1 accept_memory= [MM] 2 Format: { eager | lazy } 3 default: lazy 4 By default, unaccepted memory is accepted lazily to 5 avoid prolonged boot times. The lazy option will add 6 some runtime overhead until all memory is eventually 7 accepted. In most cases the overhead is negligible. 8 For some workloads or for debugging purposes 9 accept_memory=eager can be used to accept all memory 10 at once during boot. 11 12 acpi= [HW,ACPI,X86,ARM64,RISCV64,EARLY] 13 Advanced Configuration and Power Interface 14 Format: { force | on | off | strict | noirq | rsdt | 15 copy_dsdt | nospcr } 16 force -- enable ACPI if default was off 17 on -- enable ACPI but allow fallback to DT [arm64,riscv64] 18 off -- disable ACPI if default was on 19 noirq -- do not use ACPI for IRQ routing 20 strict -- Be less tolerant of platforms that are not 21 strictly ACPI specification compliant. 22 rsdt -- prefer RSDT over (default) XSDT 23 copy_dsdt -- copy DSDT to memory 24 nocmcff -- Disable firmware first mode for corrected 25 errors. This disables parsing the HEST CMC error 26 source to check if firmware has set the FF flag. This 27 may result in duplicate corrected error reports. 28 nospcr -- disable console in ACPI SPCR table as 29 default _serial_ console on ARM64 30 For ARM64, ONLY "acpi=off", "acpi=on", "acpi=force" or 31 "acpi=nospcr" are available 32 For RISCV64, ONLY "acpi=off", "acpi=on" or "acpi=force" 33 are available 34 35 See also Documentation/power/runtime_pm.rst, pci=noacpi 36 37 acpi_apic_instance= [ACPI,IOAPIC,EARLY] 38 Format: <int> 39 2: use 2nd APIC table, if available 40 1,0: use 1st APIC table 41 default: 0 42 43 acpi_backlight= [HW,ACPI] 44 { vendor | video | native | none } 45 If set to vendor, prefer vendor-specific driver 46 (e.g. thinkpad_acpi, sony_acpi, etc.) instead 47 of the ACPI video.ko driver. 48 If set to video, use the ACPI video.ko driver. 49 If set to native, use the device's native backlight mode. 50 If set to none, disable the ACPI backlight interface. 51 52 acpi_force_32bit_fadt_addr [ACPI,EARLY] 53 force FADT to use 32 bit addresses rather than the 54 64 bit X_* addresses. Some firmware have broken 64 55 bit addresses for force ACPI ignore these and use 56 the older legacy 32 bit addresses. 57 58 acpica_no_return_repair [HW, ACPI] 59 Disable AML predefined validation mechanism 60 This mechanism can repair the evaluation result to make 61 the return objects more ACPI specification compliant. 62 This option is useful for developers to identify the 63 root cause of an AML interpreter issue when the issue 64 has something to do with the repair mechanism. 65 66 acpi.debug_layer= [HW,ACPI,ACPI_DEBUG] 67 acpi.debug_level= [HW,ACPI,ACPI_DEBUG] 68 Format: <int> 69 CONFIG_ACPI_DEBUG must be enabled to produce any ACPI 70 debug output. Bits in debug_layer correspond to a 71 _COMPONENT in an ACPI source file, e.g., 72 #define _COMPONENT ACPI_EVENTS 73 Bits in debug_level correspond to a level in 74 ACPI_DEBUG_PRINT statements, e.g., 75 ACPI_DEBUG_PRINT((ACPI_DB_INFO, ... 76 The debug_level mask defaults to "info". See 77 Documentation/firmware-guide/acpi/debug.rst for more information about 78 debug layers and levels. 79 80 Enable processor driver info messages: 81 acpi.debug_layer=0x20000000 82 Enable AML "Debug" output, i.e., stores to the Debug 83 object while interpreting AML: 84 acpi.debug_layer=0xffffffff acpi.debug_level=0x2 85 Enable all messages related to ACPI hardware: 86 acpi.debug_layer=0x2 acpi.debug_level=0xffffffff 87 88 Some values produce so much output that the system is 89 unusable. The "log_buf_len" parameter may be useful 90 if you need to capture more output. 91 92 acpi_enforce_resources= [ACPI] 93 { strict | lax | no } 94 Check for resource conflicts between native drivers 95 and ACPI OperationRegions (SystemIO and SystemMemory 96 only). IO ports and memory declared in ACPI might be 97 used by the ACPI subsystem in arbitrary AML code and 98 can interfere with legacy drivers. 99 strict (default): access to resources claimed by ACPI 100 is denied; legacy drivers trying to access reserved 101 resources will fail to bind to device using them. 102 lax: access to resources claimed by ACPI is allowed; 103 legacy drivers trying to access reserved resources 104 will bind successfully but a warning message is logged. 105 no: ACPI OperationRegions are not marked as reserved, 106 no further checks are performed. 107 108 acpi_force_table_verification [HW,ACPI,EARLY] 109 Enable table checksum verification during early stage. 110 By default, this is disabled due to x86 early mapping 111 size limitation. 112 113 acpi_irq_balance [HW,ACPI] 114 ACPI will balance active IRQs 115 default in APIC mode 116 117 acpi_irq_nobalance [HW,ACPI] 118 ACPI will not move active IRQs (default) 119 default in PIC mode 120 121 acpi_irq_isa= [HW,ACPI] If irq_balance, mark listed IRQs used by ISA 122 Format: <irq>,<irq>... 123 124 acpi_irq_pci= [HW,ACPI] If irq_balance, clear listed IRQs for 125 use by PCI 126 Format: <irq>,<irq>... 127 128 acpi_mask_gpe= [HW,ACPI] 129 Due to the existence of _Lxx/_Exx, some GPEs triggered 130 by unsupported hardware/firmware features can result in 131 GPE floodings that cannot be automatically disabled by 132 the GPE dispatcher. 133 This facility can be used to prevent such uncontrolled 134 GPE floodings. 135 Format: <byte> or <bitmap-list> 136 137 acpi_no_auto_serialize [HW,ACPI] 138 Disable auto-serialization of AML methods 139 AML control methods that contain the opcodes to create 140 named objects will be marked as "Serialized" by the 141 auto-serialization feature. 142 This feature is enabled by default. 143 This option allows to turn off the feature. 144 145 acpi_no_memhotplug [ACPI] Disable memory hotplug. Useful for kdump 146 kernels. 147 148 acpi_no_static_ssdt [HW,ACPI,EARLY] 149 Disable installation of static SSDTs at early boot time 150 By default, SSDTs contained in the RSDT/XSDT will be 151 installed automatically and they will appear under 152 /sys/firmware/acpi/tables. 153 This option turns off this feature. 154 Note that specifying this option does not affect 155 dynamic table installation which will install SSDT 156 tables to /sys/firmware/acpi/tables/dynamic. 157 158 acpi_no_watchdog [HW,ACPI,WDT] 159 Ignore the ACPI-based watchdog interface (WDAT) and let 160 a native driver control the watchdog device instead. 161 162 acpi_rsdp= [ACPI,EFI,KEXEC,EARLY] 163 Pass the RSDP address to the kernel, mostly used 164 on machines running EFI runtime service to boot the 165 second kernel for kdump. 166 167 acpi_os_name= [HW,ACPI] Tell ACPI BIOS the name of the OS 168 Format: To spoof as Windows 98: ="Microsoft Windows" 169 170 acpi_rev_override [ACPI] Override the _REV object to return 5 (instead 171 of 2 which is mandated by ACPI 6) as the supported ACPI 172 specification revision (when using this switch, it may 173 be necessary to carry out a cold reboot _twice_ in a 174 row to make it take effect on the platform firmware). 175 176 acpi_osi= [HW,ACPI] Modify list of supported OS interface strings 177 acpi_osi="string1" # add string1 178 acpi_osi="!string2" # remove string2 179 acpi_osi=!* # remove all strings 180 acpi_osi=! # disable all built-in OS vendor 181 strings 182 acpi_osi=!! # enable all built-in OS vendor 183 strings 184 acpi_osi= # disable all strings 185 186 'acpi_osi=!' can be used in combination with single or 187 multiple 'acpi_osi="string1"' to support specific OS 188 vendor string(s). Note that such command can only 189 affect the default state of the OS vendor strings, thus 190 it cannot affect the default state of the feature group 191 strings and the current state of the OS vendor strings, 192 specifying it multiple times through kernel command line 193 is meaningless. This command is useful when one do not 194 care about the state of the feature group strings which 195 should be controlled by the OSPM. 196 Examples: 197 1. 'acpi_osi=! acpi_osi="Windows 2000"' is equivalent 198 to 'acpi_osi="Windows 2000" acpi_osi=!', they all 199 can make '_OSI("Windows 2000")' TRUE. 200 201 'acpi_osi=' cannot be used in combination with other 202 'acpi_osi=' command lines, the _OSI method will not 203 exist in the ACPI namespace. NOTE that such command can 204 only affect the _OSI support state, thus specifying it 205 multiple times through kernel command line is also 206 meaningless. 207 Examples: 208 1. 'acpi_osi=' can make 'CondRefOf(_OSI, Local1)' 209 FALSE. 210 211 'acpi_osi=!*' can be used in combination with single or 212 multiple 'acpi_osi="string1"' to support specific 213 string(s). Note that such command can affect the 214 current state of both the OS vendor strings and the 215 feature group strings, thus specifying it multiple times 216 through kernel command line is meaningful. But it may 217 still not able to affect the final state of a string if 218 there are quirks related to this string. This command 219 is useful when one want to control the state of the 220 feature group strings to debug BIOS issues related to 221 the OSPM features. 222 Examples: 223 1. 'acpi_osi="Module Device" acpi_osi=!*' can make 224 '_OSI("Module Device")' FALSE. 225 2. 'acpi_osi=!* acpi_osi="Module Device"' can make 226 '_OSI("Module Device")' TRUE. 227 3. 'acpi_osi=! acpi_osi=!* acpi_osi="Windows 2000"' is 228 equivalent to 229 'acpi_osi=!* acpi_osi=! acpi_osi="Windows 2000"' 230 and 231 'acpi_osi=!* acpi_osi="Windows 2000" acpi_osi=!', 232 they all will make '_OSI("Windows 2000")' TRUE. 233 234 acpi_pm_good [X86] 235 Override the pmtimer bug detection: force the kernel 236 to assume that this machine's pmtimer latches its value 237 and always returns good values. 238 239 acpi_sci= [HW,ACPI,EARLY] ACPI System Control Interrupt trigger mode 240 Format: { level | edge | high | low } 241 242 acpi_skip_timer_override [HW,ACPI,EARLY] 243 Recognize and ignore IRQ0/pin2 Interrupt Override. 244 For broken nForce2 BIOS resulting in XT-PIC timer. 245 246 acpi_sleep= [HW,ACPI] Sleep options 247 Format: { s3_bios, s3_mode, s3_beep, s4_hwsig, 248 s4_nohwsig, old_ordering, nonvs, 249 sci_force_enable, nobl } 250 See Documentation/power/video.rst for information on 251 s3_bios and s3_mode. 252 s3_beep is for debugging; it makes the PC's speaker beep 253 as soon as the kernel's real-mode entry point is called. 254 s4_hwsig causes the kernel to check the ACPI hardware 255 signature during resume from hibernation, and gracefully 256 refuse to resume if it has changed. This complies with 257 the ACPI specification but not with reality, since 258 Windows does not do this and many laptops do change it 259 on docking. So the default behaviour is to allow resume 260 and simply warn when the signature changes, unless the 261 s4_hwsig option is enabled. 262 s4_nohwsig prevents ACPI hardware signature from being 263 used (or even warned about) during resume. 264 old_ordering causes the ACPI 1.0 ordering of the _PTS 265 control method, with respect to putting devices into 266 low power states, to be enforced (the ACPI 2.0 ordering 267 of _PTS is used by default). 268 nonvs prevents the kernel from saving/restoring the 269 ACPI NVS memory during suspend/hibernation and resume. 270 sci_force_enable causes the kernel to set SCI_EN directly 271 on resume from S1/S3 (which is against the ACPI spec, 272 but some broken systems don't work without it). 273 nobl causes the internal blacklist of systems known to 274 behave incorrectly in some ways with respect to system 275 suspend and resume to be ignored (use wisely). 276 277 acpi_use_timer_override [HW,ACPI,EARLY] 278 Use timer override. For some broken Nvidia NF5 boards 279 that require a timer override, but don't have HPET 280 281 add_efi_memmap [EFI,X86,EARLY] Include EFI memory map in 282 kernel's map of available physical RAM. 283 284 agp= [AGP] 285 { off | try_unsupported } 286 off: disable AGP support 287 try_unsupported: try to drive unsupported chipsets 288 (may crash computer or cause data corruption) 289 290 ALSA [HW,ALSA] 291 See Documentation/sound/alsa-configuration.rst 292 293 alignment= [KNL,ARM] 294 Allow the default userspace alignment fault handler 295 behaviour to be specified. Bit 0 enables warnings, 296 bit 1 enables fixups, and bit 2 sends a segfault. 297 298 align_va_addr= [X86-64] 299 Align virtual addresses by clearing slice [14:12] when 300 allocating a VMA at process creation time. This option 301 gives you up to 3% performance improvement on AMD F15h 302 machines (where it is enabled by default) for a 303 CPU-intensive style benchmark, and it can vary highly in 304 a microbenchmark depending on workload and compiler. 305 306 32: only for 32-bit processes 307 64: only for 64-bit processes 308 on: enable for both 32- and 64-bit processes 309 off: disable for both 32- and 64-bit processes 310 311 alloc_snapshot [FTRACE] 312 Allocate the ftrace snapshot buffer on boot up when the 313 main buffer is allocated. This is handy if debugging 314 and you need to use tracing_snapshot() on boot up, and 315 do not want to use tracing_snapshot_alloc() as it needs 316 to be done where GFP_KERNEL allocations are allowed. 317 318 allow_mismatched_32bit_el0 [ARM64,EARLY] 319 Allow execve() of 32-bit applications and setting of the 320 PER_LINUX32 personality on systems where only a strict 321 subset of the CPUs support 32-bit EL0. When this 322 parameter is present, the set of CPUs supporting 32-bit 323 EL0 is indicated by /sys/devices/system/cpu/aarch32_el0 324 and hot-unplug operations may be restricted. 325 326 See Documentation/arch/arm64/asymmetric-32bit.rst for more 327 information. 328 329 amd_iommu= [HW,X86-64] 330 Pass parameters to the AMD IOMMU driver in the system. 331 Possible values are: 332 fullflush - Deprecated, equivalent to iommu.strict=1 333 off - do not initialize any AMD IOMMU found in 334 the system 335 force_isolation - Force device isolation for all 336 devices. The IOMMU driver is not 337 allowed anymore to lift isolation 338 requirements as needed. This option 339 does not override iommu=pt 340 force_enable - Force enable the IOMMU on platforms known 341 to be buggy with IOMMU enabled. Use this 342 option with care. 343 pgtbl_v1 - Use v1 page table for DMA-API (Default). 344 pgtbl_v2 - Use v2 page table for DMA-API. 345 irtcachedis - Disable Interrupt Remapping Table (IRT) caching. 346 nohugepages - Limit page-sizes used for v1 page-tables 347 to 4 KiB. 348 v2_pgsizes_only - Limit page-sizes used for v1 page-tables 349 to 4KiB/2Mib/1GiB. 350 351 352 amd_iommu_dump= [HW,X86-64] 353 Enable AMD IOMMU driver option to dump the ACPI table 354 for AMD IOMMU. With this option enabled, AMD IOMMU 355 driver will print ACPI tables for AMD IOMMU during 356 IOMMU initialization. 357 358 amd_iommu_intr= [HW,X86-64] 359 Specifies one of the following AMD IOMMU interrupt 360 remapping modes: 361 legacy - Use legacy interrupt remapping mode. 362 vapic - Use virtual APIC mode, which allows IOMMU 363 to inject interrupts directly into guest. 364 This mode requires kvm-amd.avic=1. 365 (Default when IOMMU HW support is present.) 366 367 amd_pstate= [X86,EARLY] 368 disable 369 Do not enable amd_pstate as the default 370 scaling driver for the supported processors 371 passive 372 Use amd_pstate with passive mode as a scaling driver. 373 In this mode autonomous selection is disabled. 374 Driver requests a desired performance level and platform 375 tries to match the same performance level if it is 376 satisfied by guaranteed performance level. 377 active 378 Use amd_pstate_epp driver instance as the scaling driver, 379 driver provides a hint to the hardware if software wants 380 to bias toward performance (0x0) or energy efficiency (0xff) 381 to the CPPC firmware. then CPPC power algorithm will 382 calculate the runtime workload and adjust the realtime cores 383 frequency. 384 guided 385 Activate guided autonomous mode. Driver requests minimum and 386 maximum performance level and the platform autonomously 387 selects a performance level in this range and appropriate 388 to the current workload. 389 390 amd_prefcore= 391 [X86] 392 disable 393 Disable amd-pstate preferred core. 394 395 amijoy.map= [HW,JOY] Amiga joystick support 396 Map of devices attached to JOY0DAT and JOY1DAT 397 Format: <a>,<b> 398 See also Documentation/input/joydev/joystick.rst 399 400 analog.map= [HW,JOY] Analog joystick and gamepad support 401 Specifies type or capabilities of an analog joystick 402 connected to one of 16 gameports 403 Format: <type1>,<type2>,..<type16> 404 405 apc= [HW,SPARC] 406 Power management functions (SPARCstation-4/5 + deriv.) 407 Format: noidle 408 Disable APC CPU standby support. SPARCstation-Fox does 409 not play well with APC CPU idle - disable it if you have 410 APC and your system crashes randomly. 411 412 apic [APIC,X86-64] Use IO-APIC. Default. 413 414 apic= [APIC,X86,EARLY] Advanced Programmable Interrupt Controller 415 Change the output verbosity while booting 416 Format: { quiet (default) | verbose | debug } 417 Change the amount of debugging information output 418 when initialising the APIC and IO-APIC components. 419 For X86-32, this can also be used to specify an APIC 420 driver name. 421 Format: apic=driver_name 422 Examples: apic=bigsmp 423 424 apic_extnmi= [APIC,X86,EARLY] External NMI delivery setting 425 Format: { bsp (default) | all | none } 426 bsp: External NMI is delivered only to CPU 0 427 all: External NMIs are broadcast to all CPUs as a 428 backup of CPU 0 429 none: External NMI is masked for all CPUs. This is 430 useful so that a dump capture kernel won't be 431 shot down by NMI 432 433 apicpmtimer Do APIC timer calibration using the pmtimer. Implies 434 apicmaintimer. Useful when your PIT timer is totally 435 broken. 436 437 autoconf= [IPV6] 438 See Documentation/networking/ipv6.rst. 439 440 apm= [APM] Advanced Power Management 441 See header of arch/x86/kernel/apm_32.c. 442 443 apparmor= [APPARMOR] Disable or enable AppArmor at boot time 444 Format: { "0" | "1" } 445 See security/apparmor/Kconfig help text 446 0 -- disable. 447 1 -- enable. 448 Default value is set via kernel config option. 449 450 arcrimi= [HW,NET] ARCnet - "RIM I" (entirely mem-mapped) cards 451 Format: <io>,<irq>,<nodeID> 452 453 arm64.no32bit_el0 [ARM64] Unconditionally disable the execution of 454 32 bit applications. 455 456 arm64.nobti [ARM64] Unconditionally disable Branch Target 457 Identification support 458 459 arm64.nogcs [ARM64] Unconditionally disable Guarded Control Stack 460 support 461 462 arm64.nomops [ARM64] Unconditionally disable Memory Copy and Memory 463 Set instructions support 464 465 arm64.nomte [ARM64] Unconditionally disable Memory Tagging Extension 466 support 467 468 arm64.nopauth [ARM64] Unconditionally disable Pointer Authentication 469 support 470 471 arm64.nosme [ARM64] Unconditionally disable Scalable Matrix 472 Extension support 473 474 arm64.nosve [ARM64] Unconditionally disable Scalable Vector 475 Extension support 476 477 ataflop= [HW,M68k] 478 479 atarimouse= [HW,MOUSE] Atari Mouse 480 481 atkbd.extra= [HW] Enable extra LEDs and keys on IBM RapidAccess, 482 EzKey and similar keyboards 483 484 atkbd.reset= [HW] Reset keyboard during initialization 485 486 atkbd.set= [HW] Select keyboard code set 487 Format: <int> (2 = AT (default), 3 = PS/2) 488 489 atkbd.scroll= [HW] Enable scroll wheel on MS Office and similar 490 keyboards 491 492 atkbd.softraw= [HW] Choose between synthetic and real raw mode 493 Format: <bool> (0 = real, 1 = synthetic (default)) 494 495 atkbd.softrepeat= [HW] 496 Use software keyboard repeat 497 498 audit= [KNL] Enable the audit sub-system 499 Format: { "0" | "1" | "off" | "on" } 500 0 | off - kernel audit is disabled and can not be 501 enabled until the next reboot 502 unset - kernel audit is initialized but disabled and 503 will be fully enabled by the userspace auditd. 504 1 | on - kernel audit is initialized and partially 505 enabled, storing at most audit_backlog_limit 506 messages in RAM until it is fully enabled by the 507 userspace auditd. 508 Default: unset 509 510 audit_backlog_limit= [KNL] Set the audit queue size limit. 511 Format: <int> (must be >=0) 512 Default: 64 513 514 bau= [X86_UV] Enable the BAU on SGI UV. The default 515 behavior is to disable the BAU (i.e. bau=0). 516 Format: { "0" | "1" } 517 0 - Disable the BAU. 518 1 - Enable the BAU. 519 unset - Disable the BAU. 520 521 baycom_epp= [HW,AX25] 522 Format: <io>,<mode> 523 524 baycom_par= [HW,AX25] BayCom Parallel Port AX.25 Modem 525 Format: <io>,<mode> 526 See header of drivers/net/hamradio/baycom_par.c. 527 528 baycom_ser_fdx= [HW,AX25] 529 BayCom Serial Port AX.25 Modem (Full Duplex Mode) 530 Format: <io>,<irq>,<mode>[,<baud>] 531 See header of drivers/net/hamradio/baycom_ser_fdx.c. 532 533 baycom_ser_hdx= [HW,AX25] 534 BayCom Serial Port AX.25 Modem (Half Duplex Mode) 535 Format: <io>,<irq>,<mode> 536 See header of drivers/net/hamradio/baycom_ser_hdx.c. 537 538 bdev_allow_write_mounted= 539 Format: <bool> 540 Control the ability to open a mounted block device 541 for writing, i.e., allow / disallow writes that bypass 542 the FS. This was implemented as a means to prevent 543 fuzzers from crashing the kernel by overwriting the 544 metadata underneath a mounted FS without its awareness. 545 This also prevents destructive formatting of mounted 546 filesystems by naive storage tooling that don't use 547 O_EXCL. Default is Y and can be changed through the 548 Kconfig option CONFIG_BLK_DEV_WRITE_MOUNTED. 549 550 bert_disable [ACPI] 551 Disable BERT OS support on buggy BIOSes. 552 553 bgrt_disable [ACPI,X86,EARLY] 554 Disable BGRT to avoid flickering OEM logo. 555 556 blkdevparts= Manual partition parsing of block device(s) for 557 embedded devices based on command line input. 558 See Documentation/block/cmdline-partition.rst 559 560 boot_delay= [KNL,EARLY] 561 Milliseconds to delay each printk during boot. 562 Only works if CONFIG_BOOT_PRINTK_DELAY is enabled, 563 and you may also have to specify "lpj=". Boot_delay 564 values larger than 10 seconds (10000) are assumed 565 erroneous and ignored. 566 Format: integer 567 568 bootconfig [KNL,EARLY] 569 Extended command line options can be added to an initrd 570 and this will cause the kernel to look for it. 571 572 See Documentation/admin-guide/bootconfig.rst 573 574 bttv.card= [HW,V4L] bttv (bt848 + bt878 based grabber cards) 575 bttv.radio= Most important insmod options are available as 576 kernel args too. 577 bttv.pll= See Documentation/admin-guide/media/bttv.rst 578 bttv.tuner= 579 580 bulk_remove=off [PPC] This parameter disables the use of the pSeries 581 firmware feature for flushing multiple hpte entries 582 at a time. 583 584 c101= [NET] Moxa C101 synchronous serial card 585 586 cachesize= [BUGS=X86-32] Override level 2 CPU cache size detection. 587 Sometimes CPU hardware bugs make them report the cache 588 size incorrectly. The kernel will attempt work arounds 589 to fix known problems, but for some CPUs it is not 590 possible to determine what the correct size should be. 591 This option provides an override for these situations. 592 593 carrier_timeout= 594 [NET] Specifies amount of time (in seconds) that 595 the kernel should wait for a network carrier. By default 596 it waits 120 seconds. 597 598 ca_keys= [KEYS] This parameter identifies a specific key(s) on 599 the system trusted keyring to be used for certificate 600 trust validation. 601 format: { id:<keyid> | builtin } 602 603 cca= [MIPS,EARLY] Override the kernel pages' cache coherency 604 algorithm. Accepted values range from 0 to 7 605 inclusive. See arch/mips/include/asm/pgtable-bits.h 606 for platform specific values (SB1, Loongson3 and 607 others). 608 609 ccw_timeout_log [S390] 610 See Documentation/arch/s390/common_io.rst for details. 611 612 cgroup_disable= [KNL] Disable a particular controller or optional feature 613 Format: {name of the controller(s) or feature(s) to disable} 614 The effects of cgroup_disable=foo are: 615 - foo isn't auto-mounted if you mount all cgroups in 616 a single hierarchy 617 - foo isn't visible as an individually mountable 618 subsystem 619 - if foo is an optional feature then the feature is 620 disabled and corresponding cgroup files are not 621 created 622 {Currently only "memory" controller deal with this and 623 cut the overhead, others just disable the usage. So 624 only cgroup_disable=memory is actually worthy} 625 Specifying "pressure" disables per-cgroup pressure 626 stall information accounting feature 627 628 cgroup_no_v1= [KNL] Disable cgroup controllers and named hierarchies in v1 629 Format: { { controller | "all" | "named" } 630 [,{ controller | "all" | "named" }...] } 631 Like cgroup_disable, but only applies to cgroup v1; 632 the blacklisted controllers remain available in cgroup2. 633 "all" blacklists all controllers and "named" disables 634 named mounts. Specifying both "all" and "named" disables 635 all v1 hierarchies. 636 637 cgroup_favordynmods= [KNL] Enable or Disable favordynmods. 638 Format: { "true" | "false" } 639 Defaults to the value of CONFIG_CGROUP_FAVOR_DYNMODS. 640 641 cgroup.memory= [KNL] Pass options to the cgroup memory controller. 642 Format: <string> 643 nosocket -- Disable socket memory accounting. 644 nokmem -- Disable kernel memory accounting. 645 nobpf -- Disable BPF memory accounting. 646 647 checkreqprot= [SELINUX] Set initial checkreqprot flag value. 648 Format: { "0" | "1" } 649 See security/selinux/Kconfig help text. 650 0 -- check protection applied by kernel (includes 651 any implied execute protection). 652 1 -- check protection requested by application. 653 Default value is set via a kernel config option. 654 Value can be changed at runtime via 655 /sys/fs/selinux/checkreqprot. 656 Setting checkreqprot to 1 is deprecated. 657 658 cio_ignore= [S390] 659 See Documentation/arch/s390/common_io.rst for details. 660 661 clearcpuid=X[,X...] [X86] 662 Disable CPUID feature X for the kernel. See 663 arch/x86/include/asm/cpufeatures.h for the valid bit 664 numbers X. Note the Linux-specific bits are not necessarily 665 stable over kernel options, but the vendor-specific 666 ones should be. 667 X can also be a string as appearing in the flags: line 668 in /proc/cpuinfo which does not have the above 669 instability issue. However, not all features have names 670 in /proc/cpuinfo. 671 Note that using this option will taint your kernel. 672 Also note that user programs calling CPUID directly 673 or using the feature without checking anything 674 will still see it. This just prevents it from 675 being used by the kernel or shown in /proc/cpuinfo. 676 Also note the kernel might malfunction if you disable 677 some critical bits. 678 679 clk_ignore_unused 680 [CLK] 681 Prevents the clock framework from automatically gating 682 clocks that have not been explicitly enabled by a Linux 683 device driver but are enabled in hardware at reset or 684 by the bootloader/firmware. Note that this does not 685 force such clocks to be always-on nor does it reserve 686 those clocks in any way. This parameter is useful for 687 debug and development, but should not be needed on a 688 platform with proper driver support. For more 689 information, see Documentation/driver-api/clk.rst. 690 691 clock= [BUGS=X86-32, HW] gettimeofday clocksource override. 692 [Deprecated] 693 Forces specified clocksource (if available) to be used 694 when calculating gettimeofday(). If specified 695 clocksource is not available, it defaults to PIT. 696 Format: { pit | tsc | cyclone | pmtmr } 697 698 clocksource= Override the default clocksource 699 Format: <string> 700 Override the default clocksource and use the clocksource 701 with the name specified. 702 Some clocksource names to choose from, depending on 703 the platform: 704 [all] jiffies (this is the base, fallback clocksource) 705 [ACPI] acpi_pm 706 [ARM] imx_timer1,OSTS,netx_timer,mpu_timer2, 707 pxa_timer,timer3,32k_counter,timer0_1 708 [X86-32] pit,hpet,tsc; 709 scx200_hrt on Geode; cyclone on IBM x440 710 [MIPS] MIPS 711 [PARISC] cr16 712 [S390] tod 713 [SH] SuperH 714 [SPARC64] tick 715 [X86-64] hpet,tsc 716 717 clocksource.arm_arch_timer.evtstrm= 718 [ARM,ARM64,EARLY] 719 Format: <bool> 720 Enable/disable the eventstream feature of the ARM 721 architected timer so that code using WFE-based polling 722 loops can be debugged more effectively on production 723 systems. 724 725 clocksource.verify_n_cpus= [KNL] 726 Limit the number of CPUs checked for clocksources 727 marked with CLOCK_SOURCE_VERIFY_PERCPU that 728 are marked unstable due to excessive skew. 729 A negative value says to check all CPUs, while 730 zero says not to check any. Values larger than 731 nr_cpu_ids are silently truncated to nr_cpu_ids. 732 The actual CPUs are chosen randomly, with 733 no replacement if the same CPU is chosen twice. 734 735 clocksource-wdtest.holdoff= [KNL] 736 Set the time in seconds that the clocksource 737 watchdog test waits before commencing its tests. 738 Defaults to zero when built as a module and to 739 10 seconds when built into the kernel. 740 741 cma=nn[MG]@[start[MG][-end[MG]]] 742 [KNL,CMA,EARLY] 743 Sets the size of kernel global memory area for 744 contiguous memory allocations and optionally the 745 placement constraint by the physical address range of 746 memory allocations. A value of 0 disables CMA 747 altogether. For more information, see 748 kernel/dma/contiguous.c 749 750 cma_pernuma=nn[MG] 751 [KNL,CMA,EARLY] 752 Sets the size of kernel per-numa memory area for 753 contiguous memory allocations. A value of 0 disables 754 per-numa CMA altogether. And If this option is not 755 specified, the default value is 0. 756 With per-numa CMA enabled, DMA users on node nid will 757 first try to allocate buffer from the pernuma area 758 which is located in node nid, if the allocation fails, 759 they will fallback to the global default memory area. 760 761 numa_cma=<node>:nn[MG][,<node>:nn[MG]] 762 [KNL,CMA,EARLY] 763 Sets the size of kernel numa memory area for 764 contiguous memory allocations. It will reserve CMA 765 area for the specified node. 766 767 With numa CMA enabled, DMA users on node nid will 768 first try to allocate buffer from the numa area 769 which is located in node nid, if the allocation fails, 770 they will fallback to the global default memory area. 771 772 cmo_free_hint= [PPC] Format: { yes | no } 773 Specify whether pages are marked as being inactive 774 when they are freed. This is used in CMO environments 775 to determine OS memory pressure for page stealing by 776 a hypervisor. 777 Default: yes 778 779 coherent_pool=nn[KMG] [ARM,KNL,EARLY] 780 Sets the size of memory pool for coherent, atomic dma 781 allocations, by default set to 256K. 782 783 com20020= [HW,NET] ARCnet - COM20020 chipset 784 Format: 785 <io>[,<irq>[,<nodeID>[,<backplane>[,<ckp>[,<timeout>]]]]] 786 787 com90io= [HW,NET] ARCnet - COM90xx chipset (IO-mapped buffers) 788 Format: <io>[,<irq>] 789 790 com90xx= [HW,NET] 791 ARCnet - COM90xx chipset (memory-mapped buffers) 792 Format: <io>[,<irq>[,<memstart>]] 793 794 condev= [HW,S390] console device 795 conmode= 796 797 con3215_drop= [S390,EARLY] 3215 console drop mode. 798 Format: y|n|Y|N|1|0 799 When set to true, drop data on the 3215 console when 800 the console buffer is full. In this case the 801 operator using a 3270 terminal emulator (for example 802 x3270) does not have to enter the clear key for the 803 console output to advance and the kernel to continue. 804 This leads to a much faster boot time when a 3270 805 terminal emulator is active. If no 3270 terminal 806 emulator is used, this parameter has no effect. 807 808 console= [KNL] Output console device and options. 809 810 tty<n> Use the virtual console device <n>. 811 812 ttyS<n>[,options] 813 ttyUSB0[,options] 814 Use the specified serial port. The options are of 815 the form "bbbbpnf", where "bbbb" is the baud rate, 816 "p" is parity ("n", "o", or "e"), "n" is number of 817 bits, and "f" is flow control ("r" for RTS or 818 omit it). Default is "9600n8". 819 820 See Documentation/admin-guide/serial-console.rst for more 821 information. See 822 Documentation/networking/netconsole.rst for an 823 alternative. 824 825 <DEVNAME>:<n>.<n>[,options] 826 Use the specified serial port on the serial core bus. 827 The addressing uses DEVNAME of the physical serial port 828 device, followed by the serial core controller instance, 829 and the serial port instance. The options are the same 830 as documented for the ttyS addressing above. 831 832 The mapping of the serial ports to the tty instances 833 can be viewed with: 834 835 $ ls -d /sys/bus/serial-base/devices/*:*.*/tty/* 836 /sys/bus/serial-base/devices/00:04:0.0/tty/ttyS0 837 838 In the above example, the console can be addressed with 839 console=00:04:0.0. Note that a console addressed this 840 way will only get added when the related device driver 841 is ready. The use of an earlycon parameter in addition to 842 the console may be desired for console output early on. 843 844 uart[8250],io,<addr>[,options] 845 uart[8250],mmio,<addr>[,options] 846 uart[8250],mmio16,<addr>[,options] 847 uart[8250],mmio32,<addr>[,options] 848 uart[8250],0x<addr>[,options] 849 Start an early, polled-mode console on the 8250/16550 850 UART at the specified I/O port or MMIO address, 851 switching to the matching ttyS device later. 852 MMIO inter-register address stride is either 8-bit 853 (mmio), 16-bit (mmio16), or 32-bit (mmio32). 854 If none of [io|mmio|mmio16|mmio32], <addr> is assumed 855 to be equivalent to 'mmio'. 'options' are specified in 856 the same format described for ttyS above; if unspecified, 857 the h/w is not re-initialized. 858 859 hvc<n> Use the hypervisor console device <n>. This is for 860 both Xen and PowerPC hypervisors. 861 862 { null | "" } 863 Use to disable console output, i.e., to have kernel 864 console messages discarded. 865 This must be the only console= parameter used on the 866 kernel command line. 867 868 If the device connected to the port is not a TTY but a braille 869 device, prepend "brl," before the device type, for instance 870 console=brl,ttyS0 871 For now, only VisioBraille is supported. 872 873 console_msg_format= 874 [KNL] Change console messages format 875 default 876 By default we print messages on consoles in 877 "[time stamp] text\n" format (time stamp may not be 878 printed, depending on CONFIG_PRINTK_TIME or 879 `printk_time' param). 880 syslog 881 Switch to syslog format: "<%u>[time stamp] text\n" 882 IOW, each message will have a facility and loglevel 883 prefix. The format is similar to one used by syslog() 884 syscall, or to executing "dmesg -S --raw" or to reading 885 from /proc/kmsg. 886 887 consoleblank= [KNL] The console blank (screen saver) timeout in 888 seconds. A value of 0 disables the blank timer. 889 Defaults to 0. 890 891 coredump_filter= 892 [KNL] Change the default value for 893 /proc/<pid>/coredump_filter. 894 See also Documentation/filesystems/proc.rst. 895 896 coresight_cpu_debug.enable 897 [ARM,ARM64] 898 Format: <bool> 899 Enable/disable the CPU sampling based debugging. 900 0: default value, disable debugging 901 1: enable debugging at boot time 902 903 cpcihp_generic= [HW,PCI] Generic port I/O CompactPCI driver 904 Format: 905 <first_slot>,<last_slot>,<port>,<enum_bit>[,<debug>] 906 907 cpuidle.off=1 [CPU_IDLE] 908 disable the cpuidle sub-system 909 910 cpuidle.governor= 911 [CPU_IDLE] Name of the cpuidle governor to use. 912 913 cpufreq.off=1 [CPU_FREQ] 914 disable the cpufreq sub-system 915 916 cpufreq.default_governor= 917 [CPU_FREQ] Name of the default cpufreq governor or 918 policy to use. This governor must be registered in the 919 kernel before the cpufreq driver probes. 920 921 cpu_init_udelay=N 922 [X86,EARLY] Delay for N microsec between assert and de-assert 923 of APIC INIT to start processors. This delay occurs 924 on every CPU online, such as boot, and resume from suspend. 925 Default: 10000 926 927 cpuhp.parallel= 928 [SMP] Enable/disable parallel bringup of secondary CPUs 929 Format: <bool> 930 Default is enabled if CONFIG_HOTPLUG_PARALLEL=y. Otherwise 931 the parameter has no effect. 932 933 crash_kexec_post_notifiers 934 Only jump to kdump kernel after running the panic 935 notifiers and dumping kmsg. This option increases 936 the risks of a kdump failure, since some panic 937 notifiers can make the crashed kernel more unstable. 938 In configurations where kdump may not be reliable, 939 running the panic notifiers could allow collecting 940 more data on dmesg, like stack traces from other CPUS 941 or extra data dumped by panic_print. Note that some 942 configurations enable this option unconditionally, 943 like Hyper-V, PowerPC (fadump) and AMD SEV-SNP. 944 945 crashkernel=size[KMG][@offset[KMG]] 946 [KNL,EARLY] Using kexec, Linux can switch to a 'crash kernel' 947 upon panic. This parameter reserves the physical 948 memory region [offset, offset + size] for that kernel 949 image. If '@offset' is omitted, then a suitable offset 950 is selected automatically. 951 [KNL, X86-64, ARM64, RISCV, LoongArch] Select a region 952 under 4G first, and fall back to reserve region above 953 4G when '@offset' hasn't been specified. 954 See Documentation/admin-guide/kdump/kdump.rst for further details. 955 956 crashkernel=range1:size1[,range2:size2,...][@offset] 957 [KNL] Same as above, but depends on the memory 958 in the running system. The syntax of range is 959 start-[end] where start and end are both 960 a memory unit (amount[KMG]). See also 961 Documentation/admin-guide/kdump/kdump.rst for an example. 962 963 crashkernel=size[KMG],high 964 [KNL, X86-64, ARM64, RISCV, LoongArch] range could be 965 above 4G. 966 Allow kernel to allocate physical memory region from top, 967 so could be above 4G if system have more than 4G ram 968 installed. Otherwise memory region will be allocated 969 below 4G, if available. 970 It will be ignored if crashkernel=X is specified. 971 crashkernel=size[KMG],low 972 [KNL, X86-64, ARM64, RISCV, LoongArch] range under 4G. 973 When crashkernel=X,high is passed, kernel could allocate 974 physical memory region above 4G, that cause second kernel 975 crash on system that require some amount of low memory, 976 e.g. swiotlb requires at least 64M+32K low memory, also 977 enough extra low memory is needed to make sure DMA buffers 978 for 32-bit devices won't run out. Kernel would try to allocate 979 default size of memory below 4G automatically. The default 980 size is platform dependent. 981 --> x86: max(swiotlb_size_or_default() + 8MiB, 256MiB) 982 --> arm64: 128MiB 983 --> riscv: 128MiB 984 --> loongarch: 128MiB 985 This one lets the user specify own low range under 4G 986 for second kernel instead. 987 0: to disable low allocation. 988 It will be ignored when crashkernel=X,high is not used 989 or memory reserved is below 4G. 990 991 cryptomgr.notests 992 [KNL] Disable crypto self-tests 993 994 cs89x0_dma= [HW,NET] 995 Format: <dma> 996 997 cs89x0_media= [HW,NET] 998 Format: { rj45 | aui | bnc } 999 1000 csdlock_debug= [KNL] Enable or disable debug add-ons of cross-CPU 1001 function call handling. When switched on, 1002 additional debug data is printed to the console 1003 in case a hanging CPU is detected, and that 1004 CPU is pinged again in order to try to resolve 1005 the hang situation. The default value of this 1006 option depends on the CSD_LOCK_WAIT_DEBUG_DEFAULT 1007 Kconfig option. 1008 1009 dasd= [HW,NET] 1010 See header of drivers/s390/block/dasd_devmap.c. 1011 1012 db9.dev[2|3]= [HW,JOY] Multisystem joystick support via parallel port 1013 (one device per port) 1014 Format: <port#>,<type> 1015 See also Documentation/input/devices/joystick-parport.rst 1016 1017 debug [KNL,EARLY] Enable kernel debugging (events log level). 1018 1019 debug_boot_weak_hash 1020 [KNL,EARLY] Enable printing [hashed] pointers early in the 1021 boot sequence. If enabled, we use a weak hash instead 1022 of siphash to hash pointers. Use this option if you are 1023 seeing instances of '(___ptrval___)') and need to see a 1024 value (hashed pointer) instead. Cryptographically 1025 insecure, please do not use on production kernels. 1026 1027 debug_locks_verbose= 1028 [KNL] verbose locking self-tests 1029 Format: <int> 1030 Print debugging info while doing the locking API 1031 self-tests. 1032 Bitmask for the various LOCKTYPE_ tests. Defaults to 0 1033 (no extra messages), setting it to -1 (all bits set) 1034 will print _a_lot_ more information - normally only 1035 useful to lockdep developers. 1036 1037 debug_objects [KNL,EARLY] Enable object debugging 1038 1039 debug_guardpage_minorder= 1040 [KNL,EARLY] When CONFIG_DEBUG_PAGEALLOC is set, this 1041 parameter allows control of the order of pages that will 1042 be intentionally kept free (and hence protected) by the 1043 buddy allocator. Bigger value increase the probability 1044 of catching random memory corruption, but reduce the 1045 amount of memory for normal system use. The maximum 1046 possible value is MAX_PAGE_ORDER/2. Setting this 1047 parameter to 1 or 2 should be enough to identify most 1048 random memory corruption problems caused by bugs in 1049 kernel or driver code when a CPU writes to (or reads 1050 from) a random memory location. Note that there exists 1051 a class of memory corruptions problems caused by buggy 1052 H/W or F/W or by drivers badly programming DMA 1053 (basically when memory is written at bus level and the 1054 CPU MMU is bypassed) which are not detectable by 1055 CONFIG_DEBUG_PAGEALLOC, hence this option will not 1056 help tracking down these problems. 1057 1058 debug_pagealloc= 1059 [KNL,EARLY] When CONFIG_DEBUG_PAGEALLOC is set, this parameter 1060 enables the feature at boot time. By default, it is 1061 disabled and the system will work mostly the same as a 1062 kernel built without CONFIG_DEBUG_PAGEALLOC. 1063 Note: to get most of debug_pagealloc error reports, it's 1064 useful to also enable the page_owner functionality. 1065 on: enable the feature 1066 1067 debugfs= [KNL,EARLY] This parameter enables what is exposed to 1068 userspace and debugfs internal clients. 1069 Format: { on, no-mount, off } 1070 on: All functions are enabled. 1071 no-mount: 1072 Filesystem is not registered but kernel clients can 1073 access APIs and a crashkernel can be used to read 1074 its content. There is nothing to mount. 1075 off: Filesystem is not registered and clients 1076 get a -EPERM as result when trying to register files 1077 or directories within debugfs. 1078 This is equivalent of the runtime functionality if 1079 debugfs was not enabled in the kernel at all. 1080 Default value is set in build-time with a kernel configuration. 1081 1082 debugpat [X86] Enable PAT debugging 1083 1084 default_hugepagesz= 1085 [HW] The size of the default HugeTLB page. This is 1086 the size represented by the legacy /proc/ hugepages 1087 APIs. In addition, this is the default hugetlb size 1088 used for shmget(), mmap() and mounting hugetlbfs 1089 filesystems. If not specified, defaults to the 1090 architecture's default huge page size. Huge page 1091 sizes are architecture dependent. See also 1092 Documentation/admin-guide/mm/hugetlbpage.rst. 1093 Format: size[KMG] 1094 1095 deferred_probe_timeout= 1096 [KNL] Debugging option to set a timeout in seconds for 1097 deferred probe to give up waiting on dependencies to 1098 probe. Only specific dependencies (subsystems or 1099 drivers) that have opted in will be ignored. A timeout 1100 of 0 will timeout at the end of initcalls. If the time 1101 out hasn't expired, it'll be restarted by each 1102 successful driver registration. This option will also 1103 dump out devices still on the deferred probe list after 1104 retrying. 1105 1106 delayacct [KNL] Enable per-task delay accounting 1107 1108 dell_smm_hwmon.ignore_dmi= 1109 [HW] Continue probing hardware even if DMI data 1110 indicates that the driver is running on unsupported 1111 hardware. 1112 1113 dell_smm_hwmon.force= 1114 [HW] Activate driver even if SMM BIOS signature does 1115 not match list of supported models and enable otherwise 1116 blacklisted features. 1117 1118 dell_smm_hwmon.power_status= 1119 [HW] Report power status in /proc/i8k 1120 (disabled by default). 1121 1122 dell_smm_hwmon.restricted= 1123 [HW] Allow controlling fans only if SYS_ADMIN 1124 capability is set. 1125 1126 dell_smm_hwmon.fan_mult= 1127 [HW] Factor to multiply fan speed with. 1128 1129 dell_smm_hwmon.fan_max= 1130 [HW] Maximum configurable fan speed. 1131 1132 dfltcc= [HW,S390] 1133 Format: { on | off | def_only | inf_only | always } 1134 on: s390 zlib hardware support for compression on 1135 level 1 and decompression (default) 1136 off: No s390 zlib hardware support 1137 def_only: s390 zlib hardware support for deflate 1138 only (compression on level 1) 1139 inf_only: s390 zlib hardware support for inflate 1140 only (decompression) 1141 always: Same as 'on' but ignores the selected compression 1142 level always using hardware support (used for debugging) 1143 1144 dhash_entries= [KNL] 1145 Set number of hash buckets for dentry cache. 1146 1147 disable_1tb_segments [PPC,EARLY] 1148 Disables the use of 1TB hash page table segments. This 1149 causes the kernel to fall back to 256MB segments which 1150 can be useful when debugging issues that require an SLB 1151 miss to occur. 1152 1153 disable= [IPV6] 1154 See Documentation/networking/ipv6.rst. 1155 1156 disable_radix [PPC,EARLY] 1157 Disable RADIX MMU mode on POWER9 1158 1159 disable_tlbie [PPC] 1160 Disable TLBIE instruction. Currently does not work 1161 with KVM, with HASH MMU, or with coherent accelerators. 1162 1163 disable_ddw [PPC/PSERIES,EARLY] 1164 Disable Dynamic DMA Window support. Use this 1165 to workaround buggy firmware. 1166 1167 disable_ipv6= [IPV6] 1168 See Documentation/networking/ipv6.rst. 1169 1170 disable_mtrr_cleanup [X86,EARLY] 1171 The kernel tries to adjust MTRR layout from continuous 1172 to discrete, to make X server driver able to add WB 1173 entry later. This parameter disables that. 1174 1175 disable_mtrr_trim [X86, Intel and AMD only,EARLY] 1176 By default the kernel will trim any uncacheable 1177 memory out of your available memory pool based on 1178 MTRR settings. This parameter disables that behavior, 1179 possibly causing your machine to run very slowly. 1180 1181 disable_timer_pin_1 [X86,EARLY] 1182 Disable PIN 1 of APIC timer 1183 Can be useful to work around chipset bugs. 1184 1185 dis_ucode_ldr [X86] Disable the microcode loader. 1186 1187 dma_debug=off If the kernel is compiled with DMA_API_DEBUG support, 1188 this option disables the debugging code at boot. 1189 1190 dma_debug_entries=<number> 1191 This option allows to tune the number of preallocated 1192 entries for DMA-API debugging code. One entry is 1193 required per DMA-API allocation. Use this if the 1194 DMA-API debugging code disables itself because the 1195 architectural default is too low. 1196 1197 dma_debug_driver=<driver_name> 1198 With this option the DMA-API debugging driver 1199 filter feature can be enabled at boot time. Just 1200 pass the driver to filter for as the parameter. 1201 The filter can be disabled or changed to another 1202 driver later using sysfs. 1203 1204 reg_file_data_sampling= 1205 [X86] Controls mitigation for Register File Data 1206 Sampling (RFDS) vulnerability. RFDS is a CPU 1207 vulnerability which may allow userspace to infer 1208 kernel data values previously stored in floating point 1209 registers, vector registers, or integer registers. 1210 RFDS only affects Intel Atom processors. 1211 1212 on: Turns ON the mitigation. 1213 off: Turns OFF the mitigation. 1214 1215 This parameter overrides the compile time default set 1216 by CONFIG_MITIGATION_RFDS. Mitigation cannot be 1217 disabled when other VERW based mitigations (like MDS) 1218 are enabled. In order to disable RFDS mitigation all 1219 VERW based mitigations need to be disabled. 1220 1221 For details see: 1222 Documentation/admin-guide/hw-vuln/reg-file-data-sampling.rst 1223 1224 driver_async_probe= [KNL] 1225 List of driver names to be probed asynchronously. * 1226 matches with all driver names. If * is specified, the 1227 rest of the listed driver names are those that will NOT 1228 match the *. 1229 Format: <driver_name1>,<driver_name2>... 1230 1231 drm.edid_firmware=[<connector>:]<file>[,[<connector>:]<file>] 1232 Broken monitors, graphic adapters, KVMs and EDIDless 1233 panels may send no or incorrect EDID data sets. 1234 This parameter allows to specify an EDID data sets 1235 in the /lib/firmware directory that are used instead. 1236 An EDID data set will only be used for a particular 1237 connector, if its name and a colon are prepended to 1238 the EDID name. Each connector may use a unique EDID 1239 data set by separating the files with a comma. An EDID 1240 data set with no connector name will be used for 1241 any connectors not explicitly specified. 1242 1243 dscc4.setup= [NET] 1244 1245 dt_cpu_ftrs= [PPC,EARLY] 1246 Format: {"off" | "known"} 1247 Control how the dt_cpu_ftrs device-tree binding is 1248 used for CPU feature discovery and setup (if it 1249 exists). 1250 off: Do not use it, fall back to legacy cpu table. 1251 known: Do not pass through unknown features to guests 1252 or userspace, only those that the kernel is aware of. 1253 1254 dump_apple_properties [X86] 1255 Dump name and content of EFI device properties on 1256 x86 Macs. Useful for driver authors to determine 1257 what data is available or for reverse-engineering. 1258 1259 dyndbg[="val"] [KNL,DYNAMIC_DEBUG] 1260 <module>.dyndbg[="val"] 1261 Enable debug messages at boot time. See 1262 Documentation/admin-guide/dynamic-debug-howto.rst 1263 for details. 1264 1265 early_ioremap_debug [KNL,EARLY] 1266 Enable debug messages in early_ioremap support. This 1267 is useful for tracking down temporary early mappings 1268 which are not unmapped. 1269 1270 earlycon= [KNL,EARLY] Output early console device and options. 1271 1272 When used with no options, the early console is 1273 determined by stdout-path property in device tree's 1274 chosen node or the ACPI SPCR table if supported by 1275 the platform. 1276 1277 cdns,<addr>[,options] 1278 Start an early, polled-mode console on a Cadence 1279 (xuartps) serial port at the specified address. Only 1280 supported option is baud rate. If baud rate is not 1281 specified, the serial port must already be setup and 1282 configured. 1283 1284 uart[8250],io,<addr>[,options[,uartclk]] 1285 uart[8250],mmio,<addr>[,options[,uartclk]] 1286 uart[8250],mmio32,<addr>[,options[,uartclk]] 1287 uart[8250],mmio32be,<addr>[,options[,uartclk]] 1288 uart[8250],0x<addr>[,options] 1289 Start an early, polled-mode console on the 8250/16550 1290 UART at the specified I/O port or MMIO address. 1291 MMIO inter-register address stride is either 8-bit 1292 (mmio) or 32-bit (mmio32 or mmio32be). 1293 If none of [io|mmio|mmio32|mmio32be], <addr> is assumed 1294 to be equivalent to 'mmio'. 'options' are specified 1295 in the same format described for "console=ttyS<n>"; if 1296 unspecified, the h/w is not initialized. 'uartclk' is 1297 the uart clock frequency; if unspecified, it is set 1298 to 'BASE_BAUD' * 16. 1299 1300 pl011,<addr> 1301 pl011,mmio32,<addr> 1302 Start an early, polled-mode console on a pl011 serial 1303 port at the specified address. The pl011 serial port 1304 must already be setup and configured. Options are not 1305 yet supported. If 'mmio32' is specified, then only 1306 the driver will use only 32-bit accessors to read/write 1307 the device registers. 1308 1309 liteuart,<addr> 1310 Start an early console on a litex serial port at the 1311 specified address. The serial port must already be 1312 setup and configured. Options are not yet supported. 1313 1314 meson,<addr> 1315 Start an early, polled-mode console on a meson serial 1316 port at the specified address. The serial port must 1317 already be setup and configured. Options are not yet 1318 supported. 1319 1320 msm_serial,<addr> 1321 Start an early, polled-mode console on an msm serial 1322 port at the specified address. The serial port 1323 must already be setup and configured. Options are not 1324 yet supported. 1325 1326 msm_serial_dm,<addr> 1327 Start an early, polled-mode console on an msm serial 1328 dm port at the specified address. The serial port 1329 must already be setup and configured. Options are not 1330 yet supported. 1331 1332 owl,<addr> 1333 Start an early, polled-mode console on a serial port 1334 of an Actions Semi SoC, such as S500 or S900, at the 1335 specified address. The serial port must already be 1336 setup and configured. Options are not yet supported. 1337 1338 rda,<addr> 1339 Start an early, polled-mode console on a serial port 1340 of an RDA Micro SoC, such as RDA8810PL, at the 1341 specified address. The serial port must already be 1342 setup and configured. Options are not yet supported. 1343 1344 sbi 1345 Use RISC-V SBI (Supervisor Binary Interface) for early 1346 console. 1347 1348 smh Use ARM semihosting calls for early console. 1349 1350 s3c2410,<addr> 1351 s3c2412,<addr> 1352 s3c2440,<addr> 1353 s3c6400,<addr> 1354 s5pv210,<addr> 1355 exynos4210,<addr> 1356 Use early console provided by serial driver available 1357 on Samsung SoCs, requires selecting proper type and 1358 a correct base address of the selected UART port. The 1359 serial port must already be setup and configured. 1360 Options are not yet supported. 1361 1362 lantiq,<addr> 1363 Start an early, polled-mode console on a lantiq serial 1364 (lqasc) port at the specified address. The serial port 1365 must already be setup and configured. Options are not 1366 yet supported. 1367 1368 lpuart,<addr> 1369 lpuart32,<addr> 1370 Use early console provided by Freescale LP UART driver 1371 found on Freescale Vybrid and QorIQ LS1021A processors. 1372 A valid base address must be provided, and the serial 1373 port must already be setup and configured. 1374 1375 ec_imx21,<addr> 1376 ec_imx6q,<addr> 1377 Start an early, polled-mode, output-only console on the 1378 Freescale i.MX UART at the specified address. The UART 1379 must already be setup and configured. 1380 1381 ar3700_uart,<addr> 1382 Start an early, polled-mode console on the 1383 Armada 3700 serial port at the specified 1384 address. The serial port must already be setup 1385 and configured. Options are not yet supported. 1386 1387 qcom_geni,<addr> 1388 Start an early, polled-mode console on a Qualcomm 1389 Generic Interface (GENI) based serial port at the 1390 specified address. The serial port must already be 1391 setup and configured. Options are not yet supported. 1392 1393 efifb,[options] 1394 Start an early, unaccelerated console on the EFI 1395 memory mapped framebuffer (if available). On cache 1396 coherent non-x86 systems that use system memory for 1397 the framebuffer, pass the 'ram' option so that it is 1398 mapped with the correct attributes. 1399 1400 linflex,<addr> 1401 Use early console provided by Freescale LINFlexD UART 1402 serial driver for NXP S32V234 SoCs. A valid base 1403 address must be provided, and the serial port must 1404 already be setup and configured. 1405 1406 earlyprintk= [X86,SH,ARM,M68k,S390,UM,EARLY] 1407 earlyprintk=vga 1408 earlyprintk=sclp 1409 earlyprintk=xen 1410 earlyprintk=serial[,ttySn[,baudrate]] 1411 earlyprintk=serial[,0x...[,baudrate]] 1412 earlyprintk=ttySn[,baudrate] 1413 earlyprintk=dbgp[debugController#] 1414 earlyprintk=pciserial[,force],bus:device.function[,baudrate] 1415 earlyprintk=xdbc[xhciController#] 1416 earlyprintk=bios 1417 1418 earlyprintk is useful when the kernel crashes before 1419 the normal console is initialized. It is not enabled by 1420 default because it has some cosmetic problems. 1421 1422 Append ",keep" to not disable it when the real console 1423 takes over. 1424 1425 Only one of vga, serial, or usb debug port can 1426 be used at a time. 1427 1428 Currently only ttyS0 and ttyS1 may be specified by 1429 name. Other I/O ports may be explicitly specified 1430 on some architectures (x86 and arm at least) by 1431 replacing ttySn with an I/O port address, like this: 1432 earlyprintk=serial,0x1008,115200 1433 You can find the port for a given device in 1434 /proc/tty/driver/serial: 1435 2: uart:ST16650V2 port:00001008 irq:18 ... 1436 1437 Interaction with the standard serial driver is not 1438 very good. 1439 1440 The VGA output is eventually overwritten by 1441 the real console. 1442 1443 The xen option can only be used in Xen domains. 1444 1445 The sclp output can only be used on s390. 1446 1447 The bios output can only be used on SuperH. 1448 1449 The optional "force" to "pciserial" enables use of a 1450 PCI device even when its classcode is not of the 1451 UART class. 1452 1453 edac_report= [HW,EDAC] Control how to report EDAC event 1454 Format: {"on" | "off" | "force"} 1455 on: enable EDAC to report H/W event. May be overridden 1456 by other higher priority error reporting module. 1457 off: disable H/W event reporting through EDAC. 1458 force: enforce the use of EDAC to report H/W event. 1459 default: on. 1460 1461 edd= [EDD] 1462 Format: {"off" | "on" | "skip[mbr]"} 1463 1464 efi= [EFI,EARLY] 1465 Format: { "debug", "disable_early_pci_dma", 1466 "nochunk", "noruntime", "nosoftreserve", 1467 "novamap", "no_disable_early_pci_dma" } 1468 debug: enable misc debug output. 1469 disable_early_pci_dma: disable the busmaster bit on all 1470 PCI bridges while in the EFI boot stub. 1471 nochunk: disable reading files in "chunks" in the EFI 1472 boot stub, as chunking can cause problems with some 1473 firmware implementations. 1474 noruntime : disable EFI runtime services support 1475 nosoftreserve: The EFI_MEMORY_SP (Specific Purpose) 1476 attribute may cause the kernel to reserve the 1477 memory range for a memory mapping driver to 1478 claim. Specify efi=nosoftreserve to disable this 1479 reservation and treat the memory by its base type 1480 (i.e. EFI_CONVENTIONAL_MEMORY / "System RAM"). 1481 novamap: do not call SetVirtualAddressMap(). 1482 no_disable_early_pci_dma: Leave the busmaster bit set 1483 on all PCI bridges while in the EFI boot stub 1484 1485 efi_no_storage_paranoia [EFI,X86,EARLY] 1486 Using this parameter you can use more than 50% of 1487 your efi variable storage. Use this parameter only if 1488 you are really sure that your UEFI does sane gc and 1489 fulfills the spec otherwise your board may brick. 1490 1491 efivar_ssdt= [EFI; X86] Name of an EFI variable that contains an SSDT 1492 that is to be dynamically loaded by Linux. If there are 1493 multiple variables with the same name but with different 1494 vendor GUIDs, all of them will be loaded. See 1495 Documentation/admin-guide/acpi/ssdt-overlays.rst for details. 1496 1497 1498 eisa_irq_edge= [PARISC,HW] 1499 See header of drivers/parisc/eisa.c. 1500 1501 ekgdboc= [X86,KGDB,EARLY] Allow early kernel console debugging 1502 Format: ekgdboc=kbd 1503 1504 This is designed to be used in conjunction with 1505 the boot argument: earlyprintk=vga 1506 1507 This parameter works in place of the kgdboc parameter 1508 but can only be used if the backing tty is available 1509 very early in the boot process. For early debugging 1510 via a serial port see kgdboc_earlycon instead. 1511 1512 elanfreq= [X86-32] 1513 See comment before function elanfreq_setup() in 1514 arch/x86/kernel/cpu/cpufreq/elanfreq.c. 1515 1516 elfcorehdr=[size[KMG]@]offset[KMG] [PPC,SH,X86,S390,EARLY] 1517 Specifies physical address of start of kernel core 1518 image elf header and optionally the size. Generally 1519 kexec loader will pass this option to capture kernel. 1520 See Documentation/admin-guide/kdump/kdump.rst for details. 1521 1522 enable_mtrr_cleanup [X86,EARLY] 1523 The kernel tries to adjust MTRR layout from continuous 1524 to discrete, to make X server driver able to add WB 1525 entry later. This parameter enables that. 1526 1527 enable_timer_pin_1 [X86] 1528 Enable PIN 1 of APIC timer 1529 Can be useful to work around chipset bugs 1530 (in particular on some ATI chipsets). 1531 The kernel tries to set a reasonable default. 1532 1533 enforcing= [SELINUX] Set initial enforcing status. 1534 Format: {"0" | "1"} 1535 See security/selinux/Kconfig help text. 1536 0 -- permissive (log only, no denials). 1537 1 -- enforcing (deny and log). 1538 Default value is 0. 1539 Value can be changed at runtime via 1540 /sys/fs/selinux/enforce. 1541 1542 erst_disable [ACPI] 1543 Disable Error Record Serialization Table (ERST) 1544 support. 1545 1546 ether= [HW,NET] Ethernet cards parameters 1547 This option is obsoleted by the "netdev=" option, which 1548 has equivalent usage. See its documentation for details. 1549 1550 evm= [EVM] 1551 Format: { "fix" } 1552 Permit 'security.evm' to be updated regardless of 1553 current integrity status. 1554 1555 early_page_ext [KNL,EARLY] Enforces page_ext initialization to earlier 1556 stages so cover more early boot allocations. 1557 Please note that as side effect some optimizations 1558 might be disabled to achieve that (e.g. parallelized 1559 memory initialization is disabled) so the boot process 1560 might take longer, especially on systems with a lot of 1561 memory. Available with CONFIG_PAGE_EXTENSION=y. 1562 1563 failslab= 1564 fail_usercopy= 1565 fail_page_alloc= 1566 fail_skb_realloc= 1567 fail_make_request=[KNL] 1568 General fault injection mechanism. 1569 Format: <interval>,<probability>,<space>,<times> 1570 See also Documentation/fault-injection/. 1571 1572 fb_tunnels= [NET] 1573 Format: { initns | none } 1574 See Documentation/admin-guide/sysctl/net.rst for 1575 fb_tunnels_only_for_init_ns 1576 1577 floppy= [HW] 1578 See Documentation/admin-guide/blockdev/floppy.rst. 1579 1580 forcepae [X86-32] 1581 Forcefully enable Physical Address Extension (PAE). 1582 Many Pentium M systems disable PAE but may have a 1583 functionally usable PAE implementation. 1584 Warning: use of this parameter will taint the kernel 1585 and may cause unknown problems. 1586 1587 fred= [X86-64] 1588 Enable/disable Flexible Return and Event Delivery. 1589 Format: { on | off } 1590 on: enable FRED when it's present. 1591 off: disable FRED, the default setting. 1592 1593 ftrace=[tracer] 1594 [FTRACE] will set and start the specified tracer 1595 as early as possible in order to facilitate early 1596 boot debugging. 1597 1598 ftrace_boot_snapshot 1599 [FTRACE] On boot up, a snapshot will be taken of the 1600 ftrace ring buffer that can be read at: 1601 /sys/kernel/tracing/snapshot. 1602 This is useful if you need tracing information from kernel 1603 boot up that is likely to be overridden by user space 1604 start up functionality. 1605 1606 Optionally, the snapshot can also be defined for a tracing 1607 instance that was created by the trace_instance= command 1608 line parameter. 1609 1610 trace_instance=foo,sched_switch ftrace_boot_snapshot=foo 1611 1612 The above will cause the "foo" tracing instance to trigger 1613 a snapshot at the end of boot up. 1614 1615 ftrace_dump_on_oops[=2(orig_cpu) | =<instance>][,<instance> | 1616 ,<instance>=2(orig_cpu)] 1617 [FTRACE] will dump the trace buffers on oops. 1618 If no parameter is passed, ftrace will dump global 1619 buffers of all CPUs, if you pass 2 or orig_cpu, it 1620 will dump only the buffer of the CPU that triggered 1621 the oops, or the specific instance will be dumped if 1622 its name is passed. Multiple instance dump is also 1623 supported, and instances are separated by commas. Each 1624 instance supports only dump on CPU that triggered the 1625 oops by passing 2 or orig_cpu to it. 1626 1627 ftrace_dump_on_oops=foo=orig_cpu 1628 1629 The above will dump only the buffer of "foo" instance 1630 on CPU that triggered the oops. 1631 1632 ftrace_dump_on_oops,foo,bar=orig_cpu 1633 1634 The above will dump global buffer on all CPUs, the 1635 buffer of "foo" instance on all CPUs and the buffer 1636 of "bar" instance on CPU that triggered the oops. 1637 1638 ftrace_filter=[function-list] 1639 [FTRACE] Limit the functions traced by the function 1640 tracer at boot up. function-list is a comma-separated 1641 list of functions. This list can be changed at run 1642 time by the set_ftrace_filter file in the debugfs 1643 tracing directory. 1644 1645 ftrace_notrace=[function-list] 1646 [FTRACE] Do not trace the functions specified in 1647 function-list. This list can be changed at run time 1648 by the set_ftrace_notrace file in the debugfs 1649 tracing directory. 1650 1651 ftrace_graph_filter=[function-list] 1652 [FTRACE] Limit the top level callers functions traced 1653 by the function graph tracer at boot up. 1654 function-list is a comma-separated list of functions 1655 that can be changed at run time by the 1656 set_graph_function file in the debugfs tracing directory. 1657 1658 ftrace_graph_notrace=[function-list] 1659 [FTRACE] Do not trace from the functions specified in 1660 function-list. This list is a comma-separated list of 1661 functions that can be changed at run time by the 1662 set_graph_notrace file in the debugfs tracing directory. 1663 1664 ftrace_graph_max_depth=<uint> 1665 [FTRACE] Used with the function graph tracer. This is 1666 the max depth it will trace into a function. This value 1667 can be changed at run time by the max_graph_depth file 1668 in the tracefs tracing directory. default: 0 (no limit) 1669 1670 fw_devlink= [KNL,EARLY] Create device links between consumer and supplier 1671 devices by scanning the firmware to infer the 1672 consumer/supplier relationships. This feature is 1673 especially useful when drivers are loaded as modules as 1674 it ensures proper ordering of tasks like device probing 1675 (suppliers first, then consumers), supplier boot state 1676 clean up (only after all consumers have probed), 1677 suspend/resume & runtime PM (consumers first, then 1678 suppliers). 1679 Format: { off | permissive | on | rpm } 1680 off -- Don't create device links from firmware info. 1681 permissive -- Create device links from firmware info 1682 but use it only for ordering boot state clean 1683 up (sync_state() calls). 1684 on -- Create device links from firmware info and use it 1685 to enforce probe and suspend/resume ordering. 1686 rpm -- Like "on", but also use to order runtime PM. 1687 1688 fw_devlink.strict=<bool> 1689 [KNL,EARLY] Treat all inferred dependencies as mandatory 1690 dependencies. This only applies for fw_devlink=on|rpm. 1691 Format: <bool> 1692 1693 fw_devlink.sync_state = 1694 [KNL,EARLY] When all devices that could probe have finished 1695 probing, this parameter controls what to do with 1696 devices that haven't yet received their sync_state() 1697 calls. 1698 Format: { strict | timeout } 1699 strict -- Default. Continue waiting on consumers to 1700 probe successfully. 1701 timeout -- Give up waiting on consumers and call 1702 sync_state() on any devices that haven't yet 1703 received their sync_state() calls after 1704 deferred_probe_timeout has expired or by 1705 late_initcall() if !CONFIG_MODULES. 1706 1707 gamecon.map[2|3]= 1708 [HW,JOY] Multisystem joystick and NES/SNES/PSX pad 1709 support via parallel port (up to 5 devices per port) 1710 Format: <port#>,<pad1>,<pad2>,<pad3>,<pad4>,<pad5> 1711 See also Documentation/input/devices/joystick-parport.rst 1712 1713 gamma= [HW,DRM] 1714 1715 gart_fix_e820= [X86-64,EARLY] disable the fix e820 for K8 GART 1716 Format: off | on 1717 default: on 1718 1719 gather_data_sampling= 1720 [X86,INTEL,EARLY] Control the Gather Data Sampling (GDS) 1721 mitigation. 1722 1723 Gather Data Sampling is a hardware vulnerability which 1724 allows unprivileged speculative access to data which was 1725 previously stored in vector registers. 1726 1727 This issue is mitigated by default in updated microcode. 1728 The mitigation may have a performance impact but can be 1729 disabled. On systems without the microcode mitigation 1730 disabling AVX serves as a mitigation. 1731 1732 force: Disable AVX to mitigate systems without 1733 microcode mitigation. No effect if the microcode 1734 mitigation is present. Known to cause crashes in 1735 userspace with buggy AVX enumeration. 1736 1737 off: Disable GDS mitigation. 1738 1739 gbpages [X86] Use GB pages for kernel direct mappings. 1740 1741 gcov_persist= [GCOV] When non-zero (default), profiling data for 1742 kernel modules is saved and remains accessible via 1743 debugfs, even when the module is unloaded/reloaded. 1744 When zero, profiling data is discarded and associated 1745 debugfs files are removed at module unload time. 1746 1747 goldfish [X86] Enable the goldfish android emulator platform. 1748 Don't use this when you are not running on the 1749 android emulator 1750 1751 gpio-mockup.gpio_mockup_ranges 1752 [HW] Sets the ranges of gpiochip of for this device. 1753 Format: <start1>,<end1>,<start2>,<end2>... 1754 gpio-mockup.gpio_mockup_named_lines 1755 [HW] Let the driver know GPIO lines should be named. 1756 1757 gpt [EFI] Forces disk with valid GPT signature but 1758 invalid Protective MBR to be treated as GPT. If the 1759 primary GPT is corrupted, it enables the backup/alternate 1760 GPT to be used instead. 1761 1762 grcan.enable0= [HW] Configuration of physical interface 0. Determines 1763 the "Enable 0" bit of the configuration register. 1764 Format: 0 | 1 1765 Default: 0 1766 grcan.enable1= [HW] Configuration of physical interface 1. Determines 1767 the "Enable 0" bit of the configuration register. 1768 Format: 0 | 1 1769 Default: 0 1770 grcan.select= [HW] Select which physical interface to use. 1771 Format: 0 | 1 1772 Default: 0 1773 grcan.txsize= [HW] Sets the size of the tx buffer. 1774 Format: <unsigned int> such that (txsize & ~0x1fffc0) == 0. 1775 Default: 1024 1776 grcan.rxsize= [HW] Sets the size of the rx buffer. 1777 Format: <unsigned int> such that (rxsize & ~0x1fffc0) == 0. 1778 Default: 1024 1779 1780 hardened_usercopy= 1781 [KNL] Under CONFIG_HARDENED_USERCOPY, whether 1782 hardening is enabled for this boot. Hardened 1783 usercopy checking is used to protect the kernel 1784 from reading or writing beyond known memory 1785 allocation boundaries as a proactive defense 1786 against bounds-checking flaws in the kernel's 1787 copy_to_user()/copy_from_user() interface. 1788 on Perform hardened usercopy checks (default). 1789 off Disable hardened usercopy checks. 1790 1791 hardlockup_all_cpu_backtrace= 1792 [KNL] Should the hard-lockup detector generate 1793 backtraces on all cpus. 1794 Format: 0 | 1 1795 1796 hashdist= [KNL,NUMA] Large hashes allocated during boot 1797 are distributed across NUMA nodes. Defaults on 1798 for 64-bit NUMA, off otherwise. 1799 Format: 0 | 1 (for off | on) 1800 1801 hd= [EIDE] (E)IDE hard drive subsystem geometry 1802 Format: <cyl>,<head>,<sect> 1803 1804 hest_disable [ACPI] 1805 Disable Hardware Error Source Table (HEST) support; 1806 corresponding firmware-first mode error processing 1807 logic will be disabled. 1808 1809 hibernate= [HIBERNATION] 1810 noresume Don't check if there's a hibernation image 1811 present during boot. 1812 nocompress Don't compress/decompress hibernation images. 1813 no Disable hibernation and resume. 1814 protect_image Turn on image protection during restoration 1815 (that will set all pages holding image data 1816 during restoration read-only). 1817 1818 hibernate.compressor= [HIBERNATION] Compression algorithm to be 1819 used with hibernation. 1820 Format: { lzo | lz4 } 1821 Default: lzo 1822 1823 lzo: Select LZO compression algorithm to 1824 compress/decompress hibernation image. 1825 1826 lz4: Select LZ4 compression algorithm to 1827 compress/decompress hibernation image. 1828 1829 highmem=nn[KMG] [KNL,BOOT,EARLY] forces the highmem zone to have an exact 1830 size of <nn>. This works even on boxes that have no 1831 highmem otherwise. This also works to reduce highmem 1832 size on bigger boxes. 1833 1834 highres= [KNL] Enable/disable high resolution timer mode. 1835 Valid parameters: "on", "off" 1836 Default: "on" 1837 1838 hlt [BUGS=ARM,SH] 1839 1840 hostname= [KNL,EARLY] Set the hostname (aka UTS nodename). 1841 Format: <string> 1842 This allows setting the system's hostname during early 1843 startup. This sets the name returned by gethostname. 1844 Using this parameter to set the hostname makes it 1845 possible to ensure the hostname is correctly set before 1846 any userspace processes run, avoiding the possibility 1847 that a process may call gethostname before the hostname 1848 has been explicitly set, resulting in the calling 1849 process getting an incorrect result. The string must 1850 not exceed the maximum allowed hostname length (usually 1851 64 characters) and will be truncated otherwise. 1852 1853 hpet= [X86-32,HPET] option to control HPET usage 1854 Format: { enable (default) | disable | force | 1855 verbose } 1856 disable: disable HPET and use PIT instead 1857 force: allow force enabled of undocumented chips (ICH4, 1858 VIA, nVidia) 1859 verbose: show contents of HPET registers during setup 1860 1861 hpet_mmap= [X86, HPET_MMAP] Allow userspace to mmap HPET 1862 registers. Default set by CONFIG_HPET_MMAP_DEFAULT. 1863 1864 hugepages= [HW] Number of HugeTLB pages to allocate at boot. 1865 If this follows hugepagesz (below), it specifies 1866 the number of pages of hugepagesz to be allocated. 1867 If this is the first HugeTLB parameter on the command 1868 line, it specifies the number of pages to allocate for 1869 the default huge page size. If using node format, the 1870 number of pages to allocate per-node can be specified. 1871 See also Documentation/admin-guide/mm/hugetlbpage.rst. 1872 Format: <integer> or (node format) 1873 <node>:<integer>[,<node>:<integer>] 1874 1875 hugepagesz= 1876 [HW] The size of the HugeTLB pages. This is used in 1877 conjunction with hugepages (above) to allocate huge 1878 pages of a specific size at boot. The pair 1879 hugepagesz=X hugepages=Y can be specified once for 1880 each supported huge page size. Huge page sizes are 1881 architecture dependent. See also 1882 Documentation/admin-guide/mm/hugetlbpage.rst. 1883 Format: size[KMG] 1884 1885 hugetlb_cma= [HW,CMA,EARLY] The size of a CMA area used for allocation 1886 of gigantic hugepages. Or using node format, the size 1887 of a CMA area per node can be specified. 1888 Format: nn[KMGTPE] or (node format) 1889 <node>:nn[KMGTPE][,<node>:nn[KMGTPE]] 1890 1891 Reserve a CMA area of given size and allocate gigantic 1892 hugepages using the CMA allocator. If enabled, the 1893 boot-time allocation of gigantic hugepages is skipped. 1894 1895 hugetlb_free_vmemmap= 1896 [KNL] Requires CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP 1897 enabled. 1898 Control if HugeTLB Vmemmap Optimization (HVO) is enabled. 1899 Allows heavy hugetlb users to free up some more 1900 memory (7 * PAGE_SIZE for each 2MB hugetlb page). 1901 Format: { on | off (default) } 1902 1903 on: enable HVO 1904 off: disable HVO 1905 1906 Built with CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP_DEFAULT_ON=y, 1907 the default is on. 1908 1909 Note that the vmemmap pages may be allocated from the added 1910 memory block itself when memory_hotplug.memmap_on_memory is 1911 enabled, those vmemmap pages cannot be optimized even if this 1912 feature is enabled. Other vmemmap pages not allocated from 1913 the added memory block itself do not be affected. 1914 1915 hung_task_panic= 1916 [KNL] Should the hung task detector generate panics. 1917 Format: 0 | 1 1918 1919 A value of 1 instructs the kernel to panic when a 1920 hung task is detected. The default value is controlled 1921 by the CONFIG_BOOTPARAM_HUNG_TASK_PANIC build-time 1922 option. The value selected by this boot parameter can 1923 be changed later by the kernel.hung_task_panic sysctl. 1924 1925 hvc_iucv= [S390] Number of z/VM IUCV hypervisor console (HVC) 1926 terminal devices. Valid values: 0..8 1927 hvc_iucv_allow= [S390] Comma-separated list of z/VM user IDs. 1928 If specified, z/VM IUCV HVC accepts connections 1929 from listed z/VM user IDs only. 1930 1931 hv_nopvspin [X86,HYPER_V,EARLY] 1932 Disables the paravirt spinlock optimizations 1933 which allow the hypervisor to 'idle' the guest 1934 on lock contention. 1935 1936 i2c_bus= [HW] Override the default board specific I2C bus speed 1937 or register an additional I2C bus that is not 1938 registered from board initialization code. 1939 Format: 1940 <bus_id>,<clkrate> 1941 1942 i2c_touchscreen_props= [HW,ACPI,X86] 1943 Set device-properties for ACPI-enumerated I2C-attached 1944 touchscreen, to e.g. fix coordinates of upside-down 1945 mounted touchscreens. If you need this option please 1946 submit a drivers/platform/x86/touchscreen_dmi.c patch 1947 adding a DMI quirk for this. 1948 1949 Format: 1950 <ACPI_HW_ID>:<prop_name>=<val>[:prop_name=val][:...] 1951 Where <val> is one of: 1952 Omit "=<val>" entirely Set a boolean device-property 1953 Unsigned number Set a u32 device-property 1954 Anything else Set a string device-property 1955 1956 Examples (split over multiple lines): 1957 i2c_touchscreen_props=GDIX1001:touchscreen-inverted-x: 1958 touchscreen-inverted-y 1959 1960 i2c_touchscreen_props=MSSL1680:touchscreen-size-x=1920: 1961 touchscreen-size-y=1080:touchscreen-inverted-y: 1962 firmware-name=gsl1680-vendor-model.fw:silead,home-button 1963 1964 i8042.debug [HW] Toggle i8042 debug mode 1965 i8042.unmask_kbd_data 1966 [HW] Enable printing of interrupt data from the KBD port 1967 (disabled by default, and as a pre-condition 1968 requires that i8042.debug=1 be enabled) 1969 i8042.direct [HW] Put keyboard port into non-translated mode 1970 i8042.dumbkbd [HW] Pretend that controller can only read data from 1971 keyboard and cannot control its state 1972 (Don't attempt to blink the leds) 1973 i8042.noaux [HW] Don't check for auxiliary (== mouse) port 1974 i8042.nokbd [HW] Don't check/create keyboard port 1975 i8042.noloop [HW] Disable the AUX Loopback command while probing 1976 for the AUX port 1977 i8042.nomux [HW] Don't check presence of an active multiplexing 1978 controller 1979 i8042.nopnp [HW] Don't use ACPIPnP / PnPBIOS to discover KBD/AUX 1980 controllers 1981 i8042.notimeout [HW] Ignore timeout condition signalled by controller 1982 i8042.reset [HW] Reset the controller during init, cleanup and 1983 suspend-to-ram transitions, only during s2r 1984 transitions, or never reset 1985 Format: { 1 | Y | y | 0 | N | n } 1986 1, Y, y: always reset controller 1987 0, N, n: don't ever reset controller 1988 Default: only on s2r transitions on x86; most other 1989 architectures force reset to be always executed 1990 i8042.unlock [HW] Unlock (ignore) the keylock 1991 i8042.kbdreset [HW] Reset device connected to KBD port 1992 i8042.probe_defer 1993 [HW] Allow deferred probing upon i8042 probe errors 1994 1995 i810= [HW,DRM] 1996 1997 i915.invert_brightness= 1998 [DRM] Invert the sense of the variable that is used to 1999 set the brightness of the panel backlight. Normally a 2000 brightness value of 0 indicates backlight switched off, 2001 and the maximum of the brightness value sets the backlight 2002 to maximum brightness. If this parameter is set to 0 2003 (default) and the machine requires it, or this parameter 2004 is set to 1, a brightness value of 0 sets the backlight 2005 to maximum brightness, and the maximum of the brightness 2006 value switches the backlight off. 2007 -1 -- never invert brightness 2008 0 -- machine default 2009 1 -- force brightness inversion 2010 2011 ia32_emulation= [X86-64] 2012 Format: <bool> 2013 When true, allows loading 32-bit programs and executing 32-bit 2014 syscalls, essentially overriding IA32_EMULATION_DEFAULT_DISABLED at 2015 boot time. When false, unconditionally disables IA32 emulation. 2016 2017 icn= [HW,ISDN] 2018 Format: <io>[,<membase>[,<icn_id>[,<icn_id2>]]] 2019 2020 2021 idle= [X86,EARLY] 2022 Format: idle=poll, idle=halt, idle=nomwait 2023 2024 idle=poll: Don't do power saving in the idle loop 2025 using HLT, but poll for rescheduling event. This will 2026 make the CPUs eat a lot more power, but may be useful 2027 to get slightly better performance in multiprocessor 2028 benchmarks. It also makes some profiling using 2029 performance counters more accurate. Please note that 2030 on systems with MONITOR/MWAIT support (like Intel 2031 EM64T CPUs) this option has no performance advantage 2032 over the normal idle loop. It may also interact badly 2033 with hyperthreading. 2034 2035 idle=halt: Halt is forced to be used for CPU idle. 2036 In such case C2/C3 won't be used again. 2037 2038 idle=nomwait: Disable mwait for CPU C-states 2039 2040 idxd.sva= [HW] 2041 Format: <bool> 2042 Allow force disabling of Shared Virtual Memory (SVA) 2043 support for the idxd driver. By default it is set to 2044 true (1). 2045 2046 idxd.tc_override= [HW] 2047 Format: <bool> 2048 Allow override of default traffic class configuration 2049 for the device. By default it is set to false (0). 2050 2051 ieee754= [MIPS] Select IEEE Std 754 conformance mode 2052 Format: { strict | legacy | 2008 | relaxed | emulated } 2053 Default: strict 2054 2055 Choose which programs will be accepted for execution 2056 based on the IEEE 754 NaN encoding(s) supported by 2057 the FPU and the NaN encoding requested with the value 2058 of an ELF file header flag individually set by each 2059 binary. Hardware implementations are permitted to 2060 support either or both of the legacy and the 2008 NaN 2061 encoding mode. 2062 2063 Available settings are as follows: 2064 strict accept binaries that request a NaN encoding 2065 supported by the FPU 2066 legacy only accept legacy-NaN binaries, if supported 2067 by the FPU 2068 2008 only accept 2008-NaN binaries, if supported 2069 by the FPU 2070 relaxed accept any binaries regardless of whether 2071 supported by the FPU 2072 emulated accept any binaries but enable FPU emulator 2073 if binary mode is unsupported by the FPU. 2074 2075 The FPU emulator is always able to support both NaN 2076 encodings, so if no FPU hardware is present or it has 2077 been disabled with 'nofpu', then the settings of 2078 'legacy' and '2008' strap the emulator accordingly, 2079 'relaxed' straps the emulator for both legacy-NaN and 2080 2008-NaN, whereas 'strict' enables legacy-NaN only on 2081 legacy processors and both NaN encodings on MIPS32 or 2082 MIPS64 CPUs. 2083 2084 The setting for ABS.fmt/NEG.fmt instruction execution 2085 mode generally follows that for the NaN encoding, 2086 except where unsupported by hardware. 2087 2088 ignore_loglevel [KNL,EARLY] 2089 Ignore loglevel setting - this will print /all/ 2090 kernel messages to the console. Useful for debugging. 2091 We also add it as printk module parameter, so users 2092 could change it dynamically, usually by 2093 /sys/module/printk/parameters/ignore_loglevel. 2094 2095 ignore_rlimit_data 2096 Ignore RLIMIT_DATA setting for data mappings, 2097 print warning at first misuse. Can be changed via 2098 /sys/module/kernel/parameters/ignore_rlimit_data. 2099 2100 ihash_entries= [KNL] 2101 Set number of hash buckets for inode cache. 2102 2103 ima_appraise= [IMA] appraise integrity measurements 2104 Format: { "off" | "enforce" | "fix" | "log" } 2105 default: "enforce" 2106 2107 ima_appraise_tcb [IMA] Deprecated. Use ima_policy= instead. 2108 The builtin appraise policy appraises all files 2109 owned by uid=0. 2110 2111 ima_canonical_fmt [IMA] 2112 Use the canonical format for the binary runtime 2113 measurements, instead of host native format. 2114 2115 ima_hash= [IMA] 2116 Format: { md5 | sha1 | rmd160 | sha256 | sha384 2117 | sha512 | ... } 2118 default: "sha1" 2119 2120 The list of supported hash algorithms is defined 2121 in crypto/hash_info.h. 2122 2123 ima_policy= [IMA] 2124 The builtin policies to load during IMA setup. 2125 Format: "tcb | appraise_tcb | secure_boot | 2126 fail_securely | critical_data" 2127 2128 The "tcb" policy measures all programs exec'd, files 2129 mmap'd for exec, and all files opened with the read 2130 mode bit set by either the effective uid (euid=0) or 2131 uid=0. 2132 2133 The "appraise_tcb" policy appraises the integrity of 2134 all files owned by root. 2135 2136 The "secure_boot" policy appraises the integrity 2137 of files (eg. kexec kernel image, kernel modules, 2138 firmware, policy, etc) based on file signatures. 2139 2140 The "fail_securely" policy forces file signature 2141 verification failure also on privileged mounted 2142 filesystems with the SB_I_UNVERIFIABLE_SIGNATURE 2143 flag. 2144 2145 The "critical_data" policy measures kernel integrity 2146 critical data. 2147 2148 ima_tcb [IMA] Deprecated. Use ima_policy= instead. 2149 Load a policy which meets the needs of the Trusted 2150 Computing Base. This means IMA will measure all 2151 programs exec'd, files mmap'd for exec, and all files 2152 opened for read by uid=0. 2153 2154 ima_template= [IMA] 2155 Select one of defined IMA measurements template formats. 2156 Formats: { "ima" | "ima-ng" | "ima-ngv2" | "ima-sig" | 2157 "ima-sigv2" } 2158 Default: "ima-ng" 2159 2160 ima_template_fmt= 2161 [IMA] Define a custom template format. 2162 Format: { "field1|...|fieldN" } 2163 2164 ima.ahash_minsize= [IMA] Minimum file size for asynchronous hash usage 2165 Format: <min_file_size> 2166 Set the minimal file size for using asynchronous hash. 2167 If left unspecified, ahash usage is disabled. 2168 2169 ahash performance varies for different data sizes on 2170 different crypto accelerators. This option can be used 2171 to achieve the best performance for a particular HW. 2172 2173 ima.ahash_bufsize= [IMA] Asynchronous hash buffer size 2174 Format: <bufsize> 2175 Set hashing buffer size. Default: 4k. 2176 2177 ahash performance varies for different chunk sizes on 2178 different crypto accelerators. This option can be used 2179 to achieve best performance for particular HW. 2180 2181 init= [KNL] 2182 Format: <full_path> 2183 Run specified binary instead of /sbin/init as init 2184 process. 2185 2186 initcall_debug [KNL] Trace initcalls as they are executed. Useful 2187 for working out where the kernel is dying during 2188 startup. 2189 2190 initcall_blacklist= [KNL] Do not execute a comma-separated list of 2191 initcall functions. Useful for debugging built-in 2192 modules and initcalls. 2193 2194 initramfs_async= [KNL] 2195 Format: <bool> 2196 Default: 1 2197 This parameter controls whether the initramfs 2198 image is unpacked asynchronously, concurrently 2199 with devices being probed and 2200 initialized. This should normally just work, 2201 but as a debugging aid, one can get the 2202 historical behaviour of the initramfs 2203 unpacking being completed before device_ and 2204 late_ initcalls. 2205 2206 initrd= [BOOT,EARLY] Specify the location of the initial ramdisk 2207 2208 initrdmem= [KNL,EARLY] Specify a physical address and size from which to 2209 load the initrd. If an initrd is compiled in or 2210 specified in the bootparams, it takes priority over this 2211 setting. 2212 Format: ss[KMG],nn[KMG] 2213 Default is 0, 0 2214 2215 init_on_alloc= [MM,EARLY] Fill newly allocated pages and heap objects with 2216 zeroes. 2217 Format: 0 | 1 2218 Default set by CONFIG_INIT_ON_ALLOC_DEFAULT_ON. 2219 2220 init_on_free= [MM,EARLY] Fill freed pages and heap objects with zeroes. 2221 Format: 0 | 1 2222 Default set by CONFIG_INIT_ON_FREE_DEFAULT_ON. 2223 2224 init_pkru= [X86] Specify the default memory protection keys rights 2225 register contents for all processes. 0x55555554 by 2226 default (disallow access to all but pkey 0). Can 2227 override in debugfs after boot. 2228 2229 inport.irq= [HW] Inport (ATI XL and Microsoft) busmouse driver 2230 Format: <irq> 2231 2232 int_pln_enable [X86] Enable power limit notification interrupt 2233 2234 integrity_audit=[IMA] 2235 Format: { "0" | "1" } 2236 0 -- basic integrity auditing messages. (Default) 2237 1 -- additional integrity auditing messages. 2238 2239 intel_iommu= [DMAR] Intel IOMMU driver (DMAR) option 2240 on 2241 Enable intel iommu driver. 2242 off 2243 Disable intel iommu driver. 2244 igfx_off [Default Off] 2245 By default, gfx is mapped as normal device. If a gfx 2246 device has a dedicated DMAR unit, the DMAR unit is 2247 bypassed by not enabling DMAR with this option. In 2248 this case, gfx device will use physical address for 2249 DMA. 2250 strict [Default Off] 2251 Deprecated, equivalent to iommu.strict=1. 2252 sp_off [Default Off] 2253 By default, super page will be supported if Intel IOMMU 2254 has the capability. With this option, super page will 2255 not be supported. 2256 sm_on 2257 Enable the Intel IOMMU scalable mode if the hardware 2258 advertises that it has support for the scalable mode 2259 translation. 2260 sm_off 2261 Disallow use of the Intel IOMMU scalable mode. 2262 tboot_noforce [Default Off] 2263 Do not force the Intel IOMMU enabled under tboot. 2264 By default, tboot will force Intel IOMMU on, which 2265 could harm performance of some high-throughput 2266 devices like 40GBit network cards, even if identity 2267 mapping is enabled. 2268 Note that using this option lowers the security 2269 provided by tboot because it makes the system 2270 vulnerable to DMA attacks. 2271 2272 intel_idle.max_cstate= [KNL,HW,ACPI,X86] 2273 0 disables intel_idle and fall back on acpi_idle. 2274 1 to 9 specify maximum depth of C-state. 2275 2276 intel_pstate= [X86,EARLY] 2277 disable 2278 Do not enable intel_pstate as the default 2279 scaling driver for the supported processors 2280 active 2281 Use intel_pstate driver to bypass the scaling 2282 governors layer of cpufreq and provides it own 2283 algorithms for p-state selection. There are two 2284 P-state selection algorithms provided by 2285 intel_pstate in the active mode: powersave and 2286 performance. The way they both operate depends 2287 on whether or not the hardware managed P-states 2288 (HWP) feature has been enabled in the processor 2289 and possibly on the processor model. 2290 passive 2291 Use intel_pstate as a scaling driver, but configure it 2292 to work with generic cpufreq governors (instead of 2293 enabling its internal governor). This mode cannot be 2294 used along with the hardware-managed P-states (HWP) 2295 feature. 2296 force 2297 Enable intel_pstate on systems that prohibit it by default 2298 in favor of acpi-cpufreq. Forcing the intel_pstate driver 2299 instead of acpi-cpufreq may disable platform features, such 2300 as thermal controls and power capping, that rely on ACPI 2301 P-States information being indicated to OSPM and therefore 2302 should be used with caution. This option does not work with 2303 processors that aren't supported by the intel_pstate driver 2304 or on platforms that use pcc-cpufreq instead of acpi-cpufreq. 2305 no_hwp 2306 Do not enable hardware P state control (HWP) 2307 if available. 2308 hwp_only 2309 Only load intel_pstate on systems which support 2310 hardware P state control (HWP) if available. 2311 support_acpi_ppc 2312 Enforce ACPI _PPC performance limits. If the Fixed ACPI 2313 Description Table, specifies preferred power management 2314 profile as "Enterprise Server" or "Performance Server", 2315 then this feature is turned on by default. 2316 per_cpu_perf_limits 2317 Allow per-logical-CPU P-State performance control limits using 2318 cpufreq sysfs interface 2319 2320 intremap= [X86-64,Intel-IOMMU,EARLY] 2321 on enable Interrupt Remapping (default) 2322 off disable Interrupt Remapping 2323 nosid disable Source ID checking 2324 no_x2apic_optout 2325 BIOS x2APIC opt-out request will be ignored 2326 nopost disable Interrupt Posting 2327 posted_msi 2328 enable MSIs delivered as posted interrupts 2329 2330 iomem= Disable strict checking of access to MMIO memory 2331 strict regions from userspace. 2332 relaxed 2333 2334 iommu= [X86,EARLY] 2335 2336 off 2337 Don't initialize and use any kind of IOMMU. 2338 2339 force 2340 Force the use of the hardware IOMMU even when 2341 it is not actually needed (e.g. because < 3 GB 2342 memory). 2343 2344 noforce 2345 Don't force hardware IOMMU usage when it is not 2346 needed. (default). 2347 2348 biomerge 2349 panic 2350 nopanic 2351 merge 2352 nomerge 2353 2354 soft 2355 Use software bounce buffering (SWIOTLB) (default for 2356 Intel machines). This can be used to prevent the usage 2357 of an available hardware IOMMU. 2358 2359 [X86] 2360 pt 2361 [X86] 2362 nopt 2363 [PPC/POWERNV] 2364 nobypass 2365 Disable IOMMU bypass, using IOMMU for PCI devices. 2366 2367 [X86] 2368 AMD Gart HW IOMMU-specific options: 2369 2370 <size> 2371 Set the size of the remapping area in bytes. 2372 2373 allowed 2374 Overwrite iommu off workarounds for specific chipsets 2375 2376 fullflush 2377 Flush IOMMU on each allocation (default). 2378 2379 nofullflush 2380 Don't use IOMMU fullflush. 2381 2382 memaper[=<order>] 2383 Allocate an own aperture over RAM with size 2384 32MB<<order. (default: order=1, i.e. 64MB) 2385 2386 merge 2387 Do scatter-gather (SG) merging. Implies "force" 2388 (experimental). 2389 2390 nomerge 2391 Don't do scatter-gather (SG) merging. 2392 2393 noaperture 2394 Ask the IOMMU not to touch the aperture for AGP. 2395 2396 noagp 2397 Don't initialize the AGP driver and use full aperture. 2398 2399 panic 2400 Always panic when IOMMU overflows. 2401 2402 iommu.forcedac= [ARM64,X86,EARLY] Control IOVA allocation for PCI devices. 2403 Format: { "0" | "1" } 2404 0 - Try to allocate a 32-bit DMA address first, before 2405 falling back to the full range if needed. 2406 1 - Allocate directly from the full usable range, 2407 forcing Dual Address Cycle for PCI cards supporting 2408 greater than 32-bit addressing. 2409 2410 iommu.strict= [ARM64,X86,S390,EARLY] Configure TLB invalidation behaviour 2411 Format: { "0" | "1" } 2412 0 - Lazy mode. 2413 Request that DMA unmap operations use deferred 2414 invalidation of hardware TLBs, for increased 2415 throughput at the cost of reduced device isolation. 2416 Will fall back to strict mode if not supported by 2417 the relevant IOMMU driver. 2418 1 - Strict mode. 2419 DMA unmap operations invalidate IOMMU hardware TLBs 2420 synchronously. 2421 unset - Use value of CONFIG_IOMMU_DEFAULT_DMA_{LAZY,STRICT}. 2422 Note: on x86, strict mode specified via one of the 2423 legacy driver-specific options takes precedence. 2424 2425 iommu.passthrough= 2426 [ARM64,X86,EARLY] Configure DMA to bypass the IOMMU by default. 2427 Format: { "0" | "1" } 2428 0 - Use IOMMU translation for DMA. 2429 1 - Bypass the IOMMU for DMA. 2430 unset - Use value of CONFIG_IOMMU_DEFAULT_PASSTHROUGH. 2431 2432 io7= [HW] IO7 for Marvel-based Alpha systems 2433 See comment before marvel_specify_io7 in 2434 arch/alpha/kernel/core_marvel.c. 2435 2436 io_delay= [X86,EARLY] I/O delay method 2437 0x80 2438 Standard port 0x80 based delay 2439 0xed 2440 Alternate port 0xed based delay (needed on some systems) 2441 udelay 2442 Simple two microseconds delay 2443 none 2444 No delay 2445 2446 ip= [IP_PNP] 2447 See Documentation/admin-guide/nfs/nfsroot.rst. 2448 2449 ipcmni_extend [KNL,EARLY] Extend the maximum number of unique System V 2450 IPC identifiers from 32,768 to 16,777,216. 2451 2452 ipe.enforce= [IPE] 2453 Format: <bool> 2454 Determine whether IPE starts in permissive (0) or 2455 enforce (1) mode. The default is enforce. 2456 2457 ipe.success_audit= 2458 [IPE] 2459 Format: <bool> 2460 Start IPE with success auditing enabled, emitting 2461 an audit event when a binary is allowed. The default 2462 is 0. 2463 2464 irqaffinity= [SMP] Set the default irq affinity mask 2465 The argument is a cpu list, as described above. 2466 2467 irqchip.gicv2_force_probe= 2468 [ARM,ARM64,EARLY] 2469 Format: <bool> 2470 Force the kernel to look for the second 4kB page 2471 of a GICv2 controller even if the memory range 2472 exposed by the device tree is too small. 2473 2474 irqchip.gicv3_nolpi= 2475 [ARM,ARM64,EARLY] 2476 Force the kernel to ignore the availability of 2477 LPIs (and by consequence ITSs). Intended for system 2478 that use the kernel as a bootloader, and thus want 2479 to let secondary kernels in charge of setting up 2480 LPIs. 2481 2482 irqchip.gicv3_pseudo_nmi= [ARM64,EARLY] 2483 Enables support for pseudo-NMIs in the kernel. This 2484 requires the kernel to be built with 2485 CONFIG_ARM64_PSEUDO_NMI. 2486 2487 irqfixup [HW] 2488 When an interrupt is not handled search all handlers 2489 for it. Intended to get systems with badly broken 2490 firmware running. 2491 2492 irqpoll [HW] 2493 When an interrupt is not handled search all handlers 2494 for it. Also check all handlers each timer 2495 interrupt. Intended to get systems with badly broken 2496 firmware running. 2497 2498 isapnp= [ISAPNP] 2499 Format: <RDP>,<reset>,<pci_scan>,<verbosity> 2500 2501 isolcpus= [KNL,SMP,ISOL] Isolate a given set of CPUs from disturbance. 2502 [Deprecated - use cpusets instead] 2503 Format: [flag-list,]<cpu-list> 2504 2505 Specify one or more CPUs to isolate from disturbances 2506 specified in the flag list (default: domain): 2507 2508 nohz 2509 Disable the tick when a single task runs as well as 2510 disabling other kernel noises like having RCU callbacks 2511 offloaded. This is equivalent to the nohz_full parameter. 2512 2513 A residual 1Hz tick is offloaded to workqueues, which you 2514 need to affine to housekeeping through the global 2515 workqueue's affinity configured via the 2516 /sys/devices/virtual/workqueue/cpumask sysfs file, or 2517 by using the 'domain' flag described below. 2518 2519 NOTE: by default the global workqueue runs on all CPUs, 2520 so to protect individual CPUs the 'cpumask' file has to 2521 be configured manually after bootup. 2522 2523 domain 2524 Isolate from the general SMP balancing and scheduling 2525 algorithms. Note that performing domain isolation this way 2526 is irreversible: it's not possible to bring back a CPU to 2527 the domains once isolated through isolcpus. It's strongly 2528 advised to use cpusets instead to disable scheduler load 2529 balancing through the "cpuset.sched_load_balance" file. 2530 It offers a much more flexible interface where CPUs can 2531 move in and out of an isolated set anytime. 2532 2533 You can move a process onto or off an "isolated" CPU via 2534 the CPU affinity syscalls or cpuset. 2535 <cpu number> begins at 0 and the maximum value is 2536 "number of CPUs in system - 1". 2537 2538 managed_irq 2539 2540 Isolate from being targeted by managed interrupts 2541 which have an interrupt mask containing isolated 2542 CPUs. The affinity of managed interrupts is 2543 handled by the kernel and cannot be changed via 2544 the /proc/irq/* interfaces. 2545 2546 This isolation is best effort and only effective 2547 if the automatically assigned interrupt mask of a 2548 device queue contains isolated and housekeeping 2549 CPUs. If housekeeping CPUs are online then such 2550 interrupts are directed to the housekeeping CPU 2551 so that IO submitted on the housekeeping CPU 2552 cannot disturb the isolated CPU. 2553 2554 If a queue's affinity mask contains only isolated 2555 CPUs then this parameter has no effect on the 2556 interrupt routing decision, though interrupts are 2557 only delivered when tasks running on those 2558 isolated CPUs submit IO. IO submitted on 2559 housekeeping CPUs has no influence on those 2560 queues. 2561 2562 The format of <cpu-list> is described above. 2563 2564 iucv= [HW,NET] 2565 2566 ivrs_ioapic [HW,X86-64] 2567 Provide an override to the IOAPIC-ID<->DEVICE-ID 2568 mapping provided in the IVRS ACPI table. 2569 By default, PCI segment is 0, and can be omitted. 2570 2571 For example, to map IOAPIC-ID decimal 10 to 2572 PCI segment 0x1 and PCI device 00:14.0, 2573 write the parameter as: 2574 ivrs_ioapic=10@0001:00:14.0 2575 2576 Deprecated formats: 2577 * To map IOAPIC-ID decimal 10 to PCI device 00:14.0 2578 write the parameter as: 2579 ivrs_ioapic[10]=00:14.0 2580 * To map IOAPIC-ID decimal 10 to PCI segment 0x1 and 2581 PCI device 00:14.0 write the parameter as: 2582 ivrs_ioapic[10]=0001:00:14.0 2583 2584 ivrs_hpet [HW,X86-64] 2585 Provide an override to the HPET-ID<->DEVICE-ID 2586 mapping provided in the IVRS ACPI table. 2587 By default, PCI segment is 0, and can be omitted. 2588 2589 For example, to map HPET-ID decimal 10 to 2590 PCI segment 0x1 and PCI device 00:14.0, 2591 write the parameter as: 2592 ivrs_hpet=10@0001:00:14.0 2593 2594 Deprecated formats: 2595 * To map HPET-ID decimal 0 to PCI device 00:14.0 2596 write the parameter as: 2597 ivrs_hpet[0]=00:14.0 2598 * To map HPET-ID decimal 10 to PCI segment 0x1 and 2599 PCI device 00:14.0 write the parameter as: 2600 ivrs_ioapic[10]=0001:00:14.0 2601 2602 ivrs_acpihid [HW,X86-64] 2603 Provide an override to the ACPI-HID:UID<->DEVICE-ID 2604 mapping provided in the IVRS ACPI table. 2605 By default, PCI segment is 0, and can be omitted. 2606 2607 For example, to map UART-HID:UID AMD0020:0 to 2608 PCI segment 0x1 and PCI device ID 00:14.5, 2609 write the parameter as: 2610 ivrs_acpihid=AMD0020:0@0001:00:14.5 2611 2612 Deprecated formats: 2613 * To map UART-HID:UID AMD0020:0 to PCI segment is 0, 2614 PCI device ID 00:14.5, write the parameter as: 2615 ivrs_acpihid[00:14.5]=AMD0020:0 2616 * To map UART-HID:UID AMD0020:0 to PCI segment 0x1 and 2617 PCI device ID 00:14.5, write the parameter as: 2618 ivrs_acpihid[0001:00:14.5]=AMD0020:0 2619 2620 js= [HW,JOY] Analog joystick 2621 See Documentation/input/joydev/joystick.rst. 2622 2623 kasan_multi_shot 2624 [KNL] Enforce KASAN (Kernel Address Sanitizer) to print 2625 report on every invalid memory access. Without this 2626 parameter KASAN will print report only for the first 2627 invalid access. 2628 2629 keep_bootcon [KNL,EARLY] 2630 Do not unregister boot console at start. This is only 2631 useful for debugging when something happens in the window 2632 between unregistering the boot console and initializing 2633 the real console. 2634 2635 keepinitrd [HW,ARM] See retain_initrd. 2636 2637 kernelcore= [KNL,X86,PPC,EARLY] 2638 Format: nn[KMGTPE] | nn% | "mirror" 2639 This parameter specifies the amount of memory usable by 2640 the kernel for non-movable allocations. The requested 2641 amount is spread evenly throughout all nodes in the 2642 system as ZONE_NORMAL. The remaining memory is used for 2643 movable memory in its own zone, ZONE_MOVABLE. In the 2644 event, a node is too small to have both ZONE_NORMAL and 2645 ZONE_MOVABLE, kernelcore memory will take priority and 2646 other nodes will have a larger ZONE_MOVABLE. 2647 2648 ZONE_MOVABLE is used for the allocation of pages that 2649 may be reclaimed or moved by the page migration 2650 subsystem. Note that allocations like PTEs-from-HighMem 2651 still use the HighMem zone if it exists, and the Normal 2652 zone if it does not. 2653 2654 It is possible to specify the exact amount of memory in 2655 the form of "nn[KMGTPE]", a percentage of total system 2656 memory in the form of "nn%", or "mirror". If "mirror" 2657 option is specified, mirrored (reliable) memory is used 2658 for non-movable allocations and remaining memory is used 2659 for Movable pages. "nn[KMGTPE]", "nn%", and "mirror" 2660 are exclusive, so you cannot specify multiple forms. 2661 2662 kgdbdbgp= [KGDB,HW,EARLY] kgdb over EHCI usb debug port. 2663 Format: <Controller#>[,poll interval] 2664 The controller # is the number of the ehci usb debug 2665 port as it is probed via PCI. The poll interval is 2666 optional and is the number seconds in between 2667 each poll cycle to the debug port in case you need 2668 the functionality for interrupting the kernel with 2669 gdb or control-c on the dbgp connection. When 2670 not using this parameter you use sysrq-g to break into 2671 the kernel debugger. 2672 2673 kgdboc= [KGDB,HW] kgdb over consoles. 2674 Requires a tty driver that supports console polling, 2675 or a supported polling keyboard driver (non-usb). 2676 Serial only format: <serial_device>[,baud] 2677 keyboard only format: kbd 2678 keyboard and serial format: kbd,<serial_device>[,baud] 2679 Optional Kernel mode setting: 2680 kms, kbd format: kms,kbd 2681 kms, kbd and serial format: kms,kbd,<ser_dev>[,baud] 2682 2683 kgdboc_earlycon= [KGDB,HW,EARLY] 2684 If the boot console provides the ability to read 2685 characters and can work in polling mode, you can use 2686 this parameter to tell kgdb to use it as a backend 2687 until the normal console is registered. Intended to 2688 be used together with the kgdboc parameter which 2689 specifies the normal console to transition to. 2690 2691 The name of the early console should be specified 2692 as the value of this parameter. Note that the name of 2693 the early console might be different than the tty 2694 name passed to kgdboc. It's OK to leave the value 2695 blank and the first boot console that implements 2696 read() will be picked. 2697 2698 kgdbwait [KGDB,EARLY] Stop kernel execution and enter the 2699 kernel debugger at the earliest opportunity. 2700 2701 kmac= [MIPS] Korina ethernet MAC address. 2702 Configure the RouterBoard 532 series on-chip 2703 Ethernet adapter MAC address. 2704 2705 kmemleak= [KNL,EARLY] Boot-time kmemleak enable/disable 2706 Valid arguments: on, off 2707 Default: on 2708 Built with CONFIG_DEBUG_KMEMLEAK_DEFAULT_OFF=y, 2709 the default is off. 2710 2711 kprobe_event=[probe-list] 2712 [FTRACE] Add kprobe events and enable at boot time. 2713 The probe-list is a semicolon delimited list of probe 2714 definitions. Each definition is same as kprobe_events 2715 interface, but the parameters are comma delimited. 2716 For example, to add a kprobe event on vfs_read with 2717 arg1 and arg2, add to the command line; 2718 2719 kprobe_event=p,vfs_read,$arg1,$arg2 2720 2721 See also Documentation/trace/kprobetrace.rst "Kernel 2722 Boot Parameter" section. 2723 2724 kpti= [ARM64,EARLY] Control page table isolation of 2725 user and kernel address spaces. 2726 Default: enabled on cores which need mitigation. 2727 0: force disabled 2728 1: force enabled 2729 2730 kunit.enable= [KUNIT] Enable executing KUnit tests. Requires 2731 CONFIG_KUNIT to be set to be fully enabled. The 2732 default value can be overridden via 2733 KUNIT_DEFAULT_ENABLED. 2734 Default is 1 (enabled) 2735 2736 kvm.ignore_msrs=[KVM] Ignore guest accesses to unhandled MSRs. 2737 Default is 0 (don't ignore, but inject #GP) 2738 2739 kvm.eager_page_split= 2740 [KVM,X86] Controls whether or not KVM will try to 2741 proactively split all huge pages during dirty logging. 2742 Eager page splitting reduces interruptions to vCPU 2743 execution by eliminating the write-protection faults 2744 and MMU lock contention that would otherwise be 2745 required to split huge pages lazily. 2746 2747 VM workloads that rarely perform writes or that write 2748 only to a small region of VM memory may benefit from 2749 disabling eager page splitting to allow huge pages to 2750 still be used for reads. 2751 2752 The behavior of eager page splitting depends on whether 2753 KVM_DIRTY_LOG_INITIALLY_SET is enabled or disabled. If 2754 disabled, all huge pages in a memslot will be eagerly 2755 split when dirty logging is enabled on that memslot. If 2756 enabled, eager page splitting will be performed during 2757 the KVM_CLEAR_DIRTY ioctl, and only for the pages being 2758 cleared. 2759 2760 Eager page splitting is only supported when kvm.tdp_mmu=Y. 2761 2762 Default is Y (on). 2763 2764 kvm.enable_virt_at_load=[KVM,ARM64,LOONGARCH,MIPS,RISCV,X86] 2765 If enabled, KVM will enable virtualization in hardware 2766 when KVM is loaded, and disable virtualization when KVM 2767 is unloaded (if KVM is built as a module). 2768 2769 If disabled, KVM will dynamically enable and disable 2770 virtualization on-demand when creating and destroying 2771 VMs, i.e. on the 0=>1 and 1=>0 transitions of the 2772 number of VMs. 2773 2774 Enabling virtualization at module load avoids potential 2775 latency for creation of the 0=>1 VM, as KVM serializes 2776 virtualization enabling across all online CPUs. The 2777 "cost" of enabling virtualization when KVM is loaded, 2778 is that doing so may interfere with using out-of-tree 2779 hypervisors that want to "own" virtualization hardware. 2780 2781 kvm.enable_vmware_backdoor=[KVM] Support VMware backdoor PV interface. 2782 Default is false (don't support). 2783 2784 kvm.nx_huge_pages= 2785 [KVM] Controls the software workaround for the 2786 X86_BUG_ITLB_MULTIHIT bug. 2787 force : Always deploy workaround. 2788 off : Never deploy workaround. 2789 auto : Deploy workaround based on the presence of 2790 X86_BUG_ITLB_MULTIHIT. 2791 2792 Default is 'auto'. 2793 2794 If the software workaround is enabled for the host, 2795 guests do need not to enable it for nested guests. 2796 2797 kvm.nx_huge_pages_recovery_ratio= 2798 [KVM] Controls how many 4KiB pages are periodically zapped 2799 back to huge pages. 0 disables the recovery, otherwise if 2800 the value is N KVM will zap 1/Nth of the 4KiB pages every 2801 period (see below). The default is 60. 2802 2803 kvm.nx_huge_pages_recovery_period_ms= 2804 [KVM] Controls the time period at which KVM zaps 4KiB pages 2805 back to huge pages. If the value is a non-zero N, KVM will 2806 zap a portion (see ratio above) of the pages every N msecs. 2807 If the value is 0 (the default), KVM will pick a period based 2808 on the ratio, such that a page is zapped after 1 hour on average. 2809 2810 kvm-amd.nested= [KVM,AMD] Control nested virtualization feature in 2811 KVM/SVM. Default is 1 (enabled). 2812 2813 kvm-amd.npt= [KVM,AMD] Control KVM's use of Nested Page Tables, 2814 a.k.a. Two-Dimensional Page Tables. Default is 1 2815 (enabled). Disable by KVM if hardware lacks support 2816 for NPT. 2817 2818 kvm-arm.mode= 2819 [KVM,ARM,EARLY] Select one of KVM/arm64's modes of 2820 operation. 2821 2822 none: Forcefully disable KVM. 2823 2824 nvhe: Standard nVHE-based mode, without support for 2825 protected guests. 2826 2827 protected: nVHE-based mode with support for guests whose 2828 state is kept private from the host. 2829 2830 nested: VHE-based mode with support for nested 2831 virtualization. Requires at least ARMv8.3 2832 hardware. 2833 2834 Defaults to VHE/nVHE based on hardware support. Setting 2835 mode to "protected" will disable kexec and hibernation 2836 for the host. "nested" is experimental and should be 2837 used with extreme caution. 2838 2839 kvm-arm.vgic_v3_group0_trap= 2840 [KVM,ARM,EARLY] Trap guest accesses to GICv3 group-0 2841 system registers 2842 2843 kvm-arm.vgic_v3_group1_trap= 2844 [KVM,ARM,EARLY] Trap guest accesses to GICv3 group-1 2845 system registers 2846 2847 kvm-arm.vgic_v3_common_trap= 2848 [KVM,ARM,EARLY] Trap guest accesses to GICv3 common 2849 system registers 2850 2851 kvm-arm.vgic_v4_enable= 2852 [KVM,ARM,EARLY] Allow use of GICv4 for direct 2853 injection of LPIs. 2854 2855 kvm-arm.wfe_trap_policy= 2856 [KVM,ARM] Control when to set WFE instruction trap for 2857 KVM VMs. Traps are allowed but not guaranteed by the 2858 CPU architecture. 2859 2860 trap: set WFE instruction trap 2861 2862 notrap: clear WFE instruction trap 2863 2864 kvm-arm.wfi_trap_policy= 2865 [KVM,ARM] Control when to set WFI instruction trap for 2866 KVM VMs. Traps are allowed but not guaranteed by the 2867 CPU architecture. 2868 2869 trap: set WFI instruction trap 2870 2871 notrap: clear WFI instruction trap 2872 2873 kvm_cma_resv_ratio=n [PPC,EARLY] 2874 Reserves given percentage from system memory area for 2875 contiguous memory allocation for KVM hash pagetable 2876 allocation. 2877 By default it reserves 5% of total system memory. 2878 Format: <integer> 2879 Default: 5 2880 2881 kvm-intel.ept= [KVM,Intel] Control KVM's use of Extended Page Tables, 2882 a.k.a. Two-Dimensional Page Tables. Default is 1 2883 (enabled). Disable by KVM if hardware lacks support 2884 for EPT. 2885 2886 kvm-intel.emulate_invalid_guest_state= 2887 [KVM,Intel] Control whether to emulate invalid guest 2888 state. Ignored if kvm-intel.enable_unrestricted_guest=1, 2889 as guest state is never invalid for unrestricted 2890 guests. This param doesn't apply to nested guests (L2), 2891 as KVM never emulates invalid L2 guest state. 2892 Default is 1 (enabled). 2893 2894 kvm-intel.flexpriority= 2895 [KVM,Intel] Control KVM's use of FlexPriority feature 2896 (TPR shadow). Default is 1 (enabled). Disable by KVM if 2897 hardware lacks support for it. 2898 2899 kvm-intel.nested= 2900 [KVM,Intel] Control nested virtualization feature in 2901 KVM/VMX. Default is 1 (enabled). 2902 2903 kvm-intel.unrestricted_guest= 2904 [KVM,Intel] Control KVM's use of unrestricted guest 2905 feature (virtualized real and unpaged mode). Default 2906 is 1 (enabled). Disable by KVM if EPT is disabled or 2907 hardware lacks support for it. 2908 2909 kvm-intel.vmentry_l1d_flush=[KVM,Intel] Mitigation for L1 Terminal Fault 2910 CVE-2018-3620. 2911 2912 Valid arguments: never, cond, always 2913 2914 always: L1D cache flush on every VMENTER. 2915 cond: Flush L1D on VMENTER only when the code between 2916 VMEXIT and VMENTER can leak host memory. 2917 never: Disables the mitigation 2918 2919 Default is cond (do L1 cache flush in specific instances) 2920 2921 kvm-intel.vpid= [KVM,Intel] Control KVM's use of Virtual Processor 2922 Identification feature (tagged TLBs). Default is 1 2923 (enabled). Disable by KVM if hardware lacks support 2924 for it. 2925 2926 l1d_flush= [X86,INTEL,EARLY] 2927 Control mitigation for L1D based snooping vulnerability. 2928 2929 Certain CPUs are vulnerable to an exploit against CPU 2930 internal buffers which can forward information to a 2931 disclosure gadget under certain conditions. 2932 2933 In vulnerable processors, the speculatively 2934 forwarded data can be used in a cache side channel 2935 attack, to access data to which the attacker does 2936 not have direct access. 2937 2938 This parameter controls the mitigation. The 2939 options are: 2940 2941 on - enable the interface for the mitigation 2942 2943 l1tf= [X86,EARLY] Control mitigation of the L1TF vulnerability on 2944 affected CPUs 2945 2946 The kernel PTE inversion protection is unconditionally 2947 enabled and cannot be disabled. 2948 2949 full 2950 Provides all available mitigations for the 2951 L1TF vulnerability. Disables SMT and 2952 enables all mitigations in the 2953 hypervisors, i.e. unconditional L1D flush. 2954 2955 SMT control and L1D flush control via the 2956 sysfs interface is still possible after 2957 boot. Hypervisors will issue a warning 2958 when the first VM is started in a 2959 potentially insecure configuration, 2960 i.e. SMT enabled or L1D flush disabled. 2961 2962 full,force 2963 Same as 'full', but disables SMT and L1D 2964 flush runtime control. Implies the 2965 'nosmt=force' command line option. 2966 (i.e. sysfs control of SMT is disabled.) 2967 2968 flush 2969 Leaves SMT enabled and enables the default 2970 hypervisor mitigation, i.e. conditional 2971 L1D flush. 2972 2973 SMT control and L1D flush control via the 2974 sysfs interface is still possible after 2975 boot. Hypervisors will issue a warning 2976 when the first VM is started in a 2977 potentially insecure configuration, 2978 i.e. SMT enabled or L1D flush disabled. 2979 2980 flush,nosmt 2981 2982 Disables SMT and enables the default 2983 hypervisor mitigation. 2984 2985 SMT control and L1D flush control via the 2986 sysfs interface is still possible after 2987 boot. Hypervisors will issue a warning 2988 when the first VM is started in a 2989 potentially insecure configuration, 2990 i.e. SMT enabled or L1D flush disabled. 2991 2992 flush,nowarn 2993 Same as 'flush', but hypervisors will not 2994 warn when a VM is started in a potentially 2995 insecure configuration. 2996 2997 off 2998 Disables hypervisor mitigations and doesn't 2999 emit any warnings. 3000 It also drops the swap size and available 3001 RAM limit restriction on both hypervisor and 3002 bare metal. 3003 3004 Default is 'flush'. 3005 3006 For details see: Documentation/admin-guide/hw-vuln/l1tf.rst 3007 3008 l2cr= [PPC] 3009 3010 l3cr= [PPC] 3011 3012 lapic [X86-32,APIC,EARLY] Enable the local APIC even if BIOS 3013 disabled it. 3014 3015 lapic= [X86,APIC] Do not use TSC deadline 3016 value for LAPIC timer one-shot implementation. Default 3017 back to the programmable timer unit in the LAPIC. 3018 Format: notscdeadline 3019 3020 lapic_timer_c2_ok [X86,APIC,EARLY] trust the local apic timer 3021 in C2 power state. 3022 3023 libata.dma= [LIBATA] DMA control 3024 libata.dma=0 Disable all PATA and SATA DMA 3025 libata.dma=1 PATA and SATA Disk DMA only 3026 libata.dma=2 ATAPI (CDROM) DMA only 3027 libata.dma=4 Compact Flash DMA only 3028 Combinations also work, so libata.dma=3 enables DMA 3029 for disks and CDROMs, but not CFs. 3030 3031 libata.ignore_hpa= [LIBATA] Ignore HPA limit 3032 libata.ignore_hpa=0 keep BIOS limits (default) 3033 libata.ignore_hpa=1 ignore limits, using full disk 3034 3035 libata.noacpi [LIBATA] Disables use of ACPI in libata suspend/resume 3036 when set. 3037 Format: <int> 3038 3039 libata.force= [LIBATA] Force configurations. The format is a comma- 3040 separated list of "[ID:]VAL" where ID is PORT[.DEVICE]. 3041 PORT and DEVICE are decimal numbers matching port, link 3042 or device. Basically, it matches the ATA ID string 3043 printed on console by libata. If the whole ID part is 3044 omitted, the last PORT and DEVICE values are used. If 3045 ID hasn't been specified yet, the configuration applies 3046 to all ports, links and devices. 3047 3048 If only DEVICE is omitted, the parameter applies to 3049 the port and all links and devices behind it. DEVICE 3050 number of 0 either selects the first device or the 3051 first fan-out link behind PMP device. It does not 3052 select the host link. DEVICE number of 15 selects the 3053 host link and device attached to it. 3054 3055 The VAL specifies the configuration to force. As long 3056 as there is no ambiguity, shortcut notation is allowed. 3057 For example, both 1.5 and 1.5G would work for 1.5Gbps. 3058 The following configurations can be forced. 3059 3060 * Cable type: 40c, 80c, short40c, unk, ign or sata. 3061 Any ID with matching PORT is used. 3062 3063 * SATA link speed limit: 1.5Gbps or 3.0Gbps. 3064 3065 * Transfer mode: pio[0-7], mwdma[0-4] and udma[0-7]. 3066 udma[/][16,25,33,44,66,100,133] notation is also 3067 allowed. 3068 3069 * nohrst, nosrst, norst: suppress hard, soft and both 3070 resets. 3071 3072 * rstonce: only attempt one reset during hot-unplug 3073 link recovery. 3074 3075 * [no]dbdelay: Enable or disable the extra 200ms delay 3076 before debouncing a link PHY and device presence 3077 detection. 3078 3079 * [no]ncq: Turn on or off NCQ. 3080 3081 * [no]ncqtrim: Enable or disable queued DSM TRIM. 3082 3083 * [no]ncqati: Enable or disable NCQ trim on ATI chipset. 3084 3085 * [no]trim: Enable or disable (unqueued) TRIM. 3086 3087 * trim_zero: Indicate that TRIM command zeroes data. 3088 3089 * max_trim_128m: Set 128M maximum trim size limit. 3090 3091 * [no]dma: Turn on or off DMA transfers. 3092 3093 * atapi_dmadir: Enable ATAPI DMADIR bridge support. 3094 3095 * atapi_mod16_dma: Enable the use of ATAPI DMA for 3096 commands that are not a multiple of 16 bytes. 3097 3098 * [no]dmalog: Enable or disable the use of the 3099 READ LOG DMA EXT command to access logs. 3100 3101 * [no]iddevlog: Enable or disable access to the 3102 identify device data log. 3103 3104 * [no]logdir: Enable or disable access to the general 3105 purpose log directory. 3106 3107 * max_sec_128: Set transfer size limit to 128 sectors. 3108 3109 * max_sec_1024: Set or clear transfer size limit to 3110 1024 sectors. 3111 3112 * max_sec_lba48: Set or clear transfer size limit to 3113 65535 sectors. 3114 3115 * [no]lpm: Enable or disable link power management. 3116 3117 * [no]setxfer: Indicate if transfer speed mode setting 3118 should be skipped. 3119 3120 * [no]fua: Disable or enable FUA (Force Unit Access) 3121 support for devices supporting this feature. 3122 3123 * dump_id: Dump IDENTIFY data. 3124 3125 * disable: Disable this device. 3126 3127 If there are multiple matching configurations changing 3128 the same attribute, the last one is used. 3129 3130 load_ramdisk= [RAM] [Deprecated] 3131 3132 lockd.nlm_grace_period=P [NFS] Assign grace period. 3133 Format: <integer> 3134 3135 lockd.nlm_tcpport=N [NFS] Assign TCP port. 3136 Format: <integer> 3137 3138 lockd.nlm_timeout=T [NFS] Assign timeout value. 3139 Format: <integer> 3140 3141 lockd.nlm_udpport=M [NFS] Assign UDP port. 3142 Format: <integer> 3143 3144 lockdown= [SECURITY,EARLY] 3145 { integrity | confidentiality } 3146 Enable the kernel lockdown feature. If set to 3147 integrity, kernel features that allow userland to 3148 modify the running kernel are disabled. If set to 3149 confidentiality, kernel features that allow userland 3150 to extract confidential information from the kernel 3151 are also disabled. 3152 3153 locktorture.acq_writer_lim= [KNL] 3154 Set the time limit in jiffies for a lock 3155 acquisition. Acquisitions exceeding this limit 3156 will result in a splat once they do complete. 3157 3158 locktorture.bind_readers= [KNL] 3159 Specify the list of CPUs to which the readers are 3160 to be bound. 3161 3162 locktorture.bind_writers= [KNL] 3163 Specify the list of CPUs to which the writers are 3164 to be bound. 3165 3166 locktorture.call_rcu_chains= [KNL] 3167 Specify the number of self-propagating call_rcu() 3168 chains to set up. These are used to ensure that 3169 there is a high probability of an RCU grace period 3170 in progress at any given time. Defaults to 0, 3171 which disables these call_rcu() chains. 3172 3173 locktorture.long_hold= [KNL] 3174 Specify the duration in milliseconds for the 3175 occasional long-duration lock hold time. Defaults 3176 to 100 milliseconds. Select 0 to disable. 3177 3178 locktorture.nested_locks= [KNL] 3179 Specify the maximum lock nesting depth that 3180 locktorture is to exercise, up to a limit of 8 3181 (MAX_NESTED_LOCKS). Specify zero to disable. 3182 Note that this parameter is ineffective on types 3183 of locks that do not support nested acquisition. 3184 3185 locktorture.nreaders_stress= [KNL] 3186 Set the number of locking read-acquisition kthreads. 3187 Defaults to being automatically set based on the 3188 number of online CPUs. 3189 3190 locktorture.nwriters_stress= [KNL] 3191 Set the number of locking write-acquisition kthreads. 3192 3193 locktorture.onoff_holdoff= [KNL] 3194 Set time (s) after boot for CPU-hotplug testing. 3195 3196 locktorture.onoff_interval= [KNL] 3197 Set time (s) between CPU-hotplug operations, or 3198 zero to disable CPU-hotplug testing. 3199 3200 locktorture.rt_boost= [KNL] 3201 Do periodic testing of real-time lock priority 3202 boosting. Select 0 to disable, 1 to boost 3203 only rt_mutex, and 2 to boost unconditionally. 3204 Defaults to 2, which might seem to be an 3205 odd choice, but which should be harmless for 3206 non-real-time spinlocks, due to their disabling 3207 of preemption. Note that non-realtime mutexes 3208 disable boosting. 3209 3210 locktorture.rt_boost_factor= [KNL] 3211 Number that determines how often and for how 3212 long priority boosting is exercised. This is 3213 scaled down by the number of writers, so that the 3214 number of boosts per unit time remains roughly 3215 constant as the number of writers increases. 3216 On the other hand, the duration of each boost 3217 increases with the number of writers. 3218 3219 locktorture.shuffle_interval= [KNL] 3220 Set task-shuffle interval (jiffies). Shuffling 3221 tasks allows some CPUs to go into dyntick-idle 3222 mode during the locktorture test. 3223 3224 locktorture.shutdown_secs= [KNL] 3225 Set time (s) after boot system shutdown. This 3226 is useful for hands-off automated testing. 3227 3228 locktorture.stat_interval= [KNL] 3229 Time (s) between statistics printk()s. 3230 3231 locktorture.stutter= [KNL] 3232 Time (s) to stutter testing, for example, 3233 specifying five seconds causes the test to run for 3234 five seconds, wait for five seconds, and so on. 3235 This tests the locking primitive's ability to 3236 transition abruptly to and from idle. 3237 3238 locktorture.torture_type= [KNL] 3239 Specify the locking implementation to test. 3240 3241 locktorture.verbose= [KNL] 3242 Enable additional printk() statements. 3243 3244 locktorture.writer_fifo= [KNL] 3245 Run the write-side locktorture kthreads at 3246 sched_set_fifo() real-time priority. 3247 3248 logibm.irq= [HW,MOUSE] Logitech Bus Mouse Driver 3249 Format: <irq> 3250 3251 loglevel= [KNL,EARLY] 3252 All Kernel Messages with a loglevel smaller than the 3253 console loglevel will be printed to the console. It can 3254 also be changed with klogd or other programs. The 3255 loglevels are defined as follows: 3256 3257 0 (KERN_EMERG) system is unusable 3258 1 (KERN_ALERT) action must be taken immediately 3259 2 (KERN_CRIT) critical conditions 3260 3 (KERN_ERR) error conditions 3261 4 (KERN_WARNING) warning conditions 3262 5 (KERN_NOTICE) normal but significant condition 3263 6 (KERN_INFO) informational 3264 7 (KERN_DEBUG) debug-level messages 3265 3266 log_buf_len=n[KMG] [KNL,EARLY] 3267 Sets the size of the printk ring buffer, in bytes. 3268 n must be a power of two and greater than the 3269 minimal size. The minimal size is defined by 3270 LOG_BUF_SHIFT kernel config parameter. There 3271 is also CONFIG_LOG_CPU_MAX_BUF_SHIFT config 3272 parameter that allows to increase the default size 3273 depending on the number of CPUs. See init/Kconfig 3274 for more details. 3275 3276 logo.nologo [FB] Disables display of the built-in Linux logo. 3277 This may be used to provide more screen space for 3278 kernel log messages and is useful when debugging 3279 kernel boot problems. 3280 3281 lp=0 [LP] Specify parallel ports to use, e.g, 3282 lp=port[,port...] lp=none,parport0 (lp0 not configured, lp1 uses 3283 lp=reset first parallel port). 'lp=0' disables the 3284 lp=auto printer driver. 'lp=reset' (which can be 3285 specified in addition to the ports) causes 3286 attached printers to be reset. Using 3287 lp=port1,port2,... specifies the parallel ports 3288 to associate lp devices with, starting with 3289 lp0. A port specification may be 'none' to skip 3290 that lp device, or a parport name such as 3291 'parport0'. Specifying 'lp=auto' instead of a 3292 port specification list means that device IDs 3293 from each port should be examined, to see if 3294 an IEEE 1284-compliant printer is attached; if 3295 so, the driver will manage that printer. 3296 See also header of drivers/char/lp.c. 3297 3298 lpj=n [KNL] 3299 Sets loops_per_jiffy to given constant, thus avoiding 3300 time-consuming boot-time autodetection (up to 250 ms per 3301 CPU). 0 enables autodetection (default). To determine 3302 the correct value for your kernel, boot with normal 3303 autodetection and see what value is printed. Note that 3304 on SMP systems the preset will be applied to all CPUs, 3305 which is likely to cause problems if your CPUs need 3306 significantly divergent settings. An incorrect value 3307 will cause delays in the kernel to be wrong, leading to 3308 unpredictable I/O errors and other breakage. Although 3309 unlikely, in the extreme case this might damage your 3310 hardware. 3311 3312 lsm.debug [SECURITY] Enable LSM initialization debugging output. 3313 3314 lsm=lsm1,...,lsmN 3315 [SECURITY] Choose order of LSM initialization. This 3316 overrides CONFIG_LSM, and the "security=" parameter. 3317 3318 machtype= [Loongson] Share the same kernel image file between 3319 different yeeloong laptops. 3320 Example: machtype=lemote-yeeloong-2f-7inch 3321 3322 maxcpus= [SMP,EARLY] Maximum number of processors that an SMP kernel 3323 will bring up during bootup. maxcpus=n : n >= 0 limits 3324 the kernel to bring up 'n' processors. Surely after 3325 bootup you can bring up the other plugged cpu by executing 3326 "echo 1 > /sys/devices/system/cpu/cpuX/online". So maxcpus 3327 only takes effect during system bootup. 3328 While n=0 is a special case, it is equivalent to "nosmp", 3329 which also disables the IO APIC. 3330 3331 max_loop= [LOOP] The number of loop block devices that get 3332 (loop.max_loop) unconditionally pre-created at init time. The default 3333 number is configured by BLK_DEV_LOOP_MIN_COUNT. Instead 3334 of statically allocating a predefined number, loop 3335 devices can be requested on-demand with the 3336 /dev/loop-control interface. 3337 3338 mce= [X86-{32,64}] 3339 3340 Please see Documentation/arch/x86/x86_64/machinecheck.rst for sysfs runtime tunables. 3341 3342 off 3343 disable machine check 3344 3345 no_cmci 3346 disable CMCI(Corrected Machine Check Interrupt) that 3347 Intel processor supports. Usually this disablement is 3348 not recommended, but it might be handy if your 3349 hardware is misbehaving. 3350 3351 Note that you'll get more problems without CMCI than 3352 with due to the shared banks, i.e. you might get 3353 duplicated error logs. 3354 3355 dont_log_ce 3356 don't make logs for corrected errors. All events 3357 reported as corrected are silently cleared by OS. This 3358 option will be useful if you have no interest in any 3359 of corrected errors. 3360 3361 ignore_ce 3362 disable features for corrected errors, e.g. 3363 polling timer and CMCI. All events reported as 3364 corrected are not cleared by OS and remained in its 3365 error banks. 3366 3367 Usually this disablement is not recommended, however 3368 if there is an agent checking/clearing corrected 3369 errors (e.g. BIOS or hardware monitoring 3370 applications), conflicting with OS's error handling, 3371 and you cannot deactivate the agent, then this option 3372 will be a help. 3373 3374 no_lmce 3375 do not opt-in to Local MCE delivery. Use legacy method 3376 to broadcast MCEs. 3377 3378 bootlog 3379 enable logging of machine checks left over from 3380 booting. Disabled by default on AMD Fam10h and older 3381 because some BIOS leave bogus ones. 3382 3383 If your BIOS doesn't do that it's a good idea to 3384 enable though to make sure you log even machine check 3385 events that result in a reboot. On Intel systems it is 3386 enabled by default. 3387 3388 nobootlog 3389 disable boot machine check logging. 3390 3391 monarchtimeout (number) 3392 sets the time in us to wait for other CPUs on machine 3393 checks. 0 to disable. 3394 3395 bios_cmci_threshold 3396 don't overwrite the bios-set CMCI threshold. This boot 3397 option prevents Linux from overwriting the CMCI 3398 threshold set by the bios. Without this option, Linux 3399 always sets the CMCI threshold to 1. Enabling this may 3400 make memory predictive failure analysis less effective 3401 if the bios sets thresholds for memory errors since we 3402 will not see details for all errors. 3403 3404 recovery 3405 force-enable recoverable machine check code paths 3406 3407 Everything else is in sysfs now. 3408 3409 3410 md= [HW] RAID subsystems devices and level 3411 See Documentation/admin-guide/md.rst. 3412 3413 mdacon= [MDA] 3414 Format: <first>,<last> 3415 Specifies range of consoles to be captured by the MDA. 3416 3417 mds= [X86,INTEL,EARLY] 3418 Control mitigation for the Micro-architectural Data 3419 Sampling (MDS) vulnerability. 3420 3421 Certain CPUs are vulnerable to an exploit against CPU 3422 internal buffers which can forward information to a 3423 disclosure gadget under certain conditions. 3424 3425 In vulnerable processors, the speculatively 3426 forwarded data can be used in a cache side channel 3427 attack, to access data to which the attacker does 3428 not have direct access. 3429 3430 This parameter controls the MDS mitigation. The 3431 options are: 3432 3433 full - Enable MDS mitigation on vulnerable CPUs 3434 full,nosmt - Enable MDS mitigation and disable 3435 SMT on vulnerable CPUs 3436 off - Unconditionally disable MDS mitigation 3437 3438 On TAA-affected machines, mds=off can be prevented by 3439 an active TAA mitigation as both vulnerabilities are 3440 mitigated with the same mechanism so in order to disable 3441 this mitigation, you need to specify tsx_async_abort=off 3442 too. 3443 3444 Not specifying this option is equivalent to 3445 mds=full. 3446 3447 For details see: Documentation/admin-guide/hw-vuln/mds.rst 3448 3449 mem=nn[KMG] [HEXAGON,EARLY] Set the memory size. 3450 Must be specified, otherwise memory size will be 0. 3451 3452 mem=nn[KMG] [KNL,BOOT,EARLY] Force usage of a specific amount 3453 of memory Amount of memory to be used in cases 3454 as follows: 3455 3456 1 for test; 3457 2 when the kernel is not able to see the whole system memory; 3458 3 memory that lies after 'mem=' boundary is excluded from 3459 the hypervisor, then assigned to KVM guests. 3460 4 to limit the memory available for kdump kernel. 3461 3462 [ARC,MICROBLAZE] - the limit applies only to low memory, 3463 high memory is not affected. 3464 3465 [ARM64] - only limits memory covered by the linear 3466 mapping. The NOMAP regions are not affected. 3467 3468 [X86] Work as limiting max address. Use together 3469 with memmap= to avoid physical address space collisions. 3470 Without memmap= PCI devices could be placed at addresses 3471 belonging to unused RAM. 3472 3473 Note that this only takes effects during boot time since 3474 in above case 3, memory may need be hot added after boot 3475 if system memory of hypervisor is not sufficient. 3476 3477 mem=nn[KMG]@ss[KMG] 3478 [ARM,MIPS,EARLY] - override the memory layout 3479 reported by firmware. 3480 Define a memory region of size nn[KMG] starting at 3481 ss[KMG]. 3482 Multiple different regions can be specified with 3483 multiple mem= parameters on the command line. 3484 3485 mem=nopentium [BUGS=X86-32] Disable usage of 4MB pages for kernel 3486 memory. 3487 3488 memblock=debug [KNL,EARLY] Enable memblock debug messages. 3489 3490 memchunk=nn[KMG] 3491 [KNL,SH] Allow user to override the default size for 3492 per-device physically contiguous DMA buffers. 3493 3494 memhp_default_state=online/offline/online_kernel/online_movable 3495 [KNL] Set the initial state for the memory hotplug 3496 onlining policy. If not specified, the default value is 3497 set according to the 3498 CONFIG_MEMORY_HOTPLUG_DEFAULT_ONLINE kernel config 3499 option. 3500 See Documentation/admin-guide/mm/memory-hotplug.rst. 3501 3502 memmap=exactmap [KNL,X86,EARLY] Enable setting of an exact 3503 E820 memory map, as specified by the user. 3504 Such memmap=exactmap lines can be constructed based on 3505 BIOS output or other requirements. See the memmap=nn@ss 3506 option description. 3507 3508 memmap=nn[KMG]@ss[KMG] 3509 [KNL, X86,MIPS,XTENSA,EARLY] Force usage of a specific region of memory. 3510 Region of memory to be used is from ss to ss+nn. 3511 If @ss[KMG] is omitted, it is equivalent to mem=nn[KMG], 3512 which limits max address to nn[KMG]. 3513 Multiple different regions can be specified, 3514 comma delimited. 3515 Example: 3516 memmap=100M@2G,100M#3G,1G!1024G 3517 3518 memmap=nn[KMG]#ss[KMG] 3519 [KNL,ACPI,EARLY] Mark specific memory as ACPI data. 3520 Region of memory to be marked is from ss to ss+nn. 3521 3522 memmap=nn[KMG]$ss[KMG] 3523 [KNL,ACPI,EARLY] Mark specific memory as reserved. 3524 Region of memory to be reserved is from ss to ss+nn. 3525 Example: Exclude memory from 0x18690000-0x1869ffff 3526 memmap=64K$0x18690000 3527 or 3528 memmap=0x10000$0x18690000 3529 Some bootloaders may need an escape character before '$', 3530 like Grub2, otherwise '$' and the following number 3531 will be eaten. 3532 3533 memmap=nn[KMG]!ss[KMG,EARLY] 3534 [KNL,X86] Mark specific memory as protected. 3535 Region of memory to be used, from ss to ss+nn. 3536 The memory region may be marked as e820 type 12 (0xc) 3537 and is NVDIMM or ADR memory. 3538 3539 memmap=<size>%<offset>-<oldtype>+<newtype> 3540 [KNL,ACPI,EARLY] Convert memory within the specified region 3541 from <oldtype> to <newtype>. If "-<oldtype>" is left 3542 out, the whole region will be marked as <newtype>, 3543 even if previously unavailable. If "+<newtype>" is left 3544 out, matching memory will be removed. Types are 3545 specified as e820 types, e.g., 1 = RAM, 2 = reserved, 3546 3 = ACPI, 12 = PRAM. 3547 3548 memory_corruption_check=0/1 [X86,EARLY] 3549 Some BIOSes seem to corrupt the first 64k of 3550 memory when doing things like suspend/resume. 3551 Setting this option will scan the memory 3552 looking for corruption. Enabling this will 3553 both detect corruption and prevent the kernel 3554 from using the memory being corrupted. 3555 However, its intended as a diagnostic tool; if 3556 repeatable BIOS-originated corruption always 3557 affects the same memory, you can use memmap= 3558 to prevent the kernel from using that memory. 3559 3560 memory_corruption_check_size=size [X86,EARLY] 3561 By default it checks for corruption in the low 3562 64k, making this memory unavailable for normal 3563 use. Use this parameter to scan for 3564 corruption in more or less memory. 3565 3566 memory_corruption_check_period=seconds [X86,EARLY] 3567 By default it checks for corruption every 60 3568 seconds. Use this parameter to check at some 3569 other rate. 0 disables periodic checking. 3570 3571 memory_hotplug.memmap_on_memory 3572 [KNL,X86,ARM] Boolean flag to enable this feature. 3573 Format: {on | off (default)} 3574 When enabled, runtime hotplugged memory will 3575 allocate its internal metadata (struct pages, 3576 those vmemmap pages cannot be optimized even 3577 if hugetlb_free_vmemmap is enabled) from the 3578 hotadded memory which will allow to hotadd a 3579 lot of memory without requiring additional 3580 memory to do so. 3581 This feature is disabled by default because it 3582 has some implication on large (e.g. GB) 3583 allocations in some configurations (e.g. small 3584 memory blocks). 3585 The state of the flag can be read in 3586 /sys/module/memory_hotplug/parameters/memmap_on_memory. 3587 Note that even when enabled, there are a few cases where 3588 the feature is not effective. 3589 3590 memtest= [KNL,X86,ARM,M68K,PPC,RISCV,EARLY] Enable memtest 3591 Format: <integer> 3592 default : 0 <disable> 3593 Specifies the number of memtest passes to be 3594 performed. Each pass selects another test 3595 pattern from a given set of patterns. Memtest 3596 fills the memory with this pattern, validates 3597 memory contents and reserves bad memory 3598 regions that are detected. 3599 3600 mem_encrypt= [X86-64] AMD Secure Memory Encryption (SME) control 3601 Valid arguments: on, off 3602 Default: off 3603 mem_encrypt=on: Activate SME 3604 mem_encrypt=off: Do not activate SME 3605 3606 Refer to Documentation/virt/kvm/x86/amd-memory-encryption.rst 3607 for details on when memory encryption can be activated. 3608 3609 mem_sleep_default= [SUSPEND] Default system suspend mode: 3610 s2idle - Suspend-To-Idle 3611 shallow - Power-On Suspend or equivalent (if supported) 3612 deep - Suspend-To-RAM or equivalent (if supported) 3613 See Documentation/admin-guide/pm/sleep-states.rst. 3614 3615 mfgptfix [X86-32] Fix MFGPT timers on AMD Geode platforms when 3616 the BIOS has incorrectly applied a workaround. TinyBIOS 3617 version 0.98 is known to be affected, 0.99 fixes the 3618 problem by letting the user disable the workaround. 3619 3620 mga= [HW,DRM] 3621 3622 microcode.force_minrev= [X86] 3623 Format: <bool> 3624 Enable or disable the microcode minimal revision 3625 enforcement for the runtime microcode loader. 3626 3627 mini2440= [ARM,HW,KNL] 3628 Format:[0..2][b][c][t] 3629 Default: "0tb" 3630 MINI2440 configuration specification: 3631 0 - The attached screen is the 3.5" TFT 3632 1 - The attached screen is the 7" TFT 3633 2 - The VGA Shield is attached (1024x768) 3634 Leaving out the screen size parameter will not load 3635 the TFT driver, and the framebuffer will be left 3636 unconfigured. 3637 b - Enable backlight. The TFT backlight pin will be 3638 linked to the kernel VESA blanking code and a GPIO 3639 LED. This parameter is not necessary when using the 3640 VGA shield. 3641 c - Enable the s3c camera interface. 3642 t - Reserved for enabling touchscreen support. The 3643 touchscreen support is not enabled in the mainstream 3644 kernel as of 2.6.30, a preliminary port can be found 3645 in the "bleeding edge" mini2440 support kernel at 3646 https://repo.or.cz/w/linux-2.6/mini2440.git 3647 3648 mitigations= 3649 [X86,PPC,S390,ARM64,EARLY] Control optional mitigations for 3650 CPU vulnerabilities. This is a set of curated, 3651 arch-independent options, each of which is an 3652 aggregation of existing arch-specific options. 3653 3654 Note, "mitigations" is supported if and only if the 3655 kernel was built with CPU_MITIGATIONS=y. 3656 3657 off 3658 Disable all optional CPU mitigations. This 3659 improves system performance, but it may also 3660 expose users to several CPU vulnerabilities. 3661 Equivalent to: if nokaslr then kpti=0 [ARM64] 3662 gather_data_sampling=off [X86] 3663 kvm.nx_huge_pages=off [X86] 3664 l1tf=off [X86] 3665 mds=off [X86] 3666 mmio_stale_data=off [X86] 3667 no_entry_flush [PPC] 3668 no_uaccess_flush [PPC] 3669 nobp=0 [S390] 3670 nopti [X86,PPC] 3671 nospectre_bhb [ARM64] 3672 nospectre_v1 [X86,PPC] 3673 nospectre_v2 [X86,PPC,S390,ARM64] 3674 reg_file_data_sampling=off [X86] 3675 retbleed=off [X86] 3676 spec_rstack_overflow=off [X86] 3677 spec_store_bypass_disable=off [X86,PPC] 3678 spectre_bhi=off [X86] 3679 spectre_v2_user=off [X86] 3680 srbds=off [X86,INTEL] 3681 ssbd=force-off [ARM64] 3682 tsx_async_abort=off [X86] 3683 3684 Exceptions: 3685 This does not have any effect on 3686 kvm.nx_huge_pages when 3687 kvm.nx_huge_pages=force. 3688 3689 auto (default) 3690 Mitigate all CPU vulnerabilities, but leave SMT 3691 enabled, even if it's vulnerable. This is for 3692 users who don't want to be surprised by SMT 3693 getting disabled across kernel upgrades, or who 3694 have other ways of avoiding SMT-based attacks. 3695 Equivalent to: (default behavior) 3696 3697 auto,nosmt 3698 Mitigate all CPU vulnerabilities, disabling SMT 3699 if needed. This is for users who always want to 3700 be fully mitigated, even if it means losing SMT. 3701 Equivalent to: l1tf=flush,nosmt [X86] 3702 mds=full,nosmt [X86] 3703 tsx_async_abort=full,nosmt [X86] 3704 mmio_stale_data=full,nosmt [X86] 3705 retbleed=auto,nosmt [X86] 3706 3707 mminit_loglevel= 3708 [KNL,EARLY] When CONFIG_DEBUG_MEMORY_INIT is set, this 3709 parameter allows control of the logging verbosity for 3710 the additional memory initialisation checks. A value 3711 of 0 disables mminit logging and a level of 4 will 3712 log everything. Information is printed at KERN_DEBUG 3713 so loglevel=8 may also need to be specified. 3714 3715 mmio_stale_data= 3716 [X86,INTEL,EARLY] Control mitigation for the Processor 3717 MMIO Stale Data vulnerabilities. 3718 3719 Processor MMIO Stale Data is a class of 3720 vulnerabilities that may expose data after an MMIO 3721 operation. Exposed data could originate or end in 3722 the same CPU buffers as affected by MDS and TAA. 3723 Therefore, similar to MDS and TAA, the mitigation 3724 is to clear the affected CPU buffers. 3725 3726 This parameter controls the mitigation. The 3727 options are: 3728 3729 full - Enable mitigation on vulnerable CPUs 3730 3731 full,nosmt - Enable mitigation and disable SMT on 3732 vulnerable CPUs. 3733 3734 off - Unconditionally disable mitigation 3735 3736 On MDS or TAA affected machines, 3737 mmio_stale_data=off can be prevented by an active 3738 MDS or TAA mitigation as these vulnerabilities are 3739 mitigated with the same mechanism so in order to 3740 disable this mitigation, you need to specify 3741 mds=off and tsx_async_abort=off too. 3742 3743 Not specifying this option is equivalent to 3744 mmio_stale_data=full. 3745 3746 For details see: 3747 Documentation/admin-guide/hw-vuln/processor_mmio_stale_data.rst 3748 3749 <module>.async_probe[=<bool>] [KNL] 3750 If no <bool> value is specified or if the value 3751 specified is not a valid <bool>, enable asynchronous 3752 probe on this module. Otherwise, enable/disable 3753 asynchronous probe on this module as indicated by the 3754 <bool> value. See also: module.async_probe 3755 3756 module.async_probe=<bool> 3757 [KNL] When set to true, modules will use async probing 3758 by default. To enable/disable async probing for a 3759 specific module, use the module specific control that 3760 is documented under <module>.async_probe. When both 3761 module.async_probe and <module>.async_probe are 3762 specified, <module>.async_probe takes precedence for 3763 the specific module. 3764 3765 module.enable_dups_trace 3766 [KNL] When CONFIG_MODULE_DEBUG_AUTOLOAD_DUPS is set, 3767 this means that duplicate request_module() calls will 3768 trigger a WARN_ON() instead of a pr_warn(). Note that 3769 if MODULE_DEBUG_AUTOLOAD_DUPS_TRACE is set, WARN_ON()s 3770 will always be issued and this option does nothing. 3771 module.sig_enforce 3772 [KNL] When CONFIG_MODULE_SIG is set, this means that 3773 modules without (valid) signatures will fail to load. 3774 Note that if CONFIG_MODULE_SIG_FORCE is set, that 3775 is always true, so this option does nothing. 3776 3777 module_blacklist= [KNL] Do not load a comma-separated list of 3778 modules. Useful for debugging problem modules. 3779 3780 mousedev.tap_time= 3781 [MOUSE] Maximum time between finger touching and 3782 leaving touchpad surface for touch to be considered 3783 a tap and be reported as a left button click (for 3784 touchpads working in absolute mode only). 3785 Format: <msecs> 3786 mousedev.xres= [MOUSE] Horizontal screen resolution, used for devices 3787 reporting absolute coordinates, such as tablets 3788 mousedev.yres= [MOUSE] Vertical screen resolution, used for devices 3789 reporting absolute coordinates, such as tablets 3790 3791 movablecore= [KNL,X86,PPC,EARLY] 3792 Format: nn[KMGTPE] | nn% 3793 This parameter is the complement to kernelcore=, it 3794 specifies the amount of memory used for migratable 3795 allocations. If both kernelcore and movablecore is 3796 specified, then kernelcore will be at *least* the 3797 specified value but may be more. If movablecore on its 3798 own is specified, the administrator must be careful 3799 that the amount of memory usable for all allocations 3800 is not too small. 3801 3802 movable_node [KNL,EARLY] Boot-time switch to make hotplugable memory 3803 NUMA nodes to be movable. This means that the memory 3804 of such nodes will be usable only for movable 3805 allocations which rules out almost all kernel 3806 allocations. Use with caution! 3807 3808 MTD_Partition= [MTD] 3809 Format: <name>,<region-number>,<size>,<offset> 3810 3811 MTD_Region= [MTD] Format: 3812 <name>,<region-number>[,<base>,<size>,<buswidth>,<altbuswidth>] 3813 3814 mtdparts= [MTD] 3815 See drivers/mtd/parsers/cmdlinepart.c 3816 3817 mtouchusb.raw_coordinates= 3818 [HW] Make the MicroTouch USB driver use raw coordinates 3819 ('y', default) or cooked coordinates ('n') 3820 3821 mtrr=debug [X86,EARLY] 3822 Enable printing debug information related to MTRR 3823 registers at boot time. 3824 3825 mtrr_chunk_size=nn[KMG,X86,EARLY] 3826 used for mtrr cleanup. It is largest continuous chunk 3827 that could hold holes aka. UC entries. 3828 3829 mtrr_gran_size=nn[KMG,X86,EARLY] 3830 Used for mtrr cleanup. It is granularity of mtrr block. 3831 Default is 1. 3832 Large value could prevent small alignment from 3833 using up MTRRs. 3834 3835 mtrr_spare_reg_nr=n [X86,EARLY] 3836 Format: <integer> 3837 Range: 0,7 : spare reg number 3838 Default : 1 3839 Used for mtrr cleanup. It is spare mtrr entries number. 3840 Set to 2 or more if your graphical card needs more. 3841 3842 multitce=off [PPC] This parameter disables the use of the pSeries 3843 firmware feature for updating multiple TCE entries 3844 at a time. 3845 3846 n2= [NET] SDL Inc. RISCom/N2 synchronous serial card 3847 3848 netdev= [NET] Network devices parameters 3849 Format: <irq>,<io>,<mem_start>,<mem_end>,<name> 3850 Note that mem_start is often overloaded to mean 3851 something different and driver-specific. 3852 This usage is only documented in each driver source 3853 file if at all. 3854 3855 netpoll.carrier_timeout= 3856 [NET] Specifies amount of time (in seconds) that 3857 netpoll should wait for a carrier. By default netpoll 3858 waits 4 seconds. 3859 3860 nf_conntrack.acct= 3861 [NETFILTER] Enable connection tracking flow accounting 3862 0 to disable accounting 3863 1 to enable accounting 3864 Default value is 0. 3865 3866 nfs.cache_getent= 3867 [NFS] sets the pathname to the program which is used 3868 to update the NFS client cache entries. 3869 3870 nfs.cache_getent_timeout= 3871 [NFS] sets the timeout after which an attempt to 3872 update a cache entry is deemed to have failed. 3873 3874 nfs.callback_nr_threads= 3875 [NFSv4] set the total number of threads that the 3876 NFS client will assign to service NFSv4 callback 3877 requests. 3878 3879 nfs.callback_tcpport= 3880 [NFS] set the TCP port on which the NFSv4 callback 3881 channel should listen. 3882 3883 nfs.delay_retrans= 3884 [NFS] specifies the number of times the NFSv4 client 3885 retries the request before returning an EAGAIN error, 3886 after a reply of NFS4ERR_DELAY from the server. 3887 Only applies if the softerr mount option is enabled, 3888 and the specified value is >= 0. 3889 3890 nfs.enable_ino64= 3891 [NFS] enable 64-bit inode numbers. 3892 If zero, the NFS client will fake up a 32-bit inode 3893 number for the readdir() and stat() syscalls instead 3894 of returning the full 64-bit number. 3895 The default is to return 64-bit inode numbers. 3896 3897 nfs.idmap_cache_timeout= 3898 [NFS] set the maximum lifetime for idmapper cache 3899 entries. 3900 3901 nfs.max_session_cb_slots= 3902 [NFSv4.1] Sets the maximum number of session 3903 slots the client will assign to the callback 3904 channel. This determines the maximum number of 3905 callbacks the client will process in parallel for 3906 a particular server. 3907 3908 nfs.max_session_slots= 3909 [NFSv4.1] Sets the maximum number of session slots 3910 the client will attempt to negotiate with the server. 3911 This limits the number of simultaneous RPC requests 3912 that the client can send to the NFSv4.1 server. 3913 Note that there is little point in setting this 3914 value higher than the max_tcp_slot_table_limit. 3915 3916 nfs.nfs4_disable_idmapping= 3917 [NFSv4] When set to the default of '1', this option 3918 ensures that both the RPC level authentication 3919 scheme and the NFS level operations agree to use 3920 numeric uids/gids if the mount is using the 3921 'sec=sys' security flavour. In effect it is 3922 disabling idmapping, which can make migration from 3923 legacy NFSv2/v3 systems to NFSv4 easier. 3924 Servers that do not support this mode of operation 3925 will be autodetected by the client, and it will fall 3926 back to using the idmapper. 3927 To turn off this behaviour, set the value to '0'. 3928 3929 nfs.nfs4_unique_id= 3930 [NFS4] Specify an additional fixed unique ident- 3931 ification string that NFSv4 clients can insert into 3932 their nfs_client_id4 string. This is typically a 3933 UUID that is generated at system install time. 3934 3935 nfs.recover_lost_locks= 3936 [NFSv4] Attempt to recover locks that were lost due 3937 to a lease timeout on the server. Please note that 3938 doing this risks data corruption, since there are 3939 no guarantees that the file will remain unchanged 3940 after the locks are lost. 3941 If you want to enable the kernel legacy behaviour of 3942 attempting to recover these locks, then set this 3943 parameter to '1'. 3944 The default parameter value of '0' causes the kernel 3945 not to attempt recovery of lost locks. 3946 3947 nfs.send_implementation_id= 3948 [NFSv4.1] Send client implementation identification 3949 information in exchange_id requests. 3950 If zero, no implementation identification information 3951 will be sent. 3952 The default is to send the implementation identification 3953 information. 3954 3955 nfs4.layoutstats_timer= 3956 [NFSv4.2] Change the rate at which the kernel sends 3957 layoutstats to the pNFS metadata server. 3958 3959 Setting this to value to 0 causes the kernel to use 3960 whatever value is the default set by the layout 3961 driver. A non-zero value sets the minimum interval 3962 in seconds between layoutstats transmissions. 3963 3964 nfsd.inter_copy_offload_enable= 3965 [NFSv4.2] When set to 1, the server will support 3966 server-to-server copies for which this server is 3967 the destination of the copy. 3968 3969 nfsd.nfs4_disable_idmapping= 3970 [NFSv4] When set to the default of '1', the NFSv4 3971 server will return only numeric uids and gids to 3972 clients using auth_sys, and will accept numeric uids 3973 and gids from such clients. This is intended to ease 3974 migration from NFSv2/v3. 3975 3976 nfsd.nfsd4_ssc_umount_timeout= 3977 [NFSv4.2] When used as the destination of a 3978 server-to-server copy, knfsd temporarily mounts 3979 the source server. It caches the mount in case 3980 it will be needed again, and discards it if not 3981 used for the number of milliseconds specified by 3982 this parameter. 3983 3984 nfsaddrs= [NFS] Deprecated. Use ip= instead. 3985 See Documentation/admin-guide/nfs/nfsroot.rst. 3986 3987 nfsroot= [NFS] nfs root filesystem for disk-less boxes. 3988 See Documentation/admin-guide/nfs/nfsroot.rst. 3989 3990 nfsrootdebug [NFS] enable nfsroot debugging messages. 3991 See Documentation/admin-guide/nfs/nfsroot.rst. 3992 3993 nmi_backtrace.backtrace_idle [KNL] 3994 Dump stacks even of idle CPUs in response to an 3995 NMI stack-backtrace request. 3996 3997 nmi_debug= [KNL,SH] Specify one or more actions to take 3998 when a NMI is triggered. 3999 Format: [state][,regs][,debounce][,die] 4000 4001 nmi_watchdog= [KNL,BUGS=X86] Debugging features for SMP kernels 4002 Format: [panic,][nopanic,][rNNN,][num] 4003 Valid num: 0 or 1 4004 0 - turn hardlockup detector in nmi_watchdog off 4005 1 - turn hardlockup detector in nmi_watchdog on 4006 rNNN - configure the watchdog with raw perf event 0xNNN 4007 4008 When panic is specified, panic when an NMI watchdog 4009 timeout occurs (or 'nopanic' to not panic on an NMI 4010 watchdog, if CONFIG_BOOTPARAM_HARDLOCKUP_PANIC is set) 4011 To disable both hard and soft lockup detectors, 4012 please see 'nowatchdog'. 4013 This is useful when you use a panic=... timeout and 4014 need the box quickly up again. 4015 4016 These settings can be accessed at runtime via 4017 the nmi_watchdog and hardlockup_panic sysctls. 4018 4019 no387 [BUGS=X86-32] Tells the kernel to use the 387 maths 4020 emulation library even if a 387 maths coprocessor 4021 is present. 4022 4023 no4lvl [RISCV,EARLY] Disable 4-level and 5-level paging modes. 4024 Forces kernel to use 3-level paging instead. 4025 4026 no5lvl [X86-64,RISCV,EARLY] Disable 5-level paging mode. Forces 4027 kernel to use 4-level paging instead. 4028 4029 noalign [KNL,ARM] 4030 4031 noapic [SMP,APIC,EARLY] Tells the kernel to not make use of any 4032 IOAPICs that may be present in the system. 4033 4034 noapictimer [APIC,X86] Don't set up the APIC timer 4035 4036 noautogroup Disable scheduler automatic task group creation. 4037 4038 nocache [ARM,EARLY] 4039 4040 no_console_suspend 4041 [HW] Never suspend the console 4042 Disable suspending of consoles during suspend and 4043 hibernate operations. Once disabled, debugging 4044 messages can reach various consoles while the rest 4045 of the system is being put to sleep (ie, while 4046 debugging driver suspend/resume hooks). This may 4047 not work reliably with all consoles, but is known 4048 to work with serial and VGA consoles. 4049 To facilitate more flexible debugging, we also add 4050 console_suspend, a printk module parameter to control 4051 it. Users could use console_suspend (usually 4052 /sys/module/printk/parameters/console_suspend) to 4053 turn on/off it dynamically. 4054 4055 no_debug_objects 4056 [KNL,EARLY] Disable object debugging 4057 4058 nodsp [SH] Disable hardware DSP at boot time. 4059 4060 noefi [EFI,EARLY] Disable EFI runtime services support. 4061 4062 no_entry_flush [PPC,EARLY] Don't flush the L1-D cache when entering the kernel. 4063 4064 noexec32 [X86-64] 4065 This affects only 32-bit executables. 4066 noexec32=on: enable non-executable mappings (default) 4067 read doesn't imply executable mappings 4068 noexec32=off: disable non-executable mappings 4069 read implies executable mappings 4070 4071 no_file_caps Tells the kernel not to honor file capabilities. The 4072 only way then for a file to be executed with privilege 4073 is to be setuid root or executed by root. 4074 4075 nofpu [MIPS,SH] Disable hardware FPU at boot time. 4076 4077 nofsgsbase [X86] Disables FSGSBASE instructions. 4078 4079 nofxsr [BUGS=X86-32] Disables x86 floating point extended 4080 register save and restore. The kernel will only save 4081 legacy floating-point registers on task switch. 4082 4083 nogbpages [X86] Do not use GB pages for kernel direct mappings. 4084 4085 no_hash_pointers 4086 [KNL,EARLY] 4087 Force pointers printed to the console or buffers to be 4088 unhashed. By default, when a pointer is printed via %p 4089 format string, that pointer is "hashed", i.e. obscured 4090 by hashing the pointer value. This is a security feature 4091 that hides actual kernel addresses from unprivileged 4092 users, but it also makes debugging the kernel more 4093 difficult since unequal pointers can no longer be 4094 compared. However, if this command-line option is 4095 specified, then all normal pointers will have their true 4096 value printed. This option should only be specified when 4097 debugging the kernel. Please do not use on production 4098 kernels. 4099 4100 nohibernate [HIBERNATION] Disable hibernation and resume. 4101 4102 nohlt [ARM,ARM64,MICROBLAZE,MIPS,PPC,RISCV,SH] Forces the kernel to 4103 busy wait in do_idle() and not use the arch_cpu_idle() 4104 implementation; requires CONFIG_GENERIC_IDLE_POLL_SETUP 4105 to be effective. This is useful on platforms where the 4106 sleep(SH) or wfi(ARM,ARM64) instructions do not work 4107 correctly or when doing power measurements to evaluate 4108 the impact of the sleep instructions. This is also 4109 useful when using JTAG debugger. 4110 4111 nohpet [X86] Don't use the HPET timer. 4112 4113 nohugeiomap [KNL,X86,PPC,ARM64,EARLY] Disable kernel huge I/O mappings. 4114 4115 nohugevmalloc [KNL,X86,PPC,ARM64,EARLY] Disable kernel huge vmalloc mappings. 4116 4117 nohz= [KNL] Boottime enable/disable dynamic ticks 4118 Valid arguments: on, off 4119 Default: on 4120 4121 nohz_full= [KNL,BOOT,SMP,ISOL] 4122 The argument is a cpu list, as described above. 4123 In kernels built with CONFIG_NO_HZ_FULL=y, set 4124 the specified list of CPUs whose tick will be stopped 4125 whenever possible. The boot CPU will be forced outside 4126 the range to maintain the timekeeping. Any CPUs 4127 in this list will have their RCU callbacks offloaded, 4128 just as if they had also been called out in the 4129 rcu_nocbs= boot parameter. 4130 4131 Note that this argument takes precedence over 4132 the CONFIG_RCU_NOCB_CPU_DEFAULT_ALL option. 4133 4134 noinitrd [RAM] Tells the kernel not to load any configured 4135 initial RAM disk. 4136 4137 nointremap [X86-64,Intel-IOMMU,EARLY] Do not enable interrupt 4138 remapping. 4139 [Deprecated - use intremap=off] 4140 4141 noinvpcid [X86,EARLY] Disable the INVPCID cpu feature. 4142 4143 noiotrap [SH] Disables trapped I/O port accesses. 4144 4145 noirqdebug [X86-32] Disables the code which attempts to detect and 4146 disable unhandled interrupt sources. 4147 4148 noisapnp [ISAPNP] Disables ISA PnP code. 4149 4150 nokaslr [KNL,EARLY] 4151 When CONFIG_RANDOMIZE_BASE is set, this disables 4152 kernel and module base offset ASLR (Address Space 4153 Layout Randomization). 4154 4155 no-kvmapf [X86,KVM,EARLY] Disable paravirtualized asynchronous page 4156 fault handling. 4157 4158 no-kvmclock [X86,KVM,EARLY] Disable paravirtualized KVM clock driver 4159 4160 nolapic [X86-32,APIC,EARLY] Do not enable or use the local APIC. 4161 4162 nolapic_timer [X86-32,APIC,EARLY] Do not use the local APIC timer. 4163 4164 nomce [X86-32] Disable Machine Check Exception 4165 4166 nomfgpt [X86-32] Disable Multi-Function General Purpose 4167 Timer usage (for AMD Geode machines). 4168 4169 nomodeset Disable kernel modesetting. Most systems' firmware 4170 sets up a display mode and provides framebuffer memory 4171 for output. With nomodeset, DRM and fbdev drivers will 4172 not load if they could possibly displace the pre- 4173 initialized output. Only the system framebuffer will 4174 be available for use. The respective drivers will not 4175 perform display-mode changes or accelerated rendering. 4176 4177 Useful as error fallback, or for testing and debugging. 4178 4179 nomodule Disable module load 4180 4181 nonmi_ipi [X86] Disable using NMI IPIs during panic/reboot to 4182 shutdown the other cpus. Instead use the REBOOT_VECTOR 4183 irq. 4184 4185 nopat [X86,EARLY] Disable PAT (page attribute table extension of 4186 pagetables) support. 4187 4188 nopcid [X86-64,EARLY] Disable the PCID cpu feature. 4189 4190 nopku [X86] Disable Memory Protection Keys CPU feature found 4191 in some Intel CPUs. 4192 4193 nopti [X86-64,EARLY] 4194 Equivalent to pti=off 4195 4196 nopv= [X86,XEN,KVM,HYPER_V,VMWARE,EARLY] 4197 Disables the PV optimizations forcing the guest to run 4198 as generic guest with no PV drivers. Currently support 4199 XEN HVM, KVM, HYPER_V and VMWARE guest. 4200 4201 nopvspin [X86,XEN,KVM,EARLY] 4202 Disables the qspinlock slow path using PV optimizations 4203 which allow the hypervisor to 'idle' the guest on lock 4204 contention. 4205 4206 norandmaps Don't use address space randomization. Equivalent to 4207 echo 0 > /proc/sys/kernel/randomize_va_space 4208 4209 noreplace-smp [X86-32,SMP] Don't replace SMP instructions 4210 with UP alternatives 4211 4212 noresume [SWSUSP] Disables resume and restores original swap 4213 space. 4214 4215 no-scroll [VGA] Disables scrollback. 4216 This is required for the Braillex ib80-piezo Braille 4217 reader made by F.H. Papenmeier (Germany). 4218 4219 nosgx [X86-64,SGX,EARLY] Disables Intel SGX kernel support. 4220 4221 nosmap [PPC,EARLY] 4222 Disable SMAP (Supervisor Mode Access Prevention) 4223 even if it is supported by processor. 4224 4225 nosmep [PPC64s,EARLY] 4226 Disable SMEP (Supervisor Mode Execution Prevention) 4227 even if it is supported by processor. 4228 4229 nosmp [SMP,EARLY] Tells an SMP kernel to act as a UP kernel, 4230 and disable the IO APIC. legacy for "maxcpus=0". 4231 4232 nosmt [KNL,MIPS,PPC,S390,EARLY] Disable symmetric multithreading (SMT). 4233 Equivalent to smt=1. 4234 4235 [KNL,X86,PPC] Disable symmetric multithreading (SMT). 4236 nosmt=force: Force disable SMT, cannot be undone 4237 via the sysfs control file. 4238 4239 nosoftlockup [KNL] Disable the soft-lockup detector. 4240 4241 nospec_store_bypass_disable 4242 [HW,EARLY] Disable all mitigations for the Speculative 4243 Store Bypass vulnerability 4244 4245 nospectre_bhb [ARM64,EARLY] Disable all mitigations for Spectre-BHB (branch 4246 history injection) vulnerability. System may allow data leaks 4247 with this option. 4248 4249 nospectre_v1 [X86,PPC,EARLY] Disable mitigations for Spectre Variant 1 4250 (bounds check bypass). With this option data leaks are 4251 possible in the system. 4252 4253 nospectre_v2 [X86,PPC_E500,ARM64,EARLY] Disable all mitigations 4254 for the Spectre variant 2 (indirect branch 4255 prediction) vulnerability. System may allow data 4256 leaks with this option. 4257 4258 no-steal-acc [X86,PV_OPS,ARM64,PPC/PSERIES,RISCV,LOONGARCH,EARLY] 4259 Disable paravirtualized steal time accounting. steal time 4260 is computed, but won't influence scheduler behaviour 4261 4262 nosync [HW,M68K] Disables sync negotiation for all devices. 4263 4264 no_timer_check [X86,APIC] Disables the code which tests for broken 4265 timer IRQ sources, i.e., the IO-APIC timer. This can 4266 work around problems with incorrect timer 4267 initialization on some boards. 4268 4269 no_uaccess_flush 4270 [PPC,EARLY] Don't flush the L1-D cache after accessing user data. 4271 4272 novmcoredd [KNL,KDUMP] 4273 Disable device dump. Device dump allows drivers to 4274 append dump data to vmcore so you can collect driver 4275 specified debug info. Drivers can append the data 4276 without any limit and this data is stored in memory, 4277 so this may cause significant memory stress. Disabling 4278 device dump can help save memory but the driver debug 4279 data will be no longer available. This parameter 4280 is only available when CONFIG_PROC_VMCORE_DEVICE_DUMP 4281 is set. 4282 4283 no-vmw-sched-clock 4284 [X86,PV_OPS,EARLY] Disable paravirtualized VMware 4285 scheduler clock and use the default one. 4286 4287 nowatchdog [KNL] Disable both lockup detectors, i.e. 4288 soft-lockup and NMI watchdog (hard-lockup). 4289 4290 nowb [ARM,EARLY] 4291 4292 nox2apic [X86-64,APIC,EARLY] Do not enable x2APIC mode. 4293 4294 NOTE: this parameter will be ignored on systems with the 4295 LEGACY_XAPIC_DISABLED bit set in the 4296 IA32_XAPIC_DISABLE_STATUS MSR. 4297 4298 noxsave [BUGS=X86] Disables x86 extended register state save 4299 and restore using xsave. The kernel will fallback to 4300 enabling legacy floating-point and sse state. 4301 4302 noxsaveopt [X86] Disables xsaveopt used in saving x86 extended 4303 register states. The kernel will fall back to use 4304 xsave to save the states. By using this parameter, 4305 performance of saving the states is degraded because 4306 xsave doesn't support modified optimization while 4307 xsaveopt supports it on xsaveopt enabled systems. 4308 4309 noxsaves [X86] Disables xsaves and xrstors used in saving and 4310 restoring x86 extended register state in compacted 4311 form of xsave area. The kernel will fall back to use 4312 xsaveopt and xrstor to save and restore the states 4313 in standard form of xsave area. By using this 4314 parameter, xsave area per process might occupy more 4315 memory on xsaves enabled systems. 4316 4317 nr_cpus= [SMP,EARLY] Maximum number of processors that an SMP kernel 4318 could support. nr_cpus=n : n >= 1 limits the kernel to 4319 support 'n' processors. It could be larger than the 4320 number of already plugged CPU during bootup, later in 4321 runtime you can physically add extra cpu until it reaches 4322 n. So during boot up some boot time memory for per-cpu 4323 variables need be pre-allocated for later physical cpu 4324 hot plugging. 4325 4326 nr_uarts= [SERIAL] maximum number of UARTs to be registered. 4327 4328 numa=off [KNL, ARM64, PPC, RISCV, SPARC, X86, EARLY] 4329 Disable NUMA, Only set up a single NUMA node 4330 spanning all memory. 4331 4332 numa=fake=<size>[MG] 4333 [KNL, ARM64, RISCV, X86, EARLY] 4334 If given as a memory unit, fills all system RAM with 4335 nodes of size interleaved over physical nodes. 4336 4337 numa=fake=<N> 4338 [KNL, ARM64, RISCV, X86, EARLY] 4339 If given as an integer, fills all system RAM with N 4340 fake nodes interleaved over physical nodes. 4341 4342 numa=fake=<N>U 4343 [KNL, ARM64, RISCV, X86, EARLY] 4344 If given as an integer followed by 'U', it will 4345 divide each physical node into N emulated nodes. 4346 4347 numa=noacpi [X86] Don't parse the SRAT table for NUMA setup 4348 4349 numa=nohmat [X86] Don't parse the HMAT table for NUMA setup, or 4350 soft-reserved memory partitioning. 4351 4352 numa_balancing= [KNL,ARM64,PPC,RISCV,S390,X86] Enable or disable automatic 4353 NUMA balancing. 4354 Allowed values are enable and disable 4355 4356 numa_zonelist_order= [KNL, BOOT] Select zonelist order for NUMA. 4357 'node', 'default' can be specified 4358 This can be set from sysctl after boot. 4359 See Documentation/admin-guide/sysctl/vm.rst for details. 4360 4361 ohci1394_dma=early [HW,EARLY] enable debugging via the ohci1394 driver. 4362 See Documentation/core-api/debugging-via-ohci1394.rst for more 4363 info. 4364 4365 olpc_ec_timeout= [OLPC] ms delay when issuing EC commands 4366 Rather than timing out after 20 ms if an EC 4367 command is not properly ACKed, override the length 4368 of the timeout. We have interrupts disabled while 4369 waiting for the ACK, so if this is set too high 4370 interrupts *may* be lost! 4371 4372 omap_mux= [OMAP] Override bootloader pin multiplexing. 4373 Format: <mux_mode0.mode_name=value>... 4374 For example, to override I2C bus2: 4375 omap_mux=i2c2_scl.i2c2_scl=0x100,i2c2_sda.i2c2_sda=0x100 4376 4377 onenand.bdry= [HW,MTD] Flex-OneNAND Boundary Configuration 4378 4379 Format: [die0_boundary][,die0_lock][,die1_boundary][,die1_lock] 4380 4381 boundary - index of last SLC block on Flex-OneNAND. 4382 The remaining blocks are configured as MLC blocks. 4383 lock - Configure if Flex-OneNAND boundary should be locked. 4384 Once locked, the boundary cannot be changed. 4385 1 indicates lock status, 0 indicates unlock status. 4386 4387 oops=panic [KNL,EARLY] 4388 Always panic on oopses. Default is to just kill the 4389 process, but there is a small probability of 4390 deadlocking the machine. 4391 This will also cause panics on machine check exceptions. 4392 Useful together with panic=30 to trigger a reboot. 4393 4394 page_alloc.shuffle= 4395 [KNL] Boolean flag to control whether the page allocator 4396 should randomize its free lists. This parameter can be 4397 used to enable/disable page randomization. The state of 4398 the flag can be read from sysfs at: 4399 /sys/module/page_alloc/parameters/shuffle. 4400 This parameter is only available if CONFIG_SHUFFLE_PAGE_ALLOCATOR=y. 4401 4402 page_owner= [KNL,EARLY] Boot-time page_owner enabling option. 4403 Storage of the information about who allocated 4404 each page is disabled in default. With this switch, 4405 we can turn it on. 4406 on: enable the feature 4407 4408 page_poison= [KNL,EARLY] Boot-time parameter changing the state of 4409 poisoning on the buddy allocator, available with 4410 CONFIG_PAGE_POISONING=y. 4411 off: turn off poisoning (default) 4412 on: turn on poisoning 4413 4414 page_reporting.page_reporting_order= 4415 [KNL] Minimal page reporting order 4416 Format: <integer> 4417 Adjust the minimal page reporting order. The page 4418 reporting is disabled when it exceeds MAX_PAGE_ORDER. 4419 4420 panic= [KNL] Kernel behaviour on panic: delay <timeout> 4421 timeout > 0: seconds before rebooting 4422 timeout = 0: wait forever 4423 timeout < 0: reboot immediately 4424 Format: <timeout> 4425 4426 panic_on_taint= [KNL,EARLY] 4427 Bitmask for conditionally calling panic() in add_taint() 4428 Format: <hex>[,nousertaint] 4429 Hexadecimal bitmask representing the set of TAINT flags 4430 that will cause the kernel to panic when add_taint() is 4431 called with any of the flags in this set. 4432 The optional switch "nousertaint" can be utilized to 4433 prevent userspace forced crashes by writing to sysctl 4434 /proc/sys/kernel/tainted any flagset matching with the 4435 bitmask set on panic_on_taint. 4436 See Documentation/admin-guide/tainted-kernels.rst for 4437 extra details on the taint flags that users can pick 4438 to compose the bitmask to assign to panic_on_taint. 4439 4440 panic_on_warn=1 panic() instead of WARN(). Useful to cause kdump 4441 on a WARN(). 4442 4443 panic_print= Bitmask for printing system info when panic happens. 4444 User can chose combination of the following bits: 4445 bit 0: print all tasks info 4446 bit 1: print system memory info 4447 bit 2: print timer info 4448 bit 3: print locks info if CONFIG_LOCKDEP is on 4449 bit 4: print ftrace buffer 4450 bit 5: print all printk messages in buffer 4451 bit 6: print all CPUs backtrace (if available in the arch) 4452 bit 7: print only tasks in uninterruptible (blocked) state 4453 *Be aware* that this option may print a _lot_ of lines, 4454 so there are risks of losing older messages in the log. 4455 Use this option carefully, maybe worth to setup a 4456 bigger log buffer with "log_buf_len" along with this. 4457 4458 parkbd.port= [HW] Parallel port number the keyboard adapter is 4459 connected to, default is 0. 4460 Format: <parport#> 4461 parkbd.mode= [HW] Parallel port keyboard adapter mode of operation, 4462 0 for XT, 1 for AT (default is AT). 4463 Format: <mode> 4464 4465 parport= [HW,PPT] Specify parallel ports. 0 disables. 4466 Format: { 0 | auto | 0xBBB[,IRQ[,DMA]] } 4467 Use 'auto' to force the driver to use any 4468 IRQ/DMA settings detected (the default is to 4469 ignore detected IRQ/DMA settings because of 4470 possible conflicts). You can specify the base 4471 address, IRQ, and DMA settings; IRQ and DMA 4472 should be numbers, or 'auto' (for using detected 4473 settings on that particular port), or 'nofifo' 4474 (to avoid using a FIFO even if it is detected). 4475 Parallel ports are assigned in the order they 4476 are specified on the command line, starting 4477 with parport0. 4478 4479 parport_init_mode= [HW,PPT] 4480 Configure VIA parallel port to operate in 4481 a specific mode. This is necessary on Pegasos 4482 computer where firmware has no options for setting 4483 up parallel port mode and sets it to spp. 4484 Currently this function knows 686a and 8231 chips. 4485 Format: [spp|ps2|epp|ecp|ecpepp] 4486 4487 pata_legacy.all= [HW,LIBATA] 4488 Format: <int> 4489 Set to non-zero to probe primary and secondary ISA 4490 port ranges on PCI systems where no PCI PATA device 4491 has been found at either range. Disabled by default. 4492 4493 pata_legacy.autospeed= [HW,LIBATA] 4494 Format: <int> 4495 Set to non-zero if a chip is present that snoops speed 4496 changes. Disabled by default. 4497 4498 pata_legacy.ht6560a= [HW,LIBATA] 4499 Format: <int> 4500 Set to 1, 2, or 3 for HT 6560A on the primary channel, 4501 the secondary channel, or both channels respectively. 4502 Disabled by default. 4503 4504 pata_legacy.ht6560b= [HW,LIBATA] 4505 Format: <int> 4506 Set to 1, 2, or 3 for HT 6560B on the primary channel, 4507 the secondary channel, or both channels respectively. 4508 Disabled by default. 4509 4510 pata_legacy.iordy_mask= [HW,LIBATA] 4511 Format: <int> 4512 IORDY enable mask. Set individual bits to allow IORDY 4513 for the respective channel. Bit 0 is for the first 4514 legacy channel handled by this driver, bit 1 is for 4515 the second channel, and so on. The sequence will often 4516 correspond to the primary legacy channel, the secondary 4517 legacy channel, and so on, but the handling of a PCI 4518 bus and the use of other driver options may interfere 4519 with the sequence. By default IORDY is allowed across 4520 all channels. 4521 4522 pata_legacy.opti82c46x= [HW,LIBATA] 4523 Format: <int> 4524 Set to 1, 2, or 3 for Opti 82c611A on the primary 4525 channel, the secondary channel, or both channels 4526 respectively. Disabled by default. 4527 4528 pata_legacy.opti82c611a= [HW,LIBATA] 4529 Format: <int> 4530 Set to 1, 2, or 3 for Opti 82c465MV on the primary 4531 channel, the secondary channel, or both channels 4532 respectively. Disabled by default. 4533 4534 pata_legacy.pio_mask= [HW,LIBATA] 4535 Format: <int> 4536 PIO mode mask for autospeed devices. Set individual 4537 bits to allow the use of the respective PIO modes. 4538 Bit 0 is for mode 0, bit 1 is for mode 1, and so on. 4539 All modes allowed by default. 4540 4541 pata_legacy.probe_all= [HW,LIBATA] 4542 Format: <int> 4543 Set to non-zero to probe tertiary and further ISA 4544 port ranges on PCI systems. Disabled by default. 4545 4546 pata_legacy.probe_mask= [HW,LIBATA] 4547 Format: <int> 4548 Probe mask for legacy ISA PATA ports. Depending on 4549 platform configuration and the use of other driver 4550 options up to 6 legacy ports are supported: 0x1f0, 4551 0x170, 0x1e8, 0x168, 0x1e0, 0x160, however probing 4552 of individual ports can be disabled by setting the 4553 corresponding bits in the mask to 1. Bit 0 is for 4554 the first port in the list above (0x1f0), and so on. 4555 By default all supported ports are probed. 4556 4557 pata_legacy.qdi= [HW,LIBATA] 4558 Format: <int> 4559 Set to non-zero to probe QDI controllers. By default 4560 set to 1 if CONFIG_PATA_QDI_MODULE, 0 otherwise. 4561 4562 pata_legacy.winbond= [HW,LIBATA] 4563 Format: <int> 4564 Set to non-zero to probe Winbond controllers. Use 4565 the standard I/O port (0x130) if 1, otherwise the 4566 value given is the I/O port to use (typically 0x1b0). 4567 By default set to 1 if CONFIG_PATA_WINBOND_VLB_MODULE, 4568 0 otherwise. 4569 4570 pata_platform.pio_mask= [HW,LIBATA] 4571 Format: <int> 4572 Supported PIO mode mask. Set individual bits to allow 4573 the use of the respective PIO modes. Bit 0 is for 4574 mode 0, bit 1 is for mode 1, and so on. Mode 0 only 4575 allowed by default. 4576 4577 pause_on_oops=<int> 4578 Halt all CPUs after the first oops has been printed for 4579 the specified number of seconds. This is to be used if 4580 your oopses keep scrolling off the screen. 4581 4582 pcbit= [HW,ISDN] 4583 4584 pci=option[,option...] [PCI,EARLY] various PCI subsystem options. 4585 4586 Some options herein operate on a specific device 4587 or a set of devices (<pci_dev>). These are 4588 specified in one of the following formats: 4589 4590 [<domain>:]<bus>:<dev>.<func>[/<dev>.<func>]* 4591 pci:<vendor>:<device>[:<subvendor>:<subdevice>] 4592 4593 Note: the first format specifies a PCI 4594 bus/device/function address which may change 4595 if new hardware is inserted, if motherboard 4596 firmware changes, or due to changes caused 4597 by other kernel parameters. If the 4598 domain is left unspecified, it is 4599 taken to be zero. Optionally, a path 4600 to a device through multiple device/function 4601 addresses can be specified after the base 4602 address (this is more robust against 4603 renumbering issues). The second format 4604 selects devices using IDs from the 4605 configuration space which may match multiple 4606 devices in the system. 4607 4608 earlydump dump PCI config space before the kernel 4609 changes anything 4610 off [X86] don't probe for the PCI bus 4611 bios [X86-32] force use of PCI BIOS, don't access 4612 the hardware directly. Use this if your machine 4613 has a non-standard PCI host bridge. 4614 nobios [X86-32] disallow use of PCI BIOS, only direct 4615 hardware access methods are allowed. Use this 4616 if you experience crashes upon bootup and you 4617 suspect they are caused by the BIOS. 4618 conf1 [X86] Force use of PCI Configuration Access 4619 Mechanism 1 (config address in IO port 0xCF8, 4620 data in IO port 0xCFC, both 32-bit). 4621 conf2 [X86] Force use of PCI Configuration Access 4622 Mechanism 2 (IO port 0xCF8 is an 8-bit port for 4623 the function, IO port 0xCFA, also 8-bit, sets 4624 bus number. The config space is then accessed 4625 through ports 0xC000-0xCFFF). 4626 See http://wiki.osdev.org/PCI for more info 4627 on the configuration access mechanisms. 4628 noaer [PCIE] If the PCIEAER kernel config parameter is 4629 enabled, this kernel boot option can be used to 4630 disable the use of PCIE advanced error reporting. 4631 nodomains [PCI] Disable support for multiple PCI 4632 root domains (aka PCI segments, in ACPI-speak). 4633 nommconf [X86] Disable use of MMCONFIG for PCI 4634 Configuration 4635 check_enable_amd_mmconf [X86] check for and enable 4636 properly configured MMIO access to PCI 4637 config space on AMD family 10h CPU 4638 nomsi [MSI] If the PCI_MSI kernel config parameter is 4639 enabled, this kernel boot option can be used to 4640 disable the use of MSI interrupts system-wide. 4641 noioapicquirk [APIC] Disable all boot interrupt quirks. 4642 Safety option to keep boot IRQs enabled. This 4643 should never be necessary. 4644 ioapicreroute [APIC] Enable rerouting of boot IRQs to the 4645 primary IO-APIC for bridges that cannot disable 4646 boot IRQs. This fixes a source of spurious IRQs 4647 when the system masks IRQs. 4648 noioapicreroute [APIC] Disable workaround that uses the 4649 boot IRQ equivalent of an IRQ that connects to 4650 a chipset where boot IRQs cannot be disabled. 4651 The opposite of ioapicreroute. 4652 biosirq [X86-32] Use PCI BIOS calls to get the interrupt 4653 routing table. These calls are known to be buggy 4654 on several machines and they hang the machine 4655 when used, but on other computers it's the only 4656 way to get the interrupt routing table. Try 4657 this option if the kernel is unable to allocate 4658 IRQs or discover secondary PCI buses on your 4659 motherboard. 4660 rom [X86] Assign address space to expansion ROMs. 4661 Use with caution as certain devices share 4662 address decoders between ROMs and other 4663 resources. 4664 norom [X86] Do not assign address space to 4665 expansion ROMs that do not already have 4666 BIOS assigned address ranges. 4667 nobar [X86] Do not assign address space to the 4668 BARs that weren't assigned by the BIOS. 4669 irqmask=0xMMMM [X86] Set a bit mask of IRQs allowed to be 4670 assigned automatically to PCI devices. You can 4671 make the kernel exclude IRQs of your ISA cards 4672 this way. 4673 pirqaddr=0xAAAAA [X86] Specify the physical address 4674 of the PIRQ table (normally generated 4675 by the BIOS) if it is outside the 4676 F0000h-100000h range. 4677 lastbus=N [X86] Scan all buses thru bus #N. Can be 4678 useful if the kernel is unable to find your 4679 secondary buses and you want to tell it 4680 explicitly which ones they are. 4681 assign-busses [X86] Always assign all PCI bus 4682 numbers ourselves, overriding 4683 whatever the firmware may have done. 4684 usepirqmask [X86] Honor the possible IRQ mask stored 4685 in the BIOS $PIR table. This is needed on 4686 some systems with broken BIOSes, notably 4687 some HP Pavilion N5400 and Omnibook XE3 4688 notebooks. This will have no effect if ACPI 4689 IRQ routing is enabled. 4690 noacpi [X86] Do not use ACPI for IRQ routing 4691 or for PCI scanning. 4692 use_crs [X86] Use PCI host bridge window information 4693 from ACPI. On BIOSes from 2008 or later, this 4694 is enabled by default. If you need to use this, 4695 please report a bug. 4696 nocrs [X86] Ignore PCI host bridge windows from ACPI. 4697 If you need to use this, please report a bug. 4698 use_e820 [X86] Use E820 reservations to exclude parts of 4699 PCI host bridge windows. This is a workaround 4700 for BIOS defects in host bridge _CRS methods. 4701 If you need to use this, please report a bug to 4702 <linux-pci@vger.kernel.org>. 4703 no_e820 [X86] Ignore E820 reservations for PCI host 4704 bridge windows. This is the default on modern 4705 hardware. If you need to use this, please report 4706 a bug to <linux-pci@vger.kernel.org>. 4707 routeirq Do IRQ routing for all PCI devices. 4708 This is normally done in pci_enable_device(), 4709 so this option is a temporary workaround 4710 for broken drivers that don't call it. 4711 skip_isa_align [X86] do not align io start addr, so can 4712 handle more pci cards 4713 noearly [X86] Don't do any early type 1 scanning. 4714 This might help on some broken boards which 4715 machine check when some devices' config space 4716 is read. But various workarounds are disabled 4717 and some IOMMU drivers will not work. 4718 bfsort Sort PCI devices into breadth-first order. 4719 This sorting is done to get a device 4720 order compatible with older (<= 2.4) kernels. 4721 nobfsort Don't sort PCI devices into breadth-first order. 4722 pcie_bus_tune_off Disable PCIe MPS (Max Payload Size) 4723 tuning and use the BIOS-configured MPS defaults. 4724 pcie_bus_safe Set every device's MPS to the largest value 4725 supported by all devices below the root complex. 4726 pcie_bus_perf Set device MPS to the largest allowable MPS 4727 based on its parent bus. Also set MRRS (Max 4728 Read Request Size) to the largest supported 4729 value (no larger than the MPS that the device 4730 or bus can support) for best performance. 4731 pcie_bus_peer2peer Set every device's MPS to 128B, which 4732 every device is guaranteed to support. This 4733 configuration allows peer-to-peer DMA between 4734 any pair of devices, possibly at the cost of 4735 reduced performance. This also guarantees 4736 that hot-added devices will work. 4737 cbiosize=nn[KMG] The fixed amount of bus space which is 4738 reserved for the CardBus bridge's IO window. 4739 The default value is 256 bytes. 4740 cbmemsize=nn[KMG] The fixed amount of bus space which is 4741 reserved for the CardBus bridge's memory 4742 window. The default value is 64 megabytes. 4743 resource_alignment= 4744 Format: 4745 [<order of align>@]<pci_dev>[; ...] 4746 Specifies alignment and device to reassign 4747 aligned memory resources. How to 4748 specify the device is described above. 4749 If <order of align> is not specified, 4750 PAGE_SIZE is used as alignment. 4751 A PCI-PCI bridge can be specified if resource 4752 windows need to be expanded. 4753 To specify the alignment for several 4754 instances of a device, the PCI vendor, 4755 device, subvendor, and subdevice may be 4756 specified, e.g., 12@pci:8086:9c22:103c:198f 4757 for 4096-byte alignment. 4758 ecrc= Enable/disable PCIe ECRC (transaction layer 4759 end-to-end CRC checking). Only effective if 4760 OS has native AER control (either granted by 4761 ACPI _OSC or forced via "pcie_ports=native") 4762 bios: Use BIOS/firmware settings. This is the 4763 the default. 4764 off: Turn ECRC off 4765 on: Turn ECRC on. 4766 hpiosize=nn[KMG] The fixed amount of bus space which is 4767 reserved for hotplug bridge's IO window. 4768 Default size is 256 bytes. 4769 hpmmiosize=nn[KMG] The fixed amount of bus space which is 4770 reserved for hotplug bridge's MMIO window. 4771 Default size is 2 megabytes. 4772 hpmmioprefsize=nn[KMG] The fixed amount of bus space which is 4773 reserved for hotplug bridge's MMIO_PREF window. 4774 Default size is 2 megabytes. 4775 hpmemsize=nn[KMG] The fixed amount of bus space which is 4776 reserved for hotplug bridge's MMIO and 4777 MMIO_PREF window. 4778 Default size is 2 megabytes. 4779 hpbussize=nn The minimum amount of additional bus numbers 4780 reserved for buses below a hotplug bridge. 4781 Default is 1. 4782 realloc= Enable/disable reallocating PCI bridge resources 4783 if allocations done by BIOS are too small to 4784 accommodate resources required by all child 4785 devices. 4786 off: Turn realloc off 4787 on: Turn realloc on 4788 realloc same as realloc=on 4789 noari do not use PCIe ARI. 4790 noats [PCIE, Intel-IOMMU, AMD-IOMMU] 4791 do not use PCIe ATS (and IOMMU device IOTLB). 4792 pcie_scan_all Scan all possible PCIe devices. Otherwise we 4793 only look for one device below a PCIe downstream 4794 port. 4795 big_root_window Try to add a big 64bit memory window to the PCIe 4796 root complex on AMD CPUs. Some GFX hardware 4797 can resize a BAR to allow access to all VRAM. 4798 Adding the window is slightly risky (it may 4799 conflict with unreported devices), so this 4800 taints the kernel. 4801 disable_acs_redir=<pci_dev>[; ...] 4802 Specify one or more PCI devices (in the format 4803 specified above) separated by semicolons. 4804 Each device specified will have the PCI ACS 4805 redirect capabilities forced off which will 4806 allow P2P traffic between devices through 4807 bridges without forcing it upstream. Note: 4808 this removes isolation between devices and 4809 may put more devices in an IOMMU group. 4810 config_acs= 4811 Format: 4812 <ACS flags>@<pci_dev>[; ...] 4813 Specify one or more PCI devices (in the format 4814 specified above) optionally prepended with flags 4815 and separated by semicolons. The respective 4816 capabilities will be enabled, disabled or 4817 unchanged based on what is specified in 4818 flags. 4819 4820 ACS Flags is defined as follows: 4821 bit-0 : ACS Source Validation 4822 bit-1 : ACS Translation Blocking 4823 bit-2 : ACS P2P Request Redirect 4824 bit-3 : ACS P2P Completion Redirect 4825 bit-4 : ACS Upstream Forwarding 4826 bit-5 : ACS P2P Egress Control 4827 bit-6 : ACS Direct Translated P2P 4828 Each bit can be marked as: 4829 '0' – force disabled 4830 '1' – force enabled 4831 'x' – unchanged 4832 For example, 4833 pci=config_acs=10x 4834 would configure all devices that support 4835 ACS to enable P2P Request Redirect, disable 4836 Translation Blocking, and leave Source 4837 Validation unchanged from whatever power-up 4838 or firmware set it to. 4839 4840 Note: this may remove isolation between devices 4841 and may put more devices in an IOMMU group. 4842 force_floating [S390] Force usage of floating interrupts. 4843 nomio [S390] Do not use MIO instructions. 4844 norid [S390] ignore the RID field and force use of 4845 one PCI domain per PCI function 4846 notph [PCIE] If the PCIE_TPH kernel config parameter 4847 is enabled, this kernel boot option can be used 4848 to disable PCIe TLP Processing Hints support 4849 system-wide. 4850 4851 pcie_aspm= [PCIE] Forcibly enable or ignore PCIe Active State Power 4852 Management. 4853 off Don't touch ASPM configuration at all. Leave any 4854 configuration done by firmware unchanged. 4855 force Enable ASPM even on devices that claim not to support it. 4856 WARNING: Forcing ASPM on may cause system lockups. 4857 4858 pcie_ports= [PCIE] PCIe port services handling: 4859 native Use native PCIe services (PME, AER, DPC, PCIe hotplug) 4860 even if the platform doesn't give the OS permission to 4861 use them. This may cause conflicts if the platform 4862 also tries to use these services. 4863 dpc-native Use native PCIe service for DPC only. May 4864 cause conflicts if firmware uses AER or DPC. 4865 compat Disable native PCIe services (PME, AER, DPC, PCIe 4866 hotplug). 4867 4868 pcie_port_pm= [PCIE] PCIe port power management handling: 4869 off Disable power management of all PCIe ports 4870 force Forcibly enable power management of all PCIe ports 4871 4872 pcie_pme= [PCIE,PM] Native PCIe PME signaling options: 4873 nomsi Do not use MSI for native PCIe PME signaling (this makes 4874 all PCIe root ports use INTx for all services). 4875 4876 pcmv= [HW,PCMCIA] BadgePAD 4 4877 4878 pd_ignore_unused 4879 [PM] 4880 Keep all power-domains already enabled by bootloader on, 4881 even if no driver has claimed them. This is useful 4882 for debug and development, but should not be 4883 needed on a platform with proper driver support. 4884 4885 pdcchassis= [PARISC,HW] Disable/Enable PDC Chassis Status codes at 4886 boot time. 4887 Format: { 0 | 1 } 4888 See arch/parisc/kernel/pdc_chassis.c 4889 4890 percpu_alloc= [MM,EARLY] 4891 Select which percpu first chunk allocator to use. 4892 Currently supported values are "embed" and "page". 4893 Archs may support subset or none of the selections. 4894 See comments in mm/percpu.c for details on each 4895 allocator. This parameter is primarily for debugging 4896 and performance comparison. 4897 4898 pirq= [SMP,APIC] Manual mp-table setup 4899 See Documentation/arch/x86/i386/IO-APIC.rst. 4900 4901 plip= [PPT,NET] Parallel port network link 4902 Format: { parport<nr> | timid | 0 } 4903 See also Documentation/admin-guide/parport.rst. 4904 4905 pmtmr= [X86] Manual setup of pmtmr I/O Port. 4906 Override pmtimer IOPort with a hex value. 4907 e.g. pmtmr=0x508 4908 4909 pmu_override= [PPC] Override the PMU. 4910 This option takes over the PMU facility, so it is no 4911 longer usable by perf. Setting this option starts the 4912 PMU counters by setting MMCR0 to 0 (the FC bit is 4913 cleared). If a number is given, then MMCR1 is set to 4914 that number, otherwise (e.g., 'pmu_override=on'), MMCR1 4915 remains 0. 4916 4917 pm_debug_messages [SUSPEND,KNL] 4918 Enable suspend/resume debug messages during boot up. 4919 4920 pnp.debug=1 [PNP] 4921 Enable PNP debug messages (depends on the 4922 CONFIG_PNP_DEBUG_MESSAGES option). Change at run-time 4923 via /sys/module/pnp/parameters/debug. We always show 4924 current resource usage; turning this on also shows 4925 possible settings and some assignment information. 4926 4927 pnpacpi= [ACPI] 4928 { off } 4929 4930 pnpbios= [ISAPNP] 4931 { on | off | curr | res | no-curr | no-res } 4932 4933 pnp_reserve_irq= 4934 [ISAPNP] Exclude IRQs for the autoconfiguration 4935 4936 pnp_reserve_dma= 4937 [ISAPNP] Exclude DMAs for the autoconfiguration 4938 4939 pnp_reserve_io= [ISAPNP] Exclude I/O ports for the autoconfiguration 4940 Ranges are in pairs (I/O port base and size). 4941 4942 pnp_reserve_mem= 4943 [ISAPNP] Exclude memory regions for the 4944 autoconfiguration. 4945 Ranges are in pairs (memory base and size). 4946 4947 ports= [IP_VS_FTP] IPVS ftp helper module 4948 Default is 21. 4949 Up to 8 (IP_VS_APP_MAX_PORTS) ports 4950 may be specified. 4951 Format: <port>,<port>.... 4952 4953 possible_cpus= [SMP,S390,X86] 4954 Format: <unsigned int> 4955 Set the number of possible CPUs, overriding the 4956 regular discovery mechanisms (such as ACPI/FW, etc). 4957 4958 powersave=off [PPC] This option disables power saving features. 4959 It specifically disables cpuidle and sets the 4960 platform machine description specific power_save 4961 function to NULL. On Idle the CPU just reduces 4962 execution priority. 4963 4964 ppc_strict_facility_enable 4965 [PPC,ENABLE] This option catches any kernel floating point, 4966 Altivec, VSX and SPE outside of regions specifically 4967 allowed (eg kernel_enable_fpu()/kernel_disable_fpu()). 4968 There is some performance impact when enabling this. 4969 4970 ppc_tm= [PPC,EARLY] 4971 Format: {"off"} 4972 Disable Hardware Transactional Memory 4973 4974 preempt= [KNL] 4975 Select preemption mode if you have CONFIG_PREEMPT_DYNAMIC 4976 none - Limited to cond_resched() calls 4977 voluntary - Limited to cond_resched() and might_sleep() calls 4978 full - Any section that isn't explicitly preempt disabled 4979 can be preempted anytime. Tasks will also yield 4980 contended spinlocks (if the critical section isn't 4981 explicitly preempt disabled beyond the lock itself). 4982 lazy - Scheduler controlled. Similar to full but instead 4983 of preempting the task immediately, the task gets 4984 one HZ tick time to yield itself before the 4985 preemption will be forced. One preemption is when the 4986 task returns to user space. 4987 4988 print-fatal-signals= 4989 [KNL] debug: print fatal signals 4990 4991 If enabled, warn about various signal handling 4992 related application anomalies: too many signals, 4993 too many POSIX.1 timers, fatal signals causing a 4994 coredump - etc. 4995 4996 If you hit the warning due to signal overflow, 4997 you might want to try "ulimit -i unlimited". 4998 4999 default: off. 5000 5001 printk.always_kmsg_dump= 5002 Trigger kmsg_dump for cases other than kernel oops or 5003 panics 5004 Format: <bool> (1/Y/y=enable, 0/N/n=disable) 5005 default: disabled 5006 5007 printk.console_no_auto_verbose= 5008 Disable console loglevel raise on oops, panic 5009 or lockdep-detected issues (only if lock debug is on). 5010 With an exception to setups with low baudrate on 5011 serial console, keeping this 0 is a good choice 5012 in order to provide more debug information. 5013 Format: <bool> 5014 default: 0 (auto_verbose is enabled) 5015 5016 printk.devkmsg={on,off,ratelimit} 5017 Control writing to /dev/kmsg. 5018 on - unlimited logging to /dev/kmsg from userspace 5019 off - logging to /dev/kmsg disabled 5020 ratelimit - ratelimit the logging 5021 Default: ratelimit 5022 5023 printk.time= Show timing data prefixed to each printk message line 5024 Format: <bool> (1/Y/y=enable, 0/N/n=disable) 5025 5026 proc_mem.force_override= [KNL] 5027 Format: {always | ptrace | never} 5028 Traditionally /proc/pid/mem allows memory permissions to be 5029 overridden without restrictions. This option may be set to 5030 restrict that. Can be one of: 5031 - 'always': traditional behavior always allows mem overrides. 5032 - 'ptrace': only allow mem overrides for active ptracers. 5033 - 'never': never allow mem overrides. 5034 If not specified, default is the CONFIG_PROC_MEM_* choice. 5035 5036 processor.max_cstate= [HW,ACPI] 5037 Limit processor to maximum C-state 5038 max_cstate=9 overrides any DMI blacklist limit. 5039 5040 processor.nocst [HW,ACPI] 5041 Ignore the _CST method to determine C-states, 5042 instead using the legacy FADT method 5043 5044 profile= [KNL] Enable kernel profiling via /proc/profile 5045 Format: [<profiletype>,]<number> 5046 Param: <profiletype>: "schedule" or "kvm" 5047 [defaults to kernel profiling] 5048 Param: "schedule" - profile schedule points. 5049 Param: "kvm" - profile VM exits. 5050 Param: <number> - step/bucket size as a power of 2 for 5051 statistical time based profiling. 5052 5053 prompt_ramdisk= [RAM] [Deprecated] 5054 5055 prot_virt= [S390] enable hosting protected virtual machines 5056 isolated from the hypervisor (if hardware supports 5057 that). If enabled, the default kernel base address 5058 might be overridden even when Kernel Address Space 5059 Layout Randomization is disabled. 5060 Format: <bool> 5061 5062 psi= [KNL] Enable or disable pressure stall information 5063 tracking. 5064 Format: <bool> 5065 5066 psmouse.proto= [HW,MOUSE] Highest PS2 mouse protocol extension to 5067 probe for; one of (bare|imps|exps|lifebook|any). 5068 psmouse.rate= [HW,MOUSE] Set desired mouse report rate, in reports 5069 per second. 5070 psmouse.resetafter= [HW,MOUSE] 5071 Try to reset the device after so many bad packets 5072 (0 = never). 5073 psmouse.resolution= 5074 [HW,MOUSE] Set desired mouse resolution, in dpi. 5075 psmouse.smartscroll= 5076 [HW,MOUSE] Controls Logitech smartscroll autorepeat. 5077 0 = disabled, 1 = enabled (default). 5078 5079 pstore.backend= Specify the name of the pstore backend to use 5080 5081 pti= [X86-64] Control Page Table Isolation of user and 5082 kernel address spaces. Disabling this feature 5083 removes hardening, but improves performance of 5084 system calls and interrupts. 5085 5086 on - unconditionally enable 5087 off - unconditionally disable 5088 auto - kernel detects whether your CPU model is 5089 vulnerable to issues that PTI mitigates 5090 5091 Not specifying this option is equivalent to pti=auto. 5092 5093 pty.legacy_count= 5094 [KNL] Number of legacy pty's. Overwrites compiled-in 5095 default number. 5096 5097 quiet [KNL,EARLY] Disable most log messages 5098 5099 r128= [HW,DRM] 5100 5101 radix_hcall_invalidate=on [PPC/PSERIES] 5102 Disable RADIX GTSE feature and use hcall for TLB 5103 invalidate. 5104 5105 raid= [HW,RAID] 5106 See Documentation/admin-guide/md.rst. 5107 5108 ramdisk_size= [RAM] Sizes of RAM disks in kilobytes 5109 See Documentation/admin-guide/blockdev/ramdisk.rst. 5110 5111 ramdisk_start= [RAM] RAM disk image start address 5112 5113 random.trust_cpu=off 5114 [KNL,EARLY] Disable trusting the use of the CPU's 5115 random number generator (if available) to 5116 initialize the kernel's RNG. 5117 5118 random.trust_bootloader=off 5119 [KNL,EARLY] Disable trusting the use of the a seed 5120 passed by the bootloader (if available) to 5121 initialize the kernel's RNG. 5122 5123 randomize_kstack_offset= 5124 [KNL,EARLY] Enable or disable kernel stack offset 5125 randomization, which provides roughly 5 bits of 5126 entropy, frustrating memory corruption attacks 5127 that depend on stack address determinism or 5128 cross-syscall address exposures. This is only 5129 available on architectures that have defined 5130 CONFIG_HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET. 5131 Format: <bool> (1/Y/y=enable, 0/N/n=disable) 5132 Default is CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT. 5133 5134 ras=option[,option,...] [KNL] RAS-specific options 5135 5136 cec_disable [X86] 5137 Disable the Correctable Errors Collector, 5138 see CONFIG_RAS_CEC help text. 5139 5140 rcu_nocbs[=cpu-list] 5141 [KNL] The optional argument is a cpu list, 5142 as described above. 5143 5144 In kernels built with CONFIG_RCU_NOCB_CPU=y, 5145 enable the no-callback CPU mode, which prevents 5146 such CPUs' callbacks from being invoked in 5147 softirq context. Invocation of such CPUs' RCU 5148 callbacks will instead be offloaded to "rcuox/N" 5149 kthreads created for that purpose, where "x" is 5150 "p" for RCU-preempt, "s" for RCU-sched, and "g" 5151 for the kthreads that mediate grace periods; and 5152 "N" is the CPU number. This reduces OS jitter on 5153 the offloaded CPUs, which can be useful for HPC 5154 and real-time workloads. It can also improve 5155 energy efficiency for asymmetric multiprocessors. 5156 5157 If a cpulist is passed as an argument, the specified 5158 list of CPUs is set to no-callback mode from boot. 5159 5160 Otherwise, if the '=' sign and the cpulist 5161 arguments are omitted, no CPU will be set to 5162 no-callback mode from boot but the mode may be 5163 toggled at runtime via cpusets. 5164 5165 Note that this argument takes precedence over 5166 the CONFIG_RCU_NOCB_CPU_DEFAULT_ALL option. 5167 5168 rcu_nocb_poll [KNL] 5169 Rather than requiring that offloaded CPUs 5170 (specified by rcu_nocbs= above) explicitly 5171 awaken the corresponding "rcuoN" kthreads, 5172 make these kthreads poll for callbacks. 5173 This improves the real-time response for the 5174 offloaded CPUs by relieving them of the need to 5175 wake up the corresponding kthread, but degrades 5176 energy efficiency by requiring that the kthreads 5177 periodically wake up to do the polling. 5178 5179 rcutree.blimit= [KNL] 5180 Set maximum number of finished RCU callbacks to 5181 process in one batch. 5182 5183 rcutree.csd_lock_suppress_rcu_stall= [KNL] 5184 Do only a one-line RCU CPU stall warning when 5185 there is an ongoing too-long CSD-lock wait. 5186 5187 rcutree.do_rcu_barrier= [KNL] 5188 Request a call to rcu_barrier(). This is 5189 throttled so that userspace tests can safely 5190 hammer on the sysfs variable if they so choose. 5191 If triggered before the RCU grace-period machinery 5192 is fully active, this will error out with EAGAIN. 5193 5194 rcutree.dump_tree= [KNL] 5195 Dump the structure of the rcu_node combining tree 5196 out at early boot. This is used for diagnostic 5197 purposes, to verify correct tree setup. 5198 5199 rcutree.gp_cleanup_delay= [KNL] 5200 Set the number of jiffies to delay each step of 5201 RCU grace-period cleanup. 5202 5203 rcutree.gp_init_delay= [KNL] 5204 Set the number of jiffies to delay each step of 5205 RCU grace-period initialization. 5206 5207 rcutree.gp_preinit_delay= [KNL] 5208 Set the number of jiffies to delay each step of 5209 RCU grace-period pre-initialization, that is, 5210 the propagation of recent CPU-hotplug changes up 5211 the rcu_node combining tree. 5212 5213 rcutree.jiffies_till_first_fqs= [KNL] 5214 Set delay from grace-period initialization to 5215 first attempt to force quiescent states. 5216 Units are jiffies, minimum value is zero, 5217 and maximum value is HZ. 5218 5219 rcutree.jiffies_till_next_fqs= [KNL] 5220 Set delay between subsequent attempts to force 5221 quiescent states. Units are jiffies, minimum 5222 value is one, and maximum value is HZ. 5223 5224 rcutree.jiffies_till_sched_qs= [KNL] 5225 Set required age in jiffies for a 5226 given grace period before RCU starts 5227 soliciting quiescent-state help from 5228 rcu_note_context_switch() and cond_resched(). 5229 If not specified, the kernel will calculate 5230 a value based on the most recent settings 5231 of rcutree.jiffies_till_first_fqs 5232 and rcutree.jiffies_till_next_fqs. 5233 This calculated value may be viewed in 5234 rcutree.jiffies_to_sched_qs. Any attempt to set 5235 rcutree.jiffies_to_sched_qs will be cheerfully 5236 overwritten. 5237 5238 rcutree.kthread_prio= [KNL,BOOT] 5239 Set the SCHED_FIFO priority of the RCU per-CPU 5240 kthreads (rcuc/N). This value is also used for 5241 the priority of the RCU boost threads (rcub/N) 5242 and for the RCU grace-period kthreads (rcu_bh, 5243 rcu_preempt, and rcu_sched). If RCU_BOOST is 5244 set, valid values are 1-99 and the default is 1 5245 (the least-favored priority). Otherwise, when 5246 RCU_BOOST is not set, valid values are 0-99 and 5247 the default is zero (non-realtime operation). 5248 When RCU_NOCB_CPU is set, also adjust the 5249 priority of NOCB callback kthreads. 5250 5251 rcutree.nocb_nobypass_lim_per_jiffy= [KNL] 5252 On callback-offloaded (rcu_nocbs) CPUs, 5253 RCU reduces the lock contention that would 5254 otherwise be caused by callback floods through 5255 use of the ->nocb_bypass list. However, in the 5256 common non-flooded case, RCU queues directly to 5257 the main ->cblist in order to avoid the extra 5258 overhead of the ->nocb_bypass list and its lock. 5259 But if there are too many callbacks queued during 5260 a single jiffy, RCU pre-queues the callbacks into 5261 the ->nocb_bypass queue. The definition of "too 5262 many" is supplied by this kernel boot parameter. 5263 5264 rcutree.nohz_full_patience_delay= [KNL] 5265 On callback-offloaded (rcu_nocbs) CPUs, avoid 5266 disturbing RCU unless the grace period has 5267 reached the specified age in milliseconds. 5268 Defaults to zero. Large values will be capped 5269 at five seconds. All values will be rounded down 5270 to the nearest value representable by jiffies. 5271 5272 rcutree.qhimark= [KNL] 5273 Set threshold of queued RCU callbacks beyond which 5274 batch limiting is disabled. 5275 5276 rcutree.qlowmark= [KNL] 5277 Set threshold of queued RCU callbacks below which 5278 batch limiting is re-enabled. 5279 5280 rcutree.qovld= [KNL] 5281 Set threshold of queued RCU callbacks beyond which 5282 RCU's force-quiescent-state scan will aggressively 5283 enlist help from cond_resched() and sched IPIs to 5284 help CPUs more quickly reach quiescent states. 5285 Set to less than zero to make this be set based 5286 on rcutree.qhimark at boot time and to zero to 5287 disable more aggressive help enlistment. 5288 5289 rcutree.rcu_delay_page_cache_fill_msec= [KNL] 5290 Set the page-cache refill delay (in milliseconds) 5291 in response to low-memory conditions. The range 5292 of permitted values is in the range 0:100000. 5293 5294 rcutree.rcu_divisor= [KNL] 5295 Set the shift-right count to use to compute 5296 the callback-invocation batch limit bl from 5297 the number of callbacks queued on this CPU. 5298 The result will be bounded below by the value of 5299 the rcutree.blimit kernel parameter. Every bl 5300 callbacks, the softirq handler will exit in 5301 order to allow the CPU to do other work. 5302 5303 Please note that this callback-invocation batch 5304 limit applies only to non-offloaded callback 5305 invocation. Offloaded callbacks are instead 5306 invoked in the context of an rcuoc kthread, which 5307 scheduler will preempt as it does any other task. 5308 5309 rcutree.rcu_fanout_exact= [KNL] 5310 Disable autobalancing of the rcu_node combining 5311 tree. This is used by rcutorture, and might 5312 possibly be useful for architectures having high 5313 cache-to-cache transfer latencies. 5314 5315 rcutree.rcu_fanout_leaf= [KNL] 5316 Change the number of CPUs assigned to each 5317 leaf rcu_node structure. Useful for very 5318 large systems, which will choose the value 64, 5319 and for NUMA systems with large remote-access 5320 latencies, which will choose a value aligned 5321 with the appropriate hardware boundaries. 5322 5323 rcutree.rcu_min_cached_objs= [KNL] 5324 Minimum number of objects which are cached and 5325 maintained per one CPU. Object size is equal 5326 to PAGE_SIZE. The cache allows to reduce the 5327 pressure to page allocator, also it makes the 5328 whole algorithm to behave better in low memory 5329 condition. 5330 5331 rcutree.rcu_nocb_gp_stride= [KNL] 5332 Set the number of NOCB callback kthreads in 5333 each group, which defaults to the square root 5334 of the number of CPUs. Larger numbers reduce 5335 the wakeup overhead on the global grace-period 5336 kthread, but increases that same overhead on 5337 each group's NOCB grace-period kthread. 5338 5339 rcutree.rcu_kick_kthreads= [KNL] 5340 Cause the grace-period kthread to get an extra 5341 wake_up() if it sleeps three times longer than 5342 it should at force-quiescent-state time. 5343 This wake_up() will be accompanied by a 5344 WARN_ONCE() splat and an ftrace_dump(). 5345 5346 rcutree.rcu_resched_ns= [KNL] 5347 Limit the time spend invoking a batch of RCU 5348 callbacks to the specified number of nanoseconds. 5349 By default, this limit is checked only once 5350 every 32 callbacks in order to limit the pain 5351 inflicted by local_clock() overhead. 5352 5353 rcutree.rcu_unlock_delay= [KNL] 5354 In CONFIG_RCU_STRICT_GRACE_PERIOD=y kernels, 5355 this specifies an rcu_read_unlock()-time delay 5356 in microseconds. This defaults to zero. 5357 Larger delays increase the probability of 5358 catching RCU pointer leaks, that is, buggy use 5359 of RCU-protected pointers after the relevant 5360 rcu_read_unlock() has completed. 5361 5362 rcutree.sysrq_rcu= [KNL] 5363 Commandeer a sysrq key to dump out Tree RCU's 5364 rcu_node tree with an eye towards determining 5365 why a new grace period has not yet started. 5366 5367 rcutree.use_softirq= [KNL] 5368 If set to zero, move all RCU_SOFTIRQ processing to 5369 per-CPU rcuc kthreads. Defaults to a non-zero 5370 value, meaning that RCU_SOFTIRQ is used by default. 5371 Specify rcutree.use_softirq=0 to use rcuc kthreads. 5372 5373 But note that CONFIG_PREEMPT_RT=y kernels disable 5374 this kernel boot parameter, forcibly setting it 5375 to zero. 5376 5377 rcutree.enable_rcu_lazy= [KNL] 5378 To save power, batch RCU callbacks and flush after 5379 delay, memory pressure or callback list growing too 5380 big. 5381 5382 rcutree.rcu_normal_wake_from_gp= [KNL] 5383 Reduces a latency of synchronize_rcu() call. This approach 5384 maintains its own track of synchronize_rcu() callers, so it 5385 does not interact with regular callbacks because it does not 5386 use a call_rcu[_hurry]() path. Please note, this is for a 5387 normal grace period. 5388 5389 How to enable it: 5390 5391 echo 1 > /sys/module/rcutree/parameters/rcu_normal_wake_from_gp 5392 or pass a boot parameter "rcutree.rcu_normal_wake_from_gp=1" 5393 5394 Default is 0. 5395 5396 rcuscale.gp_async= [KNL] 5397 Measure performance of asynchronous 5398 grace-period primitives such as call_rcu(). 5399 5400 rcuscale.gp_async_max= [KNL] 5401 Specify the maximum number of outstanding 5402 callbacks per writer thread. When a writer 5403 thread exceeds this limit, it invokes the 5404 corresponding flavor of rcu_barrier() to allow 5405 previously posted callbacks to drain. 5406 5407 rcuscale.gp_exp= [KNL] 5408 Measure performance of expedited synchronous 5409 grace-period primitives. 5410 5411 rcuscale.holdoff= [KNL] 5412 Set test-start holdoff period. The purpose of 5413 this parameter is to delay the start of the 5414 test until boot completes in order to avoid 5415 interference. 5416 5417 rcuscale.kfree_by_call_rcu= [KNL] 5418 In kernels built with CONFIG_RCU_LAZY=y, test 5419 call_rcu() instead of kfree_rcu(). 5420 5421 rcuscale.kfree_mult= [KNL] 5422 Instead of allocating an object of size kfree_obj, 5423 allocate one of kfree_mult * sizeof(kfree_obj). 5424 Defaults to 1. 5425 5426 rcuscale.kfree_rcu_test= [KNL] 5427 Set to measure performance of kfree_rcu() flooding. 5428 5429 rcuscale.kfree_rcu_test_double= [KNL] 5430 Test the double-argument variant of kfree_rcu(). 5431 If this parameter has the same value as 5432 rcuscale.kfree_rcu_test_single, both the single- 5433 and double-argument variants are tested. 5434 5435 rcuscale.kfree_rcu_test_single= [KNL] 5436 Test the single-argument variant of kfree_rcu(). 5437 If this parameter has the same value as 5438 rcuscale.kfree_rcu_test_double, both the single- 5439 and double-argument variants are tested. 5440 5441 rcuscale.kfree_nthreads= [KNL] 5442 The number of threads running loops of kfree_rcu(). 5443 5444 rcuscale.kfree_alloc_num= [KNL] 5445 Number of allocations and frees done in an iteration. 5446 5447 rcuscale.kfree_loops= [KNL] 5448 Number of loops doing rcuscale.kfree_alloc_num number 5449 of allocations and frees. 5450 5451 rcuscale.minruntime= [KNL] 5452 Set the minimum test run time in seconds. This 5453 does not affect the data-collection interval, 5454 but instead allows better measurement of things 5455 like CPU consumption. 5456 5457 rcuscale.nreaders= [KNL] 5458 Set number of RCU readers. The value -1 selects 5459 N, where N is the number of CPUs. A value 5460 "n" less than -1 selects N-n+1, where N is again 5461 the number of CPUs. For example, -2 selects N 5462 (the number of CPUs), -3 selects N+1, and so on. 5463 A value of "n" less than or equal to -N selects 5464 a single reader. 5465 5466 rcuscale.nwriters= [KNL] 5467 Set number of RCU writers. The values operate 5468 the same as for rcuscale.nreaders. 5469 N, where N is the number of CPUs 5470 5471 rcuscale.scale_type= [KNL] 5472 Specify the RCU implementation to test. 5473 5474 rcuscale.shutdown= [KNL] 5475 Shut the system down after performance tests 5476 complete. This is useful for hands-off automated 5477 testing. 5478 5479 rcuscale.verbose= [KNL] 5480 Enable additional printk() statements. 5481 5482 rcuscale.writer_holdoff= [KNL] 5483 Write-side holdoff between grace periods, 5484 in microseconds. The default of zero says 5485 no holdoff. 5486 5487 rcuscale.writer_holdoff_jiffies= [KNL] 5488 Additional write-side holdoff between grace 5489 periods, but in jiffies. The default of zero 5490 says no holdoff. 5491 5492 rcutorture.fqs_duration= [KNL] 5493 Set duration of force_quiescent_state bursts 5494 in microseconds. 5495 5496 rcutorture.fqs_holdoff= [KNL] 5497 Set holdoff time within force_quiescent_state bursts 5498 in microseconds. 5499 5500 rcutorture.fqs_stutter= [KNL] 5501 Set wait time between force_quiescent_state bursts 5502 in seconds. 5503 5504 rcutorture.fwd_progress= [KNL] 5505 Specifies the number of kthreads to be used 5506 for RCU grace-period forward-progress testing 5507 for the types of RCU supporting this notion. 5508 Defaults to 1 kthread, values less than zero or 5509 greater than the number of CPUs cause the number 5510 of CPUs to be used. 5511 5512 rcutorture.fwd_progress_div= [KNL] 5513 Specify the fraction of a CPU-stall-warning 5514 period to do tight-loop forward-progress testing. 5515 5516 rcutorture.fwd_progress_holdoff= [KNL] 5517 Number of seconds to wait between successive 5518 forward-progress tests. 5519 5520 rcutorture.fwd_progress_need_resched= [KNL] 5521 Enclose cond_resched() calls within checks for 5522 need_resched() during tight-loop forward-progress 5523 testing. 5524 5525 rcutorture.gp_cond= [KNL] 5526 Use conditional/asynchronous update-side 5527 normal-grace-period primitives, if available. 5528 5529 rcutorture.gp_cond_exp= [KNL] 5530 Use conditional/asynchronous update-side 5531 expedited-grace-period primitives, if available. 5532 5533 rcutorture.gp_cond_full= [KNL] 5534 Use conditional/asynchronous update-side 5535 normal-grace-period primitives that also take 5536 concurrent expedited grace periods into account, 5537 if available. 5538 5539 rcutorture.gp_cond_exp_full= [KNL] 5540 Use conditional/asynchronous update-side 5541 expedited-grace-period primitives that also take 5542 concurrent normal grace periods into account, 5543 if available. 5544 5545 rcutorture.gp_cond_wi= [KNL] 5546 Nominal wait interval for normal conditional 5547 grace periods (specified by rcutorture's 5548 gp_cond and gp_cond_full module parameters), 5549 in microseconds. The actual wait interval will 5550 be randomly selected to nanosecond granularity up 5551 to this wait interval. Defaults to 16 jiffies, 5552 for example, 16,000 microseconds on a system 5553 with HZ=1000. 5554 5555 rcutorture.gp_cond_wi_exp= [KNL] 5556 Nominal wait interval for expedited conditional 5557 grace periods (specified by rcutorture's 5558 gp_cond_exp and gp_cond_exp_full module 5559 parameters), in microseconds. The actual wait 5560 interval will be randomly selected to nanosecond 5561 granularity up to this wait interval. Defaults to 5562 128 microseconds. 5563 5564 rcutorture.gp_exp= [KNL] 5565 Use expedited update-side primitives, if available. 5566 5567 rcutorture.gp_normal= [KNL] 5568 Use normal (non-expedited) asynchronous 5569 update-side primitives, if available. 5570 5571 rcutorture.gp_poll= [KNL] 5572 Use polled update-side normal-grace-period 5573 primitives, if available. 5574 5575 rcutorture.gp_poll_exp= [KNL] 5576 Use polled update-side expedited-grace-period 5577 primitives, if available. 5578 5579 rcutorture.gp_poll_full= [KNL] 5580 Use polled update-side normal-grace-period 5581 primitives that also take concurrent expedited 5582 grace periods into account, if available. 5583 5584 rcutorture.gp_poll_exp_full= [KNL] 5585 Use polled update-side expedited-grace-period 5586 primitives that also take concurrent normal 5587 grace periods into account, if available. 5588 5589 rcutorture.gp_poll_wi= [KNL] 5590 Nominal wait interval for normal conditional 5591 grace periods (specified by rcutorture's 5592 gp_poll and gp_poll_full module parameters), 5593 in microseconds. The actual wait interval will 5594 be randomly selected to nanosecond granularity up 5595 to this wait interval. Defaults to 16 jiffies, 5596 for example, 16,000 microseconds on a system 5597 with HZ=1000. 5598 5599 rcutorture.gp_poll_wi_exp= [KNL] 5600 Nominal wait interval for expedited conditional 5601 grace periods (specified by rcutorture's 5602 gp_poll_exp and gp_poll_exp_full module 5603 parameters), in microseconds. The actual wait 5604 interval will be randomly selected to nanosecond 5605 granularity up to this wait interval. Defaults to 5606 128 microseconds. 5607 5608 rcutorture.gp_sync= [KNL] 5609 Use normal (non-expedited) synchronous 5610 update-side primitives, if available. If all 5611 of rcutorture.gp_cond=, rcutorture.gp_exp=, 5612 rcutorture.gp_normal=, and rcutorture.gp_sync= 5613 are zero, rcutorture acts as if is interpreted 5614 they are all non-zero. 5615 5616 rcutorture.irqreader= [KNL] 5617 Run RCU readers from irq handlers, or, more 5618 accurately, from a timer handler. Not all RCU 5619 flavors take kindly to this sort of thing. 5620 5621 rcutorture.leakpointer= [KNL] 5622 Leak an RCU-protected pointer out of the reader. 5623 This can of course result in splats, and is 5624 intended to test the ability of things like 5625 CONFIG_RCU_STRICT_GRACE_PERIOD=y to detect 5626 such leaks. 5627 5628 rcutorture.n_barrier_cbs= [KNL] 5629 Set callbacks/threads for rcu_barrier() testing. 5630 5631 rcutorture.nfakewriters= [KNL] 5632 Set number of concurrent RCU writers. These just 5633 stress RCU, they don't participate in the actual 5634 test, hence the "fake". 5635 5636 rcutorture.nocbs_nthreads= [KNL] 5637 Set number of RCU callback-offload togglers. 5638 Zero (the default) disables toggling. 5639 5640 rcutorture.nocbs_toggle= [KNL] 5641 Set the delay in milliseconds between successive 5642 callback-offload toggling attempts. 5643 5644 rcutorture.nreaders= [KNL] 5645 Set number of RCU readers. The value -1 selects 5646 N-1, where N is the number of CPUs. A value 5647 "n" less than -1 selects N-n-2, where N is again 5648 the number of CPUs. For example, -2 selects N 5649 (the number of CPUs), -3 selects N+1, and so on. 5650 5651 rcutorture.object_debug= [KNL] 5652 Enable debug-object double-call_rcu() testing. 5653 5654 rcutorture.onoff_holdoff= [KNL] 5655 Set time (s) after boot for CPU-hotplug testing. 5656 5657 rcutorture.onoff_interval= [KNL] 5658 Set time (jiffies) between CPU-hotplug operations, 5659 or zero to disable CPU-hotplug testing. 5660 5661 rcutorture.preempt_duration= [KNL] 5662 Set duration (in milliseconds) of preemptions 5663 by a high-priority FIFO real-time task. Set to 5664 zero (the default) to disable. The CPUs to 5665 preempt are selected randomly from the set that 5666 are online at a given point in time. Races with 5667 CPUs going offline are ignored, with that attempt 5668 at preemption skipped. 5669 5670 rcutorture.preempt_interval= [KNL] 5671 Set interval (in milliseconds, defaulting to one 5672 second) between preemptions by a high-priority 5673 FIFO real-time task. This delay is mediated 5674 by an hrtimer and is further fuzzed to avoid 5675 inadvertent synchronizations. 5676 5677 rcutorture.read_exit_burst= [KNL] 5678 The number of times in a given read-then-exit 5679 episode that a set of read-then-exit kthreads 5680 is spawned. 5681 5682 rcutorture.read_exit_delay= [KNL] 5683 The delay, in seconds, between successive 5684 read-then-exit testing episodes. 5685 5686 rcutorture.reader_flavor= [KNL] 5687 A bit mask indicating which readers to use. 5688 If there is more than one bit set, the readers 5689 are entered from low-order bit up, and are 5690 exited in the opposite order. For SRCU, the 5691 0x1 bit is normal readers, 0x2 NMI-safe readers, 5692 and 0x4 light-weight readers. 5693 5694 rcutorture.shuffle_interval= [KNL] 5695 Set task-shuffle interval (s). Shuffling tasks 5696 allows some CPUs to go into dyntick-idle mode 5697 during the rcutorture test. 5698 5699 rcutorture.shutdown_secs= [KNL] 5700 Set time (s) after boot system shutdown. This 5701 is useful for hands-off automated testing. 5702 5703 rcutorture.stall_cpu= [KNL] 5704 Duration of CPU stall (s) to test RCU CPU stall 5705 warnings, zero to disable. 5706 5707 rcutorture.stall_cpu_block= [KNL] 5708 Sleep while stalling if set. This will result 5709 in warnings from preemptible RCU in addition to 5710 any other stall-related activity. Note that 5711 in kernels built with CONFIG_PREEMPTION=n and 5712 CONFIG_PREEMPT_COUNT=y, this parameter will 5713 cause the CPU to pass through a quiescent state. 5714 Given CONFIG_PREEMPTION=n, this will suppress 5715 RCU CPU stall warnings, but will instead result 5716 in scheduling-while-atomic splats. 5717 5718 Use of this module parameter results in splats. 5719 5720 5721 rcutorture.stall_cpu_holdoff= [KNL] 5722 Time to wait (s) after boot before inducing stall. 5723 5724 rcutorture.stall_cpu_irqsoff= [KNL] 5725 Disable interrupts while stalling if set, but only 5726 on the first stall in the set. 5727 5728 rcutorture.stall_cpu_repeat= [KNL] 5729 Number of times to repeat the stall sequence, 5730 so that rcutorture.stall_cpu_repeat=3 will result 5731 in four stall sequences. 5732 5733 rcutorture.stall_gp_kthread= [KNL] 5734 Duration (s) of forced sleep within RCU 5735 grace-period kthread to test RCU CPU stall 5736 warnings, zero to disable. If both stall_cpu 5737 and stall_gp_kthread are specified, the 5738 kthread is starved first, then the CPU. 5739 5740 rcutorture.stat_interval= [KNL] 5741 Time (s) between statistics printk()s. 5742 5743 rcutorture.stutter= [KNL] 5744 Time (s) to stutter testing, for example, specifying 5745 five seconds causes the test to run for five seconds, 5746 wait for five seconds, and so on. This tests RCU's 5747 ability to transition abruptly to and from idle. 5748 5749 rcutorture.test_boost= [KNL] 5750 Test RCU priority boosting? 0=no, 1=maybe, 2=yes. 5751 "Maybe" means test if the RCU implementation 5752 under test support RCU priority boosting. 5753 5754 rcutorture.test_boost_duration= [KNL] 5755 Duration (s) of each individual boost test. 5756 5757 rcutorture.test_boost_interval= [KNL] 5758 Interval (s) between each boost test. 5759 5760 rcutorture.test_no_idle_hz= [KNL] 5761 Test RCU's dyntick-idle handling. See also the 5762 rcutorture.shuffle_interval parameter. 5763 5764 rcutorture.torture_type= [KNL] 5765 Specify the RCU implementation to test. 5766 5767 rcutorture.verbose= [KNL] 5768 Enable additional printk() statements. 5769 5770 rcupdate.rcu_cpu_stall_ftrace_dump= [KNL] 5771 Dump ftrace buffer after reporting RCU CPU 5772 stall warning. 5773 5774 rcupdate.rcu_cpu_stall_notifiers= [KNL] 5775 Provide RCU CPU stall notifiers, but see the 5776 warnings in the RCU_CPU_STALL_NOTIFIER Kconfig 5777 option's help text. TL;DR: You almost certainly 5778 do not want rcupdate.rcu_cpu_stall_notifiers. 5779 5780 rcupdate.rcu_cpu_stall_suppress= [KNL] 5781 Suppress RCU CPU stall warning messages. 5782 5783 rcupdate.rcu_cpu_stall_suppress_at_boot= [KNL] 5784 Suppress RCU CPU stall warning messages and 5785 rcutorture writer stall warnings that occur 5786 during early boot, that is, during the time 5787 before the init task is spawned. 5788 5789 rcupdate.rcu_cpu_stall_timeout= [KNL] 5790 Set timeout for RCU CPU stall warning messages. 5791 The value is in seconds and the maximum allowed 5792 value is 300 seconds. 5793 5794 rcupdate.rcu_exp_cpu_stall_timeout= [KNL] 5795 Set timeout for expedited RCU CPU stall warning 5796 messages. The value is in milliseconds 5797 and the maximum allowed value is 21000 5798 milliseconds. Please note that this value is 5799 adjusted to an arch timer tick resolution. 5800 Setting this to zero causes the value from 5801 rcupdate.rcu_cpu_stall_timeout to be used (after 5802 conversion from seconds to milliseconds). 5803 5804 rcupdate.rcu_cpu_stall_cputime= [KNL] 5805 Provide statistics on the cputime and count of 5806 interrupts and tasks during the sampling period. For 5807 multiple continuous RCU stalls, all sampling periods 5808 begin at half of the first RCU stall timeout. 5809 5810 rcupdate.rcu_exp_stall_task_details= [KNL] 5811 Print stack dumps of any tasks blocking the 5812 current expedited RCU grace period during an 5813 expedited RCU CPU stall warning. 5814 5815 rcupdate.rcu_expedited= [KNL] 5816 Use expedited grace-period primitives, for 5817 example, synchronize_rcu_expedited() instead 5818 of synchronize_rcu(). This reduces latency, 5819 but can increase CPU utilization, degrade 5820 real-time latency, and degrade energy efficiency. 5821 No effect on CONFIG_TINY_RCU kernels. 5822 5823 rcupdate.rcu_normal= [KNL] 5824 Use only normal grace-period primitives, 5825 for example, synchronize_rcu() instead of 5826 synchronize_rcu_expedited(). This improves 5827 real-time latency, CPU utilization, and 5828 energy efficiency, but can expose users to 5829 increased grace-period latency. This parameter 5830 overrides rcupdate.rcu_expedited. No effect on 5831 CONFIG_TINY_RCU kernels. 5832 5833 rcupdate.rcu_normal_after_boot= [KNL] 5834 Once boot has completed (that is, after 5835 rcu_end_inkernel_boot() has been invoked), use 5836 only normal grace-period primitives. No effect 5837 on CONFIG_TINY_RCU kernels. 5838 5839 But note that CONFIG_PREEMPT_RT=y kernels enables 5840 this kernel boot parameter, forcibly setting 5841 it to the value one, that is, converting any 5842 post-boot attempt at an expedited RCU grace 5843 period to instead use normal non-expedited 5844 grace-period processing. 5845 5846 rcupdate.rcu_task_collapse_lim= [KNL] 5847 Set the maximum number of callbacks present 5848 at the beginning of a grace period that allows 5849 the RCU Tasks flavors to collapse back to using 5850 a single callback queue. This switching only 5851 occurs when rcupdate.rcu_task_enqueue_lim is 5852 set to the default value of -1. 5853 5854 rcupdate.rcu_task_contend_lim= [KNL] 5855 Set the minimum number of callback-queuing-time 5856 lock-contention events per jiffy required to 5857 cause the RCU Tasks flavors to switch to per-CPU 5858 callback queuing. This switching only occurs 5859 when rcupdate.rcu_task_enqueue_lim is set to 5860 the default value of -1. 5861 5862 rcupdate.rcu_task_enqueue_lim= [KNL] 5863 Set the number of callback queues to use for the 5864 RCU Tasks family of RCU flavors. The default 5865 of -1 allows this to be automatically (and 5866 dynamically) adjusted. This parameter is intended 5867 for use in testing. 5868 5869 rcupdate.rcu_task_ipi_delay= [KNL] 5870 Set time in jiffies during which RCU tasks will 5871 avoid sending IPIs, starting with the beginning 5872 of a given grace period. Setting a large 5873 number avoids disturbing real-time workloads, 5874 but lengthens grace periods. 5875 5876 rcupdate.rcu_task_lazy_lim= [KNL] 5877 Number of callbacks on a given CPU that will 5878 cancel laziness on that CPU. Use -1 to disable 5879 cancellation of laziness, but be advised that 5880 doing so increases the danger of OOM due to 5881 callback flooding. 5882 5883 rcupdate.rcu_task_stall_info= [KNL] 5884 Set initial timeout in jiffies for RCU task stall 5885 informational messages, which give some indication 5886 of the problem for those not patient enough to 5887 wait for ten minutes. Informational messages are 5888 only printed prior to the stall-warning message 5889 for a given grace period. Disable with a value 5890 less than or equal to zero. Defaults to ten 5891 seconds. A change in value does not take effect 5892 until the beginning of the next grace period. 5893 5894 rcupdate.rcu_task_stall_info_mult= [KNL] 5895 Multiplier for time interval between successive 5896 RCU task stall informational messages for a given 5897 RCU tasks grace period. This value is clamped 5898 to one through ten, inclusive. It defaults to 5899 the value three, so that the first informational 5900 message is printed 10 seconds into the grace 5901 period, the second at 40 seconds, the third at 5902 160 seconds, and then the stall warning at 600 5903 seconds would prevent a fourth at 640 seconds. 5904 5905 rcupdate.rcu_task_stall_timeout= [KNL] 5906 Set timeout in jiffies for RCU task stall 5907 warning messages. Disable with a value less 5908 than or equal to zero. Defaults to ten minutes. 5909 A change in value does not take effect until 5910 the beginning of the next grace period. 5911 5912 rcupdate.rcu_tasks_lazy_ms= [KNL] 5913 Set timeout in milliseconds RCU Tasks asynchronous 5914 callback batching for call_rcu_tasks(). 5915 A negative value will take the default. A value 5916 of zero will disable batching. Batching is 5917 always disabled for synchronize_rcu_tasks(). 5918 5919 rcupdate.rcu_tasks_trace_lazy_ms= [KNL] 5920 Set timeout in milliseconds RCU Tasks 5921 Trace asynchronous callback batching for 5922 call_rcu_tasks_trace(). A negative value 5923 will take the default. A value of zero will 5924 disable batching. Batching is always disabled 5925 for synchronize_rcu_tasks_trace(). 5926 5927 rcupdate.rcu_self_test= [KNL] 5928 Run the RCU early boot self tests 5929 5930 rdinit= [KNL] 5931 Format: <full_path> 5932 Run specified binary instead of /init from the ramdisk, 5933 used for early userspace startup. See initrd. 5934 5935 rdrand= [X86,EARLY] 5936 force - Override the decision by the kernel to hide the 5937 advertisement of RDRAND support (this affects 5938 certain AMD processors because of buggy BIOS 5939 support, specifically around the suspend/resume 5940 path). 5941 5942 rdt= [HW,X86,RDT] 5943 Turn on/off individual RDT features. List is: 5944 cmt, mbmtotal, mbmlocal, l3cat, l3cdp, l2cat, l2cdp, 5945 mba, smba, bmec. 5946 E.g. to turn on cmt and turn off mba use: 5947 rdt=cmt,!mba 5948 5949 reboot= [KNL] 5950 Format (x86 or x86_64): 5951 [w[arm] | c[old] | h[ard] | s[oft] | g[pio]] | d[efault] \ 5952 [[,]s[mp]#### \ 5953 [[,]b[ios] | a[cpi] | k[bd] | t[riple] | e[fi] | p[ci]] \ 5954 [[,]f[orce] 5955 Where reboot_mode is one of warm (soft) or cold (hard) or gpio 5956 (prefix with 'panic_' to set mode for panic 5957 reboot only), 5958 reboot_type is one of bios, acpi, kbd, triple, efi, or pci, 5959 reboot_force is either force or not specified, 5960 reboot_cpu is s[mp]#### with #### being the processor 5961 to be used for rebooting. 5962 5963 acpi 5964 Use the ACPI RESET_REG in the FADT. If ACPI is not 5965 configured or the ACPI reset does not work, the reboot 5966 path attempts the reset using the keyboard controller. 5967 5968 bios 5969 Use the CPU reboot vector for warm reset 5970 5971 cold 5972 Set the cold reboot flag 5973 5974 default 5975 There are some built-in platform specific "quirks" 5976 - you may see: "reboot: <name> series board detected. 5977 Selecting <type> for reboots." In the case where you 5978 think the quirk is in error (e.g. you have newer BIOS, 5979 or newer board) using this option will ignore the 5980 built-in quirk table, and use the generic default 5981 reboot actions. 5982 5983 efi 5984 Use efi reset_system runtime service. If EFI is not 5985 configured or the EFI reset does not work, the reboot 5986 path attempts the reset using the keyboard controller. 5987 5988 force 5989 Don't stop other CPUs on reboot. This can make reboot 5990 more reliable in some cases. 5991 5992 kbd 5993 Use the keyboard controller. cold reset (default) 5994 5995 pci 5996 Use a write to the PCI config space register 0xcf9 to 5997 trigger reboot. 5998 5999 triple 6000 Force a triple fault (init) 6001 6002 warm 6003 Don't set the cold reboot flag 6004 6005 Using warm reset will be much faster especially on big 6006 memory systems because the BIOS will not go through 6007 the memory check. Disadvantage is that not all 6008 hardware will be completely reinitialized on reboot so 6009 there may be boot problems on some systems. 6010 6011 6012 refscale.holdoff= [KNL] 6013 Set test-start holdoff period. The purpose of 6014 this parameter is to delay the start of the 6015 test until boot completes in order to avoid 6016 interference. 6017 6018 refscale.lookup_instances= [KNL] 6019 Number of data elements to use for the forms of 6020 SLAB_TYPESAFE_BY_RCU testing. A negative number 6021 is negated and multiplied by nr_cpu_ids, while 6022 zero specifies nr_cpu_ids. 6023 6024 refscale.loops= [KNL] 6025 Set the number of loops over the synchronization 6026 primitive under test. Increasing this number 6027 reduces noise due to loop start/end overhead, 6028 but the default has already reduced the per-pass 6029 noise to a handful of picoseconds on ca. 2020 6030 x86 laptops. 6031 6032 refscale.nreaders= [KNL] 6033 Set number of readers. The default value of -1 6034 selects N, where N is roughly 75% of the number 6035 of CPUs. A value of zero is an interesting choice. 6036 6037 refscale.nruns= [KNL] 6038 Set number of runs, each of which is dumped onto 6039 the console log. 6040 6041 refscale.readdelay= [KNL] 6042 Set the read-side critical-section duration, 6043 measured in microseconds. 6044 6045 refscale.scale_type= [KNL] 6046 Specify the read-protection implementation to test. 6047 6048 refscale.shutdown= [KNL] 6049 Shut down the system at the end of the performance 6050 test. This defaults to 1 (shut it down) when 6051 refscale is built into the kernel and to 0 (leave 6052 it running) when refscale is built as a module. 6053 6054 refscale.verbose= [KNL] 6055 Enable additional printk() statements. 6056 6057 refscale.verbose_batched= [KNL] 6058 Batch the additional printk() statements. If zero 6059 (the default) or negative, print everything. Otherwise, 6060 print every Nth verbose statement, where N is the value 6061 specified. 6062 6063 regulator_ignore_unused 6064 [REGULATOR] 6065 Prevents regulator framework from disabling regulators 6066 that are unused, due no driver claiming them. This may 6067 be useful for debug and development, but should not be 6068 needed on a platform with proper driver support. 6069 6070 relax_domain_level= 6071 [KNL, SMP] Set scheduler's default relax_domain_level. 6072 See Documentation/admin-guide/cgroup-v1/cpusets.rst. 6073 6074 reserve= [KNL,BUGS] Force kernel to ignore I/O ports or memory 6075 Format: <base1>,<size1>[,<base2>,<size2>,...] 6076 Reserve I/O ports or memory so the kernel won't use 6077 them. If <base> is less than 0x10000, the region 6078 is assumed to be I/O ports; otherwise it is memory. 6079 6080 reserve_mem= [RAM] 6081 Format: nn[KNG]:<align>:<label> 6082 Reserve physical memory and label it with a name that 6083 other subsystems can use to access it. This is typically 6084 used for systems that do not wipe the RAM, and this command 6085 line will try to reserve the same physical memory on 6086 soft reboots. Note, it is not guaranteed to be the same 6087 location. For example, if anything about the system changes 6088 or if booting a different kernel. It can also fail if KASLR 6089 places the kernel at the location of where the RAM reservation 6090 was from a previous boot, the new reservation will be at a 6091 different location. 6092 Any subsystem using this feature must add a way to verify 6093 that the contents of the physical memory is from a previous 6094 boot, as there may be cases where the memory will not be 6095 located at the same location. 6096 6097 The format is size:align:label for example, to request 6098 12 megabytes of 4096 alignment for ramoops: 6099 6100 reserve_mem=12M:4096:oops ramoops.mem_name=oops 6101 6102 reservetop= [X86-32,EARLY] 6103 Format: nn[KMG] 6104 Reserves a hole at the top of the kernel virtual 6105 address space. 6106 6107 reset_devices [KNL] Force drivers to reset the underlying device 6108 during initialization. 6109 6110 resume= [SWSUSP] 6111 Specify the partition device for software suspend 6112 Format: 6113 {/dev/<dev> | PARTUUID=<uuid> | <int>:<int> | <hex>} 6114 6115 resume_offset= [SWSUSP] 6116 Specify the offset from the beginning of the partition 6117 given by "resume=" at which the swap header is located, 6118 in <PAGE_SIZE> units (needed only for swap files). 6119 See Documentation/power/swsusp-and-swap-files.rst 6120 6121 resumedelay= [HIBERNATION] Delay (in seconds) to pause before attempting to 6122 read the resume files 6123 6124 resumewait [HIBERNATION] Wait (indefinitely) for resume device to show up. 6125 Useful for devices that are detected asynchronously 6126 (e.g. USB and MMC devices). 6127 6128 retain_initrd [RAM] Keep initrd memory after extraction. After boot, it will 6129 be accessible via /sys/firmware/initrd. 6130 6131 retbleed= [X86] Control mitigation of RETBleed (Arbitrary 6132 Speculative Code Execution with Return Instructions) 6133 vulnerability. 6134 6135 AMD-based UNRET and IBPB mitigations alone do not stop 6136 sibling threads from influencing the predictions of other 6137 sibling threads. For that reason, STIBP is used on pro- 6138 cessors that support it, and mitigate SMT on processors 6139 that don't. 6140 6141 off - no mitigation 6142 auto - automatically select a migitation 6143 auto,nosmt - automatically select a mitigation, 6144 disabling SMT if necessary for 6145 the full mitigation (only on Zen1 6146 and older without STIBP). 6147 ibpb - On AMD, mitigate short speculation 6148 windows on basic block boundaries too. 6149 Safe, highest perf impact. It also 6150 enables STIBP if present. Not suitable 6151 on Intel. 6152 ibpb,nosmt - Like "ibpb" above but will disable SMT 6153 when STIBP is not available. This is 6154 the alternative for systems which do not 6155 have STIBP. 6156 unret - Force enable untrained return thunks, 6157 only effective on AMD f15h-f17h based 6158 systems. 6159 unret,nosmt - Like unret, but will disable SMT when STIBP 6160 is not available. This is the alternative for 6161 systems which do not have STIBP. 6162 6163 Selecting 'auto' will choose a mitigation method at run 6164 time according to the CPU. 6165 6166 Not specifying this option is equivalent to retbleed=auto. 6167 6168 rfkill.default_state= 6169 0 "airplane mode". All wifi, bluetooth, wimax, gps, fm, 6170 etc. communication is blocked by default. 6171 1 Unblocked. 6172 6173 rfkill.master_switch_mode= 6174 0 The "airplane mode" button does nothing. 6175 1 The "airplane mode" button toggles between everything 6176 blocked and the previous configuration. 6177 2 The "airplane mode" button toggles between everything 6178 blocked and everything unblocked. 6179 6180 ring3mwait=disable 6181 [KNL] Disable ring 3 MONITOR/MWAIT feature on supported 6182 CPUs. 6183 6184 riscv_isa_fallback [RISCV,EARLY] 6185 When CONFIG_RISCV_ISA_FALLBACK is not enabled, permit 6186 falling back to detecting extension support by parsing 6187 "riscv,isa" property on devicetree systems when the 6188 replacement properties are not found. See the Kconfig 6189 entry for RISCV_ISA_FALLBACK. 6190 6191 ro [KNL] Mount root device read-only on boot 6192 6193 rodata= [KNL,EARLY] 6194 on Mark read-only kernel memory as read-only (default). 6195 off Leave read-only kernel memory writable for debugging. 6196 full Mark read-only kernel memory and aliases as read-only 6197 [arm64] 6198 6199 rockchip.usb_uart 6200 [EARLY] 6201 Enable the uart passthrough on the designated usb port 6202 on Rockchip SoCs. When active, the signals of the 6203 debug-uart get routed to the D+ and D- pins of the usb 6204 port and the regular usb controller gets disabled. 6205 6206 root= [KNL] Root filesystem 6207 Usually this a a block device specifier of some kind, 6208 see the early_lookup_bdev comment in 6209 block/early-lookup.c for details. 6210 Alternatively this can be "ram" for the legacy initial 6211 ramdisk, "nfs" and "cifs" for root on a network file 6212 system, or "mtd" and "ubi" for mounting from raw flash. 6213 6214 rootdelay= [KNL] Delay (in seconds) to pause before attempting to 6215 mount the root filesystem 6216 6217 rootflags= [KNL] Set root filesystem mount option string 6218 6219 rootfstype= [KNL] Set root filesystem type 6220 6221 rootwait [KNL] Wait (indefinitely) for root device to show up. 6222 Useful for devices that are detected asynchronously 6223 (e.g. USB and MMC devices). 6224 6225 rootwait= [KNL] Maximum time (in seconds) to wait for root device 6226 to show up before attempting to mount the root 6227 filesystem. 6228 6229 rproc_mem=nn[KMG][@address] 6230 [KNL,ARM,CMA] Remoteproc physical memory block. 6231 Memory area to be used by remote processor image, 6232 managed by CMA. 6233 6234 rw [KNL] Mount root device read-write on boot 6235 6236 S [KNL] Run init in single mode 6237 6238 s390_iommu= [HW,S390] 6239 Set s390 IOTLB flushing mode 6240 strict 6241 With strict flushing every unmap operation will result 6242 in an IOTLB flush. Default is lazy flushing before 6243 reuse, which is faster. Deprecated, equivalent to 6244 iommu.strict=1. 6245 6246 s390_iommu_aperture= [KNL,S390] 6247 Specifies the size of the per device DMA address space 6248 accessible through the DMA and IOMMU APIs as a decimal 6249 factor of the size of main memory. 6250 The default is 1 meaning that one can concurrently use 6251 as many DMA addresses as physical memory is installed, 6252 if supported by hardware, and thus map all of memory 6253 once. With a value of 2 one can map all of memory twice 6254 and so on. As a special case a factor of 0 imposes no 6255 restrictions other than those given by hardware at the 6256 cost of significant additional memory use for tables. 6257 6258 sa1100ir [NET] 6259 See drivers/net/irda/sa1100_ir.c. 6260 6261 sched_verbose [KNL,EARLY] Enables verbose scheduler debug messages. 6262 6263 schedstats= [KNL,X86] Enable or disable scheduled statistics. 6264 Allowed values are enable and disable. This feature 6265 incurs a small amount of overhead in the scheduler 6266 but is useful for debugging and performance tuning. 6267 6268 sched_thermal_decay_shift= 6269 [Deprecated] 6270 [KNL, SMP] Set a decay shift for scheduler thermal 6271 pressure signal. Thermal pressure signal follows the 6272 default decay period of other scheduler pelt 6273 signals(usually 32 ms but configurable). Setting 6274 sched_thermal_decay_shift will left shift the decay 6275 period for the thermal pressure signal by the shift 6276 value. 6277 i.e. with the default pelt decay period of 32 ms 6278 sched_thermal_decay_shift thermal pressure decay pr 6279 1 64 ms 6280 2 128 ms 6281 and so on. 6282 Format: integer between 0 and 10 6283 Default is 0. 6284 6285 scftorture.holdoff= [KNL] 6286 Number of seconds to hold off before starting 6287 test. Defaults to zero for module insertion and 6288 to 10 seconds for built-in smp_call_function() 6289 tests. 6290 6291 scftorture.longwait= [KNL] 6292 Request ridiculously long waits randomly selected 6293 up to the chosen limit in seconds. Zero (the 6294 default) disables this feature. Please note 6295 that requesting even small non-zero numbers of 6296 seconds can result in RCU CPU stall warnings, 6297 softlockup complaints, and so on. 6298 6299 scftorture.nthreads= [KNL] 6300 Number of kthreads to spawn to invoke the 6301 smp_call_function() family of functions. 6302 The default of -1 specifies a number of kthreads 6303 equal to the number of CPUs. 6304 6305 scftorture.onoff_holdoff= [KNL] 6306 Number seconds to wait after the start of the 6307 test before initiating CPU-hotplug operations. 6308 6309 scftorture.onoff_interval= [KNL] 6310 Number seconds to wait between successive 6311 CPU-hotplug operations. Specifying zero (which 6312 is the default) disables CPU-hotplug operations. 6313 6314 scftorture.shutdown_secs= [KNL] 6315 The number of seconds following the start of the 6316 test after which to shut down the system. The 6317 default of zero avoids shutting down the system. 6318 Non-zero values are useful for automated tests. 6319 6320 scftorture.stat_interval= [KNL] 6321 The number of seconds between outputting the 6322 current test statistics to the console. A value 6323 of zero disables statistics output. 6324 6325 scftorture.stutter_cpus= [KNL] 6326 The number of jiffies to wait between each change 6327 to the set of CPUs under test. 6328 6329 scftorture.use_cpus_read_lock= [KNL] 6330 Use use_cpus_read_lock() instead of the default 6331 preempt_disable() to disable CPU hotplug 6332 while invoking one of the smp_call_function*() 6333 functions. 6334 6335 scftorture.verbose= [KNL] 6336 Enable additional printk() statements. 6337 6338 scftorture.weight_single= [KNL] 6339 The probability weighting to use for the 6340 smp_call_function_single() function with a zero 6341 "wait" parameter. A value of -1 selects the 6342 default if all other weights are -1. However, 6343 if at least one weight has some other value, a 6344 value of -1 will instead select a weight of zero. 6345 6346 scftorture.weight_single_wait= [KNL] 6347 The probability weighting to use for the 6348 smp_call_function_single() function with a 6349 non-zero "wait" parameter. See weight_single. 6350 6351 scftorture.weight_many= [KNL] 6352 The probability weighting to use for the 6353 smp_call_function_many() function with a zero 6354 "wait" parameter. See weight_single. 6355 Note well that setting a high probability for 6356 this weighting can place serious IPI load 6357 on the system. 6358 6359 scftorture.weight_many_wait= [KNL] 6360 The probability weighting to use for the 6361 smp_call_function_many() function with a 6362 non-zero "wait" parameter. See weight_single 6363 and weight_many. 6364 6365 scftorture.weight_all= [KNL] 6366 The probability weighting to use for the 6367 smp_call_function_all() function with a zero 6368 "wait" parameter. See weight_single and 6369 weight_many. 6370 6371 scftorture.weight_all_wait= [KNL] 6372 The probability weighting to use for the 6373 smp_call_function_all() function with a 6374 non-zero "wait" parameter. See weight_single 6375 and weight_many. 6376 6377 sdw_mclk_divider=[SDW] 6378 Specify the MCLK divider for Intel SoundWire buses in 6379 case the BIOS does not provide the clock rate properly. 6380 6381 skew_tick= [KNL,EARLY] Offset the periodic timer tick per cpu to mitigate 6382 xtime_lock contention on larger systems, and/or RCU lock 6383 contention on all systems with CONFIG_MAXSMP set. 6384 Format: { "0" | "1" } 6385 0 -- disable. (may be 1 via CONFIG_CMDLINE="skew_tick=1" 6386 1 -- enable. 6387 Note: increases power consumption, thus should only be 6388 enabled if running jitter sensitive (HPC/RT) workloads. 6389 6390 security= [SECURITY] Choose a legacy "major" security module to 6391 enable at boot. This has been deprecated by the 6392 "lsm=" parameter. 6393 6394 selinux= [SELINUX] Disable or enable SELinux at boot time. 6395 Format: { "0" | "1" } 6396 See security/selinux/Kconfig help text. 6397 0 -- disable. 6398 1 -- enable. 6399 Default value is 1. 6400 6401 serialnumber [BUGS=X86-32] 6402 6403 sev=option[,option...] [X86-64] 6404 6405 debug 6406 Enable debug messages. 6407 6408 nosnp 6409 Do not enable SEV-SNP (applies to host/hypervisor 6410 only). Setting 'nosnp' avoids the RMP check overhead 6411 in memory accesses when users do not want to run 6412 SEV-SNP guests. 6413 6414 shapers= [NET] 6415 Maximal number of shapers. 6416 6417 show_lapic= [APIC,X86] Advanced Programmable Interrupt Controller 6418 Limit apic dumping. The parameter defines the maximal 6419 number of local apics being dumped. Also it is possible 6420 to set it to "all" by meaning -- no limit here. 6421 Format: { 1 (default) | 2 | ... | all }. 6422 The parameter valid if only apic=debug or 6423 apic=verbose is specified. 6424 Example: apic=debug show_lapic=all 6425 6426 slab_debug[=options[,slabs][;[options[,slabs]]...] [MM] 6427 Enabling slab_debug allows one to determine the 6428 culprit if slab objects become corrupted. Enabling 6429 slab_debug can create guard zones around objects and 6430 may poison objects when not in use. Also tracks the 6431 last alloc / free. For more information see 6432 Documentation/mm/slub.rst. 6433 (slub_debug legacy name also accepted for now) 6434 6435 slab_max_order= [MM] 6436 Determines the maximum allowed order for slabs. 6437 A high setting may cause OOMs due to memory 6438 fragmentation. For more information see 6439 Documentation/mm/slub.rst. 6440 (slub_max_order legacy name also accepted for now) 6441 6442 slab_merge [MM] 6443 Enable merging of slabs with similar size when the 6444 kernel is built without CONFIG_SLAB_MERGE_DEFAULT. 6445 (slub_merge legacy name also accepted for now) 6446 6447 slab_min_objects= [MM] 6448 The minimum number of objects per slab. SLUB will 6449 increase the slab order up to slab_max_order to 6450 generate a sufficiently large slab able to contain 6451 the number of objects indicated. The higher the number 6452 of objects the smaller the overhead of tracking slabs 6453 and the less frequently locks need to be acquired. 6454 For more information see Documentation/mm/slub.rst. 6455 (slub_min_objects legacy name also accepted for now) 6456 6457 slab_min_order= [MM] 6458 Determines the minimum page order for slabs. Must be 6459 lower or equal to slab_max_order. For more information see 6460 Documentation/mm/slub.rst. 6461 (slub_min_order legacy name also accepted for now) 6462 6463 slab_nomerge [MM] 6464 Disable merging of slabs with similar size. May be 6465 necessary if there is some reason to distinguish 6466 allocs to different slabs, especially in hardened 6467 environments where the risk of heap overflows and 6468 layout control by attackers can usually be 6469 frustrated by disabling merging. This will reduce 6470 most of the exposure of a heap attack to a single 6471 cache (risks via metadata attacks are mostly 6472 unchanged). Debug options disable merging on their 6473 own. 6474 For more information see Documentation/mm/slub.rst. 6475 (slub_nomerge legacy name also accepted for now) 6476 6477 slab_strict_numa [MM] 6478 Support memory policies on a per object level 6479 in the slab allocator. The default is for memory 6480 policies to be applied at the folio level when 6481 a new folio is needed or a partial folio is 6482 retrieved from the lists. Increases overhead 6483 in the slab fastpaths but gains more accurate 6484 NUMA kernel object placement which helps with slow 6485 interconnects in NUMA systems. 6486 6487 slram= [HW,MTD] 6488 6489 smart2= [HW] 6490 Format: <io1>[,<io2>[,...,<io8>]] 6491 6492 smp.csd_lock_timeout= [KNL] 6493 Specify the period of time in milliseconds 6494 that smp_call_function() and friends will wait 6495 for a CPU to release the CSD lock. This is 6496 useful when diagnosing bugs involving CPUs 6497 disabling interrupts for extended periods 6498 of time. Defaults to 5,000 milliseconds, and 6499 setting a value of zero disables this feature. 6500 This feature may be more efficiently disabled 6501 using the csdlock_debug- kernel parameter. 6502 6503 smp.panic_on_ipistall= [KNL] 6504 If a csd_lock_timeout extends for more than 6505 the specified number of milliseconds, panic the 6506 system. By default, let CSD-lock acquisition 6507 take as long as they take. Specifying 300,000 6508 for this value provides a 5-minute timeout. 6509 6510 smsc-ircc2.nopnp [HW] Don't use PNP to discover SMC devices 6511 smsc-ircc2.ircc_cfg= [HW] Device configuration I/O port 6512 smsc-ircc2.ircc_sir= [HW] SIR base I/O port 6513 smsc-ircc2.ircc_fir= [HW] FIR base I/O port 6514 smsc-ircc2.ircc_irq= [HW] IRQ line 6515 smsc-ircc2.ircc_dma= [HW] DMA channel 6516 smsc-ircc2.ircc_transceiver= [HW] Transceiver type: 6517 0: Toshiba Satellite 1800 (GP data pin select) 6518 1: Fast pin select (default) 6519 2: ATC IRMode 6520 6521 smt= [KNL,MIPS,S390,EARLY] Set the maximum number of threads 6522 (logical CPUs) to use per physical CPU on systems 6523 capable of symmetric multithreading (SMT). Will 6524 be capped to the actual hardware limit. 6525 Format: <integer> 6526 Default: -1 (no limit) 6527 6528 softlockup_panic= 6529 [KNL] Should the soft-lockup detector generate panics. 6530 Format: 0 | 1 6531 6532 A value of 1 instructs the soft-lockup detector 6533 to panic the machine when a soft-lockup occurs. It is 6534 also controlled by the kernel.softlockup_panic sysctl 6535 and CONFIG_BOOTPARAM_SOFTLOCKUP_PANIC, which is the 6536 respective build-time switch to that functionality. 6537 6538 softlockup_all_cpu_backtrace= 6539 [KNL] Should the soft-lockup detector generate 6540 backtraces on all cpus. 6541 Format: 0 | 1 6542 6543 sonypi.*= [HW] Sony Programmable I/O Control Device driver 6544 See Documentation/admin-guide/laptops/sonypi.rst 6545 6546 spectre_bhi= [X86] Control mitigation of Branch History Injection 6547 (BHI) vulnerability. This setting affects the 6548 deployment of the HW BHI control and the SW BHB 6549 clearing sequence. 6550 6551 on - (default) Enable the HW or SW mitigation as 6552 needed. This protects the kernel from 6553 both syscalls and VMs. 6554 vmexit - On systems which don't have the HW mitigation 6555 available, enable the SW mitigation on vmexit 6556 ONLY. On such systems, the host kernel is 6557 protected from VM-originated BHI attacks, but 6558 may still be vulnerable to syscall attacks. 6559 off - Disable the mitigation. 6560 6561 spectre_v2= [X86,EARLY] Control mitigation of Spectre variant 2 6562 (indirect branch speculation) vulnerability. 6563 The default operation protects the kernel from 6564 user space attacks. 6565 6566 on - unconditionally enable, implies 6567 spectre_v2_user=on 6568 off - unconditionally disable, implies 6569 spectre_v2_user=off 6570 auto - kernel detects whether your CPU model is 6571 vulnerable 6572 6573 Selecting 'on' will, and 'auto' may, choose a 6574 mitigation method at run time according to the 6575 CPU, the available microcode, the setting of the 6576 CONFIG_MITIGATION_RETPOLINE configuration option, 6577 and the compiler with which the kernel was built. 6578 6579 Selecting 'on' will also enable the mitigation 6580 against user space to user space task attacks. 6581 6582 Selecting 'off' will disable both the kernel and 6583 the user space protections. 6584 6585 Specific mitigations can also be selected manually: 6586 6587 retpoline - replace indirect branches 6588 retpoline,generic - Retpolines 6589 retpoline,lfence - LFENCE; indirect branch 6590 retpoline,amd - alias for retpoline,lfence 6591 eibrs - Enhanced/Auto IBRS 6592 eibrs,retpoline - Enhanced/Auto IBRS + Retpolines 6593 eibrs,lfence - Enhanced/Auto IBRS + LFENCE 6594 ibrs - use IBRS to protect kernel 6595 6596 Not specifying this option is equivalent to 6597 spectre_v2=auto. 6598 6599 spectre_v2_user= 6600 [X86] Control mitigation of Spectre variant 2 6601 (indirect branch speculation) vulnerability between 6602 user space tasks 6603 6604 on - Unconditionally enable mitigations. Is 6605 enforced by spectre_v2=on 6606 6607 off - Unconditionally disable mitigations. Is 6608 enforced by spectre_v2=off 6609 6610 prctl - Indirect branch speculation is enabled, 6611 but mitigation can be enabled via prctl 6612 per thread. The mitigation control state 6613 is inherited on fork. 6614 6615 prctl,ibpb 6616 - Like "prctl" above, but only STIBP is 6617 controlled per thread. IBPB is issued 6618 always when switching between different user 6619 space processes. 6620 6621 seccomp 6622 - Same as "prctl" above, but all seccomp 6623 threads will enable the mitigation unless 6624 they explicitly opt out. 6625 6626 seccomp,ibpb 6627 - Like "seccomp" above, but only STIBP is 6628 controlled per thread. IBPB is issued 6629 always when switching between different 6630 user space processes. 6631 6632 auto - Kernel selects the mitigation depending on 6633 the available CPU features and vulnerability. 6634 6635 Default mitigation: "prctl" 6636 6637 Not specifying this option is equivalent to 6638 spectre_v2_user=auto. 6639 6640 spec_rstack_overflow= 6641 [X86,EARLY] Control RAS overflow mitigation on AMD Zen CPUs 6642 6643 off - Disable mitigation 6644 microcode - Enable microcode mitigation only 6645 safe-ret - Enable sw-only safe RET mitigation (default) 6646 ibpb - Enable mitigation by issuing IBPB on 6647 kernel entry 6648 ibpb-vmexit - Issue IBPB only on VMEXIT 6649 (cloud-specific mitigation) 6650 6651 spec_store_bypass_disable= 6652 [HW,EARLY] Control Speculative Store Bypass (SSB) Disable mitigation 6653 (Speculative Store Bypass vulnerability) 6654 6655 Certain CPUs are vulnerable to an exploit against a 6656 a common industry wide performance optimization known 6657 as "Speculative Store Bypass" in which recent stores 6658 to the same memory location may not be observed by 6659 later loads during speculative execution. The idea 6660 is that such stores are unlikely and that they can 6661 be detected prior to instruction retirement at the 6662 end of a particular speculation execution window. 6663 6664 In vulnerable processors, the speculatively forwarded 6665 store can be used in a cache side channel attack, for 6666 example to read memory to which the attacker does not 6667 directly have access (e.g. inside sandboxed code). 6668 6669 This parameter controls whether the Speculative Store 6670 Bypass optimization is used. 6671 6672 On x86 the options are: 6673 6674 on - Unconditionally disable Speculative Store Bypass 6675 off - Unconditionally enable Speculative Store Bypass 6676 auto - Kernel detects whether the CPU model contains an 6677 implementation of Speculative Store Bypass and 6678 picks the most appropriate mitigation. If the 6679 CPU is not vulnerable, "off" is selected. If the 6680 CPU is vulnerable the default mitigation is 6681 architecture and Kconfig dependent. See below. 6682 prctl - Control Speculative Store Bypass per thread 6683 via prctl. Speculative Store Bypass is enabled 6684 for a process by default. The state of the control 6685 is inherited on fork. 6686 seccomp - Same as "prctl" above, but all seccomp threads 6687 will disable SSB unless they explicitly opt out. 6688 6689 Default mitigations: 6690 X86: "prctl" 6691 6692 On powerpc the options are: 6693 6694 on,auto - On Power8 and Power9 insert a store-forwarding 6695 barrier on kernel entry and exit. On Power7 6696 perform a software flush on kernel entry and 6697 exit. 6698 off - No action. 6699 6700 Not specifying this option is equivalent to 6701 spec_store_bypass_disable=auto. 6702 6703 split_lock_detect= 6704 [X86] Enable split lock detection or bus lock detection 6705 6706 When enabled (and if hardware support is present), atomic 6707 instructions that access data across cache line 6708 boundaries will result in an alignment check exception 6709 for split lock detection or a debug exception for 6710 bus lock detection. 6711 6712 off - not enabled 6713 6714 warn - the kernel will emit rate-limited warnings 6715 about applications triggering the #AC 6716 exception or the #DB exception. This mode is 6717 the default on CPUs that support split lock 6718 detection or bus lock detection. Default 6719 behavior is by #AC if both features are 6720 enabled in hardware. 6721 6722 fatal - the kernel will send SIGBUS to applications 6723 that trigger the #AC exception or the #DB 6724 exception. Default behavior is by #AC if 6725 both features are enabled in hardware. 6726 6727 ratelimit:N - 6728 Set system wide rate limit to N bus locks 6729 per second for bus lock detection. 6730 0 < N <= 1000. 6731 6732 N/A for split lock detection. 6733 6734 6735 If an #AC exception is hit in the kernel or in 6736 firmware (i.e. not while executing in user mode) 6737 the kernel will oops in either "warn" or "fatal" 6738 mode. 6739 6740 #DB exception for bus lock is triggered only when 6741 CPL > 0. 6742 6743 srbds= [X86,INTEL,EARLY] 6744 Control the Special Register Buffer Data Sampling 6745 (SRBDS) mitigation. 6746 6747 Certain CPUs are vulnerable to an MDS-like 6748 exploit which can leak bits from the random 6749 number generator. 6750 6751 By default, this issue is mitigated by 6752 microcode. However, the microcode fix can cause 6753 the RDRAND and RDSEED instructions to become 6754 much slower. Among other effects, this will 6755 result in reduced throughput from /dev/urandom. 6756 6757 The microcode mitigation can be disabled with 6758 the following option: 6759 6760 off: Disable mitigation and remove 6761 performance impact to RDRAND and RDSEED 6762 6763 srcutree.big_cpu_lim [KNL] 6764 Specifies the number of CPUs constituting a 6765 large system, such that srcu_struct structures 6766 should immediately allocate an srcu_node array. 6767 This kernel-boot parameter defaults to 128, 6768 but takes effect only when the low-order four 6769 bits of srcutree.convert_to_big is equal to 3 6770 (decide at boot). 6771 6772 srcutree.convert_to_big [KNL] 6773 Specifies under what conditions an SRCU tree 6774 srcu_struct structure will be converted to big 6775 form, that is, with an rcu_node tree: 6776 6777 0: Never. 6778 1: At init_srcu_struct() time. 6779 2: When rcutorture decides to. 6780 3: Decide at boot time (default). 6781 0x1X: Above plus if high contention. 6782 6783 Either way, the srcu_node tree will be sized based 6784 on the actual runtime number of CPUs (nr_cpu_ids) 6785 instead of the compile-time CONFIG_NR_CPUS. 6786 6787 srcutree.counter_wrap_check [KNL] 6788 Specifies how frequently to check for 6789 grace-period sequence counter wrap for the 6790 srcu_data structure's ->srcu_gp_seq_needed field. 6791 The greater the number of bits set in this kernel 6792 parameter, the less frequently counter wrap will 6793 be checked for. Note that the bottom two bits 6794 are ignored. 6795 6796 srcutree.exp_holdoff [KNL] 6797 Specifies how many nanoseconds must elapse 6798 since the end of the last SRCU grace period for 6799 a given srcu_struct until the next normal SRCU 6800 grace period will be considered for automatic 6801 expediting. Set to zero to disable automatic 6802 expediting. 6803 6804 srcutree.srcu_max_nodelay [KNL] 6805 Specifies the number of no-delay instances 6806 per jiffy for which the SRCU grace period 6807 worker thread will be rescheduled with zero 6808 delay. Beyond this limit, worker thread will 6809 be rescheduled with a sleep delay of one jiffy. 6810 6811 srcutree.srcu_max_nodelay_phase [KNL] 6812 Specifies the per-grace-period phase, number of 6813 non-sleeping polls of readers. Beyond this limit, 6814 grace period worker thread will be rescheduled 6815 with a sleep delay of one jiffy, between each 6816 rescan of the readers, for a grace period phase. 6817 6818 srcutree.srcu_retry_check_delay [KNL] 6819 Specifies number of microseconds of non-sleeping 6820 delay between each non-sleeping poll of readers. 6821 6822 srcutree.small_contention_lim [KNL] 6823 Specifies the number of update-side contention 6824 events per jiffy will be tolerated before 6825 initiating a conversion of an srcu_struct 6826 structure to big form. Note that the value of 6827 srcutree.convert_to_big must have the 0x10 bit 6828 set for contention-based conversions to occur. 6829 6830 ssbd= [ARM64,HW,EARLY] 6831 Speculative Store Bypass Disable control 6832 6833 On CPUs that are vulnerable to the Speculative 6834 Store Bypass vulnerability and offer a 6835 firmware based mitigation, this parameter 6836 indicates how the mitigation should be used: 6837 6838 force-on: Unconditionally enable mitigation for 6839 for both kernel and userspace 6840 force-off: Unconditionally disable mitigation for 6841 for both kernel and userspace 6842 kernel: Always enable mitigation in the 6843 kernel, and offer a prctl interface 6844 to allow userspace to register its 6845 interest in being mitigated too. 6846 6847 stack_guard_gap= [MM] 6848 override the default stack gap protection. The value 6849 is in page units and it defines how many pages prior 6850 to (for stacks growing down) resp. after (for stacks 6851 growing up) the main stack are reserved for no other 6852 mapping. Default value is 256 pages. 6853 6854 stack_depot_disable= [KNL,EARLY] 6855 Setting this to true through kernel command line will 6856 disable the stack depot thereby saving the static memory 6857 consumed by the stack hash table. By default this is set 6858 to false. 6859 6860 stacktrace [FTRACE] 6861 Enabled the stack tracer on boot up. 6862 6863 stacktrace_filter=[function-list] 6864 [FTRACE] Limit the functions that the stack tracer 6865 will trace at boot up. function-list is a comma-separated 6866 list of functions. This list can be changed at run 6867 time by the stack_trace_filter file in the debugfs 6868 tracing directory. Note, this enables stack tracing 6869 and the stacktrace above is not needed. 6870 6871 sti= [PARISC,HW] 6872 Format: <num> 6873 Set the STI (builtin display/keyboard on the HP-PARISC 6874 machines) console (graphic card) which should be used 6875 as the initial boot-console. 6876 See also comment in drivers/video/console/sticore.c. 6877 6878 sti_font= [HW] 6879 See comment in drivers/video/console/sticore.c. 6880 6881 stifb= [HW] 6882 Format: bpp:<bpp1>[:<bpp2>[:<bpp3>...]] 6883 6884 strict_sas_size= 6885 [X86] 6886 Format: <bool> 6887 Enable or disable strict sigaltstack size checks 6888 against the required signal frame size which 6889 depends on the supported FPU features. This can 6890 be used to filter out binaries which have 6891 not yet been made aware of AT_MINSIGSTKSZ. 6892 6893 stress_hpt [PPC,EARLY] 6894 Limits the number of kernel HPT entries in the hash 6895 page table to increase the rate of hash page table 6896 faults on kernel addresses. 6897 6898 stress_slb [PPC,EARLY] 6899 Limits the number of kernel SLB entries, and flushes 6900 them frequently to increase the rate of SLB faults 6901 on kernel addresses. 6902 6903 sunrpc.min_resvport= 6904 sunrpc.max_resvport= 6905 [NFS,SUNRPC] 6906 SunRPC servers often require that client requests 6907 originate from a privileged port (i.e. a port in the 6908 range 0 < portnr < 1024). 6909 An administrator who wishes to reserve some of these 6910 ports for other uses may adjust the range that the 6911 kernel's sunrpc client considers to be privileged 6912 using these two parameters to set the minimum and 6913 maximum port values. 6914 6915 sunrpc.svc_rpc_per_connection_limit= 6916 [NFS,SUNRPC] 6917 Limit the number of requests that the server will 6918 process in parallel from a single connection. 6919 The default value is 0 (no limit). 6920 6921 sunrpc.pool_mode= 6922 [NFS] 6923 Control how the NFS server code allocates CPUs to 6924 service thread pools. Depending on how many NICs 6925 you have and where their interrupts are bound, this 6926 option will affect which CPUs will do NFS serving. 6927 Note: this parameter cannot be changed while the 6928 NFS server is running. 6929 6930 auto the server chooses an appropriate mode 6931 automatically using heuristics 6932 global a single global pool contains all CPUs 6933 percpu one pool for each CPU 6934 pernode one pool for each NUMA node (equivalent 6935 to global on non-NUMA machines) 6936 6937 sunrpc.tcp_slot_table_entries= 6938 sunrpc.udp_slot_table_entries= 6939 [NFS,SUNRPC] 6940 Sets the upper limit on the number of simultaneous 6941 RPC calls that can be sent from the client to a 6942 server. Increasing these values may allow you to 6943 improve throughput, but will also increase the 6944 amount of memory reserved for use by the client. 6945 6946 suspend.pm_test_delay= 6947 [SUSPEND] 6948 Sets the number of seconds to remain in a suspend test 6949 mode before resuming the system (see 6950 /sys/power/pm_test). Only available when CONFIG_PM_DEBUG 6951 is set. Default value is 5. 6952 6953 svm= [PPC] 6954 Format: { on | off | y | n | 1 | 0 } 6955 This parameter controls use of the Protected 6956 Execution Facility on pSeries. 6957 6958 swiotlb= [ARM,PPC,MIPS,X86,S390,EARLY] 6959 Format: { <int> [,<int>] | force | noforce } 6960 <int> -- Number of I/O TLB slabs 6961 <int> -- Second integer after comma. Number of swiotlb 6962 areas with their own lock. Will be rounded up 6963 to a power of 2. 6964 force -- force using of bounce buffers even if they 6965 wouldn't be automatically used by the kernel 6966 noforce -- Never use bounce buffers (for debugging) 6967 6968 switches= [HW,M68k,EARLY] 6969 6970 sysctl.*= [KNL] 6971 Set a sysctl parameter, right before loading the init 6972 process, as if the value was written to the respective 6973 /proc/sys/... file. Both '.' and '/' are recognized as 6974 separators. Unrecognized parameters and invalid values 6975 are reported in the kernel log. Sysctls registered 6976 later by a loaded module cannot be set this way. 6977 Example: sysctl.vm.swappiness=40 6978 6979 sysrq_always_enabled 6980 [KNL] 6981 Ignore sysrq setting - this boot parameter will 6982 neutralize any effect of /proc/sys/kernel/sysrq. 6983 Useful for debugging. 6984 6985 tcpmhash_entries= [KNL,NET] 6986 Set the number of tcp_metrics_hash slots. 6987 Default value is 8192 or 16384 depending on total 6988 ram pages. This is used to specify the TCP metrics 6989 cache size. See Documentation/networking/ip-sysctl.rst 6990 "tcp_no_metrics_save" section for more details. 6991 6992 tdfx= [HW,DRM] 6993 6994 test_suspend= [SUSPEND] 6995 Format: { "mem" | "standby" | "freeze" }[,N] 6996 Specify "mem" (for Suspend-to-RAM) or "standby" (for 6997 standby suspend) or "freeze" (for suspend type freeze) 6998 as the system sleep state during system startup with 6999 the optional capability to repeat N number of times. 7000 The system is woken from this state using a 7001 wakeup-capable RTC alarm. 7002 7003 thash_entries= [KNL,NET] 7004 Set number of hash buckets for TCP connection 7005 7006 thermal.act= [HW,ACPI] 7007 -1: disable all active trip points in all thermal zones 7008 <degrees C>: override all lowest active trip points 7009 7010 thermal.crt= [HW,ACPI] 7011 -1: disable all critical trip points in all thermal zones 7012 <degrees C>: override all critical trip points 7013 7014 thermal.off= [HW,ACPI] 7015 1: disable ACPI thermal control 7016 7017 thermal.psv= [HW,ACPI] 7018 -1: disable all passive trip points 7019 <degrees C>: override all passive trip points to this 7020 value 7021 7022 thermal.tzp= [HW,ACPI] 7023 Specify global default ACPI thermal zone polling rate 7024 <deci-seconds>: poll all this frequency 7025 0: no polling (default) 7026 7027 thp_anon= [KNL] 7028 Format: <size>[KMG],<size>[KMG]:<state>;<size>[KMG]-<size>[KMG]:<state> 7029 state is one of "always", "madvise", "never" or "inherit". 7030 Control the default behavior of the system with respect 7031 to anonymous transparent hugepages. 7032 Can be used multiple times for multiple anon THP sizes. 7033 See Documentation/admin-guide/mm/transhuge.rst for more 7034 details. 7035 7036 threadirqs [KNL,EARLY] 7037 Force threading of all interrupt handlers except those 7038 marked explicitly IRQF_NO_THREAD. 7039 7040 thp_shmem= [KNL] 7041 Format: <size>[KMG],<size>[KMG]:<policy>;<size>[KMG]-<size>[KMG]:<policy> 7042 Control the default policy of each hugepage size for the 7043 internal shmem mount. <policy> is one of policies available 7044 for the shmem mount ("always", "inherit", "never", "within_size", 7045 and "advise"). 7046 It can be used multiple times for multiple shmem THP sizes. 7047 See Documentation/admin-guide/mm/transhuge.rst for more 7048 details. 7049 7050 topology= [S390,EARLY] 7051 Format: {off | on} 7052 Specify if the kernel should make use of the cpu 7053 topology information if the hardware supports this. 7054 The scheduler will make use of this information and 7055 e.g. base its process migration decisions on it. 7056 Default is on. 7057 7058 torture.disable_onoff_at_boot= [KNL] 7059 Prevent the CPU-hotplug component of torturing 7060 until after init has spawned. 7061 7062 torture.ftrace_dump_at_shutdown= [KNL] 7063 Dump the ftrace buffer at torture-test shutdown, 7064 even if there were no errors. This can be a 7065 very costly operation when many torture tests 7066 are running concurrently, especially on systems 7067 with rotating-rust storage. 7068 7069 torture.verbose_sleep_frequency= [KNL] 7070 Specifies how many verbose printk()s should be 7071 emitted between each sleep. The default of zero 7072 disables verbose-printk() sleeping. 7073 7074 torture.verbose_sleep_duration= [KNL] 7075 Duration of each verbose-printk() sleep in jiffies. 7076 7077 tpm.disable_pcr_integrity= [HW,TPM] 7078 Do not protect PCR registers from unintended physical 7079 access, or interposers in the bus by the means of 7080 having an integrity protected session wrapped around 7081 TPM2_PCR_Extend command. Consider this in a situation 7082 where TPM is heavily utilized by IMA, thus protection 7083 causing a major performance hit, and the space where 7084 machines are deployed is by other means guarded. 7085 7086 tpm_suspend_pcr=[HW,TPM] 7087 Format: integer pcr id 7088 Specify that at suspend time, the tpm driver 7089 should extend the specified pcr with zeros, 7090 as a workaround for some chips which fail to 7091 flush the last written pcr on TPM_SaveState. 7092 This will guarantee that all the other pcrs 7093 are saved. 7094 7095 tpm_tis.interrupts= [HW,TPM] 7096 Enable interrupts for the MMIO based physical layer 7097 for the FIFO interface. By default it is set to false 7098 (0). For more information about TPM hardware interfaces 7099 defined by Trusted Computing Group (TCG) see 7100 https://trustedcomputinggroup.org/resource/pc-client-platform-tpm-profile-ptp-specification/ 7101 7102 tp_printk [FTRACE] 7103 Have the tracepoints sent to printk as well as the 7104 tracing ring buffer. This is useful for early boot up 7105 where the system hangs or reboots and does not give the 7106 option for reading the tracing buffer or performing a 7107 ftrace_dump_on_oops. 7108 7109 To turn off having tracepoints sent to printk, 7110 echo 0 > /proc/sys/kernel/tracepoint_printk 7111 Note, echoing 1 into this file without the 7112 tp_printk kernel cmdline option has no effect. 7113 7114 The tp_printk_stop_on_boot (see below) can also be used 7115 to stop the printing of events to console at 7116 late_initcall_sync. 7117 7118 ** CAUTION ** 7119 7120 Having tracepoints sent to printk() and activating high 7121 frequency tracepoints such as irq or sched, can cause 7122 the system to live lock. 7123 7124 tp_printk_stop_on_boot [FTRACE] 7125 When tp_printk (above) is set, it can cause a lot of noise 7126 on the console. It may be useful to only include the 7127 printing of events during boot up, as user space may 7128 make the system inoperable. 7129 7130 This command line option will stop the printing of events 7131 to console at the late_initcall_sync() time frame. 7132 7133 trace_buf_size=nn[KMG] 7134 [FTRACE] will set tracing buffer size on each cpu. 7135 7136 trace_clock= [FTRACE] Set the clock used for tracing events 7137 at boot up. 7138 local - Use the per CPU time stamp counter 7139 (converted into nanoseconds). Fast, but 7140 depending on the architecture, may not be 7141 in sync between CPUs. 7142 global - Event time stamps are synchronize across 7143 CPUs. May be slower than the local clock, 7144 but better for some race conditions. 7145 counter - Simple counting of events (1, 2, ..) 7146 note, some counts may be skipped due to the 7147 infrastructure grabbing the clock more than 7148 once per event. 7149 uptime - Use jiffies as the time stamp. 7150 perf - Use the same clock that perf uses. 7151 mono - Use ktime_get_mono_fast_ns() for time stamps. 7152 mono_raw - Use ktime_get_raw_fast_ns() for time 7153 stamps. 7154 boot - Use ktime_get_boot_fast_ns() for time stamps. 7155 Architectures may add more clocks. See 7156 Documentation/trace/ftrace.rst for more details. 7157 7158 trace_event=[event-list] 7159 [FTRACE] Set and start specified trace events in order 7160 to facilitate early boot debugging. The event-list is a 7161 comma-separated list of trace events to enable. See 7162 also Documentation/trace/events.rst 7163 7164 trace_instance=[instance-info] 7165 [FTRACE] Create a ring buffer instance early in boot up. 7166 This will be listed in: 7167 7168 /sys/kernel/tracing/instances 7169 7170 Events can be enabled at the time the instance is created 7171 via: 7172 7173 trace_instance=<name>,<system1>:<event1>,<system2>:<event2> 7174 7175 Note, the "<system*>:" portion is optional if the event is 7176 unique. 7177 7178 trace_instance=foo,sched:sched_switch,irq_handler_entry,initcall 7179 7180 will enable the "sched_switch" event (note, the "sched:" is optional, and 7181 the same thing would happen if it was left off). The irq_handler_entry 7182 event, and all events under the "initcall" system. 7183 7184 Flags can be added to the instance to modify its behavior when it is 7185 created. The flags are separated by '^'. 7186 7187 The available flags are: 7188 7189 traceoff - Have the tracing instance tracing disabled after it is created. 7190 traceprintk - Have trace_printk() write into this trace instance 7191 (note, "printk" and "trace_printk" can also be used) 7192 7193 trace_instance=foo^traceoff^traceprintk,sched,irq 7194 7195 The flags must come before the defined events. 7196 7197 If memory has been reserved (see memmap for x86), the instance 7198 can use that memory: 7199 7200 memmap=12M$0x284500000 trace_instance=boot_map@0x284500000:12M 7201 7202 The above will create a "boot_map" instance that uses the physical 7203 memory at 0x284500000 that is 12Megs. The per CPU buffers of that 7204 instance will be split up accordingly. 7205 7206 Alternatively, the memory can be reserved by the reserve_mem option: 7207 7208 reserve_mem=12M:4096:trace trace_instance=boot_map@trace 7209 7210 This will reserve 12 megabytes at boot up with a 4096 byte alignment 7211 and place the ring buffer in this memory. Note that due to KASLR, the 7212 memory may not be the same location each time, which will not preserve 7213 the buffer content. 7214 7215 Also note that the layout of the ring buffer data may change between 7216 kernel versions where the validator will fail and reset the ring buffer 7217 if the layout is not the same as the previous kernel. 7218 7219 If the ring buffer is used for persistent bootups and has events enabled, 7220 it is recommend to disable tracing so that events from a previous boot do not 7221 mix with events of the current boot (unless you are debugging a random crash 7222 at boot up). 7223 7224 reserve_mem=12M:4096:trace trace_instance=boot_map^traceoff^traceprintk@trace,sched,irq 7225 7226 Note, saving the trace buffer across reboots does require that the system 7227 is set up to not wipe memory. For instance, CONFIG_RESET_ATTACK_MITIGATION 7228 can force a memory reset on boot which will clear any trace that was stored. 7229 This is just one of many ways that can clear memory. Make sure your system 7230 keeps the content of memory across reboots before relying on this option. 7231 7232 See also Documentation/trace/debugging.rst 7233 7234 7235 trace_options=[option-list] 7236 [FTRACE] Enable or disable tracer options at boot. 7237 The option-list is a comma delimited list of options 7238 that can be enabled or disabled just as if you were 7239 to echo the option name into 7240 7241 /sys/kernel/tracing/trace_options 7242 7243 For example, to enable stacktrace option (to dump the 7244 stack trace of each event), add to the command line: 7245 7246 trace_options=stacktrace 7247 7248 See also Documentation/trace/ftrace.rst "trace options" 7249 section. 7250 7251 trace_trigger=[trigger-list] 7252 [FTRACE] Add a event trigger on specific events. 7253 Set a trigger on top of a specific event, with an optional 7254 filter. 7255 7256 The format is is "trace_trigger=<event>.<trigger>[ if <filter>],..." 7257 Where more than one trigger may be specified that are comma deliminated. 7258 7259 For example: 7260 7261 trace_trigger="sched_switch.stacktrace if prev_state == 2" 7262 7263 The above will enable the "stacktrace" trigger on the "sched_switch" 7264 event but only trigger it if the "prev_state" of the "sched_switch" 7265 event is "2" (TASK_UNINTERUPTIBLE). 7266 7267 See also "Event triggers" in Documentation/trace/events.rst 7268 7269 7270 traceoff_on_warning 7271 [FTRACE] enable this option to disable tracing when a 7272 warning is hit. This turns off "tracing_on". Tracing can 7273 be enabled again by echoing '1' into the "tracing_on" 7274 file located in /sys/kernel/tracing/ 7275 7276 This option is useful, as it disables the trace before 7277 the WARNING dump is called, which prevents the trace to 7278 be filled with content caused by the warning output. 7279 7280 This option can also be set at run time via the sysctl 7281 option: kernel/traceoff_on_warning 7282 7283 transparent_hugepage= 7284 [KNL] 7285 Format: [always|madvise|never] 7286 Can be used to control the default behavior of the system 7287 with respect to transparent hugepages. 7288 See Documentation/admin-guide/mm/transhuge.rst 7289 for more details. 7290 7291 transparent_hugepage_shmem= [KNL] 7292 Format: [always|within_size|advise|never|deny|force] 7293 Can be used to control the hugepage allocation policy for 7294 the internal shmem mount. 7295 See Documentation/admin-guide/mm/transhuge.rst 7296 for more details. 7297 7298 trusted.source= [KEYS] 7299 Format: <string> 7300 This parameter identifies the trust source as a backend 7301 for trusted keys implementation. Supported trust 7302 sources: 7303 - "tpm" 7304 - "tee" 7305 - "caam" 7306 - "dcp" 7307 If not specified then it defaults to iterating through 7308 the trust source list starting with TPM and assigns the 7309 first trust source as a backend which is initialized 7310 successfully during iteration. 7311 7312 trusted.rng= [KEYS] 7313 Format: <string> 7314 The RNG used to generate key material for trusted keys. 7315 Can be one of: 7316 - "kernel" 7317 - the same value as trusted.source: "tpm" or "tee" 7318 - "default" 7319 If not specified, "default" is used. In this case, 7320 the RNG's choice is left to each individual trust source. 7321 7322 trusted.dcp_use_otp_key 7323 This is intended to be used in combination with 7324 trusted.source=dcp and will select the DCP OTP key 7325 instead of the DCP UNIQUE key blob encryption. 7326 7327 trusted.dcp_skip_zk_test 7328 This is intended to be used in combination with 7329 trusted.source=dcp and will disable the check if the 7330 blob key is all zeros. This is helpful for situations where 7331 having this key zero'ed is acceptable. E.g. in testing 7332 scenarios. 7333 7334 tsc= Disable clocksource stability checks for TSC. 7335 Format: <string> 7336 [x86] reliable: mark tsc clocksource as reliable, this 7337 disables clocksource verification at runtime, as well 7338 as the stability checks done at bootup. Used to enable 7339 high-resolution timer mode on older hardware, and in 7340 virtualized environment. 7341 [x86] noirqtime: Do not use TSC to do irq accounting. 7342 Used to run time disable IRQ_TIME_ACCOUNTING on any 7343 platforms where RDTSC is slow and this accounting 7344 can add overhead. 7345 [x86] unstable: mark the TSC clocksource as unstable, this 7346 marks the TSC unconditionally unstable at bootup and 7347 avoids any further wobbles once the TSC watchdog notices. 7348 [x86] nowatchdog: disable clocksource watchdog. Used 7349 in situations with strict latency requirements (where 7350 interruptions from clocksource watchdog are not 7351 acceptable). 7352 [x86] recalibrate: force recalibration against a HW timer 7353 (HPET or PM timer) on systems whose TSC frequency was 7354 obtained from HW or FW using either an MSR or CPUID(0x15). 7355 Warn if the difference is more than 500 ppm. 7356 [x86] watchdog: Use TSC as the watchdog clocksource with 7357 which to check other HW timers (HPET or PM timer), but 7358 only on systems where TSC has been deemed trustworthy. 7359 This will be suppressed by an earlier tsc=nowatchdog and 7360 can be overridden by a later tsc=nowatchdog. A console 7361 message will flag any such suppression or overriding. 7362 7363 tsc_early_khz= [X86,EARLY] Skip early TSC calibration and use the given 7364 value instead. Useful when the early TSC frequency discovery 7365 procedure is not reliable, such as on overclocked systems 7366 with CPUID.16h support and partial CPUID.15h support. 7367 Format: <unsigned int> 7368 7369 tsx= [X86] Control Transactional Synchronization 7370 Extensions (TSX) feature in Intel processors that 7371 support TSX control. 7372 7373 This parameter controls the TSX feature. The options are: 7374 7375 on - Enable TSX on the system. Although there are 7376 mitigations for all known security vulnerabilities, 7377 TSX has been known to be an accelerator for 7378 several previous speculation-related CVEs, and 7379 so there may be unknown security risks associated 7380 with leaving it enabled. 7381 7382 off - Disable TSX on the system. (Note that this 7383 option takes effect only on newer CPUs which are 7384 not vulnerable to MDS, i.e., have 7385 MSR_IA32_ARCH_CAPABILITIES.MDS_NO=1 and which get 7386 the new IA32_TSX_CTRL MSR through a microcode 7387 update. This new MSR allows for the reliable 7388 deactivation of the TSX functionality.) 7389 7390 auto - Disable TSX if X86_BUG_TAA is present, 7391 otherwise enable TSX on the system. 7392 7393 Not specifying this option is equivalent to tsx=off. 7394 7395 See Documentation/admin-guide/hw-vuln/tsx_async_abort.rst 7396 for more details. 7397 7398 tsx_async_abort= [X86,INTEL,EARLY] Control mitigation for the TSX Async 7399 Abort (TAA) vulnerability. 7400 7401 Similar to Micro-architectural Data Sampling (MDS) 7402 certain CPUs that support Transactional 7403 Synchronization Extensions (TSX) are vulnerable to an 7404 exploit against CPU internal buffers which can forward 7405 information to a disclosure gadget under certain 7406 conditions. 7407 7408 In vulnerable processors, the speculatively forwarded 7409 data can be used in a cache side channel attack, to 7410 access data to which the attacker does not have direct 7411 access. 7412 7413 This parameter controls the TAA mitigation. The 7414 options are: 7415 7416 full - Enable TAA mitigation on vulnerable CPUs 7417 if TSX is enabled. 7418 7419 full,nosmt - Enable TAA mitigation and disable SMT on 7420 vulnerable CPUs. If TSX is disabled, SMT 7421 is not disabled because CPU is not 7422 vulnerable to cross-thread TAA attacks. 7423 off - Unconditionally disable TAA mitigation 7424 7425 On MDS-affected machines, tsx_async_abort=off can be 7426 prevented by an active MDS mitigation as both vulnerabilities 7427 are mitigated with the same mechanism so in order to disable 7428 this mitigation, you need to specify mds=off too. 7429 7430 Not specifying this option is equivalent to 7431 tsx_async_abort=full. On CPUs which are MDS affected 7432 and deploy MDS mitigation, TAA mitigation is not 7433 required and doesn't provide any additional 7434 mitigation. 7435 7436 For details see: 7437 Documentation/admin-guide/hw-vuln/tsx_async_abort.rst 7438 7439 turbografx.map[2|3]= [HW,JOY] 7440 TurboGraFX parallel port interface 7441 Format: 7442 <port#>,<js1>,<js2>,<js3>,<js4>,<js5>,<js6>,<js7> 7443 See also Documentation/input/devices/joystick-parport.rst 7444 7445 udbg-immortal [PPC] When debugging early kernel crashes that 7446 happen after console_init() and before a proper 7447 console driver takes over, this boot options might 7448 help "seeing" what's going on. 7449 7450 uhash_entries= [KNL,NET] 7451 Set number of hash buckets for UDP/UDP-Lite connections 7452 7453 uhci-hcd.ignore_oc= 7454 [USB] Ignore overcurrent events (default N). 7455 Some badly-designed motherboards generate lots of 7456 bogus events, for ports that aren't wired to 7457 anything. Set this parameter to avoid log spamming. 7458 Note that genuine overcurrent events won't be 7459 reported either. 7460 7461 unknown_nmi_panic 7462 [X86] Cause panic on unknown NMI. 7463 7464 unwind_debug [X86-64,EARLY] 7465 Enable unwinder debug output. This can be 7466 useful for debugging certain unwinder error 7467 conditions, including corrupt stacks and 7468 bad/missing unwinder metadata. 7469 7470 usbcore.authorized_default= 7471 [USB] Default USB device authorization: 7472 (default -1 = authorized (same as 1), 7473 0 = not authorized, 1 = authorized, 2 = authorized 7474 if device connected to internal port) 7475 7476 usbcore.autosuspend= 7477 [USB] The autosuspend time delay (in seconds) used 7478 for newly-detected USB devices (default 2). This 7479 is the time required before an idle device will be 7480 autosuspended. Devices for which the delay is set 7481 to a negative value won't be autosuspended at all. 7482 7483 usbcore.usbfs_snoop= 7484 [USB] Set to log all usbfs traffic (default 0 = off). 7485 7486 usbcore.usbfs_snoop_max= 7487 [USB] Maximum number of bytes to snoop in each URB 7488 (default = 65536). 7489 7490 usbcore.blinkenlights= 7491 [USB] Set to cycle leds on hubs (default 0 = off). 7492 7493 usbcore.old_scheme_first= 7494 [USB] Start with the old device initialization 7495 scheme (default 0 = off). 7496 7497 usbcore.usbfs_memory_mb= 7498 [USB] Memory limit (in MB) for buffers allocated by 7499 usbfs (default = 16, 0 = max = 2047). 7500 7501 usbcore.use_both_schemes= 7502 [USB] Try the other device initialization scheme 7503 if the first one fails (default 1 = enabled). 7504 7505 usbcore.initial_descriptor_timeout= 7506 [USB] Specifies timeout for the initial 64-byte 7507 USB_REQ_GET_DESCRIPTOR request in milliseconds 7508 (default 5000 = 5.0 seconds). 7509 7510 usbcore.nousb [USB] Disable the USB subsystem 7511 7512 usbcore.quirks= 7513 [USB] A list of quirk entries to augment the built-in 7514 usb core quirk list. List entries are separated by 7515 commas. Each entry has the form 7516 VendorID:ProductID:Flags. The IDs are 4-digit hex 7517 numbers and Flags is a set of letters. Each letter 7518 will change the built-in quirk; setting it if it is 7519 clear and clearing it if it is set. The letters have 7520 the following meanings: 7521 a = USB_QUIRK_STRING_FETCH_255 (string 7522 descriptors must not be fetched using 7523 a 255-byte read); 7524 b = USB_QUIRK_RESET_RESUME (device can't resume 7525 correctly so reset it instead); 7526 c = USB_QUIRK_NO_SET_INTF (device can't handle 7527 Set-Interface requests); 7528 d = USB_QUIRK_CONFIG_INTF_STRINGS (device can't 7529 handle its Configuration or Interface 7530 strings); 7531 e = USB_QUIRK_RESET (device can't be reset 7532 (e.g morph devices), don't use reset); 7533 f = USB_QUIRK_HONOR_BNUMINTERFACES (device has 7534 more interface descriptions than the 7535 bNumInterfaces count, and can't handle 7536 talking to these interfaces); 7537 g = USB_QUIRK_DELAY_INIT (device needs a pause 7538 during initialization, after we read 7539 the device descriptor); 7540 h = USB_QUIRK_LINEAR_UFRAME_INTR_BINTERVAL (For 7541 high speed and super speed interrupt 7542 endpoints, the USB 2.0 and USB 3.0 spec 7543 require the interval in microframes (1 7544 microframe = 125 microseconds) to be 7545 calculated as interval = 2 ^ 7546 (bInterval-1). 7547 Devices with this quirk report their 7548 bInterval as the result of this 7549 calculation instead of the exponent 7550 variable used in the calculation); 7551 i = USB_QUIRK_DEVICE_QUALIFIER (device can't 7552 handle device_qualifier descriptor 7553 requests); 7554 j = USB_QUIRK_IGNORE_REMOTE_WAKEUP (device 7555 generates spurious wakeup, ignore 7556 remote wakeup capability); 7557 k = USB_QUIRK_NO_LPM (device can't handle Link 7558 Power Management); 7559 l = USB_QUIRK_LINEAR_FRAME_INTR_BINTERVAL 7560 (Device reports its bInterval as linear 7561 frames instead of the USB 2.0 7562 calculation); 7563 m = USB_QUIRK_DISCONNECT_SUSPEND (Device needs 7564 to be disconnected before suspend to 7565 prevent spurious wakeup); 7566 n = USB_QUIRK_DELAY_CTRL_MSG (Device needs a 7567 pause after every control message); 7568 o = USB_QUIRK_HUB_SLOW_RESET (Hub needs extra 7569 delay after resetting its port); 7570 p = USB_QUIRK_SHORT_SET_ADDRESS_REQ_TIMEOUT 7571 (Reduce timeout of the SET_ADDRESS 7572 request from 5000 ms to 500 ms); 7573 Example: quirks=0781:5580:bk,0a5c:5834:gij 7574 7575 usbhid.mousepoll= 7576 [USBHID] The interval which mice are to be polled at. 7577 7578 usbhid.jspoll= 7579 [USBHID] The interval which joysticks are to be polled at. 7580 7581 usbhid.kbpoll= 7582 [USBHID] The interval which keyboards are to be polled at. 7583 7584 usb-storage.delay_use= 7585 [UMS] The delay in seconds before a new device is 7586 scanned for Logical Units (default 1). 7587 Optionally the delay in milliseconds if the value has 7588 suffix with "ms". 7589 Example: delay_use=2567ms 7590 7591 usb-storage.quirks= 7592 [UMS] A list of quirks entries to supplement or 7593 override the built-in unusual_devs list. List 7594 entries are separated by commas. Each entry has 7595 the form VID:PID:Flags where VID and PID are Vendor 7596 and Product ID values (4-digit hex numbers) and 7597 Flags is a set of characters, each corresponding 7598 to a common usb-storage quirk flag as follows: 7599 a = SANE_SENSE (collect more than 18 bytes 7600 of sense data, not on uas); 7601 b = BAD_SENSE (don't collect more than 18 7602 bytes of sense data, not on uas); 7603 c = FIX_CAPACITY (decrease the reported 7604 device capacity by one sector); 7605 d = NO_READ_DISC_INFO (don't use 7606 READ_DISC_INFO command, not on uas); 7607 e = NO_READ_CAPACITY_16 (don't use 7608 READ_CAPACITY_16 command); 7609 f = NO_REPORT_OPCODES (don't use report opcodes 7610 command, uas only); 7611 g = MAX_SECTORS_240 (don't transfer more than 7612 240 sectors at a time, uas only); 7613 h = CAPACITY_HEURISTICS (decrease the 7614 reported device capacity by one 7615 sector if the number is odd); 7616 i = IGNORE_DEVICE (don't bind to this 7617 device); 7618 j = NO_REPORT_LUNS (don't use report luns 7619 command, uas only); 7620 k = NO_SAME (do not use WRITE_SAME, uas only) 7621 l = NOT_LOCKABLE (don't try to lock and 7622 unlock ejectable media, not on uas); 7623 m = MAX_SECTORS_64 (don't transfer more 7624 than 64 sectors = 32 KB at a time, 7625 not on uas); 7626 n = INITIAL_READ10 (force a retry of the 7627 initial READ(10) command, not on uas); 7628 o = CAPACITY_OK (accept the capacity 7629 reported by the device, not on uas); 7630 p = WRITE_CACHE (the device cache is ON 7631 by default, not on uas); 7632 r = IGNORE_RESIDUE (the device reports 7633 bogus residue values, not on uas); 7634 s = SINGLE_LUN (the device has only one 7635 Logical Unit); 7636 t = NO_ATA_1X (don't allow ATA(12) and ATA(16) 7637 commands, uas only); 7638 u = IGNORE_UAS (don't bind to the uas driver); 7639 w = NO_WP_DETECT (don't test whether the 7640 medium is write-protected). 7641 y = ALWAYS_SYNC (issue a SYNCHRONIZE_CACHE 7642 even if the device claims no cache, 7643 not on uas) 7644 Example: quirks=0419:aaf5:rl,0421:0433:rc 7645 7646 user_debug= [KNL,ARM] 7647 Format: <int> 7648 See arch/arm/Kconfig.debug help text. 7649 1 - undefined instruction events 7650 2 - system calls 7651 4 - invalid data aborts 7652 8 - SIGSEGV faults 7653 16 - SIGBUS faults 7654 Example: user_debug=31 7655 7656 userpte= 7657 [X86,EARLY] Flags controlling user PTE allocations. 7658 7659 nohigh = do not allocate PTE pages in 7660 HIGHMEM regardless of setting 7661 of CONFIG_HIGHPTE. 7662 7663 vdso= [X86,SH,SPARC] 7664 On X86_32, this is an alias for vdso32=. Otherwise: 7665 7666 vdso=1: enable VDSO (the default) 7667 vdso=0: disable VDSO mapping 7668 7669 vdso32= [X86] Control the 32-bit vDSO 7670 vdso32=1: enable 32-bit VDSO 7671 vdso32=0 or vdso32=2: disable 32-bit VDSO 7672 7673 See the help text for CONFIG_COMPAT_VDSO for more 7674 details. If CONFIG_COMPAT_VDSO is set, the default is 7675 vdso32=0; otherwise, the default is vdso32=1. 7676 7677 For compatibility with older kernels, vdso32=2 is an 7678 alias for vdso32=0. 7679 7680 Try vdso32=0 if you encounter an error that says: 7681 dl_main: Assertion `(void *) ph->p_vaddr == _rtld_local._dl_sysinfo_dso' failed! 7682 7683 video= [FB,EARLY] Frame buffer configuration 7684 See Documentation/fb/modedb.rst. 7685 7686 video.brightness_switch_enabled= [ACPI] 7687 Format: [0|1] 7688 If set to 1, on receiving an ACPI notify event 7689 generated by hotkey, video driver will adjust brightness 7690 level and then send out the event to user space through 7691 the allocated input device. If set to 0, video driver 7692 will only send out the event without touching backlight 7693 brightness level. 7694 default: 1 7695 7696 virtio_mmio.device= 7697 [VMMIO] Memory mapped virtio (platform) device. 7698 7699 <size>@<baseaddr>:<irq>[:<id>] 7700 where: 7701 <size> := size (can use standard suffixes 7702 like K, M and G) 7703 <baseaddr> := physical base address 7704 <irq> := interrupt number (as passed to 7705 request_irq()) 7706 <id> := (optional) platform device id 7707 example: 7708 virtio_mmio.device=1K@0x100b0000:48:7 7709 7710 Can be used multiple times for multiple devices. 7711 7712 vga= [BOOT,X86-32] Select a particular video mode 7713 See Documentation/arch/x86/boot.rst and 7714 Documentation/admin-guide/svga.rst. 7715 Use vga=ask for menu. 7716 This is actually a boot loader parameter; the value is 7717 passed to the kernel using a special protocol. 7718 7719 vm_debug[=options] [KNL] Available with CONFIG_DEBUG_VM=y. 7720 May slow down system boot speed, especially when 7721 enabled on systems with a large amount of memory. 7722 All options are enabled by default, and this 7723 interface is meant to allow for selectively 7724 enabling or disabling specific virtual memory 7725 debugging features. 7726 7727 Available options are: 7728 P Enable page structure init time poisoning 7729 - Disable all of the above options 7730 7731 vmalloc=nn[KMG] [KNL,BOOT,EARLY] Forces the vmalloc area to have an 7732 exact size of <nn>. This can be used to increase 7733 the minimum size (128MB on x86, arm32 platforms). 7734 It can also be used to decrease the size and leave more room 7735 for directly mapped kernel RAM. Note that this parameter does 7736 not exist on many other platforms (including arm64, alpha, 7737 loongarch, arc, csky, hexagon, microblaze, mips, nios2, openrisc, 7738 parisc, m64k, powerpc, riscv, sh, um, xtensa, s390, sparc). 7739 7740 vmcp_cma=nn[MG] [KNL,S390,EARLY] 7741 Sets the memory size reserved for contiguous memory 7742 allocations for the vmcp device driver. 7743 7744 vmhalt= [KNL,S390] Perform z/VM CP command after system halt. 7745 Format: <command> 7746 7747 vmpanic= [KNL,S390] Perform z/VM CP command after kernel panic. 7748 Format: <command> 7749 7750 vmpoff= [KNL,S390] Perform z/VM CP command after power off. 7751 Format: <command> 7752 7753 vsyscall= [X86-64,EARLY] 7754 Controls the behavior of vsyscalls (i.e. calls to 7755 fixed addresses of 0xffffffffff600x00 from legacy 7756 code). Most statically-linked binaries and older 7757 versions of glibc use these calls. Because these 7758 functions are at fixed addresses, they make nice 7759 targets for exploits that can control RIP. 7760 7761 emulate Vsyscalls turn into traps and are emulated 7762 reasonably safely. The vsyscall page is 7763 readable. 7764 7765 xonly [default] Vsyscalls turn into traps and are 7766 emulated reasonably safely. The vsyscall 7767 page is not readable. 7768 7769 none Vsyscalls don't work at all. This makes 7770 them quite hard to use for exploits but 7771 might break your system. 7772 7773 vt.color= [VT] Default text color. 7774 Format: 0xYX, X = foreground, Y = background. 7775 Default: 0x07 = light gray on black. 7776 7777 vt.cur_default= [VT] Default cursor shape. 7778 Format: 0xCCBBAA, where AA, BB, and CC are the same as 7779 the parameters of the <Esc>[?A;B;Cc escape sequence; 7780 see vga-softcursor.rst. Default: 2 = underline. 7781 7782 vt.default_blu= [VT] 7783 Format: <blue0>,<blue1>,<blue2>,...,<blue15> 7784 Change the default blue palette of the console. 7785 This is a 16-member array composed of values 7786 ranging from 0-255. 7787 7788 vt.default_grn= [VT] 7789 Format: <green0>,<green1>,<green2>,...,<green15> 7790 Change the default green palette of the console. 7791 This is a 16-member array composed of values 7792 ranging from 0-255. 7793 7794 vt.default_red= [VT] 7795 Format: <red0>,<red1>,<red2>,...,<red15> 7796 Change the default red palette of the console. 7797 This is a 16-member array composed of values 7798 ranging from 0-255. 7799 7800 vt.default_utf8= 7801 [VT] 7802 Format=<0|1> 7803 Set system-wide default UTF-8 mode for all tty's. 7804 Default is 1, i.e. UTF-8 mode is enabled for all 7805 newly opened terminals. 7806 7807 vt.global_cursor_default= 7808 [VT] 7809 Format=<-1|0|1> 7810 Set system-wide default for whether a cursor 7811 is shown on new VTs. Default is -1, 7812 i.e. cursors will be created by default unless 7813 overridden by individual drivers. 0 will hide 7814 cursors, 1 will display them. 7815 7816 vt.italic= [VT] Default color for italic text; 0-15. 7817 Default: 2 = green. 7818 7819 vt.underline= [VT] Default color for underlined text; 0-15. 7820 Default: 3 = cyan. 7821 7822 watchdog timers [HW,WDT] For information on watchdog timers, 7823 see Documentation/watchdog/watchdog-parameters.rst 7824 or other driver-specific files in the 7825 Documentation/watchdog/ directory. 7826 7827 watchdog_thresh= 7828 [KNL] 7829 Set the hard lockup detector stall duration 7830 threshold in seconds. The soft lockup detector 7831 threshold is set to twice the value. A value of 0 7832 disables both lockup detectors. Default is 10 7833 seconds. 7834 7835 workqueue.unbound_cpus= 7836 [KNL,SMP] Specify to constrain one or some CPUs 7837 to use in unbound workqueues. 7838 Format: <cpu-list> 7839 By default, all online CPUs are available for 7840 unbound workqueues. 7841 7842 workqueue.watchdog_thresh= 7843 If CONFIG_WQ_WATCHDOG is configured, workqueue can 7844 warn stall conditions and dump internal state to 7845 help debugging. 0 disables workqueue stall 7846 detection; otherwise, it's the stall threshold 7847 duration in seconds. The default value is 30 and 7848 it can be updated at runtime by writing to the 7849 corresponding sysfs file. 7850 7851 workqueue.panic_on_stall=<uint> 7852 Panic when workqueue stall is detected by 7853 CONFIG_WQ_WATCHDOG. It sets the number times of the 7854 stall to trigger panic. 7855 7856 The default is 0, which disables the panic on stall. 7857 7858 workqueue.cpu_intensive_thresh_us= 7859 Per-cpu work items which run for longer than this 7860 threshold are automatically considered CPU intensive 7861 and excluded from concurrency management to prevent 7862 them from noticeably delaying other per-cpu work 7863 items. Default is 10000 (10ms). 7864 7865 If CONFIG_WQ_CPU_INTENSIVE_REPORT is set, the kernel 7866 will report the work functions which violate this 7867 threshold repeatedly. They are likely good 7868 candidates for using WQ_UNBOUND workqueues instead. 7869 7870 workqueue.cpu_intensive_warning_thresh=<uint> 7871 If CONFIG_WQ_CPU_INTENSIVE_REPORT is set, the kernel 7872 will report the work functions which violate the 7873 intensive_threshold_us repeatedly. In order to prevent 7874 spurious warnings, start printing only after a work 7875 function has violated this threshold number of times. 7876 7877 The default is 4 times. 0 disables the warning. 7878 7879 workqueue.power_efficient 7880 Per-cpu workqueues are generally preferred because 7881 they show better performance thanks to cache 7882 locality; unfortunately, per-cpu workqueues tend to 7883 be more power hungry than unbound workqueues. 7884 7885 Enabling this makes the per-cpu workqueues which 7886 were observed to contribute significantly to power 7887 consumption unbound, leading to measurably lower 7888 power usage at the cost of small performance 7889 overhead. 7890 7891 The default value of this parameter is determined by 7892 the config option CONFIG_WQ_POWER_EFFICIENT_DEFAULT. 7893 7894 workqueue.default_affinity_scope= 7895 Select the default affinity scope to use for unbound 7896 workqueues. Can be one of "cpu", "smt", "cache", 7897 "numa" and "system". Default is "cache". For more 7898 information, see the Affinity Scopes section in 7899 Documentation/core-api/workqueue.rst. 7900 7901 This can be changed after boot by writing to the 7902 matching /sys/module/workqueue/parameters file. All 7903 workqueues with the "default" affinity scope will be 7904 updated accordingly. 7905 7906 workqueue.debug_force_rr_cpu 7907 Workqueue used to implicitly guarantee that work 7908 items queued without explicit CPU specified are put 7909 on the local CPU. This guarantee is no longer true 7910 and while local CPU is still preferred work items 7911 may be put on foreign CPUs. This debug option 7912 forces round-robin CPU selection to flush out 7913 usages which depend on the now broken guarantee. 7914 When enabled, memory and cache locality will be 7915 impacted. 7916 7917 writecombine= [LOONGARCH,EARLY] Control the MAT (Memory Access 7918 Type) of ioremap_wc(). 7919 7920 on - Enable writecombine, use WUC for ioremap_wc() 7921 off - Disable writecombine, use SUC for ioremap_wc() 7922 7923 x2apic_phys [X86-64,APIC,EARLY] Use x2apic physical mode instead of 7924 default x2apic cluster mode on platforms 7925 supporting x2apic. 7926 7927 xen_512gb_limit [KNL,X86-64,XEN] 7928 Restricts the kernel running paravirtualized under Xen 7929 to use only up to 512 GB of RAM. The reason to do so is 7930 crash analysis tools and Xen tools for doing domain 7931 save/restore/migration must be enabled to handle larger 7932 domains. 7933 7934 xen_emul_unplug= [HW,X86,XEN,EARLY] 7935 Unplug Xen emulated devices 7936 Format: [unplug0,][unplug1] 7937 ide-disks -- unplug primary master IDE devices 7938 aux-ide-disks -- unplug non-primary-master IDE devices 7939 nics -- unplug network devices 7940 all -- unplug all emulated devices (NICs and IDE disks) 7941 unnecessary -- unplugging emulated devices is 7942 unnecessary even if the host did not respond to 7943 the unplug protocol 7944 never -- do not unplug even if version check succeeds 7945 7946 xen_legacy_crash [X86,XEN,EARLY] 7947 Crash from Xen panic notifier, without executing late 7948 panic() code such as dumping handler. 7949 7950 xen_mc_debug [X86,XEN,EARLY] 7951 Enable multicall debugging when running as a Xen PV guest. 7952 Enabling this feature will reduce performance a little 7953 bit, so it should only be enabled for obtaining extended 7954 debug data in case of multicall errors. 7955 7956 xen_msr_safe= [X86,XEN,EARLY] 7957 Format: <bool> 7958 Select whether to always use non-faulting (safe) MSR 7959 access functions when running as Xen PV guest. The 7960 default value is controlled by CONFIG_XEN_PV_MSR_SAFE. 7961 7962 xen_nopv [X86] 7963 Disables the PV optimizations forcing the HVM guest to 7964 run as generic HVM guest with no PV drivers. 7965 This option is obsoleted by the "nopv" option, which 7966 has equivalent effect for XEN platform. 7967 7968 xen_no_vector_callback 7969 [KNL,X86,XEN,EARLY] Disable the vector callback for Xen 7970 event channel interrupts. 7971 7972 xen_scrub_pages= [XEN] 7973 Boolean option to control scrubbing pages before giving them back 7974 to Xen, for use by other domains. Can be also changed at runtime 7975 with /sys/devices/system/xen_memory/xen_memory0/scrub_pages. 7976 Default value controlled with CONFIG_XEN_SCRUB_PAGES_DEFAULT. 7977 7978 xen_timer_slop= [X86-64,XEN,EARLY] 7979 Set the timer slop (in nanoseconds) for the virtual Xen 7980 timers (default is 100000). This adjusts the minimum 7981 delta of virtualized Xen timers, where lower values 7982 improve timer resolution at the expense of processing 7983 more timer interrupts. 7984 7985 xen.balloon_boot_timeout= [XEN] 7986 The time (in seconds) to wait before giving up to boot 7987 in case initial ballooning fails to free enough memory. 7988 Applies only when running as HVM or PVH guest and 7989 started with less memory configured than allowed at 7990 max. Default is 180. 7991 7992 xen.event_eoi_delay= [XEN] 7993 How long to delay EOI handling in case of event 7994 storms (jiffies). Default is 10. 7995 7996 xen.event_loop_timeout= [XEN] 7997 After which time (jiffies) the event handling loop 7998 should start to delay EOI handling. Default is 2. 7999 8000 xen.fifo_events= [XEN] 8001 Boolean parameter to disable using fifo event handling 8002 even if available. Normally fifo event handling is 8003 preferred over the 2-level event handling, as it is 8004 fairer and the number of possible event channels is 8005 much higher. Default is on (use fifo events). 8006 8007 xirc2ps_cs= [NET,PCMCIA] 8008 Format: 8009 <irq>,<irq_mask>,<io>,<full_duplex>,<do_sound>,<lockup_hack>[,<irq2>[,<irq3>[,<irq4>]]] 8010 8011 xive= [PPC] 8012 By default on POWER9 and above, the kernel will 8013 natively use the XIVE interrupt controller. This option 8014 allows the fallback firmware mode to be used: 8015 8016 off Fallback to firmware control of XIVE interrupt 8017 controller on both pseries and powernv 8018 platforms. Only useful on POWER9 and above. 8019 8020 xive.store-eoi=off [PPC] 8021 By default on POWER10 and above, the kernel will use 8022 stores for EOI handling when the XIVE interrupt mode 8023 is active. This option allows the XIVE driver to use 8024 loads instead, as on POWER9. 8025 8026 xhci-hcd.quirks [USB,KNL] 8027 A hex value specifying bitmask with supplemental xhci 8028 host controller quirks. Meaning of each bit can be 8029 consulted in header drivers/usb/host/xhci.h. 8030 8031 xmon [PPC,EARLY] 8032 Format: { early | on | rw | ro | off } 8033 Controls if xmon debugger is enabled. Default is off. 8034 Passing only "xmon" is equivalent to "xmon=early". 8035 early Call xmon as early as possible on boot; xmon 8036 debugger is called from setup_arch(). 8037 on xmon debugger hooks will be installed so xmon 8038 is only called on a kernel crash. Default mode, 8039 i.e. either "ro" or "rw" mode, is controlled 8040 with CONFIG_XMON_DEFAULT_RO_MODE. 8041 rw xmon debugger hooks will be installed so xmon 8042 is called only on a kernel crash, mode is write, 8043 meaning SPR registers, memory and, other data 8044 can be written using xmon commands. 8045 ro same as "rw" option above but SPR registers, 8046 memory, and other data can't be written using 8047 xmon commands. 8048 off xmon is disabled. 8049