| blob | dc663c0ca67067d041cf9a3767117eec765ccca8 |
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.
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 nospcr -- disable console in ACPI SPCR table as
25 default _serial_ console on ARM64
26 For ARM64, ONLY "acpi=off", "acpi=on", "acpi=force" or
27 "acpi=nospcr" are available
28 For RISCV64, ONLY "acpi=off", "acpi=on" or "acpi=force"
29 are available
31 See also Documentation/power/runtime_pm.rst, pci=noacpi
33 acpi_apic_instance= [ACPI,IOAPIC,EARLY]
34 Format: <int>
35 2: use 2nd APIC table, if available
36 1,0: use 1st APIC table
37 default: 0
39 acpi_backlight= [HW,ACPI]
40 { vendor | video | native | none }
41 If set to vendor, prefer vendor-specific driver
42 (e.g. thinkpad_acpi, sony_acpi, etc.) instead
43 of the ACPI video.ko driver.
44 If set to video, use the ACPI video.ko driver.
45 If set to native, use the device's native backlight mode.
46 If set to none, disable the ACPI backlight interface.
48 acpi_force_32bit_fadt_addr [ACPI,EARLY]
49 force FADT to use 32 bit addresses rather than the
50 64 bit X_* addresses. Some firmware have broken 64
51 bit addresses for force ACPI ignore these and use
52 the older legacy 32 bit addresses.
54 acpica_no_return_repair [HW, ACPI]
55 Disable AML predefined validation mechanism
56 This mechanism can repair the evaluation result to make
57 the return objects more ACPI specification compliant.
58 This option is useful for developers to identify the
59 root cause of an AML interpreter issue when the issue
60 has something to do with the repair mechanism.
62 acpi.debug_layer= [HW,ACPI,ACPI_DEBUG]
63 acpi.debug_level= [HW,ACPI,ACPI_DEBUG]
64 Format: <int>
65 CONFIG_ACPI_DEBUG must be enabled to produce any ACPI
66 debug output. Bits in debug_layer correspond to a
67 _COMPONENT in an ACPI source file, e.g.,
68 #define _COMPONENT ACPI_EVENTS
69 Bits in debug_level correspond to a level in
70 ACPI_DEBUG_PRINT statements, e.g.,
71 ACPI_DEBUG_PRINT((ACPI_DB_INFO, ...
72 The debug_level mask defaults to "info". See
73 Documentation/firmware-guide/acpi/debug.rst for more information about
74 debug layers and levels.
76 Enable processor driver info messages:
77 acpi.debug_layer=0x20000000
78 Enable AML "Debug" output, i.e., stores to the Debug
79 object while interpreting AML:
80 acpi.debug_layer=0xffffffff acpi.debug_level=0x2
81 Enable all messages related to ACPI hardware:
82 acpi.debug_layer=0x2 acpi.debug_level=0xffffffff
84 Some values produce so much output that the system is
85 unusable. The "log_buf_len" parameter may be useful
86 if you need to capture more output.
88 acpi_enforce_resources= [ACPI]
89 { strict | lax | no }
90 Check for resource conflicts between native drivers
91 and ACPI OperationRegions (SystemIO and SystemMemory
92 only). IO ports and memory declared in ACPI might be
93 used by the ACPI subsystem in arbitrary AML code and
94 can interfere with legacy drivers.
95 strict (default): access to resources claimed by ACPI
96 is denied; legacy drivers trying to access reserved
97 resources will fail to bind to device using them.
98 lax: access to resources claimed by ACPI is allowed;
99 legacy drivers trying to access reserved resources
100 will bind successfully but a warning message is logged.
101 no: ACPI OperationRegions are not marked as reserved,
102 no further checks are performed.
104 acpi_force_table_verification [HW,ACPI,EARLY]
105 Enable table checksum verification during early stage.
106 By default, this is disabled due to x86 early mapping
107 size limitation.
109 acpi_irq_balance [HW,ACPI]
110 ACPI will balance active IRQs
111 default in APIC mode
113 acpi_irq_nobalance [HW,ACPI]
114 ACPI will not move active IRQs (default)
115 default in PIC mode
117 acpi_irq_isa= [HW,ACPI] If irq_balance, mark listed IRQs used by ISA
118 Format: <irq>,<irq>...
120 acpi_irq_pci= [HW,ACPI] If irq_balance, clear listed IRQs for
121 use by PCI
122 Format: <irq>,<irq>...
124 acpi_mask_gpe= [HW,ACPI]
125 Due to the existence of _Lxx/_Exx, some GPEs triggered
126 by unsupported hardware/firmware features can result in
127 GPE floodings that cannot be automatically disabled by
128 the GPE dispatcher.
129 This facility can be used to prevent such uncontrolled
130 GPE floodings.
131 Format: <byte> or <bitmap-list>
133 acpi_no_auto_serialize [HW,ACPI]
134 Disable auto-serialization of AML methods
135 AML control methods that contain the opcodes to create
136 named objects will be marked as "Serialized" by the
137 auto-serialization feature.
138 This feature is enabled by default.
139 This option allows to turn off the feature.
141 acpi_no_memhotplug [ACPI] Disable memory hotplug. Useful for kdump
142 kernels.
144 acpi_no_static_ssdt [HW,ACPI,EARLY]
145 Disable installation of static SSDTs at early boot time
146 By default, SSDTs contained in the RSDT/XSDT will be
147 installed automatically and they will appear under
148 /sys/firmware/acpi/tables.
149 This option turns off this feature.
150 Note that specifying this option does not affect
151 dynamic table installation which will install SSDT
152 tables to /sys/firmware/acpi/tables/dynamic.
154 acpi_no_watchdog [HW,ACPI,WDT]
155 Ignore the ACPI-based watchdog interface (WDAT) and let
156 a native driver control the watchdog device instead.
158 acpi_rsdp= [ACPI,EFI,KEXEC,EARLY]
159 Pass the RSDP address to the kernel, mostly used
160 on machines running EFI runtime service to boot the
161 second kernel for kdump.
163 acpi_os_name= [HW,ACPI] Tell ACPI BIOS the name of the OS
164 Format: To spoof as Windows 98: ="Microsoft Windows"
166 acpi_rev_override [ACPI] Override the _REV object to return 5 (instead
167 of 2 which is mandated by ACPI 6) as the supported ACPI
168 specification revision (when using this switch, it may
169 be necessary to carry out a cold reboot _twice_ in a
170 row to make it take effect on the platform firmware).
172 acpi_osi= [HW,ACPI] Modify list of supported OS interface strings
173 acpi_osi="string1" # add string1
174 acpi_osi="!string2" # remove string2
175 acpi_osi=!* # remove all strings
176 acpi_osi=! # disable all built-in OS vendor
177 strings
178 acpi_osi=!! # enable all built-in OS vendor
179 strings
180 acpi_osi= # disable all strings
182 'acpi_osi=!' can be used in combination with single or
183 multiple 'acpi_osi="string1"' to support specific OS
184 vendor string(s). Note that such command can only
185 affect the default state of the OS vendor strings, thus
186 it cannot affect the default state of the feature group
187 strings and the current state of the OS vendor strings,
188 specifying it multiple times through kernel command line
189 is meaningless. This command is useful when one do not
190 care about the state of the feature group strings which
191 should be controlled by the OSPM.
192 Examples:
193 1. 'acpi_osi=! acpi_osi="Windows 2000"' is equivalent
194 to 'acpi_osi="Windows 2000" acpi_osi=!', they all
195 can make '_OSI("Windows 2000")' TRUE.
197 'acpi_osi=' cannot be used in combination with other
198 'acpi_osi=' command lines, the _OSI method will not
199 exist in the ACPI namespace. NOTE that such command can
200 only affect the _OSI support state, thus specifying it
201 multiple times through kernel command line is also
202 meaningless.
203 Examples:
204 1. 'acpi_osi=' can make 'CondRefOf(_OSI, Local1)'
205 FALSE.
207 'acpi_osi=!*' can be used in combination with single or
208 multiple 'acpi_osi="string1"' to support specific
209 string(s). Note that such command can affect the
210 current state of both the OS vendor strings and the
211 feature group strings, thus specifying it multiple times
212 through kernel command line is meaningful. But it may
213 still not able to affect the final state of a string if
214 there are quirks related to this string. This command
215 is useful when one want to control the state of the
216 feature group strings to debug BIOS issues related to
217 the OSPM features.
218 Examples:
219 1. 'acpi_osi="Module Device" acpi_osi=!*' can make
220 '_OSI("Module Device")' FALSE.
221 2. 'acpi_osi=!* acpi_osi="Module Device"' can make
222 '_OSI("Module Device")' TRUE.
223 3. 'acpi_osi=! acpi_osi=!* acpi_osi="Windows 2000"' is
224 equivalent to
225 'acpi_osi=!* acpi_osi=! acpi_osi="Windows 2000"'
226 and
227 'acpi_osi=!* acpi_osi="Windows 2000" acpi_osi=!',
228 they all will make '_OSI("Windows 2000")' TRUE.
230 acpi_pm_good [X86]
231 Override the pmtimer bug detection: force the kernel
232 to assume that this machine's pmtimer latches its value
233 and always returns good values.
235 acpi_sci= [HW,ACPI,EARLY] ACPI System Control Interrupt trigger mode
236 Format: { level | edge | high | low }
238 acpi_skip_timer_override [HW,ACPI,EARLY]
239 Recognize and ignore IRQ0/pin2 Interrupt Override.
240 For broken nForce2 BIOS resulting in XT-PIC timer.
242 acpi_sleep= [HW,ACPI] Sleep options
243 Format: { s3_bios, s3_mode, s3_beep, s4_hwsig,
244 s4_nohwsig, old_ordering, nonvs,
245 sci_force_enable, nobl }
246 See Documentation/power/video.rst for information on
247 s3_bios and s3_mode.
248 s3_beep is for debugging; it makes the PC's speaker beep
249 as soon as the kernel's real-mode entry point is called.
250 s4_hwsig causes the kernel to check the ACPI hardware
251 signature during resume from hibernation, and gracefully
252 refuse to resume if it has changed. This complies with
253 the ACPI specification but not with reality, since
254 Windows does not do this and many laptops do change it
255 on docking. So the default behaviour is to allow resume
256 and simply warn when the signature changes, unless the
257 s4_hwsig option is enabled.
258 s4_nohwsig prevents ACPI hardware signature from being
259 used (or even warned about) during resume.
260 old_ordering causes the ACPI 1.0 ordering of the _PTS
261 control method, with respect to putting devices into
262 low power states, to be enforced (the ACPI 2.0 ordering
263 of _PTS is used by default).
264 nonvs prevents the kernel from saving/restoring the
265 ACPI NVS memory during suspend/hibernation and resume.
266 sci_force_enable causes the kernel to set SCI_EN directly
267 on resume from S1/S3 (which is against the ACPI spec,
268 but some broken systems don't work without it).
269 nobl causes the internal blacklist of systems known to
270 behave incorrectly in some ways with respect to system
271 suspend and resume to be ignored (use wisely).
273 acpi_use_timer_override [HW,ACPI,EARLY]
274 Use timer override. For some broken Nvidia NF5 boards
275 that require a timer override, but don't have HPET
277 add_efi_memmap [EFI,X86,EARLY] Include EFI memory map in
278 kernel's map of available physical RAM.
280 agp= [AGP]
281 { off | try_unsupported }
282 off: disable AGP support
283 try_unsupported: try to drive unsupported chipsets
284 (may crash computer or cause data corruption)
286 ALSA [HW,ALSA]
287 See Documentation/sound/alsa-configuration.rst
289 alignment= [KNL,ARM]
290 Allow the default userspace alignment fault handler
291 behaviour to be specified. Bit 0 enables warnings,
292 bit 1 enables fixups, and bit 2 sends a segfault.
294 align_va_addr= [X86-64]
295 Align virtual addresses by clearing slice [14:12] when
296 allocating a VMA at process creation time. This option
297 gives you up to 3% performance improvement on AMD F15h
298 machines (where it is enabled by default) for a
299 CPU-intensive style benchmark, and it can vary highly in
300 a microbenchmark depending on workload and compiler.
302 32: only for 32-bit processes
303 64: only for 64-bit processes
304 on: enable for both 32- and 64-bit processes
305 off: disable for both 32- and 64-bit processes
307 alloc_snapshot [FTRACE]
308 Allocate the ftrace snapshot buffer on boot up when the
309 main buffer is allocated. This is handy if debugging
310 and you need to use tracing_snapshot() on boot up, and
311 do not want to use tracing_snapshot_alloc() as it needs
312 to be done where GFP_KERNEL allocations are allowed.
314 allow_mismatched_32bit_el0 [ARM64,EARLY]
315 Allow execve() of 32-bit applications and setting of the
316 PER_LINUX32 personality on systems where only a strict
317 subset of the CPUs support 32-bit EL0. When this
318 parameter is present, the set of CPUs supporting 32-bit
319 EL0 is indicated by /sys/devices/system/cpu/aarch32_el0
320 and hot-unplug operations may be restricted.
322 See Documentation/arch/arm64/asymmetric-32bit.rst for more
323 information.
325 amd_iommu= [HW,X86-64]
326 Pass parameters to the AMD IOMMU driver in the system.
327 Possible values are:
328 fullflush - Deprecated, equivalent to iommu.strict=1
329 off - do not initialize any AMD IOMMU found in
330 the system
331 force_isolation - Force device isolation for all
332 devices. The IOMMU driver is not
333 allowed anymore to lift isolation
334 requirements as needed. This option
335 does not override iommu=pt
336 force_enable - Force enable the IOMMU on platforms known
337 to be buggy with IOMMU enabled. Use this
338 option with care.
339 pgtbl_v1 - Use v1 page table for DMA-API (Default).
340 pgtbl_v2 - Use v2 page table for DMA-API.
341 irtcachedis - Disable Interrupt Remapping Table (IRT) caching.
342 nohugepages - Limit page-sizes used for v1 page-tables
343 to 4 KiB.
344 v2_pgsizes_only - Limit page-sizes used for v1 page-tables
345 to 4KiB/2Mib/1GiB.
348 amd_iommu_dump= [HW,X86-64]
349 Enable AMD IOMMU driver option to dump the ACPI table
350 for AMD IOMMU. With this option enabled, AMD IOMMU
351 driver will print ACPI tables for AMD IOMMU during
352 IOMMU initialization.
354 amd_iommu_intr= [HW,X86-64]
355 Specifies one of the following AMD IOMMU interrupt
356 remapping modes:
357 legacy - Use legacy interrupt remapping mode.
358 vapic - Use virtual APIC mode, which allows IOMMU
359 to inject interrupts directly into guest.
360 This mode requires kvm-amd.avic=1.
361 (Default when IOMMU HW support is present.)
363 amd_pstate= [X86,EARLY]
364 disable
365 Do not enable amd_pstate as the default
366 scaling driver for the supported processors
367 passive
368 Use amd_pstate with passive mode as a scaling driver.
369 In this mode autonomous selection is disabled.
370 Driver requests a desired performance level and platform
371 tries to match the same performance level if it is
372 satisfied by guaranteed performance level.
373 active
374 Use amd_pstate_epp driver instance as the scaling driver,
375 driver provides a hint to the hardware if software wants
376 to bias toward performance (0x0) or energy efficiency (0xff)
377 to the CPPC firmware. then CPPC power algorithm will
378 calculate the runtime workload and adjust the realtime cores
379 frequency.
380 guided
381 Activate guided autonomous mode. Driver requests minimum and
382 maximum performance level and the platform autonomously
383 selects a performance level in this range and appropriate
384 to the current workload.
386 amd_prefcore=
387 [X86]
388 disable
389 Disable amd-pstate preferred core.
391 amijoy.map= [HW,JOY] Amiga joystick support
392 Map of devices attached to JOY0DAT and JOY1DAT
393 Format: <a>,<b>
394 See also Documentation/input/joydev/joystick.rst
396 analog.map= [HW,JOY] Analog joystick and gamepad support
397 Specifies type or capabilities of an analog joystick
398 connected to one of 16 gameports
399 Format: <type1>,<type2>,..<type16>
401 apc= [HW,SPARC]
402 Power management functions (SPARCstation-4/5 + deriv.)
403 Format: noidle
404 Disable APC CPU standby support. SPARCstation-Fox does
405 not play well with APC CPU idle - disable it if you have
406 APC and your system crashes randomly.
408 apic= [APIC,X86,EARLY] Advanced Programmable Interrupt Controller
409 Change the output verbosity while booting
410 Format: { quiet (default) | verbose | debug }
411 Change the amount of debugging information output
412 when initialising the APIC and IO-APIC components.
413 For X86-32, this can also be used to specify an APIC
414 driver name.
415 Format: apic=driver_name
416 Examples: apic=bigsmp
418 apic_extnmi= [APIC,X86,EARLY] External NMI delivery setting
419 Format: { bsp (default) | all | none }
420 bsp: External NMI is delivered only to CPU 0
421 all: External NMIs are broadcast to all CPUs as a
422 backup of CPU 0
423 none: External NMI is masked for all CPUs. This is
424 useful so that a dump capture kernel won't be
425 shot down by NMI
427 autoconf= [IPV6]
428 See Documentation/networking/ipv6.rst.
430 apm= [APM] Advanced Power Management
431 See header of arch/x86/kernel/apm_32.c.
433 apparmor= [APPARMOR] Disable or enable AppArmor at boot time
434 Format: { "0" | "1" }
435 See security/apparmor/Kconfig help text
436 0 -- disable.
437 1 -- enable.
438 Default value is set via kernel config option.
440 arcrimi= [HW,NET] ARCnet - "RIM I" (entirely mem-mapped) cards
441 Format: <io>,<irq>,<nodeID>
443 arm64.no32bit_el0 [ARM64] Unconditionally disable the execution of
444 32 bit applications.
446 arm64.nobti [ARM64] Unconditionally disable Branch Target
447 Identification support
449 arm64.nogcs [ARM64] Unconditionally disable Guarded Control Stack
450 support
452 arm64.nomops [ARM64] Unconditionally disable Memory Copy and Memory
453 Set instructions support
455 arm64.nomte [ARM64] Unconditionally disable Memory Tagging Extension
456 support
458 arm64.nopauth [ARM64] Unconditionally disable Pointer Authentication
459 support
461 arm64.nosme [ARM64] Unconditionally disable Scalable Matrix
462 Extension support
464 arm64.nosve [ARM64] Unconditionally disable Scalable Vector
465 Extension support
467 ataflop= [HW,M68k]
469 atarimouse= [HW,MOUSE] Atari Mouse
471 atkbd.extra= [HW] Enable extra LEDs and keys on IBM RapidAccess,
472 EzKey and similar keyboards
474 atkbd.reset= [HW] Reset keyboard during initialization
476 atkbd.set= [HW] Select keyboard code set
477 Format: <int> (2 = AT (default), 3 = PS/2)
479 atkbd.scroll= [HW] Enable scroll wheel on MS Office and similar
480 keyboards
482 atkbd.softraw= [HW] Choose between synthetic and real raw mode
483 Format: <bool> (0 = real, 1 = synthetic (default))
485 atkbd.softrepeat= [HW]
486 Use software keyboard repeat
488 audit= [KNL] Enable the audit sub-system
489 Format: { "0" | "1" | "off" | "on" }
490 0 | off - kernel audit is disabled and can not be
491 enabled until the next reboot
492 unset - kernel audit is initialized but disabled and
493 will be fully enabled by the userspace auditd.
494 1 | on - kernel audit is initialized and partially
495 enabled, storing at most audit_backlog_limit
496 messages in RAM until it is fully enabled by the
497 userspace auditd.
498 Default: unset
500 audit_backlog_limit= [KNL] Set the audit queue size limit.
501 Format: <int> (must be >=0)
502 Default: 64
504 bau= [X86_UV] Enable the BAU on SGI UV. The default
505 behavior is to disable the BAU (i.e. bau=0).
506 Format: { "0" | "1" }
507 0 - Disable the BAU.
508 1 - Enable the BAU.
509 unset - Disable the BAU.
511 baycom_epp= [HW,AX25]
512 Format: <io>,<mode>
514 baycom_par= [HW,AX25] BayCom Parallel Port AX.25 Modem
515 Format: <io>,<mode>
516 See header of drivers/net/hamradio/baycom_par.c.
518 baycom_ser_fdx= [HW,AX25]
519 BayCom Serial Port AX.25 Modem (Full Duplex Mode)
520 Format: <io>,<irq>,<mode>[,<baud>]
521 See header of drivers/net/hamradio/baycom_ser_fdx.c.
523 baycom_ser_hdx= [HW,AX25]
524 BayCom Serial Port AX.25 Modem (Half Duplex Mode)
525 Format: <io>,<irq>,<mode>
526 See header of drivers/net/hamradio/baycom_ser_hdx.c.
528 bdev_allow_write_mounted=
529 Format: <bool>
530 Control the ability to open a mounted block device
531 for writing, i.e., allow / disallow writes that bypass
532 the FS. This was implemented as a means to prevent
533 fuzzers from crashing the kernel by overwriting the
534 metadata underneath a mounted FS without its awareness.
535 This also prevents destructive formatting of mounted
536 filesystems by naive storage tooling that don't use
537 O_EXCL. Default is Y and can be changed through the
538 Kconfig option CONFIG_BLK_DEV_WRITE_MOUNTED.
540 bert_disable [ACPI]
541 Disable BERT OS support on buggy BIOSes.
543 bgrt_disable [ACPI,X86,EARLY]
544 Disable BGRT to avoid flickering OEM logo.
546 blkdevparts= Manual partition parsing of block device(s) for
547 embedded devices based on command line input.
548 See Documentation/block/cmdline-partition.rst
550 boot_delay= [KNL,EARLY]
551 Milliseconds to delay each printk during boot.
552 Only works if CONFIG_BOOT_PRINTK_DELAY is enabled,
553 and you may also have to specify "lpj=". Boot_delay
554 values larger than 10 seconds (10000) are assumed
555 erroneous and ignored.
556 Format: integer
558 bootconfig [KNL,EARLY]
559 Extended command line options can be added to an initrd
560 and this will cause the kernel to look for it.
562 See Documentation/admin-guide/bootconfig.rst
564 bttv.card= [HW,V4L] bttv (bt848 + bt878 based grabber cards)
565 bttv.radio= Most important insmod options are available as
566 kernel args too.
567 bttv.pll= See Documentation/admin-guide/media/bttv.rst
568 bttv.tuner=
570 bulk_remove=off [PPC] This parameter disables the use of the pSeries
571 firmware feature for flushing multiple hpte entries
572 at a time.
574 c101= [NET] Moxa C101 synchronous serial card
576 cachesize= [BUGS=X86-32] Override level 2 CPU cache size detection.
577 Sometimes CPU hardware bugs make them report the cache
578 size incorrectly. The kernel will attempt work arounds
579 to fix known problems, but for some CPUs it is not
580 possible to determine what the correct size should be.
581 This option provides an override for these situations.
583 carrier_timeout=
584 [NET] Specifies amount of time (in seconds) that
585 the kernel should wait for a network carrier. By default
586 it waits 120 seconds.
588 ca_keys= [KEYS] This parameter identifies a specific key(s) on
589 the system trusted keyring to be used for certificate
590 trust validation.
591 format: { id:<keyid> | builtin }
593 cca= [MIPS,EARLY] Override the kernel pages' cache coherency
594 algorithm. Accepted values range from 0 to 7
595 inclusive. See arch/mips/include/asm/pgtable-bits.h
596 for platform specific values (SB1, Loongson3 and
597 others).
599 ccw_timeout_log [S390]
600 See Documentation/arch/s390/common_io.rst for details.
602 cgroup_disable= [KNL] Disable a particular controller or optional feature
603 Format: {name of the controller(s) or feature(s) to disable}
604 The effects of cgroup_disable=foo are:
605 - foo isn't auto-mounted if you mount all cgroups in
606 a single hierarchy
607 - foo isn't visible as an individually mountable
608 subsystem
609 - if foo is an optional feature then the feature is
610 disabled and corresponding cgroup files are not
611 created
612 {Currently only "memory" controller deal with this and
613 cut the overhead, others just disable the usage. So
614 only cgroup_disable=memory is actually worthy}
615 Specifying "pressure" disables per-cgroup pressure
616 stall information accounting feature
618 cgroup_no_v1= [KNL] Disable cgroup controllers and named hierarchies in v1
619 Format: { { controller | "all" | "named" }
620 [,{ controller | "all" | "named" }...] }
621 Like cgroup_disable, but only applies to cgroup v1;
622 the blacklisted controllers remain available in cgroup2.
623 "all" blacklists all controllers and "named" disables
624 named mounts. Specifying both "all" and "named" disables
625 all v1 hierarchies.
627 cgroup_favordynmods= [KNL] Enable or Disable favordynmods.
628 Format: { "true" | "false" }
629 Defaults to the value of CONFIG_CGROUP_FAVOR_DYNMODS.
631 cgroup.memory= [KNL] Pass options to the cgroup memory controller.
632 Format: <string>
633 nosocket -- Disable socket memory accounting.
634 nokmem -- Disable kernel memory accounting.
635 nobpf -- Disable BPF memory accounting.
637 checkreqprot= [SELINUX] Set initial checkreqprot flag value.
638 Format: { "0" | "1" }
639 See security/selinux/Kconfig help text.
640 0 -- check protection applied by kernel (includes
641 any implied execute protection).
642 1 -- check protection requested by application.
643 Default value is set via a kernel config option.
644 Value can be changed at runtime via
645 /sys/fs/selinux/checkreqprot.
646 Setting checkreqprot to 1 is deprecated.
648 cio_ignore= [S390]
649 See Documentation/arch/s390/common_io.rst for details.
651 clearcpuid=X[,X...] [X86]
652 Disable CPUID feature X for the kernel. See
653 arch/x86/include/asm/cpufeatures.h for the valid bit
654 numbers X. Note the Linux-specific bits are not necessarily
655 stable over kernel options, but the vendor-specific
656 ones should be.
657 X can also be a string as appearing in the flags: line
658 in /proc/cpuinfo which does not have the above
659 instability issue. However, not all features have names
660 in /proc/cpuinfo.
661 Note that using this option will taint your kernel.
662 Also note that user programs calling CPUID directly
663 or using the feature without checking anything
664 will still see it. This just prevents it from
665 being used by the kernel or shown in /proc/cpuinfo.
666 Also note the kernel might malfunction if you disable
667 some critical bits.
669 clk_ignore_unused
670 [CLK]
671 Prevents the clock framework from automatically gating
672 clocks that have not been explicitly enabled by a Linux
673 device driver but are enabled in hardware at reset or
674 by the bootloader/firmware. Note that this does not
675 force such clocks to be always-on nor does it reserve
676 those clocks in any way. This parameter is useful for
677 debug and development, but should not be needed on a
678 platform with proper driver support. For more
679 information, see Documentation/driver-api/clk.rst.
681 clock= [BUGS=X86-32, HW] gettimeofday clocksource override.
682 [Deprecated]
683 Forces specified clocksource (if available) to be used
684 when calculating gettimeofday(). If specified
685 clocksource is not available, it defaults to PIT.
686 Format: { pit | tsc | cyclone | pmtmr }
688 clocksource= Override the default clocksource
689 Format: <string>
690 Override the default clocksource and use the clocksource
691 with the name specified.
692 Some clocksource names to choose from, depending on
693 the platform:
694 [all] jiffies (this is the base, fallback clocksource)
695 [ACPI] acpi_pm
696 [ARM] imx_timer1,OSTS,netx_timer,mpu_timer2,
697 pxa_timer,timer3,32k_counter,timer0_1
698 [X86-32] pit,hpet,tsc;
699 scx200_hrt on Geode; cyclone on IBM x440
700 [MIPS] MIPS
701 [PARISC] cr16
702 [S390] tod
703 [SH] SuperH
704 [SPARC64] tick
705 [X86-64] hpet,tsc
707 clocksource.arm_arch_timer.evtstrm=
708 [ARM,ARM64,EARLY]
709 Format: <bool>
710 Enable/disable the eventstream feature of the ARM
711 architected timer so that code using WFE-based polling
712 loops can be debugged more effectively on production
713 systems.
715 clocksource.verify_n_cpus= [KNL]
716 Limit the number of CPUs checked for clocksources
717 marked with CLOCK_SOURCE_VERIFY_PERCPU that
718 are marked unstable due to excessive skew.
719 A negative value says to check all CPUs, while
720 zero says not to check any. Values larger than
721 nr_cpu_ids are silently truncated to nr_cpu_ids.
722 The actual CPUs are chosen randomly, with
723 no replacement if the same CPU is chosen twice.
725 clocksource-wdtest.holdoff= [KNL]
726 Set the time in seconds that the clocksource
727 watchdog test waits before commencing its tests.
728 Defaults to zero when built as a module and to
729 10 seconds when built into the kernel.
731 cma=nn[MG]@[start[MG][-end[MG]]]
732 [KNL,CMA,EARLY]
733 Sets the size of kernel global memory area for
734 contiguous memory allocations and optionally the
735 placement constraint by the physical address range of
736 memory allocations. A value of 0 disables CMA
737 altogether. For more information, see
738 kernel/dma/contiguous.c
740 cma_pernuma=nn[MG]
741 [KNL,CMA,EARLY]
742 Sets the size of kernel per-numa memory area for
743 contiguous memory allocations. A value of 0 disables
744 per-numa CMA altogether. And If this option is not
745 specified, the default value is 0.
746 With per-numa CMA enabled, DMA users on node nid will
747 first try to allocate buffer from the pernuma area
748 which is located in node nid, if the allocation fails,
749 they will fallback to the global default memory area.
751 numa_cma=<node>:nn[MG][,<node>:nn[MG]]
752 [KNL,CMA,EARLY]
753 Sets the size of kernel numa memory area for
754 contiguous memory allocations. It will reserve CMA
755 area for the specified node.
757 With numa CMA enabled, DMA users on node nid will
758 first try to allocate buffer from the numa area
759 which is located in node nid, if the allocation fails,
760 they will fallback to the global default memory area.
762 cmo_free_hint= [PPC] Format: { yes | no }
763 Specify whether pages are marked as being inactive
764 when they are freed. This is used in CMO environments
765 to determine OS memory pressure for page stealing by
766 a hypervisor.
767 Default: yes
769 coherent_pool=nn[KMG] [ARM,KNL,EARLY]
770 Sets the size of memory pool for coherent, atomic dma
771 allocations, by default set to 256K.
773 com20020= [HW,NET] ARCnet - COM20020 chipset
774 Format:
775 <io>[,<irq>[,<nodeID>[,<backplane>[,<ckp>[,<timeout>]]]]]
777 com90io= [HW,NET] ARCnet - COM90xx chipset (IO-mapped buffers)
778 Format: <io>[,<irq>]
780 com90xx= [HW,NET]
781 ARCnet - COM90xx chipset (memory-mapped buffers)
782 Format: <io>[,<irq>[,<memstart>]]
784 condev= [HW,S390] console device
785 conmode=
787 con3215_drop= [S390,EARLY] 3215 console drop mode.
788 Format: y|n|Y|N|1|0
789 When set to true, drop data on the 3215 console when
790 the console buffer is full. In this case the
791 operator using a 3270 terminal emulator (for example
792 x3270) does not have to enter the clear key for the
793 console output to advance and the kernel to continue.
794 This leads to a much faster boot time when a 3270
795 terminal emulator is active. If no 3270 terminal
796 emulator is used, this parameter has no effect.
798 console= [KNL] Output console device and options.
800 tty<n> Use the virtual console device <n>.
802 ttyS<n>[,options]
803 ttyUSB0[,options]
804 Use the specified serial port. The options are of
805 the form "bbbbpnf", where "bbbb" is the baud rate,
806 "p" is parity ("n", "o", or "e"), "n" is number of
807 bits, and "f" is flow control ("r" for RTS or
808 omit it). Default is "9600n8".
810 See Documentation/admin-guide/serial-console.rst for more
811 information. See
812 Documentation/networking/netconsole.rst for an
813 alternative.
815 <DEVNAME>:<n>.<n>[,options]
816 Use the specified serial port on the serial core bus.
817 The addressing uses DEVNAME of the physical serial port
818 device, followed by the serial core controller instance,
819 and the serial port instance. The options are the same
820 as documented for the ttyS addressing above.
822 The mapping of the serial ports to the tty instances
823 can be viewed with:
825 $ ls -d /sys/bus/serial-base/devices/*:*.*/tty/*
826 /sys/bus/serial-base/devices/00:04:0.0/tty/ttyS0
828 In the above example, the console can be addressed with
829 console=00:04:0.0. Note that a console addressed this
830 way will only get added when the related device driver
831 is ready. The use of an earlycon parameter in addition to
832 the console may be desired for console output early on.
834 uart[8250],io,<addr>[,options]
835 uart[8250],mmio,<addr>[,options]
836 uart[8250],mmio16,<addr>[,options]
837 uart[8250],mmio32,<addr>[,options]
838 uart[8250],0x<addr>[,options]
839 Start an early, polled-mode console on the 8250/16550
840 UART at the specified I/O port or MMIO address,
841 switching to the matching ttyS device later.
842 MMIO inter-register address stride is either 8-bit
843 (mmio), 16-bit (mmio16), or 32-bit (mmio32).
844 If none of [io|mmio|mmio16|mmio32], <addr> is assumed
845 to be equivalent to 'mmio'. 'options' are specified in
846 the same format described for ttyS above; if unspecified,
847 the h/w is not re-initialized.
849 hvc<n> Use the hypervisor console device <n>. This is for
850 both Xen and PowerPC hypervisors.
852 { null | "" }
853 Use to disable console output, i.e., to have kernel
854 console messages discarded.
855 This must be the only console= parameter used on the
856 kernel command line.
858 If the device connected to the port is not a TTY but a braille
859 device, prepend "brl," before the device type, for instance
860 console=brl,ttyS0
861 For now, only VisioBraille is supported.
863 console_msg_format=
864 [KNL] Change console messages format
865 default
866 By default we print messages on consoles in
867 "[time stamp] text\n" format (time stamp may not be
868 printed, depending on CONFIG_PRINTK_TIME or
869 `printk_time' param).
870 syslog
871 Switch to syslog format: "<%u>[time stamp] text\n"
872 IOW, each message will have a facility and loglevel
873 prefix. The format is similar to one used by syslog()
874 syscall, or to executing "dmesg -S --raw" or to reading
875 from /proc/kmsg.
877 consoleblank= [KNL] The console blank (screen saver) timeout in
878 seconds. A value of 0 disables the blank timer.
879 Defaults to 0.
881 coredump_filter=
882 [KNL] Change the default value for
883 /proc/<pid>/coredump_filter.
884 See also Documentation/filesystems/proc.rst.
886 coresight_cpu_debug.enable
887 [ARM,ARM64]
888 Format: <bool>
889 Enable/disable the CPU sampling based debugging.
890 0: default value, disable debugging
891 1: enable debugging at boot time
893 cpcihp_generic= [HW,PCI] Generic port I/O CompactPCI driver
894 Format:
895 <first_slot>,<last_slot>,<port>,<enum_bit>[,<debug>]
897 cpuidle.off=1 [CPU_IDLE]
898 disable the cpuidle sub-system
900 cpuidle.governor=
901 [CPU_IDLE] Name of the cpuidle governor to use.
903 cpufreq.off=1 [CPU_FREQ]
904 disable the cpufreq sub-system
906 cpufreq.default_governor=
907 [CPU_FREQ] Name of the default cpufreq governor or
908 policy to use. This governor must be registered in the
909 kernel before the cpufreq driver probes.
911 cpu_init_udelay=N
912 [X86,EARLY] Delay for N microsec between assert and de-assert
913 of APIC INIT to start processors. This delay occurs
914 on every CPU online, such as boot, and resume from suspend.
915 Default: 10000
917 cpuhp.parallel=
918 [SMP] Enable/disable parallel bringup of secondary CPUs
919 Format: <bool>
920 Default is enabled if CONFIG_HOTPLUG_PARALLEL=y. Otherwise
921 the parameter has no effect.
923 crash_kexec_post_notifiers
924 Only jump to kdump kernel after running the panic
925 notifiers and dumping kmsg. This option increases
926 the risks of a kdump failure, since some panic
927 notifiers can make the crashed kernel more unstable.
928 In configurations where kdump may not be reliable,
929 running the panic notifiers could allow collecting
930 more data on dmesg, like stack traces from other CPUS
931 or extra data dumped by panic_print. Note that some
932 configurations enable this option unconditionally,
933 like Hyper-V, PowerPC (fadump) and AMD SEV-SNP.
935 crashkernel=size[KMG][@offset[KMG]]
936 [KNL,EARLY] Using kexec, Linux can switch to a 'crash kernel'
937 upon panic. This parameter reserves the physical
938 memory region [offset, offset + size] for that kernel
939 image. If '@offset' is omitted, then a suitable offset
940 is selected automatically.
941 [KNL, X86-64, ARM64, RISCV, LoongArch] Select a region
942 under 4G first, and fall back to reserve region above
943 4G when '@offset' hasn't been specified.
944 See Documentation/admin-guide/kdump/kdump.rst for further details.
946 crashkernel=range1:size1[,range2:size2,...][@offset]
947 [KNL] Same as above, but depends on the memory
948 in the running system. The syntax of range is
949 start-[end] where start and end are both
950 a memory unit (amount[KMG]). See also
951 Documentation/admin-guide/kdump/kdump.rst for an example.
953 crashkernel=size[KMG],high
954 [KNL, X86-64, ARM64, RISCV, LoongArch] range could be
955 above 4G.
956 Allow kernel to allocate physical memory region from top,
957 so could be above 4G if system have more than 4G ram
958 installed. Otherwise memory region will be allocated
959 below 4G, if available.
960 It will be ignored if crashkernel=X is specified.
961 crashkernel=size[KMG],low
962 [KNL, X86-64, ARM64, RISCV, LoongArch] range under 4G.
963 When crashkernel=X,high is passed, kernel could allocate
964 physical memory region above 4G, that cause second kernel
965 crash on system that require some amount of low memory,
966 e.g. swiotlb requires at least 64M+32K low memory, also
967 enough extra low memory is needed to make sure DMA buffers
968 for 32-bit devices won't run out. Kernel would try to allocate
969 default size of memory below 4G automatically. The default
970 size is platform dependent.
971 --> x86: max(swiotlb_size_or_default() + 8MiB, 256MiB)
972 --> arm64: 128MiB
973 --> riscv: 128MiB
974 --> loongarch: 128MiB
975 This one lets the user specify own low range under 4G
976 for second kernel instead.
977 0: to disable low allocation.
978 It will be ignored when crashkernel=X,high is not used
979 or memory reserved is below 4G.
981 cryptomgr.notests
982 [KNL] Disable crypto self-tests
984 cs89x0_dma= [HW,NET]
985 Format: <dma>
987 cs89x0_media= [HW,NET]
988 Format: { rj45 | aui | bnc }
990 csdlock_debug= [KNL] Enable or disable debug add-ons of cross-CPU
991 function call handling. When switched on,
992 additional debug data is printed to the console
993 in case a hanging CPU is detected, and that
994 CPU is pinged again in order to try to resolve
995 the hang situation. The default value of this
996 option depends on the CSD_LOCK_WAIT_DEBUG_DEFAULT
997 Kconfig option.
999 dasd= [HW,NET]
1000 See header of drivers/s390/block/dasd_devmap.c.
1002 db9.dev[2|3]= [HW,JOY] Multisystem joystick support via parallel port
1003 (one device per port)
1004 Format: <port#>,<type>
1005 See also Documentation/input/devices/joystick-parport.rst
1007 debug [KNL,EARLY] Enable kernel debugging (events log level).
1009 debug_boot_weak_hash
1010 [KNL,EARLY] Enable printing [hashed] pointers early in the
1011 boot sequence. If enabled, we use a weak hash instead
1012 of siphash to hash pointers. Use this option if you are
1013 seeing instances of '(___ptrval___)') and need to see a
1014 value (hashed pointer) instead. Cryptographically
1015 insecure, please do not use on production kernels.
1017 debug_locks_verbose=
1018 [KNL] verbose locking self-tests
1019 Format: <int>
1020 Print debugging info while doing the locking API
1021 self-tests.
1022 Bitmask for the various LOCKTYPE_ tests. Defaults to 0
1023 (no extra messages), setting it to -1 (all bits set)
1024 will print _a_lot_ more information - normally only
1025 useful to lockdep developers.
1027 debug_objects [KNL,EARLY] Enable object debugging
1029 debug_guardpage_minorder=
1030 [KNL,EARLY] When CONFIG_DEBUG_PAGEALLOC is set, this
1031 parameter allows control of the order of pages that will
1032 be intentionally kept free (and hence protected) by the
1033 buddy allocator. Bigger value increase the probability
1034 of catching random memory corruption, but reduce the
1035 amount of memory for normal system use. The maximum
1036 possible value is MAX_PAGE_ORDER/2. Setting this
1037 parameter to 1 or 2 should be enough to identify most
1038 random memory corruption problems caused by bugs in
1039 kernel or driver code when a CPU writes to (or reads
1040 from) a random memory location. Note that there exists
1041 a class of memory corruptions problems caused by buggy
1042 H/W or F/W or by drivers badly programming DMA
1043 (basically when memory is written at bus level and the
1044 CPU MMU is bypassed) which are not detectable by
1045 CONFIG_DEBUG_PAGEALLOC, hence this option will not
1046 help tracking down these problems.
1048 debug_pagealloc=
1049 [KNL,EARLY] When CONFIG_DEBUG_PAGEALLOC is set, this parameter
1050 enables the feature at boot time. By default, it is
1051 disabled and the system will work mostly the same as a
1052 kernel built without CONFIG_DEBUG_PAGEALLOC.
1053 Note: to get most of debug_pagealloc error reports, it's
1054 useful to also enable the page_owner functionality.
1055 on: enable the feature
1057 debugfs= [KNL,EARLY] This parameter enables what is exposed to
1058 userspace and debugfs internal clients.
1059 Format: { on, no-mount, off }
1060 on: All functions are enabled.
1061 no-mount:
1062 Filesystem is not registered but kernel clients can
1063 access APIs and a crashkernel can be used to read
1064 its content. There is nothing to mount.
1065 off: Filesystem is not registered and clients
1066 get a -EPERM as result when trying to register files
1067 or directories within debugfs.
1068 This is equivalent of the runtime functionality if
1069 debugfs was not enabled in the kernel at all.
1070 Default value is set in build-time with a kernel configuration.
1072 debugpat [X86] Enable PAT debugging
1074 default_hugepagesz=
1075 [HW] The size of the default HugeTLB page. This is
1076 the size represented by the legacy /proc/ hugepages
1077 APIs. In addition, this is the default hugetlb size
1078 used for shmget(), mmap() and mounting hugetlbfs
1079 filesystems. If not specified, defaults to the
1080 architecture's default huge page size. Huge page
1081 sizes are architecture dependent. See also
1082 Documentation/admin-guide/mm/hugetlbpage.rst.
1083 Format: size[KMG]
1085 deferred_probe_timeout=
1086 [KNL] Debugging option to set a timeout in seconds for
1087 deferred probe to give up waiting on dependencies to
1088 probe. Only specific dependencies (subsystems or
1089 drivers) that have opted in will be ignored. A timeout
1090 of 0 will timeout at the end of initcalls. If the time
1091 out hasn't expired, it'll be restarted by each
1092 successful driver registration. This option will also
1093 dump out devices still on the deferred probe list after
1094 retrying.
1096 delayacct [KNL] Enable per-task delay accounting
1098 dell_smm_hwmon.ignore_dmi=
1099 [HW] Continue probing hardware even if DMI data
1100 indicates that the driver is running on unsupported
1101 hardware.
1103 dell_smm_hwmon.force=
1104 [HW] Activate driver even if SMM BIOS signature does
1105 not match list of supported models and enable otherwise
1106 blacklisted features.
1108 dell_smm_hwmon.power_status=
1109 [HW] Report power status in /proc/i8k
1110 (disabled by default).
1112 dell_smm_hwmon.restricted=
1113 [HW] Allow controlling fans only if SYS_ADMIN
1114 capability is set.
1116 dell_smm_hwmon.fan_mult=
1117 [HW] Factor to multiply fan speed with.
1119 dell_smm_hwmon.fan_max=
1120 [HW] Maximum configurable fan speed.
1122 dfltcc= [HW,S390]
1123 Format: { on | off | def_only | inf_only | always }
1124 on: s390 zlib hardware support for compression on
1125 level 1 and decompression (default)
1126 off: No s390 zlib hardware support
1127 def_only: s390 zlib hardware support for deflate
1128 only (compression on level 1)
1129 inf_only: s390 zlib hardware support for inflate
1130 only (decompression)
1131 always: Same as 'on' but ignores the selected compression
1132 level always using hardware support (used for debugging)
1134 dhash_entries= [KNL]
1135 Set number of hash buckets for dentry cache.
1137 disable_1tb_segments [PPC,EARLY]
1138 Disables the use of 1TB hash page table segments. This
1139 causes the kernel to fall back to 256MB segments which
1140 can be useful when debugging issues that require an SLB
1141 miss to occur.
1143 disable= [IPV6]
1144 See Documentation/networking/ipv6.rst.
1146 disable_radix [PPC,EARLY]
1147 Disable RADIX MMU mode on POWER9
1149 disable_tlbie [PPC]
1150 Disable TLBIE instruction. Currently does not work
1151 with KVM, with HASH MMU, or with coherent accelerators.
1153 disable_ddw [PPC/PSERIES,EARLY]
1154 Disable Dynamic DMA Window support. Use this
1155 to workaround buggy firmware.
1157 disable_ipv6= [IPV6]
1158 See Documentation/networking/ipv6.rst.
1160 disable_mtrr_cleanup [X86,EARLY]
1161 The kernel tries to adjust MTRR layout from continuous
1162 to discrete, to make X server driver able to add WB
1163 entry later. This parameter disables that.
1165 disable_mtrr_trim [X86, Intel and AMD only,EARLY]
1166 By default the kernel will trim any uncacheable
1167 memory out of your available memory pool based on
1168 MTRR settings. This parameter disables that behavior,
1169 possibly causing your machine to run very slowly.
1171 disable_timer_pin_1 [X86,EARLY]
1172 Disable PIN 1 of APIC timer
1173 Can be useful to work around chipset bugs.
1175 dis_ucode_ldr [X86] Disable the microcode loader.
1177 dma_debug=off If the kernel is compiled with DMA_API_DEBUG support,
1178 this option disables the debugging code at boot.
1180 dma_debug_entries=<number>
1181 This option allows to tune the number of preallocated
1182 entries for DMA-API debugging code. One entry is
1183 required per DMA-API allocation. Use this if the
1184 DMA-API debugging code disables itself because the
1185 architectural default is too low.
1187 dma_debug_driver=<driver_name>
1188 With this option the DMA-API debugging driver
1189 filter feature can be enabled at boot time. Just
1190 pass the driver to filter for as the parameter.
1191 The filter can be disabled or changed to another
1192 driver later using sysfs.
1194 reg_file_data_sampling=
1195 [X86] Controls mitigation for Register File Data
1196 Sampling (RFDS) vulnerability. RFDS is a CPU
1197 vulnerability which may allow userspace to infer
1198 kernel data values previously stored in floating point
1199 registers, vector registers, or integer registers.
1200 RFDS only affects Intel Atom processors.
1202 on: Turns ON the mitigation.
1203 off: Turns OFF the mitigation.
1205 This parameter overrides the compile time default set
1206 by CONFIG_MITIGATION_RFDS. Mitigation cannot be
1207 disabled when other VERW based mitigations (like MDS)
1208 are enabled. In order to disable RFDS mitigation all
1209 VERW based mitigations need to be disabled.
1211 For details see:
1212 Documentation/admin-guide/hw-vuln/reg-file-data-sampling.rst
1214 driver_async_probe= [KNL]
1215 List of driver names to be probed asynchronously. *
1216 matches with all driver names. If * is specified, the
1217 rest of the listed driver names are those that will NOT
1218 match the *.
1219 Format: <driver_name1>,<driver_name2>...
1221 drm.edid_firmware=[<connector>:]<file>[,[<connector>:]<file>]
1222 Broken monitors, graphic adapters, KVMs and EDIDless
1223 panels may send no or incorrect EDID data sets.
1224 This parameter allows to specify an EDID data sets
1225 in the /lib/firmware directory that are used instead.
1226 An EDID data set will only be used for a particular
1227 connector, if its name and a colon are prepended to
1228 the EDID name. Each connector may use a unique EDID
1229 data set by separating the files with a comma. An EDID
1230 data set with no connector name will be used for
1231 any connectors not explicitly specified.
1233 dscc4.setup= [NET]
1235 dt_cpu_ftrs= [PPC,EARLY]
1236 Format: {"off" | "known"}
1237 Control how the dt_cpu_ftrs device-tree binding is
1238 used for CPU feature discovery and setup (if it
1239 exists).
1240 off: Do not use it, fall back to legacy cpu table.
1241 known: Do not pass through unknown features to guests
1242 or userspace, only those that the kernel is aware of.
1244 dump_apple_properties [X86]
1245 Dump name and content of EFI device properties on
1246 x86 Macs. Useful for driver authors to determine
1247 what data is available or for reverse-engineering.
1249 dyndbg[="val"] [KNL,DYNAMIC_DEBUG]
1250 <module>.dyndbg[="val"]
1251 Enable debug messages at boot time. See
1252 Documentation/admin-guide/dynamic-debug-howto.rst
1253 for details.
1255 early_ioremap_debug [KNL,EARLY]
1256 Enable debug messages in early_ioremap support. This
1257 is useful for tracking down temporary early mappings
1258 which are not unmapped.
1260 earlycon= [KNL,EARLY] Output early console device and options.
1262 When used with no options, the early console is
1263 determined by stdout-path property in device tree's
1264 chosen node or the ACPI SPCR table if supported by
1265 the platform.
1267 cdns,<addr>[,options]
1268 Start an early, polled-mode console on a Cadence
1269 (xuartps) serial port at the specified address. Only
1270 supported option is baud rate. If baud rate is not
1271 specified, the serial port must already be setup and
1272 configured.
1274 uart[8250],io,<addr>[,options[,uartclk]]
1275 uart[8250],mmio,<addr>[,options[,uartclk]]
1276 uart[8250],mmio32,<addr>[,options[,uartclk]]
1277 uart[8250],mmio32be,<addr>[,options[,uartclk]]
1278 uart[8250],0x<addr>[,options]
1279 Start an early, polled-mode console on the 8250/16550
1280 UART at the specified I/O port or MMIO address.
1281 MMIO inter-register address stride is either 8-bit
1282 (mmio) or 32-bit (mmio32 or mmio32be).
1283 If none of [io|mmio|mmio32|mmio32be], <addr> is assumed
1284 to be equivalent to 'mmio'. 'options' are specified
1285 in the same format described for "console=ttyS<n>"; if
1286 unspecified, the h/w is not initialized. 'uartclk' is
1287 the uart clock frequency; if unspecified, it is set
1288 to 'BASE_BAUD' * 16.
1290 pl011,<addr>
1291 pl011,mmio32,<addr>
1292 Start an early, polled-mode console on a pl011 serial
1293 port at the specified address. The pl011 serial port
1294 must already be setup and configured. Options are not
1295 yet supported. If 'mmio32' is specified, then only
1296 the driver will use only 32-bit accessors to read/write
1297 the device registers.
1299 liteuart,<addr>
1300 Start an early console on a litex serial port at the
1301 specified address. The serial port must already be
1302 setup and configured. Options are not yet supported.
1304 meson,<addr>
1305 Start an early, polled-mode console on a meson serial
1306 port at the specified address. The serial port must
1307 already be setup and configured. Options are not yet
1308 supported.
1310 msm_serial,<addr>
1311 Start an early, polled-mode console on an msm serial
1312 port at the specified address. The serial port
1313 must already be setup and configured. Options are not
1314 yet supported.
1316 msm_serial_dm,<addr>
1317 Start an early, polled-mode console on an msm serial
1318 dm port at the specified address. The serial port
1319 must already be setup and configured. Options are not
1320 yet supported.
1322 owl,<addr>
1323 Start an early, polled-mode console on a serial port
1324 of an Actions Semi SoC, such as S500 or S900, at the
1325 specified address. The serial port must already be
1326 setup and configured. Options are not yet supported.
1328 rda,<addr>
1329 Start an early, polled-mode console on a serial port
1330 of an RDA Micro SoC, such as RDA8810PL, at the
1331 specified address. The serial port must already be
1332 setup and configured. Options are not yet supported.
1334 sbi
1335 Use RISC-V SBI (Supervisor Binary Interface) for early
1336 console.
1338 smh Use ARM semihosting calls for early console.
1340 s3c2410,<addr>
1341 s3c2412,<addr>
1342 s3c2440,<addr>
1343 s3c6400,<addr>
1344 s5pv210,<addr>
1345 exynos4210,<addr>
1346 Use early console provided by serial driver available
1347 on Samsung SoCs, requires selecting proper type and
1348 a correct base address of the selected UART port. The
1349 serial port must already be setup and configured.
1350 Options are not yet supported.
1352 lantiq,<addr>
1353 Start an early, polled-mode console on a lantiq serial
1354 (lqasc) port at the specified address. The serial port
1355 must already be setup and configured. Options are not
1356 yet supported.
1358 lpuart,<addr>
1359 lpuart32,<addr>
1360 Use early console provided by Freescale LP UART driver
1361 found on Freescale Vybrid and QorIQ LS1021A processors.
1362 A valid base address must be provided, and the serial
1363 port must already be setup and configured.
1365 ec_imx21,<addr>
1366 ec_imx6q,<addr>
1367 Start an early, polled-mode, output-only console on the
1368 Freescale i.MX UART at the specified address. The UART
1369 must already be setup and configured.
1371 ar3700_uart,<addr>
1372 Start an early, polled-mode console on the
1373 Armada 3700 serial port at the specified
1374 address. The serial port must already be setup
1375 and configured. Options are not yet supported.
1377 qcom_geni,<addr>
1378 Start an early, polled-mode console on a Qualcomm
1379 Generic Interface (GENI) based serial port at the
1380 specified address. The serial port must already be
1381 setup and configured. Options are not yet supported.
1383 efifb,[options]
1384 Start an early, unaccelerated console on the EFI
1385 memory mapped framebuffer (if available). On cache
1386 coherent non-x86 systems that use system memory for
1387 the framebuffer, pass the 'ram' option so that it is
1388 mapped with the correct attributes.
1390 linflex,<addr>
1391 Use early console provided by Freescale LINFlexD UART
1392 serial driver for NXP S32V234 SoCs. A valid base
1393 address must be provided, and the serial port must
1394 already be setup and configured.
1396 earlyprintk= [X86,SH,ARM,M68k,S390,UM,EARLY]
1397 earlyprintk=vga
1398 earlyprintk=sclp
1399 earlyprintk=xen
1400 earlyprintk=serial[,ttySn[,baudrate]]
1401 earlyprintk=serial[,0x...[,baudrate]]
1402 earlyprintk=ttySn[,baudrate]
1403 earlyprintk=dbgp[debugController#]
1404 earlyprintk=pciserial[,force],bus:device.function[,baudrate]
1405 earlyprintk=xdbc[xhciController#]
1406 earlyprintk=bios
1408 earlyprintk is useful when the kernel crashes before
1409 the normal console is initialized. It is not enabled by
1410 default because it has some cosmetic problems.
1412 Append ",keep" to not disable it when the real console
1413 takes over.
1415 Only one of vga, serial, or usb debug port can
1416 be used at a time.
1418 Currently only ttyS0 and ttyS1 may be specified by
1419 name. Other I/O ports may be explicitly specified
1420 on some architectures (x86 and arm at least) by
1421 replacing ttySn with an I/O port address, like this:
1422 earlyprintk=serial,0x1008,115200
1423 You can find the port for a given device in
1424 /proc/tty/driver/serial:
1425 2: uart:ST16650V2 port:00001008 irq:18 ...
1427 Interaction with the standard serial driver is not
1428 very good.
1430 The VGA output is eventually overwritten by
1431 the real console.
1433 The xen option can only be used in Xen domains.
1435 The sclp output can only be used on s390.
1437 The bios output can only be used on SuperH.
1439 The optional "force" to "pciserial" enables use of a
1440 PCI device even when its classcode is not of the
1441 UART class.
1443 edac_report= [HW,EDAC] Control how to report EDAC event
1444 Format: {"on" | "off" | "force"}
1445 on: enable EDAC to report H/W event. May be overridden
1446 by other higher priority error reporting module.
1447 off: disable H/W event reporting through EDAC.
1448 force: enforce the use of EDAC to report H/W event.
1449 default: on.
1451 edd= [EDD]
1452 Format: {"off" | "on" | "skip[mbr]"}
1454 efi= [EFI,EARLY]
1455 Format: { "debug", "disable_early_pci_dma",
1456 "nochunk", "noruntime", "nosoftreserve",
1457 "novamap", "no_disable_early_pci_dma" }
1458 debug: enable misc debug output.
1459 disable_early_pci_dma: disable the busmaster bit on all
1460 PCI bridges while in the EFI boot stub.
1461 nochunk: disable reading files in "chunks" in the EFI
1462 boot stub, as chunking can cause problems with some
1463 firmware implementations.
1464 noruntime : disable EFI runtime services support
1465 nosoftreserve: The EFI_MEMORY_SP (Specific Purpose)
1466 attribute may cause the kernel to reserve the
1467 memory range for a memory mapping driver to
1468 claim. Specify efi=nosoftreserve to disable this
1469 reservation and treat the memory by its base type
1470 (i.e. EFI_CONVENTIONAL_MEMORY / "System RAM").
1471 novamap: do not call SetVirtualAddressMap().
1472 no_disable_early_pci_dma: Leave the busmaster bit set
1473 on all PCI bridges while in the EFI boot stub
1475 efi_no_storage_paranoia [EFI,X86,EARLY]
1476 Using this parameter you can use more than 50% of
1477 your efi variable storage. Use this parameter only if
1478 you are really sure that your UEFI does sane gc and
1479 fulfills the spec otherwise your board may brick.
1481 efivar_ssdt= [EFI; X86] Name of an EFI variable that contains an SSDT
1482 that is to be dynamically loaded by Linux. If there are
1483 multiple variables with the same name but with different
1484 vendor GUIDs, all of them will be loaded. See
1485 Documentation/admin-guide/acpi/ssdt-overlays.rst for details.
1488 eisa_irq_edge= [PARISC,HW]
1489 See header of drivers/parisc/eisa.c.
1491 ekgdboc= [X86,KGDB,EARLY] Allow early kernel console debugging
1492 Format: ekgdboc=kbd
1494 This is designed to be used in conjunction with
1495 the boot argument: earlyprintk=vga
1497 This parameter works in place of the kgdboc parameter
1498 but can only be used if the backing tty is available
1499 very early in the boot process. For early debugging
1500 via a serial port see kgdboc_earlycon instead.
1502 elanfreq= [X86-32]
1503 See comment before function elanfreq_setup() in
1504 arch/x86/kernel/cpu/cpufreq/elanfreq.c.
1506 elfcorehdr=[size[KMG]@]offset[KMG] [PPC,SH,X86,S390,EARLY]
1507 Specifies physical address of start of kernel core
1508 image elf header and optionally the size. Generally
1509 kexec loader will pass this option to capture kernel.
1510 See Documentation/admin-guide/kdump/kdump.rst for details.
1512 enable_mtrr_cleanup [X86,EARLY]
1513 The kernel tries to adjust MTRR layout from continuous
1514 to discrete, to make X server driver able to add WB
1515 entry later. This parameter enables that.
1517 enable_timer_pin_1 [X86]
1518 Enable PIN 1 of APIC timer
1519 Can be useful to work around chipset bugs
1520 (in particular on some ATI chipsets).
1521 The kernel tries to set a reasonable default.
1523 enforcing= [SELINUX] Set initial enforcing status.
1524 Format: {"0" | "1"}
1525 See security/selinux/Kconfig help text.
1526 0 -- permissive (log only, no denials).
1527 1 -- enforcing (deny and log).
1528 Default value is 0.
1529 Value can be changed at runtime via
1530 /sys/fs/selinux/enforce.
1532 erst_disable [ACPI]
1533 Disable Error Record Serialization Table (ERST)
1534 support.
1536 ether= [HW,NET] Ethernet cards parameters
1537 This option is obsoleted by the "netdev=" option, which
1538 has equivalent usage. See its documentation for details.
1540 evm= [EVM]
1541 Format: { "fix" }
1542 Permit 'security.evm' to be updated regardless of
1543 current integrity status.
1545 early_page_ext [KNL,EARLY] Enforces page_ext initialization to earlier
1546 stages so cover more early boot allocations.
1547 Please note that as side effect some optimizations
1548 might be disabled to achieve that (e.g. parallelized
1549 memory initialization is disabled) so the boot process
1550 might take longer, especially on systems with a lot of
1551 memory. Available with CONFIG_PAGE_EXTENSION=y.
1553 failslab=
1554 fail_usercopy=
1555 fail_page_alloc=
1556 fail_skb_realloc=
1557 fail_make_request=[KNL]
1558 General fault injection mechanism.
1559 Format: <interval>,<probability>,<space>,<times>
1560 See also Documentation/fault-injection/.
1562 fb_tunnels= [NET]
1563 Format: { initns | none }
1564 See Documentation/admin-guide/sysctl/net.rst for
1565 fb_tunnels_only_for_init_ns
1567 floppy= [HW]
1568 See Documentation/admin-guide/blockdev/floppy.rst.
1570 forcepae [X86-32]
1571 Forcefully enable Physical Address Extension (PAE).
1572 Many Pentium M systems disable PAE but may have a
1573 functionally usable PAE implementation.
1574 Warning: use of this parameter will taint the kernel
1575 and may cause unknown problems.
1577 fred= [X86-64]
1578 Enable/disable Flexible Return and Event Delivery.
1579 Format: { on | off }
1580 on: enable FRED when it's present.
1581 off: disable FRED, the default setting.
1583 ftrace=[tracer]
1584 [FTRACE] will set and start the specified tracer
1585 as early as possible in order to facilitate early
1586 boot debugging.
1588 ftrace_boot_snapshot
1589 [FTRACE] On boot up, a snapshot will be taken of the
1590 ftrace ring buffer that can be read at:
1591 /sys/kernel/tracing/snapshot.
1592 This is useful if you need tracing information from kernel
1593 boot up that is likely to be overridden by user space
1594 start up functionality.
1596 Optionally, the snapshot can also be defined for a tracing
1597 instance that was created by the trace_instance= command
1598 line parameter.
1600 trace_instance=foo,sched_switch ftrace_boot_snapshot=foo
1602 The above will cause the "foo" tracing instance to trigger
1603 a snapshot at the end of boot up.
1605 ftrace_dump_on_oops[=2(orig_cpu) | =<instance>][,<instance> |
1606 ,<instance>=2(orig_cpu)]
1607 [FTRACE] will dump the trace buffers on oops.
1608 If no parameter is passed, ftrace will dump global
1609 buffers of all CPUs, if you pass 2 or orig_cpu, it
1610 will dump only the buffer of the CPU that triggered
1611 the oops, or the specific instance will be dumped if
1612 its name is passed. Multiple instance dump is also
1613 supported, and instances are separated by commas. Each
1614 instance supports only dump on CPU that triggered the
1615 oops by passing 2 or orig_cpu to it.
1617 ftrace_dump_on_oops=foo=orig_cpu
1619 The above will dump only the buffer of "foo" instance
1620 on CPU that triggered the oops.
1622 ftrace_dump_on_oops,foo,bar=orig_cpu
1624 The above will dump global buffer on all CPUs, the
1625 buffer of "foo" instance on all CPUs and the buffer
1626 of "bar" instance on CPU that triggered the oops.
1628 ftrace_filter=[function-list]
1629 [FTRACE] Limit the functions traced by the function
1630 tracer at boot up. function-list is a comma-separated
1631 list of functions. This list can be changed at run
1632 time by the set_ftrace_filter file in the debugfs
1633 tracing directory.
1635 ftrace_notrace=[function-list]
1636 [FTRACE] Do not trace the functions specified in
1637 function-list. This list can be changed at run time
1638 by the set_ftrace_notrace file in the debugfs
1639 tracing directory.
1641 ftrace_graph_filter=[function-list]
1642 [FTRACE] Limit the top level callers functions traced
1643 by the function graph tracer at boot up.
1644 function-list is a comma-separated list of functions
1645 that can be changed at run time by the
1646 set_graph_function file in the debugfs tracing directory.
1648 ftrace_graph_notrace=[function-list]
1649 [FTRACE] Do not trace from the functions specified in
1650 function-list. This list is a comma-separated list of
1651 functions that can be changed at run time by the
1652 set_graph_notrace file in the debugfs tracing directory.
1654 ftrace_graph_max_depth=<uint>
1655 [FTRACE] Used with the function graph tracer. This is
1656 the max depth it will trace into a function. This value
1657 can be changed at run time by the max_graph_depth file
1658 in the tracefs tracing directory. default: 0 (no limit)
1660 fw_devlink= [KNL,EARLY] Create device links between consumer and supplier
1661 devices by scanning the firmware to infer the
1662 consumer/supplier relationships. This feature is
1663 especially useful when drivers are loaded as modules as
1664 it ensures proper ordering of tasks like device probing
1665 (suppliers first, then consumers), supplier boot state
1666 clean up (only after all consumers have probed),
1667 suspend/resume & runtime PM (consumers first, then
1668 suppliers).
1669 Format: { off | permissive | on | rpm }
1670 off -- Don't create device links from firmware info.
1671 permissive -- Create device links from firmware info
1672 but use it only for ordering boot state clean
1673 up (sync_state() calls).
1674 on -- Create device links from firmware info and use it
1675 to enforce probe and suspend/resume ordering.
1676 rpm -- Like "on", but also use to order runtime PM.
1678 fw_devlink.strict=<bool>
1679 [KNL,EARLY] Treat all inferred dependencies as mandatory
1680 dependencies. This only applies for fw_devlink=on|rpm.
1681 Format: <bool>
1683 fw_devlink.sync_state =
1684 [KNL,EARLY] When all devices that could probe have finished
1685 probing, this parameter controls what to do with
1686 devices that haven't yet received their sync_state()
1687 calls.
1688 Format: { strict | timeout }
1689 strict -- Default. Continue waiting on consumers to
1690 probe successfully.
1691 timeout -- Give up waiting on consumers and call
1692 sync_state() on any devices that haven't yet
1693 received their sync_state() calls after
1694 deferred_probe_timeout has expired or by
1695 late_initcall() if !CONFIG_MODULES.
1697 gamecon.map[2|3]=
1698 [HW,JOY] Multisystem joystick and NES/SNES/PSX pad
1699 support via parallel port (up to 5 devices per port)
1700 Format: <port#>,<pad1>,<pad2>,<pad3>,<pad4>,<pad5>
1701 See also Documentation/input/devices/joystick-parport.rst
1703 gamma= [HW,DRM]
1705 gart_fix_e820= [X86-64,EARLY] disable the fix e820 for K8 GART
1706 Format: off | on
1707 default: on
1709 gather_data_sampling=
1710 [X86,INTEL,EARLY] Control the Gather Data Sampling (GDS)
1711 mitigation.
1713 Gather Data Sampling is a hardware vulnerability which
1714 allows unprivileged speculative access to data which was
1715 previously stored in vector registers.
1717 This issue is mitigated by default in updated microcode.
1718 The mitigation may have a performance impact but can be
1719 disabled. On systems without the microcode mitigation
1720 disabling AVX serves as a mitigation.
1722 force: Disable AVX to mitigate systems without
1723 microcode mitigation. No effect if the microcode
1724 mitigation is present. Known to cause crashes in
1725 userspace with buggy AVX enumeration.
1727 off: Disable GDS mitigation.
1729 gcov_persist= [GCOV] When non-zero (default), profiling data for
1730 kernel modules is saved and remains accessible via
1731 debugfs, even when the module is unloaded/reloaded.
1732 When zero, profiling data is discarded and associated
1733 debugfs files are removed at module unload time.
1735 goldfish [X86] Enable the goldfish android emulator platform.
1736 Don't use this when you are not running on the
1737 android emulator
1739 gpio-mockup.gpio_mockup_ranges
1740 [HW] Sets the ranges of gpiochip of for this device.
1741 Format: <start1>,<end1>,<start2>,<end2>...
1742 gpio-mockup.gpio_mockup_named_lines
1743 [HW] Let the driver know GPIO lines should be named.
1745 gpt [EFI] Forces disk with valid GPT signature but
1746 invalid Protective MBR to be treated as GPT. If the
1747 primary GPT is corrupted, it enables the backup/alternate
1748 GPT to be used instead.
1750 grcan.enable0= [HW] Configuration of physical interface 0. Determines
1751 the "Enable 0" bit of the configuration register.
1752 Format: 0 | 1
1753 Default: 0
1754 grcan.enable1= [HW] Configuration of physical interface 1. Determines
1755 the "Enable 0" bit of the configuration register.
1756 Format: 0 | 1
1757 Default: 0
1758 grcan.select= [HW] Select which physical interface to use.
1759 Format: 0 | 1
1760 Default: 0
1761 grcan.txsize= [HW] Sets the size of the tx buffer.
1762 Format: <unsigned int> such that (txsize & ~0x1fffc0) == 0.
1763 Default: 1024
1764 grcan.rxsize= [HW] Sets the size of the rx buffer.
1765 Format: <unsigned int> such that (rxsize & ~0x1fffc0) == 0.
1766 Default: 1024
1768 hardened_usercopy=
1769 [KNL] Under CONFIG_HARDENED_USERCOPY, whether
1770 hardening is enabled for this boot. Hardened
1771 usercopy checking is used to protect the kernel
1772 from reading or writing beyond known memory
1773 allocation boundaries as a proactive defense
1774 against bounds-checking flaws in the kernel's
1775 copy_to_user()/copy_from_user() interface.
1776 on Perform hardened usercopy checks (default).
1777 off Disable hardened usercopy checks.
1779 hardlockup_all_cpu_backtrace=
1780 [KNL] Should the hard-lockup detector generate
1781 backtraces on all cpus.
1782 Format: 0 | 1
1784 hashdist= [KNL,NUMA] Large hashes allocated during boot
1785 are distributed across NUMA nodes. Defaults on
1786 for 64-bit NUMA, off otherwise.
1787 Format: 0 | 1 (for off | on)
1789 hd= [EIDE] (E)IDE hard drive subsystem geometry
1790 Format: <cyl>,<head>,<sect>
1792 hest_disable [ACPI]
1793 Disable Hardware Error Source Table (HEST) support;
1794 corresponding firmware-first mode error processing
1795 logic will be disabled.
1797 hibernate= [HIBERNATION]
1798 noresume Don't check if there's a hibernation image
1799 present during boot.
1800 nocompress Don't compress/decompress hibernation images.
1801 no Disable hibernation and resume.
1802 protect_image Turn on image protection during restoration
1803 (that will set all pages holding image data
1804 during restoration read-only).
1806 hibernate.compressor= [HIBERNATION] Compression algorithm to be
1807 used with hibernation.
1808 Format: { lzo | lz4 }
1809 Default: lzo
1811 lzo: Select LZO compression algorithm to
1812 compress/decompress hibernation image.
1814 lz4: Select LZ4 compression algorithm to
1815 compress/decompress hibernation image.
1817 highmem=nn[KMG] [KNL,BOOT,EARLY] forces the highmem zone to have an exact
1818 size of <nn>. This works even on boxes that have no
1819 highmem otherwise. This also works to reduce highmem
1820 size on bigger boxes.
1822 highres= [KNL] Enable/disable high resolution timer mode.
1823 Valid parameters: "on", "off"
1824 Default: "on"
1826 hlt [BUGS=ARM,SH]
1828 hostname= [KNL,EARLY] Set the hostname (aka UTS nodename).
1829 Format: <string>
1830 This allows setting the system's hostname during early
1831 startup. This sets the name returned by gethostname.
1832 Using this parameter to set the hostname makes it
1833 possible to ensure the hostname is correctly set before
1834 any userspace processes run, avoiding the possibility
1835 that a process may call gethostname before the hostname
1836 has been explicitly set, resulting in the calling
1837 process getting an incorrect result. The string must
1838 not exceed the maximum allowed hostname length (usually
1839 64 characters) and will be truncated otherwise.
1841 hpet= [X86-32,HPET] option to control HPET usage
1842 Format: { enable (default) | disable | force |
1843 verbose }
1844 disable: disable HPET and use PIT instead
1845 force: allow force enabled of undocumented chips (ICH4,
1846 VIA, nVidia)
1847 verbose: show contents of HPET registers during setup
1849 hpet_mmap= [X86, HPET_MMAP] Allow userspace to mmap HPET
1850 registers. Default set by CONFIG_HPET_MMAP_DEFAULT.
1852 hugepages= [HW] Number of HugeTLB pages to allocate at boot.
1853 If this follows hugepagesz (below), it specifies
1854 the number of pages of hugepagesz to be allocated.
1855 If this is the first HugeTLB parameter on the command
1856 line, it specifies the number of pages to allocate for
1857 the default huge page size. If using node format, the
1858 number of pages to allocate per-node can be specified.
1859 See also Documentation/admin-guide/mm/hugetlbpage.rst.
1860 Format: <integer> or (node format)
1861 <node>:<integer>[,<node>:<integer>]
1863 hugepagesz=
1864 [HW] The size of the HugeTLB pages. This is used in
1865 conjunction with hugepages (above) to allocate huge
1866 pages of a specific size at boot. The pair
1867 hugepagesz=X hugepages=Y can be specified once for
1868 each supported huge page size. Huge page sizes are
1869 architecture dependent. See also
1870 Documentation/admin-guide/mm/hugetlbpage.rst.
1871 Format: size[KMG]
1873 hugetlb_cma= [HW,CMA,EARLY] The size of a CMA area used for allocation
1874 of gigantic hugepages. Or using node format, the size
1875 of a CMA area per node can be specified.
1876 Format: nn[KMGTPE] or (node format)
1877 <node>:nn[KMGTPE][,<node>:nn[KMGTPE]]
1879 Reserve a CMA area of given size and allocate gigantic
1880 hugepages using the CMA allocator. If enabled, the
1881 boot-time allocation of gigantic hugepages is skipped.
1883 hugetlb_free_vmemmap=
1884 [KNL] Requires CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP
1885 enabled.
1886 Control if HugeTLB Vmemmap Optimization (HVO) is enabled.
1887 Allows heavy hugetlb users to free up some more
1888 memory (7 * PAGE_SIZE for each 2MB hugetlb page).
1889 Format: { on | off (default) }
1891 on: enable HVO
1892 off: disable HVO
1894 Built with CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP_DEFAULT_ON=y,
1895 the default is on.
1897 Note that the vmemmap pages may be allocated from the added
1898 memory block itself when memory_hotplug.memmap_on_memory is
1899 enabled, those vmemmap pages cannot be optimized even if this
1900 feature is enabled. Other vmemmap pages not allocated from
1901 the added memory block itself do not be affected.
1903 hung_task_panic=
1904 [KNL] Should the hung task detector generate panics.
1905 Format: 0 | 1
1907 A value of 1 instructs the kernel to panic when a
1908 hung task is detected. The default value is controlled
1909 by the CONFIG_BOOTPARAM_HUNG_TASK_PANIC build-time
1910 option. The value selected by this boot parameter can
1911 be changed later by the kernel.hung_task_panic sysctl.
1913 hvc_iucv= [S390] Number of z/VM IUCV hypervisor console (HVC)
1914 terminal devices. Valid values: 0..8
1915 hvc_iucv_allow= [S390] Comma-separated list of z/VM user IDs.
1916 If specified, z/VM IUCV HVC accepts connections
1917 from listed z/VM user IDs only.
1919 hv_nopvspin [X86,HYPER_V,EARLY]
1920 Disables the paravirt spinlock optimizations
1921 which allow the hypervisor to 'idle' the guest
1922 on lock contention.
1924 i2c_bus= [HW] Override the default board specific I2C bus speed
1925 or register an additional I2C bus that is not
1926 registered from board initialization code.
1927 Format:
1928 <bus_id>,<clkrate>
1930 i2c_touchscreen_props= [HW,ACPI,X86]
1931 Set device-properties for ACPI-enumerated I2C-attached
1932 touchscreen, to e.g. fix coordinates of upside-down
1933 mounted touchscreens. If you need this option please
1934 submit a drivers/platform/x86/touchscreen_dmi.c patch
1935 adding a DMI quirk for this.
1937 Format:
1938 <ACPI_HW_ID>:<prop_name>=<val>[:prop_name=val][:...]
1939 Where <val> is one of:
1940 Omit "=<val>" entirely Set a boolean device-property
1941 Unsigned number Set a u32 device-property
1942 Anything else Set a string device-property
1944 Examples (split over multiple lines):
1945 i2c_touchscreen_props=GDIX1001:touchscreen-inverted-x:
1946 touchscreen-inverted-y
1948 i2c_touchscreen_props=MSSL1680:touchscreen-size-x=1920:
1949 touchscreen-size-y=1080:touchscreen-inverted-y:
1950 firmware-name=gsl1680-vendor-model.fw:silead,home-button
1952 i8042.debug [HW] Toggle i8042 debug mode
1953 i8042.unmask_kbd_data
1954 [HW] Enable printing of interrupt data from the KBD port
1955 (disabled by default, and as a pre-condition
1956 requires that i8042.debug=1 be enabled)
1957 i8042.direct [HW] Put keyboard port into non-translated mode
1958 i8042.dumbkbd [HW] Pretend that controller can only read data from
1959 keyboard and cannot control its state
1960 (Don't attempt to blink the leds)
1961 i8042.noaux [HW] Don't check for auxiliary (== mouse) port
1962 i8042.nokbd [HW] Don't check/create keyboard port
1963 i8042.noloop [HW] Disable the AUX Loopback command while probing
1964 for the AUX port
1965 i8042.nomux [HW] Don't check presence of an active multiplexing
1966 controller
1967 i8042.nopnp [HW] Don't use ACPIPnP / PnPBIOS to discover KBD/AUX
1968 controllers
1969 i8042.notimeout [HW] Ignore timeout condition signalled by controller
1970 i8042.reset [HW] Reset the controller during init, cleanup and
1971 suspend-to-ram transitions, only during s2r
1972 transitions, or never reset
1973 Format: { 1 | Y | y | 0 | N | n }
1974 1, Y, y: always reset controller
1975 0, N, n: don't ever reset controller
1976 Default: only on s2r transitions on x86; most other
1977 architectures force reset to be always executed
1978 i8042.unlock [HW] Unlock (ignore) the keylock
1979 i8042.kbdreset [HW] Reset device connected to KBD port
1980 i8042.probe_defer
1981 [HW] Allow deferred probing upon i8042 probe errors
1983 i810= [HW,DRM]
1985 i915.invert_brightness=
1986 [DRM] Invert the sense of the variable that is used to
1987 set the brightness of the panel backlight. Normally a
1988 brightness value of 0 indicates backlight switched off,
1989 and the maximum of the brightness value sets the backlight
1990 to maximum brightness. If this parameter is set to 0
1991 (default) and the machine requires it, or this parameter
1992 is set to 1, a brightness value of 0 sets the backlight
1993 to maximum brightness, and the maximum of the brightness
1994 value switches the backlight off.
1995 -1 -- never invert brightness
1996 0 -- machine default
1997 1 -- force brightness inversion
1999 ia32_emulation= [X86-64]
2000 Format: <bool>
2001 When true, allows loading 32-bit programs and executing 32-bit
2002 syscalls, essentially overriding IA32_EMULATION_DEFAULT_DISABLED at
2003 boot time. When false, unconditionally disables IA32 emulation.
2005 icn= [HW,ISDN]
2006 Format: <io>[,<membase>[,<icn_id>[,<icn_id2>]]]
2009 idle= [X86,EARLY]
2010 Format: idle=poll, idle=halt, idle=nomwait
2011 Poll forces a polling idle loop that can slightly
2012 improve the performance of waking up a idle CPU, but
2013 will use a lot of power and make the system run hot.
2014 Not recommended.
2015 idle=halt: Halt is forced to be used for CPU idle.
2016 In such case C2/C3 won't be used again.
2017 idle=nomwait: Disable mwait for CPU C-states
2019 idxd.sva= [HW]
2020 Format: <bool>
2021 Allow force disabling of Shared Virtual Memory (SVA)
2022 support for the idxd driver. By default it is set to
2023 true (1).
2025 idxd.tc_override= [HW]
2026 Format: <bool>
2027 Allow override of default traffic class configuration
2028 for the device. By default it is set to false (0).
2030 ieee754= [MIPS] Select IEEE Std 754 conformance mode
2031 Format: { strict | legacy | 2008 | relaxed | emulated }
2032 Default: strict
2034 Choose which programs will be accepted for execution
2035 based on the IEEE 754 NaN encoding(s) supported by
2036 the FPU and the NaN encoding requested with the value
2037 of an ELF file header flag individually set by each
2038 binary. Hardware implementations are permitted to
2039 support either or both of the legacy and the 2008 NaN
2040 encoding mode.
2042 Available settings are as follows:
2043 strict accept binaries that request a NaN encoding
2044 supported by the FPU
2045 legacy only accept legacy-NaN binaries, if supported
2046 by the FPU
2047 2008 only accept 2008-NaN binaries, if supported
2048 by the FPU
2049 relaxed accept any binaries regardless of whether
2050 supported by the FPU
2051 emulated accept any binaries but enable FPU emulator
2052 if binary mode is unsupported by the FPU.
2054 The FPU emulator is always able to support both NaN
2055 encodings, so if no FPU hardware is present or it has
2056 been disabled with 'nofpu', then the settings of
2057 'legacy' and '2008' strap the emulator accordingly,
2058 'relaxed' straps the emulator for both legacy-NaN and
2059 2008-NaN, whereas 'strict' enables legacy-NaN only on
2060 legacy processors and both NaN encodings on MIPS32 or
2061 MIPS64 CPUs.
2063 The setting for ABS.fmt/NEG.fmt instruction execution
2064 mode generally follows that for the NaN encoding,
2065 except where unsupported by hardware.
2067 ignore_loglevel [KNL,EARLY]
2068 Ignore loglevel setting - this will print /all/
2069 kernel messages to the console. Useful for debugging.
2070 We also add it as printk module parameter, so users
2071 could change it dynamically, usually by
2072 /sys/module/printk/parameters/ignore_loglevel.
2074 ignore_rlimit_data
2075 Ignore RLIMIT_DATA setting for data mappings,
2076 print warning at first misuse. Can be changed via
2077 /sys/module/kernel/parameters/ignore_rlimit_data.
2079 ihash_entries= [KNL]
2080 Set number of hash buckets for inode cache.
2082 ima_appraise= [IMA] appraise integrity measurements
2083 Format: { "off" | "enforce" | "fix" | "log" }
2084 default: "enforce"
2086 ima_appraise_tcb [IMA] Deprecated. Use ima_policy= instead.
2087 The builtin appraise policy appraises all files
2088 owned by uid=0.
2090 ima_canonical_fmt [IMA]
2091 Use the canonical format for the binary runtime
2092 measurements, instead of host native format.
2094 ima_hash= [IMA]
2095 Format: { md5 | sha1 | rmd160 | sha256 | sha384
2096 | sha512 | ... }
2097 default: "sha1"
2099 The list of supported hash algorithms is defined
2100 in crypto/hash_info.h.
2102 ima_policy= [IMA]
2103 The builtin policies to load during IMA setup.
2104 Format: "tcb | appraise_tcb | secure_boot |
2105 fail_securely | critical_data"
2107 The "tcb" policy measures all programs exec'd, files
2108 mmap'd for exec, and all files opened with the read
2109 mode bit set by either the effective uid (euid=0) or
2110 uid=0.
2112 The "appraise_tcb" policy appraises the integrity of
2113 all files owned by root.
2115 The "secure_boot" policy appraises the integrity
2116 of files (eg. kexec kernel image, kernel modules,
2117 firmware, policy, etc) based on file signatures.
2119 The "fail_securely" policy forces file signature
2120 verification failure also on privileged mounted
2121 filesystems with the SB_I_UNVERIFIABLE_SIGNATURE
2122 flag.
2124 The "critical_data" policy measures kernel integrity
2125 critical data.
2127 ima_tcb [IMA] Deprecated. Use ima_policy= instead.
2128 Load a policy which meets the needs of the Trusted
2129 Computing Base. This means IMA will measure all
2130 programs exec'd, files mmap'd for exec, and all files
2131 opened for read by uid=0.
2133 ima_template= [IMA]
2134 Select one of defined IMA measurements template formats.
2135 Formats: { "ima" | "ima-ng" | "ima-ngv2" | "ima-sig" |
2136 "ima-sigv2" }
2137 Default: "ima-ng"
2139 ima_template_fmt=
2140 [IMA] Define a custom template format.
2141 Format: { "field1|...|fieldN" }
2143 ima.ahash_minsize= [IMA] Minimum file size for asynchronous hash usage
2144 Format: <min_file_size>
2145 Set the minimal file size for using asynchronous hash.
2146 If left unspecified, ahash usage is disabled.
2148 ahash performance varies for different data sizes on
2149 different crypto accelerators. This option can be used
2150 to achieve the best performance for a particular HW.
2152 ima.ahash_bufsize= [IMA] Asynchronous hash buffer size
2153 Format: <bufsize>
2154 Set hashing buffer size. Default: 4k.
2156 ahash performance varies for different chunk sizes on
2157 different crypto accelerators. This option can be used
2158 to achieve best performance for particular HW.
2160 init= [KNL]
2161 Format: <full_path>
2162 Run specified binary instead of /sbin/init as init
2163 process.
2165 initcall_debug [KNL] Trace initcalls as they are executed. Useful
2166 for working out where the kernel is dying during
2167 startup.
2169 initcall_blacklist= [KNL] Do not execute a comma-separated list of
2170 initcall functions. Useful for debugging built-in
2171 modules and initcalls.
2173 initramfs_async= [KNL]
2174 Format: <bool>
2175 Default: 1
2176 This parameter controls whether the initramfs
2177 image is unpacked asynchronously, concurrently
2178 with devices being probed and
2179 initialized. This should normally just work,
2180 but as a debugging aid, one can get the
2181 historical behaviour of the initramfs
2182 unpacking being completed before device_ and
2183 late_ initcalls.
2185 initrd= [BOOT,EARLY] Specify the location of the initial ramdisk
2187 initrdmem= [KNL,EARLY] Specify a physical address and size from which to
2188 load the initrd. If an initrd is compiled in or
2189 specified in the bootparams, it takes priority over this
2190 setting.
2191 Format: ss[KMG],nn[KMG]
2192 Default is 0, 0
2194 init_on_alloc= [MM,EARLY] Fill newly allocated pages and heap objects with
2195 zeroes.
2196 Format: 0 | 1
2197 Default set by CONFIG_INIT_ON_ALLOC_DEFAULT_ON.
2199 init_on_free= [MM,EARLY] Fill freed pages and heap objects with zeroes.
2200 Format: 0 | 1
2201 Default set by CONFIG_INIT_ON_FREE_DEFAULT_ON.
2203 init_pkru= [X86] Specify the default memory protection keys rights
2204 register contents for all processes. 0x55555554 by
2205 default (disallow access to all but pkey 0). Can
2206 override in debugfs after boot.
2208 inport.irq= [HW] Inport (ATI XL and Microsoft) busmouse driver
2209 Format: <irq>
2211 int_pln_enable [X86] Enable power limit notification interrupt
2213 integrity_audit=[IMA]
2214 Format: { "0" | "1" }
2215 0 -- basic integrity auditing messages. (Default)
2216 1 -- additional integrity auditing messages.
2218 intel_iommu= [DMAR] Intel IOMMU driver (DMAR) option
2219 on
2220 Enable intel iommu driver.
2221 off
2222 Disable intel iommu driver.
2223 igfx_off [Default Off]
2224 By default, gfx is mapped as normal device. If a gfx
2225 device has a dedicated DMAR unit, the DMAR unit is
2226 bypassed by not enabling DMAR with this option. In
2227 this case, gfx device will use physical address for
2228 DMA.
2229 strict [Default Off]
2230 Deprecated, equivalent to iommu.strict=1.
2231 sp_off [Default Off]
2232 By default, super page will be supported if Intel IOMMU
2233 has the capability. With this option, super page will
2234 not be supported.
2235 sm_on
2236 Enable the Intel IOMMU scalable mode if the hardware
2237 advertises that it has support for the scalable mode
2238 translation.
2239 sm_off
2240 Disallow use of the Intel IOMMU scalable mode.
2241 tboot_noforce [Default Off]
2242 Do not force the Intel IOMMU enabled under tboot.
2243 By default, tboot will force Intel IOMMU on, which
2244 could harm performance of some high-throughput
2245 devices like 40GBit network cards, even if identity
2246 mapping is enabled.
2247 Note that using this option lowers the security
2248 provided by tboot because it makes the system
2249 vulnerable to DMA attacks.
2251 intel_idle.max_cstate= [KNL,HW,ACPI,X86]
2252 0 disables intel_idle and fall back on acpi_idle.
2253 1 to 9 specify maximum depth of C-state.
2255 intel_pstate= [X86,EARLY]
2256 disable
2257 Do not enable intel_pstate as the default
2258 scaling driver for the supported processors
2259 active
2260 Use intel_pstate driver to bypass the scaling
2261 governors layer of cpufreq and provides it own
2262 algorithms for p-state selection. There are two
2263 P-state selection algorithms provided by
2264 intel_pstate in the active mode: powersave and
2265 performance. The way they both operate depends
2266 on whether or not the hardware managed P-states
2267 (HWP) feature has been enabled in the processor
2268 and possibly on the processor model.
2269 passive
2270 Use intel_pstate as a scaling driver, but configure it
2271 to work with generic cpufreq governors (instead of
2272 enabling its internal governor). This mode cannot be
2273 used along with the hardware-managed P-states (HWP)
2274 feature.
2275 force
2276 Enable intel_pstate on systems that prohibit it by default
2277 in favor of acpi-cpufreq. Forcing the intel_pstate driver
2278 instead of acpi-cpufreq may disable platform features, such
2279 as thermal controls and power capping, that rely on ACPI
2280 P-States information being indicated to OSPM and therefore
2281 should be used with caution. This option does not work with
2282 processors that aren't supported by the intel_pstate driver
2283 or on platforms that use pcc-cpufreq instead of acpi-cpufreq.
2284 no_hwp
2285 Do not enable hardware P state control (HWP)
2286 if available.
2287 hwp_only
2288 Only load intel_pstate on systems which support
2289 hardware P state control (HWP) if available.
2290 support_acpi_ppc
2291 Enforce ACPI _PPC performance limits. If the Fixed ACPI
2292 Description Table, specifies preferred power management
2293 profile as "Enterprise Server" or "Performance Server",
2294 then this feature is turned on by default.
2295 per_cpu_perf_limits
2296 Allow per-logical-CPU P-State performance control limits using
2297 cpufreq sysfs interface
2299 intremap= [X86-64,Intel-IOMMU,EARLY]
2300 on enable Interrupt Remapping (default)
2301 off disable Interrupt Remapping
2302 nosid disable Source ID checking
2303 no_x2apic_optout
2304 BIOS x2APIC opt-out request will be ignored
2305 nopost disable Interrupt Posting
2306 posted_msi
2307 enable MSIs delivered as posted interrupts
2309 iomem= Disable strict checking of access to MMIO memory
2310 strict regions from userspace.
2311 relaxed
2313 iommu= [X86,EARLY]
2314 off
2315 force
2316 noforce
2317 biomerge
2318 panic
2319 nopanic
2320 merge
2321 nomerge
2322 soft
2323 pt [X86]
2324 nopt [X86]
2325 nobypass [PPC/POWERNV]
2326 Disable IOMMU bypass, using IOMMU for PCI devices.
2328 iommu.forcedac= [ARM64,X86,EARLY] Control IOVA allocation for PCI devices.
2329 Format: { "0" | "1" }
2330 0 - Try to allocate a 32-bit DMA address first, before
2331 falling back to the full range if needed.
2332 1 - Allocate directly from the full usable range,
2333 forcing Dual Address Cycle for PCI cards supporting
2334 greater than 32-bit addressing.
2336 iommu.strict= [ARM64,X86,S390,EARLY] Configure TLB invalidation behaviour
2337 Format: { "0" | "1" }
2338 0 - Lazy mode.
2339 Request that DMA unmap operations use deferred
2340 invalidation of hardware TLBs, for increased
2341 throughput at the cost of reduced device isolation.
2342 Will fall back to strict mode if not supported by
2343 the relevant IOMMU driver.
2344 1 - Strict mode.
2345 DMA unmap operations invalidate IOMMU hardware TLBs
2346 synchronously.
2347 unset - Use value of CONFIG_IOMMU_DEFAULT_DMA_{LAZY,STRICT}.
2348 Note: on x86, strict mode specified via one of the
2349 legacy driver-specific options takes precedence.
2351 iommu.passthrough=
2352 [ARM64,X86,EARLY] Configure DMA to bypass the IOMMU by default.
2353 Format: { "0" | "1" }
2354 0 - Use IOMMU translation for DMA.
2355 1 - Bypass the IOMMU for DMA.
2356 unset - Use value of CONFIG_IOMMU_DEFAULT_PASSTHROUGH.
2358 io7= [HW] IO7 for Marvel-based Alpha systems
2359 See comment before marvel_specify_io7 in
2360 arch/alpha/kernel/core_marvel.c.
2362 io_delay= [X86,EARLY] I/O delay method
2363 0x80
2364 Standard port 0x80 based delay
2365 0xed
2366 Alternate port 0xed based delay (needed on some systems)
2367 udelay
2368 Simple two microseconds delay
2369 none
2370 No delay
2372 ip= [IP_PNP]
2373 See Documentation/admin-guide/nfs/nfsroot.rst.
2375 ipcmni_extend [KNL,EARLY] Extend the maximum number of unique System V
2376 IPC identifiers from 32,768 to 16,777,216.
2378 ipe.enforce= [IPE]
2379 Format: <bool>
2380 Determine whether IPE starts in permissive (0) or
2381 enforce (1) mode. The default is enforce.
2383 ipe.success_audit=
2384 [IPE]
2385 Format: <bool>
2386 Start IPE with success auditing enabled, emitting
2387 an audit event when a binary is allowed. The default
2388 is 0.
2390 irqaffinity= [SMP] Set the default irq affinity mask
2391 The argument is a cpu list, as described above.
2393 irqchip.gicv2_force_probe=
2394 [ARM,ARM64,EARLY]
2395 Format: <bool>
2396 Force the kernel to look for the second 4kB page
2397 of a GICv2 controller even if the memory range
2398 exposed by the device tree is too small.
2400 irqchip.gicv3_nolpi=
2401 [ARM,ARM64,EARLY]
2402 Force the kernel to ignore the availability of
2403 LPIs (and by consequence ITSs). Intended for system
2404 that use the kernel as a bootloader, and thus want
2405 to let secondary kernels in charge of setting up
2406 LPIs.
2408 irqchip.gicv3_pseudo_nmi= [ARM64,EARLY]
2409 Enables support for pseudo-NMIs in the kernel. This
2410 requires the kernel to be built with
2411 CONFIG_ARM64_PSEUDO_NMI.
2413 irqfixup [HW]
2414 When an interrupt is not handled search all handlers
2415 for it. Intended to get systems with badly broken
2416 firmware running.
2418 irqpoll [HW]
2419 When an interrupt is not handled search all handlers
2420 for it. Also check all handlers each timer
2421 interrupt. Intended to get systems with badly broken
2422 firmware running.
2424 isapnp= [ISAPNP]
2425 Format: <RDP>,<reset>,<pci_scan>,<verbosity>
2427 isolcpus= [KNL,SMP,ISOL] Isolate a given set of CPUs from disturbance.
2428 [Deprecated - use cpusets instead]
2429 Format: [flag-list,]<cpu-list>
2431 Specify one or more CPUs to isolate from disturbances
2432 specified in the flag list (default: domain):
2434 nohz
2435 Disable the tick when a single task runs.
2437 A residual 1Hz tick is offloaded to workqueues, which you
2438 need to affine to housekeeping through the global
2439 workqueue's affinity configured via the
2440 /sys/devices/virtual/workqueue/cpumask sysfs file, or
2441 by using the 'domain' flag described below.
2443 NOTE: by default the global workqueue runs on all CPUs,
2444 so to protect individual CPUs the 'cpumask' file has to
2445 be configured manually after bootup.
2447 domain
2448 Isolate from the general SMP balancing and scheduling
2449 algorithms. Note that performing domain isolation this way
2450 is irreversible: it's not possible to bring back a CPU to
2451 the domains once isolated through isolcpus. It's strongly
2452 advised to use cpusets instead to disable scheduler load
2453 balancing through the "cpuset.sched_load_balance" file.
2454 It offers a much more flexible interface where CPUs can
2455 move in and out of an isolated set anytime.
2457 You can move a process onto or off an "isolated" CPU via
2458 the CPU affinity syscalls or cpuset.
2459 <cpu number> begins at 0 and the maximum value is
2460 "number of CPUs in system - 1".
2462 managed_irq
2464 Isolate from being targeted by managed interrupts
2465 which have an interrupt mask containing isolated
2466 CPUs. The affinity of managed interrupts is
2467 handled by the kernel and cannot be changed via
2468 the /proc/irq/* interfaces.
2470 This isolation is best effort and only effective
2471 if the automatically assigned interrupt mask of a
2472 device queue contains isolated and housekeeping
2473 CPUs. If housekeeping CPUs are online then such
2474 interrupts are directed to the housekeeping CPU
2475 so that IO submitted on the housekeeping CPU
2476 cannot disturb the isolated CPU.
2478 If a queue's affinity mask contains only isolated
2479 CPUs then this parameter has no effect on the
2480 interrupt routing decision, though interrupts are
2481 only delivered when tasks running on those
2482 isolated CPUs submit IO. IO submitted on
2483 housekeeping CPUs has no influence on those
2484 queues.
2486 The format of <cpu-list> is described above.
2488 iucv= [HW,NET]
2490 ivrs_ioapic [HW,X86-64]
2491 Provide an override to the IOAPIC-ID<->DEVICE-ID
2492 mapping provided in the IVRS ACPI table.
2493 By default, PCI segment is 0, and can be omitted.
2495 For example, to map IOAPIC-ID decimal 10 to
2496 PCI segment 0x1 and PCI device 00:14.0,
2497 write the parameter as:
2498 ivrs_ioapic=10@0001:00:14.0
2500 Deprecated formats:
2501 * To map IOAPIC-ID decimal 10 to PCI device 00:14.0
2502 write the parameter as:
2503 ivrs_ioapic[10]=00:14.0
2504 * To map IOAPIC-ID decimal 10 to PCI segment 0x1 and
2505 PCI device 00:14.0 write the parameter as:
2506 ivrs_ioapic[10]=0001:00:14.0
2508 ivrs_hpet [HW,X86-64]
2509 Provide an override to the HPET-ID<->DEVICE-ID
2510 mapping provided in the IVRS ACPI table.
2511 By default, PCI segment is 0, and can be omitted.
2513 For example, to map HPET-ID decimal 10 to
2514 PCI segment 0x1 and PCI device 00:14.0,
2515 write the parameter as:
2516 ivrs_hpet=10@0001:00:14.0
2518 Deprecated formats:
2519 * To map HPET-ID decimal 0 to PCI device 00:14.0
2520 write the parameter as:
2521 ivrs_hpet[0]=00:14.0
2522 * To map HPET-ID decimal 10 to PCI segment 0x1 and
2523 PCI device 00:14.0 write the parameter as:
2524 ivrs_ioapic[10]=0001:00:14.0
2526 ivrs_acpihid [HW,X86-64]
2527 Provide an override to the ACPI-HID:UID<->DEVICE-ID
2528 mapping provided in the IVRS ACPI table.
2529 By default, PCI segment is 0, and can be omitted.
2531 For example, to map UART-HID:UID AMD0020:0 to
2532 PCI segment 0x1 and PCI device ID 00:14.5,
2533 write the parameter as:
2534 ivrs_acpihid=AMD0020:0@0001:00:14.5
2536 Deprecated formats:
2537 * To map UART-HID:UID AMD0020:0 to PCI segment is 0,
2538 PCI device ID 00:14.5, write the parameter as:
2539 ivrs_acpihid[00:14.5]=AMD0020:0
2540 * To map UART-HID:UID AMD0020:0 to PCI segment 0x1 and
2541 PCI device ID 00:14.5, write the parameter as:
2542 ivrs_acpihid[0001:00:14.5]=AMD0020:0
2544 js= [HW,JOY] Analog joystick
2545 See Documentation/input/joydev/joystick.rst.
2547 kasan_multi_shot
2548 [KNL] Enforce KASAN (Kernel Address Sanitizer) to print
2549 report on every invalid memory access. Without this
2550 parameter KASAN will print report only for the first
2551 invalid access.
2553 keep_bootcon [KNL,EARLY]
2554 Do not unregister boot console at start. This is only
2555 useful for debugging when something happens in the window
2556 between unregistering the boot console and initializing
2557 the real console.
2559 keepinitrd [HW,ARM] See retain_initrd.
2561 kernelcore= [KNL,X86,PPC,EARLY]
2562 Format: nn[KMGTPE] | nn% | "mirror"
2563 This parameter specifies the amount of memory usable by
2564 the kernel for non-movable allocations. The requested
2565 amount is spread evenly throughout all nodes in the
2566 system as ZONE_NORMAL. The remaining memory is used for
2567 movable memory in its own zone, ZONE_MOVABLE. In the
2568 event, a node is too small to have both ZONE_NORMAL and
2569 ZONE_MOVABLE, kernelcore memory will take priority and
2570 other nodes will have a larger ZONE_MOVABLE.
2572 ZONE_MOVABLE is used for the allocation of pages that
2573 may be reclaimed or moved by the page migration
2574 subsystem. Note that allocations like PTEs-from-HighMem
2575 still use the HighMem zone if it exists, and the Normal
2576 zone if it does not.
2578 It is possible to specify the exact amount of memory in
2579 the form of "nn[KMGTPE]", a percentage of total system
2580 memory in the form of "nn%", or "mirror". If "mirror"
2581 option is specified, mirrored (reliable) memory is used
2582 for non-movable allocations and remaining memory is used
2583 for Movable pages. "nn[KMGTPE]", "nn%", and "mirror"
2584 are exclusive, so you cannot specify multiple forms.
2586 kgdbdbgp= [KGDB,HW,EARLY] kgdb over EHCI usb debug port.
2587 Format: <Controller#>[,poll interval]
2588 The controller # is the number of the ehci usb debug
2589 port as it is probed via PCI. The poll interval is
2590 optional and is the number seconds in between
2591 each poll cycle to the debug port in case you need
2592 the functionality for interrupting the kernel with
2593 gdb or control-c on the dbgp connection. When
2594 not using this parameter you use sysrq-g to break into
2595 the kernel debugger.
2597 kgdboc= [KGDB,HW] kgdb over consoles.
2598 Requires a tty driver that supports console polling,
2599 or a supported polling keyboard driver (non-usb).
2600 Serial only format: <serial_device>[,baud]
2601 keyboard only format: kbd
2602 keyboard and serial format: kbd,<serial_device>[,baud]
2603 Optional Kernel mode setting:
2604 kms, kbd format: kms,kbd
2605 kms, kbd and serial format: kms,kbd,<ser_dev>[,baud]
2607 kgdboc_earlycon= [KGDB,HW,EARLY]
2608 If the boot console provides the ability to read
2609 characters and can work in polling mode, you can use
2610 this parameter to tell kgdb to use it as a backend
2611 until the normal console is registered. Intended to
2612 be used together with the kgdboc parameter which
2613 specifies the normal console to transition to.
2615 The name of the early console should be specified
2616 as the value of this parameter. Note that the name of
2617 the early console might be different than the tty
2618 name passed to kgdboc. It's OK to leave the value
2619 blank and the first boot console that implements
2620 read() will be picked.
2622 kgdbwait [KGDB,EARLY] Stop kernel execution and enter the
2623 kernel debugger at the earliest opportunity.
2625 kmac= [MIPS] Korina ethernet MAC address.
2626 Configure the RouterBoard 532 series on-chip
2627 Ethernet adapter MAC address.
2629 kmemleak= [KNL,EARLY] Boot-time kmemleak enable/disable
2630 Valid arguments: on, off
2631 Default: on
2632 Built with CONFIG_DEBUG_KMEMLEAK_DEFAULT_OFF=y,
2633 the default is off.
2635 kprobe_event=[probe-list]
2636 [FTRACE] Add kprobe events and enable at boot time.
2637 The probe-list is a semicolon delimited list of probe
2638 definitions. Each definition is same as kprobe_events
2639 interface, but the parameters are comma delimited.
2640 For example, to add a kprobe event on vfs_read with
2641 arg1 and arg2, add to the command line;
2643 kprobe_event=p,vfs_read,$arg1,$arg2
2645 See also Documentation/trace/kprobetrace.rst "Kernel
2646 Boot Parameter" section.
2648 kpti= [ARM64,EARLY] Control page table isolation of
2649 user and kernel address spaces.
2650 Default: enabled on cores which need mitigation.
2651 0: force disabled
2652 1: force enabled
2654 kunit.enable= [KUNIT] Enable executing KUnit tests. Requires
2655 CONFIG_KUNIT to be set to be fully enabled. The
2656 default value can be overridden via
2657 KUNIT_DEFAULT_ENABLED.
2658 Default is 1 (enabled)
2660 kvm.ignore_msrs=[KVM] Ignore guest accesses to unhandled MSRs.
2661 Default is 0 (don't ignore, but inject #GP)
2663 kvm.eager_page_split=
2664 [KVM,X86] Controls whether or not KVM will try to
2665 proactively split all huge pages during dirty logging.
2666 Eager page splitting reduces interruptions to vCPU
2667 execution by eliminating the write-protection faults
2668 and MMU lock contention that would otherwise be
2669 required to split huge pages lazily.
2671 VM workloads that rarely perform writes or that write
2672 only to a small region of VM memory may benefit from
2673 disabling eager page splitting to allow huge pages to
2674 still be used for reads.
2676 The behavior of eager page splitting depends on whether
2677 KVM_DIRTY_LOG_INITIALLY_SET is enabled or disabled. If
2678 disabled, all huge pages in a memslot will be eagerly
2679 split when dirty logging is enabled on that memslot. If
2680 enabled, eager page splitting will be performed during
2681 the KVM_CLEAR_DIRTY ioctl, and only for the pages being
2682 cleared.
2684 Eager page splitting is only supported when kvm.tdp_mmu=Y.
2686 Default is Y (on).
2688 kvm.enable_virt_at_load=[KVM,ARM64,LOONGARCH,MIPS,RISCV,X86]
2689 If enabled, KVM will enable virtualization in hardware
2690 when KVM is loaded, and disable virtualization when KVM
2691 is unloaded (if KVM is built as a module).
2693 If disabled, KVM will dynamically enable and disable
2694 virtualization on-demand when creating and destroying
2695 VMs, i.e. on the 0=>1 and 1=>0 transitions of the
2696 number of VMs.
2698 Enabling virtualization at module lode avoids potential
2699 latency for creation of the 0=>1 VM, as KVM serializes
2700 virtualization enabling across all online CPUs. The
2701 "cost" of enabling virtualization when KVM is loaded,
2702 is that doing so may interfere with using out-of-tree
2703 hypervisors that want to "own" virtualization hardware.
2705 kvm.enable_vmware_backdoor=[KVM] Support VMware backdoor PV interface.
2706 Default is false (don't support).
2708 kvm.nx_huge_pages=
2709 [KVM] Controls the software workaround for the
2710 X86_BUG_ITLB_MULTIHIT bug.
2711 force : Always deploy workaround.
2712 off : Never deploy workaround.
2713 auto : Deploy workaround based on the presence of
2714 X86_BUG_ITLB_MULTIHIT.
2716 Default is 'auto'.
2718 If the software workaround is enabled for the host,
2719 guests do need not to enable it for nested guests.
2721 kvm.nx_huge_pages_recovery_ratio=
2722 [KVM] Controls how many 4KiB pages are periodically zapped
2723 back to huge pages. 0 disables the recovery, otherwise if
2724 the value is N KVM will zap 1/Nth of the 4KiB pages every
2725 period (see below). The default is 60.
2727 kvm.nx_huge_pages_recovery_period_ms=
2728 [KVM] Controls the time period at which KVM zaps 4KiB pages
2729 back to huge pages. If the value is a non-zero N, KVM will
2730 zap a portion (see ratio above) of the pages every N msecs.
2731 If the value is 0 (the default), KVM will pick a period based
2732 on the ratio, such that a page is zapped after 1 hour on average.
2734 kvm-amd.nested= [KVM,AMD] Control nested virtualization feature in
2735 KVM/SVM. Default is 1 (enabled).
2737 kvm-amd.npt= [KVM,AMD] Control KVM's use of Nested Page Tables,
2738 a.k.a. Two-Dimensional Page Tables. Default is 1
2739 (enabled). Disable by KVM if hardware lacks support
2740 for NPT.
2742 kvm-arm.mode=
2743 [KVM,ARM,EARLY] Select one of KVM/arm64's modes of
2744 operation.
2746 none: Forcefully disable KVM.
2748 nvhe: Standard nVHE-based mode, without support for
2749 protected guests.
2751 protected: nVHE-based mode with support for guests whose
2752 state is kept private from the host.
2754 nested: VHE-based mode with support for nested
2755 virtualization. Requires at least ARMv8.3
2756 hardware.
2758 Defaults to VHE/nVHE based on hardware support. Setting
2759 mode to "protected" will disable kexec and hibernation
2760 for the host. "nested" is experimental and should be
2761 used with extreme caution.
2763 kvm-arm.vgic_v3_group0_trap=
2764 [KVM,ARM,EARLY] Trap guest accesses to GICv3 group-0
2765 system registers
2767 kvm-arm.vgic_v3_group1_trap=
2768 [KVM,ARM,EARLY] Trap guest accesses to GICv3 group-1
2769 system registers
2771 kvm-arm.vgic_v3_common_trap=
2772 [KVM,ARM,EARLY] Trap guest accesses to GICv3 common
2773 system registers
2775 kvm-arm.vgic_v4_enable=
2776 [KVM,ARM,EARLY] Allow use of GICv4 for direct
2777 injection of LPIs.
2779 kvm-arm.wfe_trap_policy=
2780 [KVM,ARM] Control when to set WFE instruction trap for
2781 KVM VMs. Traps are allowed but not guaranteed by the
2782 CPU architecture.
2784 trap: set WFE instruction trap
2786 notrap: clear WFE instruction trap
2788 kvm-arm.wfi_trap_policy=
2789 [KVM,ARM] Control when to set WFI instruction trap for
2790 KVM VMs. Traps are allowed but not guaranteed by the
2791 CPU architecture.
2793 trap: set WFI instruction trap
2795 notrap: clear WFI instruction trap
2797 kvm_cma_resv_ratio=n [PPC,EARLY]
2798 Reserves given percentage from system memory area for
2799 contiguous memory allocation for KVM hash pagetable
2800 allocation.
2801 By default it reserves 5% of total system memory.
2802 Format: <integer>
2803 Default: 5
2805 kvm-intel.ept= [KVM,Intel] Control KVM's use of Extended Page Tables,
2806 a.k.a. Two-Dimensional Page Tables. Default is 1
2807 (enabled). Disable by KVM if hardware lacks support
2808 for EPT.
2810 kvm-intel.emulate_invalid_guest_state=
2811 [KVM,Intel] Control whether to emulate invalid guest
2812 state. Ignored if kvm-intel.enable_unrestricted_guest=1,
2813 as guest state is never invalid for unrestricted
2814 guests. This param doesn't apply to nested guests (L2),
2815 as KVM never emulates invalid L2 guest state.
2816 Default is 1 (enabled).
2818 kvm-intel.flexpriority=
2819 [KVM,Intel] Control KVM's use of FlexPriority feature
2820 (TPR shadow). Default is 1 (enabled). Disable by KVM if
2821 hardware lacks support for it.
2823 kvm-intel.nested=
2824 [KVM,Intel] Control nested virtualization feature in
2825 KVM/VMX. Default is 1 (enabled).
2827 kvm-intel.unrestricted_guest=
2828 [KVM,Intel] Control KVM's use of unrestricted guest
2829 feature (virtualized real and unpaged mode). Default
2830 is 1 (enabled). Disable by KVM if EPT is disabled or
2831 hardware lacks support for it.
2833 kvm-intel.vmentry_l1d_flush=[KVM,Intel] Mitigation for L1 Terminal Fault
2834 CVE-2018-3620.
2836 Valid arguments: never, cond, always
2838 always: L1D cache flush on every VMENTER.
2839 cond: Flush L1D on VMENTER only when the code between
2840 VMEXIT and VMENTER can leak host memory.
2841 never: Disables the mitigation
2843 Default is cond (do L1 cache flush in specific instances)
2845 kvm-intel.vpid= [KVM,Intel] Control KVM's use of Virtual Processor
2846 Identification feature (tagged TLBs). Default is 1
2847 (enabled). Disable by KVM if hardware lacks support
2848 for it.
2850 l1d_flush= [X86,INTEL,EARLY]
2851 Control mitigation for L1D based snooping vulnerability.
2853 Certain CPUs are vulnerable to an exploit against CPU
2854 internal buffers which can forward information to a
2855 disclosure gadget under certain conditions.
2857 In vulnerable processors, the speculatively
2858 forwarded data can be used in a cache side channel
2859 attack, to access data to which the attacker does
2860 not have direct access.
2862 This parameter controls the mitigation. The
2863 options are:
2865 on - enable the interface for the mitigation
2867 l1tf= [X86,EARLY] Control mitigation of the L1TF vulnerability on
2868 affected CPUs
2870 The kernel PTE inversion protection is unconditionally
2871 enabled and cannot be disabled.
2873 full
2874 Provides all available mitigations for the
2875 L1TF vulnerability. Disables SMT and
2876 enables all mitigations in the
2877 hypervisors, i.e. unconditional L1D flush.
2879 SMT control and L1D flush control via the
2880 sysfs interface is still possible after
2881 boot. Hypervisors will issue a warning
2882 when the first VM is started in a
2883 potentially insecure configuration,
2884 i.e. SMT enabled or L1D flush disabled.
2886 full,force
2887 Same as 'full', but disables SMT and L1D
2888 flush runtime control. Implies the
2889 'nosmt=force' command line option.
2890 (i.e. sysfs control of SMT is disabled.)
2892 flush
2893 Leaves SMT enabled and enables the default
2894 hypervisor mitigation, i.e. conditional
2895 L1D flush.
2897 SMT control and L1D flush control via the
2898 sysfs interface is still possible after
2899 boot. Hypervisors will issue a warning
2900 when the first VM is started in a
2901 potentially insecure configuration,
2902 i.e. SMT enabled or L1D flush disabled.
2904 flush,nosmt
2906 Disables SMT and enables the default
2907 hypervisor mitigation.
2909 SMT control and L1D flush control via the
2910 sysfs interface is still possible after
2911 boot. Hypervisors will issue a warning
2912 when the first VM is started in a
2913 potentially insecure configuration,
2914 i.e. SMT enabled or L1D flush disabled.
2916 flush,nowarn
2917 Same as 'flush', but hypervisors will not
2918 warn when a VM is started in a potentially
2919 insecure configuration.
2921 off
2922 Disables hypervisor mitigations and doesn't
2923 emit any warnings.
2924 It also drops the swap size and available
2925 RAM limit restriction on both hypervisor and
2926 bare metal.
2928 Default is 'flush'.
2930 For details see: Documentation/admin-guide/hw-vuln/l1tf.rst
2932 l2cr= [PPC]
2934 l3cr= [PPC]
2936 lapic [X86-32,APIC,EARLY] Enable the local APIC even if BIOS
2937 disabled it.
2939 lapic= [X86,APIC] Do not use TSC deadline
2940 value for LAPIC timer one-shot implementation. Default
2941 back to the programmable timer unit in the LAPIC.
2942 Format: notscdeadline
2944 lapic_timer_c2_ok [X86,APIC,EARLY] trust the local apic timer
2945 in C2 power state.
2947 libata.dma= [LIBATA] DMA control
2948 libata.dma=0 Disable all PATA and SATA DMA
2949 libata.dma=1 PATA and SATA Disk DMA only
2950 libata.dma=2 ATAPI (CDROM) DMA only
2951 libata.dma=4 Compact Flash DMA only
2952 Combinations also work, so libata.dma=3 enables DMA
2953 for disks and CDROMs, but not CFs.
2955 libata.ignore_hpa= [LIBATA] Ignore HPA limit
2956 libata.ignore_hpa=0 keep BIOS limits (default)
2957 libata.ignore_hpa=1 ignore limits, using full disk
2959 libata.noacpi [LIBATA] Disables use of ACPI in libata suspend/resume
2960 when set.
2961 Format: <int>
2963 libata.force= [LIBATA] Force configurations. The format is a comma-
2964 separated list of "[ID:]VAL" where ID is PORT[.DEVICE].
2965 PORT and DEVICE are decimal numbers matching port, link
2966 or device. Basically, it matches the ATA ID string
2967 printed on console by libata. If the whole ID part is
2968 omitted, the last PORT and DEVICE values are used. If
2969 ID hasn't been specified yet, the configuration applies
2970 to all ports, links and devices.
2972 If only DEVICE is omitted, the parameter applies to
2973 the port and all links and devices behind it. DEVICE
2974 number of 0 either selects the first device or the
2975 first fan-out link behind PMP device. It does not
2976 select the host link. DEVICE number of 15 selects the
2977 host link and device attached to it.
2979 The VAL specifies the configuration to force. As long
2980 as there is no ambiguity, shortcut notation is allowed.
2981 For example, both 1.5 and 1.5G would work for 1.5Gbps.
2982 The following configurations can be forced.
2984 * Cable type: 40c, 80c, short40c, unk, ign or sata.
2985 Any ID with matching PORT is used.
2987 * SATA link speed limit: 1.5Gbps or 3.0Gbps.
2989 * Transfer mode: pio[0-7], mwdma[0-4] and udma[0-7].
2990 udma[/][16,25,33,44,66,100,133] notation is also
2991 allowed.
2993 * nohrst, nosrst, norst: suppress hard, soft and both
2994 resets.
2996 * rstonce: only attempt one reset during hot-unplug
2997 link recovery.
2999 * [no]dbdelay: Enable or disable the extra 200ms delay
3000 before debouncing a link PHY and device presence
3001 detection.
3003 * [no]ncq: Turn on or off NCQ.
3005 * [no]ncqtrim: Enable or disable queued DSM TRIM.
3007 * [no]ncqati: Enable or disable NCQ trim on ATI chipset.
3009 * [no]trim: Enable or disable (unqueued) TRIM.
3011 * trim_zero: Indicate that TRIM command zeroes data.
3013 * max_trim_128m: Set 128M maximum trim size limit.
3015 * [no]dma: Turn on or off DMA transfers.
3017 * atapi_dmadir: Enable ATAPI DMADIR bridge support.
3019 * atapi_mod16_dma: Enable the use of ATAPI DMA for
3020 commands that are not a multiple of 16 bytes.
3022 * [no]dmalog: Enable or disable the use of the
3023 READ LOG DMA EXT command to access logs.
3025 * [no]iddevlog: Enable or disable access to the
3026 identify device data log.
3028 * [no]logdir: Enable or disable access to the general
3029 purpose log directory.
3031 * max_sec_128: Set transfer size limit to 128 sectors.
3033 * max_sec_1024: Set or clear transfer size limit to
3034 1024 sectors.
3036 * max_sec_lba48: Set or clear transfer size limit to
3037 65535 sectors.
3039 * [no]lpm: Enable or disable link power management.
3041 * [no]setxfer: Indicate if transfer speed mode setting
3042 should be skipped.
3044 * [no]fua: Disable or enable FUA (Force Unit Access)
3045 support for devices supporting this feature.
3047 * dump_id: Dump IDENTIFY data.
3049 * disable: Disable this device.
3051 If there are multiple matching configurations changing
3052 the same attribute, the last one is used.
3054 load_ramdisk= [RAM] [Deprecated]
3056 lockd.nlm_grace_period=P [NFS] Assign grace period.
3057 Format: <integer>
3059 lockd.nlm_tcpport=N [NFS] Assign TCP port.
3060 Format: <integer>
3062 lockd.nlm_timeout=T [NFS] Assign timeout value.
3063 Format: <integer>
3065 lockd.nlm_udpport=M [NFS] Assign UDP port.
3066 Format: <integer>
3068 lockdown= [SECURITY,EARLY]
3069 { integrity | confidentiality }
3070 Enable the kernel lockdown feature. If set to
3071 integrity, kernel features that allow userland to
3072 modify the running kernel are disabled. If set to
3073 confidentiality, kernel features that allow userland
3074 to extract confidential information from the kernel
3075 are also disabled.
3077 locktorture.acq_writer_lim= [KNL]
3078 Set the time limit in jiffies for a lock
3079 acquisition. Acquisitions exceeding this limit
3080 will result in a splat once they do complete.
3082 locktorture.bind_readers= [KNL]
3083 Specify the list of CPUs to which the readers are
3084 to be bound.
3086 locktorture.bind_writers= [KNL]
3087 Specify the list of CPUs to which the writers are
3088 to be bound.
3090 locktorture.call_rcu_chains= [KNL]
3091 Specify the number of self-propagating call_rcu()
3092 chains to set up. These are used to ensure that
3093 there is a high probability of an RCU grace period
3094 in progress at any given time. Defaults to 0,
3095 which disables these call_rcu() chains.
3097 locktorture.long_hold= [KNL]
3098 Specify the duration in milliseconds for the
3099 occasional long-duration lock hold time. Defaults
3100 to 100 milliseconds. Select 0 to disable.
3102 locktorture.nested_locks= [KNL]
3103 Specify the maximum lock nesting depth that
3104 locktorture is to exercise, up to a limit of 8
3105 (MAX_NESTED_LOCKS). Specify zero to disable.
3106 Note that this parameter is ineffective on types
3107 of locks that do not support nested acquisition.
3109 locktorture.nreaders_stress= [KNL]
3110 Set the number of locking read-acquisition kthreads.
3111 Defaults to being automatically set based on the
3112 number of online CPUs.
3114 locktorture.nwriters_stress= [KNL]
3115 Set the number of locking write-acquisition kthreads.
3117 locktorture.onoff_holdoff= [KNL]
3118 Set time (s) after boot for CPU-hotplug testing.
3120 locktorture.onoff_interval= [KNL]
3121 Set time (s) between CPU-hotplug operations, or
3122 zero to disable CPU-hotplug testing.
3124 locktorture.rt_boost= [KNL]
3125 Do periodic testing of real-time lock priority
3126 boosting. Select 0 to disable, 1 to boost
3127 only rt_mutex, and 2 to boost unconditionally.
3128 Defaults to 2, which might seem to be an
3129 odd choice, but which should be harmless for
3130 non-real-time spinlocks, due to their disabling
3131 of preemption. Note that non-realtime mutexes
3132 disable boosting.
3134 locktorture.rt_boost_factor= [KNL]
3135 Number that determines how often and for how
3136 long priority boosting is exercised. This is
3137 scaled down by the number of writers, so that the
3138 number of boosts per unit time remains roughly
3139 constant as the number of writers increases.
3140 On the other hand, the duration of each boost
3141 increases with the number of writers.
3143 locktorture.shuffle_interval= [KNL]
3144 Set task-shuffle interval (jiffies). Shuffling
3145 tasks allows some CPUs to go into dyntick-idle
3146 mode during the locktorture test.
3148 locktorture.shutdown_secs= [KNL]
3149 Set time (s) after boot system shutdown. This
3150 is useful for hands-off automated testing.
3152 locktorture.stat_interval= [KNL]
3153 Time (s) between statistics printk()s.
3155 locktorture.stutter= [KNL]
3156 Time (s) to stutter testing, for example,
3157 specifying five seconds causes the test to run for
3158 five seconds, wait for five seconds, and so on.
3159 This tests the locking primitive's ability to
3160 transition abruptly to and from idle.
3162 locktorture.torture_type= [KNL]
3163 Specify the locking implementation to test.
3165 locktorture.verbose= [KNL]
3166 Enable additional printk() statements.
3168 locktorture.writer_fifo= [KNL]
3169 Run the write-side locktorture kthreads at
3170 sched_set_fifo() real-time priority.
3172 logibm.irq= [HW,MOUSE] Logitech Bus Mouse Driver
3173 Format: <irq>
3175 loglevel= [KNL,EARLY]
3176 All Kernel Messages with a loglevel smaller than the
3177 console loglevel will be printed to the console. It can
3178 also be changed with klogd or other programs. The
3179 loglevels are defined as follows:
3181 0 (KERN_EMERG) system is unusable
3182 1 (KERN_ALERT) action must be taken immediately
3183 2 (KERN_CRIT) critical conditions
3184 3 (KERN_ERR) error conditions
3185 4 (KERN_WARNING) warning conditions
3186 5 (KERN_NOTICE) normal but significant condition
3187 6 (KERN_INFO) informational
3188 7 (KERN_DEBUG) debug-level messages
3190 log_buf_len=n[KMG] [KNL,EARLY]
3191 Sets the size of the printk ring buffer, in bytes.
3192 n must be a power of two and greater than the
3193 minimal size. The minimal size is defined by
3194 LOG_BUF_SHIFT kernel config parameter. There
3195 is also CONFIG_LOG_CPU_MAX_BUF_SHIFT config
3196 parameter that allows to increase the default size
3197 depending on the number of CPUs. See init/Kconfig
3198 for more details.
3200 logo.nologo [FB] Disables display of the built-in Linux logo.
3201 This may be used to provide more screen space for
3202 kernel log messages and is useful when debugging
3203 kernel boot problems.
3205 lp=0 [LP] Specify parallel ports to use, e.g,
3206 lp=port[,port...] lp=none,parport0 (lp0 not configured, lp1 uses
3207 lp=reset first parallel port). 'lp=0' disables the
3208 lp=auto printer driver. 'lp=reset' (which can be
3209 specified in addition to the ports) causes
3210 attached printers to be reset. Using
3211 lp=port1,port2,... specifies the parallel ports
3212 to associate lp devices with, starting with
3213 lp0. A port specification may be 'none' to skip
3214 that lp device, or a parport name such as
3215 'parport0'. Specifying 'lp=auto' instead of a
3216 port specification list means that device IDs
3217 from each port should be examined, to see if
3218 an IEEE 1284-compliant printer is attached; if
3219 so, the driver will manage that printer.
3220 See also header of drivers/char/lp.c.
3222 lpj=n [KNL]
3223 Sets loops_per_jiffy to given constant, thus avoiding
3224 time-consuming boot-time autodetection (up to 250 ms per
3225 CPU). 0 enables autodetection (default). To determine
3226 the correct value for your kernel, boot with normal
3227 autodetection and see what value is printed. Note that
3228 on SMP systems the preset will be applied to all CPUs,
3229 which is likely to cause problems if your CPUs need
3230 significantly divergent settings. An incorrect value
3231 will cause delays in the kernel to be wrong, leading to
3232 unpredictable I/O errors and other breakage. Although
3233 unlikely, in the extreme case this might damage your
3234 hardware.
3236 lsm.debug [SECURITY] Enable LSM initialization debugging output.
3238 lsm=lsm1,...,lsmN
3239 [SECURITY] Choose order of LSM initialization. This
3240 overrides CONFIG_LSM, and the "security=" parameter.
3242 machtype= [Loongson] Share the same kernel image file between
3243 different yeeloong laptops.
3244 Example: machtype=lemote-yeeloong-2f-7inch
3246 maxcpus= [SMP,EARLY] Maximum number of processors that an SMP kernel
3247 will bring up during bootup. maxcpus=n : n >= 0 limits
3248 the kernel to bring up 'n' processors. Surely after
3249 bootup you can bring up the other plugged cpu by executing
3250 "echo 1 > /sys/devices/system/cpu/cpuX/online". So maxcpus
3251 only takes effect during system bootup.
3252 While n=0 is a special case, it is equivalent to "nosmp",
3253 which also disables the IO APIC.
3255 max_loop= [LOOP] The number of loop block devices that get
3256 (loop.max_loop) unconditionally pre-created at init time. The default
3257 number is configured by BLK_DEV_LOOP_MIN_COUNT. Instead
3258 of statically allocating a predefined number, loop
3259 devices can be requested on-demand with the
3260 /dev/loop-control interface.
3262 mce [X86-32] Machine Check Exception
3264 mce=option [X86-64] See Documentation/arch/x86/x86_64/boot-options.rst
3266 md= [HW] RAID subsystems devices and level
3267 See Documentation/admin-guide/md.rst.
3269 mdacon= [MDA]
3270 Format: <first>,<last>
3271 Specifies range of consoles to be captured by the MDA.
3273 mds= [X86,INTEL,EARLY]
3274 Control mitigation for the Micro-architectural Data
3275 Sampling (MDS) vulnerability.
3277 Certain CPUs are vulnerable to an exploit against CPU
3278 internal buffers which can forward information to a
3279 disclosure gadget under certain conditions.
3281 In vulnerable processors, the speculatively
3282 forwarded data can be used in a cache side channel
3283 attack, to access data to which the attacker does
3284 not have direct access.
3286 This parameter controls the MDS mitigation. The
3287 options are:
3289 full - Enable MDS mitigation on vulnerable CPUs
3290 full,nosmt - Enable MDS mitigation and disable
3291 SMT on vulnerable CPUs
3292 off - Unconditionally disable MDS mitigation
3294 On TAA-affected machines, mds=off can be prevented by
3295 an active TAA mitigation as both vulnerabilities are
3296 mitigated with the same mechanism so in order to disable
3297 this mitigation, you need to specify tsx_async_abort=off
3298 too.
3300 Not specifying this option is equivalent to
3301 mds=full.
3303 For details see: Documentation/admin-guide/hw-vuln/mds.rst
3305 mem=nn[KMG] [HEXAGON,EARLY] Set the memory size.
3306 Must be specified, otherwise memory size will be 0.
3308 mem=nn[KMG] [KNL,BOOT,EARLY] Force usage of a specific amount
3309 of memory Amount of memory to be used in cases
3310 as follows:
3312 1 for test;
3313 2 when the kernel is not able to see the whole system memory;
3314 3 memory that lies after 'mem=' boundary is excluded from
3315 the hypervisor, then assigned to KVM guests.
3316 4 to limit the memory available for kdump kernel.
3318 [ARC,MICROBLAZE] - the limit applies only to low memory,
3319 high memory is not affected.
3321 [ARM64] - only limits memory covered by the linear
3322 mapping. The NOMAP regions are not affected.
3324 [X86] Work as limiting max address. Use together
3325 with memmap= to avoid physical address space collisions.
3326 Without memmap= PCI devices could be placed at addresses
3327 belonging to unused RAM.
3329 Note that this only takes effects during boot time since
3330 in above case 3, memory may need be hot added after boot
3331 if system memory of hypervisor is not sufficient.
3333 mem=nn[KMG]@ss[KMG]
3334 [ARM,MIPS,EARLY] - override the memory layout
3335 reported by firmware.
3336 Define a memory region of size nn[KMG] starting at
3337 ss[KMG].
3338 Multiple different regions can be specified with
3339 multiple mem= parameters on the command line.
3341 mem=nopentium [BUGS=X86-32] Disable usage of 4MB pages for kernel
3342 memory.
3344 memblock=debug [KNL,EARLY] Enable memblock debug messages.
3346 memchunk=nn[KMG]
3347 [KNL,SH] Allow user to override the default size for
3348 per-device physically contiguous DMA buffers.
3350 memhp_default_state=online/offline/online_kernel/online_movable
3351 [KNL] Set the initial state for the memory hotplug
3352 onlining policy. If not specified, the default value is
3353 set according to the
3354 CONFIG_MEMORY_HOTPLUG_DEFAULT_ONLINE kernel config
3355 option.
3356 See Documentation/admin-guide/mm/memory-hotplug.rst.
3358 memmap=exactmap [KNL,X86,EARLY] Enable setting of an exact
3359 E820 memory map, as specified by the user.
3360 Such memmap=exactmap lines can be constructed based on
3361 BIOS output or other requirements. See the memmap=nn@ss
3362 option description.
3364 memmap=nn[KMG]@ss[KMG]
3365 [KNL, X86,MIPS,XTENSA,EARLY] Force usage of a specific region of memory.
3366 Region of memory to be used is from ss to ss+nn.
3367 If @ss[KMG] is omitted, it is equivalent to mem=nn[KMG],
3368 which limits max address to nn[KMG].
3369 Multiple different regions can be specified,
3370 comma delimited.
3371 Example:
3372 memmap=100M@2G,100M#3G,1G!1024G
3374 memmap=nn[KMG]#ss[KMG]
3375 [KNL,ACPI,EARLY] Mark specific memory as ACPI data.
3376 Region of memory to be marked is from ss to ss+nn.
3378 memmap=nn[KMG]$ss[KMG]
3379 [KNL,ACPI,EARLY] Mark specific memory as reserved.
3380 Region of memory to be reserved is from ss to ss+nn.
3381 Example: Exclude memory from 0x18690000-0x1869ffff
3382 memmap=64K$0x18690000
3383 or
3384 memmap=0x10000$0x18690000
3385 Some bootloaders may need an escape character before '$',
3386 like Grub2, otherwise '$' and the following number
3387 will be eaten.
3389 memmap=nn[KMG]!ss[KMG,EARLY]
3390 [KNL,X86] Mark specific memory as protected.
3391 Region of memory to be used, from ss to ss+nn.
3392 The memory region may be marked as e820 type 12 (0xc)
3393 and is NVDIMM or ADR memory.
3395 memmap=<size>%<offset>-<oldtype>+<newtype>
3396 [KNL,ACPI,EARLY] Convert memory within the specified region
3397 from <oldtype> to <newtype>. If "-<oldtype>" is left
3398 out, the whole region will be marked as <newtype>,
3399 even if previously unavailable. If "+<newtype>" is left
3400 out, matching memory will be removed. Types are
3401 specified as e820 types, e.g., 1 = RAM, 2 = reserved,
3402 3 = ACPI, 12 = PRAM.
3404 memory_corruption_check=0/1 [X86,EARLY]
3405 Some BIOSes seem to corrupt the first 64k of
3406 memory when doing things like suspend/resume.
3407 Setting this option will scan the memory
3408 looking for corruption. Enabling this will
3409 both detect corruption and prevent the kernel
3410 from using the memory being corrupted.
3411 However, its intended as a diagnostic tool; if
3412 repeatable BIOS-originated corruption always
3413 affects the same memory, you can use memmap=
3414 to prevent the kernel from using that memory.
3416 memory_corruption_check_size=size [X86,EARLY]
3417 By default it checks for corruption in the low
3418 64k, making this memory unavailable for normal
3419 use. Use this parameter to scan for
3420 corruption in more or less memory.
3422 memory_corruption_check_period=seconds [X86,EARLY]
3423 By default it checks for corruption every 60
3424 seconds. Use this parameter to check at some
3425 other rate. 0 disables periodic checking.
3427 memory_hotplug.memmap_on_memory
3428 [KNL,X86,ARM] Boolean flag to enable this feature.
3429 Format: {on | off (default)}
3430 When enabled, runtime hotplugged memory will
3431 allocate its internal metadata (struct pages,
3432 those vmemmap pages cannot be optimized even
3433 if hugetlb_free_vmemmap is enabled) from the
3434 hotadded memory which will allow to hotadd a
3435 lot of memory without requiring additional
3436 memory to do so.
3437 This feature is disabled by default because it
3438 has some implication on large (e.g. GB)
3439 allocations in some configurations (e.g. small
3440 memory blocks).
3441 The state of the flag can be read in
3442 /sys/module/memory_hotplug/parameters/memmap_on_memory.
3443 Note that even when enabled, there are a few cases where
3444 the feature is not effective.
3446 memtest= [KNL,X86,ARM,M68K,PPC,RISCV,EARLY] Enable memtest
3447 Format: <integer>
3448 default : 0 <disable>
3449 Specifies the number of memtest passes to be
3450 performed. Each pass selects another test
3451 pattern from a given set of patterns. Memtest
3452 fills the memory with this pattern, validates
3453 memory contents and reserves bad memory
3454 regions that are detected.
3456 mem_encrypt= [X86-64] AMD Secure Memory Encryption (SME) control
3457 Valid arguments: on, off
3458 Default: off
3459 mem_encrypt=on: Activate SME
3460 mem_encrypt=off: Do not activate SME
3462 Refer to Documentation/virt/kvm/x86/amd-memory-encryption.rst
3463 for details on when memory encryption can be activated.
3465 mem_sleep_default= [SUSPEND] Default system suspend mode:
3466 s2idle - Suspend-To-Idle
3467 shallow - Power-On Suspend or equivalent (if supported)
3468 deep - Suspend-To-RAM or equivalent (if supported)
3469 See Documentation/admin-guide/pm/sleep-states.rst.
3471 mfgptfix [X86-32] Fix MFGPT timers on AMD Geode platforms when
3472 the BIOS has incorrectly applied a workaround. TinyBIOS
3473 version 0.98 is known to be affected, 0.99 fixes the
3474 problem by letting the user disable the workaround.
3476 mga= [HW,DRM]
3478 microcode.force_minrev= [X86]
3479 Format: <bool>
3480 Enable or disable the microcode minimal revision
3481 enforcement for the runtime microcode loader.
3483 mini2440= [ARM,HW,KNL]
3484 Format:[0..2][b][c][t]
3485 Default: "0tb"
3486 MINI2440 configuration specification:
3487 0 - The attached screen is the 3.5" TFT
3488 1 - The attached screen is the 7" TFT
3489 2 - The VGA Shield is attached (1024x768)
3490 Leaving out the screen size parameter will not load
3491 the TFT driver, and the framebuffer will be left
3492 unconfigured.
3493 b - Enable backlight. The TFT backlight pin will be
3494 linked to the kernel VESA blanking code and a GPIO
3495 LED. This parameter is not necessary when using the
3496 VGA shield.
3497 c - Enable the s3c camera interface.
3498 t - Reserved for enabling touchscreen support. The
3499 touchscreen support is not enabled in the mainstream
3500 kernel as of 2.6.30, a preliminary port can be found
3501 in the "bleeding edge" mini2440 support kernel at
3502 https://repo.or.cz/w/linux-2.6/mini2440.git
3504 mitigations=
3505 [X86,PPC,S390,ARM64,EARLY] Control optional mitigations for
3506 CPU vulnerabilities. This is a set of curated,
3507 arch-independent options, each of which is an
3508 aggregation of existing arch-specific options.
3510 Note, "mitigations" is supported if and only if the
3511 kernel was built with CPU_MITIGATIONS=y.
3513 off
3514 Disable all optional CPU mitigations. This
3515 improves system performance, but it may also
3516 expose users to several CPU vulnerabilities.
3517 Equivalent to: if nokaslr then kpti=0 [ARM64]
3518 gather_data_sampling=off [X86]
3519 kvm.nx_huge_pages=off [X86]
3520 l1tf=off [X86]
3521 mds=off [X86]
3522 mmio_stale_data=off [X86]
3523 no_entry_flush [PPC]
3524 no_uaccess_flush [PPC]
3525 nobp=0 [S390]
3526 nopti [X86,PPC]
3527 nospectre_bhb [ARM64]
3528 nospectre_v1 [X86,PPC]
3529 nospectre_v2 [X86,PPC,S390,ARM64]
3530 reg_file_data_sampling=off [X86]
3531 retbleed=off [X86]
3532 spec_rstack_overflow=off [X86]
3533 spec_store_bypass_disable=off [X86,PPC]
3534 spectre_bhi=off [X86]
3535 spectre_v2_user=off [X86]
3536 srbds=off [X86,INTEL]
3537 ssbd=force-off [ARM64]
3538 tsx_async_abort=off [X86]
3540 Exceptions:
3541 This does not have any effect on
3542 kvm.nx_huge_pages when
3543 kvm.nx_huge_pages=force.
3545 auto (default)
3546 Mitigate all CPU vulnerabilities, but leave SMT
3547 enabled, even if it's vulnerable. This is for
3548 users who don't want to be surprised by SMT
3549 getting disabled across kernel upgrades, or who
3550 have other ways of avoiding SMT-based attacks.
3551 Equivalent to: (default behavior)
3553 auto,nosmt
3554 Mitigate all CPU vulnerabilities, disabling SMT
3555 if needed. This is for users who always want to
3556 be fully mitigated, even if it means losing SMT.
3557 Equivalent to: l1tf=flush,nosmt [X86]
3558 mds=full,nosmt [X86]
3559 tsx_async_abort=full,nosmt [X86]
3560 mmio_stale_data=full,nosmt [X86]
3561 retbleed=auto,nosmt [X86]
3563 mminit_loglevel=
3564 [KNL,EARLY] When CONFIG_DEBUG_MEMORY_INIT is set, this
3565 parameter allows control of the logging verbosity for
3566 the additional memory initialisation checks. A value
3567 of 0 disables mminit logging and a level of 4 will
3568 log everything. Information is printed at KERN_DEBUG
3569 so loglevel=8 may also need to be specified.
3571 mmio_stale_data=
3572 [X86,INTEL,EARLY] Control mitigation for the Processor
3573 MMIO Stale Data vulnerabilities.
3575 Processor MMIO Stale Data is a class of
3576 vulnerabilities that may expose data after an MMIO
3577 operation. Exposed data could originate or end in
3578 the same CPU buffers as affected by MDS and TAA.
3579 Therefore, similar to MDS and TAA, the mitigation
3580 is to clear the affected CPU buffers.
3582 This parameter controls the mitigation. The
3583 options are:
3585 full - Enable mitigation on vulnerable CPUs
3587 full,nosmt - Enable mitigation and disable SMT on
3588 vulnerable CPUs.
3590 off - Unconditionally disable mitigation
3592 On MDS or TAA affected machines,
3593 mmio_stale_data=off can be prevented by an active
3594 MDS or TAA mitigation as these vulnerabilities are
3595 mitigated with the same mechanism so in order to
3596 disable this mitigation, you need to specify
3597 mds=off and tsx_async_abort=off too.
3599 Not specifying this option is equivalent to
3600 mmio_stale_data=full.
3602 For details see:
3603 Documentation/admin-guide/hw-vuln/processor_mmio_stale_data.rst
3605 <module>.async_probe[=<bool>] [KNL]
3606 If no <bool> value is specified or if the value
3607 specified is not a valid <bool>, enable asynchronous
3608 probe on this module. Otherwise, enable/disable
3609 asynchronous probe on this module as indicated by the
3610 <bool> value. See also: module.async_probe
3612 module.async_probe=<bool>
3613 [KNL] When set to true, modules will use async probing
3614 by default. To enable/disable async probing for a
3615 specific module, use the module specific control that
3616 is documented under <module>.async_probe. When both
3617 module.async_probe and <module>.async_probe are
3618 specified, <module>.async_probe takes precedence for
3619 the specific module.
3621 module.enable_dups_trace
3622 [KNL] When CONFIG_MODULE_DEBUG_AUTOLOAD_DUPS is set,
3623 this means that duplicate request_module() calls will
3624 trigger a WARN_ON() instead of a pr_warn(). Note that
3625 if MODULE_DEBUG_AUTOLOAD_DUPS_TRACE is set, WARN_ON()s
3626 will always be issued and this option does nothing.
3627 module.sig_enforce
3628 [KNL] When CONFIG_MODULE_SIG is set, this means that
3629 modules without (valid) signatures will fail to load.
3630 Note that if CONFIG_MODULE_SIG_FORCE is set, that
3631 is always true, so this option does nothing.
3633 module_blacklist= [KNL] Do not load a comma-separated list of
3634 modules. Useful for debugging problem modules.
3636 mousedev.tap_time=
3637 [MOUSE] Maximum time between finger touching and
3638 leaving touchpad surface for touch to be considered
3639 a tap and be reported as a left button click (for
3640 touchpads working in absolute mode only).
3641 Format: <msecs>
3642 mousedev.xres= [MOUSE] Horizontal screen resolution, used for devices
3643 reporting absolute coordinates, such as tablets
3644 mousedev.yres= [MOUSE] Vertical screen resolution, used for devices
3645 reporting absolute coordinates, such as tablets
3647 movablecore= [KNL,X86,PPC,EARLY]
3648 Format: nn[KMGTPE] | nn%
3649 This parameter is the complement to kernelcore=, it
3650 specifies the amount of memory used for migratable
3651 allocations. If both kernelcore and movablecore is
3652 specified, then kernelcore will be at *least* the
3653 specified value but may be more. If movablecore on its
3654 own is specified, the administrator must be careful
3655 that the amount of memory usable for all allocations
3656 is not too small.
3658 movable_node [KNL,EARLY] Boot-time switch to make hotplugable memory
3659 NUMA nodes to be movable. This means that the memory
3660 of such nodes will be usable only for movable
3661 allocations which rules out almost all kernel
3662 allocations. Use with caution!
3664 MTD_Partition= [MTD]
3665 Format: <name>,<region-number>,<size>,<offset>
3667 MTD_Region= [MTD] Format:
3668 <name>,<region-number>[,<base>,<size>,<buswidth>,<altbuswidth>]
3670 mtdparts= [MTD]
3671 See drivers/mtd/parsers/cmdlinepart.c
3673 mtouchusb.raw_coordinates=
3674 [HW] Make the MicroTouch USB driver use raw coordinates
3675 ('y', default) or cooked coordinates ('n')
3677 mtrr=debug [X86,EARLY]
3678 Enable printing debug information related to MTRR
3679 registers at boot time.
3681 mtrr_chunk_size=nn[KMG,X86,EARLY]
3682 used for mtrr cleanup. It is largest continuous chunk
3683 that could hold holes aka. UC entries.
3685 mtrr_gran_size=nn[KMG,X86,EARLY]
3686 Used for mtrr cleanup. It is granularity of mtrr block.
3687 Default is 1.
3688 Large value could prevent small alignment from
3689 using up MTRRs.
3691 mtrr_spare_reg_nr=n [X86,EARLY]
3692 Format: <integer>
3693 Range: 0,7 : spare reg number
3694 Default : 1
3695 Used for mtrr cleanup. It is spare mtrr entries number.
3696 Set to 2 or more if your graphical card needs more.
3698 multitce=off [PPC] This parameter disables the use of the pSeries
3699 firmware feature for updating multiple TCE entries
3700 at a time.
3702 n2= [NET] SDL Inc. RISCom/N2 synchronous serial card
3704 netdev= [NET] Network devices parameters
3705 Format: <irq>,<io>,<mem_start>,<mem_end>,<name>
3706 Note that mem_start is often overloaded to mean
3707 something different and driver-specific.
3708 This usage is only documented in each driver source
3709 file if at all.
3711 netpoll.carrier_timeout=
3712 [NET] Specifies amount of time (in seconds) that
3713 netpoll should wait for a carrier. By default netpoll
3714 waits 4 seconds.
3716 nf_conntrack.acct=
3717 [NETFILTER] Enable connection tracking flow accounting
3718 0 to disable accounting
3719 1 to enable accounting
3720 Default value is 0.
3722 nfs.cache_getent=
3723 [NFS] sets the pathname to the program which is used
3724 to update the NFS client cache entries.
3726 nfs.cache_getent_timeout=
3727 [NFS] sets the timeout after which an attempt to
3728 update a cache entry is deemed to have failed.
3730 nfs.callback_nr_threads=
3731 [NFSv4] set the total number of threads that the
3732 NFS client will assign to service NFSv4 callback
3733 requests.
3735 nfs.callback_tcpport=
3736 [NFS] set the TCP port on which the NFSv4 callback
3737 channel should listen.
3739 nfs.delay_retrans=
3740 [NFS] specifies the number of times the NFSv4 client
3741 retries the request before returning an EAGAIN error,
3742 after a reply of NFS4ERR_DELAY from the server.
3743 Only applies if the softerr mount option is enabled,
3744 and the specified value is >= 0.
3746 nfs.enable_ino64=
3747 [NFS] enable 64-bit inode numbers.
3748 If zero, the NFS client will fake up a 32-bit inode
3749 number for the readdir() and stat() syscalls instead
3750 of returning the full 64-bit number.
3751 The default is to return 64-bit inode numbers.
3753 nfs.idmap_cache_timeout=
3754 [NFS] set the maximum lifetime for idmapper cache
3755 entries.
3757 nfs.max_session_cb_slots=
3758 [NFSv4.1] Sets the maximum number of session
3759 slots the client will assign to the callback
3760 channel. This determines the maximum number of
3761 callbacks the client will process in parallel for
3762 a particular server.
3764 nfs.max_session_slots=
3765 [NFSv4.1] Sets the maximum number of session slots
3766 the client will attempt to negotiate with the server.
3767 This limits the number of simultaneous RPC requests
3768 that the client can send to the NFSv4.1 server.
3769 Note that there is little point in setting this
3770 value higher than the max_tcp_slot_table_limit.
3772 nfs.nfs4_disable_idmapping=
3773 [NFSv4] When set to the default of '1', this option
3774 ensures that both the RPC level authentication
3775 scheme and the NFS level operations agree to use
3776 numeric uids/gids if the mount is using the
3777 'sec=sys' security flavour. In effect it is
3778 disabling idmapping, which can make migration from
3779 legacy NFSv2/v3 systems to NFSv4 easier.
3780 Servers that do not support this mode of operation
3781 will be autodetected by the client, and it will fall
3782 back to using the idmapper.
3783 To turn off this behaviour, set the value to '0'.
3785 nfs.nfs4_unique_id=
3786 [NFS4] Specify an additional fixed unique ident-
3787 ification string that NFSv4 clients can insert into
3788 their nfs_client_id4 string. This is typically a
3789 UUID that is generated at system install time.
3791 nfs.recover_lost_locks=
3792 [NFSv4] Attempt to recover locks that were lost due
3793 to a lease timeout on the server. Please note that
3794 doing this risks data corruption, since there are
3795 no guarantees that the file will remain unchanged
3796 after the locks are lost.
3797 If you want to enable the kernel legacy behaviour of
3798 attempting to recover these locks, then set this
3799 parameter to '1'.
3800 The default parameter value of '0' causes the kernel
3801 not to attempt recovery of lost locks.
3803 nfs.send_implementation_id=
3804 [NFSv4.1] Send client implementation identification
3805 information in exchange_id requests.
3806 If zero, no implementation identification information
3807 will be sent.
3808 The default is to send the implementation identification
3809 information.
3811 nfs4.layoutstats_timer=
3812 [NFSv4.2] Change the rate at which the kernel sends
3813 layoutstats to the pNFS metadata server.
3815 Setting this to value to 0 causes the kernel to use
3816 whatever value is the default set by the layout
3817 driver. A non-zero value sets the minimum interval
3818 in seconds between layoutstats transmissions.
3820 nfsd.inter_copy_offload_enable=
3821 [NFSv4.2] When set to 1, the server will support
3822 server-to-server copies for which this server is
3823 the destination of the copy.
3825 nfsd.nfs4_disable_idmapping=
3826 [NFSv4] When set to the default of '1', the NFSv4
3827 server will return only numeric uids and gids to
3828 clients using auth_sys, and will accept numeric uids
3829 and gids from such clients. This is intended to ease
3830 migration from NFSv2/v3.
3832 nfsd.nfsd4_ssc_umount_timeout=
3833 [NFSv4.2] When used as the destination of a
3834 server-to-server copy, knfsd temporarily mounts
3835 the source server. It caches the mount in case
3836 it will be needed again, and discards it if not
3837 used for the number of milliseconds specified by
3838 this parameter.
3840 nfsaddrs= [NFS] Deprecated. Use ip= instead.
3841 See Documentation/admin-guide/nfs/nfsroot.rst.
3843 nfsroot= [NFS] nfs root filesystem for disk-less boxes.
3844 See Documentation/admin-guide/nfs/nfsroot.rst.
3846 nfsrootdebug [NFS] enable nfsroot debugging messages.
3847 See Documentation/admin-guide/nfs/nfsroot.rst.
3849 nmi_backtrace.backtrace_idle [KNL]
3850 Dump stacks even of idle CPUs in response to an
3851 NMI stack-backtrace request.
3853 nmi_debug= [KNL,SH] Specify one or more actions to take
3854 when a NMI is triggered.
3855 Format: [state][,regs][,debounce][,die]
3857 nmi_watchdog= [KNL,BUGS=X86] Debugging features for SMP kernels
3858 Format: [panic,][nopanic,][rNNN,][num]
3859 Valid num: 0 or 1
3860 0 - turn hardlockup detector in nmi_watchdog off
3861 1 - turn hardlockup detector in nmi_watchdog on
3862 rNNN - configure the watchdog with raw perf event 0xNNN
3864 When panic is specified, panic when an NMI watchdog
3865 timeout occurs (or 'nopanic' to not panic on an NMI
3866 watchdog, if CONFIG_BOOTPARAM_HARDLOCKUP_PANIC is set)
3867 To disable both hard and soft lockup detectors,
3868 please see 'nowatchdog'.
3869 This is useful when you use a panic=... timeout and
3870 need the box quickly up again.
3872 These settings can be accessed at runtime via
3873 the nmi_watchdog and hardlockup_panic sysctls.
3875 no387 [BUGS=X86-32] Tells the kernel to use the 387 maths
3876 emulation library even if a 387 maths coprocessor
3877 is present.
3879 no4lvl [RISCV,EARLY] Disable 4-level and 5-level paging modes.
3880 Forces kernel to use 3-level paging instead.
3882 no5lvl [X86-64,RISCV,EARLY] Disable 5-level paging mode. Forces
3883 kernel to use 4-level paging instead.
3885 noalign [KNL,ARM]
3887 noapic [SMP,APIC,EARLY] Tells the kernel to not make use of any
3888 IOAPICs that may be present in the system.
3890 noautogroup Disable scheduler automatic task group creation.
3892 nocache [ARM,EARLY]
3894 no_console_suspend
3895 [HW] Never suspend the console
3896 Disable suspending of consoles during suspend and
3897 hibernate operations. Once disabled, debugging
3898 messages can reach various consoles while the rest
3899 of the system is being put to sleep (ie, while
3900 debugging driver suspend/resume hooks). This may
3901 not work reliably with all consoles, but is known
3902 to work with serial and VGA consoles.
3903 To facilitate more flexible debugging, we also add
3904 console_suspend, a printk module parameter to control
3905 it. Users could use console_suspend (usually
3906 /sys/module/printk/parameters/console_suspend) to
3907 turn on/off it dynamically.
3909 no_debug_objects
3910 [KNL,EARLY] Disable object debugging
3912 nodsp [SH] Disable hardware DSP at boot time.
3914 noefi [EFI,EARLY] Disable EFI runtime services support.
3916 no_entry_flush [PPC,EARLY] Don't flush the L1-D cache when entering the kernel.
3918 noexec32 [X86-64]
3919 This affects only 32-bit executables.
3920 noexec32=on: enable non-executable mappings (default)
3921 read doesn't imply executable mappings
3922 noexec32=off: disable non-executable mappings
3923 read implies executable mappings
3925 no_file_caps Tells the kernel not to honor file capabilities. The
3926 only way then for a file to be executed with privilege
3927 is to be setuid root or executed by root.
3929 nofpu [MIPS,SH] Disable hardware FPU at boot time.
3931 nofsgsbase [X86] Disables FSGSBASE instructions.
3933 nofxsr [BUGS=X86-32] Disables x86 floating point extended
3934 register save and restore. The kernel will only save
3935 legacy floating-point registers on task switch.
3937 no_hash_pointers
3938 [KNL,EARLY]
3939 Force pointers printed to the console or buffers to be
3940 unhashed. By default, when a pointer is printed via %p
3941 format string, that pointer is "hashed", i.e. obscured
3942 by hashing the pointer value. This is a security feature
3943 that hides actual kernel addresses from unprivileged
3944 users, but it also makes debugging the kernel more
3945 difficult since unequal pointers can no longer be
3946 compared. However, if this command-line option is
3947 specified, then all normal pointers will have their true
3948 value printed. This option should only be specified when
3949 debugging the kernel. Please do not use on production
3950 kernels.
3952 nohibernate [HIBERNATION] Disable hibernation and resume.
3954 nohlt [ARM,ARM64,MICROBLAZE,MIPS,PPC,RISCV,SH] Forces the kernel to
3955 busy wait in do_idle() and not use the arch_cpu_idle()
3956 implementation; requires CONFIG_GENERIC_IDLE_POLL_SETUP
3957 to be effective. This is useful on platforms where the
3958 sleep(SH) or wfi(ARM,ARM64) instructions do not work
3959 correctly or when doing power measurements to evaluate
3960 the impact of the sleep instructions. This is also
3961 useful when using JTAG debugger.
3963 nohugeiomap [KNL,X86,PPC,ARM64,EARLY] Disable kernel huge I/O mappings.
3965 nohugevmalloc [KNL,X86,PPC,ARM64,EARLY] Disable kernel huge vmalloc mappings.
3967 nohz= [KNL] Boottime enable/disable dynamic ticks
3968 Valid arguments: on, off
3969 Default: on
3971 nohz_full= [KNL,BOOT,SMP,ISOL]
3972 The argument is a cpu list, as described above.
3973 In kernels built with CONFIG_NO_HZ_FULL=y, set
3974 the specified list of CPUs whose tick will be stopped
3975 whenever possible. The boot CPU will be forced outside
3976 the range to maintain the timekeeping. Any CPUs
3977 in this list will have their RCU callbacks offloaded,
3978 just as if they had also been called out in the
3979 rcu_nocbs= boot parameter.
3981 Note that this argument takes precedence over
3982 the CONFIG_RCU_NOCB_CPU_DEFAULT_ALL option.
3984 noinitrd [RAM] Tells the kernel not to load any configured
3985 initial RAM disk.
3987 nointremap [X86-64,Intel-IOMMU,EARLY] Do not enable interrupt
3988 remapping.
3989 [Deprecated - use intremap=off]
3991 noinvpcid [X86,EARLY] Disable the INVPCID cpu feature.
3993 noiotrap [SH] Disables trapped I/O port accesses.
3995 noirqdebug [X86-32] Disables the code which attempts to detect and
3996 disable unhandled interrupt sources.
3998 noisapnp [ISAPNP] Disables ISA PnP code.
4000 nokaslr [KNL,EARLY]
4001 When CONFIG_RANDOMIZE_BASE is set, this disables
4002 kernel and module base offset ASLR (Address Space
4003 Layout Randomization).
4005 no-kvmapf [X86,KVM,EARLY] Disable paravirtualized asynchronous page
4006 fault handling.
4008 no-kvmclock [X86,KVM,EARLY] Disable paravirtualized KVM clock driver
4010 nolapic [X86-32,APIC,EARLY] Do not enable or use the local APIC.
4012 nolapic_timer [X86-32,APIC,EARLY] Do not use the local APIC timer.
4014 nomce [X86-32] Disable Machine Check Exception
4016 nomfgpt [X86-32] Disable Multi-Function General Purpose
4017 Timer usage (for AMD Geode machines).
4019 nomodeset Disable kernel modesetting. Most systems' firmware
4020 sets up a display mode and provides framebuffer memory
4021 for output. With nomodeset, DRM and fbdev drivers will
4022 not load if they could possibly displace the pre-
4023 initialized output. Only the system framebuffer will
4024 be available for use. The respective drivers will not
4025 perform display-mode changes or accelerated rendering.
4027 Useful as error fallback, or for testing and debugging.
4029 nomodule Disable module load
4031 nonmi_ipi [X86] Disable using NMI IPIs during panic/reboot to
4032 shutdown the other cpus. Instead use the REBOOT_VECTOR
4033 irq.
4035 nopat [X86,EARLY] Disable PAT (page attribute table extension of
4036 pagetables) support.
4038 nopcid [X86-64,EARLY] Disable the PCID cpu feature.
4040 nopku [X86] Disable Memory Protection Keys CPU feature found
4041 in some Intel CPUs.
4043 nopti [X86-64,EARLY]
4044 Equivalent to pti=off
4046 nopv= [X86,XEN,KVM,HYPER_V,VMWARE,EARLY]
4047 Disables the PV optimizations forcing the guest to run
4048 as generic guest with no PV drivers. Currently support
4049 XEN HVM, KVM, HYPER_V and VMWARE guest.
4051 nopvspin [X86,XEN,KVM,EARLY]
4052 Disables the qspinlock slow path using PV optimizations
4053 which allow the hypervisor to 'idle' the guest on lock
4054 contention.
4056 norandmaps Don't use address space randomization. Equivalent to
4057 echo 0 > /proc/sys/kernel/randomize_va_space
4059 noreplace-smp [X86-32,SMP] Don't replace SMP instructions
4060 with UP alternatives
4062 noresume [SWSUSP] Disables resume and restores original swap
4063 space.
4065 no-scroll [VGA] Disables scrollback.
4066 This is required for the Braillex ib80-piezo Braille
4067 reader made by F.H. Papenmeier (Germany).
4069 nosgx [X86-64,SGX,EARLY] Disables Intel SGX kernel support.
4071 nosmap [PPC,EARLY]
4072 Disable SMAP (Supervisor Mode Access Prevention)
4073 even if it is supported by processor.
4075 nosmep [PPC64s,EARLY]
4076 Disable SMEP (Supervisor Mode Execution Prevention)
4077 even if it is supported by processor.
4079 nosmp [SMP,EARLY] Tells an SMP kernel to act as a UP kernel,
4080 and disable the IO APIC. legacy for "maxcpus=0".
4082 nosmt [KNL,MIPS,PPC,S390,EARLY] Disable symmetric multithreading (SMT).
4083 Equivalent to smt=1.
4085 [KNL,X86,PPC] Disable symmetric multithreading (SMT).
4086 nosmt=force: Force disable SMT, cannot be undone
4087 via the sysfs control file.
4089 nosoftlockup [KNL] Disable the soft-lockup detector.
4091 nospec_store_bypass_disable
4092 [HW,EARLY] Disable all mitigations for the Speculative
4093 Store Bypass vulnerability
4095 nospectre_bhb [ARM64,EARLY] Disable all mitigations for Spectre-BHB (branch
4096 history injection) vulnerability. System may allow data leaks
4097 with this option.
4099 nospectre_v1 [X86,PPC,EARLY] Disable mitigations for Spectre Variant 1
4100 (bounds check bypass). With this option data leaks are
4101 possible in the system.
4103 nospectre_v2 [X86,PPC_E500,ARM64,EARLY] Disable all mitigations
4104 for the Spectre variant 2 (indirect branch
4105 prediction) vulnerability. System may allow data
4106 leaks with this option.
4108 no-steal-acc [X86,PV_OPS,ARM64,PPC/PSERIES,RISCV,LOONGARCH,EARLY]
4109 Disable paravirtualized steal time accounting. steal time
4110 is computed, but won't influence scheduler behaviour
4112 nosync [HW,M68K] Disables sync negotiation for all devices.
4114 no_timer_check [X86,APIC] Disables the code which tests for
4115 broken timer IRQ sources.
4117 no_uaccess_flush
4118 [PPC,EARLY] Don't flush the L1-D cache after accessing user data.
4120 novmcoredd [KNL,KDUMP]
4121 Disable device dump. Device dump allows drivers to
4122 append dump data to vmcore so you can collect driver
4123 specified debug info. Drivers can append the data
4124 without any limit and this data is stored in memory,
4125 so this may cause significant memory stress. Disabling
4126 device dump can help save memory but the driver debug
4127 data will be no longer available. This parameter
4128 is only available when CONFIG_PROC_VMCORE_DEVICE_DUMP
4129 is set.
4131 no-vmw-sched-clock
4132 [X86,PV_OPS,EARLY] Disable paravirtualized VMware
4133 scheduler clock and use the default one.
4135 nowatchdog [KNL] Disable both lockup detectors, i.e.
4136 soft-lockup and NMI watchdog (hard-lockup).
4138 nowb [ARM,EARLY]
4140 nox2apic [X86-64,APIC,EARLY] Do not enable x2APIC mode.
4142 NOTE: this parameter will be ignored on systems with the
4143 LEGACY_XAPIC_DISABLED bit set in the
4144 IA32_XAPIC_DISABLE_STATUS MSR.
4146 noxsave [BUGS=X86] Disables x86 extended register state save
4147 and restore using xsave. The kernel will fallback to
4148 enabling legacy floating-point and sse state.
4150 noxsaveopt [X86] Disables xsaveopt used in saving x86 extended
4151 register states. The kernel will fall back to use
4152 xsave to save the states. By using this parameter,
4153 performance of saving the states is degraded because
4154 xsave doesn't support modified optimization while
4155 xsaveopt supports it on xsaveopt enabled systems.
4157 noxsaves [X86] Disables xsaves and xrstors used in saving and
4158 restoring x86 extended register state in compacted
4159 form of xsave area. The kernel will fall back to use
4160 xsaveopt and xrstor to save and restore the states
4161 in standard form of xsave area. By using this
4162 parameter, xsave area per process might occupy more
4163 memory on xsaves enabled systems.
4165 nr_cpus= [SMP,EARLY] Maximum number of processors that an SMP kernel
4166 could support. nr_cpus=n : n >= 1 limits the kernel to
4167 support 'n' processors. It could be larger than the
4168 number of already plugged CPU during bootup, later in
4169 runtime you can physically add extra cpu until it reaches
4170 n. So during boot up some boot time memory for per-cpu
4171 variables need be pre-allocated for later physical cpu
4172 hot plugging.
4174 nr_uarts= [SERIAL] maximum number of UARTs to be registered.
4176 numa=off [KNL, ARM64, PPC, RISCV, SPARC, X86, EARLY]
4177 Disable NUMA, Only set up a single NUMA node
4178 spanning all memory.
4180 numa=fake=<size>[MG]
4181 [KNL, ARM64, RISCV, X86, EARLY]
4182 If given as a memory unit, fills all system RAM with
4183 nodes of size interleaved over physical nodes.
4185 numa=fake=<N>
4186 [KNL, ARM64, RISCV, X86, EARLY]
4187 If given as an integer, fills all system RAM with N
4188 fake nodes interleaved over physical nodes.
4190 numa=fake=<N>U
4191 [KNL, ARM64, RISCV, X86, EARLY]
4192 If given as an integer followed by 'U', it will
4193 divide each physical node into N emulated nodes.
4195 numa_balancing= [KNL,ARM64,PPC,RISCV,S390,X86] Enable or disable automatic
4196 NUMA balancing.
4197 Allowed values are enable and disable
4199 numa_zonelist_order= [KNL, BOOT] Select zonelist order for NUMA.
4200 'node', 'default' can be specified
4201 This can be set from sysctl after boot.
4202 See Documentation/admin-guide/sysctl/vm.rst for details.
4204 ohci1394_dma=early [HW,EARLY] enable debugging via the ohci1394 driver.
4205 See Documentation/core-api/debugging-via-ohci1394.rst for more
4206 info.
4208 olpc_ec_timeout= [OLPC] ms delay when issuing EC commands
4209 Rather than timing out after 20 ms if an EC
4210 command is not properly ACKed, override the length
4211 of the timeout. We have interrupts disabled while
4212 waiting for the ACK, so if this is set too high
4213 interrupts *may* be lost!
4215 omap_mux= [OMAP] Override bootloader pin multiplexing.
4216 Format: <mux_mode0.mode_name=value>...
4217 For example, to override I2C bus2:
4218 omap_mux=i2c2_scl.i2c2_scl=0x100,i2c2_sda.i2c2_sda=0x100
4220 onenand.bdry= [HW,MTD] Flex-OneNAND Boundary Configuration
4222 Format: [die0_boundary][,die0_lock][,die1_boundary][,die1_lock]
4224 boundary - index of last SLC block on Flex-OneNAND.
4225 The remaining blocks are configured as MLC blocks.
4226 lock - Configure if Flex-OneNAND boundary should be locked.
4227 Once locked, the boundary cannot be changed.
4228 1 indicates lock status, 0 indicates unlock status.
4230 oops=panic [KNL,EARLY]
4231 Always panic on oopses. Default is to just kill the
4232 process, but there is a small probability of
4233 deadlocking the machine.
4234 This will also cause panics on machine check exceptions.
4235 Useful together with panic=30 to trigger a reboot.
4237 page_alloc.shuffle=
4238 [KNL] Boolean flag to control whether the page allocator
4239 should randomize its free lists. This parameter can be
4240 used to enable/disable page randomization. The state of
4241 the flag can be read from sysfs at:
4242 /sys/module/page_alloc/parameters/shuffle.
4243 This parameter is only available if CONFIG_SHUFFLE_PAGE_ALLOCATOR=y.
4245 page_owner= [KNL,EARLY] Boot-time page_owner enabling option.
4246 Storage of the information about who allocated
4247 each page is disabled in default. With this switch,
4248 we can turn it on.
4249 on: enable the feature
4251 page_poison= [KNL,EARLY] Boot-time parameter changing the state of
4252 poisoning on the buddy allocator, available with
4253 CONFIG_PAGE_POISONING=y.
4254 off: turn off poisoning (default)
4255 on: turn on poisoning
4257 page_reporting.page_reporting_order=
4258 [KNL] Minimal page reporting order
4259 Format: <integer>
4260 Adjust the minimal page reporting order. The page
4261 reporting is disabled when it exceeds MAX_PAGE_ORDER.
4263 panic= [KNL] Kernel behaviour on panic: delay <timeout>
4264 timeout > 0: seconds before rebooting
4265 timeout = 0: wait forever
4266 timeout < 0: reboot immediately
4267 Format: <timeout>
4269 panic_on_taint= [KNL,EARLY]
4270 Bitmask for conditionally calling panic() in add_taint()
4271 Format: <hex>[,nousertaint]
4272 Hexadecimal bitmask representing the set of TAINT flags
4273 that will cause the kernel to panic when add_taint() is
4274 called with any of the flags in this set.
4275 The optional switch "nousertaint" can be utilized to
4276 prevent userspace forced crashes by writing to sysctl
4277 /proc/sys/kernel/tainted any flagset matching with the
4278 bitmask set on panic_on_taint.
4279 See Documentation/admin-guide/tainted-kernels.rst for
4280 extra details on the taint flags that users can pick
4281 to compose the bitmask to assign to panic_on_taint.
4283 panic_on_warn=1 panic() instead of WARN(). Useful to cause kdump
4284 on a WARN().
4286 panic_print= Bitmask for printing system info when panic happens.
4287 User can chose combination of the following bits:
4288 bit 0: print all tasks info
4289 bit 1: print system memory info
4290 bit 2: print timer info
4291 bit 3: print locks info if CONFIG_LOCKDEP is on
4292 bit 4: print ftrace buffer
4293 bit 5: print all printk messages in buffer
4294 bit 6: print all CPUs backtrace (if available in the arch)
4295 bit 7: print only tasks in uninterruptible (blocked) state
4296 *Be aware* that this option may print a _lot_ of lines,
4297 so there are risks of losing older messages in the log.
4298 Use this option carefully, maybe worth to setup a
4299 bigger log buffer with "log_buf_len" along with this.
4301 parkbd.port= [HW] Parallel port number the keyboard adapter is
4302 connected to, default is 0.
4303 Format: <parport#>
4304 parkbd.mode= [HW] Parallel port keyboard adapter mode of operation,
4305 0 for XT, 1 for AT (default is AT).
4306 Format: <mode>
4308 parport= [HW,PPT] Specify parallel ports. 0 disables.
4309 Format: { 0 | auto | 0xBBB[,IRQ[,DMA]] }
4310 Use 'auto' to force the driver to use any
4311 IRQ/DMA settings detected (the default is to
4312 ignore detected IRQ/DMA settings because of
4313 possible conflicts). You can specify the base
4314 address, IRQ, and DMA settings; IRQ and DMA
4315 should be numbers, or 'auto' (for using detected
4316 settings on that particular port), or 'nofifo'
4317 (to avoid using a FIFO even if it is detected).
4318 Parallel ports are assigned in the order they
4319 are specified on the command line, starting
4320 with parport0.
4322 parport_init_mode= [HW,PPT]
4323 Configure VIA parallel port to operate in
4324 a specific mode. This is necessary on Pegasos
4325 computer where firmware has no options for setting
4326 up parallel port mode and sets it to spp.
4327 Currently this function knows 686a and 8231 chips.
4328 Format: [spp|ps2|epp|ecp|ecpepp]
4330 pata_legacy.all= [HW,LIBATA]
4331 Format: <int>
4332 Set to non-zero to probe primary and secondary ISA
4333 port ranges on PCI systems where no PCI PATA device
4334 has been found at either range. Disabled by default.
4336 pata_legacy.autospeed= [HW,LIBATA]
4337 Format: <int>
4338 Set to non-zero if a chip is present that snoops speed
4339 changes. Disabled by default.
4341 pata_legacy.ht6560a= [HW,LIBATA]
4342 Format: <int>
4343 Set to 1, 2, or 3 for HT 6560A on the primary channel,
4344 the secondary channel, or both channels respectively.
4345 Disabled by default.
4347 pata_legacy.ht6560b= [HW,LIBATA]
4348 Format: <int>
4349 Set to 1, 2, or 3 for HT 6560B on the primary channel,
4350 the secondary channel, or both channels respectively.
4351 Disabled by default.
4353 pata_legacy.iordy_mask= [HW,LIBATA]
4354 Format: <int>
4355 IORDY enable mask. Set individual bits to allow IORDY
4356 for the respective channel. Bit 0 is for the first
4357 legacy channel handled by this driver, bit 1 is for
4358 the second channel, and so on. The sequence will often
4359 correspond to the primary legacy channel, the secondary
4360 legacy channel, and so on, but the handling of a PCI
4361 bus and the use of other driver options may interfere
4362 with the sequence. By default IORDY is allowed across
4363 all channels.
4365 pata_legacy.opti82c46x= [HW,LIBATA]
4366 Format: <int>
4367 Set to 1, 2, or 3 for Opti 82c611A on the primary
4368 channel, the secondary channel, or both channels
4369 respectively. Disabled by default.
4371 pata_legacy.opti82c611a= [HW,LIBATA]
4372 Format: <int>
4373 Set to 1, 2, or 3 for Opti 82c465MV on the primary
4374 channel, the secondary channel, or both channels
4375 respectively. Disabled by default.
4377 pata_legacy.pio_mask= [HW,LIBATA]
4378 Format: <int>
4379 PIO mode mask for autospeed devices. Set individual
4380 bits to allow the use of the respective PIO modes.
4381 Bit 0 is for mode 0, bit 1 is for mode 1, and so on.
4382 All modes allowed by default.
4384 pata_legacy.probe_all= [HW,LIBATA]
4385 Format: <int>
4386 Set to non-zero to probe tertiary and further ISA
4387 port ranges on PCI systems. Disabled by default.
4389 pata_legacy.probe_mask= [HW,LIBATA]
4390 Format: <int>
4391 Probe mask for legacy ISA PATA ports. Depending on
4392 platform configuration and the use of other driver
4393 options up to 6 legacy ports are supported: 0x1f0,
4394 0x170, 0x1e8, 0x168, 0x1e0, 0x160, however probing
4395 of individual ports can be disabled by setting the
4396 corresponding bits in the mask to 1. Bit 0 is for
4397 the first port in the list above (0x1f0), and so on.
4398 By default all supported ports are probed.
4400 pata_legacy.qdi= [HW,LIBATA]
4401 Format: <int>
4402 Set to non-zero to probe QDI controllers. By default
4403 set to 1 if CONFIG_PATA_QDI_MODULE, 0 otherwise.
4405 pata_legacy.winbond= [HW,LIBATA]
4406 Format: <int>
4407 Set to non-zero to probe Winbond controllers. Use
4408 the standard I/O port (0x130) if 1, otherwise the
4409 value given is the I/O port to use (typically 0x1b0).
4410 By default set to 1 if CONFIG_PATA_WINBOND_VLB_MODULE,
4411 0 otherwise.
4413 pata_platform.pio_mask= [HW,LIBATA]
4414 Format: <int>
4415 Supported PIO mode mask. Set individual bits to allow
4416 the use of the respective PIO modes. Bit 0 is for
4417 mode 0, bit 1 is for mode 1, and so on. Mode 0 only
4418 allowed by default.
4420 pause_on_oops=<int>
4421 Halt all CPUs after the first oops has been printed for
4422 the specified number of seconds. This is to be used if
4423 your oopses keep scrolling off the screen.
4425 pcbit= [HW,ISDN]
4427 pci=option[,option...] [PCI,EARLY] various PCI subsystem options.
4429 Some options herein operate on a specific device
4430 or a set of devices (<pci_dev>). These are
4431 specified in one of the following formats:
4433 [<domain>:]<bus>:<dev>.<func>[/<dev>.<func>]*
4434 pci:<vendor>:<device>[:<subvendor>:<subdevice>]
4436 Note: the first format specifies a PCI
4437 bus/device/function address which may change
4438 if new hardware is inserted, if motherboard
4439 firmware changes, or due to changes caused
4440 by other kernel parameters. If the
4441 domain is left unspecified, it is
4442 taken to be zero. Optionally, a path
4443 to a device through multiple device/function
4444 addresses can be specified after the base
4445 address (this is more robust against
4446 renumbering issues). The second format
4447 selects devices using IDs from the
4448 configuration space which may match multiple
4449 devices in the system.
4451 earlydump dump PCI config space before the kernel
4452 changes anything
4453 off [X86] don't probe for the PCI bus
4454 bios [X86-32] force use of PCI BIOS, don't access
4455 the hardware directly. Use this if your machine
4456 has a non-standard PCI host bridge.
4457 nobios [X86-32] disallow use of PCI BIOS, only direct
4458 hardware access methods are allowed. Use this
4459 if you experience crashes upon bootup and you
4460 suspect they are caused by the BIOS.
4461 conf1 [X86] Force use of PCI Configuration Access
4462 Mechanism 1 (config address in IO port 0xCF8,
4463 data in IO port 0xCFC, both 32-bit).
4464 conf2 [X86] Force use of PCI Configuration Access
4465 Mechanism 2 (IO port 0xCF8 is an 8-bit port for
4466 the function, IO port 0xCFA, also 8-bit, sets
4467 bus number. The config space is then accessed
4468 through ports 0xC000-0xCFFF).
4469 See http://wiki.osdev.org/PCI for more info
4470 on the configuration access mechanisms.
4471 noaer [PCIE] If the PCIEAER kernel config parameter is
4472 enabled, this kernel boot option can be used to
4473 disable the use of PCIE advanced error reporting.
4474 nodomains [PCI] Disable support for multiple PCI
4475 root domains (aka PCI segments, in ACPI-speak).
4476 nommconf [X86] Disable use of MMCONFIG for PCI
4477 Configuration
4478 check_enable_amd_mmconf [X86] check for and enable
4479 properly configured MMIO access to PCI
4480 config space on AMD family 10h CPU
4481 nomsi [MSI] If the PCI_MSI kernel config parameter is
4482 enabled, this kernel boot option can be used to
4483 disable the use of MSI interrupts system-wide.
4484 noioapicquirk [APIC] Disable all boot interrupt quirks.
4485 Safety option to keep boot IRQs enabled. This
4486 should never be necessary.
4487 ioapicreroute [APIC] Enable rerouting of boot IRQs to the
4488 primary IO-APIC for bridges that cannot disable
4489 boot IRQs. This fixes a source of spurious IRQs
4490 when the system masks IRQs.
4491 noioapicreroute [APIC] Disable workaround that uses the
4492 boot IRQ equivalent of an IRQ that connects to
4493 a chipset where boot IRQs cannot be disabled.
4494 The opposite of ioapicreroute.
4495 biosirq [X86-32] Use PCI BIOS calls to get the interrupt
4496 routing table. These calls are known to be buggy
4497 on several machines and they hang the machine
4498 when used, but on other computers it's the only
4499 way to get the interrupt routing table. Try
4500 this option if the kernel is unable to allocate
4501 IRQs or discover secondary PCI buses on your
4502 motherboard.
4503 rom [X86] Assign address space to expansion ROMs.
4504 Use with caution as certain devices share
4505 address decoders between ROMs and other
4506 resources.
4507 norom [X86] Do not assign address space to
4508 expansion ROMs that do not already have
4509 BIOS assigned address ranges.
4510 nobar [X86] Do not assign address space to the
4511 BARs that weren't assigned by the BIOS.
4512 irqmask=0xMMMM [X86] Set a bit mask of IRQs allowed to be
4513 assigned automatically to PCI devices. You can
4514 make the kernel exclude IRQs of your ISA cards
4515 this way.
4516 pirqaddr=0xAAAAA [X86] Specify the physical address
4517 of the PIRQ table (normally generated
4518 by the BIOS) if it is outside the
4519 F0000h-100000h range.
4520 lastbus=N [X86] Scan all buses thru bus #N. Can be
4521 useful if the kernel is unable to find your
4522 secondary buses and you want to tell it
4523 explicitly which ones they are.
4524 assign-busses [X86] Always assign all PCI bus
4525 numbers ourselves, overriding
4526 whatever the firmware may have done.
4527 usepirqmask [X86] Honor the possible IRQ mask stored
4528 in the BIOS $PIR table. This is needed on
4529 some systems with broken BIOSes, notably
4530 some HP Pavilion N5400 and Omnibook XE3
4531 notebooks. This will have no effect if ACPI
4532 IRQ routing is enabled.
4533 noacpi [X86] Do not use ACPI for IRQ routing
4534 or for PCI scanning.
4535 use_crs [X86] Use PCI host bridge window information
4536 from ACPI. On BIOSes from 2008 or later, this
4537 is enabled by default. If you need to use this,
4538 please report a bug.
4539 nocrs [X86] Ignore PCI host bridge windows from ACPI.
4540 If you need to use this, please report a bug.
4541 use_e820 [X86] Use E820 reservations to exclude parts of
4542 PCI host bridge windows. This is a workaround
4543 for BIOS defects in host bridge _CRS methods.
4544 If you need to use this, please report a bug to
4545 <linux-pci@vger.kernel.org>.
4546 no_e820 [X86] Ignore E820 reservations for PCI host
4547 bridge windows. This is the default on modern
4548 hardware. If you need to use this, please report
4549 a bug to <linux-pci@vger.kernel.org>.
4550 routeirq Do IRQ routing for all PCI devices.
4551 This is normally done in pci_enable_device(),
4552 so this option is a temporary workaround
4553 for broken drivers that don't call it.
4554 skip_isa_align [X86] do not align io start addr, so can
4555 handle more pci cards
4556 noearly [X86] Don't do any early type 1 scanning.
4557 This might help on some broken boards which
4558 machine check when some devices' config space
4559 is read. But various workarounds are disabled
4560 and some IOMMU drivers will not work.
4561 bfsort Sort PCI devices into breadth-first order.
4562 This sorting is done to get a device
4563 order compatible with older (<= 2.4) kernels.
4564 nobfsort Don't sort PCI devices into breadth-first order.
4565 pcie_bus_tune_off Disable PCIe MPS (Max Payload Size)
4566 tuning and use the BIOS-configured MPS defaults.
4567 pcie_bus_safe Set every device's MPS to the largest value
4568 supported by all devices below the root complex.
4569 pcie_bus_perf Set device MPS to the largest allowable MPS
4570 based on its parent bus. Also set MRRS (Max
4571 Read Request Size) to the largest supported
4572 value (no larger than the MPS that the device
4573 or bus can support) for best performance.
4574 pcie_bus_peer2peer Set every device's MPS to 128B, which
4575 every device is guaranteed to support. This
4576 configuration allows peer-to-peer DMA between
4577 any pair of devices, possibly at the cost of
4578 reduced performance. This also guarantees
4579 that hot-added devices will work.
4580 cbiosize=nn[KMG] The fixed amount of bus space which is
4581 reserved for the CardBus bridge's IO window.
4582 The default value is 256 bytes.
4583 cbmemsize=nn[KMG] The fixed amount of bus space which is
4584 reserved for the CardBus bridge's memory
4585 window. The default value is 64 megabytes.
4586 resource_alignment=
4587 Format:
4588 [<order of align>@]<pci_dev>[; ...]
4589 Specifies alignment and device to reassign
4590 aligned memory resources. How to
4591 specify the device is described above.
4592 If <order of align> is not specified,
4593 PAGE_SIZE is used as alignment.
4594 A PCI-PCI bridge can be specified if resource
4595 windows need to be expanded.
4596 To specify the alignment for several
4597 instances of a device, the PCI vendor,
4598 device, subvendor, and subdevice may be
4599 specified, e.g., 12@pci:8086:9c22:103c:198f
4600 for 4096-byte alignment.
4601 ecrc= Enable/disable PCIe ECRC (transaction layer
4602 end-to-end CRC checking). Only effective if
4603 OS has native AER control (either granted by
4604 ACPI _OSC or forced via "pcie_ports=native")
4605 bios: Use BIOS/firmware settings. This is the
4606 the default.
4607 off: Turn ECRC off
4608 on: Turn ECRC on.
4609 hpiosize=nn[KMG] The fixed amount of bus space which is
4610 reserved for hotplug bridge's IO window.
4611 Default size is 256 bytes.
4612 hpmmiosize=nn[KMG] The fixed amount of bus space which is
4613 reserved for hotplug bridge's MMIO window.
4614 Default size is 2 megabytes.
4615 hpmmioprefsize=nn[KMG] The fixed amount of bus space which is
4616 reserved for hotplug bridge's MMIO_PREF window.
4617 Default size is 2 megabytes.
4618 hpmemsize=nn[KMG] The fixed amount of bus space which is
4619 reserved for hotplug bridge's MMIO and
4620 MMIO_PREF window.
4621 Default size is 2 megabytes.
4622 hpbussize=nn The minimum amount of additional bus numbers
4623 reserved for buses below a hotplug bridge.
4624 Default is 1.
4625 realloc= Enable/disable reallocating PCI bridge resources
4626 if allocations done by BIOS are too small to
4627 accommodate resources required by all child
4628 devices.
4629 off: Turn realloc off
4630 on: Turn realloc on
4631 realloc same as realloc=on
4632 noari do not use PCIe ARI.
4633 noats [PCIE, Intel-IOMMU, AMD-IOMMU]
4634 do not use PCIe ATS (and IOMMU device IOTLB).
4635 pcie_scan_all Scan all possible PCIe devices. Otherwise we
4636 only look for one device below a PCIe downstream
4637 port.
4638 big_root_window Try to add a big 64bit memory window to the PCIe
4639 root complex on AMD CPUs. Some GFX hardware
4640 can resize a BAR to allow access to all VRAM.
4641 Adding the window is slightly risky (it may
4642 conflict with unreported devices), so this
4643 taints the kernel.
4644 disable_acs_redir=<pci_dev>[; ...]
4645 Specify one or more PCI devices (in the format
4646 specified above) separated by semicolons.
4647 Each device specified will have the PCI ACS
4648 redirect capabilities forced off which will
4649 allow P2P traffic between devices through
4650 bridges without forcing it upstream. Note:
4651 this removes isolation between devices and
4652 may put more devices in an IOMMU group.
4653 config_acs=
4654 Format:
4655 <ACS flags>@<pci_dev>[; ...]
4656 Specify one or more PCI devices (in the format
4657 specified above) optionally prepended with flags
4658 and separated by semicolons. The respective
4659 capabilities will be enabled, disabled or
4660 unchanged based on what is specified in
4661 flags.
4663 ACS Flags is defined as follows:
4664 bit-0 : ACS Source Validation
4665 bit-1 : ACS Translation Blocking
4666 bit-2 : ACS P2P Request Redirect
4667 bit-3 : ACS P2P Completion Redirect
4668 bit-4 : ACS Upstream Forwarding
4669 bit-5 : ACS P2P Egress Control
4670 bit-6 : ACS Direct Translated P2P
4671 Each bit can be marked as:
4672 '0' – force disabled
4673 '1' – force enabled
4674 'x' – unchanged
4675 For example,
4676 pci=config_acs=10x
4677 would configure all devices that support
4678 ACS to enable P2P Request Redirect, disable
4679 Translation Blocking, and leave Source
4680 Validation unchanged from whatever power-up
4681 or firmware set it to.
4683 Note: this may remove isolation between devices
4684 and may put more devices in an IOMMU group.
4685 force_floating [S390] Force usage of floating interrupts.
4686 nomio [S390] Do not use MIO instructions.
4687 norid [S390] ignore the RID field and force use of
4688 one PCI domain per PCI function
4689 notph [PCIE] If the PCIE_TPH kernel config parameter
4690 is enabled, this kernel boot option can be used
4691 to disable PCIe TLP Processing Hints support
4692 system-wide.
4694 pcie_aspm= [PCIE] Forcibly enable or ignore PCIe Active State Power
4695 Management.
4696 off Don't touch ASPM configuration at all. Leave any
4697 configuration done by firmware unchanged.
4698 force Enable ASPM even on devices that claim not to support it.
4699 WARNING: Forcing ASPM on may cause system lockups.
4701 pcie_ports= [PCIE] PCIe port services handling:
4702 native Use native PCIe services (PME, AER, DPC, PCIe hotplug)
4703 even if the platform doesn't give the OS permission to
4704 use them. This may cause conflicts if the platform
4705 also tries to use these services.
4706 dpc-native Use native PCIe service for DPC only. May
4707 cause conflicts if firmware uses AER or DPC.
4708 compat Disable native PCIe services (PME, AER, DPC, PCIe
4709 hotplug).
4711 pcie_port_pm= [PCIE] PCIe port power management handling:
4712 off Disable power management of all PCIe ports
4713 force Forcibly enable power management of all PCIe ports
4715 pcie_pme= [PCIE,PM] Native PCIe PME signaling options:
4716 nomsi Do not use MSI for native PCIe PME signaling (this makes
4717 all PCIe root ports use INTx for all services).
4719 pcmv= [HW,PCMCIA] BadgePAD 4
4721 pd_ignore_unused
4722 [PM]
4723 Keep all power-domains already enabled by bootloader on,
4724 even if no driver has claimed them. This is useful
4725 for debug and development, but should not be
4726 needed on a platform with proper driver support.
4728 pdcchassis= [PARISC,HW] Disable/Enable PDC Chassis Status codes at
4729 boot time.
4730 Format: { 0 | 1 }
4731 See arch/parisc/kernel/pdc_chassis.c
4733 percpu_alloc= [MM,EARLY]
4734 Select which percpu first chunk allocator to use.
4735 Currently supported values are "embed" and "page".
4736 Archs may support subset or none of the selections.
4737 See comments in mm/percpu.c for details on each
4738 allocator. This parameter is primarily for debugging
4739 and performance comparison.
4741 pirq= [SMP,APIC] Manual mp-table setup
4742 See Documentation/arch/x86/i386/IO-APIC.rst.
4744 plip= [PPT,NET] Parallel port network link
4745 Format: { parport<nr> | timid | 0 }
4746 See also Documentation/admin-guide/parport.rst.
4748 pmtmr= [X86] Manual setup of pmtmr I/O Port.
4749 Override pmtimer IOPort with a hex value.
4750 e.g. pmtmr=0x508
4752 pmu_override= [PPC] Override the PMU.
4753 This option takes over the PMU facility, so it is no
4754 longer usable by perf. Setting this option starts the
4755 PMU counters by setting MMCR0 to 0 (the FC bit is
4756 cleared). If a number is given, then MMCR1 is set to
4757 that number, otherwise (e.g., 'pmu_override=on'), MMCR1
4758 remains 0.
4760 pm_debug_messages [SUSPEND,KNL]
4761 Enable suspend/resume debug messages during boot up.
4763 pnp.debug=1 [PNP]
4764 Enable PNP debug messages (depends on the
4765 CONFIG_PNP_DEBUG_MESSAGES option). Change at run-time
4766 via /sys/module/pnp/parameters/debug. We always show
4767 current resource usage; turning this on also shows
4768 possible settings and some assignment information.
4770 pnpacpi= [ACPI]
4771 { off }
4773 pnpbios= [ISAPNP]
4774 { on | off | curr | res | no-curr | no-res }
4776 pnp_reserve_irq=
4777 [ISAPNP] Exclude IRQs for the autoconfiguration
4779 pnp_reserve_dma=
4780 [ISAPNP] Exclude DMAs for the autoconfiguration
4782 pnp_reserve_io= [ISAPNP] Exclude I/O ports for the autoconfiguration
4783 Ranges are in pairs (I/O port base and size).
4785 pnp_reserve_mem=
4786 [ISAPNP] Exclude memory regions for the
4787 autoconfiguration.
4788 Ranges are in pairs (memory base and size).
4790 ports= [IP_VS_FTP] IPVS ftp helper module
4791 Default is 21.
4792 Up to 8 (IP_VS_APP_MAX_PORTS) ports
4793 may be specified.
4794 Format: <port>,<port>....
4796 possible_cpus= [SMP,S390,X86]
4797 Format: <unsigned int>
4798 Set the number of possible CPUs, overriding the
4799 regular discovery mechanisms (such as ACPI/FW, etc).
4801 powersave=off [PPC] This option disables power saving features.
4802 It specifically disables cpuidle and sets the
4803 platform machine description specific power_save
4804 function to NULL. On Idle the CPU just reduces
4805 execution priority.
4807 ppc_strict_facility_enable
4808 [PPC,ENABLE] This option catches any kernel floating point,
4809 Altivec, VSX and SPE outside of regions specifically
4810 allowed (eg kernel_enable_fpu()/kernel_disable_fpu()).
4811 There is some performance impact when enabling this.
4813 ppc_tm= [PPC,EARLY]
4814 Format: {"off"}
4815 Disable Hardware Transactional Memory
4817 preempt= [KNL]
4818 Select preemption mode if you have CONFIG_PREEMPT_DYNAMIC
4819 none - Limited to cond_resched() calls
4820 voluntary - Limited to cond_resched() and might_sleep() calls
4821 full - Any section that isn't explicitly preempt disabled
4822 can be preempted anytime. Tasks will also yield
4823 contended spinlocks (if the critical section isn't
4824 explicitly preempt disabled beyond the lock itself).
4826 print-fatal-signals=
4827 [KNL] debug: print fatal signals
4829 If enabled, warn about various signal handling
4830 related application anomalies: too many signals,
4831 too many POSIX.1 timers, fatal signals causing a
4832 coredump - etc.
4834 If you hit the warning due to signal overflow,
4835 you might want to try "ulimit -i unlimited".
4837 default: off.
4839 printk.always_kmsg_dump=
4840 Trigger kmsg_dump for cases other than kernel oops or
4841 panics
4842 Format: <bool> (1/Y/y=enable, 0/N/n=disable)
4843 default: disabled
4845 printk.console_no_auto_verbose=
4846 Disable console loglevel raise on oops, panic
4847 or lockdep-detected issues (only if lock debug is on).
4848 With an exception to setups with low baudrate on
4849 serial console, keeping this 0 is a good choice
4850 in order to provide more debug information.
4851 Format: <bool>
4852 default: 0 (auto_verbose is enabled)
4854 printk.devkmsg={on,off,ratelimit}
4855 Control writing to /dev/kmsg.
4856 on - unlimited logging to /dev/kmsg from userspace
4857 off - logging to /dev/kmsg disabled
4858 ratelimit - ratelimit the logging
4859 Default: ratelimit
4861 printk.time= Show timing data prefixed to each printk message line
4862 Format: <bool> (1/Y/y=enable, 0/N/n=disable)
4864 proc_mem.force_override= [KNL]
4865 Format: {always | ptrace | never}
4866 Traditionally /proc/pid/mem allows memory permissions to be
4867 overridden without restrictions. This option may be set to
4868 restrict that. Can be one of:
4869 - 'always': traditional behavior always allows mem overrides.
4870 - 'ptrace': only allow mem overrides for active ptracers.
4871 - 'never': never allow mem overrides.
4872 If not specified, default is the CONFIG_PROC_MEM_* choice.
4874 processor.max_cstate= [HW,ACPI]
4875 Limit processor to maximum C-state
4876 max_cstate=9 overrides any DMI blacklist limit.
4878 processor.nocst [HW,ACPI]
4879 Ignore the _CST method to determine C-states,
4880 instead using the legacy FADT method
4882 profile= [KNL] Enable kernel profiling via /proc/profile
4883 Format: [<profiletype>,]<number>
4884 Param: <profiletype>: "schedule" or "kvm"
4885 [defaults to kernel profiling]
4886 Param: "schedule" - profile schedule points.
4887 Param: "kvm" - profile VM exits.
4888 Param: <number> - step/bucket size as a power of 2 for
4889 statistical time based profiling.
4891 prompt_ramdisk= [RAM] [Deprecated]
4893 prot_virt= [S390] enable hosting protected virtual machines
4894 isolated from the hypervisor (if hardware supports
4895 that). If enabled, the default kernel base address
4896 might be overridden even when Kernel Address Space
4897 Layout Randomization is disabled.
4898 Format: <bool>
4900 psi= [KNL] Enable or disable pressure stall information
4901 tracking.
4902 Format: <bool>
4904 psmouse.proto= [HW,MOUSE] Highest PS2 mouse protocol extension to
4905 probe for; one of (bare|imps|exps|lifebook|any).
4906 psmouse.rate= [HW,MOUSE] Set desired mouse report rate, in reports
4907 per second.
4908 psmouse.resetafter= [HW,MOUSE]
4909 Try to reset the device after so many bad packets
4910 (0 = never).
4911 psmouse.resolution=
4912 [HW,MOUSE] Set desired mouse resolution, in dpi.
4913 psmouse.smartscroll=
4914 [HW,MOUSE] Controls Logitech smartscroll autorepeat.
4915 0 = disabled, 1 = enabled (default).
4917 pstore.backend= Specify the name of the pstore backend to use
4919 pti= [X86-64] Control Page Table Isolation of user and
4920 kernel address spaces. Disabling this feature
4921 removes hardening, but improves performance of
4922 system calls and interrupts.
4924 on - unconditionally enable
4925 off - unconditionally disable
4926 auto - kernel detects whether your CPU model is
4927 vulnerable to issues that PTI mitigates
4929 Not specifying this option is equivalent to pti=auto.
4931 pty.legacy_count=
4932 [KNL] Number of legacy pty's. Overwrites compiled-in
4933 default number.
4935 quiet [KNL,EARLY] Disable most log messages
4937 r128= [HW,DRM]
4939 radix_hcall_invalidate=on [PPC/PSERIES]
4940 Disable RADIX GTSE feature and use hcall for TLB
4941 invalidate.
4943 raid= [HW,RAID]
4944 See Documentation/admin-guide/md.rst.
4946 ramdisk_size= [RAM] Sizes of RAM disks in kilobytes
4947 See Documentation/admin-guide/blockdev/ramdisk.rst.
4949 ramdisk_start= [RAM] RAM disk image start address
4951 random.trust_cpu=off
4952 [KNL,EARLY] Disable trusting the use of the CPU's
4953 random number generator (if available) to
4954 initialize the kernel's RNG.
4956 random.trust_bootloader=off
4957 [KNL,EARLY] Disable trusting the use of the a seed
4958 passed by the bootloader (if available) to
4959 initialize the kernel's RNG.
4961 randomize_kstack_offset=
4962 [KNL,EARLY] Enable or disable kernel stack offset
4963 randomization, which provides roughly 5 bits of
4964 entropy, frustrating memory corruption attacks
4965 that depend on stack address determinism or
4966 cross-syscall address exposures. This is only
4967 available on architectures that have defined
4968 CONFIG_HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET.
4969 Format: <bool> (1/Y/y=enable, 0/N/n=disable)
4970 Default is CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT.
4972 ras=option[,option,...] [KNL] RAS-specific options
4974 cec_disable [X86]
4975 Disable the Correctable Errors Collector,
4976 see CONFIG_RAS_CEC help text.
4978 rcu_nocbs[=cpu-list]
4979 [KNL] The optional argument is a cpu list,
4980 as described above.
4982 In kernels built with CONFIG_RCU_NOCB_CPU=y,
4983 enable the no-callback CPU mode, which prevents
4984 such CPUs' callbacks from being invoked in
4985 softirq context. Invocation of such CPUs' RCU
4986 callbacks will instead be offloaded to "rcuox/N"
4987 kthreads created for that purpose, where "x" is
4988 "p" for RCU-preempt, "s" for RCU-sched, and "g"
4989 for the kthreads that mediate grace periods; and
4990 "N" is the CPU number. This reduces OS jitter on
4991 the offloaded CPUs, which can be useful for HPC
4992 and real-time workloads. It can also improve
4993 energy efficiency for asymmetric multiprocessors.
4995 If a cpulist is passed as an argument, the specified
4996 list of CPUs is set to no-callback mode from boot.
4998 Otherwise, if the '=' sign and the cpulist
4999 arguments are omitted, no CPU will be set to
5000 no-callback mode from boot but the mode may be
5001 toggled at runtime via cpusets.
5003 Note that this argument takes precedence over
5004 the CONFIG_RCU_NOCB_CPU_DEFAULT_ALL option.
5006 rcu_nocb_poll [KNL]
5007 Rather than requiring that offloaded CPUs
5008 (specified by rcu_nocbs= above) explicitly
5009 awaken the corresponding "rcuoN" kthreads,
5010 make these kthreads poll for callbacks.
5011 This improves the real-time response for the
5012 offloaded CPUs by relieving them of the need to
5013 wake up the corresponding kthread, but degrades
5014 energy efficiency by requiring that the kthreads
5015 periodically wake up to do the polling.
5017 rcutree.blimit= [KNL]
5018 Set maximum number of finished RCU callbacks to
5019 process in one batch.
5021 rcutree.csd_lock_suppress_rcu_stall= [KNL]
5022 Do only a one-line RCU CPU stall warning when
5023 there is an ongoing too-long CSD-lock wait.
5025 rcutree.do_rcu_barrier= [KNL]
5026 Request a call to rcu_barrier(). This is
5027 throttled so that userspace tests can safely
5028 hammer on the sysfs variable if they so choose.
5029 If triggered before the RCU grace-period machinery
5030 is fully active, this will error out with EAGAIN.
5032 rcutree.dump_tree= [KNL]
5033 Dump the structure of the rcu_node combining tree
5034 out at early boot. This is used for diagnostic
5035 purposes, to verify correct tree setup.
5037 rcutree.gp_cleanup_delay= [KNL]
5038 Set the number of jiffies to delay each step of
5039 RCU grace-period cleanup.
5041 rcutree.gp_init_delay= [KNL]
5042 Set the number of jiffies to delay each step of
5043 RCU grace-period initialization.
5045 rcutree.gp_preinit_delay= [KNL]
5046 Set the number of jiffies to delay each step of
5047 RCU grace-period pre-initialization, that is,
5048 the propagation of recent CPU-hotplug changes up
5049 the rcu_node combining tree.
5051 rcutree.jiffies_till_first_fqs= [KNL]
5052 Set delay from grace-period initialization to
5053 first attempt to force quiescent states.
5054 Units are jiffies, minimum value is zero,
5055 and maximum value is HZ.
5057 rcutree.jiffies_till_next_fqs= [KNL]
5058 Set delay between subsequent attempts to force
5059 quiescent states. Units are jiffies, minimum
5060 value is one, and maximum value is HZ.
5062 rcutree.jiffies_till_sched_qs= [KNL]
5063 Set required age in jiffies for a
5064 given grace period before RCU starts
5065 soliciting quiescent-state help from
5066 rcu_note_context_switch() and cond_resched().
5067 If not specified, the kernel will calculate
5068 a value based on the most recent settings
5069 of rcutree.jiffies_till_first_fqs
5070 and rcutree.jiffies_till_next_fqs.
5071 This calculated value may be viewed in
5072 rcutree.jiffies_to_sched_qs. Any attempt to set
5073 rcutree.jiffies_to_sched_qs will be cheerfully
5074 overwritten.
5076 rcutree.kthread_prio= [KNL,BOOT]
5077 Set the SCHED_FIFO priority of the RCU per-CPU
5078 kthreads (rcuc/N). This value is also used for
5079 the priority of the RCU boost threads (rcub/N)
5080 and for the RCU grace-period kthreads (rcu_bh,
5081 rcu_preempt, and rcu_sched). If RCU_BOOST is
5082 set, valid values are 1-99 and the default is 1
5083 (the least-favored priority). Otherwise, when
5084 RCU_BOOST is not set, valid values are 0-99 and
5085 the default is zero (non-realtime operation).
5086 When RCU_NOCB_CPU is set, also adjust the
5087 priority of NOCB callback kthreads.
5089 rcutree.nocb_nobypass_lim_per_jiffy= [KNL]
5090 On callback-offloaded (rcu_nocbs) CPUs,
5091 RCU reduces the lock contention that would
5092 otherwise be caused by callback floods through
5093 use of the ->nocb_bypass list. However, in the
5094 common non-flooded case, RCU queues directly to
5095 the main ->cblist in order to avoid the extra
5096 overhead of the ->nocb_bypass list and its lock.
5097 But if there are too many callbacks queued during
5098 a single jiffy, RCU pre-queues the callbacks into
5099 the ->nocb_bypass queue. The definition of "too
5100 many" is supplied by this kernel boot parameter.
5102 rcutree.nohz_full_patience_delay= [KNL]
5103 On callback-offloaded (rcu_nocbs) CPUs, avoid
5104 disturbing RCU unless the grace period has
5105 reached the specified age in milliseconds.
5106 Defaults to zero. Large values will be capped
5107 at five seconds. All values will be rounded down
5108 to the nearest value representable by jiffies.
5110 rcutree.qhimark= [KNL]
5111 Set threshold of queued RCU callbacks beyond which
5112 batch limiting is disabled.
5114 rcutree.qlowmark= [KNL]
5115 Set threshold of queued RCU callbacks below which
5116 batch limiting is re-enabled.
5118 rcutree.qovld= [KNL]
5119 Set threshold of queued RCU callbacks beyond which
5120 RCU's force-quiescent-state scan will aggressively
5121 enlist help from cond_resched() and sched IPIs to
5122 help CPUs more quickly reach quiescent states.
5123 Set to less than zero to make this be set based
5124 on rcutree.qhimark at boot time and to zero to
5125 disable more aggressive help enlistment.
5127 rcutree.rcu_delay_page_cache_fill_msec= [KNL]
5128 Set the page-cache refill delay (in milliseconds)
5129 in response to low-memory conditions. The range
5130 of permitted values is in the range 0:100000.
5132 rcutree.rcu_divisor= [KNL]
5133 Set the shift-right count to use to compute
5134 the callback-invocation batch limit bl from
5135 the number of callbacks queued on this CPU.
5136 The result will be bounded below by the value of
5137 the rcutree.blimit kernel parameter. Every bl
5138 callbacks, the softirq handler will exit in
5139 order to allow the CPU to do other work.
5141 Please note that this callback-invocation batch
5142 limit applies only to non-offloaded callback
5143 invocation. Offloaded callbacks are instead
5144 invoked in the context of an rcuoc kthread, which
5145 scheduler will preempt as it does any other task.
5147 rcutree.rcu_fanout_exact= [KNL]
5148 Disable autobalancing of the rcu_node combining
5149 tree. This is used by rcutorture, and might
5150 possibly be useful for architectures having high
5151 cache-to-cache transfer latencies.
5153 rcutree.rcu_fanout_leaf= [KNL]
5154 Change the number of CPUs assigned to each
5155 leaf rcu_node structure. Useful for very
5156 large systems, which will choose the value 64,
5157 and for NUMA systems with large remote-access
5158 latencies, which will choose a value aligned
5159 with the appropriate hardware boundaries.
5161 rcutree.rcu_min_cached_objs= [KNL]
5162 Minimum number of objects which are cached and
5163 maintained per one CPU. Object size is equal
5164 to PAGE_SIZE. The cache allows to reduce the
5165 pressure to page allocator, also it makes the
5166 whole algorithm to behave better in low memory
5167 condition.
5169 rcutree.rcu_nocb_gp_stride= [KNL]
5170 Set the number of NOCB callback kthreads in
5171 each group, which defaults to the square root
5172 of the number of CPUs. Larger numbers reduce
5173 the wakeup overhead on the global grace-period
5174 kthread, but increases that same overhead on
5175 each group's NOCB grace-period kthread.
5177 rcutree.rcu_kick_kthreads= [KNL]
5178 Cause the grace-period kthread to get an extra
5179 wake_up() if it sleeps three times longer than
5180 it should at force-quiescent-state time.
5181 This wake_up() will be accompanied by a
5182 WARN_ONCE() splat and an ftrace_dump().
5184 rcutree.rcu_resched_ns= [KNL]
5185 Limit the time spend invoking a batch of RCU
5186 callbacks to the specified number of nanoseconds.
5187 By default, this limit is checked only once
5188 every 32 callbacks in order to limit the pain
5189 inflicted by local_clock() overhead.
5191 rcutree.rcu_unlock_delay= [KNL]
5192 In CONFIG_RCU_STRICT_GRACE_PERIOD=y kernels,
5193 this specifies an rcu_read_unlock()-time delay
5194 in microseconds. This defaults to zero.
5195 Larger delays increase the probability of
5196 catching RCU pointer leaks, that is, buggy use
5197 of RCU-protected pointers after the relevant
5198 rcu_read_unlock() has completed.
5200 rcutree.sysrq_rcu= [KNL]
5201 Commandeer a sysrq key to dump out Tree RCU's
5202 rcu_node tree with an eye towards determining
5203 why a new grace period has not yet started.
5205 rcutree.use_softirq= [KNL]
5206 If set to zero, move all RCU_SOFTIRQ processing to
5207 per-CPU rcuc kthreads. Defaults to a non-zero
5208 value, meaning that RCU_SOFTIRQ is used by default.
5209 Specify rcutree.use_softirq=0 to use rcuc kthreads.
5211 But note that CONFIG_PREEMPT_RT=y kernels disable
5212 this kernel boot parameter, forcibly setting it
5213 to zero.
5215 rcutree.enable_rcu_lazy= [KNL]
5216 To save power, batch RCU callbacks and flush after
5217 delay, memory pressure or callback list growing too
5218 big.
5220 rcutree.rcu_normal_wake_from_gp= [KNL]
5221 Reduces a latency of synchronize_rcu() call. This approach
5222 maintains its own track of synchronize_rcu() callers, so it
5223 does not interact with regular callbacks because it does not
5224 use a call_rcu[_hurry]() path. Please note, this is for a
5225 normal grace period.
5227 How to enable it:
5229 echo 1 > /sys/module/rcutree/parameters/rcu_normal_wake_from_gp
5230 or pass a boot parameter "rcutree.rcu_normal_wake_from_gp=1"
5232 Default is 0.
5234 rcuscale.gp_async= [KNL]
5235 Measure performance of asynchronous
5236 grace-period primitives such as call_rcu().
5238 rcuscale.gp_async_max= [KNL]
5239 Specify the maximum number of outstanding
5240 callbacks per writer thread. When a writer
5241 thread exceeds this limit, it invokes the
5242 corresponding flavor of rcu_barrier() to allow
5243 previously posted callbacks to drain.
5245 rcuscale.gp_exp= [KNL]
5246 Measure performance of expedited synchronous
5247 grace-period primitives.
5249 rcuscale.holdoff= [KNL]
5250 Set test-start holdoff period. The purpose of
5251 this parameter is to delay the start of the
5252 test until boot completes in order to avoid
5253 interference.
5255 rcuscale.kfree_by_call_rcu= [KNL]
5256 In kernels built with CONFIG_RCU_LAZY=y, test
5257 call_rcu() instead of kfree_rcu().
5259 rcuscale.kfree_mult= [KNL]
5260 Instead of allocating an object of size kfree_obj,
5261 allocate one of kfree_mult * sizeof(kfree_obj).
5262 Defaults to 1.
5264 rcuscale.kfree_rcu_test= [KNL]
5265 Set to measure performance of kfree_rcu() flooding.
5267 rcuscale.kfree_rcu_test_double= [KNL]
5268 Test the double-argument variant of kfree_rcu().
5269 If this parameter has the same value as
5270 rcuscale.kfree_rcu_test_single, both the single-
5271 and double-argument variants are tested.
5273 rcuscale.kfree_rcu_test_single= [KNL]
5274 Test the single-argument variant of kfree_rcu().
5275 If this parameter has the same value as
5276 rcuscale.kfree_rcu_test_double, both the single-
5277 and double-argument variants are tested.
5279 rcuscale.kfree_nthreads= [KNL]
5280 The number of threads running loops of kfree_rcu().
5282 rcuscale.kfree_alloc_num= [KNL]
5283 Number of allocations and frees done in an iteration.
5285 rcuscale.kfree_loops= [KNL]
5286 Number of loops doing rcuscale.kfree_alloc_num number
5287 of allocations and frees.
5289 rcuscale.minruntime= [KNL]
5290 Set the minimum test run time in seconds. This
5291 does not affect the data-collection interval,
5292 but instead allows better measurement of things
5293 like CPU consumption.
5295 rcuscale.nreaders= [KNL]
5296 Set number of RCU readers. The value -1 selects
5297 N, where N is the number of CPUs. A value
5298 "n" less than -1 selects N-n+1, where N is again
5299 the number of CPUs. For example, -2 selects N
5300 (the number of CPUs), -3 selects N+1, and so on.
5301 A value of "n" less than or equal to -N selects
5302 a single reader.
5304 rcuscale.nwriters= [KNL]
5305 Set number of RCU writers. The values operate
5306 the same as for rcuscale.nreaders.
5307 N, where N is the number of CPUs
5309 rcuscale.scale_type= [KNL]
5310 Specify the RCU implementation to test.
5312 rcuscale.shutdown= [KNL]
5313 Shut the system down after performance tests
5314 complete. This is useful for hands-off automated
5315 testing.
5317 rcuscale.verbose= [KNL]
5318 Enable additional printk() statements.
5320 rcuscale.writer_holdoff= [KNL]
5321 Write-side holdoff between grace periods,
5322 in microseconds. The default of zero says
5323 no holdoff.
5325 rcuscale.writer_holdoff_jiffies= [KNL]
5326 Additional write-side holdoff between grace
5327 periods, but in jiffies. The default of zero
5328 says no holdoff.
5330 rcutorture.fqs_duration= [KNL]
5331 Set duration of force_quiescent_state bursts
5332 in microseconds.
5334 rcutorture.fqs_holdoff= [KNL]
5335 Set holdoff time within force_quiescent_state bursts
5336 in microseconds.
5338 rcutorture.fqs_stutter= [KNL]
5339 Set wait time between force_quiescent_state bursts
5340 in seconds.
5342 rcutorture.fwd_progress= [KNL]
5343 Specifies the number of kthreads to be used
5344 for RCU grace-period forward-progress testing
5345 for the types of RCU supporting this notion.
5346 Defaults to 1 kthread, values less than zero or
5347 greater than the number of CPUs cause the number
5348 of CPUs to be used.
5350 rcutorture.fwd_progress_div= [KNL]
5351 Specify the fraction of a CPU-stall-warning
5352 period to do tight-loop forward-progress testing.
5354 rcutorture.fwd_progress_holdoff= [KNL]
5355 Number of seconds to wait between successive
5356 forward-progress tests.
5358 rcutorture.fwd_progress_need_resched= [KNL]
5359 Enclose cond_resched() calls within checks for
5360 need_resched() during tight-loop forward-progress
5361 testing.
5363 rcutorture.gp_cond= [KNL]
5364 Use conditional/asynchronous update-side
5365 primitives, if available.
5367 rcutorture.gp_exp= [KNL]
5368 Use expedited update-side primitives, if available.
5370 rcutorture.gp_normal= [KNL]
5371 Use normal (non-expedited) asynchronous
5372 update-side primitives, if available.
5374 rcutorture.gp_sync= [KNL]
5375 Use normal (non-expedited) synchronous
5376 update-side primitives, if available. If all
5377 of rcutorture.gp_cond=, rcutorture.gp_exp=,
5378 rcutorture.gp_normal=, and rcutorture.gp_sync=
5379 are zero, rcutorture acts as if is interpreted
5380 they are all non-zero.
5382 rcutorture.irqreader= [KNL]
5383 Run RCU readers from irq handlers, or, more
5384 accurately, from a timer handler. Not all RCU
5385 flavors take kindly to this sort of thing.
5387 rcutorture.leakpointer= [KNL]
5388 Leak an RCU-protected pointer out of the reader.
5389 This can of course result in splats, and is
5390 intended to test the ability of things like
5391 CONFIG_RCU_STRICT_GRACE_PERIOD=y to detect
5392 such leaks.
5394 rcutorture.n_barrier_cbs= [KNL]
5395 Set callbacks/threads for rcu_barrier() testing.
5397 rcutorture.nfakewriters= [KNL]
5398 Set number of concurrent RCU writers. These just
5399 stress RCU, they don't participate in the actual
5400 test, hence the "fake".
5402 rcutorture.nocbs_nthreads= [KNL]
5403 Set number of RCU callback-offload togglers.
5404 Zero (the default) disables toggling.
5406 rcutorture.nocbs_toggle= [KNL]
5407 Set the delay in milliseconds between successive
5408 callback-offload toggling attempts.
5410 rcutorture.nreaders= [KNL]
5411 Set number of RCU readers. The value -1 selects
5412 N-1, where N is the number of CPUs. A value
5413 "n" less than -1 selects N-n-2, where N is again
5414 the number of CPUs. For example, -2 selects N
5415 (the number of CPUs), -3 selects N+1, and so on.
5417 rcutorture.object_debug= [KNL]
5418 Enable debug-object double-call_rcu() testing.
5420 rcutorture.onoff_holdoff= [KNL]
5421 Set time (s) after boot for CPU-hotplug testing.
5423 rcutorture.onoff_interval= [KNL]
5424 Set time (jiffies) between CPU-hotplug operations,
5425 or zero to disable CPU-hotplug testing.
5427 rcutorture.read_exit_burst= [KNL]
5428 The number of times in a given read-then-exit
5429 episode that a set of read-then-exit kthreads
5430 is spawned.
5432 rcutorture.read_exit_delay= [KNL]
5433 The delay, in seconds, between successive
5434 read-then-exit testing episodes.
5436 rcutorture.reader_flavor= [KNL]
5437 A bit mask indicating which readers to use.
5438 If there is more than one bit set, the readers
5439 are entered from low-order bit up, and are
5440 exited in the opposite order. For SRCU, the
5441 0x1 bit is normal readers, 0x2 NMI-safe readers,
5442 and 0x4 light-weight readers.
5444 rcutorture.shuffle_interval= [KNL]
5445 Set task-shuffle interval (s). Shuffling tasks
5446 allows some CPUs to go into dyntick-idle mode
5447 during the rcutorture test.
5449 rcutorture.shutdown_secs= [KNL]
5450 Set time (s) after boot system shutdown. This
5451 is useful for hands-off automated testing.
5453 rcutorture.stall_cpu= [KNL]
5454 Duration of CPU stall (s) to test RCU CPU stall
5455 warnings, zero to disable.
5457 rcutorture.stall_cpu_block= [KNL]
5458 Sleep while stalling if set. This will result
5459 in warnings from preemptible RCU in addition to
5460 any other stall-related activity. Note that
5461 in kernels built with CONFIG_PREEMPTION=n and
5462 CONFIG_PREEMPT_COUNT=y, this parameter will
5463 cause the CPU to pass through a quiescent state.
5464 Given CONFIG_PREEMPTION=n, this will suppress
5465 RCU CPU stall warnings, but will instead result
5466 in scheduling-while-atomic splats.
5468 Use of this module parameter results in splats.
5471 rcutorture.stall_cpu_holdoff= [KNL]
5472 Time to wait (s) after boot before inducing stall.
5474 rcutorture.stall_cpu_irqsoff= [KNL]
5475 Disable interrupts while stalling if set, but only
5476 on the first stall in the set.
5478 rcutorture.stall_cpu_repeat= [KNL]
5479 Number of times to repeat the stall sequence,
5480 so that rcutorture.stall_cpu_repeat=3 will result
5481 in four stall sequences.
5483 rcutorture.stall_gp_kthread= [KNL]
5484 Duration (s) of forced sleep within RCU
5485 grace-period kthread to test RCU CPU stall
5486 warnings, zero to disable. If both stall_cpu
5487 and stall_gp_kthread are specified, the
5488 kthread is starved first, then the CPU.
5490 rcutorture.stat_interval= [KNL]
5491 Time (s) between statistics printk()s.
5493 rcutorture.stutter= [KNL]
5494 Time (s) to stutter testing, for example, specifying
5495 five seconds causes the test to run for five seconds,
5496 wait for five seconds, and so on. This tests RCU's
5497 ability to transition abruptly to and from idle.
5499 rcutorture.test_boost= [KNL]
5500 Test RCU priority boosting? 0=no, 1=maybe, 2=yes.
5501 "Maybe" means test if the RCU implementation
5502 under test support RCU priority boosting.
5504 rcutorture.test_boost_duration= [KNL]
5505 Duration (s) of each individual boost test.
5507 rcutorture.test_boost_interval= [KNL]
5508 Interval (s) between each boost test.
5510 rcutorture.test_no_idle_hz= [KNL]
5511 Test RCU's dyntick-idle handling. See also the
5512 rcutorture.shuffle_interval parameter.
5514 rcutorture.torture_type= [KNL]
5515 Specify the RCU implementation to test.
5517 rcutorture.verbose= [KNL]
5518 Enable additional printk() statements.
5520 rcupdate.rcu_cpu_stall_ftrace_dump= [KNL]
5521 Dump ftrace buffer after reporting RCU CPU
5522 stall warning.
5524 rcupdate.rcu_cpu_stall_notifiers= [KNL]
5525 Provide RCU CPU stall notifiers, but see the
5526 warnings in the RCU_CPU_STALL_NOTIFIER Kconfig
5527 option's help text. TL;DR: You almost certainly
5528 do not want rcupdate.rcu_cpu_stall_notifiers.
5530 rcupdate.rcu_cpu_stall_suppress= [KNL]
5531 Suppress RCU CPU stall warning messages.
5533 rcupdate.rcu_cpu_stall_suppress_at_boot= [KNL]
5534 Suppress RCU CPU stall warning messages and
5535 rcutorture writer stall warnings that occur
5536 during early boot, that is, during the time
5537 before the init task is spawned.
5539 rcupdate.rcu_cpu_stall_timeout= [KNL]
5540 Set timeout for RCU CPU stall warning messages.
5541 The value is in seconds and the maximum allowed
5542 value is 300 seconds.
5544 rcupdate.rcu_exp_cpu_stall_timeout= [KNL]
5545 Set timeout for expedited RCU CPU stall warning
5546 messages. The value is in milliseconds
5547 and the maximum allowed value is 21000
5548 milliseconds. Please note that this value is
5549 adjusted to an arch timer tick resolution.
5550 Setting this to zero causes the value from
5551 rcupdate.rcu_cpu_stall_timeout to be used (after
5552 conversion from seconds to milliseconds).
5554 rcupdate.rcu_cpu_stall_cputime= [KNL]
5555 Provide statistics on the cputime and count of
5556 interrupts and tasks during the sampling period. For
5557 multiple continuous RCU stalls, all sampling periods
5558 begin at half of the first RCU stall timeout.
5560 rcupdate.rcu_exp_stall_task_details= [KNL]
5561 Print stack dumps of any tasks blocking the
5562 current expedited RCU grace period during an
5563 expedited RCU CPU stall warning.
5565 rcupdate.rcu_expedited= [KNL]
5566 Use expedited grace-period primitives, for
5567 example, synchronize_rcu_expedited() instead
5568 of synchronize_rcu(). This reduces latency,
5569 but can increase CPU utilization, degrade
5570 real-time latency, and degrade energy efficiency.
5571 No effect on CONFIG_TINY_RCU kernels.
5573 rcupdate.rcu_normal= [KNL]
5574 Use only normal grace-period primitives,
5575 for example, synchronize_rcu() instead of
5576 synchronize_rcu_expedited(). This improves
5577 real-time latency, CPU utilization, and
5578 energy efficiency, but can expose users to
5579 increased grace-period latency. This parameter
5580 overrides rcupdate.rcu_expedited. No effect on
5581 CONFIG_TINY_RCU kernels.
5583 rcupdate.rcu_normal_after_boot= [KNL]
5584 Once boot has completed (that is, after
5585 rcu_end_inkernel_boot() has been invoked), use
5586 only normal grace-period primitives. No effect
5587 on CONFIG_TINY_RCU kernels.
5589 But note that CONFIG_PREEMPT_RT=y kernels enables
5590 this kernel boot parameter, forcibly setting
5591 it to the value one, that is, converting any
5592 post-boot attempt at an expedited RCU grace
5593 period to instead use normal non-expedited
5594 grace-period processing.
5596 rcupdate.rcu_task_collapse_lim= [KNL]
5597 Set the maximum number of callbacks present
5598 at the beginning of a grace period that allows
5599 the RCU Tasks flavors to collapse back to using
5600 a single callback queue. This switching only
5601 occurs when rcupdate.rcu_task_enqueue_lim is
5602 set to the default value of -1.
5604 rcupdate.rcu_task_contend_lim= [KNL]
5605 Set the minimum number of callback-queuing-time
5606 lock-contention events per jiffy required to
5607 cause the RCU Tasks flavors to switch to per-CPU
5608 callback queuing. This switching only occurs
5609 when rcupdate.rcu_task_enqueue_lim is set to
5610 the default value of -1.
5612 rcupdate.rcu_task_enqueue_lim= [KNL]
5613 Set the number of callback queues to use for the
5614 RCU Tasks family of RCU flavors. The default
5615 of -1 allows this to be automatically (and
5616 dynamically) adjusted. This parameter is intended
5617 for use in testing.
5619 rcupdate.rcu_task_ipi_delay= [KNL]
5620 Set time in jiffies during which RCU tasks will
5621 avoid sending IPIs, starting with the beginning
5622 of a given grace period. Setting a large
5623 number avoids disturbing real-time workloads,
5624 but lengthens grace periods.
5626 rcupdate.rcu_task_lazy_lim= [KNL]
5627 Number of callbacks on a given CPU that will
5628 cancel laziness on that CPU. Use -1 to disable
5629 cancellation of laziness, but be advised that
5630 doing so increases the danger of OOM due to
5631 callback flooding.
5633 rcupdate.rcu_task_stall_info= [KNL]
5634 Set initial timeout in jiffies for RCU task stall
5635 informational messages, which give some indication
5636 of the problem for those not patient enough to
5637 wait for ten minutes. Informational messages are
5638 only printed prior to the stall-warning message
5639 for a given grace period. Disable with a value
5640 less than or equal to zero. Defaults to ten
5641 seconds. A change in value does not take effect
5642 until the beginning of the next grace period.
5644 rcupdate.rcu_task_stall_info_mult= [KNL]
5645 Multiplier for time interval between successive
5646 RCU task stall informational messages for a given
5647 RCU tasks grace period. This value is clamped
5648 to one through ten, inclusive. It defaults to
5649 the value three, so that the first informational
5650 message is printed 10 seconds into the grace
5651 period, the second at 40 seconds, the third at
5652 160 seconds, and then the stall warning at 600
5653 seconds would prevent a fourth at 640 seconds.
5655 rcupdate.rcu_task_stall_timeout= [KNL]
5656 Set timeout in jiffies for RCU task stall
5657 warning messages. Disable with a value less
5658 than or equal to zero. Defaults to ten minutes.
5659 A change in value does not take effect until
5660 the beginning of the next grace period.
5662 rcupdate.rcu_tasks_lazy_ms= [KNL]
5663 Set timeout in milliseconds RCU Tasks asynchronous
5664 callback batching for call_rcu_tasks().
5665 A negative value will take the default. A value
5666 of zero will disable batching. Batching is
5667 always disabled for synchronize_rcu_tasks().
5669 rcupdate.rcu_tasks_trace_lazy_ms= [KNL]
5670 Set timeout in milliseconds RCU Tasks
5671 Trace asynchronous callback batching for
5672 call_rcu_tasks_trace(). A negative value
5673 will take the default. A value of zero will
5674 disable batching. Batching is always disabled
5675 for synchronize_rcu_tasks_trace().
5677 rcupdate.rcu_self_test= [KNL]
5678 Run the RCU early boot self tests
5680 rdinit= [KNL]
5681 Format: <full_path>
5682 Run specified binary instead of /init from the ramdisk,
5683 used for early userspace startup. See initrd.
5685 rdrand= [X86,EARLY]
5686 force - Override the decision by the kernel to hide the
5687 advertisement of RDRAND support (this affects
5688 certain AMD processors because of buggy BIOS
5689 support, specifically around the suspend/resume
5690 path).
5692 rdt= [HW,X86,RDT]
5693 Turn on/off individual RDT features. List is:
5694 cmt, mbmtotal, mbmlocal, l3cat, l3cdp, l2cat, l2cdp,
5695 mba, smba, bmec.
5696 E.g. to turn on cmt and turn off mba use:
5697 rdt=cmt,!mba
5699 reboot= [KNL]
5700 Format (x86 or x86_64):
5701 [w[arm] | c[old] | h[ard] | s[oft] | g[pio]] | d[efault] \
5702 [[,]s[mp]#### \
5703 [[,]b[ios] | a[cpi] | k[bd] | t[riple] | e[fi] | p[ci]] \
5704 [[,]f[orce]
5705 Where reboot_mode is one of warm (soft) or cold (hard) or gpio
5706 (prefix with 'panic_' to set mode for panic
5707 reboot only),
5708 reboot_type is one of bios, acpi, kbd, triple, efi, or pci,
5709 reboot_force is either force or not specified,
5710 reboot_cpu is s[mp]#### with #### being the processor
5711 to be used for rebooting.
5713 refscale.holdoff= [KNL]
5714 Set test-start holdoff period. The purpose of
5715 this parameter is to delay the start of the
5716 test until boot completes in order to avoid
5717 interference.
5719 refscale.lookup_instances= [KNL]
5720 Number of data elements to use for the forms of
5721 SLAB_TYPESAFE_BY_RCU testing. A negative number
5722 is negated and multiplied by nr_cpu_ids, while
5723 zero specifies nr_cpu_ids.
5725 refscale.loops= [KNL]
5726 Set the number of loops over the synchronization
5727 primitive under test. Increasing this number
5728 reduces noise due to loop start/end overhead,
5729 but the default has already reduced the per-pass
5730 noise to a handful of picoseconds on ca. 2020
5731 x86 laptops.
5733 refscale.nreaders= [KNL]
5734 Set number of readers. The default value of -1
5735 selects N, where N is roughly 75% of the number
5736 of CPUs. A value of zero is an interesting choice.
5738 refscale.nruns= [KNL]
5739 Set number of runs, each of which is dumped onto
5740 the console log.
5742 refscale.readdelay= [KNL]
5743 Set the read-side critical-section duration,
5744 measured in microseconds.
5746 refscale.scale_type= [KNL]
5747 Specify the read-protection implementation to test.
5749 refscale.shutdown= [KNL]
5750 Shut down the system at the end of the performance
5751 test. This defaults to 1 (shut it down) when
5752 refscale is built into the kernel and to 0 (leave
5753 it running) when refscale is built as a module.
5755 refscale.verbose= [KNL]
5756 Enable additional printk() statements.
5758 refscale.verbose_batched= [KNL]
5759 Batch the additional printk() statements. If zero
5760 (the default) or negative, print everything. Otherwise,
5761 print every Nth verbose statement, where N is the value
5762 specified.
5764 regulator_ignore_unused
5765 [REGULATOR]
5766 Prevents regulator framework from disabling regulators
5767 that are unused, due no driver claiming them. This may
5768 be useful for debug and development, but should not be
5769 needed on a platform with proper driver support.
5771 relax_domain_level=
5772 [KNL, SMP] Set scheduler's default relax_domain_level.
5773 See Documentation/admin-guide/cgroup-v1/cpusets.rst.
5775 reserve= [KNL,BUGS] Force kernel to ignore I/O ports or memory
5776 Format: <base1>,<size1>[,<base2>,<size2>,...]
5777 Reserve I/O ports or memory so the kernel won't use
5778 them. If <base> is less than 0x10000, the region
5779 is assumed to be I/O ports; otherwise it is memory.
5781 reserve_mem= [RAM]
5782 Format: nn[KNG]:<align>:<label>
5783 Reserve physical memory and label it with a name that
5784 other subsystems can use to access it. This is typically
5785 used for systems that do not wipe the RAM, and this command
5786 line will try to reserve the same physical memory on
5787 soft reboots. Note, it is not guaranteed to be the same
5788 location. For example, if anything about the system changes
5789 or if booting a different kernel. It can also fail if KASLR
5790 places the kernel at the location of where the RAM reservation
5791 was from a previous boot, the new reservation will be at a
5792 different location.
5793 Any subsystem using this feature must add a way to verify
5794 that the contents of the physical memory is from a previous
5795 boot, as there may be cases where the memory will not be
5796 located at the same location.
5798 The format is size:align:label for example, to request
5799 12 megabytes of 4096 alignment for ramoops:
5801 reserve_mem=12M:4096:oops ramoops.mem_name=oops
5803 reservetop= [X86-32,EARLY]
5804 Format: nn[KMG]
5805 Reserves a hole at the top of the kernel virtual
5806 address space.
5808 reset_devices [KNL] Force drivers to reset the underlying device
5809 during initialization.
5811 resume= [SWSUSP]
5812 Specify the partition device for software suspend
5813 Format:
5814 {/dev/<dev> | PARTUUID=<uuid> | <int>:<int> | <hex>}
5816 resume_offset= [SWSUSP]
5817 Specify the offset from the beginning of the partition
5818 given by "resume=" at which the swap header is located,
5819 in <PAGE_SIZE> units (needed only for swap files).
5820 See Documentation/power/swsusp-and-swap-files.rst
5822 resumedelay= [HIBERNATION] Delay (in seconds) to pause before attempting to
5823 read the resume files
5825 resumewait [HIBERNATION] Wait (indefinitely) for resume device to show up.
5826 Useful for devices that are detected asynchronously
5827 (e.g. USB and MMC devices).
5829 retain_initrd [RAM] Keep initrd memory after extraction. After boot, it will
5830 be accessible via /sys/firmware/initrd.
5832 retbleed= [X86] Control mitigation of RETBleed (Arbitrary
5833 Speculative Code Execution with Return Instructions)
5834 vulnerability.
5836 AMD-based UNRET and IBPB mitigations alone do not stop
5837 sibling threads from influencing the predictions of other
5838 sibling threads. For that reason, STIBP is used on pro-
5839 cessors that support it, and mitigate SMT on processors
5840 that don't.
5842 off - no mitigation
5843 auto - automatically select a migitation
5844 auto,nosmt - automatically select a mitigation,
5845 disabling SMT if necessary for
5846 the full mitigation (only on Zen1
5847 and older without STIBP).
5848 ibpb - On AMD, mitigate short speculation
5849 windows on basic block boundaries too.
5850 Safe, highest perf impact. It also
5851 enables STIBP if present. Not suitable
5852 on Intel.
5853 ibpb,nosmt - Like "ibpb" above but will disable SMT
5854 when STIBP is not available. This is
5855 the alternative for systems which do not
5856 have STIBP.
5857 unret - Force enable untrained return thunks,
5858 only effective on AMD f15h-f17h based
5859 systems.
5860 unret,nosmt - Like unret, but will disable SMT when STIBP
5861 is not available. This is the alternative for
5862 systems which do not have STIBP.
5864 Selecting 'auto' will choose a mitigation method at run
5865 time according to the CPU.
5867 Not specifying this option is equivalent to retbleed=auto.
5869 rfkill.default_state=
5870 0 "airplane mode". All wifi, bluetooth, wimax, gps, fm,
5871 etc. communication is blocked by default.
5872 1 Unblocked.
5874 rfkill.master_switch_mode=
5875 0 The "airplane mode" button does nothing.
5876 1 The "airplane mode" button toggles between everything
5877 blocked and the previous configuration.
5878 2 The "airplane mode" button toggles between everything
5879 blocked and everything unblocked.
5881 ring3mwait=disable
5882 [KNL] Disable ring 3 MONITOR/MWAIT feature on supported
5883 CPUs.
5885 riscv_isa_fallback [RISCV,EARLY]
5886 When CONFIG_RISCV_ISA_FALLBACK is not enabled, permit
5887 falling back to detecting extension support by parsing
5888 "riscv,isa" property on devicetree systems when the
5889 replacement properties are not found. See the Kconfig
5890 entry for RISCV_ISA_FALLBACK.
5892 ro [KNL] Mount root device read-only on boot
5894 rodata= [KNL,EARLY]
5895 on Mark read-only kernel memory as read-only (default).
5896 off Leave read-only kernel memory writable for debugging.
5897 full Mark read-only kernel memory and aliases as read-only
5898 [arm64]
5900 rockchip.usb_uart
5901 [EARLY]
5902 Enable the uart passthrough on the designated usb port
5903 on Rockchip SoCs. When active, the signals of the
5904 debug-uart get routed to the D+ and D- pins of the usb
5905 port and the regular usb controller gets disabled.
5907 root= [KNL] Root filesystem
5908 Usually this a a block device specifier of some kind,
5909 see the early_lookup_bdev comment in
5910 block/early-lookup.c for details.
5911 Alternatively this can be "ram" for the legacy initial
5912 ramdisk, "nfs" and "cifs" for root on a network file
5913 system, or "mtd" and "ubi" for mounting from raw flash.
5915 rootdelay= [KNL] Delay (in seconds) to pause before attempting to
5916 mount the root filesystem
5918 rootflags= [KNL] Set root filesystem mount option string
5920 rootfstype= [KNL] Set root filesystem type
5922 rootwait [KNL] Wait (indefinitely) for root device to show up.
5923 Useful for devices that are detected asynchronously
5924 (e.g. USB and MMC devices).
5926 rootwait= [KNL] Maximum time (in seconds) to wait for root device
5927 to show up before attempting to mount the root
5928 filesystem.
5930 rproc_mem=nn[KMG][@address]
5931 [KNL,ARM,CMA] Remoteproc physical memory block.
5932 Memory area to be used by remote processor image,
5933 managed by CMA.
5935 rw [KNL] Mount root device read-write on boot
5937 S [KNL] Run init in single mode
5939 s390_iommu= [HW,S390]
5940 Set s390 IOTLB flushing mode
5941 strict
5942 With strict flushing every unmap operation will result
5943 in an IOTLB flush. Default is lazy flushing before
5944 reuse, which is faster. Deprecated, equivalent to
5945 iommu.strict=1.
5947 s390_iommu_aperture= [KNL,S390]
5948 Specifies the size of the per device DMA address space
5949 accessible through the DMA and IOMMU APIs as a decimal
5950 factor of the size of main memory.
5951 The default is 1 meaning that one can concurrently use
5952 as many DMA addresses as physical memory is installed,
5953 if supported by hardware, and thus map all of memory
5954 once. With a value of 2 one can map all of memory twice
5955 and so on. As a special case a factor of 0 imposes no
5956 restrictions other than those given by hardware at the
5957 cost of significant additional memory use for tables.
5959 sa1100ir [NET]
5960 See drivers/net/irda/sa1100_ir.c.
5962 sched_verbose [KNL,EARLY] Enables verbose scheduler debug messages.
5964 schedstats= [KNL,X86] Enable or disable scheduled statistics.
5965 Allowed values are enable and disable. This feature
5966 incurs a small amount of overhead in the scheduler
5967 but is useful for debugging and performance tuning.
5969 sched_thermal_decay_shift=
5970 [Deprecated]
5971 [KNL, SMP] Set a decay shift for scheduler thermal
5972 pressure signal. Thermal pressure signal follows the
5973 default decay period of other scheduler pelt
5974 signals(usually 32 ms but configurable). Setting
5975 sched_thermal_decay_shift will left shift the decay
5976 period for the thermal pressure signal by the shift
5977 value.
5978 i.e. with the default pelt decay period of 32 ms
5979 sched_thermal_decay_shift thermal pressure decay pr
5980 1 64 ms
5981 2 128 ms
5982 and so on.
5983 Format: integer between 0 and 10
5984 Default is 0.
5986 scftorture.holdoff= [KNL]
5987 Number of seconds to hold off before starting
5988 test. Defaults to zero for module insertion and
5989 to 10 seconds for built-in smp_call_function()
5990 tests.
5992 scftorture.longwait= [KNL]
5993 Request ridiculously long waits randomly selected
5994 up to the chosen limit in seconds. Zero (the
5995 default) disables this feature. Please note
5996 that requesting even small non-zero numbers of
5997 seconds can result in RCU CPU stall warnings,
5998 softlockup complaints, and so on.
6000 scftorture.nthreads= [KNL]
6001 Number of kthreads to spawn to invoke the
6002 smp_call_function() family of functions.
6003 The default of -1 specifies a number of kthreads
6004 equal to the number of CPUs.
6006 scftorture.onoff_holdoff= [KNL]
6007 Number seconds to wait after the start of the
6008 test before initiating CPU-hotplug operations.
6010 scftorture.onoff_interval= [KNL]
6011 Number seconds to wait between successive
6012 CPU-hotplug operations. Specifying zero (which
6013 is the default) disables CPU-hotplug operations.
6015 scftorture.shutdown_secs= [KNL]
6016 The number of seconds following the start of the
6017 test after which to shut down the system. The
6018 default of zero avoids shutting down the system.
6019 Non-zero values are useful for automated tests.
6021 scftorture.stat_interval= [KNL]
6022 The number of seconds between outputting the
6023 current test statistics to the console. A value
6024 of zero disables statistics output.
6026 scftorture.stutter_cpus= [KNL]
6027 The number of jiffies to wait between each change
6028 to the set of CPUs under test.
6030 scftorture.use_cpus_read_lock= [KNL]
6031 Use use_cpus_read_lock() instead of the default
6032 preempt_disable() to disable CPU hotplug
6033 while invoking one of the smp_call_function*()
6034 functions.
6036 scftorture.verbose= [KNL]
6037 Enable additional printk() statements.
6039 scftorture.weight_single= [KNL]
6040 The probability weighting to use for the
6041 smp_call_function_single() function with a zero
6042 "wait" parameter. A value of -1 selects the
6043 default if all other weights are -1. However,
6044 if at least one weight has some other value, a
6045 value of -1 will instead select a weight of zero.
6047 scftorture.weight_single_wait= [KNL]
6048 The probability weighting to use for the
6049 smp_call_function_single() function with a
6050 non-zero "wait" parameter. See weight_single.
6052 scftorture.weight_many= [KNL]
6053 The probability weighting to use for the
6054 smp_call_function_many() function with a zero
6055 "wait" parameter. See weight_single.
6056 Note well that setting a high probability for
6057 this weighting can place serious IPI load
6058 on the system.
6060 scftorture.weight_many_wait= [KNL]
6061 The probability weighting to use for the
6062 smp_call_function_many() function with a
6063 non-zero "wait" parameter. See weight_single
6064 and weight_many.
6066 scftorture.weight_all= [KNL]
6067 The probability weighting to use for the
6068 smp_call_function_all() function with a zero
6069 "wait" parameter. See weight_single and
6070 weight_many.
6072 scftorture.weight_all_wait= [KNL]
6073 The probability weighting to use for the
6074 smp_call_function_all() function with a
6075 non-zero "wait" parameter. See weight_single
6076 and weight_many.
6078 sdw_mclk_divider=[SDW]
6079 Specify the MCLK divider for Intel SoundWire buses in
6080 case the BIOS does not provide the clock rate properly.
6082 skew_tick= [KNL,EARLY] Offset the periodic timer tick per cpu to mitigate
6083 xtime_lock contention on larger systems, and/or RCU lock
6084 contention on all systems with CONFIG_MAXSMP set.
6085 Format: { "0" | "1" }
6086 0 -- disable. (may be 1 via CONFIG_CMDLINE="skew_tick=1"
6087 1 -- enable.
6088 Note: increases power consumption, thus should only be
6089 enabled if running jitter sensitive (HPC/RT) workloads.
6091 security= [SECURITY] Choose a legacy "major" security module to
6092 enable at boot. This has been deprecated by the
6093 "lsm=" parameter.
6095 selinux= [SELINUX] Disable or enable SELinux at boot time.
6096 Format: { "0" | "1" }
6097 See security/selinux/Kconfig help text.
6098 0 -- disable.
6099 1 -- enable.
6100 Default value is 1.
6102 serialnumber [BUGS=X86-32]
6104 sev=option[,option...] [X86-64] See Documentation/arch/x86/x86_64/boot-options.rst
6106 shapers= [NET]
6107 Maximal number of shapers.
6109 show_lapic= [APIC,X86] Advanced Programmable Interrupt Controller
6110 Limit apic dumping. The parameter defines the maximal
6111 number of local apics being dumped. Also it is possible
6112 to set it to "all" by meaning -- no limit here.
6113 Format: { 1 (default) | 2 | ... | all }.
6114 The parameter valid if only apic=debug or
6115 apic=verbose is specified.
6116 Example: apic=debug show_lapic=all
6118 slab_debug[=options[,slabs][;[options[,slabs]]...] [MM]
6119 Enabling slab_debug allows one to determine the
6120 culprit if slab objects become corrupted. Enabling
6121 slab_debug can create guard zones around objects and
6122 may poison objects when not in use. Also tracks the
6123 last alloc / free. For more information see
6124 Documentation/mm/slub.rst.
6125 (slub_debug legacy name also accepted for now)
6127 slab_max_order= [MM]
6128 Determines the maximum allowed order for slabs.
6129 A high setting may cause OOMs due to memory
6130 fragmentation. For more information see
6131 Documentation/mm/slub.rst.
6132 (slub_max_order legacy name also accepted for now)
6134 slab_merge [MM]
6135 Enable merging of slabs with similar size when the
6136 kernel is built without CONFIG_SLAB_MERGE_DEFAULT.
6137 (slub_merge legacy name also accepted for now)
6139 slab_min_objects= [MM]
6140 The minimum number of objects per slab. SLUB will
6141 increase the slab order up to slab_max_order to
6142 generate a sufficiently large slab able to contain
6143 the number of objects indicated. The higher the number
6144 of objects the smaller the overhead of tracking slabs
6145 and the less frequently locks need to be acquired.
6146 For more information see Documentation/mm/slub.rst.
6147 (slub_min_objects legacy name also accepted for now)
6149 slab_min_order= [MM]
6150 Determines the minimum page order for slabs. Must be
6151 lower or equal to slab_max_order. For more information see
6152 Documentation/mm/slub.rst.
6153 (slub_min_order legacy name also accepted for now)
6155 slab_nomerge [MM]
6156 Disable merging of slabs with similar size. May be
6157 necessary if there is some reason to distinguish
6158 allocs to different slabs, especially in hardened
6159 environments where the risk of heap overflows and
6160 layout control by attackers can usually be
6161 frustrated by disabling merging. This will reduce
6162 most of the exposure of a heap attack to a single
6163 cache (risks via metadata attacks are mostly
6164 unchanged). Debug options disable merging on their
6165 own.
6166 For more information see Documentation/mm/slub.rst.
6167 (slub_nomerge legacy name also accepted for now)
6169 slab_strict_numa [MM]
6170 Support memory policies on a per object level
6171 in the slab allocator. The default is for memory
6172 policies to be applied at the folio level when
6173 a new folio is needed or a partial folio is
6174 retrieved from the lists. Increases overhead
6175 in the slab fastpaths but gains more accurate
6176 NUMA kernel object placement which helps with slow
6177 interconnects in NUMA systems.
6179 slram= [HW,MTD]
6181 smart2= [HW]
6182 Format: <io1>[,<io2>[,...,<io8>]]
6184 smp.csd_lock_timeout= [KNL]
6185 Specify the period of time in milliseconds
6186 that smp_call_function() and friends will wait
6187 for a CPU to release the CSD lock. This is
6188 useful when diagnosing bugs involving CPUs
6189 disabling interrupts for extended periods
6190 of time. Defaults to 5,000 milliseconds, and
6191 setting a value of zero disables this feature.
6192 This feature may be more efficiently disabled
6193 using the csdlock_debug- kernel parameter.
6195 smp.panic_on_ipistall= [KNL]
6196 If a csd_lock_timeout extends for more than
6197 the specified number of milliseconds, panic the
6198 system. By default, let CSD-lock acquisition
6199 take as long as they take. Specifying 300,000
6200 for this value provides a 5-minute timeout.
6202 smsc-ircc2.nopnp [HW] Don't use PNP to discover SMC devices
6203 smsc-ircc2.ircc_cfg= [HW] Device configuration I/O port
6204 smsc-ircc2.ircc_sir= [HW] SIR base I/O port
6205 smsc-ircc2.ircc_fir= [HW] FIR base I/O port
6206 smsc-ircc2.ircc_irq= [HW] IRQ line
6207 smsc-ircc2.ircc_dma= [HW] DMA channel
6208 smsc-ircc2.ircc_transceiver= [HW] Transceiver type:
6209 0: Toshiba Satellite 1800 (GP data pin select)
6210 1: Fast pin select (default)
6211 2: ATC IRMode
6213 smt= [KNL,MIPS,S390,EARLY] Set the maximum number of threads
6214 (logical CPUs) to use per physical CPU on systems
6215 capable of symmetric multithreading (SMT). Will
6216 be capped to the actual hardware limit.
6217 Format: <integer>
6218 Default: -1 (no limit)
6220 softlockup_panic=
6221 [KNL] Should the soft-lockup detector generate panics.
6222 Format: 0 | 1
6224 A value of 1 instructs the soft-lockup detector
6225 to panic the machine when a soft-lockup occurs. It is
6226 also controlled by the kernel.softlockup_panic sysctl
6227 and CONFIG_BOOTPARAM_SOFTLOCKUP_PANIC, which is the
6228 respective build-time switch to that functionality.
6230 softlockup_all_cpu_backtrace=
6231 [KNL] Should the soft-lockup detector generate
6232 backtraces on all cpus.
6233 Format: 0 | 1
6235 sonypi.*= [HW] Sony Programmable I/O Control Device driver
6236 See Documentation/admin-guide/laptops/sonypi.rst
6238 spectre_bhi= [X86] Control mitigation of Branch History Injection
6239 (BHI) vulnerability. This setting affects the
6240 deployment of the HW BHI control and the SW BHB
6241 clearing sequence.
6243 on - (default) Enable the HW or SW mitigation as
6244 needed. This protects the kernel from
6245 both syscalls and VMs.
6246 vmexit - On systems which don't have the HW mitigation
6247 available, enable the SW mitigation on vmexit
6248 ONLY. On such systems, the host kernel is
6249 protected from VM-originated BHI attacks, but
6250 may still be vulnerable to syscall attacks.
6251 off - Disable the mitigation.
6253 spectre_v2= [X86,EARLY] Control mitigation of Spectre variant 2
6254 (indirect branch speculation) vulnerability.
6255 The default operation protects the kernel from
6256 user space attacks.
6258 on - unconditionally enable, implies
6259 spectre_v2_user=on
6260 off - unconditionally disable, implies
6261 spectre_v2_user=off
6262 auto - kernel detects whether your CPU model is
6263 vulnerable
6265 Selecting 'on' will, and 'auto' may, choose a
6266 mitigation method at run time according to the
6267 CPU, the available microcode, the setting of the
6268 CONFIG_MITIGATION_RETPOLINE configuration option,
6269 and the compiler with which the kernel was built.
6271 Selecting 'on' will also enable the mitigation
6272 against user space to user space task attacks.
6274 Selecting 'off' will disable both the kernel and
6275 the user space protections.
6277 Specific mitigations can also be selected manually:
6279 retpoline - replace indirect branches
6280 retpoline,generic - Retpolines
6281 retpoline,lfence - LFENCE; indirect branch
6282 retpoline,amd - alias for retpoline,lfence
6283 eibrs - Enhanced/Auto IBRS
6284 eibrs,retpoline - Enhanced/Auto IBRS + Retpolines
6285 eibrs,lfence - Enhanced/Auto IBRS + LFENCE
6286 ibrs - use IBRS to protect kernel
6288 Not specifying this option is equivalent to
6289 spectre_v2=auto.
6291 spectre_v2_user=
6292 [X86] Control mitigation of Spectre variant 2
6293 (indirect branch speculation) vulnerability between
6294 user space tasks
6296 on - Unconditionally enable mitigations. Is
6297 enforced by spectre_v2=on
6299 off - Unconditionally disable mitigations. Is
6300 enforced by spectre_v2=off
6302 prctl - Indirect branch speculation is enabled,
6303 but mitigation can be enabled via prctl
6304 per thread. The mitigation control state
6305 is inherited on fork.
6307 prctl,ibpb
6308 - Like "prctl" above, but only STIBP is
6309 controlled per thread. IBPB is issued
6310 always when switching between different user
6311 space processes.
6313 seccomp
6314 - Same as "prctl" above, but all seccomp
6315 threads will enable the mitigation unless
6316 they explicitly opt out.
6318 seccomp,ibpb
6319 - Like "seccomp" above, but only STIBP is
6320 controlled per thread. IBPB is issued
6321 always when switching between different
6322 user space processes.
6324 auto - Kernel selects the mitigation depending on
6325 the available CPU features and vulnerability.
6327 Default mitigation: "prctl"
6329 Not specifying this option is equivalent to
6330 spectre_v2_user=auto.
6332 spec_rstack_overflow=
6333 [X86,EARLY] Control RAS overflow mitigation on AMD Zen CPUs
6335 off - Disable mitigation
6336 microcode - Enable microcode mitigation only
6337 safe-ret - Enable sw-only safe RET mitigation (default)
6338 ibpb - Enable mitigation by issuing IBPB on
6339 kernel entry
6340 ibpb-vmexit - Issue IBPB only on VMEXIT
6341 (cloud-specific mitigation)
6343 spec_store_bypass_disable=
6344 [HW,EARLY] Control Speculative Store Bypass (SSB) Disable mitigation
6345 (Speculative Store Bypass vulnerability)
6347 Certain CPUs are vulnerable to an exploit against a
6348 a common industry wide performance optimization known
6349 as "Speculative Store Bypass" in which recent stores
6350 to the same memory location may not be observed by
6351 later loads during speculative execution. The idea
6352 is that such stores are unlikely and that they can
6353 be detected prior to instruction retirement at the
6354 end of a particular speculation execution window.
6356 In vulnerable processors, the speculatively forwarded
6357 store can be used in a cache side channel attack, for
6358 example to read memory to which the attacker does not
6359 directly have access (e.g. inside sandboxed code).
6361 This parameter controls whether the Speculative Store
6362 Bypass optimization is used.
6364 On x86 the options are:
6366 on - Unconditionally disable Speculative Store Bypass
6367 off - Unconditionally enable Speculative Store Bypass
6368 auto - Kernel detects whether the CPU model contains an
6369 implementation of Speculative Store Bypass and
6370 picks the most appropriate mitigation. If the
6371 CPU is not vulnerable, "off" is selected. If the
6372 CPU is vulnerable the default mitigation is
6373 architecture and Kconfig dependent. See below.
6374 prctl - Control Speculative Store Bypass per thread
6375 via prctl. Speculative Store Bypass is enabled
6376 for a process by default. The state of the control
6377 is inherited on fork.
6378 seccomp - Same as "prctl" above, but all seccomp threads
6379 will disable SSB unless they explicitly opt out.
6381 Default mitigations:
6382 X86: "prctl"
6384 On powerpc the options are:
6386 on,auto - On Power8 and Power9 insert a store-forwarding
6387 barrier on kernel entry and exit. On Power7
6388 perform a software flush on kernel entry and
6389 exit.
6390 off - No action.
6392 Not specifying this option is equivalent to
6393 spec_store_bypass_disable=auto.
6395 split_lock_detect=
6396 [X86] Enable split lock detection or bus lock detection
6398 When enabled (and if hardware support is present), atomic
6399 instructions that access data across cache line
6400 boundaries will result in an alignment check exception
6401 for split lock detection or a debug exception for
6402 bus lock detection.
6404 off - not enabled
6406 warn - the kernel will emit rate-limited warnings
6407 about applications triggering the #AC
6408 exception or the #DB exception. This mode is
6409 the default on CPUs that support split lock
6410 detection or bus lock detection. Default
6411 behavior is by #AC if both features are
6412 enabled in hardware.
6414 fatal - the kernel will send SIGBUS to applications
6415 that trigger the #AC exception or the #DB
6416 exception. Default behavior is by #AC if
6417 both features are enabled in hardware.
6419 ratelimit:N -
6420 Set system wide rate limit to N bus locks
6421 per second for bus lock detection.
6422 0 < N <= 1000.
6424 N/A for split lock detection.
6427 If an #AC exception is hit in the kernel or in
6428 firmware (i.e. not while executing in user mode)
6429 the kernel will oops in either "warn" or "fatal"
6430 mode.
6432 #DB exception for bus lock is triggered only when
6433 CPL > 0.
6435 srbds= [X86,INTEL,EARLY]
6436 Control the Special Register Buffer Data Sampling
6437 (SRBDS) mitigation.
6439 Certain CPUs are vulnerable to an MDS-like
6440 exploit which can leak bits from the random
6441 number generator.
6443 By default, this issue is mitigated by
6444 microcode. However, the microcode fix can cause
6445 the RDRAND and RDSEED instructions to become
6446 much slower. Among other effects, this will
6447 result in reduced throughput from /dev/urandom.
6449 The microcode mitigation can be disabled with
6450 the following option:
6452 off: Disable mitigation and remove
6453 performance impact to RDRAND and RDSEED
6455 srcutree.big_cpu_lim [KNL]
6456 Specifies the number of CPUs constituting a
6457 large system, such that srcu_struct structures
6458 should immediately allocate an srcu_node array.
6459 This kernel-boot parameter defaults to 128,
6460 but takes effect only when the low-order four
6461 bits of srcutree.convert_to_big is equal to 3
6462 (decide at boot).
6464 srcutree.convert_to_big [KNL]
6465 Specifies under what conditions an SRCU tree
6466 srcu_struct structure will be converted to big
6467 form, that is, with an rcu_node tree:
6469 0: Never.
6470 1: At init_srcu_struct() time.
6471 2: When rcutorture decides to.
6472 3: Decide at boot time (default).
6473 0x1X: Above plus if high contention.
6475 Either way, the srcu_node tree will be sized based
6476 on the actual runtime number of CPUs (nr_cpu_ids)
6477 instead of the compile-time CONFIG_NR_CPUS.
6479 srcutree.counter_wrap_check [KNL]
6480 Specifies how frequently to check for
6481 grace-period sequence counter wrap for the
6482 srcu_data structure's ->srcu_gp_seq_needed field.
6483 The greater the number of bits set in this kernel
6484 parameter, the less frequently counter wrap will
6485 be checked for. Note that the bottom two bits
6486 are ignored.
6488 srcutree.exp_holdoff [KNL]
6489 Specifies how many nanoseconds must elapse
6490 since the end of the last SRCU grace period for
6491 a given srcu_struct until the next normal SRCU
6492 grace period will be considered for automatic
6493 expediting. Set to zero to disable automatic
6494 expediting.
6496 srcutree.srcu_max_nodelay [KNL]
6497 Specifies the number of no-delay instances
6498 per jiffy for which the SRCU grace period
6499 worker thread will be rescheduled with zero
6500 delay. Beyond this limit, worker thread will
6501 be rescheduled with a sleep delay of one jiffy.
6503 srcutree.srcu_max_nodelay_phase [KNL]
6504 Specifies the per-grace-period phase, number of
6505 non-sleeping polls of readers. Beyond this limit,
6506 grace period worker thread will be rescheduled
6507 with a sleep delay of one jiffy, between each
6508 rescan of the readers, for a grace period phase.
6510 srcutree.srcu_retry_check_delay [KNL]
6511 Specifies number of microseconds of non-sleeping
6512 delay between each non-sleeping poll of readers.
6514 srcutree.small_contention_lim [KNL]
6515 Specifies the number of update-side contention
6516 events per jiffy will be tolerated before
6517 initiating a conversion of an srcu_struct
6518 structure to big form. Note that the value of
6519 srcutree.convert_to_big must have the 0x10 bit
6520 set for contention-based conversions to occur.
6522 ssbd= [ARM64,HW,EARLY]
6523 Speculative Store Bypass Disable control
6525 On CPUs that are vulnerable to the Speculative
6526 Store Bypass vulnerability and offer a
6527 firmware based mitigation, this parameter
6528 indicates how the mitigation should be used:
6530 force-on: Unconditionally enable mitigation for
6531 for both kernel and userspace
6532 force-off: Unconditionally disable mitigation for
6533 for both kernel and userspace
6534 kernel: Always enable mitigation in the
6535 kernel, and offer a prctl interface
6536 to allow userspace to register its
6537 interest in being mitigated too.
6539 stack_guard_gap= [MM]
6540 override the default stack gap protection. The value
6541 is in page units and it defines how many pages prior
6542 to (for stacks growing down) resp. after (for stacks
6543 growing up) the main stack are reserved for no other
6544 mapping. Default value is 256 pages.
6546 stack_depot_disable= [KNL,EARLY]
6547 Setting this to true through kernel command line will
6548 disable the stack depot thereby saving the static memory
6549 consumed by the stack hash table. By default this is set
6550 to false.
6552 stacktrace [FTRACE]
6553 Enabled the stack tracer on boot up.
6555 stacktrace_filter=[function-list]
6556 [FTRACE] Limit the functions that the stack tracer
6557 will trace at boot up. function-list is a comma-separated
6558 list of functions. This list can be changed at run
6559 time by the stack_trace_filter file in the debugfs
6560 tracing directory. Note, this enables stack tracing
6561 and the stacktrace above is not needed.
6563 sti= [PARISC,HW]
6564 Format: <num>
6565 Set the STI (builtin display/keyboard on the HP-PARISC
6566 machines) console (graphic card) which should be used
6567 as the initial boot-console.
6568 See also comment in drivers/video/console/sticore.c.
6570 sti_font= [HW]
6571 See comment in drivers/video/console/sticore.c.
6573 stifb= [HW]
6574 Format: bpp:<bpp1>[:<bpp2>[:<bpp3>...]]
6576 strict_sas_size=
6577 [X86]
6578 Format: <bool>
6579 Enable or disable strict sigaltstack size checks
6580 against the required signal frame size which
6581 depends on the supported FPU features. This can
6582 be used to filter out binaries which have
6583 not yet been made aware of AT_MINSIGSTKSZ.
6585 stress_hpt [PPC,EARLY]
6586 Limits the number of kernel HPT entries in the hash
6587 page table to increase the rate of hash page table
6588 faults on kernel addresses.
6590 stress_slb [PPC,EARLY]
6591 Limits the number of kernel SLB entries, and flushes
6592 them frequently to increase the rate of SLB faults
6593 on kernel addresses.
6595 sunrpc.min_resvport=
6596 sunrpc.max_resvport=
6597 [NFS,SUNRPC]
6598 SunRPC servers often require that client requests
6599 originate from a privileged port (i.e. a port in the
6600 range 0 < portnr < 1024).
6601 An administrator who wishes to reserve some of these
6602 ports for other uses may adjust the range that the
6603 kernel's sunrpc client considers to be privileged
6604 using these two parameters to set the minimum and
6605 maximum port values.
6607 sunrpc.svc_rpc_per_connection_limit=
6608 [NFS,SUNRPC]
6609 Limit the number of requests that the server will
6610 process in parallel from a single connection.
6611 The default value is 0 (no limit).
6613 sunrpc.pool_mode=
6614 [NFS]
6615 Control how the NFS server code allocates CPUs to
6616 service thread pools. Depending on how many NICs
6617 you have and where their interrupts are bound, this
6618 option will affect which CPUs will do NFS serving.
6619 Note: this parameter cannot be changed while the
6620 NFS server is running.
6622 auto the server chooses an appropriate mode
6623 automatically using heuristics
6624 global a single global pool contains all CPUs
6625 percpu one pool for each CPU
6626 pernode one pool for each NUMA node (equivalent
6627 to global on non-NUMA machines)
6629 sunrpc.tcp_slot_table_entries=
6630 sunrpc.udp_slot_table_entries=
6631 [NFS,SUNRPC]
6632 Sets the upper limit on the number of simultaneous
6633 RPC calls that can be sent from the client to a
6634 server. Increasing these values may allow you to
6635 improve throughput, but will also increase the
6636 amount of memory reserved for use by the client.
6638 suspend.pm_test_delay=
6639 [SUSPEND]
6640 Sets the number of seconds to remain in a suspend test
6641 mode before resuming the system (see
6642 /sys/power/pm_test). Only available when CONFIG_PM_DEBUG
6643 is set. Default value is 5.
6645 svm= [PPC]
6646 Format: { on | off | y | n | 1 | 0 }
6647 This parameter controls use of the Protected
6648 Execution Facility on pSeries.
6650 swiotlb= [ARM,PPC,MIPS,X86,S390,EARLY]
6651 Format: { <int> [,<int>] | force | noforce }
6652 <int> -- Number of I/O TLB slabs
6653 <int> -- Second integer after comma. Number of swiotlb
6654 areas with their own lock. Will be rounded up
6655 to a power of 2.
6656 force -- force using of bounce buffers even if they
6657 wouldn't be automatically used by the kernel
6658 noforce -- Never use bounce buffers (for debugging)
6660 switches= [HW,M68k,EARLY]
6662 sysctl.*= [KNL]
6663 Set a sysctl parameter, right before loading the init
6664 process, as if the value was written to the respective
6665 /proc/sys/... file. Both '.' and '/' are recognized as
6666 separators. Unrecognized parameters and invalid values
6667 are reported in the kernel log. Sysctls registered
6668 later by a loaded module cannot be set this way.
6669 Example: sysctl.vm.swappiness=40
6671 sysrq_always_enabled
6672 [KNL]
6673 Ignore sysrq setting - this boot parameter will
6674 neutralize any effect of /proc/sys/kernel/sysrq.
6675 Useful for debugging.
6677 tcpmhash_entries= [KNL,NET]
6678 Set the number of tcp_metrics_hash slots.
6679 Default value is 8192 or 16384 depending on total
6680 ram pages. This is used to specify the TCP metrics
6681 cache size. See Documentation/networking/ip-sysctl.rst
6682 "tcp_no_metrics_save" section for more details.
6684 tdfx= [HW,DRM]
6686 test_suspend= [SUSPEND]
6687 Format: { "mem" | "standby" | "freeze" }[,N]
6688 Specify "mem" (for Suspend-to-RAM) or "standby" (for
6689 standby suspend) or "freeze" (for suspend type freeze)
6690 as the system sleep state during system startup with
6691 the optional capability to repeat N number of times.
6692 The system is woken from this state using a
6693 wakeup-capable RTC alarm.
6695 thash_entries= [KNL,NET]
6696 Set number of hash buckets for TCP connection
6698 thermal.act= [HW,ACPI]
6699 -1: disable all active trip points in all thermal zones
6700 <degrees C>: override all lowest active trip points
6702 thermal.crt= [HW,ACPI]
6703 -1: disable all critical trip points in all thermal zones
6704 <degrees C>: override all critical trip points
6706 thermal.off= [HW,ACPI]
6707 1: disable ACPI thermal control
6709 thermal.psv= [HW,ACPI]
6710 -1: disable all passive trip points
6711 <degrees C>: override all passive trip points to this
6712 value
6714 thermal.tzp= [HW,ACPI]
6715 Specify global default ACPI thermal zone polling rate
6716 <deci-seconds>: poll all this frequency
6717 0: no polling (default)
6719 thp_anon= [KNL]
6720 Format: <size>[KMG],<size>[KMG]:<state>;<size>[KMG]-<size>[KMG]:<state>
6721 state is one of "always", "madvise", "never" or "inherit".
6722 Control the default behavior of the system with respect
6723 to anonymous transparent hugepages.
6724 Can be used multiple times for multiple anon THP sizes.
6725 See Documentation/admin-guide/mm/transhuge.rst for more
6726 details.
6728 threadirqs [KNL,EARLY]
6729 Force threading of all interrupt handlers except those
6730 marked explicitly IRQF_NO_THREAD.
6732 thp_shmem= [KNL]
6733 Format: <size>[KMG],<size>[KMG]:<policy>;<size>[KMG]-<size>[KMG]:<policy>
6734 Control the default policy of each hugepage size for the
6735 internal shmem mount. <policy> is one of policies available
6736 for the shmem mount ("always", "inherit", "never", "within_size",
6737 and "advise").
6738 It can be used multiple times for multiple shmem THP sizes.
6739 See Documentation/admin-guide/mm/transhuge.rst for more
6740 details.
6742 topology= [S390,EARLY]
6743 Format: {off | on}
6744 Specify if the kernel should make use of the cpu
6745 topology information if the hardware supports this.
6746 The scheduler will make use of this information and
6747 e.g. base its process migration decisions on it.
6748 Default is on.
6750 torture.disable_onoff_at_boot= [KNL]
6751 Prevent the CPU-hotplug component of torturing
6752 until after init has spawned.
6754 torture.ftrace_dump_at_shutdown= [KNL]
6755 Dump the ftrace buffer at torture-test shutdown,
6756 even if there were no errors. This can be a
6757 very costly operation when many torture tests
6758 are running concurrently, especially on systems
6759 with rotating-rust storage.
6761 torture.verbose_sleep_frequency= [KNL]
6762 Specifies how many verbose printk()s should be
6763 emitted between each sleep. The default of zero
6764 disables verbose-printk() sleeping.
6766 torture.verbose_sleep_duration= [KNL]
6767 Duration of each verbose-printk() sleep in jiffies.
6769 tpm.disable_pcr_integrity= [HW,TPM]
6770 Do not protect PCR registers from unintended physical
6771 access, or interposers in the bus by the means of
6772 having an integrity protected session wrapped around
6773 TPM2_PCR_Extend command. Consider this in a situation
6774 where TPM is heavily utilized by IMA, thus protection
6775 causing a major performance hit, and the space where
6776 machines are deployed is by other means guarded.
6778 tpm_suspend_pcr=[HW,TPM]
6779 Format: integer pcr id
6780 Specify that at suspend time, the tpm driver
6781 should extend the specified pcr with zeros,
6782 as a workaround for some chips which fail to
6783 flush the last written pcr on TPM_SaveState.
6784 This will guarantee that all the other pcrs
6785 are saved.
6787 tpm_tis.interrupts= [HW,TPM]
6788 Enable interrupts for the MMIO based physical layer
6789 for the FIFO interface. By default it is set to false
6790 (0). For more information about TPM hardware interfaces
6791 defined by Trusted Computing Group (TCG) see
6792 https://trustedcomputinggroup.org/resource/pc-client-platform-tpm-profile-ptp-specification/
6794 tp_printk [FTRACE]
6795 Have the tracepoints sent to printk as well as the
6796 tracing ring buffer. This is useful for early boot up
6797 where the system hangs or reboots and does not give the
6798 option for reading the tracing buffer or performing a
6799 ftrace_dump_on_oops.
6801 To turn off having tracepoints sent to printk,
6802 echo 0 > /proc/sys/kernel/tracepoint_printk
6803 Note, echoing 1 into this file without the
6804 tp_printk kernel cmdline option has no effect.
6806 The tp_printk_stop_on_boot (see below) can also be used
6807 to stop the printing of events to console at
6808 late_initcall_sync.
6810 ** CAUTION **
6812 Having tracepoints sent to printk() and activating high
6813 frequency tracepoints such as irq or sched, can cause
6814 the system to live lock.
6816 tp_printk_stop_on_boot [FTRACE]
6817 When tp_printk (above) is set, it can cause a lot of noise
6818 on the console. It may be useful to only include the
6819 printing of events during boot up, as user space may
6820 make the system inoperable.
6822 This command line option will stop the printing of events
6823 to console at the late_initcall_sync() time frame.
6825 trace_buf_size=nn[KMG]
6826 [FTRACE] will set tracing buffer size on each cpu.
6828 trace_clock= [FTRACE] Set the clock used for tracing events
6829 at boot up.
6830 local - Use the per CPU time stamp counter
6831 (converted into nanoseconds). Fast, but
6832 depending on the architecture, may not be
6833 in sync between CPUs.
6834 global - Event time stamps are synchronize across
6835 CPUs. May be slower than the local clock,
6836 but better for some race conditions.
6837 counter - Simple counting of events (1, 2, ..)
6838 note, some counts may be skipped due to the
6839 infrastructure grabbing the clock more than
6840 once per event.
6841 uptime - Use jiffies as the time stamp.
6842 perf - Use the same clock that perf uses.
6843 mono - Use ktime_get_mono_fast_ns() for time stamps.
6844 mono_raw - Use ktime_get_raw_fast_ns() for time
6845 stamps.
6846 boot - Use ktime_get_boot_fast_ns() for time stamps.
6847 Architectures may add more clocks. See
6848 Documentation/trace/ftrace.rst for more details.
6850 trace_event=[event-list]
6851 [FTRACE] Set and start specified trace events in order
6852 to facilitate early boot debugging. The event-list is a
6853 comma-separated list of trace events to enable. See
6854 also Documentation/trace/events.rst
6856 trace_instance=[instance-info]
6857 [FTRACE] Create a ring buffer instance early in boot up.
6858 This will be listed in:
6860 /sys/kernel/tracing/instances
6862 Events can be enabled at the time the instance is created
6863 via:
6865 trace_instance=<name>,<system1>:<event1>,<system2>:<event2>
6867 Note, the "<system*>:" portion is optional if the event is
6868 unique.
6870 trace_instance=foo,sched:sched_switch,irq_handler_entry,initcall
6872 will enable the "sched_switch" event (note, the "sched:" is optional, and
6873 the same thing would happen if it was left off). The irq_handler_entry
6874 event, and all events under the "initcall" system.
6876 Flags can be added to the instance to modify its behavior when it is
6877 created. The flags are separated by '^'.
6879 The available flags are:
6881 traceoff - Have the tracing instance tracing disabled after it is created.
6882 traceprintk - Have trace_printk() write into this trace instance
6883 (note, "printk" and "trace_printk" can also be used)
6885 trace_instance=foo^traceoff^traceprintk,sched,irq
6887 The flags must come before the defined events.
6889 If memory has been reserved (see memmap for x86), the instance
6890 can use that memory:
6892 memmap=12M$0x284500000 trace_instance=boot_map@0x284500000:12M
6894 The above will create a "boot_map" instance that uses the physical
6895 memory at 0x284500000 that is 12Megs. The per CPU buffers of that
6896 instance will be split up accordingly.
6898 Alternatively, the memory can be reserved by the reserve_mem option:
6900 reserve_mem=12M:4096:trace trace_instance=boot_map@trace
6902 This will reserve 12 megabytes at boot up with a 4096 byte alignment
6903 and place the ring buffer in this memory. Note that due to KASLR, the
6904 memory may not be the same location each time, which will not preserve
6905 the buffer content.
6907 Also note that the layout of the ring buffer data may change between
6908 kernel versions where the validator will fail and reset the ring buffer
6909 if the layout is not the same as the previous kernel.
6911 If the ring buffer is used for persistent bootups and has events enabled,
6912 it is recommend to disable tracing so that events from a previous boot do not
6913 mix with events of the current boot (unless you are debugging a random crash
6914 at boot up).
6916 reserve_mem=12M:4096:trace trace_instance=boot_map^traceoff^traceprintk@trace,sched,irq
6918 Note, saving the trace buffer across reboots does require that the system
6919 is set up to not wipe memory. For instance, CONFIG_RESET_ATTACK_MITIGATION
6920 can force a memory reset on boot which will clear any trace that was stored.
6921 This is just one of many ways that can clear memory. Make sure your system
6922 keeps the content of memory across reboots before relying on this option.
6924 See also Documentation/trace/debugging.rst
6927 trace_options=[option-list]
6928 [FTRACE] Enable or disable tracer options at boot.
6929 The option-list is a comma delimited list of options
6930 that can be enabled or disabled just as if you were
6931 to echo the option name into
6933 /sys/kernel/tracing/trace_options
6935 For example, to enable stacktrace option (to dump the
6936 stack trace of each event), add to the command line:
6938 trace_options=stacktrace
6940 See also Documentation/trace/ftrace.rst "trace options"
6941 section.
6943 trace_trigger=[trigger-list]
6944 [FTRACE] Add a event trigger on specific events.
6945 Set a trigger on top of a specific event, with an optional
6946 filter.
6948 The format is is "trace_trigger=<event>.<trigger>[ if <filter>],..."
6949 Where more than one trigger may be specified that are comma deliminated.
6951 For example:
6953 trace_trigger="sched_switch.stacktrace if prev_state == 2"
6955 The above will enable the "stacktrace" trigger on the "sched_switch"
6956 event but only trigger it if the "prev_state" of the "sched_switch"
6957 event is "2" (TASK_UNINTERUPTIBLE).
6959 See also "Event triggers" in Documentation/trace/events.rst
6962 traceoff_on_warning
6963 [FTRACE] enable this option to disable tracing when a
6964 warning is hit. This turns off "tracing_on". Tracing can
6965 be enabled again by echoing '1' into the "tracing_on"
6966 file located in /sys/kernel/tracing/
6968 This option is useful, as it disables the trace before
6969 the WARNING dump is called, which prevents the trace to
6970 be filled with content caused by the warning output.
6972 This option can also be set at run time via the sysctl
6973 option: kernel/traceoff_on_warning
6975 transparent_hugepage=
6976 [KNL]
6977 Format: [always|madvise|never]
6978 Can be used to control the default behavior of the system
6979 with respect to transparent hugepages.
6980 See Documentation/admin-guide/mm/transhuge.rst
6981 for more details.
6983 transparent_hugepage_shmem= [KNL]
6984 Format: [always|within_size|advise|never|deny|force]
6985 Can be used to control the hugepage allocation policy for
6986 the internal shmem mount.
6987 See Documentation/admin-guide/mm/transhuge.rst
6988 for more details.
6990 trusted.source= [KEYS]
6991 Format: <string>
6992 This parameter identifies the trust source as a backend
6993 for trusted keys implementation. Supported trust
6994 sources:
6995 - "tpm"
6996 - "tee"
6997 - "caam"
6998 - "dcp"
6999 If not specified then it defaults to iterating through
7000 the trust source list starting with TPM and assigns the
7001 first trust source as a backend which is initialized
7002 successfully during iteration.
7004 trusted.rng= [KEYS]
7005 Format: <string>
7006 The RNG used to generate key material for trusted keys.
7007 Can be one of:
7008 - "kernel"
7009 - the same value as trusted.source: "tpm" or "tee"
7010 - "default"
7011 If not specified, "default" is used. In this case,
7012 the RNG's choice is left to each individual trust source.
7014 trusted.dcp_use_otp_key
7015 This is intended to be used in combination with
7016 trusted.source=dcp and will select the DCP OTP key
7017 instead of the DCP UNIQUE key blob encryption.
7019 trusted.dcp_skip_zk_test
7020 This is intended to be used in combination with
7021 trusted.source=dcp and will disable the check if the
7022 blob key is all zeros. This is helpful for situations where
7023 having this key zero'ed is acceptable. E.g. in testing
7024 scenarios.
7026 tsc= Disable clocksource stability checks for TSC.
7027 Format: <string>
7028 [x86] reliable: mark tsc clocksource as reliable, this
7029 disables clocksource verification at runtime, as well
7030 as the stability checks done at bootup. Used to enable
7031 high-resolution timer mode on older hardware, and in
7032 virtualized environment.
7033 [x86] noirqtime: Do not use TSC to do irq accounting.
7034 Used to run time disable IRQ_TIME_ACCOUNTING on any
7035 platforms where RDTSC is slow and this accounting
7036 can add overhead.
7037 [x86] unstable: mark the TSC clocksource as unstable, this
7038 marks the TSC unconditionally unstable at bootup and
7039 avoids any further wobbles once the TSC watchdog notices.
7040 [x86] nowatchdog: disable clocksource watchdog. Used
7041 in situations with strict latency requirements (where
7042 interruptions from clocksource watchdog are not
7043 acceptable).
7044 [x86] recalibrate: force recalibration against a HW timer
7045 (HPET or PM timer) on systems whose TSC frequency was
7046 obtained from HW or FW using either an MSR or CPUID(0x15).
7047 Warn if the difference is more than 500 ppm.
7048 [x86] watchdog: Use TSC as the watchdog clocksource with
7049 which to check other HW timers (HPET or PM timer), but
7050 only on systems where TSC has been deemed trustworthy.
7051 This will be suppressed by an earlier tsc=nowatchdog and
7052 can be overridden by a later tsc=nowatchdog. A console
7053 message will flag any such suppression or overriding.
7055 tsc_early_khz= [X86,EARLY] Skip early TSC calibration and use the given
7056 value instead. Useful when the early TSC frequency discovery
7057 procedure is not reliable, such as on overclocked systems
7058 with CPUID.16h support and partial CPUID.15h support.
7059 Format: <unsigned int>
7061 tsx= [X86] Control Transactional Synchronization
7062 Extensions (TSX) feature in Intel processors that
7063 support TSX control.
7065 This parameter controls the TSX feature. The options are:
7067 on - Enable TSX on the system. Although there are
7068 mitigations for all known security vulnerabilities,
7069 TSX has been known to be an accelerator for
7070 several previous speculation-related CVEs, and
7071 so there may be unknown security risks associated
7072 with leaving it enabled.
7074 off - Disable TSX on the system. (Note that this
7075 option takes effect only on newer CPUs which are
7076 not vulnerable to MDS, i.e., have
7077 MSR_IA32_ARCH_CAPABILITIES.MDS_NO=1 and which get
7078 the new IA32_TSX_CTRL MSR through a microcode
7079 update. This new MSR allows for the reliable
7080 deactivation of the TSX functionality.)
7082 auto - Disable TSX if X86_BUG_TAA is present,
7083 otherwise enable TSX on the system.
7085 Not specifying this option is equivalent to tsx=off.
7087 See Documentation/admin-guide/hw-vuln/tsx_async_abort.rst
7088 for more details.
7090 tsx_async_abort= [X86,INTEL,EARLY] Control mitigation for the TSX Async
7091 Abort (TAA) vulnerability.
7093 Similar to Micro-architectural Data Sampling (MDS)
7094 certain CPUs that support Transactional
7095 Synchronization Extensions (TSX) are vulnerable to an
7096 exploit against CPU internal buffers which can forward
7097 information to a disclosure gadget under certain
7098 conditions.
7100 In vulnerable processors, the speculatively forwarded
7101 data can be used in a cache side channel attack, to
7102 access data to which the attacker does not have direct
7103 access.
7105 This parameter controls the TAA mitigation. The
7106 options are:
7108 full - Enable TAA mitigation on vulnerable CPUs
7109 if TSX is enabled.
7111 full,nosmt - Enable TAA mitigation and disable SMT on
7112 vulnerable CPUs. If TSX is disabled, SMT
7113 is not disabled because CPU is not
7114 vulnerable to cross-thread TAA attacks.
7115 off - Unconditionally disable TAA mitigation
7117 On MDS-affected machines, tsx_async_abort=off can be
7118 prevented by an active MDS mitigation as both vulnerabilities
7119 are mitigated with the same mechanism so in order to disable
7120 this mitigation, you need to specify mds=off too.
7122 Not specifying this option is equivalent to
7123 tsx_async_abort=full. On CPUs which are MDS affected
7124 and deploy MDS mitigation, TAA mitigation is not
7125 required and doesn't provide any additional
7126 mitigation.
7128 For details see:
7129 Documentation/admin-guide/hw-vuln/tsx_async_abort.rst
7131 turbografx.map[2|3]= [HW,JOY]
7132 TurboGraFX parallel port interface
7133 Format:
7134 <port#>,<js1>,<js2>,<js3>,<js4>,<js5>,<js6>,<js7>
7135 See also Documentation/input/devices/joystick-parport.rst
7137 udbg-immortal [PPC] When debugging early kernel crashes that
7138 happen after console_init() and before a proper
7139 console driver takes over, this boot options might
7140 help "seeing" what's going on.
7142 uhash_entries= [KNL,NET]
7143 Set number of hash buckets for UDP/UDP-Lite connections
7145 uhci-hcd.ignore_oc=
7146 [USB] Ignore overcurrent events (default N).
7147 Some badly-designed motherboards generate lots of
7148 bogus events, for ports that aren't wired to
7149 anything. Set this parameter to avoid log spamming.
7150 Note that genuine overcurrent events won't be
7151 reported either.
7153 unknown_nmi_panic
7154 [X86] Cause panic on unknown NMI.
7156 unwind_debug [X86-64,EARLY]
7157 Enable unwinder debug output. This can be
7158 useful for debugging certain unwinder error
7159 conditions, including corrupt stacks and
7160 bad/missing unwinder metadata.
7162 usbcore.authorized_default=
7163 [USB] Default USB device authorization:
7164 (default -1 = authorized (same as 1),
7165 0 = not authorized, 1 = authorized, 2 = authorized
7166 if device connected to internal port)
7168 usbcore.autosuspend=
7169 [USB] The autosuspend time delay (in seconds) used
7170 for newly-detected USB devices (default 2). This
7171 is the time required before an idle device will be
7172 autosuspended. Devices for which the delay is set
7173 to a negative value won't be autosuspended at all.
7175 usbcore.usbfs_snoop=
7176 [USB] Set to log all usbfs traffic (default 0 = off).
7178 usbcore.usbfs_snoop_max=
7179 [USB] Maximum number of bytes to snoop in each URB
7180 (default = 65536).
7182 usbcore.blinkenlights=
7183 [USB] Set to cycle leds on hubs (default 0 = off).
7185 usbcore.old_scheme_first=
7186 [USB] Start with the old device initialization
7187 scheme (default 0 = off).
7189 usbcore.usbfs_memory_mb=
7190 [USB] Memory limit (in MB) for buffers allocated by
7191 usbfs (default = 16, 0 = max = 2047).
7193 usbcore.use_both_schemes=
7194 [USB] Try the other device initialization scheme
7195 if the first one fails (default 1 = enabled).
7197 usbcore.initial_descriptor_timeout=
7198 [USB] Specifies timeout for the initial 64-byte
7199 USB_REQ_GET_DESCRIPTOR request in milliseconds
7200 (default 5000 = 5.0 seconds).
7202 usbcore.nousb [USB] Disable the USB subsystem
7204 usbcore.quirks=
7205 [USB] A list of quirk entries to augment the built-in
7206 usb core quirk list. List entries are separated by
7207 commas. Each entry has the form
7208 VendorID:ProductID:Flags. The IDs are 4-digit hex
7209 numbers and Flags is a set of letters. Each letter
7210 will change the built-in quirk; setting it if it is
7211 clear and clearing it if it is set. The letters have
7212 the following meanings:
7213 a = USB_QUIRK_STRING_FETCH_255 (string
7214 descriptors must not be fetched using
7215 a 255-byte read);
7216 b = USB_QUIRK_RESET_RESUME (device can't resume
7217 correctly so reset it instead);
7218 c = USB_QUIRK_NO_SET_INTF (device can't handle
7219 Set-Interface requests);
7220 d = USB_QUIRK_CONFIG_INTF_STRINGS (device can't
7221 handle its Configuration or Interface
7222 strings);
7223 e = USB_QUIRK_RESET (device can't be reset
7224 (e.g morph devices), don't use reset);
7225 f = USB_QUIRK_HONOR_BNUMINTERFACES (device has
7226 more interface descriptions than the
7227 bNumInterfaces count, and can't handle
7228 talking to these interfaces);
7229 g = USB_QUIRK_DELAY_INIT (device needs a pause
7230 during initialization, after we read
7231 the device descriptor);
7232 h = USB_QUIRK_LINEAR_UFRAME_INTR_BINTERVAL (For
7233 high speed and super speed interrupt
7234 endpoints, the USB 2.0 and USB 3.0 spec
7235 require the interval in microframes (1
7236 microframe = 125 microseconds) to be
7237 calculated as interval = 2 ^
7238 (bInterval-1).
7239 Devices with this quirk report their
7240 bInterval as the result of this
7241 calculation instead of the exponent
7242 variable used in the calculation);
7243 i = USB_QUIRK_DEVICE_QUALIFIER (device can't
7244 handle device_qualifier descriptor
7245 requests);
7246 j = USB_QUIRK_IGNORE_REMOTE_WAKEUP (device
7247 generates spurious wakeup, ignore
7248 remote wakeup capability);
7249 k = USB_QUIRK_NO_LPM (device can't handle Link
7250 Power Management);
7251 l = USB_QUIRK_LINEAR_FRAME_INTR_BINTERVAL
7252 (Device reports its bInterval as linear
7253 frames instead of the USB 2.0
7254 calculation);
7255 m = USB_QUIRK_DISCONNECT_SUSPEND (Device needs
7256 to be disconnected before suspend to
7257 prevent spurious wakeup);
7258 n = USB_QUIRK_DELAY_CTRL_MSG (Device needs a
7259 pause after every control message);
7260 o = USB_QUIRK_HUB_SLOW_RESET (Hub needs extra
7261 delay after resetting its port);
7262 p = USB_QUIRK_SHORT_SET_ADDRESS_REQ_TIMEOUT
7263 (Reduce timeout of the SET_ADDRESS
7264 request from 5000 ms to 500 ms);
7265 Example: quirks=0781:5580:bk,0a5c:5834:gij
7267 usbhid.mousepoll=
7268 [USBHID] The interval which mice are to be polled at.
7270 usbhid.jspoll=
7271 [USBHID] The interval which joysticks are to be polled at.
7273 usbhid.kbpoll=
7274 [USBHID] The interval which keyboards are to be polled at.
7276 usb-storage.delay_use=
7277 [UMS] The delay in seconds before a new device is
7278 scanned for Logical Units (default 1).
7279 Optionally the delay in milliseconds if the value has
7280 suffix with "ms".
7281 Example: delay_use=2567ms
7283 usb-storage.quirks=
7284 [UMS] A list of quirks entries to supplement or
7285 override the built-in unusual_devs list. List
7286 entries are separated by commas. Each entry has
7287 the form VID:PID:Flags where VID and PID are Vendor
7288 and Product ID values (4-digit hex numbers) and
7289 Flags is a set of characters, each corresponding
7290 to a common usb-storage quirk flag as follows:
7291 a = SANE_SENSE (collect more than 18 bytes
7292 of sense data, not on uas);
7293 b = BAD_SENSE (don't collect more than 18
7294 bytes of sense data, not on uas);
7295 c = FIX_CAPACITY (decrease the reported
7296 device capacity by one sector);
7297 d = NO_READ_DISC_INFO (don't use
7298 READ_DISC_INFO command, not on uas);
7299 e = NO_READ_CAPACITY_16 (don't use
7300 READ_CAPACITY_16 command);
7301 f = NO_REPORT_OPCODES (don't use report opcodes
7302 command, uas only);
7303 g = MAX_SECTORS_240 (don't transfer more than
7304 240 sectors at a time, uas only);
7305 h = CAPACITY_HEURISTICS (decrease the
7306 reported device capacity by one
7307 sector if the number is odd);
7308 i = IGNORE_DEVICE (don't bind to this
7309 device);
7310 j = NO_REPORT_LUNS (don't use report luns
7311 command, uas only);
7312 k = NO_SAME (do not use WRITE_SAME, uas only)
7313 l = NOT_LOCKABLE (don't try to lock and
7314 unlock ejectable media, not on uas);
7315 m = MAX_SECTORS_64 (don't transfer more
7316 than 64 sectors = 32 KB at a time,
7317 not on uas);
7318 n = INITIAL_READ10 (force a retry of the
7319 initial READ(10) command, not on uas);
7320 o = CAPACITY_OK (accept the capacity
7321 reported by the device, not on uas);
7322 p = WRITE_CACHE (the device cache is ON
7323 by default, not on uas);
7324 r = IGNORE_RESIDUE (the device reports
7325 bogus residue values, not on uas);
7326 s = SINGLE_LUN (the device has only one
7327 Logical Unit);
7328 t = NO_ATA_1X (don't allow ATA(12) and ATA(16)
7329 commands, uas only);
7330 u = IGNORE_UAS (don't bind to the uas driver);
7331 w = NO_WP_DETECT (don't test whether the
7332 medium is write-protected).
7333 y = ALWAYS_SYNC (issue a SYNCHRONIZE_CACHE
7334 even if the device claims no cache,
7335 not on uas)
7336 Example: quirks=0419:aaf5:rl,0421:0433:rc
7338 user_debug= [KNL,ARM]
7339 Format: <int>
7340 See arch/arm/Kconfig.debug help text.
7341 1 - undefined instruction events
7342 2 - system calls
7343 4 - invalid data aborts
7344 8 - SIGSEGV faults
7345 16 - SIGBUS faults
7346 Example: user_debug=31
7348 userpte=
7349 [X86,EARLY] Flags controlling user PTE allocations.
7351 nohigh = do not allocate PTE pages in
7352 HIGHMEM regardless of setting
7353 of CONFIG_HIGHPTE.
7355 vdso= [X86,SH,SPARC]
7356 On X86_32, this is an alias for vdso32=. Otherwise:
7358 vdso=1: enable VDSO (the default)
7359 vdso=0: disable VDSO mapping
7361 vdso32= [X86] Control the 32-bit vDSO
7362 vdso32=1: enable 32-bit VDSO
7363 vdso32=0 or vdso32=2: disable 32-bit VDSO
7365 See the help text for CONFIG_COMPAT_VDSO for more
7366 details. If CONFIG_COMPAT_VDSO is set, the default is
7367 vdso32=0; otherwise, the default is vdso32=1.
7369 For compatibility with older kernels, vdso32=2 is an
7370 alias for vdso32=0.
7372 Try vdso32=0 if you encounter an error that says:
7373 dl_main: Assertion `(void *) ph->p_vaddr == _rtld_local._dl_sysinfo_dso' failed!
7375 video= [FB,EARLY] Frame buffer configuration
7376 See Documentation/fb/modedb.rst.
7378 video.brightness_switch_enabled= [ACPI]
7379 Format: [0|1]
7380 If set to 1, on receiving an ACPI notify event
7381 generated by hotkey, video driver will adjust brightness
7382 level and then send out the event to user space through
7383 the allocated input device. If set to 0, video driver
7384 will only send out the event without touching backlight
7385 brightness level.
7386 default: 1
7388 virtio_mmio.device=
7389 [VMMIO] Memory mapped virtio (platform) device.
7391 <size>@<baseaddr>:<irq>[:<id>]
7392 where:
7393 <size> := size (can use standard suffixes
7394 like K, M and G)
7395 <baseaddr> := physical base address
7396 <irq> := interrupt number (as passed to
7397 request_irq())
7398 <id> := (optional) platform device id
7399 example:
7400 virtio_mmio.device=1K@0x100b0000:48:7
7402 Can be used multiple times for multiple devices.
7404 vga= [BOOT,X86-32] Select a particular video mode
7405 See Documentation/arch/x86/boot.rst and
7406 Documentation/admin-guide/svga.rst.
7407 Use vga=ask for menu.
7408 This is actually a boot loader parameter; the value is
7409 passed to the kernel using a special protocol.
7411 vm_debug[=options] [KNL] Available with CONFIG_DEBUG_VM=y.
7412 May slow down system boot speed, especially when
7413 enabled on systems with a large amount of memory.
7414 All options are enabled by default, and this
7415 interface is meant to allow for selectively
7416 enabling or disabling specific virtual memory
7417 debugging features.
7419 Available options are:
7420 P Enable page structure init time poisoning
7421 - Disable all of the above options
7423 vmalloc=nn[KMG] [KNL,BOOT,EARLY] Forces the vmalloc area to have an
7424 exact size of <nn>. This can be used to increase
7425 the minimum size (128MB on x86, arm32 platforms).
7426 It can also be used to decrease the size and leave more room
7427 for directly mapped kernel RAM. Note that this parameter does
7428 not exist on many other platforms (including arm64, alpha,
7429 loongarch, arc, csky, hexagon, microblaze, mips, nios2, openrisc,
7430 parisc, m64k, powerpc, riscv, sh, um, xtensa, s390, sparc).
7432 vmcp_cma=nn[MG] [KNL,S390,EARLY]
7433 Sets the memory size reserved for contiguous memory
7434 allocations for the vmcp device driver.
7436 vmhalt= [KNL,S390] Perform z/VM CP command after system halt.
7437 Format: <command>
7439 vmpanic= [KNL,S390] Perform z/VM CP command after kernel panic.
7440 Format: <command>
7442 vmpoff= [KNL,S390] Perform z/VM CP command after power off.
7443 Format: <command>
7445 vsyscall= [X86-64,EARLY]
7446 Controls the behavior of vsyscalls (i.e. calls to
7447 fixed addresses of 0xffffffffff600x00 from legacy
7448 code). Most statically-linked binaries and older
7449 versions of glibc use these calls. Because these
7450 functions are at fixed addresses, they make nice
7451 targets for exploits that can control RIP.
7453 emulate Vsyscalls turn into traps and are emulated
7454 reasonably safely. The vsyscall page is
7455 readable.
7457 xonly [default] Vsyscalls turn into traps and are
7458 emulated reasonably safely. The vsyscall
7459 page is not readable.
7461 none Vsyscalls don't work at all. This makes
7462 them quite hard to use for exploits but
7463 might break your system.
7465 vt.color= [VT] Default text color.
7466 Format: 0xYX, X = foreground, Y = background.
7467 Default: 0x07 = light gray on black.
7469 vt.cur_default= [VT] Default cursor shape.
7470 Format: 0xCCBBAA, where AA, BB, and CC are the same as
7471 the parameters of the <Esc>[?A;B;Cc escape sequence;
7472 see VGA-softcursor.txt. Default: 2 = underline.
7474 vt.default_blu= [VT]
7475 Format: <blue0>,<blue1>,<blue2>,...,<blue15>
7476 Change the default blue palette of the console.
7477 This is a 16-member array composed of values
7478 ranging from 0-255.
7480 vt.default_grn= [VT]
7481 Format: <green0>,<green1>,<green2>,...,<green15>
7482 Change the default green palette of the console.
7483 This is a 16-member array composed of values
7484 ranging from 0-255.
7486 vt.default_red= [VT]
7487 Format: <red0>,<red1>,<red2>,...,<red15>
7488 Change the default red palette of the console.
7489 This is a 16-member array composed of values
7490 ranging from 0-255.
7492 vt.default_utf8=
7493 [VT]
7494 Format=<0|1>
7495 Set system-wide default UTF-8 mode for all tty's.
7496 Default is 1, i.e. UTF-8 mode is enabled for all
7497 newly opened terminals.
7499 vt.global_cursor_default=
7500 [VT]
7501 Format=<-1|0|1>
7502 Set system-wide default for whether a cursor
7503 is shown on new VTs. Default is -1,
7504 i.e. cursors will be created by default unless
7505 overridden by individual drivers. 0 will hide
7506 cursors, 1 will display them.
7508 vt.italic= [VT] Default color for italic text; 0-15.
7509 Default: 2 = green.
7511 vt.underline= [VT] Default color for underlined text; 0-15.
7512 Default: 3 = cyan.
7514 watchdog timers [HW,WDT] For information on watchdog timers,
7515 see Documentation/watchdog/watchdog-parameters.rst
7516 or other driver-specific files in the
7517 Documentation/watchdog/ directory.
7519 watchdog_thresh=
7520 [KNL]
7521 Set the hard lockup detector stall duration
7522 threshold in seconds. The soft lockup detector
7523 threshold is set to twice the value. A value of 0
7524 disables both lockup detectors. Default is 10
7525 seconds.
7527 workqueue.unbound_cpus=
7528 [KNL,SMP] Specify to constrain one or some CPUs
7529 to use in unbound workqueues.
7530 Format: <cpu-list>
7531 By default, all online CPUs are available for
7532 unbound workqueues.
7534 workqueue.watchdog_thresh=
7535 If CONFIG_WQ_WATCHDOG is configured, workqueue can
7536 warn stall conditions and dump internal state to
7537 help debugging. 0 disables workqueue stall
7538 detection; otherwise, it's the stall threshold
7539 duration in seconds. The default value is 30 and
7540 it can be updated at runtime by writing to the
7541 corresponding sysfs file.
7543 workqueue.panic_on_stall=<uint>
7544 Panic when workqueue stall is detected by
7545 CONFIG_WQ_WATCHDOG. It sets the number times of the
7546 stall to trigger panic.
7548 The default is 0, which disables the panic on stall.
7550 workqueue.cpu_intensive_thresh_us=
7551 Per-cpu work items which run for longer than this
7552 threshold are automatically considered CPU intensive
7553 and excluded from concurrency management to prevent
7554 them from noticeably delaying other per-cpu work
7555 items. Default is 10000 (10ms).
7557 If CONFIG_WQ_CPU_INTENSIVE_REPORT is set, the kernel
7558 will report the work functions which violate this
7559 threshold repeatedly. They are likely good
7560 candidates for using WQ_UNBOUND workqueues instead.
7562 workqueue.cpu_intensive_warning_thresh=<uint>
7563 If CONFIG_WQ_CPU_INTENSIVE_REPORT is set, the kernel
7564 will report the work functions which violate the
7565 intensive_threshold_us repeatedly. In order to prevent
7566 spurious warnings, start printing only after a work
7567 function has violated this threshold number of times.
7569 The default is 4 times. 0 disables the warning.
7571 workqueue.power_efficient
7572 Per-cpu workqueues are generally preferred because
7573 they show better performance thanks to cache
7574 locality; unfortunately, per-cpu workqueues tend to
7575 be more power hungry than unbound workqueues.
7577 Enabling this makes the per-cpu workqueues which
7578 were observed to contribute significantly to power
7579 consumption unbound, leading to measurably lower
7580 power usage at the cost of small performance
7581 overhead.
7583 The default value of this parameter is determined by
7584 the config option CONFIG_WQ_POWER_EFFICIENT_DEFAULT.
7586 workqueue.default_affinity_scope=
7587 Select the default affinity scope to use for unbound
7588 workqueues. Can be one of "cpu", "smt", "cache",
7589 "numa" and "system". Default is "cache". For more
7590 information, see the Affinity Scopes section in
7591 Documentation/core-api/workqueue.rst.
7593 This can be changed after boot by writing to the
7594 matching /sys/module/workqueue/parameters file. All
7595 workqueues with the "default" affinity scope will be
7596 updated accordingly.
7598 workqueue.debug_force_rr_cpu
7599 Workqueue used to implicitly guarantee that work
7600 items queued without explicit CPU specified are put
7601 on the local CPU. This guarantee is no longer true
7602 and while local CPU is still preferred work items
7603 may be put on foreign CPUs. This debug option
7604 forces round-robin CPU selection to flush out
7605 usages which depend on the now broken guarantee.
7606 When enabled, memory and cache locality will be
7607 impacted.
7609 writecombine= [LOONGARCH,EARLY] Control the MAT (Memory Access
7610 Type) of ioremap_wc().
7612 on - Enable writecombine, use WUC for ioremap_wc()
7613 off - Disable writecombine, use SUC for ioremap_wc()
7615 x2apic_phys [X86-64,APIC,EARLY] Use x2apic physical mode instead of
7616 default x2apic cluster mode on platforms
7617 supporting x2apic.
7619 xen_512gb_limit [KNL,X86-64,XEN]
7620 Restricts the kernel running paravirtualized under Xen
7621 to use only up to 512 GB of RAM. The reason to do so is
7622 crash analysis tools and Xen tools for doing domain
7623 save/restore/migration must be enabled to handle larger
7624 domains.
7626 xen_emul_unplug= [HW,X86,XEN,EARLY]
7627 Unplug Xen emulated devices
7628 Format: [unplug0,][unplug1]
7629 ide-disks -- unplug primary master IDE devices
7630 aux-ide-disks -- unplug non-primary-master IDE devices
7631 nics -- unplug network devices
7632 all -- unplug all emulated devices (NICs and IDE disks)
7633 unnecessary -- unplugging emulated devices is
7634 unnecessary even if the host did not respond to
7635 the unplug protocol
7636 never -- do not unplug even if version check succeeds
7638 xen_legacy_crash [X86,XEN,EARLY]
7639 Crash from Xen panic notifier, without executing late
7640 panic() code such as dumping handler.
7642 xen_mc_debug [X86,XEN,EARLY]
7643 Enable multicall debugging when running as a Xen PV guest.
7644 Enabling this feature will reduce performance a little
7645 bit, so it should only be enabled for obtaining extended
7646 debug data in case of multicall errors.
7648 xen_msr_safe= [X86,XEN,EARLY]
7649 Format: <bool>
7650 Select whether to always use non-faulting (safe) MSR
7651 access functions when running as Xen PV guest. The
7652 default value is controlled by CONFIG_XEN_PV_MSR_SAFE.
7654 xen_nopv [X86]
7655 Disables the PV optimizations forcing the HVM guest to
7656 run as generic HVM guest with no PV drivers.
7657 This option is obsoleted by the "nopv" option, which
7658 has equivalent effect for XEN platform.
7660 xen_no_vector_callback
7661 [KNL,X86,XEN,EARLY] Disable the vector callback for Xen
7662 event channel interrupts.
7664 xen_scrub_pages= [XEN]
7665 Boolean option to control scrubbing pages before giving them back
7666 to Xen, for use by other domains. Can be also changed at runtime
7667 with /sys/devices/system/xen_memory/xen_memory0/scrub_pages.
7668 Default value controlled with CONFIG_XEN_SCRUB_PAGES_DEFAULT.
7670 xen_timer_slop= [X86-64,XEN,EARLY]
7671 Set the timer slop (in nanoseconds) for the virtual Xen
7672 timers (default is 100000). This adjusts the minimum
7673 delta of virtualized Xen timers, where lower values
7674 improve timer resolution at the expense of processing
7675 more timer interrupts.
7677 xen.balloon_boot_timeout= [XEN]
7678 The time (in seconds) to wait before giving up to boot
7679 in case initial ballooning fails to free enough memory.
7680 Applies only when running as HVM or PVH guest and
7681 started with less memory configured than allowed at
7682 max. Default is 180.
7684 xen.event_eoi_delay= [XEN]
7685 How long to delay EOI handling in case of event
7686 storms (jiffies). Default is 10.
7688 xen.event_loop_timeout= [XEN]
7689 After which time (jiffies) the event handling loop
7690 should start to delay EOI handling. Default is 2.
7692 xen.fifo_events= [XEN]
7693 Boolean parameter to disable using fifo event handling
7694 even if available. Normally fifo event handling is
7695 preferred over the 2-level event handling, as it is
7696 fairer and the number of possible event channels is
7697 much higher. Default is on (use fifo events).
7699 xirc2ps_cs= [NET,PCMCIA]
7700 Format:
7701 <irq>,<irq_mask>,<io>,<full_duplex>,<do_sound>,<lockup_hack>[,<irq2>[,<irq3>[,<irq4>]]]
7703 xive= [PPC]
7704 By default on POWER9 and above, the kernel will
7705 natively use the XIVE interrupt controller. This option
7706 allows the fallback firmware mode to be used:
7708 off Fallback to firmware control of XIVE interrupt
7709 controller on both pseries and powernv
7710 platforms. Only useful on POWER9 and above.
7712 xive.store-eoi=off [PPC]
7713 By default on POWER10 and above, the kernel will use
7714 stores for EOI handling when the XIVE interrupt mode
7715 is active. This option allows the XIVE driver to use
7716 loads instead, as on POWER9.
7718 xhci-hcd.quirks [USB,KNL]
7719 A hex value specifying bitmask with supplemental xhci
7720 host controller quirks. Meaning of each bit can be
7721 consulted in header drivers/usb/host/xhci.h.
7723 xmon [PPC,EARLY]
7724 Format: { early | on | rw | ro | off }
7725 Controls if xmon debugger is enabled. Default is off.
7726 Passing only "xmon" is equivalent to "xmon=early".
7727 early Call xmon as early as possible on boot; xmon
7728 debugger is called from setup_arch().
7729 on xmon debugger hooks will be installed so xmon
7730 is only called on a kernel crash. Default mode,
7731 i.e. either "ro" or "rw" mode, is controlled
7732 with CONFIG_XMON_DEFAULT_RO_MODE.
7733 rw xmon debugger hooks will be installed so xmon
7734 is called only on a kernel crash, mode is write,
7735 meaning SPR registers, memory and, other data
7736 can be written using xmon commands.
7737 ro same as "rw" option above but SPR registers,
7738 memory, and other data can't be written using
7739 xmon commands.
7740 off xmon is disabled.