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
324 pgtbl_v1 - Use v1 page table for DMA-API (Default).
325 pgtbl_v2 - Use v2 page table for DMA-API.
327 amd_iommu_dump= [HW,X86-64]
328 Enable AMD IOMMU driver option to dump the ACPI table
329 for AMD IOMMU. With this option enabled, AMD IOMMU
330 driver will print ACPI tables for AMD IOMMU during
331 IOMMU initialization.
333 amd_iommu_intr= [HW,X86-64]
334 Specifies one of the following AMD IOMMU interrupt
336 legacy - Use legacy interrupt remapping mode.
337 vapic - Use virtual APIC mode, which allows IOMMU
338 to inject interrupts directly into guest.
339 This mode requires kvm-amd.avic=1.
340 (Default when IOMMU HW support is present.)
344 Do not enable amd_pstate as the default
345 scaling driver for the supported processors
347 Use amd_pstate with passive mode as a scaling driver.
348 In this mode autonomous selection is disabled.
349 Driver requests a desired performance level and platform
350 tries to match the same performance level if it is
351 satisfied by guaranteed performance level.
353 Use amd_pstate_epp driver instance as the scaling driver,
354 driver provides a hint to the hardware if software wants
355 to bias toward performance (0x0) or energy efficiency (0xff)
356 to the CPPC firmware. then CPPC power algorithm will
357 calculate the runtime workload and adjust the realtime cores
360 Activate guided autonomous mode. Driver requests minimum and
361 maximum performance level and the platform autonomously
362 selects a performance level in this range and appropriate
363 to the current workload.
365 amijoy.map= [HW,JOY] Amiga joystick support
366 Map of devices attached to JOY0DAT and JOY1DAT
368 See also Documentation/input/joydev/joystick.rst
370 analog.map= [HW,JOY] Analog joystick and gamepad support
371 Specifies type or capabilities of an analog joystick
372 connected to one of 16 gameports
373 Format: <type1>,<type2>,..<type16>
376 Power management functions (SPARCstation-4/5 + deriv.)
378 Disable APC CPU standby support. SPARCstation-Fox does
379 not play well with APC CPU idle - disable it if you have
380 APC and your system crashes randomly.
382 apic= [APIC,X86] Advanced Programmable Interrupt Controller
383 Change the output verbosity while booting
384 Format: { quiet (default) | verbose | debug }
385 Change the amount of debugging information output
386 when initialising the APIC and IO-APIC components.
387 For X86-32, this can also be used to specify an APIC
389 Format: apic=driver_name
390 Examples: apic=bigsmp
392 apic_extnmi= [APIC,X86] External NMI delivery setting
393 Format: { bsp (default) | all | none }
394 bsp: External NMI is delivered only to CPU 0
395 all: External NMIs are broadcast to all CPUs as a
397 none: External NMI is masked for all CPUs. This is
398 useful so that a dump capture kernel won't be
402 See Documentation/networking/ipv6.rst.
404 apm= [APM] Advanced Power Management
405 See header of arch/x86/kernel/apm_32.c.
407 apparmor= [APPARMOR] Disable or enable AppArmor at boot time
408 Format: { "0" | "1" }
409 See security/apparmor/Kconfig help text
412 Default value is set via kernel config option.
414 arcrimi= [HW,NET] ARCnet - "RIM I" (entirely mem-mapped) cards
415 Format: <io>,<irq>,<nodeID>
417 arm64.nobti [ARM64] Unconditionally disable Branch Target
418 Identification support
420 arm64.nopauth [ARM64] Unconditionally disable Pointer Authentication
423 arm64.nomte [ARM64] Unconditionally disable Memory Tagging Extension
426 arm64.nosve [ARM64] Unconditionally disable Scalable Vector
429 arm64.nosme [ARM64] Unconditionally disable Scalable Matrix
434 atarimouse= [HW,MOUSE] Atari Mouse
436 atkbd.extra= [HW] Enable extra LEDs and keys on IBM RapidAccess,
437 EzKey and similar keyboards
439 atkbd.reset= [HW] Reset keyboard during initialization
441 atkbd.set= [HW] Select keyboard code set
442 Format: <int> (2 = AT (default), 3 = PS/2)
444 atkbd.scroll= [HW] Enable scroll wheel on MS Office and similar
447 atkbd.softraw= [HW] Choose between synthetic and real raw mode
448 Format: <bool> (0 = real, 1 = synthetic (default))
450 atkbd.softrepeat= [HW]
451 Use software keyboard repeat
453 audit= [KNL] Enable the audit sub-system
454 Format: { "0" | "1" | "off" | "on" }
455 0 | off - kernel audit is disabled and can not be
456 enabled until the next reboot
457 unset - kernel audit is initialized but disabled and
458 will be fully enabled by the userspace auditd.
459 1 | on - kernel audit is initialized and partially
460 enabled, storing at most audit_backlog_limit
461 messages in RAM until it is fully enabled by the
465 audit_backlog_limit= [KNL] Set the audit queue size limit.
466 Format: <int> (must be >=0)
469 bau= [X86_UV] Enable the BAU on SGI UV. The default
470 behavior is to disable the BAU (i.e. bau=0).
471 Format: { "0" | "1" }
474 unset - Disable the BAU.
476 baycom_epp= [HW,AX25]
479 baycom_par= [HW,AX25] BayCom Parallel Port AX.25 Modem
481 See header of drivers/net/hamradio/baycom_par.c.
483 baycom_ser_fdx= [HW,AX25]
484 BayCom Serial Port AX.25 Modem (Full Duplex Mode)
485 Format: <io>,<irq>,<mode>[,<baud>]
486 See header of drivers/net/hamradio/baycom_ser_fdx.c.
488 baycom_ser_hdx= [HW,AX25]
489 BayCom Serial Port AX.25 Modem (Half Duplex Mode)
490 Format: <io>,<irq>,<mode>
491 See header of drivers/net/hamradio/baycom_ser_hdx.c.
494 Disable BERT OS support on buggy BIOSes.
496 bgrt_disable [ACPI][X86]
497 Disable BGRT to avoid flickering OEM logo.
499 blkdevparts= Manual partition parsing of block device(s) for
500 embedded devices based on command line input.
501 See Documentation/block/cmdline-partition.rst
503 boot_delay= Milliseconds to delay each printk during boot.
504 Only works if CONFIG_BOOT_PRINTK_DELAY is enabled,
505 and you may also have to specify "lpj=". Boot_delay
506 values larger than 10 seconds (10000) are assumed
507 erroneous and ignored.
511 Extended command line options can be added to an initrd
512 and this will cause the kernel to look for it.
514 See Documentation/admin-guide/bootconfig.rst
516 bttv.card= [HW,V4L] bttv (bt848 + bt878 based grabber cards)
517 bttv.radio= Most important insmod options are available as
519 bttv.pll= See Documentation/admin-guide/media/bttv.rst
522 bulk_remove=off [PPC] This parameter disables the use of the pSeries
523 firmware feature for flushing multiple hpte entries
526 c101= [NET] Moxa C101 synchronous serial card
528 cachesize= [BUGS=X86-32] Override level 2 CPU cache size detection.
529 Sometimes CPU hardware bugs make them report the cache
530 size incorrectly. The kernel will attempt work arounds
531 to fix known problems, but for some CPUs it is not
532 possible to determine what the correct size should be.
533 This option provides an override for these situations.
536 [NET] Specifies amount of time (in seconds) that
537 the kernel should wait for a network carrier. By default
538 it waits 120 seconds.
540 ca_keys= [KEYS] This parameter identifies a specific key(s) on
541 the system trusted keyring to be used for certificate
543 format: { id:<keyid> | builtin }
545 cca= [MIPS] Override the kernel pages' cache coherency
546 algorithm. Accepted values range from 0 to 7
547 inclusive. See arch/mips/include/asm/pgtable-bits.h
548 for platform specific values (SB1, Loongson3 and
551 ccw_timeout_log [S390]
552 See Documentation/s390/common_io.rst for details.
554 cgroup_disable= [KNL] Disable a particular controller or optional feature
555 Format: {name of the controller(s) or feature(s) to disable}
556 The effects of cgroup_disable=foo are:
557 - foo isn't auto-mounted if you mount all cgroups in
559 - foo isn't visible as an individually mountable
561 - if foo is an optional feature then the feature is
562 disabled and corresponding cgroup files are not
564 {Currently only "memory" controller deal with this and
565 cut the overhead, others just disable the usage. So
566 only cgroup_disable=memory is actually worthy}
567 Specifying "pressure" disables per-cgroup pressure
568 stall information accounting feature
570 cgroup_no_v1= [KNL] Disable cgroup controllers and named hierarchies in v1
571 Format: { { controller | "all" | "named" }
572 [,{ controller | "all" | "named" }...] }
573 Like cgroup_disable, but only applies to cgroup v1;
574 the blacklisted controllers remain available in cgroup2.
575 "all" blacklists all controllers and "named" disables
576 named mounts. Specifying both "all" and "named" disables
579 cgroup.memory= [KNL] Pass options to the cgroup memory controller.
581 nosocket -- Disable socket memory accounting.
582 nokmem -- Disable kernel memory accounting.
583 nobpf -- Disable BPF memory accounting.
585 checkreqprot= [SELINUX] Set initial checkreqprot flag value.
586 Format: { "0" | "1" }
587 See security/selinux/Kconfig help text.
588 0 -- check protection applied by kernel (includes
589 any implied execute protection).
590 1 -- check protection requested by application.
591 Default value is set via a kernel config option.
592 Value can be changed at runtime via
593 /sys/fs/selinux/checkreqprot.
594 Setting checkreqprot to 1 is deprecated.
597 See Documentation/s390/common_io.rst for details.
599 clearcpuid=X[,X...] [X86]
600 Disable CPUID feature X for the kernel. See
601 arch/x86/include/asm/cpufeatures.h for the valid bit
602 numbers X. Note the Linux-specific bits are not necessarily
603 stable over kernel options, but the vendor-specific
605 X can also be a string as appearing in the flags: line
606 in /proc/cpuinfo which does not have the above
607 instability issue. However, not all features have names
609 Note that using this option will taint your kernel.
610 Also note that user programs calling CPUID directly
611 or using the feature without checking anything
612 will still see it. This just prevents it from
613 being used by the kernel or shown in /proc/cpuinfo.
614 Also note the kernel might malfunction if you disable
619 Prevents the clock framework from automatically gating
620 clocks that have not been explicitly enabled by a Linux
621 device driver but are enabled in hardware at reset or
622 by the bootloader/firmware. Note that this does not
623 force such clocks to be always-on nor does it reserve
624 those clocks in any way. This parameter is useful for
625 debug and development, but should not be needed on a
626 platform with proper driver support. For more
627 information, see Documentation/driver-api/clk.rst.
629 clock= [BUGS=X86-32, HW] gettimeofday clocksource override.
631 Forces specified clocksource (if available) to be used
632 when calculating gettimeofday(). If specified
633 clocksource is not available, it defaults to PIT.
634 Format: { pit | tsc | cyclone | pmtmr }
636 clocksource= Override the default clocksource
638 Override the default clocksource and use the clocksource
639 with the name specified.
640 Some clocksource names to choose from, depending on
642 [all] jiffies (this is the base, fallback clocksource)
644 [ARM] imx_timer1,OSTS,netx_timer,mpu_timer2,
645 pxa_timer,timer3,32k_counter,timer0_1
646 [X86-32] pit,hpet,tsc;
647 scx200_hrt on Geode; cyclone on IBM x440
655 clocksource.arm_arch_timer.evtstrm=
658 Enable/disable the eventstream feature of the ARM
659 architected timer so that code using WFE-based polling
660 loops can be debugged more effectively on production
663 clocksource.max_cswd_read_retries= [KNL]
664 Number of clocksource_watchdog() retries due to
665 external delays before the clock will be marked
666 unstable. Defaults to two retries, that is,
667 three attempts to read the clock under test.
669 clocksource.verify_n_cpus= [KNL]
670 Limit the number of CPUs checked for clocksources
671 marked with CLOCK_SOURCE_VERIFY_PERCPU that
672 are marked unstable due to excessive skew.
673 A negative value says to check all CPUs, while
674 zero says not to check any. Values larger than
675 nr_cpu_ids are silently truncated to nr_cpu_ids.
676 The actual CPUs are chosen randomly, with
677 no replacement if the same CPU is chosen twice.
679 clocksource-wdtest.holdoff= [KNL]
680 Set the time in seconds that the clocksource
681 watchdog test waits before commencing its tests.
682 Defaults to zero when built as a module and to
683 10 seconds when built into the kernel.
685 cma=nn[MG]@[start[MG][-end[MG]]]
687 Sets the size of kernel global memory area for
688 contiguous memory allocations and optionally the
689 placement constraint by the physical address range of
690 memory allocations. A value of 0 disables CMA
691 altogether. For more information, see
692 kernel/dma/contiguous.c
696 Sets the size of kernel per-numa memory area for
697 contiguous memory allocations. A value of 0 disables
698 per-numa CMA altogether. And If this option is not
699 specified, the default value is 0.
700 With per-numa CMA enabled, DMA users on node nid will
701 first try to allocate buffer from the pernuma area
702 which is located in node nid, if the allocation fails,
703 they will fallback to the global default memory area.
705 cmo_free_hint= [PPC] Format: { yes | no }
706 Specify whether pages are marked as being inactive
707 when they are freed. This is used in CMO environments
708 to determine OS memory pressure for page stealing by
712 coherent_pool=nn[KMG] [ARM,KNL]
713 Sets the size of memory pool for coherent, atomic dma
714 allocations, by default set to 256K.
716 com20020= [HW,NET] ARCnet - COM20020 chipset
718 <io>[,<irq>[,<nodeID>[,<backplane>[,<ckp>[,<timeout>]]]]]
720 com90io= [HW,NET] ARCnet - COM90xx chipset (IO-mapped buffers)
724 ARCnet - COM90xx chipset (memory-mapped buffers)
725 Format: <io>[,<irq>[,<memstart>]]
727 condev= [HW,S390] console device
730 con3215_drop= [S390] 3215 console drop mode.
732 When set to true, drop data on the 3215 console when
733 the console buffer is full. In this case the
734 operator using a 3270 terminal emulator (for example
735 x3270) does not have to enter the clear key for the
736 console output to advance and the kernel to continue.
737 This leads to a much faster boot time when a 3270
738 terminal emulator is active. If no 3270 terminal
739 emulator is used, this parameter has no effect.
741 console= [KNL] Output console device and options.
743 tty<n> Use the virtual console device <n>.
747 Use the specified serial port. The options are of
748 the form "bbbbpnf", where "bbbb" is the baud rate,
749 "p" is parity ("n", "o", or "e"), "n" is number of
750 bits, and "f" is flow control ("r" for RTS or
751 omit it). Default is "9600n8".
753 See Documentation/admin-guide/serial-console.rst for more
755 Documentation/networking/netconsole.rst for an
758 uart[8250],io,<addr>[,options]
759 uart[8250],mmio,<addr>[,options]
760 uart[8250],mmio16,<addr>[,options]
761 uart[8250],mmio32,<addr>[,options]
762 uart[8250],0x<addr>[,options]
763 Start an early, polled-mode console on the 8250/16550
764 UART at the specified I/O port or MMIO address,
765 switching to the matching ttyS device later.
766 MMIO inter-register address stride is either 8-bit
767 (mmio), 16-bit (mmio16), or 32-bit (mmio32).
768 If none of [io|mmio|mmio16|mmio32], <addr> is assumed
769 to be equivalent to 'mmio'. 'options' are specified in
770 the same format described for ttyS above; if unspecified,
771 the h/w is not re-initialized.
773 hvc<n> Use the hypervisor console device <n>. This is for
774 both Xen and PowerPC hypervisors.
777 Use to disable console output, i.e., to have kernel
778 console messages discarded.
779 This must be the only console= parameter used on the
782 If the device connected to the port is not a TTY but a braille
783 device, prepend "brl," before the device type, for instance
785 For now, only VisioBraille is supported.
788 [KNL] Change console messages format
790 By default we print messages on consoles in
791 "[time stamp] text\n" format (time stamp may not be
792 printed, depending on CONFIG_PRINTK_TIME or
793 `printk_time' param).
795 Switch to syslog format: "<%u>[time stamp] text\n"
796 IOW, each message will have a facility and loglevel
797 prefix. The format is similar to one used by syslog()
798 syscall, or to executing "dmesg -S --raw" or to reading
801 consoleblank= [KNL] The console blank (screen saver) timeout in
802 seconds. A value of 0 disables the blank timer.
806 [KNL] Change the default value for
807 /proc/<pid>/coredump_filter.
808 See also Documentation/filesystems/proc.rst.
810 coresight_cpu_debug.enable
813 Enable/disable the CPU sampling based debugging.
814 0: default value, disable debugging
815 1: enable debugging at boot time
817 cpcihp_generic= [HW,PCI] Generic port I/O CompactPCI driver
819 <first_slot>,<last_slot>,<port>,<enum_bit>[,<debug>]
821 cpu0_hotplug [X86] Turn on CPU0 hotplug feature when
822 CONFIG_BOOTPARAM_HOTPLUG_CPU0 is off.
823 Some features depend on CPU0. Known dependencies are:
824 1. Resume from suspend/hibernate depends on CPU0.
825 Suspend/hibernate will fail if CPU0 is offline and you
826 need to online CPU0 before suspend/hibernate.
827 2. PIC interrupts also depend on CPU0. CPU0 can't be
828 removed if a PIC interrupt is detected.
829 It's said poweroff/reboot may depend on CPU0 on some
830 machines although I haven't seen such issues so far
831 after CPU0 is offline on a few tested machines.
832 If the dependencies are under your control, you can
833 turn on cpu0_hotplug.
835 cpuidle.off=1 [CPU_IDLE]
836 disable the cpuidle sub-system
839 [CPU_IDLE] Name of the cpuidle governor to use.
841 cpufreq.off=1 [CPU_FREQ]
842 disable the cpufreq sub-system
844 cpufreq.default_governor=
845 [CPU_FREQ] Name of the default cpufreq governor or
846 policy to use. This governor must be registered in the
847 kernel before the cpufreq driver probes.
850 [X86] Delay for N microsec between assert and de-assert
851 of APIC INIT to start processors. This delay occurs
852 on every CPU online, such as boot, and resume from suspend.
855 crash_kexec_post_notifiers
856 Run kdump after running panic-notifiers and dumping
857 kmsg. This only for the users who doubt kdump always
858 succeeds in any situation.
859 Note that this also increases risks of kdump failure,
860 because some panic notifiers can make the crashed
861 kernel more unstable.
863 crashkernel=size[KMG][@offset[KMG]]
864 [KNL] Using kexec, Linux can switch to a 'crash kernel'
865 upon panic. This parameter reserves the physical
866 memory region [offset, offset + size] for that kernel
867 image. If '@offset' is omitted, then a suitable offset
868 is selected automatically.
869 [KNL, X86-64, ARM64] Select a region under 4G first, and
870 fall back to reserve region above 4G when '@offset'
871 hasn't been specified.
872 See Documentation/admin-guide/kdump/kdump.rst for further details.
874 crashkernel=range1:size1[,range2:size2,...][@offset]
875 [KNL] Same as above, but depends on the memory
876 in the running system. The syntax of range is
877 start-[end] where start and end are both
878 a memory unit (amount[KMG]). See also
879 Documentation/admin-guide/kdump/kdump.rst for an example.
881 crashkernel=size[KMG],high
882 [KNL, X86-64, ARM64] range could be above 4G. Allow kernel
883 to allocate physical memory region from top, so could
884 be above 4G if system have more than 4G ram installed.
885 Otherwise memory region will be allocated below 4G, if
887 It will be ignored if crashkernel=X is specified.
888 crashkernel=size[KMG],low
889 [KNL, X86-64, ARM64] range under 4G. When crashkernel=X,high
890 is passed, kernel could allocate physical memory region
891 above 4G, that cause second kernel crash on system
892 that require some amount of low memory, e.g. swiotlb
893 requires at least 64M+32K low memory, also enough extra
894 low memory is needed to make sure DMA buffers for 32-bit
895 devices won't run out. Kernel would try to allocate
896 default size of memory below 4G automatically. The default
897 size is platform dependent.
898 --> x86: max(swiotlb_size_or_default() + 8MiB, 256MiB)
900 This one lets the user specify own low range under 4G
901 for second kernel instead.
902 0: to disable low allocation.
903 It will be ignored when crashkernel=X,high is not used
904 or memory reserved is below 4G.
907 [KNL] Disable crypto self-tests
912 cs89x0_media= [HW,NET]
913 Format: { rj45 | aui | bnc }
915 csdlock_debug= [KNL] Enable or disable debug add-ons of cross-CPU
916 function call handling. When switched on,
917 additional debug data is printed to the console
918 in case a hanging CPU is detected, and that
919 CPU is pinged again in order to try to resolve
920 the hang situation. The default value of this
921 option depends on the CSD_LOCK_WAIT_DEBUG_DEFAULT
925 See header of drivers/s390/block/dasd_devmap.c.
927 db9.dev[2|3]= [HW,JOY] Multisystem joystick support via parallel port
928 (one device per port)
929 Format: <port#>,<type>
930 See also Documentation/input/devices/joystick-parport.rst
932 debug [KNL] Enable kernel debugging (events log level).
935 [KNL] Enable printing [hashed] pointers early in the
936 boot sequence. If enabled, we use a weak hash instead
937 of siphash to hash pointers. Use this option if you are
938 seeing instances of '(___ptrval___)') and need to see a
939 value (hashed pointer) instead. Cryptographically
940 insecure, please do not use on production kernels.
943 [KNL] verbose locking self-tests
945 Print debugging info while doing the locking API
947 Bitmask for the various LOCKTYPE_ tests. Defaults to 0
948 (no extra messages), setting it to -1 (all bits set)
949 will print _a_lot_ more information - normally only
950 useful to lockdep developers.
952 debug_objects [KNL] Enable object debugging
954 debug_guardpage_minorder=
955 [KNL] When CONFIG_DEBUG_PAGEALLOC is set, this
956 parameter allows control of the order of pages that will
957 be intentionally kept free (and hence protected) by the
958 buddy allocator. Bigger value increase the probability
959 of catching random memory corruption, but reduce the
960 amount of memory for normal system use. The maximum
961 possible value is MAX_ORDER/2. Setting this parameter
962 to 1 or 2 should be enough to identify most random
963 memory corruption problems caused by bugs in kernel or
964 driver code when a CPU writes to (or reads from) a
965 random memory location. Note that there exists a class
966 of memory corruptions problems caused by buggy H/W or
967 F/W or by drivers badly programming DMA (basically when
968 memory is written at bus level and the CPU MMU is
969 bypassed) which are not detectable by
970 CONFIG_DEBUG_PAGEALLOC, hence this option will not help
971 tracking down these problems.
974 [KNL] When CONFIG_DEBUG_PAGEALLOC is set, this parameter
975 enables the feature at boot time. By default, it is
976 disabled and the system will work mostly the same as a
977 kernel built without CONFIG_DEBUG_PAGEALLOC.
978 Note: to get most of debug_pagealloc error reports, it's
979 useful to also enable the page_owner functionality.
980 on: enable the feature
982 debugfs= [KNL] This parameter enables what is exposed to userspace
983 and debugfs internal clients.
984 Format: { on, no-mount, off }
985 on: All functions are enabled.
987 Filesystem is not registered but kernel clients can
988 access APIs and a crashkernel can be used to read
989 its content. There is nothing to mount.
990 off: Filesystem is not registered and clients
991 get a -EPERM as result when trying to register files
992 or directories within debugfs.
993 This is equivalent of the runtime functionality if
994 debugfs was not enabled in the kernel at all.
995 Default value is set in build-time with a kernel configuration.
997 debugpat [X86] Enable PAT debugging
1000 [HW] The size of the default HugeTLB page. This is
1001 the size represented by the legacy /proc/ hugepages
1002 APIs. In addition, this is the default hugetlb size
1003 used for shmget(), mmap() and mounting hugetlbfs
1004 filesystems. If not specified, defaults to the
1005 architecture's default huge page size. Huge page
1006 sizes are architecture dependent. See also
1007 Documentation/admin-guide/mm/hugetlbpage.rst.
1010 deferred_probe_timeout=
1011 [KNL] Debugging option to set a timeout in seconds for
1012 deferred probe to give up waiting on dependencies to
1013 probe. Only specific dependencies (subsystems or
1014 drivers) that have opted in will be ignored. A timeout
1015 of 0 will timeout at the end of initcalls. If the time
1016 out hasn't expired, it'll be restarted by each
1017 successful driver registration. This option will also
1018 dump out devices still on the deferred probe list after
1021 delayacct [KNL] Enable per-task delay accounting
1023 dell_smm_hwmon.ignore_dmi=
1024 [HW] Continue probing hardware even if DMI data
1025 indicates that the driver is running on unsupported
1028 dell_smm_hwmon.force=
1029 [HW] Activate driver even if SMM BIOS signature does
1030 not match list of supported models and enable otherwise
1031 blacklisted features.
1033 dell_smm_hwmon.power_status=
1034 [HW] Report power status in /proc/i8k
1035 (disabled by default).
1037 dell_smm_hwmon.restricted=
1038 [HW] Allow controlling fans only if SYS_ADMIN
1041 dell_smm_hwmon.fan_mult=
1042 [HW] Factor to multiply fan speed with.
1044 dell_smm_hwmon.fan_max=
1045 [HW] Maximum configurable fan speed.
1048 Format: { on | off | def_only | inf_only | always }
1049 on: s390 zlib hardware support for compression on
1050 level 1 and decompression (default)
1051 off: No s390 zlib hardware support
1052 def_only: s390 zlib hardware support for deflate
1053 only (compression on level 1)
1054 inf_only: s390 zlib hardware support for inflate
1055 only (decompression)
1056 always: Same as 'on' but ignores the selected compression
1057 level always using hardware support (used for debugging)
1059 dhash_entries= [KNL]
1060 Set number of hash buckets for dentry cache.
1062 disable_1tb_segments [PPC]
1063 Disables the use of 1TB hash page table segments. This
1064 causes the kernel to fall back to 256MB segments which
1065 can be useful when debugging issues that require an SLB
1069 See Documentation/networking/ipv6.rst.
1072 Disable RADIX MMU mode on POWER9
1075 Disable TLBIE instruction. Currently does not work
1076 with KVM, with HASH MMU, or with coherent accelerators.
1078 disable_cpu_apicid= [X86,APIC,SMP]
1080 The number of initial APIC ID for the
1081 corresponding CPU to be disabled at boot,
1082 mostly used for the kdump 2nd kernel to
1083 disable BSP to wake up multiple CPUs without
1084 causing system reset or hang due to sending
1085 INIT from AP to BSP.
1087 disable_ddw [PPC/PSERIES]
1088 Disable Dynamic DMA Window support. Use this
1089 to workaround buggy firmware.
1091 disable_ipv6= [IPV6]
1092 See Documentation/networking/ipv6.rst.
1094 disable_mtrr_cleanup [X86]
1095 The kernel tries to adjust MTRR layout from continuous
1096 to discrete, to make X server driver able to add WB
1097 entry later. This parameter disables that.
1099 disable_mtrr_trim [X86, Intel and AMD only]
1100 By default the kernel will trim any uncacheable
1101 memory out of your available memory pool based on
1102 MTRR settings. This parameter disables that behavior,
1103 possibly causing your machine to run very slowly.
1105 disable_timer_pin_1 [X86]
1106 Disable PIN 1 of APIC timer
1107 Can be useful to work around chipset bugs.
1109 dis_ucode_ldr [X86] Disable the microcode loader.
1111 dma_debug=off If the kernel is compiled with DMA_API_DEBUG support,
1112 this option disables the debugging code at boot.
1114 dma_debug_entries=<number>
1115 This option allows to tune the number of preallocated
1116 entries for DMA-API debugging code. One entry is
1117 required per DMA-API allocation. Use this if the
1118 DMA-API debugging code disables itself because the
1119 architectural default is too low.
1121 dma_debug_driver=<driver_name>
1122 With this option the DMA-API debugging driver
1123 filter feature can be enabled at boot time. Just
1124 pass the driver to filter for as the parameter.
1125 The filter can be disabled or changed to another
1126 driver later using sysfs.
1128 driver_async_probe= [KNL]
1129 List of driver names to be probed asynchronously. *
1130 matches with all driver names. If * is specified, the
1131 rest of the listed driver names are those that will NOT
1133 Format: <driver_name1>,<driver_name2>...
1135 drm.edid_firmware=[<connector>:]<file>[,[<connector>:]<file>]
1136 Broken monitors, graphic adapters, KVMs and EDIDless
1137 panels may send no or incorrect EDID data sets.
1138 This parameter allows to specify an EDID data sets
1139 in the /lib/firmware directory that are used instead.
1140 Generic built-in EDID data sets are used, if one of
1141 edid/1024x768.bin, edid/1280x1024.bin,
1142 edid/1680x1050.bin, or edid/1920x1080.bin is given
1143 and no file with the same name exists. Details and
1144 instructions how to build your own EDID data are
1145 available in Documentation/admin-guide/edid.rst. An EDID
1146 data set will only be used for a particular connector,
1147 if its name and a colon are prepended to the EDID
1148 name. Each connector may use a unique EDID data
1149 set by separating the files with a comma. An EDID
1150 data set with no connector name will be used for
1151 any connectors not explicitly specified.
1156 Format: {"off" | "known"}
1157 Control how the dt_cpu_ftrs device-tree binding is
1158 used for CPU feature discovery and setup (if it
1160 off: Do not use it, fall back to legacy cpu table.
1161 known: Do not pass through unknown features to guests
1162 or userspace, only those that the kernel is aware of.
1164 dump_apple_properties [X86]
1165 Dump name and content of EFI device properties on
1166 x86 Macs. Useful for driver authors to determine
1167 what data is available or for reverse-engineering.
1169 dyndbg[="val"] [KNL,DYNAMIC_DEBUG]
1170 <module>.dyndbg[="val"]
1171 Enable debug messages at boot time. See
1172 Documentation/admin-guide/dynamic-debug-howto.rst
1175 early_ioremap_debug [KNL]
1176 Enable debug messages in early_ioremap support. This
1177 is useful for tracking down temporary early mappings
1178 which are not unmapped.
1180 earlycon= [KNL] Output early console device and options.
1182 When used with no options, the early console is
1183 determined by stdout-path property in device tree's
1184 chosen node or the ACPI SPCR table if supported by
1187 cdns,<addr>[,options]
1188 Start an early, polled-mode console on a Cadence
1189 (xuartps) serial port at the specified address. Only
1190 supported option is baud rate. If baud rate is not
1191 specified, the serial port must already be setup and
1194 uart[8250],io,<addr>[,options[,uartclk]]
1195 uart[8250],mmio,<addr>[,options[,uartclk]]
1196 uart[8250],mmio32,<addr>[,options[,uartclk]]
1197 uart[8250],mmio32be,<addr>[,options[,uartclk]]
1198 uart[8250],0x<addr>[,options]
1199 Start an early, polled-mode console on the 8250/16550
1200 UART at the specified I/O port or MMIO address.
1201 MMIO inter-register address stride is either 8-bit
1202 (mmio) or 32-bit (mmio32 or mmio32be).
1203 If none of [io|mmio|mmio32|mmio32be], <addr> is assumed
1204 to be equivalent to 'mmio'. 'options' are specified
1205 in the same format described for "console=ttyS<n>"; if
1206 unspecified, the h/w is not initialized. 'uartclk' is
1207 the uart clock frequency; if unspecified, it is set
1208 to 'BASE_BAUD' * 16.
1212 Start an early, polled-mode console on a pl011 serial
1213 port at the specified address. The pl011 serial port
1214 must already be setup and configured. Options are not
1215 yet supported. If 'mmio32' is specified, then only
1216 the driver will use only 32-bit accessors to read/write
1217 the device registers.
1220 Start an early console on a litex serial port at the
1221 specified address. The serial port must already be
1222 setup and configured. Options are not yet supported.
1225 Start an early, polled-mode console on a meson serial
1226 port at the specified address. The serial port must
1227 already be setup and configured. Options are not yet
1231 Start an early, polled-mode console on an msm serial
1232 port at the specified address. The serial port
1233 must already be setup and configured. Options are not
1236 msm_serial_dm,<addr>
1237 Start an early, polled-mode console on an msm serial
1238 dm port at the specified address. The serial port
1239 must already be setup and configured. Options are not
1243 Start an early, polled-mode console on a serial port
1244 of an Actions Semi SoC, such as S500 or S900, at the
1245 specified address. The serial port must already be
1246 setup and configured. Options are not yet supported.
1249 Start an early, polled-mode console on a serial port
1250 of an RDA Micro SoC, such as RDA8810PL, at the
1251 specified address. The serial port must already be
1252 setup and configured. Options are not yet supported.
1255 Use RISC-V SBI (Supervisor Binary Interface) for early
1258 smh Use ARM semihosting calls for early console.
1266 Use early console provided by serial driver available
1267 on Samsung SoCs, requires selecting proper type and
1268 a correct base address of the selected UART port. The
1269 serial port must already be setup and configured.
1270 Options are not yet supported.
1273 Start an early, polled-mode console on a lantiq serial
1274 (lqasc) port at the specified address. The serial port
1275 must already be setup and configured. Options are not
1280 Use early console provided by Freescale LP UART driver
1281 found on Freescale Vybrid and QorIQ LS1021A processors.
1282 A valid base address must be provided, and the serial
1283 port must already be setup and configured.
1287 Start an early, polled-mode, output-only console on the
1288 Freescale i.MX UART at the specified address. The UART
1289 must already be setup and configured.
1292 Start an early, polled-mode console on the
1293 Armada 3700 serial port at the specified
1294 address. The serial port must already be setup
1295 and configured. Options are not yet supported.
1298 Start an early, polled-mode console on a Qualcomm
1299 Generic Interface (GENI) based serial port at the
1300 specified address. The serial port must already be
1301 setup and configured. Options are not yet supported.
1304 Start an early, unaccelerated console on the EFI
1305 memory mapped framebuffer (if available). On cache
1306 coherent non-x86 systems that use system memory for
1307 the framebuffer, pass the 'ram' option so that it is
1308 mapped with the correct attributes.
1311 Use early console provided by Freescale LINFlexD UART
1312 serial driver for NXP S32V234 SoCs. A valid base
1313 address must be provided, and the serial port must
1314 already be setup and configured.
1316 earlyprintk= [X86,SH,ARM,M68k,S390]
1320 earlyprintk=serial[,ttySn[,baudrate]]
1321 earlyprintk=serial[,0x...[,baudrate]]
1322 earlyprintk=ttySn[,baudrate]
1323 earlyprintk=dbgp[debugController#]
1324 earlyprintk=pciserial[,force],bus:device.function[,baudrate]
1325 earlyprintk=xdbc[xhciController#]
1327 earlyprintk is useful when the kernel crashes before
1328 the normal console is initialized. It is not enabled by
1329 default because it has some cosmetic problems.
1331 Append ",keep" to not disable it when the real console
1334 Only one of vga, serial, or usb debug port can
1337 Currently only ttyS0 and ttyS1 may be specified by
1338 name. Other I/O ports may be explicitly specified
1339 on some architectures (x86 and arm at least) by
1340 replacing ttySn with an I/O port address, like this:
1341 earlyprintk=serial,0x1008,115200
1342 You can find the port for a given device in
1343 /proc/tty/driver/serial:
1344 2: uart:ST16650V2 port:00001008 irq:18 ...
1346 Interaction with the standard serial driver is not
1349 The VGA output is eventually overwritten by
1352 The xen option can only be used in Xen domains.
1354 The sclp output can only be used on s390.
1356 The optional "force" to "pciserial" enables use of a
1357 PCI device even when its classcode is not of the
1360 edac_report= [HW,EDAC] Control how to report EDAC event
1361 Format: {"on" | "off" | "force"}
1362 on: enable EDAC to report H/W event. May be overridden
1363 by other higher priority error reporting module.
1364 off: disable H/W event reporting through EDAC.
1365 force: enforce the use of EDAC to report H/W event.
1369 Format: {"off" | "on" | "skip[mbr]"}
1372 Format: { "debug", "disable_early_pci_dma",
1373 "nochunk", "noruntime", "nosoftreserve",
1374 "novamap", "no_disable_early_pci_dma" }
1375 debug: enable misc debug output.
1376 disable_early_pci_dma: disable the busmaster bit on all
1377 PCI bridges while in the EFI boot stub.
1378 nochunk: disable reading files in "chunks" in the EFI
1379 boot stub, as chunking can cause problems with some
1380 firmware implementations.
1381 noruntime : disable EFI runtime services support
1382 nosoftreserve: The EFI_MEMORY_SP (Specific Purpose)
1383 attribute may cause the kernel to reserve the
1384 memory range for a memory mapping driver to
1385 claim. Specify efi=nosoftreserve to disable this
1386 reservation and treat the memory by its base type
1387 (i.e. EFI_CONVENTIONAL_MEMORY / "System RAM").
1388 novamap: do not call SetVirtualAddressMap().
1389 no_disable_early_pci_dma: Leave the busmaster bit set
1390 on all PCI bridges while in the EFI boot stub
1392 efi_no_storage_paranoia [EFI; X86]
1393 Using this parameter you can use more than 50% of
1394 your efi variable storage. Use this parameter only if
1395 you are really sure that your UEFI does sane gc and
1396 fulfills the spec otherwise your board may brick.
1398 efi_fake_mem= nn[KMG]@ss[KMG]:aa[,nn[KMG]@ss[KMG]:aa,..] [EFI; X86]
1399 Add arbitrary attribute to specific memory range by
1400 updating original EFI memory map.
1401 Region of memory which aa attribute is added to is
1404 If efi_fake_mem=2G@4G:0x10000,2G@0x10a0000000:0x10000
1405 is specified, EFI_MEMORY_MORE_RELIABLE(0x10000)
1406 attribute is added to range 0x100000000-0x180000000 and
1407 0x10a0000000-0x1120000000.
1409 If efi_fake_mem=8G@9G:0x40000 is specified, the
1410 EFI_MEMORY_SP(0x40000) attribute is added to
1411 range 0x240000000-0x43fffffff.
1413 Using this parameter you can do debugging of EFI memmap
1414 related features. For example, you can do debugging of
1415 Address Range Mirroring feature even if your box
1416 doesn't support it, or mark specific memory as
1419 efivar_ssdt= [EFI; X86] Name of an EFI variable that contains an SSDT
1420 that is to be dynamically loaded by Linux. If there are
1421 multiple variables with the same name but with different
1422 vendor GUIDs, all of them will be loaded. See
1423 Documentation/admin-guide/acpi/ssdt-overlays.rst for details.
1426 eisa_irq_edge= [PARISC,HW]
1427 See header of drivers/parisc/eisa.c.
1429 ekgdboc= [X86,KGDB] Allow early kernel console debugging
1432 This is designed to be used in conjunction with
1433 the boot argument: earlyprintk=vga
1435 This parameter works in place of the kgdboc parameter
1436 but can only be used if the backing tty is available
1437 very early in the boot process. For early debugging
1438 via a serial port see kgdboc_earlycon instead.
1441 See comment before function elanfreq_setup() in
1442 arch/x86/kernel/cpu/cpufreq/elanfreq.c.
1444 elfcorehdr=[size[KMG]@]offset[KMG] [IA64,PPC,SH,X86,S390]
1445 Specifies physical address of start of kernel core
1446 image elf header and optionally the size. Generally
1447 kexec loader will pass this option to capture kernel.
1448 See Documentation/admin-guide/kdump/kdump.rst for details.
1450 enable_mtrr_cleanup [X86]
1451 The kernel tries to adjust MTRR layout from continuous
1452 to discrete, to make X server driver able to add WB
1453 entry later. This parameter enables that.
1455 enable_timer_pin_1 [X86]
1456 Enable PIN 1 of APIC timer
1457 Can be useful to work around chipset bugs
1458 (in particular on some ATI chipsets).
1459 The kernel tries to set a reasonable default.
1461 enforcing= [SELINUX] Set initial enforcing status.
1463 See security/selinux/Kconfig help text.
1464 0 -- permissive (log only, no denials).
1465 1 -- enforcing (deny and log).
1467 Value can be changed at runtime via
1468 /sys/fs/selinux/enforce.
1471 Disable Error Record Serialization Table (ERST)
1474 ether= [HW,NET] Ethernet cards parameters
1475 This option is obsoleted by the "netdev=" option, which
1476 has equivalent usage. See its documentation for details.
1480 Permit 'security.evm' to be updated regardless of
1481 current integrity status.
1483 early_page_ext [KNL] Enforces page_ext initialization to earlier
1484 stages so cover more early boot allocations.
1485 Please note that as side effect some optimizations
1486 might be disabled to achieve that (e.g. parallelized
1487 memory initialization is disabled) so the boot process
1488 might take longer, especially on systems with a lot of
1489 memory. Available with CONFIG_PAGE_EXTENSION=y.
1494 fail_make_request=[KNL]
1495 General fault injection mechanism.
1496 Format: <interval>,<probability>,<space>,<times>
1497 See also Documentation/fault-injection/.
1500 Format: { initns | none }
1501 See Documentation/admin-guide/sysctl/net.rst for
1502 fb_tunnels_only_for_init_ns
1505 See Documentation/admin-guide/blockdev/floppy.rst.
1507 force_pal_cache_flush
1508 [IA-64] Avoid check_sal_cache_flush which may hang on
1509 buggy SAL_CACHE_FLUSH implementations. Using this
1510 parameter will force ia64_sal_cache_flush to call
1511 ia64_pal_cache_flush instead of SAL_CACHE_FLUSH.
1514 Forcefully enable Physical Address Extension (PAE).
1515 Many Pentium M systems disable PAE but may have a
1516 functionally usable PAE implementation.
1517 Warning: use of this parameter will taint the kernel
1518 and may cause unknown problems.
1521 [FTRACE] will set and start the specified tracer
1522 as early as possible in order to facilitate early
1525 ftrace_boot_snapshot
1526 [FTRACE] On boot up, a snapshot will be taken of the
1527 ftrace ring buffer that can be read at:
1528 /sys/kernel/tracing/snapshot.
1529 This is useful if you need tracing information from kernel
1530 boot up that is likely to be overridden by user space
1531 start up functionality.
1533 Optionally, the snapshot can also be defined for a tracing
1534 instance that was created by the trace_instance= command
1537 trace_instance=foo,sched_switch ftrace_boot_snapshot=foo
1539 The above will cause the "foo" tracing instance to trigger
1540 a snapshot at the end of boot up.
1542 ftrace_dump_on_oops[=orig_cpu]
1543 [FTRACE] will dump the trace buffers on oops.
1544 If no parameter is passed, ftrace will dump
1545 buffers of all CPUs, but if you pass orig_cpu, it will
1546 dump only the buffer of the CPU that triggered the
1549 ftrace_filter=[function-list]
1550 [FTRACE] Limit the functions traced by the function
1551 tracer at boot up. function-list is a comma-separated
1552 list of functions. This list can be changed at run
1553 time by the set_ftrace_filter file in the debugfs
1556 ftrace_notrace=[function-list]
1557 [FTRACE] Do not trace the functions specified in
1558 function-list. This list can be changed at run time
1559 by the set_ftrace_notrace file in the debugfs
1562 ftrace_graph_filter=[function-list]
1563 [FTRACE] Limit the top level callers functions traced
1564 by the function graph tracer at boot up.
1565 function-list is a comma-separated list of functions
1566 that can be changed at run time by the
1567 set_graph_function file in the debugfs tracing directory.
1569 ftrace_graph_notrace=[function-list]
1570 [FTRACE] Do not trace from the functions specified in
1571 function-list. This list is a comma-separated list of
1572 functions that can be changed at run time by the
1573 set_graph_notrace file in the debugfs tracing directory.
1575 ftrace_graph_max_depth=<uint>
1576 [FTRACE] Used with the function graph tracer. This is
1577 the max depth it will trace into a function. This value
1578 can be changed at run time by the max_graph_depth file
1579 in the tracefs tracing directory. default: 0 (no limit)
1581 fw_devlink= [KNL] Create device links between consumer and supplier
1582 devices by scanning the firmware to infer the
1583 consumer/supplier relationships. This feature is
1584 especially useful when drivers are loaded as modules as
1585 it ensures proper ordering of tasks like device probing
1586 (suppliers first, then consumers), supplier boot state
1587 clean up (only after all consumers have probed),
1588 suspend/resume & runtime PM (consumers first, then
1590 Format: { off | permissive | on | rpm }
1591 off -- Don't create device links from firmware info.
1592 permissive -- Create device links from firmware info
1593 but use it only for ordering boot state clean
1594 up (sync_state() calls).
1595 on -- Create device links from firmware info and use it
1596 to enforce probe and suspend/resume ordering.
1597 rpm -- Like "on", but also use to order runtime PM.
1599 fw_devlink.strict=<bool>
1600 [KNL] Treat all inferred dependencies as mandatory
1601 dependencies. This only applies for fw_devlink=on|rpm.
1604 fw_devlink.sync_state =
1605 [KNL] When all devices that could probe have finished
1606 probing, this parameter controls what to do with
1607 devices that haven't yet received their sync_state()
1609 Format: { strict | timeout }
1610 strict -- Default. Continue waiting on consumers to
1612 timeout -- Give up waiting on consumers and call
1613 sync_state() on any devices that haven't yet
1614 received their sync_state() calls after
1615 deferred_probe_timeout has expired or by
1616 late_initcall() if !CONFIG_MODULES.
1619 [HW,JOY] Multisystem joystick and NES/SNES/PSX pad
1620 support via parallel port (up to 5 devices per port)
1621 Format: <port#>,<pad1>,<pad2>,<pad3>,<pad4>,<pad5>
1622 See also Documentation/input/devices/joystick-parport.rst
1626 gart_fix_e820= [X86-64] disable the fix e820 for K8 GART
1630 gcov_persist= [GCOV] When non-zero (default), profiling data for
1631 kernel modules is saved and remains accessible via
1632 debugfs, even when the module is unloaded/reloaded.
1633 When zero, profiling data is discarded and associated
1634 debugfs files are removed at module unload time.
1636 goldfish [X86] Enable the goldfish android emulator platform.
1637 Don't use this when you are not running on the
1640 gpio-mockup.gpio_mockup_ranges
1641 [HW] Sets the ranges of gpiochip of for this device.
1642 Format: <start1>,<end1>,<start2>,<end2>...
1643 gpio-mockup.gpio_mockup_named_lines
1644 [HW] Let the driver know GPIO lines should be named.
1646 gpt [EFI] Forces disk with valid GPT signature but
1647 invalid Protective MBR to be treated as GPT. If the
1648 primary GPT is corrupted, it enables the backup/alternate
1649 GPT to be used instead.
1651 grcan.enable0= [HW] Configuration of physical interface 0. Determines
1652 the "Enable 0" bit of the configuration register.
1655 grcan.enable1= [HW] Configuration of physical interface 1. Determines
1656 the "Enable 0" bit of the configuration register.
1659 grcan.select= [HW] Select which physical interface to use.
1662 grcan.txsize= [HW] Sets the size of the tx buffer.
1663 Format: <unsigned int> such that (txsize & ~0x1fffc0) == 0.
1665 grcan.rxsize= [HW] Sets the size of the rx buffer.
1666 Format: <unsigned int> such that (rxsize & ~0x1fffc0) == 0.
1670 [KNL] Under CONFIG_HARDENED_USERCOPY, whether
1671 hardening is enabled for this boot. Hardened
1672 usercopy checking is used to protect the kernel
1673 from reading or writing beyond known memory
1674 allocation boundaries as a proactive defense
1675 against bounds-checking flaws in the kernel's
1676 copy_to_user()/copy_from_user() interface.
1677 on Perform hardened usercopy checks (default).
1678 off Disable hardened usercopy checks.
1680 hardlockup_all_cpu_backtrace=
1681 [KNL] Should the hard-lockup detector generate
1682 backtraces on all cpus.
1685 hashdist= [KNL,NUMA] Large hashes allocated during boot
1686 are distributed across NUMA nodes. Defaults on
1687 for 64-bit NUMA, off otherwise.
1688 Format: 0 | 1 (for off | on)
1690 hcl= [IA-64] SGI's Hardware Graph compatibility layer
1692 hd= [EIDE] (E)IDE hard drive subsystem geometry
1693 Format: <cyl>,<head>,<sect>
1696 Disable Hardware Error Source Table (HEST) support;
1697 corresponding firmware-first mode error processing
1698 logic will be disabled.
1700 hibernate= [HIBERNATION]
1701 noresume Don't check if there's a hibernation image
1702 present during boot.
1703 nocompress Don't compress/decompress hibernation images.
1704 no Disable hibernation and resume.
1705 protect_image Turn on image protection during restoration
1706 (that will set all pages holding image data
1707 during restoration read-only).
1709 highmem=nn[KMG] [KNL,BOOT] forces the highmem zone to have an exact
1710 size of <nn>. This works even on boxes that have no
1711 highmem otherwise. This also works to reduce highmem
1712 size on bigger boxes.
1714 highres= [KNL] Enable/disable high resolution timer mode.
1715 Valid parameters: "on", "off"
1720 hostname= [KNL] Set the hostname (aka UTS nodename).
1722 This allows setting the system's hostname during early
1723 startup. This sets the name returned by gethostname.
1724 Using this parameter to set the hostname makes it
1725 possible to ensure the hostname is correctly set before
1726 any userspace processes run, avoiding the possibility
1727 that a process may call gethostname before the hostname
1728 has been explicitly set, resulting in the calling
1729 process getting an incorrect result. The string must
1730 not exceed the maximum allowed hostname length (usually
1731 64 characters) and will be truncated otherwise.
1733 hpet= [X86-32,HPET] option to control HPET usage
1734 Format: { enable (default) | disable | force |
1736 disable: disable HPET and use PIT instead
1737 force: allow force enabled of undocumented chips (ICH4,
1739 verbose: show contents of HPET registers during setup
1741 hpet_mmap= [X86, HPET_MMAP] Allow userspace to mmap HPET
1742 registers. Default set by CONFIG_HPET_MMAP_DEFAULT.
1744 hugepages= [HW] Number of HugeTLB pages to allocate at boot.
1745 If this follows hugepagesz (below), it specifies
1746 the number of pages of hugepagesz to be allocated.
1747 If this is the first HugeTLB parameter on the command
1748 line, it specifies the number of pages to allocate for
1749 the default huge page size. If using node format, the
1750 number of pages to allocate per-node can be specified.
1751 See also Documentation/admin-guide/mm/hugetlbpage.rst.
1752 Format: <integer> or (node format)
1753 <node>:<integer>[,<node>:<integer>]
1756 [HW] The size of the HugeTLB pages. This is used in
1757 conjunction with hugepages (above) to allocate huge
1758 pages of a specific size at boot. The pair
1759 hugepagesz=X hugepages=Y can be specified once for
1760 each supported huge page size. Huge page sizes are
1761 architecture dependent. See also
1762 Documentation/admin-guide/mm/hugetlbpage.rst.
1765 hugetlb_cma= [HW,CMA] The size of a CMA area used for allocation
1766 of gigantic hugepages. Or using node format, the size
1767 of a CMA area per node can be specified.
1768 Format: nn[KMGTPE] or (node format)
1769 <node>:nn[KMGTPE][,<node>:nn[KMGTPE]]
1771 Reserve a CMA area of given size and allocate gigantic
1772 hugepages using the CMA allocator. If enabled, the
1773 boot-time allocation of gigantic hugepages is skipped.
1775 hugetlb_free_vmemmap=
1776 [KNL] Requires CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP
1778 Control if HugeTLB Vmemmap Optimization (HVO) is enabled.
1779 Allows heavy hugetlb users to free up some more
1780 memory (7 * PAGE_SIZE for each 2MB hugetlb page).
1781 Format: { on | off (default) }
1786 Built with CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP_DEFAULT_ON=y,
1789 Note that the vmemmap pages may be allocated from the added
1790 memory block itself when memory_hotplug.memmap_on_memory is
1791 enabled, those vmemmap pages cannot be optimized even if this
1792 feature is enabled. Other vmemmap pages not allocated from
1793 the added memory block itself do not be affected.
1796 [KNL] Should the hung task detector generate panics.
1799 A value of 1 instructs the kernel to panic when a
1800 hung task is detected. The default value is controlled
1801 by the CONFIG_BOOTPARAM_HUNG_TASK_PANIC build-time
1802 option. The value selected by this boot parameter can
1803 be changed later by the kernel.hung_task_panic sysctl.
1805 hvc_iucv= [S390] Number of z/VM IUCV hypervisor console (HVC)
1806 terminal devices. Valid values: 0..8
1807 hvc_iucv_allow= [S390] Comma-separated list of z/VM user IDs.
1808 If specified, z/VM IUCV HVC accepts connections
1809 from listed z/VM user IDs only.
1811 hv_nopvspin [X86,HYPER_V] Disables the paravirt spinlock optimizations
1812 which allow the hypervisor to 'idle' the
1813 guest on lock contention.
1815 i2c_bus= [HW] Override the default board specific I2C bus speed
1816 or register an additional I2C bus that is not
1817 registered from board initialization code.
1821 i8042.debug [HW] Toggle i8042 debug mode
1822 i8042.unmask_kbd_data
1823 [HW] Enable printing of interrupt data from the KBD port
1824 (disabled by default, and as a pre-condition
1825 requires that i8042.debug=1 be enabled)
1826 i8042.direct [HW] Put keyboard port into non-translated mode
1827 i8042.dumbkbd [HW] Pretend that controller can only read data from
1828 keyboard and cannot control its state
1829 (Don't attempt to blink the leds)
1830 i8042.noaux [HW] Don't check for auxiliary (== mouse) port
1831 i8042.nokbd [HW] Don't check/create keyboard port
1832 i8042.noloop [HW] Disable the AUX Loopback command while probing
1834 i8042.nomux [HW] Don't check presence of an active multiplexing
1836 i8042.nopnp [HW] Don't use ACPIPnP / PnPBIOS to discover KBD/AUX
1838 i8042.notimeout [HW] Ignore timeout condition signalled by controller
1839 i8042.reset [HW] Reset the controller during init, cleanup and
1840 suspend-to-ram transitions, only during s2r
1841 transitions, or never reset
1842 Format: { 1 | Y | y | 0 | N | n }
1843 1, Y, y: always reset controller
1844 0, N, n: don't ever reset controller
1845 Default: only on s2r transitions on x86; most other
1846 architectures force reset to be always executed
1847 i8042.unlock [HW] Unlock (ignore) the keylock
1848 i8042.kbdreset [HW] Reset device connected to KBD port
1850 [HW] Allow deferred probing upon i8042 probe errors
1854 i915.invert_brightness=
1855 [DRM] Invert the sense of the variable that is used to
1856 set the brightness of the panel backlight. Normally a
1857 brightness value of 0 indicates backlight switched off,
1858 and the maximum of the brightness value sets the backlight
1859 to maximum brightness. If this parameter is set to 0
1860 (default) and the machine requires it, or this parameter
1861 is set to 1, a brightness value of 0 sets the backlight
1862 to maximum brightness, and the maximum of the brightness
1863 value switches the backlight off.
1864 -1 -- never invert brightness
1865 0 -- machine default
1866 1 -- force brightness inversion
1869 Format: <io>[,<membase>[,<icn_id>[,<icn_id2>]]]
1873 Format: idle=poll, idle=halt, idle=nomwait
1874 Poll forces a polling idle loop that can slightly
1875 improve the performance of waking up a idle CPU, but
1876 will use a lot of power and make the system run hot.
1878 idle=halt: Halt is forced to be used for CPU idle.
1879 In such case C2/C3 won't be used again.
1880 idle=nomwait: Disable mwait for CPU C-states
1884 Allow force disabling of Shared Virtual Memory (SVA)
1885 support for the idxd driver. By default it is set to
1888 idxd.tc_override= [HW]
1890 Allow override of default traffic class configuration
1891 for the device. By default it is set to false (0).
1893 ieee754= [MIPS] Select IEEE Std 754 conformance mode
1894 Format: { strict | legacy | 2008 | relaxed }
1897 Choose which programs will be accepted for execution
1898 based on the IEEE 754 NaN encoding(s) supported by
1899 the FPU and the NaN encoding requested with the value
1900 of an ELF file header flag individually set by each
1901 binary. Hardware implementations are permitted to
1902 support either or both of the legacy and the 2008 NaN
1905 Available settings are as follows:
1906 strict accept binaries that request a NaN encoding
1907 supported by the FPU
1908 legacy only accept legacy-NaN binaries, if supported
1910 2008 only accept 2008-NaN binaries, if supported
1912 relaxed accept any binaries regardless of whether
1913 supported by the FPU
1915 The FPU emulator is always able to support both NaN
1916 encodings, so if no FPU hardware is present or it has
1917 been disabled with 'nofpu', then the settings of
1918 'legacy' and '2008' strap the emulator accordingly,
1919 'relaxed' straps the emulator for both legacy-NaN and
1920 2008-NaN, whereas 'strict' enables legacy-NaN only on
1921 legacy processors and both NaN encodings on MIPS32 or
1924 The setting for ABS.fmt/NEG.fmt instruction execution
1925 mode generally follows that for the NaN encoding,
1926 except where unsupported by hardware.
1928 ignore_loglevel [KNL]
1929 Ignore loglevel setting - this will print /all/
1930 kernel messages to the console. Useful for debugging.
1931 We also add it as printk module parameter, so users
1932 could change it dynamically, usually by
1933 /sys/module/printk/parameters/ignore_loglevel.
1936 Ignore RLIMIT_DATA setting for data mappings,
1937 print warning at first misuse. Can be changed via
1938 /sys/module/kernel/parameters/ignore_rlimit_data.
1940 ihash_entries= [KNL]
1941 Set number of hash buckets for inode cache.
1943 ima_appraise= [IMA] appraise integrity measurements
1944 Format: { "off" | "enforce" | "fix" | "log" }
1947 ima_appraise_tcb [IMA] Deprecated. Use ima_policy= instead.
1948 The builtin appraise policy appraises all files
1951 ima_canonical_fmt [IMA]
1952 Use the canonical format for the binary runtime
1953 measurements, instead of host native format.
1956 Format: { md5 | sha1 | rmd160 | sha256 | sha384
1960 The list of supported hash algorithms is defined
1961 in crypto/hash_info.h.
1964 The builtin policies to load during IMA setup.
1965 Format: "tcb | appraise_tcb | secure_boot |
1966 fail_securely | critical_data"
1968 The "tcb" policy measures all programs exec'd, files
1969 mmap'd for exec, and all files opened with the read
1970 mode bit set by either the effective uid (euid=0) or
1973 The "appraise_tcb" policy appraises the integrity of
1974 all files owned by root.
1976 The "secure_boot" policy appraises the integrity
1977 of files (eg. kexec kernel image, kernel modules,
1978 firmware, policy, etc) based on file signatures.
1980 The "fail_securely" policy forces file signature
1981 verification failure also on privileged mounted
1982 filesystems with the SB_I_UNVERIFIABLE_SIGNATURE
1985 The "critical_data" policy measures kernel integrity
1988 ima_tcb [IMA] Deprecated. Use ima_policy= instead.
1989 Load a policy which meets the needs of the Trusted
1990 Computing Base. This means IMA will measure all
1991 programs exec'd, files mmap'd for exec, and all files
1992 opened for read by uid=0.
1995 Select one of defined IMA measurements template formats.
1996 Formats: { "ima" | "ima-ng" | "ima-ngv2" | "ima-sig" |
2001 [IMA] Define a custom template format.
2002 Format: { "field1|...|fieldN" }
2004 ima.ahash_minsize= [IMA] Minimum file size for asynchronous hash usage
2005 Format: <min_file_size>
2006 Set the minimal file size for using asynchronous hash.
2007 If left unspecified, ahash usage is disabled.
2009 ahash performance varies for different data sizes on
2010 different crypto accelerators. This option can be used
2011 to achieve the best performance for a particular HW.
2013 ima.ahash_bufsize= [IMA] Asynchronous hash buffer size
2015 Set hashing buffer size. Default: 4k.
2017 ahash performance varies for different chunk sizes on
2018 different crypto accelerators. This option can be used
2019 to achieve best performance for particular HW.
2023 Run specified binary instead of /sbin/init as init
2026 initcall_debug [KNL] Trace initcalls as they are executed. Useful
2027 for working out where the kernel is dying during
2030 initcall_blacklist= [KNL] Do not execute a comma-separated list of
2031 initcall functions. Useful for debugging built-in
2032 modules and initcalls.
2034 initramfs_async= [KNL]
2037 This parameter controls whether the initramfs
2038 image is unpacked asynchronously, concurrently
2039 with devices being probed and
2040 initialized. This should normally just work,
2041 but as a debugging aid, one can get the
2042 historical behaviour of the initramfs
2043 unpacking being completed before device_ and
2046 initrd= [BOOT] Specify the location of the initial ramdisk
2048 initrdmem= [KNL] Specify a physical address and size from which to
2049 load the initrd. If an initrd is compiled in or
2050 specified in the bootparams, it takes priority over this
2052 Format: ss[KMG],nn[KMG]
2055 init_on_alloc= [MM] Fill newly allocated pages and heap objects with
2058 Default set by CONFIG_INIT_ON_ALLOC_DEFAULT_ON.
2060 init_on_free= [MM] Fill freed pages and heap objects with zeroes.
2062 Default set by CONFIG_INIT_ON_FREE_DEFAULT_ON.
2064 init_pkru= [X86] Specify the default memory protection keys rights
2065 register contents for all processes. 0x55555554 by
2066 default (disallow access to all but pkey 0). Can
2067 override in debugfs after boot.
2069 inport.irq= [HW] Inport (ATI XL and Microsoft) busmouse driver
2072 int_pln_enable [X86] Enable power limit notification interrupt
2074 integrity_audit=[IMA]
2075 Format: { "0" | "1" }
2076 0 -- basic integrity auditing messages. (Default)
2077 1 -- additional integrity auditing messages.
2079 intel_iommu= [DMAR] Intel IOMMU driver (DMAR) option
2081 Enable intel iommu driver.
2083 Disable intel iommu driver.
2084 igfx_off [Default Off]
2085 By default, gfx is mapped as normal device. If a gfx
2086 device has a dedicated DMAR unit, the DMAR unit is
2087 bypassed by not enabling DMAR with this option. In
2088 this case, gfx device will use physical address for
2090 strict [Default Off]
2091 Deprecated, equivalent to iommu.strict=1.
2092 sp_off [Default Off]
2093 By default, super page will be supported if Intel IOMMU
2094 has the capability. With this option, super page will
2097 Enable the Intel IOMMU scalable mode if the hardware
2098 advertises that it has support for the scalable mode
2101 Disallow use of the Intel IOMMU scalable mode.
2102 tboot_noforce [Default Off]
2103 Do not force the Intel IOMMU enabled under tboot.
2104 By default, tboot will force Intel IOMMU on, which
2105 could harm performance of some high-throughput
2106 devices like 40GBit network cards, even if identity
2108 Note that using this option lowers the security
2109 provided by tboot because it makes the system
2110 vulnerable to DMA attacks.
2112 intel_idle.max_cstate= [KNL,HW,ACPI,X86]
2113 0 disables intel_idle and fall back on acpi_idle.
2114 1 to 9 specify maximum depth of C-state.
2118 Do not enable intel_pstate as the default
2119 scaling driver for the supported processors
2121 Use intel_pstate as a scaling driver, but configure it
2122 to work with generic cpufreq governors (instead of
2123 enabling its internal governor). This mode cannot be
2124 used along with the hardware-managed P-states (HWP)
2127 Enable intel_pstate on systems that prohibit it by default
2128 in favor of acpi-cpufreq. Forcing the intel_pstate driver
2129 instead of acpi-cpufreq may disable platform features, such
2130 as thermal controls and power capping, that rely on ACPI
2131 P-States information being indicated to OSPM and therefore
2132 should be used with caution. This option does not work with
2133 processors that aren't supported by the intel_pstate driver
2134 or on platforms that use pcc-cpufreq instead of acpi-cpufreq.
2136 Do not enable hardware P state control (HWP)
2139 Only load intel_pstate on systems which support
2140 hardware P state control (HWP) if available.
2142 Enforce ACPI _PPC performance limits. If the Fixed ACPI
2143 Description Table, specifies preferred power management
2144 profile as "Enterprise Server" or "Performance Server",
2145 then this feature is turned on by default.
2147 Allow per-logical-CPU P-State performance control limits using
2148 cpufreq sysfs interface
2150 intremap= [X86-64, Intel-IOMMU]
2151 on enable Interrupt Remapping (default)
2152 off disable Interrupt Remapping
2153 nosid disable Source ID checking
2155 BIOS x2APIC opt-out request will be ignored
2156 nopost disable Interrupt Posting
2158 iomem= Disable strict checking of access to MMIO memory
2159 strict regions from userspace.
2174 nobypass [PPC/POWERNV]
2175 Disable IOMMU bypass, using IOMMU for PCI devices.
2177 iommu.forcedac= [ARM64, X86] Control IOVA allocation for PCI devices.
2178 Format: { "0" | "1" }
2179 0 - Try to allocate a 32-bit DMA address first, before
2180 falling back to the full range if needed.
2181 1 - Allocate directly from the full usable range,
2182 forcing Dual Address Cycle for PCI cards supporting
2183 greater than 32-bit addressing.
2185 iommu.strict= [ARM64, X86] Configure TLB invalidation behaviour
2186 Format: { "0" | "1" }
2188 Request that DMA unmap operations use deferred
2189 invalidation of hardware TLBs, for increased
2190 throughput at the cost of reduced device isolation.
2191 Will fall back to strict mode if not supported by
2192 the relevant IOMMU driver.
2194 DMA unmap operations invalidate IOMMU hardware TLBs
2196 unset - Use value of CONFIG_IOMMU_DEFAULT_DMA_{LAZY,STRICT}.
2197 Note: on x86, strict mode specified via one of the
2198 legacy driver-specific options takes precedence.
2201 [ARM64, X86] Configure DMA to bypass the IOMMU by default.
2202 Format: { "0" | "1" }
2203 0 - Use IOMMU translation for DMA.
2204 1 - Bypass the IOMMU for DMA.
2205 unset - Use value of CONFIG_IOMMU_DEFAULT_PASSTHROUGH.
2207 io7= [HW] IO7 for Marvel-based Alpha systems
2208 See comment before marvel_specify_io7 in
2209 arch/alpha/kernel/core_marvel.c.
2211 io_delay= [X86] I/O delay method
2213 Standard port 0x80 based delay
2215 Alternate port 0xed based delay (needed on some systems)
2217 Simple two microseconds delay
2222 See Documentation/admin-guide/nfs/nfsroot.rst.
2224 ipcmni_extend [KNL] Extend the maximum number of unique System V
2225 IPC identifiers from 32,768 to 16,777,216.
2227 irqaffinity= [SMP] Set the default irq affinity mask
2228 The argument is a cpu list, as described above.
2230 irqchip.gicv2_force_probe=
2233 Force the kernel to look for the second 4kB page
2234 of a GICv2 controller even if the memory range
2235 exposed by the device tree is too small.
2237 irqchip.gicv3_nolpi=
2239 Force the kernel to ignore the availability of
2240 LPIs (and by consequence ITSs). Intended for system
2241 that use the kernel as a bootloader, and thus want
2242 to let secondary kernels in charge of setting up
2245 irqchip.gicv3_pseudo_nmi= [ARM64]
2246 Enables support for pseudo-NMIs in the kernel. This
2247 requires the kernel to be built with
2248 CONFIG_ARM64_PSEUDO_NMI.
2251 When an interrupt is not handled search all handlers
2252 for it. Intended to get systems with badly broken
2256 When an interrupt is not handled search all handlers
2257 for it. Also check all handlers each timer
2258 interrupt. Intended to get systems with badly broken
2262 Format: <RDP>,<reset>,<pci_scan>,<verbosity>
2264 isolcpus= [KNL,SMP,ISOL] Isolate a given set of CPUs from disturbance.
2265 [Deprecated - use cpusets instead]
2266 Format: [flag-list,]<cpu-list>
2268 Specify one or more CPUs to isolate from disturbances
2269 specified in the flag list (default: domain):
2272 Disable the tick when a single task runs.
2274 A residual 1Hz tick is offloaded to workqueues, which you
2275 need to affine to housekeeping through the global
2276 workqueue's affinity configured via the
2277 /sys/devices/virtual/workqueue/cpumask sysfs file, or
2278 by using the 'domain' flag described below.
2280 NOTE: by default the global workqueue runs on all CPUs,
2281 so to protect individual CPUs the 'cpumask' file has to
2282 be configured manually after bootup.
2285 Isolate from the general SMP balancing and scheduling
2286 algorithms. Note that performing domain isolation this way
2287 is irreversible: it's not possible to bring back a CPU to
2288 the domains once isolated through isolcpus. It's strongly
2289 advised to use cpusets instead to disable scheduler load
2290 balancing through the "cpuset.sched_load_balance" file.
2291 It offers a much more flexible interface where CPUs can
2292 move in and out of an isolated set anytime.
2294 You can move a process onto or off an "isolated" CPU via
2295 the CPU affinity syscalls or cpuset.
2296 <cpu number> begins at 0 and the maximum value is
2297 "number of CPUs in system - 1".
2301 Isolate from being targeted by managed interrupts
2302 which have an interrupt mask containing isolated
2303 CPUs. The affinity of managed interrupts is
2304 handled by the kernel and cannot be changed via
2305 the /proc/irq/* interfaces.
2307 This isolation is best effort and only effective
2308 if the automatically assigned interrupt mask of a
2309 device queue contains isolated and housekeeping
2310 CPUs. If housekeeping CPUs are online then such
2311 interrupts are directed to the housekeeping CPU
2312 so that IO submitted on the housekeeping CPU
2313 cannot disturb the isolated CPU.
2315 If a queue's affinity mask contains only isolated
2316 CPUs then this parameter has no effect on the
2317 interrupt routing decision, though interrupts are
2318 only delivered when tasks running on those
2319 isolated CPUs submit IO. IO submitted on
2320 housekeeping CPUs has no influence on those
2323 The format of <cpu-list> is described above.
2327 ivrs_ioapic [HW,X86-64]
2328 Provide an override to the IOAPIC-ID<->DEVICE-ID
2329 mapping provided in the IVRS ACPI table.
2330 By default, PCI segment is 0, and can be omitted.
2332 For example, to map IOAPIC-ID decimal 10 to
2333 PCI segment 0x1 and PCI device 00:14.0,
2334 write the parameter as:
2335 ivrs_ioapic=10@0001:00:14.0
2338 * To map IOAPIC-ID decimal 10 to PCI device 00:14.0
2339 write the parameter as:
2340 ivrs_ioapic[10]=00:14.0
2341 * To map IOAPIC-ID decimal 10 to PCI segment 0x1 and
2342 PCI device 00:14.0 write the parameter as:
2343 ivrs_ioapic[10]=0001:00:14.0
2345 ivrs_hpet [HW,X86-64]
2346 Provide an override to the HPET-ID<->DEVICE-ID
2347 mapping provided in the IVRS ACPI table.
2348 By default, PCI segment is 0, and can be omitted.
2350 For example, to map HPET-ID decimal 10 to
2351 PCI segment 0x1 and PCI device 00:14.0,
2352 write the parameter as:
2353 ivrs_hpet=10@0001:00:14.0
2356 * To map HPET-ID decimal 0 to PCI device 00:14.0
2357 write the parameter as:
2358 ivrs_hpet[0]=00:14.0
2359 * To map HPET-ID decimal 10 to PCI segment 0x1 and
2360 PCI device 00:14.0 write the parameter as:
2361 ivrs_ioapic[10]=0001:00:14.0
2363 ivrs_acpihid [HW,X86-64]
2364 Provide an override to the ACPI-HID:UID<->DEVICE-ID
2365 mapping provided in the IVRS ACPI table.
2366 By default, PCI segment is 0, and can be omitted.
2368 For example, to map UART-HID:UID AMD0020:0 to
2369 PCI segment 0x1 and PCI device ID 00:14.5,
2370 write the parameter as:
2371 ivrs_acpihid=AMD0020:0@0001:00:14.5
2374 * To map UART-HID:UID AMD0020:0 to PCI segment is 0,
2375 PCI device ID 00:14.5, write the parameter as:
2376 ivrs_acpihid[00:14.5]=AMD0020:0
2377 * To map UART-HID:UID AMD0020:0 to PCI segment 0x1 and
2378 PCI device ID 00:14.5, write the parameter as:
2379 ivrs_acpihid[0001:00:14.5]=AMD0020:0
2381 js= [HW,JOY] Analog joystick
2382 See Documentation/input/joydev/joystick.rst.
2385 [KNL] Enforce KASAN (Kernel Address Sanitizer) to print
2386 report on every invalid memory access. Without this
2387 parameter KASAN will print report only for the first
2391 Do not unregister boot console at start. This is only
2392 useful for debugging when something happens in the window
2393 between unregistering the boot console and initializing
2398 kernelcore= [KNL,X86,IA-64,PPC]
2399 Format: nn[KMGTPE] | nn% | "mirror"
2400 This parameter specifies the amount of memory usable by
2401 the kernel for non-movable allocations. The requested
2402 amount is spread evenly throughout all nodes in the
2403 system as ZONE_NORMAL. The remaining memory is used for
2404 movable memory in its own zone, ZONE_MOVABLE. In the
2405 event, a node is too small to have both ZONE_NORMAL and
2406 ZONE_MOVABLE, kernelcore memory will take priority and
2407 other nodes will have a larger ZONE_MOVABLE.
2409 ZONE_MOVABLE is used for the allocation of pages that
2410 may be reclaimed or moved by the page migration
2411 subsystem. Note that allocations like PTEs-from-HighMem
2412 still use the HighMem zone if it exists, and the Normal
2413 zone if it does not.
2415 It is possible to specify the exact amount of memory in
2416 the form of "nn[KMGTPE]", a percentage of total system
2417 memory in the form of "nn%", or "mirror". If "mirror"
2418 option is specified, mirrored (reliable) memory is used
2419 for non-movable allocations and remaining memory is used
2420 for Movable pages. "nn[KMGTPE]", "nn%", and "mirror"
2421 are exclusive, so you cannot specify multiple forms.
2423 kgdbdbgp= [KGDB,HW] kgdb over EHCI usb debug port.
2424 Format: <Controller#>[,poll interval]
2425 The controller # is the number of the ehci usb debug
2426 port as it is probed via PCI. The poll interval is
2427 optional and is the number seconds in between
2428 each poll cycle to the debug port in case you need
2429 the functionality for interrupting the kernel with
2430 gdb or control-c on the dbgp connection. When
2431 not using this parameter you use sysrq-g to break into
2432 the kernel debugger.
2434 kgdboc= [KGDB,HW] kgdb over consoles.
2435 Requires a tty driver that supports console polling,
2436 or a supported polling keyboard driver (non-usb).
2437 Serial only format: <serial_device>[,baud]
2438 keyboard only format: kbd
2439 keyboard and serial format: kbd,<serial_device>[,baud]
2440 Optional Kernel mode setting:
2441 kms, kbd format: kms,kbd
2442 kms, kbd and serial format: kms,kbd,<ser_dev>[,baud]
2444 kgdboc_earlycon= [KGDB,HW]
2445 If the boot console provides the ability to read
2446 characters and can work in polling mode, you can use
2447 this parameter to tell kgdb to use it as a backend
2448 until the normal console is registered. Intended to
2449 be used together with the kgdboc parameter which
2450 specifies the normal console to transition to.
2452 The name of the early console should be specified
2453 as the value of this parameter. Note that the name of
2454 the early console might be different than the tty
2455 name passed to kgdboc. It's OK to leave the value
2456 blank and the first boot console that implements
2457 read() will be picked.
2459 kgdbwait [KGDB] Stop kernel execution and enter the
2460 kernel debugger at the earliest opportunity.
2462 kmac= [MIPS] Korina ethernet MAC address.
2463 Configure the RouterBoard 532 series on-chip
2464 Ethernet adapter MAC address.
2466 kmemleak= [KNL] Boot-time kmemleak enable/disable
2467 Valid arguments: on, off
2469 Built with CONFIG_DEBUG_KMEMLEAK_DEFAULT_OFF=y,
2472 kprobe_event=[probe-list]
2473 [FTRACE] Add kprobe events and enable at boot time.
2474 The probe-list is a semicolon delimited list of probe
2475 definitions. Each definition is same as kprobe_events
2476 interface, but the parameters are comma delimited.
2477 For example, to add a kprobe event on vfs_read with
2478 arg1 and arg2, add to the command line;
2480 kprobe_event=p,vfs_read,$arg1,$arg2
2482 See also Documentation/trace/kprobetrace.rst "Kernel
2483 Boot Parameter" section.
2485 kpti= [ARM64] Control page table isolation of user
2486 and kernel address spaces.
2487 Default: enabled on cores which need mitigation.
2491 kunit.enable= [KUNIT] Enable executing KUnit tests. Requires
2492 CONFIG_KUNIT to be set to be fully enabled. The
2493 default value can be overridden via
2494 KUNIT_DEFAULT_ENABLED.
2495 Default is 1 (enabled)
2497 kvm.ignore_msrs=[KVM] Ignore guest accesses to unhandled MSRs.
2498 Default is 0 (don't ignore, but inject #GP)
2500 kvm.eager_page_split=
2501 [KVM,X86] Controls whether or not KVM will try to
2502 proactively split all huge pages during dirty logging.
2503 Eager page splitting reduces interruptions to vCPU
2504 execution by eliminating the write-protection faults
2505 and MMU lock contention that would otherwise be
2506 required to split huge pages lazily.
2508 VM workloads that rarely perform writes or that write
2509 only to a small region of VM memory may benefit from
2510 disabling eager page splitting to allow huge pages to
2511 still be used for reads.
2513 The behavior of eager page splitting depends on whether
2514 KVM_DIRTY_LOG_INITIALLY_SET is enabled or disabled. If
2515 disabled, all huge pages in a memslot will be eagerly
2516 split when dirty logging is enabled on that memslot. If
2517 enabled, eager page splitting will be performed during
2518 the KVM_CLEAR_DIRTY ioctl, and only for the pages being
2521 Eager page splitting is only supported when kvm.tdp_mmu=Y.
2525 kvm.enable_vmware_backdoor=[KVM] Support VMware backdoor PV interface.
2526 Default is false (don't support).
2529 [KVM] Controls the software workaround for the
2530 X86_BUG_ITLB_MULTIHIT bug.
2531 force : Always deploy workaround.
2532 off : Never deploy workaround.
2533 auto : Deploy workaround based on the presence of
2534 X86_BUG_ITLB_MULTIHIT.
2538 If the software workaround is enabled for the host,
2539 guests do need not to enable it for nested guests.
2541 kvm.nx_huge_pages_recovery_ratio=
2542 [KVM] Controls how many 4KiB pages are periodically zapped
2543 back to huge pages. 0 disables the recovery, otherwise if
2544 the value is N KVM will zap 1/Nth of the 4KiB pages every
2545 period (see below). The default is 60.
2547 kvm.nx_huge_pages_recovery_period_ms=
2548 [KVM] Controls the time period at which KVM zaps 4KiB pages
2549 back to huge pages. If the value is a non-zero N, KVM will
2550 zap a portion (see ratio above) of the pages every N msecs.
2551 If the value is 0 (the default), KVM will pick a period based
2552 on the ratio, such that a page is zapped after 1 hour on average.
2554 kvm-amd.nested= [KVM,AMD] Allow nested virtualization in KVM/SVM.
2555 Default is 1 (enabled)
2557 kvm-amd.npt= [KVM,AMD] Disable nested paging (virtualized MMU)
2559 Default is 1 (enabled) if in 64-bit or 32-bit PAE mode.
2562 [KVM,ARM] Select one of KVM/arm64's modes of operation.
2564 none: Forcefully disable KVM.
2566 nvhe: Standard nVHE-based mode, without support for
2569 protected: nVHE-based mode with support for guests whose
2570 state is kept private from the host.
2572 nested: VHE-based mode with support for nested
2573 virtualization. Requires at least ARMv8.3
2576 Defaults to VHE/nVHE based on hardware support. Setting
2577 mode to "protected" will disable kexec and hibernation
2578 for the host. "nested" is experimental and should be
2579 used with extreme caution.
2581 kvm-arm.vgic_v3_group0_trap=
2582 [KVM,ARM] Trap guest accesses to GICv3 group-0
2585 kvm-arm.vgic_v3_group1_trap=
2586 [KVM,ARM] Trap guest accesses to GICv3 group-1
2589 kvm-arm.vgic_v3_common_trap=
2590 [KVM,ARM] Trap guest accesses to GICv3 common
2593 kvm-arm.vgic_v4_enable=
2594 [KVM,ARM] Allow use of GICv4 for direct injection of
2597 kvm_cma_resv_ratio=n [PPC]
2598 Reserves given percentage from system memory area for
2599 contiguous memory allocation for KVM hash pagetable
2601 By default it reserves 5% of total system memory.
2605 kvm-intel.ept= [KVM,Intel] Disable extended page tables
2606 (virtualized MMU) support on capable Intel chips.
2607 Default is 1 (enabled)
2609 kvm-intel.emulate_invalid_guest_state=
2610 [KVM,Intel] Disable emulation of invalid guest state.
2611 Ignored if kvm-intel.enable_unrestricted_guest=1, as
2612 guest state is never invalid for unrestricted guests.
2613 This param doesn't apply to nested guests (L2), as KVM
2614 never emulates invalid L2 guest state.
2615 Default is 1 (enabled)
2617 kvm-intel.flexpriority=
2618 [KVM,Intel] Disable FlexPriority feature (TPR shadow).
2619 Default is 1 (enabled)
2622 [KVM,Intel] Enable VMX nesting (nVMX).
2623 Default is 0 (disabled)
2625 kvm-intel.unrestricted_guest=
2626 [KVM,Intel] Disable unrestricted guest feature
2627 (virtualized real and unpaged mode) on capable
2628 Intel chips. Default is 1 (enabled)
2630 kvm-intel.vmentry_l1d_flush=[KVM,Intel] Mitigation for L1 Terminal Fault
2633 Valid arguments: never, cond, always
2635 always: L1D cache flush on every VMENTER.
2636 cond: Flush L1D on VMENTER only when the code between
2637 VMEXIT and VMENTER can leak host memory.
2638 never: Disables the mitigation
2640 Default is cond (do L1 cache flush in specific instances)
2642 kvm-intel.vpid= [KVM,Intel] Disable Virtual Processor Identification
2643 feature (tagged TLBs) on capable Intel chips.
2644 Default is 1 (enabled)
2646 l1d_flush= [X86,INTEL]
2647 Control mitigation for L1D based snooping vulnerability.
2649 Certain CPUs are vulnerable to an exploit against CPU
2650 internal buffers which can forward information to a
2651 disclosure gadget under certain conditions.
2653 In vulnerable processors, the speculatively
2654 forwarded data can be used in a cache side channel
2655 attack, to access data to which the attacker does
2656 not have direct access.
2658 This parameter controls the mitigation. The
2661 on - enable the interface for the mitigation
2663 l1tf= [X86] Control mitigation of the L1TF vulnerability on
2666 The kernel PTE inversion protection is unconditionally
2667 enabled and cannot be disabled.
2670 Provides all available mitigations for the
2671 L1TF vulnerability. Disables SMT and
2672 enables all mitigations in the
2673 hypervisors, i.e. unconditional L1D flush.
2675 SMT control and L1D flush control via the
2676 sysfs interface is still possible after
2677 boot. Hypervisors will issue a warning
2678 when the first VM is started in a
2679 potentially insecure configuration,
2680 i.e. SMT enabled or L1D flush disabled.
2683 Same as 'full', but disables SMT and L1D
2684 flush runtime control. Implies the
2685 'nosmt=force' command line option.
2686 (i.e. sysfs control of SMT is disabled.)
2689 Leaves SMT enabled and enables the default
2690 hypervisor mitigation, i.e. conditional
2693 SMT control and L1D flush control via the
2694 sysfs interface is still possible after
2695 boot. Hypervisors will issue a warning
2696 when the first VM is started in a
2697 potentially insecure configuration,
2698 i.e. SMT enabled or L1D flush disabled.
2702 Disables SMT and enables the default
2703 hypervisor mitigation.
2705 SMT control and L1D flush control via the
2706 sysfs interface is still possible after
2707 boot. Hypervisors will issue a warning
2708 when the first VM is started in a
2709 potentially insecure configuration,
2710 i.e. SMT enabled or L1D flush disabled.
2713 Same as 'flush', but hypervisors will not
2714 warn when a VM is started in a potentially
2715 insecure configuration.
2718 Disables hypervisor mitigations and doesn't
2720 It also drops the swap size and available
2721 RAM limit restriction on both hypervisor and
2726 For details see: Documentation/admin-guide/hw-vuln/l1tf.rst
2732 lapic [X86-32,APIC] Enable the local APIC even if BIOS
2735 lapic= [X86,APIC] Do not use TSC deadline
2736 value for LAPIC timer one-shot implementation. Default
2737 back to the programmable timer unit in the LAPIC.
2738 Format: notscdeadline
2740 lapic_timer_c2_ok [X86,APIC] trust the local apic timer
2743 libata.dma= [LIBATA] DMA control
2744 libata.dma=0 Disable all PATA and SATA DMA
2745 libata.dma=1 PATA and SATA Disk DMA only
2746 libata.dma=2 ATAPI (CDROM) DMA only
2747 libata.dma=4 Compact Flash DMA only
2748 Combinations also work, so libata.dma=3 enables DMA
2749 for disks and CDROMs, but not CFs.
2751 libata.ignore_hpa= [LIBATA] Ignore HPA limit
2752 libata.ignore_hpa=0 keep BIOS limits (default)
2753 libata.ignore_hpa=1 ignore limits, using full disk
2755 libata.noacpi [LIBATA] Disables use of ACPI in libata suspend/resume
2759 libata.force= [LIBATA] Force configurations. The format is a comma-
2760 separated list of "[ID:]VAL" where ID is PORT[.DEVICE].
2761 PORT and DEVICE are decimal numbers matching port, link
2762 or device. Basically, it matches the ATA ID string
2763 printed on console by libata. If the whole ID part is
2764 omitted, the last PORT and DEVICE values are used. If
2765 ID hasn't been specified yet, the configuration applies
2766 to all ports, links and devices.
2768 If only DEVICE is omitted, the parameter applies to
2769 the port and all links and devices behind it. DEVICE
2770 number of 0 either selects the first device or the
2771 first fan-out link behind PMP device. It does not
2772 select the host link. DEVICE number of 15 selects the
2773 host link and device attached to it.
2775 The VAL specifies the configuration to force. As long
2776 as there is no ambiguity, shortcut notation is allowed.
2777 For example, both 1.5 and 1.5G would work for 1.5Gbps.
2778 The following configurations can be forced.
2780 * Cable type: 40c, 80c, short40c, unk, ign or sata.
2781 Any ID with matching PORT is used.
2783 * SATA link speed limit: 1.5Gbps or 3.0Gbps.
2785 * Transfer mode: pio[0-7], mwdma[0-4] and udma[0-7].
2786 udma[/][16,25,33,44,66,100,133] notation is also
2789 * nohrst, nosrst, norst: suppress hard, soft and both
2792 * rstonce: only attempt one reset during hot-unplug
2795 * [no]dbdelay: Enable or disable the extra 200ms delay
2796 before debouncing a link PHY and device presence
2799 * [no]ncq: Turn on or off NCQ.
2801 * [no]ncqtrim: Enable or disable queued DSM TRIM.
2803 * [no]ncqati: Enable or disable NCQ trim on ATI chipset.
2805 * [no]trim: Enable or disable (unqueued) TRIM.
2807 * trim_zero: Indicate that TRIM command zeroes data.
2809 * max_trim_128m: Set 128M maximum trim size limit.
2811 * [no]dma: Turn on or off DMA transfers.
2813 * atapi_dmadir: Enable ATAPI DMADIR bridge support.
2815 * atapi_mod16_dma: Enable the use of ATAPI DMA for
2816 commands that are not a multiple of 16 bytes.
2818 * [no]dmalog: Enable or disable the use of the
2819 READ LOG DMA EXT command to access logs.
2821 * [no]iddevlog: Enable or disable access to the
2822 identify device data log.
2824 * [no]logdir: Enable or disable access to the general
2825 purpose log directory.
2827 * max_sec_128: Set transfer size limit to 128 sectors.
2829 * max_sec_1024: Set or clear transfer size limit to
2832 * max_sec_lba48: Set or clear transfer size limit to
2835 * [no]lpm: Enable or disable link power management.
2837 * [no]setxfer: Indicate if transfer speed mode setting
2840 * [no]fua: Disable or enable FUA (Force Unit Access)
2841 support for devices supporting this feature.
2843 * dump_id: Dump IDENTIFY data.
2845 * disable: Disable this device.
2847 If there are multiple matching configurations changing
2848 the same attribute, the last one is used.
2850 load_ramdisk= [RAM] [Deprecated]
2852 lockd.nlm_grace_period=P [NFS] Assign grace period.
2855 lockd.nlm_tcpport=N [NFS] Assign TCP port.
2858 lockd.nlm_timeout=T [NFS] Assign timeout value.
2861 lockd.nlm_udpport=M [NFS] Assign UDP port.
2864 lockdown= [SECURITY]
2865 { integrity | confidentiality }
2866 Enable the kernel lockdown feature. If set to
2867 integrity, kernel features that allow userland to
2868 modify the running kernel are disabled. If set to
2869 confidentiality, kernel features that allow userland
2870 to extract confidential information from the kernel
2873 locktorture.nreaders_stress= [KNL]
2874 Set the number of locking read-acquisition kthreads.
2875 Defaults to being automatically set based on the
2876 number of online CPUs.
2878 locktorture.nwriters_stress= [KNL]
2879 Set the number of locking write-acquisition kthreads.
2881 locktorture.onoff_holdoff= [KNL]
2882 Set time (s) after boot for CPU-hotplug testing.
2884 locktorture.onoff_interval= [KNL]
2885 Set time (s) between CPU-hotplug operations, or
2886 zero to disable CPU-hotplug testing.
2888 locktorture.shuffle_interval= [KNL]
2889 Set task-shuffle interval (jiffies). Shuffling
2890 tasks allows some CPUs to go into dyntick-idle
2891 mode during the locktorture test.
2893 locktorture.shutdown_secs= [KNL]
2894 Set time (s) after boot system shutdown. This
2895 is useful for hands-off automated testing.
2897 locktorture.stat_interval= [KNL]
2898 Time (s) between statistics printk()s.
2900 locktorture.stutter= [KNL]
2901 Time (s) to stutter testing, for example,
2902 specifying five seconds causes the test to run for
2903 five seconds, wait for five seconds, and so on.
2904 This tests the locking primitive's ability to
2905 transition abruptly to and from idle.
2907 locktorture.torture_type= [KNL]
2908 Specify the locking implementation to test.
2910 locktorture.verbose= [KNL]
2911 Enable additional printk() statements.
2913 logibm.irq= [HW,MOUSE] Logitech Bus Mouse Driver
2916 loglevel= All Kernel Messages with a loglevel smaller than the
2917 console loglevel will be printed to the console. It can
2918 also be changed with klogd or other programs. The
2919 loglevels are defined as follows:
2921 0 (KERN_EMERG) system is unusable
2922 1 (KERN_ALERT) action must be taken immediately
2923 2 (KERN_CRIT) critical conditions
2924 3 (KERN_ERR) error conditions
2925 4 (KERN_WARNING) warning conditions
2926 5 (KERN_NOTICE) normal but significant condition
2927 6 (KERN_INFO) informational
2928 7 (KERN_DEBUG) debug-level messages
2930 log_buf_len=n[KMG] Sets the size of the printk ring buffer,
2931 in bytes. n must be a power of two and greater
2932 than the minimal size. The minimal size is defined
2933 by LOG_BUF_SHIFT kernel config parameter. There is
2934 also CONFIG_LOG_CPU_MAX_BUF_SHIFT config parameter
2935 that allows to increase the default size depending on
2936 the number of CPUs. See init/Kconfig for more details.
2938 logo.nologo [FB] Disables display of the built-in Linux logo.
2939 This may be used to provide more screen space for
2940 kernel log messages and is useful when debugging
2941 kernel boot problems.
2943 lp=0 [LP] Specify parallel ports to use, e.g,
2944 lp=port[,port...] lp=none,parport0 (lp0 not configured, lp1 uses
2945 lp=reset first parallel port). 'lp=0' disables the
2946 lp=auto printer driver. 'lp=reset' (which can be
2947 specified in addition to the ports) causes
2948 attached printers to be reset. Using
2949 lp=port1,port2,... specifies the parallel ports
2950 to associate lp devices with, starting with
2951 lp0. A port specification may be 'none' to skip
2952 that lp device, or a parport name such as
2953 'parport0'. Specifying 'lp=auto' instead of a
2954 port specification list means that device IDs
2955 from each port should be examined, to see if
2956 an IEEE 1284-compliant printer is attached; if
2957 so, the driver will manage that printer.
2958 See also header of drivers/char/lp.c.
2961 Sets loops_per_jiffy to given constant, thus avoiding
2962 time-consuming boot-time autodetection (up to 250 ms per
2963 CPU). 0 enables autodetection (default). To determine
2964 the correct value for your kernel, boot with normal
2965 autodetection and see what value is printed. Note that
2966 on SMP systems the preset will be applied to all CPUs,
2967 which is likely to cause problems if your CPUs need
2968 significantly divergent settings. An incorrect value
2969 will cause delays in the kernel to be wrong, leading to
2970 unpredictable I/O errors and other breakage. Although
2971 unlikely, in the extreme case this might damage your
2975 Format: <io>,<irq>,<dma>
2977 lsm.debug [SECURITY] Enable LSM initialization debugging output.
2980 [SECURITY] Choose order of LSM initialization. This
2981 overrides CONFIG_LSM, and the "security=" parameter.
2983 machvec= [IA-64] Force the use of a particular machine-vector
2984 (machvec) in a generic kernel.
2985 Example: machvec=hpzx1
2987 machtype= [Loongson] Share the same kernel image file between
2988 different yeeloong laptops.
2989 Example: machtype=lemote-yeeloong-2f-7inch
2991 max_addr=nn[KMG] [KNL,BOOT,IA-64] All physical memory greater
2992 than or equal to this physical address is ignored.
2994 maxcpus= [SMP] Maximum number of processors that an SMP kernel
2995 will bring up during bootup. maxcpus=n : n >= 0 limits
2996 the kernel to bring up 'n' processors. Surely after
2997 bootup you can bring up the other plugged cpu by executing
2998 "echo 1 > /sys/devices/system/cpu/cpuX/online". So maxcpus
2999 only takes effect during system bootup.
3000 While n=0 is a special case, it is equivalent to "nosmp",
3001 which also disables the IO APIC.
3003 max_loop= [LOOP] The number of loop block devices that get
3004 (loop.max_loop) unconditionally pre-created at init time. The default
3005 number is configured by BLK_DEV_LOOP_MIN_COUNT. Instead
3006 of statically allocating a predefined number, loop
3007 devices can be requested on-demand with the
3008 /dev/loop-control interface.
3010 mce [X86-32] Machine Check Exception
3012 mce=option [X86-64] See Documentation/arch/x86/x86_64/boot-options.rst
3014 md= [HW] RAID subsystems devices and level
3015 See Documentation/admin-guide/md.rst.
3018 Format: <first>,<last>
3019 Specifies range of consoles to be captured by the MDA.
3022 Control mitigation for the Micro-architectural Data
3023 Sampling (MDS) vulnerability.
3025 Certain CPUs are vulnerable to an exploit against CPU
3026 internal buffers which can forward information to a
3027 disclosure gadget under certain conditions.
3029 In vulnerable processors, the speculatively
3030 forwarded data can be used in a cache side channel
3031 attack, to access data to which the attacker does
3032 not have direct access.
3034 This parameter controls the MDS mitigation. The
3037 full - Enable MDS mitigation on vulnerable CPUs
3038 full,nosmt - Enable MDS mitigation and disable
3039 SMT on vulnerable CPUs
3040 off - Unconditionally disable MDS mitigation
3042 On TAA-affected machines, mds=off can be prevented by
3043 an active TAA mitigation as both vulnerabilities are
3044 mitigated with the same mechanism so in order to disable
3045 this mitigation, you need to specify tsx_async_abort=off
3048 Not specifying this option is equivalent to
3051 For details see: Documentation/admin-guide/hw-vuln/mds.rst
3053 mem=nn[KMG] [HEXAGON] Set the memory size.
3054 Must be specified, otherwise memory size will be 0.
3056 mem=nn[KMG] [KNL,BOOT] Force usage of a specific amount of memory
3057 Amount of memory to be used in cases as follows:
3060 2 when the kernel is not able to see the whole system memory;
3061 3 memory that lies after 'mem=' boundary is excluded from
3062 the hypervisor, then assigned to KVM guests.
3063 4 to limit the memory available for kdump kernel.
3065 [ARC,MICROBLAZE] - the limit applies only to low memory,
3066 high memory is not affected.
3068 [ARM64] - only limits memory covered by the linear
3069 mapping. The NOMAP regions are not affected.
3071 [X86] Work as limiting max address. Use together
3072 with memmap= to avoid physical address space collisions.
3073 Without memmap= PCI devices could be placed at addresses
3074 belonging to unused RAM.
3076 Note that this only takes effects during boot time since
3077 in above case 3, memory may need be hot added after boot
3078 if system memory of hypervisor is not sufficient.
3081 [ARM,MIPS] - override the memory layout reported by
3083 Define a memory region of size nn[KMG] starting at
3085 Multiple different regions can be specified with
3086 multiple mem= parameters on the command line.
3088 mem=nopentium [BUGS=X86-32] Disable usage of 4MB pages for kernel
3091 memblock=debug [KNL] Enable memblock debug messages.
3094 [KNL,SH] Allow user to override the default size for
3095 per-device physically contiguous DMA buffers.
3097 memhp_default_state=online/offline
3098 [KNL] Set the initial state for the memory hotplug
3099 onlining policy. If not specified, the default value is
3100 set according to the
3101 CONFIG_MEMORY_HOTPLUG_DEFAULT_ONLINE kernel config
3103 See Documentation/admin-guide/mm/memory-hotplug.rst.
3105 memmap=exactmap [KNL,X86] Enable setting of an exact
3106 E820 memory map, as specified by the user.
3107 Such memmap=exactmap lines can be constructed based on
3108 BIOS output or other requirements. See the memmap=nn@ss
3111 memmap=nn[KMG]@ss[KMG]
3112 [KNL, X86, MIPS, XTENSA] Force usage of a specific region of memory.
3113 Region of memory to be used is from ss to ss+nn.
3114 If @ss[KMG] is omitted, it is equivalent to mem=nn[KMG],
3115 which limits max address to nn[KMG].
3116 Multiple different regions can be specified,
3119 memmap=100M@2G,100M#3G,1G!1024G
3121 memmap=nn[KMG]#ss[KMG]
3122 [KNL,ACPI] Mark specific memory as ACPI data.
3123 Region of memory to be marked is from ss to ss+nn.
3125 memmap=nn[KMG]$ss[KMG]
3126 [KNL,ACPI] Mark specific memory as reserved.
3127 Region of memory to be reserved is from ss to ss+nn.
3128 Example: Exclude memory from 0x18690000-0x1869ffff
3129 memmap=64K$0x18690000
3131 memmap=0x10000$0x18690000
3132 Some bootloaders may need an escape character before '$',
3133 like Grub2, otherwise '$' and the following number
3136 memmap=nn[KMG]!ss[KMG]
3137 [KNL,X86] Mark specific memory as protected.
3138 Region of memory to be used, from ss to ss+nn.
3139 The memory region may be marked as e820 type 12 (0xc)
3140 and is NVDIMM or ADR memory.
3142 memmap=<size>%<offset>-<oldtype>+<newtype>
3143 [KNL,ACPI] Convert memory within the specified region
3144 from <oldtype> to <newtype>. If "-<oldtype>" is left
3145 out, the whole region will be marked as <newtype>,
3146 even if previously unavailable. If "+<newtype>" is left
3147 out, matching memory will be removed. Types are
3148 specified as e820 types, e.g., 1 = RAM, 2 = reserved,
3149 3 = ACPI, 12 = PRAM.
3151 memory_corruption_check=0/1 [X86]
3152 Some BIOSes seem to corrupt the first 64k of
3153 memory when doing things like suspend/resume.
3154 Setting this option will scan the memory
3155 looking for corruption. Enabling this will
3156 both detect corruption and prevent the kernel
3157 from using the memory being corrupted.
3158 However, its intended as a diagnostic tool; if
3159 repeatable BIOS-originated corruption always
3160 affects the same memory, you can use memmap=
3161 to prevent the kernel from using that memory.
3163 memory_corruption_check_size=size [X86]
3164 By default it checks for corruption in the low
3165 64k, making this memory unavailable for normal
3166 use. Use this parameter to scan for
3167 corruption in more or less memory.
3169 memory_corruption_check_period=seconds [X86]
3170 By default it checks for corruption every 60
3171 seconds. Use this parameter to check at some
3172 other rate. 0 disables periodic checking.
3174 memory_hotplug.memmap_on_memory
3175 [KNL,X86,ARM] Boolean flag to enable this feature.
3176 Format: {on | off (default)}
3177 When enabled, runtime hotplugged memory will
3178 allocate its internal metadata (struct pages,
3179 those vmemmap pages cannot be optimized even
3180 if hugetlb_free_vmemmap is enabled) from the
3181 hotadded memory which will allow to hotadd a
3182 lot of memory without requiring additional
3184 This feature is disabled by default because it
3185 has some implication on large (e.g. GB)
3186 allocations in some configurations (e.g. small
3188 The state of the flag can be read in
3189 /sys/module/memory_hotplug/parameters/memmap_on_memory.
3190 Note that even when enabled, there are a few cases where
3191 the feature is not effective.
3193 memtest= [KNL,X86,ARM,M68K,PPC,RISCV] Enable memtest
3195 default : 0 <disable>
3196 Specifies the number of memtest passes to be
3197 performed. Each pass selects another test
3198 pattern from a given set of patterns. Memtest
3199 fills the memory with this pattern, validates
3200 memory contents and reserves bad memory
3201 regions that are detected.
3203 mem_encrypt= [X86-64] AMD Secure Memory Encryption (SME) control
3204 Valid arguments: on, off
3205 Default (depends on kernel configuration option):
3206 on (CONFIG_AMD_MEM_ENCRYPT_ACTIVE_BY_DEFAULT=y)
3207 off (CONFIG_AMD_MEM_ENCRYPT_ACTIVE_BY_DEFAULT=n)
3208 mem_encrypt=on: Activate SME
3209 mem_encrypt=off: Do not activate SME
3211 Refer to Documentation/virt/kvm/x86/amd-memory-encryption.rst
3212 for details on when memory encryption can be activated.
3214 mem_sleep_default= [SUSPEND] Default system suspend mode:
3215 s2idle - Suspend-To-Idle
3216 shallow - Power-On Suspend or equivalent (if supported)
3217 deep - Suspend-To-RAM or equivalent (if supported)
3218 See Documentation/admin-guide/pm/sleep-states.rst.
3220 mfgpt_irq= [IA-32] Specify the IRQ to use for the
3221 Multi-Function General Purpose Timers on AMD Geode
3224 mfgptfix [X86-32] Fix MFGPT timers on AMD Geode platforms when
3225 the BIOS has incorrectly applied a workaround. TinyBIOS
3226 version 0.98 is known to be affected, 0.99 fixes the
3227 problem by letting the user disable the workaround.
3231 min_addr=nn[KMG] [KNL,BOOT,IA-64] All physical memory below this
3232 physical address is ignored.
3234 mini2440= [ARM,HW,KNL]
3235 Format:[0..2][b][c][t]
3237 MINI2440 configuration specification:
3238 0 - The attached screen is the 3.5" TFT
3239 1 - The attached screen is the 7" TFT
3240 2 - The VGA Shield is attached (1024x768)
3241 Leaving out the screen size parameter will not load
3242 the TFT driver, and the framebuffer will be left
3244 b - Enable backlight. The TFT backlight pin will be
3245 linked to the kernel VESA blanking code and a GPIO
3246 LED. This parameter is not necessary when using the
3248 c - Enable the s3c camera interface.
3249 t - Reserved for enabling touchscreen support. The
3250 touchscreen support is not enabled in the mainstream
3251 kernel as of 2.6.30, a preliminary port can be found
3252 in the "bleeding edge" mini2440 support kernel at
3253 https://repo.or.cz/w/linux-2.6/mini2440.git
3256 [X86,PPC,S390,ARM64] Control optional mitigations for
3257 CPU vulnerabilities. This is a set of curated,
3258 arch-independent options, each of which is an
3259 aggregation of existing arch-specific options.
3262 Disable all optional CPU mitigations. This
3263 improves system performance, but it may also
3264 expose users to several CPU vulnerabilities.
3265 Equivalent to: nopti [X86,PPC]
3266 if nokaslr then kpti=0 [ARM64]
3267 nospectre_v1 [X86,PPC]
3269 nospectre_v2 [X86,PPC,S390,ARM64]
3270 spectre_v2_user=off [X86]
3271 spec_store_bypass_disable=off [X86,PPC]
3272 ssbd=force-off [ARM64]
3273 nospectre_bhb [ARM64]
3276 tsx_async_abort=off [X86]
3277 kvm.nx_huge_pages=off [X86]
3278 srbds=off [X86,INTEL]
3279 no_entry_flush [PPC]
3280 no_uaccess_flush [PPC]
3281 mmio_stale_data=off [X86]
3285 This does not have any effect on
3286 kvm.nx_huge_pages when
3287 kvm.nx_huge_pages=force.
3290 Mitigate all CPU vulnerabilities, but leave SMT
3291 enabled, even if it's vulnerable. This is for
3292 users who don't want to be surprised by SMT
3293 getting disabled across kernel upgrades, or who
3294 have other ways of avoiding SMT-based attacks.
3295 Equivalent to: (default behavior)
3298 Mitigate all CPU vulnerabilities, disabling SMT
3299 if needed. This is for users who always want to
3300 be fully mitigated, even if it means losing SMT.
3301 Equivalent to: l1tf=flush,nosmt [X86]
3302 mds=full,nosmt [X86]
3303 tsx_async_abort=full,nosmt [X86]
3304 mmio_stale_data=full,nosmt [X86]
3305 retbleed=auto,nosmt [X86]
3308 [KNL] When CONFIG_DEBUG_MEMORY_INIT is set, this
3309 parameter allows control of the logging verbosity for
3310 the additional memory initialisation checks. A value
3311 of 0 disables mminit logging and a level of 4 will
3312 log everything. Information is printed at KERN_DEBUG
3313 so loglevel=8 may also need to be specified.
3316 [X86,INTEL] Control mitigation for the Processor
3317 MMIO Stale Data vulnerabilities.
3319 Processor MMIO Stale Data is a class of
3320 vulnerabilities that may expose data after an MMIO
3321 operation. Exposed data could originate or end in
3322 the same CPU buffers as affected by MDS and TAA.
3323 Therefore, similar to MDS and TAA, the mitigation
3324 is to clear the affected CPU buffers.
3326 This parameter controls the mitigation. The
3329 full - Enable mitigation on vulnerable CPUs
3331 full,nosmt - Enable mitigation and disable SMT on
3334 off - Unconditionally disable mitigation
3336 On MDS or TAA affected machines,
3337 mmio_stale_data=off can be prevented by an active
3338 MDS or TAA mitigation as these vulnerabilities are
3339 mitigated with the same mechanism so in order to
3340 disable this mitigation, you need to specify
3341 mds=off and tsx_async_abort=off too.
3343 Not specifying this option is equivalent to
3344 mmio_stale_data=full.
3347 Documentation/admin-guide/hw-vuln/processor_mmio_stale_data.rst
3349 <module>.async_probe[=<bool>] [KNL]
3350 If no <bool> value is specified or if the value
3351 specified is not a valid <bool>, enable asynchronous
3352 probe on this module. Otherwise, enable/disable
3353 asynchronous probe on this module as indicated by the
3354 <bool> value. See also: module.async_probe
3356 module.async_probe=<bool>
3357 [KNL] When set to true, modules will use async probing
3358 by default. To enable/disable async probing for a
3359 specific module, use the module specific control that
3360 is documented under <module>.async_probe. When both
3361 module.async_probe and <module>.async_probe are
3362 specified, <module>.async_probe takes precedence for
3363 the specific module.
3365 module.enable_dups_trace
3366 [KNL] When CONFIG_MODULE_DEBUG_AUTOLOAD_DUPS is set,
3367 this means that duplicate request_module() calls will
3368 trigger a WARN_ON() instead of a pr_warn(). Note that
3369 if MODULE_DEBUG_AUTOLOAD_DUPS_TRACE is set, WARN_ON()s
3370 will always be issued and this option does nothing.
3372 [KNL] When CONFIG_MODULE_SIG is set, this means that
3373 modules without (valid) signatures will fail to load.
3374 Note that if CONFIG_MODULE_SIG_FORCE is set, that
3375 is always true, so this option does nothing.
3377 module_blacklist= [KNL] Do not load a comma-separated list of
3378 modules. Useful for debugging problem modules.
3381 [MOUSE] Maximum time between finger touching and
3382 leaving touchpad surface for touch to be considered
3383 a tap and be reported as a left button click (for
3384 touchpads working in absolute mode only).
3386 mousedev.xres= [MOUSE] Horizontal screen resolution, used for devices
3387 reporting absolute coordinates, such as tablets
3388 mousedev.yres= [MOUSE] Vertical screen resolution, used for devices
3389 reporting absolute coordinates, such as tablets
3391 movablecore= [KNL,X86,IA-64,PPC]
3392 Format: nn[KMGTPE] | nn%
3393 This parameter is the complement to kernelcore=, it
3394 specifies the amount of memory used for migratable
3395 allocations. If both kernelcore and movablecore is
3396 specified, then kernelcore will be at *least* the
3397 specified value but may be more. If movablecore on its
3398 own is specified, the administrator must be careful
3399 that the amount of memory usable for all allocations
3402 movable_node [KNL] Boot-time switch to make hotplugable memory
3403 NUMA nodes to be movable. This means that the memory
3404 of such nodes will be usable only for movable
3405 allocations which rules out almost all kernel
3406 allocations. Use with caution!
3408 MTD_Partition= [MTD]
3409 Format: <name>,<region-number>,<size>,<offset>
3411 MTD_Region= [MTD] Format:
3412 <name>,<region-number>[,<base>,<size>,<buswidth>,<altbuswidth>]
3415 See drivers/mtd/parsers/cmdlinepart.c
3418 ARM/S3C2412 JIVE boot control
3420 See arch/arm/mach-s3c/mach-jive.c
3422 mtouchusb.raw_coordinates=
3423 [HW] Make the MicroTouch USB driver use raw coordinates
3424 ('y', default) or cooked coordinates ('n')
3426 mtrr_chunk_size=nn[KMG] [X86]
3427 used for mtrr cleanup. It is largest continuous chunk
3428 that could hold holes aka. UC entries.
3430 mtrr_gran_size=nn[KMG] [X86]
3431 Used for mtrr cleanup. It is granularity of mtrr block.
3433 Large value could prevent small alignment from
3436 mtrr_spare_reg_nr=n [X86]
3438 Range: 0,7 : spare reg number
3440 Used for mtrr cleanup. It is spare mtrr entries number.
3441 Set to 2 or more if your graphical card needs more.
3443 multitce=off [PPC] This parameter disables the use of the pSeries
3444 firmware feature for updating multiple TCE entries
3447 n2= [NET] SDL Inc. RISCom/N2 synchronous serial card
3449 netdev= [NET] Network devices parameters
3450 Format: <irq>,<io>,<mem_start>,<mem_end>,<name>
3451 Note that mem_start is often overloaded to mean
3452 something different and driver-specific.
3453 This usage is only documented in each driver source
3456 netpoll.carrier_timeout=
3457 [NET] Specifies amount of time (in seconds) that
3458 netpoll should wait for a carrier. By default netpoll
3462 [NETFILTER] Enable connection tracking flow accounting
3463 0 to disable accounting
3464 1 to enable accounting
3468 [NFS] sets the pathname to the program which is used
3469 to update the NFS client cache entries.
3471 nfs.cache_getent_timeout=
3472 [NFS] sets the timeout after which an attempt to
3473 update a cache entry is deemed to have failed.
3475 nfs.callback_nr_threads=
3476 [NFSv4] set the total number of threads that the
3477 NFS client will assign to service NFSv4 callback
3480 nfs.callback_tcpport=
3481 [NFS] set the TCP port on which the NFSv4 callback
3482 channel should listen.
3485 [NFS] enable 64-bit inode numbers.
3486 If zero, the NFS client will fake up a 32-bit inode
3487 number for the readdir() and stat() syscalls instead
3488 of returning the full 64-bit number.
3489 The default is to return 64-bit inode numbers.
3491 nfs.idmap_cache_timeout=
3492 [NFS] set the maximum lifetime for idmapper cache
3495 nfs.max_session_cb_slots=
3496 [NFSv4.1] Sets the maximum number of session
3497 slots the client will assign to the callback
3498 channel. This determines the maximum number of
3499 callbacks the client will process in parallel for
3500 a particular server.
3502 nfs.max_session_slots=
3503 [NFSv4.1] Sets the maximum number of session slots
3504 the client will attempt to negotiate with the server.
3505 This limits the number of simultaneous RPC requests
3506 that the client can send to the NFSv4.1 server.
3507 Note that there is little point in setting this
3508 value higher than the max_tcp_slot_table_limit.
3510 nfs.nfs4_disable_idmapping=
3511 [NFSv4] When set to the default of '1', this option
3512 ensures that both the RPC level authentication
3513 scheme and the NFS level operations agree to use
3514 numeric uids/gids if the mount is using the
3515 'sec=sys' security flavour. In effect it is
3516 disabling idmapping, which can make migration from
3517 legacy NFSv2/v3 systems to NFSv4 easier.
3518 Servers that do not support this mode of operation
3519 will be autodetected by the client, and it will fall
3520 back to using the idmapper.
3521 To turn off this behaviour, set the value to '0'.
3524 [NFS4] Specify an additional fixed unique ident-
3525 ification string that NFSv4 clients can insert into
3526 their nfs_client_id4 string. This is typically a
3527 UUID that is generated at system install time.
3529 nfs.recover_lost_locks=
3530 [NFSv4] Attempt to recover locks that were lost due
3531 to a lease timeout on the server. Please note that
3532 doing this risks data corruption, since there are
3533 no guarantees that the file will remain unchanged
3534 after the locks are lost.
3535 If you want to enable the kernel legacy behaviour of
3536 attempting to recover these locks, then set this
3538 The default parameter value of '0' causes the kernel
3539 not to attempt recovery of lost locks.
3541 nfs.send_implementation_id=
3542 [NFSv4.1] Send client implementation identification
3543 information in exchange_id requests.
3544 If zero, no implementation identification information
3546 The default is to send the implementation identification
3549 nfs4.layoutstats_timer=
3550 [NFSv4.2] Change the rate at which the kernel sends
3551 layoutstats to the pNFS metadata server.
3553 Setting this to value to 0 causes the kernel to use
3554 whatever value is the default set by the layout
3555 driver. A non-zero value sets the minimum interval
3556 in seconds between layoutstats transmissions.
3558 nfsd.inter_copy_offload_enable=
3559 [NFSv4.2] When set to 1, the server will support
3560 server-to-server copies for which this server is
3561 the destination of the copy.
3563 nfsd.nfs4_disable_idmapping=
3564 [NFSv4] When set to the default of '1', the NFSv4
3565 server will return only numeric uids and gids to
3566 clients using auth_sys, and will accept numeric uids
3567 and gids from such clients. This is intended to ease
3568 migration from NFSv2/v3.
3570 nfsd.nfsd4_ssc_umount_timeout=
3571 [NFSv4.2] When used as the destination of a
3572 server-to-server copy, knfsd temporarily mounts
3573 the source server. It caches the mount in case
3574 it will be needed again, and discards it if not
3575 used for the number of milliseconds specified by
3578 nfsaddrs= [NFS] Deprecated. Use ip= instead.
3579 See Documentation/admin-guide/nfs/nfsroot.rst.
3581 nfsroot= [NFS] nfs root filesystem for disk-less boxes.
3582 See Documentation/admin-guide/nfs/nfsroot.rst.
3584 nfsrootdebug [NFS] enable nfsroot debugging messages.
3585 See Documentation/admin-guide/nfs/nfsroot.rst.
3587 nmi_backtrace.backtrace_idle [KNL]
3588 Dump stacks even of idle CPUs in response to an
3589 NMI stack-backtrace request.
3591 nmi_debug= [KNL,SH] Specify one or more actions to take
3592 when a NMI is triggered.
3593 Format: [state][,regs][,debounce][,die]
3595 nmi_watchdog= [KNL,BUGS=X86] Debugging features for SMP kernels
3596 Format: [panic,][nopanic,][num]
3598 0 - turn hardlockup detector in nmi_watchdog off
3599 1 - turn hardlockup detector in nmi_watchdog on
3600 When panic is specified, panic when an NMI watchdog
3601 timeout occurs (or 'nopanic' to not panic on an NMI
3602 watchdog, if CONFIG_BOOTPARAM_HARDLOCKUP_PANIC is set)
3603 To disable both hard and soft lockup detectors,
3604 please see 'nowatchdog'.
3605 This is useful when you use a panic=... timeout and
3606 need the box quickly up again.
3608 These settings can be accessed at runtime via
3609 the nmi_watchdog and hardlockup_panic sysctls.
3611 no387 [BUGS=X86-32] Tells the kernel to use the 387 maths
3612 emulation library even if a 387 maths coprocessor
3615 no4lvl [RISCV] Disable 4-level and 5-level paging modes. Forces
3616 kernel to use 3-level paging instead.
3618 no5lvl [X86-64,RISCV] Disable 5-level paging mode. Forces
3619 kernel to use 4-level paging instead.
3621 noaliencache [MM, NUMA, SLAB] Disables the allocation of alien
3622 caches in the slab allocator. Saves per-node memory,
3623 but will impact performance.
3627 noaltinstr [S390] Disables alternative instructions patching
3628 (CPU alternatives feature).
3630 noapic [SMP,APIC] Tells the kernel to not make use of any
3631 IOAPICs that may be present in the system.
3633 noautogroup Disable scheduler automatic task group creation.
3638 [HW] Never suspend the console
3639 Disable suspending of consoles during suspend and
3640 hibernate operations. Once disabled, debugging
3641 messages can reach various consoles while the rest
3642 of the system is being put to sleep (ie, while
3643 debugging driver suspend/resume hooks). This may
3644 not work reliably with all consoles, but is known
3645 to work with serial and VGA consoles.
3646 To facilitate more flexible debugging, we also add
3647 console_suspend, a printk module parameter to control
3648 it. Users could use console_suspend (usually
3649 /sys/module/printk/parameters/console_suspend) to
3650 turn on/off it dynamically.
3653 [KNL] Disable object debugging
3655 nodsp [SH] Disable hardware DSP at boot time.
3657 noefi Disable EFI runtime services support.
3659 no_entry_flush [PPC] Don't flush the L1-D cache when entering the kernel.
3664 This affects only 32-bit executables.
3665 noexec32=on: enable non-executable mappings (default)
3666 read doesn't imply executable mappings
3667 noexec32=off: disable non-executable mappings
3668 read implies executable mappings
3670 no_file_caps Tells the kernel not to honor file capabilities. The
3671 only way then for a file to be executed with privilege
3672 is to be setuid root or executed by root.
3674 nofpu [MIPS,SH] Disable hardware FPU at boot time.
3676 nofsgsbase [X86] Disables FSGSBASE instructions.
3678 nofxsr [BUGS=X86-32] Disables x86 floating point extended
3679 register save and restore. The kernel will only save
3680 legacy floating-point registers on task switch.
3682 nohalt [IA-64] Tells the kernel not to use the power saving
3683 function PAL_HALT_LIGHT when idle. This increases
3684 power-consumption. On the positive side, it reduces
3685 interrupt wake-up latency, which may improve performance
3686 in certain environments such as networked servers or
3690 Force pointers printed to the console or buffers to be
3691 unhashed. By default, when a pointer is printed via %p
3692 format string, that pointer is "hashed", i.e. obscured
3693 by hashing the pointer value. This is a security feature
3694 that hides actual kernel addresses from unprivileged
3695 users, but it also makes debugging the kernel more
3696 difficult since unequal pointers can no longer be
3697 compared. However, if this command-line option is
3698 specified, then all normal pointers will have their true
3699 value printed. This option should only be specified when
3700 debugging the kernel. Please do not use on production
3703 nohibernate [HIBERNATION] Disable hibernation and resume.
3705 nohlt [ARM,ARM64,MICROBLAZE,SH] Forces the kernel to busy wait
3706 in do_idle() and not use the arch_cpu_idle()
3707 implementation; requires CONFIG_GENERIC_IDLE_POLL_SETUP
3708 to be effective. This is useful on platforms where the
3709 sleep(SH) or wfi(ARM,ARM64) instructions do not work
3710 correctly or when doing power measurements to evaluate
3711 the impact of the sleep instructions. This is also
3712 useful when using JTAG debugger.
3714 nohugeiomap [KNL,X86,PPC,ARM64] Disable kernel huge I/O mappings.
3716 nohugevmalloc [KNL,X86,PPC,ARM64] Disable kernel huge vmalloc mappings.
3718 nohz= [KNL] Boottime enable/disable dynamic ticks
3719 Valid arguments: on, off
3722 nohz_full= [KNL,BOOT,SMP,ISOL]
3723 The argument is a cpu list, as described above.
3724 In kernels built with CONFIG_NO_HZ_FULL=y, set
3725 the specified list of CPUs whose tick will be stopped
3726 whenever possible. The boot CPU will be forced outside
3727 the range to maintain the timekeeping. Any CPUs
3728 in this list will have their RCU callbacks offloaded,
3729 just as if they had also been called out in the
3730 rcu_nocbs= boot parameter.
3732 Note that this argument takes precedence over
3733 the CONFIG_RCU_NOCB_CPU_DEFAULT_ALL option.
3735 noinitrd [RAM] Tells the kernel not to load any configured
3738 nointremap [X86-64, Intel-IOMMU] Do not enable interrupt
3740 [Deprecated - use intremap=off]
3744 noinvpcid [X86] Disable the INVPCID cpu feature.
3746 noiotrap [SH] Disables trapped I/O port accesses.
3748 noirqdebug [X86-32] Disables the code which attempts to detect and
3749 disable unhandled interrupt sources.
3751 noisapnp [ISAPNP] Disables ISA PnP code.
3753 nojitter [IA-64] Disables jitter checking for ITC timers.
3756 When CONFIG_RANDOMIZE_BASE is set, this disables
3757 kernel and module base offset ASLR (Address Space
3758 Layout Randomization).
3760 no-kvmapf [X86,KVM] Disable paravirtualized asynchronous page
3763 no-kvmclock [X86,KVM] Disable paravirtualized KVM clock driver
3765 nolapic [X86-32,APIC] Do not enable or use the local APIC.
3767 nolapic_timer [X86-32,APIC] Do not use the local APIC timer.
3769 nomca [IA-64] Disable machine check abort handling
3771 nomce [X86-32] Disable Machine Check Exception
3773 nomfgpt [X86-32] Disable Multi-Function General Purpose
3774 Timer usage (for AMD Geode machines).
3776 nomodeset Disable kernel modesetting. Most systems' firmware
3777 sets up a display mode and provides framebuffer memory
3778 for output. With nomodeset, DRM and fbdev drivers will
3779 not load if they could possibly displace the pre-
3780 initialized output. Only the system framebuffer will
3781 be available for use. The respective drivers will not
3782 perform display-mode changes or accelerated rendering.
3784 Useful as error fallback, or for testing and debugging.
3786 nomodule Disable module load
3788 nonmi_ipi [X86] Disable using NMI IPIs during panic/reboot to
3789 shutdown the other cpus. Instead use the REBOOT_VECTOR
3792 nopat [X86] Disable PAT (page attribute table extension of
3793 pagetables) support.
3795 nopcid [X86-64] Disable the PCID cpu feature.
3797 nopku [X86] Disable Memory Protection Keys CPU feature found
3801 Equivalent to pti=off
3803 nopv= [X86,XEN,KVM,HYPER_V,VMWARE]
3804 Disables the PV optimizations forcing the guest to run
3805 as generic guest with no PV drivers. Currently support
3806 XEN HVM, KVM, HYPER_V and VMWARE guest.
3808 nopvspin [X86,XEN,KVM]
3809 Disables the qspinlock slow path using PV optimizations
3810 which allow the hypervisor to 'idle' the guest on lock
3813 norandmaps Don't use address space randomization. Equivalent to
3814 echo 0 > /proc/sys/kernel/randomize_va_space
3816 noreplace-smp [X86-32,SMP] Don't replace SMP instructions
3817 with UP alternatives
3819 noresume [SWSUSP] Disables resume and restores original swap
3824 no-scroll [VGA] Disables scrollback.
3825 This is required for the Braillex ib80-piezo Braille
3826 reader made by F.H. Papenmeier (Germany).
3828 nosgx [X86-64,SGX] Disables Intel SGX kernel support.
3831 Disable SMAP (Supervisor Mode Access Prevention)
3832 even if it is supported by processor.
3835 Disable SMEP (Supervisor Mode Execution Prevention)
3836 even if it is supported by processor.
3838 nosmp [SMP] Tells an SMP kernel to act as a UP kernel,
3839 and disable the IO APIC. legacy for "maxcpus=0".
3841 nosmt [KNL,S390] Disable symmetric multithreading (SMT).
3842 Equivalent to smt=1.
3844 [KNL,X86] Disable symmetric multithreading (SMT).
3845 nosmt=force: Force disable SMT, cannot be undone
3846 via the sysfs control file.
3848 nosoftlockup [KNL] Disable the soft-lockup detector.
3850 nospec_store_bypass_disable
3851 [HW] Disable all mitigations for the Speculative Store Bypass vulnerability
3853 nospectre_bhb [ARM64] Disable all mitigations for Spectre-BHB (branch
3854 history injection) vulnerability. System may allow data leaks
3857 nospectre_v1 [X86,PPC] Disable mitigations for Spectre Variant 1
3858 (bounds check bypass). With this option data leaks are
3859 possible in the system.
3861 nospectre_v2 [X86,PPC_E500,ARM64] Disable all mitigations for
3862 the Spectre variant 2 (indirect branch prediction)
3863 vulnerability. System may allow data leaks with this
3866 no-steal-acc [X86,PV_OPS,ARM64,PPC/PSERIES] Disable paravirtualized
3867 steal time accounting. steal time is computed, but
3868 won't influence scheduler behaviour
3870 nosync [HW,M68K] Disables sync negotiation for all devices.
3872 no_timer_check [X86,APIC] Disables the code which tests for
3873 broken timer IRQ sources.
3876 [PPC] Don't flush the L1-D cache after accessing user data.
3878 novmcoredd [KNL,KDUMP]
3879 Disable device dump. Device dump allows drivers to
3880 append dump data to vmcore so you can collect driver
3881 specified debug info. Drivers can append the data
3882 without any limit and this data is stored in memory,
3883 so this may cause significant memory stress. Disabling
3884 device dump can help save memory but the driver debug
3885 data will be no longer available. This parameter
3886 is only available when CONFIG_PROC_VMCORE_DEVICE_DUMP
3890 [X86,PV_OPS] Disable paravirtualized VMware scheduler
3891 clock and use the default one.
3893 nowatchdog [KNL] Disable both lockup detectors, i.e.
3894 soft-lockup and NMI watchdog (hard-lockup).
3898 nox2apic [X86-64,APIC] Do not enable x2APIC mode.
3900 NOTE: this parameter will be ignored on systems with the
3901 LEGACY_XAPIC_DISABLED bit set in the
3902 IA32_XAPIC_DISABLE_STATUS MSR.
3904 noxsave [BUGS=X86] Disables x86 extended register state save
3905 and restore using xsave. The kernel will fallback to
3906 enabling legacy floating-point and sse state.
3908 noxsaveopt [X86] Disables xsaveopt used in saving x86 extended
3909 register states. The kernel will fall back to use
3910 xsave to save the states. By using this parameter,
3911 performance of saving the states is degraded because
3912 xsave doesn't support modified optimization while
3913 xsaveopt supports it on xsaveopt enabled systems.
3915 noxsaves [X86] Disables xsaves and xrstors used in saving and
3916 restoring x86 extended register state in compacted
3917 form of xsave area. The kernel will fall back to use
3918 xsaveopt and xrstor to save and restore the states
3919 in standard form of xsave area. By using this
3920 parameter, xsave area per process might occupy more
3921 memory on xsaves enabled systems.
3923 nps_mtm_hs_ctr= [KNL,ARC]
3924 This parameter sets the maximum duration, in
3925 cycles, each HW thread of the CTOP can run
3926 without interruptions, before HW switches it.
3927 The actual maximum duration is 16 times this
3929 Format: integer between 1 and 255
3932 nptcg= [IA-64] Override max number of concurrent global TLB
3933 purges which is reported from either PAL_VM_SUMMARY or
3936 nr_cpus= [SMP] Maximum number of processors that an SMP kernel
3937 could support. nr_cpus=n : n >= 1 limits the kernel to
3938 support 'n' processors. It could be larger than the
3939 number of already plugged CPU during bootup, later in
3940 runtime you can physically add extra cpu until it reaches
3941 n. So during boot up some boot time memory for per-cpu
3942 variables need be pre-allocated for later physical cpu
3945 nr_uarts= [SERIAL] maximum number of UARTs to be registered.
3947 numa=off [KNL, ARM64, PPC, RISCV, SPARC, X86] Disable NUMA, Only
3948 set up a single NUMA node spanning all memory.
3950 numa_balancing= [KNL,ARM64,PPC,RISCV,S390,X86] Enable or disable automatic
3952 Allowed values are enable and disable
3954 numa_zonelist_order= [KNL, BOOT] Select zonelist order for NUMA.
3955 'node', 'default' can be specified
3956 This can be set from sysctl after boot.
3957 See Documentation/admin-guide/sysctl/vm.rst for details.
3959 ohci1394_dma=early [HW] enable debugging via the ohci1394 driver.
3960 See Documentation/core-api/debugging-via-ohci1394.rst for more
3963 olpc_ec_timeout= [OLPC] ms delay when issuing EC commands
3964 Rather than timing out after 20 ms if an EC
3965 command is not properly ACKed, override the length
3966 of the timeout. We have interrupts disabled while
3967 waiting for the ACK, so if this is set too high
3968 interrupts *may* be lost!
3970 omap_mux= [OMAP] Override bootloader pin multiplexing.
3971 Format: <mux_mode0.mode_name=value>...
3972 For example, to override I2C bus2:
3973 omap_mux=i2c2_scl.i2c2_scl=0x100,i2c2_sda.i2c2_sda=0x100
3975 onenand.bdry= [HW,MTD] Flex-OneNAND Boundary Configuration
3977 Format: [die0_boundary][,die0_lock][,die1_boundary][,die1_lock]
3979 boundary - index of last SLC block on Flex-OneNAND.
3980 The remaining blocks are configured as MLC blocks.
3981 lock - Configure if Flex-OneNAND boundary should be locked.
3982 Once locked, the boundary cannot be changed.
3983 1 indicates lock status, 0 indicates unlock status.
3985 oops=panic Always panic on oopses. Default is to just kill the
3986 process, but there is a small probability of
3987 deadlocking the machine.
3988 This will also cause panics on machine check exceptions.
3989 Useful together with panic=30 to trigger a reboot.
3992 [KNL] Boolean flag to control whether the page allocator
3993 should randomize its free lists. The randomization may
3994 be automatically enabled if the kernel detects it is
3995 running on a platform with a direct-mapped memory-side
3996 cache, and this parameter can be used to
3997 override/disable that behavior. The state of the flag
3998 can be read from sysfs at:
3999 /sys/module/page_alloc/parameters/shuffle.
4001 page_owner= [KNL] Boot-time page_owner enabling option.
4002 Storage of the information about who allocated
4003 each page is disabled in default. With this switch,
4005 on: enable the feature
4007 page_poison= [KNL] Boot-time parameter changing the state of
4008 poisoning on the buddy allocator, available with
4009 CONFIG_PAGE_POISONING=y.
4010 off: turn off poisoning (default)
4011 on: turn on poisoning
4013 page_reporting.page_reporting_order=
4014 [KNL] Minimal page reporting order
4016 Adjust the minimal page reporting order. The page
4017 reporting is disabled when it exceeds MAX_ORDER.
4019 panic= [KNL] Kernel behaviour on panic: delay <timeout>
4020 timeout > 0: seconds before rebooting
4021 timeout = 0: wait forever
4022 timeout < 0: reboot immediately
4025 panic_print= Bitmask for printing system info when panic happens.
4026 User can chose combination of the following bits:
4027 bit 0: print all tasks info
4028 bit 1: print system memory info
4029 bit 2: print timer info
4030 bit 3: print locks info if CONFIG_LOCKDEP is on
4031 bit 4: print ftrace buffer
4032 bit 5: print all printk messages in buffer
4033 bit 6: print all CPUs backtrace (if available in the arch)
4034 *Be aware* that this option may print a _lot_ of lines,
4035 so there are risks of losing older messages in the log.
4036 Use this option carefully, maybe worth to setup a
4037 bigger log buffer with "log_buf_len" along with this.
4039 panic_on_taint= Bitmask for conditionally calling panic() in add_taint()
4040 Format: <hex>[,nousertaint]
4041 Hexadecimal bitmask representing the set of TAINT flags
4042 that will cause the kernel to panic when add_taint() is
4043 called with any of the flags in this set.
4044 The optional switch "nousertaint" can be utilized to
4045 prevent userspace forced crashes by writing to sysctl
4046 /proc/sys/kernel/tainted any flagset matching with the
4047 bitmask set on panic_on_taint.
4048 See Documentation/admin-guide/tainted-kernels.rst for
4049 extra details on the taint flags that users can pick
4050 to compose the bitmask to assign to panic_on_taint.
4052 panic_on_warn panic() instead of WARN(). Useful to cause kdump
4055 parkbd.port= [HW] Parallel port number the keyboard adapter is
4056 connected to, default is 0.
4058 parkbd.mode= [HW] Parallel port keyboard adapter mode of operation,
4059 0 for XT, 1 for AT (default is AT).
4062 parport= [HW,PPT] Specify parallel ports. 0 disables.
4063 Format: { 0 | auto | 0xBBB[,IRQ[,DMA]] }
4064 Use 'auto' to force the driver to use any
4065 IRQ/DMA settings detected (the default is to
4066 ignore detected IRQ/DMA settings because of
4067 possible conflicts). You can specify the base
4068 address, IRQ, and DMA settings; IRQ and DMA
4069 should be numbers, or 'auto' (for using detected
4070 settings on that particular port), or 'nofifo'
4071 (to avoid using a FIFO even if it is detected).
4072 Parallel ports are assigned in the order they
4073 are specified on the command line, starting
4076 parport_init_mode= [HW,PPT]
4077 Configure VIA parallel port to operate in
4078 a specific mode. This is necessary on Pegasos
4079 computer where firmware has no options for setting
4080 up parallel port mode and sets it to spp.
4081 Currently this function knows 686a and 8231 chips.
4082 Format: [spp|ps2|epp|ecp|ecpepp]
4084 pata_legacy.all= [HW,LIBATA]
4086 Set to non-zero to probe primary and secondary ISA
4087 port ranges on PCI systems where no PCI PATA device
4088 has been found at either range. Disabled by default.
4090 pata_legacy.autospeed= [HW,LIBATA]
4092 Set to non-zero if a chip is present that snoops speed
4093 changes. Disabled by default.
4095 pata_legacy.ht6560a= [HW,LIBATA]
4097 Set to 1, 2, or 3 for HT 6560A on the primary channel,
4098 the secondary channel, or both channels respectively.
4099 Disabled by default.
4101 pata_legacy.ht6560b= [HW,LIBATA]
4103 Set to 1, 2, or 3 for HT 6560B on the primary channel,
4104 the secondary channel, or both channels respectively.
4105 Disabled by default.
4107 pata_legacy.iordy_mask= [HW,LIBATA]
4109 IORDY enable mask. Set individual bits to allow IORDY
4110 for the respective channel. Bit 0 is for the first
4111 legacy channel handled by this driver, bit 1 is for
4112 the second channel, and so on. The sequence will often
4113 correspond to the primary legacy channel, the secondary
4114 legacy channel, and so on, but the handling of a PCI
4115 bus and the use of other driver options may interfere
4116 with the sequence. By default IORDY is allowed across
4119 pata_legacy.opti82c46x= [HW,LIBATA]
4121 Set to 1, 2, or 3 for Opti 82c611A on the primary
4122 channel, the secondary channel, or both channels
4123 respectively. Disabled by default.
4125 pata_legacy.opti82c611a= [HW,LIBATA]
4127 Set to 1, 2, or 3 for Opti 82c465MV on the primary
4128 channel, the secondary channel, or both channels
4129 respectively. Disabled by default.
4131 pata_legacy.pio_mask= [HW,LIBATA]
4133 PIO mode mask for autospeed devices. Set individual
4134 bits to allow the use of the respective PIO modes.
4135 Bit 0 is for mode 0, bit 1 is for mode 1, and so on.
4136 All modes allowed by default.
4138 pata_legacy.probe_all= [HW,LIBATA]
4140 Set to non-zero to probe tertiary and further ISA
4141 port ranges on PCI systems. Disabled by default.
4143 pata_legacy.probe_mask= [HW,LIBATA]
4145 Probe mask for legacy ISA PATA ports. Depending on
4146 platform configuration and the use of other driver
4147 options up to 6 legacy ports are supported: 0x1f0,
4148 0x170, 0x1e8, 0x168, 0x1e0, 0x160, however probing
4149 of individual ports can be disabled by setting the
4150 corresponding bits in the mask to 1. Bit 0 is for
4151 the first port in the list above (0x1f0), and so on.
4152 By default all supported ports are probed.
4154 pata_legacy.qdi= [HW,LIBATA]
4156 Set to non-zero to probe QDI controllers. By default
4157 set to 1 if CONFIG_PATA_QDI_MODULE, 0 otherwise.
4159 pata_legacy.winbond= [HW,LIBATA]
4161 Set to non-zero to probe Winbond controllers. Use
4162 the standard I/O port (0x130) if 1, otherwise the
4163 value given is the I/O port to use (typically 0x1b0).
4164 By default set to 1 if CONFIG_PATA_WINBOND_VLB_MODULE,
4167 pata_platform.pio_mask= [HW,LIBATA]
4169 Supported PIO mode mask. Set individual bits to allow
4170 the use of the respective PIO modes. Bit 0 is for
4171 mode 0, bit 1 is for mode 1, and so on. Mode 0 only
4175 Halt all CPUs after the first oops has been printed for
4176 the specified number of seconds. This is to be used if
4177 your oopses keep scrolling off the screen.
4181 pci=option[,option...] [PCI] various PCI subsystem options.
4183 Some options herein operate on a specific device
4184 or a set of devices (<pci_dev>). These are
4185 specified in one of the following formats:
4187 [<domain>:]<bus>:<dev>.<func>[/<dev>.<func>]*
4188 pci:<vendor>:<device>[:<subvendor>:<subdevice>]
4190 Note: the first format specifies a PCI
4191 bus/device/function address which may change
4192 if new hardware is inserted, if motherboard
4193 firmware changes, or due to changes caused
4194 by other kernel parameters. If the
4195 domain is left unspecified, it is
4196 taken to be zero. Optionally, a path
4197 to a device through multiple device/function
4198 addresses can be specified after the base
4199 address (this is more robust against
4200 renumbering issues). The second format
4201 selects devices using IDs from the
4202 configuration space which may match multiple
4203 devices in the system.
4205 earlydump dump PCI config space before the kernel
4207 off [X86] don't probe for the PCI bus
4208 bios [X86-32] force use of PCI BIOS, don't access
4209 the hardware directly. Use this if your machine
4210 has a non-standard PCI host bridge.
4211 nobios [X86-32] disallow use of PCI BIOS, only direct
4212 hardware access methods are allowed. Use this
4213 if you experience crashes upon bootup and you
4214 suspect they are caused by the BIOS.
4215 conf1 [X86] Force use of PCI Configuration Access
4216 Mechanism 1 (config address in IO port 0xCF8,
4217 data in IO port 0xCFC, both 32-bit).
4218 conf2 [X86] Force use of PCI Configuration Access
4219 Mechanism 2 (IO port 0xCF8 is an 8-bit port for
4220 the function, IO port 0xCFA, also 8-bit, sets
4221 bus number. The config space is then accessed
4222 through ports 0xC000-0xCFFF).
4223 See http://wiki.osdev.org/PCI for more info
4224 on the configuration access mechanisms.
4225 noaer [PCIE] If the PCIEAER kernel config parameter is
4226 enabled, this kernel boot option can be used to
4227 disable the use of PCIE advanced error reporting.
4228 nodomains [PCI] Disable support for multiple PCI
4229 root domains (aka PCI segments, in ACPI-speak).
4230 nommconf [X86] Disable use of MMCONFIG for PCI
4232 check_enable_amd_mmconf [X86] check for and enable
4233 properly configured MMIO access to PCI
4234 config space on AMD family 10h CPU
4235 nomsi [MSI] If the PCI_MSI kernel config parameter is
4236 enabled, this kernel boot option can be used to
4237 disable the use of MSI interrupts system-wide.
4238 noioapicquirk [APIC] Disable all boot interrupt quirks.
4239 Safety option to keep boot IRQs enabled. This
4240 should never be necessary.
4241 ioapicreroute [APIC] Enable rerouting of boot IRQs to the
4242 primary IO-APIC for bridges that cannot disable
4243 boot IRQs. This fixes a source of spurious IRQs
4244 when the system masks IRQs.
4245 noioapicreroute [APIC] Disable workaround that uses the
4246 boot IRQ equivalent of an IRQ that connects to
4247 a chipset where boot IRQs cannot be disabled.
4248 The opposite of ioapicreroute.
4249 biosirq [X86-32] Use PCI BIOS calls to get the interrupt
4250 routing table. These calls are known to be buggy
4251 on several machines and they hang the machine
4252 when used, but on other computers it's the only
4253 way to get the interrupt routing table. Try
4254 this option if the kernel is unable to allocate
4255 IRQs or discover secondary PCI buses on your
4257 rom [X86] Assign address space to expansion ROMs.
4258 Use with caution as certain devices share
4259 address decoders between ROMs and other
4261 norom [X86] Do not assign address space to
4262 expansion ROMs that do not already have
4263 BIOS assigned address ranges.
4264 nobar [X86] Do not assign address space to the
4265 BARs that weren't assigned by the BIOS.
4266 irqmask=0xMMMM [X86] Set a bit mask of IRQs allowed to be
4267 assigned automatically to PCI devices. You can
4268 make the kernel exclude IRQs of your ISA cards
4270 pirqaddr=0xAAAAA [X86] Specify the physical address
4271 of the PIRQ table (normally generated
4272 by the BIOS) if it is outside the
4273 F0000h-100000h range.
4274 lastbus=N [X86] Scan all buses thru bus #N. Can be
4275 useful if the kernel is unable to find your
4276 secondary buses and you want to tell it
4277 explicitly which ones they are.
4278 assign-busses [X86] Always assign all PCI bus
4279 numbers ourselves, overriding
4280 whatever the firmware may have done.
4281 usepirqmask [X86] Honor the possible IRQ mask stored
4282 in the BIOS $PIR table. This is needed on
4283 some systems with broken BIOSes, notably
4284 some HP Pavilion N5400 and Omnibook XE3
4285 notebooks. This will have no effect if ACPI
4286 IRQ routing is enabled.
4287 noacpi [X86] Do not use ACPI for IRQ routing
4288 or for PCI scanning.
4289 use_crs [X86] Use PCI host bridge window information
4290 from ACPI. On BIOSes from 2008 or later, this
4291 is enabled by default. If you need to use this,
4292 please report a bug.
4293 nocrs [X86] Ignore PCI host bridge windows from ACPI.
4294 If you need to use this, please report a bug.
4295 use_e820 [X86] Use E820 reservations to exclude parts of
4296 PCI host bridge windows. This is a workaround
4297 for BIOS defects in host bridge _CRS methods.
4298 If you need to use this, please report a bug to
4299 <linux-pci@vger.kernel.org>.
4300 no_e820 [X86] Ignore E820 reservations for PCI host
4301 bridge windows. This is the default on modern
4302 hardware. If you need to use this, please report
4303 a bug to <linux-pci@vger.kernel.org>.
4304 routeirq Do IRQ routing for all PCI devices.
4305 This is normally done in pci_enable_device(),
4306 so this option is a temporary workaround
4307 for broken drivers that don't call it.
4308 skip_isa_align [X86] do not align io start addr, so can
4309 handle more pci cards
4310 noearly [X86] Don't do any early type 1 scanning.
4311 This might help on some broken boards which
4312 machine check when some devices' config space
4313 is read. But various workarounds are disabled
4314 and some IOMMU drivers will not work.
4315 bfsort Sort PCI devices into breadth-first order.
4316 This sorting is done to get a device
4317 order compatible with older (<= 2.4) kernels.
4318 nobfsort Don't sort PCI devices into breadth-first order.
4319 pcie_bus_tune_off Disable PCIe MPS (Max Payload Size)
4320 tuning and use the BIOS-configured MPS defaults.
4321 pcie_bus_safe Set every device's MPS to the largest value
4322 supported by all devices below the root complex.
4323 pcie_bus_perf Set device MPS to the largest allowable MPS
4324 based on its parent bus. Also set MRRS (Max
4325 Read Request Size) to the largest supported
4326 value (no larger than the MPS that the device
4327 or bus can support) for best performance.
4328 pcie_bus_peer2peer Set every device's MPS to 128B, which
4329 every device is guaranteed to support. This
4330 configuration allows peer-to-peer DMA between
4331 any pair of devices, possibly at the cost of
4332 reduced performance. This also guarantees
4333 that hot-added devices will work.
4334 cbiosize=nn[KMG] The fixed amount of bus space which is
4335 reserved for the CardBus bridge's IO window.
4336 The default value is 256 bytes.
4337 cbmemsize=nn[KMG] The fixed amount of bus space which is
4338 reserved for the CardBus bridge's memory
4339 window. The default value is 64 megabytes.
4342 [<order of align>@]<pci_dev>[; ...]
4343 Specifies alignment and device to reassign
4344 aligned memory resources. How to
4345 specify the device is described above.
4346 If <order of align> is not specified,
4347 PAGE_SIZE is used as alignment.
4348 A PCI-PCI bridge can be specified if resource
4349 windows need to be expanded.
4350 To specify the alignment for several
4351 instances of a device, the PCI vendor,
4352 device, subvendor, and subdevice may be
4353 specified, e.g., 12@pci:8086:9c22:103c:198f
4354 for 4096-byte alignment.
4355 ecrc= Enable/disable PCIe ECRC (transaction layer
4356 end-to-end CRC checking). Only effective if
4357 OS has native AER control (either granted by
4358 ACPI _OSC or forced via "pcie_ports=native")
4359 bios: Use BIOS/firmware settings. This is the
4363 hpiosize=nn[KMG] The fixed amount of bus space which is
4364 reserved for hotplug bridge's IO window.
4365 Default size is 256 bytes.
4366 hpmmiosize=nn[KMG] The fixed amount of bus space which is
4367 reserved for hotplug bridge's MMIO window.
4368 Default size is 2 megabytes.
4369 hpmmioprefsize=nn[KMG] The fixed amount of bus space which is
4370 reserved for hotplug bridge's MMIO_PREF window.
4371 Default size is 2 megabytes.
4372 hpmemsize=nn[KMG] The fixed amount of bus space which is
4373 reserved for hotplug bridge's MMIO and
4375 Default size is 2 megabytes.
4376 hpbussize=nn The minimum amount of additional bus numbers
4377 reserved for buses below a hotplug bridge.
4379 realloc= Enable/disable reallocating PCI bridge resources
4380 if allocations done by BIOS are too small to
4381 accommodate resources required by all child
4383 off: Turn realloc off
4385 realloc same as realloc=on
4386 noari do not use PCIe ARI.
4387 noats [PCIE, Intel-IOMMU, AMD-IOMMU]
4388 do not use PCIe ATS (and IOMMU device IOTLB).
4389 pcie_scan_all Scan all possible PCIe devices. Otherwise we
4390 only look for one device below a PCIe downstream
4392 big_root_window Try to add a big 64bit memory window to the PCIe
4393 root complex on AMD CPUs. Some GFX hardware
4394 can resize a BAR to allow access to all VRAM.
4395 Adding the window is slightly risky (it may
4396 conflict with unreported devices), so this
4398 disable_acs_redir=<pci_dev>[; ...]
4399 Specify one or more PCI devices (in the format
4400 specified above) separated by semicolons.
4401 Each device specified will have the PCI ACS
4402 redirect capabilities forced off which will
4403 allow P2P traffic between devices through
4404 bridges without forcing it upstream. Note:
4405 this removes isolation between devices and
4406 may put more devices in an IOMMU group.
4407 force_floating [S390] Force usage of floating interrupts.
4408 nomio [S390] Do not use MIO instructions.
4409 norid [S390] ignore the RID field and force use of
4410 one PCI domain per PCI function
4412 pcie_aspm= [PCIE] Forcibly enable or disable PCIe Active State Power
4415 force Enable ASPM even on devices that claim not to support it.
4416 WARNING: Forcing ASPM on may cause system lockups.
4418 pcie_ports= [PCIE] PCIe port services handling:
4419 native Use native PCIe services (PME, AER, DPC, PCIe hotplug)
4420 even if the platform doesn't give the OS permission to
4421 use them. This may cause conflicts if the platform
4422 also tries to use these services.
4423 dpc-native Use native PCIe service for DPC only. May
4424 cause conflicts if firmware uses AER or DPC.
4425 compat Disable native PCIe services (PME, AER, DPC, PCIe
4428 pcie_port_pm= [PCIE] PCIe port power management handling:
4429 off Disable power management of all PCIe ports
4430 force Forcibly enable power management of all PCIe ports
4432 pcie_pme= [PCIE,PM] Native PCIe PME signaling options:
4433 nomsi Do not use MSI for native PCIe PME signaling (this makes
4434 all PCIe root ports use INTx for all services).
4436 pcmv= [HW,PCMCIA] BadgePAD 4
4440 Keep all power-domains already enabled by bootloader on,
4441 even if no driver has claimed them. This is useful
4442 for debug and development, but should not be
4443 needed on a platform with proper driver support.
4445 pdcchassis= [PARISC,HW] Disable/Enable PDC Chassis Status codes at
4448 See arch/parisc/kernel/pdc_chassis.c
4450 percpu_alloc= Select which percpu first chunk allocator to use.
4451 Currently supported values are "embed" and "page".
4452 Archs may support subset or none of the selections.
4453 See comments in mm/percpu.c for details on each
4454 allocator. This parameter is primarily for debugging
4455 and performance comparison.
4457 pirq= [SMP,APIC] Manual mp-table setup
4458 See Documentation/arch/x86/i386/IO-APIC.rst.
4460 plip= [PPT,NET] Parallel port network link
4461 Format: { parport<nr> | timid | 0 }
4462 See also Documentation/admin-guide/parport.rst.
4464 pmtmr= [X86] Manual setup of pmtmr I/O Port.
4465 Override pmtimer IOPort with a hex value.
4468 pmu_override= [PPC] Override the PMU.
4469 This option takes over the PMU facility, so it is no
4470 longer usable by perf. Setting this option starts the
4471 PMU counters by setting MMCR0 to 0 (the FC bit is
4472 cleared). If a number is given, then MMCR1 is set to
4473 that number, otherwise (e.g., 'pmu_override=on'), MMCR1
4476 pm_debug_messages [SUSPEND,KNL]
4477 Enable suspend/resume debug messages during boot up.
4480 Enable PNP debug messages (depends on the
4481 CONFIG_PNP_DEBUG_MESSAGES option). Change at run-time
4482 via /sys/module/pnp/parameters/debug. We always show
4483 current resource usage; turning this on also shows
4484 possible settings and some assignment information.
4490 { on | off | curr | res | no-curr | no-res }
4493 [ISAPNP] Exclude IRQs for the autoconfiguration
4496 [ISAPNP] Exclude DMAs for the autoconfiguration
4498 pnp_reserve_io= [ISAPNP] Exclude I/O ports for the autoconfiguration
4499 Ranges are in pairs (I/O port base and size).
4502 [ISAPNP] Exclude memory regions for the
4504 Ranges are in pairs (memory base and size).
4506 ports= [IP_VS_FTP] IPVS ftp helper module
4508 Up to 8 (IP_VS_APP_MAX_PORTS) ports
4510 Format: <port>,<port>....
4512 powersave=off [PPC] This option disables power saving features.
4513 It specifically disables cpuidle and sets the
4514 platform machine description specific power_save
4515 function to NULL. On Idle the CPU just reduces
4518 ppc_strict_facility_enable
4519 [PPC] This option catches any kernel floating point,
4520 Altivec, VSX and SPE outside of regions specifically
4521 allowed (eg kernel_enable_fpu()/kernel_disable_fpu()).
4522 There is some performance impact when enabling this.
4526 Disable Hardware Transactional Memory
4529 Select preemption mode if you have CONFIG_PREEMPT_DYNAMIC
4530 none - Limited to cond_resched() calls
4531 voluntary - Limited to cond_resched() and might_sleep() calls
4532 full - Any section that isn't explicitly preempt disabled
4533 can be preempted anytime.
4535 print-fatal-signals=
4536 [KNL] debug: print fatal signals
4538 If enabled, warn about various signal handling
4539 related application anomalies: too many signals,
4540 too many POSIX.1 timers, fatal signals causing a
4543 If you hit the warning due to signal overflow,
4544 you might want to try "ulimit -i unlimited".
4548 printk.always_kmsg_dump=
4549 Trigger kmsg_dump for cases other than kernel oops or
4551 Format: <bool> (1/Y/y=enable, 0/N/n=disable)
4554 printk.console_no_auto_verbose=
4555 Disable console loglevel raise on oops, panic
4556 or lockdep-detected issues (only if lock debug is on).
4557 With an exception to setups with low baudrate on
4558 serial console, keeping this 0 is a good choice
4559 in order to provide more debug information.
4561 default: 0 (auto_verbose is enabled)
4563 printk.devkmsg={on,off,ratelimit}
4564 Control writing to /dev/kmsg.
4565 on - unlimited logging to /dev/kmsg from userspace
4566 off - logging to /dev/kmsg disabled
4567 ratelimit - ratelimit the logging
4570 printk.time= Show timing data prefixed to each printk message line
4571 Format: <bool> (1/Y/y=enable, 0/N/n=disable)
4573 processor.max_cstate= [HW,ACPI]
4574 Limit processor to maximum C-state
4575 max_cstate=9 overrides any DMI blacklist limit.
4577 processor.nocst [HW,ACPI]
4578 Ignore the _CST method to determine C-states,
4579 instead using the legacy FADT method
4581 profile= [KNL] Enable kernel profiling via /proc/profile
4582 Format: [<profiletype>,]<number>
4583 Param: <profiletype>: "schedule", "sleep", or "kvm"
4584 [defaults to kernel profiling]
4585 Param: "schedule" - profile schedule points.
4586 Param: "sleep" - profile D-state sleeping (millisecs).
4587 Requires CONFIG_SCHEDSTATS
4588 Param: "kvm" - profile VM exits.
4589 Param: <number> - step/bucket size as a power of 2 for
4590 statistical time based profiling.
4592 prompt_ramdisk= [RAM] [Deprecated]
4594 prot_virt= [S390] enable hosting protected virtual machines
4595 isolated from the hypervisor (if hardware supports
4599 psi= [KNL] Enable or disable pressure stall information
4603 psmouse.proto= [HW,MOUSE] Highest PS2 mouse protocol extension to
4604 probe for; one of (bare|imps|exps|lifebook|any).
4605 psmouse.rate= [HW,MOUSE] Set desired mouse report rate, in reports
4607 psmouse.resetafter= [HW,MOUSE]
4608 Try to reset the device after so many bad packets
4611 [HW,MOUSE] Set desired mouse resolution, in dpi.
4612 psmouse.smartscroll=
4613 [HW,MOUSE] Controls Logitech smartscroll autorepeat.
4614 0 = disabled, 1 = enabled (default).
4616 pstore.backend= Specify the name of the pstore backend to use
4618 pti= [X86-64] Control Page Table Isolation of user and
4619 kernel address spaces. Disabling this feature
4620 removes hardening, but improves performance of
4621 system calls and interrupts.
4623 on - unconditionally enable
4624 off - unconditionally disable
4625 auto - kernel detects whether your CPU model is
4626 vulnerable to issues that PTI mitigates
4628 Not specifying this option is equivalent to pti=auto.
4631 [KNL] Number of legacy pty's. Overwrites compiled-in
4634 quiet [KNL] Disable most log messages
4638 radix_hcall_invalidate=on [PPC/PSERIES]
4639 Disable RADIX GTSE feature and use hcall for TLB
4643 See Documentation/admin-guide/md.rst.
4645 ramdisk_size= [RAM] Sizes of RAM disks in kilobytes
4646 See Documentation/admin-guide/blockdev/ramdisk.rst.
4648 ramdisk_start= [RAM] RAM disk image start address
4650 random.trust_cpu=off
4651 [KNL] Disable trusting the use of the CPU's
4652 random number generator (if available) to
4653 initialize the kernel's RNG.
4655 random.trust_bootloader=off
4656 [KNL] Disable trusting the use of the a seed
4657 passed by the bootloader (if available) to
4658 initialize the kernel's RNG.
4660 randomize_kstack_offset=
4661 [KNL] Enable or disable kernel stack offset
4662 randomization, which provides roughly 5 bits of
4663 entropy, frustrating memory corruption attacks
4664 that depend on stack address determinism or
4665 cross-syscall address exposures. This is only
4666 available on architectures that have defined
4667 CONFIG_HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET.
4668 Format: <bool> (1/Y/y=enable, 0/N/n=disable)
4669 Default is CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT.
4671 ras=option[,option,...] [KNL] RAS-specific options
4674 Disable the Correctable Errors Collector,
4675 see CONFIG_RAS_CEC help text.
4677 rcu_nocbs[=cpu-list]
4678 [KNL] The optional argument is a cpu list,
4681 In kernels built with CONFIG_RCU_NOCB_CPU=y,
4682 enable the no-callback CPU mode, which prevents
4683 such CPUs' callbacks from being invoked in
4684 softirq context. Invocation of such CPUs' RCU
4685 callbacks will instead be offloaded to "rcuox/N"
4686 kthreads created for that purpose, where "x" is
4687 "p" for RCU-preempt, "s" for RCU-sched, and "g"
4688 for the kthreads that mediate grace periods; and
4689 "N" is the CPU number. This reduces OS jitter on
4690 the offloaded CPUs, which can be useful for HPC
4691 and real-time workloads. It can also improve
4692 energy efficiency for asymmetric multiprocessors.
4694 If a cpulist is passed as an argument, the specified
4695 list of CPUs is set to no-callback mode from boot.
4697 Otherwise, if the '=' sign and the cpulist
4698 arguments are omitted, no CPU will be set to
4699 no-callback mode from boot but the mode may be
4700 toggled at runtime via cpusets.
4702 Note that this argument takes precedence over
4703 the CONFIG_RCU_NOCB_CPU_DEFAULT_ALL option.
4706 Rather than requiring that offloaded CPUs
4707 (specified by rcu_nocbs= above) explicitly
4708 awaken the corresponding "rcuoN" kthreads,
4709 make these kthreads poll for callbacks.
4710 This improves the real-time response for the
4711 offloaded CPUs by relieving them of the need to
4712 wake up the corresponding kthread, but degrades
4713 energy efficiency by requiring that the kthreads
4714 periodically wake up to do the polling.
4716 rcutree.blimit= [KNL]
4717 Set maximum number of finished RCU callbacks to
4718 process in one batch.
4720 rcutree.dump_tree= [KNL]
4721 Dump the structure of the rcu_node combining tree
4722 out at early boot. This is used for diagnostic
4723 purposes, to verify correct tree setup.
4725 rcutree.gp_cleanup_delay= [KNL]
4726 Set the number of jiffies to delay each step of
4727 RCU grace-period cleanup.
4729 rcutree.gp_init_delay= [KNL]
4730 Set the number of jiffies to delay each step of
4731 RCU grace-period initialization.
4733 rcutree.gp_preinit_delay= [KNL]
4734 Set the number of jiffies to delay each step of
4735 RCU grace-period pre-initialization, that is,
4736 the propagation of recent CPU-hotplug changes up
4737 the rcu_node combining tree.
4739 rcutree.use_softirq= [KNL]
4740 If set to zero, move all RCU_SOFTIRQ processing to
4741 per-CPU rcuc kthreads. Defaults to a non-zero
4742 value, meaning that RCU_SOFTIRQ is used by default.
4743 Specify rcutree.use_softirq=0 to use rcuc kthreads.
4745 But note that CONFIG_PREEMPT_RT=y kernels disable
4746 this kernel boot parameter, forcibly setting it
4749 rcutree.rcu_fanout_exact= [KNL]
4750 Disable autobalancing of the rcu_node combining
4751 tree. This is used by rcutorture, and might
4752 possibly be useful for architectures having high
4753 cache-to-cache transfer latencies.
4755 rcutree.rcu_fanout_leaf= [KNL]
4756 Change the number of CPUs assigned to each
4757 leaf rcu_node structure. Useful for very
4758 large systems, which will choose the value 64,
4759 and for NUMA systems with large remote-access
4760 latencies, which will choose a value aligned
4761 with the appropriate hardware boundaries.
4763 rcutree.rcu_min_cached_objs= [KNL]
4764 Minimum number of objects which are cached and
4765 maintained per one CPU. Object size is equal
4766 to PAGE_SIZE. The cache allows to reduce the
4767 pressure to page allocator, also it makes the
4768 whole algorithm to behave better in low memory
4771 rcutree.rcu_delay_page_cache_fill_msec= [KNL]
4772 Set the page-cache refill delay (in milliseconds)
4773 in response to low-memory conditions. The range
4774 of permitted values is in the range 0:100000.
4776 rcutree.jiffies_till_first_fqs= [KNL]
4777 Set delay from grace-period initialization to
4778 first attempt to force quiescent states.
4779 Units are jiffies, minimum value is zero,
4780 and maximum value is HZ.
4782 rcutree.jiffies_till_next_fqs= [KNL]
4783 Set delay between subsequent attempts to force
4784 quiescent states. Units are jiffies, minimum
4785 value is one, and maximum value is HZ.
4787 rcutree.jiffies_till_sched_qs= [KNL]
4788 Set required age in jiffies for a
4789 given grace period before RCU starts
4790 soliciting quiescent-state help from
4791 rcu_note_context_switch() and cond_resched().
4792 If not specified, the kernel will calculate
4793 a value based on the most recent settings
4794 of rcutree.jiffies_till_first_fqs
4795 and rcutree.jiffies_till_next_fqs.
4796 This calculated value may be viewed in
4797 rcutree.jiffies_to_sched_qs. Any attempt to set
4798 rcutree.jiffies_to_sched_qs will be cheerfully
4801 rcutree.kthread_prio= [KNL,BOOT]
4802 Set the SCHED_FIFO priority of the RCU per-CPU
4803 kthreads (rcuc/N). This value is also used for
4804 the priority of the RCU boost threads (rcub/N)
4805 and for the RCU grace-period kthreads (rcu_bh,
4806 rcu_preempt, and rcu_sched). If RCU_BOOST is
4807 set, valid values are 1-99 and the default is 1
4808 (the least-favored priority). Otherwise, when
4809 RCU_BOOST is not set, valid values are 0-99 and
4810 the default is zero (non-realtime operation).
4811 When RCU_NOCB_CPU is set, also adjust the
4812 priority of NOCB callback kthreads.
4814 rcutree.rcu_divisor= [KNL]
4815 Set the shift-right count to use to compute
4816 the callback-invocation batch limit bl from
4817 the number of callbacks queued on this CPU.
4818 The result will be bounded below by the value of
4819 the rcutree.blimit kernel parameter. Every bl
4820 callbacks, the softirq handler will exit in
4821 order to allow the CPU to do other work.
4823 Please note that this callback-invocation batch
4824 limit applies only to non-offloaded callback
4825 invocation. Offloaded callbacks are instead
4826 invoked in the context of an rcuoc kthread, which
4827 scheduler will preempt as it does any other task.
4829 rcutree.nocb_nobypass_lim_per_jiffy= [KNL]
4830 On callback-offloaded (rcu_nocbs) CPUs,
4831 RCU reduces the lock contention that would
4832 otherwise be caused by callback floods through
4833 use of the ->nocb_bypass list. However, in the
4834 common non-flooded case, RCU queues directly to
4835 the main ->cblist in order to avoid the extra
4836 overhead of the ->nocb_bypass list and its lock.
4837 But if there are too many callbacks queued during
4838 a single jiffy, RCU pre-queues the callbacks into
4839 the ->nocb_bypass queue. The definition of "too
4840 many" is supplied by this kernel boot parameter.
4842 rcutree.rcu_nocb_gp_stride= [KNL]
4843 Set the number of NOCB callback kthreads in
4844 each group, which defaults to the square root
4845 of the number of CPUs. Larger numbers reduce
4846 the wakeup overhead on the global grace-period
4847 kthread, but increases that same overhead on
4848 each group's NOCB grace-period kthread.
4850 rcutree.qhimark= [KNL]
4851 Set threshold of queued RCU callbacks beyond which
4852 batch limiting is disabled.
4854 rcutree.qlowmark= [KNL]
4855 Set threshold of queued RCU callbacks below which
4856 batch limiting is re-enabled.
4858 rcutree.qovld= [KNL]
4859 Set threshold of queued RCU callbacks beyond which
4860 RCU's force-quiescent-state scan will aggressively
4861 enlist help from cond_resched() and sched IPIs to
4862 help CPUs more quickly reach quiescent states.
4863 Set to less than zero to make this be set based
4864 on rcutree.qhimark at boot time and to zero to
4865 disable more aggressive help enlistment.
4867 rcutree.rcu_kick_kthreads= [KNL]
4868 Cause the grace-period kthread to get an extra
4869 wake_up() if it sleeps three times longer than
4870 it should at force-quiescent-state time.
4871 This wake_up() will be accompanied by a
4872 WARN_ONCE() splat and an ftrace_dump().
4874 rcutree.rcu_unlock_delay= [KNL]
4875 In CONFIG_RCU_STRICT_GRACE_PERIOD=y kernels,
4876 this specifies an rcu_read_unlock()-time delay
4877 in microseconds. This defaults to zero.
4878 Larger delays increase the probability of
4879 catching RCU pointer leaks, that is, buggy use
4880 of RCU-protected pointers after the relevant
4881 rcu_read_unlock() has completed.
4883 rcutree.sysrq_rcu= [KNL]
4884 Commandeer a sysrq key to dump out Tree RCU's
4885 rcu_node tree with an eye towards determining
4886 why a new grace period has not yet started.
4888 rcuscale.gp_async= [KNL]
4889 Measure performance of asynchronous
4890 grace-period primitives such as call_rcu().
4892 rcuscale.gp_async_max= [KNL]
4893 Specify the maximum number of outstanding
4894 callbacks per writer thread. When a writer
4895 thread exceeds this limit, it invokes the
4896 corresponding flavor of rcu_barrier() to allow
4897 previously posted callbacks to drain.
4899 rcuscale.gp_exp= [KNL]
4900 Measure performance of expedited synchronous
4901 grace-period primitives.
4903 rcuscale.holdoff= [KNL]
4904 Set test-start holdoff period. The purpose of
4905 this parameter is to delay the start of the
4906 test until boot completes in order to avoid
4909 rcuscale.kfree_rcu_test= [KNL]
4910 Set to measure performance of kfree_rcu() flooding.
4912 rcuscale.kfree_rcu_test_double= [KNL]
4913 Test the double-argument variant of kfree_rcu().
4914 If this parameter has the same value as
4915 rcuscale.kfree_rcu_test_single, both the single-
4916 and double-argument variants are tested.
4918 rcuscale.kfree_rcu_test_single= [KNL]
4919 Test the single-argument variant of kfree_rcu().
4920 If this parameter has the same value as
4921 rcuscale.kfree_rcu_test_double, both the single-
4922 and double-argument variants are tested.
4924 rcuscale.kfree_nthreads= [KNL]
4925 The number of threads running loops of kfree_rcu().
4927 rcuscale.kfree_alloc_num= [KNL]
4928 Number of allocations and frees done in an iteration.
4930 rcuscale.kfree_loops= [KNL]
4931 Number of loops doing rcuscale.kfree_alloc_num number
4932 of allocations and frees.
4934 rcuscale.nreaders= [KNL]
4935 Set number of RCU readers. The value -1 selects
4936 N, where N is the number of CPUs. A value
4937 "n" less than -1 selects N-n+1, where N is again
4938 the number of CPUs. For example, -2 selects N
4939 (the number of CPUs), -3 selects N+1, and so on.
4940 A value of "n" less than or equal to -N selects
4943 rcuscale.nwriters= [KNL]
4944 Set number of RCU writers. The values operate
4945 the same as for rcuscale.nreaders.
4946 N, where N is the number of CPUs
4948 rcuscale.perf_type= [KNL]
4949 Specify the RCU implementation to test.
4951 rcuscale.shutdown= [KNL]
4952 Shut the system down after performance tests
4953 complete. This is useful for hands-off automated
4956 rcuscale.verbose= [KNL]
4957 Enable additional printk() statements.
4959 rcuscale.writer_holdoff= [KNL]
4960 Write-side holdoff between grace periods,
4961 in microseconds. The default of zero says
4964 rcutorture.fqs_duration= [KNL]
4965 Set duration of force_quiescent_state bursts
4968 rcutorture.fqs_holdoff= [KNL]
4969 Set holdoff time within force_quiescent_state bursts
4972 rcutorture.fqs_stutter= [KNL]
4973 Set wait time between force_quiescent_state bursts
4976 rcutorture.fwd_progress= [KNL]
4977 Specifies the number of kthreads to be used
4978 for RCU grace-period forward-progress testing
4979 for the types of RCU supporting this notion.
4980 Defaults to 1 kthread, values less than zero or
4981 greater than the number of CPUs cause the number
4984 rcutorture.fwd_progress_div= [KNL]
4985 Specify the fraction of a CPU-stall-warning
4986 period to do tight-loop forward-progress testing.
4988 rcutorture.fwd_progress_holdoff= [KNL]
4989 Number of seconds to wait between successive
4990 forward-progress tests.
4992 rcutorture.fwd_progress_need_resched= [KNL]
4993 Enclose cond_resched() calls within checks for
4994 need_resched() during tight-loop forward-progress
4997 rcutorture.gp_cond= [KNL]
4998 Use conditional/asynchronous update-side
4999 primitives, if available.
5001 rcutorture.gp_exp= [KNL]
5002 Use expedited update-side primitives, if available.
5004 rcutorture.gp_normal= [KNL]
5005 Use normal (non-expedited) asynchronous
5006 update-side primitives, if available.
5008 rcutorture.gp_sync= [KNL]
5009 Use normal (non-expedited) synchronous
5010 update-side primitives, if available. If all
5011 of rcutorture.gp_cond=, rcutorture.gp_exp=,
5012 rcutorture.gp_normal=, and rcutorture.gp_sync=
5013 are zero, rcutorture acts as if is interpreted
5014 they are all non-zero.
5016 rcutorture.irqreader= [KNL]
5017 Run RCU readers from irq handlers, or, more
5018 accurately, from a timer handler. Not all RCU
5019 flavors take kindly to this sort of thing.
5021 rcutorture.leakpointer= [KNL]
5022 Leak an RCU-protected pointer out of the reader.
5023 This can of course result in splats, and is
5024 intended to test the ability of things like
5025 CONFIG_RCU_STRICT_GRACE_PERIOD=y to detect
5028 rcutorture.n_barrier_cbs= [KNL]
5029 Set callbacks/threads for rcu_barrier() testing.
5031 rcutorture.nfakewriters= [KNL]
5032 Set number of concurrent RCU writers. These just
5033 stress RCU, they don't participate in the actual
5034 test, hence the "fake".
5036 rcutorture.nocbs_nthreads= [KNL]
5037 Set number of RCU callback-offload togglers.
5038 Zero (the default) disables toggling.
5040 rcutorture.nocbs_toggle= [KNL]
5041 Set the delay in milliseconds between successive
5042 callback-offload toggling attempts.
5044 rcutorture.nreaders= [KNL]
5045 Set number of RCU readers. The value -1 selects
5046 N-1, where N is the number of CPUs. A value
5047 "n" less than -1 selects N-n-2, where N is again
5048 the number of CPUs. For example, -2 selects N
5049 (the number of CPUs), -3 selects N+1, and so on.
5051 rcutorture.object_debug= [KNL]
5052 Enable debug-object double-call_rcu() testing.
5054 rcutorture.onoff_holdoff= [KNL]
5055 Set time (s) after boot for CPU-hotplug testing.
5057 rcutorture.onoff_interval= [KNL]
5058 Set time (jiffies) between CPU-hotplug operations,
5059 or zero to disable CPU-hotplug testing.
5061 rcutorture.read_exit= [KNL]
5062 Set the number of read-then-exit kthreads used
5063 to test the interaction of RCU updaters and
5064 task-exit processing.
5066 rcutorture.read_exit_burst= [KNL]
5067 The number of times in a given read-then-exit
5068 episode that a set of read-then-exit kthreads
5071 rcutorture.read_exit_delay= [KNL]
5072 The delay, in seconds, between successive
5073 read-then-exit testing episodes.
5075 rcutorture.shuffle_interval= [KNL]
5076 Set task-shuffle interval (s). Shuffling tasks
5077 allows some CPUs to go into dyntick-idle mode
5078 during the rcutorture test.
5080 rcutorture.shutdown_secs= [KNL]
5081 Set time (s) after boot system shutdown. This
5082 is useful for hands-off automated testing.
5084 rcutorture.stall_cpu= [KNL]
5085 Duration of CPU stall (s) to test RCU CPU stall
5086 warnings, zero to disable.
5088 rcutorture.stall_cpu_block= [KNL]
5089 Sleep while stalling if set. This will result
5090 in warnings from preemptible RCU in addition
5091 to any other stall-related activity.
5093 rcutorture.stall_cpu_holdoff= [KNL]
5094 Time to wait (s) after boot before inducing stall.
5096 rcutorture.stall_cpu_irqsoff= [KNL]
5097 Disable interrupts while stalling if set.
5099 rcutorture.stall_gp_kthread= [KNL]
5100 Duration (s) of forced sleep within RCU
5101 grace-period kthread to test RCU CPU stall
5102 warnings, zero to disable. If both stall_cpu
5103 and stall_gp_kthread are specified, the
5104 kthread is starved first, then the CPU.
5106 rcutorture.stat_interval= [KNL]
5107 Time (s) between statistics printk()s.
5109 rcutorture.stutter= [KNL]
5110 Time (s) to stutter testing, for example, specifying
5111 five seconds causes the test to run for five seconds,
5112 wait for five seconds, and so on. This tests RCU's
5113 ability to transition abruptly to and from idle.
5115 rcutorture.test_boost= [KNL]
5116 Test RCU priority boosting? 0=no, 1=maybe, 2=yes.
5117 "Maybe" means test if the RCU implementation
5118 under test support RCU priority boosting.
5120 rcutorture.test_boost_duration= [KNL]
5121 Duration (s) of each individual boost test.
5123 rcutorture.test_boost_interval= [KNL]
5124 Interval (s) between each boost test.
5126 rcutorture.test_no_idle_hz= [KNL]
5127 Test RCU's dyntick-idle handling. See also the
5128 rcutorture.shuffle_interval parameter.
5130 rcutorture.torture_type= [KNL]
5131 Specify the RCU implementation to test.
5133 rcutorture.verbose= [KNL]
5134 Enable additional printk() statements.
5136 rcupdate.rcu_cpu_stall_ftrace_dump= [KNL]
5137 Dump ftrace buffer after reporting RCU CPU
5140 rcupdate.rcu_cpu_stall_suppress= [KNL]
5141 Suppress RCU CPU stall warning messages.
5143 rcupdate.rcu_cpu_stall_suppress_at_boot= [KNL]
5144 Suppress RCU CPU stall warning messages and
5145 rcutorture writer stall warnings that occur
5146 during early boot, that is, during the time
5147 before the init task is spawned.
5149 rcupdate.rcu_cpu_stall_timeout= [KNL]
5150 Set timeout for RCU CPU stall warning messages.
5151 The value is in seconds and the maximum allowed
5152 value is 300 seconds.
5154 rcupdate.rcu_exp_cpu_stall_timeout= [KNL]
5155 Set timeout for expedited RCU CPU stall warning
5156 messages. The value is in milliseconds
5157 and the maximum allowed value is 21000
5158 milliseconds. Please note that this value is
5159 adjusted to an arch timer tick resolution.
5160 Setting this to zero causes the value from
5161 rcupdate.rcu_cpu_stall_timeout to be used (after
5162 conversion from seconds to milliseconds).
5164 rcupdate.rcu_cpu_stall_cputime= [KNL]
5165 Provide statistics on the cputime and count of
5166 interrupts and tasks during the sampling period. For
5167 multiple continuous RCU stalls, all sampling periods
5168 begin at half of the first RCU stall timeout.
5170 rcupdate.rcu_exp_stall_task_details= [KNL]
5171 Print stack dumps of any tasks blocking the
5172 current expedited RCU grace period during an
5173 expedited RCU CPU stall warning.
5175 rcupdate.rcu_expedited= [KNL]
5176 Use expedited grace-period primitives, for
5177 example, synchronize_rcu_expedited() instead
5178 of synchronize_rcu(). This reduces latency,
5179 but can increase CPU utilization, degrade
5180 real-time latency, and degrade energy efficiency.
5181 No effect on CONFIG_TINY_RCU kernels.
5183 rcupdate.rcu_normal= [KNL]
5184 Use only normal grace-period primitives,
5185 for example, synchronize_rcu() instead of
5186 synchronize_rcu_expedited(). This improves
5187 real-time latency, CPU utilization, and
5188 energy efficiency, but can expose users to
5189 increased grace-period latency. This parameter
5190 overrides rcupdate.rcu_expedited. No effect on
5191 CONFIG_TINY_RCU kernels.
5193 rcupdate.rcu_normal_after_boot= [KNL]
5194 Once boot has completed (that is, after
5195 rcu_end_inkernel_boot() has been invoked), use
5196 only normal grace-period primitives. No effect
5197 on CONFIG_TINY_RCU kernels.
5199 But note that CONFIG_PREEMPT_RT=y kernels enables
5200 this kernel boot parameter, forcibly setting
5201 it to the value one, that is, converting any
5202 post-boot attempt at an expedited RCU grace
5203 period to instead use normal non-expedited
5204 grace-period processing.
5206 rcupdate.rcu_task_collapse_lim= [KNL]
5207 Set the maximum number of callbacks present
5208 at the beginning of a grace period that allows
5209 the RCU Tasks flavors to collapse back to using
5210 a single callback queue. This switching only
5211 occurs when rcupdate.rcu_task_enqueue_lim is
5212 set to the default value of -1.
5214 rcupdate.rcu_task_contend_lim= [KNL]
5215 Set the minimum number of callback-queuing-time
5216 lock-contention events per jiffy required to
5217 cause the RCU Tasks flavors to switch to per-CPU
5218 callback queuing. This switching only occurs
5219 when rcupdate.rcu_task_enqueue_lim is set to
5220 the default value of -1.
5222 rcupdate.rcu_task_enqueue_lim= [KNL]
5223 Set the number of callback queues to use for the
5224 RCU Tasks family of RCU flavors. The default
5225 of -1 allows this to be automatically (and
5226 dynamically) adjusted. This parameter is intended
5229 rcupdate.rcu_task_ipi_delay= [KNL]
5230 Set time in jiffies during which RCU tasks will
5231 avoid sending IPIs, starting with the beginning
5232 of a given grace period. Setting a large
5233 number avoids disturbing real-time workloads,
5234 but lengthens grace periods.
5236 rcupdate.rcu_task_stall_info= [KNL]
5237 Set initial timeout in jiffies for RCU task stall
5238 informational messages, which give some indication
5239 of the problem for those not patient enough to
5240 wait for ten minutes. Informational messages are
5241 only printed prior to the stall-warning message
5242 for a given grace period. Disable with a value
5243 less than or equal to zero. Defaults to ten
5244 seconds. A change in value does not take effect
5245 until the beginning of the next grace period.
5247 rcupdate.rcu_task_stall_info_mult= [KNL]
5248 Multiplier for time interval between successive
5249 RCU task stall informational messages for a given
5250 RCU tasks grace period. This value is clamped
5251 to one through ten, inclusive. It defaults to
5252 the value three, so that the first informational
5253 message is printed 10 seconds into the grace
5254 period, the second at 40 seconds, the third at
5255 160 seconds, and then the stall warning at 600
5256 seconds would prevent a fourth at 640 seconds.
5258 rcupdate.rcu_task_stall_timeout= [KNL]
5259 Set timeout in jiffies for RCU task stall
5260 warning messages. Disable with a value less
5261 than or equal to zero. Defaults to ten minutes.
5262 A change in value does not take effect until
5263 the beginning of the next grace period.
5265 rcupdate.rcu_self_test= [KNL]
5266 Run the RCU early boot self tests
5270 Run specified binary instead of /init from the ramdisk,
5271 used for early userspace startup. See initrd.
5274 force - Override the decision by the kernel to hide the
5275 advertisement of RDRAND support (this affects
5276 certain AMD processors because of buggy BIOS
5277 support, specifically around the suspend/resume
5281 Turn on/off individual RDT features. List is:
5282 cmt, mbmtotal, mbmlocal, l3cat, l3cdp, l2cat, l2cdp,
5284 E.g. to turn on cmt and turn off mba use:
5288 Format (x86 or x86_64):
5289 [w[arm] | c[old] | h[ard] | s[oft] | g[pio]] | d[efault] \
5291 [[,]b[ios] | a[cpi] | k[bd] | t[riple] | e[fi] | p[ci]] \
5293 Where reboot_mode is one of warm (soft) or cold (hard) or gpio
5294 (prefix with 'panic_' to set mode for panic
5296 reboot_type is one of bios, acpi, kbd, triple, efi, or pci,
5297 reboot_force is either force or not specified,
5298 reboot_cpu is s[mp]#### with #### being the processor
5299 to be used for rebooting.
5301 refscale.holdoff= [KNL]
5302 Set test-start holdoff period. The purpose of
5303 this parameter is to delay the start of the
5304 test until boot completes in order to avoid
5307 refscale.loops= [KNL]
5308 Set the number of loops over the synchronization
5309 primitive under test. Increasing this number
5310 reduces noise due to loop start/end overhead,
5311 but the default has already reduced the per-pass
5312 noise to a handful of picoseconds on ca. 2020
5315 refscale.nreaders= [KNL]
5316 Set number of readers. The default value of -1
5317 selects N, where N is roughly 75% of the number
5318 of CPUs. A value of zero is an interesting choice.
5320 refscale.nruns= [KNL]
5321 Set number of runs, each of which is dumped onto
5324 refscale.readdelay= [KNL]
5325 Set the read-side critical-section duration,
5326 measured in microseconds.
5328 refscale.scale_type= [KNL]
5329 Specify the read-protection implementation to test.
5331 refscale.shutdown= [KNL]
5332 Shut down the system at the end of the performance
5333 test. This defaults to 1 (shut it down) when
5334 refscale is built into the kernel and to 0 (leave
5335 it running) when refscale is built as a module.
5337 refscale.verbose= [KNL]
5338 Enable additional printk() statements.
5340 refscale.verbose_batched= [KNL]
5341 Batch the additional printk() statements. If zero
5342 (the default) or negative, print everything. Otherwise,
5343 print every Nth verbose statement, where N is the value
5347 [KNL, SMP] Set scheduler's default relax_domain_level.
5348 See Documentation/admin-guide/cgroup-v1/cpusets.rst.
5350 reserve= [KNL,BUGS] Force kernel to ignore I/O ports or memory
5351 Format: <base1>,<size1>[,<base2>,<size2>,...]
5352 Reserve I/O ports or memory so the kernel won't use
5353 them. If <base> is less than 0x10000, the region
5354 is assumed to be I/O ports; otherwise it is memory.
5356 reservetop= [X86-32]
5358 Reserves a hole at the top of the kernel virtual
5361 reset_devices [KNL] Force drivers to reset the underlying device
5362 during initialization.
5365 Specify the partition device for software suspend
5367 {/dev/<dev> | PARTUUID=<uuid> | <int>:<int> | <hex>}
5369 resume_offset= [SWSUSP]
5370 Specify the offset from the beginning of the partition
5371 given by "resume=" at which the swap header is located,
5372 in <PAGE_SIZE> units (needed only for swap files).
5373 See Documentation/power/swsusp-and-swap-files.rst
5375 resumedelay= [HIBERNATION] Delay (in seconds) to pause before attempting to
5376 read the resume files
5378 resumewait [HIBERNATION] Wait (indefinitely) for resume device to show up.
5379 Useful for devices that are detected asynchronously
5380 (e.g. USB and MMC devices).
5382 retain_initrd [RAM] Keep initrd memory after extraction
5384 retbleed= [X86] Control mitigation of RETBleed (Arbitrary
5385 Speculative Code Execution with Return Instructions)
5388 AMD-based UNRET and IBPB mitigations alone do not stop
5389 sibling threads from influencing the predictions of other
5390 sibling threads. For that reason, STIBP is used on pro-
5391 cessors that support it, and mitigate SMT on processors
5395 auto - automatically select a migitation
5396 auto,nosmt - automatically select a mitigation,
5397 disabling SMT if necessary for
5398 the full mitigation (only on Zen1
5399 and older without STIBP).
5400 ibpb - On AMD, mitigate short speculation
5401 windows on basic block boundaries too.
5402 Safe, highest perf impact. It also
5403 enables STIBP if present. Not suitable
5405 ibpb,nosmt - Like "ibpb" above but will disable SMT
5406 when STIBP is not available. This is
5407 the alternative for systems which do not
5409 unret - Force enable untrained return thunks,
5410 only effective on AMD f15h-f17h based
5412 unret,nosmt - Like unret, but will disable SMT when STIBP
5413 is not available. This is the alternative for
5414 systems which do not have STIBP.
5416 Selecting 'auto' will choose a mitigation method at run
5417 time according to the CPU.
5419 Not specifying this option is equivalent to retbleed=auto.
5421 rfkill.default_state=
5422 0 "airplane mode". All wifi, bluetooth, wimax, gps, fm,
5423 etc. communication is blocked by default.
5426 rfkill.master_switch_mode=
5427 0 The "airplane mode" button does nothing.
5428 1 The "airplane mode" button toggles between everything
5429 blocked and the previous configuration.
5430 2 The "airplane mode" button toggles between everything
5431 blocked and everything unblocked.
5433 rhash_entries= [KNL,NET]
5434 Set number of hash buckets for route cache
5437 [KNL] Disable ring 3 MONITOR/MWAIT feature on supported
5440 ro [KNL] Mount root device read-only on boot
5443 on Mark read-only kernel memory as read-only (default).
5444 off Leave read-only kernel memory writable for debugging.
5445 full Mark read-only kernel memory and aliases as read-only
5449 Enable the uart passthrough on the designated usb port
5450 on Rockchip SoCs. When active, the signals of the
5451 debug-uart get routed to the D+ and D- pins of the usb
5452 port and the regular usb controller gets disabled.
5454 root= [KNL] Root filesystem
5455 See name_to_dev_t comment in init/do_mounts.c.
5457 rootdelay= [KNL] Delay (in seconds) to pause before attempting to
5458 mount the root filesystem
5460 rootflags= [KNL] Set root filesystem mount option string
5462 rootfstype= [KNL] Set root filesystem type
5464 rootwait [KNL] Wait (indefinitely) for root device to show up.
5465 Useful for devices that are detected asynchronously
5466 (e.g. USB and MMC devices).
5468 rproc_mem=nn[KMG][@address]
5469 [KNL,ARM,CMA] Remoteproc physical memory block.
5470 Memory area to be used by remote processor image,
5473 rw [KNL] Mount root device read-write on boot
5475 S [KNL] Run init in single mode
5477 s390_iommu= [HW,S390]
5478 Set s390 IOTLB flushing mode
5480 With strict flushing every unmap operation will result in
5481 an IOTLB flush. Default is lazy flushing before reuse,
5484 s390_iommu_aperture= [KNL,S390]
5485 Specifies the size of the per device DMA address space
5486 accessible through the DMA and IOMMU APIs as a decimal
5487 factor of the size of main memory.
5488 The default is 1 meaning that one can concurrently use
5489 as many DMA addresses as physical memory is installed,
5490 if supported by hardware, and thus map all of memory
5491 once. With a value of 2 one can map all of memory twice
5492 and so on. As a special case a factor of 0 imposes no
5493 restrictions other than those given by hardware at the
5494 cost of significant additional memory use for tables.
5497 See drivers/net/irda/sa1100_ir.c.
5499 sched_verbose [KNL] Enables verbose scheduler debug messages.
5501 schedstats= [KNL,X86] Enable or disable scheduled statistics.
5502 Allowed values are enable and disable. This feature
5503 incurs a small amount of overhead in the scheduler
5504 but is useful for debugging and performance tuning.
5506 sched_thermal_decay_shift=
5507 [KNL, SMP] Set a decay shift for scheduler thermal
5508 pressure signal. Thermal pressure signal follows the
5509 default decay period of other scheduler pelt
5510 signals(usually 32 ms but configurable). Setting
5511 sched_thermal_decay_shift will left shift the decay
5512 period for the thermal pressure signal by the shift
5514 i.e. with the default pelt decay period of 32 ms
5515 sched_thermal_decay_shift thermal pressure decay pr
5519 Format: integer between 0 and 10
5522 scftorture.holdoff= [KNL]
5523 Number of seconds to hold off before starting
5524 test. Defaults to zero for module insertion and
5525 to 10 seconds for built-in smp_call_function()
5528 scftorture.longwait= [KNL]
5529 Request ridiculously long waits randomly selected
5530 up to the chosen limit in seconds. Zero (the
5531 default) disables this feature. Please note
5532 that requesting even small non-zero numbers of
5533 seconds can result in RCU CPU stall warnings,
5534 softlockup complaints, and so on.
5536 scftorture.nthreads= [KNL]
5537 Number of kthreads to spawn to invoke the
5538 smp_call_function() family of functions.
5539 The default of -1 specifies a number of kthreads
5540 equal to the number of CPUs.
5542 scftorture.onoff_holdoff= [KNL]
5543 Number seconds to wait after the start of the
5544 test before initiating CPU-hotplug operations.
5546 scftorture.onoff_interval= [KNL]
5547 Number seconds to wait between successive
5548 CPU-hotplug operations. Specifying zero (which
5549 is the default) disables CPU-hotplug operations.
5551 scftorture.shutdown_secs= [KNL]
5552 The number of seconds following the start of the
5553 test after which to shut down the system. The
5554 default of zero avoids shutting down the system.
5555 Non-zero values are useful for automated tests.
5557 scftorture.stat_interval= [KNL]
5558 The number of seconds between outputting the
5559 current test statistics to the console. A value
5560 of zero disables statistics output.
5562 scftorture.stutter_cpus= [KNL]
5563 The number of jiffies to wait between each change
5564 to the set of CPUs under test.
5566 scftorture.use_cpus_read_lock= [KNL]
5567 Use use_cpus_read_lock() instead of the default
5568 preempt_disable() to disable CPU hotplug
5569 while invoking one of the smp_call_function*()
5572 scftorture.verbose= [KNL]
5573 Enable additional printk() statements.
5575 scftorture.weight_single= [KNL]
5576 The probability weighting to use for the
5577 smp_call_function_single() function with a zero
5578 "wait" parameter. A value of -1 selects the
5579 default if all other weights are -1. However,
5580 if at least one weight has some other value, a
5581 value of -1 will instead select a weight of zero.
5583 scftorture.weight_single_wait= [KNL]
5584 The probability weighting to use for the
5585 smp_call_function_single() function with a
5586 non-zero "wait" parameter. See weight_single.
5588 scftorture.weight_many= [KNL]
5589 The probability weighting to use for the
5590 smp_call_function_many() function with a zero
5591 "wait" parameter. See weight_single.
5592 Note well that setting a high probability for
5593 this weighting can place serious IPI load
5596 scftorture.weight_many_wait= [KNL]
5597 The probability weighting to use for the
5598 smp_call_function_many() function with a
5599 non-zero "wait" parameter. See weight_single
5602 scftorture.weight_all= [KNL]
5603 The probability weighting to use for the
5604 smp_call_function_all() function with a zero
5605 "wait" parameter. See weight_single and
5608 scftorture.weight_all_wait= [KNL]
5609 The probability weighting to use for the
5610 smp_call_function_all() function with a
5611 non-zero "wait" parameter. See weight_single
5614 skew_tick= [KNL] Offset the periodic timer tick per cpu to mitigate
5615 xtime_lock contention on larger systems, and/or RCU lock
5616 contention on all systems with CONFIG_MAXSMP set.
5617 Format: { "0" | "1" }
5618 0 -- disable. (may be 1 via CONFIG_CMDLINE="skew_tick=1"
5620 Note: increases power consumption, thus should only be
5621 enabled if running jitter sensitive (HPC/RT) workloads.
5623 security= [SECURITY] Choose a legacy "major" security module to
5624 enable at boot. This has been deprecated by the
5627 selinux= [SELINUX] Disable or enable SELinux at boot time.
5628 Format: { "0" | "1" }
5629 See security/selinux/Kconfig help text.
5634 serialnumber [BUGS=X86-32]
5636 sev=option[,option...] [X86-64] See Documentation/arch/x86/x86_64/boot-options.rst
5639 Maximal number of shapers.
5641 show_lapic= [APIC,X86] Advanced Programmable Interrupt Controller
5642 Limit apic dumping. The parameter defines the maximal
5643 number of local apics being dumped. Also it is possible
5644 to set it to "all" by meaning -- no limit here.
5645 Format: { 1 (default) | 2 | ... | all }.
5646 The parameter valid if only apic=debug or
5647 apic=verbose is specified.
5648 Example: apic=debug show_lapic=all
5656 Enable merging of slabs with similar size when the
5657 kernel is built without CONFIG_SLAB_MERGE_DEFAULT.
5660 Disable merging of slabs with similar size. May be
5661 necessary if there is some reason to distinguish
5662 allocs to different slabs, especially in hardened
5663 environments where the risk of heap overflows and
5664 layout control by attackers can usually be
5665 frustrated by disabling merging. This will reduce
5666 most of the exposure of a heap attack to a single
5667 cache (risks via metadata attacks are mostly
5668 unchanged). Debug options disable merging on their
5670 For more information see Documentation/mm/slub.rst.
5672 slab_max_order= [MM, SLAB]
5673 Determines the maximum allowed order for slabs.
5674 A high setting may cause OOMs due to memory
5675 fragmentation. Defaults to 1 for systems with
5676 more than 32MB of RAM, 0 otherwise.
5678 slub_debug[=options[,slabs][;[options[,slabs]]...] [MM, SLUB]
5679 Enabling slub_debug allows one to determine the
5680 culprit if slab objects become corrupted. Enabling
5681 slub_debug can create guard zones around objects and
5682 may poison objects when not in use. Also tracks the
5683 last alloc / free. For more information see
5684 Documentation/mm/slub.rst.
5686 slub_max_order= [MM, SLUB]
5687 Determines the maximum allowed order for slabs.
5688 A high setting may cause OOMs due to memory
5689 fragmentation. For more information see
5690 Documentation/mm/slub.rst.
5692 slub_min_objects= [MM, SLUB]
5693 The minimum number of objects per slab. SLUB will
5694 increase the slab order up to slub_max_order to
5695 generate a sufficiently large slab able to contain
5696 the number of objects indicated. The higher the number
5697 of objects the smaller the overhead of tracking slabs
5698 and the less frequently locks need to be acquired.
5699 For more information see Documentation/mm/slub.rst.
5701 slub_min_order= [MM, SLUB]
5702 Determines the minimum page order for slabs. Must be
5703 lower than slub_max_order.
5704 For more information see Documentation/mm/slub.rst.
5706 slub_merge [MM, SLUB]
5707 Same with slab_merge.
5709 slub_nomerge [MM, SLUB]
5710 Same with slab_nomerge. This is supported for legacy.
5711 See slab_nomerge for more information.
5714 Format: <io1>[,<io2>[,...,<io8>]]
5716 smp.csd_lock_timeout= [KNL]
5717 Specify the period of time in milliseconds
5718 that smp_call_function() and friends will wait
5719 for a CPU to release the CSD lock. This is
5720 useful when diagnosing bugs involving CPUs
5721 disabling interrupts for extended periods
5722 of time. Defaults to 5,000 milliseconds, and
5723 setting a value of zero disables this feature.
5724 This feature may be more efficiently disabled
5725 using the csdlock_debug- kernel parameter.
5727 smsc-ircc2.nopnp [HW] Don't use PNP to discover SMC devices
5728 smsc-ircc2.ircc_cfg= [HW] Device configuration I/O port
5729 smsc-ircc2.ircc_sir= [HW] SIR base I/O port
5730 smsc-ircc2.ircc_fir= [HW] FIR base I/O port
5731 smsc-ircc2.ircc_irq= [HW] IRQ line
5732 smsc-ircc2.ircc_dma= [HW] DMA channel
5733 smsc-ircc2.ircc_transceiver= [HW] Transceiver type:
5734 0: Toshiba Satellite 1800 (GP data pin select)
5735 1: Fast pin select (default)
5738 smt= [KNL,S390] Set the maximum number of threads (logical
5739 CPUs) to use per physical CPU on systems capable of
5740 symmetric multithreading (SMT). Will be capped to the
5741 actual hardware limit.
5743 Default: -1 (no limit)
5746 [KNL] Should the soft-lockup detector generate panics.
5749 A value of 1 instructs the soft-lockup detector
5750 to panic the machine when a soft-lockup occurs. It is
5751 also controlled by the kernel.softlockup_panic sysctl
5752 and CONFIG_BOOTPARAM_SOFTLOCKUP_PANIC, which is the
5753 respective build-time switch to that functionality.
5755 softlockup_all_cpu_backtrace=
5756 [KNL] Should the soft-lockup detector generate
5757 backtraces on all cpus.
5760 sonypi.*= [HW] Sony Programmable I/O Control Device driver
5761 See Documentation/admin-guide/laptops/sonypi.rst
5763 spectre_v2= [X86] Control mitigation of Spectre variant 2
5764 (indirect branch speculation) vulnerability.
5765 The default operation protects the kernel from
5768 on - unconditionally enable, implies
5770 off - unconditionally disable, implies
5772 auto - kernel detects whether your CPU model is
5775 Selecting 'on' will, and 'auto' may, choose a
5776 mitigation method at run time according to the
5777 CPU, the available microcode, the setting of the
5778 CONFIG_RETPOLINE configuration option, and the
5779 compiler with which the kernel was built.
5781 Selecting 'on' will also enable the mitigation
5782 against user space to user space task attacks.
5784 Selecting 'off' will disable both the kernel and
5785 the user space protections.
5787 Specific mitigations can also be selected manually:
5789 retpoline - replace indirect branches
5790 retpoline,generic - Retpolines
5791 retpoline,lfence - LFENCE; indirect branch
5792 retpoline,amd - alias for retpoline,lfence
5793 eibrs - Enhanced/Auto IBRS
5794 eibrs,retpoline - Enhanced/Auto IBRS + Retpolines
5795 eibrs,lfence - Enhanced/Auto IBRS + LFENCE
5796 ibrs - use IBRS to protect kernel
5798 Not specifying this option is equivalent to
5802 [X86] Control mitigation of Spectre variant 2
5803 (indirect branch speculation) vulnerability between
5806 on - Unconditionally enable mitigations. Is
5807 enforced by spectre_v2=on
5809 off - Unconditionally disable mitigations. Is
5810 enforced by spectre_v2=off
5812 prctl - Indirect branch speculation is enabled,
5813 but mitigation can be enabled via prctl
5814 per thread. The mitigation control state
5815 is inherited on fork.
5818 - Like "prctl" above, but only STIBP is
5819 controlled per thread. IBPB is issued
5820 always when switching between different user
5824 - Same as "prctl" above, but all seccomp
5825 threads will enable the mitigation unless
5826 they explicitly opt out.
5829 - Like "seccomp" above, but only STIBP is
5830 controlled per thread. IBPB is issued
5831 always when switching between different
5832 user space processes.
5834 auto - Kernel selects the mitigation depending on
5835 the available CPU features and vulnerability.
5837 Default mitigation: "prctl"
5839 Not specifying this option is equivalent to
5840 spectre_v2_user=auto.
5842 spec_store_bypass_disable=
5843 [HW] Control Speculative Store Bypass (SSB) Disable mitigation
5844 (Speculative Store Bypass vulnerability)
5846 Certain CPUs are vulnerable to an exploit against a
5847 a common industry wide performance optimization known
5848 as "Speculative Store Bypass" in which recent stores
5849 to the same memory location may not be observed by
5850 later loads during speculative execution. The idea
5851 is that such stores are unlikely and that they can
5852 be detected prior to instruction retirement at the
5853 end of a particular speculation execution window.
5855 In vulnerable processors, the speculatively forwarded
5856 store can be used in a cache side channel attack, for
5857 example to read memory to which the attacker does not
5858 directly have access (e.g. inside sandboxed code).
5860 This parameter controls whether the Speculative Store
5861 Bypass optimization is used.
5863 On x86 the options are:
5865 on - Unconditionally disable Speculative Store Bypass
5866 off - Unconditionally enable Speculative Store Bypass
5867 auto - Kernel detects whether the CPU model contains an
5868 implementation of Speculative Store Bypass and
5869 picks the most appropriate mitigation. If the
5870 CPU is not vulnerable, "off" is selected. If the
5871 CPU is vulnerable the default mitigation is
5872 architecture and Kconfig dependent. See below.
5873 prctl - Control Speculative Store Bypass per thread
5874 via prctl. Speculative Store Bypass is enabled
5875 for a process by default. The state of the control
5876 is inherited on fork.
5877 seccomp - Same as "prctl" above, but all seccomp threads
5878 will disable SSB unless they explicitly opt out.
5880 Default mitigations:
5883 On powerpc the options are:
5885 on,auto - On Power8 and Power9 insert a store-forwarding
5886 barrier on kernel entry and exit. On Power7
5887 perform a software flush on kernel entry and
5891 Not specifying this option is equivalent to
5892 spec_store_bypass_disable=auto.
5894 spia_io_base= [HW,MTD]
5900 [X86] Enable split lock detection or bus lock detection
5902 When enabled (and if hardware support is present), atomic
5903 instructions that access data across cache line
5904 boundaries will result in an alignment check exception
5905 for split lock detection or a debug exception for
5910 warn - the kernel will emit rate-limited warnings
5911 about applications triggering the #AC
5912 exception or the #DB exception. This mode is
5913 the default on CPUs that support split lock
5914 detection or bus lock detection. Default
5915 behavior is by #AC if both features are
5916 enabled in hardware.
5918 fatal - the kernel will send SIGBUS to applications
5919 that trigger the #AC exception or the #DB
5920 exception. Default behavior is by #AC if
5921 both features are enabled in hardware.
5924 Set system wide rate limit to N bus locks
5925 per second for bus lock detection.
5928 N/A for split lock detection.
5931 If an #AC exception is hit in the kernel or in
5932 firmware (i.e. not while executing in user mode)
5933 the kernel will oops in either "warn" or "fatal"
5936 #DB exception for bus lock is triggered only when
5940 Control the Special Register Buffer Data Sampling
5943 Certain CPUs are vulnerable to an MDS-like
5944 exploit which can leak bits from the random
5947 By default, this issue is mitigated by
5948 microcode. However, the microcode fix can cause
5949 the RDRAND and RDSEED instructions to become
5950 much slower. Among other effects, this will
5951 result in reduced throughput from /dev/urandom.
5953 The microcode mitigation can be disabled with
5954 the following option:
5956 off: Disable mitigation and remove
5957 performance impact to RDRAND and RDSEED
5959 srcutree.big_cpu_lim [KNL]
5960 Specifies the number of CPUs constituting a
5961 large system, such that srcu_struct structures
5962 should immediately allocate an srcu_node array.
5963 This kernel-boot parameter defaults to 128,
5964 but takes effect only when the low-order four
5965 bits of srcutree.convert_to_big is equal to 3
5968 srcutree.convert_to_big [KNL]
5969 Specifies under what conditions an SRCU tree
5970 srcu_struct structure will be converted to big
5971 form, that is, with an rcu_node tree:
5974 1: At init_srcu_struct() time.
5975 2: When rcutorture decides to.
5976 3: Decide at boot time (default).
5977 0x1X: Above plus if high contention.
5979 Either way, the srcu_node tree will be sized based
5980 on the actual runtime number of CPUs (nr_cpu_ids)
5981 instead of the compile-time CONFIG_NR_CPUS.
5983 srcutree.counter_wrap_check [KNL]
5984 Specifies how frequently to check for
5985 grace-period sequence counter wrap for the
5986 srcu_data structure's ->srcu_gp_seq_needed field.
5987 The greater the number of bits set in this kernel
5988 parameter, the less frequently counter wrap will
5989 be checked for. Note that the bottom two bits
5992 srcutree.exp_holdoff [KNL]
5993 Specifies how many nanoseconds must elapse
5994 since the end of the last SRCU grace period for
5995 a given srcu_struct until the next normal SRCU
5996 grace period will be considered for automatic
5997 expediting. Set to zero to disable automatic
6000 srcutree.srcu_max_nodelay [KNL]
6001 Specifies the number of no-delay instances
6002 per jiffy for which the SRCU grace period
6003 worker thread will be rescheduled with zero
6004 delay. Beyond this limit, worker thread will
6005 be rescheduled with a sleep delay of one jiffy.
6007 srcutree.srcu_max_nodelay_phase [KNL]
6008 Specifies the per-grace-period phase, number of
6009 non-sleeping polls of readers. Beyond this limit,
6010 grace period worker thread will be rescheduled
6011 with a sleep delay of one jiffy, between each
6012 rescan of the readers, for a grace period phase.
6014 srcutree.srcu_retry_check_delay [KNL]
6015 Specifies number of microseconds of non-sleeping
6016 delay between each non-sleeping poll of readers.
6018 srcutree.small_contention_lim [KNL]
6019 Specifies the number of update-side contention
6020 events per jiffy will be tolerated before
6021 initiating a conversion of an srcu_struct
6022 structure to big form. Note that the value of
6023 srcutree.convert_to_big must have the 0x10 bit
6024 set for contention-based conversions to occur.
6027 Speculative Store Bypass Disable control
6029 On CPUs that are vulnerable to the Speculative
6030 Store Bypass vulnerability and offer a
6031 firmware based mitigation, this parameter
6032 indicates how the mitigation should be used:
6034 force-on: Unconditionally enable mitigation for
6035 for both kernel and userspace
6036 force-off: Unconditionally disable mitigation for
6037 for both kernel and userspace
6038 kernel: Always enable mitigation in the
6039 kernel, and offer a prctl interface
6040 to allow userspace to register its
6041 interest in being mitigated too.
6043 stack_guard_gap= [MM]
6044 override the default stack gap protection. The value
6045 is in page units and it defines how many pages prior
6046 to (for stacks growing down) resp. after (for stacks
6047 growing up) the main stack are reserved for no other
6048 mapping. Default value is 256 pages.
6050 stack_depot_disable= [KNL]
6051 Setting this to true through kernel command line will
6052 disable the stack depot thereby saving the static memory
6053 consumed by the stack hash table. By default this is set
6057 Enabled the stack tracer on boot up.
6059 stacktrace_filter=[function-list]
6060 [FTRACE] Limit the functions that the stack tracer
6061 will trace at boot up. function-list is a comma-separated
6062 list of functions. This list can be changed at run
6063 time by the stack_trace_filter file in the debugfs
6064 tracing directory. Note, this enables stack tracing
6065 and the stacktrace above is not needed.
6069 Set the STI (builtin display/keyboard on the HP-PARISC
6070 machines) console (graphic card) which should be used
6071 as the initial boot-console.
6072 See also comment in drivers/video/console/sticore.c.
6075 See comment in drivers/video/console/sticore.c.
6078 Format: bpp:<bpp1>[:<bpp2>[:<bpp3>...]]
6083 Enable or disable strict sigaltstack size checks
6084 against the required signal frame size which
6085 depends on the supported FPU features. This can
6086 be used to filter out binaries which have
6087 not yet been made aware of AT_MINSIGSTKSZ.
6090 Limits the number of kernel HPT entries in the hash
6091 page table to increase the rate of hash page table
6092 faults on kernel addresses.
6095 Limits the number of kernel SLB entries, and flushes
6096 them frequently to increase the rate of SLB faults
6097 on kernel addresses.
6099 sunrpc.min_resvport=
6100 sunrpc.max_resvport=
6102 SunRPC servers often require that client requests
6103 originate from a privileged port (i.e. a port in the
6104 range 0 < portnr < 1024).
6105 An administrator who wishes to reserve some of these
6106 ports for other uses may adjust the range that the
6107 kernel's sunrpc client considers to be privileged
6108 using these two parameters to set the minimum and
6109 maximum port values.
6111 sunrpc.svc_rpc_per_connection_limit=
6113 Limit the number of requests that the server will
6114 process in parallel from a single connection.
6115 The default value is 0 (no limit).
6119 Control how the NFS server code allocates CPUs to
6120 service thread pools. Depending on how many NICs
6121 you have and where their interrupts are bound, this
6122 option will affect which CPUs will do NFS serving.
6123 Note: this parameter cannot be changed while the
6124 NFS server is running.
6126 auto the server chooses an appropriate mode
6127 automatically using heuristics
6128 global a single global pool contains all CPUs
6129 percpu one pool for each CPU
6130 pernode one pool for each NUMA node (equivalent
6131 to global on non-NUMA machines)
6133 sunrpc.tcp_slot_table_entries=
6134 sunrpc.udp_slot_table_entries=
6136 Sets the upper limit on the number of simultaneous
6137 RPC calls that can be sent from the client to a
6138 server. Increasing these values may allow you to
6139 improve throughput, but will also increase the
6140 amount of memory reserved for use by the client.
6142 suspend.pm_test_delay=
6144 Sets the number of seconds to remain in a suspend test
6145 mode before resuming the system (see
6146 /sys/power/pm_test). Only available when CONFIG_PM_DEBUG
6147 is set. Default value is 5.
6150 Format: { on | off | y | n | 1 | 0 }
6151 This parameter controls use of the Protected
6152 Execution Facility on pSeries.
6154 swiotlb= [ARM,IA-64,PPC,MIPS,X86]
6155 Format: { <int> [,<int>] | force | noforce }
6156 <int> -- Number of I/O TLB slabs
6157 <int> -- Second integer after comma. Number of swiotlb
6158 areas with their own lock. Will be rounded up
6160 force -- force using of bounce buffers even if they
6161 wouldn't be automatically used by the kernel
6162 noforce -- Never use bounce buffers (for debugging)
6167 Set a sysctl parameter, right before loading the init
6168 process, as if the value was written to the respective
6169 /proc/sys/... file. Both '.' and '/' are recognized as
6170 separators. Unrecognized parameters and invalid values
6171 are reported in the kernel log. Sysctls registered
6172 later by a loaded module cannot be set this way.
6173 Example: sysctl.vm.swappiness=40
6175 sysrq_always_enabled
6177 Ignore sysrq setting - this boot parameter will
6178 neutralize any effect of /proc/sys/kernel/sysrq.
6179 Useful for debugging.
6181 tcpmhash_entries= [KNL,NET]
6182 Set the number of tcp_metrics_hash slots.
6183 Default value is 8192 or 16384 depending on total
6184 ram pages. This is used to specify the TCP metrics
6185 cache size. See Documentation/networking/ip-sysctl.rst
6186 "tcp_no_metrics_save" section for more details.
6190 test_suspend= [SUSPEND]
6191 Format: { "mem" | "standby" | "freeze" }[,N]
6192 Specify "mem" (for Suspend-to-RAM) or "standby" (for
6193 standby suspend) or "freeze" (for suspend type freeze)
6194 as the system sleep state during system startup with
6195 the optional capability to repeat N number of times.
6196 The system is woken from this state using a
6197 wakeup-capable RTC alarm.
6199 thash_entries= [KNL,NET]
6200 Set number of hash buckets for TCP connection
6202 thermal.act= [HW,ACPI]
6203 -1: disable all active trip points in all thermal zones
6204 <degrees C>: override all lowest active trip points
6206 thermal.crt= [HW,ACPI]
6207 -1: disable all critical trip points in all thermal zones
6208 <degrees C>: override all critical trip points
6210 thermal.nocrt= [HW,ACPI]
6211 Set to disable actions on ACPI thermal zone
6212 critical and hot trip points.
6214 thermal.off= [HW,ACPI]
6215 1: disable ACPI thermal control
6217 thermal.psv= [HW,ACPI]
6218 -1: disable all passive trip points
6219 <degrees C>: override all passive trip points to this
6222 thermal.tzp= [HW,ACPI]
6223 Specify global default ACPI thermal zone polling rate
6224 <deci-seconds>: poll all this frequency
6225 0: no polling (default)
6228 Force threading of all interrupt handlers except those
6229 marked explicitly IRQF_NO_THREAD.
6233 Specify if the kernel should make use of the cpu
6234 topology information if the hardware supports this.
6235 The scheduler will make use of this information and
6236 e.g. base its process migration decisions on it.
6239 topology_updates= [KNL, PPC, NUMA]
6241 Specify if the kernel should ignore (off)
6242 topology updates sent by the hypervisor to this
6245 torture.disable_onoff_at_boot= [KNL]
6246 Prevent the CPU-hotplug component of torturing
6247 until after init has spawned.
6249 torture.ftrace_dump_at_shutdown= [KNL]
6250 Dump the ftrace buffer at torture-test shutdown,
6251 even if there were no errors. This can be a
6252 very costly operation when many torture tests
6253 are running concurrently, especially on systems
6254 with rotating-rust storage.
6256 torture.verbose_sleep_frequency= [KNL]
6257 Specifies how many verbose printk()s should be
6258 emitted between each sleep. The default of zero
6259 disables verbose-printk() sleeping.
6261 torture.verbose_sleep_duration= [KNL]
6262 Duration of each verbose-printk() sleep in jiffies.
6266 tpm_suspend_pcr=[HW,TPM]
6267 Format: integer pcr id
6268 Specify that at suspend time, the tpm driver
6269 should extend the specified pcr with zeros,
6270 as a workaround for some chips which fail to
6271 flush the last written pcr on TPM_SaveState.
6272 This will guarantee that all the other pcrs
6276 Have the tracepoints sent to printk as well as the
6277 tracing ring buffer. This is useful for early boot up
6278 where the system hangs or reboots and does not give the
6279 option for reading the tracing buffer or performing a
6280 ftrace_dump_on_oops.
6282 To turn off having tracepoints sent to printk,
6283 echo 0 > /proc/sys/kernel/tracepoint_printk
6284 Note, echoing 1 into this file without the
6285 tracepoint_printk kernel cmdline option has no effect.
6287 The tp_printk_stop_on_boot (see below) can also be used
6288 to stop the printing of events to console at
6293 Having tracepoints sent to printk() and activating high
6294 frequency tracepoints such as irq or sched, can cause
6295 the system to live lock.
6297 tp_printk_stop_on_boot [FTRACE]
6298 When tp_printk (above) is set, it can cause a lot of noise
6299 on the console. It may be useful to only include the
6300 printing of events during boot up, as user space may
6301 make the system inoperable.
6303 This command line option will stop the printing of events
6304 to console at the late_initcall_sync() time frame.
6306 trace_buf_size=nn[KMG]
6307 [FTRACE] will set tracing buffer size on each cpu.
6309 trace_clock= [FTRACE] Set the clock used for tracing events
6311 local - Use the per CPU time stamp counter
6312 (converted into nanoseconds). Fast, but
6313 depending on the architecture, may not be
6314 in sync between CPUs.
6315 global - Event time stamps are synchronize across
6316 CPUs. May be slower than the local clock,
6317 but better for some race conditions.
6318 counter - Simple counting of events (1, 2, ..)
6319 note, some counts may be skipped due to the
6320 infrastructure grabbing the clock more than
6322 uptime - Use jiffies as the time stamp.
6323 perf - Use the same clock that perf uses.
6324 mono - Use ktime_get_mono_fast_ns() for time stamps.
6325 mono_raw - Use ktime_get_raw_fast_ns() for time
6327 boot - Use ktime_get_boot_fast_ns() for time stamps.
6328 Architectures may add more clocks. See
6329 Documentation/trace/ftrace.rst for more details.
6331 trace_event=[event-list]
6332 [FTRACE] Set and start specified trace events in order
6333 to facilitate early boot debugging. The event-list is a
6334 comma-separated list of trace events to enable. See
6335 also Documentation/trace/events.rst
6337 trace_instance=[instance-info]
6338 [FTRACE] Create a ring buffer instance early in boot up.
6339 This will be listed in:
6341 /sys/kernel/tracing/instances
6343 Events can be enabled at the time the instance is created
6346 trace_instance=<name>,<system1>:<event1>,<system2>:<event2>
6348 Note, the "<system*>:" portion is optional if the event is
6351 trace_instance=foo,sched:sched_switch,irq_handler_entry,initcall
6353 will enable the "sched_switch" event (note, the "sched:" is optional, and
6354 the same thing would happen if it was left off). The irq_handler_entry
6355 event, and all events under the "initcall" system.
6357 trace_options=[option-list]
6358 [FTRACE] Enable or disable tracer options at boot.
6359 The option-list is a comma delimited list of options
6360 that can be enabled or disabled just as if you were
6361 to echo the option name into
6363 /sys/kernel/tracing/trace_options
6365 For example, to enable stacktrace option (to dump the
6366 stack trace of each event), add to the command line:
6368 trace_options=stacktrace
6370 See also Documentation/trace/ftrace.rst "trace options"
6373 trace_trigger=[trigger-list]
6374 [FTRACE] Add a event trigger on specific events.
6375 Set a trigger on top of a specific event, with an optional
6378 The format is is "trace_trigger=<event>.<trigger>[ if <filter>],..."
6379 Where more than one trigger may be specified that are comma deliminated.
6383 trace_trigger="sched_switch.stacktrace if prev_state == 2"
6385 The above will enable the "stacktrace" trigger on the "sched_switch"
6386 event but only trigger it if the "prev_state" of the "sched_switch"
6387 event is "2" (TASK_UNINTERUPTIBLE).
6389 See also "Event triggers" in Documentation/trace/events.rst
6393 [FTRACE] enable this option to disable tracing when a
6394 warning is hit. This turns off "tracing_on". Tracing can
6395 be enabled again by echoing '1' into the "tracing_on"
6396 file located in /sys/kernel/tracing/
6398 This option is useful, as it disables the trace before
6399 the WARNING dump is called, which prevents the trace to
6400 be filled with content caused by the warning output.
6402 This option can also be set at run time via the sysctl
6403 option: kernel/traceoff_on_warning
6405 transparent_hugepage=
6407 Format: [always|madvise|never]
6408 Can be used to control the default behavior of the system
6409 with respect to transparent hugepages.
6410 See Documentation/admin-guide/mm/transhuge.rst
6413 trusted.source= [KEYS]
6415 This parameter identifies the trust source as a backend
6416 for trusted keys implementation. Supported trust
6421 If not specified then it defaults to iterating through
6422 the trust source list starting with TPM and assigns the
6423 first trust source as a backend which is initialized
6424 successfully during iteration.
6428 The RNG used to generate key material for trusted keys.
6431 - the same value as trusted.source: "tpm" or "tee"
6433 If not specified, "default" is used. In this case,
6434 the RNG's choice is left to each individual trust source.
6436 tsc= Disable clocksource stability checks for TSC.
6438 [x86] reliable: mark tsc clocksource as reliable, this
6439 disables clocksource verification at runtime, as well
6440 as the stability checks done at bootup. Used to enable
6441 high-resolution timer mode on older hardware, and in
6442 virtualized environment.
6443 [x86] noirqtime: Do not use TSC to do irq accounting.
6444 Used to run time disable IRQ_TIME_ACCOUNTING on any
6445 platforms where RDTSC is slow and this accounting
6447 [x86] unstable: mark the TSC clocksource as unstable, this
6448 marks the TSC unconditionally unstable at bootup and
6449 avoids any further wobbles once the TSC watchdog notices.
6450 [x86] nowatchdog: disable clocksource watchdog. Used
6451 in situations with strict latency requirements (where
6452 interruptions from clocksource watchdog are not
6454 [x86] recalibrate: force recalibration against a HW timer
6455 (HPET or PM timer) on systems whose TSC frequency was
6456 obtained from HW or FW using either an MSR or CPUID(0x15).
6457 Warn if the difference is more than 500 ppm.
6458 [x86] watchdog: Use TSC as the watchdog clocksource with
6459 which to check other HW timers (HPET or PM timer), but
6460 only on systems where TSC has been deemed trustworthy.
6461 This will be suppressed by an earlier tsc=nowatchdog and
6462 can be overridden by a later tsc=nowatchdog. A console
6463 message will flag any such suppression or overriding.
6465 tsc_early_khz= [X86] Skip early TSC calibration and use the given
6466 value instead. Useful when the early TSC frequency discovery
6467 procedure is not reliable, such as on overclocked systems
6468 with CPUID.16h support and partial CPUID.15h support.
6469 Format: <unsigned int>
6471 tsx= [X86] Control Transactional Synchronization
6472 Extensions (TSX) feature in Intel processors that
6473 support TSX control.
6475 This parameter controls the TSX feature. The options are:
6477 on - Enable TSX on the system. Although there are
6478 mitigations for all known security vulnerabilities,
6479 TSX has been known to be an accelerator for
6480 several previous speculation-related CVEs, and
6481 so there may be unknown security risks associated
6482 with leaving it enabled.
6484 off - Disable TSX on the system. (Note that this
6485 option takes effect only on newer CPUs which are
6486 not vulnerable to MDS, i.e., have
6487 MSR_IA32_ARCH_CAPABILITIES.MDS_NO=1 and which get
6488 the new IA32_TSX_CTRL MSR through a microcode
6489 update. This new MSR allows for the reliable
6490 deactivation of the TSX functionality.)
6492 auto - Disable TSX if X86_BUG_TAA is present,
6493 otherwise enable TSX on the system.
6495 Not specifying this option is equivalent to tsx=off.
6497 See Documentation/admin-guide/hw-vuln/tsx_async_abort.rst
6500 tsx_async_abort= [X86,INTEL] Control mitigation for the TSX Async
6501 Abort (TAA) vulnerability.
6503 Similar to Micro-architectural Data Sampling (MDS)
6504 certain CPUs that support Transactional
6505 Synchronization Extensions (TSX) are vulnerable to an
6506 exploit against CPU internal buffers which can forward
6507 information to a disclosure gadget under certain
6510 In vulnerable processors, the speculatively forwarded
6511 data can be used in a cache side channel attack, to
6512 access data to which the attacker does not have direct
6515 This parameter controls the TAA mitigation. The
6518 full - Enable TAA mitigation on vulnerable CPUs
6521 full,nosmt - Enable TAA mitigation and disable SMT on
6522 vulnerable CPUs. If TSX is disabled, SMT
6523 is not disabled because CPU is not
6524 vulnerable to cross-thread TAA attacks.
6525 off - Unconditionally disable TAA mitigation
6527 On MDS-affected machines, tsx_async_abort=off can be
6528 prevented by an active MDS mitigation as both vulnerabilities
6529 are mitigated with the same mechanism so in order to disable
6530 this mitigation, you need to specify mds=off too.
6532 Not specifying this option is equivalent to
6533 tsx_async_abort=full. On CPUs which are MDS affected
6534 and deploy MDS mitigation, TAA mitigation is not
6535 required and doesn't provide any additional
6539 Documentation/admin-guide/hw-vuln/tsx_async_abort.rst
6541 turbografx.map[2|3]= [HW,JOY]
6542 TurboGraFX parallel port interface
6544 <port#>,<js1>,<js2>,<js3>,<js4>,<js5>,<js6>,<js7>
6545 See also Documentation/input/devices/joystick-parport.rst
6547 udbg-immortal [PPC] When debugging early kernel crashes that
6548 happen after console_init() and before a proper
6549 console driver takes over, this boot options might
6550 help "seeing" what's going on.
6552 uhash_entries= [KNL,NET]
6553 Set number of hash buckets for UDP/UDP-Lite connections
6556 [USB] Ignore overcurrent events (default N).
6557 Some badly-designed motherboards generate lots of
6558 bogus events, for ports that aren't wired to
6559 anything. Set this parameter to avoid log spamming.
6560 Note that genuine overcurrent events won't be
6564 [X86] Cause panic on unknown NMI.
6566 usbcore.authorized_default=
6567 [USB] Default USB device authorization:
6568 (default -1 = authorized except for wireless USB,
6569 0 = not authorized, 1 = authorized, 2 = authorized
6570 if device connected to internal port)
6572 usbcore.autosuspend=
6573 [USB] The autosuspend time delay (in seconds) used
6574 for newly-detected USB devices (default 2). This
6575 is the time required before an idle device will be
6576 autosuspended. Devices for which the delay is set
6577 to a negative value won't be autosuspended at all.
6579 usbcore.usbfs_snoop=
6580 [USB] Set to log all usbfs traffic (default 0 = off).
6582 usbcore.usbfs_snoop_max=
6583 [USB] Maximum number of bytes to snoop in each URB
6586 usbcore.blinkenlights=
6587 [USB] Set to cycle leds on hubs (default 0 = off).
6589 usbcore.old_scheme_first=
6590 [USB] Start with the old device initialization
6591 scheme (default 0 = off).
6593 usbcore.usbfs_memory_mb=
6594 [USB] Memory limit (in MB) for buffers allocated by
6595 usbfs (default = 16, 0 = max = 2047).
6597 usbcore.use_both_schemes=
6598 [USB] Try the other device initialization scheme
6599 if the first one fails (default 1 = enabled).
6601 usbcore.initial_descriptor_timeout=
6602 [USB] Specifies timeout for the initial 64-byte
6603 USB_REQ_GET_DESCRIPTOR request in milliseconds
6604 (default 5000 = 5.0 seconds).
6606 usbcore.nousb [USB] Disable the USB subsystem
6609 [USB] A list of quirk entries to augment the built-in
6610 usb core quirk list. List entries are separated by
6611 commas. Each entry has the form
6612 VendorID:ProductID:Flags. The IDs are 4-digit hex
6613 numbers and Flags is a set of letters. Each letter
6614 will change the built-in quirk; setting it if it is
6615 clear and clearing it if it is set. The letters have
6616 the following meanings:
6617 a = USB_QUIRK_STRING_FETCH_255 (string
6618 descriptors must not be fetched using
6620 b = USB_QUIRK_RESET_RESUME (device can't resume
6621 correctly so reset it instead);
6622 c = USB_QUIRK_NO_SET_INTF (device can't handle
6623 Set-Interface requests);
6624 d = USB_QUIRK_CONFIG_INTF_STRINGS (device can't
6625 handle its Configuration or Interface
6627 e = USB_QUIRK_RESET (device can't be reset
6628 (e.g morph devices), don't use reset);
6629 f = USB_QUIRK_HONOR_BNUMINTERFACES (device has
6630 more interface descriptions than the
6631 bNumInterfaces count, and can't handle
6632 talking to these interfaces);
6633 g = USB_QUIRK_DELAY_INIT (device needs a pause
6634 during initialization, after we read
6635 the device descriptor);
6636 h = USB_QUIRK_LINEAR_UFRAME_INTR_BINTERVAL (For
6637 high speed and super speed interrupt
6638 endpoints, the USB 2.0 and USB 3.0 spec
6639 require the interval in microframes (1
6640 microframe = 125 microseconds) to be
6641 calculated as interval = 2 ^
6643 Devices with this quirk report their
6644 bInterval as the result of this
6645 calculation instead of the exponent
6646 variable used in the calculation);
6647 i = USB_QUIRK_DEVICE_QUALIFIER (device can't
6648 handle device_qualifier descriptor
6650 j = USB_QUIRK_IGNORE_REMOTE_WAKEUP (device
6651 generates spurious wakeup, ignore
6652 remote wakeup capability);
6653 k = USB_QUIRK_NO_LPM (device can't handle Link
6655 l = USB_QUIRK_LINEAR_FRAME_INTR_BINTERVAL
6656 (Device reports its bInterval as linear
6657 frames instead of the USB 2.0
6659 m = USB_QUIRK_DISCONNECT_SUSPEND (Device needs
6660 to be disconnected before suspend to
6661 prevent spurious wakeup);
6662 n = USB_QUIRK_DELAY_CTRL_MSG (Device needs a
6663 pause after every control message);
6664 o = USB_QUIRK_HUB_SLOW_RESET (Hub needs extra
6665 delay after resetting its port);
6666 Example: quirks=0781:5580:bk,0a5c:5834:gij
6669 [USBHID] The interval which mice are to be polled at.
6672 [USBHID] The interval which joysticks are to be polled at.
6675 [USBHID] The interval which keyboards are to be polled at.
6677 usb-storage.delay_use=
6678 [UMS] The delay in seconds before a new device is
6679 scanned for Logical Units (default 1).
6682 [UMS] A list of quirks entries to supplement or
6683 override the built-in unusual_devs list. List
6684 entries are separated by commas. Each entry has
6685 the form VID:PID:Flags where VID and PID are Vendor
6686 and Product ID values (4-digit hex numbers) and
6687 Flags is a set of characters, each corresponding
6688 to a common usb-storage quirk flag as follows:
6689 a = SANE_SENSE (collect more than 18 bytes
6690 of sense data, not on uas);
6691 b = BAD_SENSE (don't collect more than 18
6692 bytes of sense data, not on uas);
6693 c = FIX_CAPACITY (decrease the reported
6694 device capacity by one sector);
6695 d = NO_READ_DISC_INFO (don't use
6696 READ_DISC_INFO command, not on uas);
6697 e = NO_READ_CAPACITY_16 (don't use
6698 READ_CAPACITY_16 command);
6699 f = NO_REPORT_OPCODES (don't use report opcodes
6701 g = MAX_SECTORS_240 (don't transfer more than
6702 240 sectors at a time, uas only);
6703 h = CAPACITY_HEURISTICS (decrease the
6704 reported device capacity by one
6705 sector if the number is odd);
6706 i = IGNORE_DEVICE (don't bind to this
6708 j = NO_REPORT_LUNS (don't use report luns
6710 k = NO_SAME (do not use WRITE_SAME, uas only)
6711 l = NOT_LOCKABLE (don't try to lock and
6712 unlock ejectable media, not on uas);
6713 m = MAX_SECTORS_64 (don't transfer more
6714 than 64 sectors = 32 KB at a time,
6716 n = INITIAL_READ10 (force a retry of the
6717 initial READ(10) command, not on uas);
6718 o = CAPACITY_OK (accept the capacity
6719 reported by the device, not on uas);
6720 p = WRITE_CACHE (the device cache is ON
6721 by default, not on uas);
6722 r = IGNORE_RESIDUE (the device reports
6723 bogus residue values, not on uas);
6724 s = SINGLE_LUN (the device has only one
6726 t = NO_ATA_1X (don't allow ATA(12) and ATA(16)
6727 commands, uas only);
6728 u = IGNORE_UAS (don't bind to the uas driver);
6729 w = NO_WP_DETECT (don't test whether the
6730 medium is write-protected).
6731 y = ALWAYS_SYNC (issue a SYNCHRONIZE_CACHE
6732 even if the device claims no cache,
6734 Example: quirks=0419:aaf5:rl,0421:0433:rc
6736 user_debug= [KNL,ARM]
6738 See arch/arm/Kconfig.debug help text.
6739 1 - undefined instruction events
6741 4 - invalid data aborts
6744 Example: user_debug=31
6747 [X86] Flags controlling user PTE allocations.
6749 nohigh = do not allocate PTE pages in
6750 HIGHMEM regardless of setting
6753 vdso= [X86,SH,SPARC]
6754 On X86_32, this is an alias for vdso32=. Otherwise:
6756 vdso=1: enable VDSO (the default)
6757 vdso=0: disable VDSO mapping
6759 vdso32= [X86] Control the 32-bit vDSO
6760 vdso32=1: enable 32-bit VDSO
6761 vdso32=0 or vdso32=2: disable 32-bit VDSO
6763 See the help text for CONFIG_COMPAT_VDSO for more
6764 details. If CONFIG_COMPAT_VDSO is set, the default is
6765 vdso32=0; otherwise, the default is vdso32=1.
6767 For compatibility with older kernels, vdso32=2 is an
6770 Try vdso32=0 if you encounter an error that says:
6771 dl_main: Assertion `(void *) ph->p_vaddr == _rtld_local._dl_sysinfo_dso' failed!
6774 vector=percpu: enable percpu vector domain
6776 video= [FB] Frame buffer configuration
6777 See Documentation/fb/modedb.rst.
6779 video.brightness_switch_enabled= [ACPI]
6781 If set to 1, on receiving an ACPI notify event
6782 generated by hotkey, video driver will adjust brightness
6783 level and then send out the event to user space through
6784 the allocated input device. If set to 0, video driver
6785 will only send out the event without touching backlight
6790 [VMMIO] Memory mapped virtio (platform) device.
6792 <size>@<baseaddr>:<irq>[:<id>]
6794 <size> := size (can use standard suffixes
6796 <baseaddr> := physical base address
6797 <irq> := interrupt number (as passed to
6799 <id> := (optional) platform device id
6801 virtio_mmio.device=1K@0x100b0000:48:7
6803 Can be used multiple times for multiple devices.
6805 vga= [BOOT,X86-32] Select a particular video mode
6806 See Documentation/arch/x86/boot.rst and
6807 Documentation/admin-guide/svga.rst.
6808 Use vga=ask for menu.
6809 This is actually a boot loader parameter; the value is
6810 passed to the kernel using a special protocol.
6812 vm_debug[=options] [KNL] Available with CONFIG_DEBUG_VM=y.
6813 May slow down system boot speed, especially when
6814 enabled on systems with a large amount of memory.
6815 All options are enabled by default, and this
6816 interface is meant to allow for selectively
6817 enabling or disabling specific virtual memory
6820 Available options are:
6821 P Enable page structure init time poisoning
6822 - Disable all of the above options
6824 vmalloc=nn[KMG] [KNL,BOOT] Forces the vmalloc area to have an exact
6825 size of <nn>. This can be used to increase the
6826 minimum size (128MB on x86). It can also be used to
6827 decrease the size and leave more room for directly
6830 vmcp_cma=nn[MG] [KNL,S390]
6831 Sets the memory size reserved for contiguous memory
6832 allocations for the vmcp device driver.
6834 vmhalt= [KNL,S390] Perform z/VM CP command after system halt.
6837 vmpanic= [KNL,S390] Perform z/VM CP command after kernel panic.
6840 vmpoff= [KNL,S390] Perform z/VM CP command after power off.
6844 Controls the behavior of vsyscalls (i.e. calls to
6845 fixed addresses of 0xffffffffff600x00 from legacy
6846 code). Most statically-linked binaries and older
6847 versions of glibc use these calls. Because these
6848 functions are at fixed addresses, they make nice
6849 targets for exploits that can control RIP.
6851 emulate Vsyscalls turn into traps and are emulated
6852 reasonably safely. The vsyscall page is
6855 xonly [default] Vsyscalls turn into traps and are
6856 emulated reasonably safely. The vsyscall
6857 page is not readable.
6859 none Vsyscalls don't work at all. This makes
6860 them quite hard to use for exploits but
6861 might break your system.
6863 vt.color= [VT] Default text color.
6864 Format: 0xYX, X = foreground, Y = background.
6865 Default: 0x07 = light gray on black.
6867 vt.cur_default= [VT] Default cursor shape.
6868 Format: 0xCCBBAA, where AA, BB, and CC are the same as
6869 the parameters of the <Esc>[?A;B;Cc escape sequence;
6870 see VGA-softcursor.txt. Default: 2 = underline.
6872 vt.default_blu= [VT]
6873 Format: <blue0>,<blue1>,<blue2>,...,<blue15>
6874 Change the default blue palette of the console.
6875 This is a 16-member array composed of values
6878 vt.default_grn= [VT]
6879 Format: <green0>,<green1>,<green2>,...,<green15>
6880 Change the default green palette of the console.
6881 This is a 16-member array composed of values
6884 vt.default_red= [VT]
6885 Format: <red0>,<red1>,<red2>,...,<red15>
6886 Change the default red palette of the console.
6887 This is a 16-member array composed of values
6893 Set system-wide default UTF-8 mode for all tty's.
6894 Default is 1, i.e. UTF-8 mode is enabled for all
6895 newly opened terminals.
6897 vt.global_cursor_default=
6900 Set system-wide default for whether a cursor
6901 is shown on new VTs. Default is -1,
6902 i.e. cursors will be created by default unless
6903 overridden by individual drivers. 0 will hide
6904 cursors, 1 will display them.
6906 vt.italic= [VT] Default color for italic text; 0-15.
6909 vt.underline= [VT] Default color for underlined text; 0-15.
6912 watchdog timers [HW,WDT] For information on watchdog timers,
6913 see Documentation/watchdog/watchdog-parameters.rst
6914 or other driver-specific files in the
6915 Documentation/watchdog/ directory.
6919 Set the hard lockup detector stall duration
6920 threshold in seconds. The soft lockup detector
6921 threshold is set to twice the value. A value of 0
6922 disables both lockup detectors. Default is 10
6925 workqueue.watchdog_thresh=
6926 If CONFIG_WQ_WATCHDOG is configured, workqueue can
6927 warn stall conditions and dump internal state to
6928 help debugging. 0 disables workqueue stall
6929 detection; otherwise, it's the stall threshold
6930 duration in seconds. The default value is 30 and
6931 it can be updated at runtime by writing to the
6932 corresponding sysfs file.
6934 workqueue.disable_numa
6935 By default, all work items queued to unbound
6936 workqueues are affine to the NUMA nodes they're
6937 issued on, which results in better behavior in
6938 general. If NUMA affinity needs to be disabled for
6939 whatever reason, this option can be used. Note
6940 that this also can be controlled per-workqueue for
6941 workqueues visible under /sys/bus/workqueue/.
6943 workqueue.power_efficient
6944 Per-cpu workqueues are generally preferred because
6945 they show better performance thanks to cache
6946 locality; unfortunately, per-cpu workqueues tend to
6947 be more power hungry than unbound workqueues.
6949 Enabling this makes the per-cpu workqueues which
6950 were observed to contribute significantly to power
6951 consumption unbound, leading to measurably lower
6952 power usage at the cost of small performance
6955 The default value of this parameter is determined by
6956 the config option CONFIG_WQ_POWER_EFFICIENT_DEFAULT.
6958 workqueue.debug_force_rr_cpu
6959 Workqueue used to implicitly guarantee that work
6960 items queued without explicit CPU specified are put
6961 on the local CPU. This guarantee is no longer true
6962 and while local CPU is still preferred work items
6963 may be put on foreign CPUs. This debug option
6964 forces round-robin CPU selection to flush out
6965 usages which depend on the now broken guarantee.
6966 When enabled, memory and cache locality will be
6969 writecombine= [LOONGARCH] Control the MAT (Memory Access Type) of
6972 on - Enable writecombine, use WUC for ioremap_wc()
6973 off - Disable writecombine, use SUC for ioremap_wc()
6975 x2apic_phys [X86-64,APIC] Use x2apic physical mode instead of
6976 default x2apic cluster mode on platforms
6979 xen_512gb_limit [KNL,X86-64,XEN]
6980 Restricts the kernel running paravirtualized under Xen
6981 to use only up to 512 GB of RAM. The reason to do so is
6982 crash analysis tools and Xen tools for doing domain
6983 save/restore/migration must be enabled to handle larger
6986 xen_emul_unplug= [HW,X86,XEN]
6987 Unplug Xen emulated devices
6988 Format: [unplug0,][unplug1]
6989 ide-disks -- unplug primary master IDE devices
6990 aux-ide-disks -- unplug non-primary-master IDE devices
6991 nics -- unplug network devices
6992 all -- unplug all emulated devices (NICs and IDE disks)
6993 unnecessary -- unplugging emulated devices is
6994 unnecessary even if the host did not respond to
6996 never -- do not unplug even if version check succeeds
6998 xen_legacy_crash [X86,XEN]
6999 Crash from Xen panic notifier, without executing late
7000 panic() code such as dumping handler.
7002 xen_msr_safe= [X86,XEN]
7004 Select whether to always use non-faulting (safe) MSR
7005 access functions when running as Xen PV guest. The
7006 default value is controlled by CONFIG_XEN_PV_MSR_SAFE.
7008 xen_nopvspin [X86,XEN]
7009 Disables the qspinlock slowpath using Xen PV optimizations.
7010 This parameter is obsoleted by "nopvspin" parameter, which
7011 has equivalent effect for XEN platform.
7014 Disables the PV optimizations forcing the HVM guest to
7015 run as generic HVM guest with no PV drivers.
7016 This option is obsoleted by the "nopv" option, which
7017 has equivalent effect for XEN platform.
7019 xen_no_vector_callback
7020 [KNL,X86,XEN] Disable the vector callback for Xen
7021 event channel interrupts.
7023 xen_scrub_pages= [XEN]
7024 Boolean option to control scrubbing pages before giving them back
7025 to Xen, for use by other domains. Can be also changed at runtime
7026 with /sys/devices/system/xen_memory/xen_memory0/scrub_pages.
7027 Default value controlled with CONFIG_XEN_SCRUB_PAGES_DEFAULT.
7029 xen_timer_slop= [X86-64,XEN]
7030 Set the timer slop (in nanoseconds) for the virtual Xen
7031 timers (default is 100000). This adjusts the minimum
7032 delta of virtualized Xen timers, where lower values
7033 improve timer resolution at the expense of processing
7034 more timer interrupts.
7036 xen.balloon_boot_timeout= [XEN]
7037 The time (in seconds) to wait before giving up to boot
7038 in case initial ballooning fails to free enough memory.
7039 Applies only when running as HVM or PVH guest and
7040 started with less memory configured than allowed at
7041 max. Default is 180.
7043 xen.event_eoi_delay= [XEN]
7044 How long to delay EOI handling in case of event
7045 storms (jiffies). Default is 10.
7047 xen.event_loop_timeout= [XEN]
7048 After which time (jiffies) the event handling loop
7049 should start to delay EOI handling. Default is 2.
7051 xen.fifo_events= [XEN]
7052 Boolean parameter to disable using fifo event handling
7053 even if available. Normally fifo event handling is
7054 preferred over the 2-level event handling, as it is
7055 fairer and the number of possible event channels is
7056 much higher. Default is on (use fifo events).
7058 xirc2ps_cs= [NET,PCMCIA]
7060 <irq>,<irq_mask>,<io>,<full_duplex>,<do_sound>,<lockup_hack>[,<irq2>[,<irq3>[,<irq4>]]]
7063 By default on POWER9 and above, the kernel will
7064 natively use the XIVE interrupt controller. This option
7065 allows the fallback firmware mode to be used:
7067 off Fallback to firmware control of XIVE interrupt
7068 controller on both pseries and powernv
7069 platforms. Only useful on POWER9 and above.
7071 xive.store-eoi=off [PPC]
7072 By default on POWER10 and above, the kernel will use
7073 stores for EOI handling when the XIVE interrupt mode
7074 is active. This option allows the XIVE driver to use
7075 loads instead, as on POWER9.
7077 xhci-hcd.quirks [USB,KNL]
7078 A hex value specifying bitmask with supplemental xhci
7079 host controller quirks. Meaning of each bit can be
7080 consulted in header drivers/usb/host/xhci.h.
7083 Format: { early | on | rw | ro | off }
7084 Controls if xmon debugger is enabled. Default is off.
7085 Passing only "xmon" is equivalent to "xmon=early".
7086 early Call xmon as early as possible on boot; xmon
7087 debugger is called from setup_arch().
7088 on xmon debugger hooks will be installed so xmon
7089 is only called on a kernel crash. Default mode,
7090 i.e. either "ro" or "rw" mode, is controlled
7091 with CONFIG_XMON_DEFAULT_RO_MODE.
7092 rw xmon debugger hooks will be installed so xmon
7093 is called only on a kernel crash, mode is write,
7094 meaning SPR registers, memory and, other data
7095 can be written using xmon commands.
7096 ro same as "rw" option above but SPR registers,
7097 memory, and other data can't be written using
7099 off xmon is disabled.
projects / linux-2.6-microblaze.git / blob