1 ACPI ACPI support is enabled. 2 AGP AGP (Accelerated Graphics Port) is enabled. 3 ALSA ALSA sound support is enabled. 4 APIC APIC support is enabled. 5 APM Advanced Power Management support is enabled. 6 APPARMOR AppArmor support is enabled. 7 ARM ARM architecture is enabled. 8 ARM64 ARM64 architecture is enabled. 9 CLK Common clock infrastructure is enabled. 10 CMA Contiguous Memory Area support is enabled. 11 CPU_FREQ CPU frequency scaling is enabled. 12 CPU_IDLE CPU idle time management is enabled. 13 DRM Direct Rendering Management support is enabled. 14 DYNAMIC_DEBUG Build in debug messages and enable them at runtime 15 EARLY Parameter processed too early to be embedded in initrd. 16 EDD BIOS Enhanced Disk Drive Services (EDD) is enabled 17 EFI EFI Partitioning (GPT) is enabled 18 EVM Extended Verification Module 19 FB The frame buffer device is enabled. 20 FTRACE Function tracing enabled. 21 GCOV GCOV profiling is enabled. 22 HIBERNATION HIBERNATION is enabled. 23 HW Appropriate hardware is enabled. 24 HYPER_V HYPERV support is enabled. 25 IMA Integrity measurement architecture is enabled. 26 IP_PNP IP DHCP, BOOTP, or RARP is enabled. 27 IPV6 IPv6 support is enabled. 28 ISAPNP ISA PnP code is enabled. 29 ISOL CPU Isolation is enabled. 30 JOY Appropriate joystick support is enabled. 31 KGDB Kernel debugger support is enabled. 32 KVM Kernel Virtual Machine support is enabled. 33 LIBATA Libata driver is enabled 34 LOONGARCH LoongArch architecture is enabled. 35 LOOP Loopback device support is enabled. 36 LP Printer support is enabled. 37 M68k M68k architecture is enabled. 38 These options have more detailed description inside of 39 Documentation/arch/m68k/kernel-options.rst. 40 MIPS MIPS architecture is enabled. 41 MOUSE Appropriate mouse support is enabled. 42 MSI Message Signaled Interrupts (PCI). 43 MTD MTD (Memory Technology Device) support is enabled. 44 NET Appropriate network support is enabled. 45 NFS Appropriate NFS support is enabled. 46 NUMA NUMA support is enabled. 47 OF Devicetree is enabled. 48 PARISC The PA-RISC architecture is enabled. 49 PCI PCI bus support is enabled. 50 PCIE PCI Express support is enabled. 51 PCMCIA The PCMCIA subsystem is enabled. 52 PM Power Management support is enabled. 53 PNP Plug & Play support is enabled. 54 PPC PowerPC architecture is enabled. 55 PPT Parallel port support is enabled. 56 PS2 Appropriate PS/2 support is enabled. 57 PV_OPS A paravirtualized kernel is enabled. 58 RAM RAM disk support is enabled. 59 RDT Intel Resource Director Technology. 60 RISCV RISCV architecture is enabled. 61 S390 S390 architecture is enabled. 62 SCSI Appropriate SCSI support is enabled. 63 A lot of drivers have their options described inside 64 the Documentation/scsi/ sub-directory. 65 SDW SoundWire support is enabled. 66 SECURITY Different security models are enabled. 67 SELINUX SELinux support is enabled. 68 SERIAL Serial support is enabled. 69 SH SuperH architecture is enabled. 70 SMP The kernel is an SMP kernel. 71 SPARC Sparc architecture is enabled. 72 SUSPEND System suspend states are enabled. 73 SWSUSP Software suspend (hibernation) is enabled. 74 TPM TPM drivers are enabled. 75 UMS USB Mass Storage support is enabled. 76 USB USB support is enabled. 77 NVME NVMe support is enabled 78 USBHID USB Human Interface Device support is enabled. 79 V4L Video For Linux support is enabled. 80 VGA The VGA console has been enabled. 81 VMMIO Driver for memory mapped virtio devices is enabled. 82 VT Virtual terminal support is enabled. 83 WDT Watchdog support is enabled. 84 X86-32 X86-32, aka i386 architecture is enabled. 85 X86-64 X86-64 architecture is enabled. 86 X86 Either 32-bit or 64-bit x86 (same as X86-32+X86-64) 87 X86_UV SGI UV support is enabled. 88 XEN Xen support is enabled 89 XTENSA xtensa architecture is enabled. 90 91In addition, the following text indicates that the option 92 93 BOOT Is a boot loader parameter. 94 BUGS= Relates to possible processor bugs on the said processor. 95 KNL Is a kernel start-up parameter. 96 97 98Kernel parameters 99 100 accept_memory= [MM] 101 Format: { eager | lazy } 102 default: lazy 103 By default, unaccepted memory is accepted lazily to 104 avoid prolonged boot times. The lazy option will add 105 some runtime overhead until all memory is eventually 106 accepted. In most cases the overhead is negligible. 107 For some workloads or for debugging purposes 108 accept_memory=eager can be used to accept all memory 109 at once during boot. 110 111 acpi= [HW,ACPI,X86,ARM64,RISCV64,EARLY] 112 Advanced Configuration and Power Interface 113 Format: { force | on | off | strict | noirq | rsdt | 114 copy_dsdt | nospcr } 115 force -- enable ACPI if default was off 116 on -- enable ACPI but allow fallback to DT [arm64,riscv64] 117 off -- disable ACPI if default was on 118 noirq -- do not use ACPI for IRQ routing 119 strict -- Be less tolerant of platforms that are not 120 strictly ACPI specification compliant. 121 rsdt -- prefer RSDT over (default) XSDT 122 copy_dsdt -- copy DSDT to memory 123 nocmcff -- Disable firmware first mode for corrected 124 errors. This disables parsing the HEST CMC error 125 source to check if firmware has set the FF flag. This 126 may result in duplicate corrected error reports. 127 nospcr -- disable console in ACPI SPCR table as 128 default _serial_ console on ARM64 129 spcr -- enable console in ACPI SPCR table as 130 default _serial_ console on x86 131 For ARM64, ONLY "acpi=off", "acpi=on", "acpi=force" or 132 "acpi=nospcr" are available 133 For RISCV64, ONLY "acpi=off", "acpi=on" or "acpi=force" 134 are available 135 136 See also Documentation/power/runtime_pm.rst, pci=noacpi 137 138 acpi_apic_instance= [ACPI,IOAPIC,EARLY] 139 Format: <int> 140 2: use 2nd APIC table, if available 141 1,0: use 1st APIC table 142 default: 0 143 144 acpi_backlight= [HW,ACPI] 145 { vendor | video | native | none } 146 If set to vendor, prefer vendor-specific driver 147 (e.g. thinkpad_acpi, sony_acpi, etc.) instead 148 of the ACPI video.ko driver. 149 If set to video, use the ACPI video.ko driver. 150 If set to native, use the device's native backlight mode. 151 If set to none, disable the ACPI backlight interface. 152 153 acpi_force_32bit_fadt_addr [ACPI,EARLY] 154 force FADT to use 32 bit addresses rather than the 155 64 bit X_* addresses. Some firmware have broken 64 156 bit addresses for force ACPI ignore these and use 157 the older legacy 32 bit addresses. 158 159 acpica_no_return_repair [HW, ACPI] 160 Disable AML predefined validation mechanism 161 This mechanism can repair the evaluation result to make 162 the return objects more ACPI specification compliant. 163 This option is useful for developers to identify the 164 root cause of an AML interpreter issue when the issue 165 has something to do with the repair mechanism. 166 167 acpi.debug_layer= [HW,ACPI,ACPI_DEBUG] 168 acpi.debug_level= [HW,ACPI,ACPI_DEBUG] 169 Format: <int> 170 CONFIG_ACPI_DEBUG must be enabled to produce any ACPI 171 debug output. Bits in debug_layer correspond to a 172 _COMPONENT in an ACPI source file, e.g., 173 #define _COMPONENT ACPI_EVENTS 174 Bits in debug_level correspond to a level in 175 ACPI_DEBUG_PRINT statements, e.g., 176 ACPI_DEBUG_PRINT((ACPI_DB_INFO, ... 177 The debug_level mask defaults to "info". See 178 Documentation/firmware-guide/acpi/debug.rst for more information about 179 debug layers and levels. 180 181 Enable processor driver info messages: 182 acpi.debug_layer=0x20000000 183 Enable AML "Debug" output, i.e., stores to the Debug 184 object while interpreting AML: 185 acpi.debug_layer=0xffffffff acpi.debug_level=0x2 186 Enable all messages related to ACPI hardware: 187 acpi.debug_layer=0x2 acpi.debug_level=0xffffffff 188 189 Some values produce so much output that the system is 190 unusable. The "log_buf_len" parameter may be useful 191 if you need to capture more output. 192 193 acpi.poweroff_on_fatal= [ACPI] 194 {0 | 1} 195 Causes the system to poweroff when the ACPI bytecode signals 196 a fatal error. The default value of this setting is 1. 197 Overriding this value should only be done for diagnosing 198 ACPI firmware problems, as the system might behave erratically 199 after having encountered a fatal ACPI error. 200 201 acpi_enforce_resources= [ACPI] 202 { strict | lax | no } 203 Check for resource conflicts between native drivers 204 and ACPI OperationRegions (SystemIO and SystemMemory 205 only). IO ports and memory declared in ACPI might be 206 used by the ACPI subsystem in arbitrary AML code and 207 can interfere with legacy drivers. 208 strict (default): access to resources claimed by ACPI 209 is denied; legacy drivers trying to access reserved 210 resources will fail to bind to device using them. 211 lax: access to resources claimed by ACPI is allowed; 212 legacy drivers trying to access reserved resources 213 will bind successfully but a warning message is logged. 214 no: ACPI OperationRegions are not marked as reserved, 215 no further checks are performed. 216 217 acpi_force_table_verification [HW,ACPI,EARLY] 218 Enable table checksum verification during early stage. 219 By default, this is disabled due to x86 early mapping 220 size limitation. 221 222 acpi_irq_balance [HW,ACPI] 223 ACPI will balance active IRQs 224 default in APIC mode 225 226 acpi_irq_nobalance [HW,ACPI] 227 ACPI will not move active IRQs (default) 228 default in PIC mode 229 230 acpi_irq_isa= [HW,ACPI] If irq_balance, mark listed IRQs used by ISA 231 Format: <irq>,<irq>... 232 233 acpi_irq_pci= [HW,ACPI] If irq_balance, clear listed IRQs for 234 use by PCI 235 Format: <irq>,<irq>... 236 237 acpi_mask_gpe= [HW,ACPI] 238 Due to the existence of _Lxx/_Exx, some GPEs triggered 239 by unsupported hardware/firmware features can result in 240 GPE floodings that cannot be automatically disabled by 241 the GPE dispatcher. 242 This facility can be used to prevent such uncontrolled 243 GPE floodings. 244 Format: <byte> or <bitmap-list> 245 246 acpi_no_auto_serialize [HW,ACPI] 247 Disable auto-serialization of AML methods 248 AML control methods that contain the opcodes to create 249 named objects will be marked as "Serialized" by the 250 auto-serialization feature. 251 This feature is enabled by default. 252 This option allows to turn off the feature. 253 254 acpi_no_memhotplug [ACPI] Disable memory hotplug. Useful for kdump 255 kernels. 256 257 acpi_no_static_ssdt [HW,ACPI,EARLY] 258 Disable installation of static SSDTs at early boot time 259 By default, SSDTs contained in the RSDT/XSDT will be 260 installed automatically and they will appear under 261 /sys/firmware/acpi/tables. 262 This option turns off this feature. 263 Note that specifying this option does not affect 264 dynamic table installation which will install SSDT 265 tables to /sys/firmware/acpi/tables/dynamic. 266 267 acpi_no_watchdog [HW,ACPI,WDT] 268 Ignore the ACPI-based watchdog interface (WDAT) and let 269 a native driver control the watchdog device instead. 270 271 acpi_rsdp= [ACPI,EFI,KEXEC,EARLY] 272 Pass the RSDP address to the kernel, mostly used 273 on machines running EFI runtime service to boot the 274 second kernel for kdump. 275 276 acpi_os_name= [HW,ACPI] Tell ACPI BIOS the name of the OS 277 Format: To spoof as Windows 98: ="Microsoft Windows" 278 279 acpi_rev_override [ACPI] Override the _REV object to return 5 (instead 280 of 2 which is mandated by ACPI 6) as the supported ACPI 281 specification revision (when using this switch, it may 282 be necessary to carry out a cold reboot _twice_ in a 283 row to make it take effect on the platform firmware). 284 285 acpi_osi= [HW,ACPI] Modify list of supported OS interface strings 286 acpi_osi="string1" # add string1 287 acpi_osi="!string2" # remove string2 288 acpi_osi=!* # remove all strings 289 acpi_osi=! # disable all built-in OS vendor 290 strings 291 acpi_osi=!! # enable all built-in OS vendor 292 strings 293 acpi_osi= # disable all strings 294 295 'acpi_osi=!' can be used in combination with single or 296 multiple 'acpi_osi="string1"' to support specific OS 297 vendor string(s). Note that such command can only 298 affect the default state of the OS vendor strings, thus 299 it cannot affect the default state of the feature group 300 strings and the current state of the OS vendor strings, 301 specifying it multiple times through kernel command line 302 is meaningless. This command is useful when one do not 303 care about the state of the feature group strings which 304 should be controlled by the OSPM. 305 Examples: 306 1. 'acpi_osi=! acpi_osi="Windows 2000"' is equivalent 307 to 'acpi_osi="Windows 2000" acpi_osi=!', they all 308 can make '_OSI("Windows 2000")' TRUE. 309 310 'acpi_osi=' cannot be used in combination with other 311 'acpi_osi=' command lines, the _OSI method will not 312 exist in the ACPI namespace. NOTE that such command can 313 only affect the _OSI support state, thus specifying it 314 multiple times through kernel command line is also 315 meaningless. 316 Examples: 317 1. 'acpi_osi=' can make 'CondRefOf(_OSI, Local1)' 318 FALSE. 319 320 'acpi_osi=!*' can be used in combination with single or 321 multiple 'acpi_osi="string1"' to support specific 322 string(s). Note that such command can affect the 323 current state of both the OS vendor strings and the 324 feature group strings, thus specifying it multiple times 325 through kernel command line is meaningful. But it may 326 still not able to affect the final state of a string if 327 there are quirks related to this string. This command 328 is useful when one want to control the state of the 329 feature group strings to debug BIOS issues related to 330 the OSPM features. 331 Examples: 332 1. 'acpi_osi="Module Device" acpi_osi=!*' can make 333 '_OSI("Module Device")' FALSE. 334 2. 'acpi_osi=!* acpi_osi="Module Device"' can make 335 '_OSI("Module Device")' TRUE. 336 3. 'acpi_osi=! acpi_osi=!* acpi_osi="Windows 2000"' is 337 equivalent to 338 'acpi_osi=!* acpi_osi=! acpi_osi="Windows 2000"' 339 and 340 'acpi_osi=!* acpi_osi="Windows 2000" acpi_osi=!', 341 they all will make '_OSI("Windows 2000")' TRUE. 342 343 acpi_pm_good [X86] 344 Override the pmtimer bug detection: force the kernel 345 to assume that this machine's pmtimer latches its value 346 and always returns good values. 347 348 acpi_sci= [HW,ACPI,EARLY] ACPI System Control Interrupt trigger mode 349 Format: { level | edge | high | low } 350 351 acpi_skip_timer_override [HW,ACPI,EARLY] 352 Recognize and ignore IRQ0/pin2 Interrupt Override. 353 For broken nForce2 BIOS resulting in XT-PIC timer. 354 355 acpi_sleep= [HW,ACPI] Sleep options 356 Format: { s3_bios, s3_mode, s3_beep, s4_hwsig, 357 s4_nohwsig, old_ordering, nonvs, 358 sci_force_enable, nobl } 359 See Documentation/power/video.rst for information on 360 s3_bios and s3_mode. 361 s3_beep is for debugging; it makes the PC's speaker beep 362 as soon as the kernel's real-mode entry point is called. 363 s4_hwsig causes the kernel to check the ACPI hardware 364 signature during resume from hibernation, and gracefully 365 refuse to resume if it has changed. This complies with 366 the ACPI specification but not with reality, since 367 Windows does not do this and many laptops do change it 368 on docking. So the default behaviour is to allow resume 369 and simply warn when the signature changes, unless the 370 s4_hwsig option is enabled. 371 s4_nohwsig prevents ACPI hardware signature from being 372 used (or even warned about) during resume. 373 old_ordering causes the ACPI 1.0 ordering of the _PTS 374 control method, with respect to putting devices into 375 low power states, to be enforced (the ACPI 2.0 ordering 376 of _PTS is used by default). 377 nonvs prevents the kernel from saving/restoring the 378 ACPI NVS memory during suspend/hibernation and resume. 379 sci_force_enable causes the kernel to set SCI_EN directly 380 on resume from S1/S3 (which is against the ACPI spec, 381 but some broken systems don't work without it). 382 nobl causes the internal blacklist of systems known to 383 behave incorrectly in some ways with respect to system 384 suspend and resume to be ignored (use wisely). 385 386 acpi_use_timer_override [HW,ACPI,EARLY] 387 Use timer override. For some broken Nvidia NF5 boards 388 that require a timer override, but don't have HPET 389 390 add_efi_memmap [EFI,X86,EARLY] Include EFI memory map in 391 kernel's map of available physical RAM. 392 393 agp= [AGP] 394 { off | try_unsupported } 395 off: disable AGP support 396 try_unsupported: try to drive unsupported chipsets 397 (may crash computer or cause data corruption) 398 399 ALSA [HW,ALSA] 400 See Documentation/sound/alsa-configuration.rst 401 402 alignment= [KNL,ARM] 403 Allow the default userspace alignment fault handler 404 behaviour to be specified. Bit 0 enables warnings, 405 bit 1 enables fixups, and bit 2 sends a segfault. 406 407 align_va_addr= [X86-64] 408 Align virtual addresses by clearing slice [14:12] when 409 allocating a VMA at process creation time. This option 410 gives you up to 3% performance improvement on AMD F15h 411 machines (where it is enabled by default) for a 412 CPU-intensive style benchmark, and it can vary highly in 413 a microbenchmark depending on workload and compiler. 414 415 32: only for 32-bit processes 416 64: only for 64-bit processes 417 on: enable for both 32- and 64-bit processes 418 off: disable for both 32- and 64-bit processes 419 420 alloc_snapshot [FTRACE] 421 Allocate the ftrace snapshot buffer on boot up when the 422 main buffer is allocated. This is handy if debugging 423 and you need to use tracing_snapshot() on boot up, and 424 do not want to use tracing_snapshot_alloc() as it needs 425 to be done where GFP_KERNEL allocations are allowed. 426 427 allow_mismatched_32bit_el0 [ARM64,EARLY] 428 Allow execve() of 32-bit applications and setting of the 429 PER_LINUX32 personality on systems where only a strict 430 subset of the CPUs support 32-bit EL0. When this 431 parameter is present, the set of CPUs supporting 32-bit 432 EL0 is indicated by /sys/devices/system/cpu/aarch32_el0 433 and hot-unplug operations may be restricted. 434 435 See Documentation/arch/arm64/asymmetric-32bit.rst for more 436 information. 437 438 amd_iommu= [HW,X86-64] 439 Pass parameters to the AMD IOMMU driver in the system. 440 Possible values are: 441 fullflush - Deprecated, equivalent to iommu.strict=1 442 off - do not initialize any AMD IOMMU found in 443 the system 444 force_isolation - Force device isolation for all 445 devices. The IOMMU driver is not 446 allowed anymore to lift isolation 447 requirements as needed. This option 448 does not override iommu=pt 449 force_enable - Force enable the IOMMU on platforms known 450 to be buggy with IOMMU enabled. Use this 451 option with care. 452 pgtbl_v1 - Use v1 page table for DMA-API (Default). 453 pgtbl_v2 - Use v2 page table for DMA-API. 454 irtcachedis - Disable Interrupt Remapping Table (IRT) caching. 455 nohugepages - Limit page-sizes used for v1 page-tables 456 to 4 KiB. 457 v2_pgsizes_only - Limit page-sizes used for v1 page-tables 458 to 4KiB/2Mib/1GiB. 459 460 461 amd_iommu_dump= [HW,X86-64] 462 Enable AMD IOMMU driver option to dump the ACPI table 463 for AMD IOMMU. With this option enabled, AMD IOMMU 464 driver will print ACPI tables for AMD IOMMU during 465 IOMMU initialization. 466 467 amd_iommu_intr= [HW,X86-64] 468 Specifies one of the following AMD IOMMU interrupt 469 remapping modes: 470 legacy - Use legacy interrupt remapping mode. 471 vapic - Use virtual APIC mode, which allows IOMMU 472 to inject interrupts directly into guest. 473 This mode requires kvm-amd.avic=1. 474 (Default when IOMMU HW support is present.) 475 476 amd_pstate= [X86,EARLY] 477 disable 478 Do not enable amd_pstate as the default 479 scaling driver for the supported processors 480 passive 481 Use amd_pstate with passive mode as a scaling driver. 482 In this mode autonomous selection is disabled. 483 Driver requests a desired performance level and platform 484 tries to match the same performance level if it is 485 satisfied by guaranteed performance level. 486 active 487 Use amd_pstate_epp driver instance as the scaling driver, 488 driver provides a hint to the hardware if software wants 489 to bias toward performance (0x0) or energy efficiency (0xff) 490 to the CPPC firmware. then CPPC power algorithm will 491 calculate the runtime workload and adjust the realtime cores 492 frequency. 493 guided 494 Activate guided autonomous mode. Driver requests minimum and 495 maximum performance level and the platform autonomously 496 selects a performance level in this range and appropriate 497 to the current workload. 498 499 amd_prefcore= 500 [X86] 501 disable 502 Disable amd-pstate preferred core. 503 504 amd_dynamic_epp= 505 [X86] 506 disable 507 Disable amd-pstate dynamic EPP. 508 enable 509 Enable amd-pstate dynamic EPP. 510 511 amijoy.map= [HW,JOY] Amiga joystick support 512 Map of devices attached to JOY0DAT and JOY1DAT 513 Format: <a>,<b> 514 See also Documentation/input/joydev/joystick.rst 515 516 analog.map= [HW,JOY] Analog joystick and gamepad support 517 Specifies type or capabilities of an analog joystick 518 connected to one of 16 gameports 519 Format: <type1>,<type2>,..<type16> 520 521 apc= [HW,SPARC] 522 Power management functions (SPARCstation-4/5 + deriv.) 523 Format: noidle 524 Disable APC CPU standby support. SPARCstation-Fox does 525 not play well with APC CPU idle - disable it if you have 526 APC and your system crashes randomly. 527 528 apic [APIC,X86-64] Use IO-APIC. Default. 529 530 apic= [APIC,X86,EARLY] Advanced Programmable Interrupt Controller 531 Change the output verbosity while booting 532 Format: { quiet (default) | verbose | debug } 533 Change the amount of debugging information output 534 when initialising the APIC and IO-APIC components. 535 536 apic_extnmi= [APIC,X86,EARLY] External NMI delivery setting 537 Format: { bsp (default) | all | none } 538 bsp: External NMI is delivered only to CPU 0 539 all: External NMIs are broadcast to all CPUs as a 540 backup of CPU 0 541 none: External NMI is masked for all CPUs. This is 542 useful so that a dump capture kernel won't be 543 shot down by NMI 544 545 apicpmtimer Do APIC timer calibration using the pmtimer. Implies 546 apicmaintimer. Useful when your PIT timer is totally 547 broken. 548 549 autoconf= [IPV6] 550 See Documentation/networking/ipv6.rst. 551 552 apm= [APM] Advanced Power Management 553 See header of arch/x86/kernel/apm_32.c. 554 555 apparmor= [APPARMOR] Disable or enable AppArmor at boot time 556 Format: { "0" | "1" } 557 See security/apparmor/Kconfig help text 558 0 -- disable. 559 1 -- enable. 560 Default value is set via kernel config option. 561 562 arm64.no32bit_el0 [ARM64] Unconditionally disable the execution of 563 32 bit applications. 564 565 arm64.nobti [ARM64] Unconditionally disable Branch Target 566 Identification support 567 568 arm64.nogcs [ARM64] Unconditionally disable Guarded Control Stack 569 support 570 571 arm64.nohaft [ARM64] Unconditionally disable Hardware managed Access 572 Flag for Table descriptors support 573 574 arm64.nomops [ARM64] Unconditionally disable Memory Copy and Memory 575 Set instructions support 576 577 arm64.nompam [ARM64] Unconditionally disable Memory Partitioning And 578 Monitoring support 579 580 arm64.nomte [ARM64] Unconditionally disable Memory Tagging Extension 581 support 582 583 arm64.nopauth [ARM64] Unconditionally disable Pointer Authentication 584 support 585 586 arm64.nosme [ARM64] Unconditionally disable Scalable Matrix 587 Extension support 588 589 arm64.nosve [ARM64] Unconditionally disable Scalable Vector 590 Extension support 591 592 ataflop= [HW,M68k] 593 594 atkbd.extra= [HW] Enable extra LEDs and keys on IBM RapidAccess, 595 EzKey and similar keyboards 596 597 atkbd.reset= [HW] Reset keyboard during initialization 598 599 atkbd.set= [HW] Select keyboard code set 600 Format: <int> (2 = AT (default), 3 = PS/2) 601 602 atkbd.scroll= [HW] Enable scroll wheel on MS Office and similar 603 keyboards 604 605 atkbd.softraw= [HW] Choose between synthetic and real raw mode 606 Format: <bool> (0 = real, 1 = synthetic (default)) 607 608 atkbd.softrepeat= [HW] 609 Use software keyboard repeat 610 611 audit= [KNL] Enable the audit sub-system 612 Format: { "0" | "1" | "off" | "on" } 613 0 | off - kernel audit is disabled and can not be 614 enabled until the next reboot 615 unset - kernel audit is initialized but disabled and 616 will be fully enabled by the userspace auditd. 617 1 | on - kernel audit is initialized and partially 618 enabled, storing at most audit_backlog_limit 619 messages in RAM until it is fully enabled by the 620 userspace auditd. 621 Default: unset 622 623 audit_backlog_limit= [KNL] Set the audit queue size limit. 624 Format: <int> (must be >=0) 625 Default: 64 626 627 bdev_allow_write_mounted= 628 Format: <bool> 629 Control the ability to open a mounted block device 630 for writing, i.e., allow / disallow writes that bypass 631 the FS. This was implemented as a means to prevent 632 fuzzers from crashing the kernel by overwriting the 633 metadata underneath a mounted FS without its awareness. 634 This also prevents destructive formatting of mounted 635 filesystems by naive storage tooling that don't use 636 O_EXCL. Default is Y and can be changed through the 637 Kconfig option CONFIG_BLK_DEV_WRITE_MOUNTED. 638 639 bert_disable [ACPI] 640 Disable BERT OS support on buggy BIOSes. 641 642 bgrt_disable [ACPI,X86,EARLY] 643 Disable BGRT to avoid flickering OEM logo. 644 645 blkdevparts= Manual partition parsing of block device(s) for 646 embedded devices based on command line input. 647 See Documentation/block/cmdline-partition.rst 648 649 boot_delay= [KNL,EARLY] 650 Milliseconds to delay each printk during boot. 651 Only works if CONFIG_BOOT_PRINTK_DELAY is enabled, 652 and you may also have to specify "lpj=". Boot_delay 653 values larger than 10 seconds (10000) are assumed 654 erroneous and ignored. 655 Format: integer 656 657 bootconfig [KNL,EARLY] 658 Extended command line options can be added to an initrd 659 and this will cause the kernel to look for it. 660 661 See Documentation/admin-guide/bootconfig.rst 662 663 bttv.card= [HW,V4L] bttv (bt848 + bt878 based grabber cards) 664 bttv.radio= Most important insmod options are available as 665 kernel args too. 666 bttv.pll= See Documentation/admin-guide/media/bttv.rst 667 bttv.tuner= 668 669 bulk_remove=off [PPC] This parameter disables the use of the pSeries 670 firmware feature for flushing multiple hpte entries 671 at a time. 672 673 c101= [NET] Moxa C101 synchronous serial card 674 675 cachesize= [BUGS=X86-32] Override level 2 CPU cache size detection. 676 Sometimes CPU hardware bugs make them report the cache 677 size incorrectly. The kernel will attempt work arounds 678 to fix known problems, but for some CPUs it is not 679 possible to determine what the correct size should be. 680 This option provides an override for these situations. 681 682 carrier_timeout= 683 [NET] Specifies amount of time (in seconds) that 684 the kernel should wait for a network carrier. By default 685 it waits 120 seconds. 686 687 ca_keys= [KEYS] This parameter identifies a specific key(s) on 688 the system trusted keyring to be used for certificate 689 trust validation. 690 format: { id:<keyid> | builtin } 691 692 cca= [MIPS,EARLY] Override the kernel pages' cache coherency 693 algorithm. Accepted values range from 0 to 7 694 inclusive. See arch/mips/include/asm/pgtable-bits.h 695 for platform specific values (SB1, Loongson3 and 696 others). 697 698 ccw_timeout_log [S390] 699 See Documentation/arch/s390/common_io.rst for details. 700 701 cfi= [X86-64] Set Control Flow Integrity checking features 702 when CONFIG_FINEIBT is enabled. 703 Format: feature[,feature...] 704 Default: auto 705 706 auto: Use FineIBT if IBT available, otherwise kCFI. 707 Under FineIBT, enable "paranoid" mode when 708 FRED is not available. 709 off: Turn off CFI checking. 710 kcfi: Use kCFI (disable FineIBT). 711 fineibt: Use FineIBT (even if IBT not available). 712 norand: Do not re-randomize CFI hashes. 713 paranoid: Add caller hash checking under FineIBT. 714 bhi: Enable register poisoning to stop speculation 715 across FineIBT. (Disabled by default.) 716 warn: Do not enforce CFI checking: warn only. 717 debug: Report CFI initialization details. 718 719 cgroup_disable= [KNL] Disable a particular controller or optional feature 720 Format: {name of the controller(s) or feature(s) to disable} 721 The effects of cgroup_disable=foo are: 722 - foo isn't auto-mounted if you mount all cgroups in 723 a single hierarchy 724 - foo isn't visible as an individually mountable 725 subsystem 726 - if foo is an optional feature then the feature is 727 disabled and corresponding cgroup files are not 728 created 729 {Currently only "memory" controller deal with this and 730 cut the overhead, others just disable the usage. So 731 only cgroup_disable=memory is actually worthy} 732 Specifying "pressure" disables per-cgroup pressure 733 stall information accounting feature 734 735 cgroup_no_v1= [KNL] Disable cgroup controllers and named hierarchies in v1 736 Format: { { controller | "all" | "named" } 737 [,{ controller | "all" | "named" }...] } 738 Like cgroup_disable, but only applies to cgroup v1; 739 the blacklisted controllers remain available in cgroup2. 740 "all" blacklists all controllers and "named" disables 741 named mounts. Specifying both "all" and "named" disables 742 all v1 hierarchies. 743 744 cgroup_v1_proc= [KNL] Show also missing controllers in /proc/cgroups 745 Format: { "true" | "false" } 746 /proc/cgroups lists only v1 controllers by default. 747 This compatibility option enables listing also v2 748 controllers (whose v1 code is not compiled!), so that 749 semi-legacy software can check this file to decide 750 about usage of v2 (sic) controllers. 751 752 cgroup_favordynmods= [KNL] Enable or Disable favordynmods. 753 Format: { "true" | "false" } 754 Defaults to the value of CONFIG_CGROUP_FAVOR_DYNMODS. 755 756 cgroup.memory= [KNL] Pass options to the cgroup memory controller. 757 Format: <string> 758 nosocket -- Disable socket memory accounting. 759 nokmem -- Disable kernel memory accounting. 760 nobpf -- Disable BPF memory accounting. 761 762 check_pages= [MM,EARLY] Enable sanity checking of pages after 763 allocations / before freeing. This adds checks to catch 764 double-frees, use-after-frees, and other sources of 765 page corruption by inspecting page internals (flags, 766 mapcount/refcount, memcg_data, etc.). 767 Format: { "0" | "1" } 768 Default: 0 (1 if CONFIG_DEBUG_VM is set) 769 770 checkreqprot= [SELINUX] Set initial checkreqprot flag value. 771 Format: { "0" | "1" } 772 See security/selinux/Kconfig help text. 773 0 -- check protection applied by kernel (includes 774 any implied execute protection). 775 1 -- check protection requested by application. 776 Default value is set via a kernel config option. 777 Value can be changed at runtime via 778 /sys/fs/selinux/checkreqprot. 779 Setting checkreqprot to 1 is deprecated. 780 781 cio_ignore= [S390] 782 See Documentation/arch/s390/common_io.rst for details. 783 784 clk_ignore_unused 785 [CLK] 786 Prevents the clock framework from automatically gating 787 clocks that have not been explicitly enabled by a Linux 788 device driver but are enabled in hardware at reset or 789 by the bootloader/firmware. Note that this does not 790 force such clocks to be always-on nor does it reserve 791 those clocks in any way. This parameter is useful for 792 debug and development, but should not be needed on a 793 platform with proper driver support. For more 794 information, see Documentation/driver-api/clk.rst. 795 796 clock= [BUGS=X86-32, HW] gettimeofday clocksource override. 797 [Deprecated] 798 Forces specified clocksource (if available) to be used 799 when calculating gettimeofday(). If specified 800 clocksource is not available, it defaults to PIT. 801 Format: { pit | tsc | cyclone | pmtmr } 802 803 clocksource= Override the default clocksource 804 Format: <string> 805 Override the default clocksource and use the clocksource 806 with the name specified. 807 Some clocksource names to choose from, depending on 808 the platform: 809 [all] jiffies (this is the base, fallback clocksource) 810 [ACPI] acpi_pm 811 [ARM] imx_timer1,OSTS,netx_timer,mpu_timer2, 812 pxa_timer,timer3,32k_counter,timer0_1 813 [X86-32] pit,hpet,tsc; 814 scx200_hrt on Geode; cyclone on IBM x440 815 [MIPS] MIPS 816 [PARISC] cr16 817 [S390] tod 818 [SH] SuperH 819 [SPARC64] tick 820 [X86-64] hpet,tsc 821 822 clocksource.arm_arch_timer.evtstrm= 823 [ARM,ARM64,EARLY] 824 Format: <bool> 825 Enable/disable the eventstream feature of the ARM 826 architected timer so that code using WFE-based polling 827 loops can be debugged more effectively on production 828 systems. 829 830 clocksource.verify_n_cpus= [KNL] 831 Limit the number of CPUs checked for clocksources 832 marked with CLOCK_SOURCE_VERIFY_PERCPU that 833 are marked unstable due to excessive skew. 834 A negative value says to check all CPUs, while 835 zero says not to check any. Values larger than 836 nr_cpu_ids are silently truncated to nr_cpu_ids. 837 The actual CPUs are chosen randomly, with 838 no replacement if the same CPU is chosen twice. 839 840 clocksource-wdtest.holdoff= [KNL] 841 Set the time in seconds that the clocksource 842 watchdog test waits before commencing its tests. 843 Defaults to zero when built as a module and to 844 10 seconds when built into the kernel. 845 846 cma=nn[MG]@[start[MG][-end[MG]]] 847 [KNL,CMA,EARLY] 848 Sets the size of kernel global memory area for 849 contiguous memory allocations and optionally the 850 placement constraint by the physical address range of 851 memory allocations. A value of 0 disables CMA 852 altogether. For more information, see 853 kernel/dma/contiguous.c 854 855 cma_pernuma=nn[MG] 856 [KNL,CMA,EARLY] 857 Sets the size of kernel per-numa memory area for 858 contiguous memory allocations. A value of 0 disables 859 per-numa CMA altogether. And If this option is not 860 specified, the default value is 0. 861 With per-numa CMA enabled, DMA users on node nid will 862 first try to allocate buffer from the pernuma area 863 which is located in node nid, if the allocation fails, 864 they will fallback to the global default memory area. 865 866 numa_cma=<node>:nn[MG][,<node>:nn[MG]] 867 [KNL,CMA,EARLY] 868 Sets the size of kernel numa memory area for 869 contiguous memory allocations. It will reserve CMA 870 area for the specified node. 871 872 If it is setup together with upper 'cmd_pernuma=' 873 (unlikely), its size setting takes priority for the 874 specified numa nodes. 875 876 With numa CMA enabled, DMA users on node nid will 877 first try to allocate buffer from the numa area 878 which is located in node nid, if the allocation fails, 879 they will fallback to the global default memory area. 880 881 cmo_free_hint= [PPC] Format: { yes | no } 882 Specify whether pages are marked as being inactive 883 when they are freed. This is used in CMO environments 884 to determine OS memory pressure for page stealing by 885 a hypervisor. 886 Default: yes 887 888 coherent_pool=nn[KMG] [ARM,KNL,EARLY] 889 Sets the size of memory pool for coherent, atomic dma 890 allocations, by default set to 256K. 891 892 condev= [HW,S390] console device 893 conmode= 894 895 con3215_drop= [S390,EARLY] 3215 console drop mode. 896 Format: y|n|Y|N|1|0 897 When set to true, drop data on the 3215 console when 898 the console buffer is full. In this case the 899 operator using a 3270 terminal emulator (for example 900 x3270) does not have to enter the clear key for the 901 console output to advance and the kernel to continue. 902 This leads to a much faster boot time when a 3270 903 terminal emulator is active. If no 3270 terminal 904 emulator is used, this parameter has no effect. 905 906 console= [KNL] Output console device and options. 907 908 tty<n> Use the virtual console device <n>. 909 910 ttyS<n>[,options] 911 ttyUSB0[,options] 912 Use the specified serial port. The options are of 913 the form "bbbbpnf", where "bbbb" is the baud rate, 914 "p" is parity ("n", "o", or "e"), "n" is number of 915 bits, and "f" is flow control ("r" for RTS or 916 omit it). Default is "9600n8". 917 918 See Documentation/admin-guide/serial-console.rst for more 919 information. See 920 Documentation/networking/netconsole.rst for an 921 alternative. 922 923 <DEVNAME>:<n>.<n>[,options] 924 Use the specified serial port on the serial core bus. 925 The addressing uses DEVNAME of the physical serial port 926 device, followed by the serial core controller instance, 927 and the serial port instance. The options are the same 928 as documented for the ttyS addressing above. 929 930 The mapping of the serial ports to the tty instances 931 can be viewed with: 932 933 $ ls -d /sys/bus/serial-base/devices/*:*.*/tty/* 934 /sys/bus/serial-base/devices/00:04:0.0/tty/ttyS0 935 936 In the above example, the console can be addressed with 937 console=00:04:0.0. Note that a console addressed this 938 way will only get added when the related device driver 939 is ready. The use of an earlycon parameter in addition to 940 the console may be desired for console output early on. 941 942 uart[8250],io,<addr>[,options] 943 uart[8250],mmio,<addr>[,options] 944 uart[8250],mmio16,<addr>[,options] 945 uart[8250],mmio32,<addr>[,options] 946 uart[8250],0x<addr>[,options] 947 Start an early, polled-mode console on the 8250/16550 948 UART at the specified I/O port or MMIO address, 949 switching to the matching ttyS device later. 950 MMIO inter-register address stride is either 8-bit 951 (mmio), 16-bit (mmio16), or 32-bit (mmio32). 952 If none of [io|mmio|mmio16|mmio32], <addr> is assumed 953 to be equivalent to 'mmio'. 'options' are specified in 954 the same format described for ttyS above; if unspecified, 955 the h/w is not re-initialized. 956 957 hvc<n> Use the hypervisor console device <n>. This is for 958 both Xen and PowerPC hypervisors. 959 960 { null | "" } 961 Use to disable console output, i.e., to have kernel 962 console messages discarded. 963 This must be the only console= parameter used on the 964 kernel command line. 965 966 If the device connected to the port is not a TTY but a braille 967 device, prepend "brl," before the device type, for instance 968 console=brl,ttyS0 969 For now, only VisioBraille is supported. 970 971 console_msg_format= 972 [KNL] Change console messages format 973 default 974 By default we print messages on consoles in 975 "[time stamp] text\n" format (time stamp may not be 976 printed, depending on CONFIG_PRINTK_TIME or 977 `printk_time' param). 978 syslog 979 Switch to syslog format: "<%u>[time stamp] text\n" 980 IOW, each message will have a facility and loglevel 981 prefix. The format is similar to one used by syslog() 982 syscall, or to executing "dmesg -S --raw" or to reading 983 from /proc/kmsg. 984 985 consoleblank= [KNL] The console blank (screen saver) timeout in 986 seconds. A value of 0 disables the blank timer. 987 Defaults to 0. 988 989 coredump_filter= 990 [KNL] Change the default value for 991 /proc/<pid>/coredump_filter. 992 See also Documentation/filesystems/proc.rst. 993 994 coresight_cpu_debug.enable 995 [ARM,ARM64] 996 Format: <bool> 997 Enable/disable the CPU sampling based debugging. 998 0: default value, disable debugging 999 1: enable debugging at boot time 1000 1001 cpcihp_generic= [HW,PCI] Generic port I/O CompactPCI driver 1002 Format: 1003 <first_slot>,<last_slot>,<port>,<enum_bit>[,<debug>] 1004 1005 cpuidle.off=1 [CPU_IDLE] 1006 disable the cpuidle sub-system 1007 1008 cpuidle.governor= 1009 [CPU_IDLE] Name of the cpuidle governor to use. 1010 1011 cpufreq.off=1 [CPU_FREQ] 1012 disable the cpufreq sub-system 1013 1014 cpufreq.default_governor= 1015 [CPU_FREQ] Name of the default cpufreq governor or 1016 policy to use. This governor must be registered in the 1017 kernel before the cpufreq driver probes. 1018 1019 cpu_init_udelay=N 1020 [X86,EARLY] Delay for N microsec between assert and de-assert 1021 of APIC INIT to start processors. This delay occurs 1022 on every CPU online, such as boot, and resume from suspend. 1023 Default: 10000 1024 1025 cpuhp.parallel= 1026 [SMP] Enable/disable parallel bringup of secondary CPUs 1027 Format: <bool> 1028 Default is enabled if CONFIG_HOTPLUG_PARALLEL=y. Otherwise 1029 the parameter has no effect. 1030 1031 crash_kexec_post_notifiers 1032 Only jump to kdump kernel after running the panic 1033 notifiers and dumping kmsg. This option increases 1034 the risks of a kdump failure, since some panic 1035 notifiers can make the crashed kernel more unstable. 1036 In configurations where kdump may not be reliable, 1037 running the panic notifiers could allow collecting 1038 more data on dmesg, like stack traces from other CPUS 1039 or extra data dumped by panic_print. Note that some 1040 configurations enable this option unconditionally, 1041 like Hyper-V, PowerPC (fadump) and AMD SEV-SNP. 1042 1043 crashkernel=size[KMG][@offset[KMG]] 1044 [KNL,EARLY] Using kexec, Linux can switch to a 'crash kernel' 1045 upon panic. This parameter reserves the physical 1046 memory region [offset, offset + size] for that kernel 1047 image. If '@offset' is omitted, then a suitable offset 1048 is selected automatically. 1049 [KNL, X86-64, ARM64, RISCV, LoongArch] Select a region 1050 under 4G first, and fall back to reserve region above 1051 4G when '@offset' hasn't been specified. 1052 See Documentation/admin-guide/kdump/kdump.rst for further details. 1053 1054 crashkernel=range1:size1[,range2:size2,...][@offset] 1055 [KNL] Same as above, but depends on the memory 1056 in the running system. The syntax of range is 1057 start-[end] where start and end are both 1058 a memory unit (amount[KMG]). See also 1059 Documentation/admin-guide/kdump/kdump.rst for an example. 1060 1061 crashkernel=size[KMG],high 1062 [KNL, X86-64, ARM64, RISCV, LoongArch] range could be 1063 above 4G. 1064 Allow kernel to allocate physical memory region from top, 1065 so could be above 4G if system have more than 4G ram 1066 installed. Otherwise memory region will be allocated 1067 below 4G, if available. 1068 It will be ignored if crashkernel=X is specified. 1069 crashkernel=size[KMG],low 1070 [KNL, X86-64, ARM64, RISCV, LoongArch] range under 4G. 1071 When crashkernel=X,high is passed, kernel could allocate 1072 physical memory region above 4G, that cause second kernel 1073 crash on system that require some amount of low memory, 1074 e.g. swiotlb requires at least 64M+32K low memory, also 1075 enough extra low memory is needed to make sure DMA buffers 1076 for 32-bit devices won't run out. Kernel would try to allocate 1077 default size of memory below 4G automatically. The default 1078 size is platform dependent. 1079 --> x86: max(swiotlb_size_or_default() + 8MiB, 256MiB) 1080 --> arm64: 128MiB 1081 --> riscv: 128MiB 1082 --> loongarch: 128MiB 1083 This one lets the user specify own low range under 4G 1084 for second kernel instead. 1085 0: to disable low allocation. 1086 It will be ignored when crashkernel=X,high is not used 1087 or memory reserved is below 4G. 1088 crashkernel=size[KMG],cma 1089 [KNL, X86, ARM64, RISCV, PPC] Reserve additional crash 1090 kernel memory from CMA. This reservation is usable by 1091 the first system's userspace memory and kernel movable 1092 allocations (memory balloon, zswap). Pages allocated 1093 from this memory range will not be included in the vmcore 1094 so this should not be used if dumping of userspace memory 1095 is intended and it has to be expected that some movable 1096 kernel pages may be missing from the dump. 1097 1098 A standard crashkernel reservation, as described above, 1099 is still needed to hold the crash kernel and initrd. 1100 1101 This option increases the risk of a kdump failure: DMA 1102 transfers configured by the first kernel may end up 1103 corrupting the second kernel's memory. 1104 1105 This reservation method is intended for systems that 1106 can't afford to sacrifice enough memory for standard 1107 crashkernel reservation and where less reliable and 1108 possibly incomplete kdump is preferable to no kdump at 1109 all. 1110 1111 cryptomgr.notests 1112 [KNL] Disable crypto self-tests 1113 1114 cs89x0_dma= [HW,NET] 1115 Format: <dma> 1116 1117 cs89x0_media= [HW,NET] 1118 Format: { rj45 | aui | bnc } 1119 1120 csdlock_debug= [KNL] Enable or disable debug add-ons of cross-CPU 1121 function call handling. When switched on, 1122 additional debug data is printed to the console 1123 in case a hanging CPU is detected, and that 1124 CPU is pinged again in order to try to resolve 1125 the hang situation. The default value of this 1126 option depends on the CSD_LOCK_WAIT_DEBUG_DEFAULT 1127 Kconfig option. 1128 1129 dasd= [HW,NET] 1130 See header of drivers/s390/block/dasd_devmap.c. 1131 1132 db9.dev[2|3]= [HW,JOY] Multisystem joystick support via parallel port 1133 (one device per port) 1134 Format: <port#>,<type> 1135 See also Documentation/input/devices/joystick-parport.rst 1136 1137 debug [KNL,EARLY] Enable kernel debugging (events log level). 1138 1139 debug_boot_weak_hash 1140 [KNL,EARLY] Enable printing [hashed] pointers early in the 1141 boot sequence. If enabled, we use a weak hash instead 1142 of siphash to hash pointers. Use this option if you are 1143 seeing instances of '(___ptrval___)') and need to see a 1144 value (hashed pointer) instead. Cryptographically 1145 insecure, please do not use on production kernels. 1146 1147 debug_locks_verbose= 1148 [KNL] verbose locking self-tests 1149 Format: <int> 1150 Print debugging info while doing the locking API 1151 self-tests. 1152 Bitmask for the various LOCKTYPE_ tests. Defaults to 0 1153 (no extra messages), setting it to -1 (all bits set) 1154 will print _a_lot_ more information - normally only 1155 useful to lockdep developers. 1156 1157 debug_objects [KNL,EARLY] Enable object debugging 1158 1159 debug_guardpage_minorder= 1160 [KNL,EARLY] When CONFIG_DEBUG_PAGEALLOC is set, this 1161 parameter allows control of the order of pages that will 1162 be intentionally kept free (and hence protected) by the 1163 buddy allocator. Bigger value increase the probability 1164 of catching random memory corruption, but reduce the 1165 amount of memory for normal system use. The maximum 1166 possible value is MAX_PAGE_ORDER/2. Setting this 1167 parameter to 1 or 2 should be enough to identify most 1168 random memory corruption problems caused by bugs in 1169 kernel or driver code when a CPU writes to (or reads 1170 from) a random memory location. Note that there exists 1171 a class of memory corruptions problems caused by buggy 1172 H/W or F/W or by drivers badly programming DMA 1173 (basically when memory is written at bus level and the 1174 CPU MMU is bypassed) which are not detectable by 1175 CONFIG_DEBUG_PAGEALLOC, hence this option will not 1176 help tracking down these problems. 1177 1178 debug_pagealloc= 1179 [KNL,EARLY] When CONFIG_DEBUG_PAGEALLOC is set, this parameter 1180 enables the feature at boot time. By default, it is 1181 disabled and the system will work mostly the same as a 1182 kernel built without CONFIG_DEBUG_PAGEALLOC. 1183 Note: to get most of debug_pagealloc error reports, it's 1184 useful to also enable the page_owner functionality. 1185 on: enable the feature 1186 1187 debugfs= [KNL,EARLY] This parameter enables what is exposed to 1188 userspace and debugfs internal clients. 1189 Format: { on, off } 1190 on: All functions are enabled. 1191 off: Filesystem is not registered and clients 1192 get a -EPERM as result when trying to register files 1193 or directories within debugfs. 1194 This is equivalent of the runtime functionality if 1195 debugfs was not enabled in the kernel at all. 1196 Default value is set in build-time with a kernel configuration. 1197 1198 debugpat [X86] Enable PAT debugging 1199 1200 default_hugepagesz= 1201 [HW] The size of the default HugeTLB page. This is 1202 the size represented by the legacy /proc/ hugepages 1203 APIs. In addition, this is the default hugetlb size 1204 used for shmget(), mmap() and mounting hugetlbfs 1205 filesystems. If not specified, defaults to the 1206 architecture's default huge page size. Huge page 1207 sizes are architecture dependent. See also 1208 Documentation/admin-guide/mm/hugetlbpage.rst. 1209 Format: size[KMG] 1210 1211 deferred_probe_timeout= 1212 [KNL] Debugging option to set a timeout in seconds for 1213 deferred probe to give up waiting on dependencies to 1214 probe. Only specific dependencies (subsystems or 1215 drivers) that have opted in will be ignored. A timeout 1216 of 0 will timeout at the end of initcalls; a negative 1217 value is treated as an infinite timeout value. If the 1218 timeout hasn't expired, it'll be restarted by each 1219 successful driver registration. This option will also 1220 dump out devices still on the deferred probe list after 1221 retrying. 1222 1223 delayacct [KNL] Enable per-task delay accounting 1224 1225 dell_smm_hwmon.ignore_dmi= 1226 [HW] Continue probing hardware even if DMI data 1227 indicates that the driver is running on unsupported 1228 hardware. 1229 1230 dell_smm_hwmon.force= 1231 [HW] Activate driver even if SMM BIOS signature does 1232 not match list of supported models and enable otherwise 1233 blacklisted features. 1234 1235 dell_smm_hwmon.power_status= 1236 [HW] Report power status in /proc/i8k 1237 (disabled by default). 1238 1239 dell_smm_hwmon.restricted= 1240 [HW] Allow controlling fans only if SYS_ADMIN 1241 capability is set. 1242 1243 dell_smm_hwmon.fan_mult= 1244 [HW] Factor to multiply fan speed with. 1245 1246 dell_smm_hwmon.fan_max= 1247 [HW] Maximum configurable fan speed. 1248 1249 dfltcc= [HW,S390] 1250 Format: { on | off | def_only | inf_only | always } 1251 on: s390 zlib hardware support for compression on 1252 level 1 and decompression (default) 1253 off: No s390 zlib hardware support 1254 def_only: s390 zlib hardware support for deflate 1255 only (compression on level 1) 1256 inf_only: s390 zlib hardware support for inflate 1257 only (decompression) 1258 always: Same as 'on' but ignores the selected compression 1259 level always using hardware support (used for debugging) 1260 1261 dhash_entries= [KNL] 1262 Set number of hash buckets for dentry cache. 1263 1264 disable_1tb_segments [PPC,EARLY] 1265 Disables the use of 1TB hash page table segments. This 1266 causes the kernel to fall back to 256MB segments which 1267 can be useful when debugging issues that require an SLB 1268 miss to occur. 1269 1270 disable= [IPV6] 1271 See Documentation/networking/ipv6.rst. 1272 1273 disable_radix [PPC,EARLY] 1274 Disable RADIX MMU mode on POWER9 1275 1276 disable_tlbie [PPC] 1277 Disable TLBIE instruction. Currently does not work 1278 with KVM, with HASH MMU, or with coherent accelerators. 1279 1280 disable_ddw [PPC/PSERIES,EARLY] 1281 Disable Dynamic DMA Window support. Use this 1282 to workaround buggy firmware. 1283 1284 disable_ipv6= [IPV6] 1285 See Documentation/networking/ipv6.rst. 1286 1287 disable_mtrr_cleanup [X86,EARLY] 1288 The kernel tries to adjust MTRR layout from continuous 1289 to discrete, to make X server driver able to add WB 1290 entry later. This parameter disables that. 1291 1292 disable_mtrr_trim [X86, Intel and AMD only,EARLY] 1293 By default the kernel will trim any uncacheable 1294 memory out of your available memory pool based on 1295 MTRR settings. This parameter disables that behavior, 1296 possibly causing your machine to run very slowly. 1297 1298 disable_timer_pin_1 [X86,EARLY] 1299 Disable PIN 1 of APIC timer 1300 Can be useful to work around chipset bugs. 1301 1302 dis_ucode_ldr [X86] Disable the microcode loader. 1303 1304 dma_debug=off If the kernel is compiled with DMA_API_DEBUG support, 1305 this option disables the debugging code at boot. 1306 1307 dma_debug_entries=<number> 1308 This option allows to tune the number of preallocated 1309 entries for DMA-API debugging code. One entry is 1310 required per DMA-API allocation. Use this if the 1311 DMA-API debugging code disables itself because the 1312 architectural default is too low. 1313 1314 dma_debug_driver=<driver_name> 1315 With this option the DMA-API debugging driver 1316 filter feature can be enabled at boot time. Just 1317 pass the driver to filter for as the parameter. 1318 The filter can be disabled or changed to another 1319 driver later using sysfs. 1320 1321 reg_file_data_sampling= 1322 [X86] Controls mitigation for Register File Data 1323 Sampling (RFDS) vulnerability. RFDS is a CPU 1324 vulnerability which may allow userspace to infer 1325 kernel data values previously stored in floating point 1326 registers, vector registers, or integer registers. 1327 RFDS only affects Intel Atom processors. 1328 1329 on: Turns ON the mitigation. 1330 off: Turns OFF the mitigation. 1331 1332 This parameter overrides the compile time default set 1333 by CONFIG_MITIGATION_RFDS. Mitigation cannot be 1334 disabled when other VERW based mitigations (like MDS) 1335 are enabled. In order to disable RFDS mitigation all 1336 VERW based mitigations need to be disabled. 1337 1338 For details see: 1339 Documentation/admin-guide/hw-vuln/reg-file-data-sampling.rst 1340 1341 dm_verity.keyring_unsealed= 1342 [KNL] When set to 1, leave the dm-verity keyring 1343 unsealed after initialization so userspace can 1344 provision keys. Once the keyring is restricted 1345 it becomes active and is searched during signature 1346 verification. 1347 1348 driver_async_probe= [KNL] 1349 List of driver names to be probed asynchronously. * 1350 matches with all driver names. If * is specified, the 1351 rest of the listed driver names are those that will NOT 1352 match the *. 1353 Format: <driver_name1>,<driver_name2>... 1354 1355 drm.edid_firmware=[<connector>:]<file>[,[<connector>:]<file>] 1356 Broken monitors, graphic adapters, KVMs and EDIDless 1357 panels may send no or incorrect EDID data sets. 1358 This parameter allows to specify an EDID data sets 1359 in the /lib/firmware directory that are used instead. 1360 An EDID data set will only be used for a particular 1361 connector, if its name and a colon are prepended to 1362 the EDID name. Each connector may use a unique EDID 1363 data set by separating the files with a comma. An EDID 1364 data set with no connector name will be used for 1365 any connectors not explicitly specified. 1366 1367 dscc4.setup= [NET] 1368 1369 dt_cpu_ftrs= [PPC,EARLY] 1370 Format: {"off" | "known"} 1371 Control how the dt_cpu_ftrs device-tree binding is 1372 used for CPU feature discovery and setup (if it 1373 exists). 1374 off: Do not use it, fall back to legacy cpu table. 1375 known: Do not pass through unknown features to guests 1376 or userspace, only those that the kernel is aware of. 1377 1378 dump_apple_properties [X86] 1379 Dump name and content of EFI device properties on 1380 x86 Macs. Useful for driver authors to determine 1381 what data is available or for reverse-engineering. 1382 1383 dyndbg[="val"] [KNL,DYNAMIC_DEBUG] 1384 <module>.dyndbg[="val"] 1385 Enable debug messages at boot time. See 1386 Documentation/admin-guide/dynamic-debug-howto.rst 1387 for details. 1388 1389 early_ioremap_debug [KNL,EARLY] 1390 Enable debug messages in early_ioremap support. This 1391 is useful for tracking down temporary early mappings 1392 which are not unmapped. 1393 1394 earlycon= [KNL,EARLY] Output early console device and options. 1395 1396 When used with no options, the early console is 1397 determined by stdout-path property in device tree's 1398 chosen node or the ACPI SPCR table if supported by 1399 the platform. 1400 1401 cdns,<addr>[,options] 1402 Start an early, polled-mode console on a Cadence 1403 (xuartps) serial port at the specified address. Only 1404 supported option is baud rate. If baud rate is not 1405 specified, the serial port must already be setup and 1406 configured. 1407 1408 uart[8250],io,<addr>[,options[,uartclk]] 1409 uart[8250],mmio,<addr>[,options[,uartclk]] 1410 uart[8250],mmio32,<addr>[,options[,uartclk]] 1411 uart[8250],mmio32be,<addr>[,options[,uartclk]] 1412 uart[8250],0x<addr>[,options] 1413 Start an early, polled-mode console on the 8250/16550 1414 UART at the specified I/O port or MMIO address. 1415 MMIO inter-register address stride is either 8-bit 1416 (mmio) or 32-bit (mmio32 or mmio32be). 1417 If none of [io|mmio|mmio32|mmio32be], <addr> is assumed 1418 to be equivalent to 'mmio'. 'options' are specified 1419 in the same format described for "console=ttyS<n>"; if 1420 unspecified, the h/w is not initialized. 'uartclk' is 1421 the uart clock frequency; if unspecified, it is set 1422 to 'BASE_BAUD' * 16. 1423 1424 pl011,<addr> 1425 pl011,mmio32,<addr> 1426 Start an early, polled-mode console on a pl011 serial 1427 port at the specified address. The pl011 serial port 1428 must already be setup and configured. Options are not 1429 yet supported. If 'mmio32' is specified, then only 1430 the driver will use only 32-bit accessors to read/write 1431 the device registers. 1432 1433 liteuart,<addr> 1434 Start an early console on a litex serial port at the 1435 specified address. The serial port must already be 1436 setup and configured. Options are not yet supported. 1437 1438 meson,<addr> 1439 Start an early, polled-mode console on a meson serial 1440 port at the specified address. The serial port must 1441 already be setup and configured. Options are not yet 1442 supported. 1443 1444 msm_serial,<addr> 1445 Start an early, polled-mode console on an msm serial 1446 port at the specified address. The serial port 1447 must already be setup and configured. Options are not 1448 yet supported. 1449 1450 msm_serial_dm,<addr> 1451 Start an early, polled-mode console on an msm serial 1452 dm port at the specified address. The serial port 1453 must already be setup and configured. Options are not 1454 yet supported. 1455 1456 owl,<addr> 1457 Start an early, polled-mode console on a serial port 1458 of an Actions Semi SoC, such as S500 or S900, at the 1459 specified address. The serial port must already be 1460 setup and configured. Options are not yet supported. 1461 1462 rda,<addr> 1463 Start an early, polled-mode console on a serial port 1464 of an RDA Micro SoC, such as RDA8810PL, at the 1465 specified address. The serial port must already be 1466 setup and configured. Options are not yet supported. 1467 1468 sbi 1469 Use RISC-V SBI (Supervisor Binary Interface) for early 1470 console. 1471 1472 smh Use ARM semihosting calls for early console. 1473 1474 s3c2410,<addr> 1475 s3c2412,<addr> 1476 s3c2440,<addr> 1477 s3c6400,<addr> 1478 s5pv210,<addr> 1479 exynos4210,<addr> 1480 Use early console provided by serial driver available 1481 on Samsung SoCs, requires selecting proper type and 1482 a correct base address of the selected UART port. The 1483 serial port must already be setup and configured. 1484 Options are not yet supported. 1485 1486 lantiq,<addr> 1487 Start an early, polled-mode console on a lantiq serial 1488 (lqasc) port at the specified address. The serial port 1489 must already be setup and configured. Options are not 1490 yet supported. 1491 1492 lpuart,<addr> 1493 lpuart32,<addr> 1494 Use early console provided by Freescale LP UART driver 1495 found on Freescale Vybrid and QorIQ LS1021A processors. 1496 A valid base address must be provided, and the serial 1497 port must already be setup and configured. 1498 1499 ec_imx21,<addr> 1500 ec_imx6q,<addr> 1501 Start an early, polled-mode, output-only console on the 1502 Freescale i.MX UART at the specified address. The UART 1503 must already be setup and configured. 1504 1505 ar3700_uart,<addr> 1506 Start an early, polled-mode console on the 1507 Armada 3700 serial port at the specified 1508 address. The serial port must already be setup 1509 and configured. Options are not yet supported. 1510 1511 qcom_geni,<addr> 1512 Start an early, polled-mode console on a Qualcomm 1513 Generic Interface (GENI) based serial port at the 1514 specified address. The serial port must already be 1515 setup and configured. Options are not yet supported. 1516 1517 efifb,[options] 1518 Start an early, unaccelerated console on the EFI 1519 memory mapped framebuffer (if available). On cache 1520 coherent non-x86 systems that use system memory for 1521 the framebuffer, pass the 'ram' option so that it is 1522 mapped with the correct attributes. 1523 1524 linflex,<addr> 1525 Use early console provided by Freescale LINFlexD UART 1526 serial driver for NXP S32V234 SoCs. A valid base 1527 address must be provided, and the serial port must 1528 already be setup and configured. 1529 1530 earlyprintk= [X86,SH,ARM,M68k,S390,UM,EARLY] 1531 earlyprintk=vga 1532 earlyprintk=sclp 1533 earlyprintk=xen 1534 earlyprintk=serial[,ttySn[,baudrate]] 1535 earlyprintk=serial[,0x...[,baudrate]] 1536 earlyprintk=ttySn[,baudrate] 1537 earlyprintk=dbgp[debugController#] 1538 earlyprintk=mmio32,membase[,{nocfg|baudrate}] 1539 earlyprintk=pciserial[,force],bus:device.function[,{nocfg|baudrate}] 1540 earlyprintk=xdbc[xhciController#] 1541 earlyprintk=bios 1542 1543 earlyprintk is useful when the kernel crashes before 1544 the normal console is initialized. It is not enabled by 1545 default because it has some cosmetic problems. 1546 1547 Use "nocfg" to skip UART configuration, assume 1548 BIOS/firmware has configured UART correctly. 1549 1550 Append ",keep" to not disable it when the real console 1551 takes over. 1552 1553 Only one of vga, serial, or usb debug port can 1554 be used at a time. 1555 1556 Currently only ttyS0 and ttyS1 may be specified by 1557 name. Other I/O ports may be explicitly specified 1558 on some architectures (x86 and arm at least) by 1559 replacing ttySn with an I/O port address, like this: 1560 earlyprintk=serial,0x1008,115200 1561 You can find the port for a given device in 1562 /proc/tty/driver/serial: 1563 2: uart:ST16650V2 port:00001008 irq:18 ... 1564 1565 Interaction with the standard serial driver is not 1566 very good. 1567 1568 The VGA output is eventually overwritten by 1569 the real console. 1570 1571 The xen option can only be used in Xen domains. 1572 1573 The sclp output can only be used on s390. 1574 1575 The bios output can only be used on SuperH. 1576 1577 The optional "force" to "pciserial" enables use of a 1578 PCI device even when its classcode is not of the 1579 UART class. 1580 1581 edd= [EDD] 1582 Format: {"off" | "on" | "skip[mbr]"} 1583 1584 efi= [EFI,EARLY] 1585 Format: { "debug", "disable_early_pci_dma", 1586 "nochunk", "noruntime", "nosoftreserve", 1587 "novamap", "no_disable_early_pci_dma" } 1588 debug: enable misc debug output. 1589 disable_early_pci_dma: disable the busmaster bit on all 1590 PCI bridges while in the EFI boot stub. 1591 nochunk: disable reading files in "chunks" in the EFI 1592 boot stub, as chunking can cause problems with some 1593 firmware implementations. 1594 noruntime : disable EFI runtime services support 1595 nosoftreserve: The EFI_MEMORY_SP (Specific Purpose) 1596 attribute may cause the kernel to reserve the 1597 memory range for a memory mapping driver to 1598 claim. Specify efi=nosoftreserve to disable this 1599 reservation and treat the memory by its base type 1600 (i.e. EFI_CONVENTIONAL_MEMORY / "System RAM"). 1601 novamap: do not call SetVirtualAddressMap(). 1602 no_disable_early_pci_dma: Leave the busmaster bit set 1603 on all PCI bridges while in the EFI boot stub 1604 1605 efi_no_storage_paranoia [EFI,X86,EARLY] 1606 The kernel reserves 5KB of EFI variable storage for 1607 safety, because some UEFI implementation may fail to 1608 boot if there's insufficient space in the EFI variable 1609 storage. 1610 1611 Using this parameter, you can use the 5KB reservation 1612 in the EFI variable storage. 1613 1614 However, Use this parameter only if you are really 1615 sure that your UEFI does sane gc and fulfills the spec 1616 otherwise your board may brick. 1617 1618 efivar_ssdt= [EFI; X86] Name of an EFI variable that contains an SSDT 1619 that is to be dynamically loaded by Linux. If there are 1620 multiple variables with the same name but with different 1621 vendor GUIDs, all of them will be loaded. See 1622 Documentation/admin-guide/acpi/ssdt-overlays.rst for details. 1623 1624 1625 eisa_irq_edge= [PARISC,HW] 1626 See header of drivers/parisc/eisa.c. 1627 1628 ekgdboc= [X86,KGDB,EARLY] Allow early kernel console debugging 1629 Format: ekgdboc=kbd 1630 1631 This is designed to be used in conjunction with 1632 the boot argument: earlyprintk=vga 1633 1634 This parameter works in place of the kgdboc parameter 1635 but can only be used if the backing tty is available 1636 very early in the boot process. For early debugging 1637 via a serial port see kgdboc_earlycon instead. 1638 1639 elfcorehdr=[size[KMG]@]offset[KMG] [PPC,SH,X86,S390,EARLY] 1640 Specifies physical address of start of kernel core 1641 image elf header and optionally the size. Generally 1642 kexec loader will pass this option to capture kernel. 1643 See Documentation/admin-guide/kdump/kdump.rst for details. 1644 1645 enable_mtrr_cleanup [X86,EARLY] 1646 The kernel tries to adjust MTRR layout from continuous 1647 to discrete, to make X server driver able to add WB 1648 entry later. This parameter enables that. 1649 1650 enforcing= [SELINUX] Set initial enforcing status. 1651 Format: {"0" | "1"} 1652 See security/selinux/Kconfig help text. 1653 0 -- permissive (log only, no denials). 1654 1 -- enforcing (deny and log). 1655 Default value is 0. 1656 Value can be changed at runtime via 1657 /sys/fs/selinux/enforce. 1658 1659 erst_disable [ACPI] 1660 Disable Error Record Serialization Table (ERST) 1661 support. 1662 1663 ether= [HW,NET] Ethernet cards parameters 1664 This option is obsoleted by the "netdev=" option, which 1665 has equivalent usage. See its documentation for details. 1666 1667 evm= [EVM] 1668 Format: { "fix" } 1669 Permit 'security.evm' to be updated regardless of 1670 current integrity status. 1671 1672 early_page_ext [KNL,EARLY] Enforces page_ext initialization to earlier 1673 stages so cover more early boot allocations. 1674 Please note that as side effect some optimizations 1675 might be disabled to achieve that (e.g. parallelized 1676 memory initialization is disabled) so the boot process 1677 might take longer, especially on systems with a lot of 1678 memory. Available with CONFIG_PAGE_EXTENSION=y. 1679 1680 failslab= 1681 fail_usercopy= 1682 fail_page_alloc= 1683 fail_skb_realloc= 1684 fail_make_request=[KNL] 1685 General fault injection mechanism. 1686 Format: <interval>,<probability>,<space>,<times> 1687 See also Documentation/fault-injection/. 1688 1689 fb_tunnels= [NET] 1690 Format: { initns | none } 1691 See Documentation/admin-guide/sysctl/net.rst for 1692 fb_tunnels_only_for_init_ns 1693 1694 floppy= [HW] 1695 See Documentation/admin-guide/blockdev/floppy.rst. 1696 1697 forcepae [X86-32] 1698 Forcefully enable Physical Address Extension (PAE). 1699 Many Pentium M systems disable PAE but may have a 1700 functionally usable PAE implementation. 1701 Warning: use of this parameter will taint the kernel 1702 and may cause unknown problems. 1703 1704 fred= [X86-64] 1705 Enable/disable Flexible Return and Event Delivery. 1706 Format: { on | off } 1707 on: enable FRED when it's present, the default setting. 1708 off: disable FRED. 1709 1710 ftrace=[tracer] 1711 [FTRACE] will set and start the specified tracer 1712 as early as possible in order to facilitate early 1713 boot debugging. 1714 1715 ftrace_boot_snapshot 1716 [FTRACE] On boot up, a snapshot will be taken of the 1717 ftrace ring buffer that can be read at: 1718 /sys/kernel/tracing/snapshot. 1719 This is useful if you need tracing information from kernel 1720 boot up that is likely to be overridden by user space 1721 start up functionality. 1722 1723 Optionally, the snapshot can also be defined for a tracing 1724 instance that was created by the trace_instance= command 1725 line parameter. 1726 1727 trace_instance=foo,sched_switch ftrace_boot_snapshot=foo 1728 1729 The above will cause the "foo" tracing instance to trigger 1730 a snapshot at the end of boot up. 1731 1732 ftrace_dump_on_oops[=2(orig_cpu) | =<instance>][,<instance> | 1733 ,<instance>=2(orig_cpu)] 1734 [FTRACE] will dump the trace buffers on oops. 1735 If no parameter is passed, ftrace will dump global 1736 buffers of all CPUs, if you pass 2 or orig_cpu, it 1737 will dump only the buffer of the CPU that triggered 1738 the oops, or the specific instance will be dumped if 1739 its name is passed. Multiple instance dump is also 1740 supported, and instances are separated by commas. Each 1741 instance supports only dump on CPU that triggered the 1742 oops by passing 2 or orig_cpu to it. 1743 1744 ftrace_dump_on_oops=foo=orig_cpu 1745 1746 The above will dump only the buffer of "foo" instance 1747 on CPU that triggered the oops. 1748 1749 ftrace_dump_on_oops,foo,bar=orig_cpu 1750 1751 The above will dump global buffer on all CPUs, the 1752 buffer of "foo" instance on all CPUs and the buffer 1753 of "bar" instance on CPU that triggered the oops. 1754 1755 ftrace_filter=[function-list] 1756 [FTRACE] Limit the functions traced by the function 1757 tracer at boot up. function-list is a comma-separated 1758 list of functions. This list can be changed at run 1759 time by the set_ftrace_filter file in the debugfs 1760 tracing directory. 1761 1762 ftrace_notrace=[function-list] 1763 [FTRACE] Do not trace the functions specified in 1764 function-list. This list can be changed at run time 1765 by the set_ftrace_notrace file in the debugfs 1766 tracing directory. 1767 1768 ftrace_graph_filter=[function-list] 1769 [FTRACE] Limit the top level callers functions traced 1770 by the function graph tracer at boot up. 1771 function-list is a comma-separated list of functions 1772 that can be changed at run time by the 1773 set_graph_function file in the debugfs tracing directory. 1774 1775 ftrace_graph_notrace=[function-list] 1776 [FTRACE] Do not trace from the functions specified in 1777 function-list. This list is a comma-separated list of 1778 functions that can be changed at run time by the 1779 set_graph_notrace file in the debugfs tracing directory. 1780 1781 ftrace_graph_max_depth=<uint> 1782 [FTRACE] Used with the function graph tracer. This is 1783 the max depth it will trace into a function. This value 1784 can be changed at run time by the max_graph_depth file 1785 in the tracefs tracing directory. default: 0 (no limit) 1786 1787 fw_devlink= [KNL,EARLY] Create device links between consumer and supplier 1788 devices by scanning the firmware to infer the 1789 consumer/supplier relationships. This feature is 1790 especially useful when drivers are loaded as modules as 1791 it ensures proper ordering of tasks like device probing 1792 (suppliers first, then consumers), supplier boot state 1793 clean up (only after all consumers have probed), 1794 suspend/resume & runtime PM (consumers first, then 1795 suppliers). 1796 Format: { off | permissive | on | rpm } 1797 off -- Don't create device links from firmware info. 1798 permissive -- Create device links from firmware info 1799 but use it only for ordering boot state clean 1800 up (sync_state() calls). 1801 on -- Create device links from firmware info and use it 1802 to enforce probe and suspend/resume ordering. 1803 rpm -- Like "on", but also use to order runtime PM. 1804 1805 fw_devlink.strict=<bool> 1806 [KNL,EARLY] Treat all inferred dependencies as mandatory 1807 dependencies. This only applies for fw_devlink=on|rpm. 1808 Format: <bool> 1809 1810 fw_devlink.sync_state = 1811 [KNL,EARLY] When all devices that could probe have finished 1812 probing, this parameter controls what to do with 1813 devices that haven't yet received their sync_state() 1814 calls. 1815 Format: { strict | timeout } 1816 strict -- Default. Continue waiting on consumers to 1817 probe successfully. 1818 timeout -- Give up waiting on consumers and call 1819 sync_state() on any devices that haven't yet 1820 received their sync_state() calls after 1821 deferred_probe_timeout has expired or by 1822 late_initcall() if !CONFIG_MODULES. 1823 1824 gamecon.map[2|3]= 1825 [HW,JOY] Multisystem joystick and NES/SNES/PSX pad 1826 support via parallel port (up to 5 devices per port) 1827 Format: <port#>,<pad1>,<pad2>,<pad3>,<pad4>,<pad5> 1828 See also Documentation/input/devices/joystick-parport.rst 1829 1830 gamma= [HW,DRM] 1831 1832 gart_fix_e820= [X86-64,EARLY] disable the fix e820 for K8 GART 1833 Format: off | on 1834 default: on 1835 1836 gather_data_sampling= 1837 [X86,INTEL,EARLY] Control the Gather Data Sampling (GDS) 1838 mitigation. 1839 1840 Gather Data Sampling is a hardware vulnerability which 1841 allows unprivileged speculative access to data which was 1842 previously stored in vector registers. 1843 1844 This issue is mitigated by default in updated microcode. 1845 The mitigation may have a performance impact but can be 1846 disabled. On systems without the microcode mitigation 1847 disabling AVX serves as a mitigation. 1848 1849 force: Disable AVX to mitigate systems without 1850 microcode mitigation. No effect if the microcode 1851 mitigation is present. Known to cause crashes in 1852 userspace with buggy AVX enumeration. 1853 1854 off: Disable GDS mitigation. 1855 1856 gbpages [X86] Use GB pages for kernel direct mappings. 1857 1858 gcov_persist= [GCOV] When non-zero (default), profiling data for 1859 kernel modules is saved and remains accessible via 1860 debugfs, even when the module is unloaded/reloaded. 1861 When zero, profiling data is discarded and associated 1862 debugfs files are removed at module unload time. 1863 1864 goldfish [X86] Enable the goldfish android emulator platform. 1865 Don't use this when you are not running on the 1866 android emulator 1867 1868 gpio-mockup.gpio_mockup_ranges 1869 [HW] Sets the ranges of gpiochip of for this device. 1870 Format: <start1>,<end1>,<start2>,<end2>... 1871 gpio-mockup.gpio_mockup_named_lines 1872 [HW] Let the driver know GPIO lines should be named. 1873 1874 gpt [EFI] Forces disk with valid GPT signature but 1875 invalid Protective MBR to be treated as GPT. If the 1876 primary GPT is corrupted, it enables the backup/alternate 1877 GPT to be used instead. 1878 1879 grcan.enable0= [HW] Configuration of physical interface 0. Determines 1880 the "Enable 0" bit of the configuration register. 1881 Format: 0 | 1 1882 Default: 0 1883 grcan.enable1= [HW] Configuration of physical interface 1. Determines 1884 the "Enable 0" bit of the configuration register. 1885 Format: 0 | 1 1886 Default: 0 1887 grcan.select= [HW] Select which physical interface to use. 1888 Format: 0 | 1 1889 Default: 0 1890 grcan.txsize= [HW] Sets the size of the tx buffer. 1891 Format: <unsigned int> such that (txsize & ~0x1fffc0) == 0. 1892 Default: 1024 1893 grcan.rxsize= [HW] Sets the size of the rx buffer. 1894 Format: <unsigned int> such that (rxsize & ~0x1fffc0) == 0. 1895 Default: 1024 1896 1897 hardened_usercopy= 1898 [KNL] Under CONFIG_HARDENED_USERCOPY, whether 1899 hardening is enabled for this boot. Hardened 1900 usercopy checking is used to protect the kernel 1901 from reading or writing beyond known memory 1902 allocation boundaries as a proactive defense 1903 against bounds-checking flaws in the kernel's 1904 copy_to_user()/copy_from_user() interface. 1905 The default is determined by 1906 CONFIG_HARDENED_USERCOPY_DEFAULT_ON. 1907 on Perform hardened usercopy checks. 1908 off Disable hardened usercopy checks. 1909 1910 hardlockup_all_cpu_backtrace= 1911 [KNL] Should the hard-lockup detector generate 1912 backtraces on all cpus. 1913 Format: 0 | 1 1914 1915 hash_pointers= 1916 [KNL,EARLY] 1917 By default, when pointers are printed to the console 1918 or buffers via the %p format string, that pointer is 1919 "hashed", i.e. obscured by hashing the pointer value. 1920 This is a security feature that hides actual kernel 1921 addresses from unprivileged users, but it also makes 1922 debugging the kernel more difficult since unequal 1923 pointers can no longer be compared. The choices are: 1924 Format: { auto | always | never } 1925 Default: auto 1926 1927 auto - Hash pointers unless slab_debug is enabled. 1928 always - Always hash pointers (even if slab_debug is 1929 enabled). 1930 never - Never hash pointers. This option should only 1931 be specified when debugging the kernel. Do 1932 not use on production kernels. The boot 1933 param "no_hash_pointers" is an alias for 1934 this mode. 1935 1936 For controlling hashing dynamically at runtime, 1937 use the "kernel.kptr_restrict" sysctl instead. 1938 1939 hashdist= [KNL,NUMA] Large hashes allocated during boot 1940 are distributed across NUMA nodes. Defaults on 1941 for 64-bit NUMA, off otherwise. 1942 Format: 0 | 1 (for off | on) 1943 1944 hd= [EIDE] (E)IDE hard drive subsystem geometry 1945 Format: <cyl>,<head>,<sect> 1946 1947 hest_disable [ACPI] 1948 Disable Hardware Error Source Table (HEST) support; 1949 corresponding firmware-first mode error processing 1950 logic will be disabled. 1951 1952 hibernate= [HIBERNATION] 1953 noresume Don't check if there's a hibernation image 1954 present during boot. 1955 nocompress Don't compress/decompress hibernation images. 1956 no Disable hibernation and resume. 1957 protect_image Turn on image protection during restoration 1958 (that will set all pages holding image data 1959 during restoration read-only). 1960 1961 hibernate.compressor= [HIBERNATION] Compression algorithm to be 1962 used with hibernation. 1963 Format: { lzo | lz4 } 1964 Default: lzo 1965 1966 lzo: Select LZO compression algorithm to 1967 compress/decompress hibernation image. 1968 1969 lz4: Select LZ4 compression algorithm to 1970 compress/decompress hibernation image. 1971 1972 hibernate.pm_test_delay= 1973 [HIBERNATION] 1974 Sets the number of seconds to remain in a hibernation test 1975 mode before resuming the system (see 1976 /sys/power/pm_test). Only available when CONFIG_PM_DEBUG 1977 is set. Default value is 5. 1978 1979 hibernate_compression_threads= 1980 [HIBERNATION] 1981 Set the number of threads used for compressing or decompressing 1982 hibernation images. 1983 1984 Format: <integer> 1985 Default: 3 1986 Minimum: 1 1987 Example: hibernate_compression_threads=4 1988 1989 highmem=nn[KMG] [KNL,BOOT,EARLY] forces the highmem zone to have an exact 1990 size of <nn>. This works even on boxes that have no 1991 highmem otherwise. This also works to reduce highmem 1992 size on bigger boxes. 1993 1994 highres= [KNL] Enable/disable high resolution timer mode. 1995 Valid parameters: "on", "off" 1996 Default: "on" 1997 1998 hlt [BUGS=ARM,SH] 1999 2000 hostname= [KNL,EARLY] Set the hostname (aka UTS nodename). 2001 Format: <string> 2002 This allows setting the system's hostname during early 2003 startup. This sets the name returned by gethostname. 2004 Using this parameter to set the hostname makes it 2005 possible to ensure the hostname is correctly set before 2006 any userspace processes run, avoiding the possibility 2007 that a process may call gethostname before the hostname 2008 has been explicitly set, resulting in the calling 2009 process getting an incorrect result. The string must 2010 not exceed the maximum allowed hostname length (usually 2011 64 characters) and will be truncated otherwise. 2012 2013 hpet= [X86-32,HPET] option to control HPET usage 2014 Format: { enable (default) | disable | force | 2015 verbose } 2016 disable: disable HPET and use PIT instead 2017 force: allow force enabled of undocumented chips (ICH4, 2018 VIA, nVidia) 2019 verbose: show contents of HPET registers during setup 2020 2021 hpet_mmap= [X86, HPET_MMAP] Allow userspace to mmap HPET 2022 registers. Default set by CONFIG_HPET_MMAP_DEFAULT. 2023 2024 hugepages= [HW,EARLY] Number of HugeTLB pages to allocate at boot. 2025 If this follows hugepagesz (below), it specifies 2026 the number of pages of hugepagesz to be allocated. 2027 If this is the first HugeTLB parameter on the command 2028 line, it specifies the number of pages to allocate for 2029 the default huge page size. If using node format, the 2030 number of pages to allocate per-node can be specified. 2031 See also Documentation/admin-guide/mm/hugetlbpage.rst. 2032 Format: <integer> or (node format) 2033 <node>:<integer>[,<node>:<integer>] 2034 2035 hugepagesz= 2036 [HW,EARLY] The size of the HugeTLB pages. This is 2037 used in conjunction with hugepages (above) to 2038 allocate huge pages of a specific size at boot. The 2039 pair hugepagesz=X hugepages=Y can be specified once 2040 for each supported huge page size. Huge page sizes 2041 are architecture dependent. See also 2042 Documentation/admin-guide/mm/hugetlbpage.rst. 2043 Format: size[KMG] 2044 2045 hugepage_alloc_threads= 2046 [HW] The number of threads that should be used to 2047 allocate hugepages during boot. This option can be 2048 used to improve system bootup time when allocating 2049 a large amount of huge pages. 2050 The default value is 25% of the available hardware threads. 2051 2052 Note that this parameter only applies to non-gigantic huge pages. 2053 2054 hugetlb_cma= [HW,CMA,EARLY] The size of a CMA area used for allocation 2055 of gigantic hugepages. Or using node format, the size 2056 of a CMA area per node can be specified. 2057 The size can be an absolute value (e.g., 2G) or a 2058 percentage of the total memory or node memory (e.g., 20%). 2059 Percentage-derived sizes are rounded down to a multiple of 2060 the architecture's gigantic hugepage size and may become 2061 zero. 2062 Format: nn[KMGTPE] or nn% or (node format) 2063 <node>:nn[KMGTPE][,<node>:nn[KMGTPE]] or 2064 <node>:nn%[,<node>:nn%] 2065 2066 The size must be a multiple of the gigantic page size. 2067 When using node format, this applies to each per-node size. 2068 Missaligned values are dropped with a warning. 2069 2070 Reserve a CMA area of given size and allocate gigantic 2071 hugepages using the CMA allocator. If enabled, the 2072 boot-time allocation of gigantic hugepages is skipped. 2073 2074 hugetlb_cma_only= 2075 [HW,CMA,EARLY] When allocating new HugeTLB pages, only 2076 try to allocate from the CMA areas. 2077 2078 This option does nothing if hugetlb_cma= is not also 2079 specified. 2080 2081 hugetlb_free_vmemmap= 2082 [KNL] Requires CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP 2083 enabled. 2084 Control if HugeTLB Vmemmap Optimization (HVO) is enabled. 2085 Allows heavy hugetlb users to free up some more 2086 memory (7 * PAGE_SIZE for each 2MB hugetlb page). 2087 Format: { on | off (default) } 2088 2089 on: enable HVO 2090 off: disable HVO 2091 2092 Built with CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP_DEFAULT_ON=y, 2093 the default is on. 2094 2095 Note that the vmemmap pages may be allocated from the added 2096 memory block itself when memory_hotplug.memmap_on_memory is 2097 enabled, those vmemmap pages cannot be optimized even if this 2098 feature is enabled. Other vmemmap pages not allocated from 2099 the added memory block itself do not be affected. 2100 2101 hung_task_panic= 2102 [KNL] Number of hung tasks to trigger kernel panic. 2103 Format: <int> 2104 2105 When set to a non-zero value, a kernel panic will be triggered if 2106 the number of detected hung tasks reaches this value. 2107 2108 0: don't panic 2109 1: panic immediately on first hung task 2110 N: panic after N hung tasks are detected in a single scan 2111 2112 The default value is controlled by the 2113 CONFIG_BOOTPARAM_HUNG_TASK_PANIC build-time option. The value 2114 selected by this boot parameter can be changed later by the 2115 kernel.hung_task_panic sysctl. 2116 2117 hvc_iucv= [S390] Number of z/VM IUCV hypervisor console (HVC) 2118 terminal devices. Valid values: 0..8 2119 hvc_iucv_allow= [S390] Comma-separated list of z/VM user IDs. 2120 If specified, z/VM IUCV HVC accepts connections 2121 from listed z/VM user IDs only. 2122 2123 hv_nopvspin [X86,HYPER_V,EARLY] 2124 Disables the paravirt spinlock optimizations 2125 which allow the hypervisor to 'idle' the guest 2126 on lock contention. 2127 2128 hw_protection= [HW] 2129 Format: reboot | shutdown 2130 2131 Hardware protection action taken on critical events like 2132 overtemperature or imminent voltage loss. 2133 2134 i2c_bus= [HW] Override the default board specific I2C bus speed 2135 or register an additional I2C bus that is not 2136 registered from board initialization code. 2137 Format: 2138 <bus_id>,<clkrate> 2139 2140 i2c_touchscreen_props= [HW,ACPI,X86] 2141 Set device-properties for ACPI-enumerated I2C-attached 2142 touchscreen, to e.g. fix coordinates of upside-down 2143 mounted touchscreens. If you need this option please 2144 submit a drivers/platform/x86/touchscreen_dmi.c patch 2145 adding a DMI quirk for this. 2146 2147 Format: 2148 <ACPI_HW_ID>:<prop_name>=<val>[:prop_name=val][:...] 2149 Where <val> is one of: 2150 Omit "=<val>" entirely Set a boolean device-property 2151 Unsigned number Set a u32 device-property 2152 Anything else Set a string device-property 2153 2154 Examples (split over multiple lines): 2155 i2c_touchscreen_props=GDIX1001:touchscreen-inverted-x: 2156 touchscreen-inverted-y 2157 2158 i2c_touchscreen_props=MSSL1680:touchscreen-size-x=1920: 2159 touchscreen-size-y=1080:touchscreen-inverted-y: 2160 firmware-name=gsl1680-vendor-model.fw:silead,home-button 2161 2162 i8042.debug [HW] Toggle i8042 debug mode 2163 i8042.unmask_kbd_data 2164 [HW] Enable printing of interrupt data from the KBD port 2165 (disabled by default, and as a pre-condition 2166 requires that i8042.debug=1 be enabled) 2167 i8042.direct [HW] Put keyboard port into non-translated mode 2168 i8042.dumbkbd [HW] Pretend that controller can only read data from 2169 keyboard and cannot control its state 2170 (Don't attempt to blink the leds) 2171 i8042.noaux [HW] Don't check for auxiliary (== mouse) port 2172 i8042.nokbd [HW] Don't check/create keyboard port 2173 i8042.noloop [HW] Disable the AUX Loopback command while probing 2174 for the AUX port 2175 i8042.nomux [HW] Don't check presence of an active multiplexing 2176 controller 2177 i8042.nopnp [HW] Don't use ACPIPnP / PnPBIOS to discover KBD/AUX 2178 controllers 2179 i8042.notimeout [HW] Ignore timeout condition signalled by controller 2180 i8042.reset [HW] Reset the controller during init, cleanup and 2181 suspend-to-ram transitions, only during s2r 2182 transitions, or never reset 2183 Format: { 1 | Y | y | 0 | N | n } 2184 1, Y, y: always reset controller 2185 0, N, n: don't ever reset controller 2186 Default: only on s2r transitions on x86; most other 2187 architectures force reset to be always executed 2188 i8042.unlock [HW] Unlock (ignore) the keylock 2189 i8042.kbdreset [HW] Reset device connected to KBD port 2190 i8042.probe_defer 2191 [HW] Allow deferred probing upon i8042 probe errors 2192 2193 i810= [HW,DRM] 2194 2195 i915.invert_brightness= 2196 [DRM] Invert the sense of the variable that is used to 2197 set the brightness of the panel backlight. Normally a 2198 brightness value of 0 indicates backlight switched off, 2199 and the maximum of the brightness value sets the backlight 2200 to maximum brightness. If this parameter is set to 0 2201 (default) and the machine requires it, or this parameter 2202 is set to 1, a brightness value of 0 sets the backlight 2203 to maximum brightness, and the maximum of the brightness 2204 value switches the backlight off. 2205 -1 -- never invert brightness 2206 0 -- machine default 2207 1 -- force brightness inversion 2208 2209 ia32_emulation= [X86-64] 2210 Format: <bool> 2211 When true, allows loading 32-bit programs and executing 32-bit 2212 syscalls, essentially overriding IA32_EMULATION_DEFAULT_DISABLED at 2213 boot time. When false, unconditionally disables IA32 emulation. 2214 2215 2216 idle= [X86,EARLY] 2217 Format: idle=poll, idle=halt, idle=nomwait 2218 2219 idle=poll: Don't do power saving in the idle loop 2220 using HLT, but poll for rescheduling event. This will 2221 make the CPUs eat a lot more power, but may be useful 2222 to get slightly better performance in multiprocessor 2223 benchmarks. It also makes some profiling using 2224 performance counters more accurate. Please note that 2225 on systems with MONITOR/MWAIT support (like Intel 2226 EM64T CPUs) this option has no performance advantage 2227 over the normal idle loop. It may also interact badly 2228 with hyperthreading. 2229 2230 idle=halt: Halt is forced to be used for CPU idle. 2231 In such case C2/C3 won't be used again. 2232 2233 idle=nomwait: Disable mwait for CPU C-states 2234 2235 [ARM64,EARLY] 2236 Format: idle=wfi, idle=yield, idle=nop 2237 2238 idle=wfi: Use the WFI (Wait For Interrupt) hint 2239 instruction in the idle loop. This is the default and 2240 allows the CPU to enter a low-power state until an 2241 interrupt arrives. 2242 2243 idle=yield: Use the YIELD hint instruction instead of 2244 WFI. CPUs supporting simultaneous multi-threading (SMT), 2245 can continue executing another thread when the current 2246 thread reaches the idle loop. This will make the CPUs 2247 eat more power, but may be useful to get slightly better 2248 performance in some applications, since the CPUs will 2249 not enter a low-power state. 2250 2251 idle=nop: Do not execute any idle instruction in the 2252 idle loop. This is useful on platforms where WFI 2253 misbehaves, leading to system instability or loss of CPU 2254 state. This will make the CPUs eat more power, but may 2255 give slightly better performance in some applications, 2256 since the CPUs will not enter a low-power state. 2257 2258 idxd.sva= [HW] 2259 Format: <bool> 2260 Allow force disabling of Shared Virtual Memory (SVA) 2261 support for the idxd driver. By default it is set to 2262 true (1). 2263 2264 idxd.tc_override= [HW] 2265 Format: <bool> 2266 Allow override of default traffic class configuration 2267 for the device. By default it is set to false (0). 2268 2269 ieee754= [MIPS] Select IEEE Std 754 conformance mode 2270 Format: { strict | legacy | 2008 | relaxed | emulated } 2271 Default: strict 2272 2273 Choose which programs will be accepted for execution 2274 based on the IEEE 754 NaN encoding(s) supported by 2275 the FPU and the NaN encoding requested with the value 2276 of an ELF file header flag individually set by each 2277 binary. Hardware implementations are permitted to 2278 support either or both of the legacy and the 2008 NaN 2279 encoding mode. 2280 2281 Available settings are as follows: 2282 strict accept binaries that request a NaN encoding 2283 supported by the FPU 2284 legacy only accept legacy-NaN binaries, if supported 2285 by the FPU 2286 2008 only accept 2008-NaN binaries, if supported 2287 by the FPU 2288 relaxed accept any binaries regardless of whether 2289 supported by the FPU 2290 emulated accept any binaries but enable FPU emulator 2291 if binary mode is unsupported by the FPU. 2292 2293 The FPU emulator is always able to support both NaN 2294 encodings, so if no FPU hardware is present or it has 2295 been disabled with 'nofpu', then the settings of 2296 'legacy' and '2008' strap the emulator accordingly, 2297 'relaxed' straps the emulator for both legacy-NaN and 2298 2008-NaN, whereas 'strict' enables legacy-NaN only on 2299 legacy processors and both NaN encodings on MIPS32 or 2300 MIPS64 CPUs. 2301 2302 The setting for ABS.fmt/NEG.fmt instruction execution 2303 mode generally follows that for the NaN encoding, 2304 except where unsupported by hardware. 2305 2306 ignore_loglevel [KNL,EARLY] 2307 Ignore loglevel setting - this will print /all/ 2308 kernel messages to the console. Useful for debugging. 2309 We also add it as printk module parameter, so users 2310 could change it dynamically, usually by 2311 /sys/module/printk/parameters/ignore_loglevel. 2312 2313 ignore_rlimit_data 2314 Ignore RLIMIT_DATA setting for data mappings, 2315 print warning at first misuse. Can be changed via 2316 /sys/module/kernel/parameters/ignore_rlimit_data. 2317 2318 ihash_entries= [KNL] 2319 Set number of hash buckets for inode cache. 2320 2321 ima_appraise= [IMA] appraise integrity measurements 2322 Format: { "off" | "enforce" | "fix" | "log" } 2323 default: "enforce" 2324 2325 ima_appraise_tcb [IMA] Deprecated. Use ima_policy= instead. 2326 The builtin appraise policy appraises all files 2327 owned by uid=0. 2328 2329 ima_canonical_fmt [IMA] 2330 Use the canonical format for the binary runtime 2331 measurements, instead of host native format. 2332 2333 ima_flush_htable [IMA] 2334 Flush the IMA hash table when deleting all the 2335 staged measurement records, to achieve maximum 2336 memory saving at the cost of having duplicate 2337 records across the staged measurement lists. 2338 2339 ima_hash= [IMA] 2340 Format: { md5 | sha1 | rmd160 | sha256 | sha384 2341 | sha512 | ... } 2342 default: "sha1" 2343 2344 The list of supported hash algorithms is defined 2345 in crypto/hash_info.h. 2346 2347 ima_policy= [IMA] 2348 The builtin policies to load during IMA setup. 2349 Format: "tcb | appraise_tcb | secure_boot | 2350 fail_securely | critical_data" 2351 2352 The "tcb" policy measures all programs exec'd, files 2353 mmap'd for exec, and all files opened with the read 2354 mode bit set by either the effective uid (euid=0) or 2355 uid=0. 2356 2357 The "appraise_tcb" policy appraises the integrity of 2358 all files owned by root. 2359 2360 The "secure_boot" policy appraises the integrity 2361 of files (eg. kexec kernel image, kernel modules, 2362 firmware, policy, etc) based on file signatures. 2363 2364 The "fail_securely" policy forces file signature 2365 verification failure also on privileged mounted 2366 filesystems with the SB_I_UNVERIFIABLE_SIGNATURE 2367 flag. 2368 2369 The "critical_data" policy measures kernel integrity 2370 critical data. 2371 2372 ima_tcb [IMA] Deprecated. Use ima_policy= instead. 2373 Load a policy which meets the needs of the Trusted 2374 Computing Base. This means IMA will measure all 2375 programs exec'd, files mmap'd for exec, and all files 2376 opened for read by uid=0. 2377 2378 ima_template= [IMA] 2379 Select one of defined IMA measurements template formats. 2380 Formats: { "ima" | "ima-ng" | "ima-ngv2" | "ima-sig" | 2381 "ima-sigv2" } 2382 Default: "ima-ng" 2383 2384 ima_template_fmt= 2385 [IMA] Define a custom template format. 2386 Format: { "field1|...|fieldN" } 2387 2388 ima= [IMA] Enable or disable IMA 2389 Format: { "off" | "on" } 2390 Default: "on" 2391 Note that disabling IMA is limited to kdump kernel. 2392 2393 indirect_target_selection= [X86,Intel] Mitigation control for Indirect 2394 Target Selection(ITS) bug in Intel CPUs. Updated 2395 microcode is also required for a fix in IBPB. 2396 2397 on: Enable mitigation (default). 2398 off: Disable mitigation. 2399 force: Force the ITS bug and deploy default 2400 mitigation. 2401 vmexit: Only deploy mitigation if CPU is affected by 2402 guest/host isolation part of ITS. 2403 stuff: Deploy RSB-fill mitigation when retpoline is 2404 also deployed. Otherwise, deploy the default 2405 mitigation. 2406 2407 For details see: 2408 Documentation/admin-guide/hw-vuln/indirect-target-selection.rst 2409 2410 init= [KNL] 2411 Format: <full_path> 2412 Run specified binary instead of /sbin/init as init 2413 process. 2414 2415 initcall_debug [KNL] Trace initcalls as they are executed. Useful 2416 for working out where the kernel is dying during 2417 startup. 2418 2419 initcall_blacklist= [KNL] Do not execute a comma-separated list of 2420 initcall functions. Useful for debugging built-in 2421 modules and initcalls. 2422 2423 initramfs_async= [KNL] 2424 Format: <bool> 2425 Default: 1 2426 This parameter controls whether the initramfs 2427 image is unpacked asynchronously, concurrently 2428 with devices being probed and 2429 initialized. This should normally just work, 2430 but as a debugging aid, one can get the 2431 historical behaviour of the initramfs 2432 unpacking being completed before device_ and 2433 late_ initcalls. 2434 2435 initrd= [BOOT,EARLY] Specify the location of the initial ramdisk 2436 2437 initrdmem= [KNL,EARLY] Specify a physical address and size from which to 2438 load the initrd. If an initrd is compiled in or 2439 specified in the bootparams, it takes priority over this 2440 setting. 2441 Format: ss[KMG],nn[KMG] 2442 Default is 0, 0 2443 2444 init_on_alloc= [MM,EARLY] Fill newly allocated pages and heap objects with 2445 zeroes. 2446 Format: 0 | 1 2447 Default set by CONFIG_INIT_ON_ALLOC_DEFAULT_ON. 2448 2449 init_on_free= [MM,EARLY] Fill freed pages and heap objects with zeroes. 2450 Format: 0 | 1 2451 Default set by CONFIG_INIT_ON_FREE_DEFAULT_ON. 2452 2453 init_pkru= [X86] Specify the default memory protection keys rights 2454 register contents for all processes. 0x55555554 by 2455 default (disallow access to all but pkey 0). Can 2456 override in debugfs after boot. 2457 2458 inport.irq= [HW] Inport (ATI XL and Microsoft) busmouse driver 2459 Format: <irq> 2460 2461 int_pln_enable [X86] Enable power limit notification interrupt 2462 2463 integrity_audit=[IMA] 2464 Format: { "0" | "1" } 2465 0 -- basic integrity auditing messages. (Default) 2466 1 -- additional integrity auditing messages. 2467 2468 intel_iommu= [DMAR] Intel IOMMU driver (DMAR) option 2469 on 2470 Enable intel iommu driver. 2471 off 2472 Disable intel iommu driver. 2473 igfx_off [Default Off] 2474 By default, gfx is mapped as normal device. If a gfx 2475 device has a dedicated DMAR unit, the DMAR unit is 2476 bypassed by not enabling DMAR with this option. In 2477 this case, gfx device will use physical address for 2478 DMA. 2479 strict [Default Off] 2480 Deprecated, equivalent to iommu.strict=1. 2481 sp_off [Default Off] 2482 By default, super page will be supported if Intel IOMMU 2483 has the capability. With this option, super page will 2484 not be supported. 2485 sm_on 2486 Enable the Intel IOMMU scalable mode if the hardware 2487 advertises that it has support for the scalable mode 2488 translation. 2489 sm_off 2490 Disallow use of the Intel IOMMU scalable mode. 2491 tboot_noforce [Default Off] 2492 Do not force the Intel IOMMU enabled under tboot. 2493 By default, tboot will force Intel IOMMU on, which 2494 could harm performance of some high-throughput 2495 devices like 40GBit network cards, even if identity 2496 mapping is enabled. 2497 Note that using this option lowers the security 2498 provided by tboot because it makes the system 2499 vulnerable to DMA attacks. 2500 2501 intel_idle.max_cstate= [KNL,HW,ACPI,X86] 2502 0 disables intel_idle and fall back on acpi_idle. 2503 1 to 9 specify maximum depth of C-state. 2504 2505 intel_pstate= [X86,EARLY] 2506 disable 2507 Do not enable intel_pstate as the default 2508 scaling driver for the supported processors 2509 active 2510 Use intel_pstate driver to bypass the scaling 2511 governors layer of cpufreq and provides it own 2512 algorithms for p-state selection. There are two 2513 P-state selection algorithms provided by 2514 intel_pstate in the active mode: powersave and 2515 performance. The way they both operate depends 2516 on whether or not the hardware managed P-states 2517 (HWP) feature has been enabled in the processor 2518 and possibly on the processor model. 2519 passive 2520 Use intel_pstate as a scaling driver, but configure it 2521 to work with generic cpufreq governors (instead of 2522 enabling its internal governor). This mode cannot be 2523 used along with the hardware-managed P-states (HWP) 2524 feature. 2525 force 2526 Enable intel_pstate on systems that prohibit it by default 2527 in favor of acpi-cpufreq. Forcing the intel_pstate driver 2528 instead of acpi-cpufreq may disable platform features, such 2529 as thermal controls and power capping, that rely on ACPI 2530 P-States information being indicated to OSPM and therefore 2531 should be used with caution. This option does not work with 2532 processors that aren't supported by the intel_pstate driver 2533 or on platforms that use pcc-cpufreq instead of acpi-cpufreq. 2534 no_hwp 2535 Do not enable hardware P state control (HWP) 2536 if available. 2537 hwp_only 2538 Only load intel_pstate on systems which support 2539 hardware P state control (HWP) if available. 2540 support_acpi_ppc 2541 Enforce ACPI _PPC performance limits. If the Fixed ACPI 2542 Description Table, specifies preferred power management 2543 profile as "Enterprise Server" or "Performance Server", 2544 then this feature is turned on by default. 2545 per_cpu_perf_limits 2546 Allow per-logical-CPU P-State performance control limits using 2547 cpufreq sysfs interface 2548 no_cas 2549 Do not enable capacity-aware scheduling (CAS) on 2550 hybrid systems 2551 2552 intremap= [X86-64,Intel-IOMMU,EARLY] 2553 on enable Interrupt Remapping (default) 2554 off disable Interrupt Remapping 2555 nosid disable Source ID checking 2556 no_x2apic_optout 2557 BIOS x2APIC opt-out request will be ignored 2558 nopost disable Interrupt Posting 2559 posted_msi 2560 enable MSIs delivered as posted interrupts 2561 2562 iomem= Disable strict checking of access to MMIO memory 2563 strict regions from userspace. 2564 relaxed 2565 2566 iommu= [X86,EARLY] 2567 2568 off 2569 Don't initialize and use any kind of IOMMU. 2570 2571 force 2572 Force the use of the hardware IOMMU even when 2573 it is not actually needed (e.g. because < 3 GB 2574 memory). 2575 2576 noforce 2577 Don't force hardware IOMMU usage when it is not 2578 needed. (default). 2579 2580 merge 2581 Do scatter-gather (SG) merging. Implies "force" 2582 (experimental). 2583 2584 nomerge 2585 Don't do scatter-gather (SG) merging. 2586 2587 biomerge 2588 Do scatter-gather (SG) merging. Implies "force" 2589 (experimental). [same as "merge"] 2590 2591 panic 2592 Always panic when IOMMU overflows. 2593 2594 nopanic 2595 Don't panic on IOMMU overflows. 2596 2597 pt 2598 Use passththrough mode by default 2599 (Equivalent to iommu.passthrough=1) 2600 2601 nopt 2602 Use translated mode for DMA by default 2603 (Equivalent to iommu.passthrough=0) 2604 2605 soft 2606 Use software bounce buffering (SWIOTLB) (default for 2607 Intel machines). This can be used to prevent the usage 2608 of an available hardware IOMMU. 2609 2610 usedac 2611 Use the DAC on VIA PCI bridge 2612 (default: disable the VIA PCI bridge DAC) 2613 2614 AMD Gart HW IOMMU-specific options: (CONFIG_GART_IOMMU) 2615 2616 <size> 2617 Set the size of the remapping area in bytes. 2618 2619 allowed 2620 Overwrite iommu off workarounds for specific chipsets 2621 2622 force 2623 Overwrite iommu off workarounds for specific chipsets 2624 2625 fullflush 2626 Flush IOMMU on each allocation (default). 2627 2628 nofullflush 2629 Don't use IOMMU fullflush. 2630 2631 memaper[=<order>] 2632 Allocate an own aperture over RAM with size 2633 32MB<<order. (default: order=1, i.e. 64MB) 2634 2635 noaperture 2636 Ask the IOMMU not to touch the aperture for AGP. 2637 2638 noagp 2639 Don't initialize the AGP driver and use full aperture. 2640 2641 iommu= [PPC/POWERNV] 2642 2643 nobypass 2644 Disable IOMMU bypass, using IOMMU for PCI devices. 2645 2646 iommu.forcedac= [ARM64,X86,EARLY] Control IOVA allocation for PCI devices. 2647 Format: { "0" | "1" } 2648 0 - Try to allocate a 32-bit DMA address first, before 2649 falling back to the full range if needed. 2650 1 - Allocate directly from the full usable range, 2651 forcing Dual Address Cycle for PCI cards supporting 2652 greater than 32-bit addressing. 2653 2654 iommu.strict= [ARM64,X86,S390,EARLY] Configure TLB invalidation behaviour 2655 Format: { "0" | "1" } 2656 0 - Lazy mode. 2657 Request that DMA unmap operations use deferred 2658 invalidation of hardware TLBs, for increased 2659 throughput at the cost of reduced device isolation. 2660 Will fall back to strict mode if not supported by 2661 the relevant IOMMU driver. 2662 1 - Strict mode. 2663 DMA unmap operations invalidate IOMMU hardware TLBs 2664 synchronously. 2665 unset - Use value of CONFIG_IOMMU_DEFAULT_DMA_{LAZY,STRICT}. 2666 Note: on x86, strict mode specified via one of the 2667 legacy driver-specific options takes precedence. 2668 2669 iommu.passthrough= 2670 [ARM64,X86,EARLY] Configure DMA to bypass the IOMMU by default. 2671 Format: { "0" | "1" } 2672 0 - Use IOMMU translation for DMA. 2673 1 - Bypass the IOMMU for DMA. 2674 unset - Use value of CONFIG_IOMMU_DEFAULT_PASSTHROUGH. 2675 2676 iommu.debug_pagealloc= 2677 [KNL,EARLY] When CONFIG_IOMMU_DEBUG_PAGEALLOC is set, this 2678 parameter enables the feature at boot time. By default, it 2679 is disabled and the system behaves the same way as a kernel 2680 built without CONFIG_IOMMU_DEBUG_PAGEALLOC. 2681 Format: { "0" | "1" } 2682 0 - Sanitizer disabled. 2683 1 - Sanitizer enabled, expect runtime overhead. 2684 2685 io7= [HW] IO7 for Marvel-based Alpha systems 2686 See comment before marvel_specify_io7 in 2687 arch/alpha/kernel/core_marvel.c. 2688 2689 io_delay= [X86,EARLY] I/O delay method 2690 0x80 2691 Standard port 0x80 based delay 2692 0xed 2693 Alternate port 0xed based delay (needed on some systems) 2694 udelay 2695 Simple two microseconds delay 2696 none 2697 No delay 2698 2699 ip= [IP_PNP] 2700 See Documentation/admin-guide/nfs/nfsroot.rst. 2701 2702 ipcmni_extend [KNL,EARLY] Extend the maximum number of unique System V 2703 IPC identifiers from 32,768 to 16,777,216. 2704 2705 ipe.enforce= [IPE] 2706 Format: <bool> 2707 Determine whether IPE starts in permissive (0) or 2708 enforce (1) mode. The default is enforce. 2709 2710 ipe.success_audit= 2711 [IPE] 2712 Format: <bool> 2713 Start IPE with success auditing enabled, emitting 2714 an audit event when a binary is allowed. The default 2715 is 0. 2716 2717 irqaffinity= [SMP] Set the default irq affinity mask 2718 The argument is a cpu list, as described above. 2719 2720 irqchip.gicv2_force_probe= 2721 [ARM,ARM64,EARLY] 2722 Format: <bool> 2723 Force the kernel to look for the second 4kB page 2724 of a GICv2 controller even if the memory range 2725 exposed by the device tree is too small. 2726 2727 irqchip.gicv3_nolpi= 2728 [ARM,ARM64,EARLY] 2729 Force the kernel to ignore the availability of 2730 LPIs (and by consequence ITSs). Intended for system 2731 that use the kernel as a bootloader, and thus want 2732 to let secondary kernels in charge of setting up 2733 LPIs. 2734 2735 irqchip.gicv3_pseudo_nmi= [ARM64,EARLY] 2736 Enables support for pseudo-NMIs in the kernel. This 2737 requires the kernel to be built with 2738 CONFIG_ARM64_PSEUDO_NMI. 2739 2740 irqchip.riscv_imsic_noipi 2741 [RISC-V,EARLY] 2742 Force the kernel to not use IMSIC software injected MSIs 2743 as IPIs. Intended for system where IMSIC is trap-n-emulated, 2744 and thus want to reduce MMIO traps when triggering IPIs 2745 to multiple harts. 2746 2747 irqfixup [HW] 2748 When an interrupt is not handled search all handlers 2749 for it. Intended to get systems with badly broken 2750 firmware running. 2751 2752 irqhandler.duration_warn_us= [KNL] 2753 Warn if an IRQ handler exceeds the specified duration 2754 threshold in microseconds. Useful for identifying 2755 long-running IRQs in the system. 2756 2757 irqpoll [HW] 2758 When an interrupt is not handled search all handlers 2759 for it. Also check all handlers each timer 2760 interrupt. Intended to get systems with badly broken 2761 firmware running. 2762 2763 isapnp= [ISAPNP] 2764 Format: <RDP>,<reset>,<pci_scan>,<verbosity> 2765 2766 isolcpus= [KNL,SMP,ISOL] Isolate a given set of CPUs from disturbance. 2767 Format: [flag-list,]<cpu-list> 2768 2769 Specify one or more CPUs to isolate from disturbances 2770 specified in the flag list (default: domain): 2771 2772 nohz 2773 Disable the tick when a single task runs as well as 2774 disabling other kernel noises like having RCU callbacks 2775 offloaded. This is equivalent to the nohz_full parameter. 2776 2777 A residual 1Hz tick is offloaded to workqueues, which you 2778 need to affine to housekeeping through the global 2779 workqueue's affinity configured via the 2780 /sys/devices/virtual/workqueue/cpumask sysfs file, or 2781 by using the 'domain' flag described below. 2782 2783 NOTE: by default the global workqueue runs on all CPUs, 2784 so to protect individual CPUs the 'cpumask' file has to 2785 be configured manually after bootup. 2786 2787 domain 2788 Isolate from the general SMP balancing and scheduling 2789 algorithms. Note that performing domain isolation this way 2790 is irreversible: it's not possible to bring back a CPU to 2791 the domains once isolated through this boot time 2792 configuration. Use cpusets for a dynamic configuration 2793 which can be altered at runtime. For details see 2794 Documentation/admin-guide/cpu-isolation.rst. 2795 2796 You can move a process onto or off an "isolated" CPU via 2797 the CPU affinity syscalls or cpuset. 2798 <cpu number> begins at 0 and the maximum value is 2799 "number of CPUs in system - 1". 2800 2801 managed_irq 2802 2803 Isolate from being targeted by managed interrupts 2804 which have an interrupt mask containing isolated 2805 CPUs. The affinity of managed interrupts is 2806 handled by the kernel and cannot be changed via 2807 the /proc/irq/* interfaces. 2808 2809 This isolation is best effort and only effective 2810 if the automatically assigned interrupt mask of a 2811 device queue contains isolated and housekeeping 2812 CPUs. If housekeeping CPUs are online then such 2813 interrupts are directed to the housekeeping CPU 2814 so that IO submitted on the housekeeping CPU 2815 cannot disturb the isolated CPU. 2816 2817 If a queue's affinity mask contains only isolated 2818 CPUs then this parameter has no effect on the 2819 interrupt routing decision, though interrupts are 2820 only delivered when tasks running on those 2821 isolated CPUs submit IO. IO submitted on 2822 housekeeping CPUs has no influence on those 2823 queues. 2824 2825 The format of <cpu-list> is described above. 2826 2827 iucv= [HW,NET] 2828 2829 ivrs_ioapic [HW,X86-64] 2830 Provide an override to the IOAPIC-ID<->DEVICE-ID 2831 mapping provided in the IVRS ACPI table. 2832 By default, PCI segment is 0, and can be omitted. 2833 2834 For example, to map IOAPIC-ID decimal 10 to 2835 PCI segment 0x1 and PCI device 00:14.0, 2836 write the parameter as: 2837 ivrs_ioapic=10@0001:00:14.0 2838 2839 Deprecated formats: 2840 * To map IOAPIC-ID decimal 10 to PCI device 00:14.0 2841 write the parameter as: 2842 ivrs_ioapic[10]=00:14.0 2843 * To map IOAPIC-ID decimal 10 to PCI segment 0x1 and 2844 PCI device 00:14.0 write the parameter as: 2845 ivrs_ioapic[10]=0001:00:14.0 2846 2847 ivrs_hpet [HW,X86-64] 2848 Provide an override to the HPET-ID<->DEVICE-ID 2849 mapping provided in the IVRS ACPI table. 2850 By default, PCI segment is 0, and can be omitted. 2851 2852 For example, to map HPET-ID decimal 10 to 2853 PCI segment 0x1 and PCI device 00:14.0, 2854 write the parameter as: 2855 ivrs_hpet=10@0001:00:14.0 2856 2857 Deprecated formats: 2858 * To map HPET-ID decimal 0 to PCI device 00:14.0 2859 write the parameter as: 2860 ivrs_hpet[0]=00:14.0 2861 * To map HPET-ID decimal 10 to PCI segment 0x1 and 2862 PCI device 00:14.0 write the parameter as: 2863 ivrs_ioapic[10]=0001:00:14.0 2864 2865 ivrs_acpihid [HW,X86-64] 2866 Provide an override to the ACPI-HID:UID<->DEVICE-ID 2867 mapping provided in the IVRS ACPI table. 2868 By default, PCI segment is 0, and can be omitted. 2869 2870 For example, to map UART-HID:UID AMD0020:0 to 2871 PCI segment 0x1 and PCI device ID 00:14.5, 2872 write the parameter as: 2873 ivrs_acpihid=AMD0020:0@0001:00:14.5 2874 2875 Deprecated formats: 2876 * To map UART-HID:UID AMD0020:0 to PCI segment is 0, 2877 PCI device ID 00:14.5, write the parameter as: 2878 ivrs_acpihid[00:14.5]=AMD0020:0 2879 * To map UART-HID:UID AMD0020:0 to PCI segment 0x1 and 2880 PCI device ID 00:14.5, write the parameter as: 2881 ivrs_acpihid[0001:00:14.5]=AMD0020:0 2882 2883 js= [HW,JOY] Analog joystick 2884 See Documentation/input/joydev/joystick.rst. 2885 2886 kasan_multi_shot 2887 [KNL] Enforce KASAN (Kernel Address Sanitizer) to print 2888 report on every invalid memory access. Without this 2889 parameter KASAN will print report only for the first 2890 invalid access. 2891 2892 keep_bootcon [KNL,EARLY] 2893 Do not unregister boot console at start. This is only 2894 useful for debugging when something happens in the window 2895 between unregistering the boot console and initializing 2896 the real console. 2897 2898 keepinitrd [HW,ARM] See retain_initrd. 2899 2900 kernelcore= [KNL,X86,PPC,EARLY] 2901 Format: nn[KMGTPE] | nn% | "mirror" 2902 This parameter specifies the amount of memory usable by 2903 the kernel for non-movable allocations. The requested 2904 amount is spread evenly throughout all nodes in the 2905 system as ZONE_NORMAL. The remaining memory is used for 2906 movable memory in its own zone, ZONE_MOVABLE. In the 2907 event, a node is too small to have both ZONE_NORMAL and 2908 ZONE_MOVABLE, kernelcore memory will take priority and 2909 other nodes will have a larger ZONE_MOVABLE. 2910 2911 ZONE_MOVABLE is used for the allocation of pages that 2912 may be reclaimed or moved by the page migration 2913 subsystem. Note that allocations like PTEs-from-HighMem 2914 still use the HighMem zone if it exists, and the Normal 2915 zone if it does not. 2916 2917 It is possible to specify the exact amount of memory in 2918 the form of "nn[KMGTPE]", a percentage of total system 2919 memory in the form of "nn%", or "mirror". If "mirror" 2920 option is specified, mirrored (reliable) memory is used 2921 for non-movable allocations and remaining memory is used 2922 for Movable pages. "nn[KMGTPE]", "nn%", and "mirror" 2923 are exclusive, so you cannot specify multiple forms. 2924 2925 kfence.burst= [MM,KFENCE] The number of additional successive 2926 allocations to be attempted through KFENCE for each 2927 sample interval. 2928 Format: <unsigned integer> 2929 Default: 0 2930 2931 kfence.check_on_panic= 2932 [MM,KFENCE] Whether to check all KFENCE-managed objects' 2933 canaries on panic. 2934 Format: <bool> 2935 Default: false 2936 2937 kfence.deferrable= 2938 [MM,KFENCE] Whether to use a deferrable timer to trigger 2939 allocations. This avoids forcing CPU wake-ups if the 2940 system is idle, at the risk of a less predictable 2941 sample interval. 2942 Format: <bool> 2943 Default: CONFIG_KFENCE_DEFERRABLE 2944 2945 kfence.fault= [MM,KFENCE] Controls the behavior when a KFENCE 2946 error is detected. 2947 report - print the error report and continue (default). 2948 oops - print the error report and oops. 2949 panic - print the error report and panic. 2950 2951 kfence.sample_interval= 2952 [MM,KFENCE] KFENCE's sample interval in milliseconds. 2953 Format: <unsigned integer> 2954 0 - Disable KFENCE. 2955 >0 - Enabled KFENCE with given sample interval. 2956 Default: CONFIG_KFENCE_SAMPLE_INTERVAL 2957 2958 kfence.skip_covered_thresh= 2959 [MM,KFENCE] If pool utilization reaches this threshold 2960 (pool usage%), KFENCE limits currently covered 2961 allocations of the same source from further filling 2962 up the pool. 2963 Format: <unsigned integer> 2964 Default: 75 2965 2966 kgdbdbgp= [KGDB,HW,EARLY] kgdb over EHCI usb debug port. 2967 Format: <Controller#>[,poll interval] 2968 The controller # is the number of the ehci usb debug 2969 port as it is probed via PCI. The poll interval is 2970 optional and is the number seconds in between 2971 each poll cycle to the debug port in case you need 2972 the functionality for interrupting the kernel with 2973 gdb or control-c on the dbgp connection. When 2974 not using this parameter you use sysrq-g to break into 2975 the kernel debugger. 2976 2977 kgdboc= [KGDB,HW] kgdb over consoles. 2978 Requires a tty driver that supports console polling, 2979 or a supported polling keyboard driver (non-usb). 2980 Serial only format: <serial_device>[,baud] 2981 keyboard only format: kbd 2982 keyboard and serial format: kbd,<serial_device>[,baud] 2983 Optional Kernel mode setting: 2984 kms, kbd format: kms,kbd 2985 kms, kbd and serial format: kms,kbd,<ser_dev>[,baud] 2986 2987 kgdboc_earlycon= [KGDB,HW,EARLY] 2988 If the boot console provides the ability to read 2989 characters and can work in polling mode, you can use 2990 this parameter to tell kgdb to use it as a backend 2991 until the normal console is registered. Intended to 2992 be used together with the kgdboc parameter which 2993 specifies the normal console to transition to. 2994 2995 The name of the early console should be specified 2996 as the value of this parameter. Note that the name of 2997 the early console might be different than the tty 2998 name passed to kgdboc. It's OK to leave the value 2999 blank and the first boot console that implements 3000 read() will be picked. 3001 3002 kgdbwait [KGDB,EARLY] Stop kernel execution and enter the 3003 kernel debugger at the earliest opportunity. 3004 3005 kho= [KEXEC,EARLY] 3006 Format: { "0" | "1" | "off" | "on" | "y" | "n" } 3007 Enables or disables Kexec HandOver. 3008 "0" | "off" | "n" - kexec handover is disabled 3009 "1" | "on" | "y" - kexec handover is enabled 3010 3011 kho_scratch= [KEXEC,EARLY] 3012 Format: ll[KMG],mm[KMG],nn[KMG] | nn% 3013 Defines the size of the KHO scratch region. The KHO 3014 scratch regions are physically contiguous memory 3015 ranges that can only be used for non-kernel 3016 allocations. That way, even when memory is heavily 3017 fragmented with handed over memory, the kexeced 3018 kernel will always have enough contiguous ranges to 3019 bootstrap itself. 3020 3021 It is possible to specify the exact amount of 3022 memory in the form of "ll[KMG],mm[KMG],nn[KMG]" 3023 where the first parameter defines the size of a low 3024 memory scratch area, the second parameter defines 3025 the size of a global scratch area and the third 3026 parameter defines the size of additional per-node 3027 scratch areas. The form "nn%" defines scale factor 3028 (in percents) of memory that was used during boot. 3029 3030 kmac= [MIPS] Korina ethernet MAC address. 3031 Configure the RouterBoard 532 series on-chip 3032 Ethernet adapter MAC address. 3033 3034 kmemleak= [KNL,EARLY] Boot-time kmemleak enable/disable 3035 Valid arguments: on, off 3036 Default: on 3037 Built with CONFIG_DEBUG_KMEMLEAK_DEFAULT_OFF=y, 3038 the default is off. 3039 3040 kprobe_event=[probe-list] 3041 [FTRACE] Add kprobe events and enable at boot time. 3042 The probe-list is a semicolon delimited list of probe 3043 definitions. Each definition is same as kprobe_events 3044 interface, but the parameters are comma delimited. 3045 For example, to add a kprobe event on vfs_read with 3046 arg1 and arg2, add to the command line; 3047 3048 kprobe_event=p,vfs_read,$arg1,$arg2 3049 3050 See also Documentation/trace/kprobetrace.rst "Kernel 3051 Boot Parameter" section. 3052 3053 kpti= [ARM64,EARLY] Control page table isolation of 3054 user and kernel address spaces. 3055 Default: enabled on cores which need mitigation. 3056 0: force disabled 3057 1: force enabled 3058 3059 kunit.enable= [KUNIT] Enable executing KUnit tests. Requires 3060 CONFIG_KUNIT to be set to be fully enabled. The 3061 default value can be overridden via 3062 KUNIT_DEFAULT_ENABLED. 3063 Default is 1 (enabled) 3064 3065 kvm.ignore_msrs=[KVM] Ignore guest accesses to unhandled MSRs. 3066 Default is 0 (don't ignore, but inject #GP) 3067 3068 kvm.eager_page_split= 3069 [KVM,X86,RISCV] Controls whether or not KVM will try to 3070 proactively split all huge pages during dirty logging. 3071 Eager page splitting reduces interruptions to vCPU 3072 execution by eliminating the write-protection faults 3073 and MMU lock contention that would otherwise be 3074 required to split huge pages lazily. 3075 3076 VM workloads that rarely perform writes or that write 3077 only to a small region of VM memory may benefit from 3078 disabling eager page splitting to allow huge pages to 3079 still be used for reads. 3080 3081 The behavior of eager page splitting depends on whether 3082 KVM_DIRTY_LOG_INITIALLY_SET is enabled or disabled. If 3083 disabled, all huge pages in a memslot will be eagerly 3084 split when dirty logging is enabled on that memslot. If 3085 enabled, eager page splitting will be performed during 3086 the KVM_CLEAR_DIRTY ioctl, and only for the pages being 3087 cleared. 3088 3089 On x86, eager page splitting is only supported when 3090 kvm.tdp_mmu=Y. 3091 3092 On RISCV, eager page splitting is supported by default. 3093 3094 Default is Y (on). 3095 3096 kvm.enable_pmu=[KVM,X86] 3097 If enabled, KVM will virtualize PMU functionality based 3098 on the virtual CPU model defined by userspace. This 3099 can be overridden on a per-VM basis via 3100 KVM_CAP_PMU_CAPABILITY. 3101 3102 If disabled, KVM will not virtualize PMU functionality, 3103 e.g. MSRs, PMCs, PMIs, etc., even if userspace defines 3104 a virtual CPU model that contains PMU assets. 3105 3106 Note, KVM's vPMU support implicitly requires running 3107 with an in-kernel local APIC, e.g. to deliver PMIs to 3108 the guest. Running without an in-kernel local APIC is 3109 not supported, though KVM will allow such a combination 3110 (with severely degraded functionality). 3111 3112 See also enable_mediated_pmu. 3113 3114 Default is Y (on). 3115 3116 kvm.enable_virt_at_load=[KVM,ARM64,LOONGARCH,MIPS,RISCV,X86] 3117 If enabled, KVM will enable virtualization in hardware 3118 when KVM is loaded, and disable virtualization when KVM 3119 is unloaded (if KVM is built as a module). 3120 3121 If disabled, KVM will dynamically enable and disable 3122 virtualization on-demand when creating and destroying 3123 VMs, i.e. on the 0=>1 and 1=>0 transitions of the 3124 number of VMs. 3125 3126 Enabling virtualization at module load avoids potential 3127 latency for creation of the 0=>1 VM, as KVM serializes 3128 virtualization enabling across all online CPUs. The 3129 "cost" of enabling virtualization when KVM is loaded, 3130 is that doing so may interfere with using out-of-tree 3131 hypervisors that want to "own" virtualization hardware. 3132 3133 kvm.enable_vmware_backdoor=[KVM] Support VMware backdoor PV interface. 3134 Default is false (don't support). 3135 3136 kvm.nx_huge_pages= 3137 [KVM] Controls the software workaround for the 3138 X86_BUG_ITLB_MULTIHIT bug. 3139 force : Always deploy workaround. 3140 off : Never deploy workaround. 3141 auto : Deploy workaround based on the presence of 3142 X86_BUG_ITLB_MULTIHIT. 3143 3144 Default is 'auto'. 3145 3146 If the software workaround is enabled for the host, 3147 guests do need not to enable it for nested guests. 3148 3149 kvm.nx_huge_pages_recovery_ratio= 3150 [KVM] Controls how many 4KiB pages are periodically zapped 3151 back to huge pages. 0 disables the recovery, otherwise if 3152 the value is N KVM will zap 1/Nth of the 4KiB pages every 3153 period (see below). The default is 60. 3154 3155 kvm.nx_huge_pages_recovery_period_ms= 3156 [KVM] Controls the time period at which KVM zaps 4KiB pages 3157 back to huge pages. If the value is a non-zero N, KVM will 3158 zap a portion (see ratio above) of the pages every N msecs. 3159 If the value is 0 (the default), KVM will pick a period based 3160 on the ratio, such that a page is zapped after 1 hour on average. 3161 3162 kvm-{amd,intel}.enable_mediated_pmu=[KVM,AMD,INTEL] 3163 If enabled, KVM will provide a mediated virtual PMU, 3164 instead of the default perf-based virtual PMU (if 3165 kvm.enable_pmu is true and PMU is enumerated via the 3166 virtual CPU model). 3167 3168 With a perf-based vPMU, KVM operates as a user of perf, 3169 i.e. emulates guest PMU counters using perf events. 3170 KVM-created perf events are managed by perf as regular 3171 (guest-only) events, e.g. are scheduled in/out, contend 3172 for hardware resources, etc. Using a perf-based vPMU 3173 allows guest and host usage of the PMU to co-exist, but 3174 incurs non-trivial overhead and can result in silently 3175 dropped guest events (due to resource contention). 3176 3177 With a mediated vPMU, hardware PMU state is context 3178 switched around the world switch to/from the guest. 3179 KVM mediates which events the guest can utilize, but 3180 gives the guest direct access to all other PMU assets 3181 when possible (KVM may intercept some accesses if the 3182 virtual CPU model provides a subset of hardware PMU 3183 functionality). Using a mediated vPMU significantly 3184 reduces PMU virtualization overhead and eliminates lost 3185 guest events, but is mutually exclusive with using perf 3186 to profile KVM guests and adds latency to most VM-Exits 3187 (to context switch PMU state). 3188 3189 Default is N (off). 3190 3191 kvm-amd.nested= [KVM,AMD] Control nested virtualization feature in 3192 KVM/SVM. Default is 1 (enabled). 3193 3194 kvm-amd.npt= [KVM,AMD] Control KVM's use of Nested Page Tables, 3195 a.k.a. Two-Dimensional Page Tables. Default is 1 3196 (enabled). Disable by KVM if hardware lacks support 3197 for NPT. 3198 3199 kvm-amd.ciphertext_hiding_asids= 3200 [KVM,AMD] Ciphertext hiding prevents disallowed accesses 3201 to SNP private memory from reading ciphertext. Instead, 3202 reads will see constant default values (0xff). 3203 3204 If ciphertext hiding is enabled, the joint SEV-ES and 3205 SEV-SNP ASID space is partitioned into separate SEV-ES 3206 and SEV-SNP ASID ranges, with the SEV-SNP range being 3207 [1..max_snp_asid] and the SEV-ES range being 3208 (max_snp_asid..min_sev_asid), where min_sev_asid is 3209 enumerated by CPUID.0x.8000_001F[EDX]. 3210 3211 A non-zero value enables SEV-SNP ciphertext hiding and 3212 adjusts the ASID ranges for SEV-ES and SEV-SNP guests. 3213 KVM caps the number of SEV-SNP ASIDs at the maximum 3214 possible value, e.g. specifying -1u will assign all 3215 joint SEV-ES and SEV-SNP ASIDs to SEV-SNP. Note, 3216 assigning all joint ASIDs to SEV-SNP, i.e. configuring 3217 max_snp_asid == min_sev_asid-1, will effectively make 3218 SEV-ES unusable. 3219 3220 kvm-arm.mode= 3221 [KVM,ARM,EARLY] Select one of KVM/arm64's modes of 3222 operation. 3223 3224 none: Forcefully disable KVM. 3225 3226 nvhe: Standard nVHE-based mode, without support for 3227 protected guests. 3228 3229 protected: Mode with support for guests whose state is 3230 kept private from the host, using VHE or 3231 nVHE depending on HW support. 3232 3233 nested: VHE-based mode with support for nested 3234 virtualization. Requires at least ARMv8.4 3235 hardware (with FEAT_NV2). 3236 3237 Defaults to VHE/nVHE based on hardware support. Setting 3238 mode to "protected" will disable kexec and hibernation 3239 for the host. To force nVHE on VHE hardware, add 3240 "arm64_sw.hvhe=0 id_aa64mmfr1.vh=0" to the 3241 command-line. 3242 "nested" and "protected" are experimental and should be 3243 used with extreme caution. 3244 3245 kvm-arm.vgic_v3_group0_trap= 3246 [KVM,ARM,EARLY] Trap guest accesses to GICv3 group-0 3247 system registers 3248 3249 kvm-arm.vgic_v3_group1_trap= 3250 [KVM,ARM,EARLY] Trap guest accesses to GICv3 group-1 3251 system registers 3252 3253 kvm-arm.vgic_v3_common_trap= 3254 [KVM,ARM,EARLY] Trap guest accesses to GICv3 common 3255 system registers 3256 3257 kvm-arm.vgic_v4_enable= 3258 [KVM,ARM,EARLY] Allow use of GICv4 for direct 3259 injection of LPIs. 3260 3261 kvm-arm.wfe_trap_policy= 3262 [KVM,ARM] Control when to set WFE instruction trap for 3263 KVM VMs. Traps are allowed but not guaranteed by the 3264 CPU architecture. 3265 3266 trap: set WFE instruction trap 3267 3268 notrap: clear WFE instruction trap 3269 3270 kvm-arm.wfi_trap_policy= 3271 [KVM,ARM] Control when to set WFI instruction trap for 3272 KVM VMs. Traps are allowed but not guaranteed by the 3273 CPU architecture. 3274 3275 trap: set WFI instruction trap 3276 3277 notrap: clear WFI instruction trap 3278 3279 kvm_cma_resv_ratio=n [PPC,EARLY] 3280 Reserves given percentage from system memory area for 3281 contiguous memory allocation for KVM hash pagetable 3282 allocation. 3283 By default it reserves 5% of total system memory. 3284 Format: <integer> 3285 Default: 5 3286 3287 kvm-intel.ept= [KVM,Intel] Control KVM's use of Extended Page Tables, 3288 a.k.a. Two-Dimensional Page Tables. Default is 1 3289 (enabled). Disable by KVM if hardware lacks support 3290 for EPT. 3291 3292 kvm-intel.emulate_invalid_guest_state= 3293 [KVM,Intel] Control whether to emulate invalid guest 3294 state. Ignored if kvm-intel.enable_unrestricted_guest=1, 3295 as guest state is never invalid for unrestricted 3296 guests. This param doesn't apply to nested guests (L2), 3297 as KVM never emulates invalid L2 guest state. 3298 Default is 1 (enabled). 3299 3300 kvm-intel.flexpriority= 3301 [KVM,Intel] Control KVM's use of FlexPriority feature 3302 (TPR shadow). Default is 1 (enabled). Disable by KVM if 3303 hardware lacks support for it. 3304 3305 kvm-intel.nested= 3306 [KVM,Intel] Control nested virtualization feature in 3307 KVM/VMX. Default is 1 (enabled). 3308 3309 kvm-intel.unrestricted_guest= 3310 [KVM,Intel] Control KVM's use of unrestricted guest 3311 feature (virtualized real and unpaged mode). Default 3312 is 1 (enabled). Disable by KVM if EPT is disabled or 3313 hardware lacks support for it. 3314 3315 kvm-intel.vmentry_l1d_flush=[KVM,Intel] Mitigation for L1 Terminal Fault 3316 CVE-2018-3620. 3317 3318 Valid arguments: never, cond, always 3319 3320 always: L1D cache flush on every VMENTER. 3321 cond: Flush L1D on VMENTER only when the code between 3322 VMEXIT and VMENTER can leak host memory. 3323 never: Disables the mitigation 3324 3325 Default is cond (do L1 cache flush in specific instances) 3326 3327 kvm-intel.vpid= [KVM,Intel] Control KVM's use of Virtual Processor 3328 Identification feature (tagged TLBs). Default is 1 3329 (enabled). Disable by KVM if hardware lacks support 3330 for it. 3331 3332 l1d_flush= [X86,INTEL,EARLY] 3333 Control mitigation for L1D based snooping vulnerability. 3334 3335 Certain CPUs are vulnerable to an exploit against CPU 3336 internal buffers which can forward information to a 3337 disclosure gadget under certain conditions. 3338 3339 In vulnerable processors, the speculatively 3340 forwarded data can be used in a cache side channel 3341 attack, to access data to which the attacker does 3342 not have direct access. 3343 3344 This parameter controls the mitigation. The 3345 options are: 3346 3347 on - enable the interface for the mitigation 3348 3349 l1tf= [X86,EARLY] Control mitigation of the L1TF vulnerability on 3350 affected CPUs 3351 3352 The kernel PTE inversion protection is unconditionally 3353 enabled and cannot be disabled. 3354 3355 full 3356 Provides all available mitigations for the 3357 L1TF vulnerability. Disables SMT and 3358 enables all mitigations in the 3359 hypervisors, i.e. unconditional L1D flush. 3360 3361 SMT control and L1D flush control via the 3362 sysfs interface is still possible after 3363 boot. Hypervisors will issue a warning 3364 when the first VM is started in a 3365 potentially insecure configuration, 3366 i.e. SMT enabled or L1D flush disabled. 3367 3368 full,force 3369 Same as 'full', but disables SMT and L1D 3370 flush runtime control. Implies the 3371 'nosmt=force' command line option. 3372 (i.e. sysfs control of SMT is disabled.) 3373 3374 flush 3375 Leaves SMT enabled and enables the default 3376 hypervisor mitigation, i.e. conditional 3377 L1D flush. 3378 3379 SMT control and L1D flush control via the 3380 sysfs interface is still possible after 3381 boot. Hypervisors will issue a warning 3382 when the first VM is started in a 3383 potentially insecure configuration, 3384 i.e. SMT enabled or L1D flush disabled. 3385 3386 flush,nosmt 3387 3388 Disables SMT and enables the default 3389 hypervisor mitigation. 3390 3391 SMT control and L1D flush control via the 3392 sysfs interface is still possible after 3393 boot. Hypervisors will issue a warning 3394 when the first VM is started in a 3395 potentially insecure configuration, 3396 i.e. SMT enabled or L1D flush disabled. 3397 3398 flush,nowarn 3399 Same as 'flush', but hypervisors will not 3400 warn when a VM is started in a potentially 3401 insecure configuration. 3402 3403 off 3404 Disables hypervisor mitigations and doesn't 3405 emit any warnings. 3406 It also drops the swap size and available 3407 RAM limit restriction on both hypervisor and 3408 bare metal. 3409 3410 Default is 'flush'. 3411 3412 For details see: Documentation/admin-guide/hw-vuln/l1tf.rst 3413 3414 l2cr= [PPC] 3415 3416 l3cr= [PPC] 3417 3418 lapic [X86-32,APIC,EARLY] Enable the local APIC even if BIOS 3419 disabled it. 3420 3421 lapic= [X86,APIC] Do not use TSC deadline 3422 value for LAPIC timer one-shot implementation. Default 3423 back to the programmable timer unit in the LAPIC. 3424 Format: notscdeadline 3425 3426 lapic_timer_c2_ok [X86,APIC,EARLY] trust the local apic timer 3427 in C2 power state. 3428 3429 libata.dma= [LIBATA] DMA control 3430 libata.dma=0 Disable all PATA and SATA DMA 3431 libata.dma=1 PATA and SATA Disk DMA only 3432 libata.dma=2 ATAPI (CDROM) DMA only 3433 libata.dma=4 Compact Flash DMA only 3434 Combinations also work, so libata.dma=3 enables DMA 3435 for disks and CDROMs, but not CFs. 3436 3437 libata.ignore_hpa= [LIBATA] Ignore HPA limit 3438 libata.ignore_hpa=0 keep BIOS limits (default) 3439 libata.ignore_hpa=1 ignore limits, using full disk 3440 3441 libata.noacpi [LIBATA] Disables use of ACPI in libata suspend/resume 3442 when set. 3443 Format: <int> 3444 3445 libata.force= [LIBATA] Force configurations. The format is a comma- 3446 separated list of "[ID:]VAL" where ID is PORT[.DEVICE]. 3447 PORT and DEVICE are decimal numbers matching port, link 3448 or device. Basically, it matches the ATA ID string 3449 printed on console by libata. If the whole ID part is 3450 omitted, the last PORT and DEVICE values are used. If 3451 ID hasn't been specified yet, the configuration applies 3452 to all ports, links and devices. 3453 3454 If only DEVICE is omitted, the parameter applies to 3455 the port and all links and devices behind it. DEVICE 3456 number of 0 either selects the first device or the 3457 first fan-out link behind PMP device. It does not 3458 select the host link. DEVICE number of 15 selects the 3459 host link and device attached to it. 3460 3461 The VAL specifies the configuration to force. As long 3462 as there is no ambiguity, shortcut notation is allowed. 3463 For example, both 1.5 and 1.5G would work for 1.5Gbps. 3464 The following configurations can be forced. 3465 3466 * Cable type: 40c, 80c, short40c, unk, ign or sata. 3467 Any ID with matching PORT is used. 3468 3469 * SATA link speed limit: 1.5Gbps or 3.0Gbps. 3470 3471 * Transfer mode: pio[0-7], mwdma[0-4] and udma[0-7]. 3472 udma[/][16,25,33,44,66,100,133] notation is also 3473 allowed. 3474 3475 * nohrst, nosrst, norst: suppress hard, soft and both 3476 resets. 3477 3478 * rstonce: only attempt one reset during hot-unplug 3479 link recovery. 3480 3481 * [no]dbdelay: Enable or disable the extra 200ms delay 3482 before debouncing a link PHY and device presence 3483 detection. 3484 3485 * [no]ncq: Turn on or off NCQ. 3486 3487 * [no]ncqtrim: Enable or disable queued DSM TRIM. 3488 3489 * [no]ncqati: Enable or disable NCQ trim on ATI chipset. 3490 3491 * [no]trim: Enable or disable (unqueued) TRIM. 3492 3493 * trim_zero: Indicate that TRIM command zeroes data. 3494 3495 * max_trim_128m: Set 128M maximum trim size limit. 3496 3497 * [no]dma: Turn on or off DMA transfers. 3498 3499 * atapi_dmadir: Enable ATAPI DMADIR bridge support. 3500 3501 * atapi_mod16_dma: Enable the use of ATAPI DMA for 3502 commands that are not a multiple of 16 bytes. 3503 3504 * [no]dmalog: Enable or disable the use of the 3505 READ LOG DMA EXT command to access logs. 3506 3507 * [no]iddevlog: Enable or disable access to the 3508 identify device data log. 3509 3510 * [no]logdir: Enable or disable access to the general 3511 purpose log directory. 3512 3513 * max_sec=<sectors>: Set the transfer size limit, in 3514 number of 512-byte sectors, to the value specified in 3515 <sectors>. The value specified in <sectors> has to be 3516 a non-zero positive integer. 3517 3518 * max_sec_128: Set transfer size limit to 128 sectors. 3519 3520 * max_sec_1024: Set or clear transfer size limit to 3521 1024 sectors. 3522 3523 * max_sec_lba48: Set or clear transfer size limit to 3524 65535 sectors. 3525 3526 * external: Mark port as external (hotplug-capable). 3527 3528 * [no]lpm: Enable or disable link power management. 3529 3530 * [no]setxfer: Indicate if transfer speed mode setting 3531 should be skipped. 3532 3533 * [no]fua: Disable or enable FUA (Force Unit Access) 3534 support for devices supporting this feature. 3535 3536 * dump_id: Dump IDENTIFY data. 3537 3538 * disable: Disable this device. 3539 3540 If there are multiple matching configurations changing 3541 the same attribute, the last one is used. 3542 3543 liveupdate= [KNL,EARLY] 3544 Format: <bool> 3545 Enable Live Update Orchestrator (LUO). 3546 Default: off. 3547 3548 lockd.nlm_grace_period=P [NFS] Assign grace period. 3549 Format: <integer> 3550 3551 lockd.nlm_tcpport=N [NFS] Assign TCP port. 3552 Format: <integer> 3553 3554 lockd.nlm_timeout=T [NFS] Assign timeout value. 3555 Format: <integer> 3556 3557 lockd.nlm_udpport=M [NFS] Assign UDP port. 3558 Format: <integer> 3559 3560 lockdown= [SECURITY,EARLY] 3561 { integrity | confidentiality } 3562 Enable the kernel lockdown feature. If set to 3563 integrity, kernel features that allow userland to 3564 modify the running kernel are disabled. If set to 3565 confidentiality, kernel features that allow userland 3566 to extract confidential information from the kernel 3567 are also disabled. 3568 3569 locktorture.acq_writer_lim= [KNL] 3570 Set the time limit in jiffies for a lock 3571 acquisition. Acquisitions exceeding this limit 3572 will result in a splat once they do complete. 3573 3574 locktorture.bind_readers= [KNL] 3575 Specify the list of CPUs to which the readers are 3576 to be bound. 3577 3578 locktorture.bind_writers= [KNL] 3579 Specify the list of CPUs to which the writers are 3580 to be bound. 3581 3582 locktorture.call_rcu_chains= [KNL] 3583 Specify the number of self-propagating call_rcu() 3584 chains to set up. These are used to ensure that 3585 there is a high probability of an RCU grace period 3586 in progress at any given time. Defaults to 0, 3587 which disables these call_rcu() chains. 3588 3589 locktorture.long_hold= [KNL] 3590 Specify the duration in milliseconds for the 3591 occasional long-duration lock hold time. Defaults 3592 to 100 milliseconds. Select 0 to disable. 3593 3594 locktorture.nested_locks= [KNL] 3595 Specify the maximum lock nesting depth that 3596 locktorture is to exercise, up to a limit of 8 3597 (MAX_NESTED_LOCKS). Specify zero to disable. 3598 Note that this parameter is ineffective on types 3599 of locks that do not support nested acquisition. 3600 3601 locktorture.nreaders_stress= [KNL] 3602 Set the number of locking read-acquisition kthreads. 3603 Defaults to being automatically set based on the 3604 number of online CPUs. 3605 3606 locktorture.nwriters_stress= [KNL] 3607 Set the number of locking write-acquisition kthreads. 3608 3609 locktorture.onoff_holdoff= [KNL] 3610 Set time (s) after boot for CPU-hotplug testing. 3611 3612 locktorture.onoff_interval= [KNL] 3613 Set time (s) between CPU-hotplug operations, or 3614 zero to disable CPU-hotplug testing. 3615 3616 locktorture.rt_boost= [KNL] 3617 Do periodic testing of real-time lock priority 3618 boosting. Select 0 to disable, 1 to boost 3619 only rt_mutex, and 2 to boost unconditionally. 3620 Defaults to 2, which might seem to be an 3621 odd choice, but which should be harmless for 3622 non-real-time spinlocks, due to their disabling 3623 of preemption. Note that non-realtime mutexes 3624 disable boosting. 3625 3626 locktorture.rt_boost_factor= [KNL] 3627 Number that determines how often and for how 3628 long priority boosting is exercised. This is 3629 scaled down by the number of writers, so that the 3630 number of boosts per unit time remains roughly 3631 constant as the number of writers increases. 3632 On the other hand, the duration of each boost 3633 increases with the number of writers. 3634 3635 locktorture.shuffle_interval= [KNL] 3636 Set task-shuffle interval (jiffies). Shuffling 3637 tasks allows some CPUs to go into dyntick-idle 3638 mode during the locktorture test. 3639 3640 locktorture.shutdown_secs= [KNL] 3641 Set time (s) after boot system shutdown. This 3642 is useful for hands-off automated testing. 3643 3644 locktorture.stat_interval= [KNL] 3645 Time (s) between statistics printk()s. 3646 3647 locktorture.stutter= [KNL] 3648 Time (s) to stutter testing, for example, 3649 specifying five seconds causes the test to run for 3650 five seconds, wait for five seconds, and so on. 3651 This tests the locking primitive's ability to 3652 transition abruptly to and from idle. 3653 3654 locktorture.torture_type= [KNL] 3655 Specify the locking implementation to test. 3656 3657 locktorture.verbose= [KNL] 3658 Enable additional printk() statements. 3659 3660 locktorture.writer_fifo= [KNL] 3661 Run the write-side locktorture kthreads at 3662 sched_set_fifo() real-time priority. 3663 3664 logibm.irq= [HW,MOUSE] Logitech Bus Mouse Driver 3665 Format: <irq> 3666 3667 loglevel= [KNL,EARLY] 3668 All Kernel Messages with a loglevel smaller than the 3669 console loglevel will be printed to the console. It can 3670 also be changed with klogd or other programs. The 3671 loglevels are defined as follows: 3672 3673 0 (KERN_EMERG) system is unusable 3674 1 (KERN_ALERT) action must be taken immediately 3675 2 (KERN_CRIT) critical conditions 3676 3 (KERN_ERR) error conditions 3677 4 (KERN_WARNING) warning conditions 3678 5 (KERN_NOTICE) normal but significant condition 3679 6 (KERN_INFO) informational 3680 7 (KERN_DEBUG) debug-level messages 3681 3682 log_buf_len=n[KMG] [KNL,EARLY] 3683 Sets the size of the printk ring buffer, in bytes. 3684 n must be a power of two and greater than the 3685 minimal size. The minimal size is defined by 3686 LOG_BUF_SHIFT kernel config parameter. There 3687 is also CONFIG_LOG_CPU_MAX_BUF_SHIFT config 3688 parameter that allows to increase the default size 3689 depending on the number of CPUs. See init/Kconfig 3690 for more details. 3691 3692 logo.nologo [FB] Disables display of the built-in Linux logo. 3693 This may be used to provide more screen space for 3694 kernel log messages and is useful when debugging 3695 kernel boot problems. 3696 3697 lp=0 [LP] Specify parallel ports to use, e.g, 3698 lp=port[,port...] lp=none,parport0 (lp0 not configured, lp1 uses 3699 lp=reset first parallel port). 'lp=0' disables the 3700 lp=auto printer driver. 'lp=reset' (which can be 3701 specified in addition to the ports) causes 3702 attached printers to be reset. Using 3703 lp=port1,port2,... specifies the parallel ports 3704 to associate lp devices with, starting with 3705 lp0. A port specification may be 'none' to skip 3706 that lp device, or a parport name such as 3707 'parport0'. Specifying 'lp=auto' instead of a 3708 port specification list means that device IDs 3709 from each port should be examined, to see if 3710 an IEEE 1284-compliant printer is attached; if 3711 so, the driver will manage that printer. 3712 See also header of drivers/char/lp.c. 3713 3714 lpj=n [KNL] 3715 Sets loops_per_jiffy to given constant, thus avoiding 3716 time-consuming boot-time autodetection (up to 250 ms per 3717 CPU). 0 enables autodetection (default). To determine 3718 the correct value for your kernel, boot with normal 3719 autodetection and see what value is printed. Note that 3720 on SMP systems the preset will be applied to all CPUs, 3721 which is likely to cause problems if your CPUs need 3722 significantly divergent settings. An incorrect value 3723 will cause delays in the kernel to be wrong, leading to 3724 unpredictable I/O errors and other breakage. Although 3725 unlikely, in the extreme case this might damage your 3726 hardware. 3727 3728 lsm.debug [SECURITY] Enable LSM initialization debugging output. 3729 3730 lsm=lsm1,...,lsmN 3731 [SECURITY] Choose order of LSM initialization. This 3732 overrides CONFIG_LSM, and the "security=" parameter. 3733 3734 machtype= [Loongson] Share the same kernel image file between 3735 different yeeloong laptops. 3736 Example: machtype=lemote-yeeloong-2f-7inch 3737 3738 maxcpus= [SMP,EARLY] Maximum number of processors that an SMP kernel 3739 will bring up during bootup. maxcpus=n : n >= 0 limits 3740 the kernel to bring up 'n' processors. Surely after 3741 bootup you can bring up the other plugged cpu by executing 3742 "echo 1 > /sys/devices/system/cpu/cpuX/online". So maxcpus 3743 only takes effect during system bootup. 3744 While n=0 is a special case, it is equivalent to "nosmp", 3745 which also disables the IO APIC. 3746 3747 max_loop= [LOOP] The number of loop block devices that get 3748 (loop.max_loop) unconditionally pre-created at init time. The default 3749 number is configured by BLK_DEV_LOOP_MIN_COUNT. Instead 3750 of statically allocating a predefined number, loop 3751 devices can be requested on-demand with the 3752 /dev/loop-control interface. 3753 3754 mce= [X86-{32,64}] 3755 3756 Please see Documentation/arch/x86/x86_64/machinecheck.rst for sysfs runtime tunables. 3757 3758 off 3759 disable machine check 3760 3761 no_cmci 3762 disable CMCI(Corrected Machine Check Interrupt) that 3763 Intel processor supports. Usually this disablement is 3764 not recommended, but it might be handy if your 3765 hardware is misbehaving. 3766 3767 Note that you'll get more problems without CMCI than 3768 with due to the shared banks, i.e. you might get 3769 duplicated error logs. 3770 3771 dont_log_ce 3772 don't make logs for corrected errors. All events 3773 reported as corrected are silently cleared by OS. This 3774 option will be useful if you have no interest in any 3775 of corrected errors. 3776 3777 ignore_ce 3778 disable features for corrected errors, e.g. 3779 polling timer and CMCI. All events reported as 3780 corrected are not cleared by OS and remained in its 3781 error banks. 3782 3783 Usually this disablement is not recommended, however 3784 if there is an agent checking/clearing corrected 3785 errors (e.g. BIOS or hardware monitoring 3786 applications), conflicting with OS's error handling, 3787 and you cannot deactivate the agent, then this option 3788 will be a help. 3789 3790 no_lmce 3791 do not opt-in to Local MCE delivery. Use legacy method 3792 to broadcast MCEs. 3793 3794 bootlog 3795 enable logging of machine checks left over from 3796 booting. Disabled by default on AMD Fam10h and older 3797 because some BIOS leave bogus ones. 3798 3799 If your BIOS doesn't do that it's a good idea to 3800 enable though to make sure you log even machine check 3801 events that result in a reboot. On Intel systems it is 3802 enabled by default. 3803 3804 nobootlog 3805 disable boot machine check logging. 3806 3807 print_all 3808 print all machine check logs to the console. 3809 3810 monarchtimeout (number) 3811 sets the time in us to wait for other CPUs on machine 3812 checks. 0 to disable. 3813 3814 bios_cmci_threshold 3815 don't overwrite the bios-set CMCI threshold. This boot 3816 option prevents Linux from overwriting the CMCI 3817 threshold set by the bios. Without this option, Linux 3818 always sets the CMCI threshold to 1. Enabling this may 3819 make memory predictive failure analysis less effective 3820 if the bios sets thresholds for memory errors since we 3821 will not see details for all errors. 3822 3823 recovery 3824 force-enable recoverable machine check code paths 3825 3826 Everything else is in sysfs now. 3827 3828 3829 md= [HW] RAID subsystems devices and level 3830 See Documentation/admin-guide/md.rst. 3831 3832 mds= [X86,INTEL,EARLY] 3833 Control mitigation for the Micro-architectural Data 3834 Sampling (MDS) vulnerability. 3835 3836 Certain CPUs are vulnerable to an exploit against CPU 3837 internal buffers which can forward information to a 3838 disclosure gadget under certain conditions. 3839 3840 In vulnerable processors, the speculatively 3841 forwarded data can be used in a cache side channel 3842 attack, to access data to which the attacker does 3843 not have direct access. 3844 3845 This parameter controls the MDS mitigation. The 3846 options are: 3847 3848 full - Enable MDS mitigation on vulnerable CPUs 3849 full,nosmt - Enable MDS mitigation and disable 3850 SMT on vulnerable CPUs 3851 off - Unconditionally disable MDS mitigation 3852 3853 On TAA-affected machines, mds=off can be prevented by 3854 an active TAA mitigation as both vulnerabilities are 3855 mitigated with the same mechanism so in order to disable 3856 this mitigation, you need to specify tsx_async_abort=off 3857 too. 3858 3859 Not specifying this option is equivalent to 3860 mds=full. 3861 3862 For details see: Documentation/admin-guide/hw-vuln/mds.rst 3863 3864 mem=nn[KMG] [HEXAGON,EARLY] Set the memory size. 3865 Must be specified, otherwise memory size will be 0. 3866 3867 mem=nn[KMG] [KNL,BOOT,EARLY] Force usage of a specific amount 3868 of memory Amount of memory to be used in cases 3869 as follows: 3870 3871 1 for test; 3872 2 when the kernel is not able to see the whole system memory; 3873 3 memory that lies after 'mem=' boundary is excluded from 3874 the hypervisor, then assigned to KVM guests. 3875 4 to limit the memory available for kdump kernel. 3876 3877 [ARC,MICROBLAZE] - the limit applies only to low memory, 3878 high memory is not affected. 3879 3880 [ARM64] - only limits memory covered by the linear 3881 mapping. The NOMAP regions are not affected. 3882 3883 [X86] Work as limiting max address. Use together 3884 with memmap= to avoid physical address space collisions. 3885 Without memmap= PCI devices could be placed at addresses 3886 belonging to unused RAM. 3887 3888 Note that this only takes effects during boot time since 3889 in above case 3, memory may need be hot added after boot 3890 if system memory of hypervisor is not sufficient. 3891 3892 mem=nn[KMG]@ss[KMG] 3893 [ARM,MIPS,EARLY] - override the memory layout 3894 reported by firmware. 3895 Define a memory region of size nn[KMG] starting at 3896 ss[KMG]. 3897 Multiple different regions can be specified with 3898 multiple mem= parameters on the command line. 3899 3900 mem=nopentium [BUGS=X86-32] Disable usage of 4MB pages for kernel 3901 memory. 3902 3903 memblock=debug [KNL,EARLY] Enable memblock debug messages. 3904 3905 memchunk=nn[KMG] 3906 [KNL,SH] Allow user to override the default size for 3907 per-device physically contiguous DMA buffers. 3908 3909 memhp_default_state=online/offline/online_kernel/online_movable 3910 [KNL] Set the initial state for the memory hotplug 3911 onlining policy. If not specified, the default value is 3912 set according to the 3913 CONFIG_MHP_DEFAULT_ONLINE_TYPE kernel config 3914 options. 3915 See Documentation/admin-guide/mm/memory-hotplug.rst. 3916 3917 memmap=exactmap [KNL,X86,EARLY] Enable setting of an exact 3918 E820 memory map, as specified by the user. 3919 Such memmap=exactmap lines can be constructed based on 3920 BIOS output or other requirements. See the memmap=nn@ss 3921 option description. 3922 3923 memmap=nn[KMG]@ss[KMG] 3924 [KNL, X86,MIPS,XTENSA,EARLY] Force usage of a specific region of memory. 3925 Region of memory to be used is from ss to ss+nn. 3926 If @ss[KMG] is omitted, it is equivalent to mem=nn[KMG], 3927 which limits max address to nn[KMG]. 3928 Multiple different regions can be specified, 3929 comma delimited. 3930 Example: 3931 memmap=100M@2G,100M#3G,1G!1024G 3932 3933 memmap=nn[KMG]#ss[KMG] 3934 [KNL,ACPI,EARLY] Mark specific memory as ACPI data. 3935 Region of memory to be marked is from ss to ss+nn. 3936 3937 memmap=nn[KMG]$ss[KMG] 3938 [KNL,ACPI,EARLY] Mark specific memory as reserved. 3939 Region of memory to be reserved is from ss to ss+nn. 3940 Example: Exclude memory from 0x18690000-0x1869ffff 3941 memmap=64K$0x18690000 3942 or 3943 memmap=0x10000$0x18690000 3944 Some bootloaders may need an escape character before '$', 3945 like Grub2, otherwise '$' and the following number 3946 will be eaten. 3947 3948 memmap=nn[KMG]!ss[KMG,EARLY] 3949 [KNL,X86] Mark specific memory as protected. 3950 Region of memory to be used, from ss to ss+nn. 3951 The memory region may be marked as e820 type 12 (0xc) 3952 and is NVDIMM or ADR memory. 3953 3954 memmap=<size>%<offset>-<oldtype>+<newtype> 3955 [KNL,ACPI,EARLY] Convert memory within the specified region 3956 from <oldtype> to <newtype>. If "-<oldtype>" is left 3957 out, the whole region will be marked as <newtype>, 3958 even if previously unavailable. If "+<newtype>" is left 3959 out, matching memory will be removed. Types are 3960 specified as e820 types, e.g., 1 = RAM, 2 = reserved, 3961 3 = ACPI, 12 = PRAM. 3962 3963 memory_corruption_check=0/1 [X86,EARLY] 3964 Some BIOSes seem to corrupt the first 64k of 3965 memory when doing things like suspend/resume. 3966 Setting this option will scan the memory 3967 looking for corruption. Enabling this will 3968 both detect corruption and prevent the kernel 3969 from using the memory being corrupted. 3970 However, it's intended as a diagnostic tool; if 3971 repeatable BIOS-originated corruption always 3972 affects the same memory, you can use memmap= 3973 to prevent the kernel from using that memory. 3974 3975 memory_corruption_check_size=size [X86,EARLY] 3976 By default it checks for corruption in the low 3977 64k, making this memory unavailable for normal 3978 use. Use this parameter to scan for 3979 corruption in more or less memory. 3980 3981 memory_corruption_check_period=seconds [X86,EARLY] 3982 By default it checks for corruption every 60 3983 seconds. Use this parameter to check at some 3984 other rate. 0 disables periodic checking. 3985 3986 memory_hotplug.memmap_on_memory 3987 [KNL,X86,ARM] Boolean flag to enable this feature. 3988 Format: {on | off (default)} 3989 When enabled, runtime hotplugged memory will 3990 allocate its internal metadata (struct pages, 3991 those vmemmap pages cannot be optimized even 3992 if hugetlb_free_vmemmap is enabled) from the 3993 hotadded memory which will allow to hotadd a 3994 lot of memory without requiring additional 3995 memory to do so. 3996 This feature is disabled by default because it 3997 has some implication on large (e.g. GB) 3998 allocations in some configurations (e.g. small 3999 memory blocks). 4000 The state of the flag can be read in 4001 /sys/module/memory_hotplug/parameters/memmap_on_memory. 4002 Note that even when enabled, there are a few cases where 4003 the feature is not effective. 4004 4005 mempool_debug [MM] 4006 Enable mempool debugging. This enables element 4007 poison checking when freeing elements back to the 4008 pool. Useful for debugging mempool corruption. 4009 4010 memtest= [KNL,X86,ARM,M68K,PPC,RISCV,EARLY] Enable memtest 4011 Format: <integer> 4012 default : 0 <disable> 4013 Specifies the number of memtest passes to be 4014 performed. Each pass selects another test 4015 pattern from a given set of patterns. Memtest 4016 fills the memory with this pattern, validates 4017 memory contents and reserves bad memory 4018 regions that are detected. 4019 4020 mem_encrypt= [X86-64] AMD Secure Memory Encryption (SME) control 4021 Valid arguments: on, off 4022 Default: off 4023 mem_encrypt=on: Activate SME 4024 mem_encrypt=off: Do not activate SME 4025 4026 Refer to Documentation/virt/kvm/x86/amd-memory-encryption.rst 4027 for details on when memory encryption can be activated. 4028 4029 mem_sleep_default= [SUSPEND] Default system suspend mode: 4030 s2idle - Suspend-To-Idle 4031 shallow - Power-On Suspend or equivalent (if supported) 4032 deep - Suspend-To-RAM or equivalent (if supported) 4033 See Documentation/admin-guide/pm/sleep-states.rst. 4034 4035 mfgptfix [X86-32] Fix MFGPT timers on AMD Geode platforms when 4036 the BIOS has incorrectly applied a workaround. TinyBIOS 4037 version 0.98 is known to be affected, 0.99 fixes the 4038 problem by letting the user disable the workaround. 4039 4040 mga= [HW,DRM] 4041 4042 microcode= [X86] Control the behavior of the microcode loader. 4043 Available options, comma separated: 4044 4045 base_rev=X - with <X> with format: <u32> 4046 Set the base microcode revision of each thread when in 4047 debug mode. 4048 4049 dis_ucode_ldr: disable the microcode loader 4050 4051 force_minrev: 4052 Enable or disable the microcode minimal revision 4053 enforcement for the runtime microcode loader. 4054 4055 mitigations= 4056 [X86,PPC,S390,ARM64,EARLY] Control optional mitigations for 4057 CPU vulnerabilities. This is a set of curated, 4058 arch-independent options, each of which is an 4059 aggregation of existing arch-specific options. 4060 4061 Note, "mitigations" is supported if and only if the 4062 kernel was built with CPU_MITIGATIONS=y. 4063 4064 off 4065 Disable all optional CPU mitigations. This 4066 improves system performance, but it may also 4067 expose users to several CPU vulnerabilities. 4068 Equivalent to: if nokaslr then kpti=0 [ARM64] 4069 gather_data_sampling=off [X86] 4070 indirect_target_selection=off [X86] 4071 kvm.nx_huge_pages=off [X86] 4072 l1tf=off [X86] 4073 mds=off [X86] 4074 mmio_stale_data=off [X86] 4075 no_entry_flush [PPC] 4076 no_uaccess_flush [PPC] 4077 nobp=0 [S390] 4078 nopti [X86,PPC] 4079 nospectre_bhb [ARM64] 4080 nospectre_v1 [X86,PPC] 4081 nospectre_v2 [X86,PPC,S390,ARM64] 4082 reg_file_data_sampling=off [X86] 4083 retbleed=off [X86] 4084 spec_rstack_overflow=off [X86] 4085 spec_store_bypass_disable=off [X86,PPC] 4086 spectre_bhi=off [X86] 4087 spectre_v2_user=off [X86] 4088 srbds=off [X86,INTEL] 4089 ssbd=force-off [ARM64] 4090 tsa=off [X86,AMD] 4091 tsx_async_abort=off [X86] 4092 vmscape=off [X86] 4093 4094 Exceptions: 4095 This does not have any effect on 4096 kvm.nx_huge_pages when 4097 kvm.nx_huge_pages=force. 4098 4099 auto (default) 4100 Mitigate all CPU vulnerabilities, but leave SMT 4101 enabled, even if it's vulnerable. This is for 4102 users who don't want to be surprised by SMT 4103 getting disabled across kernel upgrades, or who 4104 have other ways of avoiding SMT-based attacks. 4105 Equivalent to: (default behavior) 4106 4107 auto,nosmt 4108 Mitigate all CPU vulnerabilities, disabling SMT 4109 if needed. This is for users who always want to 4110 be fully mitigated, even if it means losing SMT. 4111 Equivalent to: l1tf=flush,nosmt [X86] 4112 mds=full,nosmt [X86] 4113 tsx_async_abort=full,nosmt [X86] 4114 mmio_stale_data=full,nosmt [X86] 4115 retbleed=auto,nosmt [X86] 4116 4117 [X86] After one of the above options, additionally 4118 supports attack-vector based controls as documented in 4119 Documentation/admin-guide/hw-vuln/attack_vector_controls.rst 4120 4121 mminit_loglevel= 4122 [KNL,EARLY] When CONFIG_DEBUG_MEMORY_INIT is set, this 4123 parameter allows control of the logging verbosity for 4124 the additional memory initialisation checks. A value 4125 of 0 disables mminit logging and a level of 4 will 4126 log everything. Information is printed at KERN_DEBUG 4127 so loglevel=8 may also need to be specified. 4128 4129 mmio_stale_data= 4130 [X86,INTEL,EARLY] Control mitigation for the Processor 4131 MMIO Stale Data vulnerabilities. 4132 4133 Processor MMIO Stale Data is a class of 4134 vulnerabilities that may expose data after an MMIO 4135 operation. Exposed data could originate or end in 4136 the same CPU buffers as affected by MDS and TAA. 4137 Therefore, similar to MDS and TAA, the mitigation 4138 is to clear the affected CPU buffers. 4139 4140 This parameter controls the mitigation. The 4141 options are: 4142 4143 full - Enable mitigation on vulnerable CPUs 4144 4145 full,nosmt - Enable mitigation and disable SMT on 4146 vulnerable CPUs. 4147 4148 off - Unconditionally disable mitigation 4149 4150 On MDS or TAA affected machines, 4151 mmio_stale_data=off can be prevented by an active 4152 MDS or TAA mitigation as these vulnerabilities are 4153 mitigated with the same mechanism so in order to 4154 disable this mitigation, you need to specify 4155 mds=off and tsx_async_abort=off too. 4156 4157 Not specifying this option is equivalent to 4158 mmio_stale_data=full. 4159 4160 For details see: 4161 Documentation/admin-guide/hw-vuln/processor_mmio_stale_data.rst 4162 4163 <module>.async_probe[=<bool>] [KNL] 4164 If no <bool> value is specified or if the value 4165 specified is not a valid <bool>, enable asynchronous 4166 probe on this module. Otherwise, enable/disable 4167 asynchronous probe on this module as indicated by the 4168 <bool> value. See also: module.async_probe 4169 4170 module.async_probe=<bool> 4171 [KNL] When set to true, modules will use async probing 4172 by default. To enable/disable async probing for a 4173 specific module, use the module specific control that 4174 is documented under <module>.async_probe. When both 4175 module.async_probe and <module>.async_probe are 4176 specified, <module>.async_probe takes precedence for 4177 the specific module. 4178 4179 module.enable_dups_trace 4180 [KNL] When CONFIG_MODULE_DEBUG_AUTOLOAD_DUPS is set, 4181 this means that duplicate request_module() calls will 4182 trigger a WARN_ON() instead of a pr_warn(). Note that 4183 if MODULE_DEBUG_AUTOLOAD_DUPS_TRACE is set, WARN_ON()s 4184 will always be issued and this option does nothing. 4185 module.sig_enforce 4186 [KNL] When CONFIG_MODULE_SIG is set, this means that 4187 modules without (valid) signatures will fail to load. 4188 Note that if CONFIG_MODULE_SIG_FORCE is set, that 4189 is always true, so this option does nothing. 4190 4191 module_blacklist= [KNL] Do not load a comma-separated list of 4192 modules. Useful for debugging problem modules. 4193 4194 mousedev.tap_time= 4195 [MOUSE] Maximum time between finger touching and 4196 leaving touchpad surface for touch to be considered 4197 a tap and be reported as a left button click (for 4198 touchpads working in absolute mode only). 4199 Format: <msecs> 4200 mousedev.xres= [MOUSE] Horizontal screen resolution, used for devices 4201 reporting absolute coordinates, such as tablets 4202 mousedev.yres= [MOUSE] Vertical screen resolution, used for devices 4203 reporting absolute coordinates, such as tablets 4204 4205 movablecore= [KNL,X86,PPC,EARLY] 4206 Format: nn[KMGTPE] | nn% 4207 This parameter is the complement to kernelcore=, it 4208 specifies the amount of memory used for migratable 4209 allocations. If both kernelcore and movablecore is 4210 specified, then kernelcore will be at *least* the 4211 specified value but may be more. If movablecore on its 4212 own is specified, the administrator must be careful 4213 that the amount of memory usable for all allocations 4214 is not too small. 4215 4216 movable_node [KNL,EARLY] Boot-time switch to make hotplugable memory 4217 NUMA nodes to be movable. This means that the memory 4218 of such nodes will be usable only for movable 4219 allocations which rules out almost all kernel 4220 allocations. Use with caution! 4221 4222 MTD_Partition= [MTD] 4223 Format: <name>,<region-number>,<size>,<offset> 4224 4225 MTD_Region= [MTD] Format: 4226 <name>,<region-number>[,<base>,<size>,<buswidth>,<altbuswidth>] 4227 4228 mtdparts= [MTD] 4229 See drivers/mtd/parsers/cmdlinepart.c 4230 4231 mtouchusb.raw_coordinates= 4232 [HW] Make the MicroTouch USB driver use raw coordinates 4233 ('y', default) or cooked coordinates ('n') 4234 4235 mtrr=debug [X86,EARLY] 4236 Enable printing debug information related to MTRR 4237 registers at boot time. 4238 4239 mtrr_chunk_size=nn[KMG,X86,EARLY] 4240 used for mtrr cleanup. It is largest continuous chunk 4241 that could hold holes aka. UC entries. 4242 4243 mtrr_gran_size=nn[KMG,X86,EARLY] 4244 Used for mtrr cleanup. It is granularity of mtrr block. 4245 Default is 1. 4246 Large value could prevent small alignment from 4247 using up MTRRs. 4248 4249 mtrr_spare_reg_nr=n [X86,EARLY] 4250 Format: <integer> 4251 Range: 0,7 : spare reg number 4252 Default : 1 4253 Used for mtrr cleanup. It is spare mtrr entries number. 4254 Set to 2 or more if your graphical card needs more. 4255 4256 multitce=off [PPC] This parameter disables the use of the pSeries 4257 firmware feature for updating multiple TCE entries 4258 at a time. 4259 4260 n2= [NET] SDL Inc. RISCom/N2 synchronous serial card 4261 4262 netdev= [NET] NE2000 ISA network devices parameters 4263 Format: <irq>,<io>,<mem_start>,<mem_end>,<name> 4264 4265 netpoll.carrier_timeout= 4266 [NET] Specifies amount of time (in seconds) that 4267 netpoll should wait for a carrier. By default netpoll 4268 waits 4 seconds. 4269 4270 nf_conntrack.acct= 4271 [NETFILTER] Enable connection tracking flow accounting 4272 0 to disable accounting 4273 1 to enable accounting 4274 Default value is 0. 4275 4276 nfs.cache_getent= 4277 [NFS] sets the pathname to the program which is used 4278 to update the NFS client cache entries. 4279 4280 nfs.cache_getent_timeout= 4281 [NFS] sets the timeout after which an attempt to 4282 update a cache entry is deemed to have failed. 4283 4284 nfs.callback_nr_threads= 4285 [NFSv4] set the total number of threads that the 4286 NFS client will assign to service NFSv4 callback 4287 requests. 4288 4289 nfs.callback_tcpport= 4290 [NFS] set the TCP port on which the NFSv4 callback 4291 channel should listen. 4292 4293 nfs.delay_retrans= 4294 [NFS] specifies the number of times the NFSv4 client 4295 retries the request before returning an EAGAIN error, 4296 after a reply of NFS4ERR_DELAY from the server. 4297 Only applies if the softerr mount option is enabled, 4298 and the specified value is >= 0. 4299 4300 nfs.idmap_cache_timeout= 4301 [NFS] set the maximum lifetime for idmapper cache 4302 entries. 4303 4304 nfs.max_session_cb_slots= 4305 [NFSv4.1] Sets the maximum number of session 4306 slots the client will assign to the callback 4307 channel. This determines the maximum number of 4308 callbacks the client will process in parallel for 4309 a particular server. 4310 4311 nfs.max_session_slots= 4312 [NFSv4.1] Sets the maximum number of session slots 4313 the client will attempt to negotiate with the server. 4314 This limits the number of simultaneous RPC requests 4315 that the client can send to the NFSv4.1 server. 4316 Note that there is little point in setting this 4317 value higher than the max_tcp_slot_table_limit. 4318 4319 nfs.nfs4_disable_idmapping= 4320 [NFSv4] When set to the default of '1', this option 4321 ensures that both the RPC level authentication 4322 scheme and the NFS level operations agree to use 4323 numeric uids/gids if the mount is using the 4324 'sec=sys' security flavour. In effect it is 4325 disabling idmapping, which can make migration from 4326 legacy NFSv2/v3 systems to NFSv4 easier. 4327 Servers that do not support this mode of operation 4328 will be autodetected by the client, and it will fall 4329 back to using the idmapper. 4330 To turn off this behaviour, set the value to '0'. 4331 4332 nfs.nfs4_unique_id= 4333 [NFS4] Specify an additional fixed unique ident- 4334 ification string that NFSv4 clients can insert into 4335 their nfs_client_id4 string. This is typically a 4336 UUID that is generated at system install time. 4337 4338 nfs.recover_lost_locks= 4339 [NFSv4] Attempt to recover locks that were lost due 4340 to a lease timeout on the server. Please note that 4341 doing this risks data corruption, since there are 4342 no guarantees that the file will remain unchanged 4343 after the locks are lost. 4344 If you want to enable the kernel legacy behaviour of 4345 attempting to recover these locks, then set this 4346 parameter to '1'. 4347 The default parameter value of '0' causes the kernel 4348 not to attempt recovery of lost locks. 4349 4350 nfs.send_implementation_id= 4351 [NFSv4.1] Send client implementation identification 4352 information in exchange_id requests. 4353 If zero, no implementation identification information 4354 will be sent. 4355 The default is to send the implementation identification 4356 information. 4357 4358 nfs4.layoutstats_timer= 4359 [NFSv4.2] Change the rate at which the kernel sends 4360 layoutstats to the pNFS metadata server. 4361 4362 Setting this to value to 0 causes the kernel to use 4363 whatever value is the default set by the layout 4364 driver. A non-zero value sets the minimum interval 4365 in seconds between layoutstats transmissions. 4366 4367 nfsd.inter_copy_offload_enable= 4368 [NFSv4.2] When set to 1, the server will support 4369 server-to-server copies for which this server is 4370 the destination of the copy. 4371 4372 nfsd.nfs4_disable_idmapping= 4373 [NFSv4] When set to the default of '1', the NFSv4 4374 server will return only numeric uids and gids to 4375 clients using auth_sys, and will accept numeric uids 4376 and gids from such clients. This is intended to ease 4377 migration from NFSv2/v3. 4378 4379 nfsd.nfsd4_ssc_umount_timeout= 4380 [NFSv4.2] When used as the destination of a 4381 server-to-server copy, knfsd temporarily mounts 4382 the source server. It caches the mount in case 4383 it will be needed again, and discards it if not 4384 used for the number of milliseconds specified by 4385 this parameter. 4386 4387 nfsaddrs= [NFS] Deprecated. Use ip= instead. 4388 See Documentation/admin-guide/nfs/nfsroot.rst. 4389 4390 nfsroot= [NFS] nfs root filesystem for disk-less boxes. 4391 See Documentation/admin-guide/nfs/nfsroot.rst. 4392 4393 nfsrootdebug [NFS] enable nfsroot debugging messages. 4394 See Documentation/admin-guide/nfs/nfsroot.rst. 4395 4396 nmi_backtrace.backtrace_idle [KNL] 4397 Dump stacks even of idle CPUs in response to an 4398 NMI stack-backtrace request. 4399 4400 nmi_debug= [KNL,SH] Specify one or more actions to take 4401 when a NMI is triggered. 4402 Format: [state][,regs][,debounce][,die] 4403 4404 nmi_watchdog= [KNL,BUGS=X86] Debugging features for SMP kernels 4405 Format: [panic,][nopanic,][rNNN,][num] 4406 Valid num: 0 or 1 4407 0 - turn hardlockup detector in nmi_watchdog off 4408 1 - turn hardlockup detector in nmi_watchdog on 4409 rNNN - configure the watchdog with raw perf event 0xNNN 4410 4411 When panic is specified, panic when an NMI watchdog 4412 timeout occurs (or 'nopanic' to not panic on an NMI 4413 watchdog, if CONFIG_BOOTPARAM_HARDLOCKUP_PANIC is set) 4414 To disable both hard and soft lockup detectors, 4415 please see 'nowatchdog'. 4416 This is useful when you use a panic=... timeout and 4417 need the box quickly up again. 4418 4419 These settings can be accessed at runtime via 4420 the nmi_watchdog and hardlockup_panic sysctls. 4421 4422 no4lvl [RISCV,EARLY] Disable 4-level and 5-level paging modes. 4423 Forces kernel to use 3-level paging instead. 4424 4425 no5lvl [X86-64,RISCV,EARLY] Disable 5-level paging mode. Forces 4426 kernel to use 4-level paging instead. 4427 4428 noalign [KNL,ARM] 4429 4430 noapic [SMP,APIC,EARLY] Tells the kernel to not make use of any 4431 IOAPICs that may be present in the system. 4432 4433 noapictimer [APIC,X86] Don't set up the APIC timer 4434 4435 noautogroup Disable scheduler automatic task group creation. 4436 4437 nocache [ARM,EARLY] 4438 4439 no_console_suspend 4440 [HW] Never suspend the console 4441 Disable suspending of consoles during suspend and 4442 hibernate operations. Once disabled, debugging 4443 messages can reach various consoles while the rest 4444 of the system is being put to sleep (ie, while 4445 debugging driver suspend/resume hooks). This may 4446 not work reliably with all consoles, but is known 4447 to work with serial and VGA consoles. 4448 To facilitate more flexible debugging, we also add 4449 console_suspend, a printk module parameter to control 4450 it. Users could use console_suspend (usually 4451 /sys/module/printk/parameters/console_suspend) to 4452 turn on/off it dynamically. 4453 4454 no_debug_objects 4455 [KNL,EARLY] Disable object debugging 4456 4457 nodsp [SH] Disable hardware DSP at boot time. 4458 4459 noefi [EFI,EARLY] Disable EFI runtime services support. 4460 4461 no_entry_flush [PPC,EARLY] Don't flush the L1-D cache when entering the kernel. 4462 4463 noexec32 [X86-64] 4464 This affects only 32-bit executables. 4465 noexec32=on: enable non-executable mappings (default) 4466 read doesn't imply executable mappings 4467 noexec32=off: disable non-executable mappings 4468 read implies executable mappings 4469 4470 no_file_caps Tells the kernel not to honor file capabilities. The 4471 only way then for a file to be executed with privilege 4472 is to be setuid root or executed by root. 4473 4474 nofpu [MIPS,SH] Disable hardware FPU at boot time. 4475 4476 nofsgsbase [X86] Disables FSGSBASE instructions. 4477 4478 nofxsr [BUGS=X86-32] Disables x86 floating point extended 4479 register save and restore. The kernel will only save 4480 legacy floating-point registers on task switch. 4481 4482 nogbpages [X86] Do not use GB pages for kernel direct mappings. 4483 4484 no_hash_pointers 4485 [KNL,EARLY] 4486 Alias for "hash_pointers=never". 4487 4488 nohibernate [HIBERNATION] Disable hibernation and resume. 4489 4490 nohlt [ARM,ARM64,MICROBLAZE,MIPS,PPC,RISCV,SH] Forces the kernel to 4491 busy wait in do_idle() and not use the arch_cpu_idle() 4492 implementation; requires CONFIG_GENERIC_IDLE_POLL_SETUP 4493 to be effective. This is useful on platforms where the 4494 sleep(SH) or wfi(ARM,ARM64) instructions do not work 4495 correctly or when doing power measurements to evaluate 4496 the impact of the sleep instructions. This is also 4497 useful when using JTAG debugger. 4498 4499 nohpet [X86] Don't use the HPET timer. 4500 4501 nohugeiomap [KNL,X86,PPC,ARM64,EARLY] Disable kernel huge I/O mappings. 4502 4503 nohugevmalloc [KNL,X86,PPC,ARM64,EARLY] Disable kernel huge vmalloc mappings. 4504 4505 nohz= [KNL] Boottime enable/disable dynamic ticks 4506 Valid arguments: on, off 4507 Default: on 4508 4509 nohz_full= [KNL,BOOT,SMP,ISOL] 4510 The argument is a cpu list, as described above. 4511 In kernels built with CONFIG_NO_HZ_FULL=y, set 4512 the specified list of CPUs whose tick will be stopped 4513 whenever possible. The boot CPU will be forced outside 4514 the range to maintain the timekeeping. Any CPUs 4515 in this list will have their RCU callbacks offloaded, 4516 just as if they had also been called out in the 4517 rcu_nocbs= boot parameter. 4518 4519 Note that this argument takes precedence over 4520 the CONFIG_RCU_NOCB_CPU_DEFAULT_ALL option. 4521 4522 noinitrd [Deprecated,RAM] Tells the kernel not to load any configured 4523 initial RAM disk. Currently this parameter applies to 4524 initrd only, not to initramfs. But it applies to both 4525 in EFI mode. 4526 4527 nointremap [X86-64,Intel-IOMMU,EARLY] Do not enable interrupt 4528 remapping. 4529 [Deprecated - use intremap=off] 4530 4531 noinvpcid [X86,EARLY] Disable the INVPCID cpu feature. 4532 4533 noiotrap [SH] Disables trapped I/O port accesses. 4534 4535 noirqdebug [X86-32] Disables the code which attempts to detect and 4536 disable unhandled interrupt sources. 4537 4538 noisapnp [ISAPNP] Disables ISA PnP code. 4539 4540 nokaslr [KNL,EARLY] 4541 When CONFIG_RANDOMIZE_BASE is set, this disables 4542 kernel and module base offset ASLR (Address Space 4543 Layout Randomization). 4544 4545 no-kvmapf [X86,KVM,EARLY] Disable paravirtualized asynchronous page 4546 fault handling. 4547 4548 no-kvmclock [X86,KVM,EARLY] Disable paravirtualized KVM clock driver 4549 4550 nolapic [X86-32,APIC,EARLY] Do not enable or use the local APIC. 4551 4552 nolapic_timer [X86-32,APIC,EARLY] Do not use the local APIC timer. 4553 4554 nomce [X86-32] Disable Machine Check Exception 4555 4556 nomodeset Disable kernel modesetting. Most systems' firmware 4557 sets up a display mode and provides framebuffer memory 4558 for output. With nomodeset, DRM and fbdev drivers will 4559 not load if they could possibly displace the pre- 4560 initialized output. Only the system framebuffer will 4561 be available for use. The respective drivers will not 4562 perform display-mode changes or accelerated rendering. 4563 4564 Useful as error fallback, or for testing and debugging. 4565 4566 nomodule Disable module load 4567 4568 nonmi_ipi [X86] Disable using NMI IPIs during panic/reboot to 4569 shutdown the other cpus. Instead use the REBOOT_VECTOR 4570 irq. 4571 4572 nopat [X86,EARLY] Disable PAT (page attribute table extension of 4573 pagetables) support. 4574 4575 nopcid [X86-64,EARLY] Disable the PCID cpu feature. 4576 4577 nopku [X86] Disable Memory Protection Keys CPU feature found 4578 in some Intel CPUs. 4579 4580 nopti [X86-64,EARLY] 4581 Equivalent to pti=off 4582 4583 nopv= [X86,XEN,KVM,HYPER_V,VMWARE,EARLY] 4584 Disables the PV optimizations forcing the guest to run 4585 as generic guest with no PV drivers. Currently support 4586 XEN HVM, KVM, HYPER_V and VMWARE guest. 4587 4588 nopvspin [X86,XEN,KVM,EARLY] 4589 Disables the qspinlock slow path using PV optimizations 4590 which allow the hypervisor to 'idle' the guest on lock 4591 contention. 4592 4593 norandmaps Don't use address space randomization. Equivalent to 4594 echo 0 > /proc/sys/kernel/randomize_va_space 4595 4596 noresume [SWSUSP] Disables resume and restores original swap 4597 space. 4598 4599 no-scroll [VGA] Disables scrollback. 4600 This is required for the Braillex ib80-piezo Braille 4601 reader made by F.H. Papenmeier (Germany). 4602 4603 nosgx [X86-64,SGX,EARLY] Disables Intel SGX kernel support. 4604 4605 nosmap [PPC,EARLY] 4606 Disable SMAP (Supervisor Mode Access Prevention) 4607 even if it is supported by processor. 4608 4609 nosmep [PPC64s,EARLY] 4610 Disable SMEP (Supervisor Mode Execution Prevention) 4611 even if it is supported by processor. 4612 4613 nosmp [SMP,EARLY] Tells an SMP kernel to act as a UP kernel, 4614 and disable the IO APIC. legacy for "maxcpus=0". 4615 4616 nosmt [KNL,MIPS,PPC,EARLY] Disable symmetric multithreading (SMT). 4617 Equivalent to smt=1. 4618 4619 [KNL,LOONGARCH,X86,ARM64,PPC,S390] Disable symmetric multithreading (SMT). 4620 nosmt=force: Force disable SMT, cannot be undone 4621 via the sysfs control file. 4622 4623 nosoftlockup [KNL] Disable the soft-lockup detector. 4624 4625 nospec_store_bypass_disable 4626 [HW,EARLY] Disable all mitigations for the Speculative 4627 Store Bypass vulnerability 4628 4629 nospectre_bhb [ARM64,EARLY] Disable all mitigations for Spectre-BHB (branch 4630 history injection) vulnerability. System may allow data leaks 4631 with this option. 4632 4633 nospectre_v1 [X86,PPC,EARLY] Disable mitigations for Spectre Variant 1 4634 (bounds check bypass). With this option data leaks are 4635 possible in the system. 4636 4637 nospectre_v2 [X86,PPC_E500,ARM64,EARLY] Disable all mitigations 4638 for the Spectre variant 2 (indirect branch 4639 prediction) vulnerability. System may allow data 4640 leaks with this option. 4641 4642 no-steal-acc [X86,PV_OPS,ARM64,PPC/PSERIES,RISCV,LOONGARCH,EARLY] 4643 Disable paravirtualized steal time accounting. steal time 4644 is computed, but won't influence scheduler behaviour 4645 4646 nosync [HW,M68K] Disables sync negotiation for all devices. 4647 4648 no_timer_check [X86,APIC] Disables the code which tests for broken 4649 timer IRQ sources, i.e., the IO-APIC timer. This can 4650 work around problems with incorrect timer 4651 initialization on some boards. 4652 4653 no_uaccess_flush 4654 [PPC,EARLY] Don't flush the L1-D cache after accessing user data. 4655 4656 novmcoredd [KNL,KDUMP] 4657 Disable device dump. Device dump allows drivers to 4658 append dump data to vmcore so you can collect driver 4659 specified debug info. Drivers can append the data 4660 without any limit and this data is stored in memory, 4661 so this may cause significant memory stress. Disabling 4662 device dump can help save memory but the driver debug 4663 data will be no longer available. This parameter 4664 is only available when CONFIG_PROC_VMCORE_DEVICE_DUMP 4665 is set. 4666 4667 no-vmw-sched-clock 4668 [X86,PV_OPS,EARLY] Disable paravirtualized VMware 4669 scheduler clock and use the default one. 4670 4671 nowatchdog [KNL] Disable both lockup detectors, i.e. 4672 soft-lockup and NMI watchdog (hard-lockup). 4673 4674 nowb [ARM,EARLY] 4675 4676 nox2apic [X86-64,APIC,EARLY] Do not enable x2APIC mode. 4677 4678 NOTE: this parameter will be ignored on systems with the 4679 LEGACY_XAPIC_DISABLED bit set in the 4680 IA32_XAPIC_DISABLE_STATUS MSR. 4681 4682 noxsave [BUGS=X86] Disables x86 extended register state save 4683 and restore using xsave. The kernel will fallback to 4684 enabling legacy floating-point and sse state. 4685 4686 noxsaveopt [X86] Disables xsaveopt used in saving x86 extended 4687 register states. The kernel will fall back to use 4688 xsave to save the states. By using this parameter, 4689 performance of saving the states is degraded because 4690 xsave doesn't support modified optimization while 4691 xsaveopt supports it on xsaveopt enabled systems. 4692 4693 noxsaves [X86] Disables xsaves and xrstors used in saving and 4694 restoring x86 extended register state in compacted 4695 form of xsave area. The kernel will fall back to use 4696 xsaveopt and xrstor to save and restore the states 4697 in standard form of xsave area. By using this 4698 parameter, xsave area per process might occupy more 4699 memory on xsaves enabled systems. 4700 4701 nr_cpus= [SMP,EARLY] Maximum number of processors that an SMP kernel 4702 could support. nr_cpus=n : n >= 1 limits the kernel to 4703 support 'n' processors. It could be larger than the 4704 number of already plugged CPU during bootup, later in 4705 runtime you can physically add extra cpu until it reaches 4706 n. So during boot up some boot time memory for per-cpu 4707 variables need be pre-allocated for later physical cpu 4708 hot plugging. 4709 4710 nr_uarts= [SERIAL] maximum number of UARTs to be registered. 4711 4712 numa=off [KNL, ARM64, PPC, RISCV, SPARC, X86, EARLY] 4713 Disable NUMA, Only set up a single NUMA node 4714 spanning all memory. 4715 4716 numa=fake=<size>[MG] 4717 [KNL, ARM64, RISCV, X86, EARLY] 4718 If given as a memory unit, fills all system RAM with 4719 nodes of size interleaved over physical nodes. 4720 4721 numa=fake=<N> 4722 [KNL, ARM64, RISCV, X86, EARLY] 4723 If given as an integer, fills all system RAM with N 4724 fake nodes interleaved over physical nodes. 4725 4726 numa=fake=<N>U 4727 [KNL, ARM64, RISCV, X86, EARLY] 4728 If given as an integer followed by 'U', it will 4729 divide each physical node into N emulated nodes. 4730 4731 numa=noacpi [X86] Don't parse the SRAT table for NUMA setup 4732 4733 numa=nohmat [X86] Don't parse the HMAT table for NUMA setup, or 4734 soft-reserved memory partitioning. 4735 4736 numa_balancing= [KNL,ARM64,PPC,RISCV,S390,X86] Enable or disable automatic 4737 NUMA balancing. 4738 Allowed values are enable and disable 4739 4740 numa_zonelist_order= [KNL, BOOT] Select zonelist order for NUMA. 4741 'node', 'default' can be specified 4742 This can be set from sysctl after boot. 4743 See Documentation/admin-guide/sysctl/vm.rst for details. 4744 4745 nvme.quirks= [NVME] A list of quirk entries to augment the built-in 4746 nvme quirk list. List entries are separated by a 4747 '-' character. 4748 Each entry has the form VendorID:ProductID:quirk_names. 4749 The IDs are 4-digits hex numbers and quirk_names is a 4750 list of quirk names separated by commas. A quirk name 4751 can be prefixed by '^', meaning that the specified 4752 quirk must be disabled. 4753 4754 Example: 4755 nvme.quirks=7710:2267:bogus_nid,^identify_cns-9900:7711:broken_msi 4756 4757 ohci1394_dma=early [HW,EARLY] enable debugging via the ohci1394 driver. 4758 See Documentation/core-api/debugging-via-ohci1394.rst for more 4759 info. 4760 4761 olpc_ec_timeout= [OLPC] ms delay when issuing EC commands 4762 Rather than timing out after 20 ms if an EC 4763 command is not properly ACKed, override the length 4764 of the timeout. We have interrupts disabled while 4765 waiting for the ACK, so if this is set too high 4766 interrupts *may* be lost! 4767 4768 onenand.bdry= [HW,MTD] Flex-OneNAND Boundary Configuration 4769 4770 Format: [die0_boundary][,die0_lock][,die1_boundary][,die1_lock] 4771 4772 boundary - index of last SLC block on Flex-OneNAND. 4773 The remaining blocks are configured as MLC blocks. 4774 lock - Configure if Flex-OneNAND boundary should be locked. 4775 Once locked, the boundary cannot be changed. 4776 1 indicates lock status, 0 indicates unlock status. 4777 4778 oops=panic [KNL,EARLY] 4779 Always panic on oopses. Default is to just kill the 4780 process, but there is a small probability of 4781 deadlocking the machine. 4782 This will also cause panics on machine check exceptions. 4783 Useful together with panic=30 to trigger a reboot. 4784 4785 page_alloc.shuffle= 4786 [KNL] Boolean flag to control whether the page allocator 4787 should randomize its free lists. This parameter can be 4788 used to enable/disable page randomization. The state of 4789 the flag can be read from sysfs at: 4790 /sys/module/page_alloc/parameters/shuffle. 4791 This parameter is only available if CONFIG_SHUFFLE_PAGE_ALLOCATOR=y. 4792 4793 page_owner= [KNL,EARLY] Boot-time page_owner enabling option. 4794 Storage of the information about who allocated 4795 each page is disabled in default. With this switch, 4796 we can turn it on. 4797 on: enable the feature 4798 4799 page_poison= [KNL,EARLY] Boot-time parameter changing the state of 4800 poisoning on the buddy allocator, available with 4801 CONFIG_PAGE_POISONING=y. 4802 off: turn off poisoning (default) 4803 on: turn on poisoning 4804 4805 page_reporting.page_reporting_order= 4806 [KNL] Minimal page reporting order 4807 Format: <integer> 4808 Adjust the minimal page reporting order. The page 4809 reporting is disabled when it exceeds MAX_PAGE_ORDER. 4810 4811 page_reporting.page_reporting_delay_ms= 4812 [KNL] Free page reporting delay in milliseconds 4813 Format: <unsigned integer> 4814 Adjust the delay in milliseconds between free page 4815 reporting intervals. Default is 2000 (2 seconds). 4816 4817 panic= [KNL] Kernel behaviour on panic: delay <timeout> 4818 timeout > 0: seconds before rebooting 4819 timeout = 0: wait forever 4820 timeout < 0: reboot immediately 4821 Format: <timeout> 4822 4823 panic_on_taint= [KNL,EARLY] 4824 Bitmask for conditionally calling panic() in add_taint() 4825 Format: <hex>[,nousertaint] 4826 Hexadecimal bitmask representing the set of TAINT flags 4827 that will cause the kernel to panic when add_taint() is 4828 called with any of the flags in this set. 4829 The optional switch "nousertaint" can be utilized to 4830 prevent userspace forced crashes by writing to sysctl 4831 /proc/sys/kernel/tainted any flagset matching with the 4832 bitmask set on panic_on_taint. 4833 See Documentation/admin-guide/tainted-kernels.rst for 4834 extra details on the taint flags that users can pick 4835 to compose the bitmask to assign to panic_on_taint. 4836 4837 panic_on_warn=1 panic() instead of WARN(). Useful to cause kdump 4838 on a WARN(). 4839 4840 panic_force_cpu= 4841 [KNL,SMP] Force panic handling to execute on a specific CPU. 4842 Format: <cpu number> 4843 Some platforms require panic handling to occur on a 4844 specific CPU for the crash kernel to function correctly. 4845 This can be due to firmware limitations, interrupt routing 4846 constraints, or platform-specific requirements where only 4847 a particular CPU can safely enter the crash kernel. 4848 When set, panic() will redirect execution to the specified 4849 CPU before proceeding with the normal panic and kexec flow. 4850 If the target CPU is offline or unavailable, panic proceeds 4851 on the current CPU. 4852 This option should only be used for systems with the above 4853 constraints as it might cause the panic operation to be less reliable. 4854 4855 panic_print= Bitmask for printing system info when panic happens. 4856 User can chose combination of the following bits: 4857 4858 bit 0: print all tasks info 4859 bit 1: print system memory info 4860 bit 2: print timer info 4861 bit 3: print locks info if CONFIG_LOCKDEP is on 4862 bit 4: print ftrace buffer 4863 bit 5: replay all kernel messages on consoles at the end of panic 4864 bit 6: print all CPUs backtrace (if available in the arch) 4865 bit 7: print only tasks in uninterruptible (blocked) state 4866 4867 *Be aware* that this option may print a _lot_ of lines. 4868 4869 There is an increased risk of losing older messages in 4870 the log. Maybe worth to setup a bigger log buffer with 4871 "log_buf_len" along with this. 4872 4873 Also consider using "printk.console_no_auto_verbose=Y" 4874 if using this along with pstore, to avoid extra delays 4875 due to increased console verbosity during panic. 4876 4877 panic_sys_info= A comma separated list of extra information to be dumped 4878 on panic. 4879 Format: val[,val...] 4880 Where @val can be any of the following: 4881 4882 tasks: print all tasks info 4883 mem: print system memory info 4884 timers: print timers info 4885 locks: print locks info if CONFIG_LOCKDEP is on 4886 ftrace: print ftrace buffer 4887 all_bt: print all CPUs backtrace (if available in the arch) 4888 blocked_tasks: print only tasks in uninterruptible (blocked) state 4889 4890 This is a human readable alternative to the 'panic_print' 4891 option above. Please be aware of the caveats using this, 4892 check the 'panic_print' parameter for more details. 4893 4894 panic_console_replay 4895 When panic happens, replay all kernel messages on 4896 consoles at the end of panic. 4897 4898 parkbd.port= [HW] Parallel port number the keyboard adapter is 4899 connected to, default is 0. 4900 Format: <parport#> 4901 parkbd.mode= [HW] Parallel port keyboard adapter mode of operation, 4902 0 for XT, 1 for AT (default is AT). 4903 Format: <mode> 4904 4905 parport= [HW,PPT] Specify parallel ports. 0 disables. 4906 Format: { 0 | auto | 0xBBB[,IRQ[,DMA]] } 4907 Use 'auto' to force the driver to use any 4908 IRQ/DMA settings detected (the default is to 4909 ignore detected IRQ/DMA settings because of 4910 possible conflicts). You can specify the base 4911 address, IRQ, and DMA settings; IRQ and DMA 4912 should be numbers, or 'auto' (for using detected 4913 settings on that particular port), or 'nofifo' 4914 (to avoid using a FIFO even if it is detected). 4915 Parallel ports are assigned in the order they 4916 are specified on the command line, starting 4917 with parport0. 4918 4919 parport_init_mode= [HW,PPT] 4920 Configure VIA parallel port to operate in 4921 a specific mode. This is necessary on Pegasos 4922 computer where firmware has no options for setting 4923 up parallel port mode and sets it to spp. 4924 Currently this function knows 686a and 8231 chips. 4925 Format: [spp|ps2|epp|ecp|ecpepp] 4926 4927 pata_legacy.all= [HW,LIBATA] 4928 Format: <int> 4929 Set to non-zero to probe primary and secondary ISA 4930 port ranges on PCI systems where no PCI PATA device 4931 has been found at either range. Disabled by default. 4932 4933 pata_legacy.autospeed= [HW,LIBATA] 4934 Format: <int> 4935 Set to non-zero if a chip is present that snoops speed 4936 changes. Disabled by default. 4937 4938 pata_legacy.ht6560a= [HW,LIBATA] 4939 Format: <int> 4940 Set to 1, 2, or 3 for HT 6560A on the primary channel, 4941 the secondary channel, or both channels respectively. 4942 Disabled by default. 4943 4944 pata_legacy.ht6560b= [HW,LIBATA] 4945 Format: <int> 4946 Set to 1, 2, or 3 for HT 6560B on the primary channel, 4947 the secondary channel, or both channels respectively. 4948 Disabled by default. 4949 4950 pata_legacy.iordy_mask= [HW,LIBATA] 4951 Format: <int> 4952 IORDY enable mask. Set individual bits to allow IORDY 4953 for the respective channel. Bit 0 is for the first 4954 legacy channel handled by this driver, bit 1 is for 4955 the second channel, and so on. The sequence will often 4956 correspond to the primary legacy channel, the secondary 4957 legacy channel, and so on, but the handling of a PCI 4958 bus and the use of other driver options may interfere 4959 with the sequence. By default IORDY is allowed across 4960 all channels. 4961 4962 pata_legacy.opti82c46x= [HW,LIBATA] 4963 Format: <int> 4964 Set to 1, 2, or 3 for Opti 82c611A on the primary 4965 channel, the secondary channel, or both channels 4966 respectively. Disabled by default. 4967 4968 pata_legacy.opti82c611a= [HW,LIBATA] 4969 Format: <int> 4970 Set to 1, 2, or 3 for Opti 82c465MV on the primary 4971 channel, the secondary channel, or both channels 4972 respectively. Disabled by default. 4973 4974 pata_legacy.pio_mask= [HW,LIBATA] 4975 Format: <int> 4976 PIO mode mask for autospeed devices. Set individual 4977 bits to allow the use of the respective PIO modes. 4978 Bit 0 is for mode 0, bit 1 is for mode 1, and so on. 4979 All modes allowed by default. 4980 4981 pata_legacy.probe_all= [HW,LIBATA] 4982 Format: <int> 4983 Set to non-zero to probe tertiary and further ISA 4984 port ranges on PCI systems. Disabled by default. 4985 4986 pata_legacy.probe_mask= [HW,LIBATA] 4987 Format: <int> 4988 Probe mask for legacy ISA PATA ports. Depending on 4989 platform configuration and the use of other driver 4990 options up to 6 legacy ports are supported: 0x1f0, 4991 0x170, 0x1e8, 0x168, 0x1e0, 0x160, however probing 4992 of individual ports can be disabled by setting the 4993 corresponding bits in the mask to 1. Bit 0 is for 4994 the first port in the list above (0x1f0), and so on. 4995 By default all supported ports are probed. 4996 4997 pata_legacy.qdi= [HW,LIBATA] 4998 Format: <int> 4999 Set to non-zero to probe QDI controllers. By default 5000 set to 1 if CONFIG_PATA_QDI_MODULE, 0 otherwise. 5001 5002 pata_legacy.winbond= [HW,LIBATA] 5003 Format: <int> 5004 Set to non-zero to probe Winbond controllers. Use 5005 the standard I/O port (0x130) if 1, otherwise the 5006 value given is the I/O port to use (typically 0x1b0). 5007 By default set to 1 if CONFIG_PATA_WINBOND_VLB_MODULE, 5008 0 otherwise. 5009 5010 pata_platform.pio_mask= [HW,LIBATA] 5011 Format: <int> 5012 Supported PIO mode mask. Set individual bits to allow 5013 the use of the respective PIO modes. Bit 0 is for 5014 mode 0, bit 1 is for mode 1, and so on. Mode 0 only 5015 allowed by default. 5016 5017 pause_on_oops=<int> 5018 Halt all CPUs after the first oops has been printed for 5019 the specified number of seconds. This is to be used if 5020 your oopses keep scrolling off the screen. 5021 5022 pci=option[,option...] [PCI,EARLY] various PCI subsystem options. 5023 5024 Some options herein operate on a specific device 5025 or a set of devices (<pci_dev>). These are 5026 specified in one of the following formats: 5027 5028 [<domain>:]<bus>:<dev>.<func>[/<dev>.<func>]* 5029 pci:<vendor>:<device>[:<subvendor>:<subdevice>] 5030 5031 Note: the first format specifies a PCI 5032 bus/device/function address which may change 5033 if new hardware is inserted, if motherboard 5034 firmware changes, or due to changes caused 5035 by other kernel parameters. If the 5036 domain is left unspecified, it is 5037 taken to be zero. Optionally, a path 5038 to a device through multiple device/function 5039 addresses can be specified after the base 5040 address (this is more robust against 5041 renumbering issues). The second format 5042 selects devices using IDs from the 5043 configuration space which may match multiple 5044 devices in the system. 5045 5046 earlydump dump PCI config space before the kernel 5047 changes anything 5048 off [X86] don't probe for the PCI bus 5049 bios [X86-32] force use of PCI BIOS, don't access 5050 the hardware directly. Use this if your machine 5051 has a non-standard PCI host bridge. 5052 nobios [X86-32] disallow use of PCI BIOS, only direct 5053 hardware access methods are allowed. Use this 5054 if you experience crashes upon bootup and you 5055 suspect they are caused by the BIOS. 5056 conf1 [X86] Force use of PCI Configuration Access 5057 Mechanism 1 (config address in IO port 0xCF8, 5058 data in IO port 0xCFC, both 32-bit). 5059 conf2 [X86] Force use of PCI Configuration Access 5060 Mechanism 2 (IO port 0xCF8 is an 8-bit port for 5061 the function, IO port 0xCFA, also 8-bit, sets 5062 bus number. The config space is then accessed 5063 through ports 0xC000-0xCFFF). 5064 See http://wiki.osdev.org/PCI for more info 5065 on the configuration access mechanisms. 5066 noaer [PCIE] If the PCIEAER kernel config parameter is 5067 enabled, this kernel boot option can be used to 5068 disable the use of PCIE advanced error reporting. 5069 nodomains [PCI] Disable support for multiple PCI 5070 root domains (aka PCI segments, in ACPI-speak). 5071 nommconf [X86] Disable use of MMCONFIG for PCI 5072 Configuration 5073 check_enable_amd_mmconf [X86] check for and enable 5074 properly configured MMIO access to PCI 5075 config space on AMD family 10h CPU 5076 nomsi [MSI] If the PCI_MSI kernel config parameter is 5077 enabled, this kernel boot option can be used to 5078 disable the use of MSI interrupts system-wide. 5079 noioapicquirk [APIC] Disable all boot interrupt quirks. 5080 Safety option to keep boot IRQs enabled. This 5081 should never be necessary. 5082 ioapicreroute [APIC] Enable rerouting of boot IRQs to the 5083 primary IO-APIC for bridges that cannot disable 5084 boot IRQs. This fixes a source of spurious IRQs 5085 when the system masks IRQs. 5086 noioapicreroute [APIC] Disable workaround that uses the 5087 boot IRQ equivalent of an IRQ that connects to 5088 a chipset where boot IRQs cannot be disabled. 5089 The opposite of ioapicreroute. 5090 biosirq [X86-32] Use PCI BIOS calls to get the interrupt 5091 routing table. These calls are known to be buggy 5092 on several machines and they hang the machine 5093 when used, but on other computers it's the only 5094 way to get the interrupt routing table. Try 5095 this option if the kernel is unable to allocate 5096 IRQs or discover secondary PCI buses on your 5097 motherboard. 5098 rom [X86] Assign address space to expansion ROMs. 5099 Use with caution as certain devices share 5100 address decoders between ROMs and other 5101 resources. 5102 norom [X86] Do not assign address space to 5103 expansion ROMs that do not already have 5104 BIOS assigned address ranges. 5105 nobar [X86] Do not assign address space to the 5106 BARs that weren't assigned by the BIOS. 5107 irqmask=0xMMMM [X86] Set a bit mask of IRQs allowed to be 5108 assigned automatically to PCI devices. You can 5109 make the kernel exclude IRQs of your ISA cards 5110 this way. 5111 pirqaddr=0xAAAAA [X86] Specify the physical address 5112 of the PIRQ table (normally generated 5113 by the BIOS) if it is outside the 5114 F0000h-100000h range. 5115 lastbus=N [X86] Scan all buses thru bus #N. Can be 5116 useful if the kernel is unable to find your 5117 secondary buses and you want to tell it 5118 explicitly which ones they are. 5119 assign-busses [X86] Always assign all PCI bus 5120 numbers ourselves, overriding 5121 whatever the firmware may have done. 5122 usepirqmask [X86] Honor the possible IRQ mask stored 5123 in the BIOS $PIR table. This is needed on 5124 some systems with broken BIOSes, notably 5125 some HP Pavilion N5400 and Omnibook XE3 5126 notebooks. This will have no effect if ACPI 5127 IRQ routing is enabled. 5128 noacpi [X86] Do not use ACPI for IRQ routing 5129 or for PCI scanning. 5130 use_crs [X86] Use PCI host bridge window information 5131 from ACPI. On BIOSes from 2008 or later, this 5132 is enabled by default. If you need to use this, 5133 please report a bug. 5134 nocrs [X86] Ignore PCI host bridge windows from ACPI. 5135 If you need to use this, please report a bug. 5136 use_e820 [X86] Use E820 reservations to exclude parts of 5137 PCI host bridge windows. This is a workaround 5138 for BIOS defects in host bridge _CRS methods. 5139 If you need to use this, please report a bug to 5140 <linux-pci@vger.kernel.org>. 5141 no_e820 [X86] Ignore E820 reservations for PCI host 5142 bridge windows. This is the default on modern 5143 hardware. If you need to use this, please report 5144 a bug to <linux-pci@vger.kernel.org>. 5145 routeirq Do IRQ routing for all PCI devices. 5146 This is normally done in pci_enable_device(), 5147 so this option is a temporary workaround 5148 for broken drivers that don't call it. 5149 skip_isa_align [X86] do not align io start addr, so can 5150 handle more pci cards 5151 noearly [X86] Don't do any early type 1 scanning. 5152 This might help on some broken boards which 5153 machine check when some devices' config space 5154 is read. But various workarounds are disabled 5155 and some IOMMU drivers will not work. 5156 bfsort Sort PCI devices into breadth-first order. 5157 This sorting is done to get a device 5158 order compatible with older (<= 2.4) kernels. 5159 nobfsort Don't sort PCI devices into breadth-first order. 5160 pcie_bus_tune_off Disable PCIe MPS (Max Payload Size) 5161 tuning and use the BIOS-configured MPS defaults. 5162 pcie_bus_safe Set every device's MPS to the largest value 5163 supported by all devices below the root complex. 5164 pcie_bus_perf Set device MPS to the largest allowable MPS 5165 based on its parent bus. Also set MRRS (Max 5166 Read Request Size) to the largest supported 5167 value (no larger than the MPS that the device 5168 or bus can support) for best performance. 5169 pcie_bus_peer2peer Set every device's MPS to 128B, which 5170 every device is guaranteed to support. This 5171 configuration allows peer-to-peer DMA between 5172 any pair of devices, possibly at the cost of 5173 reduced performance. This also guarantees 5174 that hot-added devices will work. 5175 cbiosize=nn[KMG] The fixed amount of bus space which is 5176 reserved for the CardBus bridge's IO window. 5177 The default value is 256 bytes. 5178 cbmemsize=nn[KMG] The fixed amount of bus space which is 5179 reserved for the CardBus bridge's memory 5180 window. The default value is 64 megabytes. 5181 resource_alignment= 5182 Format: 5183 [<order of align>@]<pci_dev>[; ...] 5184 Specifies alignment and device to reassign 5185 aligned memory resources. How to 5186 specify the device is described above. 5187 If <order of align> is not specified, 5188 PAGE_SIZE is used as alignment. 5189 A PCI-PCI bridge can be specified if resource 5190 windows need to be expanded. 5191 To specify the alignment for several 5192 instances of a device, the PCI vendor, 5193 device, subvendor, and subdevice may be 5194 specified, e.g., 12@pci:8086:9c22:103c:198f 5195 for 4096-byte alignment. 5196 ecrc= Enable/disable PCIe ECRC (transaction layer 5197 end-to-end CRC checking). Only effective if 5198 OS has native AER control (either granted by 5199 ACPI _OSC or forced via "pcie_ports=native") 5200 bios: Use BIOS/firmware settings. This is the 5201 the default. 5202 off: Turn ECRC off 5203 on: Turn ECRC on. 5204 hpiosize=nn[KMG] The fixed amount of bus space which is 5205 reserved for hotplug bridge's IO window. 5206 Default size is 256 bytes. 5207 hpmmiosize=nn[KMG] The fixed amount of bus space which is 5208 reserved for hotplug bridge's MMIO window. 5209 Default size is 2 megabytes. 5210 hpmmioprefsize=nn[KMG] The fixed amount of bus space which is 5211 reserved for hotplug bridge's MMIO_PREF window. 5212 Default size is 2 megabytes. 5213 hpmemsize=nn[KMG] The fixed amount of bus space which is 5214 reserved for hotplug bridge's MMIO and 5215 MMIO_PREF window. 5216 Default size is 2 megabytes. 5217 hpbussize=nn The minimum amount of additional bus numbers 5218 reserved for buses below a hotplug bridge. 5219 Default is 1. 5220 realloc= Enable/disable reallocating PCI bridge resources 5221 if allocations done by BIOS are too small to 5222 accommodate resources required by all child 5223 devices. 5224 off: Turn realloc off 5225 on: Turn realloc on 5226 realloc same as realloc=on 5227 noari do not use PCIe ARI. 5228 noats [PCIE, Intel-IOMMU, AMD-IOMMU] 5229 do not use PCIe ATS (and IOMMU device IOTLB). 5230 pcie_scan_all Scan all possible PCIe devices. Otherwise we 5231 only look for one device below a PCIe downstream 5232 port. 5233 big_root_window Try to add a big 64bit memory window to the PCIe 5234 root complex on AMD CPUs. Some GFX hardware 5235 can resize a BAR to allow access to all VRAM. 5236 Adding the window is slightly risky (it may 5237 conflict with unreported devices), so this 5238 taints the kernel. 5239 disable_acs_redir=<pci_dev>[; ...] 5240 Specify one or more PCI devices (in the format 5241 specified above) separated by semicolons. 5242 Each device specified will have the PCI ACS 5243 redirect capabilities forced off which will 5244 allow P2P traffic between devices through 5245 bridges without forcing it upstream. Note: 5246 this removes isolation between devices and 5247 may put more devices in an IOMMU group. 5248 config_acs= 5249 Format: 5250 <ACS flags>@<pci_dev>[; ...] 5251 Specify one or more PCI devices (in the format 5252 specified above) optionally prepended with flags 5253 and separated by semicolons. The respective 5254 capabilities will be enabled, disabled or 5255 unchanged based on what is specified in 5256 flags. 5257 5258 ACS Flags is defined as follows: 5259 bit-0 : ACS Source Validation 5260 bit-1 : ACS Translation Blocking 5261 bit-2 : ACS P2P Request Redirect 5262 bit-3 : ACS P2P Completion Redirect 5263 bit-4 : ACS Upstream Forwarding 5264 bit-5 : ACS P2P Egress Control 5265 bit-6 : ACS Direct Translated P2P 5266 Each bit can be marked as: 5267 '0' – force disabled 5268 '1' – force enabled 5269 'x' – unchanged 5270 For example, 5271 pci=config_acs=10x@pci:0:0 5272 would configure all devices that support 5273 ACS to enable P2P Request Redirect, disable 5274 Translation Blocking, and leave Source 5275 Validation unchanged from whatever power-up 5276 or firmware set it to. 5277 5278 Note: this may remove isolation between devices 5279 and may put more devices in an IOMMU group. 5280 force_floating [S390] Force usage of floating interrupts. 5281 nomio [S390] Do not use MIO instructions. 5282 norid [S390] ignore the RID field and force use of 5283 one PCI domain per PCI function 5284 notph [PCIE] If the PCIE_TPH kernel config parameter 5285 is enabled, this kernel boot option can be used 5286 to disable PCIe TLP Processing Hints support 5287 system-wide. 5288 5289 pcie_aspm= [PCIE] Forcibly enable or ignore PCIe Active State Power 5290 Management. 5291 off Don't touch ASPM configuration at all. Leave any 5292 configuration done by firmware unchanged. 5293 force Enable ASPM even on devices that claim not to support it. 5294 WARNING: Forcing ASPM on may cause system lockups. 5295 5296 pcie_ports= [PCIE] PCIe port services handling: 5297 native Use native PCIe services (PME, AER, DPC, PCIe hotplug) 5298 even if the platform doesn't give the OS permission to 5299 use them. This may cause conflicts if the platform 5300 also tries to use these services. 5301 dpc-native Use native PCIe service for DPC only. May 5302 cause conflicts if firmware uses AER or DPC. 5303 compat Disable native PCIe services (PME, AER, DPC, PCIe 5304 hotplug). 5305 5306 pcie_port_pm= [PCIE] PCIe port power management handling: 5307 off Disable power management of all PCIe ports 5308 force Forcibly enable power management of all PCIe ports 5309 5310 pcie_pme= [PCIE,PM] Native PCIe PME signaling options: 5311 nomsi Do not use MSI for native PCIe PME signaling (this makes 5312 all PCIe root ports use INTx for all services). 5313 5314 pd_ignore_unused 5315 [PM] 5316 Keep all power-domains already enabled by bootloader on, 5317 even if no driver has claimed them. This is useful 5318 for debug and development, but should not be 5319 needed on a platform with proper driver support. 5320 5321 pdcchassis= [PARISC,HW] Disable/Enable PDC Chassis Status codes at 5322 boot time. 5323 Format: { 0 | 1 } 5324 See arch/parisc/kernel/pdc_chassis.c 5325 5326 percpu_alloc= [MM,EARLY] 5327 Select which percpu first chunk allocator to use. 5328 Currently supported values are "embed" and "page". 5329 Archs may support subset or none of the selections. 5330 See comments in mm/percpu.c for details on each 5331 allocator. This parameter is primarily for debugging 5332 and performance comparison. 5333 5334 pirq= [SMP,APIC] Manual mp-table setup 5335 See Documentation/arch/x86/i386/IO-APIC.rst. 5336 5337 plip= [PPT,NET] Parallel port network link 5338 Format: { parport<nr> | timid | 0 } 5339 See also Documentation/admin-guide/parport.rst. 5340 5341 pmtmr= [X86] Manual setup of pmtmr I/O Port. 5342 Override pmtimer IOPort with a hex value. 5343 e.g. pmtmr=0x508 5344 5345 pmu_override= [PPC] Override the PMU. 5346 This option takes over the PMU facility, so it is no 5347 longer usable by perf. Setting this option starts the 5348 PMU counters by setting MMCR0 to 0 (the FC bit is 5349 cleared). If a number is given, then MMCR1 is set to 5350 that number, otherwise (e.g., 'pmu_override=on'), MMCR1 5351 remains 0. 5352 5353 pm_async= [PM] 5354 Format: off 5355 This parameter sets the initial value of the 5356 /sys/power/pm_async sysfs knob at boot time. 5357 If set to "off", disables asynchronous suspend and 5358 resume of devices during system-wide power transitions. 5359 This can be useful on platforms where device 5360 dependencies are not well-defined, or for debugging 5361 power management issues. Asynchronous operations are 5362 enabled by default. 5363 5364 5365 pm_debug_messages [SUSPEND,KNL] 5366 Enable suspend/resume debug messages during boot up. 5367 5368 pm.dpm_watchdog_enabled= 5369 [PM] Enable or disable the DPM watchdog. Requires 5370 CONFIG_PM_SLEEP and CONFIG_DPM_WATCHDOG enabled. 5371 Format: <bool> 5372 Default value is set by CONFIG_DPM_WATCHDOG_ENABLED. 5373 5374 pnp.debug=1 [PNP] 5375 Enable PNP debug messages (depends on the 5376 CONFIG_PNP_DEBUG_MESSAGES option). Change at run-time 5377 via /sys/module/pnp/parameters/debug. We always show 5378 current resource usage; turning this on also shows 5379 possible settings and some assignment information. 5380 5381 pnpacpi= [ACPI] 5382 { off } 5383 5384 pnpbios= [ISAPNP] 5385 { on | off | curr | res | no-curr | no-res } 5386 5387 pnp_reserve_irq= 5388 [ISAPNP] Exclude IRQs for the autoconfiguration 5389 5390 pnp_reserve_dma= 5391 [ISAPNP] Exclude DMAs for the autoconfiguration 5392 5393 pnp_reserve_io= [ISAPNP] Exclude I/O ports for the autoconfiguration 5394 Ranges are in pairs (I/O port base and size). 5395 5396 pnp_reserve_mem= 5397 [ISAPNP] Exclude memory regions for the 5398 autoconfiguration. 5399 Ranges are in pairs (memory base and size). 5400 5401 ports= [IP_VS_FTP] IPVS ftp helper module 5402 Default is 21. 5403 Up to 8 (IP_VS_APP_MAX_PORTS) ports 5404 may be specified. 5405 Format: <port>,<port>.... 5406 5407 possible_cpus= [SMP,S390,X86] 5408 Format: <unsigned int> 5409 Set the number of possible CPUs, overriding the 5410 regular discovery mechanisms (such as ACPI/FW, etc). 5411 5412 powersave=off [PPC] This option disables power saving features. 5413 It specifically disables cpuidle and sets the 5414 platform machine description specific power_save 5415 function to NULL. On Idle the CPU just reduces 5416 execution priority. 5417 5418 ppc_strict_facility_enable 5419 [PPC,ENABLE] This option catches any kernel floating point, 5420 Altivec, VSX and SPE outside of regions specifically 5421 allowed (eg kernel_enable_fpu()/kernel_disable_fpu()). 5422 There is some performance impact when enabling this. 5423 5424 ppc_tm= [PPC,EARLY] 5425 Format: {"off"} 5426 Disable Hardware Transactional Memory 5427 5428 preempt= [KNL] 5429 Select preemption mode if you have CONFIG_PREEMPT_DYNAMIC 5430 none - Limited to cond_resched() calls 5431 voluntary - Limited to cond_resched() and might_sleep() calls 5432 full - Any section that isn't explicitly preempt disabled 5433 can be preempted anytime. Tasks will also yield 5434 contended spinlocks (if the critical section isn't 5435 explicitly preempt disabled beyond the lock itself). 5436 lazy - Scheduler controlled. Similar to full but instead 5437 of preempting the task immediately, the task gets 5438 one HZ tick time to yield itself before the 5439 preemption will be forced. One preemption is when the 5440 task returns to user space. 5441 5442 print-fatal-signals= 5443 [KNL] debug: print fatal signals 5444 5445 If enabled, warn about various signal handling 5446 related application anomalies: too many signals, 5447 too many POSIX.1 timers, fatal signals causing a 5448 coredump - etc. 5449 5450 If you hit the warning due to signal overflow, 5451 you might want to try "ulimit -i unlimited". 5452 5453 default: off. 5454 5455 printk.always_kmsg_dump= 5456 Trigger kmsg_dump for cases other than kernel oops or 5457 panics 5458 Format: <bool> (1/Y/y=enable, 0/N/n=disable) 5459 default: disabled 5460 5461 printk.console_no_auto_verbose= 5462 Disable console loglevel raise on oops, panic 5463 or lockdep-detected issues (only if lock debug is on). 5464 With an exception to setups with low baudrate on 5465 serial console, keeping this 0 is a good choice 5466 in order to provide more debug information. 5467 Format: <bool> 5468 default: 0 (auto_verbose is enabled) 5469 5470 printk.debug_non_panic_cpus= 5471 Allows storing messages from non-panic CPUs into 5472 the printk log buffer during panic(). They are 5473 flushed to consoles by the panic-CPU on 5474 a best-effort basis. 5475 Format: <bool> (1/Y/y=enable, 0/N/n=disable) 5476 Default: disabled 5477 5478 printk.devkmsg={on,off,ratelimit} 5479 Control writing to /dev/kmsg. 5480 on - unlimited logging to /dev/kmsg from userspace 5481 off - logging to /dev/kmsg disabled 5482 ratelimit - ratelimit the logging 5483 Default: ratelimit 5484 5485 printk.time= Show timing data prefixed to each printk message line 5486 Format: <bool> (1/Y/y=enable, 0/N/n=disable) 5487 5488 proc_mem.force_override= [KNL] 5489 Format: {always | ptrace | never} 5490 Traditionally /proc/pid/mem allows memory permissions to be 5491 overridden without restrictions. This option may be set to 5492 restrict that. Can be one of: 5493 - 'always': traditional behavior always allows mem overrides. 5494 - 'ptrace': only allow mem overrides for active ptracers. 5495 - 'never': never allow mem overrides. 5496 If not specified, default is the CONFIG_PROC_MEM_* choice. 5497 5498 processor.max_cstate= [HW,ACPI] 5499 Limit processor to maximum C-state 5500 max_cstate=9 overrides any DMI blacklist limit. 5501 5502 processor.nocst [HW,ACPI] 5503 Ignore the _CST method to determine C-states, 5504 instead using the legacy FADT method 5505 5506 profile= [KNL] Enable kernel profiling via /proc/profile 5507 Format: [<profiletype>,]<number> 5508 Param: <profiletype>: "schedule" or "kvm" 5509 [defaults to kernel profiling] 5510 Param: "schedule" - profile schedule points. 5511 Param: "kvm" - profile VM exits. 5512 Param: <number> - step/bucket size as a power of 2 for 5513 statistical time based profiling. 5514 5515 prot_virt= [S390] enable hosting protected virtual machines 5516 isolated from the hypervisor (if hardware supports 5517 that). If enabled, the default kernel base address 5518 might be overridden even when Kernel Address Space 5519 Layout Randomization is disabled. 5520 Format: <bool> 5521 5522 psi= [KNL] Enable or disable pressure stall information 5523 tracking. 5524 Format: <bool> 5525 5526 psmouse.proto= [HW,MOUSE] Highest PS2 mouse protocol extension to 5527 probe for; one of (bare|imps|exps|lifebook|any). 5528 psmouse.rate= [HW,MOUSE] Set desired mouse report rate, in reports 5529 per second. 5530 psmouse.resetafter= [HW,MOUSE] 5531 Try to reset the device after so many bad packets 5532 (0 = never). 5533 psmouse.resolution= 5534 [HW,MOUSE] Set desired mouse resolution, in dpi. 5535 psmouse.smartscroll= 5536 [HW,MOUSE] Controls Logitech smartscroll autorepeat. 5537 0 = disabled, 1 = enabled (default). 5538 5539 pstore.backend= Specify the name of the pstore backend to use 5540 5541 pti= [X86-64] Control Page Table Isolation of user and 5542 kernel address spaces. Disabling this feature 5543 removes hardening, but improves performance of 5544 system calls and interrupts. 5545 5546 on - unconditionally enable 5547 off - unconditionally disable 5548 auto - kernel detects whether your CPU model is 5549 vulnerable to issues that PTI mitigates 5550 5551 Not specifying this option is equivalent to pti=auto. 5552 5553 pty.legacy_count= 5554 [KNL] Number of legacy pty's. Overwrites compiled-in 5555 default number. 5556 5557 quiet [KNL,EARLY] Disable most log messages 5558 5559 r128= [HW,DRM] 5560 5561 radix_hcall_invalidate=on [PPC/PSERIES] 5562 Disable RADIX GTSE feature and use hcall for TLB 5563 invalidate. 5564 5565 raid= [HW,RAID] 5566 See Documentation/admin-guide/md.rst. 5567 5568 ramdisk_size= [RAM] Sizes of RAM disks in kilobytes 5569 See Documentation/admin-guide/blockdev/ramdisk.rst. 5570 5571 ramdisk_start= [Deprecated,RAM] RAM disk image start address 5572 5573 random.trust_cpu=off 5574 [KNL,EARLY] Disable trusting the use of the CPU's 5575 random number generator (if available) to 5576 initialize the kernel's RNG. 5577 5578 random.trust_bootloader=off 5579 [KNL,EARLY] Disable trusting the use of the a seed 5580 passed by the bootloader (if available) to 5581 initialize the kernel's RNG. 5582 5583 randomize_kstack_offset= 5584 [KNL,EARLY] Enable or disable kernel stack offset 5585 randomization, which provides roughly 5 bits of 5586 entropy, frustrating memory corruption attacks 5587 that depend on stack address determinism or 5588 cross-syscall address exposures. This is only 5589 available on architectures that have defined 5590 CONFIG_HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET. 5591 Format: <bool> (1/Y/y=enable, 0/N/n=disable) 5592 Default is CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT. 5593 5594 ras=option[,option,...] [KNL] RAS-specific options 5595 5596 cec_disable [X86] 5597 Disable the Correctable Errors Collector, 5598 see CONFIG_RAS_CEC help text. 5599 5600 rcu_nocbs[=cpu-list] 5601 [KNL] The optional argument is a cpu list, 5602 as described above. 5603 5604 In kernels built with CONFIG_RCU_NOCB_CPU=y, 5605 enable the no-callback CPU mode, which prevents 5606 such CPUs' callbacks from being invoked in 5607 softirq context. Invocation of such CPUs' RCU 5608 callbacks will instead be offloaded to "rcuox/N" 5609 kthreads created for that purpose, where "x" is 5610 "p" for RCU-preempt, "s" for RCU-sched, and "g" 5611 for the kthreads that mediate grace periods; and 5612 "N" is the CPU number. This reduces OS jitter on 5613 the offloaded CPUs, which can be useful for HPC 5614 and real-time workloads. It can also improve 5615 energy efficiency for asymmetric multiprocessors. 5616 5617 If a cpulist is passed as an argument, the specified 5618 list of CPUs is set to no-callback mode from boot. 5619 5620 Otherwise, if the '=' sign and the cpulist 5621 arguments are omitted, no CPU will be set to 5622 no-callback mode from boot but the mode may be 5623 toggled at runtime via cpusets. 5624 5625 Note that this argument takes precedence over 5626 the CONFIG_RCU_NOCB_CPU_DEFAULT_ALL option. 5627 5628 rcu_nocb_poll [KNL] 5629 Rather than requiring that offloaded CPUs 5630 (specified by rcu_nocbs= above) explicitly 5631 awaken the corresponding "rcuoN" kthreads, 5632 make these kthreads poll for callbacks. 5633 This improves the real-time response for the 5634 offloaded CPUs by relieving them of the need to 5635 wake up the corresponding kthread, but degrades 5636 energy efficiency by requiring that the kthreads 5637 periodically wake up to do the polling. 5638 5639 rcutree.blimit= [KNL] 5640 Set maximum number of finished RCU callbacks to 5641 process in one batch. 5642 5643 rcutree.csd_lock_suppress_rcu_stall= [KNL] 5644 Do only a one-line RCU CPU stall warning when 5645 there is an ongoing too-long CSD-lock wait. 5646 5647 rcutree.do_rcu_barrier= [KNL] 5648 Request a call to rcu_barrier(). This is 5649 throttled so that userspace tests can safely 5650 hammer on the sysfs variable if they so choose. 5651 If triggered before the RCU grace-period machinery 5652 is fully active, this will error out with EAGAIN. 5653 5654 rcutree.dump_tree= [KNL] 5655 Dump the structure of the rcu_node combining tree 5656 out at early boot. This is used for diagnostic 5657 purposes, to verify correct tree setup. 5658 5659 rcutree.gp_cleanup_delay= [KNL] 5660 Set the number of jiffies to delay each step of 5661 RCU grace-period cleanup. 5662 5663 rcutree.gp_init_delay= [KNL] 5664 Set the number of jiffies to delay each step of 5665 RCU grace-period initialization. 5666 5667 rcutree.gp_preinit_delay= [KNL] 5668 Set the number of jiffies to delay each step of 5669 RCU grace-period pre-initialization, that is, 5670 the propagation of recent CPU-hotplug changes up 5671 the rcu_node combining tree. 5672 5673 rcutree.jiffies_till_first_fqs= [KNL] 5674 Set delay from grace-period initialization to 5675 first attempt to force quiescent states. 5676 Units are jiffies, minimum value is zero, 5677 and maximum value is HZ. 5678 5679 rcutree.jiffies_till_next_fqs= [KNL] 5680 Set delay between subsequent attempts to force 5681 quiescent states. Units are jiffies, minimum 5682 value is one, and maximum value is HZ. 5683 5684 rcutree.jiffies_till_sched_qs= [KNL] 5685 Set required age in jiffies for a 5686 given grace period before RCU starts 5687 soliciting quiescent-state help from 5688 rcu_note_context_switch() and cond_resched(). 5689 If not specified, the kernel will calculate 5690 a value based on the most recent settings 5691 of rcutree.jiffies_till_first_fqs 5692 and rcutree.jiffies_till_next_fqs. 5693 This calculated value may be viewed in 5694 rcutree.jiffies_to_sched_qs. Any attempt to set 5695 rcutree.jiffies_to_sched_qs will be cheerfully 5696 overwritten. 5697 5698 rcutree.kthread_prio= [KNL,BOOT] 5699 Set the SCHED_FIFO priority of the RCU per-CPU 5700 kthreads (rcuc/N). This value is also used for 5701 the priority of the RCU boost threads (rcub/N) 5702 and for the RCU grace-period kthreads (rcu_bh, 5703 rcu_preempt, and rcu_sched). If RCU_BOOST is 5704 set, valid values are 1-99 and the default is 1 5705 (the least-favored priority). Otherwise, when 5706 RCU_BOOST is not set, valid values are 0-99 and 5707 the default is zero (non-realtime operation). 5708 When RCU_NOCB_CPU is set, also adjust the 5709 priority of NOCB callback kthreads. 5710 5711 rcutree.nocb_nobypass_lim_per_jiffy= [KNL] 5712 On callback-offloaded (rcu_nocbs) CPUs, 5713 RCU reduces the lock contention that would 5714 otherwise be caused by callback floods through 5715 use of the ->nocb_bypass list. However, in the 5716 common non-flooded case, RCU queues directly to 5717 the main ->cblist in order to avoid the extra 5718 overhead of the ->nocb_bypass list and its lock. 5719 But if there are too many callbacks queued during 5720 a single jiffy, RCU pre-queues the callbacks into 5721 the ->nocb_bypass queue. The definition of "too 5722 many" is supplied by this kernel boot parameter. 5723 5724 rcutree.nohz_full_patience_delay= [KNL] 5725 On callback-offloaded (rcu_nocbs) CPUs, avoid 5726 disturbing RCU unless the grace period has 5727 reached the specified age in milliseconds. 5728 Defaults to zero. Large values will be capped 5729 at five seconds. All values will be rounded down 5730 to the nearest value representable by jiffies. 5731 5732 rcutree.qhimark= [KNL] 5733 Set threshold of queued RCU callbacks beyond which 5734 batch limiting is disabled. 5735 5736 rcutree.qlowmark= [KNL] 5737 Set threshold of queued RCU callbacks below which 5738 batch limiting is re-enabled. 5739 5740 rcutree.qovld= [KNL] 5741 Set threshold of queued RCU callbacks beyond which 5742 RCU's force-quiescent-state scan will aggressively 5743 enlist help from cond_resched() and sched IPIs to 5744 help CPUs more quickly reach quiescent states. 5745 Set to less than zero to make this be set based 5746 on rcutree.qhimark at boot time and to zero to 5747 disable more aggressive help enlistment. 5748 5749 rcutree.rcu_delay_page_cache_fill_msec= [KNL] 5750 Set the page-cache refill delay (in milliseconds) 5751 in response to low-memory conditions. The range 5752 of permitted values is in the range 0:100000. 5753 5754 rcutree.rcu_divisor= [KNL] 5755 Set the shift-right count to use to compute 5756 the callback-invocation batch limit bl from 5757 the number of callbacks queued on this CPU. 5758 The result will be bounded below by the value of 5759 the rcutree.blimit kernel parameter. Every bl 5760 callbacks, the softirq handler will exit in 5761 order to allow the CPU to do other work. 5762 5763 Please note that this callback-invocation batch 5764 limit applies only to non-offloaded callback 5765 invocation. Offloaded callbacks are instead 5766 invoked in the context of an rcuoc kthread, which 5767 scheduler will preempt as it does any other task. 5768 5769 rcutree.rcu_fanout_exact= [KNL] 5770 Disable autobalancing of the rcu_node combining 5771 tree. This is used by rcutorture, and might 5772 possibly be useful for architectures having high 5773 cache-to-cache transfer latencies. 5774 5775 rcutree.rcu_fanout_leaf= [KNL] 5776 Change the number of CPUs assigned to each 5777 leaf rcu_node structure. Useful for very 5778 large systems, which will choose the value 64, 5779 and for NUMA systems with large remote-access 5780 latencies, which will choose a value aligned 5781 with the appropriate hardware boundaries. 5782 5783 rcutree.rcu_min_cached_objs= [KNL] 5784 Minimum number of objects which are cached and 5785 maintained per one CPU. Object size is equal 5786 to PAGE_SIZE. The cache allows to reduce the 5787 pressure to page allocator, also it makes the 5788 whole algorithm to behave better in low memory 5789 condition. 5790 5791 rcutree.rcu_nocb_gp_stride= [KNL] 5792 Set the number of NOCB callback kthreads in 5793 each group, which defaults to the square root 5794 of the number of CPUs. Larger numbers reduce 5795 the wakeup overhead on the global grace-period 5796 kthread, but increases that same overhead on 5797 each group's NOCB grace-period kthread. 5798 5799 rcutree.rcu_kick_kthreads= [KNL] 5800 Cause the grace-period kthread to get an extra 5801 wake_up() if it sleeps three times longer than 5802 it should at force-quiescent-state time. 5803 This wake_up() will be accompanied by a 5804 WARN_ONCE() splat and an ftrace_dump(). 5805 5806 rcutree.rcu_resched_ns= [KNL] 5807 Limit the time spend invoking a batch of RCU 5808 callbacks to the specified number of nanoseconds. 5809 By default, this limit is checked only once 5810 every 32 callbacks in order to limit the pain 5811 inflicted by local_clock() overhead. 5812 5813 rcutree.rcu_unlock_delay= [KNL] 5814 In CONFIG_RCU_STRICT_GRACE_PERIOD=y kernels, 5815 this specifies an rcu_read_unlock()-time delay 5816 in microseconds. This defaults to zero. 5817 Larger delays increase the probability of 5818 catching RCU pointer leaks, that is, buggy use 5819 of RCU-protected pointers after the relevant 5820 rcu_read_unlock() has completed. 5821 5822 rcutree.sysrq_rcu= [KNL] 5823 Commandeer a sysrq key to dump out Tree RCU's 5824 rcu_node tree with an eye towards determining 5825 why a new grace period has not yet started. 5826 5827 rcutree.use_softirq= [KNL] 5828 If set to zero, move all RCU_SOFTIRQ processing to 5829 per-CPU rcuc kthreads. Defaults to a non-zero 5830 value, meaning that RCU_SOFTIRQ is used by default. 5831 Specify rcutree.use_softirq=0 to use rcuc kthreads. 5832 5833 But note that CONFIG_PREEMPT_RT=y kernels disable 5834 this kernel boot parameter, forcibly setting it 5835 to zero. 5836 5837 rcutree.enable_rcu_lazy= [KNL] 5838 To save power, batch RCU callbacks and flush after 5839 delay, memory pressure or callback list growing too 5840 big. 5841 5842 rcutree.rcu_normal_wake_from_gp= [KNL] 5843 Reduces a latency of synchronize_rcu() call. This approach 5844 maintains its own track of synchronize_rcu() callers, so it 5845 does not interact with regular callbacks because it does not 5846 use a call_rcu[_hurry]() path. Please note, this is for a 5847 normal grace period. 5848 5849 How to disable it: 5850 5851 echo 0 > /sys/module/rcutree/parameters/rcu_normal_wake_from_gp 5852 or pass a boot parameter "rcutree.rcu_normal_wake_from_gp=0" 5853 5854 Default is 1 if it is not explicitly disabled by the boot parameter 5855 passing 0. 5856 5857 rcuscale.gp_async= [KNL] 5858 Measure performance of asynchronous 5859 grace-period primitives such as call_rcu(). 5860 5861 rcuscale.gp_async_max= [KNL] 5862 Specify the maximum number of outstanding 5863 callbacks per writer thread. When a writer 5864 thread exceeds this limit, it invokes the 5865 corresponding flavor of rcu_barrier() to allow 5866 previously posted callbacks to drain. 5867 5868 rcuscale.gp_exp= [KNL] 5869 Measure performance of expedited synchronous 5870 grace-period primitives. 5871 5872 rcuscale.holdoff= [KNL] 5873 Set test-start holdoff period. The purpose of 5874 this parameter is to delay the start of the 5875 test until boot completes in order to avoid 5876 interference. 5877 5878 rcuscale.kfree_by_call_rcu= [KNL] 5879 In kernels built with CONFIG_RCU_LAZY=y, test 5880 call_rcu() instead of kfree_rcu(). 5881 5882 rcuscale.kfree_mult= [KNL] 5883 Instead of allocating an object of size kfree_obj, 5884 allocate one of kfree_mult * sizeof(kfree_obj). 5885 Defaults to 1. 5886 5887 rcuscale.kfree_rcu_test= [KNL] 5888 Set to measure performance of kfree_rcu() flooding. 5889 5890 rcuscale.kfree_rcu_test_double= [KNL] 5891 Test the double-argument variant of kfree_rcu(). 5892 If this parameter has the same value as 5893 rcuscale.kfree_rcu_test_single, both the single- 5894 and double-argument variants are tested. 5895 5896 rcuscale.kfree_rcu_test_single= [KNL] 5897 Test the single-argument variant of kfree_rcu(). 5898 If this parameter has the same value as 5899 rcuscale.kfree_rcu_test_double, both the single- 5900 and double-argument variants are tested. 5901 5902 rcuscale.kfree_nthreads= [KNL] 5903 The number of threads running loops of kfree_rcu(). 5904 5905 rcuscale.kfree_alloc_num= [KNL] 5906 Number of allocations and frees done in an iteration. 5907 5908 rcuscale.kfree_loops= [KNL] 5909 Number of loops doing rcuscale.kfree_alloc_num number 5910 of allocations and frees. 5911 5912 rcuscale.minruntime= [KNL] 5913 Set the minimum test run time in seconds. This 5914 does not affect the data-collection interval, 5915 but instead allows better measurement of things 5916 like CPU consumption. 5917 5918 rcuscale.nreaders= [KNL] 5919 Set number of RCU readers. The value -1 selects 5920 N, where N is the number of CPUs. A value 5921 "n" less than -1 selects N-n+1, where N is again 5922 the number of CPUs. For example, -2 selects N 5923 (the number of CPUs), -3 selects N+1, and so on. 5924 A value of "n" less than or equal to -N selects 5925 a single reader. 5926 5927 rcuscale.nwriters= [KNL] 5928 Set number of RCU writers. The values operate 5929 the same as for rcuscale.nreaders. 5930 N, where N is the number of CPUs 5931 5932 rcuscale.scale_type= [KNL] 5933 Specify the RCU implementation to test. 5934 5935 rcuscale.shutdown= [KNL] 5936 Shut the system down after performance tests 5937 complete. This is useful for hands-off automated 5938 testing. 5939 5940 rcuscale.verbose= [KNL] 5941 Enable additional printk() statements. 5942 5943 rcuscale.writer_holdoff= [KNL] 5944 Write-side holdoff between grace periods, 5945 in microseconds. The default of zero says 5946 no holdoff. 5947 5948 rcuscale.writer_holdoff_jiffies= [KNL] 5949 Additional write-side holdoff between grace 5950 periods, but in jiffies. The default of zero 5951 says no holdoff. 5952 5953 rcutorture.fqs_duration= [KNL] 5954 Set duration of force_quiescent_state bursts 5955 in microseconds. 5956 5957 rcutorture.fqs_holdoff= [KNL] 5958 Set holdoff time within force_quiescent_state bursts 5959 in microseconds. 5960 5961 rcutorture.fqs_stutter= [KNL] 5962 Set wait time between force_quiescent_state bursts 5963 in seconds. 5964 5965 rcutorture.fwd_progress= [KNL] 5966 Specifies the number of kthreads to be used 5967 for RCU grace-period forward-progress testing 5968 for the types of RCU supporting this notion. 5969 Defaults to 1 kthread, values less than zero or 5970 greater than the number of CPUs cause the number 5971 of CPUs to be used. 5972 5973 rcutorture.fwd_progress_div= [KNL] 5974 Specify the fraction of a CPU-stall-warning 5975 period to do tight-loop forward-progress testing. 5976 5977 rcutorture.fwd_progress_holdoff= [KNL] 5978 Number of seconds to wait between successive 5979 forward-progress tests. 5980 5981 rcutorture.fwd_progress_need_resched= [KNL] 5982 Enclose cond_resched() calls within checks for 5983 need_resched() during tight-loop forward-progress 5984 testing. 5985 5986 rcutorture.gp_cond= [KNL] 5987 Use conditional/asynchronous update-side 5988 normal-grace-period primitives, if available. 5989 5990 rcutorture.gp_cond_exp= [KNL] 5991 Use conditional/asynchronous update-side 5992 expedited-grace-period primitives, if available. 5993 5994 rcutorture.gp_cond_full= [KNL] 5995 Use conditional/asynchronous update-side 5996 normal-grace-period primitives that also take 5997 concurrent expedited grace periods into account, 5998 if available. 5999 6000 rcutorture.gp_cond_exp_full= [KNL] 6001 Use conditional/asynchronous update-side 6002 expedited-grace-period primitives that also take 6003 concurrent normal grace periods into account, 6004 if available. 6005 6006 rcutorture.gp_cond_wi= [KNL] 6007 Nominal wait interval for normal conditional 6008 grace periods (specified by rcutorture's 6009 gp_cond and gp_cond_full module parameters), 6010 in microseconds. The actual wait interval will 6011 be randomly selected to nanosecond granularity up 6012 to this wait interval. Defaults to 16 jiffies, 6013 for example, 16,000 microseconds on a system 6014 with HZ=1000. 6015 6016 rcutorture.gp_cond_wi_exp= [KNL] 6017 Nominal wait interval for expedited conditional 6018 grace periods (specified by rcutorture's 6019 gp_cond_exp and gp_cond_exp_full module 6020 parameters), in microseconds. The actual wait 6021 interval will be randomly selected to nanosecond 6022 granularity up to this wait interval. Defaults to 6023 128 microseconds. 6024 6025 rcutorture.gp_exp= [KNL] 6026 Use expedited update-side primitives, if available. 6027 6028 rcutorture.gp_normal= [KNL] 6029 Use normal (non-expedited) asynchronous 6030 update-side primitives, if available. 6031 6032 rcutorture.gp_poll= [KNL] 6033 Use polled update-side normal-grace-period 6034 primitives, if available. 6035 6036 rcutorture.gp_poll_exp= [KNL] 6037 Use polled update-side expedited-grace-period 6038 primitives, if available. 6039 6040 rcutorture.gp_poll_full= [KNL] 6041 Use polled update-side normal-grace-period 6042 primitives that also take concurrent expedited 6043 grace periods into account, if available. 6044 6045 rcutorture.gp_poll_exp_full= [KNL] 6046 Use polled update-side expedited-grace-period 6047 primitives that also take concurrent normal 6048 grace periods into account, if available. 6049 6050 rcutorture.gp_poll_wi= [KNL] 6051 Nominal wait interval for normal conditional 6052 grace periods (specified by rcutorture's 6053 gp_poll and gp_poll_full module parameters), 6054 in microseconds. The actual wait interval will 6055 be randomly selected to nanosecond granularity up 6056 to this wait interval. Defaults to 16 jiffies, 6057 for example, 16,000 microseconds on a system 6058 with HZ=1000. 6059 6060 rcutorture.gp_poll_wi_exp= [KNL] 6061 Nominal wait interval for expedited conditional 6062 grace periods (specified by rcutorture's 6063 gp_poll_exp and gp_poll_exp_full module 6064 parameters), in microseconds. The actual wait 6065 interval will be randomly selected to nanosecond 6066 granularity up to this wait interval. Defaults to 6067 128 microseconds. 6068 6069 rcutorture.gp_sync= [KNL] 6070 Use normal (non-expedited) synchronous 6071 update-side primitives, if available. If all 6072 of rcutorture.gp_cond=, rcutorture.gp_exp=, 6073 rcutorture.gp_normal=, and rcutorture.gp_sync= 6074 are zero, rcutorture acts as if is interpreted 6075 they are all non-zero. 6076 6077 rcutorture.gpwrap_lag= [KNL] 6078 Enable grace-period wrap lag testing. Setting 6079 to false prevents the gpwrap lag test from 6080 running. Default is true. 6081 6082 rcutorture.gpwrap_lag_gps= [KNL] 6083 Set the value for grace-period wrap lag during 6084 active lag testing periods. This controls how many 6085 grace periods differences we tolerate between 6086 rdp and rnp's gp_seq before setting overflow flag. 6087 The default is always set to 8. 6088 6089 rcutorture.gpwrap_lag_cycle_mins= [KNL] 6090 Set the total cycle duration for gpwrap lag 6091 testing in minutes. This is the total time for 6092 one complete cycle of active and inactive 6093 testing periods. Default is 30 minutes. 6094 6095 rcutorture.gpwrap_lag_active_mins= [KNL] 6096 Set the duration for which gpwrap lag is active 6097 within each cycle, in minutes. During this time, 6098 the grace-period wrap lag will be set to the 6099 value specified by gpwrap_lag_gps. Default is 6100 5 minutes. 6101 6102 rcutorture.irqreader= [KNL] 6103 Run RCU readers from irq handlers, or, more 6104 accurately, from a timer handler. Not all RCU 6105 flavors take kindly to this sort of thing. 6106 6107 rcutorture.leakpointer= [KNL] 6108 Leak an RCU-protected pointer out of the reader. 6109 This can of course result in splats, and is 6110 intended to test the ability of things like 6111 CONFIG_RCU_STRICT_GRACE_PERIOD=y to detect 6112 such leaks. 6113 6114 rcutorture.n_barrier_cbs= [KNL] 6115 Set callbacks/threads for rcu_barrier() testing. 6116 6117 rcutorture.nfakewriters= [KNL] 6118 Set number of concurrent RCU writers. These just 6119 stress RCU, they don't participate in the actual 6120 test, hence the "fake". 6121 6122 rcutorture.nocbs_nthreads= [KNL] 6123 Set number of RCU callback-offload togglers. 6124 Zero (the default) disables toggling. 6125 6126 rcutorture.nocbs_toggle= [KNL] 6127 Set the delay in milliseconds between successive 6128 callback-offload toggling attempts. 6129 6130 rcutorture.nreaders= [KNL] 6131 Set number of RCU readers. The value -1 selects 6132 N-1, where N is the number of CPUs. A value 6133 "n" less than -1 selects N-n-2, where N is again 6134 the number of CPUs. For example, -2 selects N 6135 (the number of CPUs), -3 selects N+1, and so on. 6136 6137 rcutorture.nwriters= [KNL] 6138 Set number of RCU writers, which must be either 6139 zero or one. For additional writers, use instead 6140 the rcutorture.nfakewriters parameter. 6141 6142 rcutorture.object_debug= [KNL] 6143 Enable debug-object double-call_rcu() testing. 6144 6145 rcutorture.onoff_holdoff= [KNL] 6146 Set time (s) after boot for CPU-hotplug testing. 6147 6148 rcutorture.onoff_interval= [KNL] 6149 Set time (jiffies) between CPU-hotplug operations, 6150 or zero to disable CPU-hotplug testing. 6151 6152 rcutorture.preempt_duration= [KNL] 6153 Set duration (in milliseconds) of preemptions 6154 by a high-priority FIFO real-time task. Set to 6155 zero (the default) to disable. The CPUs to 6156 preempt are selected randomly from the set that 6157 are online at a given point in time. Races with 6158 CPUs going offline are ignored, with that attempt 6159 at preemption skipped. 6160 6161 rcutorture.preempt_interval= [KNL] 6162 Set interval (in milliseconds, defaulting to one 6163 second) between preemptions by a high-priority 6164 FIFO real-time task. This delay is mediated 6165 by an hrtimer and is further fuzzed to avoid 6166 inadvertent synchronizations. 6167 6168 rcutorture.read_exit_burst= [KNL] 6169 The number of times in a given read-then-exit 6170 episode that a set of read-then-exit kthreads 6171 is spawned. 6172 6173 rcutorture.read_exit_delay= [KNL] 6174 The delay, in seconds, between successive 6175 read-then-exit testing episodes. 6176 6177 rcutorture.reader_flavor= [KNL] 6178 A bit mask indicating which readers to use. 6179 If there is more than one bit set, the readers 6180 are entered from low-order bit up, and are 6181 exited in the opposite order. For SRCU, the 6182 0x1 bit is normal readers, 0x2 NMI-safe readers, 6183 and 0x4 light-weight readers. 6184 6185 rcutorture.shuffle_interval= [KNL] 6186 Set task-shuffle interval (s). Shuffling tasks 6187 allows some CPUs to go into dyntick-idle mode 6188 during the rcutorture test. 6189 6190 rcutorture.shutdown_secs= [KNL] 6191 Set time (s) after boot system shutdown. This 6192 is useful for hands-off automated testing. 6193 6194 rcutorture.stall_cpu= [KNL] 6195 Duration of CPU stall (s) to test RCU CPU stall 6196 warnings, zero to disable. 6197 6198 rcutorture.stall_cpu_block= [KNL] 6199 Sleep while stalling if set. This will result 6200 in warnings from preemptible RCU in addition to 6201 any other stall-related activity. Note that 6202 in kernels built with CONFIG_PREEMPTION=n and 6203 CONFIG_PREEMPT_COUNT=y, this parameter will 6204 cause the CPU to pass through a quiescent state. 6205 Given CONFIG_PREEMPTION=n, this will suppress 6206 RCU CPU stall warnings, but will instead result 6207 in scheduling-while-atomic splats. 6208 6209 Use of this module parameter results in splats. 6210 6211 6212 rcutorture.stall_cpu_holdoff= [KNL] 6213 Time to wait (s) after boot before inducing stall. 6214 6215 rcutorture.stall_cpu_irqsoff= [KNL] 6216 Disable interrupts while stalling if set, but only 6217 on the first stall in the set. 6218 6219 rcutorture.stall_cpu_repeat= [KNL] 6220 Number of times to repeat the stall sequence, 6221 so that rcutorture.stall_cpu_repeat=3 will result 6222 in four stall sequences. 6223 6224 rcutorture.stall_gp_kthread= [KNL] 6225 Duration (s) of forced sleep within RCU 6226 grace-period kthread to test RCU CPU stall 6227 warnings, zero to disable. If both stall_cpu 6228 and stall_gp_kthread are specified, the 6229 kthread is starved first, then the CPU. 6230 6231 rcutorture.stall_only= [KNL] 6232 Shut off all rcutorture kthreads other than the 6233 RCU CPU stall-warning test kthreads. The purpose 6234 of this is to test production applictions' 6235 reactions to CPU stalls, and with minimal 6236 additional overhead. Or you can omit the 6237 stall-warning tests as well and get a heavy 6238 no-op, your choice! 6239 6240 rcutorture.stat_interval= [KNL] 6241 Time (s) between statistics printk()s. 6242 6243 rcutorture.stutter= [KNL] 6244 Time (s) to stutter testing, for example, specifying 6245 five seconds causes the test to run for five seconds, 6246 wait for five seconds, and so on. This tests RCU's 6247 ability to transition abruptly to and from idle. 6248 6249 rcutorture.test_boost= [KNL] 6250 Test RCU priority boosting? 0=no, 1=maybe, 2=yes. 6251 "Maybe" means test if the RCU implementation 6252 under test support RCU priority boosting. 6253 6254 rcutorture.test_boost_duration= [KNL] 6255 Duration (s) of each individual boost test. 6256 6257 rcutorture.test_boost_holdoff= [KNL] 6258 Holdoff time (s) from start of test to the start 6259 of RCU priority-boost testing. Defaults to zero, 6260 that is, no holdoff. 6261 6262 rcutorture.test_boost_interval= [KNL] 6263 Interval (s) between each boost test. 6264 6265 rcutorture.test_no_idle_hz= [KNL] 6266 Test RCU's dyntick-idle handling. See also the 6267 rcutorture.shuffle_interval parameter. 6268 6269 rcutorture.torture_type= [KNL] 6270 Specify the RCU implementation to test. 6271 6272 rcutorture.verbose= [KNL] 6273 Enable additional printk() statements. 6274 6275 rcupdate.rcu_cpu_stall_ftrace_dump= [KNL] 6276 Dump ftrace buffer after reporting RCU CPU 6277 stall warning. 6278 6279 rcupdate.rcu_cpu_stall_notifiers= [KNL] 6280 Provide RCU CPU stall notifiers, but see the 6281 warnings in the RCU_CPU_STALL_NOTIFIER Kconfig 6282 option's help text. TL;DR: You almost certainly 6283 do not want rcupdate.rcu_cpu_stall_notifiers. 6284 6285 rcupdate.rcu_cpu_stall_suppress= [KNL] 6286 Suppress RCU CPU stall warning messages. 6287 6288 rcupdate.rcu_cpu_stall_suppress_at_boot= [KNL] 6289 Suppress RCU CPU stall warning messages and 6290 rcutorture writer stall warnings that occur 6291 during early boot, that is, during the time 6292 before the init task is spawned. 6293 6294 rcupdate.rcu_cpu_stall_timeout= [KNL] 6295 Set timeout for RCU CPU stall warning messages. 6296 The value is in seconds and the maximum allowed 6297 value is 300 seconds. 6298 6299 rcupdate.rcu_exp_cpu_stall_timeout= [KNL] 6300 Set timeout for expedited RCU CPU stall warning 6301 messages. The value is in milliseconds 6302 and the maximum allowed value is 21000 6303 milliseconds. Please note that this value is 6304 adjusted to an arch timer tick resolution. 6305 Setting this to zero causes the value from 6306 rcupdate.rcu_cpu_stall_timeout to be used (after 6307 conversion from seconds to milliseconds). 6308 6309 rcupdate.rcu_cpu_stall_cputime= [KNL] 6310 Provide statistics on the cputime and count of 6311 interrupts and tasks during the sampling period. For 6312 multiple continuous RCU stalls, all sampling periods 6313 begin at half of the first RCU stall timeout. 6314 6315 rcupdate.rcu_exp_stall_task_details= [KNL] 6316 Print stack dumps of any tasks blocking the 6317 current expedited RCU grace period during an 6318 expedited RCU CPU stall warning. 6319 6320 rcupdate.rcu_expedited= [KNL] 6321 Use expedited grace-period primitives, for 6322 example, synchronize_rcu_expedited() instead 6323 of synchronize_rcu(). This reduces latency, 6324 but can increase CPU utilization, degrade 6325 real-time latency, and degrade energy efficiency. 6326 No effect on CONFIG_TINY_RCU kernels. 6327 6328 rcupdate.rcu_normal= [KNL] 6329 Use only normal grace-period primitives, 6330 for example, synchronize_rcu() instead of 6331 synchronize_rcu_expedited(). This improves 6332 real-time latency, CPU utilization, and 6333 energy efficiency, but can expose users to 6334 increased grace-period latency. This parameter 6335 overrides rcupdate.rcu_expedited. No effect on 6336 CONFIG_TINY_RCU kernels. 6337 6338 rcupdate.rcu_normal_after_boot= [KNL] 6339 Once boot has completed (that is, after 6340 rcu_end_inkernel_boot() has been invoked), use 6341 only normal grace-period primitives. No effect 6342 on CONFIG_TINY_RCU kernels. 6343 6344 But note that CONFIG_PREEMPT_RT=y kernels enables 6345 this kernel boot parameter, forcibly setting 6346 it to the value one, that is, converting any 6347 post-boot attempt at an expedited RCU grace 6348 period to instead use normal non-expedited 6349 grace-period processing. 6350 6351 rcupdate.rcu_task_collapse_lim= [KNL] 6352 Set the maximum number of callbacks present 6353 at the beginning of a grace period that allows 6354 the RCU Tasks flavors to collapse back to using 6355 a single callback queue. This switching only 6356 occurs when rcupdate.rcu_task_enqueue_lim is 6357 set to the default value of -1. 6358 6359 rcupdate.rcu_task_contend_lim= [KNL] 6360 Set the minimum number of callback-queuing-time 6361 lock-contention events per jiffy required to 6362 cause the RCU Tasks flavors to switch to per-CPU 6363 callback queuing. This switching only occurs 6364 when rcupdate.rcu_task_enqueue_lim is set to 6365 the default value of -1. 6366 6367 rcupdate.rcu_task_enqueue_lim= [KNL] 6368 Set the number of callback queues to use for the 6369 RCU Tasks family of RCU flavors. The default 6370 of -1 allows this to be automatically (and 6371 dynamically) adjusted. This parameter is intended 6372 for use in testing. 6373 6374 rcupdate.rcu_task_lazy_lim= [KNL] 6375 Number of callbacks on a given CPU that will 6376 cancel laziness on that CPU. Use -1 to disable 6377 cancellation of laziness, but be advised that 6378 doing so increases the danger of OOM due to 6379 callback flooding. 6380 6381 rcupdate.rcu_task_stall_info= [KNL] 6382 Set initial timeout in jiffies for RCU task stall 6383 informational messages, which give some indication 6384 of the problem for those not patient enough to 6385 wait for ten minutes. Informational messages are 6386 only printed prior to the stall-warning message 6387 for a given grace period. Disable with a value 6388 less than or equal to zero. Defaults to ten 6389 seconds. A change in value does not take effect 6390 until the beginning of the next grace period. 6391 6392 rcupdate.rcu_task_stall_info_mult= [KNL] 6393 Multiplier for time interval between successive 6394 RCU task stall informational messages for a given 6395 RCU tasks grace period. This value is clamped 6396 to one through ten, inclusive. It defaults to 6397 the value three, so that the first informational 6398 message is printed 10 seconds into the grace 6399 period, the second at 40 seconds, the third at 6400 160 seconds, and then the stall warning at 600 6401 seconds would prevent a fourth at 640 seconds. 6402 6403 rcupdate.rcu_task_stall_timeout= [KNL] 6404 Set timeout in jiffies for RCU task stall 6405 warning messages. Disable with a value less 6406 than or equal to zero. Defaults to ten minutes. 6407 A change in value does not take effect until 6408 the beginning of the next grace period. 6409 6410 rcupdate.rcu_tasks_lazy_ms= [KNL] 6411 Set timeout in milliseconds RCU Tasks asynchronous 6412 callback batching for call_rcu_tasks(). 6413 A negative value will take the default. A value 6414 of zero will disable batching. Batching is 6415 always disabled for synchronize_rcu_tasks(). 6416 6417 rcupdate.rcu_self_test= [KNL] 6418 Run the RCU early boot self tests 6419 6420 rdinit= [KNL] 6421 Format: <full_path> 6422 Run specified binary instead of /init from the ramdisk, 6423 used for early userspace startup. See initrd. 6424 6425 rdrand= [X86,EARLY] 6426 force - Override the decision by the kernel to hide the 6427 advertisement of RDRAND support (this affects 6428 certain AMD processors because of buggy BIOS 6429 support, specifically around the suspend/resume 6430 path). 6431 6432 rdt= [HW,X86,RDT] 6433 Turn on/off individual RDT features. List is: 6434 cmt, mbmtotal, mbmlocal, l3cat, l3cdp, l2cat, l2cdp, 6435 mba, smba, bmec, abmc, sdciae, energy[:guid], 6436 perf[:guid]. 6437 E.g. to turn on cmt and turn off mba use: 6438 rdt=cmt,!mba 6439 To turn off all energy telemetry monitoring and ensure that 6440 perf telemetry monitoring associated with guid 0x12345 6441 is enabled use: 6442 rdt=!energy,perf:0x12345 6443 6444 reboot= [KNL] 6445 Format (x86 or x86_64): 6446 [w[arm] | c[old] | h[ard] | s[oft] | g[pio]] | d[efault] \ 6447 [[,]s[mp]####] \ 6448 [[,]b[ios] | a[cpi] | k[bd] | t[riple] | e[fi] | p[ci]] \ 6449 [[,]f[orce]] 6450 Where reboot_mode is one of warm (soft) or cold (hard) or gpio 6451 (prefix with 'panic_' to set mode for panic 6452 reboot only), 6453 reboot_type is one of bios, acpi, kbd, triple, efi, or pci, 6454 reboot_force is either force or not specified, 6455 reboot_cpu is s[mp]#### with #### being the processor 6456 to be used for rebooting. 6457 6458 acpi 6459 Use the ACPI RESET_REG in the FADT. If ACPI is not 6460 configured or the ACPI reset does not work, the reboot 6461 path attempts the reset using the keyboard controller. 6462 6463 bios 6464 Use the CPU reboot vector for warm reset 6465 6466 cold 6467 Set the cold reboot flag 6468 6469 default 6470 There are some built-in platform specific "quirks" 6471 - you may see: "reboot: <name> series board detected. 6472 Selecting <type> for reboots." In the case where you 6473 think the quirk is in error (e.g. you have newer BIOS, 6474 or newer board) using this option will ignore the 6475 built-in quirk table, and use the generic default 6476 reboot actions. 6477 6478 efi 6479 Use efi reset_system runtime service. If EFI is not 6480 configured or the EFI reset does not work, the reboot 6481 path attempts the reset using the keyboard controller. 6482 6483 force 6484 Don't stop other CPUs on reboot. This can make reboot 6485 more reliable in some cases. 6486 6487 kbd 6488 Use the keyboard controller. cold reset (default) 6489 6490 pci 6491 Use a write to the PCI config space register 0xcf9 to 6492 trigger reboot. 6493 6494 triple 6495 Force a triple fault (init) 6496 6497 warm 6498 Don't set the cold reboot flag 6499 6500 Using warm reset will be much faster especially on big 6501 memory systems because the BIOS will not go through 6502 the memory check. Disadvantage is that not all 6503 hardware will be completely reinitialized on reboot so 6504 there may be boot problems on some systems. 6505 6506 6507 refscale.holdoff= [KNL] 6508 Set test-start holdoff period. The purpose of 6509 this parameter is to delay the start of the 6510 test until boot completes in order to avoid 6511 interference. 6512 6513 refscale.lookup_instances= [KNL] 6514 Number of data elements to use for the forms of 6515 SLAB_TYPESAFE_BY_RCU testing. A negative number 6516 is negated and multiplied by nr_cpu_ids, while 6517 zero specifies nr_cpu_ids. 6518 6519 refscale.loops= [KNL] 6520 Set the number of loops over the synchronization 6521 primitive under test. Increasing this number 6522 reduces noise due to loop start/end overhead, 6523 but the default has already reduced the per-pass 6524 noise to a handful of picoseconds on ca. 2020 6525 x86 laptops. 6526 6527 refscale.nreaders= [KNL] 6528 Set number of readers. The default value of -1 6529 selects N, where N is roughly 75% of the number 6530 of CPUs. A value of zero is an interesting choice. 6531 6532 refscale.nruns= [KNL] 6533 Set number of runs, each of which is dumped onto 6534 the console log. 6535 6536 refscale.readdelay= [KNL] 6537 Set the read-side critical-section duration, 6538 measured in microseconds. 6539 6540 refscale.scale_type= [KNL] 6541 Specify the read-protection implementation to test. 6542 6543 refscale.shutdown= [KNL] 6544 Shut down the system at the end of the performance 6545 test. This defaults to 1 (shut it down) when 6546 refscale is built into the kernel and to 0 (leave 6547 it running) when refscale is built as a module. 6548 6549 refscale.verbose= [KNL] 6550 Enable additional printk() statements. 6551 6552 refscale.verbose_batched= [KNL] 6553 Batch the additional printk() statements. If zero 6554 (the default) or negative, print everything. Otherwise, 6555 print every Nth verbose statement, where N is the value 6556 specified. 6557 6558 regulator_ignore_unused 6559 [REGULATOR] 6560 Prevents regulator framework from disabling regulators 6561 that are unused, due no driver claiming them. This may 6562 be useful for debug and development, but should not be 6563 needed on a platform with proper driver support. 6564 6565 relax_domain_level= 6566 [KNL, SMP] Set scheduler's default relax_domain_level. 6567 See Documentation/admin-guide/cgroup-v1/cpusets.rst. 6568 6569 reserve= [KNL,BUGS] Force kernel to ignore I/O ports or memory 6570 Format: <base1>,<size1>[,<base2>,<size2>,...] 6571 Reserve I/O ports or memory so the kernel won't use 6572 them. If <base> is less than 0x10000, the region 6573 is assumed to be I/O ports; otherwise it is memory. 6574 6575 reserve_mem= [RAM] 6576 Format: nn[KMG]:<align>:<label> 6577 Reserve physical memory and label it with a name that 6578 other subsystems can use to access it. This is typically 6579 used for systems that do not wipe the RAM, and this command 6580 line will try to reserve the same physical memory on 6581 soft reboots. Note, it is not guaranteed to be the same 6582 location. For example, if anything about the system changes 6583 or if booting a different kernel. It can also fail if KASLR 6584 places the kernel at the location of where the RAM reservation 6585 was from a previous boot, the new reservation will be at a 6586 different location. 6587 Any subsystem using this feature must add a way to verify 6588 that the contents of the physical memory is from a previous 6589 boot, as there may be cases where the memory will not be 6590 located at the same location. 6591 6592 The format is size:align:label for example, to request 6593 12 megabytes of 4096 alignment for ramoops: 6594 6595 reserve_mem=12M:4096:oops ramoops.mem_name=oops 6596 6597 reservetop= [X86-32,EARLY] 6598 Format: nn[KMG] 6599 Reserves a hole at the top of the kernel virtual 6600 address space. 6601 6602 reset_devices [KNL] Force drivers to reset the underlying device 6603 during initialization. 6604 6605 resume= [SWSUSP] 6606 Specify the partition device for software suspend 6607 Format: 6608 {/dev/<dev> | PARTUUID=<uuid> | <int>:<int> | <hex>} 6609 6610 resume_offset= [SWSUSP] 6611 Specify the offset from the beginning of the partition 6612 given by "resume=" at which the swap header is located, 6613 in <PAGE_SIZE> units (needed only for swap files). 6614 See Documentation/power/swsusp-and-swap-files.rst 6615 6616 resumedelay= [HIBERNATION] Delay (in seconds) to pause before attempting to 6617 read the resume files 6618 6619 resumewait [HIBERNATION] Wait (indefinitely) for resume device to show up. 6620 Useful for devices that are detected asynchronously 6621 (e.g. USB and MMC devices). 6622 6623 retain_initrd [RAM] Keep initrd memory after extraction. After boot, it will 6624 be accessible via /sys/firmware/initrd. 6625 6626 retbleed= [X86] Control mitigation of RETBleed (Arbitrary 6627 Speculative Code Execution with Return Instructions) 6628 vulnerability. 6629 6630 AMD-based UNRET and IBPB mitigations alone do not stop 6631 sibling threads from influencing the predictions of other 6632 sibling threads. For that reason, STIBP is used on pro- 6633 cessors that support it, and mitigate SMT on processors 6634 that don't. 6635 6636 off - no mitigation 6637 auto - automatically select a mitigation 6638 auto,nosmt - automatically select a mitigation, 6639 disabling SMT if necessary for 6640 the full mitigation (only on Zen1 6641 and older without STIBP). 6642 ibpb - On AMD, mitigate short speculation 6643 windows on basic block boundaries too. 6644 Safe, highest perf impact. It also 6645 enables STIBP if present. Not suitable 6646 on Intel. 6647 ibpb,nosmt - Like "ibpb" above but will disable SMT 6648 when STIBP is not available. This is 6649 the alternative for systems which do not 6650 have STIBP. 6651 unret - Force enable untrained return thunks, 6652 only effective on AMD f15h-f17h based 6653 systems. 6654 unret,nosmt - Like unret, but will disable SMT when STIBP 6655 is not available. This is the alternative for 6656 systems which do not have STIBP. 6657 6658 Selecting 'auto' will choose a mitigation method at run 6659 time according to the CPU. 6660 6661 Not specifying this option is equivalent to retbleed=auto. 6662 6663 rfkill.default_state= 6664 0 "airplane mode". All wifi, bluetooth, wimax, gps, fm, 6665 etc. communication is blocked by default. 6666 1 Unblocked. 6667 6668 rfkill.master_switch_mode= 6669 0 The "airplane mode" button does nothing. 6670 1 The "airplane mode" button toggles between everything 6671 blocked and the previous configuration. 6672 2 The "airplane mode" button toggles between everything 6673 blocked and everything unblocked. 6674 6675 ring3mwait=disable 6676 [KNL] Disable ring 3 MONITOR/MWAIT feature on supported 6677 CPUs. 6678 6679 riscv_isa_fallback [RISCV,EARLY] 6680 When CONFIG_RISCV_ISA_FALLBACK is not enabled, permit 6681 falling back to detecting extension support by parsing 6682 "riscv,isa" property on devicetree systems when the 6683 replacement properties are not found. See the Kconfig 6684 entry for RISCV_ISA_FALLBACK. 6685 6686 riscv_nousercfi= 6687 all Disable user CFI ABI to userspace even if cpu extension 6688 are available. 6689 bcfi Disable user backward CFI ABI to userspace even if 6690 the shadow stack extension is available. 6691 fcfi Disable user forward CFI ABI to userspace even if the 6692 landing pad extension is available. 6693 6694 ro [KNL] Mount root device read-only on boot 6695 6696 rodata= [KNL,EARLY] 6697 on Mark read-only kernel memory as read-only (default). 6698 off Leave read-only kernel memory writable for debugging. 6699 noalias Mark read-only kernel memory as read-only but retain 6700 writable aliases in the direct map for regions outside 6701 of the kernel image. [arm64] 6702 6703 rockchip.usb_uart 6704 [EARLY] 6705 Enable the uart passthrough on the designated usb port 6706 on Rockchip SoCs. When active, the signals of the 6707 debug-uart get routed to the D+ and D- pins of the usb 6708 port and the regular usb controller gets disabled. 6709 6710 root= [KNL] Root filesystem 6711 Usually this is a block device specifier of some kind, 6712 see the early_lookup_bdev comment in 6713 block/early-lookup.c for details. 6714 Alternatively this can be "ram" for the legacy initial 6715 ramdisk, "nfs" and "cifs" for root on a network file 6716 system, or "mtd" and "ubi" for mounting from raw flash. 6717 6718 rootdelay= [KNL] Delay (in seconds) to pause before attempting to 6719 mount the root filesystem 6720 6721 rootflags= [KNL] Set root filesystem mount option string 6722 6723 rseq_slice_ext= [KNL] RSEQ based time slice extension 6724 Format: boolean 6725 Control enablement of RSEQ based time slice extension. 6726 Default is 'on'. 6727 6728 initramfs_options= [KNL] 6729 Specify mount options for the initramfs mount. 6730 6731 rootfstype= [KNL] Set root filesystem type 6732 6733 rootwait [KNL] Wait (indefinitely) for root device to show up. 6734 Useful for devices that are detected asynchronously 6735 (e.g. USB and MMC devices). 6736 6737 rootwait= [KNL] Maximum time (in seconds) to wait for root device 6738 to show up before attempting to mount the root 6739 filesystem. 6740 6741 rproc_mem=nn[KMG][@address] 6742 [KNL,ARM,CMA] Remoteproc physical memory block. 6743 Memory area to be used by remote processor image, 6744 managed by CMA. 6745 6746 rseq_debug= [KNL] Enable or disable restartable sequence 6747 debug mode. Defaults to CONFIG_RSEQ_DEBUG_DEFAULT_ENABLE. 6748 Format: <bool> 6749 6750 rt_group_sched= [KNL] Enable or disable SCHED_RR/FIFO group scheduling 6751 when CONFIG_RT_GROUP_SCHED=y. Defaults to 6752 !CONFIG_RT_GROUP_SCHED_DEFAULT_DISABLED. 6753 Format: <bool> 6754 6755 rw [KNL] Mount root device read-write on boot 6756 6757 S [KNL] Run init in single mode 6758 6759 s390_iommu= [HW,S390] 6760 Set s390 IOTLB flushing mode 6761 strict 6762 With strict flushing every unmap operation will result 6763 in an IOTLB flush. Default is lazy flushing before 6764 reuse, which is faster. Deprecated, equivalent to 6765 iommu.strict=1. 6766 6767 s390_iommu_aperture= [KNL,S390] 6768 Specifies the size of the per device DMA address space 6769 accessible through the DMA and IOMMU APIs as a decimal 6770 factor of the size of main memory. 6771 The default is 1 meaning that one can concurrently use 6772 as many DMA addresses as physical memory is installed, 6773 if supported by hardware, and thus map all of memory 6774 once. With a value of 2 one can map all of memory twice 6775 and so on. As a special case a factor of 0 imposes no 6776 restrictions other than those given by hardware at the 6777 cost of significant additional memory use for tables. 6778 6779 6780 sched_proxy_exec= [KNL] 6781 Enables or disables "proxy execution" style 6782 solution to mutex-based priority inversion. 6783 Format: <bool> 6784 6785 sched_verbose [KNL,EARLY] Enables verbose scheduler debug messages. 6786 6787 schedstats= [KNL,X86] Enable or disable scheduled statistics. 6788 Allowed values are enable and disable. This feature 6789 incurs a small amount of overhead in the scheduler 6790 but is useful for debugging and performance tuning. 6791 6792 sched_thermal_decay_shift= 6793 [Deprecated] 6794 [KNL, SMP] Set a decay shift for scheduler thermal 6795 pressure signal. Thermal pressure signal follows the 6796 default decay period of other scheduler pelt 6797 signals(usually 32 ms but configurable). Setting 6798 sched_thermal_decay_shift will left shift the decay 6799 period for the thermal pressure signal by the shift 6800 value. 6801 i.e. with the default pelt decay period of 32 ms 6802 sched_thermal_decay_shift thermal pressure decay pr 6803 1 64 ms 6804 2 128 ms 6805 and so on. 6806 Format: integer between 0 and 10 6807 Default is 0. 6808 6809 scftorture.holdoff= [KNL] 6810 Number of seconds to hold off before starting 6811 test. Defaults to zero for module insertion and 6812 to 10 seconds for built-in smp_call_function() 6813 tests. 6814 6815 scftorture.longwait= [KNL] 6816 Request ridiculously long waits randomly selected 6817 up to the chosen limit in seconds. Zero (the 6818 default) disables this feature. Please note 6819 that requesting even small non-zero numbers of 6820 seconds can result in RCU CPU stall warnings, 6821 softlockup complaints, and so on. 6822 6823 scftorture.nthreads= [KNL] 6824 Number of kthreads to spawn to invoke the 6825 smp_call_function() family of functions. 6826 The default of -1 specifies a number of kthreads 6827 equal to the number of CPUs. 6828 6829 scftorture.onoff_holdoff= [KNL] 6830 Number seconds to wait after the start of the 6831 test before initiating CPU-hotplug operations. 6832 6833 scftorture.onoff_interval= [KNL] 6834 Number seconds to wait between successive 6835 CPU-hotplug operations. Specifying zero (which 6836 is the default) disables CPU-hotplug operations. 6837 6838 scftorture.shutdown_secs= [KNL] 6839 The number of seconds following the start of the 6840 test after which to shut down the system. The 6841 default of zero avoids shutting down the system. 6842 Non-zero values are useful for automated tests. 6843 6844 scftorture.stat_interval= [KNL] 6845 The number of seconds between outputting the 6846 current test statistics to the console. A value 6847 of zero disables statistics output. 6848 6849 scftorture.stutter_cpus= [KNL] 6850 The number of jiffies to wait between each change 6851 to the set of CPUs under test. 6852 6853 scftorture.use_cpus_read_lock= [KNL] 6854 Use use_cpus_read_lock() instead of the default 6855 preempt_disable() to disable CPU hotplug 6856 while invoking one of the smp_call_function*() 6857 functions. 6858 6859 scftorture.verbose= [KNL] 6860 Enable additional printk() statements. 6861 6862 scftorture.weight_single= [KNL] 6863 The probability weighting to use for the 6864 smp_call_function_single() function with a zero 6865 "wait" parameter. A value of -1 selects the 6866 default if all other weights are -1. However, 6867 if at least one weight has some other value, a 6868 value of -1 will instead select a weight of zero. 6869 6870 scftorture.weight_single_wait= [KNL] 6871 The probability weighting to use for the 6872 smp_call_function_single() function with a 6873 non-zero "wait" parameter. See weight_single. 6874 6875 scftorture.weight_many= [KNL] 6876 The probability weighting to use for the 6877 smp_call_function_many() function with a zero 6878 "wait" parameter. See weight_single. 6879 Note well that setting a high probability for 6880 this weighting can place serious IPI load 6881 on the system. 6882 6883 scftorture.weight_many_wait= [KNL] 6884 The probability weighting to use for the 6885 smp_call_function_many() function with a 6886 non-zero "wait" parameter. See weight_single 6887 and weight_many. 6888 6889 scftorture.weight_all= [KNL] 6890 The probability weighting to use for the 6891 smp_call_function_all() function with a zero 6892 "wait" parameter. See weight_single and 6893 weight_many. 6894 6895 scftorture.weight_all_wait= [KNL] 6896 The probability weighting to use for the 6897 smp_call_function_all() function with a 6898 non-zero "wait" parameter. See weight_single 6899 and weight_many. 6900 6901 sdw_mclk_divider=[SDW] 6902 Specify the MCLK divider for Intel SoundWire buses in 6903 case the BIOS does not provide the clock rate properly. 6904 6905 skew_tick= [KNL,EARLY] Offset the periodic timer tick per cpu to mitigate 6906 xtime_lock contention on larger systems, and/or RCU lock 6907 contention on all systems with CONFIG_MAXSMP set. 6908 Format: { "0" | "1" } 6909 0 -- disable. (may be 1 via CONFIG_CMDLINE="skew_tick=1") 6910 1 -- enable. 6911 Note: increases power consumption, thus should only be 6912 enabled if running jitter sensitive (HPC/RT) workloads. 6913 6914 security= [SECURITY] Choose a legacy "major" security module to 6915 enable at boot. This has been deprecated by the 6916 "lsm=" parameter. 6917 6918 selinux= [SELINUX] Disable or enable SELinux at boot time. 6919 Format: { "0" | "1" } 6920 See security/selinux/Kconfig help text. 6921 0 -- disable. 6922 1 -- enable. 6923 Default value is 1. 6924 6925 serialnumber [BUGS=X86-32] 6926 6927 sev=option[,option...] [X86-64] 6928 6929 debug 6930 Enable debug messages. 6931 6932 nosnp 6933 Do not enable SEV-SNP (applies to host/hypervisor 6934 only). Setting 'nosnp' avoids the RMP check overhead 6935 in memory accesses when users do not want to run 6936 SEV-SNP guests. 6937 6938 shapers= [NET] 6939 Maximal number of shapers. 6940 6941 show_lapic= [APIC,X86] Advanced Programmable Interrupt Controller 6942 Limit apic dumping. The parameter defines the maximal 6943 number of local apics being dumped. Also it is possible 6944 to set it to "all" by meaning -- no limit here. 6945 Format: { 1 (default) | 2 | ... | all }. 6946 The parameter valid if only apic=debug or 6947 apic=verbose is specified. 6948 Example: apic=debug show_lapic=all 6949 6950 slab_debug[=options[,slabs][;[options[,slabs]]...]] [MM] 6951 Enabling slab_debug allows one to determine the 6952 culprit if slab objects become corrupted. Enabling 6953 slab_debug can create guard zones around objects and 6954 may poison objects when not in use. Also tracks the 6955 last alloc / free. For more information see 6956 Documentation/admin-guide/mm/slab.rst. 6957 (slub_debug legacy name also accepted for now) 6958 6959 Using this option implies the "no_hash_pointers" 6960 option which can be undone by adding the 6961 "hash_pointers=always" option. 6962 6963 slab_max_order= [MM] 6964 Determines the maximum allowed order for slabs. 6965 A high setting may cause OOMs due to memory 6966 fragmentation. For more information see 6967 Documentation/admin-guide/mm/slab.rst. 6968 (slub_max_order legacy name also accepted for now) 6969 6970 slab_merge [MM] 6971 Enable merging of slabs with similar size when the 6972 kernel is built without CONFIG_SLAB_MERGE_DEFAULT. 6973 (slub_merge legacy name also accepted for now) 6974 6975 slab_min_objects= [MM] 6976 The minimum number of objects per slab. SLUB will 6977 increase the slab order up to slab_max_order to 6978 generate a sufficiently large slab able to contain 6979 the number of objects indicated. The higher the number 6980 of objects the smaller the overhead of tracking slabs 6981 and the less frequently locks need to be acquired. 6982 For more information see 6983 Documentation/admin-guide/mm/slab.rst. 6984 (slub_min_objects legacy name also accepted for now) 6985 6986 slab_min_order= [MM] 6987 Determines the minimum page order for slabs. Must be 6988 lower or equal to slab_max_order. For more information see 6989 Documentation/admin-guide/mm/slab.rst. 6990 (slub_min_order legacy name also accepted for now) 6991 6992 slab_nomerge [MM] 6993 Disable merging of slabs with similar size. May be 6994 necessary if there is some reason to distinguish 6995 allocs to different slabs, especially in hardened 6996 environments where the risk of heap overflows and 6997 layout control by attackers can usually be 6998 frustrated by disabling merging. This will reduce 6999 most of the exposure of a heap attack to a single 7000 cache (risks via metadata attacks are mostly 7001 unchanged). Debug options disable merging on their 7002 own. 7003 For more information see 7004 Documentation/admin-guide/mm/slab.rst. 7005 (slub_nomerge legacy name also accepted for now) 7006 7007 slab_strict_numa [MM] 7008 Support memory policies on a per object level 7009 in the slab allocator. The default is for memory 7010 policies to be applied at the folio level when 7011 a new folio is needed or a partial folio is 7012 retrieved from the lists. Increases overhead 7013 in the slab fastpaths but gains more accurate 7014 NUMA kernel object placement which helps with slow 7015 interconnects in NUMA systems. 7016 7017 slram= [HW,MTD] 7018 7019 smart2= [HW] 7020 Format: <io1>[,<io2>[,...,<io8>]] 7021 7022 smp.csd_lock_timeout= [KNL] 7023 Specify the period of time in milliseconds 7024 that smp_call_function() and friends will wait 7025 for a CPU to release the CSD lock. This is 7026 useful when diagnosing bugs involving CPUs 7027 disabling interrupts for extended periods 7028 of time. Defaults to 5,000 milliseconds, and 7029 setting a value of zero disables this feature. 7030 This feature may be more efficiently disabled 7031 using the csdlock_debug- kernel parameter. 7032 7033 smp.panic_on_ipistall= [KNL] 7034 If a csd_lock_timeout extends for more than 7035 the specified number of milliseconds, panic the 7036 system. By default, let CSD-lock acquisition 7037 take as long as they take. Specifying 300,000 7038 for this value provides a 5-minute timeout. 7039 7040 smt= [KNL,MIPS,S390,EARLY] Set the maximum number of threads 7041 (logical CPUs) to use per physical CPU on systems 7042 capable of symmetric multithreading (SMT). Will 7043 be capped to the actual hardware limit. 7044 Format: <integer> 7045 Default: -1 (no limit) 7046 7047 softlockup_panic= 7048 [KNL] Should the soft-lockup detector generate panics. 7049 Format: <int> 7050 7051 A value of non-zero instructs the soft-lockup detector 7052 to panic the machine when a soft-lockup duration exceeds 7053 N thresholds. It is also controlled by the kernel.softlockup_panic 7054 sysctl and CONFIG_BOOTPARAM_SOFTLOCKUP_PANIC, which is the 7055 respective build-time switch to that functionality. 7056 7057 softlockup_all_cpu_backtrace= 7058 [KNL] Should the soft-lockup detector generate 7059 backtraces on all cpus. 7060 Format: 0 | 1 7061 7062 sonypi.*= [HW] Sony Programmable I/O Control Device driver 7063 See Documentation/admin-guide/laptops/sonypi.rst 7064 7065 spectre_bhi= [X86] Control mitigation of Branch History Injection 7066 (BHI) vulnerability. This setting affects the 7067 deployment of the HW BHI control and the SW BHB 7068 clearing sequence. 7069 7070 on - (default) Enable the HW or SW mitigation as 7071 needed. This protects the kernel from 7072 both syscalls and VMs. 7073 vmexit - On systems which don't have the HW mitigation 7074 available, enable the SW mitigation on vmexit 7075 ONLY. On such systems, the host kernel is 7076 protected from VM-originated BHI attacks, but 7077 may still be vulnerable to syscall attacks. 7078 off - Disable the mitigation. 7079 7080 spectre_v2= [X86,EARLY] Control mitigation of Spectre variant 2 7081 (indirect branch speculation) vulnerability. 7082 The default operation protects the kernel from 7083 user space attacks. 7084 7085 on - unconditionally enable, implies 7086 spectre_v2_user=on 7087 off - unconditionally disable, implies 7088 spectre_v2_user=off 7089 auto - kernel detects whether your CPU model is 7090 vulnerable 7091 7092 Selecting 'on' will, and 'auto' may, choose a 7093 mitigation method at run time according to the 7094 CPU, the available microcode, the setting of the 7095 CONFIG_MITIGATION_RETPOLINE configuration option, 7096 and the compiler with which the kernel was built. 7097 7098 Selecting 'on' will also enable the mitigation 7099 against user space to user space task attacks. 7100 Selecting specific mitigation does not force enable 7101 user mitigations. 7102 7103 Selecting 'off' will disable both the kernel and 7104 the user space protections. 7105 7106 Specific mitigations can also be selected manually: 7107 7108 retpoline - replace indirect branches 7109 retpoline,generic - Retpolines 7110 retpoline,lfence - LFENCE; indirect branch 7111 retpoline,amd - alias for retpoline,lfence 7112 eibrs - Enhanced/Auto IBRS 7113 eibrs,retpoline - Enhanced/Auto IBRS + Retpolines 7114 eibrs,lfence - Enhanced/Auto IBRS + LFENCE 7115 ibrs - use IBRS to protect kernel 7116 7117 Not specifying this option is equivalent to 7118 spectre_v2=auto. 7119 7120 spectre_v2_user= 7121 [X86] Control mitigation of Spectre variant 2 7122 (indirect branch speculation) vulnerability between 7123 user space tasks 7124 7125 on - Unconditionally enable mitigations. Is 7126 enforced by spectre_v2=on 7127 7128 off - Unconditionally disable mitigations. Is 7129 enforced by spectre_v2=off 7130 7131 prctl - Indirect branch speculation is enabled, 7132 but mitigation can be enabled via prctl 7133 per thread. The mitigation control state 7134 is inherited on fork. 7135 7136 prctl,ibpb 7137 - Like "prctl" above, but only STIBP is 7138 controlled per thread. IBPB is issued 7139 always when switching between different user 7140 space processes. 7141 7142 seccomp 7143 - Same as "prctl" above, but all seccomp 7144 threads will enable the mitigation unless 7145 they explicitly opt out. 7146 7147 seccomp,ibpb 7148 - Like "seccomp" above, but only STIBP is 7149 controlled per thread. IBPB is issued 7150 always when switching between different 7151 user space processes. 7152 7153 auto - Kernel selects the mitigation depending on 7154 the available CPU features and vulnerability. 7155 7156 Default mitigation: "prctl" 7157 7158 Not specifying this option is equivalent to 7159 spectre_v2_user=auto. 7160 7161 spec_rstack_overflow= 7162 [X86,EARLY] Control RAS overflow mitigation on AMD Zen CPUs 7163 7164 off - Disable mitigation 7165 microcode - Enable microcode mitigation only 7166 safe-ret - Enable sw-only safe RET mitigation (default) 7167 ibpb - Enable mitigation by issuing IBPB on 7168 kernel entry 7169 ibpb-vmexit - Issue IBPB only on VMEXIT 7170 (cloud-specific mitigation) 7171 7172 spec_store_bypass_disable= 7173 [HW,EARLY] Control Speculative Store Bypass (SSB) Disable mitigation 7174 (Speculative Store Bypass vulnerability) 7175 7176 Certain CPUs are vulnerable to an exploit against a 7177 a common industry wide performance optimization known 7178 as "Speculative Store Bypass" in which recent stores 7179 to the same memory location may not be observed by 7180 later loads during speculative execution. The idea 7181 is that such stores are unlikely and that they can 7182 be detected prior to instruction retirement at the 7183 end of a particular speculation execution window. 7184 7185 In vulnerable processors, the speculatively forwarded 7186 store can be used in a cache side channel attack, for 7187 example to read memory to which the attacker does not 7188 directly have access (e.g. inside sandboxed code). 7189 7190 This parameter controls whether the Speculative Store 7191 Bypass optimization is used. 7192 7193 On x86 the options are: 7194 7195 on - Unconditionally disable Speculative Store Bypass 7196 off - Unconditionally enable Speculative Store Bypass 7197 auto - Kernel detects whether the CPU model contains an 7198 implementation of Speculative Store Bypass and 7199 picks the most appropriate mitigation. If the 7200 CPU is not vulnerable, "off" is selected. If the 7201 CPU is vulnerable the default mitigation is 7202 architecture and Kconfig dependent. See below. 7203 prctl - Control Speculative Store Bypass per thread 7204 via prctl. Speculative Store Bypass is enabled 7205 for a process by default. The state of the control 7206 is inherited on fork. 7207 seccomp - Same as "prctl" above, but all seccomp threads 7208 will disable SSB unless they explicitly opt out. 7209 7210 Default mitigations: 7211 X86: "prctl" 7212 7213 On powerpc the options are: 7214 7215 on,auto - On Power8 and Power9 insert a store-forwarding 7216 barrier on kernel entry and exit. On Power7 7217 perform a software flush on kernel entry and 7218 exit. 7219 off - No action. 7220 7221 Not specifying this option is equivalent to 7222 spec_store_bypass_disable=auto. 7223 7224 split_llc= 7225 [X86,EARLY] Split the LLC N-ways 7226 7227 When set, the LLC is split this many ways by matching 7228 'core_id % n'. This is setup before SMP bringup and 7229 used during SMP bringup before it knows the full 7230 topology. If your core count doesn't nicely divide by 7231 the number given, you get to keep the pieces. 7232 7233 This is mostly a debug feature to emulate multiple LLCs 7234 on hardware that only have a single LLC. 7235 7236 split_lock_detect= 7237 [X86] Enable split lock detection or bus lock detection 7238 7239 When enabled (and if hardware support is present), atomic 7240 instructions that access data across cache line 7241 boundaries will result in an alignment check exception 7242 for split lock detection or a debug exception for 7243 bus lock detection. 7244 7245 off - not enabled 7246 7247 warn - the kernel will emit rate-limited warnings 7248 about applications triggering the #AC 7249 exception or the #DB exception. This mode is 7250 the default on CPUs that support split lock 7251 detection or bus lock detection. Default 7252 behavior is by #AC if both features are 7253 enabled in hardware. 7254 7255 fatal - the kernel will send SIGBUS to applications 7256 that trigger the #AC exception or the #DB 7257 exception. Default behavior is by #AC if 7258 both features are enabled in hardware. 7259 7260 ratelimit:N - 7261 Set system wide rate limit to N bus locks 7262 per second for bus lock detection. 7263 0 < N <= 1000. 7264 7265 N/A for split lock detection. 7266 7267 7268 If an #AC exception is hit in the kernel or in 7269 firmware (i.e. not while executing in user mode) 7270 the kernel will oops in either "warn" or "fatal" 7271 mode. 7272 7273 #DB exception for bus lock is triggered only when 7274 CPL > 0. 7275 7276 srbds= [X86,INTEL,EARLY] 7277 Control the Special Register Buffer Data Sampling 7278 (SRBDS) mitigation. 7279 7280 Certain CPUs are vulnerable to an MDS-like 7281 exploit which can leak bits from the random 7282 number generator. 7283 7284 By default, this issue is mitigated by 7285 microcode. However, the microcode fix can cause 7286 the RDRAND and RDSEED instructions to become 7287 much slower. Among other effects, this will 7288 result in reduced throughput from /dev/urandom. 7289 7290 The microcode mitigation can be disabled with 7291 the following option: 7292 7293 off: Disable mitigation and remove 7294 performance impact to RDRAND and RDSEED 7295 7296 srcutree.big_cpu_lim [KNL] 7297 Specifies the number of CPUs constituting a 7298 large system, such that srcu_struct structures 7299 should immediately allocate an srcu_node array. 7300 This kernel-boot parameter defaults to 128, 7301 but takes effect only when the low-order four 7302 bits of srcutree.convert_to_big is equal to 3 7303 (decide at boot). 7304 7305 srcutree.convert_to_big [KNL] 7306 Specifies under what conditions an SRCU tree 7307 srcu_struct structure will be converted to big 7308 form, that is, with an rcu_node tree: 7309 7310 0: Never. 7311 1: At init_srcu_struct() time. 7312 2: When rcutorture decides to. 7313 3: Decide at boot time (default). 7314 0x1X: Above plus if high contention. 7315 7316 Either way, the srcu_node tree will be sized based 7317 on the actual runtime number of CPUs (nr_cpu_ids) 7318 instead of the compile-time CONFIG_NR_CPUS. 7319 7320 srcutree.counter_wrap_check [KNL] 7321 Specifies how frequently to check for 7322 grace-period sequence counter wrap for the 7323 srcu_data structure's ->srcu_gp_seq_needed field. 7324 The greater the number of bits set in this kernel 7325 parameter, the less frequently counter wrap will 7326 be checked for. Note that the bottom two bits 7327 are ignored. 7328 7329 srcutree.exp_holdoff [KNL] 7330 Specifies how many nanoseconds must elapse 7331 since the end of the last SRCU grace period for 7332 a given srcu_struct until the next normal SRCU 7333 grace period will be considered for automatic 7334 expediting. Set to zero to disable automatic 7335 expediting. 7336 7337 srcutree.srcu_max_nodelay [KNL] 7338 Specifies the number of no-delay instances 7339 per jiffy for which the SRCU grace period 7340 worker thread will be rescheduled with zero 7341 delay. Beyond this limit, worker thread will 7342 be rescheduled with a sleep delay of one jiffy. 7343 7344 srcutree.srcu_max_nodelay_phase [KNL] 7345 Specifies the per-grace-period phase, number of 7346 non-sleeping polls of readers. Beyond this limit, 7347 grace period worker thread will be rescheduled 7348 with a sleep delay of one jiffy, between each 7349 rescan of the readers, for a grace period phase. 7350 7351 srcutree.srcu_retry_check_delay [KNL] 7352 Specifies number of microseconds of non-sleeping 7353 delay between each non-sleeping poll of readers. 7354 7355 srcutree.small_contention_lim [KNL] 7356 Specifies the number of update-side contention 7357 events per jiffy will be tolerated before 7358 initiating a conversion of an srcu_struct 7359 structure to big form. Note that the value of 7360 srcutree.convert_to_big must have the 0x10 bit 7361 set for contention-based conversions to occur. 7362 7363 ssbd= [ARM64,HW,EARLY] 7364 Speculative Store Bypass Disable control 7365 7366 On CPUs that are vulnerable to the Speculative 7367 Store Bypass vulnerability and offer a 7368 firmware based mitigation, this parameter 7369 indicates how the mitigation should be used: 7370 7371 force-on: Unconditionally enable mitigation for 7372 for both kernel and userspace 7373 force-off: Unconditionally disable mitigation for 7374 for both kernel and userspace 7375 kernel: Always enable mitigation in the 7376 kernel, and offer a prctl interface 7377 to allow userspace to register its 7378 interest in being mitigated too. 7379 7380 stack_guard_gap= [MM] 7381 override the default stack gap protection. The value 7382 is in page units and it defines how many pages prior 7383 to (for stacks growing down) resp. after (for stacks 7384 growing up) the main stack are reserved for no other 7385 mapping. Default value is 256 pages. 7386 7387 stack_depot_disable= [KNL,EARLY] 7388 Setting this to true through kernel command line will 7389 disable the stack depot thereby saving the static memory 7390 consumed by the stack hash table. By default this is set 7391 to false. 7392 7393 stack_depot_max_pools= [KNL,EARLY] 7394 Specify the maximum number of pools to use for storing 7395 stack traces. Pools are allocated on-demand up to this 7396 limit. Default value is 8191 pools. 7397 7398 stacktrace [FTRACE] 7399 Enable the stack tracer on boot up. 7400 7401 stacktrace_filter=[function-list] 7402 [FTRACE] Limit the functions that the stack tracer 7403 will trace at boot up. function-list is a comma-separated 7404 list of functions. This list can be changed at run 7405 time by the stack_trace_filter file in the debugfs 7406 tracing directory. Note, this enables stack tracing 7407 and the stacktrace above is not needed. 7408 7409 sti= [PARISC,HW] 7410 Format: <num> 7411 Set the STI (builtin display/keyboard on the HP-PARISC 7412 machines) console (graphic card) which should be used 7413 as the initial boot-console. 7414 See also comment in drivers/video/sticore.c. 7415 7416 sti_font= [HW] 7417 See comment in drivers/video/sticore.c. 7418 7419 stifb= [HW] 7420 Format: bpp:<bpp1>[:<bpp2>[:<bpp3>...]] 7421 7422 strict_sas_size= 7423 [X86] 7424 Format: <bool> 7425 Enable or disable strict sigaltstack size checks 7426 against the required signal frame size which 7427 depends on the supported FPU features. This can 7428 be used to filter out binaries which have 7429 not yet been made aware of AT_MINSIGSTKSZ. 7430 7431 stress_hpt [PPC,EARLY] 7432 Limits the number of kernel HPT entries in the hash 7433 page table to increase the rate of hash page table 7434 faults on kernel addresses. 7435 7436 stress_slb [PPC,EARLY] 7437 Limits the number of kernel SLB entries, and flushes 7438 them frequently to increase the rate of SLB faults 7439 on kernel addresses. 7440 7441 no_slb_preload [PPC,EARLY] 7442 Disables slb preloading for userspace. 7443 7444 sunrpc.min_resvport= 7445 sunrpc.max_resvport= 7446 [NFS,SUNRPC] 7447 SunRPC servers often require that client requests 7448 originate from a privileged port (i.e. a port in the 7449 range 0 < portnr < 1024). 7450 An administrator who wishes to reserve some of these 7451 ports for other uses may adjust the range that the 7452 kernel's sunrpc client considers to be privileged 7453 using these two parameters to set the minimum and 7454 maximum port values. 7455 7456 sunrpc.svc_rpc_per_connection_limit= 7457 [NFS,SUNRPC] 7458 Limit the number of requests that the server will 7459 process in parallel from a single connection. 7460 The default value is 0 (no limit). 7461 7462 sunrpc.pool_mode= 7463 [NFS] 7464 Deprecated. The NFS server now always uses one 7465 service thread pool per NUMA node (equivalent to a 7466 single global pool on non-NUMA machines). All of 7467 the previously accepted values (auto, global, 7468 percpu, pernode) are still accepted for backward 7469 compatibility but are ignored: the mode is always 7470 pernode, and reads always return "pernode". 7471 7472 sunrpc.tcp_slot_table_entries= 7473 sunrpc.udp_slot_table_entries= 7474 [NFS,SUNRPC] 7475 Sets the upper limit on the number of simultaneous 7476 RPC calls that can be sent from the client to a 7477 server. Increasing these values may allow you to 7478 improve throughput, but will also increase the 7479 amount of memory reserved for use by the client. 7480 7481 suspend.pm_test_delay= 7482 [SUSPEND] 7483 Sets the number of seconds to remain in a suspend test 7484 mode before resuming the system (see 7485 /sys/power/pm_test). Only available when CONFIG_PM_DEBUG 7486 is set. Default value is 5. 7487 7488 svm= [PPC] 7489 Format: { on | off | y | n | 1 | 0 } 7490 This parameter controls use of the Protected 7491 Execution Facility on pSeries. 7492 7493 swiotlb= [ARM,PPC,MIPS,X86,S390,EARLY] 7494 Format: { <int> [,<int>] | force | noforce | track_hiwater} 7495 <int> -- Number of I/O TLB slabs 7496 <int> -- Second integer after comma. Number of swiotlb 7497 areas with their own lock. Will be rounded up 7498 to a power of 2. 7499 force -- force using of bounce buffers even if they 7500 wouldn't be automatically used by the kernel 7501 noforce -- Never use bounce buffers (for debugging) 7502 track_hiwater -- Track high watermark of swiotlb buffers. 7503 Only available when CONFIG_DEBUG_FS is set. 7504 7505 switches= [HW,M68k,EARLY] 7506 7507 sysctl.*= [KNL] 7508 Set a sysctl parameter, right before loading the init 7509 process, as if the value was written to the respective 7510 /proc/sys/... file. Both '.' and '/' are recognized as 7511 separators. Unrecognized parameters and invalid values 7512 are reported in the kernel log. Sysctls registered 7513 later by a loaded module cannot be set this way. 7514 Example: sysctl.vm.swappiness=40 7515 7516 sysrq_always_enabled 7517 [KNL] 7518 Ignore sysrq setting - this boot parameter will 7519 neutralize any effect of /proc/sys/kernel/sysrq. 7520 Useful for debugging. 7521 7522 tcpmhash_entries= [KNL,NET] 7523 Set the number of tcp_metrics_hash slots. 7524 Default value is 8192 or 16384 depending on total 7525 ram pages. This is used to specify the TCP metrics 7526 cache size. See Documentation/networking/ip-sysctl.rst 7527 "tcp_no_metrics_save" section for more details. 7528 7529 tdfx= [HW,DRM] 7530 7531 test_suspend= [SUSPEND] 7532 Format: { "mem" | "standby" | "freeze" }[,N] 7533 Specify "mem" (for Suspend-to-RAM) or "standby" (for 7534 standby suspend) or "freeze" (for suspend type freeze) 7535 as the system sleep state during system startup with 7536 the optional capability to repeat N number of times. 7537 The system is woken from this state using a 7538 wakeup-capable RTC alarm. 7539 7540 thash_entries= [KNL,NET] 7541 Set number of hash buckets for TCP connection 7542 7543 thermal.act= [HW,ACPI] 7544 -1: disable all active trip points in all thermal zones 7545 <degrees C>: override all lowest active trip points 7546 7547 thermal.crt= [HW,ACPI] 7548 -1: disable all critical trip points in all thermal zones 7549 <degrees C>: override all critical trip points 7550 7551 thermal.off= [HW,ACPI] 7552 1: disable ACPI thermal control 7553 7554 thermal.psv= [HW,ACPI] 7555 -1: disable all passive trip points 7556 <degrees C>: override all passive trip points to this 7557 value 7558 7559 thermal.tzp= [HW,ACPI] 7560 Specify global default ACPI thermal zone polling rate 7561 <deci-seconds>: poll all this frequency 7562 0: no polling (default) 7563 7564 thp_anon= [KNL] 7565 Format: <size>[KMG],<size>[KMG]:<state>;<size>[KMG]-<size>[KMG]:<state> 7566 state is one of "always", "madvise", "never" or "inherit". 7567 Control the default behavior of the system with respect 7568 to anonymous transparent hugepages. 7569 Can be used multiple times for multiple anon THP sizes. 7570 See Documentation/admin-guide/mm/transhuge.rst for more 7571 details. 7572 7573 threadirqs [KNL,EARLY] 7574 Force threading of all interrupt handlers except those 7575 marked explicitly IRQF_NO_THREAD. 7576 7577 thp_shmem= [KNL] 7578 Format: <size>[KMG],<size>[KMG]:<policy>;<size>[KMG]-<size>[KMG]:<policy> 7579 Control the default policy of each hugepage size for the 7580 internal shmem mount. <policy> is one of policies available 7581 for the shmem mount ("always", "inherit", "never", "within_size", 7582 and "advise"). 7583 It can be used multiple times for multiple shmem THP sizes. 7584 See Documentation/admin-guide/mm/transhuge.rst for more 7585 details. 7586 7587 tlbi= [X86-64] 7588 Format: {ipi} 7589 ipi: switch to IPI-based TLB flushing 7590 7591 topology= [S390,EARLY] 7592 Format: {off | on} 7593 Specify if the kernel should make use of the cpu 7594 topology information if the hardware supports this. 7595 The scheduler will make use of this information and 7596 e.g. base its process migration decisions on it. 7597 Default is on. 7598 7599 torture.disable_onoff_at_boot= [KNL] 7600 Prevent the CPU-hotplug component of torturing 7601 until after init has spawned. 7602 7603 torture.ftrace_dump_at_shutdown= [KNL] 7604 Dump the ftrace buffer at torture-test shutdown, 7605 even if there were no errors. This can be a 7606 very costly operation when many torture tests 7607 are running concurrently, especially on systems 7608 with rotating-rust storage. 7609 7610 torture.verbose_sleep_frequency= [KNL] 7611 Specifies how many verbose printk()s should be 7612 emitted between each sleep. The default of zero 7613 disables verbose-printk() sleeping. 7614 7615 torture.verbose_sleep_duration= [KNL] 7616 Duration of each verbose-printk() sleep in jiffies. 7617 7618 tpm.disable_pcr_integrity= [HW,TPM] 7619 Do not protect PCR registers from unintended physical 7620 access, or interposers in the bus by the means of 7621 having an integrity protected session wrapped around 7622 TPM2_PCR_Extend command. Consider this in a situation 7623 where TPM is heavily utilized by IMA, thus protection 7624 causing a major performance hit, and the space where 7625 machines are deployed is by other means guarded. 7626 7627 tpm_crb_ffa.busy_timeout_ms= [ARM64,TPM] 7628 Maximum time in milliseconds to retry sending a message 7629 to the TPM service before giving up. This parameter controls 7630 how long the system will continue retrying when the TPM 7631 service is busy. 7632 Format: <unsigned int> 7633 Default: 2000 (2 seconds) 7634 7635 tpm_suspend_pcr=[HW,TPM] 7636 Format: integer pcr id 7637 Specify that at suspend time, the tpm driver 7638 should extend the specified pcr with zeros, 7639 as a workaround for some chips which fail to 7640 flush the last written pcr on TPM_SaveState. 7641 This will guarantee that all the other pcrs 7642 are saved. 7643 7644 tpm_tis.interrupts= [HW,TPM] 7645 Enable interrupts for the MMIO based physical layer 7646 for the FIFO interface. By default it is set to false 7647 (0). For more information about TPM hardware interfaces 7648 defined by Trusted Computing Group (TCG) see 7649 https://trustedcomputinggroup.org/resource/pc-client-platform-tpm-profile-ptp-specification/ 7650 7651 tp_printk [FTRACE] 7652 Have the tracepoints sent to printk as well as the 7653 tracing ring buffer. This is useful for early boot up 7654 where the system hangs or reboots and does not give the 7655 option for reading the tracing buffer or performing a 7656 ftrace_dump_on_oops. 7657 7658 To turn off having tracepoints sent to printk, 7659 echo 0 > /proc/sys/kernel/tracepoint_printk 7660 Note, echoing 1 into this file without the 7661 tp_printk kernel cmdline option has no effect. 7662 7663 The tp_printk_stop_on_boot (see below) can also be used 7664 to stop the printing of events to console at 7665 late_initcall_sync. 7666 7667 ** CAUTION ** 7668 7669 Having tracepoints sent to printk() and activating high 7670 frequency tracepoints such as irq or sched, can cause 7671 the system to live lock. 7672 7673 tp_printk_stop_on_boot [FTRACE] 7674 When tp_printk (above) is set, it can cause a lot of noise 7675 on the console. It may be useful to only include the 7676 printing of events during boot up, as user space may 7677 make the system inoperable. 7678 7679 This command line option will stop the printing of events 7680 to console at the late_initcall_sync() time frame. 7681 7682 trace_buf_size=nn[KMG] 7683 [FTRACE] will set tracing buffer size on each cpu. 7684 7685 trace_clock= [FTRACE] Set the clock used for tracing events 7686 at boot up. 7687 local - Use the per CPU time stamp counter 7688 (converted into nanoseconds). Fast, but 7689 depending on the architecture, may not be 7690 in sync between CPUs. 7691 global - Event time stamps are synchronized across 7692 CPUs. May be slower than the local clock, 7693 but better for some race conditions. 7694 counter - Simple counting of events (1, 2, ..) 7695 note, some counts may be skipped due to the 7696 infrastructure grabbing the clock more than 7697 once per event. 7698 uptime - Use jiffies as the time stamp. 7699 perf - Use the same clock that perf uses. 7700 mono - Use ktime_get_mono_fast_ns() for time stamps. 7701 mono_raw - Use ktime_get_raw_fast_ns() for time 7702 stamps. 7703 boot - Use ktime_get_boot_fast_ns() for time stamps. 7704 Architectures may add more clocks. See 7705 Documentation/trace/ftrace.rst for more details. 7706 7707 trace_event=[event-list] 7708 [FTRACE] Set and start specified trace events in order 7709 to facilitate early boot debugging. The event-list is a 7710 comma-separated list of trace events to enable. See 7711 also Documentation/trace/events.rst 7712 7713 To enable modules, use :mod: keyword: 7714 7715 trace_event=:mod:<module> 7716 7717 The value before :mod: will only enable specific events 7718 that are part of the module. See the above mentioned 7719 document for more information. 7720 7721 trace_instance=[instance-info] 7722 [FTRACE] Create a ring buffer instance early in boot up. 7723 This will be listed in: 7724 7725 /sys/kernel/tracing/instances 7726 7727 Events can be enabled at the time the instance is created 7728 via: 7729 7730 trace_instance=<name>,<system1>:<event1>,<system2>:<event2> 7731 7732 Note, the "<system*>:" portion is optional if the event is 7733 unique. 7734 7735 trace_instance=foo,sched:sched_switch,irq_handler_entry,initcall 7736 7737 will enable the "sched_switch" event (note, the "sched:" is optional, and 7738 the same thing would happen if it was left off). The irq_handler_entry 7739 event, and all events under the "initcall" system. 7740 7741 Flags can be added to the instance to modify its behavior when it is 7742 created. The flags are separated by '^'. 7743 7744 The available flags are: 7745 7746 traceoff - Have the tracing instance tracing disabled after it is created. 7747 traceprintk - Have trace_printk() write into this trace instance 7748 (note, "printk" and "trace_printk" can also be used) 7749 7750 trace_instance=foo^traceoff^traceprintk,sched,irq 7751 7752 The flags must come before the defined events. 7753 7754 If memory has been reserved (see memmap for x86), the instance 7755 can use that memory: 7756 7757 memmap=12M$0x284500000 trace_instance=boot_map@0x284500000:12M 7758 7759 The above will create a "boot_map" instance that uses the physical 7760 memory at 0x284500000 that is 12Megs. The per CPU buffers of that 7761 instance will be split up accordingly. 7762 7763 Alternatively, the memory can be reserved by the reserve_mem option: 7764 7765 reserve_mem=12M:4096:trace trace_instance=boot_map@trace 7766 7767 This will reserve 12 megabytes at boot up with a 4096 byte alignment 7768 and place the ring buffer in this memory. Note that due to KASLR, the 7769 memory may not be the same location each time, which will not preserve 7770 the buffer content. 7771 7772 Also note that the layout of the ring buffer data may change between 7773 kernel versions where the validator will fail and reset the ring buffer 7774 if the layout is not the same as the previous kernel. 7775 7776 If the ring buffer is used for persistent bootups and has events enabled, 7777 it is recommend to disable tracing so that events from a previous boot do not 7778 mix with events of the current boot (unless you are debugging a random crash 7779 at boot up). 7780 7781 reserve_mem=12M:4096:trace trace_instance=boot_map^traceoff^traceprintk@trace,sched,irq 7782 7783 Note, saving the trace buffer across reboots does require that the system 7784 is set up to not wipe memory. For instance, CONFIG_RESET_ATTACK_MITIGATION 7785 can force a memory reset on boot which will clear any trace that was stored. 7786 This is just one of many ways that can clear memory. Make sure your system 7787 keeps the content of memory across reboots before relying on this option. 7788 7789 NB: Both the mapped address and size must be page aligned for the architecture. 7790 7791 See also Documentation/trace/debugging.rst 7792 7793 7794 trace_options=[option-list] 7795 [FTRACE] Enable or disable tracer options at boot. 7796 The option-list is a comma delimited list of options 7797 that can be enabled or disabled just as if you were 7798 to echo the option name into 7799 7800 /sys/kernel/tracing/trace_options 7801 7802 For example, to enable stacktrace option (to dump the 7803 stack trace of each event), add to the command line: 7804 7805 trace_options=stacktrace 7806 7807 See also Documentation/trace/ftrace.rst "trace options" 7808 section. 7809 7810 trace_trigger=[trigger-list] 7811 [FTRACE] Add an event trigger on specific events. 7812 Set a trigger on top of a specific event, with an optional 7813 filter. 7814 7815 The format is "trace_trigger=<event>.<trigger>[ if <filter>],..." 7816 Where more than one trigger may be specified that are comma delimited. 7817 7818 For example: 7819 7820 trace_trigger="sched_switch.stacktrace if prev_state == 2" 7821 7822 The above will enable the "stacktrace" trigger on the "sched_switch" 7823 event but only trigger it if the "prev_state" of the "sched_switch" 7824 event is "2" (TASK_UNINTERRUPTIBLE). 7825 7826 See also "Event triggers" in Documentation/trace/events.rst 7827 7828 7829 traceoff_after_boot 7830 [FTRACE] Sometimes tracing is used to debug issues 7831 during the boot process. Since the trace buffer has a 7832 limited amount of storage, it may be prudent to 7833 disable tracing after the boot is finished, otherwise 7834 the critical information may be overwritten. With this 7835 option, the main tracing buffer will be turned off at 7836 the end of the boot process. 7837 7838 traceoff_on_warning 7839 [FTRACE] enable this option to disable tracing when a 7840 warning is hit. This turns off "tracing_on". Tracing can 7841 be enabled again by echoing '1' into the "tracing_on" 7842 file located in /sys/kernel/tracing/ 7843 7844 This option is useful, as it disables the trace before 7845 the WARNING dump is called, which prevents the trace to 7846 be filled with content caused by the warning output. 7847 7848 This option can also be set at run time via the sysctl 7849 option: kernel/traceoff_on_warning 7850 7851 transparent_hugepage= 7852 [KNL] 7853 Format: [always|madvise|never] 7854 Can be used to control the default behavior of the system 7855 with respect to transparent hugepages. 7856 See Documentation/admin-guide/mm/transhuge.rst 7857 for more details. 7858 7859 transparent_hugepage_shmem= [KNL] 7860 Format: [always|within_size|advise|never|deny|force] 7861 Can be used to control the hugepage allocation policy for 7862 the internal shmem mount. 7863 See Documentation/admin-guide/mm/transhuge.rst 7864 for more details. 7865 7866 transparent_hugepage_tmpfs= [KNL] 7867 Format: [always|within_size|advise|never] 7868 Can be used to control the default hugepage allocation policy 7869 for the tmpfs mount. 7870 See Documentation/admin-guide/mm/transhuge.rst 7871 for more details. 7872 7873 trusted.source= [KEYS] 7874 Format: <string> 7875 This parameter identifies the trust source as a backend 7876 for trusted keys implementation. Supported trust 7877 sources: 7878 - "tpm" 7879 - "tee" 7880 - "caam" 7881 - "dcp" 7882 - "pkwm" 7883 If not specified then it defaults to iterating through 7884 the trust source list starting with TPM and assigns the 7885 first trust source as a backend which is initialized 7886 successfully during iteration. 7887 7888 trusted.debug= [KEYS] 7889 Format: <bool> 7890 Enable trusted keys debug traces at runtime when 7891 CONFIG_TRUSTED_KEYS_DEBUG=y. 7892 7893 To make the traces visible after enabling the option, 7894 use trusted.dyndbg='+p' as needed. By convention, 7895 the subsystem uses pr_debug() for these traces. 7896 7897 SAFETY: The traces can leak sensitive data, so be 7898 cautious before enabling this. They remain inactive 7899 unless this parameter is set this option to a true 7900 value. 7901 7902 Default: false 7903 7904 trusted.rng= [KEYS] 7905 Format: <string> 7906 The RNG used to generate key material for trusted keys. 7907 Can be one of: 7908 - "kernel" 7909 - the same value as trusted.source: "tpm" or "tee" 7910 - "default" 7911 If not specified, "default" is used. In this case, 7912 the RNG's choice is left to each individual trust source. 7913 7914 trusted.dcp_use_otp_key 7915 This is intended to be used in combination with 7916 trusted.source=dcp and will select the DCP OTP key 7917 instead of the DCP UNIQUE key blob encryption. 7918 7919 trusted.dcp_skip_zk_test 7920 This is intended to be used in combination with 7921 trusted.source=dcp and will disable the check if the 7922 blob key is all zeros. This is helpful for situations where 7923 having this key zero'ed is acceptable. E.g. in testing 7924 scenarios. 7925 7926 tsa= [X86] Control mitigation for Transient Scheduler 7927 Attacks on AMD CPUs. Search the following in your 7928 favourite search engine for more details: 7929 7930 "Technical guidance for mitigating transient scheduler 7931 attacks". 7932 7933 off - disable the mitigation 7934 on - enable the mitigation (default) 7935 user - mitigate only user/kernel transitions 7936 vm - mitigate only guest/host transitions 7937 7938 7939 tsc= Disable clocksource stability checks for TSC. 7940 Format: <string> 7941 [x86] reliable: mark tsc clocksource as reliable, this 7942 disables clocksource verification at runtime, as well 7943 as the stability checks done at bootup. Used to enable 7944 high-resolution timer mode on older hardware, and in 7945 virtualized environment. 7946 [x86] noirqtime: Do not use TSC to do irq accounting. 7947 Used to run time disable IRQ_TIME_ACCOUNTING on any 7948 platforms where RDTSC is slow and this accounting 7949 can add overhead. 7950 [x86] unstable: mark the TSC clocksource as unstable, this 7951 marks the TSC unconditionally unstable at bootup and 7952 avoids any further wobbles once the TSC watchdog notices. 7953 [x86] nowatchdog: disable clocksource watchdog. Used 7954 in situations with strict latency requirements (where 7955 interruptions from clocksource watchdog are not 7956 acceptable). 7957 [x86] recalibrate: force recalibration against a HW timer 7958 (HPET or PM timer) on systems whose TSC frequency was 7959 obtained from HW or FW using either an MSR or CPUID(0x15). 7960 Warn if the difference is more than 500 ppm. 7961 [x86] watchdog: Enforce the clocksource watchdog on TSC 7962 7963 tsc_early_khz= [X86,EARLY] Skip early TSC calibration and use the given 7964 value instead. Useful when the early TSC frequency discovery 7965 procedure is not reliable, such as on overclocked systems 7966 with CPUID.16h support and partial CPUID.15h support. 7967 Format: <unsigned int> 7968 7969 tsx= [X86] Control Transactional Synchronization 7970 Extensions (TSX) feature in Intel processors that 7971 support TSX control. 7972 7973 This parameter controls the TSX feature. The options are: 7974 7975 on - Enable TSX on the system. Although there are 7976 mitigations for all known security vulnerabilities, 7977 TSX has been known to be an accelerator for 7978 several previous speculation-related CVEs, and 7979 so there may be unknown security risks associated 7980 with leaving it enabled. 7981 7982 off - Disable TSX on the system. (Note that this 7983 option takes effect only on newer CPUs which are 7984 not vulnerable to MDS, i.e., have 7985 MSR_IA32_ARCH_CAPABILITIES.MDS_NO=1 and which get 7986 the new IA32_TSX_CTRL MSR through a microcode 7987 update. This new MSR allows for the reliable 7988 deactivation of the TSX functionality.) 7989 7990 auto - Disable TSX if X86_BUG_TAA is present, 7991 otherwise enable TSX on the system. 7992 7993 Not specifying this option is equivalent to tsx=off. 7994 7995 See Documentation/admin-guide/hw-vuln/tsx_async_abort.rst 7996 for more details. 7997 7998 tsx_async_abort= [X86,INTEL,EARLY] Control mitigation for the TSX Async 7999 Abort (TAA) vulnerability. 8000 8001 Similar to Micro-architectural Data Sampling (MDS) 8002 certain CPUs that support Transactional 8003 Synchronization Extensions (TSX) are vulnerable to an 8004 exploit against CPU internal buffers which can forward 8005 information to a disclosure gadget under certain 8006 conditions. 8007 8008 In vulnerable processors, the speculatively forwarded 8009 data can be used in a cache side channel attack, to 8010 access data to which the attacker does not have direct 8011 access. 8012 8013 This parameter controls the TAA mitigation. The 8014 options are: 8015 8016 full - Enable TAA mitigation on vulnerable CPUs 8017 if TSX is enabled. 8018 8019 full,nosmt - Enable TAA mitigation and disable SMT on 8020 vulnerable CPUs. If TSX is disabled, SMT 8021 is not disabled because CPU is not 8022 vulnerable to cross-thread TAA attacks. 8023 off - Unconditionally disable TAA mitigation 8024 8025 On MDS-affected machines, tsx_async_abort=off can be 8026 prevented by an active MDS mitigation as both vulnerabilities 8027 are mitigated with the same mechanism so in order to disable 8028 this mitigation, you need to specify mds=off too. 8029 8030 Not specifying this option is equivalent to 8031 tsx_async_abort=full. On CPUs which are MDS affected 8032 and deploy MDS mitigation, TAA mitigation is not 8033 required and doesn't provide any additional 8034 mitigation. 8035 8036 For details see: 8037 Documentation/admin-guide/hw-vuln/tsx_async_abort.rst 8038 8039 turbografx.map[2|3]= [HW,JOY] 8040 TurboGraFX parallel port interface 8041 Format: 8042 <port#>,<js1>,<js2>,<js3>,<js4>,<js5>,<js6>,<js7> 8043 See also Documentation/input/devices/joystick-parport.rst 8044 8045 udbg-immortal [PPC] When debugging early kernel crashes that 8046 happen after console_init() and before a proper 8047 console driver takes over, this boot options might 8048 help "seeing" what's going on. 8049 8050 uhash_entries= [KNL,NET] 8051 Set number of hash buckets for UDP/UDP-Lite connections 8052 8053 uhci-hcd.ignore_oc= 8054 [USB] Ignore overcurrent events (default N). 8055 Some badly-designed motherboards generate lots of 8056 bogus events, for ports that aren't wired to 8057 anything. Set this parameter to avoid log spamming. 8058 Note that genuine overcurrent events won't be 8059 reported either. 8060 8061 unaligned_scalar_speed= 8062 [RISCV] 8063 Format: {slow | fast | unsupported} 8064 Allow skipping scalar unaligned access speed tests. This 8065 is useful for testing alternative code paths and to skip 8066 the tests in environments where they run too slowly. All 8067 CPUs must have the same scalar unaligned access speed. 8068 8069 unaligned_vector_speed= 8070 [RISCV] 8071 Format: {slow | fast | unsupported} 8072 Allow skipping vector unaligned access speed tests. This 8073 is useful for testing alternative code paths and to skip 8074 the tests in environments where they run too slowly. All 8075 CPUs must have the same vector unaligned access speed. 8076 8077 unknown_nmi_panic 8078 [X86] Cause panic on unknown NMI. 8079 8080 unwind_debug [X86-64,EARLY] 8081 Enable unwinder debug output. This can be 8082 useful for debugging certain unwinder error 8083 conditions, including corrupt stacks and 8084 bad/missing unwinder metadata. 8085 8086 usbcore.authorized_default= 8087 [USB] Default USB device authorization: 8088 (default -1 = authorized (same as 1), 8089 0 = not authorized, 1 = authorized, 2 = authorized 8090 if device connected to internal port) 8091 8092 usbcore.autosuspend= 8093 [USB] The autosuspend time delay (in seconds) used 8094 for newly-detected USB devices (default 2). This 8095 is the time required before an idle device will be 8096 autosuspended. Devices for which the delay is set 8097 to a negative value won't be autosuspended at all. 8098 8099 usbcore.usbfs_snoop= 8100 [USB] Set to log all usbfs traffic (default 0 = off). 8101 8102 usbcore.usbfs_snoop_max= 8103 [USB] Maximum number of bytes to snoop in each URB 8104 (default = 65536). 8105 8106 usbcore.blinkenlights= 8107 [USB] Set to cycle leds on hubs (default 0 = off). 8108 8109 usbcore.old_scheme_first= 8110 [USB] Start with the old device initialization 8111 scheme (default 0 = off). 8112 8113 usbcore.usbfs_memory_mb= 8114 [USB] Memory limit (in MB) for buffers allocated by 8115 usbfs (default = 16, 0 = max = 2047). 8116 8117 usbcore.use_both_schemes= 8118 [USB] Try the other device initialization scheme 8119 if the first one fails (default 1 = enabled). 8120 8121 usbcore.initial_descriptor_timeout= 8122 [USB] Specifies timeout for the initial 64-byte 8123 USB_REQ_GET_DESCRIPTOR request in milliseconds 8124 (default 5000 = 5.0 seconds). 8125 8126 usbcore.nousb [USB] Disable the USB subsystem 8127 8128 usbcore.quirks= 8129 [USB] A list of quirk entries to augment the built-in 8130 usb core quirk list. List entries are separated by 8131 commas. Each entry has the form 8132 VendorID:ProductID:Flags. The IDs are 4-digit hex 8133 numbers and Flags is a set of letters. Each letter 8134 will change the built-in quirk; setting it if it is 8135 clear and clearing it if it is set. The letters have 8136 the following meanings: 8137 a = USB_QUIRK_STRING_FETCH_255 (string 8138 descriptors must not be fetched using 8139 a 255-byte read); 8140 b = USB_QUIRK_RESET_RESUME (device can't resume 8141 correctly so reset it instead); 8142 c = USB_QUIRK_NO_SET_INTF (device can't handle 8143 Set-Interface requests); 8144 d = USB_QUIRK_CONFIG_INTF_STRINGS (device can't 8145 handle its Configuration or Interface 8146 strings); 8147 e = USB_QUIRK_RESET (device can't be reset 8148 (e.g morph devices), don't use reset); 8149 f = USB_QUIRK_HONOR_BNUMINTERFACES (device has 8150 more interface descriptions than the 8151 bNumInterfaces count, and can't handle 8152 talking to these interfaces); 8153 g = USB_QUIRK_DELAY_INIT (device needs a pause 8154 during initialization, after we read 8155 the device descriptor); 8156 h = USB_QUIRK_LINEAR_UFRAME_INTR_BINTERVAL (For 8157 high speed and super speed interrupt 8158 endpoints, the USB 2.0 and USB 3.0 spec 8159 require the interval in microframes (1 8160 microframe = 125 microseconds) to be 8161 calculated as interval = 2 ^ 8162 (bInterval-1). 8163 Devices with this quirk report their 8164 bInterval as the result of this 8165 calculation instead of the exponent 8166 variable used in the calculation); 8167 i = USB_QUIRK_DEVICE_QUALIFIER (device can't 8168 handle device_qualifier descriptor 8169 requests); 8170 j = USB_QUIRK_IGNORE_REMOTE_WAKEUP (device 8171 generates spurious wakeup, ignore 8172 remote wakeup capability); 8173 k = USB_QUIRK_NO_LPM (device can't handle Link 8174 Power Management); 8175 l = USB_QUIRK_LINEAR_FRAME_INTR_BINTERVAL 8176 (Device reports its bInterval as linear 8177 frames instead of the USB 2.0 8178 calculation); 8179 m = USB_QUIRK_DISCONNECT_SUSPEND (Device needs 8180 to be disconnected before suspend to 8181 prevent spurious wakeup); 8182 n = USB_QUIRK_DELAY_CTRL_MSG (Device needs a 8183 pause after every control message); 8184 o = USB_QUIRK_HUB_SLOW_RESET (Hub needs extra 8185 delay after resetting its port); 8186 p = USB_QUIRK_SHORT_SET_ADDRESS_REQ_TIMEOUT 8187 (Reduce timeout of the SET_ADDRESS 8188 request from 5000 ms to 500 ms); 8189 q = USB_QUIRK_FORCE_ONE_CONFIG (Device 8190 claims zero configurations, 8191 forcing to 1); 8192 r = USB_QUIRK_WINDOWS_CONFIG_REQ_SIZE (Device 8193 fails during initialization when asked for 8194 9-bytes configuration descriptor request. 8195 Ask for 255-bytes request instead to mirror 8196 Windows' behavior); 8197 Example: quirks=0781:5580:bk,0a5c:5834:gij 8198 8199 usbhid.mousepoll= 8200 [USBHID] The interval which mice are to be polled at. 8201 8202 usbhid.jspoll= 8203 [USBHID] The interval which joysticks are to be polled at. 8204 8205 usbhid.kbpoll= 8206 [USBHID] The interval which keyboards are to be polled at. 8207 8208 usb-storage.delay_use= 8209 [UMS] The delay in seconds before a new device is 8210 scanned for Logical Units (default 1). 8211 Optionally the delay in milliseconds if the value has 8212 suffix with "ms". 8213 Example: delay_use=2567ms 8214 8215 usb-storage.quirks= 8216 [UMS] A list of quirks entries to supplement or 8217 override the built-in unusual_devs list. List 8218 entries are separated by commas. Each entry has 8219 the form VID:PID:Flags where VID and PID are Vendor 8220 and Product ID values (4-digit hex numbers) and 8221 Flags is a set of characters, each corresponding 8222 to a common usb-storage quirk flag as follows: 8223 a = SANE_SENSE (collect more than 18 bytes 8224 of sense data, not on uas); 8225 b = BAD_SENSE (don't collect more than 18 8226 bytes of sense data, not on uas); 8227 c = FIX_CAPACITY (decrease the reported 8228 device capacity by one sector); 8229 d = NO_READ_DISC_INFO (don't use 8230 READ_DISC_INFO command, not on uas); 8231 e = NO_READ_CAPACITY_16 (don't use 8232 READ_CAPACITY_16 command); 8233 f = NO_REPORT_OPCODES (don't use report opcodes 8234 command, uas only); 8235 g = MAX_SECTORS_240 (don't transfer more than 8236 240 sectors at a time, uas only); 8237 h = CAPACITY_HEURISTICS (decrease the 8238 reported device capacity by one 8239 sector if the number is odd); 8240 i = IGNORE_DEVICE (don't bind to this 8241 device); 8242 j = NO_REPORT_LUNS (don't use report luns 8243 command, uas only); 8244 k = NO_SAME (do not use WRITE_SAME, uas only) 8245 l = NOT_LOCKABLE (don't try to lock and 8246 unlock ejectable media, not on uas); 8247 m = MAX_SECTORS_64 (don't transfer more 8248 than 64 sectors = 32 KB at a time, 8249 not on uas); 8250 n = INITIAL_READ10 (force a retry of the 8251 initial READ(10) command, not on uas); 8252 o = CAPACITY_OK (accept the capacity 8253 reported by the device, not on uas); 8254 p = WRITE_CACHE (the device cache is ON 8255 by default, not on uas); 8256 r = IGNORE_RESIDUE (the device reports 8257 bogus residue values, not on uas); 8258 s = SINGLE_LUN (the device has only one 8259 Logical Unit); 8260 t = NO_ATA_1X (don't allow ATA(12) and ATA(16) 8261 commands, uas only); 8262 u = IGNORE_UAS (don't bind to the uas driver); 8263 w = NO_WP_DETECT (don't test whether the 8264 medium is write-protected). 8265 y = ALWAYS_SYNC (issue a SYNCHRONIZE_CACHE 8266 even if the device claims no cache, 8267 not on uas) 8268 Example: quirks=0419:aaf5:rl,0421:0433:rc 8269 8270 user_debug= [KNL,ARM] 8271 Format: <int> 8272 See arch/arm/Kconfig.debug help text. 8273 1 - undefined instruction events 8274 2 - system calls 8275 4 - invalid data aborts 8276 8 - SIGSEGV faults 8277 16 - SIGBUS faults 8278 Example: user_debug=31 8279 8280 vdso= [X86,SH,SPARC] 8281 On X86_32, this is an alias for vdso32=. Otherwise: 8282 8283 vdso=1: enable VDSO (the default) 8284 vdso=0: disable VDSO mapping 8285 8286 vdso32= [X86] Control the 32-bit vDSO 8287 vdso32=1: enable 32-bit VDSO 8288 vdso32=0 or vdso32=2: disable 32-bit VDSO 8289 8290 See the help text for CONFIG_COMPAT_VDSO for more 8291 details. If CONFIG_COMPAT_VDSO is set, the default is 8292 vdso32=0; otherwise, the default is vdso32=1. 8293 8294 For compatibility with older kernels, vdso32=2 is an 8295 alias for vdso32=0. 8296 8297 Try vdso32=0 if you encounter an error that says: 8298 dl_main: Assertion `(void *) ph->p_vaddr == _rtld_local._dl_sysinfo_dso' failed! 8299 8300 video= [FB,EARLY] Frame buffer configuration 8301 See Documentation/fb/modedb.rst. 8302 8303 video.brightness_switch_enabled= [ACPI] 8304 Format: [0|1] 8305 If set to 1, on receiving an ACPI notify event 8306 generated by hotkey, video driver will adjust brightness 8307 level and then send out the event to user space through 8308 the allocated input device. If set to 0, video driver 8309 will only send out the event without touching backlight 8310 brightness level. 8311 default: 1 8312 8313 virtio_mmio.device= 8314 [VMMIO] Memory mapped virtio (platform) device. 8315 8316 <size>@<baseaddr>:<irq>[:<id>] 8317 where: 8318 <size> := size (can use standard suffixes 8319 like K, M and G) 8320 <baseaddr> := physical base address 8321 <irq> := interrupt number (as passed to 8322 request_irq()) 8323 <id> := (optional) platform device id 8324 example: 8325 virtio_mmio.device=1K@0x100b0000:48:7 8326 8327 Can be used multiple times for multiple devices. 8328 8329 vga= [BOOT,X86-32] Select a particular video mode 8330 See Documentation/arch/x86/boot.rst and 8331 Documentation/admin-guide/svga.rst. 8332 Use vga=ask for menu. 8333 This is actually a boot loader parameter; the value is 8334 passed to the kernel using a special protocol. 8335 8336 vm_debug[=options] [KNL] Available with CONFIG_DEBUG_VM=y. 8337 May slow down system boot speed, especially when 8338 enabled on systems with a large amount of memory. 8339 All options are enabled by default, and this 8340 interface is meant to allow for selectively 8341 enabling or disabling specific virtual memory 8342 debugging features. 8343 8344 Available options are: 8345 P Enable page structure init time poisoning 8346 - Disable all of the above options 8347 8348 vmalloc=nn[KMG] [KNL,BOOT,EARLY] Forces the vmalloc area to have an 8349 exact size of <nn>. This can be used to increase 8350 the minimum size (128MB on x86, arm32 platforms). 8351 It can also be used to decrease the size and leave more room 8352 for directly mapped kernel RAM. Note that this parameter does 8353 not exist on many other platforms (including arm64, alpha, 8354 loongarch, arc, csky, hexagon, microblaze, mips, nios2, openrisc, 8355 parisc, m64k, powerpc, riscv, sh, um, xtensa, s390, sparc). 8356 8357 vmcp_cma=nn[MG] [KNL,S390,EARLY] 8358 Sets the memory size reserved for contiguous memory 8359 allocations for the vmcp device driver. 8360 8361 vmhalt= [KNL,S390] Perform z/VM CP command after system halt. 8362 Format: <command> 8363 8364 vmpanic= [KNL,S390] Perform z/VM CP command after kernel panic. 8365 Format: <command> 8366 8367 vmpoff= [KNL,S390] Perform z/VM CP command after power off. 8368 Format: <command> 8369 8370 vmscape= [X86] Controls mitigation for VMscape attacks. 8371 VMscape attacks can leak information from a userspace 8372 hypervisor to a guest via speculative side-channels. 8373 8374 off - disable the mitigation 8375 ibpb - use Indirect Branch Prediction Barrier 8376 (IBPB) mitigation (default) 8377 force - force vulnerability detection even on 8378 unaffected processors 8379 8380 vsyscall= [X86-64,EARLY] 8381 Controls the behavior of vsyscalls (i.e. calls to 8382 fixed addresses of 0xffffffffff600x00 from legacy 8383 code). Most statically-linked binaries and older 8384 versions of glibc use these calls. Because these 8385 functions are at fixed addresses, they make nice 8386 targets for exploits that can control RIP. 8387 8388 emulate Vsyscalls turn into traps and are emulated 8389 reasonably safely. The vsyscall page is 8390 readable. This disables the Linear 8391 Address Space Separation (LASS) security 8392 feature and makes the system less secure. 8393 8394 xonly [default] Vsyscalls turn into traps and are 8395 emulated reasonably safely. The vsyscall 8396 page is not readable. 8397 8398 none Vsyscalls don't work at all. This makes 8399 them quite hard to use for exploits but 8400 might break your system. 8401 8402 vt.color= [VT] Default text color. 8403 Format: 0xYX, X = foreground, Y = background. 8404 Default: 0x07 = light gray on black. 8405 8406 vt.cur_default= [VT] Default cursor shape. 8407 Format: 0xCCBBAA, where AA, BB, and CC are the same as 8408 the parameters of the <Esc>[?A;B;Cc escape sequence; 8409 see vga-softcursor.rst. Default: 2 = underline. 8410 8411 vt.default_blu= [VT] 8412 Format: <blue0>,<blue1>,<blue2>,...,<blue15> 8413 Change the default blue palette of the console. 8414 This is a 16-member array composed of values 8415 ranging from 0-255. 8416 8417 vt.default_grn= [VT] 8418 Format: <green0>,<green1>,<green2>,...,<green15> 8419 Change the default green palette of the console. 8420 This is a 16-member array composed of values 8421 ranging from 0-255. 8422 8423 vt.default_red= [VT] 8424 Format: <red0>,<red1>,<red2>,...,<red15> 8425 Change the default red palette of the console. 8426 This is a 16-member array composed of values 8427 ranging from 0-255. 8428 8429 vt.default_utf8= 8430 [VT] 8431 Format=<0|1> 8432 Set system-wide default UTF-8 mode for all tty's. 8433 Default is 1, i.e. UTF-8 mode is enabled for all 8434 newly opened terminals. 8435 8436 vt.global_cursor_default= 8437 [VT] 8438 Format=<-1|0|1> 8439 Set system-wide default for whether a cursor 8440 is shown on new VTs. Default is -1, 8441 i.e. cursors will be created by default unless 8442 overridden by individual drivers. 0 will hide 8443 cursors, 1 will display them. 8444 8445 vt.italic= [VT] Default color for italic text; 0-15. 8446 Default: 2 = green. 8447 8448 vt.underline= [VT] Default color for underlined text; 0-15. 8449 Default: 3 = cyan. 8450 8451 watchdog timers [HW,WDT] For information on watchdog timers, 8452 see Documentation/watchdog/watchdog-parameters.rst 8453 or other driver-specific files in the 8454 Documentation/watchdog/ directory. 8455 8456 watchdog_thresh= 8457 [KNL] 8458 Set the hard lockup detector stall duration 8459 threshold in seconds. The soft lockup detector 8460 threshold is set to twice the value. A value of 0 8461 disables both lockup detectors. Default is 10 8462 seconds. 8463 8464 workqueue.unbound_cpus= 8465 [KNL,SMP] Specify to constrain one or some CPUs 8466 to use in unbound workqueues. 8467 Format: <cpu-list> 8468 By default, all online CPUs are available for 8469 unbound workqueues. 8470 8471 workqueue.watchdog_thresh= 8472 If CONFIG_WQ_WATCHDOG is configured, workqueue can 8473 warn stall conditions and dump internal state to 8474 help debugging. 0 disables workqueue stall 8475 detection; otherwise, it's the stall threshold 8476 duration in seconds. The default value is 30 and 8477 it can be updated at runtime by writing to the 8478 corresponding sysfs file. 8479 8480 workqueue.panic_on_stall=<uint> 8481 Panic when workqueue stall is detected by 8482 CONFIG_WQ_WATCHDOG. It sets the number times of the 8483 stall to trigger panic. 8484 8485 The default is set by CONFIG_BOOTPARAM_WQ_STALL_PANIC, 8486 which is 0 (disabled) if not configured. 8487 8488 workqueue.panic_on_stall_time=<uint> 8489 Panic when a workqueue stall has been continuous for 8490 the specified number of seconds. Unlike panic_on_stall 8491 which counts accumulated stall events, this triggers 8492 based on the duration of a single continuous stall. 8493 8494 The default is 0, which disables the time-based panic. 8495 8496 workqueue.cpu_intensive_thresh_us= 8497 Per-cpu work items which run for longer than this 8498 threshold are automatically considered CPU intensive 8499 and excluded from concurrency management to prevent 8500 them from noticeably delaying other per-cpu work 8501 items. Default is 10000 (10ms). 8502 8503 If CONFIG_WQ_CPU_INTENSIVE_REPORT is set, the kernel 8504 will report the work functions which violate this 8505 threshold repeatedly. They are likely good 8506 candidates for using WQ_UNBOUND workqueues instead. 8507 8508 workqueue.cpu_intensive_warning_thresh=<uint> 8509 If CONFIG_WQ_CPU_INTENSIVE_REPORT is set, the kernel 8510 will report the work functions which violate the 8511 intensive_threshold_us repeatedly. In order to prevent 8512 spurious warnings, start printing only after a work 8513 function has violated this threshold number of times. 8514 8515 The default is 4 times. 0 disables the warning. 8516 8517 workqueue.power_efficient 8518 Per-cpu workqueues are generally preferred because 8519 they show better performance thanks to cache 8520 locality; unfortunately, per-cpu workqueues tend to 8521 be more power hungry than unbound workqueues. 8522 8523 Enabling this makes the per-cpu workqueues which 8524 were observed to contribute significantly to power 8525 consumption unbound, leading to measurably lower 8526 power usage at the cost of small performance 8527 overhead. 8528 8529 The default value of this parameter is determined by 8530 the config option CONFIG_WQ_POWER_EFFICIENT_DEFAULT. 8531 8532 workqueue.default_affinity_scope= 8533 Select the default affinity scope to use for unbound 8534 workqueues. Can be one of "cpu", "smt", "cache", 8535 "cache_shard", "numa" and "system". Default is 8536 "cache_shard". For more 8537 information, see the Affinity Scopes section in 8538 Documentation/core-api/workqueue.rst. 8539 8540 This can be changed after boot by writing to the 8541 matching /sys/module/workqueue/parameters file. All 8542 workqueues with the "default" affinity scope will be 8543 updated accordingly. 8544 8545 workqueue.debug_force_rr_cpu 8546 Workqueue used to implicitly guarantee that work 8547 items queued without explicit CPU specified are put 8548 on the local CPU. This guarantee is no longer true 8549 and while local CPU is still preferred work items 8550 may be put on foreign CPUs. This debug option 8551 forces round-robin CPU selection to flush out 8552 usages which depend on the now broken guarantee. 8553 When enabled, memory and cache locality will be 8554 impacted. 8555 8556 writecombine= [LOONGARCH,EARLY] Control the MAT (Memory Access 8557 Type) of ioremap_wc(). 8558 8559 on - Enable writecombine, use WUC for ioremap_wc() 8560 off - Disable writecombine, use SUC for ioremap_wc() 8561 8562 x2apic_phys [X86-64,APIC,EARLY] Use x2apic physical mode instead of 8563 default x2apic cluster mode on platforms 8564 supporting x2apic. 8565 8566 xen_512gb_limit [KNL,X86-64,XEN] 8567 Restricts the kernel running paravirtualized under Xen 8568 to use only up to 512 GB of RAM. The reason to do so is 8569 crash analysis tools and Xen tools for doing domain 8570 save/restore/migration must be enabled to handle larger 8571 domains. 8572 8573 xen_console_io [XEN,EARLY] 8574 Boolean option to enable/disable the usage of the Xen 8575 console_io hypercalls to read and write to the console. 8576 Mostly useful for debugging and development. 8577 8578 xen_emul_unplug= [HW,X86,XEN,EARLY] 8579 Unplug Xen emulated devices 8580 Format: [unplug0,][unplug1] 8581 ide-disks -- unplug primary master IDE devices 8582 aux-ide-disks -- unplug non-primary-master IDE devices 8583 nics -- unplug network devices 8584 all -- unplug all emulated devices (NICs and IDE disks) 8585 unnecessary -- unplugging emulated devices is 8586 unnecessary even if the host did not respond to 8587 the unplug protocol 8588 never -- do not unplug even if version check succeeds 8589 8590 xen_legacy_crash [X86,XEN,EARLY] 8591 Crash from Xen panic notifier, without executing late 8592 panic() code such as dumping handler. 8593 8594 xen_mc_debug [X86,XEN,EARLY] 8595 Enable multicall debugging when running as a Xen PV guest. 8596 Enabling this feature will reduce performance a little 8597 bit, so it should only be enabled for obtaining extended 8598 debug data in case of multicall errors. 8599 8600 xen_msr_safe= [X86,XEN,EARLY] 8601 Format: <bool> 8602 Select whether to always use non-faulting (safe) MSR 8603 access functions when running as Xen PV guest. The 8604 default value is controlled by CONFIG_XEN_PV_MSR_SAFE. 8605 8606 xen_nopv [X86] 8607 Disables the PV optimizations forcing the HVM guest to 8608 run as generic HVM guest with no PV drivers. 8609 This option is obsoleted by the "nopv" option, which 8610 has equivalent effect for XEN platform. 8611 8612 xen_no_vector_callback 8613 [KNL,X86,XEN,EARLY] Disable the vector callback for Xen 8614 event channel interrupts. 8615 8616 xen_scrub_pages= [XEN] 8617 Boolean option to control scrubbing pages before giving them back 8618 to Xen, for use by other domains. Can be also changed at runtime 8619 with /sys/devices/system/xen_memory/xen_memory0/scrub_pages. 8620 Default value controlled with CONFIG_XEN_SCRUB_PAGES_DEFAULT. 8621 8622 xen_timer_slop= [X86-64,XEN,EARLY] 8623 Set the timer slop (in nanoseconds) for the virtual Xen 8624 timers (default is 100000). This adjusts the minimum 8625 delta of virtualized Xen timers, where lower values 8626 improve timer resolution at the expense of processing 8627 more timer interrupts. 8628 8629 xen.balloon_boot_timeout= [XEN] 8630 The time (in seconds) to wait before giving up to boot 8631 in case initial ballooning fails to free enough memory. 8632 Applies only when running as HVM or PVH guest and 8633 started with less memory configured than allowed at 8634 max. Default is 180. 8635 8636 xen.event_eoi_delay= [XEN] 8637 How long to delay EOI handling in case of event 8638 storms (jiffies). Default is 10. 8639 8640 xen.event_loop_timeout= [XEN] 8641 After which time (jiffies) the event handling loop 8642 should start to delay EOI handling. Default is 2. 8643 8644 xen.fifo_events= [XEN] 8645 Boolean parameter to disable using fifo event handling 8646 even if available. Normally fifo event handling is 8647 preferred over the 2-level event handling, as it is 8648 fairer and the number of possible event channels is 8649 much higher. Default is on (use fifo events). 8650 8651 xirc2ps_cs= [NET,PCMCIA] 8652 Format: 8653 <irq>,<irq_mask>,<io>,<full_duplex>,<do_sound>,<lockup_hack>[,<irq2>[,<irq3>[,<irq4>]]] 8654 8655 xive= [PPC] 8656 By default on POWER9 and above, the kernel will 8657 natively use the XIVE interrupt controller. This option 8658 allows the fallback firmware mode to be used: 8659 8660 off Fallback to firmware control of XIVE interrupt 8661 controller on both pseries and powernv 8662 platforms. Only useful on POWER9 and above. 8663 8664 xive.store-eoi=off [PPC] 8665 By default on POWER10 and above, the kernel will use 8666 stores for EOI handling when the XIVE interrupt mode 8667 is active. This option allows the XIVE driver to use 8668 loads instead, as on POWER9. 8669 8670 xhci-hcd.quirks [USB,KNL] 8671 A hex value specifying bitmask with supplemental xhci 8672 host controller quirks. Meaning of each bit can be 8673 consulted in header drivers/usb/host/xhci.h. 8674 8675 xmon [PPC,EARLY] 8676 Format: { early | on | rw | ro | off } 8677 Controls if xmon debugger is enabled. Default is off. 8678 Passing only "xmon" is equivalent to "xmon=early". 8679 early Call xmon as early as possible on boot; xmon 8680 debugger is called from setup_arch(). 8681 on xmon debugger hooks will be installed so xmon 8682 is only called on a kernel crash. Default mode, 8683 i.e. either "ro" or "rw" mode, is controlled 8684 with CONFIG_XMON_DEFAULT_RO_MODE. 8685 rw xmon debugger hooks will be installed so xmon 8686 is called only on a kernel crash, mode is write, 8687 meaning SPR registers, memory and, other data 8688 can be written using xmon commands. 8689 ro same as "rw" option above but SPR registers, 8690 memory, and other data can't be written using 8691 xmon commands. 8692 off xmon is disabled. 8693