enum_bit>[,<debug>] 1037 1038 cpuidle.off=1 [CPU_IDLE] 1039 disable the cpuidle sub-system 1040 1041 cpuidle.governor= 1042 [CPU_IDLE] Name of the cpuidle governor to use. 1043 1044 cpufreq.off=1 [CPU_FREQ] 1045 disable the cpufreq sub-system 1046 1047 cpufreq.default_governor= 1048 [CPU_FREQ] Name of the default cpufreq governor or 1049 policy to use. This governor must be registered in the 1050 kernel before the cpufreq driver probes. 1051 1052 cpu_init_udelay=N 1053 [X86,EARLY] Delay for N microsec between assert and de-assert 1054 of APIC INIT to start processors. This delay occurs 1055 on every CPU online, such as boot, and resume from suspend. 1056 Default: 10000 1057 1058 cpuhp.parallel= 1059 [SMP] Enable/disable parallel bringup of secondary CPUs 1060 Format: <bool> 1061 Default is enabled if CONFIG_HOTPLUG_PARALLEL=y. Otherwise 1062 the parameter has no effect. 1063 1064 crash_kexec_post_notifiers 1065 Only jump to kdump kernel after running the panic 1066 notifiers and dumping kmsg. This option increases 1067 the risks of a kdump failure, since some panic 1068 notifiers can make the crashed kernel more unstable. 1069 In configurations where kdump may not be reliable, 1070 running the panic notifiers could allow collecting 1071 more data on dmesg, like stack traces from other CPUS 1072 or extra data dumped by panic_print. Note that some 1073 configurations enable this option unconditionally, 1074 like Hyper-V, PowerPC (fadump) and AMD SEV-SNP. 1075 1076 crashkernel=size[KMG][@offset[KMG]] 1077 [KNL,EARLY] Using kexec, Linux can switch to a 'crash kernel' 1078 upon panic. This parameter reserves the physical 1079 memory region [offset, offset + size] for that kernel 1080 image. If '@offset' is omitted, then a suitable offset 1081 is selected automatically. 1082 [KNL, X86-64, ARM64, RISCV, LoongArch] Select a region 1083 under 4G first, and fall back to reserve region above 1084 4G when '@offset' hasn't been specified. 1085 See Documentation/admin-guide/kdump/kdump.rst for further details. 1086 1087 crashkernel=range1:size1[,range2:size2,...][@offset] 1088 [KNL] Same as above, but depends on the memory 1089 in the running system. The syntax of range is 1090 start-[end] where start and end are both 1091 a memory unit (amount[KMG]). See also 1092 Documentation/admin-guide/kdump/kdump.rst for an example. 1093 1094 crashkernel=size[KMG],high 1095 [KNL, X86-64, ARM64, RISCV, LoongArch] range could be 1096 above 4G. 1097 Allow kernel to allocate physical memory region from top, 1098 so could be above 4G if system have more than 4G ram 1099 installed. Otherwise memory region will be allocated 1100 below 4G, if available. 1101 It will be ignored if crashkernel=X is specified. 1102 crashkernel=size[KMG],low 1103 [KNL, X86-64, ARM64, RISCV, LoongArch] range under 4G. 1104 When crashkernel=X,high is passed, kernel could allocate 1105 physical memory region above 4G, that cause second kernel 1106 crash on system that require some amount of low memory, 1107 e.g. swiotlb requires at least 64M+32K low memory, also 1108 enough extra low memory is needed to make sure DMA buffers 1109 for 32-bit devices won't run out. Kernel would try to allocate 1110 default size of memory below 4G automatically. The default 1111 size is platform dependent. 1112 --> x86: max(swiotlb_size_or_default() + 8MiB, 256MiB) 1113 --> arm64: 128MiB 1114 --> riscv: 128MiB 1115 --> loongarch: 128MiB 1116 This one lets the user specify own low range under 4G 1117 for second kernel instead. 1118 0: to disable low allocation. 1119 It will be ignored when crashkernel=X,high is not used 1120 or memory reserved is below 4G. 1121 crashkernel=size[KMG],cma 1122 [KNL, X86, ppc] Reserve additional crash kernel memory from 1123 CMA. This reservation is usable by the first system's 1124 userspace memory and kernel movable allocations (memory 1125 balloon, zswap). Pages allocated from this memory range 1126 will not be included in the vmcore so this should not 1127 be used if dumping of userspace memory is intended and 1128 it has to be expected that some movable kernel pages 1129 may be missing from the dump. 1130 1131 A standard crashkernel reservation, as described above, 1132 is still needed to hold the crash kernel and initrd. 1133 1134 This option increases the risk of a kdump failure: DMA 1135 transfers configured by the first kernel may end up 1136 corrupting the second kernel's memory. 1137 1138 This reservation method is intended for systems that 1139 can't afford to sacrifice enough memory for standard 1140 crashkernel reservation and where less reliable and 1141 possibly incomplete kdump is preferable to no kdump at 1142 all. 1143 1144 cryptomgr.notests 1145 [KNL] Disable crypto self-tests 1146 1147 cs89x0_dma= [HW,NET] 1148 Format: <dma> 1149 1150 cs89x0_media= [HW,NET] 1151 Format: { rj45 | aui | bnc } 1152 1153 csdlock_debug= [KNL] Enable or disable debug add-ons of cross-CPU 1154 function call handling. When switched on, 1155 additional debug data is printed to the console 1156 in case a hanging CPU is detected, and that 1157 CPU is pinged again in order to try to resolve 1158 the hang situation. The default value of this 1159 option depends on the CSD_LOCK_WAIT_DEBUG_DEFAULT 1160 Kconfig option. 1161 1162 dasd= [HW,NET] 1163 See header of drivers/s390/block/dasd_devmap.c. 1164 1165 db9.dev[2|3]= [HW,JOY] Multisystem joystick support via parallel port 1166 (one device per port) 1167 Format: <port#>,<type> 1168 See also Documentation/input/devices/joystick-parport.rst 1169 1170 debug [KNL,EARLY] Enable kernel debugging (events log level). 1171 1172 debug_boot_weak_hash 1173 [KNL,EARLY] Enable printing [hashed] pointers early in the 1174 boot sequence. If enabled, we use a weak hash instead 1175 of siphash to hash pointers. Use this option if you are 1176 seeing instances of '(___ptrval___)') and need to see a 1177 value (hashed pointer) instead. Cryptographically 1178 insecure, please do not use on production kernels. 1179 1180 debug_locks_verbose= 1181 [KNL] verbose locking self-tests 1182 Format: <int> 1183 Print debugging info while doing the locking API 1184 self-tests. 1185 Bitmask for the various LOCKTYPE_ tests. Defaults to 0 1186 (no extra messages), setting it to -1 (all bits set) 1187 will print _a_lot_ more information - normally only 1188 useful to lockdep developers. 1189 1190 debug_objects [KNL,EARLY] Enable object debugging 1191 1192 debug_guardpage_minorder= 1193 [KNL,EARLY] When CONFIG_DEBUG_PAGEALLOC is set, this 1194 parameter allows control of the order of pages that will 1195 be intentionally kept free (and hence protected) by the 1196 buddy allocator. Bigger value increase the probability 1197 of catching random memory corruption, but reduce the 1198 amount of memory for normal system use. The maximum 1199 possible value is MAX_PAGE_ORDER/2. Setting this 1200 parameter to 1 or 2 should be enough to identify most 1201 random memory corruption problems caused by bugs in 1202 kernel or driver code when a CPU writes to (or reads 1203 from) a random memory location. Note that there exists 1204 a class of memory corruptions problems caused by buggy 1205 H/W or F/W or by drivers badly programming DMA 1206 (basically when memory is written at bus level and the 1207 CPU MMU is bypassed) which are not detectable by 1208 CONFIG_DEBUG_PAGEALLOC, hence this option will not 1209 help tracking down these problems. 1210 1211 debug_pagealloc= 1212 [KNL,EARLY] When CONFIG_DEBUG_PAGEALLOC is set, this parameter 1213 enables the feature at boot time. By default, it is 1214 disabled and the system will work mostly the same as a 1215 kernel built without CONFIG_DEBUG_PAGEALLOC. 1216 Note: to get most of debug_pagealloc error reports, it's 1217 useful to also enable the page_owner functionality. 1218 on: enable the feature 1219 1220 debugfs= [KNL,EARLY] This parameter enables what is exposed to 1221 userspace and debugfs internal clients. 1222 Format: { on, off } 1223 on: All functions are enabled. 1224 off: Filesystem is not registered and clients 1225 get a -EPERM as result when trying to register files 1226 or directories within debugfs. 1227 This is equivalent of the runtime functionality if 1228 debugfs was not enabled in the kernel at all. 1229 Default value is set in build-time with a kernel configuration. 1230 1231 debugpat [X86] Enable PAT debugging 1232 1233 default_hugepagesz= 1234 [HW] The size of the default HugeTLB page. This is 1235 the size represented by the legacy /proc/ hugepages 1236 APIs. In addition, this is the default hugetlb size 1237 used for shmget(), mmap() and mounting hugetlbfs 1238 filesystems. If not specified, defaults to the 1239 architecture's default huge page size. Huge page 1240 sizes are architecture dependent. See also 1241 Documentation/admin-guide/mm/hugetlbpage.rst. 1242 Format: size[KMG] 1243 1244 deferred_probe_timeout= 1245 [KNL] Debugging option to set a timeout in seconds for 1246 deferred probe to give up waiting on dependencies to 1247 probe. Only specific dependencies (subsystems or 1248 drivers) that have opted in will be ignored. A timeout 1249 of 0 will timeout at the end of initcalls. If the time 1250 out hasn't expired, it'll be restarted by each 1251 successful driver registration. This option will also 1252 dump out devices still on the deferred probe list after 1253 retrying. 1254 1255 delayacct [KNL] Enable per-task delay accounting 1256 1257 dell_smm_hwmon.ignore_dmi= 1258 [HW] Continue probing hardware even if DMI data 1259 indicates that the driver is running on unsupported 1260 hardware. 1261 1262 dell_smm_hwmon.force= 1263 [HW] Activate driver even if SMM BIOS signature does 1264 not match list of supported models and enable otherwise 1265 blacklisted features. 1266 1267 dell_smm_hwmon.power_status= 1268 [HW] Report power status in /proc/i8k 1269 (disabled by default). 1270 1271 dell_smm_hwmon.restricted= 1272 [HW] Allow controlling fans only if SYS_ADMIN 1273 capability is set. 1274 1275 dell_smm_hwmon.fan_mult= 1276 [HW] Factor to multiply fan speed with. 1277 1278 dell_smm_hwmon.fan_max= 1279 [HW] Maximum configurable fan speed. 1280 1281 dfltcc= [HW,S390] 1282 Format: { on | off | def_only | inf_only | always } 1283 on: s390 zlib hardware support for compression on 1284 level 1 and decompression (default) 1285 off: No s390 zlib hardware support 1286 def_only: s390 zlib hardware support for deflate 1287 only (compression on level 1) 1288 inf_only: s390 zlib hardware support for inflate 1289 only (decompression) 1290 always: Same as 'on' but ignores the selected compression 1291 level always using hardware support (used for debugging) 1292 1293 dhash_entries= [KNL] 1294 Set number of hash buckets for dentry cache. 1295 1296 disable_1tb_segments [PPC,EARLY] 1297 Disables the use of 1TB hash page table segments. This 1298 causes the kernel to fall back to 256MB segments which 1299 can be useful when debugging issues that require an SLB 1300 miss to occur. 1301 1302 disable= [IPV6] 1303 See Documentation/networking/ipv6.rst. 1304 1305 disable_radix [PPC,EARLY] 1306 Disable RADIX MMU mode on POWER9 1307 1308 disable_tlbie [PPC] 1309 Disable TLBIE instruction. Currently does not work 1310 with KVM, with HASH MMU, or with coherent accelerators. 1311 1312 disable_ddw [PPC/PSERIES,EARLY] 1313 Disable Dynamic DMA Window support. Use this 1314 to workaround buggy firmware. 1315 1316 disable_ipv6= [IPV6] 1317 See Documentation/networking/ipv6.rst. 1318 1319 disable_mtrr_cleanup [X86,EARLY] 1320 The kernel tries to adjust MTRR layout from continuous 1321 to discrete, to make X server driver able to add WB 1322 entry later. This parameter disables that. 1323 1324 disable_mtrr_trim [X86, Intel and AMD only,EARLY] 1325 By default the kernel will trim any uncacheable 1326 memory out of your available memory pool based on 1327 MTRR settings. This parameter disables that behavior, 1328 possibly causing your machine to run very slowly. 1329 1330 disable_timer_pin_1 [X86,EARLY] 1331 Disable PIN 1 of APIC timer 1332 Can be useful to work around chipset bugs. 1333 1334 dis_ucode_ldr [X86] Disable the microcode loader. 1335 1336 dma_debug=off If the kernel is compiled with DMA_API_DEBUG support, 1337 this option disables the debugging code at boot. 1338 1339 dma_debug_entries=<number> 1340 This option allows to tune the number of preallocated 1341 entries for DMA-API debugging code. One entry is 1342 required per DMA-API allocation. Use this if the 1343 DMA-API debugging code disables itself because the 1344 architectural default is too low. 1345 1346 dma_debug_driver=<driver_name> 1347 With this option the DMA-API debugging driver 1348 filter feature can be enabled at boot time. Just 1349 pass the driver to filter for as the parameter. 1350 The filter can be disabled or changed to another 1351 driver later using sysfs. 1352 1353 reg_file_data_sampling= 1354 [X86] Controls mitigation for Register File Data 1355 Sampling (RFDS) vulnerability. RFDS is a CPU 1356 vulnerability which may allow userspace to infer 1357 kernel data values previously stored in floating point 1358 registers, vector registers, or integer registers. 1359 RFDS only affects Intel Atom processors. 1360 1361 on: Turns ON the mitigation. 1362 off: Turns OFF the mitigation. 1363 1364 This parameter overrides the compile time default set 1365 by CONFIG_MITIGATION_RFDS. Mitigation cannot be 1366 disabled when other VERW based mitigations (like MDS) 1367 are enabled. In order to disable RFDS mitigation all 1368 VERW based mitigations need to be disabled. 1369 1370 For details see: 1371 Documentation/admin-guide/hw-vuln/reg-file-data-sampling.rst 1372 1373 dm_verity.keyring_unsealed= 1374 [KNL] When set to 1, leave the dm-verity keyring 1375 unsealed after initialization so userspace can 1376 provision keys. Once the keyring is restricted 1377 it becomes active and is searched during signature 1378 verification. 1379 1380 driver_async_probe= [KNL] 1381 List of driver names to be probed asynchronously. * 1382 matches with all driver names. If * is specified, the 1383 rest of the listed driver names are those that will NOT 1384 match the *. 1385 Format: <driver_name1>,<driver_name2>... 1386 1387 drm.edid_firmware=[<connector>:]<file>[,[<connector>:]<file>] 1388 Broken monitors, graphic adapters, KVMs and EDIDless 1389 panels may send no or incorrect EDID data sets. 1390 This parameter allows to specify an EDID data sets 1391 in the /lib/firmware directory that are used instead. 1392 An EDID data set will only be used for a particular 1393 connector, if its name and a colon are prepended to 1394 the EDID name. Each connector may use a unique EDID 1395 data set by separating the files with a comma. An EDID 1396 data set with no connector name will be used for 1397 any connectors not explicitly specified. 1398 1399 dscc4.setup= [NET] 1400 1401 dt_cpu_ftrs= [PPC,EARLY] 1402 Format: {"off" | "known"} 1403 Control how the dt_cpu_ftrs device-tree binding is 1404 used for CPU feature discovery and setup (if it 1405 exists). 1406 off: Do not use it, fall back to legacy cpu table. 1407 known: Do not pass through unknown features to guests 1408 or userspace, only those that the kernel is aware of. 1409 1410 dump_apple_properties [X86] 1411 Dump name and content of EFI device properties on 1412 x86 Macs. Useful for driver authors to determine 1413 what data is available or for reverse-engineering. 1414 1415 dyndbg[="val"] [KNL,DYNAMIC_DEBUG] 1416 <module>.dyndbg[="val"] 1417 Enable debug messages at boot time. See 1418 Documentation/admin-guide/dynamic-debug-howto.rst 1419 for details. 1420 1421 early_ioremap_debug [KNL,EARLY] 1422 Enable debug messages in early_ioremap support. This 1423 is useful for tracking down temporary early mappings 1424 which are not unmapped. 1425 1426 earlycon= [KNL,EARLY] Output early console device and options. 1427 1428 When used with no options, the early console is 1429 determined by stdout-path property in device tree's 1430 chosen node or the ACPI SPCR table if supported by 1431 the platform. 1432 1433 cdns,<addr>[,options] 1434 Start an early, polled-mode console on a Cadence 1435 (xuartps) serial port at the specified address. Only 1436 supported option is baud rate. If baud rate is not 1437 specified, the serial port must already be setup and 1438 configured. 1439 1440 uart[8250],io,<addr>[,options[,uartclk]] 1441 uart[8250],mmio,<addr>[,options[,uartclk]] 1442 uart[8250],mmio32,<addr>[,options[,uartclk]] 1443 uart[8250],mmio32be,<addr>[,options[,uartclk]] 1444 uart[8250],0x<addr>[,options] 1445 Start an early, polled-mode console on the 8250/16550 1446 UART at the specified I/O port or MMIO address. 1447 MMIO inter-register address stride is either 8-bit 1448 (mmio) or 32-bit (mmio32 or mmio32be). 1449 If none of [io|mmio|mmio32|mmio32be], <addr> is assumed 1450 to be equivalent to 'mmio'. 'options' are specified 1451 in the same format described for "console=ttyS<n>"; if 1452 unspecified, the h/w is not initialized. 'uartclk' is 1453 the uart clock frequency; if unspecified, it is set 1454 to 'BASE_BAUD' * 16. 1455 1456 pl011,<addr> 1457 pl011,mmio32,<addr> 1458 Start an early, polled-mode console on a pl011 serial 1459 port at the specified address. The pl011 serial port 1460 must already be setup and configured. Options are not 1461 yet supported. If 'mmio32' is specified, then only 1462 the driver will use only 32-bit accessors to read/write 1463 the device registers. 1464 1465 liteuart,<addr> 1466 Start an early console on a litex serial port at the 1467 specified address. The serial port must already be 1468 setup and configured. Options are not yet supported. 1469 1470 meson,<addr> 1471 Start an early, polled-mode console on a meson serial 1472 port at the specified address. The serial port must 1473 already be setup and configured. Options are not yet 1474 supported. 1475 1476 msm_serial,<addr> 1477 Start an early, polled-mode console on an msm serial 1478 port at the specified address. The serial port 1479 must already be setup and configured. Options are not 1480 yet supported. 1481 1482 msm_serial_dm,<addr> 1483 Start an early, polled-mode console on an msm serial 1484 dm port at the specified address. The serial port 1485 must already be setup and configured. Options are not 1486 yet supported. 1487 1488 owl,<addr> 1489 Start an early, polled-mode console on a serial port 1490 of an Actions Semi SoC, such as S500 or S900, at the 1491 specified address. The serial port must already be 1492 setup and configured. Options are not yet supported. 1493 1494 rda,<addr> 1495 Start an early, polled-mode console on a serial port 1496 of an RDA Micro SoC, such as RDA8810PL, at the 1497 specified address. The serial port must already be 1498 setup and configured. Options are not yet supported. 1499 1500 sbi 1501 Use RISC-V SBI (Supervisor Binary Interface) for early 1502 console. 1503 1504 smh Use ARM semihosting calls for early console. 1505 1506 s3c2410,<addr> 1507 s3c2412,<addr> 1508 s3c2440,<addr> 1509 s3c6400,<addr> 1510 s5pv210,<addr> 1511 exynos4210,<addr> 1512 Use early console provided by serial driver available 1513 on Samsung SoCs, requires selecting proper type and 1514 a correct base address of the selected UART port. The 1515 serial port must already be setup and configured. 1516 Options are not yet supported. 1517 1518 lantiq,<addr> 1519 Start an early, polled-mode console on a lantiq serial 1520 (lqasc) port at the specified address. The serial port 1521 must already be setup and configured. Options are not 1522 yet supported. 1523 1524 lpuart,<addr> 1525 lpuart32,<addr> 1526 Use early console provided by Freescale LP UART driver 1527 found on Freescale Vybrid and QorIQ LS1021A processors. 1528 A valid base address must be provided, and the serial 1529 port must already be setup and configured. 1530 1531 ec_imx21,<addr> 1532 ec_imx6q,<addr> 1533 Start an early, polled-mode, output-only console on the 1534 Freescale i.MX UART at the specified address. The UART 1535 must already be setup and configured. 1536 1537 ar3700_uart,<addr> 1538 Start an early, polled-mode console on the 1539 Armada 3700 serial port at the specified 1540 address. The serial port must already be setup 1541 and configured. Options are not yet supported. 1542 1543 qcom_geni,<addr> 1544 Start an early, polled-mode console on a Qualcomm 1545 Generic Interface (GENI) based serial port at the 1546 specified address. The serial port must already be 1547 setup and configured. Options are not yet supported. 1548 1549 efifb,[options] 1550 Start an early, unaccelerated console on the EFI 1551 memory mapped framebuffer (if available). On cache 1552 coherent non-x86 systems that use system memory for 1553 the framebuffer, pass the 'ram' option so that it is 1554 mapped with the correct attributes. 1555 1556 linflex,<addr> 1557 Use early console provided by Freescale LINFlexD UART 1558 serial driver for NXP S32V234 SoCs. A valid base 1559 address must be provided, and the serial port must 1560 already be setup and configured. 1561 1562 earlyprintk= [X86,SH,ARM,M68k,S390,UM,EARLY] 1563 earlyprintk=vga 1564 earlyprintk=sclp 1565 earlyprintk=xen 1566 earlyprintk=serial[,ttySn[,baudrate]] 1567 earlyprintk=serial[,0x...[,baudrate]] 1568 earlyprintk=ttySn[,baudrate] 1569 earlyprintk=dbgp[debugController#] 1570 earlyprintk=mmio32,membase[,{nocfg|baudrate}] 1571 earlyprintk=pciserial[,force],bus:device.function[,{nocfg|baudrate}] 1572 earlyprintk=xdbc[xhciController#] 1573 earlyprintk=bios 1574 1575 earlyprintk is useful when the kernel crashes before 1576 the normal console is initialized. It is not enabled by 1577 default because it has some cosmetic problems. 1578 1579 Use "nocfg" to skip UART configuration, assume 1580 BIOS/firmware has configured UART correctly. 1581 1582 Append ",keep" to not disable it when the real console 1583 takes over. 1584 1585 Only one of vga, serial, or usb debug port can 1586 be used at a time. 1587 1588 Currently only ttyS0 and ttyS1 may be specified by 1589 name. Other I/O ports may be explicitly specified 1590 on some architectures (x86 and arm at least) by 1591 replacing ttySn with an I/O port address, like this: 1592 earlyprintk=serial,0x1008,115200 1593 You can find the port for a given device in 1594 /proc/tty/driver/serial: 1595 2: uart:ST16650V2 port:00001008 irq:18 ... 1596 1597 Interaction with the standard serial driver is not 1598 very good. 1599 1600 The VGA output is eventually overwritten by 1601 the real console. 1602 1603 The xen option can only be used in Xen domains. 1604 1605 The sclp output can only be used on s390. 1606 1607 The bios output can only be used on SuperH. 1608 1609 The optional "force" to "pciserial" enables use of a 1610 PCI device even when its classcode is not of the 1611 UART class. 1612 1613 edac_report= [HW,EDAC] Control how to report EDAC event 1614 Format: {"on" | "off" | "force"} 1615 on: enable EDAC to report H/W event. May be overridden 1616 by other higher priority error reporting module. 1617 off: disable H/W event reporting through EDAC. 1618 force: enforce the use of EDAC to report H/W event. 1619 default: on. 1620 1621 edd= [EDD] 1622 Format: {"off" | "on" | "skip[mbr]"} 1623 1624 efi= [EFI,EARLY] 1625 Format: { "debug", "disable_early_pci_dma", 1626 "nochunk", "noruntime", "nosoftreserve", 1627 "novamap", "no_disable_early_pci_dma" } 1628 debug: enable misc debug output. 1629 disable_early_pci_dma: disable the busmaster bit on all 1630 PCI bridges while in the EFI boot stub. 1631 nochunk: disable reading files in "chunks" in the EFI 1632 boot stub, as chunking can cause problems with some 1633 firmware implementations. 1634 noruntime : disable EFI runtime services support 1635 nosoftreserve: The EFI_MEMORY_SP (Specific Purpose) 1636 attribute may cause the kernel to reserve the 1637 memory range for a memory mapping driver to 1638 claim. Specify efi=nosoftreserve to disable this 1639 reservation and treat the memory by its base type 1640 (i.e. EFI_CONVENTIONAL_MEMORY / "System RAM"). 1641 novamap: do not call SetVirtualAddressMap(). 1642 no_disable_early_pci_dma: Leave the busmaster bit set 1643 on all PCI bridges while in the EFI boot stub 1644 1645 efi_no_storage_paranoia [EFI,X86,EARLY] 1646 Using this parameter you can use more than 50% of 1647 your efi variable storage. Use this parameter only if 1648 you are really sure that your UEFI does sane gc and 1649 fulfills the spec otherwise your board may brick. 1650 1651 efivar_ssdt= [EFI; X86] Name of an EFI variable that contains an SSDT 1652 that is to be dynamically loaded by Linux. If there are 1653 multiple variables with the same name but with different 1654 vendor GUIDs, all of them will be loaded. See 1655 Documentation/admin-guide/acpi/ssdt-overlays.rst for details. 1656 1657 1658 eisa_irq_edge= [PARISC,HW] 1659 See header of drivers/parisc/eisa.c. 1660 1661 ekgdboc= [X86,KGDB,EARLY] Allow early kernel console debugging 1662 Format: ekgdboc=kbd 1663 1664 This is designed to be used in conjunction with 1665 the boot argument: earlyprintk=vga 1666 1667 This parameter works in place of the kgdboc parameter 1668 but can only be used if the backing tty is available 1669 very early in the boot process. For early debugging 1670 via a serial port see kgdboc_earlycon instead. 1671 1672 elanfreq= [X86-32] 1673 See comment before function elanfreq_setup() in 1674 arch/x86/kernel/cpu/cpufreq/elanfreq.c. 1675 1676 elfcorehdr=[size[KMG]@]offset[KMG] [PPC,SH,X86,S390,EARLY] 1677 Specifies physical address of start of kernel core 1678 image elf header and optionally the size. Generally 1679 kexec loader will pass this option to capture kernel. 1680 See Documentation/admin-guide/kdump/kdump.rst for details. 1681 1682 enable_mtrr_cleanup [X86,EARLY] 1683 The kernel tries to adjust MTRR layout from continuous 1684 to discrete, to make X server driver able to add WB 1685 entry later. This parameter enables that. 1686 1687 enable_timer_pin_1 [X86] 1688 Enable PIN 1 of APIC timer 1689 Can be useful to work around chipset bugs 1690 (in particular on some ATI chipsets). 1691 The kernel tries to set a reasonable default. 1692 1693 enforcing= [SELINUX] Set initial enforcing status. 1694 Format: {"0" | "1"} 1695 See security/selinux/Kconfig help text. 1696 0 -- permissive (log only, no denials). 1697 1 -- enforcing (deny and log). 1698 Default value is 0. 1699 Value can be changed at runtime via 1700 /sys/fs/selinux/enforce. 1701 1702 erst_disable [ACPI] 1703 Disable Error Record Serialization Table (ERST) 1704 support. 1705 1706 ether= [HW,NET] Ethernet cards parameters 1707 This option is obsoleted by the "netdev=" option, which 1708 has equivalent usage. See its documentation for details. 1709 1710 evm= [EVM] 1711 Format: { "fix" } 1712 Permit 'security.evm' to be updated regardless of 1713 current integrity status. 1714 1715 early_page_ext [KNL,EARLY] Enforces page_ext initialization to earlier 1716 stages so cover more early boot allocations. 1717 Please note that as side effect some optimizations 1718 might be disabled to achieve that (e.g. parallelized 1719 memory initialization is disabled) so the boot process 1720 might take longer, especially on systems with a lot of 1721 memory. Available with CONFIG_PAGE_EXTENSION=y. 1722 1723 failslab= 1724 fail_usercopy= 1725 fail_page_alloc= 1726 fail_skb_realloc= 1727 fail_make_request=[KNL] 1728 General fault injection mechanism. 1729 Format: <interval>,<probability>,<space>,<times> 1730 See also Documentation/fault-injection/. 1731 1732 fb_tunnels= [NET] 1733 Format: { initns | none } 1734 See Documentation/admin-guide/sysctl/net.rst for 1735 fb_tunnels_only_for_init_ns 1736 1737 floppy= [HW] 1738 See Documentation/admin-guide/blockdev/floppy.rst. 1739 1740 forcepae [X86-32] 1741 Forcefully enable Physical Address Extension (PAE). 1742 Many Pentium M systems disable PAE but may have a 1743 functionally usable PAE implementation. 1744 Warning: use of this parameter will taint the kernel 1745 and may cause unknown problems. 1746 1747 fred= [X86-64] 1748 Enable/disable Flexible Return and Event Delivery. 1749 Format: { on | off } 1750 on: enable FRED when it's present, the default setting. 1751 off: disable FRED. 1752 1753 ftrace=[tracer] 1754 [FTRACE] will set and start the specified tracer 1755 as early as possible in order to facilitate early 1756 boot debugging. 1757 1758 ftrace_boot_snapshot 1759 [FTRACE] On boot up, a snapshot will be taken of the 1760 ftrace ring buffer that can be read at: 1761 /sys/kernel/tracing/snapshot. 1762 This is useful if you need tracing information from kernel 1763 boot up that is likely to be overridden by user space 1764 start up functionality. 1765 1766 Optionally, the snapshot can also be defined for a tracing 1767 instance that was created by the trace_instance= command 1768 line parameter. 1769 1770 trace_instance=foo,sched_switch ftrace_boot_snapshot=foo 1771 1772 The above will cause the "foo" tracing instance to trigger 1773 a snapshot at the end of boot up. 1774 1775 ftrace_dump_on_oops[=2(orig_cpu) | =<instance>][,<instance> | 1776 ,<instance>=2(orig_cpu)] 1777 [FTRACE] will dump the trace buffers on oops. 1778 If no parameter is passed, ftrace will dump global 1779 buffers of all CPUs, if you pass 2 or orig_cpu, it 1780 will dump only the buffer of the CPU that triggered 1781 the oops, or the specific instance will be dumped if 1782 its name is passed. Multiple instance dump is also 1783 supported, and instances are separated by commas. Each 1784 instance supports only dump on CPU that triggered the 1785 oops by passing 2 or orig_cpu to it. 1786 1787 ftrace_dump_on_oops=foo=orig_cpu 1788 1789 The above will dump only the buffer of "foo" instance 1790 on CPU that triggered the oops. 1791 1792 ftrace_dump_on_oops,foo,bar=orig_cpu 1793 1794 The above will dump global buffer on all CPUs, the 1795 buffer of "foo" instance on all CPUs and the buffer 1796 of "bar" instance on CPU that triggered the oops. 1797 1798 ftrace_filter=[function-list] 1799 [FTRACE] Limit the functions traced by the function 1800 tracer at boot up. function-list is a comma-separated 1801 list of functions. This list can be changed at run 1802 time by the set_ftrace_filter file in the debugfs 1803 tracing directory. 1804 1805 ftrace_notrace=[function-list] 1806 [FTRACE] Do not trace the functions specified in 1807 function-list. This list can be changed at run time 1808 by the set_ftrace_notrace file in the debugfs 1809 tracing directory. 1810 1811 ftrace_graph_filter=[function-list] 1812 [FTRACE] Limit the top level callers functions traced 1813 by the function graph tracer at boot up. 1814 function-list is a comma-separated list of functions 1815 that can be changed at run time by the 1816 set_graph_function file in the debugfs tracing directory. 1817 1818 ftrace_graph_notrace=[function-list] 1819 [FTRACE] Do not trace from the functions specified in 1820 function-list. This list is a comma-separated list of 1821 functions that can be changed at run time by the 1822 set_graph_notrace file in the debugfs tracing directory. 1823 1824 ftrace_graph_max_depth=<uint> 1825 [FTRACE] Used with the function graph tracer. This is 1826 the max depth it will trace into a function. This value 1827 can be changed at run time by the max_graph_depth file 1828 in the tracefs tracing directory. default: 0 (no limit) 1829 1830 fw_devlink= [KNL,EARLY] Create device links between consumer and supplier 1831 devices by scanning the firmware to infer the 1832 consumer/supplier relationships. This feature is 1833 especially useful when drivers are loaded as modules as 1834 it ensures proper ordering of tasks like device probing 1835 (suppliers first, then consumers), supplier boot state 1836 clean up (only after all consumers have probed), 1837 suspend/resume & runtime PM (consumers first, then 1838 suppliers). 1839 Format: { off | permissive | on | rpm } 1840 off -- Don't create device links from firmware info. 1841 permissive -- Create device links from firmware info 1842 but use it only for ordering boot state clean 1843 up (sync_state() calls). 1844 on -- Create device links from firmware info and use it 1845 to enforce probe and suspend/resume ordering. 1846 rpm -- Like "on", but also use to order runtime PM. 1847 1848 fw_devlink.strict=<bool> 1849 [KNL,EARLY] Treat all inferred dependencies as mandatory 1850 dependencies. This only applies for fw_devlink=on|rpm. 1851 Format: <bool> 1852 1853 fw_devlink.sync_state = 1854 [KNL,EARLY] When all devices that could probe have finished 1855 probing, this parameter controls what to do with 1856 devices that haven't yet received their sync_state() 1857 calls. 1858 Format: { strict | timeout } 1859 strict -- Default. Continue waiting on consumers to 1860 probe successfully. 1861 timeout -- Give up waiting on consumers and call 1862 sync_state() on any devices that haven't yet 1863 received their sync_state() calls after 1864 deferred_probe_timeout has expired or by 1865 late_initcall() if !CONFIG_MODULES. 1866 1867 gamecon.map[2|3]= 1868 [HW,JOY] Multisystem joystick and NES/SNES/PSX pad 1869 support via parallel port (up to 5 devices per port) 1870 Format: <port#>,<pad1>,<pad2>,<pad3>,<pad4>,<pad5> 1871 See also Documentation/input/devices/joystick-parport.rst 1872 1873 gamma= [HW,DRM] 1874 1875 gart_fix_e820= [X86-64,EARLY] disable the fix e820 for K8 GART 1876 Format: off | on 1877 default: on 1878 1879 gather_data_sampling= 1880 [X86,INTEL,EARLY] Control the Gather Data Sampling (GDS) 1881 mitigation. 1882 1883 Gather Data Sampling is a hardware vulnerability which 1884 allows unprivileged speculative access to data which was 1885 previously stored in vector registers. 1886 1887 This issue is mitigated by default in updated microcode. 1888 The mitigation may have a performance impact but can be 1889 disabled. On systems without the microcode mitigation 1890 disabling AVX serves as a mitigation. 1891 1892 force: Disable AVX to mitigate systems without 1893 microcode mitigation. No effect if the microcode 1894 mitigation is present. Known to cause crashes in 1895 userspace with buggy AVX enumeration. 1896 1897 off: Disable GDS mitigation. 1898 1899 gbpages [X86] Use GB pages for kernel direct mappings. 1900 1901 gcov_persist= [GCOV] When non-zero (default), profiling data for 1902 kernel modules is saved and remains accessible via 1903 debugfs, even when the module is unloaded/reloaded. 1904 When zero, profiling data is discarded and associated 1905 debugfs files are removed at module unload time. 1906 1907 goldfish [X86] Enable the goldfish android emulator platform. 1908 Don't use this when you are not running on the 1909 android emulator 1910 1911 gpio-mockup.gpio_mockup_ranges 1912 [HW] Sets the ranges of gpiochip of for this device. 1913 Format: <start1>,<end1>,<start2>,<end2>... 1914 gpio-mockup.gpio_mockup_named_lines 1915 [HW] Let the driver know GPIO lines should be named. 1916 1917 gpt [EFI] Forces disk with valid GPT signature but 1918 invalid Protective MBR to be treated as GPT. If the 1919 primary GPT is corrupted, it enables the backup/alternate 1920 GPT to be used instead. 1921 1922 grcan.enable0= [HW] Configuration of physical interface 0. Determines 1923 the "Enable 0" bit of the configuration register. 1924 Format: 0 | 1 1925 Default: 0 1926 grcan.enable1= [HW] Configuration of physical interface 1. Determines 1927 the "Enable 0" bit of the configuration register. 1928 Format: 0 | 1 1929 Default: 0 1930 grcan.select= [HW] Select which physical interface to use. 1931 Format: 0 | 1 1932 Default: 0 1933 grcan.txsize= [HW] Sets the size of the tx buffer. 1934 Format: <unsigned int> such that (txsize & ~0x1fffc0) == 0. 1935 Default: 1024 1936 grcan.rxsize= [HW] Sets the size of the rx buffer. 1937 Format: <unsigned int> such that (rxsize & ~0x1fffc0) == 0. 1938 Default: 1024 1939 1940 hardened_usercopy= 1941 [KNL] Under CONFIG_HARDENED_USERCOPY, whether 1942 hardening is enabled for this boot. Hardened 1943 usercopy checking is used to protect the kernel 1944 from reading or writing beyond known memory 1945 allocation boundaries as a proactive defense 1946 against bounds-checking flaws in the kernel's 1947 copy_to_user()/copy_from_user() interface. 1948 The default is determined by 1949 CONFIG_HARDENED_USERCOPY_DEFAULT_ON. 1950 on Perform hardened usercopy checks. 1951 off Disable hardened usercopy checks. 1952 1953 hardlockup_all_cpu_backtrace= 1954 [KNL] Should the hard-lockup detector generate 1955 backtraces on all cpus. 1956 Format: 0 | 1 1957 1958 hash_pointers= 1959 [KNL,EARLY] 1960 By default, when pointers are printed to the console 1961 or buffers via the %p format string, that pointer is 1962 "hashed", i.e. obscured by hashing the pointer value. 1963 This is a security feature that hides actual kernel 1964 addresses from unprivileged users, but it also makes 1965 debugging the kernel more difficult since unequal 1966 pointers can no longer be compared. The choices are: 1967 Format: { auto | always | never } 1968 Default: auto 1969 1970 auto - Hash pointers unless slab_debug is enabled. 1971 always - Always hash pointers (even if slab_debug is 1972 enabled). 1973 never - Never hash pointers. This option should only 1974 be specified when debugging the kernel. Do 1975 not use on production kernels. The boot 1976 param "no_hash_pointers" is an alias for 1977 this mode. 1978 1979 For controlling hashing dynamically at runtime, 1980 use the "kernel.kptr_restrict" sysctl instead. 1981 1982 hashdist= [KNL,NUMA] Large hashes allocated during boot 1983 are distributed across NUMA nodes. Defaults on 1984 for 64-bit NUMA, off otherwise. 1985 Format: 0 | 1 (for off | on) 1986 1987 hd= [EIDE] (E)IDE hard drive subsystem geometry 1988 Format: <cyl>,<head>,<sect> 1989 1990 hest_disable [ACPI] 1991 Disable Hardware Error Source Table (HEST) support; 1992 corresponding firmware-first mode error processing 1993 logic will be disabled. 1994 1995 hibernate= [HIBERNATION] 1996 noresume Don't check if there's a hibernation image 1997 present during boot. 1998 nocompress Don't compress/decompress hibernation images. 1999 no Disable hibernation and resume. 2000 protect_image Turn on image protection during restoration 2001 (that will set all pages holding image data 2002 during restoration read-only). 2003 2004 hibernate.compressor= [HIBERNATION] Compression algorithm to be 2005 used with hibernation. 2006 Format: { lzo | lz4 } 2007 Default: lzo 2008 2009 lzo: Select LZO compression algorithm to 2010 compress/decompress hibernation image. 2011 2012 lz4: Select LZ4 compression algorithm to 2013 compress/decompress hibernation image. 2014 2015 hibernate.pm_test_delay= 2016 [HIBERNATION] 2017 Sets the number of seconds to remain in a hibernation test 2018 mode before resuming the system (see 2019 /sys/power/pm_test). Only available when CONFIG_PM_DEBUG 2020 is set. Default value is 5. 2021 2022 hibernate_compression_threads= 2023 [HIBERNATION] 2024 Set the number of threads used for compressing or decompressing 2025 hibernation images. 2026 2027 Format: <integer> 2028 Default: 3 2029 Minimum: 1 2030 Example: hibernate_compression_threads=4 2031 2032 highmem=nn[KMG] [KNL,BOOT,EARLY] forces the highmem zone to have an exact 2033 size of <nn>. This works even on boxes that have no 2034 highmem otherwise. This also works to reduce highmem 2035 size on bigger boxes. 2036 2037 highres= [KNL] Enable/disable high resolution timer mode. 2038 Valid parameters: "on", "off" 2039 Default: "on" 2040 2041 hlt [BUGS=ARM,SH] 2042 2043 hostname= [KNL,EARLY] Set the hostname (aka UTS nodename). 2044 Format: <string> 2045 This allows setting the system's hostname during early 2046 startup. This sets the name returned by gethostname. 2047 Using this parameter to set the hostname makes it 2048 possible to ensure the hostname is correctly set before 2049 any userspace processes run, avoiding the possibility 2050 that a process may call gethostname before the hostname 2051 has been explicitly set, resulting in the calling 2052 process getting an incorrect result. The string must 2053 not exceed the maximum allowed hostname length (usually 2054 64 characters) and will be truncated otherwise. 2055 2056 hpet= [X86-32,HPET] option to control HPET usage 2057 Format: { enable (default) | disable | force | 2058 verbose } 2059 disable: disable HPET and use PIT instead 2060 force: allow force enabled of undocumented chips (ICH4, 2061 VIA, nVidia) 2062 verbose: show contents of HPET registers during setup 2063 2064 hpet_mmap= [X86, HPET_MMAP] Allow userspace to mmap HPET 2065 registers. Default set by CONFIG_HPET_MMAP_DEFAULT. 2066 2067 hugepages= [HW,EARLY] Number of HugeTLB pages to allocate at boot. 2068 If this follows hugepagesz (below), it specifies 2069 the number of pages of hugepagesz to be allocated. 2070 If this is the first HugeTLB parameter on the command 2071 line, it specifies the number of pages to allocate for 2072 the default huge page size. If using node format, the 2073 number of pages to allocate per-node can be specified. 2074 See also Documentation/admin-guide/mm/hugetlbpage.rst. 2075 Format: <integer> or (node format) 2076 <node>:<integer>[,<node>:<integer>] 2077 2078 hugepagesz= 2079 [HW,EARLY] The size of the HugeTLB pages. This is 2080 used in conjunction with hugepages (above) to 2081 allocate huge pages of a specific size at boot. The 2082 pair hugepagesz=X hugepages=Y can be specified once 2083 for each supported huge page size. Huge page sizes 2084 are architecture dependent. See also 2085 Documentation/admin-guide/mm/hugetlbpage.rst. 2086 Format: size[KMG] 2087 2088 hugepage_alloc_threads= 2089 [HW] The number of threads that should be used to 2090 allocate hugepages during boot. This option can be 2091 used to improve system bootup time when allocating 2092 a large amount of huge pages. 2093 The default value is 25% of the available hardware threads. 2094 2095 Note that this parameter only applies to non-gigantic huge pages. 2096 2097 hugetlb_cma= [HW,CMA,EARLY] The size of a CMA area used for allocation 2098 of gigantic hugepages. Or using node format, the size 2099 of a CMA area per node can be specified. 2100 Format: nn[KMGTPE] or (node format) 2101 <node>:nn[KMGTPE][,<node>:nn[KMGTPE]] 2102 2103 Reserve a CMA area of given size and allocate gigantic 2104 hugepages using the CMA allocator. If enabled, the 2105 boot-time allocation of gigantic hugepages is skipped. 2106 2107 hugetlb_cma_only= 2108 [HW,CMA,EARLY] When allocating new HugeTLB pages, only 2109 try to allocate from the CMA areas. 2110 2111 This option does nothing if hugetlb_cma= is not also 2112 specified. 2113 2114 hugetlb_free_vmemmap= 2115 [KNL] Requires CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP 2116 enabled. 2117 Control if HugeTLB Vmemmap Optimization (HVO) is enabled. 2118 Allows heavy hugetlb users to free up some more 2119 memory (7 * PAGE_SIZE for each 2MB hugetlb page). 2120 Format: { on | off (default) } 2121 2122 on: enable HVO 2123 off: disable HVO 2124 2125 Built with CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP_DEFAULT_ON=y, 2126 the default is on. 2127 2128 Note that the vmemmap pages may be allocated from the added 2129 memory block itself when memory_hotplug.memmap_on_memory is 2130 enabled, those vmemmap pages cannot be optimized even if this 2131 feature is enabled. Other vmemmap pages not allocated from 2132 the added memory block itself do not be affected. 2133 2134 hung_task_panic= 2135 [KNL] Number of hung tasks to trigger kernel panic. 2136 Format: <int> 2137 2138 When set to a non-zero value, a kernel panic will be triggered if 2139 the number of detected hung tasks reaches this value. 2140 2141 0: don't panic 2142 1: panic immediately on first hung task 2143 N: panic after N hung tasks are detected in a single scan 2144 2145 The default value is controlled by the 2146 CONFIG_BOOTPARAM_HUNG_TASK_PANIC build-time option. The value 2147 selected by this boot parameter can be changed later by the 2148 kernel.hung_task_panic sysctl. 2149 2150 hvc_iucv= [S390] Number of z/VM IUCV hypervisor console (HVC) 2151 terminal devices. Valid values: 0..8 2152 hvc_iucv_allow= [S390] Comma-separated list of z/VM user IDs. 2153 If specified, z/VM IUCV HVC accepts connections 2154 from listed z/VM user IDs only. 2155 2156 hv_nopvspin [X86,HYPER_V,EARLY] 2157 Disables the paravirt spinlock optimizations 2158 which allow the hypervisor to 'idle' the guest 2159 on lock contention. 2160 2161 hw_protection= [HW] 2162 Format: reboot | shutdown 2163 2164 Hardware protection action taken on critical events like 2165 overtemperature or imminent voltage loss. 2166 2167 i2c_bus= [HW] Override the default board specific I2C bus speed 2168 or register an additional I2C bus that is not 2169 registered from board initialization code. 2170 Format: 2171 <bus_id>,<clkrate> 2172 2173 i2c_touchscreen_props= [HW,ACPI,X86] 2174 Set device-properties for ACPI-enumerated I2C-attached 2175 touchscreen, to e.g. fix coordinates of upside-down 2176 mounted touchscreens. If you need this option please 2177 submit a drivers/platform/x86/touchscreen_dmi.c patch 2178 adding a DMI quirk for this. 2179 2180 Format: 2181 <ACPI_HW_ID>:<prop_name>=<val>[:prop_name=val][:...] 2182 Where <val> is one of: 2183 Omit "=<val>" entirely Set a boolean device-property 2184 Unsigned number Set a u32 device-property 2185 Anything else Set a string device-property 2186 2187 Examples (split over multiple lines): 2188 i2c_touchscreen_props=GDIX1001:touchscreen-inverted-x: 2189 touchscreen-inverted-y 2190 2191 i2c_touchscreen_props=MSSL1680:touchscreen-size-x=1920: 2192 touchscreen-size-y=1080:touchscreen-inverted-y: 2193 firmware-name=gsl1680-vendor-model.fw:silead,home-button 2194 2195 i8042.debug [HW] Toggle i8042 debug mode 2196 i8042.unmask_kbd_data 2197 [HW] Enable printing of interrupt data from the KBD port 2198 (disabled by default, and as a pre-condition 2199 requires that i8042.debug=1 be enabled) 2200 i8042.direct [HW] Put keyboard port into non-translated mode 2201 i8042.dumbkbd [HW] Pretend that controller can only read data from 2202 keyboard and cannot control its state 2203 (Don't attempt to blink the leds) 2204 i8042.noaux [HW] Don't check for auxiliary (== mouse) port 2205 i8042.nokbd [HW] Don't check/create keyboard port 2206 i8042.noloop [HW] Disable the AUX Loopback command while probing 2207 for the AUX port 2208 i8042.nomux [HW] Don't check presence of an active multiplexing 2209 controller 2210 i8042.nopnp [HW] Don't use ACPIPnP / PnPBIOS to discover KBD/AUX 2211 controllers 2212 i8042.notimeout [HW] Ignore timeout condition signalled by controller 2213 i8042.reset [HW] Reset the controller during init, cleanup and 2214 suspend-to-ram transitions, only during s2r 2215 transitions, or never reset 2216 Format: { 1 | Y | y | 0 | N | n } 2217 1, Y, y: always reset controller 2218 0, N, n: don't ever reset controller 2219 Default: only on s2r transitions on x86; most other 2220 architectures force reset to be always executed 2221 i8042.unlock [HW] Unlock (ignore) the keylock 2222 i8042.kbdreset [HW] Reset device connected to KBD port 2223 i8042.probe_defer 2224 [HW] Allow deferred probing upon i8042 probe errors 2225 2226 i810= [HW,DRM] 2227 2228 i915.invert_brightness= 2229 [DRM] Invert the sense of the variable that is used to 2230 set the brightness of the panel backlight. Normally a 2231 brightness value of 0 indicates backlight switched off, 2232 and the maximum of the brightness value sets the backlight 2233 to maximum brightness. If this parameter is set to 0 2234 (default) and the machine requires it, or this parameter 2235 is set to 1, a brightness value of 0 sets the backlight 2236 to maximum brightness, and the maximum of the brightness 2237 value switches the backlight off. 2238 -1 -- never invert brightness 2239 0 -- machine default 2240 1 -- force brightness inversion 2241 2242 ia32_emulation= [X86-64] 2243 Format: <bool> 2244 When true, allows loading 32-bit programs and executing 32-bit 2245 syscalls, essentially overriding IA32_EMULATION_DEFAULT_DISABLED at 2246 boot time. When false, unconditionally disables IA32 emulation. 2247 2248 icn= [HW,ISDN] 2249 Format: <io>[,<membase>[,<icn_id>[,<icn_id2>]]] 2250 2251 2252 idle= [X86,EARLY] 2253 Format: idle=poll, idle=halt, idle=nomwait 2254 2255 idle=poll: Don't do power saving in the idle loop 2256 using HLT, but poll for rescheduling event. This will 2257 make the CPUs eat a lot more power, but may be useful 2258 to get slightly better performance in multiprocessor 2259 benchmarks. It also makes some profiling using 2260 performance counters more accurate. Please note that 2261 on systems with MONITOR/MWAIT support (like Intel 2262 EM64T CPUs) this option has no performance advantage 2263 over the normal idle loop. It may also interact badly 2264 with hyperthreading. 2265 2266 idle=halt: Halt is forced to be used for CPU idle. 2267 In such case C2/C3 won't be used again. 2268 2269 idle=nomwait: Disable mwait for CPU C-states 2270 2271 idxd.sva= [HW] 2272 Format: <bool> 2273 Allow force disabling of Shared Virtual Memory (SVA) 2274 support for the idxd driver. By default it is set to 2275 true (1). 2276 2277 idxd.tc_override= [HW] 2278 Format: <bool> 2279 Allow override of default traffic class configuration 2280 for the device. By default it is set to false (0). 2281 2282 ieee754= [MIPS] Select IEEE Std 754 conformance mode 2283 Format: { strict | legacy | 2008 | relaxed | emulated } 2284 Default: strict 2285 2286 Choose which programs will be accepted for execution 2287 based on the IEEE 754 NaN encoding(s) supported by 2288 the FPU and the NaN encoding requested with the value 2289 of an ELF file header flag individually set by each 2290 binary. Hardware implementations are permitted to 2291 support either or both of the legacy and the 2008 NaN 2292 encoding mode. 2293 2294 Available settings are as follows: 2295 strict accept binaries that request a NaN encoding 2296 supported by the FPU 2297 legacy only accept legacy-NaN binaries, if supported 2298 by the FPU 2299 2008 only accept 2008-NaN binaries, if supported 2300 by the FPU 2301 relaxed accept any binaries regardless of whether 2302 supported by the FPU 2303 emulated accept any binaries but enable FPU emulator 2304 if binary mode is unsupported by the FPU. 2305 2306 The FPU emulator is always able to support both NaN 2307 encodings, so if no FPU hardware is present or it has 2308 been disabled with 'nofpu', then the settings of 2309 'legacy' and '2008' strap the emulator accordingly, 2310 'relaxed' straps the emulator for both legacy-NaN and 2311 2008-NaN, whereas 'strict' enables legacy-NaN only on 2312 legacy processors and both NaN encodings on MIPS32 or 2313 MIPS64 CPUs. 2314 2315 The setting for ABS.fmt/NEG.fmt instruction execution 2316 mode generally follows that for the NaN encoding, 2317 except where unsupported by hardware. 2318 2319 ignore_loglevel [KNL,EARLY] 2320 Ignore loglevel setting - this will print /all/ 2321 kernel messages to the console. Useful for debugging. 2322 We also add it as printk module parameter, so users 2323 could change it dynamically, usually by 2324 /sys/module/printk/parameters/ignore_loglevel. 2325 2326 ignore_rlimit_data 2327 Ignore RLIMIT_DATA setting for data mappings, 2328 print warning at first misuse. Can be changed via 2329 /sys/module/kernel/parameters/ignore_rlimit_data. 2330 2331 ihash_entries= [KNL] 2332 Set number of hash buckets for inode cache. 2333 2334 ima_appraise= [IMA] appraise integrity measurements 2335 Format: { "off" | "enforce" | "fix" | "log" } 2336 default: "enforce" 2337 2338 ima_appraise_tcb [IMA] Deprecated. Use ima_policy= instead. 2339 The builtin appraise policy appraises all files 2340 owned by uid=0. 2341 2342 ima_canonical_fmt [IMA] 2343 Use the canonical format for the binary runtime 2344 measurements, instead of host native format. 2345 2346 ima_hash= [IMA] 2347 Format: { md5 | sha1 | rmd160 | sha256 | sha384 2348 | sha512 | ... } 2349 default: "sha1" 2350 2351 The list of supported hash algorithms is defined 2352 in crypto/hash_info.h. 2353 2354 ima_policy= [IMA] 2355 The builtin policies to load during IMA setup. 2356 Format: "tcb | appraise_tcb | secure_boot | 2357 fail_securely | critical_data" 2358 2359 The "tcb" policy measures all programs exec'd, files 2360 mmap'd for exec, and all files opened with the read 2361 mode bit set by either the effective uid (euid=0) or 2362 uid=0. 2363 2364 The "appraise_tcb" policy appraises the integrity of 2365 all files owned by root. 2366 2367 The "secure_boot" policy appraises the integrity 2368 of files (eg. kexec kernel image, kernel modules, 2369 firmware, policy, etc) based on file signatures. 2370 2371 The "fail_securely" policy forces file signature 2372 verification failure also on privileged mounted 2373 filesystems with the SB_I_UNVERIFIABLE_SIGNATURE 2374 flag. 2375 2376 The "critical_data" policy measures kernel integrity 2377 critical data. 2378 2379 ima_tcb [IMA] Deprecated. Use ima_policy= instead. 2380 Load a policy which meets the needs of the Trusted 2381 Computing Base. This means IMA will measure all 2382 programs exec'd, files mmap'd for exec, and all files 2383 opened for read by uid=0. 2384 2385 ima_template= [IMA] 2386 Select one of defined IMA measurements template formats. 2387 Formats: { "ima" | "ima-ng" | "ima-ngv2" | "ima-sig" | 2388 "ima-sigv2" } 2389 Default: "ima-ng" 2390 2391 ima_template_fmt= 2392 [IMA] Define a custom template format. 2393 Format: { "field1|...|fieldN" } 2394 2395 ima.ahash_minsize= [IMA] Minimum file size for asynchronous hash usage 2396 Format: <min_file_size> 2397 Set the minimal file size for using asynchronous hash. 2398 If left unspecified, ahash usage is disabled. 2399 2400 ahash performance varies for different data sizes on 2401 different crypto accelerators. This option can be used 2402 to achieve the best performance for a particular HW. 2403 2404 ima.ahash_bufsize= [IMA] Asynchronous hash buffer size 2405 Format: <bufsize> 2406 Set hashing buffer size. Default: 4k. 2407 2408 ahash performance varies for different chunk sizes on 2409 different crypto accelerators. This option can be used 2410 to achieve best performance for particular HW. 2411 2412 ima= [IMA] Enable or disable IMA 2413 Format: { "off" | "on" } 2414 Default: "on" 2415 Note that disabling IMA is limited to kdump kernel. 2416 2417 indirect_target_selection= [X86,Intel] Mitigation control for Indirect 2418 Target Selection(ITS) bug in Intel CPUs. Updated 2419 microcode is also required for a fix in IBPB. 2420 2421 on: Enable mitigation (default). 2422 off: Disable mitigation. 2423 force: Force the ITS bug and deploy default 2424 mitigation. 2425 vmexit: Only deploy mitigation if CPU is affected by 2426 guest/host isolation part of ITS. 2427 stuff: Deploy RSB-fill mitigation when retpoline is 2428 also deployed. Otherwise, deploy the default 2429 mitigation. 2430 2431 For details see: 2432 Documentation/admin-guide/hw-vuln/indirect-target-selection.rst 2433 2434 init= [KNL] 2435 Format: <full_path> 2436 Run specified binary instead of /sbin/init as init 2437 process. 2438 2439 initcall_debug [KNL] Trace initcalls as they are executed. Useful 2440 for working out where the kernel is dying during 2441 startup. 2442 2443 initcall_blacklist= [KNL] Do not execute a comma-separated list of 2444 initcall functions. Useful for debugging built-in 2445 modules and initcalls. 2446 2447 initramfs_async= [KNL] 2448 Format: <bool> 2449 Default: 1 2450 This parameter controls whether the initramfs 2451 image is unpacked asynchronously, concurrently 2452 with devices being probed and 2453 initialized. This should normally just work, 2454 but as a debugging aid, one can get the 2455 historical behaviour of the initramfs 2456 unpacking being completed before device_ and 2457 late_ initcalls. 2458 2459 initrd= [BOOT,EARLY] Specify the location of the initial ramdisk 2460 2461 initrdmem= [KNL,EARLY] Specify a physical address and size from which to 2462 load the initrd. If an initrd is compiled in or 2463 specified in the bootparams, it takes priority over this 2464 setting. 2465 Format: ss[KMG],nn[KMG] 2466 Default is 0, 0 2467 2468 init_on_alloc= [MM,EARLY] Fill newly allocated pages and heap objects with 2469 zeroes. 2470 Format: 0 | 1 2471 Default set by CONFIG_INIT_ON_ALLOC_DEFAULT_ON. 2472 2473 init_on_free= [MM,EARLY] Fill freed pages and heap objects with zeroes. 2474 Format: 0 | 1 2475 Default set by CONFIG_INIT_ON_FREE_DEFAULT_ON. 2476 2477 init_pkru= [X86] Specify the default memory protection keys rights 2478 register contents for all processes. 0x55555554 by 2479 default (disallow access to all but pkey 0). Can 2480 override in debugfs after boot. 2481 2482 inport.irq= [HW] Inport (ATI XL and Microsoft) busmouse driver 2483 Format: <irq> 2484 2485 int_pln_enable [X86] Enable power limit notification interrupt 2486 2487 integrity_audit=[IMA] 2488 Format: { "0" | "1" } 2489 0 -- basic integrity auditing messages. (Default) 2490 1 -- additional integrity auditing messages. 2491 2492 intel_iommu= [DMAR] Intel IOMMU driver (DMAR) option 2493 on 2494 Enable intel iommu driver. 2495 off 2496 Disable intel iommu driver. 2497 igfx_off [Default Off] 2498 By default, gfx is mapped as normal device. If a gfx 2499 device has a dedicated DMAR unit, the DMAR unit is 2500 bypassed by not enabling DMAR with this option. In 2501 this case, gfx device will use physical address for 2502 DMA. 2503 strict [Default Off] 2504 Deprecated, equivalent to iommu.strict=1. 2505 sp_off [Default Off] 2506 By default, super page will be supported if Intel IOMMU 2507 has the capability. With this option, super page will 2508 not be supported. 2509 sm_on 2510 Enable the Intel IOMMU scalable mode if the hardware 2511 advertises that it has support for the scalable mode 2512 translation. 2513 sm_off 2514 Disallow use of the Intel IOMMU scalable mode. 2515 tboot_noforce [Default Off] 2516 Do not force the Intel IOMMU enabled under tboot. 2517 By default, tboot will force Intel IOMMU on, which 2518 could harm performance of some high-throughput 2519 devices like 40GBit network cards, even if identity 2520 mapping is enabled. 2521 Note that using this option lowers the security 2522 provided by tboot because it makes the system 2523 vulnerable to DMA attacks. 2524 2525 intel_idle.max_cstate= [KNL,HW,ACPI,X86] 2526 0 disables intel_idle and fall back on acpi_idle. 2527 1 to 9 specify maximum depth of C-state. 2528 2529 intel_pstate= [X86,EARLY] 2530 disable 2531 Do not enable intel_pstate as the default 2532 scaling driver for the supported processors 2533 active 2534 Use intel_pstate driver to bypass the scaling 2535 governors layer of cpufreq and provides it own 2536 algorithms for p-state selection. There are two 2537 P-state selection algorithms provided by 2538 intel_pstate in the active mode: powersave and 2539 performance. The way they both operate depends 2540 on whether or not the hardware managed P-states 2541 (HWP) feature has been enabled in the processor 2542 and possibly on the processor model. 2543 passive 2544 Use intel_pstate as a scaling driver, but configure it 2545 to work with generic cpufreq governors (instead of 2546 enabling its internal governor). This mode cannot be 2547 used along with the hardware-managed P-states (HWP) 2548 feature. 2549 force 2550 Enable intel_pstate on systems that prohibit it by default 2551 in favor of acpi-cpufreq. Forcing the intel_pstate driver 2552 instead of acpi-cpufreq may disable platform features, such 2553 as thermal controls and power capping, that rely on ACPI 2554 P-States information being indicated to OSPM and therefore 2555 should be used with caution. This option does not work with 2556 processors that aren't supported by the intel_pstate driver 2557 or on platforms that use pcc-cpufreq instead of acpi-cpufreq. 2558 no_hwp 2559 Do not enable hardware P state control (HWP) 2560 if available. 2561 hwp_only 2562 Only load intel_pstate on systems which support 2563 hardware P state control (HWP) if available. 2564 support_acpi_ppc 2565 Enforce ACPI _PPC performance limits. If the Fixed ACPI 2566 Description Table, specifies preferred power management 2567 profile as "Enterprise Server" or "Performance Server", 2568 then this feature is turned on by default. 2569 per_cpu_perf_limits 2570 Allow per-logical-CPU P-State performance control limits using 2571 cpufreq sysfs interface 2572 no_cas 2573 Do not enable capacity-aware scheduling (CAS) on 2574 hybrid systems 2575 2576 intremap= [X86-64,Intel-IOMMU,EARLY] 2577 on enable Interrupt Remapping (default) 2578 off disable Interrupt Remapping 2579 nosid disable Source ID checking 2580 no_x2apic_optout 2581 BIOS x2APIC opt-out request will be ignored 2582 nopost disable Interrupt Posting 2583 posted_msi 2584 enable MSIs delivered as posted interrupts 2585 2586 iomem= Disable strict checking of access to MMIO memory 2587 strict regions from userspace. 2588 relaxed 2589 2590 iommu= [X86,EARLY] 2591 2592 off 2593 Don't initialize and use any kind of IOMMU. 2594 2595 force 2596 Force the use of the hardware IOMMU even when 2597 it is not actually needed (e.g. because < 3 GB 2598 memory). 2599 2600 noforce 2601 Don't force hardware IOMMU usage when it is not 2602 needed. (default). 2603 2604 biomerge 2605 panic 2606 nopanic 2607 merge 2608 nomerge 2609 2610 soft 2611 Use software bounce buffering (SWIOTLB) (default for 2612 Intel machines). This can be used to prevent the usage 2613 of an available hardware IOMMU. 2614 2615 pt 2616 nopt 2617 nobypass [PPC/POWERNV] 2618 Disable IOMMU bypass, using IOMMU for PCI devices. 2619 2620 AMD Gart HW IOMMU-specific options: 2621 2622 <size> 2623 Set the size of the remapping area in bytes. 2624 2625 allowed 2626 Overwrite iommu off workarounds for specific chipsets 2627 2628 fullflush 2629 Flush IOMMU on each allocation (default). 2630 2631 nofullflush 2632 Don't use IOMMU fullflush. 2633 2634 memaper[=<order>] 2635 Allocate an own aperture over RAM with size 2636 32MB<<order. (default: order=1, i.e. 64MB) 2637 2638 merge 2639 Do scatter-gather (SG) merging. Implies "force" 2640 (experimental). 2641 2642 nomerge 2643 Don't do scatter-gather (SG) merging. 2644 2645 noaperture 2646 Ask the IOMMU not to touch the aperture for AGP. 2647 2648 noagp 2649 Don't initialize the AGP driver and use full aperture. 2650 2651 panic 2652 Always panic when IOMMU overflows. 2653 2654 iommu.forcedac= [ARM64,X86,EARLY] Control IOVA allocation for PCI devices. 2655 Format: { "0" | "1" } 2656 0 - Try to allocate a 32-bit DMA address first, before 2657 falling back to the full range if needed. 2658 1 - Allocate directly from the full usable range, 2659 forcing Dual Address Cycle for PCI cards supporting 2660 greater than 32-bit addressing. 2661 2662 iommu.strict= [ARM64,X86,S390,EARLY] Configure TLB invalidation behaviour 2663 Format: { "0" | "1" } 2664 0 - Lazy mode. 2665 Request that DMA unmap operations use deferred 2666 invalidation of hardware TLBs, for increased 2667 throughput at the cost of reduced device isolation. 2668 Will fall back to strict mode if not supported by 2669 the relevant IOMMU driver. 2670 1 - Strict mode. 2671 DMA unmap operations invalidate IOMMU hardware TLBs 2672 synchronously. 2673 unset - Use value of CONFIG_IOMMU_DEFAULT_DMA_{LAZY,STRICT}. 2674 Note: on x86, strict mode specified via one of the 2675 legacy driver-specific options takes precedence. 2676 2677 iommu.passthrough= 2678 [ARM64,X86,EARLY] Configure DMA to bypass the IOMMU by default. 2679 Format: { "0" | "1" } 2680 0 - Use IOMMU translation for DMA. 2681 1 - Bypass the IOMMU for DMA. 2682 unset - Use value of CONFIG_IOMMU_DEFAULT_PASSTHROUGH. 2683 2684 iommu.debug_pagealloc= 2685 [KNL,EARLY] When CONFIG_IOMMU_DEBUG_PAGEALLOC is set, this 2686 parameter enables the feature at boot time. By default, it 2687 is disabled and the system behaves the same way as a kernel 2688 built without CONFIG_IOMMU_DEBUG_PAGEALLOC. 2689 Format: { "0" | "1" } 2690 0 - Sanitizer disabled. 2691 1 - Sanitizer enabled, expect runtime overhead. 2692 2693 io7= [HW] IO7 for Marvel-based Alpha systems 2694 See comment before marvel_specify_io7 in 2695 arch/alpha/kernel/core_marvel.c. 2696 2697 io_delay= [X86,EARLY] I/O delay method 2698 0x80 2699 Standard port 0x80 based delay 2700 0xed 2701 Alternate port 0xed based delay (needed on some systems) 2702 udelay 2703 Simple two microseconds delay 2704 none 2705 No delay 2706 2707 ip= [IP_PNP] 2708 See Documentation/admin-guide/nfs/nfsroot.rst. 2709 2710 ipcmni_extend [KNL,EARLY] Extend the maximum number of unique System V 2711 IPC identifiers from 32,768 to 16,777,216. 2712 2713 ipe.enforce= [IPE] 2714 Format: <bool> 2715 Determine whether IPE starts in permissive (0) or 2716 enforce (1) mode. The default is enforce. 2717 2718 ipe.success_audit= 2719 [IPE] 2720 Format: <bool> 2721 Start IPE with success auditing enabled, emitting 2722 an audit event when a binary is allowed. The default 2723 is 0. 2724 2725 irqaffinity= [SMP] Set the default irq affinity mask 2726 The argument is a cpu list, as described above. 2727 2728 irqchip.gicv2_force_probe= 2729 [ARM,ARM64,EARLY] 2730 Format: <bool> 2731 Force the kernel to look for the second 4kB page 2732 of a GICv2 controller even if the memory range 2733 exposed by the device tree is too small. 2734 2735 irqchip.gicv3_nolpi= 2736 [ARM,ARM64,EARLY] 2737 Force the kernel to ignore the availability of 2738 LPIs (and by consequence ITSs). Intended for system 2739 that use the kernel as a bootloader, and thus want 2740 to let secondary kernels in charge of setting up 2741 LPIs. 2742 2743 irqchip.gicv3_pseudo_nmi= [ARM64,EARLY] 2744 Enables support for pseudo-NMIs in the kernel. This 2745 requires the kernel to be built with 2746 CONFIG_ARM64_PSEUDO_NMI. 2747 2748 irqchip.riscv_imsic_noipi 2749 [RISC-V,EARLY] 2750 Force the kernel to not use IMSIC software injected MSIs 2751 as IPIs. Intended for system where IMSIC is trap-n-emulated, 2752 and thus want to reduce MMIO traps when triggering IPIs 2753 to multiple harts. 2754 2755 irqfixup [HW] 2756 When an interrupt is not handled search all handlers 2757 for it. Intended to get systems with badly broken 2758 firmware running. 2759 2760 irqhandler.duration_warn_us= [KNL] 2761 Warn if an IRQ handler exceeds the specified duration 2762 threshold in microseconds. Useful for identifying 2763 long-running IRQs in the system. 2764 2765 irqpoll [HW] 2766 When an interrupt is not handled search all handlers 2767 for it. Also check all handlers each timer 2768 interrupt. Intended to get systems with badly broken 2769 firmware running. 2770 2771 isapnp= [ISAPNP] 2772 Format: <RDP>,<reset>,<pci_scan>,<verbosity> 2773 2774 isolcpus= [KNL,SMP,ISOL] Isolate a given set of CPUs from disturbance. 2775 [Deprecated - use cpusets instead] 2776 Format: [flag-list,]<cpu-list> 2777 2778 Specify one or more CPUs to isolate from disturbances 2779 specified in the flag list (default: domain): 2780 2781 nohz 2782 Disable the tick when a single task runs as well as 2783 disabling other kernel noises like having RCU callbacks 2784 offloaded. This is equivalent to the nohz_full parameter. 2785 2786 A residual 1Hz tick is offloaded to workqueues, which you 2787 need to affine to housekeeping through the global 2788 workqueue's affinity configured via the 2789 /sys/devices/virtual/workqueue/cpumask sysfs file, or 2790 by using the 'domain' flag described below. 2791 2792 NOTE: by default the global workqueue runs on all CPUs, 2793 so to protect individual CPUs the 'cpumask' file has to 2794 be configured manually after bootup. 2795 2796 domain 2797 Isolate from the general SMP balancing and scheduling 2798 algorithms. Note that performing domain isolation this way 2799 is irreversible: it's not possible to bring back a CPU to 2800 the domains once isolated through isolcpus. It's strongly 2801 advised to use cpusets instead to disable scheduler load 2802 balancing through the "cpuset.sched_load_balance" file. 2803 It offers a much more flexible interface where CPUs can 2804 move in and out of an isolated set anytime. 2805 2806 You can move a process onto or off an "isolated" CPU via 2807 the CPU affinity syscalls or cpuset. 2808 <cpu number> begins at 0 and the maximum value is 2809 "number of CPUs in system - 1". 2810 2811 managed_irq 2812 2813 Isolate from being targeted by managed interrupts 2814 which have an interrupt mask containing isolated 2815 CPUs. The affinity of managed interrupts is 2816 handled by the kernel and cannot be changed via 2817 the /proc/irq/* interfaces. 2818 2819 This isolation is best effort and only effective 2820 if the automatically assigned interrupt mask of a 2821 device queue contains isolated and housekeeping 2822 CPUs. If housekeeping CPUs are online then such 2823 interrupts are directed to the housekeeping CPU 2824 so that IO submitted on the housekeeping CPU 2825 cannot disturb the isolated CPU. 2826 2827 If a queue's affinity mask contains only isolated 2828 CPUs then this parameter has no effect on the 2829 interrupt routing decision, though interrupts are 2830 only delivered when tasks running on those 2831 isolated CPUs submit IO. IO submitted on 2832 housekeeping CPUs has no influence on those 2833 queues. 2834 2835 The format of <cpu-list> is described above. 2836 2837 iucv= [HW,NET] 2838 2839 ivrs_ioapic [HW,X86-64] 2840 Provide an override to the IOAPIC-ID<->DEVICE-ID 2841 mapping provided in the IVRS ACPI table. 2842 By default, PCI segment is 0, and can be omitted. 2843 2844 For example, to map IOAPIC-ID decimal 10 to 2845 PCI segment 0x1 and PCI device 00:14.0, 2846 write the parameter as: 2847 ivrs_ioapic=10@0001:00:14.0 2848 2849 Deprecated formats: 2850 * To map IOAPIC-ID decimal 10 to PCI device 00:14.0 2851 write the parameter as: 2852 ivrs_ioapic[10]=00:14.0 2853 * To map IOAPIC-ID decimal 10 to PCI segment 0x1 and 2854 PCI device 00:14.0 write the parameter as: 2855 ivrs_ioapic[10]=0001:00:14.0 2856 2857 ivrs_hpet [HW,X86-64] 2858 Provide an override to the HPET-ID<->DEVICE-ID 2859 mapping provided in the IVRS ACPI table. 2860 By default, PCI segment is 0, and can be omitted. 2861 2862 For example, to map HPET-ID decimal 10 to 2863 PCI segment 0x1 and PCI device 00:14.0, 2864 write the parameter as: 2865 ivrs_hpet=10@0001:00:14.0 2866 2867 Deprecated formats: 2868 * To map HPET-ID decimal 0 to PCI device 00:14.0 2869 write the parameter as: 2870 ivrs_hpet[0]=00:14.0 2871 * To map HPET-ID decimal 10 to PCI segment 0x1 and 2872 PCI device 00:14.0 write the parameter as: 2873 ivrs_ioapic[10]=0001:00:14.0 2874 2875 ivrs_acpihid [HW,X86-64] 2876 Provide an override to the ACPI-HID:UID<->DEVICE-ID 2877 mapping provided in the IVRS ACPI table. 2878 By default, PCI segment is 0, and can be omitted. 2879 2880 For example, to map UART-HID:UID AMD0020:0 to 2881 PCI segment 0x1 and PCI device ID 00:14.5, 2882 write the parameter as: 2883 ivrs_acpihid=AMD0020:0@0001:00:14.5 2884 2885 Deprecated formats: 2886 * To map UART-HID:UID AMD0020:0 to PCI segment is 0, 2887 PCI device ID 00:14.5, write the parameter as: 2888 ivrs_acpihid[00:14.5]=AMD0020:0 2889 * To map UART-HID:UID AMD0020:0 to PCI segment 0x1 and 2890 PCI device ID 00:14.5, write the parameter as: 2891 ivrs_acpihid[0001:00:14.5]=AMD0020:0 2892 2893 js= [HW,JOY] Analog joystick 2894 See Documentation/input/joydev/joystick.rst. 2895 2896 kasan_multi_shot 2897 [KNL] Enforce KASAN (Kernel Address Sanitizer) to print 2898 report on every invalid memory access. Without this 2899 parameter KASAN will print report only for the first 2900 invalid access. 2901 2902 keep_bootcon [KNL,EARLY] 2903 Do not unregister boot console at start. This is only 2904 useful for debugging when something happens in the window 2905 between unregistering the boot console and initializing 2906 the real console. 2907 2908 keepinitrd [HW,ARM] See retain_initrd. 2909 2910 kernelcore= [KNL,X86,PPC,EARLY] 2911 Format: nn[KMGTPE] | nn% | "mirror" 2912 This parameter specifies the amount of memory usable by 2913 the kernel for non-movable allocations. The requested 2914 amount is spread evenly throughout all nodes in the 2915 system as ZONE_NORMAL. The remaining memory is used for 2916 movable memory in its own zone, ZONE_MOVABLE. In the 2917 event, a node is too small to have both ZONE_NORMAL and 2918 ZONE_MOVABLE, kernelcore memory will take priority and 2919 other nodes will have a larger ZONE_MOVABLE. 2920 2921 ZONE_MOVABLE is used for the allocation of pages that 2922 may be reclaimed or moved by the page migration 2923 subsystem. Note that allocations like PTEs-from-HighMem 2924 still use the HighMem zone if it exists, and the Normal 2925 zone if it does not. 2926 2927 It is possible to specify the exact amount of memory in 2928 the form of "nn[KMGTPE]", a percentage of total system 2929 memory in the form of "nn%", or "mirror". If "mirror" 2930 option is specified, mirrored (reliable) memory is used 2931 for non-movable allocations and remaining memory is used 2932 for Movable pages. "nn[KMGTPE]", "nn%", and "mirror" 2933 are exclusive, so you cannot specify multiple forms. 2934 2935 kfence.burst= [MM,KFENCE] The number of additional successive 2936 allocations to be attempted through KFENCE for each 2937 sample interval. 2938 Format: <unsigned integer> 2939 Default: 0 2940 2941 kfence.check_on_panic= 2942 [MM,KFENCE] Whether to check all KFENCE-managed objects' 2943 canaries on panic. 2944 Format: <bool> 2945 Default: false 2946 2947 kfence.deferrable= 2948 [MM,KFENCE] Whether to use a deferrable timer to trigger 2949 allocations. This avoids forcing CPU wake-ups if the 2950 system is idle, at the risk of a less predictable 2951 sample interval. 2952 Format: <bool> 2953 Default: CONFIG_KFENCE_DEFERRABLE 2954 2955 kfence.sample_interval= 2956 [MM,KFENCE] KFENCE's sample interval in milliseconds. 2957 Format: <unsigned integer> 2958 0 - Disable KFENCE. 2959 >0 - Enabled KFENCE with given sample interval. 2960 Default: CONFIG_KFENCE_SAMPLE_INTERVAL 2961 2962 kfence.skip_covered_thresh= 2963 [MM,KFENCE] If pool utilization reaches this threshold 2964 (pool usage%), KFENCE limits currently covered 2965 allocations of the same source from further filling 2966 up the pool. 2967 Format: <unsigned integer> 2968 Default: 75 2969 2970 kgdbdbgp= [KGDB,HW,EARLY] kgdb over EHCI usb debug port. 2971 Format: <Controller#>[,poll interval] 2972 The controller # is the number of the ehci usb debug 2973 port as it is probed via PCI. The poll interval is 2974 optional and is the number seconds in between 2975 each poll cycle to the debug port in case you need 2976 the functionality for interrupting the kernel with 2977 gdb or control-c on the dbgp connection. When 2978 not using this parameter you use sysrq-g to break into 2979 the kernel debugger. 2980 2981 kgdboc= [KGDB,HW] kgdb over consoles. 2982 Requires a tty driver that supports console polling, 2983 or a supported polling keyboard driver (non-usb). 2984 Serial only format: <serial_device>[,baud] 2985 keyboard only format: kbd 2986 keyboard and serial format: kbd,<serial_device>[,baud] 2987 Optional Kernel mode setting: 2988 kms, kbd format: kms,kbd 2989 kms, kbd and serial format: kms,kbd,<ser_dev>[,baud] 2990 2991 kgdboc_earlycon= [KGDB,HW,EARLY] 2992 If the boot console provides the ability to read 2993 characters and can work in polling mode, you can use 2994 this parameter to tell kgdb to use it as a backend 2995 until the normal console is registered. Intended to 2996 be used together with the kgdboc parameter which 2997 specifies the normal console to transition to. 2998 2999 The name of the early console should be specified 3000 as the value of this parameter. Note that the name of 3001 the early console might be different than the tty 3002 name passed to kgdboc. It's OK to leave the value 3003 blank and the first boot console that implements 3004 read() will be picked. 3005 3006 kgdbwait [KGDB,EARLY] Stop kernel execution and enter the 3007 kernel debugger at the earliest opportunity. 3008 3009 kho= [KEXEC,EARLY] 3010 Format: { "0" | "1" | "off" | "on" | "y" | "n" } 3011 Enables or disables Kexec HandOver. 3012 "0" | "off" | "n" - kexec handover is disabled 3013 "1" | "on" | "y" - kexec handover is enabled 3014 3015 kho_scratch= [KEXEC,EARLY] 3016 Format: ll[KMG],mm[KMG],nn[KMG] | nn% 3017 Defines the size of the KHO scratch region. The KHO 3018 scratch regions are physically contiguous memory 3019 ranges that can only be used for non-kernel 3020 allocations. That way, even when memory is heavily 3021 fragmented with handed over memory, the kexeced 3022 kernel will always have enough contiguous ranges to 3023 bootstrap itself. 3024 3025 It is possible to specify the exact amount of 3026 memory in the form of "ll[KMG],mm[KMG],nn[KMG]" 3027 where the first parameter defines the size of a low 3028 memory scratch area, the second parameter defines 3029 the size of a global scratch area and the third 3030 parameter defines the size of additional per-node 3031 scratch areas. The form "nn%" defines scale factor 3032 (in percents) of memory that was used during boot. 3033 3034 kmac= [MIPS] Korina ethernet MAC address. 3035 Configure the RouterBoard 532 series on-chip 3036 Ethernet adapter MAC address. 3037 3038 kmemleak= [KNL,EARLY] Boot-time kmemleak enable/disable 3039 Valid arguments: on, off 3040 Default: on 3041 Built with CONFIG_DEBUG_KMEMLEAK_DEFAULT_OFF=y, 3042 the default is off. 3043 3044 kprobe_event=[probe-list] 3045 [FTRACE] Add kprobe events and enable at boot time. 3046 The probe-list is a semicolon delimited list of probe 3047 definitions. Each definition is same as kprobe_events 3048 interface, but the parameters are comma delimited. 3049 For example, to add a kprobe event on vfs_read with 3050 arg1 and arg2, add to the command line; 3051 3052 kprobe_event=p,vfs_read,$arg1,$arg2 3053 3054 See also Documentation/trace/kprobetrace.rst "Kernel 3055 Boot Parameter" section. 3056 3057 kpti= [ARM64,EARLY] Control page table isolation of 3058 user and kernel address spaces. 3059 Default: enabled on cores which need mitigation. 3060 0: force disabled 3061 1: force enabled 3062 3063 kunit.enable= [KUNIT] Enable executing KUnit tests. Requires 3064 CONFIG_KUNIT to be set to be fully enabled. The 3065 default value can be overridden via 3066 KUNIT_DEFAULT_ENABLED. 3067 Default is 1 (enabled) 3068 3069 kvm.ignore_msrs=[KVM] Ignore guest accesses to unhandled MSRs. 3070 Default is 0 (don't ignore, but inject #GP) 3071 3072 kvm.eager_page_split= 3073 [KVM,X86] Controls whether or not KVM will try to 3074 proactively split all huge pages during dirty logging. 3075 Eager page splitting reduces interruptions to vCPU 3076 execution by eliminating the write-protection faults 3077 and MMU lock contention that would otherwise be 3078 required to split huge pages lazily. 3079 3080 VM workloads that rarely perform writes or that write 3081 only to a small region of VM memory may benefit from 3082 disabling eager page splitting to allow huge pages to 3083 still be used for reads. 3084 3085 The behavior of eager page splitting depends on whether 3086 KVM_DIRTY_LOG_INITIALLY_SET is enabled or disabled. If 3087 disabled, all huge pages in a memslot will be eagerly 3088 split when dirty logging is enabled on that memslot. If 3089 enabled, eager page splitting will be performed during 3090 the KVM_CLEAR_DIRTY ioctl, and only for the pages being 3091 cleared. 3092 3093 Eager page splitting is only supported when kvm.tdp_mmu=Y. 3094 3095 Default is Y (on). 3096 3097 kvm.enable_pmu=[KVM,X86] 3098 If enabled, KVM will virtualize PMU functionality based 3099 on the virtual CPU model defined by userspace. This 3100 can be overridden on a per-VM basis via 3101 KVM_CAP_PMU_CAPABILITY. 3102 3103 If disabled, KVM will not virtualize PMU functionality, 3104 e.g. MSRs, PMCs, PMIs, etc., even if userspace defines 3105 a virtual CPU model that contains PMU assets. 3106 3107 Note, KVM's vPMU support implicitly requires running 3108 with an in-kernel local APIC, e.g. to deliver PMIs to 3109 the guest. Running without an in-kernel local APIC is 3110 not supported, though KVM will allow such a combination 3111 (with severely degraded functionality). 3112 3113 See also enable_mediated_pmu. 3114 3115 Default is Y (on). 3116 3117 kvm.enable_virt_at_load=[KVM,ARM64,LOONGARCH,MIPS,RISCV,X86] 3118 If enabled, KVM will enable virtualization in hardware 3119 when KVM is loaded, and disable virtualization when KVM 3120 is unloaded (if KVM is built as a module). 3121 3122 If disabled, KVM will dynamically enable and disable 3123 virtualization on-demand when creating and destroying 3124 VMs, i.e. on the 0=>1 and 1=>0 transitions of the 3125 number of VMs. 3126 3127 Enabling virtualization at module load avoids potential 3128 latency for creation of the 0=>1 VM, as KVM serializes 3129 virtualization enabling across all online CPUs. The 3130 "cost" of enabling virtualization when KVM is loaded, 3131 is that doing so may interfere with using out-of-tree 3132 hypervisors that want to "own" virtualization hardware. 3133 3134 kvm.enable_vmware_backdoor=[KVM] Support VMware backdoor PV interface. 3135 Default is false (don't support). 3136 3137 kvm.nx_huge_pages= 3138 [KVM] Controls the software workaround for the 3139 X86_BUG_ITLB_MULTIHIT bug. 3140 force : Always deploy workaround. 3141 off : Never deploy workaround. 3142 auto : Deploy workaround based on the presence of 3143 X86_BUG_ITLB_MULTIHIT. 3144 3145 Default is 'auto'. 3146 3147 If the software workaround is enabled for the host, 3148 guests do need not to enable it for nested guests. 3149 3150 kvm.nx_huge_pages_recovery_ratio= 3151 [KVM] Controls how many 4KiB pages are periodically zapped 3152 back to huge pages. 0 disables the recovery, otherwise if 3153 the value is N KVM will zap 1/Nth of the 4KiB pages every 3154 period (see below). The default is 60. 3155 3156 kvm.nx_huge_pages_recovery_period_ms= 3157 [KVM] Controls the time period at which KVM zaps 4KiB pages 3158 back to huge pages. If the value is a non-zero N, KVM will 3159 zap a portion (see ratio above) of the pages every N msecs. 3160 If the value is 0 (the default), KVM will pick a period based 3161 on the ratio, such that a page is zapped after 1 hour on average. 3162 3163 kvm-{amd,intel}.enable_mediated_pmu=[KVM,AMD,INTEL] 3164 If enabled, KVM will provide a mediated virtual PMU, 3165 instead of the default perf-based virtual PMU (if 3166 kvm.enable_pmu is true and PMU is enumerated via the 3167 virtual CPU model). 3168 3169 With a perf-based vPMU, KVM operates as a user of perf, 3170 i.e. emulates guest PMU counters using perf events. 3171 KVM-created perf events are managed by perf as regular 3172 (guest-only) events, e.g. are scheduled in/out, contend 3173 for hardware resources, etc. Using a perf-based vPMU 3174 allows guest and host usage of the PMU to co-exist, but 3175 incurs non-trivial overhead and can result in silently 3176 dropped guest events (due to resource contention). 3177 3178 With a mediated vPMU, hardware PMU state is context 3179 switched around the world switch to/from the guest. 3180 KVM mediates which events the guest can utilize, but 3181 gives the guest direct access to all other PMU assets 3182 when possible (KVM may intercept some accesses if the 3183 virtual CPU model provides a subset of hardware PMU 3184 functionality). Using a mediated vPMU significantly 3185 reduces PMU virtualization overhead and eliminates lost 3186 guest events, but is mutually exclusive with using perf 3187 to profile KVM guests and adds latency to most VM-Exits 3188 (to context switch PMU state). 3189 3190 Default is N (off). 3191 3192 kvm-amd.nested= [KVM,AMD] Control nested virtualization feature in 3193 KVM/SVM. Default is 1 (enabled). 3194 3195 kvm-amd.npt= [KVM,AMD] Control KVM's use of Nested Page Tables, 3196 a.k.a. Two-Dimensional Page Tables. Default is 1 3197 (enabled). Disable by KVM if hardware lacks support 3198 for NPT. 3199 3200 kvm-amd.ciphertext_hiding_asids= 3201 [KVM,AMD] Ciphertext hiding prevents disallowed accesses 3202 to SNP private memory from reading ciphertext. Instead, 3203 reads will see constant default values (0xff). 3204 3205 If ciphertext hiding is enabled, the joint SEV-ES and 3206 SEV-SNP ASID space is partitioned into separate SEV-ES 3207 and SEV-SNP ASID ranges, with the SEV-SNP range being 3208 [1..max_snp_asid] and the SEV-ES range being 3209 (max_snp_asid..min_sev_asid), where min_sev_asid is 3210 enumerated by CPUID.0x.8000_001F[EDX]. 3211 3212 A non-zero value enables SEV-SNP ciphertext hiding and 3213 adjusts the ASID ranges for SEV-ES and SEV-SNP guests. 3214 KVM caps the number of SEV-SNP ASIDs at the maximum 3215 possible value, e.g. specifying -1u will assign all 3216 joint SEV-ES and SEV-SNP ASIDs to SEV-SNP. Note, 3217 assigning all joint ASIDs to SEV-SNP, i.e. configuring 3218 max_snp_asid == min_sev_asid-1, will effectively make 3219 SEV-ES unusable. 3220 3221 kvm-arm.mode= 3222 [KVM,ARM,EARLY] Select one of KVM/arm64's modes of 3223 operation. 3224 3225 none: Forcefully disable KVM. 3226 3227 nvhe: Standard nVHE-based mode, without support for 3228 protected guests. 3229 3230 protected: Mode with support for guests whose state is 3231 kept private from the host, using VHE or 3232 nVHE depending on HW support. 3233 3234 nested: VHE-based mode with support for nested 3235 virtualization. Requires at least ARMv8.4 3236 hardware (with FEAT_NV2). 3237 3238 Defaults to VHE/nVHE based on hardware support. Setting 3239 mode to "protected" will disable kexec and hibernation 3240 for the host. To force nVHE on VHE hardware, add 3241 "arm64_sw.hvhe=0 id_aa64mmfr1.vh=0" to the 3242 command-line. 3243 "nested" is experimental and should be used with 3244 extreme caution. 3245 3246 kvm-arm.vgic_v3_group0_trap= 3247 [KVM,ARM,EARLY] Trap guest accesses to GICv3 group-0 3248 system registers 3249 3250 kvm-arm.vgic_v3_group1_trap= 3251 [KVM,ARM,EARLY] Trap guest accesses to GICv3 group-1 3252 system registers 3253 3254 kvm-arm.vgic_v3_common_trap= 3255 [KVM,ARM,EARLY] Trap guest accesses to GICv3 common 3256 system registers 3257 3258 kvm-arm.vgic_v4_enable= 3259 [KVM,ARM,EARLY] Allow use of GICv4 for direct 3260 injection of LPIs. 3261 3262 kvm-arm.wfe_trap_policy= 3263 [KVM,ARM] Control when to set WFE instruction trap for 3264 KVM VMs. Traps are allowed but not guaranteed by the 3265 CPU architecture. 3266 3267 trap: set WFE instruction trap 3268 3269 notrap: clear WFE instruction trap 3270 3271 kvm-arm.wfi_trap_policy= 3272 [KVM,ARM] Control when to set WFI instruction trap for 3273 KVM VMs. Traps are allowed but not guaranteed by the 3274 CPU architecture. 3275 3276 trap: set WFI instruction trap 3277 3278 notrap: clear WFI instruction trap 3279 3280 kvm_cma_resv_ratio=n [PPC,EARLY] 3281 Reserves given percentage from system memory area for 3282 contiguous memory allocation for KVM hash pagetable 3283 allocation. 3284 By default it reserves 5% of total system memory. 3285 Format: <integer> 3286 Default: 5 3287 3288 kvm-intel.ept= [KVM,Intel] Control KVM's use of Extended Page Tables, 3289 a.k.a. Two-Dimensional Page Tables. Default is 1 3290 (enabled). Disable by KVM if hardware lacks support 3291 for EPT. 3292 3293 kvm-intel.emulate_invalid_guest_state= 3294 [KVM,Intel] Control whether to emulate invalid guest 3295 state. Ignored if kvm-intel.enable_unrestricted_guest=1, 3296 as guest state is never invalid for unrestricted 3297 guests. This param doesn't apply to nested guests (L2), 3298 as KVM never emulates invalid L2 guest state. 3299 Default is 1 (enabled). 3300 3301 kvm-intel.flexpriority= 3302 [KVM,Intel] Control KVM's use of FlexPriority feature 3303 (TPR shadow). Default is 1 (enabled). Disable by KVM if 3304 hardware lacks support for it. 3305 3306 kvm-intel.nested= 3307 [KVM,Intel] Control nested virtualization feature in 3308 KVM/VMX. Default is 1 (enabled). 3309 3310 kvm-intel.unrestricted_guest= 3311 [KVM,Intel] Control KVM's use of unrestricted guest 3312 feature (virtualized real and unpaged mode). Default 3313 is 1 (enabled). Disable by KVM if EPT is disabled or 3314 hardware lacks support for it. 3315 3316 kvm-intel.vmentry_l1d_flush=[KVM,Intel] Mitigation for L1 Terminal Fault 3317 CVE-2018-3620. 3318 3319 Valid arguments: never, cond, always 3320 3321 always: L1D cache flush on every VMENTER. 3322 cond: Flush L1D on VMENTER only when the code between 3323 VMEXIT and VMENTER can leak host memory. 3324 never: Disables the mitigation 3325 3326 Default is cond (do L1 cache flush in specific instances) 3327 3328 kvm-intel.vpid= [KVM,Intel] Control KVM's use of Virtual Processor 3329 Identification feature (tagged TLBs). Default is 1 3330 (enabled). Disable by KVM if hardware lacks support 3331 for it. 3332 3333 l1d_flush= [X86,INTEL,EARLY] 3334 Control mitigation for L1D based snooping vulnerability. 3335 3336 Certain CPUs are vulnerable to an exploit against CPU 3337 internal buffers which can forward information to a 3338 disclosure gadget under certain conditions. 3339 3340 In vulnerable processors, the speculatively 3341 forwarded data can be used in a cache side channel 3342 attack, to access data to which the attacker does 3343 not have direct access. 3344 3345 This parameter controls the mitigation. The 3346 options are: 3347 3348 on - enable the interface for the mitigation 3349 3350 l1tf= [X86,EARLY] Control mitigation of the L1TF vulnerability on 3351 affected CPUs 3352 3353 The kernel PTE inversion protection is unconditionally 3354 enabled and cannot be disabled. 3355 3356 full 3357 Provides all available mitigations for the 3358 L1TF vulnerability. Disables SMT and 3359 enables all mitigations in the 3360 hypervisors, i.e. unconditional L1D flush. 3361 3362 SMT control and L1D flush control via the 3363 sysfs interface is still possible after 3364 boot. Hypervisors will issue a warning 3365 when the first VM is started in a 3366 potentially insecure configuration, 3367 i.e. SMT enabled or L1D flush disabled. 3368 3369 full,force 3370 Same as 'full', but disables SMT and L1D 3371 flush runtime control. Implies the 3372 'nosmt=force' command line option. 3373 (i.e. sysfs control of SMT is disabled.) 3374 3375 flush 3376 Leaves SMT enabled and enables the default 3377 hypervisor mitigation, i.e. conditional 3378 L1D flush. 3379 3380 SMT control and L1D flush control via the 3381 sysfs interface is still possible after 3382 boot. Hypervisors will issue a warning 3383 when the first VM is started in a 3384 potentially insecure configuration, 3385 i.e. SMT enabled or L1D flush disabled. 3386 3387 flush,nosmt 3388 3389 Disables SMT and enables the default 3390 hypervisor mitigation. 3391 3392 SMT control and L1D flush control via the 3393 sysfs interface is still possible after 3394 boot. Hypervisors will issue a warning 3395 when the first VM is started in a 3396 potentially insecure configuration, 3397 i.e. SMT enabled or L1D flush disabled. 3398 3399 flush,nowarn 3400 Same as 'flush', but hypervisors will not 3401 warn when a VM is started in a potentially 3402 insecure configuration. 3403 3404 off 3405 Disables hypervisor mitigations and doesn't 3406 emit any warnings. 3407 It also drops the swap size and available 3408 RAM limit restriction on both hypervisor and 3409 bare metal. 3410 3411 Default is 'flush'. 3412 3413 For details see: Documentation/admin-guide/hw-vuln/l1tf.rst 3414 3415 l2cr= [PPC] 3416 3417 l3cr= [PPC] 3418 3419 lapic [X86-32,APIC,EARLY] Enable the local APIC even if BIOS 3420 disabled it. 3421 3422 lapic= [X86,APIC] Do not use TSC deadline 3423 value for LAPIC timer one-shot implementation. Default 3424 back to the programmable timer unit in the LAPIC. 3425 Format: notscdeadline 3426 3427 lapic_timer_c2_ok [X86,APIC,EARLY] trust the local apic timer 3428 in C2 power state. 3429 3430 libata.dma= [LIBATA] DMA control 3431 libata.dma=0 Disable all PATA and SATA DMA 3432 libata.dma=1 PATA and SATA Disk DMA only 3433 libata.dma=2 ATAPI (CDROM) DMA only 3434 libata.dma=4 Compact Flash DMA only 3435 Combinations also work, so libata.dma=3 enables DMA 3436 for disks and CDROMs, but not CFs. 3437 3438 libata.ignore_hpa= [LIBATA] Ignore HPA limit 3439 libata.ignore_hpa=0 keep BIOS limits (default) 3440 libata.ignore_hpa=1 ignore limits, using full disk 3441 3442 libata.noacpi [LIBATA] Disables use of ACPI in libata suspend/resume 3443 when set. 3444 Format: <int> 3445 3446 libata.force= [LIBATA] Force configurations. The format is a comma- 3447 separated list of "[ID:]VAL" where ID is PORT[.DEVICE]. 3448 PORT and DEVICE are decimal numbers matching port, link 3449 or device. Basically, it matches the ATA ID string 3450 printed on console by libata. If the whole ID part is 3451 omitted, the last PORT and DEVICE values are used. If 3452 ID hasn't been specified yet, the configuration applies 3453 to all ports, links and devices. 3454 3455 If only DEVICE is omitted, the parameter applies to 3456 the port and all links and devices behind it. DEVICE 3457 number of 0 either selects the first device or the 3458 first fan-out link behind PMP device. It does not 3459 select the host link. DEVICE number of 15 selects the 3460 host link and device attached to it. 3461 3462 The VAL specifies the configuration to force. As long 3463 as there is no ambiguity, shortcut notation is allowed. 3464 For example, both 1.5 and 1.5G would work for 1.5Gbps. 3465 The following configurations can be forced. 3466 3467 * Cable type: 40c, 80c, short40c, unk, ign or sata. 3468 Any ID with matching PORT is used. 3469 3470 * SATA link speed limit: 1.5Gbps or 3.0Gbps. 3471 3472 * Transfer mode: pio[0-7], mwdma[0-4] and udma[0-7]. 3473 udma[/][16,25,33,44,66,100,133] notation is also 3474 allowed. 3475 3476 * nohrst, nosrst, norst: suppress hard, soft and both 3477 resets. 3478 3479 * rstonce: only attempt one reset during hot-unplug 3480 link recovery. 3481 3482 * [no]dbdelay: Enable or disable the extra 200ms delay 3483 before debouncing a link PHY and device presence 3484 detection. 3485 3486 * [no]ncq: Turn on or off NCQ. 3487 3488 * [no]ncqtrim: Enable or disable queued DSM TRIM. 3489 3490 * [no]ncqati: Enable or disable NCQ trim on ATI chipset. 3491 3492 * [no]trim: Enable or disable (unqueued) TRIM. 3493 3494 * trim_zero: Indicate that TRIM command zeroes data. 3495 3496 * max_trim_128m: Set 128M maximum trim size limit. 3497 3498 * [no]dma: Turn on or off DMA transfers. 3499 3500 * atapi_dmadir: Enable ATAPI DMADIR bridge support. 3501 3502 * atapi_mod16_dma: Enable the use of ATAPI DMA for 3503 commands that are not a multiple of 16 bytes. 3504 3505 * [no]dmalog: Enable or disable the use of the 3506 READ LOG DMA EXT command to access logs. 3507 3508 * [no]iddevlog: Enable or disable access to the 3509 identify device data log. 3510 3511 * [no]logdir: Enable or disable access to the general 3512 purpose log directory. 3513 3514 * max_sec=<sectors>: Set the transfer size limit, in 3515 number of 512-byte sectors, to the value specified in 3516 <sectors>. The value specified in <sectors> has to be 3517 a non-zero positive integer. 3518 3519 * max_sec_128: Set transfer size limit to 128 sectors. 3520 3521 * max_sec_1024: Set or clear transfer size limit to 3522 1024 sectors. 3523 3524 * max_sec_lba48: Set or clear transfer size limit to 3525 65535 sectors. 3526 3527 * external: Mark port as external (hotplug-capable). 3528 3529 * [no]lpm: Enable or disable link power management. 3530 3531 * [no]setxfer: Indicate if transfer speed mode setting 3532 should be skipped. 3533 3534 * [no]fua: Disable or enable FUA (Force Unit Access) 3535 support for devices supporting this feature. 3536 3537 * dump_id: Dump IDENTIFY data. 3538 3539 * disable: Disable this device. 3540 3541 If there are multiple matching configurations changing 3542 the same attribute, the last one is used. 3543 3544 liveupdate= [KNL,EARLY] 3545 Format: <bool> 3546 Enable Live Update Orchestrator (LUO). 3547 Default: off. 3548 3549 lockd.nlm_grace_period=P [NFS] Assign grace period. 3550 Format: <integer> 3551 3552 lockd.nlm_tcpport=N [NFS] Assign TCP port. 3553 Format: <integer> 3554 3555 lockd.nlm_timeout=T [NFS] Assign timeout value. 3556 Format: <integer> 3557 3558 lockd.nlm_udpport=M [NFS] Assign UDP port. 3559 Format: <integer> 3560 3561 lockdown= [SECURITY,EARLY] 3562 { integrity | confidentiality } 3563 Enable the kernel lockdown feature. If set to 3564 integrity, kernel features that allow userland to 3565 modify the running kernel are disabled. If set to 3566 confidentiality, kernel features that allow userland 3567 to extract confidential information from the kernel 3568 are also disabled. 3569 3570 locktorture.acq_writer_lim= [KNL] 3571 Set the time limit in jiffies for a lock 3572 acquisition. Acquisitions exceeding this limit 3573 will result in a splat once they do complete. 3574 3575 locktorture.bind_readers= [KNL] 3576 Specify the list of CPUs to which the readers are 3577 to be bound. 3578 3579 locktorture.bind_writers= [KNL] 3580 Specify the list of CPUs to which the writers are 3581 to be bound. 3582 3583 locktorture.call_rcu_chains= [KNL] 3584 Specify the number of self-propagating call_rcu() 3585 chains to set up. These are used to ensure that 3586 there is a high probability of an RCU grace period 3587 in progress at any given time. Defaults to 0, 3588 which disables these call_rcu() chains. 3589 3590 locktorture.long_hold= [KNL] 3591 Specify the duration in milliseconds for the 3592 occasional long-duration lock hold time. Defaults 3593 to 100 milliseconds. Select 0 to disable. 3594 3595 locktorture.nested_locks= [KNL] 3596 Specify the maximum lock nesting depth that 3597 locktorture is to exercise, up to a limit of 8 3598 (MAX_NESTED_LOCKS). Specify zero to disable. 3599 Note that this parameter is ineffective on types 3600 of locks that do not support nested acquisition. 3601 3602 locktorture.nreaders_stress= [KNL] 3603 Set the number of locking read-acquisition kthreads. 3604 Defaults to being automatically set based on the 3605 number of online CPUs. 3606 3607 locktorture.nwriters_stress= [KNL] 3608 Set the number of locking write-acquisition kthreads. 3609 3610 locktorture.onoff_holdoff= [KNL] 3611 Set time (s) after boot for CPU-hotplug testing. 3612 3613 locktorture.onoff_interval= [KNL] 3614 Set time (s) between CPU-hotplug operations, or 3615 zero to disable CPU-hotplug testing. 3616 3617 locktorture.rt_boost= [KNL] 3618 Do periodic testing of real-time lock priority 3619 boosting. Select 0 to disable, 1 to boost 3620 only rt_mutex, and 2 to boost unconditionally. 3621 Defaults to 2, which might seem to be an 3622 odd choice, but which should be harmless for 3623 non-real-time spinlocks, due to their disabling 3624 of preemption. Note that non-realtime mutexes 3625 disable boosting. 3626 3627 locktorture.rt_boost_factor= [KNL] 3628 Number that determines how often and for how 3629 long priority boosting is exercised. This is 3630 scaled down by the number of writers, so that the 3631 number of boosts per unit time remains roughly 3632 constant as the number of writers increases. 3633 On the other hand, the duration of each boost 3634 increases with the number of writers. 3635 3636 locktorture.shuffle_interval= [KNL] 3637 Set task-shuffle interval (jiffies). Shuffling 3638 tasks allows some CPUs to go into dyntick-idle 3639 mode during the locktorture test. 3640 3641 locktorture.shutdown_secs= [KNL] 3642 Set time (s) after boot system shutdown. This 3643 is useful for hands-off automated testing. 3644 3645 locktorture.stat_interval= [KNL] 3646 Time (s) between statistics printk()s. 3647 3648 locktorture.stutter= [KNL] 3649 Time (s) to stutter testing, for example, 3650 specifying five seconds causes the test to run for 3651 five seconds, wait for five seconds, and so on. 3652 This tests the locking primitive's ability to 3653 transition abruptly to and from idle. 3654 3655 locktorture.torture_type= [KNL] 3656 Specify the locking implementation to test. 3657 3658 locktorture.verbose= [KNL] 3659 Enable additional printk() statements. 3660 3661 locktorture.writer_fifo= [KNL] 3662 Run the write-side locktorture kthreads at 3663 sched_set_fifo() real-time priority. 3664 3665 logibm.irq= [HW,MOUSE] Logitech Bus Mouse Driver 3666 Format: <irq> 3667 3668 loglevel= [KNL,EARLY] 3669 All Kernel Messages with a loglevel smaller than the 3670 console loglevel will be printed to the console. It can 3671 also be changed with klogd or other programs. The 3672 loglevels are defined as follows: 3673 3674 0 (KERN_EMERG) system is unusable 3675 1 (KERN_ALERT) action must be taken immediately 3676 2 (KERN_CRIT) critical conditions 3677 3 (KERN_ERR) error conditions 3678 4 (KERN_WARNING) warning conditions 3679 5 (KERN_NOTICE) normal but significant condition 3680 6 (KERN_INFO) informational 3681 7 (KERN_DEBUG) debug-level messages 3682 3683 log_buf_len=n[KMG] [KNL,EARLY] 3684 Sets the size of the printk ring buffer, in bytes. 3685 n must be a power of two and greater than the 3686 minimal size. The minimal size is defined by 3687 LOG_BUF_SHIFT kernel config parameter. There 3688 is also CONFIG_LOG_CPU_MAX_BUF_SHIFT config 3689 parameter that allows to increase the default size 3690 depending on the number of CPUs. See init/Kconfig 3691 for more details. 3692 3693 logo.nologo [FB] Disables display of the built-in Linux logo. 3694 This may be used to provide more screen space for 3695 kernel log messages and is useful when debugging 3696 kernel boot problems. 3697 3698 lp=0 [LP] Specify parallel ports to use, e.g, 3699 lp=port[,port...] lp=none,parport0 (lp0 not configured, lp1 uses 3700 lp=reset first parallel port). 'lp=0' disables the 3701 lp=auto printer driver. 'lp=reset' (which can be 3702 specified in addition to the ports) causes 3703 attached printers to be reset. Using 3704 lp=port1,port2,... specifies the parallel ports 3705 to associate lp devices with, starting with 3706 lp0. A port specification may be 'none' to skip 3707 that lp device, or a parport name such as 3708 'parport0'. Specifying 'lp=auto' instead of a 3709 port specification list means that device IDs 3710 from each port should be examined, to see if 3711 an IEEE 1284-compliant printer is attached; if 3712 so, the driver will manage that printer. 3713 See also header of drivers/char/lp.c. 3714 3715 lpj=n [KNL] 3716 Sets loops_per_jiffy to given constant, thus avoiding 3717 time-consuming boot-time autodetection (up to 250 ms per 3718 CPU). 0 enables autodetection (default). To determine 3719 the correct value for your kernel, boot with normal 3720 autodetection and see what value is printed. Note that 3721 on SMP systems the preset will be applied to all CPUs, 3722 which is likely to cause problems if your CPUs need 3723 significantly divergent settings. An incorrect value 3724 will cause delays in the kernel to be wrong, leading to 3725 unpredictable I/O errors and other breakage. Although 3726 unlikely, in the extreme case this might damage your 3727 hardware. 3728 3729 lsm.debug [SECURITY] Enable LSM initialization debugging output. 3730 3731 lsm=lsm1,...,lsmN 3732 [SECURITY] Choose order of LSM initialization. This 3733 overrides CONFIG_LSM, and the "security=" parameter. 3734 3735 machtype= [Loongson] Share the same kernel image file between 3736 different yeeloong laptops. 3737 Example: machtype=lemote-yeeloong-2f-7inch 3738 3739 maxcpus= [SMP,EARLY] Maximum number of processors that an SMP kernel 3740 will bring up during bootup. maxcpus=n : n >= 0 limits 3741 the kernel to bring up 'n' processors. Surely after 3742 bootup you can bring up the other plugged cpu by executing 3743 "echo 1 > /sys/devices/system/cpu/cpuX/online". So maxcpus 3744 only takes effect during system bootup. 3745 While n=0 is a special case, it is equivalent to "nosmp", 3746 which also disables the IO APIC. 3747 3748 max_loop= [LOOP] The number of loop block devices that get 3749 (loop.max_loop) unconditionally pre-created at init time. The default 3750 number is configured by BLK_DEV_LOOP_MIN_COUNT. Instead 3751 of statically allocating a predefined number, loop 3752 devices can be requested on-demand with the 3753 /dev/loop-control interface. 3754 3755 mce= [X86-{32,64}] 3756 3757 Please see Documentation/arch/x86/x86_64/machinecheck.rst for sysfs runtime tunables. 3758 3759 off 3760 disable machine check 3761 3762 no_cmci 3763 disable CMCI(Corrected Machine Check Interrupt) that 3764 Intel processor supports. Usually this disablement is 3765 not recommended, but it might be handy if your 3766 hardware is misbehaving. 3767 3768 Note that you'll get more problems without CMCI than 3769 with due to the shared banks, i.e. you might get 3770 duplicated error logs. 3771 3772 dont_log_ce 3773 don't make logs for corrected errors. All events 3774 reported as corrected are silently cleared by OS. This 3775 option will be useful if you have no interest in any 3776 of corrected errors. 3777 3778 ignore_ce 3779 disable features for corrected errors, e.g. 3780 polling timer and CMCI. All events reported as 3781 corrected are not cleared by OS and remained in its 3782 error banks. 3783 3784 Usually this disablement is not recommended, however 3785 if there is an agent checking/clearing corrected 3786 errors (e.g. BIOS or hardware monitoring 3787 applications), conflicting with OS's error handling, 3788 and you cannot deactivate the agent, then this option 3789 will be a help. 3790 3791 no_lmce 3792 do not opt-in to Local MCE delivery. Use legacy method 3793 to broadcast MCEs. 3794 3795 bootlog 3796 enable logging of machine checks left over from 3797 booting. Disabled by default on AMD Fam10h and older 3798 because some BIOS leave bogus ones. 3799 3800 If your BIOS doesn't do that it's a good idea to 3801 enable though to make sure you log even machine check 3802 events that result in a reboot. On Intel systems it is 3803 enabled by default. 3804 3805 nobootlog 3806 disable boot machine check logging. 3807 3808 monarchtimeout (number) 3809 sets the time in us to wait for other CPUs on machine 3810 checks. 0 to disable. 3811 3812 bios_cmci_threshold 3813 don't overwrite the bios-set CMCI threshold. This boot 3814 option prevents Linux from overwriting the CMCI 3815 threshold set by the bios. Without this option, Linux 3816 always sets the CMCI threshold to 1. Enabling this may 3817 make memory predictive failure analysis less effective 3818 if the bios sets thresholds for memory errors since we 3819 will not see details for all errors. 3820 3821 recovery 3822 force-enable recoverable machine check code paths 3823 3824 Everything else is in sysfs now. 3825 3826 3827 md= [HW] RAID subsystems devices and level 3828 See Documentation/admin-guide/md.rst. 3829 3830 mdacon= [MDA] 3831 Format: <first>,<last> 3832 Specifies range of consoles to be captured by the MDA. 3833 3834 mds= [X86,INTEL,EARLY] 3835 Control mitigation for the Micro-architectural Data 3836 Sampling (MDS) vulnerability. 3837 3838 Certain CPUs are vulnerable to an exploit against CPU 3839 internal buffers which can forward information to a 3840 disclosure gadget under certain conditions. 3841 3842 In vulnerable processors, the speculatively 3843 forwarded data can be used in a cache side channel 3844 attack, to access data to which the attacker does 3845 not have direct access. 3846 3847 This parameter controls the MDS mitigation. The 3848 options are: 3849 3850 full - Enable MDS mitigation on vulnerable CPUs 3851 full,nosmt - Enable MDS mitigation and disable 3852 SMT on vulnerable CPUs 3853 off - Unconditionally disable MDS mitigation 3854 3855 On TAA-affected machines, mds=off can be prevented by 3856 an active TAA mitigation as both vulnerabilities are 3857 mitigated with the same mechanism so in order to disable 3858 this mitigation, you need to specify tsx_async_abort=off 3859 too. 3860 3861 Not specifying this option is equivalent to 3862 mds=full. 3863 3864 For details see: Documentation/admin-guide/hw-vuln/mds.rst 3865 3866 mem=nn[KMG] [HEXAGON,EARLY] Set the memory size. 3867 Must be specified, otherwise memory size will be 0. 3868 3869 mem=nn[KMG] [KNL,BOOT,EARLY] Force usage of a specific amount 3870 of memory Amount of memory to be used in cases 3871 as follows: 3872 3873 1 for test; 3874 2 when the kernel is not able to see the whole system memory; 3875 3 memory that lies after 'mem=' boundary is excluded from 3876 the hypervisor, then assigned to KVM guests. 3877 4 to limit the memory available for kdump kernel. 3878 3879 [ARC,MICROBLAZE] - the limit applies only to low memory, 3880 high memory is not affected. 3881 3882 [ARM64] - only limits memory covered by the linear 3883 mapping. The NOMAP regions are not affected. 3884 3885 [X86] Work as limiting max address. Use together 3886 with memmap= to avoid physical address space collisions. 3887 Without memmap= PCI devices could be placed at addresses 3888 belonging to unused RAM. 3889 3890 Note that this only takes effects during boot time since 3891 in above case 3, memory may need be hot added after boot 3892 if system memory of hypervisor is not sufficient. 3893 3894 mem=nn[KMG]@ss[KMG] 3895 [ARM,MIPS,EARLY] - override the memory layout 3896 reported by firmware. 3897 Define a memory region of size nn[KMG] starting at 3898 ss[KMG]. 3899 Multiple different regions can be specified with 3900 multiple mem= parameters on the command line. 3901 3902 mem=nopentium [BUGS=X86-32] Disable usage of 4MB pages for kernel 3903 memory. 3904 3905 memblock=debug [KNL,EARLY] Enable memblock debug messages. 3906 3907 memchunk=nn[KMG] 3908 [KNL,SH] Allow user to override the default size for 3909 per-device physically contiguous DMA buffers. 3910 3911 memhp_default_state=online/offline/online_kernel/online_movable 3912 [KNL] Set the initial state for the memory hotplug 3913 onlining policy. If not specified, the default value is 3914 set according to the 3915 CONFIG_MHP_DEFAULT_ONLINE_TYPE kernel config 3916 options. 3917 See Documentation/admin-guide/mm/memory-hotplug.rst. 3918 3919 memmap=exactmap [KNL,X86,EARLY] Enable setting of an exact 3920 E820 memory map, as specified by the user. 3921 Such memmap=exactmap lines can be constructed based on 3922 BIOS output or other requirements. See the memmap=nn@ss 3923 option description. 3924 3925 memmap=nn[KMG]@ss[KMG] 3926 [KNL, X86,MIPS,XTENSA,EARLY] Force usage of a specific region of memory. 3927 Region of memory to be used is from ss to ss+nn. 3928 If @ss[KMG] is omitted, it is equivalent to mem=nn[KMG], 3929 which limits max address to nn[KMG]. 3930 Multiple different regions can be specified, 3931 comma delimited. 3932 Example: 3933 memmap=100M@2G,100M#3G,1G!1024G 3934 3935 memmap=nn[KMG]#ss[KMG] 3936 [KNL,ACPI,EARLY] Mark specific memory as ACPI data. 3937 Region of memory to be marked is from ss to ss+nn. 3938 3939 memmap=nn[KMG]$ss[KMG] 3940 [KNL,ACPI,EARLY] Mark specific memory as reserved. 3941 Region of memory to be reserved is from ss to ss+nn. 3942 Example: Exclude memory from 0x18690000-0x1869ffff 3943 memmap=64K$0x18690000 3944 or 3945 memmap=0x10000$0x18690000 3946 Some bootloaders may need an escape character before '$', 3947 like Grub2, otherwise '$' and the following number 3948 will be eaten. 3949 3950 memmap=nn[KMG]!ss[KMG,EARLY] 3951 [KNL,X86] Mark specific memory as protected. 3952 Region of memory to be used, from ss to ss+nn. 3953 The memory region may be marked as e820 type 12 (0xc) 3954 and is NVDIMM or ADR memory. 3955 3956 memmap=<size>%<offset>-<oldtype>+<newtype> 3957 [KNL,ACPI,EARLY] Convert memory within the specified region 3958 from <oldtype> to <newtype>. If "-<oldtype>" is left 3959 out, the whole region will be marked as <newtype>, 3960 even if previously unavailable. If "+<newtype>" is left 3961 out, matching memory will be removed. Types are 3962 specified as e820 types, e.g., 1 = RAM, 2 = reserved, 3963 3 = ACPI, 12 = PRAM. 3964 3965 memory_corruption_check=0/1 [X86,EARLY] 3966 Some BIOSes seem to corrupt the first 64k of 3967 memory when doing things like suspend/resume. 3968 Setting this option will scan the memory 3969 looking for corruption. Enabling this will 3970 both detect corruption and prevent the kernel 3971 from using the memory being corrupted. 3972 However, it's intended as a diagnostic tool; if 3973 repeatable BIOS-originated corruption always 3974 affects the same memory, you can use memmap= 3975 to prevent the kernel from using that memory. 3976 3977 memory_corruption_check_size=size [X86,EARLY] 3978 By default it checks for corruption in the low 3979 64k, making this memory unavailable for normal 3980 use. Use this parameter to scan for 3981 corruption in more or less memory. 3982 3983 memory_corruption_check_period=seconds [X86,EARLY] 3984 By default it checks for corruption every 60 3985 seconds. Use this parameter to check at some 3986 other rate. 0 disables periodic checking. 3987 3988 memory_hotplug.memmap_on_memory 3989 [KNL,X86,ARM] Boolean flag to enable this feature. 3990 Format: {on | off (default)} 3991 When enabled, runtime hotplugged memory will 3992 allocate its internal metadata (struct pages, 3993 those vmemmap pages cannot be optimized even 3994 if hugetlb_free_vmemmap is enabled) from the 3995 hotadded memory which will allow to hotadd a 3996 lot of memory without requiring additional 3997 memory to do so. 3998 This feature is disabled by default because it 3999 has some implication on large (e.g. GB) 4000 allocations in some configurations (e.g. small 4001 memory blocks). 4002 The state of the flag can be read in 4003 /sys/module/memory_hotplug/parameters/memmap_on_memory. 4004 Note that even when enabled, there are a few cases where 4005 the feature is not effective. 4006 4007 memtest= [KNL,X86,ARM,M68K,PPC,RISCV,EARLY] Enable memtest 4008 Format: <integer> 4009 default : 0 <disable> 4010 Specifies the number of memtest passes to be 4011 performed. Each pass selects another test 4012 pattern from a given set of patterns. Memtest 4013 fills the memory with this pattern, validates 4014 memory contents and reserves bad memory 4015 regions that are detected. 4016 4017 mem_encrypt= [X86-64] AMD Secure Memory Encryption (SME) control 4018 Valid arguments: on, off 4019 Default: off 4020 mem_encrypt=on: Activate SME 4021 mem_encrypt=off: Do not activate SME 4022 4023 Refer to Documentation/virt/kvm/x86/amd-memory-encryption.rst 4024 for details on when memory encryption can be activated. 4025 4026 mem_sleep_default= [SUSPEND] Default system suspend mode: 4027 s2idle - Suspend-To-Idle 4028 shallow - Power-On Suspend or equivalent (if supported) 4029 deep - Suspend-To-RAM or equivalent (if supported) 4030 See Documentation/admin-guide/pm/sleep-states.rst. 4031 4032 mfgptfix [X86-32] Fix MFGPT timers on AMD Geode platforms when 4033 the BIOS has incorrectly applied a workaround. TinyBIOS 4034 version 0.98 is known to be affected, 0.99 fixes the 4035 problem by letting the user disable the workaround. 4036 4037 mga= [HW,DRM] 4038 4039 microcode= [X86] Control the behavior of the microcode loader. 4040 Available options, comma separated: 4041 4042 base_rev=X - with <X> with format: <u32> 4043 Set the base microcode revision of each thread when in 4044 debug mode. 4045 4046 dis_ucode_ldr: disable the microcode loader 4047 4048 force_minrev: 4049 Enable or disable the microcode minimal revision 4050 enforcement for the runtime microcode loader. 4051 4052 mini2440= [ARM,HW,KNL] 4053 Format:[0..2][b][c][t] 4054 Default: "0tb" 4055 MINI2440 configuration specification: 4056 0 - The attached screen is the 3.5" TFT 4057 1 - The attached screen is the 7" TFT 4058 2 - The VGA Shield is attached (1024x768) 4059 Leaving out the screen size parameter will not load 4060 the TFT driver, and the framebuffer will be left 4061 unconfigured. 4062 b - Enable backlight. The TFT backlight pin will be 4063 linked to the kernel VESA blanking code and a GPIO 4064 LED. This parameter is not necessary when using the 4065 VGA shield. 4066 c - Enable the s3c camera interface. 4067 t - Reserved for enabling touchscreen support. The 4068 touchscreen support is not enabled in the mainstream 4069 kernel as of 2.6.30, a preliminary port can be found 4070 in the "bleeding edge" mini2440 support kernel at 4071 https://repo.or.cz/w/linux-2.6/mini2440.git 4072 4073 mitigations= 4074 [X86,PPC,S390,ARM64,EARLY] Control optional mitigations for 4075 CPU vulnerabilities. This is a set of curated, 4076 arch-independent options, each of which is an 4077 aggregation of existing arch-specific options. 4078 4079 Note, "mitigations" is supported if and only if the 4080 kernel was built with CPU_MITIGATIONS=y. 4081 4082 off 4083 Disable all optional CPU mitigations. This 4084 improves system performance, but it may also 4085 expose users to several CPU vulnerabilities. 4086 Equivalent to: if nokaslr then kpti=0 [ARM64] 4087 gather_data_sampling=off [X86] 4088 indirect_target_selection=off [X86] 4089 kvm.nx_huge_pages=off [X86] 4090 l1tf=off [X86] 4091 mds=off [X86] 4092 mmio_stale_data=off [X86] 4093 no_entry_flush [PPC] 4094 no_uaccess_flush [PPC] 4095 nobp=0 [S390] 4096 nopti [X86,PPC] 4097 nospectre_bhb [ARM64] 4098 nospectre_v1 [X86,PPC] 4099 nospectre_v2 [X86,PPC,S390,ARM64] 4100 reg_file_data_sampling=off [X86] 4101 retbleed=off [X86] 4102 spec_rstack_overflow=off [X86] 4103 spec_store_bypass_disable=off [X86,PPC] 4104 spectre_bhi=off [X86] 4105 spectre_v2_user=off [X86] 4106 srbds=off [X86,INTEL] 4107 ssbd=force-off [ARM64] 4108 tsa=off [X86,AMD] 4109 tsx_async_abort=off [X86] 4110 vmscape=off [X86] 4111 4112 Exceptions: 4113 This does not have any effect on 4114 kvm.nx_huge_pages when 4115 kvm.nx_huge_pages=force. 4116 4117 auto (default) 4118 Mitigate all CPU vulnerabilities, but leave SMT 4119 enabled, even if it's vulnerable. This is for 4120 users who don't want to be surprised by SMT 4121 getting disabled across kernel upgrades, or who 4122 have other ways of avoiding SMT-based attacks. 4123 Equivalent to: (default behavior) 4124 4125 auto,nosmt 4126 Mitigate all CPU vulnerabilities, disabling SMT 4127 if needed. This is for users who always want to 4128 be fully mitigated, even if it means losing SMT. 4129 Equivalent to: l1tf=flush,nosmt [X86] 4130 mds=full,nosmt [X86] 4131 tsx_async_abort=full,nosmt [X86] 4132 mmio_stale_data=full,nosmt [X86] 4133 retbleed=auto,nosmt [X86] 4134 4135 [X86] After one of the above options, additionally 4136 supports attack-vector based controls as documented in 4137 Documentation/admin-guide/hw-vuln/attack_vector_controls.rst 4138 4139 mminit_loglevel= 4140 [KNL,EARLY] When CONFIG_DEBUG_MEMORY_INIT is set, this 4141 parameter allows control of the logging verbosity for 4142 the additional memory initialisation checks. A value 4143 of 0 disables mminit logging and a level of 4 will 4144 log everything. Information is printed at KERN_DEBUG 4145 so loglevel=8 may also need to be specified. 4146 4147 mmio_stale_data= 4148 [X86,INTEL,EARLY] Control mitigation for the Processor 4149 MMIO Stale Data vulnerabilities. 4150 4151 Processor MMIO Stale Data is a class of 4152 vulnerabilities that may expose data after an MMIO 4153 operation. Exposed data could originate or end in 4154 the same CPU buffers as affected by MDS and TAA. 4155 Therefore, similar to MDS and TAA, the mitigation 4156 is to clear the affected CPU buffers. 4157 4158 This parameter controls the mitigation. The 4159 options are: 4160 4161 full - Enable mitigation on vulnerable CPUs 4162 4163 full,nosmt - Enable mitigation and disable SMT on 4164 vulnerable CPUs. 4165 4166 off - Unconditionally disable mitigation 4167 4168 On MDS or TAA affected machines, 4169 mmio_stale_data=off can be prevented by an active 4170 MDS or TAA mitigation as these vulnerabilities are 4171 mitigated with the same mechanism so in order to 4172 disable this mitigation, you need to specify 4173 mds=off and tsx_async_abort=off too. 4174 4175 Not specifying this option is equivalent to 4176 mmio_stale_data=full. 4177 4178 For details see: 4179 Documentation/admin-guide/hw-vuln/processor_mmio_stale_data.rst 4180 4181 <module>.async_probe[=<bool>] [KNL] 4182 If no <bool> value is specified or if the value 4183 specified is not a valid <bool>, enable asynchronous 4184 probe on this module. Otherwise, enable/disable 4185 asynchronous probe on this module as indicated by the 4186 <bool> value. See also: module.async_probe 4187 4188 module.async_probe=<bool> 4189 [KNL] When set to true, modules will use async probing 4190 by default. To enable/disable async probing for a 4191 specific module, use the module specific control that 4192 is documented under <module>.async_probe. When both 4193 module.async_probe and <module>.async_probe are 4194 specified, <module>.async_probe takes precedence for 4195 the specific module. 4196 4197 module.enable_dups_trace 4198 [KNL] When CONFIG_MODULE_DEBUG_AUTOLOAD_DUPS is set, 4199 this means that duplicate request_module() calls will 4200 trigger a WARN_ON() instead of a pr_warn(). Note that 4201 if MODULE_DEBUG_AUTOLOAD_DUPS_TRACE is set, WARN_ON()s 4202 will always be issued and this option does nothing. 4203 module.sig_enforce 4204 [KNL] When CONFIG_MODULE_SIG is set, this means that 4205 modules without (valid) signatures will fail to load. 4206 Note that if CONFIG_MODULE_SIG_FORCE is set, that 4207 is always true, so this option does nothing. 4208 4209 module_blacklist= [KNL] Do not load a comma-separated list of 4210 modules. Useful for debugging problem modules. 4211 4212 mousedev.tap_time= 4213 [MOUSE] Maximum time between finger touching and 4214 leaving touchpad surface for touch to be considered 4215 a tap and be reported as a left button click (for 4216 touchpads working in absolute mode only). 4217 Format: <msecs> 4218 mousedev.xres= [MOUSE] Horizontal screen resolution, used for devices 4219 reporting absolute coordinates, such as tablets 4220 mousedev.yres= [MOUSE] Vertical screen resolution, used for devices 4221 reporting absolute coordinates, such as tablets 4222 4223 movablecore= [KNL,X86,PPC,EARLY] 4224 Format: nn[KMGTPE] | nn% 4225 This parameter is the complement to kernelcore=, it 4226 specifies the amount of memory used for migratable 4227 allocations. If both kernelcore and movablecore is 4228 specified, then kernelcore will be at *least* the 4229 specified value but may be more. If movablecore on its 4230 own is specified, the administrator must be careful 4231 that the amount of memory usable for all allocations 4232 is not too small. 4233 4234 movable_node [KNL,EARLY] Boot-time switch to make hotplugable memory 4235 NUMA nodes to be movable. This means that the memory 4236 of such nodes will be usable only for movable 4237 allocations which rules out almost all kernel 4238 allocations. Use with caution! 4239 4240 MTD_Partition= [MTD] 4241 Format: <name>,<region-number>,<size>,<offset> 4242 4243 MTD_Region= [MTD] Format: 4244 <name>,<region-number>[,<base>,<size>,<buswidth>,<altbuswidth>] 4245 4246 mtdparts= [MTD] 4247 See drivers/mtd/parsers/cmdlinepart.c 4248 4249 mtouchusb.raw_coordinates= 4250 [HW] Make the MicroTouch USB driver use raw coordinates 4251 ('y', default) or cooked coordinates ('n') 4252 4253 mtrr=debug [X86,EARLY] 4254 Enable printing debug information related to MTRR 4255 registers at boot time. 4256 4257 mtrr_chunk_size=nn[KMG,X86,EARLY] 4258 used for mtrr cleanup. It is largest continuous chunk 4259 that could hold holes aka. UC entries. 4260 4261 mtrr_gran_size=nn[KMG,X86,EARLY] 4262 Used for mtrr cleanup. It is granularity of mtrr block. 4263 Default is 1. 4264 Large value could prevent small alignment from 4265 using up MTRRs. 4266 4267 mtrr_spare_reg_nr=n [X86,EARLY] 4268 Format: <integer> 4269 Range: 0,7 : spare reg number 4270 Default : 1 4271 Used for mtrr cleanup. It is spare mtrr entries number. 4272 Set to 2 or more if your graphical card needs more. 4273 4274 multitce=off [PPC] This parameter disables the use of the pSeries 4275 firmware feature for updating multiple TCE entries 4276 at a time. 4277 4278 n2= [NET] SDL Inc. RISCom/N2 synchronous serial card 4279 4280 netdev= [NET] Network devices parameters 4281 Format: <irq>,<io>,<mem_start>,<mem_end>,<name> 4282 Note that mem_start is often overloaded to mean 4283 something different and driver-specific. 4284 This usage is only documented in each driver source 4285 file if at all. 4286 4287 netpoll.carrier_timeout= 4288 [NET] Specifies amount of time (in seconds) that 4289 netpoll should wait for a carrier. By default netpoll 4290 waits 4 seconds. 4291 4292 nf_conntrack.acct= 4293 [NETFILTER] Enable connection tracking flow accounting 4294 0 to disable accounting 4295 1 to enable accounting 4296 Default value is 0. 4297 4298 nfs.cache_getent= 4299 [NFS] sets the pathname to the program which is used 4300 to update the NFS client cache entries. 4301 4302 nfs.cache_getent_timeout= 4303 [NFS] sets the timeout after which an attempt to 4304 update a cache entry is deemed to have failed. 4305 4306 nfs.callback_nr_threads= 4307 [NFSv4] set the total number of threads that the 4308 NFS client will assign to service NFSv4 callback 4309 requests. 4310 4311 nfs.callback_tcpport= 4312 [NFS] set the TCP port on which the NFSv4 callback 4313 channel should listen. 4314 4315 nfs.delay_retrans= 4316 [NFS] specifies the number of times the NFSv4 client 4317 retries the request before returning an EAGAIN error, 4318 after a reply of NFS4ERR_DELAY from the server. 4319 Only applies if the softerr mount option is enabled, 4320 and the specified value is >= 0. 4321 4322 nfs.enable_ino64= 4323 [NFS] enable 64-bit inode numbers. 4324 If zero, the NFS client will fake up a 32-bit inode 4325 number for the readdir() and stat() syscalls instead 4326 of returning the full 64-bit number. 4327 The default is to return 64-bit inode numbers. 4328 4329 nfs.idmap_cache_timeout= 4330 [NFS] set the maximum lifetime for idmapper cache 4331 entries. 4332 4333 nfs.max_session_cb_slots= 4334 [NFSv4.1] Sets the maximum number of session 4335 slots the client will assign to the callback 4336 channel. This determines the maximum number of 4337 callbacks the client will process in parallel for 4338 a particular server. 4339 4340 nfs.max_session_slots= 4341 [NFSv4.1] Sets the maximum number of session slots 4342 the client will attempt to negotiate with the server. 4343 This limits the number of simultaneous RPC requests 4344 that the client can send to the NFSv4.1 server. 4345 Note that there is little point in setting this 4346 value higher than the max_tcp_slot_table_limit. 4347 4348 nfs.nfs4_disable_idmapping= 4349 [NFSv4] When set to the default of '1', this option 4350 ensures that both the RPC level authentication 4351 scheme and the NFS level operations agree to use 4352 numeric uids/gids if the mount is using the 4353 'sec=sys' security flavour. In effect it is 4354 disabling idmapping, which can make migration from 4355 legacy NFSv2/v3 systems to NFSv4 easier. 4356 Servers that do not support this mode of operation 4357 will be autodetected by the client, and it will fall 4358 back to using the idmapper. 4359 To turn off this behaviour, set the value to '0'. 4360 4361 nfs.nfs4_unique_id= 4362 [NFS4] Specify an additional fixed unique ident- 4363 ification string that NFSv4 clients can insert into 4364 their nfs_client_id4 string. This is typically a 4365 UUID that is generated at system install time. 4366 4367 nfs.recover_lost_locks= 4368 [NFSv4] Attempt to recover locks that were lost due 4369 to a lease timeout on the server. Please note that 4370 doing this risks data corruption, since there are 4371 no guarantees that the file will remain unchanged 4372 after the locks are lost. 4373 If you want to enable the kernel legacy behaviour of 4374 attempting to recover these locks, then set this 4375 parameter to '1'. 4376 The default parameter value of '0' causes the kernel 4377 not to attempt recovery of lost locks. 4378 4379 nfs.send_implementation_id= 4380 [NFSv4.1] Send client implementation identification 4381 information in exchange_id requests. 4382 If zero, no implementation identification information 4383 will be sent. 4384 The default is to send the implementation identification 4385 information. 4386 4387 nfs4.layoutstats_timer= 4388 [NFSv4.2] Change the rate at which the kernel sends 4389 layoutstats to the pNFS metadata server. 4390 4391 Setting this to value to 0 causes the kernel to use 4392 whatever value is the default set by the layout 4393 driver. A non-zero value sets the minimum interval 4394 in seconds between layoutstats transmissions. 4395 4396 nfsd.inter_copy_offload_enable= 4397 [NFSv4.2] When set to 1, the server will support 4398 server-to-server copies for which this server is 4399 the destination of the copy. 4400 4401 nfsd.nfs4_disable_idmapping= 4402 [NFSv4] When set to the default of '1', the NFSv4 4403 server will return only numeric uids and gids to 4404 clients using auth_sys, and will accept numeric uids 4405 and gids from such clients. This is intended to ease 4406 migration from NFSv2/v3. 4407 4408 nfsd.nfsd4_ssc_umount_timeout= 4409 [NFSv4.2] When used as the destination of a 4410 server-to-server copy, knfsd temporarily mounts 4411 the source server. It caches the mount in case 4412 it will be needed again, and discards it if not 4413 used for the number of milliseconds specified by 4414 this parameter. 4415 4416 nfsaddrs= [NFS] Deprecated. Use ip= instead. 4417 See Documentation/admin-guide/nfs/nfsroot.rst. 4418 4419 nfsroot= [NFS] nfs root filesystem for disk-less boxes. 4420 See Documentation/admin-guide/nfs/nfsroot.rst. 4421 4422 nfsrootdebug [NFS] enable nfsroot debugging messages. 4423 See Documentation/admin-guide/nfs/nfsroot.rst. 4424 4425 nmi_backtrace.backtrace_idle [KNL] 4426 Dump stacks even of idle CPUs in response to an 4427 NMI stack-backtrace request. 4428 4429 nmi_debug= [KNL,SH] Specify one or more actions to take 4430 when a NMI is triggered. 4431 Format: [state][,regs][,debounce][,die] 4432 4433 nmi_watchdog= [KNL,BUGS=X86] Debugging features for SMP kernels 4434 Format: [panic,][nopanic,][rNNN,][num] 4435 Valid num: 0 or 1 4436 0 - turn hardlockup detector in nmi_watchdog off 4437 1 - turn hardlockup detector in nmi_watchdog on 4438 rNNN - configure the watchdog with raw perf event 0xNNN 4439 4440 When panic is specified, panic when an NMI watchdog 4441 timeout occurs (or 'nopanic' to not panic on an NMI 4442 watchdog, if CONFIG_BOOTPARAM_HARDLOCKUP_PANIC is set) 4443 To disable both hard and soft lockup detectors, 4444 please see 'nowatchdog'. 4445 This is useful when you use a panic=... timeout and 4446 need the box quickly up again. 4447 4448 These settings can be accessed at runtime via 4449 the nmi_watchdog and hardlockup_panic sysctls. 4450 4451 no387 [BUGS=X86-32] Tells the kernel to use the 387 maths 4452 emulation library even if a 387 maths coprocessor 4453 is present. 4454 4455 no4lvl [RISCV,EARLY] Disable 4-level and 5-level paging modes. 4456 Forces kernel to use 3-level paging instead. 4457 4458 no5lvl [X86-64,RISCV,EARLY] Disable 5-level paging mode. Forces 4459 kernel to use 4-level paging instead. 4460 4461 noalign [KNL,ARM] 4462 4463 noapic [SMP,APIC,EARLY] Tells the kernel to not make use of any 4464 IOAPICs that may be present in the system. 4465 4466 noapictimer [APIC,X86] Don't set up the APIC timer 4467 4468 noautogroup Disable scheduler automatic task group creation. 4469 4470 nocache [ARM,EARLY] 4471 4472 no_console_suspend 4473 [HW] Never suspend the console 4474 Disable suspending of consoles during suspend and 4475 hibernate operations. Once disabled, debugging 4476 messages can reach various consoles while the rest 4477 of the system is being put to sleep (ie, while 4478 debugging driver suspend/resume hooks). This may 4479 not work reliably with all consoles, but is known 4480 to work with serial and VGA consoles. 4481 To facilitate more flexible debugging, we also add 4482 console_suspend, a printk module parameter to control 4483 it. Users could use console_suspend (usually 4484 /sys/module/printk/parameters/console_suspend) to 4485 turn on/off it dynamically. 4486 4487 no_debug_objects 4488 [KNL,EARLY] Disable object debugging 4489 4490 nodsp [SH] Disable hardware DSP at boot time. 4491 4492 noefi [EFI,EARLY] Disable EFI runtime services support. 4493 4494 no_entry_flush [PPC,EARLY] Don't flush the L1-D cache when entering the kernel. 4495 4496 noexec32 [X86-64] 4497 This affects only 32-bit executables. 4498 noexec32=on: enable non-executable mappings (default) 4499 read doesn't imply executable mappings 4500 noexec32=off: disable non-executable mappings 4501 read implies executable mappings 4502 4503 no_file_caps Tells the kernel not to honor file capabilities. The 4504 only way then for a file to be executed with privilege 4505 is to be setuid root or executed by root. 4506 4507 nofpu [MIPS,SH] Disable hardware FPU at boot time. 4508 4509 nofsgsbase [X86] Disables FSGSBASE instructions. 4510 4511 nofxsr [BUGS=X86-32] Disables x86 floating point extended 4512 register save and restore. The kernel will only save 4513 legacy floating-point registers on task switch. 4514 4515 nogbpages [X86] Do not use GB pages for kernel direct mappings. 4516 4517 no_hash_pointers 4518 [KNL,EARLY] 4519 Alias for "hash_pointers=never". 4520 4521 nohibernate [HIBERNATION] Disable hibernation and resume. 4522 4523 nohlt [ARM,ARM64,MICROBLAZE,MIPS,PPC,RISCV,SH] Forces the kernel to 4524 busy wait in do_idle() and not use the arch_cpu_idle() 4525 implementation; requires CONFIG_GENERIC_IDLE_POLL_SETUP 4526 to be effective. This is useful on platforms where the 4527 sleep(SH) or wfi(ARM,ARM64) instructions do not work 4528 correctly or when doing power measurements to evaluate 4529 the impact of the sleep instructions. This is also 4530 useful when using JTAG debugger. 4531 4532 nohpet [X86] Don't use the HPET timer. 4533 4534 nohugeiomap [KNL,X86,PPC,ARM64,EARLY] Disable kernel huge I/O mappings. 4535 4536 nohugevmalloc [KNL,X86,PPC,ARM64,EARLY] Disable kernel huge vmalloc mappings. 4537 4538 nohz= [KNL] Boottime enable/disable dynamic ticks 4539 Valid arguments: on, off 4540 Default: on 4541 4542 nohz_full= [KNL,BOOT,SMP,ISOL] 4543 The argument is a cpu list, as described above. 4544 In kernels built with CONFIG_NO_HZ_FULL=y, set 4545 the specified list of CPUs whose tick will be stopped 4546 whenever possible. The boot CPU will be forced outside 4547 the range to maintain the timekeeping. Any CPUs 4548 in this list will have their RCU callbacks offloaded, 4549 just as if they had also been called out in the 4550 rcu_nocbs= boot parameter. 4551 4552 Note that this argument takes precedence over 4553 the CONFIG_RCU_NOCB_CPU_DEFAULT_ALL option. 4554 4555 noinitrd [Deprecated,RAM] Tells the kernel not to load any configured 4556 initial RAM disk. Currently this parameter applies to 4557 initrd only, not to initramfs. But it applies to both 4558 in EFI mode. 4559 4560 nointremap [X86-64,Intel-IOMMU,EARLY] Do not enable interrupt 4561 remapping. 4562 [Deprecated - use intremap=off] 4563 4564 noinvpcid [X86,EARLY] Disable the INVPCID cpu feature. 4565 4566 noiotrap [SH] Disables trapped I/O port accesses. 4567 4568 noirqdebug [X86-32] Disables the code which attempts to detect and 4569 disable unhandled interrupt sources. 4570 4571 noisapnp [ISAPNP] Disables ISA PnP code. 4572 4573 nokaslr [KNL,EARLY] 4574 When CONFIG_RANDOMIZE_BASE is set, this disables 4575 kernel and module base offset ASLR (Address Space 4576 Layout Randomization). 4577 4578 no-kvmapf [X86,KVM,EARLY] Disable paravirtualized asynchronous page 4579 fault handling. 4580 4581 no-kvmclock [X86,KVM,EARLY] Disable paravirtualized KVM clock driver 4582 4583 nolapic [X86-32,APIC,EARLY] Do not enable or use the local APIC. 4584 4585 nolapic_timer [X86-32,APIC,EARLY] Do not use the local APIC timer. 4586 4587 nomce [X86-32] Disable Machine Check Exception 4588 4589 nomfgpt [X86-32] Disable Multi-Function General Purpose 4590 Timer usage (for AMD Geode machines). 4591 4592 nomodeset Disable kernel modesetting. Most systems' firmware 4593 sets up a display mode and provides framebuffer memory 4594 for output. With nomodeset, DRM and fbdev drivers will 4595 not load if they could possibly displace the pre- 4596 initialized output. Only the system framebuffer will 4597 be available for use. The respective drivers will not 4598 perform display-mode changes or accelerated rendering. 4599 4600 Useful as error fallback, or for testing and debugging. 4601 4602 nomodule Disable module load 4603 4604 nonmi_ipi [X86] Disable using NMI IPIs during panic/reboot to 4605 shutdown the other cpus. Instead use the REBOOT_VECTOR 4606 irq. 4607 4608 nopat [X86,EARLY] Disable PAT (page attribute table extension of 4609 pagetables) support. 4610 4611 nopcid [X86-64,EARLY] Disable the PCID cpu feature. 4612 4613 nopku [X86] Disable Memory Protection Keys CPU feature found 4614 in some Intel CPUs. 4615 4616 nopti [X86-64,EARLY] 4617 Equivalent to pti=off 4618 4619 nopv= [X86,XEN,KVM,HYPER_V,VMWARE,EARLY] 4620 Disables the PV optimizations forcing the guest to run 4621 as generic guest with no PV drivers. Currently support 4622 XEN HVM, KVM, HYPER_V and VMWARE guest. 4623 4624 nopvspin [X86,XEN,KVM,EARLY] 4625 Disables the qspinlock slow path using PV optimizations 4626 which allow the hypervisor to 'idle' the guest on lock 4627 contention. 4628 4629 norandmaps Don't use address space randomization. Equivalent to 4630 echo 0 > /proc/sys/kernel/randomize_va_space 4631 4632 noreplace-smp [X86-32,SMP] Don't replace SMP instructions 4633 with UP alternatives 4634 4635 noresume [SWSUSP] Disables resume and restores original swap 4636 space. 4637 4638 no-scroll [VGA] Disables scrollback. 4639 This is required for the Braillex ib80-piezo Braille 4640 reader made by F.H. Papenmeier (Germany). 4641 4642 nosgx [X86-64,SGX,EARLY] Disables Intel SGX kernel support. 4643 4644 nosmap [PPC,EARLY] 4645 Disable SMAP (Supervisor Mode Access Prevention) 4646 even if it is supported by processor. 4647 4648 nosmep [PPC64s,EARLY] 4649 Disable SMEP (Supervisor Mode Execution Prevention) 4650 even if it is supported by processor. 4651 4652 nosmp [SMP,EARLY] Tells an SMP kernel to act as a UP kernel, 4653 and disable the IO APIC. legacy for "maxcpus=0". 4654 4655 nosmt [KNL,MIPS,PPC,EARLY] Disable symmetric multithreading (SMT). 4656 Equivalent to smt=1. 4657 4658 [KNL,LOONGARCH,X86,PPC,S390] Disable symmetric multithreading (SMT). 4659 nosmt=force: Force disable SMT, cannot be undone 4660 via the sysfs control file. 4661 4662 nosoftlockup [KNL] Disable the soft-lockup detector. 4663 4664 nospec_store_bypass_disable 4665 [HW,EARLY] Disable all mitigations for the Speculative 4666 Store Bypass vulnerability 4667 4668 nospectre_bhb [ARM64,EARLY] Disable all mitigations for Spectre-BHB (branch 4669 history injection) vulnerability. System may allow data leaks 4670 with this option. 4671 4672 nospectre_v1 [X86,PPC,EARLY] Disable mitigations for Spectre Variant 1 4673 (bounds check bypass). With this option data leaks are 4674 possible in the system. 4675 4676 nospectre_v2 [X86,PPC_E500,ARM64,EARLY] Disable all mitigations 4677 for the Spectre variant 2 (indirect branch 4678 prediction) vulnerability. System may allow data 4679 leaks with this option. 4680 4681 no-steal-acc [X86,PV_OPS,ARM64,PPC/PSERIES,RISCV,LOONGARCH,EARLY] 4682 Disable paravirtualized steal time accounting. steal time 4683 is computed, but won't influence scheduler behaviour 4684 4685 nosync [HW,M68K] Disables sync negotiation for all devices. 4686 4687 no_timer_check [X86,APIC] Disables the code which tests for broken 4688 timer IRQ sources, i.e., the IO-APIC timer. This can 4689 work around problems with incorrect timer 4690 initialization on some boards. 4691 4692 no_uaccess_flush 4693 [PPC,EARLY] Don't flush the L1-D cache after accessing user data. 4694 4695 novmcoredd [KNL,KDUMP] 4696 Disable device dump. Device dump allows drivers to 4697 append dump data to vmcore so you can collect driver 4698 specified debug info. Drivers can append the data 4699 without any limit and this data is stored in memory, 4700 so this may cause significant memory stress. Disabling 4701 device dump can help save memory but the driver debug 4702 data will be no longer available. This parameter 4703 is only available when CONFIG_PROC_VMCORE_DEVICE_DUMP 4704 is set. 4705 4706 no-vmw-sched-clock 4707 [X86,PV_OPS,EARLY] Disable paravirtualized VMware 4708 scheduler clock and use the default one. 4709 4710 nowatchdog [KNL] Disable both lockup detectors, i.e. 4711 soft-lockup and NMI watchdog (hard-lockup). 4712 4713 nowb [ARM,EARLY] 4714 4715 nox2apic [X86-64,APIC,EARLY] Do not enable x2APIC mode. 4716 4717 NOTE: this parameter will be ignored on systems with the 4718 LEGACY_XAPIC_DISABLED bit set in the 4719 IA32_XAPIC_DISABLE_STATUS MSR. 4720 4721 noxsave [BUGS=X86] Disables x86 extended register state save 4722 and restore using xsave. The kernel will fallback to 4723 enabling legacy floating-point and sse state. 4724 4725 noxsaveopt [X86] Disables xsaveopt used in saving x86 extended 4726 register states. The kernel will fall back to use 4727 xsave to save the states. By using this parameter, 4728 performance of saving the states is degraded because 4729 xsave doesn't support modified optimization while 4730 xsaveopt supports it on xsaveopt enabled systems. 4731 4732 noxsaves [X86] Disables xsaves and xrstors used in saving and 4733 restoring x86 extended register state in compacted 4734 form of xsave area. The kernel will fall back to use 4735 xsaveopt and xrstor to save and restore the states 4736 in standard form of xsave area. By using this 4737 parameter, xsave area per process might occupy more 4738 memory on xsaves enabled systems. 4739 4740 nr_cpus= [SMP,EARLY] Maximum number of processors that an SMP kernel 4741 could support. nr_cpus=n : n >= 1 limits the kernel to 4742 support 'n' processors. It could be larger than the 4743 number of already plugged CPU during bootup, later in 4744 runtime you can physically add extra cpu until it reaches 4745 n. So during boot up some boot time memory for per-cpu 4746 variables need be pre-allocated for later physical cpu 4747 hot plugging. 4748 4749 nr_uarts= [SERIAL] maximum number of UARTs to be registered. 4750 4751 numa=off [KNL, ARM64, PPC, RISCV, SPARC, X86, EARLY] 4752 Disable NUMA, Only set up a single NUMA node 4753 spanning all memory. 4754 4755 numa=fake=<size>[MG] 4756 [KNL, ARM64, RISCV, X86, EARLY] 4757 If given as a memory unit, fills all system RAM with 4758 nodes of size interleaved over physical nodes. 4759 4760 numa=fake=<N> 4761 [KNL, ARM64, RISCV, X86, EARLY] 4762 If given as an integer, fills all system RAM with N 4763 fake nodes interleaved over physical nodes. 4764 4765 numa=fake=<N>U 4766 [KNL, ARM64, RISCV, X86, EARLY] 4767 If given as an integer followed by 'U', it will 4768 divide each physical node into N emulated nodes. 4769 4770 numa=noacpi [X86] Don't parse the SRAT table for NUMA setup 4771 4772 numa=nohmat [X86] Don't parse the HMAT table for NUMA setup, or 4773 soft-reserved memory partitioning. 4774 4775 numa_balancing= [KNL,ARM64,PPC,RISCV,S390,X86] Enable or disable automatic 4776 NUMA balancing. 4777 Allowed values are enable and disable 4778 4779 numa_zonelist_order= [KNL, BOOT] Select zonelist order for NUMA. 4780 'node', 'default' can be specified 4781 This can be set from sysctl after boot. 4782 See Documentation/admin-guide/sysctl/vm.rst for details. 4783 4784 nvme.quirks= [NVME] A list of quirk entries to augment the built-in 4785 nvme quirk list. List entries are separated by a 4786 '-' character. 4787 Each entry has the form VendorID:ProductID:quirk_names. 4788 The IDs are 4-digits hex numbers and quirk_names is a 4789 list of quirk names separated by commas. A quirk name 4790 can be prefixed by '^', meaning that the specified 4791 quirk must be disabled. 4792 4793 Example: 4794 nvme.quirks=7710:2267:bogus_nid,^identify_cns-9900:7711:broken_msi 4795 4796 ohci1394_dma=early [HW,EARLY] enable debugging via the ohci1394 driver. 4797 See Documentation/core-api/debugging-via-ohci1394.rst for more 4798 info. 4799 4800 olpc_ec_timeout= [OLPC] ms delay when issuing EC commands 4801 Rather than timing out after 20 ms if an EC 4802 command is not properly ACKed, override the length 4803 of the timeout. We have interrupts disabled while 4804 waiting for the ACK, so if this is set too high 4805 interrupts *may* be lost! 4806 4807 omap_mux= [OMAP] Override bootloader pin multiplexing. 4808 Format: <mux_mode0.mode_name=value>... 4809 For example, to override I2C bus2: 4810 omap_mux=i2c2_scl.i2c2_scl=0x100,i2c2_sda.i2c2_sda=0x100 4811 4812 onenand.bdry= [HW,MTD] Flex-OneNAND Boundary Configuration 4813 4814 Format: [die0_boundary][,die0_lock][,die1_boundary][,die1_lock] 4815 4816 boundary - index of last SLC block on Flex-OneNAND. 4817 The remaining blocks are configured as MLC blocks. 4818 lock - Configure if Flex-OneNAND boundary should be locked. 4819 Once locked, the boundary cannot be changed. 4820 1 indicates lock status, 0 indicates unlock status. 4821 4822 oops=panic [KNL,EARLY] 4823 Always panic on oopses. Default is to just kill the 4824 process, but there is a small probability of 4825 deadlocking the machine. 4826 This will also cause panics on machine check exceptions. 4827 Useful together with panic=30 to trigger a reboot. 4828 4829 page_alloc.shuffle= 4830 [KNL] Boolean flag to control whether the page allocator 4831 should randomize its free lists. This parameter can be 4832 used to enable/disable page randomization. The state of 4833 the flag can be read from sysfs at: 4834 /sys/module/page_alloc/parameters/shuffle. 4835 This parameter is only available if CONFIG_SHUFFLE_PAGE_ALLOCATOR=y. 4836 4837 page_owner= [KNL,EARLY] Boot-time page_owner enabling option. 4838 Storage of the information about who allocated 4839 each page is disabled in default. With this switch, 4840 we can turn it on. 4841 on: enable the feature 4842 4843 page_poison= [KNL,EARLY] Boot-time parameter changing the state of 4844 poisoning on the buddy allocator, available with 4845 CONFIG_PAGE_POISONING=y. 4846 off: turn off poisoning (default) 4847 on: turn on poisoning 4848 4849 page_reporting.page_reporting_order= 4850 [KNL] Minimal page reporting order 4851 Format: <integer> 4852 Adjust the minimal page reporting order. The page 4853 reporting is disabled when it exceeds MAX_PAGE_ORDER. 4854 4855 panic= [KNL] Kernel behaviour on panic: delay <timeout> 4856 timeout > 0: seconds before rebooting 4857 timeout = 0: wait forever 4858 timeout < 0: reboot immediately 4859 Format: <timeout> 4860 4861 panic_on_taint= [KNL,EARLY] 4862 Bitmask for conditionally calling panic() in add_taint() 4863 Format: <hex>[,nousertaint] 4864 Hexadecimal bitmask representing the set of TAINT flags 4865 that will cause the kernel to panic when add_taint() is 4866 called with any of the flags in this set. 4867 The optional switch "nousertaint" can be utilized to 4868 prevent userspace forced crashes by writing to sysctl 4869 /proc/sys/kernel/tainted any flagset matching with the 4870 bitmask set on panic_on_taint. 4871 See Documentation/admin-guide/tainted-kernels.rst for 4872 extra details on the taint flags that users can pick 4873 to compose the bitmask to assign to panic_on_taint. 4874 4875 panic_on_warn=1 panic() instead of WARN(). Useful to cause kdump 4876 on a WARN(). 4877 4878 panic_force_cpu= 4879 [KNL,SMP] Force panic handling to execute on a specific CPU. 4880 Format: <cpu number> 4881 Some platforms require panic handling to occur on a 4882 specific CPU for the crash kernel to function correctly. 4883 This can be due to firmware limitations, interrupt routing 4884 constraints, or platform-specific requirements where only 4885 a particular CPU can safely enter the crash kernel. 4886 When set, panic() will redirect execution to the specified 4887 CPU before proceeding with the normal panic and kexec flow. 4888 If the target CPU is offline or unavailable, panic proceeds 4889 on the current CPU. 4890 This option should only be used for systems with the above 4891 constraints as it might cause the panic operation to be less reliable. 4892 4893 panic_print= Bitmask for printing system info when panic happens. 4894 User can chose combination of the following bits: 4895 bit 0: print all tasks info 4896 bit 1: print system memory info 4897 bit 2: print timer info 4898 bit 3: print locks info if CONFIG_LOCKDEP is on 4899 bit 4: print ftrace buffer 4900 bit 5: replay all kernel messages on consoles at the end of panic 4901 bit 6: print all CPUs backtrace (if available in the arch) 4902 bit 7: print only tasks in uninterruptible (blocked) state 4903 *Be aware* that this option may print a _lot_ of lines, 4904 so there are risks of losing older messages in the log. 4905 Use this option carefully, maybe worth to setup a 4906 bigger log buffer with "log_buf_len" along with this. 4907 4908 panic_sys_info= A comma separated list of extra information to be dumped 4909 on panic. 4910 Format: val[,val...] 4911 Where @val can be any of the following: 4912 4913 tasks: print all tasks info 4914 mem: print system memory info 4915 timers: print timers info 4916 locks: print locks info if CONFIG_LOCKDEP is on 4917 ftrace: print ftrace buffer 4918 all_bt: print all CPUs backtrace (if available in the arch) 4919 blocked_tasks: print only tasks in uninterruptible (blocked) state 4920 4921 This is a human readable alternative to the 'panic_print' option. 4922 4923 panic_console_replay 4924 When panic happens, replay all kernel messages on 4925 consoles at the end of panic. 4926 4927 parkbd.port= [HW] Parallel port number the keyboard adapter is 4928 connected to, default is 0. 4929 Format: <parport#> 4930 parkbd.mode= [HW] Parallel port keyboard adapter mode of operation, 4931 0 for XT, 1 for AT (default is AT). 4932 Format: <mode> 4933 4934 parport= [HW,PPT] Specify parallel ports. 0 disables. 4935 Format: { 0 | auto | 0xBBB[,IRQ[,DMA]] } 4936 Use 'auto' to force the driver to use any 4937 IRQ/DMA settings detected (the default is to 4938 ignore detected IRQ/DMA settings because of 4939 possible conflicts). You can specify the base 4940 address, IRQ, and DMA settings; IRQ and DMA 4941 should be numbers, or 'auto' (for using detected 4942 settings on that particular port), or 'nofifo' 4943 (to avoid using a FIFO even if it is detected). 4944 Parallel ports are assigned in the order they 4945 are specified on the command line, starting 4946 with parport0. 4947 4948 parport_init_mode= [HW,PPT] 4949 Configure VIA parallel port to operate in 4950 a specific mode. This is necessary on Pegasos 4951 computer where firmware has no options for setting 4952 up parallel port mode and sets it to spp. 4953 Currently this function knows 686a and 8231 chips. 4954 Format: [spp|ps2|epp|ecp|ecpepp] 4955 4956 pata_legacy.all= [HW,LIBATA] 4957 Format: <int> 4958 Set to non-zero to probe primary and secondary ISA 4959 port ranges on PCI systems where no PCI PATA device 4960 has been found at either range. Disabled by default. 4961 4962 pata_legacy.autospeed= [HW,LIBATA] 4963 Format: <int> 4964 Set to non-zero if a chip is present that snoops speed 4965 changes. Disabled by default. 4966 4967 pata_legacy.ht6560a= [HW,LIBATA] 4968 Format: <int> 4969 Set to 1, 2, or 3 for HT 6560A on the primary channel, 4970 the secondary channel, or both channels respectively. 4971 Disabled by default. 4972 4973 pata_legacy.ht6560b= [HW,LIBATA] 4974 Format: <int> 4975 Set to 1, 2, or 3 for HT 6560B on the primary channel, 4976 the secondary channel, or both channels respectively. 4977 Disabled by default. 4978 4979 pata_legacy.iordy_mask= [HW,LIBATA] 4980 Format: <int> 4981 IORDY enable mask. Set individual bits to allow IORDY 4982 for the respective channel. Bit 0 is for the first 4983 legacy channel handled by this driver, bit 1 is for 4984 the second channel, and so on. The sequence will often 4985 correspond to the primary legacy channel, the secondary 4986 legacy channel, and so on, but the handling of a PCI 4987 bus and the use of other driver options may interfere 4988 with the sequence. By default IORDY is allowed across 4989 all channels. 4990 4991 pata_legacy.opti82c46x= [HW,LIBATA] 4992 Format: <int> 4993 Set to 1, 2, or 3 for Opti 82c611A on the primary 4994 channel, the secondary channel, or both channels 4995 respectively. Disabled by default. 4996 4997 pata_legacy.opti82c611a= [HW,LIBATA] 4998 Format: <int> 4999 Set to 1, 2, or 3 for Opti 82c465MV on the primary 5000 channel, the secondary channel, or both channels 5001 respectively. Disabled by default. 5002 5003 pata_legacy.pio_mask= [HW,LIBATA] 5004 Format: <int> 5005 PIO mode mask for autospeed devices. Set individual 5006 bits to allow the use of the respective PIO modes. 5007 Bit 0 is for mode 0, bit 1 is for mode 1, and so on. 5008 All modes allowed by default. 5009 5010 pata_legacy.probe_all= [HW,LIBATA] 5011 Format: <int> 5012 Set to non-zero to probe tertiary and further ISA 5013 port ranges on PCI systems. Disabled by default. 5014 5015 pata_legacy.probe_mask= [HW,LIBATA] 5016 Format: <int> 5017 Probe mask for legacy ISA PATA ports. Depending on 5018 platform configuration and the use of other driver 5019 options up to 6 legacy ports are supported: 0x1f0, 5020 0x170, 0x1e8, 0x168, 0x1e0, 0x160, however probing 5021 of individual ports can be disabled by setting the 5022 corresponding bits in the mask to 1. Bit 0 is for 5023 the first port in the list above (0x1f0), and so on. 5024 By default all supported ports are probed. 5025 5026 pata_legacy.qdi= [HW,LIBATA] 5027 Format: <int> 5028 Set to non-zero to probe QDI controllers. By default 5029 set to 1 if CONFIG_PATA_QDI_MODULE, 0 otherwise. 5030 5031 pata_legacy.winbond= [HW,LIBATA] 5032 Format: <int> 5033 Set to non-zero to probe Winbond controllers. Use 5034 the standard I/O port (0x130) if 1, otherwise the 5035 value given is the I/O port to use (typically 0x1b0). 5036 By default set to 1 if CONFIG_PATA_WINBOND_VLB_MODULE, 5037 0 otherwise. 5038 5039 pata_platform.pio_mask= [HW,LIBATA] 5040 Format: <int> 5041 Supported PIO mode mask. Set individual bits to allow 5042 the use of the respective PIO modes. Bit 0 is for 5043 mode 0, bit 1 is for mode 1, and so on. Mode 0 only 5044 allowed by default. 5045 5046 pause_on_oops=<int> 5047 Halt all CPUs after the first oops has been printed for 5048 the specified number of seconds. This is to be used if 5049 your oopses keep scrolling off the screen. 5050 5051 pcbit= [HW,ISDN] 5052 5053 pci=option[,option...] [PCI,EARLY] various PCI subsystem options. 5054 5055 Some options herein operate on a specific device 5056 or a set of devices (<pci_dev>). These are 5057 specified in one of the following formats: 5058 5059 [<domain>:]<bus>:<dev>.<func>[/<dev>.<func>]* 5060 pci:<vendor>:<device>[:<subvendor>:<subdevice>] 5061 5062 Note: the first format specifies a PCI 5063 bus/device/function address which may change 5064 if new hardware is inserted, if motherboard 5065 firmware changes, or due to changes caused 5066 by other kernel parameters. If the 5067 domain is left unspecified, it is 5068 taken to be zero. Optionally, a path 5069 to a device through multiple device/function 5070 addresses can be specified after the base 5071 address (this is more robust against 5072 renumbering issues). The second format 5073 selects devices using IDs from the 5074 configuration space which may match multiple 5075 devices in the system. 5076 5077 earlydump dump PCI config space before the kernel 5078 changes anything 5079 off [X86] don't probe for the PCI bus 5080 bios [X86-32] force use of PCI BIOS, don't access 5081 the hardware directly. Use this if your machine 5082 has a non-standard PCI host bridge. 5083 nobios [X86-32] disallow use of PCI BIOS, only direct 5084 hardware access methods are allowed. Use this 5085 if you experience crashes upon bootup and you 5086 suspect they are caused by the BIOS. 5087 conf1 [X86] Force use of PCI Configuration Access 5088 Mechanism 1 (config address in IO port 0xCF8, 5089 data in IO port 0xCFC, both 32-bit). 5090 conf2 [X86] Force use of PCI Configuration Access 5091 Mechanism 2 (IO port 0xCF8 is an 8-bit port for 5092 the function, IO port 0xCFA, also 8-bit, sets 5093 bus number. The config space is then accessed 5094 through ports 0xC000-0xCFFF). 5095 See http://wiki.osdev.org/PCI for more info 5096 on the configuration access mechanisms. 5097 noaer [PCIE] If the PCIEAER kernel config parameter is 5098 enabled, this kernel boot option can be used to 5099 disable the use of PCIE advanced error reporting. 5100 nodomains [PCI] Disable support for multiple PCI 5101 root domains (aka PCI segments, in ACPI-speak). 5102 nommconf [X86] Disable use of MMCONFIG for PCI 5103 Configuration 5104 check_enable_amd_mmconf [X86] check for and enable 5105 properly configured MMIO access to PCI 5106 config space on AMD family 10h CPU 5107 nomsi [MSI] If the PCI_MSI kernel config parameter is 5108 enabled, this kernel boot option can be used to 5109 disable the use of MSI interrupts system-wide. 5110 noioapicquirk [APIC] Disable all boot interrupt quirks. 5111 Safety option to keep boot IRQs enabled. This 5112 should never be necessary. 5113 ioapicreroute [APIC] Enable rerouting of boot IRQs to the 5114 primary IO-APIC for bridges that cannot disable 5115 boot IRQs. This fixes a source of spurious IRQs 5116 when the system masks IRQs. 5117 noioapicreroute [APIC] Disable workaround that uses the 5118 boot IRQ equivalent of an IRQ that connects to 5119 a chipset where boot IRQs cannot be disabled. 5120 The opposite of ioapicreroute. 5121 biosirq [X86-32] Use PCI BIOS calls to get the interrupt 5122 routing table. These calls are known to be buggy 5123 on several machines and they hang the machine 5124 when used, but on other computers it's the only 5125 way to get the interrupt routing table. Try 5126 this option if the kernel is unable to allocate 5127 IRQs or discover secondary PCI buses on your 5128 motherboard. 5129 rom [X86] Assign address space to expansion ROMs. 5130 Use with caution as certain devices share 5131 address decoders between ROMs and other 5132 resources. 5133 norom [X86] Do not assign address space to 5134 expansion ROMs that do not already have 5135 BIOS assigned address ranges. 5136 nobar [X86] Do not assign address space to the 5137 BARs that weren't assigned by the BIOS. 5138 irqmask=0xMMMM [X86] Set a bit mask of IRQs allowed to be 5139 assigned automatically to PCI devices. You can 5140 make the kernel exclude IRQs of your ISA cards 5141 this way. 5142 pirqaddr=0xAAAAA [X86] Specify the physical address 5143 of the PIRQ table (normally generated 5144 by the BIOS) if it is outside the 5145 F0000h-100000h range. 5146 lastbus=N [X86] Scan all buses thru bus #N. Can be 5147 useful if the kernel is unable to find your 5148 secondary buses and you want to tell it 5149 explicitly which ones they are. 5150 assign-busses [X86] Always assign all PCI bus 5151 numbers ourselves, overriding 5152 whatever the firmware may have done. 5153 usepirqmask [X86] Honor the possible IRQ mask stored 5154 in the BIOS $PIR table. This is needed on 5155 some systems with broken BIOSes, notably 5156 some HP Pavilion N5400 and Omnibook XE3 5157 notebooks. This will have no effect if ACPI 5158 IRQ routing is enabled. 5159 noacpi [X86] Do not use ACPI for IRQ routing 5160 or for PCI scanning. 5161 use_crs [X86] Use PCI host bridge window information 5162 from ACPI. On BIOSes from 2008 or later, this 5163 is enabled by default. If you need to use this, 5164 please report a bug. 5165 nocrs [X86] Ignore PCI host bridge windows from ACPI. 5166 If you need to use this, please report a bug. 5167 use_e820 [X86] Use E820 reservations to exclude parts of 5168 PCI host bridge windows. This is a workaround 5169 for BIOS defects in host bridge _CRS methods. 5170 If you need to use this, please report a bug to 5171 <linux-pci@vger.kernel.org>. 5172 no_e820 [X86] Ignore E820 reservations for PCI host 5173 bridge windows. This is the default on modern 5174 hardware. If you need to use this, please report 5175 a bug to <linux-pci@vger.kernel.org>. 5176 routeirq Do IRQ routing for all PCI devices. 5177 This is normally done in pci_enable_device(), 5178 so this option is a temporary workaround 5179 for broken drivers that don't call it. 5180 skip_isa_align [X86] do not align io start addr, so can 5181 handle more pci cards 5182 noearly [X86] Don't do any early type 1 scanning. 5183 This might help on some broken boards which 5184 machine check when some devices' config space 5185 is read. But various workarounds are disabled 5186 and some IOMMU drivers will not work. 5187 bfsort Sort PCI devices into breadth-first order. 5188 This sorting is done to get a device 5189 order compatible with older (<= 2.4) kernels. 5190 nobfsort Don't sort PCI devices into breadth-first order. 5191 pcie_bus_tune_off Disable PCIe MPS (Max Payload Size) 5192 tuning and use the BIOS-configured MPS defaults. 5193 pcie_bus_safe Set every device's MPS to the largest value 5194 supported by all devices below the root complex. 5195 pcie_bus_perf Set device MPS to the largest allowable MPS 5196 based on its parent bus. Also set MRRS (Max 5197 Read Request Size) to the largest supported 5198 value (no larger than the MPS that the device 5199 or bus can support) for best performance. 5200 pcie_bus_peer2peer Set every device's MPS to 128B, which 5201 every device is guaranteed to support. This 5202 configuration allows peer-to-peer DMA between 5203 any pair of devices, possibly at the cost of 5204 reduced performance. This also guarantees 5205 that hot-added devices will work. 5206 cbiosize=nn[KMG] The fixed amount of bus space which is 5207 reserved for the CardBus bridge's IO window. 5208 The default value is 256 bytes. 5209 cbmemsize=nn[KMG] The fixed amount of bus space which is 5210 reserved for the CardBus bridge's memory 5211 window. The default value is 64 megabytes. 5212 resource_alignment= 5213 Format: 5214 [<order of align>@]<pci_dev>[; ...] 5215 Specifies alignment and device to reassign 5216 aligned memory resources. How to 5217 specify the device is described above. 5218 If <order of align> is not specified, 5219 PAGE_SIZE is used as alignment. 5220 A PCI-PCI bridge can be specified if resource 5221 windows need to be expanded. 5222 To specify the alignment for several 5223 instances of a device, the PCI vendor, 5224 device, subvendor, and subdevice may be 5225 specified, e.g., 12@pci:8086:9c22:103c:198f 5226 for 4096-byte alignment. 5227 ecrc= Enable/disable PCIe ECRC (transaction layer 5228 end-to-end CRC checking). Only effective if 5229 OS has native AER control (either granted by 5230 ACPI _OSC or forced via "pcie_ports=native") 5231 bios: Use BIOS/firmware settings. This is the 5232 the default. 5233 off: Turn ECRC off 5234 on: Turn ECRC on. 5235 hpiosize=nn[KMG] The fixed amount of bus space which is 5236 reserved for hotplug bridge's IO window. 5237 Default size is 256 bytes. 5238 hpmmiosize=nn[KMG] The fixed amount of bus space which is 5239 reserved for hotplug bridge's MMIO window. 5240 Default size is 2 megabytes. 5241 hpmmioprefsize=nn[KMG] The fixed amount of bus space which is 5242 reserved for hotplug bridge's MMIO_PREF window. 5243 Default size is 2 megabytes. 5244 hpmemsize=nn[KMG] The fixed amount of bus space which is 5245 reserved for hotplug bridge's MMIO and 5246 MMIO_PREF window. 5247 Default size is 2 megabytes. 5248 hpbussize=nn The minimum amount of additional bus numbers 5249 reserved for buses below a hotplug bridge. 5250 Default is 1. 5251 realloc= Enable/disable reallocating PCI bridge resources 5252 if allocations done by BIOS are too small to 5253 accommodate resources required by all child 5254 devices. 5255 off: Turn realloc off 5256 on: Turn realloc on 5257 realloc same as realloc=on 5258 noari do not use PCIe ARI. 5259 noats [PCIE, Intel-IOMMU, AMD-IOMMU] 5260 do not use PCIe ATS (and IOMMU device IOTLB). 5261 pcie_scan_all Scan all possible PCIe devices. Otherwise we 5262 only look for one device below a PCIe downstream 5263 port. 5264 big_root_window Try to add a big 64bit memory window to the PCIe 5265 root complex on AMD CPUs. Some GFX hardware 5266 can resize a BAR to allow access to all VRAM. 5267 Adding the window is slightly risky (it may 5268 conflict with unreported devices), so this 5269 taints the kernel. 5270 disable_acs_redir=<pci_dev>[; ...] 5271 Specify one or more PCI devices (in the format 5272 specified above) separated by semicolons. 5273 Each device specified will have the PCI ACS 5274 redirect capabilities forced off which will 5275 allow P2P traffic between devices through 5276 bridges without forcing it upstream. Note: 5277 this removes isolation between devices and 5278 may put more devices in an IOMMU group. 5279 config_acs= 5280 Format: 5281 <ACS flags>@<pci_dev>[; ...] 5282 Specify one or more PCI devices (in the format 5283 specified above) optionally prepended with flags 5284 and separated by semicolons. The respective 5285 capabilities will be enabled, disabled or 5286 unchanged based on what is specified in 5287 flags. 5288 5289 ACS Flags is defined as follows: 5290 bit-0 : ACS Source Validation 5291 bit-1 : ACS Translation Blocking 5292 bit-2 : ACS P2P Request Redirect 5293 bit-3 : ACS P2P Completion Redirect 5294 bit-4 : ACS Upstream Forwarding 5295 bit-5 : ACS P2P Egress Control 5296 bit-6 : ACS Direct Translated P2P 5297 Each bit can be marked as: 5298 '0' – force disabled 5299 '1' – force enabled 5300 'x' – unchanged 5301 For example, 5302 pci=config_acs=10x@pci:0:0 5303 would configure all devices that support 5304 ACS to enable P2P Request Redirect, disable 5305 Translation Blocking, and leave Source 5306 Validation unchanged from whatever power-up 5307 or firmware set it to. 5308 5309 Note: this may remove isolation between devices 5310 and may put more devices in an IOMMU group. 5311 force_floating [S390] Force usage of floating interrupts. 5312 nomio [S390] Do not use MIO instructions. 5313 norid [S390] ignore the RID field and force use of 5314 one PCI domain per PCI function 5315 notph [PCIE] If the PCIE_TPH kernel config parameter 5316 is enabled, this kernel boot option can be used 5317 to disable PCIe TLP Processing Hints support 5318 system-wide. 5319 5320 pcie_aspm= [PCIE] Forcibly enable or ignore PCIe Active State Power 5321 Management. 5322 off Don't touch ASPM configuration at all. Leave any 5323 configuration done by firmware unchanged. 5324 force Enable ASPM even on devices that claim not to support it. 5325 WARNING: Forcing ASPM on may cause system lockups. 5326 5327 pcie_ports= [PCIE] PCIe port services handling: 5328 native Use native PCIe services (PME, AER, DPC, PCIe hotplug) 5329 even if the platform doesn't give the OS permission to 5330 use them. This may cause conflicts if the platform 5331 also tries to use these services. 5332 dpc-native Use native PCIe service for DPC only. May 5333 cause conflicts if firmware uses AER or DPC. 5334 compat Disable native PCIe services (PME, AER, DPC, PCIe 5335 hotplug). 5336 5337 pcie_port_pm= [PCIE] PCIe port power management handling: 5338 off Disable power management of all PCIe ports 5339 force Forcibly enable power management of all PCIe ports 5340 5341 pcie_pme= [PCIE,PM] Native PCIe PME signaling options: 5342 nomsi Do not use MSI for native PCIe PME signaling (this makes 5343 all PCIe root ports use INTx for all services). 5344 5345 pcmv= [HW,PCMCIA] BadgePAD 4 5346 5347 pd_ignore_unused 5348 [PM] 5349 Keep all power-domains already enabled by bootloader on, 5350 even if no driver has claimed them. This is useful 5351 for debug and development, but should not be 5352 needed on a platform with proper driver support. 5353 5354 pdcchassis= [PARISC,HW] Disable/Enable PDC Chassis Status codes at 5355 boot time. 5356 Format: { 0 | 1 } 5357 See arch/parisc/kernel/pdc_chassis.c 5358 5359 percpu_alloc= [MM,EARLY] 5360 Select which percpu first chunk allocator to use. 5361 Currently supported values are "embed" and "page". 5362 Archs may support subset or none of the selections. 5363 See comments in mm/percpu.c for details on each 5364 allocator. This parameter is primarily for debugging 5365 and performance comparison. 5366 5367 pirq= [SMP,APIC] Manual mp-table setup 5368 See Documentation/arch/x86/i386/IO-APIC.rst. 5369 5370 plip= [PPT,NET] Parallel port network link 5371 Format: { parport<nr> | timid | 0 } 5372 See also Documentation/admin-guide/parport.rst. 5373 5374 pmtmr= [X86] Manual setup of pmtmr I/O Port. 5375 Override pmtimer IOPort with a hex value. 5376 e.g. pmtmr=0x508 5377 5378 pmu_override= [PPC] Override the PMU. 5379 This option takes over the PMU facility, so it is no 5380 longer usable by perf. Setting this option starts the 5381 PMU counters by setting MMCR0 to 0 (the FC bit is 5382 cleared). If a number is given, then MMCR1 is set to 5383 that number, otherwise (e.g., 'pmu_override=on'), MMCR1 5384 remains 0. 5385 5386 pm_async= [PM] 5387 Format: off 5388 This parameter sets the initial value of the 5389 /sys/power/pm_async sysfs knob at boot time. 5390 If set to "off", disables asynchronous suspend and 5391 resume of devices during system-wide power transitions. 5392 This can be useful on platforms where device 5393 dependencies are not well-defined, or for debugging 5394 power management issues. Asynchronous operations are 5395 enabled by default. 5396 5397 5398 pm_debug_messages [SUSPEND,KNL] 5399 Enable suspend/resume debug messages during boot up. 5400 5401 pnp.debug=1 [PNP] 5402 Enable PNP debug messages (depends on the 5403 CONFIG_PNP_DEBUG_MESSAGES option). Change at run-time 5404 via /sys/module/pnp/parameters/debug. We always show 5405 current resource usage; turning this on also shows 5406 possible settings and some assignment information. 5407 5408 pnpacpi= [ACPI] 5409 { off } 5410 5411 pnpbios= [ISAPNP] 5412 { on | off | curr | res | no-curr | no-res } 5413 5414 pnp_reserve_irq= 5415 [ISAPNP] Exclude IRQs for the autoconfiguration 5416 5417 pnp_reserve_dma= 5418 [ISAPNP] Exclude DMAs for the autoconfiguration 5419 5420 pnp_reserve_io= [ISAPNP] Exclude I/O ports for the autoconfiguration 5421 Ranges are in pairs (I/O port base and size). 5422 5423 pnp_reserve_mem= 5424 [ISAPNP] Exclude memory regions for the 5425 autoconfiguration. 5426 Ranges are in pairs (memory base and size). 5427 5428 ports= [IP_VS_FTP] IPVS ftp helper module 5429 Default is 21. 5430 Up to 8 (IP_VS_APP_MAX_PORTS) ports 5431 may be specified. 5432 Format: <port>,<port>.... 5433 5434 possible_cpus= [SMP,S390,X86] 5435 Format: <unsigned int> 5436 Set the number of possible CPUs, overriding the 5437 regular discovery mechanisms (such as ACPI/FW, etc). 5438 5439 powersave=off [PPC] This option disables power saving features. 5440 It specifically disables cpuidle and sets the 5441 platform machine description specific power_save 5442 function to NULL. On Idle the CPU just reduces 5443 execution priority. 5444 5445 ppc_strict_facility_enable 5446 [PPC,ENABLE] This option catches any kernel floating point, 5447 Altivec, VSX and SPE outside of regions specifically 5448 allowed (eg kernel_enable_fpu()/kernel_disable_fpu()). 5449 There is some performance impact when enabling this. 5450 5451 ppc_tm= [PPC,EARLY] 5452 Format: {"off"} 5453 Disable Hardware Transactional Memory 5454 5455 preempt= [KNL] 5456 Select preemption mode if you have CONFIG_PREEMPT_DYNAMIC 5457 none - Limited to cond_resched() calls 5458 voluntary - Limited to cond_resched() and might_sleep() calls 5459 full - Any section that isn't explicitly preempt disabled 5460 can be preempted anytime. Tasks will also yield 5461 contended spinlocks (if the critical section isn't 5462 explicitly preempt disabled beyond the lock itself). 5463 lazy - Scheduler controlled. Similar to full but instead 5464 of preempting the task immediately, the task gets 5465 one HZ tick time to yield itself before the 5466 preemption will be forced. One preemption is when the 5467 task returns to user space. 5468 5469 print-fatal-signals= 5470 [KNL] debug: print fatal signals 5471 5472 If enabled, warn about various signal handling 5473 related application anomalies: too many signals, 5474 too many POSIX.1 timers, fatal signals causing a 5475 coredump - etc. 5476 5477 If you hit the warning due to signal overflow, 5478 you might want to try "ulimit -i unlimited". 5479 5480 default: off. 5481 5482 printk.always_kmsg_dump= 5483 Trigger kmsg_dump for cases other than kernel oops or 5484 panics 5485 Format: <bool> (1/Y/y=enable, 0/N/n=disable) 5486 default: disabled 5487 5488 printk.console_no_auto_verbose= 5489 Disable console loglevel raise on oops, panic 5490 or lockdep-detected issues (only if lock debug is on). 5491 With an exception to setups with low baudrate on 5492 serial console, keeping this 0 is a good choice 5493 in order to provide more debug information. 5494 Format: <bool> 5495 default: 0 (auto_verbose is enabled) 5496 5497 printk.debug_non_panic_cpus= 5498 Allows storing messages from non-panic CPUs into 5499 the printk log buffer during panic(). They are 5500 flushed to consoles by the panic-CPU on 5501 a best-effort basis. 5502 Format: <bool> (1/Y/y=enable, 0/N/n=disable) 5503 Default: disabled 5504 5505 printk.devkmsg={on,off,ratelimit} 5506 Control writing to /dev/kmsg. 5507 on - unlimited logging to /dev/kmsg from userspace 5508 off - logging to /dev/kmsg disabled 5509 ratelimit - ratelimit the logging 5510 Default: ratelimit 5511 5512 printk.time= Show timing data prefixed to each printk message line 5513 Format: <bool> (1/Y/y=enable, 0/N/n=disable) 5514 5515 proc_mem.force_override= [KNL] 5516 Format: {always | ptrace | never} 5517 Traditionally /proc/pid/mem allows memory permissions to be 5518 overridden without restrictions. This option may be set to 5519 restrict that. Can be one of: 5520 - 'always': traditional behavior always allows mem overrides. 5521 - 'ptrace': only allow mem overrides for active ptracers. 5522 - 'never': never allow mem overrides. 5523 If not specified, default is the CONFIG_PROC_MEM_* choice. 5524 5525 processor.max_cstate= [HW,ACPI] 5526 Limit processor to maximum C-state 5527 max_cstate=9 overrides any DMI blacklist limit. 5528 5529 processor.nocst [HW,ACPI] 5530 Ignore the _CST method to determine C-states, 5531 instead using the legacy FADT method 5532 5533 profile= [KNL] Enable kernel profiling via /proc/profile 5534 Format: [<profiletype>,]<number> 5535 Param: <profiletype>: "schedule" or "kvm" 5536 [defaults to kernel profiling] 5537 Param: "schedule" - profile schedule points. 5538 Param: "kvm" - profile VM exits. 5539 Param: <number> - step/bucket size as a power of 2 for 5540 statistical time based profiling. 5541 5542 prot_virt= [S390] enable hosting protected virtual machines 5543 isolated from the hypervisor (if hardware supports 5544 that). If enabled, the default kernel base address 5545 might be overridden even when Kernel Address Space 5546 Layout Randomization is disabled. 5547 Format: <bool> 5548 5549 psi= [KNL] Enable or disable pressure stall information 5550 tracking. 5551 Format: <bool> 5552 5553 psmouse.proto= [HW,MOUSE] Highest PS2 mouse protocol extension to 5554 probe for; one of (bare|imps|exps|lifebook|any). 5555 psmouse.rate= [HW,MOUSE] Set desired mouse report rate, in reports 5556 per second. 5557 psmouse.resetafter= [HW,MOUSE] 5558 Try to reset the device after so many bad packets 5559 (0 = never). 5560 psmouse.resolution= 5561 [HW,MOUSE] Set desired mouse resolution, in dpi. 5562 psmouse.smartscroll= 5563 [HW,MOUSE] Controls Logitech smartscroll autorepeat. 5564 0 = disabled, 1 = enabled (default). 5565 5566 pstore.backend= Specify the name of the pstore backend to use 5567 5568 pti= [X86-64] Control Page Table Isolation of user and 5569 kernel address spaces. Disabling this feature 5570 removes hardening, but improves performance of 5571 system calls and interrupts. 5572 5573 on - unconditionally enable 5574 off - unconditionally disable 5575 auto - kernel detects whether your CPU model is 5576 vulnerable to issues that PTI mitigates 5577 5578 Not specifying this option is equivalent to pti=auto. 5579 5580 pty.legacy_count= 5581 [KNL] Number of legacy pty's. Overwrites compiled-in 5582 default number. 5583 5584 quiet [KNL,EARLY] Disable most log messages 5585 5586 r128= [HW,DRM] 5587 5588 radix_hcall_invalidate=on [PPC/PSERIES] 5589 Disable RADIX GTSE feature and use hcall for TLB 5590 invalidate. 5591 5592 raid= [HW,RAID] 5593 See Documentation/admin-guide/md.rst. 5594 5595 ramdisk_size= [RAM] Sizes of RAM disks in kilobytes 5596 See Documentation/admin-guide/blockdev/ramdisk.rst. 5597 5598 ramdisk_start= [Deprecated,RAM] RAM disk image start address 5599 5600 random.trust_cpu=off 5601 [KNL,EARLY] Disable trusting the use of the CPU's 5602 random number generator (if available) to 5603 initialize the kernel's RNG. 5604 5605 random.trust_bootloader=off 5606 [KNL,EARLY] Disable trusting the use of the a seed 5607 passed by the bootloader (if available) to 5608 initialize the kernel's RNG. 5609 5610 randomize_kstack_offset= 5611 [KNL,EARLY] Enable or disable kernel stack offset 5612 randomization, which provides roughly 5 bits of 5613 entropy, frustrating memory corruption attacks 5614 that depend on stack address determinism or 5615 cross-syscall address exposures. This is only 5616 available on architectures that have defined 5617 CONFIG_HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET. 5618 Format: <bool> (1/Y/y=enable, 0/N/n=disable) 5619 Default is CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT. 5620 5621 ras=option[,option,...] [KNL] RAS-specific options 5622 5623 cec_disable [X86] 5624 Disable the Correctable Errors Collector, 5625 see CONFIG_RAS_CEC help text. 5626 5627 rcu_nocbs[=cpu-list] 5628 [KNL] The optional argument is a cpu list, 5629 as described above. 5630 5631 In kernels built with CONFIG_RCU_NOCB_CPU=y, 5632 enable the no-callback CPU mode, which prevents 5633 such CPUs' callbacks from being invoked in 5634 softirq context. Invocation of such CPUs' RCU 5635 callbacks will instead be offloaded to "rcuox/N" 5636 kthreads created for that purpose, where "x" is 5637 "p" for RCU-preempt, "s" for RCU-sched, and "g" 5638 for the kthreads that mediate grace periods; and 5639 "N" is the CPU number. This reduces OS jitter on 5640 the offloaded CPUs, which can be useful for HPC 5641 and real-time workloads. It can also improve 5642 energy efficiency for asymmetric multiprocessors. 5643 5644 If a cpulist is passed as an argument, the specified 5645 list of CPUs is set to no-callback mode from boot. 5646 5647 Otherwise, if the '=' sign and the cpulist 5648 arguments are omitted, no CPU will be set to 5649 no-callback mode from boot but the mode may be 5650 toggled at runtime via cpusets. 5651 5652 Note that this argument takes precedence over 5653 the CONFIG_RCU_NOCB_CPU_DEFAULT_ALL option. 5654 5655 rcu_nocb_poll [KNL] 5656 Rather than requiring that offloaded CPUs 5657 (specified by rcu_nocbs= above) explicitly 5658 awaken the corresponding "rcuoN" kthreads, 5659 make these kthreads poll for callbacks. 5660 This improves the real-time response for the 5661 offloaded CPUs by relieving them of the need to 5662 wake up the corresponding kthread, but degrades 5663 energy efficiency by requiring that the kthreads 5664 periodically wake up to do the polling. 5665 5666 rcutree.blimit= [KNL] 5667 Set maximum number of finished RCU callbacks to 5668 process in one batch. 5669 5670 rcutree.csd_lock_suppress_rcu_stall= [KNL] 5671 Do only a one-line RCU CPU stall warning when 5672 there is an ongoing too-long CSD-lock wait. 5673 5674 rcutree.do_rcu_barrier= [KNL] 5675 Request a call to rcu_barrier(). This is 5676 throttled so that userspace tests can safely 5677 hammer on the sysfs variable if they so choose. 5678 If triggered before the RCU grace-period machinery 5679 is fully active, this will error out with EAGAIN. 5680 5681 rcutree.dump_tree= [KNL] 5682 Dump the structure of the rcu_node combining tree 5683 out at early boot. This is used for diagnostic 5684 purposes, to verify correct tree setup. 5685 5686 rcutree.gp_cleanup_delay= [KNL] 5687 Set the number of jiffies to delay each step of 5688 RCU grace-period cleanup. 5689 5690 rcutree.gp_init_delay= [KNL] 5691 Set the number of jiffies to delay each step of 5692 RCU grace-period initialization. 5693 5694 rcutree.gp_preinit_delay= [KNL] 5695 Set the number of jiffies to delay each step of 5696 RCU grace-period pre-initialization, that is, 5697 the propagation of recent CPU-hotplug changes up 5698 the rcu_node combining tree. 5699 5700 rcutree.jiffies_till_first_fqs= [KNL] 5701 Set delay from grace-period initialization to 5702 first attempt to force quiescent states. 5703 Units are jiffies, minimum value is zero, 5704 and maximum value is HZ. 5705 5706 rcutree.jiffies_till_next_fqs= [KNL] 5707 Set delay between subsequent attempts to force 5708 quiescent states. Units are jiffies, minimum 5709 value is one, and maximum value is HZ. 5710 5711 rcutree.jiffies_till_sched_qs= [KNL] 5712 Set required age in jiffies for a 5713 given grace period before RCU starts 5714 soliciting quiescent-state help from 5715 rcu_note_context_switch() and cond_resched(). 5716 If not specified, the kernel will calculate 5717 a value based on the most recent settings 5718 of rcutree.jiffies_till_first_fqs 5719 and rcutree.jiffies_till_next_fqs. 5720 This calculated value may be viewed in 5721 rcutree.jiffies_to_sched_qs. Any attempt to set 5722 rcutree.jiffies_to_sched_qs will be cheerfully 5723 overwritten. 5724 5725 rcutree.kthread_prio= [KNL,BOOT] 5726 Set the SCHED_FIFO priority of the RCU per-CPU 5727 kthreads (rcuc/N). This value is also used for 5728 the priority of the RCU boost threads (rcub/N) 5729 and for the RCU grace-period kthreads (rcu_bh, 5730 rcu_preempt, and rcu_sched). If RCU_BOOST is 5731 set, valid values are 1-99 and the default is 1 5732 (the least-favored priority). Otherwise, when 5733 RCU_BOOST is not set, valid values are 0-99 and 5734 the default is zero (non-realtime operation). 5735 When RCU_NOCB_CPU is set, also adjust the 5736 priority of NOCB callback kthreads. 5737 5738 rcutree.nocb_nobypass_lim_per_jiffy= [KNL] 5739 On callback-offloaded (rcu_nocbs) CPUs, 5740 RCU reduces the lock contention that would 5741 otherwise be caused by callback floods through 5742 use of the ->nocb_bypass list. However, in the 5743 common non-flooded case, RCU queues directly to 5744 the main ->cblist in order to avoid the extra 5745 overhead of the ->nocb_bypass list and its lock. 5746 But if there are too many callbacks queued during 5747 a single jiffy, RCU pre-queues the callbacks into 5748 the ->nocb_bypass queue. The definition of "too 5749 many" is supplied by this kernel boot parameter. 5750 5751 rcutree.nohz_full_patience_delay= [KNL] 5752 On callback-offloaded (rcu_nocbs) CPUs, avoid 5753 disturbing RCU unless the grace period has 5754 reached the specified age in milliseconds. 5755 Defaults to zero. Large values will be capped 5756 at five seconds. All values will be rounded down 5757 to the nearest value representable by jiffies. 5758 5759 rcutree.qhimark= [KNL] 5760 Set threshold of queued RCU callbacks beyond which 5761 batch limiting is disabled. 5762 5763 rcutree.qlowmark= [KNL] 5764 Set threshold of queued RCU callbacks below which 5765 batch limiting is re-enabled. 5766 5767 rcutree.qovld= [KNL] 5768 Set threshold of queued RCU callbacks beyond which 5769 RCU's force-quiescent-state scan will aggressively 5770 enlist help from cond_resched() and sched IPIs to 5771 help CPUs more quickly reach quiescent states. 5772 Set to less than zero to make this be set based 5773 on rcutree.qhimark at boot time and to zero to 5774 disable more aggressive help enlistment. 5775 5776 rcutree.rcu_delay_page_cache_fill_msec= [KNL] 5777 Set the page-cache refill delay (in milliseconds) 5778 in response to low-memory conditions. The range 5779 of permitted values is in the range 0:100000. 5780 5781 rcutree.rcu_divisor= [KNL] 5782 Set the shift-right count to use to compute 5783 the callback-invocation batch limit bl from 5784 the number of callbacks queued on this CPU. 5785 The result will be bounded below by the value of 5786 the rcutree.blimit kernel parameter. Every bl 5787 callbacks, the softirq handler will exit in 5788 order to allow the CPU to do other work. 5789 5790 Please note that this callback-invocation batch 5791 limit applies only to non-offloaded callback 5792 invocation. Offloaded callbacks are instead 5793 invoked in the context of an rcuoc kthread, which 5794 scheduler will preempt as it does any other task. 5795 5796 rcutree.rcu_fanout_exact= [KNL] 5797 Disable autobalancing of the rcu_node combining 5798 tree. This is used by rcutorture, and might 5799 possibly be useful for architectures having high 5800 cache-to-cache transfer latencies. 5801 5802 rcutree.rcu_fanout_leaf= [KNL] 5803 Change the number of CPUs assigned to each 5804 leaf rcu_node structure. Useful for very 5805 large systems, which will choose the value 64, 5806 and for NUMA systems with large remote-access 5807 latencies, which will choose a value aligned 5808 with the appropriate hardware boundaries. 5809 5810 rcutree.rcu_min_cached_objs= [KNL] 5811 Minimum number of objects which are cached and 5812 maintained per one CPU. Object size is equal 5813 to PAGE_SIZE. The cache allows to reduce the 5814 pressure to page allocator, also it makes the 5815 whole algorithm to behave better in low memory 5816 condition. 5817 5818 rcutree.rcu_nocb_gp_stride= [KNL] 5819 Set the number of NOCB callback kthreads in 5820 each group, which defaults to the square root 5821 of the number of CPUs. Larger numbers reduce 5822 the wakeup overhead on the global grace-period 5823 kthread, but increases that same overhead on 5824 each group's NOCB grace-period kthread. 5825 5826 rcutree.rcu_kick_kthreads= [KNL] 5827 Cause the grace-period kthread to get an extra 5828 wake_up() if it sleeps three times longer than 5829 it should at force-quiescent-state time. 5830 This wake_up() will be accompanied by a 5831 WARN_ONCE() splat and an ftrace_dump(). 5832 5833 rcutree.rcu_resched_ns= [KNL] 5834 Limit the time spend invoking a batch of RCU 5835 callbacks to the specified number of nanoseconds. 5836 By default, this limit is checked only once 5837 every 32 callbacks in order to limit the pain 5838 inflicted by local_clock() overhead. 5839 5840 rcutree.rcu_unlock_delay= [KNL] 5841 In CONFIG_RCU_STRICT_GRACE_PERIOD=y kernels, 5842 this specifies an rcu_read_unlock()-time delay 5843 in microseconds. This defaults to zero. 5844 Larger delays increase the probability of 5845 catching RCU pointer leaks, that is, buggy use 5846 of RCU-protected pointers after the relevant 5847 rcu_read_unlock() has completed. 5848 5849 rcutree.sysrq_rcu= [KNL] 5850 Commandeer a sysrq key to dump out Tree RCU's 5851 rcu_node tree with an eye towards determining 5852 why a new grace period has not yet started. 5853 5854 rcutree.use_softirq= [KNL] 5855 If set to zero, move all RCU_SOFTIRQ processing to 5856 per-CPU rcuc kthreads. Defaults to a non-zero 5857 value, meaning that RCU_SOFTIRQ is used by default. 5858 Specify rcutree.use_softirq=0 to use rcuc kthreads. 5859 5860 But note that CONFIG_PREEMPT_RT=y kernels disable 5861 this kernel boot parameter, forcibly setting it 5862 to zero. 5863 5864 rcutree.enable_rcu_lazy= [KNL] 5865 To save power, batch RCU callbacks and flush after 5866 delay, memory pressure or callback list growing too 5867 big. 5868 5869 rcutree.rcu_normal_wake_from_gp= [KNL] 5870 Reduces a latency of synchronize_rcu() call. This approach 5871 maintains its own track of synchronize_rcu() callers, so it 5872 does not interact with regular callbacks because it does not 5873 use a call_rcu[_hurry]() path. Please note, this is for a 5874 normal grace period. 5875 5876 How to enable it: 5877 5878 echo 1 > /sys/module/rcutree/parameters/rcu_normal_wake_from_gp 5879 or pass a boot parameter "rcutree.rcu_normal_wake_from_gp=1" 5880 5881 Default is 1 if num_possible_cpus() <= 16 and it is not explicitly 5882 disabled by the boot parameter passing 0. 5883 5884 rcuscale.gp_async= [KNL] 5885 Measure performance of asynchronous 5886 grace-period primitives such as call_rcu(). 5887 5888 rcuscale.gp_async_max= [KNL] 5889 Specify the maximum number of outstanding 5890 callbacks per writer thread. When a writer 5891 thread exceeds this limit, it invokes the 5892 corresponding flavor of rcu_barrier() to allow 5893 previously posted callbacks to drain. 5894 5895 rcuscale.gp_exp= [KNL] 5896 Measure performance of expedited synchronous 5897 grace-period primitives. 5898 5899 rcuscale.holdoff= [KNL] 5900 Set test-start holdoff period. The purpose of 5901 this parameter is to delay the start of the 5902 test until boot completes in order to avoid 5903 interference. 5904 5905 rcuscale.kfree_by_call_rcu= [KNL] 5906 In kernels built with CONFIG_RCU_LAZY=y, test 5907 call_rcu() instead of kfree_rcu(). 5908 5909 rcuscale.kfree_mult= [KNL] 5910 Instead of allocating an object of size kfree_obj, 5911 allocate one of kfree_mult * sizeof(kfree_obj). 5912 Defaults to 1. 5913 5914 rcuscale.kfree_rcu_test= [KNL] 5915 Set to measure performance of kfree_rcu() flooding. 5916 5917 rcuscale.kfree_rcu_test_double= [KNL] 5918 Test the double-argument variant of kfree_rcu(). 5919 If this parameter has the same value as 5920 rcuscale.kfree_rcu_test_single, both the single- 5921 and double-argument variants are tested. 5922 5923 rcuscale.kfree_rcu_test_single= [KNL] 5924 Test the single-argument variant of kfree_rcu(). 5925 If this parameter has the same value as 5926 rcuscale.kfree_rcu_test_double, both the single- 5927 and double-argument variants are tested. 5928 5929 rcuscale.kfree_nthreads= [KNL] 5930 The number of threads running loops of kfree_rcu(). 5931 5932 rcuscale.kfree_alloc_num= [KNL] 5933 Number of allocations and frees done in an iteration. 5934 5935 rcuscale.kfree_loops= [KNL] 5936 Number of loops doing rcuscale.kfree_alloc_num number 5937 of allocations and frees. 5938 5939 rcuscale.minruntime= [KNL] 5940 Set the minimum test run time in seconds. This 5941 does not affect the data-collection interval, 5942 but instead allows better measurement of things 5943 like CPU consumption. 5944 5945 rcuscale.nreaders= [KNL] 5946 Set number of RCU readers. The value -1 selects 5947 N, where N is the number of CPUs. A value 5948 "n" less than -1 selects N-n+1, where N is again 5949 the number of CPUs. For example, -2 selects N 5950 (the number of CPUs), -3 selects N+1, and so on. 5951 A value of "n" less than or equal to -N selects 5952 a single reader. 5953 5954 rcuscale.nwriters= [KNL] 5955 Set number of RCU writers. The values operate 5956 the same as for rcuscale.nreaders. 5957 N, where N is the number of CPUs 5958 5959 rcuscale.scale_type= [KNL] 5960 Specify the RCU implementation to test. 5961 5962 rcuscale.shutdown= [KNL] 5963 Shut the system down after performance tests 5964 complete. This is useful for hands-off automated 5965 testing. 5966 5967 rcuscale.verbose= [KNL] 5968 Enable additional printk() statements. 5969 5970 rcuscale.writer_holdoff= [KNL] 5971 Write-side holdoff between grace periods, 5972 in microseconds. The default of zero says 5973 no holdoff. 5974 5975 rcuscale.writer_holdoff_jiffies= [KNL] 5976 Additional write-side holdoff between grace 5977 periods, but in jiffies. The default of zero 5978 says no holdoff. 5979 5980 rcutorture.fqs_duration= [KNL] 5981 Set duration of force_quiescent_state bursts 5982 in microseconds. 5983 5984 rcutorture.fqs_holdoff= [KNL] 5985 Set holdoff time within force_quiescent_state bursts 5986 in microseconds. 5987 5988 rcutorture.fqs_stutter= [KNL] 5989 Set wait time between force_quiescent_state bursts 5990 in seconds. 5991 5992 rcutorture.fwd_progress= [KNL] 5993 Specifies the number of kthreads to be used 5994 for RCU grace-period forward-progress testing 5995 for the types of RCU supporting this notion. 5996 Defaults to 1 kthread, values less than zero or 5997 greater than the number of CPUs cause the number 5998 of CPUs to be used. 5999 6000 rcutorture.fwd_progress_div= [KNL] 6001 Specify the fraction of a CPU-stall-warning 6002 period to do tight-loop forward-progress testing. 6003 6004 rcutorture.fwd_progress_holdoff= [KNL] 6005 Number of seconds to wait between successive 6006 forward-progress tests. 6007 6008 rcutorture.fwd_progress_need_resched= [KNL] 6009 Enclose cond_resched() calls within checks for 6010 need_resched() during tight-loop forward-progress 6011 testing. 6012 6013 rcutorture.gp_cond= [KNL] 6014 Use conditional/asynchronous update-side 6015 normal-grace-period primitives, if available. 6016 6017 rcutorture.gp_cond_exp= [KNL] 6018 Use conditional/asynchronous update-side 6019 expedited-grace-period primitives, if available. 6020 6021 rcutorture.gp_cond_full= [KNL] 6022 Use conditional/asynchronous update-side 6023 normal-grace-period primitives that also take 6024 concurrent expedited grace periods into account, 6025 if available. 6026 6027 rcutorture.gp_cond_exp_full= [KNL] 6028 Use conditional/asynchronous update-side 6029 expedited-grace-period primitives that also take 6030 concurrent normal grace periods into account, 6031 if available. 6032 6033 rcutorture.gp_cond_wi= [KNL] 6034 Nominal wait interval for normal conditional 6035 grace periods (specified by rcutorture's 6036 gp_cond and gp_cond_full module parameters), 6037 in microseconds. The actual wait interval will 6038 be randomly selected to nanosecond granularity up 6039 to this wait interval. Defaults to 16 jiffies, 6040 for example, 16,000 microseconds on a system 6041 with HZ=1000. 6042 6043 rcutorture.gp_cond_wi_exp= [KNL] 6044 Nominal wait interval for expedited conditional 6045 grace periods (specified by rcutorture's 6046 gp_cond_exp and gp_cond_exp_full module 6047 parameters), in microseconds. The actual wait 6048 interval will be randomly selected to nanosecond 6049 granularity up to this wait interval. Defaults to 6050 128 microseconds. 6051 6052 rcutorture.gp_exp= [KNL] 6053 Use expedited update-side primitives, if available. 6054 6055 rcutorture.gp_normal= [KNL] 6056 Use normal (non-expedited) asynchronous 6057 update-side primitives, if available. 6058 6059 rcutorture.gp_poll= [KNL] 6060 Use polled update-side normal-grace-period 6061 primitives, if available. 6062 6063 rcutorture.gp_poll_exp= [KNL] 6064 Use polled update-side expedited-grace-period 6065 primitives, if available. 6066 6067 rcutorture.gp_poll_full= [KNL] 6068 Use polled update-side normal-grace-period 6069 primitives that also take concurrent expedited 6070 grace periods into account, if available. 6071 6072 rcutorture.gp_poll_exp_full= [KNL] 6073 Use polled update-side expedited-grace-period 6074 primitives that also take concurrent normal 6075 grace periods into account, if available. 6076 6077 rcutorture.gp_poll_wi= [KNL] 6078 Nominal wait interval for normal conditional 6079 grace periods (specified by rcutorture's 6080 gp_poll and gp_poll_full module parameters), 6081 in microseconds. The actual wait interval will 6082 be randomly selected to nanosecond granularity up 6083 to this wait interval. Defaults to 16 jiffies, 6084 for example, 16,000 microseconds on a system 6085 with HZ=1000. 6086 6087 rcutorture.gp_poll_wi_exp= [KNL] 6088 Nominal wait interval for expedited conditional 6089 grace periods (specified by rcutorture's 6090 gp_poll_exp and gp_poll_exp_full module 6091 parameters), in microseconds. The actual wait 6092 interval will be randomly selected to nanosecond 6093 granularity up to this wait interval. Defaults to 6094 128 microseconds. 6095 6096 rcutorture.gp_sync= [KNL] 6097 Use normal (non-expedited) synchronous 6098 update-side primitives, if available. If all 6099 of rcutorture.gp_cond=, rcutorture.gp_exp=, 6100 rcutorture.gp_normal=, and rcutorture.gp_sync= 6101 are zero, rcutorture acts as if is interpreted 6102 they are all non-zero. 6103 6104 rcutorture.gpwrap_lag= [KNL] 6105 Enable grace-period wrap lag testing. Setting 6106 to false prevents the gpwrap lag test from 6107 running. Default is true. 6108 6109 rcutorture.gpwrap_lag_gps= [KNL] 6110 Set the value for grace-period wrap lag during 6111 active lag testing periods. This controls how many 6112 grace periods differences we tolerate between 6113 rdp and rnp's gp_seq before setting overflow flag. 6114 The default is always set to 8. 6115 6116 rcutorture.gpwrap_lag_cycle_mins= [KNL] 6117 Set the total cycle duration for gpwrap lag 6118 testing in minutes. This is the total time for 6119 one complete cycle of active and inactive 6120 testing periods. Default is 30 minutes. 6121 6122 rcutorture.gpwrap_lag_active_mins= [KNL] 6123 Set the duration for which gpwrap lag is active 6124 within each cycle, in minutes. During this time, 6125 the grace-period wrap lag will be set to the 6126 value specified by gpwrap_lag_gps. Default is 6127 5 minutes. 6128 6129 rcutorture.irqreader= [KNL] 6130 Run RCU readers from irq handlers, or, more 6131 accurately, from a timer handler. Not all RCU 6132 flavors take kindly to this sort of thing. 6133 6134 rcutorture.leakpointer= [KNL] 6135 Leak an RCU-protected pointer out of the reader. 6136 This can of course result in splats, and is 6137 intended to test the ability of things like 6138 CONFIG_RCU_STRICT_GRACE_PERIOD=y to detect 6139 such leaks. 6140 6141 rcutorture.n_barrier_cbs= [KNL] 6142 Set callbacks/threads for rcu_barrier() testing. 6143 6144 rcutorture.nfakewriters= [KNL] 6145 Set number of concurrent RCU writers. These just 6146 stress RCU, they don't participate in the actual 6147 test, hence the "fake". 6148 6149 rcutorture.nocbs_nthreads= [KNL] 6150 Set number of RCU callback-offload togglers. 6151 Zero (the default) disables toggling. 6152 6153 rcutorture.nocbs_toggle= [KNL] 6154 Set the delay in milliseconds between successive 6155 callback-offload toggling attempts. 6156 6157 rcutorture.nreaders= [KNL] 6158 Set number of RCU readers. The value -1 selects 6159 N-1, where N is the number of CPUs. A value 6160 "n" less than -1 selects N-n-2, where N is again 6161 the number of CPUs. For example, -2 selects N 6162 (the number of CPUs), -3 selects N+1, and so on. 6163 6164 rcutorture.object_debug= [KNL] 6165 Enable debug-object double-call_rcu() testing. 6166 6167 rcutorture.onoff_holdoff= [KNL] 6168 Set time (s) after boot for CPU-hotplug testing. 6169 6170 rcutorture.onoff_interval= [KNL] 6171 Set time (jiffies) between CPU-hotplug operations, 6172 or zero to disable CPU-hotplug testing. 6173 6174 rcutorture.preempt_duration= [KNL] 6175 Set duration (in milliseconds) of preemptions 6176 by a high-priority FIFO real-time task. Set to 6177 zero (the default) to disable. The CPUs to 6178 preempt are selected randomly from the set that 6179 are online at a given point in time. Races with 6180 CPUs going offline are ignored, with that attempt 6181 at preemption skipped. 6182 6183 rcutorture.preempt_interval= [KNL] 6184 Set interval (in milliseconds, defaulting to one 6185 second) between preemptions by a high-priority 6186 FIFO real-time task. This delay is mediated 6187 by an hrtimer and is further fuzzed to avoid 6188 inadvertent synchronizations. 6189 6190 rcutorture.read_exit_burst= [KNL] 6191 The number of times in a given read-then-exit 6192 episode that a set of read-then-exit kthreads 6193 is spawned. 6194 6195 rcutorture.read_exit_delay= [KNL] 6196 The delay, in seconds, between successive 6197 read-then-exit testing episodes. 6198 6199 rcutorture.reader_flavor= [KNL] 6200 A bit mask indicating which readers to use. 6201 If there is more than one bit set, the readers 6202 are entered from low-order bit up, and are 6203 exited in the opposite order. For SRCU, the 6204 0x1 bit is normal readers, 0x2 NMI-safe readers, 6205 and 0x4 light-weight readers. 6206 6207 rcutorture.shuffle_interval= [KNL] 6208 Set task-shuffle interval (s). Shuffling tasks 6209 allows some CPUs to go into dyntick-idle mode 6210 during the rcutorture test. 6211 6212 rcutorture.shutdown_secs= [KNL] 6213 Set time (s) after boot system shutdown. This 6214 is useful for hands-off automated testing. 6215 6216 rcutorture.stall_cpu= [KNL] 6217 Duration of CPU stall (s) to test RCU CPU stall 6218 warnings, zero to disable. 6219 6220 rcutorture.stall_cpu_block= [KNL] 6221 Sleep while stalling if set. This will result 6222 in warnings from preemptible RCU in addition to 6223 any other stall-related activity. Note that 6224 in kernels built with CONFIG_PREEMPTION=n and 6225 CONFIG_PREEMPT_COUNT=y, this parameter will 6226 cause the CPU to pass through a quiescent state. 6227 Given CONFIG_PREEMPTION=n, this will suppress 6228 RCU CPU stall warnings, but will instead result 6229 in scheduling-while-atomic splats. 6230 6231 Use of this module parameter results in splats. 6232 6233 6234 rcutorture.stall_cpu_holdoff= [KNL] 6235 Time to wait (s) after boot before inducing stall. 6236 6237 rcutorture.stall_cpu_irqsoff= [KNL] 6238 Disable interrupts while stalling if set, but only 6239 on the first stall in the set. 6240 6241 rcutorture.stall_cpu_repeat= [KNL] 6242 Number of times to repeat the stall sequence, 6243 so that rcutorture.stall_cpu_repeat=3 will result 6244 in four stall sequences. 6245 6246 rcutorture.stall_gp_kthread= [KNL] 6247 Duration (s) of forced sleep within RCU 6248 grace-period kthread to test RCU CPU stall 6249 warnings, zero to disable. If both stall_cpu 6250 and stall_gp_kthread are specified, the 6251 kthread is starved first, then the CPU. 6252 6253 rcutorture.stat_interval= [KNL] 6254 Time (s) between statistics printk()s. 6255 6256 rcutorture.stutter= [KNL] 6257 Time (s) to stutter testing, for example, specifying 6258 five seconds causes the test to run for five seconds, 6259 wait for five seconds, and so on. This tests RCU's 6260 ability to transition abruptly to and from idle. 6261 6262 rcutorture.test_boost= [KNL] 6263 Test RCU priority boosting? 0=no, 1=maybe, 2=yes. 6264 "Maybe" means test if the RCU implementation 6265 under test support RCU priority boosting. 6266 6267 rcutorture.test_boost_duration= [KNL] 6268 Duration (s) of each individual boost test. 6269 6270 rcutorture.test_boost_holdoff= [KNL] 6271 Holdoff time (s) from start of test to the start 6272 of RCU priority-boost testing. Defaults to zero, 6273 that is, no holdoff. 6274 6275 rcutorture.test_boost_interval= [KNL] 6276 Interval (s) between each boost test. 6277 6278 rcutorture.test_no_idle_hz= [KNL] 6279 Test RCU's dyntick-idle handling. See also the 6280 rcutorture.shuffle_interval parameter. 6281 6282 rcutorture.torture_type= [KNL] 6283 Specify the RCU implementation to test. 6284 6285 rcutorture.verbose= [KNL] 6286 Enable additional printk() statements. 6287 6288 rcupdate.rcu_cpu_stall_ftrace_dump= [KNL] 6289 Dump ftrace buffer after reporting RCU CPU 6290 stall warning. 6291 6292 rcupdate.rcu_cpu_stall_notifiers= [KNL] 6293 Provide RCU CPU stall notifiers, but see the 6294 warnings in the RCU_CPU_STALL_NOTIFIER Kconfig 6295 option's help text. TL;DR: You almost certainly 6296 do not want rcupdate.rcu_cpu_stall_notifiers. 6297 6298 rcupdate.rcu_cpu_stall_suppress= [KNL] 6299 Suppress RCU CPU stall warning messages. 6300 6301 rcupdate.rcu_cpu_stall_suppress_at_boot= [KNL] 6302 Suppress RCU CPU stall warning messages and 6303 rcutorture writer stall warnings that occur 6304 during early boot, that is, during the time 6305 before the init task is spawned. 6306 6307 rcupdate.rcu_cpu_stall_timeout= [KNL] 6308 Set timeout for RCU CPU stall warning messages. 6309 The value is in seconds and the maximum allowed 6310 value is 300 seconds. 6311 6312 rcupdate.rcu_exp_cpu_stall_timeout= [KNL] 6313 Set timeout for expedited RCU CPU stall warning 6314 messages. The value is in milliseconds 6315 and the maximum allowed value is 21000 6316 milliseconds. Please note that this value is 6317 adjusted to an arch timer tick resolution. 6318 Setting this to zero causes the value from 6319 rcupdate.rcu_cpu_stall_timeout to be used (after 6320 conversion from seconds to milliseconds). 6321 6322 rcupdate.rcu_cpu_stall_cputime= [KNL] 6323 Provide statistics on the cputime and count of 6324 interrupts and tasks during the sampling period. For 6325 multiple continuous RCU stalls, all sampling periods 6326 begin at half of the first RCU stall timeout. 6327 6328 rcupdate.rcu_exp_stall_task_details= [KNL] 6329 Print stack dumps of any tasks blocking the 6330 current expedited RCU grace period during an 6331 expedited RCU CPU stall warning. 6332 6333 rcupdate.rcu_expedited= [KNL] 6334 Use expedited grace-period primitives, for 6335 example, synchronize_rcu_expedited() instead 6336 of synchronize_rcu(). This reduces latency, 6337 but can increase CPU utilization, degrade 6338 real-time latency, and degrade energy efficiency. 6339 No effect on CONFIG_TINY_RCU kernels. 6340 6341 rcupdate.rcu_normal= [KNL] 6342 Use only normal grace-period primitives, 6343 for example, synchronize_rcu() instead of 6344 synchronize_rcu_expedited(). This improves 6345 real-time latency, CPU utilization, and 6346 energy efficiency, but can expose users to 6347 increased grace-period latency. This parameter 6348 overrides rcupdate.rcu_expedited. No effect on 6349 CONFIG_TINY_RCU kernels. 6350 6351 rcupdate.rcu_normal_after_boot= [KNL] 6352 Once boot has completed (that is, after 6353 rcu_end_inkernel_boot() has been invoked), use 6354 only normal grace-period primitives. No effect 6355 on CONFIG_TINY_RCU kernels. 6356 6357 But note that CONFIG_PREEMPT_RT=y kernels enables 6358 this kernel boot parameter, forcibly setting 6359 it to the value one, that is, converting any 6360 post-boot attempt at an expedited RCU grace 6361 period to instead use normal non-expedited 6362 grace-period processing. 6363 6364 rcupdate.rcu_task_collapse_lim= [KNL] 6365 Set the maximum number of callbacks present 6366 at the beginning of a grace period that allows 6367 the RCU Tasks flavors to collapse back to using 6368 a single callback queue. This switching only 6369 occurs when rcupdate.rcu_task_enqueue_lim is 6370 set to the default value of -1. 6371 6372 rcupdate.rcu_task_contend_lim= [KNL] 6373 Set the minimum number of callback-queuing-time 6374 lock-contention events per jiffy required to 6375 cause the RCU Tasks flavors to switch to per-CPU 6376 callback queuing. This switching only occurs 6377 when rcupdate.rcu_task_enqueue_lim is set to 6378 the default value of -1. 6379 6380 rcupdate.rcu_task_enqueue_lim= [KNL] 6381 Set the number of callback queues to use for the 6382 RCU Tasks family of RCU flavors. The default 6383 of -1 allows this to be automatically (and 6384 dynamically) adjusted. This parameter is intended 6385 for use in testing. 6386 6387 rcupdate.rcu_task_lazy_lim= [KNL] 6388 Number of callbacks on a given CPU that will 6389 cancel laziness on that CPU. Use -1 to disable 6390 cancellation of laziness, but be advised that 6391 doing so increases the danger of OOM due to 6392 callback flooding. 6393 6394 rcupdate.rcu_task_stall_info= [KNL] 6395 Set initial timeout in jiffies for RCU task stall 6396 informational messages, which give some indication 6397 of the problem for those not patient enough to 6398 wait for ten minutes. Informational messages are 6399 only printed prior to the stall-warning message 6400 for a given grace period. Disable with a value 6401 less than or equal to zero. Defaults to ten 6402 seconds. A change in value does not take effect 6403 until the beginning of the next grace period. 6404 6405 rcupdate.rcu_task_stall_info_mult= [KNL] 6406 Multiplier for time interval between successive 6407 RCU task stall informational messages for a given 6408 RCU tasks grace period. This value is clamped 6409 to one through ten, inclusive. It defaults to 6410 the value three, so that the first informational 6411 message is printed 10 seconds into the grace 6412 period, the second at 40 seconds, the third at 6413 160 seconds, and then the stall warning at 600 6414 seconds would prevent a fourth at 640 seconds. 6415 6416 rcupdate.rcu_task_stall_timeout= [KNL] 6417 Set timeout in jiffies for RCU task stall 6418 warning messages. Disable with a value less 6419 than or equal to zero. Defaults to ten minutes. 6420 A change in value does not take effect until 6421 the beginning of the next grace period. 6422 6423 rcupdate.rcu_tasks_lazy_ms= [KNL] 6424 Set timeout in milliseconds RCU Tasks asynchronous 6425 callback batching for call_rcu_tasks(). 6426 A negative value will take the default. A value 6427 of zero will disable batching. Batching is 6428 always disabled for synchronize_rcu_tasks(). 6429 6430 rcupdate.rcu_self_test= [KNL] 6431 Run the RCU early boot self tests 6432 6433 rdinit= [KNL] 6434 Format: <full_path> 6435 Run specified binary instead of /init from the ramdisk, 6436 used for early userspace startup. See initrd. 6437 6438 rdrand= [X86,EARLY] 6439 force - Override the decision by the kernel to hide the 6440 advertisement of RDRAND support (this affects 6441 certain AMD processors because of buggy BIOS 6442 support, specifically around the suspend/resume 6443 path). 6444 6445 rdt= [HW,X86,RDT] 6446 Turn on/off individual RDT features. List is: 6447 cmt, mbmtotal, mbmlocal, l3cat, l3cdp, l2cat, l2cdp, 6448 mba, smba, bmec, abmc, sdciae, energy[:guid], 6449 perf[:guid]. 6450 E.g. to turn on cmt and turn off mba use: 6451 rdt=cmt,!mba 6452 To turn off all energy telemetry monitoring and ensure that 6453 perf telemetry monitoring associated with guid 0x12345 6454 is enabled use: 6455 rdt=!energy,perf:0x12345 6456 6457 reboot= [KNL] 6458 Format (x86 or x86_64): 6459 [w[arm] | c[old] | h[ard] | s[oft] | g[pio]] | d[efault] \ 6460 [[,]s[mp]#### \ 6461 [[,]b[ios] | a[cpi] | k[bd] | t[riple] | e[fi] | p[ci]] \ 6462 [[,]f[orce] 6463 Where reboot_mode is one of warm (soft) or cold (hard) or gpio 6464 (prefix with 'panic_' to set mode for panic 6465 reboot only), 6466 reboot_type is one of bios, acpi, kbd, triple, efi, or pci, 6467 reboot_force is either force or not specified, 6468 reboot_cpu is s[mp]#### with #### being the processor 6469 to be used for rebooting. 6470 6471 acpi 6472 Use the ACPI RESET_REG in the FADT. If ACPI is not 6473 configured or the ACPI reset does not work, the reboot 6474 path attempts the reset using the keyboard controller. 6475 6476 bios 6477 Use the CPU reboot vector for warm reset 6478 6479 cold 6480 Set the cold reboot flag 6481 6482 default 6483 There are some built-in platform specific "quirks" 6484 - you may see: "reboot: <name> series board detected. 6485 Selecting <type> for reboots." In the case where you 6486 think the quirk is in error (e.g. you have newer BIOS, 6487 or newer board) using this option will ignore the 6488 built-in quirk table, and use the generic default 6489 reboot actions. 6490 6491 efi 6492 Use efi reset_system runtime service. If EFI is not 6493 configured or the EFI reset does not work, the reboot 6494 path attempts the reset using the keyboard controller. 6495 6496 force 6497 Don't stop other CPUs on reboot. This can make reboot 6498 more reliable in some cases. 6499 6500 kbd 6501 Use the keyboard controller. cold reset (default) 6502 6503 pci 6504 Use a write to the PCI config space register 0xcf9 to 6505 trigger reboot. 6506 6507 triple 6508 Force a triple fault (init) 6509 6510 warm 6511 Don't set the cold reboot flag 6512 6513 Using warm reset will be much faster especially on big 6514 memory systems because the BIOS will not go through 6515 the memory check. Disadvantage is that not all 6516 hardware will be completely reinitialized on reboot so 6517 there may be boot problems on some systems. 6518 6519 6520 refscale.holdoff= [KNL] 6521 Set test-start holdoff period. The purpose of 6522 this parameter is to delay the start of the 6523 test until boot completes in order to avoid 6524 interference. 6525 6526 refscale.lookup_instances= [KNL] 6527 Number of data elements to use for the forms of 6528 SLAB_TYPESAFE_BY_RCU testing. A negative number 6529 is negated and multiplied by nr_cpu_ids, while 6530 zero specifies nr_cpu_ids. 6531 6532 refscale.loops= [KNL] 6533 Set the number of loops over the synchronization 6534 primitive under test. Increasing this number 6535 reduces noise due to loop start/end overhead, 6536 but the default has already reduced the per-pass 6537 noise to a handful of picoseconds on ca. 2020 6538 x86 laptops. 6539 6540 refscale.nreaders= [KNL] 6541 Set number of readers. The default value of -1 6542 selects N, where N is roughly 75% of the number 6543 of CPUs. A value of zero is an interesting choice. 6544 6545 refscale.nruns= [KNL] 6546 Set number of runs, each of which is dumped onto 6547 the console log. 6548 6549 refscale.readdelay= [KNL] 6550 Set the read-side critical-section duration, 6551 measured in microseconds. 6552 6553 refscale.scale_type= [KNL] 6554 Specify the read-protection implementation to test. 6555 6556 refscale.shutdown= [KNL] 6557 Shut down the system at the end of the performance 6558 test. This defaults to 1 (shut it down) when 6559 refscale is built into the kernel and to 0 (leave 6560 it running) when refscale is built as a module. 6561 6562 refscale.verbose= [KNL] 6563 Enable additional printk() statements. 6564 6565 refscale.verbose_batched= [KNL] 6566 Batch the additional printk() statements. If zero 6567 (the default) or negative, print everything. Otherwise, 6568 print every Nth verbose statement, where N is the value 6569 specified. 6570 6571 regulator_ignore_unused 6572 [REGULATOR] 6573 Prevents regulator framework from disabling regulators 6574 that are unused, due no driver claiming them. This may 6575 be useful for debug and development, but should not be 6576 needed on a platform with proper driver support. 6577 6578 relax_domain_level= 6579 [KNL, SMP] Set scheduler's default relax_domain_level. 6580 See Documentation/admin-guide/cgroup-v1/cpusets.rst. 6581 6582 reserve= [KNL,BUGS] Force kernel to ignore I/O ports or memory 6583 Format: <base1>,<size1>[,<base2>,<size2>,...] 6584 Reserve I/O ports or memory so the kernel won't use 6585 them. If <base> is less than 0x10000, the region 6586 is assumed to be I/O ports; otherwise it is memory. 6587 6588 reserve_mem= [RAM] 6589 Format: nn[KMG]:<align>:<label> 6590 Reserve physical memory and label it with a name that 6591 other subsystems can use to access it. This is typically 6592 used for systems that do not wipe the RAM, and this command 6593 line will try to reserve the same physical memory on 6594 soft reboots. Note, it is not guaranteed to be the same 6595 location. For example, if anything about the system changes 6596 or if booting a different kernel. It can also fail if KASLR 6597 places the kernel at the location of where the RAM reservation 6598 was from a previous boot, the new reservation will be at a 6599 different location. 6600 Any subsystem using this feature must add a way to verify 6601 that the contents of the physical memory is from a previous 6602 boot, as there may be cases where the memory will not be 6603 located at the same location. 6604 6605 The format is size:align:label for example, to request 6606 12 megabytes of 4096 alignment for ramoops: 6607 6608 reserve_mem=12M:4096:oops ramoops.mem_name=oops 6609 6610 reservetop= [X86-32,EARLY] 6611 Format: nn[KMG] 6612 Reserves a hole at the top of the kernel virtual 6613 address space. 6614 6615 reset_devices [KNL] Force drivers to reset the underlying device 6616 during initialization. 6617 6618 resume= [SWSUSP] 6619 Specify the partition device for software suspend 6620 Format: 6621 {/dev/<dev> | PARTUUID=<uuid> | <int>:<int> | <hex>} 6622 6623 resume_offset= [SWSUSP] 6624 Specify the offset from the beginning of the partition 6625 given by "resume=" at which the swap header is located, 6626 in <PAGE_SIZE> units (needed only for swap files). 6627 See Documentation/power/swsusp-and-swap-files.rst 6628 6629 resumedelay= [HIBERNATION] Delay (in seconds) to pause before attempting to 6630 read the resume files 6631 6632 resumewait [HIBERNATION] Wait (indefinitely) for resume device to show up. 6633 Useful for devices that are detected asynchronously 6634 (e.g. USB and MMC devices). 6635 6636 retain_initrd [RAM] Keep initrd memory after extraction. After boot, it will 6637 be accessible via /sys/firmware/initrd. 6638 6639 retbleed= [X86] Control mitigation of RETBleed (Arbitrary 6640 Speculative Code Execution with Return Instructions) 6641 vulnerability. 6642 6643 AMD-based UNRET and IBPB mitigations alone do not stop 6644 sibling threads from influencing the predictions of other 6645 sibling threads. For that reason, STIBP is used on pro- 6646 cessors that support it, and mitigate SMT on processors 6647 that don't. 6648 6649 off - no mitigation 6650 auto - automatically select a mitigation 6651 auto,nosmt - automatically select a mitigation, 6652 disabling SMT if necessary for 6653 the full mitigation (only on Zen1 6654 and older without STIBP). 6655 ibpb - On AMD, mitigate short speculation 6656 windows on basic block boundaries too. 6657 Safe, highest perf impact. It also 6658 enables STIBP if present. Not suitable 6659 on Intel. 6660 ibpb,nosmt - Like "ibpb" above but will disable SMT 6661 when STIBP is not available. This is 6662 the alternative for systems which do not 6663 have STIBP. 6664 unret - Force enable untrained return thunks, 6665 only effective on AMD f15h-f17h based 6666 systems. 6667 unret,nosmt - Like unret, but will disable SMT when STIBP 6668 is not available. This is the alternative for 6669 systems which do not have STIBP. 6670 6671 Selecting 'auto' will choose a mitigation method at run 6672 time according to the CPU. 6673 6674 Not specifying this option is equivalent to retbleed=auto. 6675 6676 rfkill.default_state= 6677 0 "airplane mode". All wifi, bluetooth, wimax, gps, fm, 6678 etc. communication is blocked by default. 6679 1 Unblocked. 6680 6681 rfkill.master_switch_mode= 6682 0 The "airplane mode" button does nothing. 6683 1 The "airplane mode" button toggles between everything 6684 blocked and the previous configuration. 6685 2 The "airplane mode" button toggles between everything 6686 blocked and everything unblocked. 6687 6688 ring3mwait=disable 6689 [KNL] Disable ring 3 MONITOR/MWAIT feature on supported 6690 CPUs. 6691 6692 riscv_isa_fallback [RISCV,EARLY] 6693 When CONFIG_RISCV_ISA_FALLBACK is not enabled, permit 6694 falling back to detecting extension support by parsing 6695 "riscv,isa" property on devicetree systems when the 6696 replacement properties are not found. See the Kconfig 6697 entry for RISCV_ISA_FALLBACK. 6698 6699 riscv_nousercfi= 6700 all Disable user CFI ABI to userspace even if cpu extension 6701 are available. 6702 bcfi Disable user backward CFI ABI to userspace even if 6703 the shadow stack extension is available. 6704 fcfi Disable user forward CFI ABI to userspace even if the 6705 landing pad extension is available. 6706 6707 ro [KNL] Mount root device read-only on boot 6708 6709 rodata= [KNL,EARLY] 6710 on Mark read-only kernel memory as read-only (default). 6711 off Leave read-only kernel memory writable for debugging. 6712 noalias Mark read-only kernel memory as read-only but retain 6713 writable aliases in the direct map for regions outside 6714 of the kernel image. [arm64] 6715 6716 rockchip.usb_uart 6717 [EARLY] 6718 Enable the uart passthrough on the designated usb port 6719 on Rockchip SoCs. When active, the signals of the 6720 debug-uart get routed to the D+ and D- pins of the usb 6721 port and the regular usb controller gets disabled. 6722 6723 root= [KNL] Root filesystem 6724 Usually this is a block device specifier of some kind, 6725 see the early_lookup_bdev comment in 6726 block/early-lookup.c for details. 6727 Alternatively this can be "ram" for the legacy initial 6728 ramdisk, "nfs" and "cifs" for root on a network file 6729 system, or "mtd" and "ubi" for mounting from raw flash. 6730 6731 rootdelay= [KNL] Delay (in seconds) to pause before attempting to 6732 mount the root filesystem 6733 6734 rootflags= [KNL] Set root filesystem mount option string 6735 6736 rseq_slice_ext= [KNL] RSEQ based time slice extension 6737 Format: boolean 6738 Control enablement of RSEQ based time slice extension. 6739 Default is 'on'. 6740 6741 initramfs_options= [KNL] 6742 Specify mount options for the initramfs mount. 6743 6744 rootfstype= [KNL] Set root filesystem type 6745 6746 rootwait [KNL] Wait (indefinitely) for root device to show up. 6747 Useful for devices that are detected asynchronously 6748 (e.g. USB and MMC devices). 6749 6750 rootwait= [KNL] Maximum time (in seconds) to wait for root device 6751 to show up before attempting to mount the root 6752 filesystem. 6753 6754 rproc_mem=nn[KMG][@address] 6755 [KNL,ARM,CMA] Remoteproc physical memory block. 6756 Memory area to be used by remote processor image, 6757 managed by CMA. 6758 6759 rseq_debug= [KNL] Enable or disable restartable sequence 6760 debug mode. Defaults to CONFIG_RSEQ_DEBUG_DEFAULT_ENABLE. 6761 Format: <bool> 6762 6763 rt_group_sched= [KNL] Enable or disable SCHED_RR/FIFO group scheduling 6764 when CONFIG_RT_GROUP_SCHED=y. Defaults to 6765 !CONFIG_RT_GROUP_SCHED_DEFAULT_DISABLED. 6766 Format: <bool> 6767 6768 rw [KNL] Mount root device read-write on boot 6769 6770 S [KNL] Run init in single mode 6771 6772 s390_iommu= [HW,S390] 6773 Set s390 IOTLB flushing mode 6774 strict 6775 With strict flushing every unmap operation will result 6776 in an IOTLB flush. Default is lazy flushing before 6777 reuse, which is faster. Deprecated, equivalent to 6778 iommu.strict=1. 6779 6780 s390_iommu_aperture= [KNL,S390] 6781 Specifies the size of the per device DMA address space 6782 accessible through the DMA and IOMMU APIs as a decimal 6783 factor of the size of main memory. 6784 The default is 1 meaning that one can concurrently use 6785 as many DMA addresses as physical memory is installed, 6786 if supported by hardware, and thus map all of memory 6787 once. With a value of 2 one can map all of memory twice 6788 and so on. As a special case a factor of 0 imposes no 6789 restrictions other than those given by hardware at the 6790 cost of significant additional memory use for tables. 6791 6792 sa1100ir [NET] 6793 See drivers/net/irda/sa1100_ir.c. 6794 6795 sched_proxy_exec= [KNL] 6796 Enables or disables "proxy execution" style 6797 solution to mutex-based priority inversion. 6798 Format: <bool> 6799 6800 sched_verbose [KNL,EARLY] Enables verbose scheduler debug messages. 6801 6802 schedstats= [KNL,X86] Enable or disable scheduled statistics. 6803 Allowed values are enable and disable. This feature 6804 incurs a small amount of overhead in the scheduler 6805 but is useful for debugging and performance tuning. 6806 6807 sched_thermal_decay_shift= 6808 [Deprecated] 6809 [KNL, SMP] Set a decay shift for scheduler thermal 6810 pressure signal. Thermal pressure signal follows the 6811 default decay period of other scheduler pelt 6812 signals(usually 32 ms but configurable). Setting 6813 sched_thermal_decay_shift will left shift the decay 6814 period for the thermal pressure signal by the shift 6815 value. 6816 i.e. with the default pelt decay period of 32 ms 6817 sched_thermal_decay_shift thermal pressure decay pr 6818 1 64 ms 6819 2 128 ms 6820 and so on. 6821 Format: integer between 0 and 10 6822 Default is 0. 6823 6824 scftorture.holdoff= [KNL] 6825 Number of seconds to hold off before starting 6826 test. Defaults to zero for module insertion and 6827 to 10 seconds for built-in smp_call_function() 6828 tests. 6829 6830 scftorture.longwait= [KNL] 6831 Request ridiculously long waits randomly selected 6832 up to the chosen limit in seconds. Zero (the 6833 default) disables this feature. Please note 6834 that requesting even small non-zero numbers of 6835 seconds can result in RCU CPU stall warnings, 6836 softlockup complaints, and so on. 6837 6838 scftorture.nthreads= [KNL] 6839 Number of kthreads to spawn to invoke the 6840 smp_call_function() family of functions. 6841 The default of -1 specifies a number of kthreads 6842 equal to the number of CPUs. 6843 6844 scftorture.onoff_holdoff= [KNL] 6845 Number seconds to wait after the start of the 6846 test before initiating CPU-hotplug operations. 6847 6848 scftorture.onoff_interval= [KNL] 6849 Number seconds to wait between successive 6850 CPU-hotplug operations. Specifying zero (which 6851 is the default) disables CPU-hotplug operations. 6852 6853 scftorture.shutdown_secs= [KNL] 6854 The number of seconds following the start of the 6855 test after which to shut down the system. The 6856 default of zero avoids shutting down the system. 6857 Non-zero values are useful for automated tests. 6858 6859 scftorture.stat_interval= [KNL] 6860 The number of seconds between outputting the 6861 current test statistics to the console. A value 6862 of zero disables statistics output. 6863 6864 scftorture.stutter_cpus= [KNL] 6865 The number of jiffies to wait between each change 6866 to the set of CPUs under test. 6867 6868 scftorture.use_cpus_read_lock= [KNL] 6869 Use use_cpus_read_lock() instead of the default 6870 preempt_disable() to disable CPU hotplug 6871 while invoking one of the smp_call_function*() 6872 functions. 6873 6874 scftorture.verbose= [KNL] 6875 Enable additional printk() statements. 6876 6877 scftorture.weight_single= [KNL] 6878 The probability weighting to use for the 6879 smp_call_function_single() function with a zero 6880 "wait" parameter. A value of -1 selects the 6881 default if all other weights are -1. However, 6882 if at least one weight has some other value, a 6883 value of -1 will instead select a weight of zero. 6884 6885 scftorture.weight_single_wait= [KNL] 6886 The probability weighting to use for the 6887 smp_call_function_single() function with a 6888 non-zero "wait" parameter. See weight_single. 6889 6890 scftorture.weight_many= [KNL] 6891 The probability weighting to use for the 6892 smp_call_function_many() function with a zero 6893 "wait" parameter. See weight_single. 6894 Note well that setting a high probability for 6895 this weighting can place serious IPI load 6896 on the system. 6897 6898 scftorture.weight_many_wait= [KNL] 6899 The probability weighting to use for the 6900 smp_call_function_many() function with a 6901 non-zero "wait" parameter. See weight_single 6902 and weight_many. 6903 6904 scftorture.weight_all= [KNL] 6905 The probability weighting to use for the 6906 smp_call_function_all() function with a zero 6907 "wait" parameter. See weight_single and 6908 weight_many. 6909 6910 scftorture.weight_all_wait= [KNL] 6911 The probability weighting to use for the 6912 smp_call_function_all() function with a 6913 non-zero "wait" parameter. See weight_single 6914 and weight_many. 6915 6916 sdw_mclk_divider=[SDW] 6917 Specify the MCLK divider for Intel SoundWire buses in 6918 case the BIOS does not provide the clock rate properly. 6919 6920 skew_tick= [KNL,EARLY] Offset the periodic timer tick per cpu to mitigate 6921 xtime_lock contention on larger systems, and/or RCU lock 6922 contention on all systems with CONFIG_MAXSMP set. 6923 Format: { "0" | "1" } 6924 0 -- disable. (may be 1 via CONFIG_CMDLINE="skew_tick=1" 6925 1 -- enable. 6926 Note: increases power consumption, thus should only be 6927 enabled if running jitter sensitive (HPC/RT) workloads. 6928 6929 security= [SECURITY] Choose a legacy "major" security module to 6930 enable at boot. This has been deprecated by the 6931 "lsm=" parameter. 6932 6933 selinux= [SELINUX] Disable or enable SELinux at boot time. 6934 Format: { "0" | "1" } 6935 See security/selinux/Kconfig help text. 6936 0 -- disable. 6937 1 -- enable. 6938 Default value is 1. 6939 6940 serialnumber [BUGS=X86-32] 6941 6942 sev=option[,option...] [X86-64] 6943 6944 debug 6945 Enable debug messages. 6946 6947 nosnp 6948 Do not enable SEV-SNP (applies to host/hypervisor 6949 only). Setting 'nosnp' avoids the RMP check overhead 6950 in memory accesses when users do not want to run 6951 SEV-SNP guests. 6952 6953 shapers= [NET] 6954 Maximal number of shapers. 6955 6956 show_lapic= [APIC,X86] Advanced Programmable Interrupt Controller 6957 Limit apic dumping. The parameter defines the maximal 6958 number of local apics being dumped. Also it is possible 6959 to set it to "all" by meaning -- no limit here. 6960 Format: { 1 (default) | 2 | ... | all }. 6961 The parameter valid if only apic=debug or 6962 apic=verbose is specified. 6963 Example: apic=debug show_lapic=all 6964 6965 slab_debug[=options[,slabs][;[options[,slabs]]...] [MM] 6966 Enabling slab_debug allows one to determine the 6967 culprit if slab objects become corrupted. Enabling 6968 slab_debug can create guard zones around objects and 6969 may poison objects when not in use. Also tracks the 6970 last alloc / free. For more information see 6971 Documentation/admin-guide/mm/slab.rst. 6972 (slub_debug legacy name also accepted for now) 6973 6974 Using this option implies the "no_hash_pointers" 6975 option which can be undone by adding the 6976 "hash_pointers=always" option. 6977 6978 slab_max_order= [MM] 6979 Determines the maximum allowed order for slabs. 6980 A high setting may cause OOMs due to memory 6981 fragmentation. For more information see 6982 Documentation/admin-guide/mm/slab.rst. 6983 (slub_max_order legacy name also accepted for now) 6984 6985 slab_merge [MM] 6986 Enable merging of slabs with similar size when the 6987 kernel is built without CONFIG_SLAB_MERGE_DEFAULT. 6988 (slub_merge legacy name also accepted for now) 6989 6990 slab_min_objects= [MM] 6991 The minimum number of objects per slab. SLUB will 6992 increase the slab order up to slab_max_order to 6993 generate a sufficiently large slab able to contain 6994 the number of objects indicated. The higher the number 6995 of objects the smaller the overhead of tracking slabs 6996 and the less frequently locks need to be acquired. 6997 For more information see 6998 Documentation/admin-guide/mm/slab.rst. 6999 (slub_min_objects legacy name also accepted for now) 7000 7001 slab_min_order= [MM] 7002 Determines the minimum page order for slabs. Must be 7003 lower or equal to slab_max_order. For more information see 7004 Documentation/admin-guide/mm/slab.rst. 7005 (slub_min_order legacy name also accepted for now) 7006 7007 slab_nomerge [MM] 7008 Disable merging of slabs with similar size. May be 7009 necessary if there is some reason to distinguish 7010 allocs to different slabs, especially in hardened 7011 environments where the risk of heap overflows and 7012 layout control by attackers can usually be 7013 frustrated by disabling merging. This will reduce 7014 most of the exposure of a heap attack to a single 7015 cache (risks via metadata attacks are mostly 7016 unchanged). Debug options disable merging on their 7017 own. 7018 For more information see 7019 Documentation/admin-guide/mm/slab.rst. 7020 (slub_nomerge legacy name also accepted for now) 7021 7022 slab_strict_numa [MM] 7023 Support memory policies on a per object level 7024 in the slab allocator. The default is for memory 7025 policies to be applied at the folio level when 7026 a new folio is needed or a partial folio is 7027 retrieved from the lists. Increases overhead 7028 in the slab fastpaths but gains more accurate 7029 NUMA kernel object placement which helps with slow 7030 interconnects in NUMA systems. 7031 7032 slram= [HW,MTD] 7033 7034 smart2= [HW] 7035 Format: <io1>[,<io2>[,...,<io8>]] 7036 7037 smp.csd_lock_timeout= [KNL] 7038 Specify the period of time in milliseconds 7039 that smp_call_function() and friends will wait 7040 for a CPU to release the CSD lock. This is 7041 useful when diagnosing bugs involving CPUs 7042 disabling interrupts for extended periods 7043 of time. Defaults to 5,000 milliseconds, and 7044 setting a value of zero disables this feature. 7045 This feature may be more efficiently disabled 7046 using the csdlock_debug- kernel parameter. 7047 7048 smp.panic_on_ipistall= [KNL] 7049 If a csd_lock_timeout extends for more than 7050 the specified number of milliseconds, panic the 7051 system. By default, let CSD-lock acquisition 7052 take as long as they take. Specifying 300,000 7053 for this value provides a 5-minute timeout. 7054 7055 smsc-ircc2.nopnp [HW] Don't use PNP to discover SMC devices 7056 smsc-ircc2.ircc_cfg= [HW] Device configuration I/O port 7057 smsc-ircc2.ircc_sir= [HW] SIR base I/O port 7058 smsc-ircc2.ircc_fir= [HW] FIR base I/O port 7059 smsc-ircc2.ircc_irq= [HW] IRQ line 7060 smsc-ircc2.ircc_dma= [HW] DMA channel 7061 smsc-ircc2.ircc_transceiver= [HW] Transceiver type: 7062 0: Toshiba Satellite 1800 (GP data pin select) 7063 1: Fast pin select (default) 7064 2: ATC IRMode 7065 7066 smt= [KNL,MIPS,S390,EARLY] Set the maximum number of threads 7067 (logical CPUs) to use per physical CPU on systems 7068 capable of symmetric multithreading (SMT). Will 7069 be capped to the actual hardware limit. 7070 Format: <integer> 7071 Default: -1 (no limit) 7072 7073 softlockup_panic= 7074 [KNL] Should the soft-lockup detector generate panics. 7075 Format: <int> 7076 7077 A value of non-zero instructs the soft-lockup detector 7078 to panic the machine when a soft-lockup duration exceeds 7079 N thresholds. It is also controlled by the kernel.softlockup_panic 7080 sysctl and CONFIG_BOOTPARAM_SOFTLOCKUP_PANIC, which is the 7081 respective build-time switch to that functionality. 7082 7083 softlockup_all_cpu_backtrace= 7084 [KNL] Should the soft-lockup detector generate 7085 backtraces on all cpus. 7086 Format: 0 | 1 7087 7088 sonypi.*= [HW] Sony Programmable I/O Control Device driver 7089 See Documentation/admin-guide/laptops/sonypi.rst 7090 7091 spectre_bhi= [X86] Control mitigation of Branch History Injection 7092 (BHI) vulnerability. This setting affects the 7093 deployment of the HW BHI control and the SW BHB 7094 clearing sequence. 7095 7096 on - (default) Enable the HW or SW mitigation as 7097 needed. This protects the kernel from 7098 both syscalls and VMs. 7099 vmexit - On systems which don't have the HW mitigation 7100 available, enable the SW mitigation on vmexit 7101 ONLY. On such systems, the host kernel is 7102 protected from VM-originated BHI attacks, but 7103 may still be vulnerable to syscall attacks. 7104 off - Disable the mitigation. 7105 7106 spectre_v2= [X86,EARLY] Control mitigation of Spectre variant 2 7107 (indirect branch speculation) vulnerability. 7108 The default operation protects the kernel from 7109 user space attacks. 7110 7111 on - unconditionally enable, implies 7112 spectre_v2_user=on 7113 off - unconditionally disable, implies 7114 spectre_v2_user=off 7115 auto - kernel detects whether your CPU model is 7116 vulnerable 7117 7118 Selecting 'on' will, and 'auto' may, choose a 7119 mitigation method at run time according to the 7120 CPU, the available microcode, the setting of the 7121 CONFIG_MITIGATION_RETPOLINE configuration option, 7122 and the compiler with which the kernel was built. 7123 7124 Selecting 'on' will also enable the mitigation 7125 against user space to user space task attacks. 7126 Selecting specific mitigation does not force enable 7127 user mitigations. 7128 7129 Selecting 'off' will disable both the kernel and 7130 the user space protections. 7131 7132 Specific mitigations can also be selected manually: 7133 7134 retpoline - replace indirect branches 7135 retpoline,generic - Retpolines 7136 retpoline,lfence - LFENCE; indirect branch 7137 retpoline,amd - alias for retpoline,lfence 7138 eibrs - Enhanced/Auto IBRS 7139 eibrs,retpoline - Enhanced/Auto IBRS + Retpolines 7140 eibrs,lfence - Enhanced/Auto IBRS + LFENCE 7141 ibrs - use IBRS to protect kernel 7142 7143 Not specifying this option is equivalent to 7144 spectre_v2=auto. 7145 7146 spectre_v2_user= 7147 [X86] Control mitigation of Spectre variant 2 7148 (indirect branch speculation) vulnerability between 7149 user space tasks 7150 7151 on - Unconditionally enable mitigations. Is 7152 enforced by spectre_v2=on 7153 7154 off - Unconditionally disable mitigations. Is 7155 enforced by spectre_v2=off 7156 7157 prctl - Indirect branch speculation is enabled, 7158 but mitigation can be enabled via prctl 7159 per thread. The mitigation control state 7160 is inherited on fork. 7161 7162 prctl,ibpb 7163 - Like "prctl" above, but only STIBP is 7164 controlled per thread. IBPB is issued 7165 always when switching between different user 7166 space processes. 7167 7168 seccomp 7169 - Same as "prctl" above, but all seccomp 7170 threads will enable the mitigation unless 7171 they explicitly opt out. 7172 7173 seccomp,ibpb 7174 - Like "seccomp" above, but only STIBP is 7175 controlled per thread. IBPB is issued 7176 always when switching between different 7177 user space processes. 7178 7179 auto - Kernel selects the mitigation depending on 7180 the available CPU features and vulnerability. 7181 7182 Default mitigation: "prctl" 7183 7184 Not specifying this option is equivalent to 7185 spectre_v2_user=auto. 7186 7187 spec_rstack_overflow= 7188 [X86,EARLY] Control RAS overflow mitigation on AMD Zen CPUs 7189 7190 off - Disable mitigation 7191 microcode - Enable microcode mitigation only 7192 safe-ret - Enable sw-only safe RET mitigation (default) 7193 ibpb - Enable mitigation by issuing IBPB on 7194 kernel entry 7195 ibpb-vmexit - Issue IBPB only on VMEXIT 7196 (cloud-specific mitigation) 7197 7198 spec_store_bypass_disable= 7199 [HW,EARLY] Control Speculative Store Bypass (SSB) Disable mitigation 7200 (Speculative Store Bypass vulnerability) 7201 7202 Certain CPUs are vulnerable to an exploit against a 7203 a common industry wide performance optimization known 7204 as "Speculative Store Bypass" in which recent stores 7205 to the same memory location may not be observed by 7206 later loads during speculative execution. The idea 7207 is that such stores are unlikely and that they can 7208 be detected prior to instruction retirement at the 7209 end of a particular speculation execution window. 7210 7211 In vulnerable processors, the speculatively forwarded 7212 store can be used in a cache side channel attack, for 7213 example to read memory to which the attacker does not 7214 directly have access (e.g. inside sandboxed code). 7215 7216 This parameter controls whether the Speculative Store 7217 Bypass optimization is used. 7218 7219 On x86 the options are: 7220 7221 on - Unconditionally disable Speculative Store Bypass 7222 off - Unconditionally enable Speculative Store Bypass 7223 auto - Kernel detects whether the CPU model contains an 7224 implementation of Speculative Store Bypass and 7225 picks the most appropriate mitigation. If the 7226 CPU is not vulnerable, "off" is selected. If the 7227 CPU is vulnerable the default mitigation is 7228 architecture and Kconfig dependent. See below. 7229 prctl - Control Speculative Store Bypass per thread 7230 via prctl. Speculative Store Bypass is enabled 7231 for a process by default. The state of the control 7232 is inherited on fork. 7233 seccomp - Same as "prctl" above, but all seccomp threads 7234 will disable SSB unless they explicitly opt out. 7235 7236 Default mitigations: 7237 X86: "prctl" 7238 7239 On powerpc the options are: 7240 7241 on,auto - On Power8 and Power9 insert a store-forwarding 7242 barrier on kernel entry and exit. On Power7 7243 perform a software flush on kernel entry and 7244 exit. 7245 off - No action. 7246 7247 Not specifying this option is equivalent to 7248 spec_store_bypass_disable=auto. 7249 7250 split_lock_detect= 7251 [X86] Enable split lock detection or bus lock detection 7252 7253 When enabled (and if hardware support is present), atomic 7254 instructions that access data across cache line 7255 boundaries will result in an alignment check exception 7256 for split lock detection or a debug exception for 7257 bus lock detection. 7258 7259 off - not enabled 7260 7261 warn - the kernel will emit rate-limited warnings 7262 about applications triggering the #AC 7263 exception or the #DB exception. This mode is 7264 the default on CPUs that support split lock 7265 detection or bus lock detection. Default 7266 behavior is by #AC if both features are 7267 enabled in hardware. 7268 7269 fatal - the kernel will send SIGBUS to applications 7270 that trigger the #AC exception or the #DB 7271 exception. Default behavior is by #AC if 7272 both features are enabled in hardware. 7273 7274 ratelimit:N - 7275 Set system wide rate limit to N bus locks 7276 per second for bus lock detection. 7277 0 < N <= 1000. 7278 7279 N/A for split lock detection. 7280 7281 7282 If an #AC exception is hit in the kernel or in 7283 firmware (i.e. not while executing in user mode) 7284 the kernel will oops in either "warn" or "fatal" 7285 mode. 7286 7287 #DB exception for bus lock is triggered only when 7288 CPL > 0. 7289 7290 srbds= [X86,INTEL,EARLY] 7291 Control the Special Register Buffer Data Sampling 7292 (SRBDS) mitigation. 7293 7294 Certain CPUs are vulnerable to an MDS-like 7295 exploit which can leak bits from the random 7296 number generator. 7297 7298 By default, this issue is mitigated by 7299 microcode. However, the microcode fix can cause 7300 the RDRAND and RDSEED instructions to become 7301 much slower. Among other effects, this will 7302 result in reduced throughput from /dev/urandom. 7303 7304 The microcode mitigation can be disabled with 7305 the following option: 7306 7307 off: Disable mitigation and remove 7308 performance impact to RDRAND and RDSEED 7309 7310 srcutree.big_cpu_lim [KNL] 7311 Specifies the number of CPUs constituting a 7312 large system, such that srcu_struct structures 7313 should immediately allocate an srcu_node array. 7314 This kernel-boot parameter defaults to 128, 7315 but takes effect only when the low-order four 7316 bits of srcutree.convert_to_big is equal to 3 7317 (decide at boot). 7318 7319 srcutree.convert_to_big [KNL] 7320 Specifies under what conditions an SRCU tree 7321 srcu_struct structure will be converted to big 7322 form, that is, with an rcu_node tree: 7323 7324 0: Never. 7325 1: At init_srcu_struct() time. 7326 2: When rcutorture decides to. 7327 3: Decide at boot time (default). 7328 0x1X: Above plus if high contention. 7329 7330 Either way, the srcu_node tree will be sized based 7331 on the actual runtime number of CPUs (nr_cpu_ids) 7332 instead of the compile-time CONFIG_NR_CPUS. 7333 7334 srcutree.counter_wrap_check [KNL] 7335 Specifies how frequently to check for 7336 grace-period sequence counter wrap for the 7337 srcu_data structure's ->srcu_gp_seq_needed field. 7338 The greater the number of bits set in this kernel 7339 parameter, the less frequently counter wrap will 7340 be checked for. Note that the bottom two bits 7341 are ignored. 7342 7343 srcutree.exp_holdoff [KNL] 7344 Specifies how many nanoseconds must elapse 7345 since the end of the last SRCU grace period for 7346 a given srcu_struct until the next normal SRCU 7347 grace period will be considered for automatic 7348 expediting. Set to zero to disable automatic 7349 expediting. 7350 7351 srcutree.srcu_max_nodelay [KNL] 7352 Specifies the number of no-delay instances 7353 per jiffy for which the SRCU grace period 7354 worker thread will be rescheduled with zero 7355 delay. Beyond this limit, worker thread will 7356 be rescheduled with a sleep delay of one jiffy. 7357 7358 srcutree.srcu_max_nodelay_phase [KNL] 7359 Specifies the per-grace-period phase, number of 7360 non-sleeping polls of readers. Beyond this limit, 7361 grace period worker thread will be rescheduled 7362 with a sleep delay of one jiffy, between each 7363 rescan of the readers, for a grace period phase. 7364 7365 srcutree.srcu_retry_check_delay [KNL] 7366 Specifies number of microseconds of non-sleeping 7367 delay between each non-sleeping poll of readers. 7368 7369 srcutree.small_contention_lim [KNL] 7370 Specifies the number of update-side contention 7371 events per jiffy will be tolerated before 7372 initiating a conversion of an srcu_struct 7373 structure to big form. Note that the value of 7374 srcutree.convert_to_big must have the 0x10 bit 7375 set for contention-based conversions to occur. 7376 7377 ssbd= [ARM64,HW,EARLY] 7378 Speculative Store Bypass Disable control 7379 7380 On CPUs that are vulnerable to the Speculative 7381 Store Bypass vulnerability and offer a 7382 firmware based mitigation, this parameter 7383 indicates how the mitigation should be used: 7384 7385 force-on: Unconditionally enable mitigation for 7386 for both kernel and userspace 7387 force-off: Unconditionally disable mitigation for 7388 for both kernel and userspace 7389 kernel: Always enable mitigation in the 7390 kernel, and offer a prctl interface 7391 to allow userspace to register its 7392 interest in being mitigated too. 7393 7394 stack_guard_gap= [MM] 7395 override the default stack gap protection. The value 7396 is in page units and it defines how many pages prior 7397 to (for stacks growing down) resp. after (for stacks 7398 growing up) the main stack are reserved for no other 7399 mapping. Default value is 256 pages. 7400 7401 stack_depot_disable= [KNL,EARLY] 7402 Setting this to true through kernel command line will 7403 disable the stack depot thereby saving the static memory 7404 consumed by the stack hash table. By default this is set 7405 to false. 7406 7407 stack_depot_max_pools= [KNL,EARLY] 7408 Specify the maximum number of pools to use for storing 7409 stack traces. Pools are allocated on-demand up to this 7410 limit. Default value is 8191 pools. 7411 7412 stacktrace [FTRACE] 7413 Enable the stack tracer on boot up. 7414 7415 stacktrace_filter=[function-list] 7416 [FTRACE] Limit the functions that the stack tracer 7417 will trace at boot up. function-list is a comma-separated 7418 list of functions. This list can be changed at run 7419 time by the stack_trace_filter file in the debugfs 7420 tracing directory. Note, this enables stack tracing 7421 and the stacktrace above is not needed. 7422 7423 sti= [PARISC,HW] 7424 Format: <num> 7425 Set the STI (builtin display/keyboard on the HP-PARISC 7426 machines) console (graphic card) which should be used 7427 as the initial boot-console. 7428 See also comment in drivers/video/console/sticore.c. 7429 7430 sti_font= [HW] 7431 See comment in drivers/video/console/sticore.c. 7432 7433 stifb= [HW] 7434 Format: bpp:<bpp1>[:<bpp2>[:<bpp3>...]] 7435 7436 strict_sas_size= 7437 [X86] 7438 Format: <bool> 7439 Enable or disable strict sigaltstack size checks 7440 against the required signal frame size which 7441 depends on the supported FPU features. This can 7442 be used to filter out binaries which have 7443 not yet been made aware of AT_MINSIGSTKSZ. 7444 7445 stress_hpt [PPC,EARLY] 7446 Limits the number of kernel HPT entries in the hash 7447 page table to increase the rate of hash page table 7448 faults on kernel addresses. 7449 7450 stress_slb [PPC,EARLY] 7451 Limits the number of kernel SLB entries, and flushes 7452 them frequently to increase the rate of SLB faults 7453 on kernel addresses. 7454 7455 no_slb_preload [PPC,EARLY] 7456 Disables slb preloading for userspace. 7457 7458 sunrpc.min_resvport= 7459 sunrpc.max_resvport= 7460 [NFS,SUNRPC] 7461 SunRPC servers often require that client requests 7462 originate from a privileged port (i.e. a port in the 7463 range 0 < portnr < 1024). 7464 An administrator who wishes to reserve some of these 7465 ports for other uses may adjust the range that the 7466 kernel's sunrpc client considers to be privileged 7467 using these two parameters to set the minimum and 7468 maximum port values. 7469 7470 sunrpc.svc_rpc_per_connection_limit= 7471 [NFS,SUNRPC] 7472 Limit the number of requests that the server will 7473 process in parallel from a single connection. 7474 The default value is 0 (no limit). 7475 7476 sunrpc.pool_mode= 7477 [NFS] 7478 Control how the NFS server code allocates CPUs to 7479 service thread pools. Depending on how many NICs 7480 you have and where their interrupts are bound, this 7481 option will affect which CPUs will do NFS serving. 7482 Note: this parameter cannot be changed while the 7483 NFS server is running. 7484 7485 auto the server chooses an appropriate mode 7486 automatically using heuristics 7487 global a single global pool contains all CPUs 7488 percpu one pool for each CPU 7489 pernode one pool for each NUMA node (equivalent 7490 to global on non-NUMA machines) 7491 7492 sunrpc.tcp_slot_table_entries= 7493 sunrpc.udp_slot_table_entries= 7494 [NFS,SUNRPC] 7495 Sets the upper limit on the number of simultaneous 7496 RPC calls that can be sent from the client to a 7497 server. Increasing these values may allow you to 7498 improve throughput, but will also increase the 7499 amount of memory reserved for use by the client. 7500 7501 suspend.pm_test_delay= 7502 [SUSPEND] 7503 Sets the number of seconds to remain in a suspend test 7504 mode before resuming the system (see 7505 /sys/power/pm_test). Only available when CONFIG_PM_DEBUG 7506 is set. Default value is 5. 7507 7508 svm= [PPC] 7509 Format: { on | off | y | n | 1 | 0 } 7510 This parameter controls use of the Protected 7511 Execution Facility on pSeries. 7512 7513 swiotlb= [ARM,PPC,MIPS,X86,S390,EARLY] 7514 Format: { <int> [,<int>] | force | noforce } 7515 <int> -- Number of I/O TLB slabs 7516 <int> -- Second integer after comma. Number of swiotlb 7517 areas with their own lock. Will be rounded up 7518 to a power of 2. 7519 force -- force using of bounce buffers even if they 7520 wouldn't be automatically used by the kernel 7521 noforce -- Never use bounce buffers (for debugging) 7522 7523 switches= [HW,M68k,EARLY] 7524 7525 sysctl.*= [KNL] 7526 Set a sysctl parameter, right before loading the init 7527 process, as if the value was written to the respective 7528 /proc/sys/... file. Both '.' and '/' are recognized as 7529 separators. Unrecognized parameters and invalid values 7530 are reported in the kernel log. Sysctls registered 7531 later by a loaded module cannot be set this way. 7532 Example: sysctl.vm.swappiness=40 7533 7534 sysrq_always_enabled 7535 [KNL] 7536 Ignore sysrq setting - this boot parameter will 7537 neutralize any effect of /proc/sys/kernel/sysrq. 7538 Useful for debugging. 7539 7540 tcpmhash_entries= [KNL,NET] 7541 Set the number of tcp_metrics_hash slots. 7542 Default value is 8192 or 16384 depending on total 7543 ram pages. This is used to specify the TCP metrics 7544 cache size. See Documentation/networking/ip-sysctl.rst 7545 "tcp_no_metrics_save" section for more details. 7546 7547 tdfx= [HW,DRM] 7548 7549 test_suspend= [SUSPEND] 7550 Format: { "mem" | "standby" | "freeze" }[,N] 7551 Specify "mem" (for Suspend-to-RAM) or "standby" (for 7552 standby suspend) or "freeze" (for suspend type freeze) 7553 as the system sleep state during system startup with 7554 the optional capability to repeat N number of times. 7555 The system is woken from this state using a 7556 wakeup-capable RTC alarm. 7557 7558 thash_entries= [KNL,NET] 7559 Set number of hash buckets for TCP connection 7560 7561 thermal.act= [HW,ACPI] 7562 -1: disable all active trip points in all thermal zones 7563 <degrees C>: override all lowest active trip points 7564 7565 thermal.crt= [HW,ACPI] 7566 -1: disable all critical trip points in all thermal zones 7567 <degrees C>: override all critical trip points 7568 7569 thermal.off= [HW,ACPI] 7570 1: disable ACPI thermal control 7571 7572 thermal.psv= [HW,ACPI] 7573 -1: disable all passive trip points 7574 <degrees C>: override all passive trip points to this 7575 value 7576 7577 thermal.tzp= [HW,ACPI] 7578 Specify global default ACPI thermal zone polling rate 7579 <deci-seconds>: poll all this frequency 7580 0: no polling (default) 7581 7582 thp_anon= [KNL] 7583 Format: <size>[KMG],<size>[KMG]:<state>;<size>[KMG]-<size>[KMG]:<state> 7584 state is one of "always", "madvise", "never" or "inherit". 7585 Control the default behavior of the system with respect 7586 to anonymous transparent hugepages. 7587 Can be used multiple times for multiple anon THP sizes. 7588 See Documentation/admin-guide/mm/transhuge.rst for more 7589 details. 7590 7591 threadirqs [KNL,EARLY] 7592 Force threading of all interrupt handlers except those 7593 marked explicitly IRQF_NO_THREAD. 7594 7595 thp_shmem= [KNL] 7596 Format: <size>[KMG],<size>[KMG]:<policy>;<size>[KMG]-<size>[KMG]:<policy> 7597 Control the default policy of each hugepage size for the 7598 internal shmem mount. <policy> is one of policies available 7599 for the shmem mount ("always", "inherit", "never", "within_size", 7600 and "advise"). 7601 It can be used multiple times for multiple shmem THP sizes. 7602 See Documentation/admin-guide/mm/transhuge.rst for more 7603 details. 7604 7605 topology= [S390,EARLY] 7606 Format: {off | on} 7607 Specify if the kernel should make use of the cpu 7608 topology information if the hardware supports this. 7609 The scheduler will make use of this information and 7610 e.g. base its process migration decisions on it. 7611 Default is on. 7612 7613 torture.disable_onoff_at_boot= [KNL] 7614 Prevent the CPU-hotplug component of torturing 7615 until after init has spawned. 7616 7617 torture.ftrace_dump_at_shutdown= [KNL] 7618 Dump the ftrace buffer at torture-test shutdown, 7619 even if there were no errors. This can be a 7620 very costly operation when many torture tests 7621 are running concurrently, especially on systems 7622 with rotating-rust storage. 7623 7624 torture.verbose_sleep_frequency= [KNL] 7625 Specifies how many verbose printk()s should be 7626 emitted between each sleep. The default of zero 7627 disables verbose-printk() sleeping. 7628 7629 torture.verbose_sleep_duration= [KNL] 7630 Duration of each verbose-printk() sleep in jiffies. 7631 7632 tpm.disable_pcr_integrity= [HW,TPM] 7633 Do not protect PCR registers from unintended physical 7634 access, or interposers in the bus by the means of 7635 having an integrity protected session wrapped around 7636 TPM2_PCR_Extend command. Consider this in a situation 7637 where TPM is heavily utilized by IMA, thus protection 7638 causing a major performance hit, and the space where 7639 machines are deployed is by other means guarded. 7640 7641 tpm_crb_ffa.busy_timeout_ms= [ARM64,TPM] 7642 Maximum time in milliseconds to retry sending a message 7643 to the TPM service before giving up. This parameter controls 7644 how long the system will continue retrying when the TPM 7645 service is busy. 7646 Format: <unsigned int> 7647 Default: 2000 (2 seconds) 7648 7649 tpm_suspend_pcr=[HW,TPM] 7650 Format: integer pcr id 7651 Specify that at suspend time, the tpm driver 7652 should extend the specified pcr with zeros, 7653 as a workaround for some chips which fail to 7654 flush the last written pcr on TPM_SaveState. 7655 This will guarantee that all the other pcrs 7656 are saved. 7657 7658 tpm_tis.interrupts= [HW,TPM] 7659 Enable interrupts for the MMIO based physical layer 7660 for the FIFO interface. By default it is set to false 7661 (0). For more information about TPM hardware interfaces 7662 defined by Trusted Computing Group (TCG) see 7663 https://trustedcomputinggroup.org/resource/pc-client-platform-tpm-profile-ptp-specification/ 7664 7665 tp_printk [FTRACE] 7666 Have the tracepoints sent to printk as well as the 7667 tracing ring buffer. This is useful for early boot up 7668 where the system hangs or reboots and does not give the 7669 option for reading the tracing buffer or performing a 7670 ftrace_dump_on_oops. 7671 7672 To turn off having tracepoints sent to printk, 7673 echo 0 > /proc/sys/kernel/tracepoint_printk 7674 Note, echoing 1 into this file without the 7675 tp_printk kernel cmdline option has no effect. 7676 7677 The tp_printk_stop_on_boot (see below) can also be used 7678 to stop the printing of events to console at 7679 late_initcall_sync. 7680 7681 ** CAUTION ** 7682 7683 Having tracepoints sent to printk() and activating high 7684 frequency tracepoints such as irq or sched, can cause 7685 the system to live lock. 7686 7687 tp_printk_stop_on_boot [FTRACE] 7688 When tp_printk (above) is set, it can cause a lot of noise 7689 on the console. It may be useful to only include the 7690 printing of events during boot up, as user space may 7691 make the system inoperable. 7692 7693 This command line option will stop the printing of events 7694 to console at the late_initcall_sync() time frame. 7695 7696 trace_buf_size=nn[KMG] 7697 [FTRACE] will set tracing buffer size on each cpu. 7698 7699 trace_clock= [FTRACE] Set the clock used for tracing events 7700 at boot up. 7701 local - Use the per CPU time stamp counter 7702 (converted into nanoseconds). Fast, but 7703 depending on the architecture, may not be 7704 in sync between CPUs. 7705 global - Event time stamps are synchronized across 7706 CPUs. May be slower than the local clock, 7707 but better for some race conditions. 7708 counter - Simple counting of events (1, 2, ..) 7709 note, some counts may be skipped due to the 7710 infrastructure grabbing the clock more than 7711 once per event. 7712 uptime - Use jiffies as the time stamp. 7713 perf - Use the same clock that perf uses. 7714 mono - Use ktime_get_mono_fast_ns() for time stamps. 7715 mono_raw - Use ktime_get_raw_fast_ns() for time 7716 stamps. 7717 boot - Use ktime_get_boot_fast_ns() for time stamps. 7718 Architectures may add more clocks. See 7719 Documentation/trace/ftrace.rst for more details. 7720 7721 trace_event=[event-list] 7722 [FTRACE] Set and start specified trace events in order 7723 to facilitate early boot debugging. The event-list is a 7724 comma-separated list of trace events to enable. See 7725 also Documentation/trace/events.rst 7726 7727 To enable modules, use :mod: keyword: 7728 7729 trace_event=:mod:<module> 7730 7731 The value before :mod: will only enable specific events 7732 that are part of the module. See the above mentioned 7733 document for more information. 7734 7735 trace_instance=[instance-info] 7736 [FTRACE] Create a ring buffer instance early in boot up. 7737 This will be listed in: 7738 7739 /sys/kernel/tracing/instances 7740 7741 Events can be enabled at the time the instance is created 7742 via: 7743 7744 trace_instance=<name>,<system1>:<event1>,<system2>:<event2> 7745 7746 Note, the "<system*>:" portion is optional if the event is 7747 unique. 7748 7749 trace_instance=foo,sched:sched_switch,irq_handler_entry,initcall 7750 7751 will enable the "sched_switch" event (note, the "sched:" is optional, and 7752 the same thing would happen if it was left off). The irq_handler_entry 7753 event, and all events under the "initcall" system. 7754 7755 Flags can be added to the instance to modify its behavior when it is 7756 created. The flags are separated by '^'. 7757 7758 The available flags are: 7759 7760 traceoff - Have the tracing instance tracing disabled after it is created. 7761 traceprintk - Have trace_printk() write into this trace instance 7762 (note, "printk" and "trace_printk" can also be used) 7763 7764 trace_instance=foo^traceoff^traceprintk,sched,irq 7765 7766 The flags must come before the defined events. 7767 7768 If memory has been reserved (see memmap for x86), the instance 7769 can use that memory: 7770 7771 memmap=12M$0x284500000 trace_instance=boot_map@0x284500000:12M 7772 7773 The above will create a "boot_map" instance that uses the physical 7774 memory at 0x284500000 that is 12Megs. The per CPU buffers of that 7775 instance will be split up accordingly. 7776 7777 Alternatively, the memory can be reserved by the reserve_mem option: 7778 7779 reserve_mem=12M:4096:trace trace_instance=boot_map@trace 7780 7781 This will reserve 12 megabytes at boot up with a 4096 byte alignment 7782 and place the ring buffer in this memory. Note that due to KASLR, the 7783 memory may not be the same location each time, which will not preserve 7784 the buffer content. 7785 7786 Also note that the layout of the ring buffer data may change between 7787 kernel versions where the validator will fail and reset the ring buffer 7788 if the layout is not the same as the previous kernel. 7789 7790 If the ring buffer is used for persistent bootups and has events enabled, 7791 it is recommend to disable tracing so that events from a previous boot do not 7792 mix with events of the current boot (unless you are debugging a random crash 7793 at boot up). 7794 7795 reserve_mem=12M:4096:trace trace_instance=boot_map^traceoff^traceprintk@trace,sched,irq 7796 7797 Note, saving the trace buffer across reboots does require that the system 7798 is set up to not wipe memory. For instance, CONFIG_RESET_ATTACK_MITIGATION 7799 can force a memory reset on boot which will clear any trace that was stored. 7800 This is just one of many ways that can clear memory. Make sure your system 7801 keeps the content of memory across reboots before relying on this option. 7802 7803 NB: Both the mapped address and size must be page aligned for the architecture. 7804 7805 See also Documentation/trace/debugging.rst 7806 7807 7808 trace_options=[option-list] 7809 [FTRACE] Enable or disable tracer options at boot. 7810 The option-list is a comma delimited list of options 7811 that can be enabled or disabled just as if you were 7812 to echo the option name into 7813 7814 /sys/kernel/tracing/trace_options 7815 7816 For example, to enable stacktrace option (to dump the 7817 stack trace of each event), add to the command line: 7818 7819 trace_options=stacktrace 7820 7821 See also Documentation/trace/ftrace.rst "trace options" 7822 section. 7823 7824 trace_trigger=[trigger-list] 7825 [FTRACE] Add an event trigger on specific events. 7826 Set a trigger on top of a specific event, with an optional 7827 filter. 7828 7829 The format is "trace_trigger=<event>.<trigger>[ if <filter>],..." 7830 Where more than one trigger may be specified that are comma delimited. 7831 7832 For example: 7833 7834 trace_trigger="sched_switch.stacktrace if prev_state == 2" 7835 7836 The above will enable the "stacktrace" trigger on the "sched_switch" 7837 event but only trigger it if the "prev_state" of the "sched_switch" 7838 event is "2" (TASK_UNINTERRUPTIBLE). 7839 7840 See also "Event triggers" in Documentation/trace/events.rst 7841 7842 7843 traceoff_after_boot 7844 [FTRACE] Sometimes tracing is used to debug issues 7845 during the boot process. Since the trace buffer has a 7846 limited amount of storage, it may be prudent to 7847 disable tracing after the boot is finished, otherwise 7848 the critical information may be overwritten. With this 7849 option, the main tracing buffer will be turned off at 7850 the end of the boot process. 7851 7852 traceoff_on_warning 7853 [FTRACE] enable this option to disable tracing when a 7854 warning is hit. This turns off "tracing_on". Tracing can 7855 be enabled again by echoing '1' into the "tracing_on" 7856 file located in /sys/kernel/tracing/ 7857 7858 This option is useful, as it disables the trace before 7859 the WARNING dump is called, which prevents the trace to 7860 be filled with content caused by the warning output. 7861 7862 This option can also be set at run time via the sysctl 7863 option: kernel/traceoff_on_warning 7864 7865 transparent_hugepage= 7866 [KNL] 7867 Format: [always|madvise|never] 7868 Can be used to control the default behavior of the system 7869 with respect to transparent hugepages. 7870 See Documentation/admin-guide/mm/transhuge.rst 7871 for more details. 7872 7873 transparent_hugepage_shmem= [KNL] 7874 Format: [always|within_size|advise|never|deny|force] 7875 Can be used to control the hugepage allocation policy for 7876 the internal shmem mount. 7877 See Documentation/admin-guide/mm/transhuge.rst 7878 for more details. 7879 7880 transparent_hugepage_tmpfs= [KNL] 7881 Format: [always|within_size|advise|never] 7882 Can be used to control the default hugepage allocation policy 7883 for the tmpfs mount. 7884 See Documentation/admin-guide/mm/transhuge.rst 7885 for more details. 7886 7887 trusted.source= [KEYS] 7888 Format: <string> 7889 This parameter identifies the trust source as a backend 7890 for trusted keys implementation. Supported trust 7891 sources: 7892 - "tpm" 7893 - "tee" 7894 - "caam" 7895 - "dcp" 7896 - "pkwm" 7897 If not specified then it defaults to iterating through 7898 the trust source list starting with TPM and assigns the 7899 first trust source as a backend which is initialized 7900 successfully during iteration. 7901 7902 trusted.rng= [KEYS] 7903 Format: <string> 7904 The RNG used to generate key material for trusted keys. 7905 Can be one of: 7906 - "kernel" 7907 - the same value as trusted.source: "tpm" or "tee" 7908 - "default" 7909 If not specified, "default" is used. In this case, 7910 the RNG's choice is left to each individual trust source. 7911 7912 trusted.dcp_use_otp_key 7913 This is intended to be used in combination with 7914 trusted.source=dcp and will select the DCP OTP key 7915 instead of the DCP UNIQUE key blob encryption. 7916 7917 trusted.dcp_skip_zk_test 7918 This is intended to be used in combination with 7919 trusted.source=dcp and will disable the check if the 7920 blob key is all zeros. This is helpful for situations where 7921 having this key zero'ed is acceptable. E.g. in testing 7922 scenarios. 7923 7924 tsa= [X86] Control mitigation for Transient Scheduler 7925 Attacks on AMD CPUs. Search the following in your 7926 favourite search engine for more details: 7927 7928 "Technical guidance for mitigating transient scheduler 7929 attacks". 7930 7931 off - disable the mitigation 7932 on - enable the mitigation (default) 7933 user - mitigate only user/kernel transitions 7934 vm - mitigate only guest/host transitions 7935 7936 7937 tsc= Disable clocksource stability checks for TSC. 7938 Format: <string> 7939 [x86] reliable: mark tsc clocksource as reliable, this 7940 disables clocksource verification at runtime, as well 7941 as the stability checks done at bootup. Used to enable 7942 high-resolution timer mode on older hardware, and in 7943 virtualized environment. 7944 [x86] noirqtime: Do not use TSC to do irq accounting. 7945 Used to run time disable IRQ_TIME_ACCOUNTING on any 7946 platforms where RDTSC is slow and this accounting 7947 can add overhead. 7948 [x86] unstable: mark the TSC clocksource as unstable, this 7949 marks the TSC unconditionally unstable at bootup and 7950 avoids any further wobbles once the TSC watchdog notices. 7951 [x86] nowatchdog: disable clocksource watchdog. Used 7952 in situations with strict latency requirements (where 7953 interruptions from clocksource watchdog are not 7954 acceptable). 7955 [x86] recalibrate: force recalibration against a HW timer 7956 (HPET or PM timer) on systems whose TSC frequency was 7957 obtained from HW or FW using either an MSR or CPUID(0x15). 7958 Warn if the difference is more than 500 ppm. 7959 [x86] watchdog: Enforce the clocksource watchdog on TSC 7960 7961 tsc_early_khz= [X86,EARLY] Skip early TSC calibration and use the given 7962 value instead. Useful when the early TSC frequency discovery 7963 procedure is not reliable, such as on overclocked systems 7964 with CPUID.16h support and partial CPUID.15h support. 7965 Format: <unsigned int> 7966 7967 tsx= [X86] Control Transactional Synchronization 7968 Extensions (TSX) feature in Intel processors that 7969 support TSX control. 7970 7971 This parameter controls the TSX feature. The options are: 7972 7973 on - Enable TSX on the system. Although there are 7974 mitigations for all known security vulnerabilities, 7975 TSX has been known to be an accelerator for 7976 several previous speculation-related CVEs, and 7977 so there may be unknown security risks associated 7978 with leaving it enabled. 7979 7980 off - Disable TSX on the system. (Note that this 7981 option takes effect only on newer CPUs which are 7982 not vulnerable to MDS, i.e., have 7983 MSR_IA32_ARCH_CAPABILITIES.MDS_NO=1 and which get 7984 the new IA32_TSX_CTRL MSR through a microcode 7985 update. This new MSR allows for the reliable 7986 deactivation of the TSX functionality.) 7987 7988 auto - Disable TSX if X86_BUG_TAA is present, 7989 otherwise enable TSX on the system. 7990 7991 Not specifying this option is equivalent to tsx=off. 7992 7993 See Documentation/admin-guide/hw-vuln/tsx_async_abort.rst 7994 for more details. 7995 7996 tsx_async_abort= [X86,INTEL,EARLY] Control mitigation for the TSX Async 7997 Abort (TAA) vulnerability. 7998 7999 Similar to Micro-architectural Data Sampling (MDS) 8000 certain CPUs that support Transactional 8001 Synchronization Extensions (TSX) are vulnerable to an 8002 exploit against CPU internal buffers which can forward 8003 information to a disclosure gadget under certain 8004 conditions. 8005 8006 In vulnerable processors, the speculatively forwarded 8007 data can be used in a cache side channel attack, to 8008 access data to which the attacker does not have direct 8009 access. 8010 8011 This parameter controls the TAA mitigation. The 8012 options are: 8013 8014 full - Enable TAA mitigation on vulnerable CPUs 8015 if TSX is enabled. 8016 8017 full,nosmt - Enable TAA mitigation and disable SMT on 8018 vulnerable CPUs. If TSX is disabled, SMT 8019 is not disabled because CPU is not 8020 vulnerable to cross-thread TAA attacks. 8021 off - Unconditionally disable TAA mitigation 8022 8023 On MDS-affected machines, tsx_async_abort=off can be 8024 prevented by an active MDS mitigation as both vulnerabilities 8025 are mitigated with the same mechanism so in order to disable 8026 this mitigation, you need to specify mds=off too. 8027 8028 Not specifying this option is equivalent to 8029 tsx_async_abort=full. On CPUs which are MDS affected 8030 and deploy MDS mitigation, TAA mitigation is not 8031 required and doesn't provide any additional 8032 mitigation. 8033 8034 For details see: 8035 Documentation/admin-guide/hw-vuln/tsx_async_abort.rst 8036 8037 turbografx.map[2|3]= [HW,JOY] 8038 TurboGraFX parallel port interface 8039 Format: 8040 <port#>,<js1>,<js2>,<js3>,<js4>,<js5>,<js6>,<js7> 8041 See also Documentation/input/devices/joystick-parport.rst 8042 8043 udbg-immortal [PPC] When debugging early kernel crashes that 8044 happen after console_init() and before a proper 8045 console driver takes over, this boot options might 8046 help "seeing" what's going on. 8047 8048 uhash_entries= [KNL,NET] 8049 Set number of hash buckets for UDP/UDP-Lite connections 8050 8051 uhci-hcd.ignore_oc= 8052 [USB] Ignore overcurrent events (default N). 8053 Some badly-designed motherboards generate lots of 8054 bogus events, for ports that aren't wired to 8055 anything. Set this parameter to avoid log spamming. 8056 Note that genuine overcurrent events won't be 8057 reported either. 8058 8059 unaligned_scalar_speed= 8060 [RISCV] 8061 Format: {slow | fast | unsupported} 8062 Allow skipping scalar unaligned access speed tests. This 8063 is useful for testing alternative code paths and to skip 8064 the tests in environments where they run too slowly. All 8065 CPUs must have the same scalar unaligned access speed. 8066 8067 unaligned_vector_speed= 8068 [RISCV] 8069 Format: {slow | fast | unsupported} 8070 Allow skipping vector unaligned access speed tests. This 8071 is useful for testing alternative code paths and to skip 8072 the tests in environments where they run too slowly. All 8073 CPUs must have the same vector unaligned access speed. 8074 8075 unknown_nmi_panic 8076 [X86] Cause panic on unknown NMI. 8077 8078 unwind_debug [X86-64,EARLY] 8079 Enable unwinder debug output. This can be 8080 useful for debugging certain unwinder error 8081 conditions, including corrupt stacks and 8082 bad/missing unwinder metadata. 8083 8084 usbcore.authorized_default= 8085 [USB] Default USB device authorization: 8086 (default -1 = authorized (same as 1), 8087 0 = not authorized, 1 = authorized, 2 = authorized 8088 if device connected to internal port) 8089 8090 usbcore.autosuspend= 8091 [USB] The autosuspend time delay (in seconds) used 8092 for newly-detected USB devices (default 2). This 8093 is the time required before an idle device will be 8094 autosuspended. Devices for which the delay is set 8095 to a negative value won't be autosuspended at all. 8096 8097 usbcore.usbfs_snoop= 8098 [USB] Set to log all usbfs traffic (default 0 = off). 8099 8100 usbcore.usbfs_snoop_max= 8101 [USB] Maximum number of bytes to snoop in each URB 8102 (default = 65536). 8103 8104 usbcore.blinkenlights= 8105 [USB] Set to cycle leds on hubs (default 0 = off). 8106 8107 usbcore.old_scheme_first= 8108 [USB] Start with the old device initialization 8109 scheme (default 0 = off). 8110 8111 usbcore.usbfs_memory_mb= 8112 [USB] Memory limit (in MB) for buffers allocated by 8113 usbfs (default = 16, 0 = max = 2047). 8114 8115 usbcore.use_both_schemes= 8116 [USB] Try the other device initialization scheme 8117 if the first one fails (default 1 = enabled). 8118 8119 usbcore.initial_descriptor_timeout= 8120 [USB] Specifies timeout for the initial 64-byte 8121 USB_REQ_GET_DESCRIPTOR request in milliseconds 8122 (default 5000 = 5.0 seconds). 8123 8124 usbcore.nousb [USB] Disable the USB subsystem 8125 8126 usbcore.quirks= 8127 [USB] A list of quirk entries to augment the built-in 8128 usb core quirk list. List entries are separated by 8129 commas. Each entry has the form 8130 VendorID:ProductID:Flags. The IDs are 4-digit hex 8131 numbers and Flags is a set of letters. Each letter 8132 will change the built-in quirk; setting it if it is 8133 clear and clearing it if it is set. The letters have 8134 the following meanings: 8135 a = USB_QUIRK_STRING_FETCH_255 (string 8136 descriptors must not be fetched using 8137 a 255-byte read); 8138 b = USB_QUIRK_RESET_RESUME (device can't resume 8139 correctly so reset it instead); 8140 c = USB_QUIRK_NO_SET_INTF (device can't handle 8141 Set-Interface requests); 8142 d = USB_QUIRK_CONFIG_INTF_STRINGS (device can't 8143 handle its Configuration or Interface 8144 strings); 8145 e = USB_QUIRK_RESET (device can't be reset 8146 (e.g morph devices), don't use reset); 8147 f = USB_QUIRK_HONOR_BNUMINTERFACES (device has 8148 more interface descriptions than the 8149 bNumInterfaces count, and can't handle 8150 talking to these interfaces); 8151 g = USB_QUIRK_DELAY_INIT (device needs a pause 8152 during initialization, after we read 8153 the device descriptor); 8154 h = USB_QUIRK_LINEAR_UFRAME_INTR_BINTERVAL (For 8155 high speed and super speed interrupt 8156 endpoints, the USB 2.0 and USB 3.0 spec 8157 require the interval in microframes (1 8158 microframe = 125 microseconds) to be 8159 calculated as interval = 2 ^ 8160 (bInterval-1). 8161 Devices with this quirk report their 8162 bInterval as the result of this 8163 calculation instead of the exponent 8164 variable used in the calculation); 8165 i = USB_QUIRK_DEVICE_QUALIFIER (device can't 8166 handle device_qualifier descriptor 8167 requests); 8168 j = USB_QUIRK_IGNORE_REMOTE_WAKEUP (device 8169 generates spurious wakeup, ignore 8170 remote wakeup capability); 8171 k = USB_QUIRK_NO_LPM (device can't handle Link 8172 Power Management); 8173 l = USB_QUIRK_LINEAR_FRAME_INTR_BINTERVAL 8174 (Device reports its bInterval as linear 8175 frames instead of the USB 2.0 8176 calculation); 8177 m = USB_QUIRK_DISCONNECT_SUSPEND (Device needs 8178 to be disconnected before suspend to 8179 prevent spurious wakeup); 8180 n = USB_QUIRK_DELAY_CTRL_MSG (Device needs a 8181 pause after every control message); 8182 o = USB_QUIRK_HUB_SLOW_RESET (Hub needs extra 8183 delay after resetting its port); 8184 p = USB_QUIRK_SHORT_SET_ADDRESS_REQ_TIMEOUT 8185 (Reduce timeout of the SET_ADDRESS 8186 request from 5000 ms to 500 ms); 8187 q = USB_QUIRK_FORCE_ONE_CONFIG (Device 8188 claims zero configurations, 8189 forcing to 1); 8190 Example: quirks=0781:5580:bk,0a5c:5834:gij 8191 8192 usbhid.mousepoll= 8193 [USBHID] The interval which mice are to be polled at. 8194 8195 usbhid.jspoll= 8196 [USBHID] The interval which joysticks are to be polled at. 8197 8198 usbhid.kbpoll= 8199 [USBHID] The interval which keyboards are to be polled at. 8200 8201 usb-storage.delay_use= 8202 [UMS] The delay in seconds before a new device is 8203 scanned for Logical Units (default 1). 8204 Optionally the delay in milliseconds if the value has 8205 suffix with "ms". 8206 Example: delay_use=2567ms 8207 8208 usb-storage.quirks= 8209 [UMS] A list of quirks entries to supplement or 8210 override the built-in unusual_devs list. List 8211 entries are separated by commas. Each entry has 8212 the form VID:PID:Flags where VID and PID are Vendor 8213 and Product ID values (4-digit hex numbers) and 8214 Flags is a set of characters, each corresponding 8215 to a common usb-storage quirk flag as follows: 8216 a = SANE_SENSE (collect more than 18 bytes 8217 of sense data, not on uas); 8218 b = BAD_SENSE (don't collect more than 18 8219 bytes of sense data, not on uas); 8220 c = FIX_CAPACITY (decrease the reported 8221 device capacity by one sector); 8222 d = NO_READ_DISC_INFO (don't use 8223 READ_DISC_INFO command, not on uas); 8224 e = NO_READ_CAPACITY_16 (don't use 8225 READ_CAPACITY_16 command); 8226 f = NO_REPORT_OPCODES (don't use report opcodes 8227 command, uas only); 8228 g = MAX_SECTORS_240 (don't transfer more than 8229 240 sectors at a time, uas only); 8230 h = CAPACITY_HEURISTICS (decrease the 8231 reported device capacity by one 8232 sector if the number is odd); 8233 i = IGNORE_DEVICE (don't bind to this 8234 device); 8235 j = NO_REPORT_LUNS (don't use report luns 8236 command, uas only); 8237 k = NO_SAME (do not use WRITE_SAME, uas only) 8238 l = NOT_LOCKABLE (don't try to lock and 8239 unlock ejectable media, not on uas); 8240 m = MAX_SECTORS_64 (don't transfer more 8241 than 64 sectors = 32 KB at a time, 8242 not on uas); 8243 n = INITIAL_READ10 (force a retry of the 8244 initial READ(10) command, not on uas); 8245 o = CAPACITY_OK (accept the capacity 8246 reported by the device, not on uas); 8247 p = WRITE_CACHE (the device cache is ON 8248 by default, not on uas); 8249 r = IGNORE_RESIDUE (the device reports 8250 bogus residue values, not on uas); 8251 s = SINGLE_LUN (the device has only one 8252 Logical Unit); 8253 t = NO_ATA_1X (don't allow ATA(12) and ATA(16) 8254 commands, uas only); 8255 u = IGNORE_UAS (don't bind to the uas driver); 8256 w = NO_WP_DETECT (don't test whether the 8257 medium is write-protected). 8258 y = ALWAYS_SYNC (issue a SYNCHRONIZE_CACHE 8259 even if the device claims no cache, 8260 not on uas) 8261 Example: quirks=0419:aaf5:rl,0421:0433:rc 8262 8263 user_debug= [KNL,ARM] 8264 Format: <int> 8265 See arch/arm/Kconfig.debug help text. 8266 1 - undefined instruction events 8267 2 - system calls 8268 4 - invalid data aborts 8269 8 - SIGSEGV faults 8270 16 - SIGBUS faults 8271 Example: user_debug=31 8272 8273 vdso= [X86,SH,SPARC] 8274 On X86_32, this is an alias for vdso32=. Otherwise: 8275 8276 vdso=1: enable VDSO (the default) 8277 vdso=0: disable VDSO mapping 8278 8279 vdso32= [X86] Control the 32-bit vDSO 8280 vdso32=1: enable 32-bit VDSO 8281 vdso32=0 or vdso32=2: disable 32-bit VDSO 8282 8283 See the help text for CONFIG_COMPAT_VDSO for more 8284 details. If CONFIG_COMPAT_VDSO is set, the default is 8285 vdso32=0; otherwise, the default is vdso32=1. 8286 8287 For compatibility with older kernels, vdso32=2 is an 8288 alias for vdso32=0. 8289 8290 Try vdso32=0 if you encounter an error that says: 8291 dl_main: Assertion `(void *) ph->p_vaddr == _rtld_local._dl_sysinfo_dso' failed! 8292 8293 video= [FB,EARLY] Frame buffer configuration 8294 See Documentation/fb/modedb.rst. 8295 8296 video.brightness_switch_enabled= [ACPI] 8297 Format: [0|1] 8298 If set to 1, on receiving an ACPI notify event 8299 generated by hotkey, video driver will adjust brightness 8300 level and then send out the event to user space through 8301 the allocated input device. If set to 0, video driver 8302 will only send out the event without touching backlight 8303 brightness level. 8304 default: 1 8305 8306 virtio_mmio.device= 8307 [VMMIO] Memory mapped virtio (platform) device. 8308 8309 <size>@<baseaddr>:<irq>[:<id>] 8310 where: 8311 <size> := size (can use standard suffixes 8312 like K, M and G) 8313 <baseaddr> := physical base address 8314 <irq> := interrupt number (as passed to 8315 request_irq()) 8316 <id> := (optional) platform device id 8317 example: 8318 virtio_mmio.device=1K@0x100b0000:48:7 8319 8320 Can be used multiple times for multiple devices. 8321 8322 vga= [BOOT,X86-32] Select a particular video mode 8323 See Documentation/arch/x86/boot.rst and 8324 Documentation/admin-guide/svga.rst. 8325 Use vga=ask for menu. 8326 This is actually a boot loader parameter; the value is 8327 passed to the kernel using a special protocol. 8328 8329 vm_debug[=options] [KNL] Available with CONFIG_DEBUG_VM=y. 8330 May slow down system boot speed, especially when 8331 enabled on systems with a large amount of memory. 8332 All options are enabled by default, and this 8333 interface is meant to allow for selectively 8334 enabling or disabling specific virtual memory 8335 debugging features. 8336 8337 Available options are: 8338 P Enable page structure init time poisoning 8339 - Disable all of the above options 8340 8341 vmalloc=nn[KMG] [KNL,BOOT,EARLY] Forces the vmalloc area to have an 8342 exact size of <nn>. This can be used to increase 8343 the minimum size (128MB on x86, arm32 platforms). 8344 It can also be used to decrease the size and leave more room 8345 for directly mapped kernel RAM. Note that this parameter does 8346 not exist on many other platforms (including arm64, alpha, 8347 loongarch, arc, csky, hexagon, microblaze, mips, nios2, openrisc, 8348 parisc, m64k, powerpc, riscv, sh, um, xtensa, s390, sparc). 8349 8350 vmcp_cma=nn[MG] [KNL,S390,EARLY] 8351 Sets the memory size reserved for contiguous memory 8352 allocations for the vmcp device driver. 8353 8354 vmhalt= [KNL,S390] Perform z/VM CP command after system halt. 8355 Format: <command> 8356 8357 vmpanic= [KNL,S390] Perform z/VM CP command after kernel panic. 8358 Format: <command> 8359 8360 vmpoff= [KNL,S390] Perform z/VM CP command after power off. 8361 Format: <command> 8362 8363 vmscape= [X86] Controls mitigation for VMscape attacks. 8364 VMscape attacks can leak information from a userspace 8365 hypervisor to a guest via speculative side-channels. 8366 8367 off - disable the mitigation 8368 ibpb - use Indirect Branch Prediction Barrier 8369 (IBPB) mitigation (default) 8370 force - force vulnerability detection even on 8371 unaffected processors 8372 8373 vsyscall= [X86-64,EARLY] 8374 Controls the behavior of vsyscalls (i.e. calls to 8375 fixed addresses of 0xffffffffff600x00 from legacy 8376 code). Most statically-linked binaries and older 8377 versions of glibc use these calls. Because these 8378 functions are at fixed addresses, they make nice 8379 targets for exploits that can control RIP. 8380 8381 emulate Vsyscalls turn into traps and are emulated 8382 reasonably safely. The vsyscall page is 8383 readable. This disables the Linear 8384 Address Space Separation (LASS) security 8385 feature and makes the system less secure. 8386 8387 xonly [default] Vsyscalls turn into traps and are 8388 emulated reasonably safely. The vsyscall 8389 page is not readable. 8390 8391 none Vsyscalls don't work at all. This makes 8392 them quite hard to use for exploits but 8393 might break your system. 8394 8395 vt.color= [VT] Default text color. 8396 Format: 0xYX, X = foreground, Y = background. 8397 Default: 0x07 = light gray on black. 8398 8399 vt.cur_default= [VT] Default cursor shape. 8400 Format: 0xCCBBAA, where AA, BB, and CC are the same as 8401 the parameters of the <Esc>[?A;B;Cc escape sequence; 8402 see vga-softcursor.rst. Default: 2 = underline. 8403 8404 vt.default_blu= [VT] 8405 Format: <blue0>,<blue1>,<blue2>,...,<blue15> 8406 Change the default blue palette of the console. 8407 This is a 16-member array composed of values 8408 ranging from 0-255. 8409 8410 vt.default_grn= [VT] 8411 Format: <green0>,<green1>,<green2>,...,<green15> 8412 Change the default green palette of the console. 8413 This is a 16-member array composed of values 8414 ranging from 0-255. 8415 8416 vt.default_red= [VT] 8417 Format: <red0>,<red1>,<red2>,...,<red15> 8418 Change the default red palette of the console. 8419 This is a 16-member array composed of values 8420 ranging from 0-255. 8421 8422 vt.default_utf8= 8423 [VT] 8424 Format=<0|1> 8425 Set system-wide default UTF-8 mode for all tty's. 8426 Default is 1, i.e. UTF-8 mode is enabled for all 8427 newly opened terminals. 8428 8429 vt.global_cursor_default= 8430 [VT] 8431 Format=<-1|0|1> 8432 Set system-wide default for whether a cursor 8433 is shown on new VTs. Default is -1, 8434 i.e. cursors will be created by default unless 8435 overridden by individual drivers. 0 will hide 8436 cursors, 1 will display them. 8437 8438 vt.italic= [VT] Default color for italic text; 0-15. 8439 Default: 2 = green. 8440 8441 vt.underline= [VT] Default color for underlined text; 0-15. 8442 Default: 3 = cyan. 8443 8444 watchdog timers [HW,WDT] For information on watchdog timers, 8445 see Documentation/watchdog/watchdog-parameters.rst 8446 or other driver-specific files in the 8447 Documentation/watchdog/ directory. 8448 8449 watchdog_thresh= 8450 [KNL] 8451 Set the hard lockup detector stall duration 8452 threshold in seconds. The soft lockup detector 8453 threshold is set to twice the value. A value of 0 8454 disables both lockup detectors. Default is 10 8455 seconds. 8456 8457 workqueue.unbound_cpus= 8458 [KNL,SMP] Specify to constrain one or some CPUs 8459 to use in unbound workqueues. 8460 Format: <cpu-list> 8461 By default, all online CPUs are available for 8462 unbound workqueues. 8463 8464 workqueue.watchdog_thresh= 8465 If CONFIG_WQ_WATCHDOG is configured, workqueue can 8466 warn stall conditions and dump internal state to 8467 help debugging. 0 disables workqueue stall 8468 detection; otherwise, it's the stall threshold 8469 duration in seconds. The default value is 30 and 8470 it can be updated at runtime by writing to the 8471 corresponding sysfs file. 8472 8473 workqueue.panic_on_stall=<uint> 8474 Panic when workqueue stall is detected by 8475 CONFIG_WQ_WATCHDOG. It sets the number times of the 8476 stall to trigger panic. 8477 8478 The default is set by CONFIG_BOOTPARAM_WQ_STALL_PANIC, 8479 which is 0 (disabled) if not configured. 8480 8481 workqueue.panic_on_stall_time=<uint> 8482 Panic when a workqueue stall has been continuous for 8483 the specified number of seconds. Unlike panic_on_stall 8484 which counts accumulated stall events, this triggers 8485 based on the duration of a single continuous stall. 8486 8487 The default is 0, which disables the time-based panic. 8488 8489 workqueue.cpu_intensive_thresh_us= 8490 Per-cpu work items which run for longer than this 8491 threshold are automatically considered CPU intensive 8492 and excluded from concurrency management to prevent 8493 them from noticeably delaying other per-cpu work 8494 items. Default is 10000 (10ms). 8495 8496 If CONFIG_WQ_CPU_INTENSIVE_REPORT is set, the kernel 8497 will report the work functions which violate this 8498 threshold repeatedly. They are likely good 8499 candidates for using WQ_UNBOUND workqueues instead. 8500 8501 workqueue.cpu_intensive_warning_thresh=<uint> 8502 If CONFIG_WQ_CPU_INTENSIVE_REPORT is set, the kernel 8503 will report the work functions which violate the 8504 intensive_threshold_us repeatedly. In order to prevent 8505 spurious warnings, start printing only after a work 8506 function has violated this threshold number of times. 8507 8508 The default is 4 times. 0 disables the warning. 8509 8510 workqueue.power_efficient 8511 Per-cpu workqueues are generally preferred because 8512 they show better performance thanks to cache 8513 locality; unfortunately, per-cpu workqueues tend to 8514 be more power hungry than unbound workqueues. 8515 8516 Enabling this makes the per-cpu workqueues which 8517 were observed to contribute significantly to power 8518 consumption unbound, leading to measurably lower 8519 power usage at the cost of small performance 8520 overhead. 8521 8522 The default value of this parameter is determined by 8523 the config option CONFIG_WQ_POWER_EFFICIENT_DEFAULT. 8524 8525 workqueue.default_affinity_scope= 8526 Select the default affinity scope to use for unbound 8527 workqueues. Can be one of "cpu", "smt", "cache", 8528 "numa" and "system". Default is "cache". For more 8529 information, see the Affinity Scopes section in 8530 Documentation/core-api/workqueue.rst. 8531 8532 This can be changed after boot by writing to the 8533 matching /sys/module/workqueue/parameters file. All 8534 workqueues with the "default" affinity scope will be 8535 updated accordingly. 8536 8537 workqueue.debug_force_rr_cpu 8538 Workqueue used to implicitly guarantee that work 8539 items queued without explicit CPU specified are put 8540 on the local CPU. This guarantee is no longer true 8541 and while local CPU is still preferred work items 8542 may be put on foreign CPUs. This debug option 8543 forces round-robin CPU selection to flush out 8544 usages which depend on the now broken guarantee. 8545 When enabled, memory and cache locality will be 8546 impacted. 8547 8548 writecombine= [LOONGARCH,EARLY] Control the MAT (Memory Access 8549 Type) of ioremap_wc(). 8550 8551 on - Enable writecombine, use WUC for ioremap_wc() 8552 off - Disable writecombine, use SUC for ioremap_wc() 8553 8554 x2apic_phys [X86-64,APIC,EARLY] Use x2apic physical mode instead of 8555 default x2apic cluster mode on platforms 8556 supporting x2apic. 8557 8558 xen_512gb_limit [KNL,X86-64,XEN] 8559 Restricts the kernel running paravirtualized under Xen 8560 to use only up to 512 GB of RAM. The reason to do so is 8561 crash analysis tools and Xen tools for doing domain 8562 save/restore/migration must be enabled to handle larger 8563 domains. 8564 8565 xen_console_io [XEN,EARLY] 8566 Boolean option to enable/disable the usage of the Xen 8567 console_io hypercalls to read and write to the console. 8568 Mostly useful for debugging and development. 8569 8570 xen_emul_unplug= [HW,X86,XEN,EARLY] 8571 Unplug Xen emulated devices 8572 Format: [unplug0,][unplug1] 8573 ide-disks -- unplug primary master IDE devices 8574 aux-ide-disks -- unplug non-primary-master IDE devices 8575 nics -- unplug network devices 8576 all -- unplug all emulated devices (NICs and IDE disks) 8577 unnecessary -- unplugging emulated devices is 8578 unnecessary even if the host did not respond to 8579 the unplug protocol 8580 never -- do not unplug even if version check succeeds 8581 8582 xen_legacy_crash [X86,XEN,EARLY] 8583 Crash from Xen panic notifier, without executing late 8584 panic() code such as dumping handler. 8585 8586 xen_mc_debug [X86,XEN,EARLY] 8587 Enable multicall debugging when running as a Xen PV guest. 8588 Enabling this feature will reduce performance a little 8589 bit, so it should only be enabled for obtaining extended 8590 debug data in case of multicall errors. 8591 8592 xen_msr_safe= [X86,XEN,EARLY] 8593 Format: <bool> 8594 Select whether to always use non-faulting (safe) MSR 8595 access functions when running as Xen PV guest. The 8596 default value is controlled by CONFIG_XEN_PV_MSR_SAFE. 8597 8598 xen_nopv [X86] 8599 Disables the PV optimizations forcing the HVM guest to 8600 run as generic HVM guest with no PV drivers. 8601 This option is obsoleted by the "nopv" option, which 8602 has equivalent effect for XEN platform. 8603 8604 xen_no_vector_callback 8605 [KNL,X86,XEN,EARLY] Disable the vector callback for Xen 8606 event channel interrupts. 8607 8608 xen_scrub_pages= [XEN] 8609 Boolean option to control scrubbing pages before giving them back 8610 to Xen, for use by other domains. Can be also changed at runtime 8611 with /sys/devices/system/xen_memory/xen_memory0/scrub_pages. 8612 Default value controlled with CONFIG_XEN_SCRUB_PAGES_DEFAULT. 8613 8614 xen_timer_slop= [X86-64,XEN,EARLY] 8615 Set the timer slop (in nanoseconds) for the virtual Xen 8616 timers (default is 100000). This adjusts the minimum 8617 delta of virtualized Xen timers, where lower values 8618 improve timer resolution at the expense of processing 8619 more timer interrupts. 8620 8621 xen.balloon_boot_timeout= [XEN] 8622 The time (in seconds) to wait before giving up to boot 8623 in case initial ballooning fails to free enough memory. 8624 Applies only when running as HVM or PVH guest and 8625 started with less memory configured than allowed at 8626 max. Default is 180. 8627 8628 xen.event_eoi_delay= [XEN] 8629 How long to delay EOI handling in case of event 8630 storms (jiffies). Default is 10. 8631 8632 xen.event_loop_timeout= [XEN] 8633 After which time (jiffies) the event handling loop 8634 should start to delay EOI handling. Default is 2. 8635 8636 xen.fifo_events= [XEN] 8637 Boolean parameter to disable using fifo event handling 8638 even if available. Normally fifo event handling is 8639 preferred over the 2-level event handling, as it is 8640 fairer and the number of possible event channels is 8641 much higher. Default is on (use fifo events). 8642 8643 xirc2ps_cs= [NET,PCMCIA] 8644 Format: 8645 <irq>,<irq_mask>,<io>,<full_duplex>,<do_sound>,<lockup_hack>[,<irq2>[,<irq3>[,<irq4>]]] 8646 8647 xive= [PPC] 8648 By default on POWER9 and above, the kernel will 8649 natively use the XIVE interrupt controller. This option 8650 allows the fallback firmware mode to be used: 8651 8652 off Fallback to firmware control of XIVE interrupt 8653 controller on both pseries and powernv 8654 platforms. Only useful on POWER9 and above. 8655 8656 xive.store-eoi=off [PPC] 8657 By default on POWER10 and above, the kernel will use 8658 stores for EOI handling when the XIVE interrupt mode 8659 is active. This option allows the XIVE driver to use 8660 loads instead, as on POWER9. 8661 8662 xhci-hcd.quirks [USB,KNL] 8663 A hex value specifying bitmask with supplemental xhci 8664 host controller quirks. Meaning of each bit can be 8665 consulted in header drivers/usb/host/xhci.h. 8666 8667 xmon [PPC,EARLY] 8668 Format: { early | on | rw | ro | off } 8669 Controls if xmon debugger is enabled. Default is off. 8670 Passing only "xmon" is equivalent to "xmon=early". 8671 early Call xmon as early as possible on boot; xmon 8672 debugger is called from setup_arch(). 8673 on xmon debugger hooks will be installed so xmon 8674 is only called on a kernel crash. Default mode, 8675 i.e. either "ro" or "rw" mode, is controlled 8676 with CONFIG_XMON_DEFAULT_RO_MODE. 8677 rw xmon debugger hooks will be installed so xmon 8678 is called only on a kernel crash, mode is write, 8679 meaning SPR registers, memory and, other data 8680 can be written using xmon commands. 8681 ro same as "rw" option above but SPR registers, 8682 memory, and other data can't be written using 8683 xmon commands. 8684 off xmon is disabled. 8685