1 acpi= [HW,ACPI,X86,ARM64]
2 Advanced Configuration and Power Interface
3 Format: { force | on | off | strict | noirq | rsdt |
5 force -- enable ACPI if default was off
6 on -- enable ACPI but allow fallback to DT [arm64]
7 off -- disable ACPI if default was on
8 noirq -- do not use ACPI for IRQ routing
9 strict -- Be less tolerant of platforms that are not
10 strictly ACPI specification compliant.
11 rsdt -- prefer RSDT over (default) XSDT
12 copy_dsdt -- copy DSDT to memory
13 For ARM64, ONLY "acpi=off", "acpi=on" or "acpi=force"
16 See also Documentation/power/runtime_pm.rst, pci=noacpi
18 acpi_apic_instance= [ACPI, IOAPIC]
20 2: use 2nd APIC table, if available
21 1,0: use 1st APIC table
24 acpi_backlight= [HW,ACPI]
25 { vendor | video | native | none }
26 If set to vendor, prefer vendor-specific driver
27 (e.g. thinkpad_acpi, sony_acpi, etc.) instead
28 of the ACPI video.ko driver.
29 If set to video, use the ACPI video.ko driver.
30 If set to native, use the device's native backlight mode.
31 If set to none, disable the ACPI backlight interface.
33 acpi_force_32bit_fadt_addr
34 force FADT to use 32 bit addresses rather than the
35 64 bit X_* addresses. Some firmware have broken 64
36 bit addresses for force ACPI ignore these and use
37 the older legacy 32 bit addresses.
39 acpica_no_return_repair [HW, ACPI]
40 Disable AML predefined validation mechanism
41 This mechanism can repair the evaluation result to make
42 the return objects more ACPI specification compliant.
43 This option is useful for developers to identify the
44 root cause of an AML interpreter issue when the issue
45 has something to do with the repair mechanism.
47 acpi.debug_layer= [HW,ACPI,ACPI_DEBUG]
48 acpi.debug_level= [HW,ACPI,ACPI_DEBUG]
50 CONFIG_ACPI_DEBUG must be enabled to produce any ACPI
51 debug output. Bits in debug_layer correspond to a
52 _COMPONENT in an ACPI source file, e.g.,
53 #define _COMPONENT ACPI_EVENTS
54 Bits in debug_level correspond to a level in
55 ACPI_DEBUG_PRINT statements, e.g.,
56 ACPI_DEBUG_PRINT((ACPI_DB_INFO, ...
57 The debug_level mask defaults to "info". See
58 Documentation/firmware-guide/acpi/debug.rst for more information about
59 debug layers and levels.
61 Enable processor driver info messages:
62 acpi.debug_layer=0x20000000
63 Enable AML "Debug" output, i.e., stores to the Debug
64 object while interpreting AML:
65 acpi.debug_layer=0xffffffff acpi.debug_level=0x2
66 Enable all messages related to ACPI hardware:
67 acpi.debug_layer=0x2 acpi.debug_level=0xffffffff
69 Some values produce so much output that the system is
70 unusable. The "log_buf_len" parameter may be useful
71 if you need to capture more output.
73 acpi_enforce_resources= [ACPI]
75 Check for resource conflicts between native drivers
76 and ACPI OperationRegions (SystemIO and SystemMemory
77 only). IO ports and memory declared in ACPI might be
78 used by the ACPI subsystem in arbitrary AML code and
79 can interfere with legacy drivers.
80 strict (default): access to resources claimed by ACPI
81 is denied; legacy drivers trying to access reserved
82 resources will fail to bind to device using them.
83 lax: access to resources claimed by ACPI is allowed;
84 legacy drivers trying to access reserved resources
85 will bind successfully but a warning message is logged.
86 no: ACPI OperationRegions are not marked as reserved,
87 no further checks are performed.
89 acpi_force_table_verification [HW,ACPI]
90 Enable table checksum verification during early stage.
91 By default, this is disabled due to x86 early mapping
94 acpi_irq_balance [HW,ACPI]
95 ACPI will balance active IRQs
98 acpi_irq_nobalance [HW,ACPI]
99 ACPI will not move active IRQs (default)
102 acpi_irq_isa= [HW,ACPI] If irq_balance, mark listed IRQs used by ISA
103 Format: <irq>,<irq>...
105 acpi_irq_pci= [HW,ACPI] If irq_balance, clear listed IRQs for
107 Format: <irq>,<irq>...
109 acpi_mask_gpe= [HW,ACPI]
110 Due to the existence of _Lxx/_Exx, some GPEs triggered
111 by unsupported hardware/firmware features can result in
112 GPE floodings that cannot be automatically disabled by
114 This facility can be used to prevent such uncontrolled
116 Format: <byte> or <bitmap-list>
118 acpi_no_auto_serialize [HW,ACPI]
119 Disable auto-serialization of AML methods
120 AML control methods that contain the opcodes to create
121 named objects will be marked as "Serialized" by the
122 auto-serialization feature.
123 This feature is enabled by default.
124 This option allows to turn off the feature.
126 acpi_no_memhotplug [ACPI] Disable memory hotplug. Useful for kdump
129 acpi_no_static_ssdt [HW,ACPI]
130 Disable installation of static SSDTs at early boot time
131 By default, SSDTs contained in the RSDT/XSDT will be
132 installed automatically and they will appear under
133 /sys/firmware/acpi/tables.
134 This option turns off this feature.
135 Note that specifying this option does not affect
136 dynamic table installation which will install SSDT
137 tables to /sys/firmware/acpi/tables/dynamic.
139 acpi_no_watchdog [HW,ACPI,WDT]
140 Ignore the ACPI-based watchdog interface (WDAT) and let
141 a native driver control the watchdog device instead.
143 acpi_rsdp= [ACPI,EFI,KEXEC]
144 Pass the RSDP address to the kernel, mostly used
145 on machines running EFI runtime service to boot the
146 second kernel for kdump.
148 acpi_os_name= [HW,ACPI] Tell ACPI BIOS the name of the OS
149 Format: To spoof as Windows 98: ="Microsoft Windows"
151 acpi_rev_override [ACPI] Override the _REV object to return 5 (instead
152 of 2 which is mandated by ACPI 6) as the supported ACPI
153 specification revision (when using this switch, it may
154 be necessary to carry out a cold reboot _twice_ in a
155 row to make it take effect on the platform firmware).
157 acpi_osi= [HW,ACPI] Modify list of supported OS interface strings
158 acpi_osi="string1" # add string1
159 acpi_osi="!string2" # remove string2
160 acpi_osi=!* # remove all strings
161 acpi_osi=! # disable all built-in OS vendor
163 acpi_osi=!! # enable all built-in OS vendor
165 acpi_osi= # disable all strings
167 'acpi_osi=!' can be used in combination with single or
168 multiple 'acpi_osi="string1"' to support specific OS
169 vendor string(s). Note that such command can only
170 affect the default state of the OS vendor strings, thus
171 it cannot affect the default state of the feature group
172 strings and the current state of the OS vendor strings,
173 specifying it multiple times through kernel command line
174 is meaningless. This command is useful when one do not
175 care about the state of the feature group strings which
176 should be controlled by the OSPM.
178 1. 'acpi_osi=! acpi_osi="Windows 2000"' is equivalent
179 to 'acpi_osi="Windows 2000" acpi_osi=!', they all
180 can make '_OSI("Windows 2000")' TRUE.
182 'acpi_osi=' cannot be used in combination with other
183 'acpi_osi=' command lines, the _OSI method will not
184 exist in the ACPI namespace. NOTE that such command can
185 only affect the _OSI support state, thus specifying it
186 multiple times through kernel command line is also
189 1. 'acpi_osi=' can make 'CondRefOf(_OSI, Local1)'
192 'acpi_osi=!*' can be used in combination with single or
193 multiple 'acpi_osi="string1"' to support specific
194 string(s). Note that such command can affect the
195 current state of both the OS vendor strings and the
196 feature group strings, thus specifying it multiple times
197 through kernel command line is meaningful. But it may
198 still not able to affect the final state of a string if
199 there are quirks related to this string. This command
200 is useful when one want to control the state of the
201 feature group strings to debug BIOS issues related to
204 1. 'acpi_osi="Module Device" acpi_osi=!*' can make
205 '_OSI("Module Device")' FALSE.
206 2. 'acpi_osi=!* acpi_osi="Module Device"' can make
207 '_OSI("Module Device")' TRUE.
208 3. 'acpi_osi=! acpi_osi=!* acpi_osi="Windows 2000"' is
210 'acpi_osi=!* acpi_osi=! acpi_osi="Windows 2000"'
212 'acpi_osi=!* acpi_osi="Windows 2000" acpi_osi=!',
213 they all will make '_OSI("Windows 2000")' TRUE.
216 Override the pmtimer bug detection: force the kernel
217 to assume that this machine's pmtimer latches its value
218 and always returns good values.
220 acpi_sci= [HW,ACPI] ACPI System Control Interrupt trigger mode
221 Format: { level | edge | high | low }
223 acpi_skip_timer_override [HW,ACPI]
224 Recognize and ignore IRQ0/pin2 Interrupt Override.
225 For broken nForce2 BIOS resulting in XT-PIC timer.
227 acpi_sleep= [HW,ACPI] Sleep options
228 Format: { s3_bios, s3_mode, s3_beep, s4_hwsig,
229 s4_nohwsig, old_ordering, nonvs,
230 sci_force_enable, nobl }
231 See Documentation/power/video.rst for information on
233 s3_beep is for debugging; it makes the PC's speaker beep
234 as soon as the kernel's real-mode entry point is called.
235 s4_hwsig causes the kernel to check the ACPI hardware
236 signature during resume from hibernation, and gracefully
237 refuse to resume if it has changed. This complies with
238 the ACPI specification but not with reality, since
239 Windows does not do this and many laptops do change it
240 on docking. So the default behaviour is to allow resume
241 and simply warn when the signature changes, unless the
242 s4_hwsig option is enabled.
243 s4_nohwsig prevents ACPI hardware signature from being
244 used (or even warned about) during resume.
245 old_ordering causes the ACPI 1.0 ordering of the _PTS
246 control method, with respect to putting devices into
247 low power states, to be enforced (the ACPI 2.0 ordering
248 of _PTS is used by default).
249 nonvs prevents the kernel from saving/restoring the
250 ACPI NVS memory during suspend/hibernation and resume.
251 sci_force_enable causes the kernel to set SCI_EN directly
252 on resume from S1/S3 (which is against the ACPI spec,
253 but some broken systems don't work without it).
254 nobl causes the internal blacklist of systems known to
255 behave incorrectly in some ways with respect to system
256 suspend and resume to be ignored (use wisely).
258 acpi_use_timer_override [HW,ACPI]
259 Use timer override. For some broken Nvidia NF5 boards
260 that require a timer override, but don't have HPET
262 add_efi_memmap [EFI; X86] Include EFI memory map in
263 kernel's map of available physical RAM.
266 { off | try_unsupported }
267 off: disable AGP support
268 try_unsupported: try to drive unsupported chipsets
269 (may crash computer or cause data corruption)
272 See Documentation/sound/alsa-configuration.rst
275 Allow the default userspace alignment fault handler
276 behaviour to be specified. Bit 0 enables warnings,
277 bit 1 enables fixups, and bit 2 sends a segfault.
279 align_va_addr= [X86-64]
280 Align virtual addresses by clearing slice [14:12] when
281 allocating a VMA at process creation time. This option
282 gives you up to 3% performance improvement on AMD F15h
283 machines (where it is enabled by default) for a
284 CPU-intensive style benchmark, and it can vary highly in
285 a microbenchmark depending on workload and compiler.
287 32: only for 32-bit processes
288 64: only for 64-bit processes
289 on: enable for both 32- and 64-bit processes
290 off: disable for both 32- and 64-bit processes
292 alloc_snapshot [FTRACE]
293 Allocate the ftrace snapshot buffer on boot up when the
294 main buffer is allocated. This is handy if debugging
295 and you need to use tracing_snapshot() on boot up, and
296 do not want to use tracing_snapshot_alloc() as it needs
297 to be done where GFP_KERNEL allocations are allowed.
299 allow_mismatched_32bit_el0 [ARM64]
300 Allow execve() of 32-bit applications and setting of the
301 PER_LINUX32 personality on systems where only a strict
302 subset of the CPUs support 32-bit EL0. When this
303 parameter is present, the set of CPUs supporting 32-bit
304 EL0 is indicated by /sys/devices/system/cpu/aarch32_el0
305 and hot-unplug operations may be restricted.
307 See Documentation/arm64/asymmetric-32bit.rst for more
310 amd_iommu= [HW,X86-64]
311 Pass parameters to the AMD IOMMU driver in the system.
313 fullflush - Deprecated, equivalent to iommu.strict=1
314 off - do not initialize any AMD IOMMU found in
316 force_isolation - Force device isolation for all
317 devices. The IOMMU driver is not
318 allowed anymore to lift isolation
319 requirements as needed. This option
320 does not override iommu=pt
321 force_enable - Force enable the IOMMU on platforms known
322 to be buggy with IOMMU enabled. Use this
325 amd_iommu_dump= [HW,X86-64]
326 Enable AMD IOMMU driver option to dump the ACPI table
327 for AMD IOMMU. With this option enabled, AMD IOMMU
328 driver will print ACPI tables for AMD IOMMU during
329 IOMMU initialization.
331 amd_iommu_intr= [HW,X86-64]
332 Specifies one of the following AMD IOMMU interrupt
334 legacy - Use legacy interrupt remapping mode.
335 vapic - Use virtual APIC mode, which allows IOMMU
336 to inject interrupts directly into guest.
337 This mode requires kvm-amd.avic=1.
338 (Default when IOMMU HW support is present.)
340 amijoy.map= [HW,JOY] Amiga joystick support
341 Map of devices attached to JOY0DAT and JOY1DAT
343 See also Documentation/input/joydev/joystick.rst
345 analog.map= [HW,JOY] Analog joystick and gamepad support
346 Specifies type or capabilities of an analog joystick
347 connected to one of 16 gameports
348 Format: <type1>,<type2>,..<type16>
351 Power management functions (SPARCstation-4/5 + deriv.)
353 Disable APC CPU standby support. SPARCstation-Fox does
354 not play well with APC CPU idle - disable it if you have
355 APC and your system crashes randomly.
357 apic= [APIC,X86] Advanced Programmable Interrupt Controller
358 Change the output verbosity while booting
359 Format: { quiet (default) | verbose | debug }
360 Change the amount of debugging information output
361 when initialising the APIC and IO-APIC components.
362 For X86-32, this can also be used to specify an APIC
364 Format: apic=driver_name
365 Examples: apic=bigsmp
367 apic_extnmi= [APIC,X86] External NMI delivery setting
368 Format: { bsp (default) | all | none }
369 bsp: External NMI is delivered only to CPU 0
370 all: External NMIs are broadcast to all CPUs as a
372 none: External NMI is masked for all CPUs. This is
373 useful so that a dump capture kernel won't be
377 See Documentation/networking/ipv6.rst.
379 show_lapic= [APIC,X86] Advanced Programmable Interrupt Controller
380 Limit apic dumping. The parameter defines the maximal
381 number of local apics being dumped. Also it is possible
382 to set it to "all" by meaning -- no limit here.
383 Format: { 1 (default) | 2 | ... | all }.
384 The parameter valid if only apic=debug or
385 apic=verbose is specified.
386 Example: apic=debug show_lapic=all
388 apm= [APM] Advanced Power Management
389 See header of arch/x86/kernel/apm_32.c.
391 arcrimi= [HW,NET] ARCnet - "RIM I" (entirely mem-mapped) cards
392 Format: <io>,<irq>,<nodeID>
394 arm64.nobti [ARM64] Unconditionally disable Branch Target
395 Identification support
397 arm64.nopauth [ARM64] Unconditionally disable Pointer Authentication
400 arm64.nomte [ARM64] Unconditionally disable Memory Tagging Extension
403 arm64.nosve [ARM64] Unconditionally disable Scalable Vector
406 arm64.nosme [ARM64] Unconditionally disable Scalable Matrix
411 atarimouse= [HW,MOUSE] Atari Mouse
413 atkbd.extra= [HW] Enable extra LEDs and keys on IBM RapidAccess,
414 EzKey and similar keyboards
416 atkbd.reset= [HW] Reset keyboard during initialization
418 atkbd.set= [HW] Select keyboard code set
419 Format: <int> (2 = AT (default), 3 = PS/2)
421 atkbd.scroll= [HW] Enable scroll wheel on MS Office and similar
424 atkbd.softraw= [HW] Choose between synthetic and real raw mode
425 Format: <bool> (0 = real, 1 = synthetic (default))
427 atkbd.softrepeat= [HW]
428 Use software keyboard repeat
430 audit= [KNL] Enable the audit sub-system
431 Format: { "0" | "1" | "off" | "on" }
432 0 | off - kernel audit is disabled and can not be
433 enabled until the next reboot
434 unset - kernel audit is initialized but disabled and
435 will be fully enabled by the userspace auditd.
436 1 | on - kernel audit is initialized and partially
437 enabled, storing at most audit_backlog_limit
438 messages in RAM until it is fully enabled by the
442 audit_backlog_limit= [KNL] Set the audit queue size limit.
443 Format: <int> (must be >=0)
446 bau= [X86_UV] Enable the BAU on SGI UV. The default
447 behavior is to disable the BAU (i.e. bau=0).
448 Format: { "0" | "1" }
451 unset - Disable the BAU.
453 baycom_epp= [HW,AX25]
456 baycom_par= [HW,AX25] BayCom Parallel Port AX.25 Modem
458 See header of drivers/net/hamradio/baycom_par.c.
460 baycom_ser_fdx= [HW,AX25]
461 BayCom Serial Port AX.25 Modem (Full Duplex Mode)
462 Format: <io>,<irq>,<mode>[,<baud>]
463 See header of drivers/net/hamradio/baycom_ser_fdx.c.
465 baycom_ser_hdx= [HW,AX25]
466 BayCom Serial Port AX.25 Modem (Half Duplex Mode)
467 Format: <io>,<irq>,<mode>
468 See header of drivers/net/hamradio/baycom_ser_hdx.c.
471 Disable BERT OS support on buggy BIOSes.
473 bgrt_disable [ACPI][X86]
474 Disable BGRT to avoid flickering OEM logo.
476 blkdevparts= Manual partition parsing of block device(s) for
477 embedded devices based on command line input.
478 See Documentation/block/cmdline-partition.rst
480 boot_delay= Milliseconds to delay each printk during boot.
481 Values larger than 10 seconds (10000) are changed to
486 Extended command line options can be added to an initrd
487 and this will cause the kernel to look for it.
489 See Documentation/admin-guide/bootconfig.rst
491 bttv.card= [HW,V4L] bttv (bt848 + bt878 based grabber cards)
492 bttv.radio= Most important insmod options are available as
494 bttv.pll= See Documentation/admin-guide/media/bttv.rst
497 bulk_remove=off [PPC] This parameter disables the use of the pSeries
498 firmware feature for flushing multiple hpte entries
501 c101= [NET] Moxa C101 synchronous serial card
503 cachesize= [BUGS=X86-32] Override level 2 CPU cache size detection.
504 Sometimes CPU hardware bugs make them report the cache
505 size incorrectly. The kernel will attempt work arounds
506 to fix known problems, but for some CPUs it is not
507 possible to determine what the correct size should be.
508 This option provides an override for these situations.
511 [NET] Specifies amount of time (in seconds) that
512 the kernel should wait for a network carrier. By default
513 it waits 120 seconds.
515 ca_keys= [KEYS] This parameter identifies a specific key(s) on
516 the system trusted keyring to be used for certificate
518 format: { id:<keyid> | builtin }
520 cca= [MIPS] Override the kernel pages' cache coherency
521 algorithm. Accepted values range from 0 to 7
522 inclusive. See arch/mips/include/asm/pgtable-bits.h
523 for platform specific values (SB1, Loongson3 and
526 ccw_timeout_log [S390]
527 See Documentation/s390/common_io.rst for details.
529 cgroup_disable= [KNL] Disable a particular controller or optional feature
530 Format: {name of the controller(s) or feature(s) to disable}
531 The effects of cgroup_disable=foo are:
532 - foo isn't auto-mounted if you mount all cgroups in
534 - foo isn't visible as an individually mountable
536 - if foo is an optional feature then the feature is
537 disabled and corresponding cgroup files are not
539 {Currently only "memory" controller deal with this and
540 cut the overhead, others just disable the usage. So
541 only cgroup_disable=memory is actually worthy}
542 Specifying "pressure" disables per-cgroup pressure
543 stall information accounting feature
545 cgroup_no_v1= [KNL] Disable cgroup controllers and named hierarchies in v1
546 Format: { { controller | "all" | "named" }
547 [,{ controller | "all" | "named" }...] }
548 Like cgroup_disable, but only applies to cgroup v1;
549 the blacklisted controllers remain available in cgroup2.
550 "all" blacklists all controllers and "named" disables
551 named mounts. Specifying both "all" and "named" disables
554 cgroup.memory= [KNL] Pass options to the cgroup memory controller.
556 nosocket -- Disable socket memory accounting.
557 nokmem -- Disable kernel memory accounting.
559 checkreqprot= [SELINUX] Set initial checkreqprot flag value.
560 Format: { "0" | "1" }
561 See security/selinux/Kconfig help text.
562 0 -- check protection applied by kernel (includes
563 any implied execute protection).
564 1 -- check protection requested by application.
565 Default value is set via a kernel config option.
566 Value can be changed at runtime via
567 /sys/fs/selinux/checkreqprot.
568 Setting checkreqprot to 1 is deprecated.
571 See Documentation/s390/common_io.rst for details.
573 clearcpuid=X[,X...] [X86]
574 Disable CPUID feature X for the kernel. See
575 arch/x86/include/asm/cpufeatures.h for the valid bit
576 numbers X. Note the Linux-specific bits are not necessarily
577 stable over kernel options, but the vendor-specific
579 X can also be a string as appearing in the flags: line
580 in /proc/cpuinfo which does not have the above
581 instability issue. However, not all features have names
583 Note that using this option will taint your kernel.
584 Also note that user programs calling CPUID directly
585 or using the feature without checking anything
586 will still see it. This just prevents it from
587 being used by the kernel or shown in /proc/cpuinfo.
588 Also note the kernel might malfunction if you disable
593 Prevents the clock framework from automatically gating
594 clocks that have not been explicitly enabled by a Linux
595 device driver but are enabled in hardware at reset or
596 by the bootloader/firmware. Note that this does not
597 force such clocks to be always-on nor does it reserve
598 those clocks in any way. This parameter is useful for
599 debug and development, but should not be needed on a
600 platform with proper driver support. For more
601 information, see Documentation/driver-api/clk.rst.
603 clock= [BUGS=X86-32, HW] gettimeofday clocksource override.
605 Forces specified clocksource (if available) to be used
606 when calculating gettimeofday(). If specified
607 clocksource is not available, it defaults to PIT.
608 Format: { pit | tsc | cyclone | pmtmr }
610 clocksource= Override the default clocksource
612 Override the default clocksource and use the clocksource
613 with the name specified.
614 Some clocksource names to choose from, depending on
616 [all] jiffies (this is the base, fallback clocksource)
618 [ARM] imx_timer1,OSTS,netx_timer,mpu_timer2,
619 pxa_timer,timer3,32k_counter,timer0_1
620 [X86-32] pit,hpet,tsc;
621 scx200_hrt on Geode; cyclone on IBM x440
629 clocksource.arm_arch_timer.evtstrm=
632 Enable/disable the eventstream feature of the ARM
633 architected timer so that code using WFE-based polling
634 loops can be debugged more effectively on production
637 clocksource.max_cswd_read_retries= [KNL]
638 Number of clocksource_watchdog() retries due to
639 external delays before the clock will be marked
640 unstable. Defaults to two retries, that is,
641 three attempts to read the clock under test.
643 clocksource.verify_n_cpus= [KNL]
644 Limit the number of CPUs checked for clocksources
645 marked with CLOCK_SOURCE_VERIFY_PERCPU that
646 are marked unstable due to excessive skew.
647 A negative value says to check all CPUs, while
648 zero says not to check any. Values larger than
649 nr_cpu_ids are silently truncated to nr_cpu_ids.
650 The actual CPUs are chosen randomly, with
651 no replacement if the same CPU is chosen twice.
653 clocksource-wdtest.holdoff= [KNL]
654 Set the time in seconds that the clocksource
655 watchdog test waits before commencing its tests.
656 Defaults to zero when built as a module and to
657 10 seconds when built into the kernel.
659 cma=nn[MG]@[start[MG][-end[MG]]]
661 Sets the size of kernel global memory area for
662 contiguous memory allocations and optionally the
663 placement constraint by the physical address range of
664 memory allocations. A value of 0 disables CMA
665 altogether. For more information, see
666 kernel/dma/contiguous.c
670 Sets the size of kernel per-numa memory area for
671 contiguous memory allocations. A value of 0 disables
672 per-numa CMA altogether. And If this option is not
673 specificed, the default value is 0.
674 With per-numa CMA enabled, DMA users on node nid will
675 first try to allocate buffer from the pernuma area
676 which is located in node nid, if the allocation fails,
677 they will fallback to the global default memory area.
679 cmo_free_hint= [PPC] Format: { yes | no }
680 Specify whether pages are marked as being inactive
681 when they are freed. This is used in CMO environments
682 to determine OS memory pressure for page stealing by
686 coherent_pool=nn[KMG] [ARM,KNL]
687 Sets the size of memory pool for coherent, atomic dma
688 allocations, by default set to 256K.
690 com20020= [HW,NET] ARCnet - COM20020 chipset
692 <io>[,<irq>[,<nodeID>[,<backplane>[,<ckp>[,<timeout>]]]]]
694 com90io= [HW,NET] ARCnet - COM90xx chipset (IO-mapped buffers)
698 ARCnet - COM90xx chipset (memory-mapped buffers)
699 Format: <io>[,<irq>[,<memstart>]]
701 condev= [HW,S390] console device
704 console= [KNL] Output console device and options.
706 tty<n> Use the virtual console device <n>.
710 Use the specified serial port. The options are of
711 the form "bbbbpnf", where "bbbb" is the baud rate,
712 "p" is parity ("n", "o", or "e"), "n" is number of
713 bits, and "f" is flow control ("r" for RTS or
714 omit it). Default is "9600n8".
716 See Documentation/admin-guide/serial-console.rst for more
718 Documentation/networking/netconsole.rst for an
721 uart[8250],io,<addr>[,options]
722 uart[8250],mmio,<addr>[,options]
723 uart[8250],mmio16,<addr>[,options]
724 uart[8250],mmio32,<addr>[,options]
725 uart[8250],0x<addr>[,options]
726 Start an early, polled-mode console on the 8250/16550
727 UART at the specified I/O port or MMIO address,
728 switching to the matching ttyS device later.
729 MMIO inter-register address stride is either 8-bit
730 (mmio), 16-bit (mmio16), or 32-bit (mmio32).
731 If none of [io|mmio|mmio16|mmio32], <addr> is assumed
732 to be equivalent to 'mmio'. 'options' are specified in
733 the same format described for ttyS above; if unspecified,
734 the h/w is not re-initialized.
736 hvc<n> Use the hypervisor console device <n>. This is for
737 both Xen and PowerPC hypervisors.
740 Use to disable console output, i.e., to have kernel
741 console messages discarded.
742 This must be the only console= parameter used on the
745 If the device connected to the port is not a TTY but a braille
746 device, prepend "brl," before the device type, for instance
748 For now, only VisioBraille is supported.
751 [KNL] Change console messages format
753 By default we print messages on consoles in
754 "[time stamp] text\n" format (time stamp may not be
755 printed, depending on CONFIG_PRINTK_TIME or
756 `printk_time' param).
758 Switch to syslog format: "<%u>[time stamp] text\n"
759 IOW, each message will have a facility and loglevel
760 prefix. The format is similar to one used by syslog()
761 syscall, or to executing "dmesg -S --raw" or to reading
764 consoleblank= [KNL] The console blank (screen saver) timeout in
765 seconds. A value of 0 disables the blank timer.
769 [KNL] Change the default value for
770 /proc/<pid>/coredump_filter.
771 See also Documentation/filesystems/proc.rst.
773 coresight_cpu_debug.enable
776 Enable/disable the CPU sampling based debugging.
777 0: default value, disable debugging
778 1: enable debugging at boot time
780 cpcihp_generic= [HW,PCI] Generic port I/O CompactPCI driver
782 <first_slot>,<last_slot>,<port>,<enum_bit>[,<debug>]
784 cpu0_hotplug [X86] Turn on CPU0 hotplug feature when
785 CONFIG_BOOTPARAM_HOTPLUG_CPU0 is off.
786 Some features depend on CPU0. Known dependencies are:
787 1. Resume from suspend/hibernate depends on CPU0.
788 Suspend/hibernate will fail if CPU0 is offline and you
789 need to online CPU0 before suspend/hibernate.
790 2. PIC interrupts also depend on CPU0. CPU0 can't be
791 removed if a PIC interrupt is detected.
792 It's said poweroff/reboot may depend on CPU0 on some
793 machines although I haven't seen such issues so far
794 after CPU0 is offline on a few tested machines.
795 If the dependencies are under your control, you can
796 turn on cpu0_hotplug.
798 cpuidle.off=1 [CPU_IDLE]
799 disable the cpuidle sub-system
802 [CPU_IDLE] Name of the cpuidle governor to use.
804 cpufreq.off=1 [CPU_FREQ]
805 disable the cpufreq sub-system
807 cpufreq.default_governor=
808 [CPU_FREQ] Name of the default cpufreq governor or
809 policy to use. This governor must be registered in the
810 kernel before the cpufreq driver probes.
813 [X86] Delay for N microsec between assert and de-assert
814 of APIC INIT to start processors. This delay occurs
815 on every CPU online, such as boot, and resume from suspend.
818 crash_kexec_post_notifiers
819 Run kdump after running panic-notifiers and dumping
820 kmsg. This only for the users who doubt kdump always
821 succeeds in any situation.
822 Note that this also increases risks of kdump failure,
823 because some panic notifiers can make the crashed
824 kernel more unstable.
826 crashkernel=size[KMG][@offset[KMG]]
827 [KNL] Using kexec, Linux can switch to a 'crash kernel'
828 upon panic. This parameter reserves the physical
829 memory region [offset, offset + size] for that kernel
830 image. If '@offset' is omitted, then a suitable offset
831 is selected automatically.
832 [KNL, X86-64] Select a region under 4G first, and
833 fall back to reserve region above 4G when '@offset'
834 hasn't been specified.
835 See Documentation/admin-guide/kdump/kdump.rst for further details.
837 crashkernel=range1:size1[,range2:size2,...][@offset]
838 [KNL] Same as above, but depends on the memory
839 in the running system. The syntax of range is
840 start-[end] where start and end are both
841 a memory unit (amount[KMG]). See also
842 Documentation/admin-guide/kdump/kdump.rst for an example.
844 crashkernel=size[KMG],high
845 [KNL, X86-64, ARM64] range could be above 4G. Allow kernel
846 to allocate physical memory region from top, so could
847 be above 4G if system have more than 4G ram installed.
848 Otherwise memory region will be allocated below 4G, if
850 It will be ignored if crashkernel=X is specified.
851 crashkernel=size[KMG],low
852 [KNL, X86-64] range under 4G. When crashkernel=X,high
853 is passed, kernel could allocate physical memory region
854 above 4G, that cause second kernel crash on system
855 that require some amount of low memory, e.g. swiotlb
856 requires at least 64M+32K low memory, also enough extra
857 low memory is needed to make sure DMA buffers for 32-bit
858 devices won't run out. Kernel would try to allocate
859 at least 256M below 4G automatically.
860 This one lets the user specify own low range under 4G
861 for second kernel instead.
862 0: to disable low allocation.
863 It will be ignored when crashkernel=X,high is not used
864 or memory reserved is below 4G.
866 [KNL, ARM64] range in low memory.
867 This one lets the user specify a low range in the
868 DMA zone for the crash dump kernel.
869 It will be ignored when crashkernel=X,high is not used
870 or memory reserved is located in the DMA zones.
873 [KNL] Disable crypto self-tests
878 cs89x0_media= [HW,NET]
879 Format: { rj45 | aui | bnc }
881 csdlock_debug= [KNL] Enable debug add-ons of cross-CPU function call
882 handling. When switched on, additional debug data is
883 printed to the console in case a hanging CPU is
884 detected, and that CPU is pinged again in order to try
885 to resolve the hang situation.
886 0: disable csdlock debugging (default)
887 1: enable basic csdlock debugging (minor impact)
888 ext: enable extended csdlock debugging (more impact,
892 See header of drivers/s390/block/dasd_devmap.c.
894 db9.dev[2|3]= [HW,JOY] Multisystem joystick support via parallel port
895 (one device per port)
896 Format: <port#>,<type>
897 See also Documentation/input/devices/joystick-parport.rst
899 debug [KNL] Enable kernel debugging (events log level).
902 [KNL] Enable printing [hashed] pointers early in the
903 boot sequence. If enabled, we use a weak hash instead
904 of siphash to hash pointers. Use this option if you are
905 seeing instances of '(___ptrval___)') and need to see a
906 value (hashed pointer) instead. Cryptographically
907 insecure, please do not use on production kernels.
910 [KNL] verbose locking self-tests
912 Print debugging info while doing the locking API
914 Bitmask for the various LOCKTYPE_ tests. Defaults to 0
915 (no extra messages), setting it to -1 (all bits set)
916 will print _a_lot_ more information - normally only
917 useful to lockdep developers.
919 debug_objects [KNL] Enable object debugging
922 [KNL] Disable object debugging
924 debug_guardpage_minorder=
925 [KNL] When CONFIG_DEBUG_PAGEALLOC is set, this
926 parameter allows control of the order of pages that will
927 be intentionally kept free (and hence protected) by the
928 buddy allocator. Bigger value increase the probability
929 of catching random memory corruption, but reduce the
930 amount of memory for normal system use. The maximum
931 possible value is MAX_ORDER/2. Setting this parameter
932 to 1 or 2 should be enough to identify most random
933 memory corruption problems caused by bugs in kernel or
934 driver code when a CPU writes to (or reads from) a
935 random memory location. Note that there exists a class
936 of memory corruptions problems caused by buggy H/W or
937 F/W or by drivers badly programing DMA (basically when
938 memory is written at bus level and the CPU MMU is
939 bypassed) which are not detectable by
940 CONFIG_DEBUG_PAGEALLOC, hence this option will not help
941 tracking down these problems.
944 [KNL] When CONFIG_DEBUG_PAGEALLOC is set, this parameter
945 enables the feature at boot time. By default, it is
946 disabled and the system will work mostly the same as a
947 kernel built without CONFIG_DEBUG_PAGEALLOC.
948 Note: to get most of debug_pagealloc error reports, it's
949 useful to also enable the page_owner functionality.
950 on: enable the feature
952 debugfs= [KNL] This parameter enables what is exposed to userspace
953 and debugfs internal clients.
954 Format: { on, no-mount, off }
955 on: All functions are enabled.
957 Filesystem is not registered but kernel clients can
958 access APIs and a crashkernel can be used to read
959 its content. There is nothing to mount.
960 off: Filesystem is not registered and clients
961 get a -EPERM as result when trying to register files
962 or directories within debugfs.
963 This is equivalent of the runtime functionality if
964 debugfs was not enabled in the kernel at all.
965 Default value is set in build-time with a kernel configuration.
967 debugpat [X86] Enable PAT debugging
969 decnet.addr= [HW,NET]
970 Format: <area>[,<node>]
971 See also Documentation/networking/decnet.rst.
974 [HW] The size of the default HugeTLB page. This is
975 the size represented by the legacy /proc/ hugepages
976 APIs. In addition, this is the default hugetlb size
977 used for shmget(), mmap() and mounting hugetlbfs
978 filesystems. If not specified, defaults to the
979 architecture's default huge page size. Huge page
980 sizes are architecture dependent. See also
981 Documentation/admin-guide/mm/hugetlbpage.rst.
984 deferred_probe_timeout=
985 [KNL] Debugging option to set a timeout in seconds for
986 deferred probe to give up waiting on dependencies to
987 probe. Only specific dependencies (subsystems or
988 drivers) that have opted in will be ignored. A timeout
989 of 0 will timeout at the end of initcalls. If the time
990 out hasn't expired, it'll be restarted by each
991 successful driver registration. This option will also
992 dump out devices still on the deferred probe list after
995 delayacct [KNL] Enable per-task delay accounting
997 dell_smm_hwmon.ignore_dmi=
998 [HW] Continue probing hardware even if DMI data
999 indicates that the driver is running on unsupported
1002 dell_smm_hwmon.force=
1003 [HW] Activate driver even if SMM BIOS signature does
1004 not match list of supported models and enable otherwise
1005 blacklisted features.
1007 dell_smm_hwmon.power_status=
1008 [HW] Report power status in /proc/i8k
1009 (disabled by default).
1011 dell_smm_hwmon.restricted=
1012 [HW] Allow controlling fans only if SYS_ADMIN
1015 dell_smm_hwmon.fan_mult=
1016 [HW] Factor to multiply fan speed with.
1018 dell_smm_hwmon.fan_max=
1019 [HW] Maximum configurable fan speed.
1022 Format: { on | off | def_only | inf_only | always }
1023 on: s390 zlib hardware support for compression on
1024 level 1 and decompression (default)
1025 off: No s390 zlib hardware support
1026 def_only: s390 zlib hardware support for deflate
1027 only (compression on level 1)
1028 inf_only: s390 zlib hardware support for inflate
1029 only (decompression)
1030 always: Same as 'on' but ignores the selected compression
1031 level always using hardware support (used for debugging)
1033 dhash_entries= [KNL]
1034 Set number of hash buckets for dentry cache.
1036 disable_1tb_segments [PPC]
1037 Disables the use of 1TB hash page table segments. This
1038 causes the kernel to fall back to 256MB segments which
1039 can be useful when debugging issues that require an SLB
1043 Limits the number of kernel SLB entries, and flushes
1044 them frequently to increase the rate of SLB faults
1045 on kernel addresses.
1048 See Documentation/networking/ipv6.rst.
1051 Disable RADIX MMU mode on POWER9
1053 radix_hcall_invalidate=on [PPC/PSERIES]
1054 Disable RADIX GTSE feature and use hcall for TLB
1058 Disable TLBIE instruction. Currently does not work
1059 with KVM, with HASH MMU, or with coherent accelerators.
1061 disable_cpu_apicid= [X86,APIC,SMP]
1063 The number of initial APIC ID for the
1064 corresponding CPU to be disabled at boot,
1065 mostly used for the kdump 2nd kernel to
1066 disable BSP to wake up multiple CPUs without
1067 causing system reset or hang due to sending
1068 INIT from AP to BSP.
1070 disable_ddw [PPC/PSERIES]
1071 Disable Dynamic DMA Window support. Use this
1072 to workaround buggy firmware.
1074 disable_ipv6= [IPV6]
1075 See Documentation/networking/ipv6.rst.
1077 disable_mtrr_cleanup [X86]
1078 The kernel tries to adjust MTRR layout from continuous
1079 to discrete, to make X server driver able to add WB
1080 entry later. This parameter disables that.
1082 disable_mtrr_trim [X86, Intel and AMD only]
1083 By default the kernel will trim any uncacheable
1084 memory out of your available memory pool based on
1085 MTRR settings. This parameter disables that behavior,
1086 possibly causing your machine to run very slowly.
1088 disable_timer_pin_1 [X86]
1089 Disable PIN 1 of APIC timer
1090 Can be useful to work around chipset bugs.
1092 dis_ucode_ldr [X86] Disable the microcode loader.
1094 dma_debug=off If the kernel is compiled with DMA_API_DEBUG support,
1095 this option disables the debugging code at boot.
1097 dma_debug_entries=<number>
1098 This option allows to tune the number of preallocated
1099 entries for DMA-API debugging code. One entry is
1100 required per DMA-API allocation. Use this if the
1101 DMA-API debugging code disables itself because the
1102 architectural default is too low.
1104 dma_debug_driver=<driver_name>
1105 With this option the DMA-API debugging driver
1106 filter feature can be enabled at boot time. Just
1107 pass the driver to filter for as the parameter.
1108 The filter can be disabled or changed to another
1109 driver later using sysfs.
1111 driver_async_probe= [KNL]
1112 List of driver names to be probed asynchronously. *
1113 matches with all driver names. If * is specified, the
1114 rest of the listed driver names are those that will NOT
1116 Format: <driver_name1>,<driver_name2>...
1118 drm.edid_firmware=[<connector>:]<file>[,[<connector>:]<file>]
1119 Broken monitors, graphic adapters, KVMs and EDIDless
1120 panels may send no or incorrect EDID data sets.
1121 This parameter allows to specify an EDID data sets
1122 in the /lib/firmware directory that are used instead.
1123 Generic built-in EDID data sets are used, if one of
1124 edid/1024x768.bin, edid/1280x1024.bin,
1125 edid/1680x1050.bin, or edid/1920x1080.bin is given
1126 and no file with the same name exists. Details and
1127 instructions how to build your own EDID data are
1128 available in Documentation/admin-guide/edid.rst. An EDID
1129 data set will only be used for a particular connector,
1130 if its name and a colon are prepended to the EDID
1131 name. Each connector may use a unique EDID data
1132 set by separating the files with a comma. An EDID
1133 data set with no connector name will be used for
1134 any connectors not explicitly specified.
1139 Format: {"off" | "known"}
1140 Control how the dt_cpu_ftrs device-tree binding is
1141 used for CPU feature discovery and setup (if it
1143 off: Do not use it, fall back to legacy cpu table.
1144 known: Do not pass through unknown features to guests
1145 or userspace, only those that the kernel is aware of.
1147 dump_apple_properties [X86]
1148 Dump name and content of EFI device properties on
1149 x86 Macs. Useful for driver authors to determine
1150 what data is available or for reverse-engineering.
1152 dyndbg[="val"] [KNL,DYNAMIC_DEBUG]
1153 <module>.dyndbg[="val"]
1154 Enable debug messages at boot time. See
1155 Documentation/admin-guide/dynamic-debug-howto.rst
1158 nopku [X86] Disable Memory Protection Keys CPU feature found
1161 <module>.async_probe [KNL]
1162 Enable asynchronous probe on this module.
1164 early_ioremap_debug [KNL]
1165 Enable debug messages in early_ioremap support. This
1166 is useful for tracking down temporary early mappings
1167 which are not unmapped.
1169 earlycon= [KNL] Output early console device and options.
1171 When used with no options, the early console is
1172 determined by stdout-path property in device tree's
1173 chosen node or the ACPI SPCR table if supported by
1176 cdns,<addr>[,options]
1177 Start an early, polled-mode console on a Cadence
1178 (xuartps) serial port at the specified address. Only
1179 supported option is baud rate. If baud rate is not
1180 specified, the serial port must already be setup and
1183 uart[8250],io,<addr>[,options]
1184 uart[8250],mmio,<addr>[,options]
1185 uart[8250],mmio32,<addr>[,options]
1186 uart[8250],mmio32be,<addr>[,options]
1187 uart[8250],0x<addr>[,options]
1188 Start an early, polled-mode console on the 8250/16550
1189 UART at the specified I/O port or MMIO address.
1190 MMIO inter-register address stride is either 8-bit
1191 (mmio) or 32-bit (mmio32 or mmio32be).
1192 If none of [io|mmio|mmio32|mmio32be], <addr> is assumed
1193 to be equivalent to 'mmio'. 'options' are specified
1194 in the same format described for "console=ttyS<n>"; if
1195 unspecified, the h/w is not initialized.
1199 Start an early, polled-mode console on a pl011 serial
1200 port at the specified address. The pl011 serial port
1201 must already be setup and configured. Options are not
1202 yet supported. If 'mmio32' is specified, then only
1203 the driver will use only 32-bit accessors to read/write
1204 the device registers.
1207 Start an early console on a litex serial port at the
1208 specified address. The serial port must already be
1209 setup and configured. Options are not yet supported.
1212 Start an early, polled-mode console on a meson serial
1213 port at the specified address. The serial port must
1214 already be setup and configured. Options are not yet
1218 Start an early, polled-mode console on an msm serial
1219 port at the specified address. The serial port
1220 must already be setup and configured. Options are not
1223 msm_serial_dm,<addr>
1224 Start an early, polled-mode console on an msm serial
1225 dm port at the specified address. The serial port
1226 must already be setup and configured. Options are not
1230 Start an early, polled-mode console on a serial port
1231 of an Actions Semi SoC, such as S500 or S900, at the
1232 specified address. The serial port must already be
1233 setup and configured. Options are not yet supported.
1236 Start an early, polled-mode console on a serial port
1237 of an RDA Micro SoC, such as RDA8810PL, at the
1238 specified address. The serial port must already be
1239 setup and configured. Options are not yet supported.
1242 Use RISC-V SBI (Supervisor Binary Interface) for early
1245 smh Use ARM semihosting calls for early console.
1253 Use early console provided by serial driver available
1254 on Samsung SoCs, requires selecting proper type and
1255 a correct base address of the selected UART port. The
1256 serial port must already be setup and configured.
1257 Options are not yet supported.
1260 Start an early, polled-mode console on a lantiq serial
1261 (lqasc) port at the specified address. The serial port
1262 must already be setup and configured. Options are not
1267 Use early console provided by Freescale LP UART driver
1268 found on Freescale Vybrid and QorIQ LS1021A processors.
1269 A valid base address must be provided, and the serial
1270 port must already be setup and configured.
1274 Start an early, polled-mode, output-only console on the
1275 Freescale i.MX UART at the specified address. The UART
1276 must already be setup and configured.
1279 Start an early, polled-mode console on the
1280 Armada 3700 serial port at the specified
1281 address. The serial port must already be setup
1282 and configured. Options are not yet supported.
1285 Start an early, polled-mode console on a Qualcomm
1286 Generic Interface (GENI) based serial port at the
1287 specified address. The serial port must already be
1288 setup and configured. Options are not yet supported.
1291 Start an early, unaccelerated console on the EFI
1292 memory mapped framebuffer (if available). On cache
1293 coherent non-x86 systems that use system memory for
1294 the framebuffer, pass the 'ram' option so that it is
1295 mapped with the correct attributes.
1298 Use early console provided by Freescale LINFlexD UART
1299 serial driver for NXP S32V234 SoCs. A valid base
1300 address must be provided, and the serial port must
1301 already be setup and configured.
1303 earlyprintk= [X86,SH,ARM,M68k,S390]
1307 earlyprintk=serial[,ttySn[,baudrate]]
1308 earlyprintk=serial[,0x...[,baudrate]]
1309 earlyprintk=ttySn[,baudrate]
1310 earlyprintk=dbgp[debugController#]
1311 earlyprintk=pciserial[,force],bus:device.function[,baudrate]
1312 earlyprintk=xdbc[xhciController#]
1314 earlyprintk is useful when the kernel crashes before
1315 the normal console is initialized. It is not enabled by
1316 default because it has some cosmetic problems.
1318 Append ",keep" to not disable it when the real console
1321 Only one of vga, serial, or usb debug port can
1324 Currently only ttyS0 and ttyS1 may be specified by
1325 name. Other I/O ports may be explicitly specified
1326 on some architectures (x86 and arm at least) by
1327 replacing ttySn with an I/O port address, like this:
1328 earlyprintk=serial,0x1008,115200
1329 You can find the port for a given device in
1330 /proc/tty/driver/serial:
1331 2: uart:ST16650V2 port:00001008 irq:18 ...
1333 Interaction with the standard serial driver is not
1336 The VGA output is eventually overwritten by
1339 The xen option can only be used in Xen domains.
1341 The sclp output can only be used on s390.
1343 The optional "force" to "pciserial" enables use of a
1344 PCI device even when its classcode is not of the
1347 edac_report= [HW,EDAC] Control how to report EDAC event
1348 Format: {"on" | "off" | "force"}
1349 on: enable EDAC to report H/W event. May be overridden
1350 by other higher priority error reporting module.
1351 off: disable H/W event reporting through EDAC.
1352 force: enforce the use of EDAC to report H/W event.
1356 Format: {"off" | "on" | "skip[mbr]"}
1359 Format: { "debug", "disable_early_pci_dma",
1360 "nochunk", "noruntime", "nosoftreserve",
1361 "novamap", "no_disable_early_pci_dma" }
1362 debug: enable misc debug output.
1363 disable_early_pci_dma: disable the busmaster bit on all
1364 PCI bridges while in the EFI boot stub.
1365 nochunk: disable reading files in "chunks" in the EFI
1366 boot stub, as chunking can cause problems with some
1367 firmware implementations.
1368 noruntime : disable EFI runtime services support
1369 nosoftreserve: The EFI_MEMORY_SP (Specific Purpose)
1370 attribute may cause the kernel to reserve the
1371 memory range for a memory mapping driver to
1372 claim. Specify efi=nosoftreserve to disable this
1373 reservation and treat the memory by its base type
1374 (i.e. EFI_CONVENTIONAL_MEMORY / "System RAM").
1375 novamap: do not call SetVirtualAddressMap().
1376 no_disable_early_pci_dma: Leave the busmaster bit set
1377 on all PCI bridges while in the EFI boot stub
1379 efi_no_storage_paranoia [EFI; X86]
1380 Using this parameter you can use more than 50% of
1381 your efi variable storage. Use this parameter only if
1382 you are really sure that your UEFI does sane gc and
1383 fulfills the spec otherwise your board may brick.
1385 efi_fake_mem= nn[KMG]@ss[KMG]:aa[,nn[KMG]@ss[KMG]:aa,..] [EFI; X86]
1386 Add arbitrary attribute to specific memory range by
1387 updating original EFI memory map.
1388 Region of memory which aa attribute is added to is
1391 If efi_fake_mem=2G@4G:0x10000,2G@0x10a0000000:0x10000
1392 is specified, EFI_MEMORY_MORE_RELIABLE(0x10000)
1393 attribute is added to range 0x100000000-0x180000000 and
1394 0x10a0000000-0x1120000000.
1396 If efi_fake_mem=8G@9G:0x40000 is specified, the
1397 EFI_MEMORY_SP(0x40000) attribute is added to
1398 range 0x240000000-0x43fffffff.
1400 Using this parameter you can do debugging of EFI memmap
1401 related features. For example, you can do debugging of
1402 Address Range Mirroring feature even if your box
1403 doesn't support it, or mark specific memory as
1406 efivar_ssdt= [EFI; X86] Name of an EFI variable that contains an SSDT
1407 that is to be dynamically loaded by Linux. If there are
1408 multiple variables with the same name but with different
1409 vendor GUIDs, all of them will be loaded. See
1410 Documentation/admin-guide/acpi/ssdt-overlays.rst for details.
1413 eisa_irq_edge= [PARISC,HW]
1414 See header of drivers/parisc/eisa.c.
1416 ekgdboc= [X86,KGDB] Allow early kernel console debugging
1419 This is designed to be used in conjunction with
1420 the boot argument: earlyprintk=vga
1422 This parameter works in place of the kgdboc parameter
1423 but can only be used if the backing tty is available
1424 very early in the boot process. For early debugging
1425 via a serial port see kgdboc_earlycon instead.
1428 See comment before function elanfreq_setup() in
1429 arch/x86/kernel/cpu/cpufreq/elanfreq.c.
1431 elfcorehdr=[size[KMG]@]offset[KMG] [IA64,PPC,SH,X86,S390]
1432 Specifies physical address of start of kernel core
1433 image elf header and optionally the size. Generally
1434 kexec loader will pass this option to capture kernel.
1435 See Documentation/admin-guide/kdump/kdump.rst for details.
1437 enable_mtrr_cleanup [X86]
1438 The kernel tries to adjust MTRR layout from continuous
1439 to discrete, to make X server driver able to add WB
1440 entry later. This parameter enables that.
1442 enable_timer_pin_1 [X86]
1443 Enable PIN 1 of APIC timer
1444 Can be useful to work around chipset bugs
1445 (in particular on some ATI chipsets).
1446 The kernel tries to set a reasonable default.
1448 enforcing= [SELINUX] Set initial enforcing status.
1450 See security/selinux/Kconfig help text.
1451 0 -- permissive (log only, no denials).
1452 1 -- enforcing (deny and log).
1454 Value can be changed at runtime via
1455 /sys/fs/selinux/enforce.
1458 Disable Error Record Serialization Table (ERST)
1461 ether= [HW,NET] Ethernet cards parameters
1462 This option is obsoleted by the "netdev=" option, which
1463 has equivalent usage. See its documentation for details.
1467 Permit 'security.evm' to be updated regardless of
1468 current integrity status.
1473 fail_make_request=[KNL]
1474 General fault injection mechanism.
1475 Format: <interval>,<probability>,<space>,<times>
1476 See also Documentation/fault-injection/.
1479 Format: { initns | none }
1480 See Documentation/admin-guide/sysctl/net.rst for
1481 fb_tunnels_only_for_init_ns
1484 See Documentation/admin-guide/blockdev/floppy.rst.
1486 force_pal_cache_flush
1487 [IA-64] Avoid check_sal_cache_flush which may hang on
1488 buggy SAL_CACHE_FLUSH implementations. Using this
1489 parameter will force ia64_sal_cache_flush to call
1490 ia64_pal_cache_flush instead of SAL_CACHE_FLUSH.
1493 Forcefully enable Physical Address Extension (PAE).
1494 Many Pentium M systems disable PAE but may have a
1495 functionally usable PAE implementation.
1496 Warning: use of this parameter will taint the kernel
1497 and may cause unknown problems.
1500 [FTRACE] will set and start the specified tracer
1501 as early as possible in order to facilitate early
1504 ftrace_boot_snapshot
1505 [FTRACE] On boot up, a snapshot will be taken of the
1506 ftrace ring buffer that can be read at:
1507 /sys/kernel/tracing/snapshot.
1508 This is useful if you need tracing information from kernel
1509 boot up that is likely to be overridden by user space
1510 start up functionality.
1512 ftrace_dump_on_oops[=orig_cpu]
1513 [FTRACE] will dump the trace buffers on oops.
1514 If no parameter is passed, ftrace will dump
1515 buffers of all CPUs, but if you pass orig_cpu, it will
1516 dump only the buffer of the CPU that triggered the
1519 ftrace_filter=[function-list]
1520 [FTRACE] Limit the functions traced by the function
1521 tracer at boot up. function-list is a comma-separated
1522 list of functions. This list can be changed at run
1523 time by the set_ftrace_filter file in the debugfs
1526 ftrace_notrace=[function-list]
1527 [FTRACE] Do not trace the functions specified in
1528 function-list. This list can be changed at run time
1529 by the set_ftrace_notrace file in the debugfs
1532 ftrace_graph_filter=[function-list]
1533 [FTRACE] Limit the top level callers functions traced
1534 by the function graph tracer at boot up.
1535 function-list is a comma-separated list of functions
1536 that can be changed at run time by the
1537 set_graph_function file in the debugfs tracing directory.
1539 ftrace_graph_notrace=[function-list]
1540 [FTRACE] Do not trace from the functions specified in
1541 function-list. This list is a comma-separated list of
1542 functions that can be changed at run time by the
1543 set_graph_notrace file in the debugfs tracing directory.
1545 ftrace_graph_max_depth=<uint>
1546 [FTRACE] Used with the function graph tracer. This is
1547 the max depth it will trace into a function. This value
1548 can be changed at run time by the max_graph_depth file
1549 in the tracefs tracing directory. default: 0 (no limit)
1551 fw_devlink= [KNL] Create device links between consumer and supplier
1552 devices by scanning the firmware to infer the
1553 consumer/supplier relationships. This feature is
1554 especially useful when drivers are loaded as modules as
1555 it ensures proper ordering of tasks like device probing
1556 (suppliers first, then consumers), supplier boot state
1557 clean up (only after all consumers have probed),
1558 suspend/resume & runtime PM (consumers first, then
1560 Format: { off | permissive | on | rpm }
1561 off -- Don't create device links from firmware info.
1562 permissive -- Create device links from firmware info
1563 but use it only for ordering boot state clean
1564 up (sync_state() calls).
1565 on -- Create device links from firmware info and use it
1566 to enforce probe and suspend/resume ordering.
1567 rpm -- Like "on", but also use to order runtime PM.
1569 fw_devlink.strict=<bool>
1570 [KNL] Treat all inferred dependencies as mandatory
1571 dependencies. This only applies for fw_devlink=on|rpm.
1575 [HW,JOY] Multisystem joystick and NES/SNES/PSX pad
1576 support via parallel port (up to 5 devices per port)
1577 Format: <port#>,<pad1>,<pad2>,<pad3>,<pad4>,<pad5>
1578 See also Documentation/input/devices/joystick-parport.rst
1582 gart_fix_e820= [X86-64] disable the fix e820 for K8 GART
1586 gcov_persist= [GCOV] When non-zero (default), profiling data for
1587 kernel modules is saved and remains accessible via
1588 debugfs, even when the module is unloaded/reloaded.
1589 When zero, profiling data is discarded and associated
1590 debugfs files are removed at module unload time.
1592 goldfish [X86] Enable the goldfish android emulator platform.
1593 Don't use this when you are not running on the
1596 gpio-mockup.gpio_mockup_ranges
1597 [HW] Sets the ranges of gpiochip of for this device.
1598 Format: <start1>,<end1>,<start2>,<end2>...
1599 gpio-mockup.gpio_mockup_named_lines
1600 [HW] Let the driver know GPIO lines should be named.
1602 gpt [EFI] Forces disk with valid GPT signature but
1603 invalid Protective MBR to be treated as GPT. If the
1604 primary GPT is corrupted, it enables the backup/alternate
1605 GPT to be used instead.
1607 grcan.enable0= [HW] Configuration of physical interface 0. Determines
1608 the "Enable 0" bit of the configuration register.
1611 grcan.enable1= [HW] Configuration of physical interface 1. Determines
1612 the "Enable 0" bit of the configuration register.
1615 grcan.select= [HW] Select which physical interface to use.
1618 grcan.txsize= [HW] Sets the size of the tx buffer.
1619 Format: <unsigned int> such that (txsize & ~0x1fffc0) == 0.
1621 grcan.rxsize= [HW] Sets the size of the rx buffer.
1622 Format: <unsigned int> such that (rxsize & ~0x1fffc0) == 0.
1626 [KNL] Under CONFIG_HARDENED_USERCOPY, whether
1627 hardening is enabled for this boot. Hardened
1628 usercopy checking is used to protect the kernel
1629 from reading or writing beyond known memory
1630 allocation boundaries as a proactive defense
1631 against bounds-checking flaws in the kernel's
1632 copy_to_user()/copy_from_user() interface.
1633 on Perform hardened usercopy checks (default).
1634 off Disable hardened usercopy checks.
1636 hardlockup_all_cpu_backtrace=
1637 [KNL] Should the hard-lockup detector generate
1638 backtraces on all cpus.
1641 hashdist= [KNL,NUMA] Large hashes allocated during boot
1642 are distributed across NUMA nodes. Defaults on
1643 for 64-bit NUMA, off otherwise.
1644 Format: 0 | 1 (for off | on)
1646 hcl= [IA-64] SGI's Hardware Graph compatibility layer
1648 hd= [EIDE] (E)IDE hard drive subsystem geometry
1649 Format: <cyl>,<head>,<sect>
1652 Disable Hardware Error Source Table (HEST) support;
1653 corresponding firmware-first mode error processing
1654 logic will be disabled.
1656 hibernate= [HIBERNATION]
1657 noresume Don't check if there's a hibernation image
1658 present during boot.
1659 nocompress Don't compress/decompress hibernation images.
1660 no Disable hibernation and resume.
1661 protect_image Turn on image protection during restoration
1662 (that will set all pages holding image data
1663 during restoration read-only).
1665 highmem=nn[KMG] [KNL,BOOT] forces the highmem zone to have an exact
1666 size of <nn>. This works even on boxes that have no
1667 highmem otherwise. This also works to reduce highmem
1668 size on bigger boxes.
1670 highres= [KNL] Enable/disable high resolution timer mode.
1671 Valid parameters: "on", "off"
1676 hpet= [X86-32,HPET] option to control HPET usage
1677 Format: { enable (default) | disable | force |
1679 disable: disable HPET and use PIT instead
1680 force: allow force enabled of undocumented chips (ICH4,
1682 verbose: show contents of HPET registers during setup
1684 hpet_mmap= [X86, HPET_MMAP] Allow userspace to mmap HPET
1685 registers. Default set by CONFIG_HPET_MMAP_DEFAULT.
1687 hugepages= [HW] Number of HugeTLB pages to allocate at boot.
1688 If this follows hugepagesz (below), it specifies
1689 the number of pages of hugepagesz to be allocated.
1690 If this is the first HugeTLB parameter on the command
1691 line, it specifies the number of pages to allocate for
1692 the default huge page size. If using node format, the
1693 number of pages to allocate per-node can be specified.
1694 See also Documentation/admin-guide/mm/hugetlbpage.rst.
1695 Format: <integer> or (node format)
1696 <node>:<integer>[,<node>:<integer>]
1699 [HW] The size of the HugeTLB pages. This is used in
1700 conjunction with hugepages (above) to allocate huge
1701 pages of a specific size at boot. The pair
1702 hugepagesz=X hugepages=Y can be specified once for
1703 each supported huge page size. Huge page sizes are
1704 architecture dependent. See also
1705 Documentation/admin-guide/mm/hugetlbpage.rst.
1708 hugetlb_cma= [HW,CMA] The size of a CMA area used for allocation
1709 of gigantic hugepages. Or using node format, the size
1710 of a CMA area per node can be specified.
1711 Format: nn[KMGTPE] or (node format)
1712 <node>:nn[KMGTPE][,<node>:nn[KMGTPE]]
1714 Reserve a CMA area of given size and allocate gigantic
1715 hugepages using the CMA allocator. If enabled, the
1716 boot-time allocation of gigantic hugepages is skipped.
1718 hugetlb_free_vmemmap=
1719 [KNL] Reguires CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP
1721 Allows heavy hugetlb users to free up some more
1722 memory (7 * PAGE_SIZE for each 2MB hugetlb page).
1723 Format: { [oO][Nn]/Y/y/1 | [oO][Ff]/N/n/0 (default) }
1725 [oO][Nn]/Y/y/1: enable the feature
1726 [oO][Ff]/N/n/0: disable the feature
1728 Built with CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP_DEFAULT_ON=y,
1731 This is not compatible with memory_hotplug.memmap_on_memory.
1732 If both parameters are enabled, hugetlb_free_vmemmap takes
1733 precedence over memory_hotplug.memmap_on_memory.
1736 [KNL] Should the hung task detector generate panics.
1739 A value of 1 instructs the kernel to panic when a
1740 hung task is detected. The default value is controlled
1741 by the CONFIG_BOOTPARAM_HUNG_TASK_PANIC build-time
1742 option. The value selected by this boot parameter can
1743 be changed later by the kernel.hung_task_panic sysctl.
1745 hvc_iucv= [S390] Number of z/VM IUCV hypervisor console (HVC)
1746 terminal devices. Valid values: 0..8
1747 hvc_iucv_allow= [S390] Comma-separated list of z/VM user IDs.
1748 If specified, z/VM IUCV HVC accepts connections
1749 from listed z/VM user IDs only.
1751 hv_nopvspin [X86,HYPER_V] Disables the paravirt spinlock optimizations
1752 which allow the hypervisor to 'idle' the
1753 guest on lock contention.
1756 Do not unregister boot console at start. This is only
1757 useful for debugging when something happens in the window
1758 between unregistering the boot console and initializing
1761 i2c_bus= [HW] Override the default board specific I2C bus speed
1762 or register an additional I2C bus that is not
1763 registered from board initialization code.
1767 i8042.debug [HW] Toggle i8042 debug mode
1768 i8042.unmask_kbd_data
1769 [HW] Enable printing of interrupt data from the KBD port
1770 (disabled by default, and as a pre-condition
1771 requires that i8042.debug=1 be enabled)
1772 i8042.direct [HW] Put keyboard port into non-translated mode
1773 i8042.dumbkbd [HW] Pretend that controller can only read data from
1774 keyboard and cannot control its state
1775 (Don't attempt to blink the leds)
1776 i8042.noaux [HW] Don't check for auxiliary (== mouse) port
1777 i8042.nokbd [HW] Don't check/create keyboard port
1778 i8042.noloop [HW] Disable the AUX Loopback command while probing
1780 i8042.nomux [HW] Don't check presence of an active multiplexing
1782 i8042.nopnp [HW] Don't use ACPIPnP / PnPBIOS to discover KBD/AUX
1784 i8042.notimeout [HW] Ignore timeout condition signalled by controller
1785 i8042.reset [HW] Reset the controller during init, cleanup and
1786 suspend-to-ram transitions, only during s2r
1787 transitions, or never reset
1788 Format: { 1 | Y | y | 0 | N | n }
1789 1, Y, y: always reset controller
1790 0, N, n: don't ever reset controller
1791 Default: only on s2r transitions on x86; most other
1792 architectures force reset to be always executed
1793 i8042.unlock [HW] Unlock (ignore) the keylock
1794 i8042.kbdreset [HW] Reset device connected to KBD port
1796 [HW] Allow deferred probing upon i8042 probe errors
1800 i915.invert_brightness=
1801 [DRM] Invert the sense of the variable that is used to
1802 set the brightness of the panel backlight. Normally a
1803 brightness value of 0 indicates backlight switched off,
1804 and the maximum of the brightness value sets the backlight
1805 to maximum brightness. If this parameter is set to 0
1806 (default) and the machine requires it, or this parameter
1807 is set to 1, a brightness value of 0 sets the backlight
1808 to maximum brightness, and the maximum of the brightness
1809 value switches the backlight off.
1810 -1 -- never invert brightness
1811 0 -- machine default
1812 1 -- force brightness inversion
1815 Format: <io>[,<membase>[,<icn_id>[,<icn_id2>]]]
1819 Format: idle=poll, idle=halt, idle=nomwait
1820 Poll forces a polling idle loop that can slightly
1821 improve the performance of waking up a idle CPU, but
1822 will use a lot of power and make the system run hot.
1824 idle=halt: Halt is forced to be used for CPU idle.
1825 In such case C2/C3 won't be used again.
1826 idle=nomwait: Disable mwait for CPU C-states
1830 Allow force disabling of Shared Virtual Memory (SVA)
1831 support for the idxd driver. By default it is set to
1834 idxd.tc_override= [HW]
1836 Allow override of default traffic class configuration
1837 for the device. By default it is set to false (0).
1839 ieee754= [MIPS] Select IEEE Std 754 conformance mode
1840 Format: { strict | legacy | 2008 | relaxed }
1843 Choose which programs will be accepted for execution
1844 based on the IEEE 754 NaN encoding(s) supported by
1845 the FPU and the NaN encoding requested with the value
1846 of an ELF file header flag individually set by each
1847 binary. Hardware implementations are permitted to
1848 support either or both of the legacy and the 2008 NaN
1851 Available settings are as follows:
1852 strict accept binaries that request a NaN encoding
1853 supported by the FPU
1854 legacy only accept legacy-NaN binaries, if supported
1856 2008 only accept 2008-NaN binaries, if supported
1858 relaxed accept any binaries regardless of whether
1859 supported by the FPU
1861 The FPU emulator is always able to support both NaN
1862 encodings, so if no FPU hardware is present or it has
1863 been disabled with 'nofpu', then the settings of
1864 'legacy' and '2008' strap the emulator accordingly,
1865 'relaxed' straps the emulator for both legacy-NaN and
1866 2008-NaN, whereas 'strict' enables legacy-NaN only on
1867 legacy processors and both NaN encodings on MIPS32 or
1870 The setting for ABS.fmt/NEG.fmt instruction execution
1871 mode generally follows that for the NaN encoding,
1872 except where unsupported by hardware.
1874 ignore_loglevel [KNL]
1875 Ignore loglevel setting - this will print /all/
1876 kernel messages to the console. Useful for debugging.
1877 We also add it as printk module parameter, so users
1878 could change it dynamically, usually by
1879 /sys/module/printk/parameters/ignore_loglevel.
1882 Ignore RLIMIT_DATA setting for data mappings,
1883 print warning at first misuse. Can be changed via
1884 /sys/module/kernel/parameters/ignore_rlimit_data.
1886 ihash_entries= [KNL]
1887 Set number of hash buckets for inode cache.
1889 ima_appraise= [IMA] appraise integrity measurements
1890 Format: { "off" | "enforce" | "fix" | "log" }
1893 ima_appraise_tcb [IMA] Deprecated. Use ima_policy= instead.
1894 The builtin appraise policy appraises all files
1897 ima_canonical_fmt [IMA]
1898 Use the canonical format for the binary runtime
1899 measurements, instead of host native format.
1902 Format: { md5 | sha1 | rmd160 | sha256 | sha384
1906 The list of supported hash algorithms is defined
1907 in crypto/hash_info.h.
1910 The builtin policies to load during IMA setup.
1911 Format: "tcb | appraise_tcb | secure_boot |
1912 fail_securely | critical_data"
1914 The "tcb" policy measures all programs exec'd, files
1915 mmap'd for exec, and all files opened with the read
1916 mode bit set by either the effective uid (euid=0) or
1919 The "appraise_tcb" policy appraises the integrity of
1920 all files owned by root.
1922 The "secure_boot" policy appraises the integrity
1923 of files (eg. kexec kernel image, kernel modules,
1924 firmware, policy, etc) based on file signatures.
1926 The "fail_securely" policy forces file signature
1927 verification failure also on privileged mounted
1928 filesystems with the SB_I_UNVERIFIABLE_SIGNATURE
1931 The "critical_data" policy measures kernel integrity
1934 ima_tcb [IMA] Deprecated. Use ima_policy= instead.
1935 Load a policy which meets the needs of the Trusted
1936 Computing Base. This means IMA will measure all
1937 programs exec'd, files mmap'd for exec, and all files
1938 opened for read by uid=0.
1941 Select one of defined IMA measurements template formats.
1942 Formats: { "ima" | "ima-ng" | "ima-ngv2" | "ima-sig" |
1947 [IMA] Define a custom template format.
1948 Format: { "field1|...|fieldN" }
1950 ima.ahash_minsize= [IMA] Minimum file size for asynchronous hash usage
1951 Format: <min_file_size>
1952 Set the minimal file size for using asynchronous hash.
1953 If left unspecified, ahash usage is disabled.
1955 ahash performance varies for different data sizes on
1956 different crypto accelerators. This option can be used
1957 to achieve the best performance for a particular HW.
1959 ima.ahash_bufsize= [IMA] Asynchronous hash buffer size
1961 Set hashing buffer size. Default: 4k.
1963 ahash performance varies for different chunk sizes on
1964 different crypto accelerators. This option can be used
1965 to achieve best performance for particular HW.
1969 Run specified binary instead of /sbin/init as init
1972 initcall_debug [KNL] Trace initcalls as they are executed. Useful
1973 for working out where the kernel is dying during
1976 initcall_blacklist= [KNL] Do not execute a comma-separated list of
1977 initcall functions. Useful for debugging built-in
1978 modules and initcalls.
1980 initramfs_async= [KNL]
1983 This parameter controls whether the initramfs
1984 image is unpacked asynchronously, concurrently
1985 with devices being probed and
1986 initialized. This should normally just work,
1987 but as a debugging aid, one can get the
1988 historical behaviour of the initramfs
1989 unpacking being completed before device_ and
1992 initrd= [BOOT] Specify the location of the initial ramdisk
1994 initrdmem= [KNL] Specify a physical address and size from which to
1995 load the initrd. If an initrd is compiled in or
1996 specified in the bootparams, it takes priority over this
1998 Format: ss[KMG],nn[KMG]
2001 init_on_alloc= [MM] Fill newly allocated pages and heap objects with
2004 Default set by CONFIG_INIT_ON_ALLOC_DEFAULT_ON.
2006 init_on_free= [MM] Fill freed pages and heap objects with zeroes.
2008 Default set by CONFIG_INIT_ON_FREE_DEFAULT_ON.
2010 init_pkru= [X86] Specify the default memory protection keys rights
2011 register contents for all processes. 0x55555554 by
2012 default (disallow access to all but pkey 0). Can
2013 override in debugfs after boot.
2015 inport.irq= [HW] Inport (ATI XL and Microsoft) busmouse driver
2018 int_pln_enable [X86] Enable power limit notification interrupt
2020 integrity_audit=[IMA]
2021 Format: { "0" | "1" }
2022 0 -- basic integrity auditing messages. (Default)
2023 1 -- additional integrity auditing messages.
2025 intel_iommu= [DMAR] Intel IOMMU driver (DMAR) option
2027 Enable intel iommu driver.
2029 Disable intel iommu driver.
2030 igfx_off [Default Off]
2031 By default, gfx is mapped as normal device. If a gfx
2032 device has a dedicated DMAR unit, the DMAR unit is
2033 bypassed by not enabling DMAR with this option. In
2034 this case, gfx device will use physical address for
2036 strict [Default Off]
2037 Deprecated, equivalent to iommu.strict=1.
2038 sp_off [Default Off]
2039 By default, super page will be supported if Intel IOMMU
2040 has the capability. With this option, super page will
2043 Enable the Intel IOMMU scalable mode if the hardware
2044 advertises that it has support for the scalable mode
2047 Disallow use of the Intel IOMMU scalable mode.
2048 tboot_noforce [Default Off]
2049 Do not force the Intel IOMMU enabled under tboot.
2050 By default, tboot will force Intel IOMMU on, which
2051 could harm performance of some high-throughput
2052 devices like 40GBit network cards, even if identity
2054 Note that using this option lowers the security
2055 provided by tboot because it makes the system
2056 vulnerable to DMA attacks.
2058 intel_idle.max_cstate= [KNL,HW,ACPI,X86]
2059 0 disables intel_idle and fall back on acpi_idle.
2060 1 to 9 specify maximum depth of C-state.
2064 Do not enable intel_pstate as the default
2065 scaling driver for the supported processors
2067 Use intel_pstate as a scaling driver, but configure it
2068 to work with generic cpufreq governors (instead of
2069 enabling its internal governor). This mode cannot be
2070 used along with the hardware-managed P-states (HWP)
2073 Enable intel_pstate on systems that prohibit it by default
2074 in favor of acpi-cpufreq. Forcing the intel_pstate driver
2075 instead of acpi-cpufreq may disable platform features, such
2076 as thermal controls and power capping, that rely on ACPI
2077 P-States information being indicated to OSPM and therefore
2078 should be used with caution. This option does not work with
2079 processors that aren't supported by the intel_pstate driver
2080 or on platforms that use pcc-cpufreq instead of acpi-cpufreq.
2082 Do not enable hardware P state control (HWP)
2085 Only load intel_pstate on systems which support
2086 hardware P state control (HWP) if available.
2088 Enforce ACPI _PPC performance limits. If the Fixed ACPI
2089 Description Table, specifies preferred power management
2090 profile as "Enterprise Server" or "Performance Server",
2091 then this feature is turned on by default.
2093 Allow per-logical-CPU P-State performance control limits using
2094 cpufreq sysfs interface
2096 intremap= [X86-64, Intel-IOMMU]
2097 on enable Interrupt Remapping (default)
2098 off disable Interrupt Remapping
2099 nosid disable Source ID checking
2101 BIOS x2APIC opt-out request will be ignored
2102 nopost disable Interrupt Posting
2104 iomem= Disable strict checking of access to MMIO memory
2105 strict regions from userspace.
2120 nobypass [PPC/POWERNV]
2121 Disable IOMMU bypass, using IOMMU for PCI devices.
2123 iommu.forcedac= [ARM64, X86] Control IOVA allocation for PCI devices.
2124 Format: { "0" | "1" }
2125 0 - Try to allocate a 32-bit DMA address first, before
2126 falling back to the full range if needed.
2127 1 - Allocate directly from the full usable range,
2128 forcing Dual Address Cycle for PCI cards supporting
2129 greater than 32-bit addressing.
2131 iommu.strict= [ARM64, X86] Configure TLB invalidation behaviour
2132 Format: { "0" | "1" }
2134 Request that DMA unmap operations use deferred
2135 invalidation of hardware TLBs, for increased
2136 throughput at the cost of reduced device isolation.
2137 Will fall back to strict mode if not supported by
2138 the relevant IOMMU driver.
2140 DMA unmap operations invalidate IOMMU hardware TLBs
2142 unset - Use value of CONFIG_IOMMU_DEFAULT_DMA_{LAZY,STRICT}.
2143 Note: on x86, strict mode specified via one of the
2144 legacy driver-specific options takes precedence.
2147 [ARM64, X86] Configure DMA to bypass the IOMMU by default.
2148 Format: { "0" | "1" }
2149 0 - Use IOMMU translation for DMA.
2150 1 - Bypass the IOMMU for DMA.
2151 unset - Use value of CONFIG_IOMMU_DEFAULT_PASSTHROUGH.
2153 io7= [HW] IO7 for Marvel-based Alpha systems
2154 See comment before marvel_specify_io7 in
2155 arch/alpha/kernel/core_marvel.c.
2157 io_delay= [X86] I/O delay method
2159 Standard port 0x80 based delay
2161 Alternate port 0xed based delay (needed on some systems)
2163 Simple two microseconds delay
2168 See Documentation/admin-guide/nfs/nfsroot.rst.
2170 ipcmni_extend [KNL] Extend the maximum number of unique System V
2171 IPC identifiers from 32,768 to 16,777,216.
2173 irqaffinity= [SMP] Set the default irq affinity mask
2174 The argument is a cpu list, as described above.
2176 irqchip.gicv2_force_probe=
2179 Force the kernel to look for the second 4kB page
2180 of a GICv2 controller even if the memory range
2181 exposed by the device tree is too small.
2183 irqchip.gicv3_nolpi=
2185 Force the kernel to ignore the availability of
2186 LPIs (and by consequence ITSs). Intended for system
2187 that use the kernel as a bootloader, and thus want
2188 to let secondary kernels in charge of setting up
2191 irqchip.gicv3_pseudo_nmi= [ARM64]
2192 Enables support for pseudo-NMIs in the kernel. This
2193 requires the kernel to be built with
2194 CONFIG_ARM64_PSEUDO_NMI.
2197 When an interrupt is not handled search all handlers
2198 for it. Intended to get systems with badly broken
2202 When an interrupt is not handled search all handlers
2203 for it. Also check all handlers each timer
2204 interrupt. Intended to get systems with badly broken
2208 Format: <RDP>,<reset>,<pci_scan>,<verbosity>
2210 isolcpus= [KNL,SMP,ISOL] Isolate a given set of CPUs from disturbance.
2211 [Deprecated - use cpusets instead]
2212 Format: [flag-list,]<cpu-list>
2214 Specify one or more CPUs to isolate from disturbances
2215 specified in the flag list (default: domain):
2218 Disable the tick when a single task runs.
2220 A residual 1Hz tick is offloaded to workqueues, which you
2221 need to affine to housekeeping through the global
2222 workqueue's affinity configured via the
2223 /sys/devices/virtual/workqueue/cpumask sysfs file, or
2224 by using the 'domain' flag described below.
2226 NOTE: by default the global workqueue runs on all CPUs,
2227 so to protect individual CPUs the 'cpumask' file has to
2228 be configured manually after bootup.
2231 Isolate from the general SMP balancing and scheduling
2232 algorithms. Note that performing domain isolation this way
2233 is irreversible: it's not possible to bring back a CPU to
2234 the domains once isolated through isolcpus. It's strongly
2235 advised to use cpusets instead to disable scheduler load
2236 balancing through the "cpuset.sched_load_balance" file.
2237 It offers a much more flexible interface where CPUs can
2238 move in and out of an isolated set anytime.
2240 You can move a process onto or off an "isolated" CPU via
2241 the CPU affinity syscalls or cpuset.
2242 <cpu number> begins at 0 and the maximum value is
2243 "number of CPUs in system - 1".
2247 Isolate from being targeted by managed interrupts
2248 which have an interrupt mask containing isolated
2249 CPUs. The affinity of managed interrupts is
2250 handled by the kernel and cannot be changed via
2251 the /proc/irq/* interfaces.
2253 This isolation is best effort and only effective
2254 if the automatically assigned interrupt mask of a
2255 device queue contains isolated and housekeeping
2256 CPUs. If housekeeping CPUs are online then such
2257 interrupts are directed to the housekeeping CPU
2258 so that IO submitted on the housekeeping CPU
2259 cannot disturb the isolated CPU.
2261 If a queue's affinity mask contains only isolated
2262 CPUs then this parameter has no effect on the
2263 interrupt routing decision, though interrupts are
2264 only delivered when tasks running on those
2265 isolated CPUs submit IO. IO submitted on
2266 housekeeping CPUs has no influence on those
2269 The format of <cpu-list> is described above.
2273 ivrs_ioapic [HW,X86-64]
2274 Provide an override to the IOAPIC-ID<->DEVICE-ID
2275 mapping provided in the IVRS ACPI table. For
2276 example, to map IOAPIC-ID decimal 10 to
2277 PCI device 00:14.0 write the parameter as:
2278 ivrs_ioapic[10]=00:14.0
2280 ivrs_hpet [HW,X86-64]
2281 Provide an override to the HPET-ID<->DEVICE-ID
2282 mapping provided in the IVRS ACPI table. For
2283 example, to map HPET-ID decimal 0 to
2284 PCI device 00:14.0 write the parameter as:
2285 ivrs_hpet[0]=00:14.0
2287 ivrs_acpihid [HW,X86-64]
2288 Provide an override to the ACPI-HID:UID<->DEVICE-ID
2289 mapping provided in the IVRS ACPI table. For
2290 example, to map UART-HID:UID AMD0020:0 to
2291 PCI device 00:14.5 write the parameter as:
2292 ivrs_acpihid[00:14.5]=AMD0020:0
2294 js= [HW,JOY] Analog joystick
2295 See Documentation/input/joydev/joystick.rst.
2298 When CONFIG_RANDOMIZE_BASE is set, this disables
2299 kernel and module base offset ASLR (Address Space
2300 Layout Randomization).
2303 [KNL] Enforce KASAN (Kernel Address Sanitizer) to print
2304 report on every invalid memory access. Without this
2305 parameter KASAN will print report only for the first
2310 kernelcore= [KNL,X86,IA-64,PPC]
2311 Format: nn[KMGTPE] | nn% | "mirror"
2312 This parameter specifies the amount of memory usable by
2313 the kernel for non-movable allocations. The requested
2314 amount is spread evenly throughout all nodes in the
2315 system as ZONE_NORMAL. The remaining memory is used for
2316 movable memory in its own zone, ZONE_MOVABLE. In the
2317 event, a node is too small to have both ZONE_NORMAL and
2318 ZONE_MOVABLE, kernelcore memory will take priority and
2319 other nodes will have a larger ZONE_MOVABLE.
2321 ZONE_MOVABLE is used for the allocation of pages that
2322 may be reclaimed or moved by the page migration
2323 subsystem. Note that allocations like PTEs-from-HighMem
2324 still use the HighMem zone if it exists, and the Normal
2325 zone if it does not.
2327 It is possible to specify the exact amount of memory in
2328 the form of "nn[KMGTPE]", a percentage of total system
2329 memory in the form of "nn%", or "mirror". If "mirror"
2330 option is specified, mirrored (reliable) memory is used
2331 for non-movable allocations and remaining memory is used
2332 for Movable pages. "nn[KMGTPE]", "nn%", and "mirror"
2333 are exclusive, so you cannot specify multiple forms.
2335 kgdbdbgp= [KGDB,HW] kgdb over EHCI usb debug port.
2336 Format: <Controller#>[,poll interval]
2337 The controller # is the number of the ehci usb debug
2338 port as it is probed via PCI. The poll interval is
2339 optional and is the number seconds in between
2340 each poll cycle to the debug port in case you need
2341 the functionality for interrupting the kernel with
2342 gdb or control-c on the dbgp connection. When
2343 not using this parameter you use sysrq-g to break into
2344 the kernel debugger.
2346 kgdboc= [KGDB,HW] kgdb over consoles.
2347 Requires a tty driver that supports console polling,
2348 or a supported polling keyboard driver (non-usb).
2349 Serial only format: <serial_device>[,baud]
2350 keyboard only format: kbd
2351 keyboard and serial format: kbd,<serial_device>[,baud]
2352 Optional Kernel mode setting:
2353 kms, kbd format: kms,kbd
2354 kms, kbd and serial format: kms,kbd,<ser_dev>[,baud]
2356 kgdboc_earlycon= [KGDB,HW]
2357 If the boot console provides the ability to read
2358 characters and can work in polling mode, you can use
2359 this parameter to tell kgdb to use it as a backend
2360 until the normal console is registered. Intended to
2361 be used together with the kgdboc parameter which
2362 specifies the normal console to transition to.
2364 The name of the early console should be specified
2365 as the value of this parameter. Note that the name of
2366 the early console might be different than the tty
2367 name passed to kgdboc. It's OK to leave the value
2368 blank and the first boot console that implements
2369 read() will be picked.
2371 kgdbwait [KGDB] Stop kernel execution and enter the
2372 kernel debugger at the earliest opportunity.
2374 kmac= [MIPS] Korina ethernet MAC address.
2375 Configure the RouterBoard 532 series on-chip
2376 Ethernet adapter MAC address.
2378 kmemleak= [KNL] Boot-time kmemleak enable/disable
2379 Valid arguments: on, off
2381 Built with CONFIG_DEBUG_KMEMLEAK_DEFAULT_OFF=y,
2384 kprobe_event=[probe-list]
2385 [FTRACE] Add kprobe events and enable at boot time.
2386 The probe-list is a semicolon delimited list of probe
2387 definitions. Each definition is same as kprobe_events
2388 interface, but the parameters are comma delimited.
2389 For example, to add a kprobe event on vfs_read with
2390 arg1 and arg2, add to the command line;
2392 kprobe_event=p,vfs_read,$arg1,$arg2
2394 See also Documentation/trace/kprobetrace.rst "Kernel
2395 Boot Parameter" section.
2397 kpti= [ARM64] Control page table isolation of user
2398 and kernel address spaces.
2399 Default: enabled on cores which need mitigation.
2403 kvm.ignore_msrs=[KVM] Ignore guest accesses to unhandled MSRs.
2404 Default is 0 (don't ignore, but inject #GP)
2406 kvm.eager_page_split=
2407 [KVM,X86] Controls whether or not KVM will try to
2408 proactively split all huge pages during dirty logging.
2409 Eager page splitting reduces interruptions to vCPU
2410 execution by eliminating the write-protection faults
2411 and MMU lock contention that would otherwise be
2412 required to split huge pages lazily.
2414 VM workloads that rarely perform writes or that write
2415 only to a small region of VM memory may benefit from
2416 disabling eager page splitting to allow huge pages to
2417 still be used for reads.
2419 The behavior of eager page splitting depends on whether
2420 KVM_DIRTY_LOG_INITIALLY_SET is enabled or disabled. If
2421 disabled, all huge pages in a memslot will be eagerly
2422 split when dirty logging is enabled on that memslot. If
2423 enabled, eager page splitting will be performed during
2424 the KVM_CLEAR_DIRTY ioctl, and only for the pages being
2427 Eager page splitting currently only supports splitting
2428 huge pages mapped by the TDP MMU.
2432 kvm.enable_vmware_backdoor=[KVM] Support VMware backdoor PV interface.
2433 Default is false (don't support).
2436 [KVM] Controls the software workaround for the
2437 X86_BUG_ITLB_MULTIHIT bug.
2438 force : Always deploy workaround.
2439 off : Never deploy workaround.
2440 auto : Deploy workaround based on the presence of
2441 X86_BUG_ITLB_MULTIHIT.
2445 If the software workaround is enabled for the host,
2446 guests do need not to enable it for nested guests.
2448 kvm.nx_huge_pages_recovery_ratio=
2449 [KVM] Controls how many 4KiB pages are periodically zapped
2450 back to huge pages. 0 disables the recovery, otherwise if
2451 the value is N KVM will zap 1/Nth of the 4KiB pages every
2452 period (see below). The default is 60.
2454 kvm.nx_huge_pages_recovery_period_ms=
2455 [KVM] Controls the time period at which KVM zaps 4KiB pages
2456 back to huge pages. If the value is a non-zero N, KVM will
2457 zap a portion (see ratio above) of the pages every N msecs.
2458 If the value is 0 (the default), KVM will pick a period based
2459 on the ratio, such that a page is zapped after 1 hour on average.
2461 kvm-amd.nested= [KVM,AMD] Allow nested virtualization in KVM/SVM.
2462 Default is 1 (enabled)
2464 kvm-amd.npt= [KVM,AMD] Disable nested paging (virtualized MMU)
2466 Default is 1 (enabled) if in 64-bit or 32-bit PAE mode.
2469 [KVM,ARM] Select one of KVM/arm64's modes of operation.
2471 none: Forcefully disable KVM.
2473 nvhe: Standard nVHE-based mode, without support for
2476 protected: nVHE-based mode with support for guests whose
2477 state is kept private from the host.
2479 Defaults to VHE/nVHE based on hardware support. Setting
2480 mode to "protected" will disable kexec and hibernation
2483 kvm-arm.vgic_v3_group0_trap=
2484 [KVM,ARM] Trap guest accesses to GICv3 group-0
2487 kvm-arm.vgic_v3_group1_trap=
2488 [KVM,ARM] Trap guest accesses to GICv3 group-1
2491 kvm-arm.vgic_v3_common_trap=
2492 [KVM,ARM] Trap guest accesses to GICv3 common
2495 kvm-arm.vgic_v4_enable=
2496 [KVM,ARM] Allow use of GICv4 for direct injection of
2499 kvm_cma_resv_ratio=n [PPC]
2500 Reserves given percentage from system memory area for
2501 contiguous memory allocation for KVM hash pagetable
2503 By default it reserves 5% of total system memory.
2507 kvm-intel.ept= [KVM,Intel] Disable extended page tables
2508 (virtualized MMU) support on capable Intel chips.
2509 Default is 1 (enabled)
2511 kvm-intel.emulate_invalid_guest_state=
2512 [KVM,Intel] Disable emulation of invalid guest state.
2513 Ignored if kvm-intel.enable_unrestricted_guest=1, as
2514 guest state is never invalid for unrestricted guests.
2515 This param doesn't apply to nested guests (L2), as KVM
2516 never emulates invalid L2 guest state.
2517 Default is 1 (enabled)
2519 kvm-intel.flexpriority=
2520 [KVM,Intel] Disable FlexPriority feature (TPR shadow).
2521 Default is 1 (enabled)
2524 [KVM,Intel] Enable VMX nesting (nVMX).
2525 Default is 0 (disabled)
2527 kvm-intel.unrestricted_guest=
2528 [KVM,Intel] Disable unrestricted guest feature
2529 (virtualized real and unpaged mode) on capable
2530 Intel chips. Default is 1 (enabled)
2532 kvm-intel.vmentry_l1d_flush=[KVM,Intel] Mitigation for L1 Terminal Fault
2535 Valid arguments: never, cond, always
2537 always: L1D cache flush on every VMENTER.
2538 cond: Flush L1D on VMENTER only when the code between
2539 VMEXIT and VMENTER can leak host memory.
2540 never: Disables the mitigation
2542 Default is cond (do L1 cache flush in specific instances)
2544 kvm-intel.vpid= [KVM,Intel] Disable Virtual Processor Identification
2545 feature (tagged TLBs) on capable Intel chips.
2546 Default is 1 (enabled)
2548 l1d_flush= [X86,INTEL]
2549 Control mitigation for L1D based snooping vulnerability.
2551 Certain CPUs are vulnerable to an exploit against CPU
2552 internal buffers which can forward information to a
2553 disclosure gadget under certain conditions.
2555 In vulnerable processors, the speculatively
2556 forwarded data can be used in a cache side channel
2557 attack, to access data to which the attacker does
2558 not have direct access.
2560 This parameter controls the mitigation. The
2563 on - enable the interface for the mitigation
2565 l1tf= [X86] Control mitigation of the L1TF vulnerability on
2568 The kernel PTE inversion protection is unconditionally
2569 enabled and cannot be disabled.
2572 Provides all available mitigations for the
2573 L1TF vulnerability. Disables SMT and
2574 enables all mitigations in the
2575 hypervisors, i.e. unconditional L1D flush.
2577 SMT control and L1D flush control via the
2578 sysfs interface is still possible after
2579 boot. Hypervisors will issue a warning
2580 when the first VM is started in a
2581 potentially insecure configuration,
2582 i.e. SMT enabled or L1D flush disabled.
2585 Same as 'full', but disables SMT and L1D
2586 flush runtime control. Implies the
2587 'nosmt=force' command line option.
2588 (i.e. sysfs control of SMT is disabled.)
2591 Leaves SMT enabled and enables the default
2592 hypervisor mitigation, i.e. conditional
2595 SMT control and L1D flush control via the
2596 sysfs interface is still possible after
2597 boot. Hypervisors will issue a warning
2598 when the first VM is started in a
2599 potentially insecure configuration,
2600 i.e. SMT enabled or L1D flush disabled.
2604 Disables SMT and enables the default
2605 hypervisor mitigation.
2607 SMT control and L1D flush control via the
2608 sysfs interface is still possible after
2609 boot. Hypervisors will issue a warning
2610 when the first VM is started in a
2611 potentially insecure configuration,
2612 i.e. SMT enabled or L1D flush disabled.
2615 Same as 'flush', but hypervisors will not
2616 warn when a VM is started in a potentially
2617 insecure configuration.
2620 Disables hypervisor mitigations and doesn't
2622 It also drops the swap size and available
2623 RAM limit restriction on both hypervisor and
2628 For details see: Documentation/admin-guide/hw-vuln/l1tf.rst
2634 lapic [X86-32,APIC] Enable the local APIC even if BIOS
2637 lapic= [X86,APIC] Do not use TSC deadline
2638 value for LAPIC timer one-shot implementation. Default
2639 back to the programmable timer unit in the LAPIC.
2640 Format: notscdeadline
2642 lapic_timer_c2_ok [X86,APIC] trust the local apic timer
2645 libata.dma= [LIBATA] DMA control
2646 libata.dma=0 Disable all PATA and SATA DMA
2647 libata.dma=1 PATA and SATA Disk DMA only
2648 libata.dma=2 ATAPI (CDROM) DMA only
2649 libata.dma=4 Compact Flash DMA only
2650 Combinations also work, so libata.dma=3 enables DMA
2651 for disks and CDROMs, but not CFs.
2653 libata.ignore_hpa= [LIBATA] Ignore HPA limit
2654 libata.ignore_hpa=0 keep BIOS limits (default)
2655 libata.ignore_hpa=1 ignore limits, using full disk
2657 libata.noacpi [LIBATA] Disables use of ACPI in libata suspend/resume
2661 libata.force= [LIBATA] Force configurations. The format is a comma-
2662 separated list of "[ID:]VAL" where ID is PORT[.DEVICE].
2663 PORT and DEVICE are decimal numbers matching port, link
2664 or device. Basically, it matches the ATA ID string
2665 printed on console by libata. If the whole ID part is
2666 omitted, the last PORT and DEVICE values are used. If
2667 ID hasn't been specified yet, the configuration applies
2668 to all ports, links and devices.
2670 If only DEVICE is omitted, the parameter applies to
2671 the port and all links and devices behind it. DEVICE
2672 number of 0 either selects the first device or the
2673 first fan-out link behind PMP device. It does not
2674 select the host link. DEVICE number of 15 selects the
2675 host link and device attached to it.
2677 The VAL specifies the configuration to force. As long
2678 as there is no ambiguity, shortcut notation is allowed.
2679 For example, both 1.5 and 1.5G would work for 1.5Gbps.
2680 The following configurations can be forced.
2682 * Cable type: 40c, 80c, short40c, unk, ign or sata.
2683 Any ID with matching PORT is used.
2685 * SATA link speed limit: 1.5Gbps or 3.0Gbps.
2687 * Transfer mode: pio[0-7], mwdma[0-4] and udma[0-7].
2688 udma[/][16,25,33,44,66,100,133] notation is also
2691 * nohrst, nosrst, norst: suppress hard, soft and both
2694 * rstonce: only attempt one reset during hot-unplug
2697 * [no]dbdelay: Enable or disable the extra 200ms delay
2698 before debouncing a link PHY and device presence
2701 * [no]ncq: Turn on or off NCQ.
2703 * [no]ncqtrim: Enable or disable queued DSM TRIM.
2705 * [no]ncqati: Enable or disable NCQ trim on ATI chipset.
2707 * [no]trim: Enable or disable (unqueued) TRIM.
2709 * trim_zero: Indicate that TRIM command zeroes data.
2711 * max_trim_128m: Set 128M maximum trim size limit.
2713 * [no]dma: Turn on or off DMA transfers.
2715 * atapi_dmadir: Enable ATAPI DMADIR bridge support.
2717 * atapi_mod16_dma: Enable the use of ATAPI DMA for
2718 commands that are not a multiple of 16 bytes.
2720 * [no]dmalog: Enable or disable the use of the
2721 READ LOG DMA EXT command to access logs.
2723 * [no]iddevlog: Enable or disable access to the
2724 identify device data log.
2726 * [no]logdir: Enable or disable access to the general
2727 purpose log directory.
2729 * max_sec_128: Set transfer size limit to 128 sectors.
2731 * max_sec_1024: Set or clear transfer size limit to
2734 * max_sec_lba48: Set or clear transfer size limit to
2737 * [no]lpm: Enable or disable link power management.
2739 * [no]setxfer: Indicate if transfer speed mode setting
2742 * dump_id: Dump IDENTIFY data.
2744 * disable: Disable this device.
2746 If there are multiple matching configurations changing
2747 the same attribute, the last one is used.
2749 load_ramdisk= [RAM] [Deprecated]
2751 lockd.nlm_grace_period=P [NFS] Assign grace period.
2754 lockd.nlm_tcpport=N [NFS] Assign TCP port.
2757 lockd.nlm_timeout=T [NFS] Assign timeout value.
2760 lockd.nlm_udpport=M [NFS] Assign UDP port.
2763 lockdown= [SECURITY]
2764 { integrity | confidentiality }
2765 Enable the kernel lockdown feature. If set to
2766 integrity, kernel features that allow userland to
2767 modify the running kernel are disabled. If set to
2768 confidentiality, kernel features that allow userland
2769 to extract confidential information from the kernel
2772 locktorture.nreaders_stress= [KNL]
2773 Set the number of locking read-acquisition kthreads.
2774 Defaults to being automatically set based on the
2775 number of online CPUs.
2777 locktorture.nwriters_stress= [KNL]
2778 Set the number of locking write-acquisition kthreads.
2780 locktorture.onoff_holdoff= [KNL]
2781 Set time (s) after boot for CPU-hotplug testing.
2783 locktorture.onoff_interval= [KNL]
2784 Set time (s) between CPU-hotplug operations, or
2785 zero to disable CPU-hotplug testing.
2787 locktorture.shuffle_interval= [KNL]
2788 Set task-shuffle interval (jiffies). Shuffling
2789 tasks allows some CPUs to go into dyntick-idle
2790 mode during the locktorture test.
2792 locktorture.shutdown_secs= [KNL]
2793 Set time (s) after boot system shutdown. This
2794 is useful for hands-off automated testing.
2796 locktorture.stat_interval= [KNL]
2797 Time (s) between statistics printk()s.
2799 locktorture.stutter= [KNL]
2800 Time (s) to stutter testing, for example,
2801 specifying five seconds causes the test to run for
2802 five seconds, wait for five seconds, and so on.
2803 This tests the locking primitive's ability to
2804 transition abruptly to and from idle.
2806 locktorture.torture_type= [KNL]
2807 Specify the locking implementation to test.
2809 locktorture.verbose= [KNL]
2810 Enable additional printk() statements.
2812 logibm.irq= [HW,MOUSE] Logitech Bus Mouse Driver
2815 loglevel= All Kernel Messages with a loglevel smaller than the
2816 console loglevel will be printed to the console. It can
2817 also be changed with klogd or other programs. The
2818 loglevels are defined as follows:
2820 0 (KERN_EMERG) system is unusable
2821 1 (KERN_ALERT) action must be taken immediately
2822 2 (KERN_CRIT) critical conditions
2823 3 (KERN_ERR) error conditions
2824 4 (KERN_WARNING) warning conditions
2825 5 (KERN_NOTICE) normal but significant condition
2826 6 (KERN_INFO) informational
2827 7 (KERN_DEBUG) debug-level messages
2829 log_buf_len=n[KMG] Sets the size of the printk ring buffer,
2830 in bytes. n must be a power of two and greater
2831 than the minimal size. The minimal size is defined
2832 by LOG_BUF_SHIFT kernel config parameter. There is
2833 also CONFIG_LOG_CPU_MAX_BUF_SHIFT config parameter
2834 that allows to increase the default size depending on
2835 the number of CPUs. See init/Kconfig for more details.
2837 logo.nologo [FB] Disables display of the built-in Linux logo.
2838 This may be used to provide more screen space for
2839 kernel log messages and is useful when debugging
2840 kernel boot problems.
2842 lp=0 [LP] Specify parallel ports to use, e.g,
2843 lp=port[,port...] lp=none,parport0 (lp0 not configured, lp1 uses
2844 lp=reset first parallel port). 'lp=0' disables the
2845 lp=auto printer driver. 'lp=reset' (which can be
2846 specified in addition to the ports) causes
2847 attached printers to be reset. Using
2848 lp=port1,port2,... specifies the parallel ports
2849 to associate lp devices with, starting with
2850 lp0. A port specification may be 'none' to skip
2851 that lp device, or a parport name such as
2852 'parport0'. Specifying 'lp=auto' instead of a
2853 port specification list means that device IDs
2854 from each port should be examined, to see if
2855 an IEEE 1284-compliant printer is attached; if
2856 so, the driver will manage that printer.
2857 See also header of drivers/char/lp.c.
2860 Sets loops_per_jiffy to given constant, thus avoiding
2861 time-consuming boot-time autodetection (up to 250 ms per
2862 CPU). 0 enables autodetection (default). To determine
2863 the correct value for your kernel, boot with normal
2864 autodetection and see what value is printed. Note that
2865 on SMP systems the preset will be applied to all CPUs,
2866 which is likely to cause problems if your CPUs need
2867 significantly divergent settings. An incorrect value
2868 will cause delays in the kernel to be wrong, leading to
2869 unpredictable I/O errors and other breakage. Although
2870 unlikely, in the extreme case this might damage your
2874 Format: <io>,<irq>,<dma>
2876 lsm.debug [SECURITY] Enable LSM initialization debugging output.
2879 [SECURITY] Choose order of LSM initialization. This
2880 overrides CONFIG_LSM, and the "security=" parameter.
2882 machvec= [IA-64] Force the use of a particular machine-vector
2883 (machvec) in a generic kernel.
2884 Example: machvec=hpzx1
2886 machtype= [Loongson] Share the same kernel image file between
2887 different yeeloong laptops.
2888 Example: machtype=lemote-yeeloong-2f-7inch
2890 max_addr=nn[KMG] [KNL,BOOT,IA-64] All physical memory greater
2891 than or equal to this physical address is ignored.
2893 maxcpus= [SMP] Maximum number of processors that an SMP kernel
2894 will bring up during bootup. maxcpus=n : n >= 0 limits
2895 the kernel to bring up 'n' processors. Surely after
2896 bootup you can bring up the other plugged cpu by executing
2897 "echo 1 > /sys/devices/system/cpu/cpuX/online". So maxcpus
2898 only takes effect during system bootup.
2899 While n=0 is a special case, it is equivalent to "nosmp",
2900 which also disables the IO APIC.
2902 max_loop= [LOOP] The number of loop block devices that get
2903 (loop.max_loop) unconditionally pre-created at init time. The default
2904 number is configured by BLK_DEV_LOOP_MIN_COUNT. Instead
2905 of statically allocating a predefined number, loop
2906 devices can be requested on-demand with the
2907 /dev/loop-control interface.
2909 mce [X86-32] Machine Check Exception
2911 mce=option [X86-64] See Documentation/x86/x86_64/boot-options.rst
2913 md= [HW] RAID subsystems devices and level
2914 See Documentation/admin-guide/md.rst.
2917 Format: <first>,<last>
2918 Specifies range of consoles to be captured by the MDA.
2921 Control mitigation for the Micro-architectural Data
2922 Sampling (MDS) vulnerability.
2924 Certain CPUs are vulnerable to an exploit against CPU
2925 internal buffers which can forward information to a
2926 disclosure gadget under certain conditions.
2928 In vulnerable processors, the speculatively
2929 forwarded data can be used in a cache side channel
2930 attack, to access data to which the attacker does
2931 not have direct access.
2933 This parameter controls the MDS mitigation. The
2936 full - Enable MDS mitigation on vulnerable CPUs
2937 full,nosmt - Enable MDS mitigation and disable
2938 SMT on vulnerable CPUs
2939 off - Unconditionally disable MDS mitigation
2941 On TAA-affected machines, mds=off can be prevented by
2942 an active TAA mitigation as both vulnerabilities are
2943 mitigated with the same mechanism so in order to disable
2944 this mitigation, you need to specify tsx_async_abort=off
2947 Not specifying this option is equivalent to
2950 For details see: Documentation/admin-guide/hw-vuln/mds.rst
2952 mem=nn[KMG] [HEXAGON] Set the memory size.
2953 Must be specified, otherwise memory size will be 0.
2955 mem=nn[KMG] [KNL,BOOT] Force usage of a specific amount of memory
2956 Amount of memory to be used in cases as follows:
2959 2 when the kernel is not able to see the whole system memory;
2960 3 memory that lies after 'mem=' boundary is excluded from
2961 the hypervisor, then assigned to KVM guests.
2962 4 to limit the memory available for kdump kernel.
2964 [ARC,MICROBLAZE] - the limit applies only to low memory,
2965 high memory is not affected.
2967 [ARM64] - only limits memory covered by the linear
2968 mapping. The NOMAP regions are not affected.
2970 [X86] Work as limiting max address. Use together
2971 with memmap= to avoid physical address space collisions.
2972 Without memmap= PCI devices could be placed at addresses
2973 belonging to unused RAM.
2975 Note that this only takes effects during boot time since
2976 in above case 3, memory may need be hot added after boot
2977 if system memory of hypervisor is not sufficient.
2980 [ARM,MIPS] - override the memory layout reported by
2982 Define a memory region of size nn[KMG] starting at
2984 Multiple different regions can be specified with
2985 multiple mem= parameters on the command line.
2987 mem=nopentium [BUGS=X86-32] Disable usage of 4MB pages for kernel
2990 memblock=debug [KNL] Enable memblock debug messages.
2993 [KNL,SH] Allow user to override the default size for
2994 per-device physically contiguous DMA buffers.
2996 memhp_default_state=online/offline
2997 [KNL] Set the initial state for the memory hotplug
2998 onlining policy. If not specified, the default value is
2999 set according to the
3000 CONFIG_MEMORY_HOTPLUG_DEFAULT_ONLINE kernel config
3002 See Documentation/admin-guide/mm/memory-hotplug.rst.
3004 memmap=exactmap [KNL,X86] Enable setting of an exact
3005 E820 memory map, as specified by the user.
3006 Such memmap=exactmap lines can be constructed based on
3007 BIOS output or other requirements. See the memmap=nn@ss
3010 memmap=nn[KMG]@ss[KMG]
3011 [KNL, X86, MIPS, XTENSA] Force usage of a specific region of memory.
3012 Region of memory to be used is from ss to ss+nn.
3013 If @ss[KMG] is omitted, it is equivalent to mem=nn[KMG],
3014 which limits max address to nn[KMG].
3015 Multiple different regions can be specified,
3018 memmap=100M@2G,100M#3G,1G!1024G
3020 memmap=nn[KMG]#ss[KMG]
3021 [KNL,ACPI] Mark specific memory as ACPI data.
3022 Region of memory to be marked is from ss to ss+nn.
3024 memmap=nn[KMG]$ss[KMG]
3025 [KNL,ACPI] Mark specific memory as reserved.
3026 Region of memory to be reserved is from ss to ss+nn.
3027 Example: Exclude memory from 0x18690000-0x1869ffff
3028 memmap=64K$0x18690000
3030 memmap=0x10000$0x18690000
3031 Some bootloaders may need an escape character before '$',
3032 like Grub2, otherwise '$' and the following number
3035 memmap=nn[KMG]!ss[KMG]
3036 [KNL,X86] Mark specific memory as protected.
3037 Region of memory to be used, from ss to ss+nn.
3038 The memory region may be marked as e820 type 12 (0xc)
3039 and is NVDIMM or ADR memory.
3041 memmap=<size>%<offset>-<oldtype>+<newtype>
3042 [KNL,ACPI] Convert memory within the specified region
3043 from <oldtype> to <newtype>. If "-<oldtype>" is left
3044 out, the whole region will be marked as <newtype>,
3045 even if previously unavailable. If "+<newtype>" is left
3046 out, matching memory will be removed. Types are
3047 specified as e820 types, e.g., 1 = RAM, 2 = reserved,
3048 3 = ACPI, 12 = PRAM.
3050 memory_corruption_check=0/1 [X86]
3051 Some BIOSes seem to corrupt the first 64k of
3052 memory when doing things like suspend/resume.
3053 Setting this option will scan the memory
3054 looking for corruption. Enabling this will
3055 both detect corruption and prevent the kernel
3056 from using the memory being corrupted.
3057 However, its intended as a diagnostic tool; if
3058 repeatable BIOS-originated corruption always
3059 affects the same memory, you can use memmap=
3060 to prevent the kernel from using that memory.
3062 memory_corruption_check_size=size [X86]
3063 By default it checks for corruption in the low
3064 64k, making this memory unavailable for normal
3065 use. Use this parameter to scan for
3066 corruption in more or less memory.
3068 memory_corruption_check_period=seconds [X86]
3069 By default it checks for corruption every 60
3070 seconds. Use this parameter to check at some
3071 other rate. 0 disables periodic checking.
3073 memory_hotplug.memmap_on_memory
3074 [KNL,X86,ARM] Boolean flag to enable this feature.
3075 Format: {on | off (default)}
3076 When enabled, runtime hotplugged memory will
3077 allocate its internal metadata (struct pages)
3078 from the hotadded memory which will allow to
3079 hotadd a lot of memory without requiring
3080 additional memory to do so.
3081 This feature is disabled by default because it
3082 has some implication on large (e.g. GB)
3083 allocations in some configurations (e.g. small
3085 The state of the flag can be read in
3086 /sys/module/memory_hotplug/parameters/memmap_on_memory.
3087 Note that even when enabled, there are a few cases where
3088 the feature is not effective.
3090 This is not compatible with hugetlb_free_vmemmap. If
3091 both parameters are enabled, hugetlb_free_vmemmap takes
3092 precedence over memory_hotplug.memmap_on_memory.
3094 memtest= [KNL,X86,ARM,M68K,PPC,RISCV] Enable memtest
3096 default : 0 <disable>
3097 Specifies the number of memtest passes to be
3098 performed. Each pass selects another test
3099 pattern from a given set of patterns. Memtest
3100 fills the memory with this pattern, validates
3101 memory contents and reserves bad memory
3102 regions that are detected.
3104 mem_encrypt= [X86-64] AMD Secure Memory Encryption (SME) control
3105 Valid arguments: on, off
3106 Default (depends on kernel configuration option):
3107 on (CONFIG_AMD_MEM_ENCRYPT_ACTIVE_BY_DEFAULT=y)
3108 off (CONFIG_AMD_MEM_ENCRYPT_ACTIVE_BY_DEFAULT=n)
3109 mem_encrypt=on: Activate SME
3110 mem_encrypt=off: Do not activate SME
3112 Refer to Documentation/virt/kvm/amd-memory-encryption.rst
3113 for details on when memory encryption can be activated.
3115 mem_sleep_default= [SUSPEND] Default system suspend mode:
3116 s2idle - Suspend-To-Idle
3117 shallow - Power-On Suspend or equivalent (if supported)
3118 deep - Suspend-To-RAM or equivalent (if supported)
3119 See Documentation/admin-guide/pm/sleep-states.rst.
3121 meye.*= [HW] Set MotionEye Camera parameters
3122 See Documentation/admin-guide/media/meye.rst.
3124 mfgpt_irq= [IA-32] Specify the IRQ to use for the
3125 Multi-Function General Purpose Timers on AMD Geode
3128 mfgptfix [X86-32] Fix MFGPT timers on AMD Geode platforms when
3129 the BIOS has incorrectly applied a workaround. TinyBIOS
3130 version 0.98 is known to be affected, 0.99 fixes the
3131 problem by letting the user disable the workaround.
3135 min_addr=nn[KMG] [KNL,BOOT,IA-64] All physical memory below this
3136 physical address is ignored.
3138 mini2440= [ARM,HW,KNL]
3139 Format:[0..2][b][c][t]
3141 MINI2440 configuration specification:
3142 0 - The attached screen is the 3.5" TFT
3143 1 - The attached screen is the 7" TFT
3144 2 - The VGA Shield is attached (1024x768)
3145 Leaving out the screen size parameter will not load
3146 the TFT driver, and the framebuffer will be left
3148 b - Enable backlight. The TFT backlight pin will be
3149 linked to the kernel VESA blanking code and a GPIO
3150 LED. This parameter is not necessary when using the
3152 c - Enable the s3c camera interface.
3153 t - Reserved for enabling touchscreen support. The
3154 touchscreen support is not enabled in the mainstream
3155 kernel as of 2.6.30, a preliminary port can be found
3156 in the "bleeding edge" mini2440 support kernel at
3157 https://repo.or.cz/w/linux-2.6/mini2440.git
3160 [X86,PPC,S390,ARM64] Control optional mitigations for
3161 CPU vulnerabilities. This is a set of curated,
3162 arch-independent options, each of which is an
3163 aggregation of existing arch-specific options.
3166 Disable all optional CPU mitigations. This
3167 improves system performance, but it may also
3168 expose users to several CPU vulnerabilities.
3169 Equivalent to: nopti [X86,PPC]
3170 if nokaslr then kpti=0 [ARM64]
3171 nospectre_v1 [X86,PPC]
3173 nospectre_v2 [X86,PPC,S390,ARM64]
3174 spectre_v2_user=off [X86]
3175 spec_store_bypass_disable=off [X86,PPC]
3176 ssbd=force-off [ARM64]
3179 tsx_async_abort=off [X86]
3180 kvm.nx_huge_pages=off [X86]
3181 srbds=off [X86,INTEL]
3182 no_entry_flush [PPC]
3183 no_uaccess_flush [PPC]
3184 mmio_stale_data=off [X86]
3188 This does not have any effect on
3189 kvm.nx_huge_pages when
3190 kvm.nx_huge_pages=force.
3193 Mitigate all CPU vulnerabilities, but leave SMT
3194 enabled, even if it's vulnerable. This is for
3195 users who don't want to be surprised by SMT
3196 getting disabled across kernel upgrades, or who
3197 have other ways of avoiding SMT-based attacks.
3198 Equivalent to: (default behavior)
3201 Mitigate all CPU vulnerabilities, disabling SMT
3202 if needed. This is for users who always want to
3203 be fully mitigated, even if it means losing SMT.
3204 Equivalent to: l1tf=flush,nosmt [X86]
3205 mds=full,nosmt [X86]
3206 tsx_async_abort=full,nosmt [X86]
3207 mmio_stale_data=full,nosmt [X86]
3208 retbleed=auto,nosmt [X86]
3211 [KNL] When CONFIG_DEBUG_MEMORY_INIT is set, this
3212 parameter allows control of the logging verbosity for
3213 the additional memory initialisation checks. A value
3214 of 0 disables mminit logging and a level of 4 will
3215 log everything. Information is printed at KERN_DEBUG
3216 so loglevel=8 may also need to be specified.
3219 [X86,INTEL] Control mitigation for the Processor
3220 MMIO Stale Data vulnerabilities.
3222 Processor MMIO Stale Data is a class of
3223 vulnerabilities that may expose data after an MMIO
3224 operation. Exposed data could originate or end in
3225 the same CPU buffers as affected by MDS and TAA.
3226 Therefore, similar to MDS and TAA, the mitigation
3227 is to clear the affected CPU buffers.
3229 This parameter controls the mitigation. The
3232 full - Enable mitigation on vulnerable CPUs
3234 full,nosmt - Enable mitigation and disable SMT on
3237 off - Unconditionally disable mitigation
3239 On MDS or TAA affected machines,
3240 mmio_stale_data=off can be prevented by an active
3241 MDS or TAA mitigation as these vulnerabilities are
3242 mitigated with the same mechanism so in order to
3243 disable this mitigation, you need to specify
3244 mds=off and tsx_async_abort=off too.
3246 Not specifying this option is equivalent to
3247 mmio_stale_data=full.
3250 Documentation/admin-guide/hw-vuln/processor_mmio_stale_data.rst
3253 [KNL] When CONFIG_MODULE_SIG is set, this means that
3254 modules without (valid) signatures will fail to load.
3255 Note that if CONFIG_MODULE_SIG_FORCE is set, that
3256 is always true, so this option does nothing.
3258 module_blacklist= [KNL] Do not load a comma-separated list of
3259 modules. Useful for debugging problem modules.
3262 [MOUSE] Maximum time between finger touching and
3263 leaving touchpad surface for touch to be considered
3264 a tap and be reported as a left button click (for
3265 touchpads working in absolute mode only).
3267 mousedev.xres= [MOUSE] Horizontal screen resolution, used for devices
3268 reporting absolute coordinates, such as tablets
3269 mousedev.yres= [MOUSE] Vertical screen resolution, used for devices
3270 reporting absolute coordinates, such as tablets
3272 movablecore= [KNL,X86,IA-64,PPC]
3273 Format: nn[KMGTPE] | nn%
3274 This parameter is the complement to kernelcore=, it
3275 specifies the amount of memory used for migratable
3276 allocations. If both kernelcore and movablecore is
3277 specified, then kernelcore will be at *least* the
3278 specified value but may be more. If movablecore on its
3279 own is specified, the administrator must be careful
3280 that the amount of memory usable for all allocations
3283 movable_node [KNL] Boot-time switch to make hotplugable memory
3284 NUMA nodes to be movable. This means that the memory
3285 of such nodes will be usable only for movable
3286 allocations which rules out almost all kernel
3287 allocations. Use with caution!
3289 MTD_Partition= [MTD]
3290 Format: <name>,<region-number>,<size>,<offset>
3292 MTD_Region= [MTD] Format:
3293 <name>,<region-number>[,<base>,<size>,<buswidth>,<altbuswidth>]
3296 See drivers/mtd/parsers/cmdlinepart.c
3299 ARM/S3C2412 JIVE boot control
3301 See arch/arm/mach-s3c/mach-jive.c
3303 mtouchusb.raw_coordinates=
3304 [HW] Make the MicroTouch USB driver use raw coordinates
3305 ('y', default) or cooked coordinates ('n')
3307 mtrr_chunk_size=nn[KMG] [X86]
3308 used for mtrr cleanup. It is largest continuous chunk
3309 that could hold holes aka. UC entries.
3311 mtrr_gran_size=nn[KMG] [X86]
3312 Used for mtrr cleanup. It is granularity of mtrr block.
3314 Large value could prevent small alignment from
3317 mtrr_spare_reg_nr=n [X86]
3319 Range: 0,7 : spare reg number
3321 Used for mtrr cleanup. It is spare mtrr entries number.
3322 Set to 2 or more if your graphical card needs more.
3324 multitce=off [PPC] This parameter disables the use of the pSeries
3325 firmware feature for updating multiple TCE entries
3328 n2= [NET] SDL Inc. RISCom/N2 synchronous serial card
3330 netdev= [NET] Network devices parameters
3331 Format: <irq>,<io>,<mem_start>,<mem_end>,<name>
3332 Note that mem_start is often overloaded to mean
3333 something different and driver-specific.
3334 This usage is only documented in each driver source
3337 netpoll.carrier_timeout=
3338 [NET] Specifies amount of time (in seconds) that
3339 netpoll should wait for a carrier. By default netpoll
3343 [NETFILTER] Enable connection tracking flow accounting
3344 0 to disable accounting
3345 1 to enable accounting
3348 nfsaddrs= [NFS] Deprecated. Use ip= instead.
3349 See Documentation/admin-guide/nfs/nfsroot.rst.
3351 nfsroot= [NFS] nfs root filesystem for disk-less boxes.
3352 See Documentation/admin-guide/nfs/nfsroot.rst.
3354 nfsrootdebug [NFS] enable nfsroot debugging messages.
3355 See Documentation/admin-guide/nfs/nfsroot.rst.
3357 nfs.callback_nr_threads=
3358 [NFSv4] set the total number of threads that the
3359 NFS client will assign to service NFSv4 callback
3362 nfs.callback_tcpport=
3363 [NFS] set the TCP port on which the NFSv4 callback
3364 channel should listen.
3367 [NFS] sets the pathname to the program which is used
3368 to update the NFS client cache entries.
3370 nfs.cache_getent_timeout=
3371 [NFS] sets the timeout after which an attempt to
3372 update a cache entry is deemed to have failed.
3374 nfs.idmap_cache_timeout=
3375 [NFS] set the maximum lifetime for idmapper cache
3379 [NFS] enable 64-bit inode numbers.
3380 If zero, the NFS client will fake up a 32-bit inode
3381 number for the readdir() and stat() syscalls instead
3382 of returning the full 64-bit number.
3383 The default is to return 64-bit inode numbers.
3385 nfs.max_session_cb_slots=
3386 [NFSv4.1] Sets the maximum number of session
3387 slots the client will assign to the callback
3388 channel. This determines the maximum number of
3389 callbacks the client will process in parallel for
3390 a particular server.
3392 nfs.max_session_slots=
3393 [NFSv4.1] Sets the maximum number of session slots
3394 the client will attempt to negotiate with the server.
3395 This limits the number of simultaneous RPC requests
3396 that the client can send to the NFSv4.1 server.
3397 Note that there is little point in setting this
3398 value higher than the max_tcp_slot_table_limit.
3400 nfs.nfs4_disable_idmapping=
3401 [NFSv4] When set to the default of '1', this option
3402 ensures that both the RPC level authentication
3403 scheme and the NFS level operations agree to use
3404 numeric uids/gids if the mount is using the
3405 'sec=sys' security flavour. In effect it is
3406 disabling idmapping, which can make migration from
3407 legacy NFSv2/v3 systems to NFSv4 easier.
3408 Servers that do not support this mode of operation
3409 will be autodetected by the client, and it will fall
3410 back to using the idmapper.
3411 To turn off this behaviour, set the value to '0'.
3413 [NFS4] Specify an additional fixed unique ident-
3414 ification string that NFSv4 clients can insert into
3415 their nfs_client_id4 string. This is typically a
3416 UUID that is generated at system install time.
3418 nfs.send_implementation_id =
3419 [NFSv4.1] Send client implementation identification
3420 information in exchange_id requests.
3421 If zero, no implementation identification information
3423 The default is to send the implementation identification
3426 nfs.recover_lost_locks =
3427 [NFSv4] Attempt to recover locks that were lost due
3428 to a lease timeout on the server. Please note that
3429 doing this risks data corruption, since there are
3430 no guarantees that the file will remain unchanged
3431 after the locks are lost.
3432 If you want to enable the kernel legacy behaviour of
3433 attempting to recover these locks, then set this
3435 The default parameter value of '0' causes the kernel
3436 not to attempt recovery of lost locks.
3438 nfs4.layoutstats_timer =
3439 [NFSv4.2] Change the rate at which the kernel sends
3440 layoutstats to the pNFS metadata server.
3442 Setting this to value to 0 causes the kernel to use
3443 whatever value is the default set by the layout
3444 driver. A non-zero value sets the minimum interval
3445 in seconds between layoutstats transmissions.
3447 nfsd.inter_copy_offload_enable =
3448 [NFSv4.2] When set to 1, the server will support
3449 server-to-server copies for which this server is
3450 the destination of the copy.
3452 nfsd.nfsd4_ssc_umount_timeout =
3453 [NFSv4.2] When used as the destination of a
3454 server-to-server copy, knfsd temporarily mounts
3455 the source server. It caches the mount in case
3456 it will be needed again, and discards it if not
3457 used for the number of milliseconds specified by
3460 nfsd.nfs4_disable_idmapping=
3461 [NFSv4] When set to the default of '1', the NFSv4
3462 server will return only numeric uids and gids to
3463 clients using auth_sys, and will accept numeric uids
3464 and gids from such clients. This is intended to ease
3465 migration from NFSv2/v3.
3468 nmi_backtrace.backtrace_idle [KNL]
3469 Dump stacks even of idle CPUs in response to an
3470 NMI stack-backtrace request.
3472 nmi_debug= [KNL,SH] Specify one or more actions to take
3473 when a NMI is triggered.
3474 Format: [state][,regs][,debounce][,die]
3476 nmi_watchdog= [KNL,BUGS=X86] Debugging features for SMP kernels
3477 Format: [panic,][nopanic,][num]
3479 0 - turn hardlockup detector in nmi_watchdog off
3480 1 - turn hardlockup detector in nmi_watchdog on
3481 When panic is specified, panic when an NMI watchdog
3482 timeout occurs (or 'nopanic' to not panic on an NMI
3483 watchdog, if CONFIG_BOOTPARAM_HARDLOCKUP_PANIC is set)
3484 To disable both hard and soft lockup detectors,
3485 please see 'nowatchdog'.
3486 This is useful when you use a panic=... timeout and
3487 need the box quickly up again.
3489 These settings can be accessed at runtime via
3490 the nmi_watchdog and hardlockup_panic sysctls.
3492 no387 [BUGS=X86-32] Tells the kernel to use the 387 maths
3493 emulation library even if a 387 maths coprocessor
3496 no5lvl [X86-64] Disable 5-level paging mode. Forces
3497 kernel to use 4-level paging instead.
3499 nofsgsbase [X86] Disables FSGSBASE instructions.
3502 [HW] Never suspend the console
3503 Disable suspending of consoles during suspend and
3504 hibernate operations. Once disabled, debugging
3505 messages can reach various consoles while the rest
3506 of the system is being put to sleep (ie, while
3507 debugging driver suspend/resume hooks). This may
3508 not work reliably with all consoles, but is known
3509 to work with serial and VGA consoles.
3510 To facilitate more flexible debugging, we also add
3511 console_suspend, a printk module parameter to control
3512 it. Users could use console_suspend (usually
3513 /sys/module/printk/parameters/console_suspend) to
3514 turn on/off it dynamically.
3516 novmcoredd [KNL,KDUMP]
3517 Disable device dump. Device dump allows drivers to
3518 append dump data to vmcore so you can collect driver
3519 specified debug info. Drivers can append the data
3520 without any limit and this data is stored in memory,
3521 so this may cause significant memory stress. Disabling
3522 device dump can help save memory but the driver debug
3523 data will be no longer available. This parameter
3524 is only available when CONFIG_PROC_VMCORE_DEVICE_DUMP
3527 noaliencache [MM, NUMA, SLAB] Disables the allocation of alien
3528 caches in the slab allocator. Saves per-node memory,
3529 but will impact performance.
3533 noaltinstr [S390] Disables alternative instructions patching
3534 (CPU alternatives feature).
3536 noapic [SMP,APIC] Tells the kernel to not make use of any
3537 IOAPICs that may be present in the system.
3539 noautogroup Disable scheduler automatic task group creation.
3541 nobats [PPC] Do not use BATs for mapping kernel lowmem
3542 on "Classic" PPC cores.
3546 nodsp [SH] Disable hardware DSP at boot time.
3548 noefi Disable EFI runtime services support.
3550 no_entry_flush [PPC] Don't flush the L1-D cache when entering the kernel.
3555 Disable SMAP (Supervisor Mode Access Prevention)
3556 even if it is supported by processor.
3559 Disable SMEP (Supervisor Mode Execution Prevention)
3560 even if it is supported by processor.
3563 This affects only 32-bit executables.
3564 noexec32=on: enable non-executable mappings (default)
3565 read doesn't imply executable mappings
3566 noexec32=off: disable non-executable mappings
3567 read implies executable mappings
3569 nofpu [MIPS,SH] Disable hardware FPU at boot time.
3571 nofxsr [BUGS=X86-32] Disables x86 floating point extended
3572 register save and restore. The kernel will only save
3573 legacy floating-point registers on task switch.
3575 nohugeiomap [KNL,X86,PPC,ARM64] Disable kernel huge I/O mappings.
3577 nohugevmalloc [PPC] Disable kernel huge vmalloc mappings.
3579 nosmt [KNL,S390] Disable symmetric multithreading (SMT).
3580 Equivalent to smt=1.
3582 [KNL,X86] Disable symmetric multithreading (SMT).
3583 nosmt=force: Force disable SMT, cannot be undone
3584 via the sysfs control file.
3586 nospectre_v1 [X86,PPC] Disable mitigations for Spectre Variant 1
3587 (bounds check bypass). With this option data leaks are
3588 possible in the system.
3590 nospectre_v2 [X86,PPC_FSL_BOOK3E,ARM64] Disable all mitigations for
3591 the Spectre variant 2 (indirect branch prediction)
3592 vulnerability. System may allow data leaks with this
3595 nospec_store_bypass_disable
3596 [HW] Disable all mitigations for the Speculative Store Bypass vulnerability
3599 [PPC] Don't flush the L1-D cache after accessing user data.
3601 noxsave [BUGS=X86] Disables x86 extended register state save
3602 and restore using xsave. The kernel will fallback to
3603 enabling legacy floating-point and sse state.
3605 noxsaveopt [X86] Disables xsaveopt used in saving x86 extended
3606 register states. The kernel will fall back to use
3607 xsave to save the states. By using this parameter,
3608 performance of saving the states is degraded because
3609 xsave doesn't support modified optimization while
3610 xsaveopt supports it on xsaveopt enabled systems.
3612 noxsaves [X86] Disables xsaves and xrstors used in saving and
3613 restoring x86 extended register state in compacted
3614 form of xsave area. The kernel will fall back to use
3615 xsaveopt and xrstor to save and restore the states
3616 in standard form of xsave area. By using this
3617 parameter, xsave area per process might occupy more
3618 memory on xsaves enabled systems.
3620 nohlt [ARM,ARM64,MICROBLAZE,SH] Forces the kernel to busy wait
3621 in do_idle() and not use the arch_cpu_idle()
3622 implementation; requires CONFIG_GENERIC_IDLE_POLL_SETUP
3623 to be effective. This is useful on platforms where the
3624 sleep(SH) or wfi(ARM,ARM64) instructions do not work
3625 correctly or when doing power measurements to evalute
3626 the impact of the sleep instructions. This is also
3627 useful when using JTAG debugger.
3629 no_file_caps Tells the kernel not to honor file capabilities. The
3630 only way then for a file to be executed with privilege
3631 is to be setuid root or executed by root.
3633 nohalt [IA-64] Tells the kernel not to use the power saving
3634 function PAL_HALT_LIGHT when idle. This increases
3635 power-consumption. On the positive side, it reduces
3636 interrupt wake-up latency, which may improve performance
3637 in certain environments such as networked servers or
3641 Force pointers printed to the console or buffers to be
3642 unhashed. By default, when a pointer is printed via %p
3643 format string, that pointer is "hashed", i.e. obscured
3644 by hashing the pointer value. This is a security feature
3645 that hides actual kernel addresses from unprivileged
3646 users, but it also makes debugging the kernel more
3647 difficult since unequal pointers can no longer be
3648 compared. However, if this command-line option is
3649 specified, then all normal pointers will have their true
3650 value printed. This option should only be specified when
3651 debugging the kernel. Please do not use on production
3654 nohibernate [HIBERNATION] Disable hibernation and resume.
3656 nohz= [KNL] Boottime enable/disable dynamic ticks
3657 Valid arguments: on, off
3660 nohz_full= [KNL,BOOT,SMP,ISOL]
3661 The argument is a cpu list, as described above.
3662 In kernels built with CONFIG_NO_HZ_FULL=y, set
3663 the specified list of CPUs whose tick will be stopped
3664 whenever possible. The boot CPU will be forced outside
3665 the range to maintain the timekeeping. Any CPUs
3666 in this list will have their RCU callbacks offloaded,
3667 just as if they had also been called out in the
3668 rcu_nocbs= boot parameter.
3670 noiotrap [SH] Disables trapped I/O port accesses.
3672 noirqdebug [X86-32] Disables the code which attempts to detect and
3673 disable unhandled interrupt sources.
3675 no_timer_check [X86,APIC] Disables the code which tests for
3676 broken timer IRQ sources.
3678 noisapnp [ISAPNP] Disables ISA PnP code.
3680 noinitrd [RAM] Tells the kernel not to load any configured
3683 nointremap [X86-64, Intel-IOMMU] Do not enable interrupt
3685 [Deprecated - use intremap=off]
3689 noinvpcid [X86] Disable the INVPCID cpu feature.
3691 nojitter [IA-64] Disables jitter checking for ITC timers.
3693 no-kvmclock [X86,KVM] Disable paravirtualized KVM clock driver
3695 no-kvmapf [X86,KVM] Disable paravirtualized asynchronous page
3699 [X86,PV_OPS] Disable paravirtualized VMware scheduler
3700 clock and use the default one.
3702 no-steal-acc [X86,PV_OPS,ARM64] Disable paravirtualized steal time
3703 accounting. steal time is computed, but won't
3704 influence scheduler behaviour
3706 nolapic [X86-32,APIC] Do not enable or use the local APIC.
3708 nolapic_timer [X86-32,APIC] Do not use the local APIC timer.
3710 noltlbs [PPC] Do not use large page/tlb entries for kernel
3711 lowmem mapping on PPC40x and PPC8xx
3713 nomca [IA-64] Disable machine check abort handling
3715 nomce [X86-32] Disable Machine Check Exception
3717 nomfgpt [X86-32] Disable Multi-Function General Purpose
3718 Timer usage (for AMD Geode machines).
3720 nonmi_ipi [X86] Disable using NMI IPIs during panic/reboot to
3721 shutdown the other cpus. Instead use the REBOOT_VECTOR
3724 nomodeset Disable kernel modesetting. DRM drivers will not perform
3725 display-mode changes or accelerated rendering. Only the
3726 system framebuffer will be available for use if this was
3727 set-up by the firmware or boot loader.
3729 Useful as fallback, or for testing and debugging.
3731 nomodule Disable module load
3733 nopat [X86] Disable PAT (page attribute table extension of
3734 pagetables) support.
3736 nopcid [X86-64] Disable the PCID cpu feature.
3738 norandmaps Don't use address space randomization. Equivalent to
3739 echo 0 > /proc/sys/kernel/randomize_va_space
3741 noreplace-smp [X86-32,SMP] Don't replace SMP instructions
3742 with UP alternatives
3744 noresume [SWSUSP] Disables resume and restores original swap
3747 no-scroll [VGA] Disables scrollback.
3748 This is required for the Braillex ib80-piezo Braille
3749 reader made by F.H. Papenmeier (Germany).
3753 nosgx [X86-64,SGX] Disables Intel SGX kernel support.
3755 nosmp [SMP] Tells an SMP kernel to act as a UP kernel,
3756 and disable the IO APIC. legacy for "maxcpus=0".
3758 nosoftlockup [KNL] Disable the soft-lockup detector.
3760 nosync [HW,M68K] Disables sync negotiation for all devices.
3762 nowatchdog [KNL] Disable both lockup detectors, i.e.
3763 soft-lockup and NMI watchdog (hard-lockup).
3767 nox2apic [X86-64,APIC] Do not enable x2APIC mode.
3769 nps_mtm_hs_ctr= [KNL,ARC]
3770 This parameter sets the maximum duration, in
3771 cycles, each HW thread of the CTOP can run
3772 without interruptions, before HW switches it.
3773 The actual maximum duration is 16 times this
3775 Format: integer between 1 and 255
3778 nptcg= [IA-64] Override max number of concurrent global TLB
3779 purges which is reported from either PAL_VM_SUMMARY or
3782 nr_cpus= [SMP] Maximum number of processors that an SMP kernel
3783 could support. nr_cpus=n : n >= 1 limits the kernel to
3784 support 'n' processors. It could be larger than the
3785 number of already plugged CPU during bootup, later in
3786 runtime you can physically add extra cpu until it reaches
3787 n. So during boot up some boot time memory for per-cpu
3788 variables need be pre-allocated for later physical cpu
3791 nr_uarts= [SERIAL] maximum number of UARTs to be registered.
3793 numa=off [KNL, ARM64, PPC, RISCV, SPARC, X86] Disable NUMA, Only
3794 set up a single NUMA node spanning all memory.
3796 numa_balancing= [KNL,ARM64,PPC,RISCV,S390,X86] Enable or disable automatic
3798 Allowed values are enable and disable
3800 numa_zonelist_order= [KNL, BOOT] Select zonelist order for NUMA.
3801 'node', 'default' can be specified
3802 This can be set from sysctl after boot.
3803 See Documentation/admin-guide/sysctl/vm.rst for details.
3805 ohci1394_dma=early [HW] enable debugging via the ohci1394 driver.
3806 See Documentation/core-api/debugging-via-ohci1394.rst for more
3809 olpc_ec_timeout= [OLPC] ms delay when issuing EC commands
3810 Rather than timing out after 20 ms if an EC
3811 command is not properly ACKed, override the length
3812 of the timeout. We have interrupts disabled while
3813 waiting for the ACK, so if this is set too high
3814 interrupts *may* be lost!
3816 omap_mux= [OMAP] Override bootloader pin multiplexing.
3817 Format: <mux_mode0.mode_name=value>...
3818 For example, to override I2C bus2:
3819 omap_mux=i2c2_scl.i2c2_scl=0x100,i2c2_sda.i2c2_sda=0x100
3821 onenand.bdry= [HW,MTD] Flex-OneNAND Boundary Configuration
3823 Format: [die0_boundary][,die0_lock][,die1_boundary][,die1_lock]
3825 boundary - index of last SLC block on Flex-OneNAND.
3826 The remaining blocks are configured as MLC blocks.
3827 lock - Configure if Flex-OneNAND boundary should be locked.
3828 Once locked, the boundary cannot be changed.
3829 1 indicates lock status, 0 indicates unlock status.
3831 oops=panic Always panic on oopses. Default is to just kill the
3832 process, but there is a small probability of
3833 deadlocking the machine.
3834 This will also cause panics on machine check exceptions.
3835 Useful together with panic=30 to trigger a reboot.
3838 [KNL] Boolean flag to control whether the page allocator
3839 should randomize its free lists. The randomization may
3840 be automatically enabled if the kernel detects it is
3841 running on a platform with a direct-mapped memory-side
3842 cache, and this parameter can be used to
3843 override/disable that behavior. The state of the flag
3844 can be read from sysfs at:
3845 /sys/module/page_alloc/parameters/shuffle.
3847 page_owner= [KNL] Boot-time page_owner enabling option.
3848 Storage of the information about who allocated
3849 each page is disabled in default. With this switch,
3851 on: enable the feature
3853 page_poison= [KNL] Boot-time parameter changing the state of
3854 poisoning on the buddy allocator, available with
3855 CONFIG_PAGE_POISONING=y.
3856 off: turn off poisoning (default)
3857 on: turn on poisoning
3859 page_reporting.page_reporting_order=
3860 [KNL] Minimal page reporting order
3862 Adjust the minimal page reporting order. The page
3863 reporting is disabled when it exceeds (MAX_ORDER-1).
3865 panic= [KNL] Kernel behaviour on panic: delay <timeout>
3866 timeout > 0: seconds before rebooting
3867 timeout = 0: wait forever
3868 timeout < 0: reboot immediately
3871 panic_print= Bitmask for printing system info when panic happens.
3872 User can chose combination of the following bits:
3873 bit 0: print all tasks info
3874 bit 1: print system memory info
3875 bit 2: print timer info
3876 bit 3: print locks info if CONFIG_LOCKDEP is on
3877 bit 4: print ftrace buffer
3878 bit 5: print all printk messages in buffer
3879 bit 6: print all CPUs backtrace (if available in the arch)
3880 *Be aware* that this option may print a _lot_ of lines,
3881 so there are risks of losing older messages in the log.
3882 Use this option carefully, maybe worth to setup a
3883 bigger log buffer with "log_buf_len" along with this.
3885 panic_on_taint= Bitmask for conditionally calling panic() in add_taint()
3886 Format: <hex>[,nousertaint]
3887 Hexadecimal bitmask representing the set of TAINT flags
3888 that will cause the kernel to panic when add_taint() is
3889 called with any of the flags in this set.
3890 The optional switch "nousertaint" can be utilized to
3891 prevent userspace forced crashes by writing to sysctl
3892 /proc/sys/kernel/tainted any flagset matching with the
3893 bitmask set on panic_on_taint.
3894 See Documentation/admin-guide/tainted-kernels.rst for
3895 extra details on the taint flags that users can pick
3896 to compose the bitmask to assign to panic_on_taint.
3898 panic_on_warn panic() instead of WARN(). Useful to cause kdump
3901 parkbd.port= [HW] Parallel port number the keyboard adapter is
3902 connected to, default is 0.
3904 parkbd.mode= [HW] Parallel port keyboard adapter mode of operation,
3905 0 for XT, 1 for AT (default is AT).
3908 parport= [HW,PPT] Specify parallel ports. 0 disables.
3909 Format: { 0 | auto | 0xBBB[,IRQ[,DMA]] }
3910 Use 'auto' to force the driver to use any
3911 IRQ/DMA settings detected (the default is to
3912 ignore detected IRQ/DMA settings because of
3913 possible conflicts). You can specify the base
3914 address, IRQ, and DMA settings; IRQ and DMA
3915 should be numbers, or 'auto' (for using detected
3916 settings on that particular port), or 'nofifo'
3917 (to avoid using a FIFO even if it is detected).
3918 Parallel ports are assigned in the order they
3919 are specified on the command line, starting
3922 parport_init_mode= [HW,PPT]
3923 Configure VIA parallel port to operate in
3924 a specific mode. This is necessary on Pegasos
3925 computer where firmware has no options for setting
3926 up parallel port mode and sets it to spp.
3927 Currently this function knows 686a and 8231 chips.
3928 Format: [spp|ps2|epp|ecp|ecpepp]
3930 pata_legacy.all= [HW,LIBATA]
3932 Set to non-zero to probe primary and secondary ISA
3933 port ranges on PCI systems where no PCI PATA device
3934 has been found at either range. Disabled by default.
3936 pata_legacy.autospeed= [HW,LIBATA]
3938 Set to non-zero if a chip is present that snoops speed
3939 changes. Disabled by default.
3941 pata_legacy.ht6560a= [HW,LIBATA]
3943 Set to 1, 2, or 3 for HT 6560A on the primary channel,
3944 the secondary channel, or both channels respectively.
3945 Disabled by default.
3947 pata_legacy.ht6560b= [HW,LIBATA]
3949 Set to 1, 2, or 3 for HT 6560B on the primary channel,
3950 the secondary channel, or both channels respectively.
3951 Disabled by default.
3953 pata_legacy.iordy_mask= [HW,LIBATA]
3955 IORDY enable mask. Set individual bits to allow IORDY
3956 for the respective channel. Bit 0 is for the first
3957 legacy channel handled by this driver, bit 1 is for
3958 the second channel, and so on. The sequence will often
3959 correspond to the primary legacy channel, the secondary
3960 legacy channel, and so on, but the handling of a PCI
3961 bus and the use of other driver options may interfere
3962 with the sequence. By default IORDY is allowed across
3965 pata_legacy.opti82c46x= [HW,LIBATA]
3967 Set to 1, 2, or 3 for Opti 82c611A on the primary
3968 channel, the secondary channel, or both channels
3969 respectively. Disabled by default.
3971 pata_legacy.opti82c611a= [HW,LIBATA]
3973 Set to 1, 2, or 3 for Opti 82c465MV on the primary
3974 channel, the secondary channel, or both channels
3975 respectively. Disabled by default.
3977 pata_legacy.pio_mask= [HW,LIBATA]
3979 PIO mode mask for autospeed devices. Set individual
3980 bits to allow the use of the respective PIO modes.
3981 Bit 0 is for mode 0, bit 1 is for mode 1, and so on.
3982 All modes allowed by default.
3984 pata_legacy.probe_all= [HW,LIBATA]
3986 Set to non-zero to probe tertiary and further ISA
3987 port ranges on PCI systems. Disabled by default.
3989 pata_legacy.probe_mask= [HW,LIBATA]
3991 Probe mask for legacy ISA PATA ports. Depending on
3992 platform configuration and the use of other driver
3993 options up to 6 legacy ports are supported: 0x1f0,
3994 0x170, 0x1e8, 0x168, 0x1e0, 0x160, however probing
3995 of individual ports can be disabled by setting the
3996 corresponding bits in the mask to 1. Bit 0 is for
3997 the first port in the list above (0x1f0), and so on.
3998 By default all supported ports are probed.
4000 pata_legacy.qdi= [HW,LIBATA]
4002 Set to non-zero to probe QDI controllers. By default
4003 set to 1 if CONFIG_PATA_QDI_MODULE, 0 otherwise.
4005 pata_legacy.winbond= [HW,LIBATA]
4007 Set to non-zero to probe Winbond controllers. Use
4008 the standard I/O port (0x130) if 1, otherwise the
4009 value given is the I/O port to use (typically 0x1b0).
4010 By default set to 1 if CONFIG_PATA_WINBOND_VLB_MODULE,
4013 pata_platform.pio_mask= [HW,LIBATA]
4015 Supported PIO mode mask. Set individual bits to allow
4016 the use of the respective PIO modes. Bit 0 is for
4017 mode 0, bit 1 is for mode 1, and so on. Mode 0 only
4021 Halt all CPUs after the first oops has been printed for
4022 the specified number of seconds. This is to be used if
4023 your oopses keep scrolling off the screen.
4028 See header of drivers/block/paride/pcd.c.
4029 See also Documentation/admin-guide/blockdev/paride.rst.
4031 pci=option[,option...] [PCI] various PCI subsystem options.
4033 Some options herein operate on a specific device
4034 or a set of devices (<pci_dev>). These are
4035 specified in one of the following formats:
4037 [<domain>:]<bus>:<dev>.<func>[/<dev>.<func>]*
4038 pci:<vendor>:<device>[:<subvendor>:<subdevice>]
4040 Note: the first format specifies a PCI
4041 bus/device/function address which may change
4042 if new hardware is inserted, if motherboard
4043 firmware changes, or due to changes caused
4044 by other kernel parameters. If the
4045 domain is left unspecified, it is
4046 taken to be zero. Optionally, a path
4047 to a device through multiple device/function
4048 addresses can be specified after the base
4049 address (this is more robust against
4050 renumbering issues). The second format
4051 selects devices using IDs from the
4052 configuration space which may match multiple
4053 devices in the system.
4055 earlydump dump PCI config space before the kernel
4057 off [X86] don't probe for the PCI bus
4058 bios [X86-32] force use of PCI BIOS, don't access
4059 the hardware directly. Use this if your machine
4060 has a non-standard PCI host bridge.
4061 nobios [X86-32] disallow use of PCI BIOS, only direct
4062 hardware access methods are allowed. Use this
4063 if you experience crashes upon bootup and you
4064 suspect they are caused by the BIOS.
4065 conf1 [X86] Force use of PCI Configuration Access
4066 Mechanism 1 (config address in IO port 0xCF8,
4067 data in IO port 0xCFC, both 32-bit).
4068 conf2 [X86] Force use of PCI Configuration Access
4069 Mechanism 2 (IO port 0xCF8 is an 8-bit port for
4070 the function, IO port 0xCFA, also 8-bit, sets
4071 bus number. The config space is then accessed
4072 through ports 0xC000-0xCFFF).
4073 See http://wiki.osdev.org/PCI for more info
4074 on the configuration access mechanisms.
4075 noaer [PCIE] If the PCIEAER kernel config parameter is
4076 enabled, this kernel boot option can be used to
4077 disable the use of PCIE advanced error reporting.
4078 nodomains [PCI] Disable support for multiple PCI
4079 root domains (aka PCI segments, in ACPI-speak).
4080 nommconf [X86] Disable use of MMCONFIG for PCI
4082 check_enable_amd_mmconf [X86] check for and enable
4083 properly configured MMIO access to PCI
4084 config space on AMD family 10h CPU
4085 nomsi [MSI] If the PCI_MSI kernel config parameter is
4086 enabled, this kernel boot option can be used to
4087 disable the use of MSI interrupts system-wide.
4088 noioapicquirk [APIC] Disable all boot interrupt quirks.
4089 Safety option to keep boot IRQs enabled. This
4090 should never be necessary.
4091 ioapicreroute [APIC] Enable rerouting of boot IRQs to the
4092 primary IO-APIC for bridges that cannot disable
4093 boot IRQs. This fixes a source of spurious IRQs
4094 when the system masks IRQs.
4095 noioapicreroute [APIC] Disable workaround that uses the
4096 boot IRQ equivalent of an IRQ that connects to
4097 a chipset where boot IRQs cannot be disabled.
4098 The opposite of ioapicreroute.
4099 biosirq [X86-32] Use PCI BIOS calls to get the interrupt
4100 routing table. These calls are known to be buggy
4101 on several machines and they hang the machine
4102 when used, but on other computers it's the only
4103 way to get the interrupt routing table. Try
4104 this option if the kernel is unable to allocate
4105 IRQs or discover secondary PCI buses on your
4107 rom [X86] Assign address space to expansion ROMs.
4108 Use with caution as certain devices share
4109 address decoders between ROMs and other
4111 norom [X86] Do not assign address space to
4112 expansion ROMs that do not already have
4113 BIOS assigned address ranges.
4114 nobar [X86] Do not assign address space to the
4115 BARs that weren't assigned by the BIOS.
4116 irqmask=0xMMMM [X86] Set a bit mask of IRQs allowed to be
4117 assigned automatically to PCI devices. You can
4118 make the kernel exclude IRQs of your ISA cards
4120 pirqaddr=0xAAAAA [X86] Specify the physical address
4121 of the PIRQ table (normally generated
4122 by the BIOS) if it is outside the
4123 F0000h-100000h range.
4124 lastbus=N [X86] Scan all buses thru bus #N. Can be
4125 useful if the kernel is unable to find your
4126 secondary buses and you want to tell it
4127 explicitly which ones they are.
4128 assign-busses [X86] Always assign all PCI bus
4129 numbers ourselves, overriding
4130 whatever the firmware may have done.
4131 usepirqmask [X86] Honor the possible IRQ mask stored
4132 in the BIOS $PIR table. This is needed on
4133 some systems with broken BIOSes, notably
4134 some HP Pavilion N5400 and Omnibook XE3
4135 notebooks. This will have no effect if ACPI
4136 IRQ routing is enabled.
4137 noacpi [X86] Do not use ACPI for IRQ routing
4138 or for PCI scanning.
4139 use_crs [X86] Use PCI host bridge window information
4140 from ACPI. On BIOSes from 2008 or later, this
4141 is enabled by default. If you need to use this,
4142 please report a bug.
4143 nocrs [X86] Ignore PCI host bridge windows from ACPI.
4144 If you need to use this, please report a bug.
4145 use_e820 [X86] Use E820 reservations to exclude parts of
4146 PCI host bridge windows. This is a workaround
4147 for BIOS defects in host bridge _CRS methods.
4148 If you need to use this, please report a bug to
4149 <linux-pci@vger.kernel.org>.
4150 no_e820 [X86] Ignore E820 reservations for PCI host
4151 bridge windows. This is the default on modern
4152 hardware. If you need to use this, please report
4153 a bug to <linux-pci@vger.kernel.org>.
4154 routeirq Do IRQ routing for all PCI devices.
4155 This is normally done in pci_enable_device(),
4156 so this option is a temporary workaround
4157 for broken drivers that don't call it.
4158 skip_isa_align [X86] do not align io start addr, so can
4159 handle more pci cards
4160 noearly [X86] Don't do any early type 1 scanning.
4161 This might help on some broken boards which
4162 machine check when some devices' config space
4163 is read. But various workarounds are disabled
4164 and some IOMMU drivers will not work.
4165 bfsort Sort PCI devices into breadth-first order.
4166 This sorting is done to get a device
4167 order compatible with older (<= 2.4) kernels.
4168 nobfsort Don't sort PCI devices into breadth-first order.
4169 pcie_bus_tune_off Disable PCIe MPS (Max Payload Size)
4170 tuning and use the BIOS-configured MPS defaults.
4171 pcie_bus_safe Set every device's MPS to the largest value
4172 supported by all devices below the root complex.
4173 pcie_bus_perf Set device MPS to the largest allowable MPS
4174 based on its parent bus. Also set MRRS (Max
4175 Read Request Size) to the largest supported
4176 value (no larger than the MPS that the device
4177 or bus can support) for best performance.
4178 pcie_bus_peer2peer Set every device's MPS to 128B, which
4179 every device is guaranteed to support. This
4180 configuration allows peer-to-peer DMA between
4181 any pair of devices, possibly at the cost of
4182 reduced performance. This also guarantees
4183 that hot-added devices will work.
4184 cbiosize=nn[KMG] The fixed amount of bus space which is
4185 reserved for the CardBus bridge's IO window.
4186 The default value is 256 bytes.
4187 cbmemsize=nn[KMG] The fixed amount of bus space which is
4188 reserved for the CardBus bridge's memory
4189 window. The default value is 64 megabytes.
4192 [<order of align>@]<pci_dev>[; ...]
4193 Specifies alignment and device to reassign
4194 aligned memory resources. How to
4195 specify the device is described above.
4196 If <order of align> is not specified,
4197 PAGE_SIZE is used as alignment.
4198 A PCI-PCI bridge can be specified if resource
4199 windows need to be expanded.
4200 To specify the alignment for several
4201 instances of a device, the PCI vendor,
4202 device, subvendor, and subdevice may be
4203 specified, e.g., 12@pci:8086:9c22:103c:198f
4204 for 4096-byte alignment.
4205 ecrc= Enable/disable PCIe ECRC (transaction layer
4206 end-to-end CRC checking).
4207 bios: Use BIOS/firmware settings. This is the
4211 hpiosize=nn[KMG] The fixed amount of bus space which is
4212 reserved for hotplug bridge's IO window.
4213 Default size is 256 bytes.
4214 hpmmiosize=nn[KMG] The fixed amount of bus space which is
4215 reserved for hotplug bridge's MMIO window.
4216 Default size is 2 megabytes.
4217 hpmmioprefsize=nn[KMG] The fixed amount of bus space which is
4218 reserved for hotplug bridge's MMIO_PREF window.
4219 Default size is 2 megabytes.
4220 hpmemsize=nn[KMG] The fixed amount of bus space which is
4221 reserved for hotplug bridge's MMIO and
4223 Default size is 2 megabytes.
4224 hpbussize=nn The minimum amount of additional bus numbers
4225 reserved for buses below a hotplug bridge.
4227 realloc= Enable/disable reallocating PCI bridge resources
4228 if allocations done by BIOS are too small to
4229 accommodate resources required by all child
4231 off: Turn realloc off
4233 realloc same as realloc=on
4234 noari do not use PCIe ARI.
4235 noats [PCIE, Intel-IOMMU, AMD-IOMMU]
4236 do not use PCIe ATS (and IOMMU device IOTLB).
4237 pcie_scan_all Scan all possible PCIe devices. Otherwise we
4238 only look for one device below a PCIe downstream
4240 big_root_window Try to add a big 64bit memory window to the PCIe
4241 root complex on AMD CPUs. Some GFX hardware
4242 can resize a BAR to allow access to all VRAM.
4243 Adding the window is slightly risky (it may
4244 conflict with unreported devices), so this
4246 disable_acs_redir=<pci_dev>[; ...]
4247 Specify one or more PCI devices (in the format
4248 specified above) separated by semicolons.
4249 Each device specified will have the PCI ACS
4250 redirect capabilities forced off which will
4251 allow P2P traffic between devices through
4252 bridges without forcing it upstream. Note:
4253 this removes isolation between devices and
4254 may put more devices in an IOMMU group.
4255 force_floating [S390] Force usage of floating interrupts.
4256 nomio [S390] Do not use MIO instructions.
4257 norid [S390] ignore the RID field and force use of
4258 one PCI domain per PCI function
4260 pcie_aspm= [PCIE] Forcibly enable or disable PCIe Active State Power
4263 force Enable ASPM even on devices that claim not to support it.
4264 WARNING: Forcing ASPM on may cause system lockups.
4266 pcie_ports= [PCIE] PCIe port services handling:
4267 native Use native PCIe services (PME, AER, DPC, PCIe hotplug)
4268 even if the platform doesn't give the OS permission to
4269 use them. This may cause conflicts if the platform
4270 also tries to use these services.
4271 dpc-native Use native PCIe service for DPC only. May
4272 cause conflicts if firmware uses AER or DPC.
4273 compat Disable native PCIe services (PME, AER, DPC, PCIe
4276 pcie_port_pm= [PCIE] PCIe port power management handling:
4277 off Disable power management of all PCIe ports
4278 force Forcibly enable power management of all PCIe ports
4280 pcie_pme= [PCIE,PM] Native PCIe PME signaling options:
4281 nomsi Do not use MSI for native PCIe PME signaling (this makes
4282 all PCIe root ports use INTx for all services).
4284 pcmv= [HW,PCMCIA] BadgePAD 4
4288 Keep all power-domains already enabled by bootloader on,
4289 even if no driver has claimed them. This is useful
4290 for debug and development, but should not be
4291 needed on a platform with proper driver support.
4294 See Documentation/admin-guide/blockdev/paride.rst.
4296 pdcchassis= [PARISC,HW] Disable/Enable PDC Chassis Status codes at
4299 See arch/parisc/kernel/pdc_chassis.c
4301 percpu_alloc= Select which percpu first chunk allocator to use.
4302 Currently supported values are "embed" and "page".
4303 Archs may support subset or none of the selections.
4304 See comments in mm/percpu.c for details on each
4305 allocator. This parameter is primarily for debugging
4306 and performance comparison.
4309 See Documentation/admin-guide/blockdev/paride.rst.
4312 See Documentation/admin-guide/blockdev/paride.rst.
4314 pirq= [SMP,APIC] Manual mp-table setup
4315 See Documentation/x86/i386/IO-APIC.rst.
4317 plip= [PPT,NET] Parallel port network link
4318 Format: { parport<nr> | timid | 0 }
4319 See also Documentation/admin-guide/parport.rst.
4321 pmtmr= [X86] Manual setup of pmtmr I/O Port.
4322 Override pmtimer IOPort with a hex value.
4325 pmu_override= [PPC] Override the PMU.
4326 This option takes over the PMU facility, so it is no
4327 longer usable by perf. Setting this option starts the
4328 PMU counters by setting MMCR0 to 0 (the FC bit is
4329 cleared). If a number is given, then MMCR1 is set to
4330 that number, otherwise (e.g., 'pmu_override=on'), MMCR1
4333 pm_debug_messages [SUSPEND,KNL]
4334 Enable suspend/resume debug messages during boot up.
4337 Enable PNP debug messages (depends on the
4338 CONFIG_PNP_DEBUG_MESSAGES option). Change at run-time
4339 via /sys/module/pnp/parameters/debug. We always show
4340 current resource usage; turning this on also shows
4341 possible settings and some assignment information.
4347 { on | off | curr | res | no-curr | no-res }
4350 [ISAPNP] Exclude IRQs for the autoconfiguration
4353 [ISAPNP] Exclude DMAs for the autoconfiguration
4355 pnp_reserve_io= [ISAPNP] Exclude I/O ports for the autoconfiguration
4356 Ranges are in pairs (I/O port base and size).
4359 [ISAPNP] Exclude memory regions for the
4361 Ranges are in pairs (memory base and size).
4363 ports= [IP_VS_FTP] IPVS ftp helper module
4365 Up to 8 (IP_VS_APP_MAX_PORTS) ports
4367 Format: <port>,<port>....
4369 powersave=off [PPC] This option disables power saving features.
4370 It specifically disables cpuidle and sets the
4371 platform machine description specific power_save
4372 function to NULL. On Idle the CPU just reduces
4375 ppc_strict_facility_enable
4376 [PPC] This option catches any kernel floating point,
4377 Altivec, VSX and SPE outside of regions specifically
4378 allowed (eg kernel_enable_fpu()/kernel_disable_fpu()).
4379 There is some performance impact when enabling this.
4383 Disable Hardware Transactional Memory
4386 Select preemption mode if you have CONFIG_PREEMPT_DYNAMIC
4387 none - Limited to cond_resched() calls
4388 voluntary - Limited to cond_resched() and might_sleep() calls
4389 full - Any section that isn't explicitly preempt disabled
4390 can be preempted anytime.
4392 print-fatal-signals=
4393 [KNL] debug: print fatal signals
4395 If enabled, warn about various signal handling
4396 related application anomalies: too many signals,
4397 too many POSIX.1 timers, fatal signals causing a
4400 If you hit the warning due to signal overflow,
4401 you might want to try "ulimit -i unlimited".
4405 printk.always_kmsg_dump=
4406 Trigger kmsg_dump for cases other than kernel oops or
4408 Format: <bool> (1/Y/y=enable, 0/N/n=disable)
4411 printk.console_no_auto_verbose=
4412 Disable console loglevel raise on oops, panic
4413 or lockdep-detected issues (only if lock debug is on).
4414 With an exception to setups with low baudrate on
4415 serial console, keeping this 0 is a good choice
4416 in order to provide more debug information.
4418 default: 0 (auto_verbose is enabled)
4420 printk.devkmsg={on,off,ratelimit}
4421 Control writing to /dev/kmsg.
4422 on - unlimited logging to /dev/kmsg from userspace
4423 off - logging to /dev/kmsg disabled
4424 ratelimit - ratelimit the logging
4427 printk.time= Show timing data prefixed to each printk message line
4428 Format: <bool> (1/Y/y=enable, 0/N/n=disable)
4430 processor.max_cstate= [HW,ACPI]
4431 Limit processor to maximum C-state
4432 max_cstate=9 overrides any DMI blacklist limit.
4434 processor.nocst [HW,ACPI]
4435 Ignore the _CST method to determine C-states,
4436 instead using the legacy FADT method
4438 profile= [KNL] Enable kernel profiling via /proc/profile
4439 Format: [<profiletype>,]<number>
4440 Param: <profiletype>: "schedule", "sleep", or "kvm"
4441 [defaults to kernel profiling]
4442 Param: "schedule" - profile schedule points.
4443 Param: "sleep" - profile D-state sleeping (millisecs).
4444 Requires CONFIG_SCHEDSTATS
4445 Param: "kvm" - profile VM exits.
4446 Param: <number> - step/bucket size as a power of 2 for
4447 statistical time based profiling.
4449 prompt_ramdisk= [RAM] [Deprecated]
4451 prot_virt= [S390] enable hosting protected virtual machines
4452 isolated from the hypervisor (if hardware supports
4456 psi= [KNL] Enable or disable pressure stall information
4460 psmouse.proto= [HW,MOUSE] Highest PS2 mouse protocol extension to
4461 probe for; one of (bare|imps|exps|lifebook|any).
4462 psmouse.rate= [HW,MOUSE] Set desired mouse report rate, in reports
4464 psmouse.resetafter= [HW,MOUSE]
4465 Try to reset the device after so many bad packets
4468 [HW,MOUSE] Set desired mouse resolution, in dpi.
4469 psmouse.smartscroll=
4470 [HW,MOUSE] Controls Logitech smartscroll autorepeat.
4471 0 = disabled, 1 = enabled (default).
4473 pstore.backend= Specify the name of the pstore backend to use
4476 See Documentation/admin-guide/blockdev/paride.rst.
4478 pti= [X86-64] Control Page Table Isolation of user and
4479 kernel address spaces. Disabling this feature
4480 removes hardening, but improves performance of
4481 system calls and interrupts.
4483 on - unconditionally enable
4484 off - unconditionally disable
4485 auto - kernel detects whether your CPU model is
4486 vulnerable to issues that PTI mitigates
4488 Not specifying this option is equivalent to pti=auto.
4491 Equivalent to pti=off
4494 [KNL] Number of legacy pty's. Overwrites compiled-in
4497 quiet [KNL] Disable most log messages
4502 See Documentation/admin-guide/md.rst.
4504 ramdisk_size= [RAM] Sizes of RAM disks in kilobytes
4505 See Documentation/admin-guide/blockdev/ramdisk.rst.
4507 ramdisk_start= [RAM] RAM disk image start address
4509 random.trust_cpu={on,off}
4510 [KNL] Enable or disable trusting the use of the
4511 CPU's random number generator (if available) to
4512 fully seed the kernel's CRNG. Default is controlled
4513 by CONFIG_RANDOM_TRUST_CPU.
4515 random.trust_bootloader={on,off}
4516 [KNL] Enable or disable trusting the use of a
4517 seed passed by the bootloader (if available) to
4518 fully seed the kernel's CRNG. Default is controlled
4519 by CONFIG_RANDOM_TRUST_BOOTLOADER.
4521 randomize_kstack_offset=
4522 [KNL] Enable or disable kernel stack offset
4523 randomization, which provides roughly 5 bits of
4524 entropy, frustrating memory corruption attacks
4525 that depend on stack address determinism or
4526 cross-syscall address exposures. This is only
4527 available on architectures that have defined
4528 CONFIG_HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET.
4529 Format: <bool> (1/Y/y=enable, 0/N/n=disable)
4530 Default is CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT.
4532 ras=option[,option,...] [KNL] RAS-specific options
4535 Disable the Correctable Errors Collector,
4536 see CONFIG_RAS_CEC help text.
4538 rcu_nocbs[=cpu-list]
4539 [KNL] The optional argument is a cpu list,
4542 In kernels built with CONFIG_RCU_NOCB_CPU=y,
4543 enable the no-callback CPU mode, which prevents
4544 such CPUs' callbacks from being invoked in
4545 softirq context. Invocation of such CPUs' RCU
4546 callbacks will instead be offloaded to "rcuox/N"
4547 kthreads created for that purpose, where "x" is
4548 "p" for RCU-preempt, "s" for RCU-sched, and "g"
4549 for the kthreads that mediate grace periods; and
4550 "N" is the CPU number. This reduces OS jitter on
4551 the offloaded CPUs, which can be useful for HPC
4552 and real-time workloads. It can also improve
4553 energy efficiency for asymmetric multiprocessors.
4555 If a cpulist is passed as an argument, the specified
4556 list of CPUs is set to no-callback mode from boot.
4558 Otherwise, if the '=' sign and the cpulist
4559 arguments are omitted, no CPU will be set to
4560 no-callback mode from boot but the mode may be
4561 toggled at runtime via cpusets.
4564 Rather than requiring that offloaded CPUs
4565 (specified by rcu_nocbs= above) explicitly
4566 awaken the corresponding "rcuoN" kthreads,
4567 make these kthreads poll for callbacks.
4568 This improves the real-time response for the
4569 offloaded CPUs by relieving them of the need to
4570 wake up the corresponding kthread, but degrades
4571 energy efficiency by requiring that the kthreads
4572 periodically wake up to do the polling.
4574 rcutree.blimit= [KNL]
4575 Set maximum number of finished RCU callbacks to
4576 process in one batch.
4578 rcutree.dump_tree= [KNL]
4579 Dump the structure of the rcu_node combining tree
4580 out at early boot. This is used for diagnostic
4581 purposes, to verify correct tree setup.
4583 rcutree.gp_cleanup_delay= [KNL]
4584 Set the number of jiffies to delay each step of
4585 RCU grace-period cleanup.
4587 rcutree.gp_init_delay= [KNL]
4588 Set the number of jiffies to delay each step of
4589 RCU grace-period initialization.
4591 rcutree.gp_preinit_delay= [KNL]
4592 Set the number of jiffies to delay each step of
4593 RCU grace-period pre-initialization, that is,
4594 the propagation of recent CPU-hotplug changes up
4595 the rcu_node combining tree.
4597 rcutree.use_softirq= [KNL]
4598 If set to zero, move all RCU_SOFTIRQ processing to
4599 per-CPU rcuc kthreads. Defaults to a non-zero
4600 value, meaning that RCU_SOFTIRQ is used by default.
4601 Specify rcutree.use_softirq=0 to use rcuc kthreads.
4603 But note that CONFIG_PREEMPT_RT=y kernels disable
4604 this kernel boot parameter, forcibly setting it
4607 rcutree.rcu_fanout_exact= [KNL]
4608 Disable autobalancing of the rcu_node combining
4609 tree. This is used by rcutorture, and might
4610 possibly be useful for architectures having high
4611 cache-to-cache transfer latencies.
4613 rcutree.rcu_fanout_leaf= [KNL]
4614 Change the number of CPUs assigned to each
4615 leaf rcu_node structure. Useful for very
4616 large systems, which will choose the value 64,
4617 and for NUMA systems with large remote-access
4618 latencies, which will choose a value aligned
4619 with the appropriate hardware boundaries.
4621 rcutree.rcu_min_cached_objs= [KNL]
4622 Minimum number of objects which are cached and
4623 maintained per one CPU. Object size is equal
4624 to PAGE_SIZE. The cache allows to reduce the
4625 pressure to page allocator, also it makes the
4626 whole algorithm to behave better in low memory
4629 rcutree.rcu_delay_page_cache_fill_msec= [KNL]
4630 Set the page-cache refill delay (in milliseconds)
4631 in response to low-memory conditions. The range
4632 of permitted values is in the range 0:100000.
4634 rcutree.jiffies_till_first_fqs= [KNL]
4635 Set delay from grace-period initialization to
4636 first attempt to force quiescent states.
4637 Units are jiffies, minimum value is zero,
4638 and maximum value is HZ.
4640 rcutree.jiffies_till_next_fqs= [KNL]
4641 Set delay between subsequent attempts to force
4642 quiescent states. Units are jiffies, minimum
4643 value is one, and maximum value is HZ.
4645 rcutree.jiffies_till_sched_qs= [KNL]
4646 Set required age in jiffies for a
4647 given grace period before RCU starts
4648 soliciting quiescent-state help from
4649 rcu_note_context_switch() and cond_resched().
4650 If not specified, the kernel will calculate
4651 a value based on the most recent settings
4652 of rcutree.jiffies_till_first_fqs
4653 and rcutree.jiffies_till_next_fqs.
4654 This calculated value may be viewed in
4655 rcutree.jiffies_to_sched_qs. Any attempt to set
4656 rcutree.jiffies_to_sched_qs will be cheerfully
4659 rcutree.kthread_prio= [KNL,BOOT]
4660 Set the SCHED_FIFO priority of the RCU per-CPU
4661 kthreads (rcuc/N). This value is also used for
4662 the priority of the RCU boost threads (rcub/N)
4663 and for the RCU grace-period kthreads (rcu_bh,
4664 rcu_preempt, and rcu_sched). If RCU_BOOST is
4665 set, valid values are 1-99 and the default is 1
4666 (the least-favored priority). Otherwise, when
4667 RCU_BOOST is not set, valid values are 0-99 and
4668 the default is zero (non-realtime operation).
4669 When RCU_NOCB_CPU is set, also adjust the
4670 priority of NOCB callback kthreads.
4672 rcutree.rcu_nocb_gp_stride= [KNL]
4673 Set the number of NOCB callback kthreads in
4674 each group, which defaults to the square root
4675 of the number of CPUs. Larger numbers reduce
4676 the wakeup overhead on the global grace-period
4677 kthread, but increases that same overhead on
4678 each group's NOCB grace-period kthread.
4680 rcutree.qhimark= [KNL]
4681 Set threshold of queued RCU callbacks beyond which
4682 batch limiting is disabled.
4684 rcutree.qlowmark= [KNL]
4685 Set threshold of queued RCU callbacks below which
4686 batch limiting is re-enabled.
4688 rcutree.qovld= [KNL]
4689 Set threshold of queued RCU callbacks beyond which
4690 RCU's force-quiescent-state scan will aggressively
4691 enlist help from cond_resched() and sched IPIs to
4692 help CPUs more quickly reach quiescent states.
4693 Set to less than zero to make this be set based
4694 on rcutree.qhimark at boot time and to zero to
4695 disable more aggressive help enlistment.
4697 rcutree.rcu_kick_kthreads= [KNL]
4698 Cause the grace-period kthread to get an extra
4699 wake_up() if it sleeps three times longer than
4700 it should at force-quiescent-state time.
4701 This wake_up() will be accompanied by a
4702 WARN_ONCE() splat and an ftrace_dump().
4704 rcutree.rcu_unlock_delay= [KNL]
4705 In CONFIG_RCU_STRICT_GRACE_PERIOD=y kernels,
4706 this specifies an rcu_read_unlock()-time delay
4707 in microseconds. This defaults to zero.
4708 Larger delays increase the probability of
4709 catching RCU pointer leaks, that is, buggy use
4710 of RCU-protected pointers after the relevant
4711 rcu_read_unlock() has completed.
4713 rcutree.sysrq_rcu= [KNL]
4714 Commandeer a sysrq key to dump out Tree RCU's
4715 rcu_node tree with an eye towards determining
4716 why a new grace period has not yet started.
4718 rcuscale.gp_async= [KNL]
4719 Measure performance of asynchronous
4720 grace-period primitives such as call_rcu().
4722 rcuscale.gp_async_max= [KNL]
4723 Specify the maximum number of outstanding
4724 callbacks per writer thread. When a writer
4725 thread exceeds this limit, it invokes the
4726 corresponding flavor of rcu_barrier() to allow
4727 previously posted callbacks to drain.
4729 rcuscale.gp_exp= [KNL]
4730 Measure performance of expedited synchronous
4731 grace-period primitives.
4733 rcuscale.holdoff= [KNL]
4734 Set test-start holdoff period. The purpose of
4735 this parameter is to delay the start of the
4736 test until boot completes in order to avoid
4739 rcuscale.kfree_rcu_test= [KNL]
4740 Set to measure performance of kfree_rcu() flooding.
4742 rcuscale.kfree_rcu_test_double= [KNL]
4743 Test the double-argument variant of kfree_rcu().
4744 If this parameter has the same value as
4745 rcuscale.kfree_rcu_test_single, both the single-
4746 and double-argument variants are tested.
4748 rcuscale.kfree_rcu_test_single= [KNL]
4749 Test the single-argument variant of kfree_rcu().
4750 If this parameter has the same value as
4751 rcuscale.kfree_rcu_test_double, both the single-
4752 and double-argument variants are tested.
4754 rcuscale.kfree_nthreads= [KNL]
4755 The number of threads running loops of kfree_rcu().
4757 rcuscale.kfree_alloc_num= [KNL]
4758 Number of allocations and frees done in an iteration.
4760 rcuscale.kfree_loops= [KNL]
4761 Number of loops doing rcuscale.kfree_alloc_num number
4762 of allocations and frees.
4764 rcuscale.nreaders= [KNL]
4765 Set number of RCU readers. The value -1 selects
4766 N, where N is the number of CPUs. A value
4767 "n" less than -1 selects N-n+1, where N is again
4768 the number of CPUs. For example, -2 selects N
4769 (the number of CPUs), -3 selects N+1, and so on.
4770 A value of "n" less than or equal to -N selects
4773 rcuscale.nwriters= [KNL]
4774 Set number of RCU writers. The values operate
4775 the same as for rcuscale.nreaders.
4776 N, where N is the number of CPUs
4778 rcuscale.perf_type= [KNL]
4779 Specify the RCU implementation to test.
4781 rcuscale.shutdown= [KNL]
4782 Shut the system down after performance tests
4783 complete. This is useful for hands-off automated
4786 rcuscale.verbose= [KNL]
4787 Enable additional printk() statements.
4789 rcuscale.writer_holdoff= [KNL]
4790 Write-side holdoff between grace periods,
4791 in microseconds. The default of zero says
4794 rcutorture.fqs_duration= [KNL]
4795 Set duration of force_quiescent_state bursts
4798 rcutorture.fqs_holdoff= [KNL]
4799 Set holdoff time within force_quiescent_state bursts
4802 rcutorture.fqs_stutter= [KNL]
4803 Set wait time between force_quiescent_state bursts
4806 rcutorture.fwd_progress= [KNL]
4807 Specifies the number of kthreads to be used
4808 for RCU grace-period forward-progress testing
4809 for the types of RCU supporting this notion.
4810 Defaults to 1 kthread, values less than zero or
4811 greater than the number of CPUs cause the number
4814 rcutorture.fwd_progress_div= [KNL]
4815 Specify the fraction of a CPU-stall-warning
4816 period to do tight-loop forward-progress testing.
4818 rcutorture.fwd_progress_holdoff= [KNL]
4819 Number of seconds to wait between successive
4820 forward-progress tests.
4822 rcutorture.fwd_progress_need_resched= [KNL]
4823 Enclose cond_resched() calls within checks for
4824 need_resched() during tight-loop forward-progress
4827 rcutorture.gp_cond= [KNL]
4828 Use conditional/asynchronous update-side
4829 primitives, if available.
4831 rcutorture.gp_exp= [KNL]
4832 Use expedited update-side primitives, if available.
4834 rcutorture.gp_normal= [KNL]
4835 Use normal (non-expedited) asynchronous
4836 update-side primitives, if available.
4838 rcutorture.gp_sync= [KNL]
4839 Use normal (non-expedited) synchronous
4840 update-side primitives, if available. If all
4841 of rcutorture.gp_cond=, rcutorture.gp_exp=,
4842 rcutorture.gp_normal=, and rcutorture.gp_sync=
4843 are zero, rcutorture acts as if is interpreted
4844 they are all non-zero.
4846 rcutorture.irqreader= [KNL]
4847 Run RCU readers from irq handlers, or, more
4848 accurately, from a timer handler. Not all RCU
4849 flavors take kindly to this sort of thing.
4851 rcutorture.leakpointer= [KNL]
4852 Leak an RCU-protected pointer out of the reader.
4853 This can of course result in splats, and is
4854 intended to test the ability of things like
4855 CONFIG_RCU_STRICT_GRACE_PERIOD=y to detect
4858 rcutorture.n_barrier_cbs= [KNL]
4859 Set callbacks/threads for rcu_barrier() testing.
4861 rcutorture.nfakewriters= [KNL]
4862 Set number of concurrent RCU writers. These just
4863 stress RCU, they don't participate in the actual
4864 test, hence the "fake".
4866 rcutorture.nocbs_nthreads= [KNL]
4867 Set number of RCU callback-offload togglers.
4868 Zero (the default) disables toggling.
4870 rcutorture.nocbs_toggle= [KNL]
4871 Set the delay in milliseconds between successive
4872 callback-offload toggling attempts.
4874 rcutorture.nreaders= [KNL]
4875 Set number of RCU readers. The value -1 selects
4876 N-1, where N is the number of CPUs. A value
4877 "n" less than -1 selects N-n-2, where N is again
4878 the number of CPUs. For example, -2 selects N
4879 (the number of CPUs), -3 selects N+1, and so on.
4881 rcutorture.object_debug= [KNL]
4882 Enable debug-object double-call_rcu() testing.
4884 rcutorture.onoff_holdoff= [KNL]
4885 Set time (s) after boot for CPU-hotplug testing.
4887 rcutorture.onoff_interval= [KNL]
4888 Set time (jiffies) between CPU-hotplug operations,
4889 or zero to disable CPU-hotplug testing.
4891 rcutorture.read_exit= [KNL]
4892 Set the number of read-then-exit kthreads used
4893 to test the interaction of RCU updaters and
4894 task-exit processing.
4896 rcutorture.read_exit_burst= [KNL]
4897 The number of times in a given read-then-exit
4898 episode that a set of read-then-exit kthreads
4901 rcutorture.read_exit_delay= [KNL]
4902 The delay, in seconds, between successive
4903 read-then-exit testing episodes.
4905 rcutorture.shuffle_interval= [KNL]
4906 Set task-shuffle interval (s). Shuffling tasks
4907 allows some CPUs to go into dyntick-idle mode
4908 during the rcutorture test.
4910 rcutorture.shutdown_secs= [KNL]
4911 Set time (s) after boot system shutdown. This
4912 is useful for hands-off automated testing.
4914 rcutorture.stall_cpu= [KNL]
4915 Duration of CPU stall (s) to test RCU CPU stall
4916 warnings, zero to disable.
4918 rcutorture.stall_cpu_block= [KNL]
4919 Sleep while stalling if set. This will result
4920 in warnings from preemptible RCU in addition
4921 to any other stall-related activity.
4923 rcutorture.stall_cpu_holdoff= [KNL]
4924 Time to wait (s) after boot before inducing stall.
4926 rcutorture.stall_cpu_irqsoff= [KNL]
4927 Disable interrupts while stalling if set.
4929 rcutorture.stall_gp_kthread= [KNL]
4930 Duration (s) of forced sleep within RCU
4931 grace-period kthread to test RCU CPU stall
4932 warnings, zero to disable. If both stall_cpu
4933 and stall_gp_kthread are specified, the
4934 kthread is starved first, then the CPU.
4936 rcutorture.stat_interval= [KNL]
4937 Time (s) between statistics printk()s.
4939 rcutorture.stutter= [KNL]
4940 Time (s) to stutter testing, for example, specifying
4941 five seconds causes the test to run for five seconds,
4942 wait for five seconds, and so on. This tests RCU's
4943 ability to transition abruptly to and from idle.
4945 rcutorture.test_boost= [KNL]
4946 Test RCU priority boosting? 0=no, 1=maybe, 2=yes.
4947 "Maybe" means test if the RCU implementation
4948 under test support RCU priority boosting.
4950 rcutorture.test_boost_duration= [KNL]
4951 Duration (s) of each individual boost test.
4953 rcutorture.test_boost_interval= [KNL]
4954 Interval (s) between each boost test.
4956 rcutorture.test_no_idle_hz= [KNL]
4957 Test RCU's dyntick-idle handling. See also the
4958 rcutorture.shuffle_interval parameter.
4960 rcutorture.torture_type= [KNL]
4961 Specify the RCU implementation to test.
4963 rcutorture.verbose= [KNL]
4964 Enable additional printk() statements.
4966 rcupdate.rcu_cpu_stall_ftrace_dump= [KNL]
4967 Dump ftrace buffer after reporting RCU CPU
4970 rcupdate.rcu_cpu_stall_suppress= [KNL]
4971 Suppress RCU CPU stall warning messages.
4973 rcupdate.rcu_cpu_stall_suppress_at_boot= [KNL]
4974 Suppress RCU CPU stall warning messages and
4975 rcutorture writer stall warnings that occur
4976 during early boot, that is, during the time
4977 before the init task is spawned.
4979 rcupdate.rcu_cpu_stall_timeout= [KNL]
4980 Set timeout for RCU CPU stall warning messages.
4981 The value is in seconds and the maximum allowed
4982 value is 300 seconds.
4984 rcupdate.rcu_exp_cpu_stall_timeout= [KNL]
4985 Set timeout for expedited RCU CPU stall warning
4986 messages. The value is in milliseconds
4987 and the maximum allowed value is 21000
4988 milliseconds. Please note that this value is
4989 adjusted to an arch timer tick resolution.
4990 Setting this to zero causes the value from
4991 rcupdate.rcu_cpu_stall_timeout to be used (after
4992 conversion from seconds to milliseconds).
4994 rcupdate.rcu_expedited= [KNL]
4995 Use expedited grace-period primitives, for
4996 example, synchronize_rcu_expedited() instead
4997 of synchronize_rcu(). This reduces latency,
4998 but can increase CPU utilization, degrade
4999 real-time latency, and degrade energy efficiency.
5000 No effect on CONFIG_TINY_RCU kernels.
5002 rcupdate.rcu_normal= [KNL]
5003 Use only normal grace-period primitives,
5004 for example, synchronize_rcu() instead of
5005 synchronize_rcu_expedited(). This improves
5006 real-time latency, CPU utilization, and
5007 energy efficiency, but can expose users to
5008 increased grace-period latency. This parameter
5009 overrides rcupdate.rcu_expedited. No effect on
5010 CONFIG_TINY_RCU kernels.
5012 rcupdate.rcu_normal_after_boot= [KNL]
5013 Once boot has completed (that is, after
5014 rcu_end_inkernel_boot() has been invoked), use
5015 only normal grace-period primitives. No effect
5016 on CONFIG_TINY_RCU kernels.
5018 But note that CONFIG_PREEMPT_RT=y kernels enables
5019 this kernel boot parameter, forcibly setting
5020 it to the value one, that is, converting any
5021 post-boot attempt at an expedited RCU grace
5022 period to instead use normal non-expedited
5023 grace-period processing.
5025 rcupdate.rcu_task_collapse_lim= [KNL]
5026 Set the maximum number of callbacks present
5027 at the beginning of a grace period that allows
5028 the RCU Tasks flavors to collapse back to using
5029 a single callback queue. This switching only
5030 occurs when rcupdate.rcu_task_enqueue_lim is
5031 set to the default value of -1.
5033 rcupdate.rcu_task_contend_lim= [KNL]
5034 Set the minimum number of callback-queuing-time
5035 lock-contention events per jiffy required to
5036 cause the RCU Tasks flavors to switch to per-CPU
5037 callback queuing. This switching only occurs
5038 when rcupdate.rcu_task_enqueue_lim is set to
5039 the default value of -1.
5041 rcupdate.rcu_task_enqueue_lim= [KNL]
5042 Set the number of callback queues to use for the
5043 RCU Tasks family of RCU flavors. The default
5044 of -1 allows this to be automatically (and
5045 dynamically) adjusted. This parameter is intended
5048 rcupdate.rcu_task_ipi_delay= [KNL]
5049 Set time in jiffies during which RCU tasks will
5050 avoid sending IPIs, starting with the beginning
5051 of a given grace period. Setting a large
5052 number avoids disturbing real-time workloads,
5053 but lengthens grace periods.
5055 rcupdate.rcu_task_stall_info= [KNL]
5056 Set initial timeout in jiffies for RCU task stall
5057 informational messages, which give some indication
5058 of the problem for those not patient enough to
5059 wait for ten minutes. Informational messages are
5060 only printed prior to the stall-warning message
5061 for a given grace period. Disable with a value
5062 less than or equal to zero. Defaults to ten
5063 seconds. A change in value does not take effect
5064 until the beginning of the next grace period.
5066 rcupdate.rcu_task_stall_info_mult= [KNL]
5067 Multiplier for time interval between successive
5068 RCU task stall informational messages for a given
5069 RCU tasks grace period. This value is clamped
5070 to one through ten, inclusive. It defaults to
5071 the value three, so that the first informational
5072 message is printed 10 seconds into the grace
5073 period, the second at 40 seconds, the third at
5074 160 seconds, and then the stall warning at 600
5075 seconds would prevent a fourth at 640 seconds.
5077 rcupdate.rcu_task_stall_timeout= [KNL]
5078 Set timeout in jiffies for RCU task stall
5079 warning messages. Disable with a value less
5080 than or equal to zero. Defaults to ten minutes.
5081 A change in value does not take effect until
5082 the beginning of the next grace period.
5084 rcupdate.rcu_self_test= [KNL]
5085 Run the RCU early boot self tests
5089 Run specified binary instead of /init from the ramdisk,
5090 used for early userspace startup. See initrd.
5093 force - Override the decision by the kernel to hide the
5094 advertisement of RDRAND support (this affects
5095 certain AMD processors because of buggy BIOS
5096 support, specifically around the suspend/resume
5100 Turn on/off individual RDT features. List is:
5101 cmt, mbmtotal, mbmlocal, l3cat, l3cdp, l2cat, l2cdp,
5103 E.g. to turn on cmt and turn off mba use:
5107 Format (x86 or x86_64):
5108 [w[arm] | c[old] | h[ard] | s[oft] | g[pio]] | d[efault] \
5110 [[,]b[ios] | a[cpi] | k[bd] | t[riple] | e[fi] | p[ci]] \
5112 Where reboot_mode is one of warm (soft) or cold (hard) or gpio
5113 (prefix with 'panic_' to set mode for panic
5115 reboot_type is one of bios, acpi, kbd, triple, efi, or pci,
5116 reboot_force is either force or not specified,
5117 reboot_cpu is s[mp]#### with #### being the processor
5118 to be used for rebooting.
5120 refscale.holdoff= [KNL]
5121 Set test-start holdoff period. The purpose of
5122 this parameter is to delay the start of the
5123 test until boot completes in order to avoid
5126 refscale.loops= [KNL]
5127 Set the number of loops over the synchronization
5128 primitive under test. Increasing this number
5129 reduces noise due to loop start/end overhead,
5130 but the default has already reduced the per-pass
5131 noise to a handful of picoseconds on ca. 2020
5134 refscale.nreaders= [KNL]
5135 Set number of readers. The default value of -1
5136 selects N, where N is roughly 75% of the number
5137 of CPUs. A value of zero is an interesting choice.
5139 refscale.nruns= [KNL]
5140 Set number of runs, each of which is dumped onto
5143 refscale.readdelay= [KNL]
5144 Set the read-side critical-section duration,
5145 measured in microseconds.
5147 refscale.scale_type= [KNL]
5148 Specify the read-protection implementation to test.
5150 refscale.shutdown= [KNL]
5151 Shut down the system at the end of the performance
5152 test. This defaults to 1 (shut it down) when
5153 refscale is built into the kernel and to 0 (leave
5154 it running) when refscale is built as a module.
5156 refscale.verbose= [KNL]
5157 Enable additional printk() statements.
5159 refscale.verbose_batched= [KNL]
5160 Batch the additional printk() statements. If zero
5161 (the default) or negative, print everything. Otherwise,
5162 print every Nth verbose statement, where N is the value
5166 [KNL, SMP] Set scheduler's default relax_domain_level.
5167 See Documentation/admin-guide/cgroup-v1/cpusets.rst.
5169 reserve= [KNL,BUGS] Force kernel to ignore I/O ports or memory
5170 Format: <base1>,<size1>[,<base2>,<size2>,...]
5171 Reserve I/O ports or memory so the kernel won't use
5172 them. If <base> is less than 0x10000, the region
5173 is assumed to be I/O ports; otherwise it is memory.
5175 reservetop= [X86-32]
5177 Reserves a hole at the top of the kernel virtual
5180 reset_devices [KNL] Force drivers to reset the underlying device
5181 during initialization.
5184 Specify the partition device for software suspend
5186 {/dev/<dev> | PARTUUID=<uuid> | <int>:<int> | <hex>}
5188 resume_offset= [SWSUSP]
5189 Specify the offset from the beginning of the partition
5190 given by "resume=" at which the swap header is located,
5191 in <PAGE_SIZE> units (needed only for swap files).
5192 See Documentation/power/swsusp-and-swap-files.rst
5194 resumedelay= [HIBERNATION] Delay (in seconds) to pause before attempting to
5195 read the resume files
5197 resumewait [HIBERNATION] Wait (indefinitely) for resume device to show up.
5198 Useful for devices that are detected asynchronously
5199 (e.g. USB and MMC devices).
5201 retain_initrd [RAM] Keep initrd memory after extraction
5203 retbleed= [X86] Control mitigation of RETBleed (Arbitrary
5204 Speculative Code Execution with Return Instructions)
5208 auto - automatically select a migitation
5209 auto,nosmt - automatically select a mitigation,
5210 disabling SMT if necessary for
5211 the full mitigation (only on Zen1
5212 and older without STIBP).
5213 ibpb - mitigate short speculation windows on
5214 basic block boundaries too. Safe, highest
5216 unret - force enable untrained return thunks,
5217 only effective on AMD f15h-f17h
5219 unret,nosmt - like unret, will disable SMT when STIBP
5222 Selecting 'auto' will choose a mitigation method at run
5223 time according to the CPU.
5225 Not specifying this option is equivalent to retbleed=auto.
5227 rfkill.default_state=
5228 0 "airplane mode". All wifi, bluetooth, wimax, gps, fm,
5229 etc. communication is blocked by default.
5232 rfkill.master_switch_mode=
5233 0 The "airplane mode" button does nothing.
5234 1 The "airplane mode" button toggles between everything
5235 blocked and the previous configuration.
5236 2 The "airplane mode" button toggles between everything
5237 blocked and everything unblocked.
5239 rhash_entries= [KNL,NET]
5240 Set number of hash buckets for route cache
5243 [KNL] Disable ring 3 MONITOR/MWAIT feature on supported
5246 ro [KNL] Mount root device read-only on boot
5249 on Mark read-only kernel memory as read-only (default).
5250 off Leave read-only kernel memory writable for debugging.
5253 Enable the uart passthrough on the designated usb port
5254 on Rockchip SoCs. When active, the signals of the
5255 debug-uart get routed to the D+ and D- pins of the usb
5256 port and the regular usb controller gets disabled.
5258 root= [KNL] Root filesystem
5259 See name_to_dev_t comment in init/do_mounts.c.
5261 rootdelay= [KNL] Delay (in seconds) to pause before attempting to
5262 mount the root filesystem
5264 rootflags= [KNL] Set root filesystem mount option string
5266 rootfstype= [KNL] Set root filesystem type
5268 rootwait [KNL] Wait (indefinitely) for root device to show up.
5269 Useful for devices that are detected asynchronously
5270 (e.g. USB and MMC devices).
5272 rproc_mem=nn[KMG][@address]
5273 [KNL,ARM,CMA] Remoteproc physical memory block.
5274 Memory area to be used by remote processor image,
5277 rw [KNL] Mount root device read-write on boot
5279 S [KNL] Run init in single mode
5281 s390_iommu= [HW,S390]
5282 Set s390 IOTLB flushing mode
5284 With strict flushing every unmap operation will result in
5285 an IOTLB flush. Default is lazy flushing before reuse,
5288 s390_iommu_aperture= [KNL,S390]
5289 Specifies the size of the per device DMA address space
5290 accessible through the DMA and IOMMU APIs as a decimal
5291 factor of the size of main memory.
5292 The default is 1 meaning that one can concurrently use
5293 as many DMA addresses as physical memory is installed,
5294 if supported by hardware, and thus map all of memory
5295 once. With a value of 2 one can map all of memory twice
5296 and so on. As a special case a factor of 0 imposes no
5297 restrictions other than those given by hardware at the
5298 cost of significant additional memory use for tables.
5301 See drivers/net/irda/sa1100_ir.c.
5303 sched_verbose [KNL] Enables verbose scheduler debug messages.
5305 schedstats= [KNL,X86] Enable or disable scheduled statistics.
5306 Allowed values are enable and disable. This feature
5307 incurs a small amount of overhead in the scheduler
5308 but is useful for debugging and performance tuning.
5310 sched_thermal_decay_shift=
5311 [KNL, SMP] Set a decay shift for scheduler thermal
5312 pressure signal. Thermal pressure signal follows the
5313 default decay period of other scheduler pelt
5314 signals(usually 32 ms but configurable). Setting
5315 sched_thermal_decay_shift will left shift the decay
5316 period for the thermal pressure signal by the shift
5318 i.e. with the default pelt decay period of 32 ms
5319 sched_thermal_decay_shift thermal pressure decay pr
5323 Format: integer between 0 and 10
5326 scftorture.holdoff= [KNL]
5327 Number of seconds to hold off before starting
5328 test. Defaults to zero for module insertion and
5329 to 10 seconds for built-in smp_call_function()
5332 scftorture.longwait= [KNL]
5333 Request ridiculously long waits randomly selected
5334 up to the chosen limit in seconds. Zero (the
5335 default) disables this feature. Please note
5336 that requesting even small non-zero numbers of
5337 seconds can result in RCU CPU stall warnings,
5338 softlockup complaints, and so on.
5340 scftorture.nthreads= [KNL]
5341 Number of kthreads to spawn to invoke the
5342 smp_call_function() family of functions.
5343 The default of -1 specifies a number of kthreads
5344 equal to the number of CPUs.
5346 scftorture.onoff_holdoff= [KNL]
5347 Number seconds to wait after the start of the
5348 test before initiating CPU-hotplug operations.
5350 scftorture.onoff_interval= [KNL]
5351 Number seconds to wait between successive
5352 CPU-hotplug operations. Specifying zero (which
5353 is the default) disables CPU-hotplug operations.
5355 scftorture.shutdown_secs= [KNL]
5356 The number of seconds following the start of the
5357 test after which to shut down the system. The
5358 default of zero avoids shutting down the system.
5359 Non-zero values are useful for automated tests.
5361 scftorture.stat_interval= [KNL]
5362 The number of seconds between outputting the
5363 current test statistics to the console. A value
5364 of zero disables statistics output.
5366 scftorture.stutter_cpus= [KNL]
5367 The number of jiffies to wait between each change
5368 to the set of CPUs under test.
5370 scftorture.use_cpus_read_lock= [KNL]
5371 Use use_cpus_read_lock() instead of the default
5372 preempt_disable() to disable CPU hotplug
5373 while invoking one of the smp_call_function*()
5376 scftorture.verbose= [KNL]
5377 Enable additional printk() statements.
5379 scftorture.weight_single= [KNL]
5380 The probability weighting to use for the
5381 smp_call_function_single() function with a zero
5382 "wait" parameter. A value of -1 selects the
5383 default if all other weights are -1. However,
5384 if at least one weight has some other value, a
5385 value of -1 will instead select a weight of zero.
5387 scftorture.weight_single_wait= [KNL]
5388 The probability weighting to use for the
5389 smp_call_function_single() function with a
5390 non-zero "wait" parameter. See weight_single.
5392 scftorture.weight_many= [KNL]
5393 The probability weighting to use for the
5394 smp_call_function_many() function with a zero
5395 "wait" parameter. See weight_single.
5396 Note well that setting a high probability for
5397 this weighting can place serious IPI load
5400 scftorture.weight_many_wait= [KNL]
5401 The probability weighting to use for the
5402 smp_call_function_many() function with a
5403 non-zero "wait" parameter. See weight_single
5406 scftorture.weight_all= [KNL]
5407 The probability weighting to use for the
5408 smp_call_function_all() function with a zero
5409 "wait" parameter. See weight_single and
5412 scftorture.weight_all_wait= [KNL]
5413 The probability weighting to use for the
5414 smp_call_function_all() function with a
5415 non-zero "wait" parameter. See weight_single
5418 skew_tick= [KNL] Offset the periodic timer tick per cpu to mitigate
5419 xtime_lock contention on larger systems, and/or RCU lock
5420 contention on all systems with CONFIG_MAXSMP set.
5421 Format: { "0" | "1" }
5422 0 -- disable. (may be 1 via CONFIG_CMDLINE="skew_tick=1"
5424 Note: increases power consumption, thus should only be
5425 enabled if running jitter sensitive (HPC/RT) workloads.
5427 security= [SECURITY] Choose a legacy "major" security module to
5428 enable at boot. This has been deprecated by the
5431 selinux= [SELINUX] Disable or enable SELinux at boot time.
5432 Format: { "0" | "1" }
5433 See security/selinux/Kconfig help text.
5438 apparmor= [APPARMOR] Disable or enable AppArmor at boot time
5439 Format: { "0" | "1" }
5440 See security/apparmor/Kconfig help text
5443 Default value is set via kernel config option.
5445 serialnumber [BUGS=X86-32]
5447 sev=option[,option...] [X86-64] See Documentation/x86/x86_64/boot-options.rst
5450 Maximal number of shapers.
5458 Enable merging of slabs with similar size when the
5459 kernel is built without CONFIG_SLAB_MERGE_DEFAULT.
5462 Disable merging of slabs with similar size. May be
5463 necessary if there is some reason to distinguish
5464 allocs to different slabs, especially in hardened
5465 environments where the risk of heap overflows and
5466 layout control by attackers can usually be
5467 frustrated by disabling merging. This will reduce
5468 most of the exposure of a heap attack to a single
5469 cache (risks via metadata attacks are mostly
5470 unchanged). Debug options disable merging on their
5472 For more information see Documentation/vm/slub.rst.
5474 slab_max_order= [MM, SLAB]
5475 Determines the maximum allowed order for slabs.
5476 A high setting may cause OOMs due to memory
5477 fragmentation. Defaults to 1 for systems with
5478 more than 32MB of RAM, 0 otherwise.
5480 slub_debug[=options[,slabs][;[options[,slabs]]...] [MM, SLUB]
5481 Enabling slub_debug allows one to determine the
5482 culprit if slab objects become corrupted. Enabling
5483 slub_debug can create guard zones around objects and
5484 may poison objects when not in use. Also tracks the
5485 last alloc / free. For more information see
5486 Documentation/vm/slub.rst.
5488 slub_max_order= [MM, SLUB]
5489 Determines the maximum allowed order for slabs.
5490 A high setting may cause OOMs due to memory
5491 fragmentation. For more information see
5492 Documentation/vm/slub.rst.
5494 slub_min_objects= [MM, SLUB]
5495 The minimum number of objects per slab. SLUB will
5496 increase the slab order up to slub_max_order to
5497 generate a sufficiently large slab able to contain
5498 the number of objects indicated. The higher the number
5499 of objects the smaller the overhead of tracking slabs
5500 and the less frequently locks need to be acquired.
5501 For more information see Documentation/vm/slub.rst.
5503 slub_min_order= [MM, SLUB]
5504 Determines the minimum page order for slabs. Must be
5505 lower than slub_max_order.
5506 For more information see Documentation/vm/slub.rst.
5508 slub_merge [MM, SLUB]
5509 Same with slab_merge.
5511 slub_nomerge [MM, SLUB]
5512 Same with slab_nomerge. This is supported for legacy.
5513 See slab_nomerge for more information.
5516 Format: <io1>[,<io2>[,...,<io8>]]
5518 smp.csd_lock_timeout= [KNL]
5519 Specify the period of time in milliseconds
5520 that smp_call_function() and friends will wait
5521 for a CPU to release the CSD lock. This is
5522 useful when diagnosing bugs involving CPUs
5523 disabling interrupts for extended periods
5524 of time. Defaults to 5,000 milliseconds, and
5525 setting a value of zero disables this feature.
5526 This feature may be more efficiently disabled
5527 using the csdlock_debug- kernel parameter.
5529 smsc-ircc2.nopnp [HW] Don't use PNP to discover SMC devices
5530 smsc-ircc2.ircc_cfg= [HW] Device configuration I/O port
5531 smsc-ircc2.ircc_sir= [HW] SIR base I/O port
5532 smsc-ircc2.ircc_fir= [HW] FIR base I/O port
5533 smsc-ircc2.ircc_irq= [HW] IRQ line
5534 smsc-ircc2.ircc_dma= [HW] DMA channel
5535 smsc-ircc2.ircc_transceiver= [HW] Transceiver type:
5536 0: Toshiba Satellite 1800 (GP data pin select)
5537 1: Fast pin select (default)
5540 smt= [KNL,S390] Set the maximum number of threads (logical
5541 CPUs) to use per physical CPU on systems capable of
5542 symmetric multithreading (SMT). Will be capped to the
5543 actual hardware limit.
5545 Default: -1 (no limit)
5548 [KNL] Should the soft-lockup detector generate panics.
5551 A value of 1 instructs the soft-lockup detector
5552 to panic the machine when a soft-lockup occurs. It is
5553 also controlled by the kernel.softlockup_panic sysctl
5554 and CONFIG_BOOTPARAM_SOFTLOCKUP_PANIC, which is the
5555 respective build-time switch to that functionality.
5557 softlockup_all_cpu_backtrace=
5558 [KNL] Should the soft-lockup detector generate
5559 backtraces on all cpus.
5562 sonypi.*= [HW] Sony Programmable I/O Control Device driver
5563 See Documentation/admin-guide/laptops/sonypi.rst
5565 spectre_v2= [X86] Control mitigation of Spectre variant 2
5566 (indirect branch speculation) vulnerability.
5567 The default operation protects the kernel from
5570 on - unconditionally enable, implies
5572 off - unconditionally disable, implies
5574 auto - kernel detects whether your CPU model is
5577 Selecting 'on' will, and 'auto' may, choose a
5578 mitigation method at run time according to the
5579 CPU, the available microcode, the setting of the
5580 CONFIG_RETPOLINE configuration option, and the
5581 compiler with which the kernel was built.
5583 Selecting 'on' will also enable the mitigation
5584 against user space to user space task attacks.
5586 Selecting 'off' will disable both the kernel and
5587 the user space protections.
5589 Specific mitigations can also be selected manually:
5591 retpoline - replace indirect branches
5592 retpoline,generic - Retpolines
5593 retpoline,lfence - LFENCE; indirect branch
5594 retpoline,amd - alias for retpoline,lfence
5595 eibrs - enhanced IBRS
5596 eibrs,retpoline - enhanced IBRS + Retpolines
5597 eibrs,lfence - enhanced IBRS + LFENCE
5598 ibrs - use IBRS to protect kernel
5600 Not specifying this option is equivalent to
5604 [X86] Control mitigation of Spectre variant 2
5605 (indirect branch speculation) vulnerability between
5608 on - Unconditionally enable mitigations. Is
5609 enforced by spectre_v2=on
5611 off - Unconditionally disable mitigations. Is
5612 enforced by spectre_v2=off
5614 prctl - Indirect branch speculation is enabled,
5615 but mitigation can be enabled via prctl
5616 per thread. The mitigation control state
5617 is inherited on fork.
5620 - Like "prctl" above, but only STIBP is
5621 controlled per thread. IBPB is issued
5622 always when switching between different user
5626 - Same as "prctl" above, but all seccomp
5627 threads will enable the mitigation unless
5628 they explicitly opt out.
5631 - Like "seccomp" above, but only STIBP is
5632 controlled per thread. IBPB is issued
5633 always when switching between different
5634 user space processes.
5636 auto - Kernel selects the mitigation depending on
5637 the available CPU features and vulnerability.
5639 Default mitigation: "prctl"
5641 Not specifying this option is equivalent to
5642 spectre_v2_user=auto.
5644 spec_store_bypass_disable=
5645 [HW] Control Speculative Store Bypass (SSB) Disable mitigation
5646 (Speculative Store Bypass vulnerability)
5648 Certain CPUs are vulnerable to an exploit against a
5649 a common industry wide performance optimization known
5650 as "Speculative Store Bypass" in which recent stores
5651 to the same memory location may not be observed by
5652 later loads during speculative execution. The idea
5653 is that such stores are unlikely and that they can
5654 be detected prior to instruction retirement at the
5655 end of a particular speculation execution window.
5657 In vulnerable processors, the speculatively forwarded
5658 store can be used in a cache side channel attack, for
5659 example to read memory to which the attacker does not
5660 directly have access (e.g. inside sandboxed code).
5662 This parameter controls whether the Speculative Store
5663 Bypass optimization is used.
5665 On x86 the options are:
5667 on - Unconditionally disable Speculative Store Bypass
5668 off - Unconditionally enable Speculative Store Bypass
5669 auto - Kernel detects whether the CPU model contains an
5670 implementation of Speculative Store Bypass and
5671 picks the most appropriate mitigation. If the
5672 CPU is not vulnerable, "off" is selected. If the
5673 CPU is vulnerable the default mitigation is
5674 architecture and Kconfig dependent. See below.
5675 prctl - Control Speculative Store Bypass per thread
5676 via prctl. Speculative Store Bypass is enabled
5677 for a process by default. The state of the control
5678 is inherited on fork.
5679 seccomp - Same as "prctl" above, but all seccomp threads
5680 will disable SSB unless they explicitly opt out.
5682 Default mitigations:
5685 On powerpc the options are:
5687 on,auto - On Power8 and Power9 insert a store-forwarding
5688 barrier on kernel entry and exit. On Power7
5689 perform a software flush on kernel entry and
5693 Not specifying this option is equivalent to
5694 spec_store_bypass_disable=auto.
5696 spia_io_base= [HW,MTD]
5702 [X86] Enable split lock detection or bus lock detection
5704 When enabled (and if hardware support is present), atomic
5705 instructions that access data across cache line
5706 boundaries will result in an alignment check exception
5707 for split lock detection or a debug exception for
5712 warn - the kernel will emit rate-limited warnings
5713 about applications triggering the #AC
5714 exception or the #DB exception. This mode is
5715 the default on CPUs that support split lock
5716 detection or bus lock detection. Default
5717 behavior is by #AC if both features are
5718 enabled in hardware.
5720 fatal - the kernel will send SIGBUS to applications
5721 that trigger the #AC exception or the #DB
5722 exception. Default behavior is by #AC if
5723 both features are enabled in hardware.
5726 Set system wide rate limit to N bus locks
5727 per second for bus lock detection.
5730 N/A for split lock detection.
5733 If an #AC exception is hit in the kernel or in
5734 firmware (i.e. not while executing in user mode)
5735 the kernel will oops in either "warn" or "fatal"
5738 #DB exception for bus lock is triggered only when
5742 Control the Special Register Buffer Data Sampling
5745 Certain CPUs are vulnerable to an MDS-like
5746 exploit which can leak bits from the random
5749 By default, this issue is mitigated by
5750 microcode. However, the microcode fix can cause
5751 the RDRAND and RDSEED instructions to become
5752 much slower. Among other effects, this will
5753 result in reduced throughput from /dev/urandom.
5755 The microcode mitigation can be disabled with
5756 the following option:
5758 off: Disable mitigation and remove
5759 performance impact to RDRAND and RDSEED
5761 srcutree.big_cpu_lim [KNL]
5762 Specifies the number of CPUs constituting a
5763 large system, such that srcu_struct structures
5764 should immediately allocate an srcu_node array.
5765 This kernel-boot parameter defaults to 128,
5766 but takes effect only when the low-order four
5767 bits of srcutree.convert_to_big is equal to 3
5770 srcutree.convert_to_big [KNL]
5771 Specifies under what conditions an SRCU tree
5772 srcu_struct structure will be converted to big
5773 form, that is, with an rcu_node tree:
5776 1: At init_srcu_struct() time.
5777 2: When rcutorture decides to.
5778 3: Decide at boot time (default).
5779 0x1X: Above plus if high contention.
5781 Either way, the srcu_node tree will be sized based
5782 on the actual runtime number of CPUs (nr_cpu_ids)
5783 instead of the compile-time CONFIG_NR_CPUS.
5785 srcutree.counter_wrap_check [KNL]
5786 Specifies how frequently to check for
5787 grace-period sequence counter wrap for the
5788 srcu_data structure's ->srcu_gp_seq_needed field.
5789 The greater the number of bits set in this kernel
5790 parameter, the less frequently counter wrap will
5791 be checked for. Note that the bottom two bits
5794 srcutree.exp_holdoff [KNL]
5795 Specifies how many nanoseconds must elapse
5796 since the end of the last SRCU grace period for
5797 a given srcu_struct until the next normal SRCU
5798 grace period will be considered for automatic
5799 expediting. Set to zero to disable automatic
5802 srcutree.srcu_max_nodelay [KNL]
5803 Specifies the number of no-delay instances
5804 per jiffy for which the SRCU grace period
5805 worker thread will be rescheduled with zero
5806 delay. Beyond this limit, worker thread will
5807 be rescheduled with a sleep delay of one jiffy.
5809 srcutree.srcu_max_nodelay_phase [KNL]
5810 Specifies the per-grace-period phase, number of
5811 non-sleeping polls of readers. Beyond this limit,
5812 grace period worker thread will be rescheduled
5813 with a sleep delay of one jiffy, between each
5814 rescan of the readers, for a grace period phase.
5816 srcutree.srcu_retry_check_delay [KNL]
5817 Specifies number of microseconds of non-sleeping
5818 delay between each non-sleeping poll of readers.
5820 srcutree.small_contention_lim [KNL]
5821 Specifies the number of update-side contention
5822 events per jiffy will be tolerated before
5823 initiating a conversion of an srcu_struct
5824 structure to big form. Note that the value of
5825 srcutree.convert_to_big must have the 0x10 bit
5826 set for contention-based conversions to occur.
5829 Speculative Store Bypass Disable control
5831 On CPUs that are vulnerable to the Speculative
5832 Store Bypass vulnerability and offer a
5833 firmware based mitigation, this parameter
5834 indicates how the mitigation should be used:
5836 force-on: Unconditionally enable mitigation for
5837 for both kernel and userspace
5838 force-off: Unconditionally disable mitigation for
5839 for both kernel and userspace
5840 kernel: Always enable mitigation in the
5841 kernel, and offer a prctl interface
5842 to allow userspace to register its
5843 interest in being mitigated too.
5845 stack_guard_gap= [MM]
5846 override the default stack gap protection. The value
5847 is in page units and it defines how many pages prior
5848 to (for stacks growing down) resp. after (for stacks
5849 growing up) the main stack are reserved for no other
5850 mapping. Default value is 256 pages.
5852 stack_depot_disable= [KNL]
5853 Setting this to true through kernel command line will
5854 disable the stack depot thereby saving the static memory
5855 consumed by the stack hash table. By default this is set
5859 Enabled the stack tracer on boot up.
5861 stacktrace_filter=[function-list]
5862 [FTRACE] Limit the functions that the stack tracer
5863 will trace at boot up. function-list is a comma-separated
5864 list of functions. This list can be changed at run
5865 time by the stack_trace_filter file in the debugfs
5866 tracing directory. Note, this enables stack tracing
5867 and the stacktrace above is not needed.
5871 Set the STI (builtin display/keyboard on the HP-PARISC
5872 machines) console (graphic card) which should be used
5873 as the initial boot-console.
5874 See also comment in drivers/video/console/sticore.c.
5877 See comment in drivers/video/console/sticore.c.
5880 Format: bpp:<bpp1>[:<bpp2>[:<bpp3>...]]
5885 Enable or disable strict sigaltstack size checks
5886 against the required signal frame size which
5887 depends on the supported FPU features. This can
5888 be used to filter out binaries which have
5889 not yet been made aware of AT_MINSIGSTKSZ.
5891 sunrpc.min_resvport=
5892 sunrpc.max_resvport=
5894 SunRPC servers often require that client requests
5895 originate from a privileged port (i.e. a port in the
5896 range 0 < portnr < 1024).
5897 An administrator who wishes to reserve some of these
5898 ports for other uses may adjust the range that the
5899 kernel's sunrpc client considers to be privileged
5900 using these two parameters to set the minimum and
5901 maximum port values.
5903 sunrpc.svc_rpc_per_connection_limit=
5905 Limit the number of requests that the server will
5906 process in parallel from a single connection.
5907 The default value is 0 (no limit).
5911 Control how the NFS server code allocates CPUs to
5912 service thread pools. Depending on how many NICs
5913 you have and where their interrupts are bound, this
5914 option will affect which CPUs will do NFS serving.
5915 Note: this parameter cannot be changed while the
5916 NFS server is running.
5918 auto the server chooses an appropriate mode
5919 automatically using heuristics
5920 global a single global pool contains all CPUs
5921 percpu one pool for each CPU
5922 pernode one pool for each NUMA node (equivalent
5923 to global on non-NUMA machines)
5925 sunrpc.tcp_slot_table_entries=
5926 sunrpc.udp_slot_table_entries=
5928 Sets the upper limit on the number of simultaneous
5929 RPC calls that can be sent from the client to a
5930 server. Increasing these values may allow you to
5931 improve throughput, but will also increase the
5932 amount of memory reserved for use by the client.
5934 suspend.pm_test_delay=
5936 Sets the number of seconds to remain in a suspend test
5937 mode before resuming the system (see
5938 /sys/power/pm_test). Only available when CONFIG_PM_DEBUG
5939 is set. Default value is 5.
5942 Format: { on | off | y | n | 1 | 0 }
5943 This parameter controls use of the Protected
5944 Execution Facility on pSeries.
5948 Enable accounting of swap in memory resource
5949 controller if no parameter or 1 is given or disable
5950 it if 0 is given (See Documentation/admin-guide/cgroup-v1/memory.rst)
5952 swiotlb= [ARM,IA-64,PPC,MIPS,X86]
5953 Format: { <int> | force | noforce }
5954 <int> -- Number of I/O TLB slabs
5955 force -- force using of bounce buffers even if they
5956 wouldn't be automatically used by the kernel
5957 noforce -- Never use bounce buffers (for debugging)
5962 Set a sysctl parameter, right before loading the init
5963 process, as if the value was written to the respective
5964 /proc/sys/... file. Both '.' and '/' are recognized as
5965 separators. Unrecognized parameters and invalid values
5966 are reported in the kernel log. Sysctls registered
5967 later by a loaded module cannot be set this way.
5968 Example: sysctl.vm.swappiness=40
5970 sysfs.deprecated=0|1 [KNL]
5971 Enable/disable old style sysfs layout for old udev
5972 on older distributions. When this option is enabled
5973 very new udev will not work anymore. When this option
5974 is disabled (or CONFIG_SYSFS_DEPRECATED not compiled)
5975 in older udev will not work anymore.
5976 Default depends on CONFIG_SYSFS_DEPRECATED_V2 set in
5977 the kernel configuration.
5979 sysrq_always_enabled
5981 Ignore sysrq setting - this boot parameter will
5982 neutralize any effect of /proc/sys/kernel/sysrq.
5983 Useful for debugging.
5985 tcpmhash_entries= [KNL,NET]
5986 Set the number of tcp_metrics_hash slots.
5987 Default value is 8192 or 16384 depending on total
5988 ram pages. This is used to specify the TCP metrics
5989 cache size. See Documentation/networking/ip-sysctl.rst
5990 "tcp_no_metrics_save" section for more details.
5994 test_suspend= [SUSPEND]
5995 Format: { "mem" | "standby" | "freeze" }[,N]
5996 Specify "mem" (for Suspend-to-RAM) or "standby" (for
5997 standby suspend) or "freeze" (for suspend type freeze)
5998 as the system sleep state during system startup with
5999 the optional capability to repeat N number of times.
6000 The system is woken from this state using a
6001 wakeup-capable RTC alarm.
6003 thash_entries= [KNL,NET]
6004 Set number of hash buckets for TCP connection
6006 thermal.act= [HW,ACPI]
6007 -1: disable all active trip points in all thermal zones
6008 <degrees C>: override all lowest active trip points
6010 thermal.crt= [HW,ACPI]
6011 -1: disable all critical trip points in all thermal zones
6012 <degrees C>: override all critical trip points
6014 thermal.nocrt= [HW,ACPI]
6015 Set to disable actions on ACPI thermal zone
6016 critical and hot trip points.
6018 thermal.off= [HW,ACPI]
6019 1: disable ACPI thermal control
6021 thermal.psv= [HW,ACPI]
6022 -1: disable all passive trip points
6023 <degrees C>: override all passive trip points to this
6026 thermal.tzp= [HW,ACPI]
6027 Specify global default ACPI thermal zone polling rate
6028 <deci-seconds>: poll all this frequency
6029 0: no polling (default)
6032 Force threading of all interrupt handlers except those
6033 marked explicitly IRQF_NO_THREAD.
6037 Specify if the kernel should make use of the cpu
6038 topology information if the hardware supports this.
6039 The scheduler will make use of this information and
6040 e.g. base its process migration decisions on it.
6043 topology_updates= [KNL, PPC, NUMA]
6045 Specify if the kernel should ignore (off)
6046 topology updates sent by the hypervisor to this
6049 torture.disable_onoff_at_boot= [KNL]
6050 Prevent the CPU-hotplug component of torturing
6051 until after init has spawned.
6053 torture.ftrace_dump_at_shutdown= [KNL]
6054 Dump the ftrace buffer at torture-test shutdown,
6055 even if there were no errors. This can be a
6056 very costly operation when many torture tests
6057 are running concurrently, especially on systems
6058 with rotating-rust storage.
6060 torture.verbose_sleep_frequency= [KNL]
6061 Specifies how many verbose printk()s should be
6062 emitted between each sleep. The default of zero
6063 disables verbose-printk() sleeping.
6065 torture.verbose_sleep_duration= [KNL]
6066 Duration of each verbose-printk() sleep in jiffies.
6070 tpm_suspend_pcr=[HW,TPM]
6071 Format: integer pcr id
6072 Specify that at suspend time, the tpm driver
6073 should extend the specified pcr with zeros,
6074 as a workaround for some chips which fail to
6075 flush the last written pcr on TPM_SaveState.
6076 This will guarantee that all the other pcrs
6080 Have the tracepoints sent to printk as well as the
6081 tracing ring buffer. This is useful for early boot up
6082 where the system hangs or reboots and does not give the
6083 option for reading the tracing buffer or performing a
6084 ftrace_dump_on_oops.
6086 To turn off having tracepoints sent to printk,
6087 echo 0 > /proc/sys/kernel/tracepoint_printk
6088 Note, echoing 1 into this file without the
6089 tracepoint_printk kernel cmdline option has no effect.
6091 The tp_printk_stop_on_boot (see below) can also be used
6092 to stop the printing of events to console at
6097 Having tracepoints sent to printk() and activating high
6098 frequency tracepoints such as irq or sched, can cause
6099 the system to live lock.
6101 tp_printk_stop_on_boot [FTRACE]
6102 When tp_printk (above) is set, it can cause a lot of noise
6103 on the console. It may be useful to only include the
6104 printing of events during boot up, as user space may
6105 make the system inoperable.
6107 This command line option will stop the printing of events
6108 to console at the late_initcall_sync() time frame.
6110 trace_buf_size=nn[KMG]
6111 [FTRACE] will set tracing buffer size on each cpu.
6113 trace_clock= [FTRACE] Set the clock used for tracing events
6115 local - Use the per CPU time stamp counter
6116 (converted into nanoseconds). Fast, but
6117 depending on the architecture, may not be
6118 in sync between CPUs.
6119 global - Event time stamps are synchronize across
6120 CPUs. May be slower than the local clock,
6121 but better for some race conditions.
6122 counter - Simple counting of events (1, 2, ..)
6123 note, some counts may be skipped due to the
6124 infrastructure grabbing the clock more than
6126 uptime - Use jiffies as the time stamp.
6127 perf - Use the same clock that perf uses.
6128 mono - Use ktime_get_mono_fast_ns() for time stamps.
6129 mono_raw - Use ktime_get_raw_fast_ns() for time
6131 boot - Use ktime_get_boot_fast_ns() for time stamps.
6132 Architectures may add more clocks. See
6133 Documentation/trace/ftrace.rst for more details.
6135 trace_event=[event-list]
6136 [FTRACE] Set and start specified trace events in order
6137 to facilitate early boot debugging. The event-list is a
6138 comma-separated list of trace events to enable. See
6139 also Documentation/trace/events.rst
6141 trace_options=[option-list]
6142 [FTRACE] Enable or disable tracer options at boot.
6143 The option-list is a comma delimited list of options
6144 that can be enabled or disabled just as if you were
6145 to echo the option name into
6147 /sys/kernel/debug/tracing/trace_options
6149 For example, to enable stacktrace option (to dump the
6150 stack trace of each event), add to the command line:
6152 trace_options=stacktrace
6154 See also Documentation/trace/ftrace.rst "trace options"
6158 [FTRACE] enable this option to disable tracing when a
6159 warning is hit. This turns off "tracing_on". Tracing can
6160 be enabled again by echoing '1' into the "tracing_on"
6161 file located in /sys/kernel/debug/tracing/
6163 This option is useful, as it disables the trace before
6164 the WARNING dump is called, which prevents the trace to
6165 be filled with content caused by the warning output.
6167 This option can also be set at run time via the sysctl
6168 option: kernel/traceoff_on_warning
6170 transparent_hugepage=
6172 Format: [always|madvise|never]
6173 Can be used to control the default behavior of the system
6174 with respect to transparent hugepages.
6175 See Documentation/admin-guide/mm/transhuge.rst
6178 trusted.source= [KEYS]
6180 This parameter identifies the trust source as a backend
6181 for trusted keys implementation. Supported trust
6186 If not specified then it defaults to iterating through
6187 the trust source list starting with TPM and assigns the
6188 first trust source as a backend which is initialized
6189 successfully during iteration.
6193 The RNG used to generate key material for trusted keys.
6196 - the same value as trusted.source: "tpm" or "tee"
6198 If not specified, "default" is used. In this case,
6199 the RNG's choice is left to each individual trust source.
6201 tsc= Disable clocksource stability checks for TSC.
6203 [x86] reliable: mark tsc clocksource as reliable, this
6204 disables clocksource verification at runtime, as well
6205 as the stability checks done at bootup. Used to enable
6206 high-resolution timer mode on older hardware, and in
6207 virtualized environment.
6208 [x86] noirqtime: Do not use TSC to do irq accounting.
6209 Used to run time disable IRQ_TIME_ACCOUNTING on any
6210 platforms where RDTSC is slow and this accounting
6212 [x86] unstable: mark the TSC clocksource as unstable, this
6213 marks the TSC unconditionally unstable at bootup and
6214 avoids any further wobbles once the TSC watchdog notices.
6215 [x86] nowatchdog: disable clocksource watchdog. Used
6216 in situations with strict latency requirements (where
6217 interruptions from clocksource watchdog are not
6220 tsc_early_khz= [X86] Skip early TSC calibration and use the given
6221 value instead. Useful when the early TSC frequency discovery
6222 procedure is not reliable, such as on overclocked systems
6223 with CPUID.16h support and partial CPUID.15h support.
6224 Format: <unsigned int>
6226 tsx= [X86] Control Transactional Synchronization
6227 Extensions (TSX) feature in Intel processors that
6228 support TSX control.
6230 This parameter controls the TSX feature. The options are:
6232 on - Enable TSX on the system. Although there are
6233 mitigations for all known security vulnerabilities,
6234 TSX has been known to be an accelerator for
6235 several previous speculation-related CVEs, and
6236 so there may be unknown security risks associated
6237 with leaving it enabled.
6239 off - Disable TSX on the system. (Note that this
6240 option takes effect only on newer CPUs which are
6241 not vulnerable to MDS, i.e., have
6242 MSR_IA32_ARCH_CAPABILITIES.MDS_NO=1 and which get
6243 the new IA32_TSX_CTRL MSR through a microcode
6244 update. This new MSR allows for the reliable
6245 deactivation of the TSX functionality.)
6247 auto - Disable TSX if X86_BUG_TAA is present,
6248 otherwise enable TSX on the system.
6250 Not specifying this option is equivalent to tsx=off.
6252 See Documentation/admin-guide/hw-vuln/tsx_async_abort.rst
6255 tsx_async_abort= [X86,INTEL] Control mitigation for the TSX Async
6256 Abort (TAA) vulnerability.
6258 Similar to Micro-architectural Data Sampling (MDS)
6259 certain CPUs that support Transactional
6260 Synchronization Extensions (TSX) are vulnerable to an
6261 exploit against CPU internal buffers which can forward
6262 information to a disclosure gadget under certain
6265 In vulnerable processors, the speculatively forwarded
6266 data can be used in a cache side channel attack, to
6267 access data to which the attacker does not have direct
6270 This parameter controls the TAA mitigation. The
6273 full - Enable TAA mitigation on vulnerable CPUs
6276 full,nosmt - Enable TAA mitigation and disable SMT on
6277 vulnerable CPUs. If TSX is disabled, SMT
6278 is not disabled because CPU is not
6279 vulnerable to cross-thread TAA attacks.
6280 off - Unconditionally disable TAA mitigation
6282 On MDS-affected machines, tsx_async_abort=off can be
6283 prevented by an active MDS mitigation as both vulnerabilities
6284 are mitigated with the same mechanism so in order to disable
6285 this mitigation, you need to specify mds=off too.
6287 Not specifying this option is equivalent to
6288 tsx_async_abort=full. On CPUs which are MDS affected
6289 and deploy MDS mitigation, TAA mitigation is not
6290 required and doesn't provide any additional
6294 Documentation/admin-guide/hw-vuln/tsx_async_abort.rst
6296 turbografx.map[2|3]= [HW,JOY]
6297 TurboGraFX parallel port interface
6299 <port#>,<js1>,<js2>,<js3>,<js4>,<js5>,<js6>,<js7>
6300 See also Documentation/input/devices/joystick-parport.rst
6302 udbg-immortal [PPC] When debugging early kernel crashes that
6303 happen after console_init() and before a proper
6304 console driver takes over, this boot options might
6305 help "seeing" what's going on.
6307 uhash_entries= [KNL,NET]
6308 Set number of hash buckets for UDP/UDP-Lite connections
6311 [USB] Ignore overcurrent events (default N).
6312 Some badly-designed motherboards generate lots of
6313 bogus events, for ports that aren't wired to
6314 anything. Set this parameter to avoid log spamming.
6315 Note that genuine overcurrent events won't be
6319 [X86] Cause panic on unknown NMI.
6321 usbcore.authorized_default=
6322 [USB] Default USB device authorization:
6323 (default -1 = authorized except for wireless USB,
6324 0 = not authorized, 1 = authorized, 2 = authorized
6325 if device connected to internal port)
6327 usbcore.autosuspend=
6328 [USB] The autosuspend time delay (in seconds) used
6329 for newly-detected USB devices (default 2). This
6330 is the time required before an idle device will be
6331 autosuspended. Devices for which the delay is set
6332 to a negative value won't be autosuspended at all.
6334 usbcore.usbfs_snoop=
6335 [USB] Set to log all usbfs traffic (default 0 = off).
6337 usbcore.usbfs_snoop_max=
6338 [USB] Maximum number of bytes to snoop in each URB
6341 usbcore.blinkenlights=
6342 [USB] Set to cycle leds on hubs (default 0 = off).
6344 usbcore.old_scheme_first=
6345 [USB] Start with the old device initialization
6346 scheme (default 0 = off).
6348 usbcore.usbfs_memory_mb=
6349 [USB] Memory limit (in MB) for buffers allocated by
6350 usbfs (default = 16, 0 = max = 2047).
6352 usbcore.use_both_schemes=
6353 [USB] Try the other device initialization scheme
6354 if the first one fails (default 1 = enabled).
6356 usbcore.initial_descriptor_timeout=
6357 [USB] Specifies timeout for the initial 64-byte
6358 USB_REQ_GET_DESCRIPTOR request in milliseconds
6359 (default 5000 = 5.0 seconds).
6361 usbcore.nousb [USB] Disable the USB subsystem
6364 [USB] A list of quirk entries to augment the built-in
6365 usb core quirk list. List entries are separated by
6366 commas. Each entry has the form
6367 VendorID:ProductID:Flags. The IDs are 4-digit hex
6368 numbers and Flags is a set of letters. Each letter
6369 will change the built-in quirk; setting it if it is
6370 clear and clearing it if it is set. The letters have
6371 the following meanings:
6372 a = USB_QUIRK_STRING_FETCH_255 (string
6373 descriptors must not be fetched using
6375 b = USB_QUIRK_RESET_RESUME (device can't resume
6376 correctly so reset it instead);
6377 c = USB_QUIRK_NO_SET_INTF (device can't handle
6378 Set-Interface requests);
6379 d = USB_QUIRK_CONFIG_INTF_STRINGS (device can't
6380 handle its Configuration or Interface
6382 e = USB_QUIRK_RESET (device can't be reset
6383 (e.g morph devices), don't use reset);
6384 f = USB_QUIRK_HONOR_BNUMINTERFACES (device has
6385 more interface descriptions than the
6386 bNumInterfaces count, and can't handle
6387 talking to these interfaces);
6388 g = USB_QUIRK_DELAY_INIT (device needs a pause
6389 during initialization, after we read
6390 the device descriptor);
6391 h = USB_QUIRK_LINEAR_UFRAME_INTR_BINTERVAL (For
6392 high speed and super speed interrupt
6393 endpoints, the USB 2.0 and USB 3.0 spec
6394 require the interval in microframes (1
6395 microframe = 125 microseconds) to be
6396 calculated as interval = 2 ^
6398 Devices with this quirk report their
6399 bInterval as the result of this
6400 calculation instead of the exponent
6401 variable used in the calculation);
6402 i = USB_QUIRK_DEVICE_QUALIFIER (device can't
6403 handle device_qualifier descriptor
6405 j = USB_QUIRK_IGNORE_REMOTE_WAKEUP (device
6406 generates spurious wakeup, ignore
6407 remote wakeup capability);
6408 k = USB_QUIRK_NO_LPM (device can't handle Link
6410 l = USB_QUIRK_LINEAR_FRAME_INTR_BINTERVAL
6411 (Device reports its bInterval as linear
6412 frames instead of the USB 2.0
6414 m = USB_QUIRK_DISCONNECT_SUSPEND (Device needs
6415 to be disconnected before suspend to
6416 prevent spurious wakeup);
6417 n = USB_QUIRK_DELAY_CTRL_MSG (Device needs a
6418 pause after every control message);
6419 o = USB_QUIRK_HUB_SLOW_RESET (Hub needs extra
6420 delay after resetting its port);
6421 Example: quirks=0781:5580:bk,0a5c:5834:gij
6424 [USBHID] The interval which mice are to be polled at.
6427 [USBHID] The interval which joysticks are to be polled at.
6430 [USBHID] The interval which keyboards are to be polled at.
6432 usb-storage.delay_use=
6433 [UMS] The delay in seconds before a new device is
6434 scanned for Logical Units (default 1).
6437 [UMS] A list of quirks entries to supplement or
6438 override the built-in unusual_devs list. List
6439 entries are separated by commas. Each entry has
6440 the form VID:PID:Flags where VID and PID are Vendor
6441 and Product ID values (4-digit hex numbers) and
6442 Flags is a set of characters, each corresponding
6443 to a common usb-storage quirk flag as follows:
6444 a = SANE_SENSE (collect more than 18 bytes
6445 of sense data, not on uas);
6446 b = BAD_SENSE (don't collect more than 18
6447 bytes of sense data, not on uas);
6448 c = FIX_CAPACITY (decrease the reported
6449 device capacity by one sector);
6450 d = NO_READ_DISC_INFO (don't use
6451 READ_DISC_INFO command, not on uas);
6452 e = NO_READ_CAPACITY_16 (don't use
6453 READ_CAPACITY_16 command);
6454 f = NO_REPORT_OPCODES (don't use report opcodes
6456 g = MAX_SECTORS_240 (don't transfer more than
6457 240 sectors at a time, uas only);
6458 h = CAPACITY_HEURISTICS (decrease the
6459 reported device capacity by one
6460 sector if the number is odd);
6461 i = IGNORE_DEVICE (don't bind to this
6463 j = NO_REPORT_LUNS (don't use report luns
6465 k = NO_SAME (do not use WRITE_SAME, uas only)
6466 l = NOT_LOCKABLE (don't try to lock and
6467 unlock ejectable media, not on uas);
6468 m = MAX_SECTORS_64 (don't transfer more
6469 than 64 sectors = 32 KB at a time,
6471 n = INITIAL_READ10 (force a retry of the
6472 initial READ(10) command, not on uas);
6473 o = CAPACITY_OK (accept the capacity
6474 reported by the device, not on uas);
6475 p = WRITE_CACHE (the device cache is ON
6476 by default, not on uas);
6477 r = IGNORE_RESIDUE (the device reports
6478 bogus residue values, not on uas);
6479 s = SINGLE_LUN (the device has only one
6481 t = NO_ATA_1X (don't allow ATA(12) and ATA(16)
6482 commands, uas only);
6483 u = IGNORE_UAS (don't bind to the uas driver);
6484 w = NO_WP_DETECT (don't test whether the
6485 medium is write-protected).
6486 y = ALWAYS_SYNC (issue a SYNCHRONIZE_CACHE
6487 even if the device claims no cache,
6489 Example: quirks=0419:aaf5:rl,0421:0433:rc
6491 user_debug= [KNL,ARM]
6493 See arch/arm/Kconfig.debug help text.
6494 1 - undefined instruction events
6496 4 - invalid data aborts
6499 Example: user_debug=31
6502 [X86] Flags controlling user PTE allocations.
6504 nohigh = do not allocate PTE pages in
6505 HIGHMEM regardless of setting
6508 vdso= [X86,SH,SPARC]
6509 On X86_32, this is an alias for vdso32=. Otherwise:
6511 vdso=1: enable VDSO (the default)
6512 vdso=0: disable VDSO mapping
6514 vdso32= [X86] Control the 32-bit vDSO
6515 vdso32=1: enable 32-bit VDSO
6516 vdso32=0 or vdso32=2: disable 32-bit VDSO
6518 See the help text for CONFIG_COMPAT_VDSO for more
6519 details. If CONFIG_COMPAT_VDSO is set, the default is
6520 vdso32=0; otherwise, the default is vdso32=1.
6522 For compatibility with older kernels, vdso32=2 is an
6525 Try vdso32=0 if you encounter an error that says:
6526 dl_main: Assertion `(void *) ph->p_vaddr == _rtld_local._dl_sysinfo_dso' failed!
6529 vector=percpu: enable percpu vector domain
6531 video= [FB] Frame buffer configuration
6532 See Documentation/fb/modedb.rst.
6534 video.brightness_switch_enabled= [ACPI]
6536 If set to 1, on receiving an ACPI notify event
6537 generated by hotkey, video driver will adjust brightness
6538 level and then send out the event to user space through
6539 the allocated input device. If set to 0, video driver
6540 will only send out the event without touching backlight
6545 [VMMIO] Memory mapped virtio (platform) device.
6547 <size>@<baseaddr>:<irq>[:<id>]
6549 <size> := size (can use standard suffixes
6551 <baseaddr> := physical base address
6552 <irq> := interrupt number (as passed to
6554 <id> := (optional) platform device id
6556 virtio_mmio.device=1K@0x100b0000:48:7
6558 Can be used multiple times for multiple devices.
6560 vga= [BOOT,X86-32] Select a particular video mode
6561 See Documentation/x86/boot.rst and
6562 Documentation/admin-guide/svga.rst.
6563 Use vga=ask for menu.
6564 This is actually a boot loader parameter; the value is
6565 passed to the kernel using a special protocol.
6567 vm_debug[=options] [KNL] Available with CONFIG_DEBUG_VM=y.
6568 May slow down system boot speed, especially when
6569 enabled on systems with a large amount of memory.
6570 All options are enabled by default, and this
6571 interface is meant to allow for selectively
6572 enabling or disabling specific virtual memory
6575 Available options are:
6576 P Enable page structure init time poisoning
6577 - Disable all of the above options
6579 vmalloc=nn[KMG] [KNL,BOOT] Forces the vmalloc area to have an exact
6580 size of <nn>. This can be used to increase the
6581 minimum size (128MB on x86). It can also be used to
6582 decrease the size and leave more room for directly
6585 vmcp_cma=nn[MG] [KNL,S390]
6586 Sets the memory size reserved for contiguous memory
6587 allocations for the vmcp device driver.
6589 vmhalt= [KNL,S390] Perform z/VM CP command after system halt.
6592 vmpanic= [KNL,S390] Perform z/VM CP command after kernel panic.
6595 vmpoff= [KNL,S390] Perform z/VM CP command after power off.
6599 Controls the behavior of vsyscalls (i.e. calls to
6600 fixed addresses of 0xffffffffff600x00 from legacy
6601 code). Most statically-linked binaries and older
6602 versions of glibc use these calls. Because these
6603 functions are at fixed addresses, they make nice
6604 targets for exploits that can control RIP.
6606 emulate [default] Vsyscalls turn into traps and are
6607 emulated reasonably safely. The vsyscall
6610 xonly Vsyscalls turn into traps and are
6611 emulated reasonably safely. The vsyscall
6612 page is not readable.
6614 none Vsyscalls don't work at all. This makes
6615 them quite hard to use for exploits but
6616 might break your system.
6618 vt.color= [VT] Default text color.
6619 Format: 0xYX, X = foreground, Y = background.
6620 Default: 0x07 = light gray on black.
6622 vt.cur_default= [VT] Default cursor shape.
6623 Format: 0xCCBBAA, where AA, BB, and CC are the same as
6624 the parameters of the <Esc>[?A;B;Cc escape sequence;
6625 see VGA-softcursor.txt. Default: 2 = underline.
6627 vt.default_blu= [VT]
6628 Format: <blue0>,<blue1>,<blue2>,...,<blue15>
6629 Change the default blue palette of the console.
6630 This is a 16-member array composed of values
6633 vt.default_grn= [VT]
6634 Format: <green0>,<green1>,<green2>,...,<green15>
6635 Change the default green palette of the console.
6636 This is a 16-member array composed of values
6639 vt.default_red= [VT]
6640 Format: <red0>,<red1>,<red2>,...,<red15>
6641 Change the default red palette of the console.
6642 This is a 16-member array composed of values
6648 Set system-wide default UTF-8 mode for all tty's.
6649 Default is 1, i.e. UTF-8 mode is enabled for all
6650 newly opened terminals.
6652 vt.global_cursor_default=
6655 Set system-wide default for whether a cursor
6656 is shown on new VTs. Default is -1,
6657 i.e. cursors will be created by default unless
6658 overridden by individual drivers. 0 will hide
6659 cursors, 1 will display them.
6661 vt.italic= [VT] Default color for italic text; 0-15.
6664 vt.underline= [VT] Default color for underlined text; 0-15.
6667 watchdog timers [HW,WDT] For information on watchdog timers,
6668 see Documentation/watchdog/watchdog-parameters.rst
6669 or other driver-specific files in the
6670 Documentation/watchdog/ directory.
6674 Set the hard lockup detector stall duration
6675 threshold in seconds. The soft lockup detector
6676 threshold is set to twice the value. A value of 0
6677 disables both lockup detectors. Default is 10
6680 workqueue.watchdog_thresh=
6681 If CONFIG_WQ_WATCHDOG is configured, workqueue can
6682 warn stall conditions and dump internal state to
6683 help debugging. 0 disables workqueue stall
6684 detection; otherwise, it's the stall threshold
6685 duration in seconds. The default value is 30 and
6686 it can be updated at runtime by writing to the
6687 corresponding sysfs file.
6689 workqueue.disable_numa
6690 By default, all work items queued to unbound
6691 workqueues are affine to the NUMA nodes they're
6692 issued on, which results in better behavior in
6693 general. If NUMA affinity needs to be disabled for
6694 whatever reason, this option can be used. Note
6695 that this also can be controlled per-workqueue for
6696 workqueues visible under /sys/bus/workqueue/.
6698 workqueue.power_efficient
6699 Per-cpu workqueues are generally preferred because
6700 they show better performance thanks to cache
6701 locality; unfortunately, per-cpu workqueues tend to
6702 be more power hungry than unbound workqueues.
6704 Enabling this makes the per-cpu workqueues which
6705 were observed to contribute significantly to power
6706 consumption unbound, leading to measurably lower
6707 power usage at the cost of small performance
6710 The default value of this parameter is determined by
6711 the config option CONFIG_WQ_POWER_EFFICIENT_DEFAULT.
6713 workqueue.debug_force_rr_cpu
6714 Workqueue used to implicitly guarantee that work
6715 items queued without explicit CPU specified are put
6716 on the local CPU. This guarantee is no longer true
6717 and while local CPU is still preferred work items
6718 may be put on foreign CPUs. This debug option
6719 forces round-robin CPU selection to flush out
6720 usages which depend on the now broken guarantee.
6721 When enabled, memory and cache locality will be
6724 x2apic_phys [X86-64,APIC] Use x2apic physical mode instead of
6725 default x2apic cluster mode on platforms
6728 xen_512gb_limit [KNL,X86-64,XEN]
6729 Restricts the kernel running paravirtualized under Xen
6730 to use only up to 512 GB of RAM. The reason to do so is
6731 crash analysis tools and Xen tools for doing domain
6732 save/restore/migration must be enabled to handle larger
6735 xen_emul_unplug= [HW,X86,XEN]
6736 Unplug Xen emulated devices
6737 Format: [unplug0,][unplug1]
6738 ide-disks -- unplug primary master IDE devices
6739 aux-ide-disks -- unplug non-primary-master IDE devices
6740 nics -- unplug network devices
6741 all -- unplug all emulated devices (NICs and IDE disks)
6742 unnecessary -- unplugging emulated devices is
6743 unnecessary even if the host did not respond to
6745 never -- do not unplug even if version check succeeds
6747 xen_legacy_crash [X86,XEN]
6748 Crash from Xen panic notifier, without executing late
6749 panic() code such as dumping handler.
6751 xen_nopvspin [X86,XEN]
6752 Disables the qspinlock slowpath using Xen PV optimizations.
6753 This parameter is obsoleted by "nopvspin" parameter, which
6754 has equivalent effect for XEN platform.
6757 Disables the PV optimizations forcing the HVM guest to
6758 run as generic HVM guest with no PV drivers.
6759 This option is obsoleted by the "nopv" option, which
6760 has equivalent effect for XEN platform.
6762 xen_no_vector_callback
6763 [KNL,X86,XEN] Disable the vector callback for Xen
6764 event channel interrupts.
6766 xen_scrub_pages= [XEN]
6767 Boolean option to control scrubbing pages before giving them back
6768 to Xen, for use by other domains. Can be also changed at runtime
6769 with /sys/devices/system/xen_memory/xen_memory0/scrub_pages.
6770 Default value controlled with CONFIG_XEN_SCRUB_PAGES_DEFAULT.
6772 xen_timer_slop= [X86-64,XEN]
6773 Set the timer slop (in nanoseconds) for the virtual Xen
6774 timers (default is 100000). This adjusts the minimum
6775 delta of virtualized Xen timers, where lower values
6776 improve timer resolution at the expense of processing
6777 more timer interrupts.
6779 xen.balloon_boot_timeout= [XEN]
6780 The time (in seconds) to wait before giving up to boot
6781 in case initial ballooning fails to free enough memory.
6782 Applies only when running as HVM or PVH guest and
6783 started with less memory configured than allowed at
6784 max. Default is 180.
6786 xen.event_eoi_delay= [XEN]
6787 How long to delay EOI handling in case of event
6788 storms (jiffies). Default is 10.
6790 xen.event_loop_timeout= [XEN]
6791 After which time (jiffies) the event handling loop
6792 should start to delay EOI handling. Default is 2.
6794 xen.fifo_events= [XEN]
6795 Boolean parameter to disable using fifo event handling
6796 even if available. Normally fifo event handling is
6797 preferred over the 2-level event handling, as it is
6798 fairer and the number of possible event channels is
6799 much higher. Default is on (use fifo events).
6801 nopv= [X86,XEN,KVM,HYPER_V,VMWARE]
6802 Disables the PV optimizations forcing the guest to run
6803 as generic guest with no PV drivers. Currently support
6804 XEN HVM, KVM, HYPER_V and VMWARE guest.
6806 nopvspin [X86,XEN,KVM]
6807 Disables the qspinlock slow path using PV optimizations
6808 which allow the hypervisor to 'idle' the guest on lock
6811 xirc2ps_cs= [NET,PCMCIA]
6813 <irq>,<irq_mask>,<io>,<full_duplex>,<do_sound>,<lockup_hack>[,<irq2>[,<irq3>[,<irq4>]]]
6816 By default on POWER9 and above, the kernel will
6817 natively use the XIVE interrupt controller. This option
6818 allows the fallback firmware mode to be used:
6820 off Fallback to firmware control of XIVE interrupt
6821 controller on both pseries and powernv
6822 platforms. Only useful on POWER9 and above.
6824 xive.store-eoi=off [PPC]
6825 By default on POWER10 and above, the kernel will use
6826 stores for EOI handling when the XIVE interrupt mode
6827 is active. This option allows the XIVE driver to use
6828 loads instead, as on POWER9.
6830 xhci-hcd.quirks [USB,KNL]
6831 A hex value specifying bitmask with supplemental xhci
6832 host controller quirks. Meaning of each bit can be
6833 consulted in header drivers/usb/host/xhci.h.
6836 Format: { early | on | rw | ro | off }
6837 Controls if xmon debugger is enabled. Default is off.
6838 Passing only "xmon" is equivalent to "xmon=early".
6839 early Call xmon as early as possible on boot; xmon
6840 debugger is called from setup_arch().
6841 on xmon debugger hooks will be installed so xmon
6842 is only called on a kernel crash. Default mode,
6843 i.e. either "ro" or "rw" mode, is controlled
6844 with CONFIG_XMON_DEFAULT_RO_MODE.
6845 rw xmon debugger hooks will be installed so xmon
6846 is called only on a kernel crash, mode is write,
6847 meaning SPR registers, memory and, other data
6848 can be written using xmon commands.
6849 ro same as "rw" option above but SPR registers,
6850 memory, and other data can't be written using
6852 off xmon is disabled.