2 Format: { eager | lazy }
4 By default, unaccepted memory is accepted lazily to
5 avoid prolonged boot times. The lazy option will add
6 some runtime overhead until all memory is eventually
7 accepted. In most cases the overhead is negligible.
8 For some workloads or for debugging purposes
9 accept_memory=eager can be used to accept all memory
12 acpi= [HW,ACPI,X86,ARM64,RISCV64]
13 Advanced Configuration and Power Interface
14 Format: { force | on | off | strict | noirq | rsdt |
16 force -- enable ACPI if default was off
17 on -- enable ACPI but allow fallback to DT [arm64,riscv64]
18 off -- disable ACPI if default was on
19 noirq -- do not use ACPI for IRQ routing
20 strict -- Be less tolerant of platforms that are not
21 strictly ACPI specification compliant.
22 rsdt -- prefer RSDT over (default) XSDT
23 copy_dsdt -- copy DSDT to memory
24 For ARM64 and RISCV64, ONLY "acpi=off", "acpi=on" or
25 "acpi=force" are available
27 See also Documentation/power/runtime_pm.rst, pci=noacpi
29 acpi_apic_instance= [ACPI, IOAPIC]
31 2: use 2nd APIC table, if available
32 1,0: use 1st APIC table
35 acpi_backlight= [HW,ACPI]
36 { vendor | video | native | none }
37 If set to vendor, prefer vendor-specific driver
38 (e.g. thinkpad_acpi, sony_acpi, etc.) instead
39 of the ACPI video.ko driver.
40 If set to video, use the ACPI video.ko driver.
41 If set to native, use the device's native backlight mode.
42 If set to none, disable the ACPI backlight interface.
44 acpi_force_32bit_fadt_addr
45 force FADT to use 32 bit addresses rather than the
46 64 bit X_* addresses. Some firmware have broken 64
47 bit addresses for force ACPI ignore these and use
48 the older legacy 32 bit addresses.
50 acpica_no_return_repair [HW, ACPI]
51 Disable AML predefined validation mechanism
52 This mechanism can repair the evaluation result to make
53 the return objects more ACPI specification compliant.
54 This option is useful for developers to identify the
55 root cause of an AML interpreter issue when the issue
56 has something to do with the repair mechanism.
58 acpi.debug_layer= [HW,ACPI,ACPI_DEBUG]
59 acpi.debug_level= [HW,ACPI,ACPI_DEBUG]
61 CONFIG_ACPI_DEBUG must be enabled to produce any ACPI
62 debug output. Bits in debug_layer correspond to a
63 _COMPONENT in an ACPI source file, e.g.,
64 #define _COMPONENT ACPI_EVENTS
65 Bits in debug_level correspond to a level in
66 ACPI_DEBUG_PRINT statements, e.g.,
67 ACPI_DEBUG_PRINT((ACPI_DB_INFO, ...
68 The debug_level mask defaults to "info". See
69 Documentation/firmware-guide/acpi/debug.rst for more information about
70 debug layers and levels.
72 Enable processor driver info messages:
73 acpi.debug_layer=0x20000000
74 Enable AML "Debug" output, i.e., stores to the Debug
75 object while interpreting AML:
76 acpi.debug_layer=0xffffffff acpi.debug_level=0x2
77 Enable all messages related to ACPI hardware:
78 acpi.debug_layer=0x2 acpi.debug_level=0xffffffff
80 Some values produce so much output that the system is
81 unusable. The "log_buf_len" parameter may be useful
82 if you need to capture more output.
84 acpi_enforce_resources= [ACPI]
86 Check for resource conflicts between native drivers
87 and ACPI OperationRegions (SystemIO and SystemMemory
88 only). IO ports and memory declared in ACPI might be
89 used by the ACPI subsystem in arbitrary AML code and
90 can interfere with legacy drivers.
91 strict (default): access to resources claimed by ACPI
92 is denied; legacy drivers trying to access reserved
93 resources will fail to bind to device using them.
94 lax: access to resources claimed by ACPI is allowed;
95 legacy drivers trying to access reserved resources
96 will bind successfully but a warning message is logged.
97 no: ACPI OperationRegions are not marked as reserved,
98 no further checks are performed.
100 acpi_force_table_verification [HW,ACPI]
101 Enable table checksum verification during early stage.
102 By default, this is disabled due to x86 early mapping
105 acpi_irq_balance [HW,ACPI]
106 ACPI will balance active IRQs
109 acpi_irq_nobalance [HW,ACPI]
110 ACPI will not move active IRQs (default)
113 acpi_irq_isa= [HW,ACPI] If irq_balance, mark listed IRQs used by ISA
114 Format: <irq>,<irq>...
116 acpi_irq_pci= [HW,ACPI] If irq_balance, clear listed IRQs for
118 Format: <irq>,<irq>...
120 acpi_mask_gpe= [HW,ACPI]
121 Due to the existence of _Lxx/_Exx, some GPEs triggered
122 by unsupported hardware/firmware features can result in
123 GPE floodings that cannot be automatically disabled by
125 This facility can be used to prevent such uncontrolled
127 Format: <byte> or <bitmap-list>
129 acpi_no_auto_serialize [HW,ACPI]
130 Disable auto-serialization of AML methods
131 AML control methods that contain the opcodes to create
132 named objects will be marked as "Serialized" by the
133 auto-serialization feature.
134 This feature is enabled by default.
135 This option allows to turn off the feature.
137 acpi_no_memhotplug [ACPI] Disable memory hotplug. Useful for kdump
140 acpi_no_static_ssdt [HW,ACPI]
141 Disable installation of static SSDTs at early boot time
142 By default, SSDTs contained in the RSDT/XSDT will be
143 installed automatically and they will appear under
144 /sys/firmware/acpi/tables.
145 This option turns off this feature.
146 Note that specifying this option does not affect
147 dynamic table installation which will install SSDT
148 tables to /sys/firmware/acpi/tables/dynamic.
150 acpi_no_watchdog [HW,ACPI,WDT]
151 Ignore the ACPI-based watchdog interface (WDAT) and let
152 a native driver control the watchdog device instead.
154 acpi_rsdp= [ACPI,EFI,KEXEC]
155 Pass the RSDP address to the kernel, mostly used
156 on machines running EFI runtime service to boot the
157 second kernel for kdump.
159 acpi_os_name= [HW,ACPI] Tell ACPI BIOS the name of the OS
160 Format: To spoof as Windows 98: ="Microsoft Windows"
162 acpi_rev_override [ACPI] Override the _REV object to return 5 (instead
163 of 2 which is mandated by ACPI 6) as the supported ACPI
164 specification revision (when using this switch, it may
165 be necessary to carry out a cold reboot _twice_ in a
166 row to make it take effect on the platform firmware).
168 acpi_osi= [HW,ACPI] Modify list of supported OS interface strings
169 acpi_osi="string1" # add string1
170 acpi_osi="!string2" # remove string2
171 acpi_osi=!* # remove all strings
172 acpi_osi=! # disable all built-in OS vendor
174 acpi_osi=!! # enable all built-in OS vendor
176 acpi_osi= # disable all strings
178 'acpi_osi=!' can be used in combination with single or
179 multiple 'acpi_osi="string1"' to support specific OS
180 vendor string(s). Note that such command can only
181 affect the default state of the OS vendor strings, thus
182 it cannot affect the default state of the feature group
183 strings and the current state of the OS vendor strings,
184 specifying it multiple times through kernel command line
185 is meaningless. This command is useful when one do not
186 care about the state of the feature group strings which
187 should be controlled by the OSPM.
189 1. 'acpi_osi=! acpi_osi="Windows 2000"' is equivalent
190 to 'acpi_osi="Windows 2000" acpi_osi=!', they all
191 can make '_OSI("Windows 2000")' TRUE.
193 'acpi_osi=' cannot be used in combination with other
194 'acpi_osi=' command lines, the _OSI method will not
195 exist in the ACPI namespace. NOTE that such command can
196 only affect the _OSI support state, thus specifying it
197 multiple times through kernel command line is also
200 1. 'acpi_osi=' can make 'CondRefOf(_OSI, Local1)'
203 'acpi_osi=!*' can be used in combination with single or
204 multiple 'acpi_osi="string1"' to support specific
205 string(s). Note that such command can affect the
206 current state of both the OS vendor strings and the
207 feature group strings, thus specifying it multiple times
208 through kernel command line is meaningful. But it may
209 still not able to affect the final state of a string if
210 there are quirks related to this string. This command
211 is useful when one want to control the state of the
212 feature group strings to debug BIOS issues related to
215 1. 'acpi_osi="Module Device" acpi_osi=!*' can make
216 '_OSI("Module Device")' FALSE.
217 2. 'acpi_osi=!* acpi_osi="Module Device"' can make
218 '_OSI("Module Device")' TRUE.
219 3. 'acpi_osi=! acpi_osi=!* acpi_osi="Windows 2000"' is
221 'acpi_osi=!* acpi_osi=! acpi_osi="Windows 2000"'
223 'acpi_osi=!* acpi_osi="Windows 2000" acpi_osi=!',
224 they all will make '_OSI("Windows 2000")' TRUE.
227 Override the pmtimer bug detection: force the kernel
228 to assume that this machine's pmtimer latches its value
229 and always returns good values.
231 acpi_sci= [HW,ACPI] ACPI System Control Interrupt trigger mode
232 Format: { level | edge | high | low }
234 acpi_skip_timer_override [HW,ACPI]
235 Recognize and ignore IRQ0/pin2 Interrupt Override.
236 For broken nForce2 BIOS resulting in XT-PIC timer.
238 acpi_sleep= [HW,ACPI] Sleep options
239 Format: { s3_bios, s3_mode, s3_beep, s4_hwsig,
240 s4_nohwsig, old_ordering, nonvs,
241 sci_force_enable, nobl }
242 See Documentation/power/video.rst for information on
244 s3_beep is for debugging; it makes the PC's speaker beep
245 as soon as the kernel's real-mode entry point is called.
246 s4_hwsig causes the kernel to check the ACPI hardware
247 signature during resume from hibernation, and gracefully
248 refuse to resume if it has changed. This complies with
249 the ACPI specification but not with reality, since
250 Windows does not do this and many laptops do change it
251 on docking. So the default behaviour is to allow resume
252 and simply warn when the signature changes, unless the
253 s4_hwsig option is enabled.
254 s4_nohwsig prevents ACPI hardware signature from being
255 used (or even warned about) during resume.
256 old_ordering causes the ACPI 1.0 ordering of the _PTS
257 control method, with respect to putting devices into
258 low power states, to be enforced (the ACPI 2.0 ordering
259 of _PTS is used by default).
260 nonvs prevents the kernel from saving/restoring the
261 ACPI NVS memory during suspend/hibernation and resume.
262 sci_force_enable causes the kernel to set SCI_EN directly
263 on resume from S1/S3 (which is against the ACPI spec,
264 but some broken systems don't work without it).
265 nobl causes the internal blacklist of systems known to
266 behave incorrectly in some ways with respect to system
267 suspend and resume to be ignored (use wisely).
269 acpi_use_timer_override [HW,ACPI]
270 Use timer override. For some broken Nvidia NF5 boards
271 that require a timer override, but don't have HPET
273 add_efi_memmap [EFI; X86] Include EFI memory map in
274 kernel's map of available physical RAM.
277 { off | try_unsupported }
278 off: disable AGP support
279 try_unsupported: try to drive unsupported chipsets
280 (may crash computer or cause data corruption)
283 See Documentation/sound/alsa-configuration.rst
286 Allow the default userspace alignment fault handler
287 behaviour to be specified. Bit 0 enables warnings,
288 bit 1 enables fixups, and bit 2 sends a segfault.
290 align_va_addr= [X86-64]
291 Align virtual addresses by clearing slice [14:12] when
292 allocating a VMA at process creation time. This option
293 gives you up to 3% performance improvement on AMD F15h
294 machines (where it is enabled by default) for a
295 CPU-intensive style benchmark, and it can vary highly in
296 a microbenchmark depending on workload and compiler.
298 32: only for 32-bit processes
299 64: only for 64-bit processes
300 on: enable for both 32- and 64-bit processes
301 off: disable for both 32- and 64-bit processes
303 alloc_snapshot [FTRACE]
304 Allocate the ftrace snapshot buffer on boot up when the
305 main buffer is allocated. This is handy if debugging
306 and you need to use tracing_snapshot() on boot up, and
307 do not want to use tracing_snapshot_alloc() as it needs
308 to be done where GFP_KERNEL allocations are allowed.
310 allow_mismatched_32bit_el0 [ARM64]
311 Allow execve() of 32-bit applications and setting of the
312 PER_LINUX32 personality on systems where only a strict
313 subset of the CPUs support 32-bit EL0. When this
314 parameter is present, the set of CPUs supporting 32-bit
315 EL0 is indicated by /sys/devices/system/cpu/aarch32_el0
316 and hot-unplug operations may be restricted.
318 See Documentation/arch/arm64/asymmetric-32bit.rst for more
321 amd_iommu= [HW,X86-64]
322 Pass parameters to the AMD IOMMU driver in the system.
324 fullflush - Deprecated, equivalent to iommu.strict=1
325 off - do not initialize any AMD IOMMU found in
327 force_isolation - Force device isolation for all
328 devices. The IOMMU driver is not
329 allowed anymore to lift isolation
330 requirements as needed. This option
331 does not override iommu=pt
332 force_enable - Force enable the IOMMU on platforms known
333 to be buggy with IOMMU enabled. Use this
335 pgtbl_v1 - Use v1 page table for DMA-API (Default).
336 pgtbl_v2 - Use v2 page table for DMA-API.
337 irtcachedis - Disable Interrupt Remapping Table (IRT) caching.
339 amd_iommu_dump= [HW,X86-64]
340 Enable AMD IOMMU driver option to dump the ACPI table
341 for AMD IOMMU. With this option enabled, AMD IOMMU
342 driver will print ACPI tables for AMD IOMMU during
343 IOMMU initialization.
345 amd_iommu_intr= [HW,X86-64]
346 Specifies one of the following AMD IOMMU interrupt
348 legacy - Use legacy interrupt remapping mode.
349 vapic - Use virtual APIC mode, which allows IOMMU
350 to inject interrupts directly into guest.
351 This mode requires kvm-amd.avic=1.
352 (Default when IOMMU HW support is present.)
356 Do not enable amd_pstate as the default
357 scaling driver for the supported processors
359 Use amd_pstate with passive mode as a scaling driver.
360 In this mode autonomous selection is disabled.
361 Driver requests a desired performance level and platform
362 tries to match the same performance level if it is
363 satisfied by guaranteed performance level.
365 Use amd_pstate_epp driver instance as the scaling driver,
366 driver provides a hint to the hardware if software wants
367 to bias toward performance (0x0) or energy efficiency (0xff)
368 to the CPPC firmware. then CPPC power algorithm will
369 calculate the runtime workload and adjust the realtime cores
372 Activate guided autonomous mode. Driver requests minimum and
373 maximum performance level and the platform autonomously
374 selects a performance level in this range and appropriate
375 to the current workload.
377 amijoy.map= [HW,JOY] Amiga joystick support
378 Map of devices attached to JOY0DAT and JOY1DAT
380 See also Documentation/input/joydev/joystick.rst
382 analog.map= [HW,JOY] Analog joystick and gamepad support
383 Specifies type or capabilities of an analog joystick
384 connected to one of 16 gameports
385 Format: <type1>,<type2>,..<type16>
388 Power management functions (SPARCstation-4/5 + deriv.)
390 Disable APC CPU standby support. SPARCstation-Fox does
391 not play well with APC CPU idle - disable it if you have
392 APC and your system crashes randomly.
394 apic= [APIC,X86] Advanced Programmable Interrupt Controller
395 Change the output verbosity while booting
396 Format: { quiet (default) | verbose | debug }
397 Change the amount of debugging information output
398 when initialising the APIC and IO-APIC components.
399 For X86-32, this can also be used to specify an APIC
401 Format: apic=driver_name
402 Examples: apic=bigsmp
404 apic_extnmi= [APIC,X86] External NMI delivery setting
405 Format: { bsp (default) | all | none }
406 bsp: External NMI is delivered only to CPU 0
407 all: External NMIs are broadcast to all CPUs as a
409 none: External NMI is masked for all CPUs. This is
410 useful so that a dump capture kernel won't be
414 See Documentation/networking/ipv6.rst.
416 apm= [APM] Advanced Power Management
417 See header of arch/x86/kernel/apm_32.c.
419 apparmor= [APPARMOR] Disable or enable AppArmor at boot time
420 Format: { "0" | "1" }
421 See security/apparmor/Kconfig help text
424 Default value is set via kernel config option.
426 arcrimi= [HW,NET] ARCnet - "RIM I" (entirely mem-mapped) cards
427 Format: <io>,<irq>,<nodeID>
429 arm64.nobti [ARM64] Unconditionally disable Branch Target
430 Identification support
432 arm64.nomops [ARM64] Unconditionally disable Memory Copy and Memory
433 Set instructions support
435 arm64.nomte [ARM64] Unconditionally disable Memory Tagging Extension
438 arm64.nopauth [ARM64] Unconditionally disable Pointer Authentication
441 arm64.nosme [ARM64] Unconditionally disable Scalable Matrix
444 arm64.nosve [ARM64] Unconditionally disable Scalable Vector
449 atarimouse= [HW,MOUSE] Atari Mouse
451 atkbd.extra= [HW] Enable extra LEDs and keys on IBM RapidAccess,
452 EzKey and similar keyboards
454 atkbd.reset= [HW] Reset keyboard during initialization
456 atkbd.set= [HW] Select keyboard code set
457 Format: <int> (2 = AT (default), 3 = PS/2)
459 atkbd.scroll= [HW] Enable scroll wheel on MS Office and similar
462 atkbd.softraw= [HW] Choose between synthetic and real raw mode
463 Format: <bool> (0 = real, 1 = synthetic (default))
465 atkbd.softrepeat= [HW]
466 Use software keyboard repeat
468 audit= [KNL] Enable the audit sub-system
469 Format: { "0" | "1" | "off" | "on" }
470 0 | off - kernel audit is disabled and can not be
471 enabled until the next reboot
472 unset - kernel audit is initialized but disabled and
473 will be fully enabled by the userspace auditd.
474 1 | on - kernel audit is initialized and partially
475 enabled, storing at most audit_backlog_limit
476 messages in RAM until it is fully enabled by the
480 audit_backlog_limit= [KNL] Set the audit queue size limit.
481 Format: <int> (must be >=0)
484 bau= [X86_UV] Enable the BAU on SGI UV. The default
485 behavior is to disable the BAU (i.e. bau=0).
486 Format: { "0" | "1" }
489 unset - Disable the BAU.
491 baycom_epp= [HW,AX25]
494 baycom_par= [HW,AX25] BayCom Parallel Port AX.25 Modem
496 See header of drivers/net/hamradio/baycom_par.c.
498 baycom_ser_fdx= [HW,AX25]
499 BayCom Serial Port AX.25 Modem (Full Duplex Mode)
500 Format: <io>,<irq>,<mode>[,<baud>]
501 See header of drivers/net/hamradio/baycom_ser_fdx.c.
503 baycom_ser_hdx= [HW,AX25]
504 BayCom Serial Port AX.25 Modem (Half Duplex Mode)
505 Format: <io>,<irq>,<mode>
506 See header of drivers/net/hamradio/baycom_ser_hdx.c.
509 Disable BERT OS support on buggy BIOSes.
511 bgrt_disable [ACPI][X86]
512 Disable BGRT to avoid flickering OEM logo.
514 blkdevparts= Manual partition parsing of block device(s) for
515 embedded devices based on command line input.
516 See Documentation/block/cmdline-partition.rst
518 boot_delay= Milliseconds to delay each printk during boot.
519 Only works if CONFIG_BOOT_PRINTK_DELAY is enabled,
520 and you may also have to specify "lpj=". Boot_delay
521 values larger than 10 seconds (10000) are assumed
522 erroneous and ignored.
526 Extended command line options can be added to an initrd
527 and this will cause the kernel to look for it.
529 See Documentation/admin-guide/bootconfig.rst
531 bttv.card= [HW,V4L] bttv (bt848 + bt878 based grabber cards)
532 bttv.radio= Most important insmod options are available as
534 bttv.pll= See Documentation/admin-guide/media/bttv.rst
537 bulk_remove=off [PPC] This parameter disables the use of the pSeries
538 firmware feature for flushing multiple hpte entries
541 c101= [NET] Moxa C101 synchronous serial card
543 cachesize= [BUGS=X86-32] Override level 2 CPU cache size detection.
544 Sometimes CPU hardware bugs make them report the cache
545 size incorrectly. The kernel will attempt work arounds
546 to fix known problems, but for some CPUs it is not
547 possible to determine what the correct size should be.
548 This option provides an override for these situations.
551 [NET] Specifies amount of time (in seconds) that
552 the kernel should wait for a network carrier. By default
553 it waits 120 seconds.
555 ca_keys= [KEYS] This parameter identifies a specific key(s) on
556 the system trusted keyring to be used for certificate
558 format: { id:<keyid> | builtin }
560 cca= [MIPS] Override the kernel pages' cache coherency
561 algorithm. Accepted values range from 0 to 7
562 inclusive. See arch/mips/include/asm/pgtable-bits.h
563 for platform specific values (SB1, Loongson3 and
566 ccw_timeout_log [S390]
567 See Documentation/arch/s390/common_io.rst for details.
569 cgroup_disable= [KNL] Disable a particular controller or optional feature
570 Format: {name of the controller(s) or feature(s) to disable}
571 The effects of cgroup_disable=foo are:
572 - foo isn't auto-mounted if you mount all cgroups in
574 - foo isn't visible as an individually mountable
576 - if foo is an optional feature then the feature is
577 disabled and corresponding cgroup files are not
579 {Currently only "memory" controller deal with this and
580 cut the overhead, others just disable the usage. So
581 only cgroup_disable=memory is actually worthy}
582 Specifying "pressure" disables per-cgroup pressure
583 stall information accounting feature
585 cgroup_no_v1= [KNL] Disable cgroup controllers and named hierarchies in v1
586 Format: { { controller | "all" | "named" }
587 [,{ controller | "all" | "named" }...] }
588 Like cgroup_disable, but only applies to cgroup v1;
589 the blacklisted controllers remain available in cgroup2.
590 "all" blacklists all controllers and "named" disables
591 named mounts. Specifying both "all" and "named" disables
594 cgroup_favordynmods= [KNL] Enable or Disable favordynmods.
595 Format: { "true" | "false" }
596 Defaults to the value of CONFIG_CGROUP_FAVOR_DYNMODS.
598 cgroup.memory= [KNL] Pass options to the cgroup memory controller.
600 nosocket -- Disable socket memory accounting.
601 nokmem -- Disable kernel memory accounting.
602 nobpf -- Disable BPF memory accounting.
604 checkreqprot= [SELINUX] Set initial checkreqprot flag value.
605 Format: { "0" | "1" }
606 See security/selinux/Kconfig help text.
607 0 -- check protection applied by kernel (includes
608 any implied execute protection).
609 1 -- check protection requested by application.
610 Default value is set via a kernel config option.
611 Value can be changed at runtime via
612 /sys/fs/selinux/checkreqprot.
613 Setting checkreqprot to 1 is deprecated.
616 See Documentation/arch/s390/common_io.rst for details.
618 clearcpuid=X[,X...] [X86]
619 Disable CPUID feature X for the kernel. See
620 arch/x86/include/asm/cpufeatures.h for the valid bit
621 numbers X. Note the Linux-specific bits are not necessarily
622 stable over kernel options, but the vendor-specific
624 X can also be a string as appearing in the flags: line
625 in /proc/cpuinfo which does not have the above
626 instability issue. However, not all features have names
628 Note that using this option will taint your kernel.
629 Also note that user programs calling CPUID directly
630 or using the feature without checking anything
631 will still see it. This just prevents it from
632 being used by the kernel or shown in /proc/cpuinfo.
633 Also note the kernel might malfunction if you disable
638 Prevents the clock framework from automatically gating
639 clocks that have not been explicitly enabled by a Linux
640 device driver but are enabled in hardware at reset or
641 by the bootloader/firmware. Note that this does not
642 force such clocks to be always-on nor does it reserve
643 those clocks in any way. This parameter is useful for
644 debug and development, but should not be needed on a
645 platform with proper driver support. For more
646 information, see Documentation/driver-api/clk.rst.
648 clock= [BUGS=X86-32, HW] gettimeofday clocksource override.
650 Forces specified clocksource (if available) to be used
651 when calculating gettimeofday(). If specified
652 clocksource is not available, it defaults to PIT.
653 Format: { pit | tsc | cyclone | pmtmr }
655 clocksource= Override the default clocksource
657 Override the default clocksource and use the clocksource
658 with the name specified.
659 Some clocksource names to choose from, depending on
661 [all] jiffies (this is the base, fallback clocksource)
663 [ARM] imx_timer1,OSTS,netx_timer,mpu_timer2,
664 pxa_timer,timer3,32k_counter,timer0_1
665 [X86-32] pit,hpet,tsc;
666 scx200_hrt on Geode; cyclone on IBM x440
674 clocksource.arm_arch_timer.evtstrm=
677 Enable/disable the eventstream feature of the ARM
678 architected timer so that code using WFE-based polling
679 loops can be debugged more effectively on production
682 clocksource.max_cswd_read_retries= [KNL]
683 Number of clocksource_watchdog() retries due to
684 external delays before the clock will be marked
685 unstable. Defaults to two retries, that is,
686 three attempts to read the clock under test.
688 clocksource.verify_n_cpus= [KNL]
689 Limit the number of CPUs checked for clocksources
690 marked with CLOCK_SOURCE_VERIFY_PERCPU that
691 are marked unstable due to excessive skew.
692 A negative value says to check all CPUs, while
693 zero says not to check any. Values larger than
694 nr_cpu_ids are silently truncated to nr_cpu_ids.
695 The actual CPUs are chosen randomly, with
696 no replacement if the same CPU is chosen twice.
698 clocksource-wdtest.holdoff= [KNL]
699 Set the time in seconds that the clocksource
700 watchdog test waits before commencing its tests.
701 Defaults to zero when built as a module and to
702 10 seconds when built into the kernel.
704 cma=nn[MG]@[start[MG][-end[MG]]]
706 Sets the size of kernel global memory area for
707 contiguous memory allocations and optionally the
708 placement constraint by the physical address range of
709 memory allocations. A value of 0 disables CMA
710 altogether. For more information, see
711 kernel/dma/contiguous.c
715 Sets the size of kernel per-numa memory area for
716 contiguous memory allocations. A value of 0 disables
717 per-numa CMA altogether. And If this option is not
718 specified, the default value is 0.
719 With per-numa CMA enabled, DMA users on node nid will
720 first try to allocate buffer from the pernuma area
721 which is located in node nid, if the allocation fails,
722 they will fallback to the global default memory area.
724 numa_cma=<node>:nn[MG][,<node>:nn[MG]]
726 Sets the size of kernel numa memory area for
727 contiguous memory allocations. It will reserve CMA
728 area for the specified node.
730 With numa CMA enabled, DMA users on node nid will
731 first try to allocate buffer from the numa area
732 which is located in node nid, if the allocation fails,
733 they will fallback to the global default memory area.
735 cmo_free_hint= [PPC] Format: { yes | no }
736 Specify whether pages are marked as being inactive
737 when they are freed. This is used in CMO environments
738 to determine OS memory pressure for page stealing by
742 coherent_pool=nn[KMG] [ARM,KNL]
743 Sets the size of memory pool for coherent, atomic dma
744 allocations, by default set to 256K.
746 com20020= [HW,NET] ARCnet - COM20020 chipset
748 <io>[,<irq>[,<nodeID>[,<backplane>[,<ckp>[,<timeout>]]]]]
750 com90io= [HW,NET] ARCnet - COM90xx chipset (IO-mapped buffers)
754 ARCnet - COM90xx chipset (memory-mapped buffers)
755 Format: <io>[,<irq>[,<memstart>]]
757 condev= [HW,S390] console device
760 con3215_drop= [S390] 3215 console drop mode.
762 When set to true, drop data on the 3215 console when
763 the console buffer is full. In this case the
764 operator using a 3270 terminal emulator (for example
765 x3270) does not have to enter the clear key for the
766 console output to advance and the kernel to continue.
767 This leads to a much faster boot time when a 3270
768 terminal emulator is active. If no 3270 terminal
769 emulator is used, this parameter has no effect.
771 console= [KNL] Output console device and options.
773 tty<n> Use the virtual console device <n>.
777 Use the specified serial port. The options are of
778 the form "bbbbpnf", where "bbbb" is the baud rate,
779 "p" is parity ("n", "o", or "e"), "n" is number of
780 bits, and "f" is flow control ("r" for RTS or
781 omit it). Default is "9600n8".
783 See Documentation/admin-guide/serial-console.rst for more
785 Documentation/networking/netconsole.rst for an
788 uart[8250],io,<addr>[,options]
789 uart[8250],mmio,<addr>[,options]
790 uart[8250],mmio16,<addr>[,options]
791 uart[8250],mmio32,<addr>[,options]
792 uart[8250],0x<addr>[,options]
793 Start an early, polled-mode console on the 8250/16550
794 UART at the specified I/O port or MMIO address,
795 switching to the matching ttyS device later.
796 MMIO inter-register address stride is either 8-bit
797 (mmio), 16-bit (mmio16), or 32-bit (mmio32).
798 If none of [io|mmio|mmio16|mmio32], <addr> is assumed
799 to be equivalent to 'mmio'. 'options' are specified in
800 the same format described for ttyS above; if unspecified,
801 the h/w is not re-initialized.
803 hvc<n> Use the hypervisor console device <n>. This is for
804 both Xen and PowerPC hypervisors.
807 Use to disable console output, i.e., to have kernel
808 console messages discarded.
809 This must be the only console= parameter used on the
812 If the device connected to the port is not a TTY but a braille
813 device, prepend "brl," before the device type, for instance
815 For now, only VisioBraille is supported.
818 [KNL] Change console messages format
820 By default we print messages on consoles in
821 "[time stamp] text\n" format (time stamp may not be
822 printed, depending on CONFIG_PRINTK_TIME or
823 `printk_time' param).
825 Switch to syslog format: "<%u>[time stamp] text\n"
826 IOW, each message will have a facility and loglevel
827 prefix. The format is similar to one used by syslog()
828 syscall, or to executing "dmesg -S --raw" or to reading
831 consoleblank= [KNL] The console blank (screen saver) timeout in
832 seconds. A value of 0 disables the blank timer.
836 [KNL] Change the default value for
837 /proc/<pid>/coredump_filter.
838 See also Documentation/filesystems/proc.rst.
840 coresight_cpu_debug.enable
843 Enable/disable the CPU sampling based debugging.
844 0: default value, disable debugging
845 1: enable debugging at boot time
847 cpcihp_generic= [HW,PCI] Generic port I/O CompactPCI driver
849 <first_slot>,<last_slot>,<port>,<enum_bit>[,<debug>]
851 cpuidle.off=1 [CPU_IDLE]
852 disable the cpuidle sub-system
855 [CPU_IDLE] Name of the cpuidle governor to use.
857 cpufreq.off=1 [CPU_FREQ]
858 disable the cpufreq sub-system
860 cpufreq.default_governor=
861 [CPU_FREQ] Name of the default cpufreq governor or
862 policy to use. This governor must be registered in the
863 kernel before the cpufreq driver probes.
866 [X86] Delay for N microsec between assert and de-assert
867 of APIC INIT to start processors. This delay occurs
868 on every CPU online, such as boot, and resume from suspend.
872 [SMP] Enable/disable parallel bringup of secondary CPUs
874 Default is enabled if CONFIG_HOTPLUG_PARALLEL=y. Otherwise
875 the parameter has no effect.
877 crash_kexec_post_notifiers
878 Run kdump after running panic-notifiers and dumping
879 kmsg. This only for the users who doubt kdump always
880 succeeds in any situation.
881 Note that this also increases risks of kdump failure,
882 because some panic notifiers can make the crashed
883 kernel more unstable.
885 crashkernel=size[KMG][@offset[KMG]]
886 [KNL] Using kexec, Linux can switch to a 'crash kernel'
887 upon panic. This parameter reserves the physical
888 memory region [offset, offset + size] for that kernel
889 image. If '@offset' is omitted, then a suitable offset
890 is selected automatically.
891 [KNL, X86-64, ARM64, RISCV] Select a region under 4G first, and
892 fall back to reserve region above 4G when '@offset'
893 hasn't been specified.
894 See Documentation/admin-guide/kdump/kdump.rst for further details.
896 crashkernel=range1:size1[,range2:size2,...][@offset]
897 [KNL] Same as above, but depends on the memory
898 in the running system. The syntax of range is
899 start-[end] where start and end are both
900 a memory unit (amount[KMG]). See also
901 Documentation/admin-guide/kdump/kdump.rst for an example.
903 crashkernel=size[KMG],high
904 [KNL, X86-64, ARM64, RISCV] range could be above 4G.
905 Allow kernel to allocate physical memory region from top,
906 so could be above 4G if system have more than 4G ram
907 installed. Otherwise memory region will be allocated
908 below 4G, if available.
909 It will be ignored if crashkernel=X is specified.
910 crashkernel=size[KMG],low
911 [KNL, X86-64, ARM64, RISCV] range under 4G. When crashkernel=X,high
912 is passed, kernel could allocate physical memory region
913 above 4G, that cause second kernel crash on system
914 that require some amount of low memory, e.g. swiotlb
915 requires at least 64M+32K low memory, also enough extra
916 low memory is needed to make sure DMA buffers for 32-bit
917 devices won't run out. Kernel would try to allocate
918 default size of memory below 4G automatically. The default
919 size is platform dependent.
920 --> x86: max(swiotlb_size_or_default() + 8MiB, 256MiB)
923 This one lets the user specify own low range under 4G
924 for second kernel instead.
925 0: to disable low allocation.
926 It will be ignored when crashkernel=X,high is not used
927 or memory reserved is below 4G.
930 [KNL] Disable crypto self-tests
935 cs89x0_media= [HW,NET]
936 Format: { rj45 | aui | bnc }
938 csdlock_debug= [KNL] Enable or disable debug add-ons of cross-CPU
939 function call handling. When switched on,
940 additional debug data is printed to the console
941 in case a hanging CPU is detected, and that
942 CPU is pinged again in order to try to resolve
943 the hang situation. The default value of this
944 option depends on the CSD_LOCK_WAIT_DEBUG_DEFAULT
948 See header of drivers/s390/block/dasd_devmap.c.
950 db9.dev[2|3]= [HW,JOY] Multisystem joystick support via parallel port
951 (one device per port)
952 Format: <port#>,<type>
953 See also Documentation/input/devices/joystick-parport.rst
955 debug [KNL] Enable kernel debugging (events log level).
958 [KNL] Enable printing [hashed] pointers early in the
959 boot sequence. If enabled, we use a weak hash instead
960 of siphash to hash pointers. Use this option if you are
961 seeing instances of '(___ptrval___)') and need to see a
962 value (hashed pointer) instead. Cryptographically
963 insecure, please do not use on production kernels.
966 [KNL] verbose locking self-tests
968 Print debugging info while doing the locking API
970 Bitmask for the various LOCKTYPE_ tests. Defaults to 0
971 (no extra messages), setting it to -1 (all bits set)
972 will print _a_lot_ more information - normally only
973 useful to lockdep developers.
975 debug_objects [KNL] Enable object debugging
977 debug_guardpage_minorder=
978 [KNL] When CONFIG_DEBUG_PAGEALLOC is set, this
979 parameter allows control of the order of pages that will
980 be intentionally kept free (and hence protected) by the
981 buddy allocator. Bigger value increase the probability
982 of catching random memory corruption, but reduce the
983 amount of memory for normal system use. The maximum
984 possible value is MAX_PAGE_ORDER/2. Setting this
985 parameter to 1 or 2 should be enough to identify most
986 random memory corruption problems caused by bugs in
987 kernel or driver code when a CPU writes to (or reads
988 from) a random memory location. Note that there exists
989 a class of memory corruptions problems caused by buggy
990 H/W or F/W or by drivers badly programming DMA
991 (basically when memory is written at bus level and the
992 CPU MMU is bypassed) which are not detectable by
993 CONFIG_DEBUG_PAGEALLOC, hence this option will not
994 help tracking down these problems.
997 [KNL] When CONFIG_DEBUG_PAGEALLOC is set, this parameter
998 enables the feature at boot time. By default, it is
999 disabled and the system will work mostly the same as a
1000 kernel built without CONFIG_DEBUG_PAGEALLOC.
1001 Note: to get most of debug_pagealloc error reports, it's
1002 useful to also enable the page_owner functionality.
1003 on: enable the feature
1005 debugfs= [KNL] This parameter enables what is exposed to userspace
1006 and debugfs internal clients.
1007 Format: { on, no-mount, off }
1008 on: All functions are enabled.
1010 Filesystem is not registered but kernel clients can
1011 access APIs and a crashkernel can be used to read
1012 its content. There is nothing to mount.
1013 off: Filesystem is not registered and clients
1014 get a -EPERM as result when trying to register files
1015 or directories within debugfs.
1016 This is equivalent of the runtime functionality if
1017 debugfs was not enabled in the kernel at all.
1018 Default value is set in build-time with a kernel configuration.
1020 debugpat [X86] Enable PAT debugging
1023 [HW] The size of the default HugeTLB page. This is
1024 the size represented by the legacy /proc/ hugepages
1025 APIs. In addition, this is the default hugetlb size
1026 used for shmget(), mmap() and mounting hugetlbfs
1027 filesystems. If not specified, defaults to the
1028 architecture's default huge page size. Huge page
1029 sizes are architecture dependent. See also
1030 Documentation/admin-guide/mm/hugetlbpage.rst.
1033 deferred_probe_timeout=
1034 [KNL] Debugging option to set a timeout in seconds for
1035 deferred probe to give up waiting on dependencies to
1036 probe. Only specific dependencies (subsystems or
1037 drivers) that have opted in will be ignored. A timeout
1038 of 0 will timeout at the end of initcalls. If the time
1039 out hasn't expired, it'll be restarted by each
1040 successful driver registration. This option will also
1041 dump out devices still on the deferred probe list after
1044 delayacct [KNL] Enable per-task delay accounting
1046 dell_smm_hwmon.ignore_dmi=
1047 [HW] Continue probing hardware even if DMI data
1048 indicates that the driver is running on unsupported
1051 dell_smm_hwmon.force=
1052 [HW] Activate driver even if SMM BIOS signature does
1053 not match list of supported models and enable otherwise
1054 blacklisted features.
1056 dell_smm_hwmon.power_status=
1057 [HW] Report power status in /proc/i8k
1058 (disabled by default).
1060 dell_smm_hwmon.restricted=
1061 [HW] Allow controlling fans only if SYS_ADMIN
1064 dell_smm_hwmon.fan_mult=
1065 [HW] Factor to multiply fan speed with.
1067 dell_smm_hwmon.fan_max=
1068 [HW] Maximum configurable fan speed.
1071 Format: { on | off | def_only | inf_only | always }
1072 on: s390 zlib hardware support for compression on
1073 level 1 and decompression (default)
1074 off: No s390 zlib hardware support
1075 def_only: s390 zlib hardware support for deflate
1076 only (compression on level 1)
1077 inf_only: s390 zlib hardware support for inflate
1078 only (decompression)
1079 always: Same as 'on' but ignores the selected compression
1080 level always using hardware support (used for debugging)
1082 dhash_entries= [KNL]
1083 Set number of hash buckets for dentry cache.
1085 disable_1tb_segments [PPC]
1086 Disables the use of 1TB hash page table segments. This
1087 causes the kernel to fall back to 256MB segments which
1088 can be useful when debugging issues that require an SLB
1092 See Documentation/networking/ipv6.rst.
1095 Disable RADIX MMU mode on POWER9
1098 Disable TLBIE instruction. Currently does not work
1099 with KVM, with HASH MMU, or with coherent accelerators.
1101 disable_cpu_apicid= [X86,APIC,SMP]
1103 The number of initial APIC ID for the
1104 corresponding CPU to be disabled at boot,
1105 mostly used for the kdump 2nd kernel to
1106 disable BSP to wake up multiple CPUs without
1107 causing system reset or hang due to sending
1108 INIT from AP to BSP.
1110 disable_ddw [PPC/PSERIES]
1111 Disable Dynamic DMA Window support. Use this
1112 to workaround buggy firmware.
1114 disable_ipv6= [IPV6]
1115 See Documentation/networking/ipv6.rst.
1117 disable_mtrr_cleanup [X86]
1118 The kernel tries to adjust MTRR layout from continuous
1119 to discrete, to make X server driver able to add WB
1120 entry later. This parameter disables that.
1122 disable_mtrr_trim [X86, Intel and AMD only]
1123 By default the kernel will trim any uncacheable
1124 memory out of your available memory pool based on
1125 MTRR settings. This parameter disables that behavior,
1126 possibly causing your machine to run very slowly.
1128 disable_timer_pin_1 [X86]
1129 Disable PIN 1 of APIC timer
1130 Can be useful to work around chipset bugs.
1132 dis_ucode_ldr [X86] Disable the microcode loader.
1134 dma_debug=off If the kernel is compiled with DMA_API_DEBUG support,
1135 this option disables the debugging code at boot.
1137 dma_debug_entries=<number>
1138 This option allows to tune the number of preallocated
1139 entries for DMA-API debugging code. One entry is
1140 required per DMA-API allocation. Use this if the
1141 DMA-API debugging code disables itself because the
1142 architectural default is too low.
1144 dma_debug_driver=<driver_name>
1145 With this option the DMA-API debugging driver
1146 filter feature can be enabled at boot time. Just
1147 pass the driver to filter for as the parameter.
1148 The filter can be disabled or changed to another
1149 driver later using sysfs.
1151 driver_async_probe= [KNL]
1152 List of driver names to be probed asynchronously. *
1153 matches with all driver names. If * is specified, the
1154 rest of the listed driver names are those that will NOT
1156 Format: <driver_name1>,<driver_name2>...
1158 drm.edid_firmware=[<connector>:]<file>[,[<connector>:]<file>]
1159 Broken monitors, graphic adapters, KVMs and EDIDless
1160 panels may send no or incorrect EDID data sets.
1161 This parameter allows to specify an EDID data sets
1162 in the /lib/firmware directory that are used instead.
1163 Generic built-in EDID data sets are used, if one of
1164 edid/1024x768.bin, edid/1280x1024.bin,
1165 edid/1680x1050.bin, or edid/1920x1080.bin is given
1166 and no file with the same name exists. Details and
1167 instructions how to build your own EDID data are
1168 available in Documentation/admin-guide/edid.rst. An EDID
1169 data set will only be used for a particular connector,
1170 if its name and a colon are prepended to the EDID
1171 name. Each connector may use a unique EDID data
1172 set by separating the files with a comma. An EDID
1173 data set with no connector name will be used for
1174 any connectors not explicitly specified.
1179 Format: {"off" | "known"}
1180 Control how the dt_cpu_ftrs device-tree binding is
1181 used for CPU feature discovery and setup (if it
1183 off: Do not use it, fall back to legacy cpu table.
1184 known: Do not pass through unknown features to guests
1185 or userspace, only those that the kernel is aware of.
1187 dump_apple_properties [X86]
1188 Dump name and content of EFI device properties on
1189 x86 Macs. Useful for driver authors to determine
1190 what data is available or for reverse-engineering.
1192 dyndbg[="val"] [KNL,DYNAMIC_DEBUG]
1193 <module>.dyndbg[="val"]
1194 Enable debug messages at boot time. See
1195 Documentation/admin-guide/dynamic-debug-howto.rst
1198 early_ioremap_debug [KNL]
1199 Enable debug messages in early_ioremap support. This
1200 is useful for tracking down temporary early mappings
1201 which are not unmapped.
1203 earlycon= [KNL] Output early console device and options.
1205 When used with no options, the early console is
1206 determined by stdout-path property in device tree's
1207 chosen node or the ACPI SPCR table if supported by
1210 cdns,<addr>[,options]
1211 Start an early, polled-mode console on a Cadence
1212 (xuartps) serial port at the specified address. Only
1213 supported option is baud rate. If baud rate is not
1214 specified, the serial port must already be setup and
1217 uart[8250],io,<addr>[,options[,uartclk]]
1218 uart[8250],mmio,<addr>[,options[,uartclk]]
1219 uart[8250],mmio32,<addr>[,options[,uartclk]]
1220 uart[8250],mmio32be,<addr>[,options[,uartclk]]
1221 uart[8250],0x<addr>[,options]
1222 Start an early, polled-mode console on the 8250/16550
1223 UART at the specified I/O port or MMIO address.
1224 MMIO inter-register address stride is either 8-bit
1225 (mmio) or 32-bit (mmio32 or mmio32be).
1226 If none of [io|mmio|mmio32|mmio32be], <addr> is assumed
1227 to be equivalent to 'mmio'. 'options' are specified
1228 in the same format described for "console=ttyS<n>"; if
1229 unspecified, the h/w is not initialized. 'uartclk' is
1230 the uart clock frequency; if unspecified, it is set
1231 to 'BASE_BAUD' * 16.
1235 Start an early, polled-mode console on a pl011 serial
1236 port at the specified address. The pl011 serial port
1237 must already be setup and configured. Options are not
1238 yet supported. If 'mmio32' is specified, then only
1239 the driver will use only 32-bit accessors to read/write
1240 the device registers.
1243 Start an early console on a litex serial port at the
1244 specified address. The serial port must already be
1245 setup and configured. Options are not yet supported.
1248 Start an early, polled-mode console on a meson serial
1249 port at the specified address. The serial port must
1250 already be setup and configured. Options are not yet
1254 Start an early, polled-mode console on an msm serial
1255 port at the specified address. The serial port
1256 must already be setup and configured. Options are not
1259 msm_serial_dm,<addr>
1260 Start an early, polled-mode console on an msm serial
1261 dm port at the specified address. The serial port
1262 must already be setup and configured. Options are not
1266 Start an early, polled-mode console on a serial port
1267 of an Actions Semi SoC, such as S500 or S900, at the
1268 specified address. The serial port must already be
1269 setup and configured. Options are not yet supported.
1272 Start an early, polled-mode console on a serial port
1273 of an RDA Micro SoC, such as RDA8810PL, at the
1274 specified address. The serial port must already be
1275 setup and configured. Options are not yet supported.
1278 Use RISC-V SBI (Supervisor Binary Interface) for early
1281 smh Use ARM semihosting calls for early console.
1289 Use early console provided by serial driver available
1290 on Samsung SoCs, requires selecting proper type and
1291 a correct base address of the selected UART port. The
1292 serial port must already be setup and configured.
1293 Options are not yet supported.
1296 Start an early, polled-mode console on a lantiq serial
1297 (lqasc) port at the specified address. The serial port
1298 must already be setup and configured. Options are not
1303 Use early console provided by Freescale LP UART driver
1304 found on Freescale Vybrid and QorIQ LS1021A processors.
1305 A valid base address must be provided, and the serial
1306 port must already be setup and configured.
1310 Start an early, polled-mode, output-only console on the
1311 Freescale i.MX UART at the specified address. The UART
1312 must already be setup and configured.
1315 Start an early, polled-mode console on the
1316 Armada 3700 serial port at the specified
1317 address. The serial port must already be setup
1318 and configured. Options are not yet supported.
1321 Start an early, polled-mode console on a Qualcomm
1322 Generic Interface (GENI) based serial port at the
1323 specified address. The serial port must already be
1324 setup and configured. Options are not yet supported.
1327 Start an early, unaccelerated console on the EFI
1328 memory mapped framebuffer (if available). On cache
1329 coherent non-x86 systems that use system memory for
1330 the framebuffer, pass the 'ram' option so that it is
1331 mapped with the correct attributes.
1334 Use early console provided by Freescale LINFlexD UART
1335 serial driver for NXP S32V234 SoCs. A valid base
1336 address must be provided, and the serial port must
1337 already be setup and configured.
1339 earlyprintk= [X86,SH,ARM,M68k,S390]
1343 earlyprintk=serial[,ttySn[,baudrate]]
1344 earlyprintk=serial[,0x...[,baudrate]]
1345 earlyprintk=ttySn[,baudrate]
1346 earlyprintk=dbgp[debugController#]
1347 earlyprintk=pciserial[,force],bus:device.function[,baudrate]
1348 earlyprintk=xdbc[xhciController#]
1351 earlyprintk is useful when the kernel crashes before
1352 the normal console is initialized. It is not enabled by
1353 default because it has some cosmetic problems.
1355 Append ",keep" to not disable it when the real console
1358 Only one of vga, serial, or usb debug port can
1361 Currently only ttyS0 and ttyS1 may be specified by
1362 name. Other I/O ports may be explicitly specified
1363 on some architectures (x86 and arm at least) by
1364 replacing ttySn with an I/O port address, like this:
1365 earlyprintk=serial,0x1008,115200
1366 You can find the port for a given device in
1367 /proc/tty/driver/serial:
1368 2: uart:ST16650V2 port:00001008 irq:18 ...
1370 Interaction with the standard serial driver is not
1373 The VGA output is eventually overwritten by
1376 The xen option can only be used in Xen domains.
1378 The sclp output can only be used on s390.
1380 The bios output can only be used on SuperH.
1382 The optional "force" to "pciserial" enables use of a
1383 PCI device even when its classcode is not of the
1386 edac_report= [HW,EDAC] Control how to report EDAC event
1387 Format: {"on" | "off" | "force"}
1388 on: enable EDAC to report H/W event. May be overridden
1389 by other higher priority error reporting module.
1390 off: disable H/W event reporting through EDAC.
1391 force: enforce the use of EDAC to report H/W event.
1395 Format: {"off" | "on" | "skip[mbr]"}
1398 Format: { "debug", "disable_early_pci_dma",
1399 "nochunk", "noruntime", "nosoftreserve",
1400 "novamap", "no_disable_early_pci_dma" }
1401 debug: enable misc debug output.
1402 disable_early_pci_dma: disable the busmaster bit on all
1403 PCI bridges while in the EFI boot stub.
1404 nochunk: disable reading files in "chunks" in the EFI
1405 boot stub, as chunking can cause problems with some
1406 firmware implementations.
1407 noruntime : disable EFI runtime services support
1408 nosoftreserve: The EFI_MEMORY_SP (Specific Purpose)
1409 attribute may cause the kernel to reserve the
1410 memory range for a memory mapping driver to
1411 claim. Specify efi=nosoftreserve to disable this
1412 reservation and treat the memory by its base type
1413 (i.e. EFI_CONVENTIONAL_MEMORY / "System RAM").
1414 novamap: do not call SetVirtualAddressMap().
1415 no_disable_early_pci_dma: Leave the busmaster bit set
1416 on all PCI bridges while in the EFI boot stub
1418 efi_no_storage_paranoia [EFI; X86]
1419 Using this parameter you can use more than 50% of
1420 your efi variable storage. Use this parameter only if
1421 you are really sure that your UEFI does sane gc and
1422 fulfills the spec otherwise your board may brick.
1424 efi_fake_mem= nn[KMG]@ss[KMG]:aa[,nn[KMG]@ss[KMG]:aa,..] [EFI; X86]
1425 Add arbitrary attribute to specific memory range by
1426 updating original EFI memory map.
1427 Region of memory which aa attribute is added to is
1430 If efi_fake_mem=2G@4G:0x10000,2G@0x10a0000000:0x10000
1431 is specified, EFI_MEMORY_MORE_RELIABLE(0x10000)
1432 attribute is added to range 0x100000000-0x180000000 and
1433 0x10a0000000-0x1120000000.
1435 If efi_fake_mem=8G@9G:0x40000 is specified, the
1436 EFI_MEMORY_SP(0x40000) attribute is added to
1437 range 0x240000000-0x43fffffff.
1439 Using this parameter you can do debugging of EFI memmap
1440 related features. For example, you can do debugging of
1441 Address Range Mirroring feature even if your box
1442 doesn't support it, or mark specific memory as
1445 efivar_ssdt= [EFI; X86] Name of an EFI variable that contains an SSDT
1446 that is to be dynamically loaded by Linux. If there are
1447 multiple variables with the same name but with different
1448 vendor GUIDs, all of them will be loaded. See
1449 Documentation/admin-guide/acpi/ssdt-overlays.rst for details.
1452 eisa_irq_edge= [PARISC,HW]
1453 See header of drivers/parisc/eisa.c.
1455 ekgdboc= [X86,KGDB] Allow early kernel console debugging
1458 This is designed to be used in conjunction with
1459 the boot argument: earlyprintk=vga
1461 This parameter works in place of the kgdboc parameter
1462 but can only be used if the backing tty is available
1463 very early in the boot process. For early debugging
1464 via a serial port see kgdboc_earlycon instead.
1467 See comment before function elanfreq_setup() in
1468 arch/x86/kernel/cpu/cpufreq/elanfreq.c.
1470 elfcorehdr=[size[KMG]@]offset[KMG] [PPC,SH,X86,S390]
1471 Specifies physical address of start of kernel core
1472 image elf header and optionally the size. Generally
1473 kexec loader will pass this option to capture kernel.
1474 See Documentation/admin-guide/kdump/kdump.rst for details.
1476 enable_mtrr_cleanup [X86]
1477 The kernel tries to adjust MTRR layout from continuous
1478 to discrete, to make X server driver able to add WB
1479 entry later. This parameter enables that.
1481 enable_timer_pin_1 [X86]
1482 Enable PIN 1 of APIC timer
1483 Can be useful to work around chipset bugs
1484 (in particular on some ATI chipsets).
1485 The kernel tries to set a reasonable default.
1487 enforcing= [SELINUX] Set initial enforcing status.
1489 See security/selinux/Kconfig help text.
1490 0 -- permissive (log only, no denials).
1491 1 -- enforcing (deny and log).
1493 Value can be changed at runtime via
1494 /sys/fs/selinux/enforce.
1497 Disable Error Record Serialization Table (ERST)
1500 ether= [HW,NET] Ethernet cards parameters
1501 This option is obsoleted by the "netdev=" option, which
1502 has equivalent usage. See its documentation for details.
1506 Permit 'security.evm' to be updated regardless of
1507 current integrity status.
1509 early_page_ext [KNL] Enforces page_ext initialization to earlier
1510 stages so cover more early boot allocations.
1511 Please note that as side effect some optimizations
1512 might be disabled to achieve that (e.g. parallelized
1513 memory initialization is disabled) so the boot process
1514 might take longer, especially on systems with a lot of
1515 memory. Available with CONFIG_PAGE_EXTENSION=y.
1520 fail_make_request=[KNL]
1521 General fault injection mechanism.
1522 Format: <interval>,<probability>,<space>,<times>
1523 See also Documentation/fault-injection/.
1526 Format: { initns | none }
1527 See Documentation/admin-guide/sysctl/net.rst for
1528 fb_tunnels_only_for_init_ns
1531 See Documentation/admin-guide/blockdev/floppy.rst.
1534 Forcefully enable Physical Address Extension (PAE).
1535 Many Pentium M systems disable PAE but may have a
1536 functionally usable PAE implementation.
1537 Warning: use of this parameter will taint the kernel
1538 and may cause unknown problems.
1541 [FTRACE] will set and start the specified tracer
1542 as early as possible in order to facilitate early
1545 ftrace_boot_snapshot
1546 [FTRACE] On boot up, a snapshot will be taken of the
1547 ftrace ring buffer that can be read at:
1548 /sys/kernel/tracing/snapshot.
1549 This is useful if you need tracing information from kernel
1550 boot up that is likely to be overridden by user space
1551 start up functionality.
1553 Optionally, the snapshot can also be defined for a tracing
1554 instance that was created by the trace_instance= command
1557 trace_instance=foo,sched_switch ftrace_boot_snapshot=foo
1559 The above will cause the "foo" tracing instance to trigger
1560 a snapshot at the end of boot up.
1562 ftrace_dump_on_oops[=orig_cpu]
1563 [FTRACE] will dump the trace buffers on oops.
1564 If no parameter is passed, ftrace will dump
1565 buffers of all CPUs, but if you pass orig_cpu, it will
1566 dump only the buffer of the CPU that triggered the
1569 ftrace_filter=[function-list]
1570 [FTRACE] Limit the functions traced by the function
1571 tracer at boot up. function-list is a comma-separated
1572 list of functions. This list can be changed at run
1573 time by the set_ftrace_filter file in the debugfs
1576 ftrace_notrace=[function-list]
1577 [FTRACE] Do not trace the functions specified in
1578 function-list. This list can be changed at run time
1579 by the set_ftrace_notrace file in the debugfs
1582 ftrace_graph_filter=[function-list]
1583 [FTRACE] Limit the top level callers functions traced
1584 by the function graph tracer at boot up.
1585 function-list is a comma-separated list of functions
1586 that can be changed at run time by the
1587 set_graph_function file in the debugfs tracing directory.
1589 ftrace_graph_notrace=[function-list]
1590 [FTRACE] Do not trace from the functions specified in
1591 function-list. This list is a comma-separated list of
1592 functions that can be changed at run time by the
1593 set_graph_notrace file in the debugfs tracing directory.
1595 ftrace_graph_max_depth=<uint>
1596 [FTRACE] Used with the function graph tracer. This is
1597 the max depth it will trace into a function. This value
1598 can be changed at run time by the max_graph_depth file
1599 in the tracefs tracing directory. default: 0 (no limit)
1601 fw_devlink= [KNL] Create device links between consumer and supplier
1602 devices by scanning the firmware to infer the
1603 consumer/supplier relationships. This feature is
1604 especially useful when drivers are loaded as modules as
1605 it ensures proper ordering of tasks like device probing
1606 (suppliers first, then consumers), supplier boot state
1607 clean up (only after all consumers have probed),
1608 suspend/resume & runtime PM (consumers first, then
1610 Format: { off | permissive | on | rpm }
1611 off -- Don't create device links from firmware info.
1612 permissive -- Create device links from firmware info
1613 but use it only for ordering boot state clean
1614 up (sync_state() calls).
1615 on -- Create device links from firmware info and use it
1616 to enforce probe and suspend/resume ordering.
1617 rpm -- Like "on", but also use to order runtime PM.
1619 fw_devlink.strict=<bool>
1620 [KNL] Treat all inferred dependencies as mandatory
1621 dependencies. This only applies for fw_devlink=on|rpm.
1624 fw_devlink.sync_state =
1625 [KNL] When all devices that could probe have finished
1626 probing, this parameter controls what to do with
1627 devices that haven't yet received their sync_state()
1629 Format: { strict | timeout }
1630 strict -- Default. Continue waiting on consumers to
1632 timeout -- Give up waiting on consumers and call
1633 sync_state() on any devices that haven't yet
1634 received their sync_state() calls after
1635 deferred_probe_timeout has expired or by
1636 late_initcall() if !CONFIG_MODULES.
1639 [HW,JOY] Multisystem joystick and NES/SNES/PSX pad
1640 support via parallel port (up to 5 devices per port)
1641 Format: <port#>,<pad1>,<pad2>,<pad3>,<pad4>,<pad5>
1642 See also Documentation/input/devices/joystick-parport.rst
1646 gart_fix_e820= [X86-64] disable the fix e820 for K8 GART
1650 gather_data_sampling=
1651 [X86,INTEL] Control the Gather Data Sampling (GDS)
1654 Gather Data Sampling is a hardware vulnerability which
1655 allows unprivileged speculative access to data which was
1656 previously stored in vector registers.
1658 This issue is mitigated by default in updated microcode.
1659 The mitigation may have a performance impact but can be
1660 disabled. On systems without the microcode mitigation
1661 disabling AVX serves as a mitigation.
1663 force: Disable AVX to mitigate systems without
1664 microcode mitigation. No effect if the microcode
1665 mitigation is present. Known to cause crashes in
1666 userspace with buggy AVX enumeration.
1668 off: Disable GDS mitigation.
1670 gcov_persist= [GCOV] When non-zero (default), profiling data for
1671 kernel modules is saved and remains accessible via
1672 debugfs, even when the module is unloaded/reloaded.
1673 When zero, profiling data is discarded and associated
1674 debugfs files are removed at module unload time.
1676 goldfish [X86] Enable the goldfish android emulator platform.
1677 Don't use this when you are not running on the
1680 gpio-mockup.gpio_mockup_ranges
1681 [HW] Sets the ranges of gpiochip of for this device.
1682 Format: <start1>,<end1>,<start2>,<end2>...
1683 gpio-mockup.gpio_mockup_named_lines
1684 [HW] Let the driver know GPIO lines should be named.
1686 gpt [EFI] Forces disk with valid GPT signature but
1687 invalid Protective MBR to be treated as GPT. If the
1688 primary GPT is corrupted, it enables the backup/alternate
1689 GPT to be used instead.
1691 grcan.enable0= [HW] Configuration of physical interface 0. Determines
1692 the "Enable 0" bit of the configuration register.
1695 grcan.enable1= [HW] Configuration of physical interface 1. Determines
1696 the "Enable 0" bit of the configuration register.
1699 grcan.select= [HW] Select which physical interface to use.
1702 grcan.txsize= [HW] Sets the size of the tx buffer.
1703 Format: <unsigned int> such that (txsize & ~0x1fffc0) == 0.
1705 grcan.rxsize= [HW] Sets the size of the rx buffer.
1706 Format: <unsigned int> such that (rxsize & ~0x1fffc0) == 0.
1710 [KNL] Under CONFIG_HARDENED_USERCOPY, whether
1711 hardening is enabled for this boot. Hardened
1712 usercopy checking is used to protect the kernel
1713 from reading or writing beyond known memory
1714 allocation boundaries as a proactive defense
1715 against bounds-checking flaws in the kernel's
1716 copy_to_user()/copy_from_user() interface.
1717 on Perform hardened usercopy checks (default).
1718 off Disable hardened usercopy checks.
1720 hardlockup_all_cpu_backtrace=
1721 [KNL] Should the hard-lockup detector generate
1722 backtraces on all cpus.
1725 hashdist= [KNL,NUMA] Large hashes allocated during boot
1726 are distributed across NUMA nodes. Defaults on
1727 for 64-bit NUMA, off otherwise.
1728 Format: 0 | 1 (for off | on)
1730 hcl= [IA-64] SGI's Hardware Graph compatibility layer
1732 hd= [EIDE] (E)IDE hard drive subsystem geometry
1733 Format: <cyl>,<head>,<sect>
1736 Disable Hardware Error Source Table (HEST) support;
1737 corresponding firmware-first mode error processing
1738 logic will be disabled.
1740 hibernate= [HIBERNATION]
1741 noresume Don't check if there's a hibernation image
1742 present during boot.
1743 nocompress Don't compress/decompress hibernation images.
1744 no Disable hibernation and resume.
1745 protect_image Turn on image protection during restoration
1746 (that will set all pages holding image data
1747 during restoration read-only).
1749 highmem=nn[KMG] [KNL,BOOT] forces the highmem zone to have an exact
1750 size of <nn>. This works even on boxes that have no
1751 highmem otherwise. This also works to reduce highmem
1752 size on bigger boxes.
1754 highres= [KNL] Enable/disable high resolution timer mode.
1755 Valid parameters: "on", "off"
1760 hostname= [KNL] Set the hostname (aka UTS nodename).
1762 This allows setting the system's hostname during early
1763 startup. This sets the name returned by gethostname.
1764 Using this parameter to set the hostname makes it
1765 possible to ensure the hostname is correctly set before
1766 any userspace processes run, avoiding the possibility
1767 that a process may call gethostname before the hostname
1768 has been explicitly set, resulting in the calling
1769 process getting an incorrect result. The string must
1770 not exceed the maximum allowed hostname length (usually
1771 64 characters) and will be truncated otherwise.
1773 hpet= [X86-32,HPET] option to control HPET usage
1774 Format: { enable (default) | disable | force |
1776 disable: disable HPET and use PIT instead
1777 force: allow force enabled of undocumented chips (ICH4,
1779 verbose: show contents of HPET registers during setup
1781 hpet_mmap= [X86, HPET_MMAP] Allow userspace to mmap HPET
1782 registers. Default set by CONFIG_HPET_MMAP_DEFAULT.
1784 hugepages= [HW] Number of HugeTLB pages to allocate at boot.
1785 If this follows hugepagesz (below), it specifies
1786 the number of pages of hugepagesz to be allocated.
1787 If this is the first HugeTLB parameter on the command
1788 line, it specifies the number of pages to allocate for
1789 the default huge page size. If using node format, the
1790 number of pages to allocate per-node can be specified.
1791 See also Documentation/admin-guide/mm/hugetlbpage.rst.
1792 Format: <integer> or (node format)
1793 <node>:<integer>[,<node>:<integer>]
1796 [HW] The size of the HugeTLB pages. This is used in
1797 conjunction with hugepages (above) to allocate huge
1798 pages of a specific size at boot. The pair
1799 hugepagesz=X hugepages=Y can be specified once for
1800 each supported huge page size. Huge page sizes are
1801 architecture dependent. See also
1802 Documentation/admin-guide/mm/hugetlbpage.rst.
1805 hugetlb_cma= [HW,CMA] The size of a CMA area used for allocation
1806 of gigantic hugepages. Or using node format, the size
1807 of a CMA area per node can be specified.
1808 Format: nn[KMGTPE] or (node format)
1809 <node>:nn[KMGTPE][,<node>:nn[KMGTPE]]
1811 Reserve a CMA area of given size and allocate gigantic
1812 hugepages using the CMA allocator. If enabled, the
1813 boot-time allocation of gigantic hugepages is skipped.
1815 hugetlb_free_vmemmap=
1816 [KNL] Requires CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP
1818 Control if HugeTLB Vmemmap Optimization (HVO) is enabled.
1819 Allows heavy hugetlb users to free up some more
1820 memory (7 * PAGE_SIZE for each 2MB hugetlb page).
1821 Format: { on | off (default) }
1826 Built with CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP_DEFAULT_ON=y,
1829 Note that the vmemmap pages may be allocated from the added
1830 memory block itself when memory_hotplug.memmap_on_memory is
1831 enabled, those vmemmap pages cannot be optimized even if this
1832 feature is enabled. Other vmemmap pages not allocated from
1833 the added memory block itself do not be affected.
1836 [KNL] Should the hung task detector generate panics.
1839 A value of 1 instructs the kernel to panic when a
1840 hung task is detected. The default value is controlled
1841 by the CONFIG_BOOTPARAM_HUNG_TASK_PANIC build-time
1842 option. The value selected by this boot parameter can
1843 be changed later by the kernel.hung_task_panic sysctl.
1845 hvc_iucv= [S390] Number of z/VM IUCV hypervisor console (HVC)
1846 terminal devices. Valid values: 0..8
1847 hvc_iucv_allow= [S390] Comma-separated list of z/VM user IDs.
1848 If specified, z/VM IUCV HVC accepts connections
1849 from listed z/VM user IDs only.
1851 hv_nopvspin [X86,HYPER_V] Disables the paravirt spinlock optimizations
1852 which allow the hypervisor to 'idle' the
1853 guest on lock contention.
1855 i2c_bus= [HW] Override the default board specific I2C bus speed
1856 or register an additional I2C bus that is not
1857 registered from board initialization code.
1861 i8042.debug [HW] Toggle i8042 debug mode
1862 i8042.unmask_kbd_data
1863 [HW] Enable printing of interrupt data from the KBD port
1864 (disabled by default, and as a pre-condition
1865 requires that i8042.debug=1 be enabled)
1866 i8042.direct [HW] Put keyboard port into non-translated mode
1867 i8042.dumbkbd [HW] Pretend that controller can only read data from
1868 keyboard and cannot control its state
1869 (Don't attempt to blink the leds)
1870 i8042.noaux [HW] Don't check for auxiliary (== mouse) port
1871 i8042.nokbd [HW] Don't check/create keyboard port
1872 i8042.noloop [HW] Disable the AUX Loopback command while probing
1874 i8042.nomux [HW] Don't check presence of an active multiplexing
1876 i8042.nopnp [HW] Don't use ACPIPnP / PnPBIOS to discover KBD/AUX
1878 i8042.notimeout [HW] Ignore timeout condition signalled by controller
1879 i8042.reset [HW] Reset the controller during init, cleanup and
1880 suspend-to-ram transitions, only during s2r
1881 transitions, or never reset
1882 Format: { 1 | Y | y | 0 | N | n }
1883 1, Y, y: always reset controller
1884 0, N, n: don't ever reset controller
1885 Default: only on s2r transitions on x86; most other
1886 architectures force reset to be always executed
1887 i8042.unlock [HW] Unlock (ignore) the keylock
1888 i8042.kbdreset [HW] Reset device connected to KBD port
1890 [HW] Allow deferred probing upon i8042 probe errors
1894 i915.invert_brightness=
1895 [DRM] Invert the sense of the variable that is used to
1896 set the brightness of the panel backlight. Normally a
1897 brightness value of 0 indicates backlight switched off,
1898 and the maximum of the brightness value sets the backlight
1899 to maximum brightness. If this parameter is set to 0
1900 (default) and the machine requires it, or this parameter
1901 is set to 1, a brightness value of 0 sets the backlight
1902 to maximum brightness, and the maximum of the brightness
1903 value switches the backlight off.
1904 -1 -- never invert brightness
1905 0 -- machine default
1906 1 -- force brightness inversion
1908 ia32_emulation= [X86-64]
1910 When true, allows loading 32-bit programs and executing 32-bit
1911 syscalls, essentially overriding IA32_EMULATION_DEFAULT_DISABLED at
1912 boot time. When false, unconditionally disables IA32 emulation.
1915 Format: <io>[,<membase>[,<icn_id>[,<icn_id2>]]]
1919 Format: idle=poll, idle=halt, idle=nomwait
1920 Poll forces a polling idle loop that can slightly
1921 improve the performance of waking up a idle CPU, but
1922 will use a lot of power and make the system run hot.
1924 idle=halt: Halt is forced to be used for CPU idle.
1925 In such case C2/C3 won't be used again.
1926 idle=nomwait: Disable mwait for CPU C-states
1930 Allow force disabling of Shared Virtual Memory (SVA)
1931 support for the idxd driver. By default it is set to
1934 idxd.tc_override= [HW]
1936 Allow override of default traffic class configuration
1937 for the device. By default it is set to false (0).
1939 ieee754= [MIPS] Select IEEE Std 754 conformance mode
1940 Format: { strict | legacy | 2008 | relaxed }
1943 Choose which programs will be accepted for execution
1944 based on the IEEE 754 NaN encoding(s) supported by
1945 the FPU and the NaN encoding requested with the value
1946 of an ELF file header flag individually set by each
1947 binary. Hardware implementations are permitted to
1948 support either or both of the legacy and the 2008 NaN
1951 Available settings are as follows:
1952 strict accept binaries that request a NaN encoding
1953 supported by the FPU
1954 legacy only accept legacy-NaN binaries, if supported
1956 2008 only accept 2008-NaN binaries, if supported
1958 relaxed accept any binaries regardless of whether
1959 supported by the FPU
1961 The FPU emulator is always able to support both NaN
1962 encodings, so if no FPU hardware is present or it has
1963 been disabled with 'nofpu', then the settings of
1964 'legacy' and '2008' strap the emulator accordingly,
1965 'relaxed' straps the emulator for both legacy-NaN and
1966 2008-NaN, whereas 'strict' enables legacy-NaN only on
1967 legacy processors and both NaN encodings on MIPS32 or
1970 The setting for ABS.fmt/NEG.fmt instruction execution
1971 mode generally follows that for the NaN encoding,
1972 except where unsupported by hardware.
1974 ignore_loglevel [KNL]
1975 Ignore loglevel setting - this will print /all/
1976 kernel messages to the console. Useful for debugging.
1977 We also add it as printk module parameter, so users
1978 could change it dynamically, usually by
1979 /sys/module/printk/parameters/ignore_loglevel.
1982 Ignore RLIMIT_DATA setting for data mappings,
1983 print warning at first misuse. Can be changed via
1984 /sys/module/kernel/parameters/ignore_rlimit_data.
1986 ihash_entries= [KNL]
1987 Set number of hash buckets for inode cache.
1989 ima_appraise= [IMA] appraise integrity measurements
1990 Format: { "off" | "enforce" | "fix" | "log" }
1993 ima_appraise_tcb [IMA] Deprecated. Use ima_policy= instead.
1994 The builtin appraise policy appraises all files
1997 ima_canonical_fmt [IMA]
1998 Use the canonical format for the binary runtime
1999 measurements, instead of host native format.
2002 Format: { md5 | sha1 | rmd160 | sha256 | sha384
2006 The list of supported hash algorithms is defined
2007 in crypto/hash_info.h.
2010 The builtin policies to load during IMA setup.
2011 Format: "tcb | appraise_tcb | secure_boot |
2012 fail_securely | critical_data"
2014 The "tcb" policy measures all programs exec'd, files
2015 mmap'd for exec, and all files opened with the read
2016 mode bit set by either the effective uid (euid=0) or
2019 The "appraise_tcb" policy appraises the integrity of
2020 all files owned by root.
2022 The "secure_boot" policy appraises the integrity
2023 of files (eg. kexec kernel image, kernel modules,
2024 firmware, policy, etc) based on file signatures.
2026 The "fail_securely" policy forces file signature
2027 verification failure also on privileged mounted
2028 filesystems with the SB_I_UNVERIFIABLE_SIGNATURE
2031 The "critical_data" policy measures kernel integrity
2034 ima_tcb [IMA] Deprecated. Use ima_policy= instead.
2035 Load a policy which meets the needs of the Trusted
2036 Computing Base. This means IMA will measure all
2037 programs exec'd, files mmap'd for exec, and all files
2038 opened for read by uid=0.
2041 Select one of defined IMA measurements template formats.
2042 Formats: { "ima" | "ima-ng" | "ima-ngv2" | "ima-sig" |
2047 [IMA] Define a custom template format.
2048 Format: { "field1|...|fieldN" }
2050 ima.ahash_minsize= [IMA] Minimum file size for asynchronous hash usage
2051 Format: <min_file_size>
2052 Set the minimal file size for using asynchronous hash.
2053 If left unspecified, ahash usage is disabled.
2055 ahash performance varies for different data sizes on
2056 different crypto accelerators. This option can be used
2057 to achieve the best performance for a particular HW.
2059 ima.ahash_bufsize= [IMA] Asynchronous hash buffer size
2061 Set hashing buffer size. Default: 4k.
2063 ahash performance varies for different chunk sizes on
2064 different crypto accelerators. This option can be used
2065 to achieve best performance for particular HW.
2069 Run specified binary instead of /sbin/init as init
2072 initcall_debug [KNL] Trace initcalls as they are executed. Useful
2073 for working out where the kernel is dying during
2076 initcall_blacklist= [KNL] Do not execute a comma-separated list of
2077 initcall functions. Useful for debugging built-in
2078 modules and initcalls.
2080 initramfs_async= [KNL]
2083 This parameter controls whether the initramfs
2084 image is unpacked asynchronously, concurrently
2085 with devices being probed and
2086 initialized. This should normally just work,
2087 but as a debugging aid, one can get the
2088 historical behaviour of the initramfs
2089 unpacking being completed before device_ and
2092 initrd= [BOOT] Specify the location of the initial ramdisk
2094 initrdmem= [KNL] Specify a physical address and size from which to
2095 load the initrd. If an initrd is compiled in or
2096 specified in the bootparams, it takes priority over this
2098 Format: ss[KMG],nn[KMG]
2101 init_on_alloc= [MM] Fill newly allocated pages and heap objects with
2104 Default set by CONFIG_INIT_ON_ALLOC_DEFAULT_ON.
2106 init_on_free= [MM] Fill freed pages and heap objects with zeroes.
2108 Default set by CONFIG_INIT_ON_FREE_DEFAULT_ON.
2110 init_pkru= [X86] Specify the default memory protection keys rights
2111 register contents for all processes. 0x55555554 by
2112 default (disallow access to all but pkey 0). Can
2113 override in debugfs after boot.
2115 inport.irq= [HW] Inport (ATI XL and Microsoft) busmouse driver
2118 int_pln_enable [X86] Enable power limit notification interrupt
2120 integrity_audit=[IMA]
2121 Format: { "0" | "1" }
2122 0 -- basic integrity auditing messages. (Default)
2123 1 -- additional integrity auditing messages.
2125 intel_iommu= [DMAR] Intel IOMMU driver (DMAR) option
2127 Enable intel iommu driver.
2129 Disable intel iommu driver.
2130 igfx_off [Default Off]
2131 By default, gfx is mapped as normal device. If a gfx
2132 device has a dedicated DMAR unit, the DMAR unit is
2133 bypassed by not enabling DMAR with this option. In
2134 this case, gfx device will use physical address for
2136 strict [Default Off]
2137 Deprecated, equivalent to iommu.strict=1.
2138 sp_off [Default Off]
2139 By default, super page will be supported if Intel IOMMU
2140 has the capability. With this option, super page will
2143 Enable the Intel IOMMU scalable mode if the hardware
2144 advertises that it has support for the scalable mode
2147 Disallow use of the Intel IOMMU scalable mode.
2148 tboot_noforce [Default Off]
2149 Do not force the Intel IOMMU enabled under tboot.
2150 By default, tboot will force Intel IOMMU on, which
2151 could harm performance of some high-throughput
2152 devices like 40GBit network cards, even if identity
2154 Note that using this option lowers the security
2155 provided by tboot because it makes the system
2156 vulnerable to DMA attacks.
2158 intel_idle.max_cstate= [KNL,HW,ACPI,X86]
2159 0 disables intel_idle and fall back on acpi_idle.
2160 1 to 9 specify maximum depth of C-state.
2164 Do not enable intel_pstate as the default
2165 scaling driver for the supported processors
2167 Use intel_pstate driver to bypass the scaling
2168 governors layer of cpufreq and provides it own
2169 algorithms for p-state selection. There are two
2170 P-state selection algorithms provided by
2171 intel_pstate in the active mode: powersave and
2172 performance. The way they both operate depends
2173 on whether or not the hardware managed P-states
2174 (HWP) feature has been enabled in the processor
2175 and possibly on the processor model.
2177 Use intel_pstate as a scaling driver, but configure it
2178 to work with generic cpufreq governors (instead of
2179 enabling its internal governor). This mode cannot be
2180 used along with the hardware-managed P-states (HWP)
2183 Enable intel_pstate on systems that prohibit it by default
2184 in favor of acpi-cpufreq. Forcing the intel_pstate driver
2185 instead of acpi-cpufreq may disable platform features, such
2186 as thermal controls and power capping, that rely on ACPI
2187 P-States information being indicated to OSPM and therefore
2188 should be used with caution. This option does not work with
2189 processors that aren't supported by the intel_pstate driver
2190 or on platforms that use pcc-cpufreq instead of acpi-cpufreq.
2192 Do not enable hardware P state control (HWP)
2195 Only load intel_pstate on systems which support
2196 hardware P state control (HWP) if available.
2198 Enforce ACPI _PPC performance limits. If the Fixed ACPI
2199 Description Table, specifies preferred power management
2200 profile as "Enterprise Server" or "Performance Server",
2201 then this feature is turned on by default.
2203 Allow per-logical-CPU P-State performance control limits using
2204 cpufreq sysfs interface
2206 intremap= [X86-64, Intel-IOMMU]
2207 on enable Interrupt Remapping (default)
2208 off disable Interrupt Remapping
2209 nosid disable Source ID checking
2211 BIOS x2APIC opt-out request will be ignored
2212 nopost disable Interrupt Posting
2214 iomem= Disable strict checking of access to MMIO memory
2215 strict regions from userspace.
2230 nobypass [PPC/POWERNV]
2231 Disable IOMMU bypass, using IOMMU for PCI devices.
2233 iommu.forcedac= [ARM64, X86] Control IOVA allocation for PCI devices.
2234 Format: { "0" | "1" }
2235 0 - Try to allocate a 32-bit DMA address first, before
2236 falling back to the full range if needed.
2237 1 - Allocate directly from the full usable range,
2238 forcing Dual Address Cycle for PCI cards supporting
2239 greater than 32-bit addressing.
2241 iommu.strict= [ARM64, X86, S390] Configure TLB invalidation behaviour
2242 Format: { "0" | "1" }
2244 Request that DMA unmap operations use deferred
2245 invalidation of hardware TLBs, for increased
2246 throughput at the cost of reduced device isolation.
2247 Will fall back to strict mode if not supported by
2248 the relevant IOMMU driver.
2250 DMA unmap operations invalidate IOMMU hardware TLBs
2252 unset - Use value of CONFIG_IOMMU_DEFAULT_DMA_{LAZY,STRICT}.
2253 Note: on x86, strict mode specified via one of the
2254 legacy driver-specific options takes precedence.
2257 [ARM64, X86] Configure DMA to bypass the IOMMU by default.
2258 Format: { "0" | "1" }
2259 0 - Use IOMMU translation for DMA.
2260 1 - Bypass the IOMMU for DMA.
2261 unset - Use value of CONFIG_IOMMU_DEFAULT_PASSTHROUGH.
2263 io7= [HW] IO7 for Marvel-based Alpha systems
2264 See comment before marvel_specify_io7 in
2265 arch/alpha/kernel/core_marvel.c.
2267 io_delay= [X86] I/O delay method
2269 Standard port 0x80 based delay
2271 Alternate port 0xed based delay (needed on some systems)
2273 Simple two microseconds delay
2278 See Documentation/admin-guide/nfs/nfsroot.rst.
2280 ipcmni_extend [KNL] Extend the maximum number of unique System V
2281 IPC identifiers from 32,768 to 16,777,216.
2283 irqaffinity= [SMP] Set the default irq affinity mask
2284 The argument is a cpu list, as described above.
2286 irqchip.gicv2_force_probe=
2289 Force the kernel to look for the second 4kB page
2290 of a GICv2 controller even if the memory range
2291 exposed by the device tree is too small.
2293 irqchip.gicv3_nolpi=
2295 Force the kernel to ignore the availability of
2296 LPIs (and by consequence ITSs). Intended for system
2297 that use the kernel as a bootloader, and thus want
2298 to let secondary kernels in charge of setting up
2301 irqchip.gicv3_pseudo_nmi= [ARM64]
2302 Enables support for pseudo-NMIs in the kernel. This
2303 requires the kernel to be built with
2304 CONFIG_ARM64_PSEUDO_NMI.
2307 When an interrupt is not handled search all handlers
2308 for it. Intended to get systems with badly broken
2312 When an interrupt is not handled search all handlers
2313 for it. Also check all handlers each timer
2314 interrupt. Intended to get systems with badly broken
2318 Format: <RDP>,<reset>,<pci_scan>,<verbosity>
2320 isolcpus= [KNL,SMP,ISOL] Isolate a given set of CPUs from disturbance.
2321 [Deprecated - use cpusets instead]
2322 Format: [flag-list,]<cpu-list>
2324 Specify one or more CPUs to isolate from disturbances
2325 specified in the flag list (default: domain):
2328 Disable the tick when a single task runs.
2330 A residual 1Hz tick is offloaded to workqueues, which you
2331 need to affine to housekeeping through the global
2332 workqueue's affinity configured via the
2333 /sys/devices/virtual/workqueue/cpumask sysfs file, or
2334 by using the 'domain' flag described below.
2336 NOTE: by default the global workqueue runs on all CPUs,
2337 so to protect individual CPUs the 'cpumask' file has to
2338 be configured manually after bootup.
2341 Isolate from the general SMP balancing and scheduling
2342 algorithms. Note that performing domain isolation this way
2343 is irreversible: it's not possible to bring back a CPU to
2344 the domains once isolated through isolcpus. It's strongly
2345 advised to use cpusets instead to disable scheduler load
2346 balancing through the "cpuset.sched_load_balance" file.
2347 It offers a much more flexible interface where CPUs can
2348 move in and out of an isolated set anytime.
2350 You can move a process onto or off an "isolated" CPU via
2351 the CPU affinity syscalls or cpuset.
2352 <cpu number> begins at 0 and the maximum value is
2353 "number of CPUs in system - 1".
2357 Isolate from being targeted by managed interrupts
2358 which have an interrupt mask containing isolated
2359 CPUs. The affinity of managed interrupts is
2360 handled by the kernel and cannot be changed via
2361 the /proc/irq/* interfaces.
2363 This isolation is best effort and only effective
2364 if the automatically assigned interrupt mask of a
2365 device queue contains isolated and housekeeping
2366 CPUs. If housekeeping CPUs are online then such
2367 interrupts are directed to the housekeeping CPU
2368 so that IO submitted on the housekeeping CPU
2369 cannot disturb the isolated CPU.
2371 If a queue's affinity mask contains only isolated
2372 CPUs then this parameter has no effect on the
2373 interrupt routing decision, though interrupts are
2374 only delivered when tasks running on those
2375 isolated CPUs submit IO. IO submitted on
2376 housekeeping CPUs has no influence on those
2379 The format of <cpu-list> is described above.
2383 ivrs_ioapic [HW,X86-64]
2384 Provide an override to the IOAPIC-ID<->DEVICE-ID
2385 mapping provided in the IVRS ACPI table.
2386 By default, PCI segment is 0, and can be omitted.
2388 For example, to map IOAPIC-ID decimal 10 to
2389 PCI segment 0x1 and PCI device 00:14.0,
2390 write the parameter as:
2391 ivrs_ioapic=10@0001:00:14.0
2394 * To map IOAPIC-ID decimal 10 to PCI device 00:14.0
2395 write the parameter as:
2396 ivrs_ioapic[10]=00:14.0
2397 * To map IOAPIC-ID decimal 10 to PCI segment 0x1 and
2398 PCI device 00:14.0 write the parameter as:
2399 ivrs_ioapic[10]=0001:00:14.0
2401 ivrs_hpet [HW,X86-64]
2402 Provide an override to the HPET-ID<->DEVICE-ID
2403 mapping provided in the IVRS ACPI table.
2404 By default, PCI segment is 0, and can be omitted.
2406 For example, to map HPET-ID decimal 10 to
2407 PCI segment 0x1 and PCI device 00:14.0,
2408 write the parameter as:
2409 ivrs_hpet=10@0001:00:14.0
2412 * To map HPET-ID decimal 0 to PCI device 00:14.0
2413 write the parameter as:
2414 ivrs_hpet[0]=00:14.0
2415 * To map HPET-ID decimal 10 to PCI segment 0x1 and
2416 PCI device 00:14.0 write the parameter as:
2417 ivrs_ioapic[10]=0001:00:14.0
2419 ivrs_acpihid [HW,X86-64]
2420 Provide an override to the ACPI-HID:UID<->DEVICE-ID
2421 mapping provided in the IVRS ACPI table.
2422 By default, PCI segment is 0, and can be omitted.
2424 For example, to map UART-HID:UID AMD0020:0 to
2425 PCI segment 0x1 and PCI device ID 00:14.5,
2426 write the parameter as:
2427 ivrs_acpihid=AMD0020:0@0001:00:14.5
2430 * To map UART-HID:UID AMD0020:0 to PCI segment is 0,
2431 PCI device ID 00:14.5, write the parameter as:
2432 ivrs_acpihid[00:14.5]=AMD0020:0
2433 * To map UART-HID:UID AMD0020:0 to PCI segment 0x1 and
2434 PCI device ID 00:14.5, write the parameter as:
2435 ivrs_acpihid[0001:00:14.5]=AMD0020:0
2437 js= [HW,JOY] Analog joystick
2438 See Documentation/input/joydev/joystick.rst.
2441 [KNL] Enforce KASAN (Kernel Address Sanitizer) to print
2442 report on every invalid memory access. Without this
2443 parameter KASAN will print report only for the first
2447 Do not unregister boot console at start. This is only
2448 useful for debugging when something happens in the window
2449 between unregistering the boot console and initializing
2452 keepinitrd [HW,ARM] See retain_initrd.
2454 kernelcore= [KNL,X86,IA-64,PPC]
2455 Format: nn[KMGTPE] | nn% | "mirror"
2456 This parameter specifies the amount of memory usable by
2457 the kernel for non-movable allocations. The requested
2458 amount is spread evenly throughout all nodes in the
2459 system as ZONE_NORMAL. The remaining memory is used for
2460 movable memory in its own zone, ZONE_MOVABLE. In the
2461 event, a node is too small to have both ZONE_NORMAL and
2462 ZONE_MOVABLE, kernelcore memory will take priority and
2463 other nodes will have a larger ZONE_MOVABLE.
2465 ZONE_MOVABLE is used for the allocation of pages that
2466 may be reclaimed or moved by the page migration
2467 subsystem. Note that allocations like PTEs-from-HighMem
2468 still use the HighMem zone if it exists, and the Normal
2469 zone if it does not.
2471 It is possible to specify the exact amount of memory in
2472 the form of "nn[KMGTPE]", a percentage of total system
2473 memory in the form of "nn%", or "mirror". If "mirror"
2474 option is specified, mirrored (reliable) memory is used
2475 for non-movable allocations and remaining memory is used
2476 for Movable pages. "nn[KMGTPE]", "nn%", and "mirror"
2477 are exclusive, so you cannot specify multiple forms.
2479 kgdbdbgp= [KGDB,HW] kgdb over EHCI usb debug port.
2480 Format: <Controller#>[,poll interval]
2481 The controller # is the number of the ehci usb debug
2482 port as it is probed via PCI. The poll interval is
2483 optional and is the number seconds in between
2484 each poll cycle to the debug port in case you need
2485 the functionality for interrupting the kernel with
2486 gdb or control-c on the dbgp connection. When
2487 not using this parameter you use sysrq-g to break into
2488 the kernel debugger.
2490 kgdboc= [KGDB,HW] kgdb over consoles.
2491 Requires a tty driver that supports console polling,
2492 or a supported polling keyboard driver (non-usb).
2493 Serial only format: <serial_device>[,baud]
2494 keyboard only format: kbd
2495 keyboard and serial format: kbd,<serial_device>[,baud]
2496 Optional Kernel mode setting:
2497 kms, kbd format: kms,kbd
2498 kms, kbd and serial format: kms,kbd,<ser_dev>[,baud]
2500 kgdboc_earlycon= [KGDB,HW]
2501 If the boot console provides the ability to read
2502 characters and can work in polling mode, you can use
2503 this parameter to tell kgdb to use it as a backend
2504 until the normal console is registered. Intended to
2505 be used together with the kgdboc parameter which
2506 specifies the normal console to transition to.
2508 The name of the early console should be specified
2509 as the value of this parameter. Note that the name of
2510 the early console might be different than the tty
2511 name passed to kgdboc. It's OK to leave the value
2512 blank and the first boot console that implements
2513 read() will be picked.
2515 kgdbwait [KGDB] Stop kernel execution and enter the
2516 kernel debugger at the earliest opportunity.
2518 kmac= [MIPS] Korina ethernet MAC address.
2519 Configure the RouterBoard 532 series on-chip
2520 Ethernet adapter MAC address.
2522 kmemleak= [KNL] Boot-time kmemleak enable/disable
2523 Valid arguments: on, off
2525 Built with CONFIG_DEBUG_KMEMLEAK_DEFAULT_OFF=y,
2528 kprobe_event=[probe-list]
2529 [FTRACE] Add kprobe events and enable at boot time.
2530 The probe-list is a semicolon delimited list of probe
2531 definitions. Each definition is same as kprobe_events
2532 interface, but the parameters are comma delimited.
2533 For example, to add a kprobe event on vfs_read with
2534 arg1 and arg2, add to the command line;
2536 kprobe_event=p,vfs_read,$arg1,$arg2
2538 See also Documentation/trace/kprobetrace.rst "Kernel
2539 Boot Parameter" section.
2541 kpti= [ARM64] Control page table isolation of user
2542 and kernel address spaces.
2543 Default: enabled on cores which need mitigation.
2547 kunit.enable= [KUNIT] Enable executing KUnit tests. Requires
2548 CONFIG_KUNIT to be set to be fully enabled. The
2549 default value can be overridden via
2550 KUNIT_DEFAULT_ENABLED.
2551 Default is 1 (enabled)
2553 kvm.ignore_msrs=[KVM] Ignore guest accesses to unhandled MSRs.
2554 Default is 0 (don't ignore, but inject #GP)
2556 kvm.eager_page_split=
2557 [KVM,X86] Controls whether or not KVM will try to
2558 proactively split all huge pages during dirty logging.
2559 Eager page splitting reduces interruptions to vCPU
2560 execution by eliminating the write-protection faults
2561 and MMU lock contention that would otherwise be
2562 required to split huge pages lazily.
2564 VM workloads that rarely perform writes or that write
2565 only to a small region of VM memory may benefit from
2566 disabling eager page splitting to allow huge pages to
2567 still be used for reads.
2569 The behavior of eager page splitting depends on whether
2570 KVM_DIRTY_LOG_INITIALLY_SET is enabled or disabled. If
2571 disabled, all huge pages in a memslot will be eagerly
2572 split when dirty logging is enabled on that memslot. If
2573 enabled, eager page splitting will be performed during
2574 the KVM_CLEAR_DIRTY ioctl, and only for the pages being
2577 Eager page splitting is only supported when kvm.tdp_mmu=Y.
2581 kvm.enable_vmware_backdoor=[KVM] Support VMware backdoor PV interface.
2582 Default is false (don't support).
2585 [KVM] Controls the software workaround for the
2586 X86_BUG_ITLB_MULTIHIT bug.
2587 force : Always deploy workaround.
2588 off : Never deploy workaround.
2589 auto : Deploy workaround based on the presence of
2590 X86_BUG_ITLB_MULTIHIT.
2594 If the software workaround is enabled for the host,
2595 guests do need not to enable it for nested guests.
2597 kvm.nx_huge_pages_recovery_ratio=
2598 [KVM] Controls how many 4KiB pages are periodically zapped
2599 back to huge pages. 0 disables the recovery, otherwise if
2600 the value is N KVM will zap 1/Nth of the 4KiB pages every
2601 period (see below). The default is 60.
2603 kvm.nx_huge_pages_recovery_period_ms=
2604 [KVM] Controls the time period at which KVM zaps 4KiB pages
2605 back to huge pages. If the value is a non-zero N, KVM will
2606 zap a portion (see ratio above) of the pages every N msecs.
2607 If the value is 0 (the default), KVM will pick a period based
2608 on the ratio, such that a page is zapped after 1 hour on average.
2610 kvm-amd.nested= [KVM,AMD] Control nested virtualization feature in
2611 KVM/SVM. Default is 1 (enabled).
2613 kvm-amd.npt= [KVM,AMD] Control KVM's use of Nested Page Tables,
2614 a.k.a. Two-Dimensional Page Tables. Default is 1
2615 (enabled). Disable by KVM if hardware lacks support
2619 [KVM,ARM] Select one of KVM/arm64's modes of operation.
2621 none: Forcefully disable KVM.
2623 nvhe: Standard nVHE-based mode, without support for
2626 protected: nVHE-based mode with support for guests whose
2627 state is kept private from the host.
2629 nested: VHE-based mode with support for nested
2630 virtualization. Requires at least ARMv8.3
2633 Defaults to VHE/nVHE based on hardware support. Setting
2634 mode to "protected" will disable kexec and hibernation
2635 for the host. "nested" is experimental and should be
2636 used with extreme caution.
2638 kvm-arm.vgic_v3_group0_trap=
2639 [KVM,ARM] Trap guest accesses to GICv3 group-0
2642 kvm-arm.vgic_v3_group1_trap=
2643 [KVM,ARM] Trap guest accesses to GICv3 group-1
2646 kvm-arm.vgic_v3_common_trap=
2647 [KVM,ARM] Trap guest accesses to GICv3 common
2650 kvm-arm.vgic_v4_enable=
2651 [KVM,ARM] Allow use of GICv4 for direct injection of
2654 kvm_cma_resv_ratio=n [PPC]
2655 Reserves given percentage from system memory area for
2656 contiguous memory allocation for KVM hash pagetable
2658 By default it reserves 5% of total system memory.
2662 kvm-intel.ept= [KVM,Intel] Control KVM's use of Extended Page Tables,
2663 a.k.a. Two-Dimensional Page Tables. Default is 1
2664 (enabled). Disable by KVM if hardware lacks support
2667 kvm-intel.emulate_invalid_guest_state=
2668 [KVM,Intel] Control whether to emulate invalid guest
2669 state. Ignored if kvm-intel.enable_unrestricted_guest=1,
2670 as guest state is never invalid for unrestricted
2671 guests. This param doesn't apply to nested guests (L2),
2672 as KVM never emulates invalid L2 guest state.
2673 Default is 1 (enabled).
2675 kvm-intel.flexpriority=
2676 [KVM,Intel] Control KVM's use of FlexPriority feature
2677 (TPR shadow). Default is 1 (enabled). Disable by KVM if
2678 hardware lacks support for it.
2681 [KVM,Intel] Control nested virtualization feature in
2682 KVM/VMX. Default is 1 (enabled).
2684 kvm-intel.unrestricted_guest=
2685 [KVM,Intel] Control KVM's use of unrestricted guest
2686 feature (virtualized real and unpaged mode). Default
2687 is 1 (enabled). Disable by KVM if EPT is disabled or
2688 hardware lacks support for it.
2690 kvm-intel.vmentry_l1d_flush=[KVM,Intel] Mitigation for L1 Terminal Fault
2693 Valid arguments: never, cond, always
2695 always: L1D cache flush on every VMENTER.
2696 cond: Flush L1D on VMENTER only when the code between
2697 VMEXIT and VMENTER can leak host memory.
2698 never: Disables the mitigation
2700 Default is cond (do L1 cache flush in specific instances)
2702 kvm-intel.vpid= [KVM,Intel] Control KVM's use of Virtual Processor
2703 Identification feature (tagged TLBs). Default is 1
2704 (enabled). Disable by KVM if hardware lacks support
2707 l1d_flush= [X86,INTEL]
2708 Control mitigation for L1D based snooping vulnerability.
2710 Certain CPUs are vulnerable to an exploit against CPU
2711 internal buffers which can forward information to a
2712 disclosure gadget under certain conditions.
2714 In vulnerable processors, the speculatively
2715 forwarded data can be used in a cache side channel
2716 attack, to access data to which the attacker does
2717 not have direct access.
2719 This parameter controls the mitigation. The
2722 on - enable the interface for the mitigation
2724 l1tf= [X86] Control mitigation of the L1TF vulnerability on
2727 The kernel PTE inversion protection is unconditionally
2728 enabled and cannot be disabled.
2731 Provides all available mitigations for the
2732 L1TF vulnerability. Disables SMT and
2733 enables all mitigations in the
2734 hypervisors, i.e. unconditional L1D flush.
2736 SMT control and L1D flush control via the
2737 sysfs interface is still possible after
2738 boot. Hypervisors will issue a warning
2739 when the first VM is started in a
2740 potentially insecure configuration,
2741 i.e. SMT enabled or L1D flush disabled.
2744 Same as 'full', but disables SMT and L1D
2745 flush runtime control. Implies the
2746 'nosmt=force' command line option.
2747 (i.e. sysfs control of SMT is disabled.)
2750 Leaves SMT enabled and enables the default
2751 hypervisor mitigation, i.e. conditional
2754 SMT control and L1D flush control via the
2755 sysfs interface is still possible after
2756 boot. Hypervisors will issue a warning
2757 when the first VM is started in a
2758 potentially insecure configuration,
2759 i.e. SMT enabled or L1D flush disabled.
2763 Disables SMT and enables the default
2764 hypervisor mitigation.
2766 SMT control and L1D flush control via the
2767 sysfs interface is still possible after
2768 boot. Hypervisors will issue a warning
2769 when the first VM is started in a
2770 potentially insecure configuration,
2771 i.e. SMT enabled or L1D flush disabled.
2774 Same as 'flush', but hypervisors will not
2775 warn when a VM is started in a potentially
2776 insecure configuration.
2779 Disables hypervisor mitigations and doesn't
2781 It also drops the swap size and available
2782 RAM limit restriction on both hypervisor and
2787 For details see: Documentation/admin-guide/hw-vuln/l1tf.rst
2793 lapic [X86-32,APIC] Enable the local APIC even if BIOS
2796 lapic= [X86,APIC] Do not use TSC deadline
2797 value for LAPIC timer one-shot implementation. Default
2798 back to the programmable timer unit in the LAPIC.
2799 Format: notscdeadline
2801 lapic_timer_c2_ok [X86,APIC] trust the local apic timer
2804 libata.dma= [LIBATA] DMA control
2805 libata.dma=0 Disable all PATA and SATA DMA
2806 libata.dma=1 PATA and SATA Disk DMA only
2807 libata.dma=2 ATAPI (CDROM) DMA only
2808 libata.dma=4 Compact Flash DMA only
2809 Combinations also work, so libata.dma=3 enables DMA
2810 for disks and CDROMs, but not CFs.
2812 libata.ignore_hpa= [LIBATA] Ignore HPA limit
2813 libata.ignore_hpa=0 keep BIOS limits (default)
2814 libata.ignore_hpa=1 ignore limits, using full disk
2816 libata.noacpi [LIBATA] Disables use of ACPI in libata suspend/resume
2820 libata.force= [LIBATA] Force configurations. The format is a comma-
2821 separated list of "[ID:]VAL" where ID is PORT[.DEVICE].
2822 PORT and DEVICE are decimal numbers matching port, link
2823 or device. Basically, it matches the ATA ID string
2824 printed on console by libata. If the whole ID part is
2825 omitted, the last PORT and DEVICE values are used. If
2826 ID hasn't been specified yet, the configuration applies
2827 to all ports, links and devices.
2829 If only DEVICE is omitted, the parameter applies to
2830 the port and all links and devices behind it. DEVICE
2831 number of 0 either selects the first device or the
2832 first fan-out link behind PMP device. It does not
2833 select the host link. DEVICE number of 15 selects the
2834 host link and device attached to it.
2836 The VAL specifies the configuration to force. As long
2837 as there is no ambiguity, shortcut notation is allowed.
2838 For example, both 1.5 and 1.5G would work for 1.5Gbps.
2839 The following configurations can be forced.
2841 * Cable type: 40c, 80c, short40c, unk, ign or sata.
2842 Any ID with matching PORT is used.
2844 * SATA link speed limit: 1.5Gbps or 3.0Gbps.
2846 * Transfer mode: pio[0-7], mwdma[0-4] and udma[0-7].
2847 udma[/][16,25,33,44,66,100,133] notation is also
2850 * nohrst, nosrst, norst: suppress hard, soft and both
2853 * rstonce: only attempt one reset during hot-unplug
2856 * [no]dbdelay: Enable or disable the extra 200ms delay
2857 before debouncing a link PHY and device presence
2860 * [no]ncq: Turn on or off NCQ.
2862 * [no]ncqtrim: Enable or disable queued DSM TRIM.
2864 * [no]ncqati: Enable or disable NCQ trim on ATI chipset.
2866 * [no]trim: Enable or disable (unqueued) TRIM.
2868 * trim_zero: Indicate that TRIM command zeroes data.
2870 * max_trim_128m: Set 128M maximum trim size limit.
2872 * [no]dma: Turn on or off DMA transfers.
2874 * atapi_dmadir: Enable ATAPI DMADIR bridge support.
2876 * atapi_mod16_dma: Enable the use of ATAPI DMA for
2877 commands that are not a multiple of 16 bytes.
2879 * [no]dmalog: Enable or disable the use of the
2880 READ LOG DMA EXT command to access logs.
2882 * [no]iddevlog: Enable or disable access to the
2883 identify device data log.
2885 * [no]logdir: Enable or disable access to the general
2886 purpose log directory.
2888 * max_sec_128: Set transfer size limit to 128 sectors.
2890 * max_sec_1024: Set or clear transfer size limit to
2893 * max_sec_lba48: Set or clear transfer size limit to
2896 * [no]lpm: Enable or disable link power management.
2898 * [no]setxfer: Indicate if transfer speed mode setting
2901 * [no]fua: Disable or enable FUA (Force Unit Access)
2902 support for devices supporting this feature.
2904 * dump_id: Dump IDENTIFY data.
2906 * disable: Disable this device.
2908 If there are multiple matching configurations changing
2909 the same attribute, the last one is used.
2911 load_ramdisk= [RAM] [Deprecated]
2913 lockd.nlm_grace_period=P [NFS] Assign grace period.
2916 lockd.nlm_tcpport=N [NFS] Assign TCP port.
2919 lockd.nlm_timeout=T [NFS] Assign timeout value.
2922 lockd.nlm_udpport=M [NFS] Assign UDP port.
2925 lockdown= [SECURITY]
2926 { integrity | confidentiality }
2927 Enable the kernel lockdown feature. If set to
2928 integrity, kernel features that allow userland to
2929 modify the running kernel are disabled. If set to
2930 confidentiality, kernel features that allow userland
2931 to extract confidential information from the kernel
2934 locktorture.acq_writer_lim= [KNL]
2935 Set the time limit in jiffies for a lock
2936 acquisition. Acquisitions exceeding this limit
2937 will result in a splat once they do complete.
2939 locktorture.bind_readers= [KNL]
2940 Specify the list of CPUs to which the readers are
2943 locktorture.bind_writers= [KNL]
2944 Specify the list of CPUs to which the writers are
2947 locktorture.call_rcu_chains= [KNL]
2948 Specify the number of self-propagating call_rcu()
2949 chains to set up. These are used to ensure that
2950 there is a high probability of an RCU grace period
2951 in progress at any given time. Defaults to 0,
2952 which disables these call_rcu() chains.
2954 locktorture.long_hold= [KNL]
2955 Specify the duration in milliseconds for the
2956 occasional long-duration lock hold time. Defaults
2957 to 100 milliseconds. Select 0 to disable.
2959 locktorture.nested_locks= [KNL]
2960 Specify the maximum lock nesting depth that
2961 locktorture is to exercise, up to a limit of 8
2962 (MAX_NESTED_LOCKS). Specify zero to disable.
2963 Note that this parameter is ineffective on types
2964 of locks that do not support nested acquisition.
2966 locktorture.nreaders_stress= [KNL]
2967 Set the number of locking read-acquisition kthreads.
2968 Defaults to being automatically set based on the
2969 number of online CPUs.
2971 locktorture.nwriters_stress= [KNL]
2972 Set the number of locking write-acquisition kthreads.
2974 locktorture.onoff_holdoff= [KNL]
2975 Set time (s) after boot for CPU-hotplug testing.
2977 locktorture.onoff_interval= [KNL]
2978 Set time (s) between CPU-hotplug operations, or
2979 zero to disable CPU-hotplug testing.
2981 locktorture.rt_boost= [KNL]
2982 Do periodic testing of real-time lock priority
2983 boosting. Select 0 to disable, 1 to boost
2984 only rt_mutex, and 2 to boost unconditionally.
2985 Defaults to 2, which might seem to be an
2986 odd choice, but which should be harmless for
2987 non-real-time spinlocks, due to their disabling
2988 of preemption. Note that non-realtime mutexes
2991 locktorture.rt_boost_factor= [KNL]
2992 Number that determines how often and for how
2993 long priority boosting is exercised. This is
2994 scaled down by the number of writers, so that the
2995 number of boosts per unit time remains roughly
2996 constant as the number of writers increases.
2997 On the other hand, the duration of each boost
2998 increases with the number of writers.
3000 locktorture.shuffle_interval= [KNL]
3001 Set task-shuffle interval (jiffies). Shuffling
3002 tasks allows some CPUs to go into dyntick-idle
3003 mode during the locktorture test.
3005 locktorture.shutdown_secs= [KNL]
3006 Set time (s) after boot system shutdown. This
3007 is useful for hands-off automated testing.
3009 locktorture.stat_interval= [KNL]
3010 Time (s) between statistics printk()s.
3012 locktorture.stutter= [KNL]
3013 Time (s) to stutter testing, for example,
3014 specifying five seconds causes the test to run for
3015 five seconds, wait for five seconds, and so on.
3016 This tests the locking primitive's ability to
3017 transition abruptly to and from idle.
3019 locktorture.torture_type= [KNL]
3020 Specify the locking implementation to test.
3022 locktorture.verbose= [KNL]
3023 Enable additional printk() statements.
3025 locktorture.writer_fifo= [KNL]
3026 Run the write-side locktorture kthreads at
3027 sched_set_fifo() real-time priority.
3029 logibm.irq= [HW,MOUSE] Logitech Bus Mouse Driver
3032 loglevel= All Kernel Messages with a loglevel smaller than the
3033 console loglevel will be printed to the console. It can
3034 also be changed with klogd or other programs. The
3035 loglevels are defined as follows:
3037 0 (KERN_EMERG) system is unusable
3038 1 (KERN_ALERT) action must be taken immediately
3039 2 (KERN_CRIT) critical conditions
3040 3 (KERN_ERR) error conditions
3041 4 (KERN_WARNING) warning conditions
3042 5 (KERN_NOTICE) normal but significant condition
3043 6 (KERN_INFO) informational
3044 7 (KERN_DEBUG) debug-level messages
3046 log_buf_len=n[KMG] Sets the size of the printk ring buffer,
3047 in bytes. n must be a power of two and greater
3048 than the minimal size. The minimal size is defined
3049 by LOG_BUF_SHIFT kernel config parameter. There is
3050 also CONFIG_LOG_CPU_MAX_BUF_SHIFT config parameter
3051 that allows to increase the default size depending on
3052 the number of CPUs. See init/Kconfig for more details.
3054 logo.nologo [FB] Disables display of the built-in Linux logo.
3055 This may be used to provide more screen space for
3056 kernel log messages and is useful when debugging
3057 kernel boot problems.
3059 lp=0 [LP] Specify parallel ports to use, e.g,
3060 lp=port[,port...] lp=none,parport0 (lp0 not configured, lp1 uses
3061 lp=reset first parallel port). 'lp=0' disables the
3062 lp=auto printer driver. 'lp=reset' (which can be
3063 specified in addition to the ports) causes
3064 attached printers to be reset. Using
3065 lp=port1,port2,... specifies the parallel ports
3066 to associate lp devices with, starting with
3067 lp0. A port specification may be 'none' to skip
3068 that lp device, or a parport name such as
3069 'parport0'. Specifying 'lp=auto' instead of a
3070 port specification list means that device IDs
3071 from each port should be examined, to see if
3072 an IEEE 1284-compliant printer is attached; if
3073 so, the driver will manage that printer.
3074 See also header of drivers/char/lp.c.
3077 Sets loops_per_jiffy to given constant, thus avoiding
3078 time-consuming boot-time autodetection (up to 250 ms per
3079 CPU). 0 enables autodetection (default). To determine
3080 the correct value for your kernel, boot with normal
3081 autodetection and see what value is printed. Note that
3082 on SMP systems the preset will be applied to all CPUs,
3083 which is likely to cause problems if your CPUs need
3084 significantly divergent settings. An incorrect value
3085 will cause delays in the kernel to be wrong, leading to
3086 unpredictable I/O errors and other breakage. Although
3087 unlikely, in the extreme case this might damage your
3091 Format: <io>,<irq>,<dma>
3093 lsm.debug [SECURITY] Enable LSM initialization debugging output.
3096 [SECURITY] Choose order of LSM initialization. This
3097 overrides CONFIG_LSM, and the "security=" parameter.
3099 machvec= [IA-64] Force the use of a particular machine-vector
3100 (machvec) in a generic kernel.
3101 Example: machvec=hpzx1
3103 machtype= [Loongson] Share the same kernel image file between
3104 different yeeloong laptops.
3105 Example: machtype=lemote-yeeloong-2f-7inch
3107 max_addr=nn[KMG] [KNL,BOOT,IA-64] All physical memory greater
3108 than or equal to this physical address is ignored.
3110 maxcpus= [SMP] Maximum number of processors that an SMP kernel
3111 will bring up during bootup. maxcpus=n : n >= 0 limits
3112 the kernel to bring up 'n' processors. Surely after
3113 bootup you can bring up the other plugged cpu by executing
3114 "echo 1 > /sys/devices/system/cpu/cpuX/online". So maxcpus
3115 only takes effect during system bootup.
3116 While n=0 is a special case, it is equivalent to "nosmp",
3117 which also disables the IO APIC.
3119 max_loop= [LOOP] The number of loop block devices that get
3120 (loop.max_loop) unconditionally pre-created at init time. The default
3121 number is configured by BLK_DEV_LOOP_MIN_COUNT. Instead
3122 of statically allocating a predefined number, loop
3123 devices can be requested on-demand with the
3124 /dev/loop-control interface.
3126 mce [X86-32] Machine Check Exception
3128 mce=option [X86-64] See Documentation/arch/x86/x86_64/boot-options.rst
3130 md= [HW] RAID subsystems devices and level
3131 See Documentation/admin-guide/md.rst.
3134 Format: <first>,<last>
3135 Specifies range of consoles to be captured by the MDA.
3138 Control mitigation for the Micro-architectural Data
3139 Sampling (MDS) vulnerability.
3141 Certain CPUs are vulnerable to an exploit against CPU
3142 internal buffers which can forward information to a
3143 disclosure gadget under certain conditions.
3145 In vulnerable processors, the speculatively
3146 forwarded data can be used in a cache side channel
3147 attack, to access data to which the attacker does
3148 not have direct access.
3150 This parameter controls the MDS mitigation. The
3153 full - Enable MDS mitigation on vulnerable CPUs
3154 full,nosmt - Enable MDS mitigation and disable
3155 SMT on vulnerable CPUs
3156 off - Unconditionally disable MDS mitigation
3158 On TAA-affected machines, mds=off can be prevented by
3159 an active TAA mitigation as both vulnerabilities are
3160 mitigated with the same mechanism so in order to disable
3161 this mitigation, you need to specify tsx_async_abort=off
3164 Not specifying this option is equivalent to
3167 For details see: Documentation/admin-guide/hw-vuln/mds.rst
3169 mem=nn[KMG] [HEXAGON] Set the memory size.
3170 Must be specified, otherwise memory size will be 0.
3172 mem=nn[KMG] [KNL,BOOT] Force usage of a specific amount of memory
3173 Amount of memory to be used in cases as follows:
3176 2 when the kernel is not able to see the whole system memory;
3177 3 memory that lies after 'mem=' boundary is excluded from
3178 the hypervisor, then assigned to KVM guests.
3179 4 to limit the memory available for kdump kernel.
3181 [ARC,MICROBLAZE] - the limit applies only to low memory,
3182 high memory is not affected.
3184 [ARM64] - only limits memory covered by the linear
3185 mapping. The NOMAP regions are not affected.
3187 [X86] Work as limiting max address. Use together
3188 with memmap= to avoid physical address space collisions.
3189 Without memmap= PCI devices could be placed at addresses
3190 belonging to unused RAM.
3192 Note that this only takes effects during boot time since
3193 in above case 3, memory may need be hot added after boot
3194 if system memory of hypervisor is not sufficient.
3197 [ARM,MIPS] - override the memory layout reported by
3199 Define a memory region of size nn[KMG] starting at
3201 Multiple different regions can be specified with
3202 multiple mem= parameters on the command line.
3204 mem=nopentium [BUGS=X86-32] Disable usage of 4MB pages for kernel
3207 memblock=debug [KNL] Enable memblock debug messages.
3210 [KNL,SH] Allow user to override the default size for
3211 per-device physically contiguous DMA buffers.
3213 memhp_default_state=online/offline/online_kernel/online_movable
3214 [KNL] Set the initial state for the memory hotplug
3215 onlining policy. If not specified, the default value is
3216 set according to the
3217 CONFIG_MEMORY_HOTPLUG_DEFAULT_ONLINE kernel config
3219 See Documentation/admin-guide/mm/memory-hotplug.rst.
3221 memmap=exactmap [KNL,X86] Enable setting of an exact
3222 E820 memory map, as specified by the user.
3223 Such memmap=exactmap lines can be constructed based on
3224 BIOS output or other requirements. See the memmap=nn@ss
3227 memmap=nn[KMG]@ss[KMG]
3228 [KNL, X86, MIPS, XTENSA] Force usage of a specific region of memory.
3229 Region of memory to be used is from ss to ss+nn.
3230 If @ss[KMG] is omitted, it is equivalent to mem=nn[KMG],
3231 which limits max address to nn[KMG].
3232 Multiple different regions can be specified,
3235 memmap=100M@2G,100M#3G,1G!1024G
3237 memmap=nn[KMG]#ss[KMG]
3238 [KNL,ACPI] Mark specific memory as ACPI data.
3239 Region of memory to be marked is from ss to ss+nn.
3241 memmap=nn[KMG]$ss[KMG]
3242 [KNL,ACPI] Mark specific memory as reserved.
3243 Region of memory to be reserved is from ss to ss+nn.
3244 Example: Exclude memory from 0x18690000-0x1869ffff
3245 memmap=64K$0x18690000
3247 memmap=0x10000$0x18690000
3248 Some bootloaders may need an escape character before '$',
3249 like Grub2, otherwise '$' and the following number
3252 memmap=nn[KMG]!ss[KMG]
3253 [KNL,X86] Mark specific memory as protected.
3254 Region of memory to be used, from ss to ss+nn.
3255 The memory region may be marked as e820 type 12 (0xc)
3256 and is NVDIMM or ADR memory.
3258 memmap=<size>%<offset>-<oldtype>+<newtype>
3259 [KNL,ACPI] Convert memory within the specified region
3260 from <oldtype> to <newtype>. If "-<oldtype>" is left
3261 out, the whole region will be marked as <newtype>,
3262 even if previously unavailable. If "+<newtype>" is left
3263 out, matching memory will be removed. Types are
3264 specified as e820 types, e.g., 1 = RAM, 2 = reserved,
3265 3 = ACPI, 12 = PRAM.
3267 memory_corruption_check=0/1 [X86]
3268 Some BIOSes seem to corrupt the first 64k of
3269 memory when doing things like suspend/resume.
3270 Setting this option will scan the memory
3271 looking for corruption. Enabling this will
3272 both detect corruption and prevent the kernel
3273 from using the memory being corrupted.
3274 However, its intended as a diagnostic tool; if
3275 repeatable BIOS-originated corruption always
3276 affects the same memory, you can use memmap=
3277 to prevent the kernel from using that memory.
3279 memory_corruption_check_size=size [X86]
3280 By default it checks for corruption in the low
3281 64k, making this memory unavailable for normal
3282 use. Use this parameter to scan for
3283 corruption in more or less memory.
3285 memory_corruption_check_period=seconds [X86]
3286 By default it checks for corruption every 60
3287 seconds. Use this parameter to check at some
3288 other rate. 0 disables periodic checking.
3290 memory_hotplug.memmap_on_memory
3291 [KNL,X86,ARM] Boolean flag to enable this feature.
3292 Format: {on | off (default)}
3293 When enabled, runtime hotplugged memory will
3294 allocate its internal metadata (struct pages,
3295 those vmemmap pages cannot be optimized even
3296 if hugetlb_free_vmemmap is enabled) from the
3297 hotadded memory which will allow to hotadd a
3298 lot of memory without requiring additional
3300 This feature is disabled by default because it
3301 has some implication on large (e.g. GB)
3302 allocations in some configurations (e.g. small
3304 The state of the flag can be read in
3305 /sys/module/memory_hotplug/parameters/memmap_on_memory.
3306 Note that even when enabled, there are a few cases where
3307 the feature is not effective.
3309 memtest= [KNL,X86,ARM,M68K,PPC,RISCV] Enable memtest
3311 default : 0 <disable>
3312 Specifies the number of memtest passes to be
3313 performed. Each pass selects another test
3314 pattern from a given set of patterns. Memtest
3315 fills the memory with this pattern, validates
3316 memory contents and reserves bad memory
3317 regions that are detected.
3319 mem_encrypt= [X86-64] AMD Secure Memory Encryption (SME) control
3320 Valid arguments: on, off
3321 Default (depends on kernel configuration option):
3322 on (CONFIG_AMD_MEM_ENCRYPT_ACTIVE_BY_DEFAULT=y)
3323 off (CONFIG_AMD_MEM_ENCRYPT_ACTIVE_BY_DEFAULT=n)
3324 mem_encrypt=on: Activate SME
3325 mem_encrypt=off: Do not activate SME
3327 Refer to Documentation/virt/kvm/x86/amd-memory-encryption.rst
3328 for details on when memory encryption can be activated.
3330 mem_sleep_default= [SUSPEND] Default system suspend mode:
3331 s2idle - Suspend-To-Idle
3332 shallow - Power-On Suspend or equivalent (if supported)
3333 deep - Suspend-To-RAM or equivalent (if supported)
3334 See Documentation/admin-guide/pm/sleep-states.rst.
3336 mfgpt_irq= [IA-32] Specify the IRQ to use for the
3337 Multi-Function General Purpose Timers on AMD Geode
3340 mfgptfix [X86-32] Fix MFGPT timers on AMD Geode platforms when
3341 the BIOS has incorrectly applied a workaround. TinyBIOS
3342 version 0.98 is known to be affected, 0.99 fixes the
3343 problem by letting the user disable the workaround.
3347 microcode.force_minrev= [X86]
3349 Enable or disable the microcode minimal revision
3350 enforcement for the runtime microcode loader.
3352 min_addr=nn[KMG] [KNL,BOOT,IA-64] All physical memory below this
3353 physical address is ignored.
3355 mini2440= [ARM,HW,KNL]
3356 Format:[0..2][b][c][t]
3358 MINI2440 configuration specification:
3359 0 - The attached screen is the 3.5" TFT
3360 1 - The attached screen is the 7" TFT
3361 2 - The VGA Shield is attached (1024x768)
3362 Leaving out the screen size parameter will not load
3363 the TFT driver, and the framebuffer will be left
3365 b - Enable backlight. The TFT backlight pin will be
3366 linked to the kernel VESA blanking code and a GPIO
3367 LED. This parameter is not necessary when using the
3369 c - Enable the s3c camera interface.
3370 t - Reserved for enabling touchscreen support. The
3371 touchscreen support is not enabled in the mainstream
3372 kernel as of 2.6.30, a preliminary port can be found
3373 in the "bleeding edge" mini2440 support kernel at
3374 https://repo.or.cz/w/linux-2.6/mini2440.git
3377 [X86,PPC,S390,ARM64] Control optional mitigations for
3378 CPU vulnerabilities. This is a set of curated,
3379 arch-independent options, each of which is an
3380 aggregation of existing arch-specific options.
3383 Disable all optional CPU mitigations. This
3384 improves system performance, but it may also
3385 expose users to several CPU vulnerabilities.
3386 Equivalent to: if nokaslr then kpti=0 [ARM64]
3387 gather_data_sampling=off [X86]
3388 kvm.nx_huge_pages=off [X86]
3391 mmio_stale_data=off [X86]
3392 no_entry_flush [PPC]
3393 no_uaccess_flush [PPC]
3396 nospectre_bhb [ARM64]
3397 nospectre_v1 [X86,PPC]
3398 nospectre_v2 [X86,PPC,S390,ARM64]
3400 spec_store_bypass_disable=off [X86,PPC]
3401 spectre_v2_user=off [X86]
3402 srbds=off [X86,INTEL]
3403 ssbd=force-off [ARM64]
3404 tsx_async_abort=off [X86]
3407 This does not have any effect on
3408 kvm.nx_huge_pages when
3409 kvm.nx_huge_pages=force.
3412 Mitigate all CPU vulnerabilities, but leave SMT
3413 enabled, even if it's vulnerable. This is for
3414 users who don't want to be surprised by SMT
3415 getting disabled across kernel upgrades, or who
3416 have other ways of avoiding SMT-based attacks.
3417 Equivalent to: (default behavior)
3420 Mitigate all CPU vulnerabilities, disabling SMT
3421 if needed. This is for users who always want to
3422 be fully mitigated, even if it means losing SMT.
3423 Equivalent to: l1tf=flush,nosmt [X86]
3424 mds=full,nosmt [X86]
3425 tsx_async_abort=full,nosmt [X86]
3426 mmio_stale_data=full,nosmt [X86]
3427 retbleed=auto,nosmt [X86]
3430 [KNL] When CONFIG_DEBUG_MEMORY_INIT is set, this
3431 parameter allows control of the logging verbosity for
3432 the additional memory initialisation checks. A value
3433 of 0 disables mminit logging and a level of 4 will
3434 log everything. Information is printed at KERN_DEBUG
3435 so loglevel=8 may also need to be specified.
3438 [X86,INTEL] Control mitigation for the Processor
3439 MMIO Stale Data vulnerabilities.
3441 Processor MMIO Stale Data is a class of
3442 vulnerabilities that may expose data after an MMIO
3443 operation. Exposed data could originate or end in
3444 the same CPU buffers as affected by MDS and TAA.
3445 Therefore, similar to MDS and TAA, the mitigation
3446 is to clear the affected CPU buffers.
3448 This parameter controls the mitigation. The
3451 full - Enable mitigation on vulnerable CPUs
3453 full,nosmt - Enable mitigation and disable SMT on
3456 off - Unconditionally disable mitigation
3458 On MDS or TAA affected machines,
3459 mmio_stale_data=off can be prevented by an active
3460 MDS or TAA mitigation as these vulnerabilities are
3461 mitigated with the same mechanism so in order to
3462 disable this mitigation, you need to specify
3463 mds=off and tsx_async_abort=off too.
3465 Not specifying this option is equivalent to
3466 mmio_stale_data=full.
3469 Documentation/admin-guide/hw-vuln/processor_mmio_stale_data.rst
3471 <module>.async_probe[=<bool>] [KNL]
3472 If no <bool> value is specified or if the value
3473 specified is not a valid <bool>, enable asynchronous
3474 probe on this module. Otherwise, enable/disable
3475 asynchronous probe on this module as indicated by the
3476 <bool> value. See also: module.async_probe
3478 module.async_probe=<bool>
3479 [KNL] When set to true, modules will use async probing
3480 by default. To enable/disable async probing for a
3481 specific module, use the module specific control that
3482 is documented under <module>.async_probe. When both
3483 module.async_probe and <module>.async_probe are
3484 specified, <module>.async_probe takes precedence for
3485 the specific module.
3487 module.enable_dups_trace
3488 [KNL] When CONFIG_MODULE_DEBUG_AUTOLOAD_DUPS is set,
3489 this means that duplicate request_module() calls will
3490 trigger a WARN_ON() instead of a pr_warn(). Note that
3491 if MODULE_DEBUG_AUTOLOAD_DUPS_TRACE is set, WARN_ON()s
3492 will always be issued and this option does nothing.
3494 [KNL] When CONFIG_MODULE_SIG is set, this means that
3495 modules without (valid) signatures will fail to load.
3496 Note that if CONFIG_MODULE_SIG_FORCE is set, that
3497 is always true, so this option does nothing.
3499 module_blacklist= [KNL] Do not load a comma-separated list of
3500 modules. Useful for debugging problem modules.
3503 [MOUSE] Maximum time between finger touching and
3504 leaving touchpad surface for touch to be considered
3505 a tap and be reported as a left button click (for
3506 touchpads working in absolute mode only).
3508 mousedev.xres= [MOUSE] Horizontal screen resolution, used for devices
3509 reporting absolute coordinates, such as tablets
3510 mousedev.yres= [MOUSE] Vertical screen resolution, used for devices
3511 reporting absolute coordinates, such as tablets
3513 movablecore= [KNL,X86,IA-64,PPC]
3514 Format: nn[KMGTPE] | nn%
3515 This parameter is the complement to kernelcore=, it
3516 specifies the amount of memory used for migratable
3517 allocations. If both kernelcore and movablecore is
3518 specified, then kernelcore will be at *least* the
3519 specified value but may be more. If movablecore on its
3520 own is specified, the administrator must be careful
3521 that the amount of memory usable for all allocations
3524 movable_node [KNL] Boot-time switch to make hotplugable memory
3525 NUMA nodes to be movable. This means that the memory
3526 of such nodes will be usable only for movable
3527 allocations which rules out almost all kernel
3528 allocations. Use with caution!
3530 MTD_Partition= [MTD]
3531 Format: <name>,<region-number>,<size>,<offset>
3533 MTD_Region= [MTD] Format:
3534 <name>,<region-number>[,<base>,<size>,<buswidth>,<altbuswidth>]
3537 See drivers/mtd/parsers/cmdlinepart.c
3540 ARM/S3C2412 JIVE boot control
3542 See arch/arm/mach-s3c/mach-jive.c
3544 mtouchusb.raw_coordinates=
3545 [HW] Make the MicroTouch USB driver use raw coordinates
3546 ('y', default) or cooked coordinates ('n')
3549 Enable printing debug information related to MTRR
3550 registers at boot time.
3552 mtrr_chunk_size=nn[KMG] [X86]
3553 used for mtrr cleanup. It is largest continuous chunk
3554 that could hold holes aka. UC entries.
3556 mtrr_gran_size=nn[KMG] [X86]
3557 Used for mtrr cleanup. It is granularity of mtrr block.
3559 Large value could prevent small alignment from
3562 mtrr_spare_reg_nr=n [X86]
3564 Range: 0,7 : spare reg number
3566 Used for mtrr cleanup. It is spare mtrr entries number.
3567 Set to 2 or more if your graphical card needs more.
3569 multitce=off [PPC] This parameter disables the use of the pSeries
3570 firmware feature for updating multiple TCE entries
3573 n2= [NET] SDL Inc. RISCom/N2 synchronous serial card
3575 netdev= [NET] Network devices parameters
3576 Format: <irq>,<io>,<mem_start>,<mem_end>,<name>
3577 Note that mem_start is often overloaded to mean
3578 something different and driver-specific.
3579 This usage is only documented in each driver source
3582 netpoll.carrier_timeout=
3583 [NET] Specifies amount of time (in seconds) that
3584 netpoll should wait for a carrier. By default netpoll
3588 [NETFILTER] Enable connection tracking flow accounting
3589 0 to disable accounting
3590 1 to enable accounting
3594 [NFS] sets the pathname to the program which is used
3595 to update the NFS client cache entries.
3597 nfs.cache_getent_timeout=
3598 [NFS] sets the timeout after which an attempt to
3599 update a cache entry is deemed to have failed.
3601 nfs.callback_nr_threads=
3602 [NFSv4] set the total number of threads that the
3603 NFS client will assign to service NFSv4 callback
3606 nfs.callback_tcpport=
3607 [NFS] set the TCP port on which the NFSv4 callback
3608 channel should listen.
3611 [NFS] specifies the number of times the NFSv4 client
3612 retries the request before returning an EAGAIN error,
3613 after a reply of NFS4ERR_DELAY from the server.
3614 Only applies if the softerr mount option is enabled,
3615 and the specified value is >= 0.
3618 [NFS] enable 64-bit inode numbers.
3619 If zero, the NFS client will fake up a 32-bit inode
3620 number for the readdir() and stat() syscalls instead
3621 of returning the full 64-bit number.
3622 The default is to return 64-bit inode numbers.
3624 nfs.idmap_cache_timeout=
3625 [NFS] set the maximum lifetime for idmapper cache
3628 nfs.max_session_cb_slots=
3629 [NFSv4.1] Sets the maximum number of session
3630 slots the client will assign to the callback
3631 channel. This determines the maximum number of
3632 callbacks the client will process in parallel for
3633 a particular server.
3635 nfs.max_session_slots=
3636 [NFSv4.1] Sets the maximum number of session slots
3637 the client will attempt to negotiate with the server.
3638 This limits the number of simultaneous RPC requests
3639 that the client can send to the NFSv4.1 server.
3640 Note that there is little point in setting this
3641 value higher than the max_tcp_slot_table_limit.
3643 nfs.nfs4_disable_idmapping=
3644 [NFSv4] When set to the default of '1', this option
3645 ensures that both the RPC level authentication
3646 scheme and the NFS level operations agree to use
3647 numeric uids/gids if the mount is using the
3648 'sec=sys' security flavour. In effect it is
3649 disabling idmapping, which can make migration from
3650 legacy NFSv2/v3 systems to NFSv4 easier.
3651 Servers that do not support this mode of operation
3652 will be autodetected by the client, and it will fall
3653 back to using the idmapper.
3654 To turn off this behaviour, set the value to '0'.
3657 [NFS4] Specify an additional fixed unique ident-
3658 ification string that NFSv4 clients can insert into
3659 their nfs_client_id4 string. This is typically a
3660 UUID that is generated at system install time.
3662 nfs.recover_lost_locks=
3663 [NFSv4] Attempt to recover locks that were lost due
3664 to a lease timeout on the server. Please note that
3665 doing this risks data corruption, since there are
3666 no guarantees that the file will remain unchanged
3667 after the locks are lost.
3668 If you want to enable the kernel legacy behaviour of
3669 attempting to recover these locks, then set this
3671 The default parameter value of '0' causes the kernel
3672 not to attempt recovery of lost locks.
3674 nfs.send_implementation_id=
3675 [NFSv4.1] Send client implementation identification
3676 information in exchange_id requests.
3677 If zero, no implementation identification information
3679 The default is to send the implementation identification
3682 nfs4.layoutstats_timer=
3683 [NFSv4.2] Change the rate at which the kernel sends
3684 layoutstats to the pNFS metadata server.
3686 Setting this to value to 0 causes the kernel to use
3687 whatever value is the default set by the layout
3688 driver. A non-zero value sets the minimum interval
3689 in seconds between layoutstats transmissions.
3691 nfsd.inter_copy_offload_enable=
3692 [NFSv4.2] When set to 1, the server will support
3693 server-to-server copies for which this server is
3694 the destination of the copy.
3696 nfsd.nfs4_disable_idmapping=
3697 [NFSv4] When set to the default of '1', the NFSv4
3698 server will return only numeric uids and gids to
3699 clients using auth_sys, and will accept numeric uids
3700 and gids from such clients. This is intended to ease
3701 migration from NFSv2/v3.
3703 nfsd.nfsd4_ssc_umount_timeout=
3704 [NFSv4.2] When used as the destination of a
3705 server-to-server copy, knfsd temporarily mounts
3706 the source server. It caches the mount in case
3707 it will be needed again, and discards it if not
3708 used for the number of milliseconds specified by
3711 nfsaddrs= [NFS] Deprecated. Use ip= instead.
3712 See Documentation/admin-guide/nfs/nfsroot.rst.
3714 nfsroot= [NFS] nfs root filesystem for disk-less boxes.
3715 See Documentation/admin-guide/nfs/nfsroot.rst.
3717 nfsrootdebug [NFS] enable nfsroot debugging messages.
3718 See Documentation/admin-guide/nfs/nfsroot.rst.
3720 nmi_backtrace.backtrace_idle [KNL]
3721 Dump stacks even of idle CPUs in response to an
3722 NMI stack-backtrace request.
3724 nmi_debug= [KNL,SH] Specify one or more actions to take
3725 when a NMI is triggered.
3726 Format: [state][,regs][,debounce][,die]
3728 nmi_watchdog= [KNL,BUGS=X86] Debugging features for SMP kernels
3729 Format: [panic,][nopanic,][num]
3731 0 - turn hardlockup detector in nmi_watchdog off
3732 1 - turn hardlockup detector in nmi_watchdog on
3733 When panic is specified, panic when an NMI watchdog
3734 timeout occurs (or 'nopanic' to not panic on an NMI
3735 watchdog, if CONFIG_BOOTPARAM_HARDLOCKUP_PANIC is set)
3736 To disable both hard and soft lockup detectors,
3737 please see 'nowatchdog'.
3738 This is useful when you use a panic=... timeout and
3739 need the box quickly up again.
3741 These settings can be accessed at runtime via
3742 the nmi_watchdog and hardlockup_panic sysctls.
3744 no387 [BUGS=X86-32] Tells the kernel to use the 387 maths
3745 emulation library even if a 387 maths coprocessor
3748 no4lvl [RISCV] Disable 4-level and 5-level paging modes. Forces
3749 kernel to use 3-level paging instead.
3751 no5lvl [X86-64,RISCV] Disable 5-level paging mode. Forces
3752 kernel to use 4-level paging instead.
3754 noaliencache [MM, NUMA, SLAB] Disables the allocation of alien
3755 caches in the slab allocator. Saves per-node memory,
3756 but will impact performance.
3760 noaltinstr [S390] Disables alternative instructions patching
3761 (CPU alternatives feature).
3763 noapic [SMP,APIC] Tells the kernel to not make use of any
3764 IOAPICs that may be present in the system.
3766 noautogroup Disable scheduler automatic task group creation.
3771 [HW] Never suspend the console
3772 Disable suspending of consoles during suspend and
3773 hibernate operations. Once disabled, debugging
3774 messages can reach various consoles while the rest
3775 of the system is being put to sleep (ie, while
3776 debugging driver suspend/resume hooks). This may
3777 not work reliably with all consoles, but is known
3778 to work with serial and VGA consoles.
3779 To facilitate more flexible debugging, we also add
3780 console_suspend, a printk module parameter to control
3781 it. Users could use console_suspend (usually
3782 /sys/module/printk/parameters/console_suspend) to
3783 turn on/off it dynamically.
3786 [KNL] Disable object debugging
3788 nodsp [SH] Disable hardware DSP at boot time.
3790 noefi Disable EFI runtime services support.
3792 no_entry_flush [PPC] Don't flush the L1-D cache when entering the kernel.
3797 This affects only 32-bit executables.
3798 noexec32=on: enable non-executable mappings (default)
3799 read doesn't imply executable mappings
3800 noexec32=off: disable non-executable mappings
3801 read implies executable mappings
3803 no_file_caps Tells the kernel not to honor file capabilities. The
3804 only way then for a file to be executed with privilege
3805 is to be setuid root or executed by root.
3807 nofpu [MIPS,SH] Disable hardware FPU at boot time.
3809 nofsgsbase [X86] Disables FSGSBASE instructions.
3811 nofxsr [BUGS=X86-32] Disables x86 floating point extended
3812 register save and restore. The kernel will only save
3813 legacy floating-point registers on task switch.
3815 nohalt [IA-64] Tells the kernel not to use the power saving
3816 function PAL_HALT_LIGHT when idle. This increases
3817 power-consumption. On the positive side, it reduces
3818 interrupt wake-up latency, which may improve performance
3819 in certain environments such as networked servers or
3823 Force pointers printed to the console or buffers to be
3824 unhashed. By default, when a pointer is printed via %p
3825 format string, that pointer is "hashed", i.e. obscured
3826 by hashing the pointer value. This is a security feature
3827 that hides actual kernel addresses from unprivileged
3828 users, but it also makes debugging the kernel more
3829 difficult since unequal pointers can no longer be
3830 compared. However, if this command-line option is
3831 specified, then all normal pointers will have their true
3832 value printed. This option should only be specified when
3833 debugging the kernel. Please do not use on production
3836 nohibernate [HIBERNATION] Disable hibernation and resume.
3838 nohlt [ARM,ARM64,MICROBLAZE,MIPS,PPC,SH] Forces the kernel to
3839 busy wait in do_idle() and not use the arch_cpu_idle()
3840 implementation; requires CONFIG_GENERIC_IDLE_POLL_SETUP
3841 to be effective. This is useful on platforms where the
3842 sleep(SH) or wfi(ARM,ARM64) instructions do not work
3843 correctly or when doing power measurements to evaluate
3844 the impact of the sleep instructions. This is also
3845 useful when using JTAG debugger.
3847 nohugeiomap [KNL,X86,PPC,ARM64] Disable kernel huge I/O mappings.
3849 nohugevmalloc [KNL,X86,PPC,ARM64] Disable kernel huge vmalloc mappings.
3851 nohz= [KNL] Boottime enable/disable dynamic ticks
3852 Valid arguments: on, off
3855 nohz_full= [KNL,BOOT,SMP,ISOL]
3856 The argument is a cpu list, as described above.
3857 In kernels built with CONFIG_NO_HZ_FULL=y, set
3858 the specified list of CPUs whose tick will be stopped
3859 whenever possible. The boot CPU will be forced outside
3860 the range to maintain the timekeeping. Any CPUs
3861 in this list will have their RCU callbacks offloaded,
3862 just as if they had also been called out in the
3863 rcu_nocbs= boot parameter.
3865 Note that this argument takes precedence over
3866 the CONFIG_RCU_NOCB_CPU_DEFAULT_ALL option.
3868 noinitrd [RAM] Tells the kernel not to load any configured
3871 nointremap [X86-64, Intel-IOMMU] Do not enable interrupt
3873 [Deprecated - use intremap=off]
3877 noinvpcid [X86] Disable the INVPCID cpu feature.
3879 noiotrap [SH] Disables trapped I/O port accesses.
3881 noirqdebug [X86-32] Disables the code which attempts to detect and
3882 disable unhandled interrupt sources.
3884 noisapnp [ISAPNP] Disables ISA PnP code.
3886 nojitter [IA-64] Disables jitter checking for ITC timers.
3889 When CONFIG_RANDOMIZE_BASE is set, this disables
3890 kernel and module base offset ASLR (Address Space
3891 Layout Randomization).
3893 no-kvmapf [X86,KVM] Disable paravirtualized asynchronous page
3896 no-kvmclock [X86,KVM] Disable paravirtualized KVM clock driver
3898 nolapic [X86-32,APIC] Do not enable or use the local APIC.
3900 nolapic_timer [X86-32,APIC] Do not use the local APIC timer.
3902 nomca [IA-64] Disable machine check abort handling
3904 nomce [X86-32] Disable Machine Check Exception
3906 nomfgpt [X86-32] Disable Multi-Function General Purpose
3907 Timer usage (for AMD Geode machines).
3909 nomodeset Disable kernel modesetting. Most systems' firmware
3910 sets up a display mode and provides framebuffer memory
3911 for output. With nomodeset, DRM and fbdev drivers will
3912 not load if they could possibly displace the pre-
3913 initialized output. Only the system framebuffer will
3914 be available for use. The respective drivers will not
3915 perform display-mode changes or accelerated rendering.
3917 Useful as error fallback, or for testing and debugging.
3919 nomodule Disable module load
3921 nonmi_ipi [X86] Disable using NMI IPIs during panic/reboot to
3922 shutdown the other cpus. Instead use the REBOOT_VECTOR
3925 nopat [X86] Disable PAT (page attribute table extension of
3926 pagetables) support.
3928 nopcid [X86-64] Disable the PCID cpu feature.
3930 nopku [X86] Disable Memory Protection Keys CPU feature found
3934 Equivalent to pti=off
3936 nopv= [X86,XEN,KVM,HYPER_V,VMWARE]
3937 Disables the PV optimizations forcing the guest to run
3938 as generic guest with no PV drivers. Currently support
3939 XEN HVM, KVM, HYPER_V and VMWARE guest.
3941 nopvspin [X86,XEN,KVM]
3942 Disables the qspinlock slow path using PV optimizations
3943 which allow the hypervisor to 'idle' the guest on lock
3946 norandmaps Don't use address space randomization. Equivalent to
3947 echo 0 > /proc/sys/kernel/randomize_va_space
3949 noreplace-smp [X86-32,SMP] Don't replace SMP instructions
3950 with UP alternatives
3952 noresume [SWSUSP] Disables resume and restores original swap
3957 no-scroll [VGA] Disables scrollback.
3958 This is required for the Braillex ib80-piezo Braille
3959 reader made by F.H. Papenmeier (Germany).
3961 nosgx [X86-64,SGX] Disables Intel SGX kernel support.
3964 Disable SMAP (Supervisor Mode Access Prevention)
3965 even if it is supported by processor.
3968 Disable SMEP (Supervisor Mode Execution Prevention)
3969 even if it is supported by processor.
3971 nosmp [SMP] Tells an SMP kernel to act as a UP kernel,
3972 and disable the IO APIC. legacy for "maxcpus=0".
3974 nosmt [KNL,MIPS,PPC,S390] Disable symmetric multithreading (SMT).
3975 Equivalent to smt=1.
3977 [KNL,X86,PPC] Disable symmetric multithreading (SMT).
3978 nosmt=force: Force disable SMT, cannot be undone
3979 via the sysfs control file.
3981 nosoftlockup [KNL] Disable the soft-lockup detector.
3983 nospec_store_bypass_disable
3984 [HW] Disable all mitigations for the Speculative Store Bypass vulnerability
3986 nospectre_bhb [ARM64] Disable all mitigations for Spectre-BHB (branch
3987 history injection) vulnerability. System may allow data leaks
3990 nospectre_v1 [X86,PPC] Disable mitigations for Spectre Variant 1
3991 (bounds check bypass). With this option data leaks are
3992 possible in the system.
3994 nospectre_v2 [X86,PPC_E500,ARM64] Disable all mitigations for
3995 the Spectre variant 2 (indirect branch prediction)
3996 vulnerability. System may allow data leaks with this
3999 no-steal-acc [X86,PV_OPS,ARM64,PPC/PSERIES,RISCV] Disable
4000 paravirtualized steal time accounting. steal time is
4001 computed, but won't influence scheduler behaviour
4003 nosync [HW,M68K] Disables sync negotiation for all devices.
4005 no_timer_check [X86,APIC] Disables the code which tests for
4006 broken timer IRQ sources.
4009 [PPC] Don't flush the L1-D cache after accessing user data.
4011 novmcoredd [KNL,KDUMP]
4012 Disable device dump. Device dump allows drivers to
4013 append dump data to vmcore so you can collect driver
4014 specified debug info. Drivers can append the data
4015 without any limit and this data is stored in memory,
4016 so this may cause significant memory stress. Disabling
4017 device dump can help save memory but the driver debug
4018 data will be no longer available. This parameter
4019 is only available when CONFIG_PROC_VMCORE_DEVICE_DUMP
4023 [X86,PV_OPS] Disable paravirtualized VMware scheduler
4024 clock and use the default one.
4026 nowatchdog [KNL] Disable both lockup detectors, i.e.
4027 soft-lockup and NMI watchdog (hard-lockup).
4031 nox2apic [X86-64,APIC] Do not enable x2APIC mode.
4033 NOTE: this parameter will be ignored on systems with the
4034 LEGACY_XAPIC_DISABLED bit set in the
4035 IA32_XAPIC_DISABLE_STATUS MSR.
4037 noxsave [BUGS=X86] Disables x86 extended register state save
4038 and restore using xsave. The kernel will fallback to
4039 enabling legacy floating-point and sse state.
4041 noxsaveopt [X86] Disables xsaveopt used in saving x86 extended
4042 register states. The kernel will fall back to use
4043 xsave to save the states. By using this parameter,
4044 performance of saving the states is degraded because
4045 xsave doesn't support modified optimization while
4046 xsaveopt supports it on xsaveopt enabled systems.
4048 noxsaves [X86] Disables xsaves and xrstors used in saving and
4049 restoring x86 extended register state in compacted
4050 form of xsave area. The kernel will fall back to use
4051 xsaveopt and xrstor to save and restore the states
4052 in standard form of xsave area. By using this
4053 parameter, xsave area per process might occupy more
4054 memory on xsaves enabled systems.
4056 nps_mtm_hs_ctr= [KNL,ARC]
4057 This parameter sets the maximum duration, in
4058 cycles, each HW thread of the CTOP can run
4059 without interruptions, before HW switches it.
4060 The actual maximum duration is 16 times this
4062 Format: integer between 1 and 255
4065 nptcg= [IA-64] Override max number of concurrent global TLB
4066 purges which is reported from either PAL_VM_SUMMARY or
4069 nr_cpus= [SMP] Maximum number of processors that an SMP kernel
4070 could support. nr_cpus=n : n >= 1 limits the kernel to
4071 support 'n' processors. It could be larger than the
4072 number of already plugged CPU during bootup, later in
4073 runtime you can physically add extra cpu until it reaches
4074 n. So during boot up some boot time memory for per-cpu
4075 variables need be pre-allocated for later physical cpu
4078 nr_uarts= [SERIAL] maximum number of UARTs to be registered.
4080 numa=off [KNL, ARM64, PPC, RISCV, SPARC, X86] Disable NUMA, Only
4081 set up a single NUMA node spanning all memory.
4083 numa_balancing= [KNL,ARM64,PPC,RISCV,S390,X86] Enable or disable automatic
4085 Allowed values are enable and disable
4087 numa_zonelist_order= [KNL, BOOT] Select zonelist order for NUMA.
4088 'node', 'default' can be specified
4089 This can be set from sysctl after boot.
4090 See Documentation/admin-guide/sysctl/vm.rst for details.
4092 ohci1394_dma=early [HW] enable debugging via the ohci1394 driver.
4093 See Documentation/core-api/debugging-via-ohci1394.rst for more
4096 olpc_ec_timeout= [OLPC] ms delay when issuing EC commands
4097 Rather than timing out after 20 ms if an EC
4098 command is not properly ACKed, override the length
4099 of the timeout. We have interrupts disabled while
4100 waiting for the ACK, so if this is set too high
4101 interrupts *may* be lost!
4103 omap_mux= [OMAP] Override bootloader pin multiplexing.
4104 Format: <mux_mode0.mode_name=value>...
4105 For example, to override I2C bus2:
4106 omap_mux=i2c2_scl.i2c2_scl=0x100,i2c2_sda.i2c2_sda=0x100
4108 onenand.bdry= [HW,MTD] Flex-OneNAND Boundary Configuration
4110 Format: [die0_boundary][,die0_lock][,die1_boundary][,die1_lock]
4112 boundary - index of last SLC block on Flex-OneNAND.
4113 The remaining blocks are configured as MLC blocks.
4114 lock - Configure if Flex-OneNAND boundary should be locked.
4115 Once locked, the boundary cannot be changed.
4116 1 indicates lock status, 0 indicates unlock status.
4118 oops=panic Always panic on oopses. Default is to just kill the
4119 process, but there is a small probability of
4120 deadlocking the machine.
4121 This will also cause panics on machine check exceptions.
4122 Useful together with panic=30 to trigger a reboot.
4125 [KNL] Boolean flag to control whether the page allocator
4126 should randomize its free lists. The randomization may
4127 be automatically enabled if the kernel detects it is
4128 running on a platform with a direct-mapped memory-side
4129 cache, and this parameter can be used to
4130 override/disable that behavior. The state of the flag
4131 can be read from sysfs at:
4132 /sys/module/page_alloc/parameters/shuffle.
4134 page_owner= [KNL] Boot-time page_owner enabling option.
4135 Storage of the information about who allocated
4136 each page is disabled in default. With this switch,
4138 on: enable the feature
4140 page_poison= [KNL] Boot-time parameter changing the state of
4141 poisoning on the buddy allocator, available with
4142 CONFIG_PAGE_POISONING=y.
4143 off: turn off poisoning (default)
4144 on: turn on poisoning
4146 page_reporting.page_reporting_order=
4147 [KNL] Minimal page reporting order
4149 Adjust the minimal page reporting order. The page
4150 reporting is disabled when it exceeds MAX_PAGE_ORDER.
4152 panic= [KNL] Kernel behaviour on panic: delay <timeout>
4153 timeout > 0: seconds before rebooting
4154 timeout = 0: wait forever
4155 timeout < 0: reboot immediately
4158 panic_on_taint= Bitmask for conditionally calling panic() in add_taint()
4159 Format: <hex>[,nousertaint]
4160 Hexadecimal bitmask representing the set of TAINT flags
4161 that will cause the kernel to panic when add_taint() is
4162 called with any of the flags in this set.
4163 The optional switch "nousertaint" can be utilized to
4164 prevent userspace forced crashes by writing to sysctl
4165 /proc/sys/kernel/tainted any flagset matching with the
4166 bitmask set on panic_on_taint.
4167 See Documentation/admin-guide/tainted-kernels.rst for
4168 extra details on the taint flags that users can pick
4169 to compose the bitmask to assign to panic_on_taint.
4171 panic_on_warn=1 panic() instead of WARN(). Useful to cause kdump
4174 panic_print= Bitmask for printing system info when panic happens.
4175 User can chose combination of the following bits:
4176 bit 0: print all tasks info
4177 bit 1: print system memory info
4178 bit 2: print timer info
4179 bit 3: print locks info if CONFIG_LOCKDEP is on
4180 bit 4: print ftrace buffer
4181 bit 5: print all printk messages in buffer
4182 bit 6: print all CPUs backtrace (if available in the arch)
4183 *Be aware* that this option may print a _lot_ of lines,
4184 so there are risks of losing older messages in the log.
4185 Use this option carefully, maybe worth to setup a
4186 bigger log buffer with "log_buf_len" along with this.
4188 parkbd.port= [HW] Parallel port number the keyboard adapter is
4189 connected to, default is 0.
4191 parkbd.mode= [HW] Parallel port keyboard adapter mode of operation,
4192 0 for XT, 1 for AT (default is AT).
4195 parport= [HW,PPT] Specify parallel ports. 0 disables.
4196 Format: { 0 | auto | 0xBBB[,IRQ[,DMA]] }
4197 Use 'auto' to force the driver to use any
4198 IRQ/DMA settings detected (the default is to
4199 ignore detected IRQ/DMA settings because of
4200 possible conflicts). You can specify the base
4201 address, IRQ, and DMA settings; IRQ and DMA
4202 should be numbers, or 'auto' (for using detected
4203 settings on that particular port), or 'nofifo'
4204 (to avoid using a FIFO even if it is detected).
4205 Parallel ports are assigned in the order they
4206 are specified on the command line, starting
4209 parport_init_mode= [HW,PPT]
4210 Configure VIA parallel port to operate in
4211 a specific mode. This is necessary on Pegasos
4212 computer where firmware has no options for setting
4213 up parallel port mode and sets it to spp.
4214 Currently this function knows 686a and 8231 chips.
4215 Format: [spp|ps2|epp|ecp|ecpepp]
4217 pata_legacy.all= [HW,LIBATA]
4219 Set to non-zero to probe primary and secondary ISA
4220 port ranges on PCI systems where no PCI PATA device
4221 has been found at either range. Disabled by default.
4223 pata_legacy.autospeed= [HW,LIBATA]
4225 Set to non-zero if a chip is present that snoops speed
4226 changes. Disabled by default.
4228 pata_legacy.ht6560a= [HW,LIBATA]
4230 Set to 1, 2, or 3 for HT 6560A on the primary channel,
4231 the secondary channel, or both channels respectively.
4232 Disabled by default.
4234 pata_legacy.ht6560b= [HW,LIBATA]
4236 Set to 1, 2, or 3 for HT 6560B on the primary channel,
4237 the secondary channel, or both channels respectively.
4238 Disabled by default.
4240 pata_legacy.iordy_mask= [HW,LIBATA]
4242 IORDY enable mask. Set individual bits to allow IORDY
4243 for the respective channel. Bit 0 is for the first
4244 legacy channel handled by this driver, bit 1 is for
4245 the second channel, and so on. The sequence will often
4246 correspond to the primary legacy channel, the secondary
4247 legacy channel, and so on, but the handling of a PCI
4248 bus and the use of other driver options may interfere
4249 with the sequence. By default IORDY is allowed across
4252 pata_legacy.opti82c46x= [HW,LIBATA]
4254 Set to 1, 2, or 3 for Opti 82c611A on the primary
4255 channel, the secondary channel, or both channels
4256 respectively. Disabled by default.
4258 pata_legacy.opti82c611a= [HW,LIBATA]
4260 Set to 1, 2, or 3 for Opti 82c465MV on the primary
4261 channel, the secondary channel, or both channels
4262 respectively. Disabled by default.
4264 pata_legacy.pio_mask= [HW,LIBATA]
4266 PIO mode mask for autospeed devices. Set individual
4267 bits to allow the use of the respective PIO modes.
4268 Bit 0 is for mode 0, bit 1 is for mode 1, and so on.
4269 All modes allowed by default.
4271 pata_legacy.probe_all= [HW,LIBATA]
4273 Set to non-zero to probe tertiary and further ISA
4274 port ranges on PCI systems. Disabled by default.
4276 pata_legacy.probe_mask= [HW,LIBATA]
4278 Probe mask for legacy ISA PATA ports. Depending on
4279 platform configuration and the use of other driver
4280 options up to 6 legacy ports are supported: 0x1f0,
4281 0x170, 0x1e8, 0x168, 0x1e0, 0x160, however probing
4282 of individual ports can be disabled by setting the
4283 corresponding bits in the mask to 1. Bit 0 is for
4284 the first port in the list above (0x1f0), and so on.
4285 By default all supported ports are probed.
4287 pata_legacy.qdi= [HW,LIBATA]
4289 Set to non-zero to probe QDI controllers. By default
4290 set to 1 if CONFIG_PATA_QDI_MODULE, 0 otherwise.
4292 pata_legacy.winbond= [HW,LIBATA]
4294 Set to non-zero to probe Winbond controllers. Use
4295 the standard I/O port (0x130) if 1, otherwise the
4296 value given is the I/O port to use (typically 0x1b0).
4297 By default set to 1 if CONFIG_PATA_WINBOND_VLB_MODULE,
4300 pata_platform.pio_mask= [HW,LIBATA]
4302 Supported PIO mode mask. Set individual bits to allow
4303 the use of the respective PIO modes. Bit 0 is for
4304 mode 0, bit 1 is for mode 1, and so on. Mode 0 only
4308 Halt all CPUs after the first oops has been printed for
4309 the specified number of seconds. This is to be used if
4310 your oopses keep scrolling off the screen.
4314 pci=option[,option...] [PCI] various PCI subsystem options.
4316 Some options herein operate on a specific device
4317 or a set of devices (<pci_dev>). These are
4318 specified in one of the following formats:
4320 [<domain>:]<bus>:<dev>.<func>[/<dev>.<func>]*
4321 pci:<vendor>:<device>[:<subvendor>:<subdevice>]
4323 Note: the first format specifies a PCI
4324 bus/device/function address which may change
4325 if new hardware is inserted, if motherboard
4326 firmware changes, or due to changes caused
4327 by other kernel parameters. If the
4328 domain is left unspecified, it is
4329 taken to be zero. Optionally, a path
4330 to a device through multiple device/function
4331 addresses can be specified after the base
4332 address (this is more robust against
4333 renumbering issues). The second format
4334 selects devices using IDs from the
4335 configuration space which may match multiple
4336 devices in the system.
4338 earlydump dump PCI config space before the kernel
4340 off [X86] don't probe for the PCI bus
4341 bios [X86-32] force use of PCI BIOS, don't access
4342 the hardware directly. Use this if your machine
4343 has a non-standard PCI host bridge.
4344 nobios [X86-32] disallow use of PCI BIOS, only direct
4345 hardware access methods are allowed. Use this
4346 if you experience crashes upon bootup and you
4347 suspect they are caused by the BIOS.
4348 conf1 [X86] Force use of PCI Configuration Access
4349 Mechanism 1 (config address in IO port 0xCF8,
4350 data in IO port 0xCFC, both 32-bit).
4351 conf2 [X86] Force use of PCI Configuration Access
4352 Mechanism 2 (IO port 0xCF8 is an 8-bit port for
4353 the function, IO port 0xCFA, also 8-bit, sets
4354 bus number. The config space is then accessed
4355 through ports 0xC000-0xCFFF).
4356 See http://wiki.osdev.org/PCI for more info
4357 on the configuration access mechanisms.
4358 noaer [PCIE] If the PCIEAER kernel config parameter is
4359 enabled, this kernel boot option can be used to
4360 disable the use of PCIE advanced error reporting.
4361 nodomains [PCI] Disable support for multiple PCI
4362 root domains (aka PCI segments, in ACPI-speak).
4363 nommconf [X86] Disable use of MMCONFIG for PCI
4365 check_enable_amd_mmconf [X86] check for and enable
4366 properly configured MMIO access to PCI
4367 config space on AMD family 10h CPU
4368 nomsi [MSI] If the PCI_MSI kernel config parameter is
4369 enabled, this kernel boot option can be used to
4370 disable the use of MSI interrupts system-wide.
4371 noioapicquirk [APIC] Disable all boot interrupt quirks.
4372 Safety option to keep boot IRQs enabled. This
4373 should never be necessary.
4374 ioapicreroute [APIC] Enable rerouting of boot IRQs to the
4375 primary IO-APIC for bridges that cannot disable
4376 boot IRQs. This fixes a source of spurious IRQs
4377 when the system masks IRQs.
4378 noioapicreroute [APIC] Disable workaround that uses the
4379 boot IRQ equivalent of an IRQ that connects to
4380 a chipset where boot IRQs cannot be disabled.
4381 The opposite of ioapicreroute.
4382 biosirq [X86-32] Use PCI BIOS calls to get the interrupt
4383 routing table. These calls are known to be buggy
4384 on several machines and they hang the machine
4385 when used, but on other computers it's the only
4386 way to get the interrupt routing table. Try
4387 this option if the kernel is unable to allocate
4388 IRQs or discover secondary PCI buses on your
4390 rom [X86] Assign address space to expansion ROMs.
4391 Use with caution as certain devices share
4392 address decoders between ROMs and other
4394 norom [X86] Do not assign address space to
4395 expansion ROMs that do not already have
4396 BIOS assigned address ranges.
4397 nobar [X86] Do not assign address space to the
4398 BARs that weren't assigned by the BIOS.
4399 irqmask=0xMMMM [X86] Set a bit mask of IRQs allowed to be
4400 assigned automatically to PCI devices. You can
4401 make the kernel exclude IRQs of your ISA cards
4403 pirqaddr=0xAAAAA [X86] Specify the physical address
4404 of the PIRQ table (normally generated
4405 by the BIOS) if it is outside the
4406 F0000h-100000h range.
4407 lastbus=N [X86] Scan all buses thru bus #N. Can be
4408 useful if the kernel is unable to find your
4409 secondary buses and you want to tell it
4410 explicitly which ones they are.
4411 assign-busses [X86] Always assign all PCI bus
4412 numbers ourselves, overriding
4413 whatever the firmware may have done.
4414 usepirqmask [X86] Honor the possible IRQ mask stored
4415 in the BIOS $PIR table. This is needed on
4416 some systems with broken BIOSes, notably
4417 some HP Pavilion N5400 and Omnibook XE3
4418 notebooks. This will have no effect if ACPI
4419 IRQ routing is enabled.
4420 noacpi [X86] Do not use ACPI for IRQ routing
4421 or for PCI scanning.
4422 use_crs [X86] Use PCI host bridge window information
4423 from ACPI. On BIOSes from 2008 or later, this
4424 is enabled by default. If you need to use this,
4425 please report a bug.
4426 nocrs [X86] Ignore PCI host bridge windows from ACPI.
4427 If you need to use this, please report a bug.
4428 use_e820 [X86] Use E820 reservations to exclude parts of
4429 PCI host bridge windows. This is a workaround
4430 for BIOS defects in host bridge _CRS methods.
4431 If you need to use this, please report a bug to
4432 <linux-pci@vger.kernel.org>.
4433 no_e820 [X86] Ignore E820 reservations for PCI host
4434 bridge windows. This is the default on modern
4435 hardware. If you need to use this, please report
4436 a bug to <linux-pci@vger.kernel.org>.
4437 routeirq Do IRQ routing for all PCI devices.
4438 This is normally done in pci_enable_device(),
4439 so this option is a temporary workaround
4440 for broken drivers that don't call it.
4441 skip_isa_align [X86] do not align io start addr, so can
4442 handle more pci cards
4443 noearly [X86] Don't do any early type 1 scanning.
4444 This might help on some broken boards which
4445 machine check when some devices' config space
4446 is read. But various workarounds are disabled
4447 and some IOMMU drivers will not work.
4448 bfsort Sort PCI devices into breadth-first order.
4449 This sorting is done to get a device
4450 order compatible with older (<= 2.4) kernels.
4451 nobfsort Don't sort PCI devices into breadth-first order.
4452 pcie_bus_tune_off Disable PCIe MPS (Max Payload Size)
4453 tuning and use the BIOS-configured MPS defaults.
4454 pcie_bus_safe Set every device's MPS to the largest value
4455 supported by all devices below the root complex.
4456 pcie_bus_perf Set device MPS to the largest allowable MPS
4457 based on its parent bus. Also set MRRS (Max
4458 Read Request Size) to the largest supported
4459 value (no larger than the MPS that the device
4460 or bus can support) for best performance.
4461 pcie_bus_peer2peer Set every device's MPS to 128B, which
4462 every device is guaranteed to support. This
4463 configuration allows peer-to-peer DMA between
4464 any pair of devices, possibly at the cost of
4465 reduced performance. This also guarantees
4466 that hot-added devices will work.
4467 cbiosize=nn[KMG] The fixed amount of bus space which is
4468 reserved for the CardBus bridge's IO window.
4469 The default value is 256 bytes.
4470 cbmemsize=nn[KMG] The fixed amount of bus space which is
4471 reserved for the CardBus bridge's memory
4472 window. The default value is 64 megabytes.
4475 [<order of align>@]<pci_dev>[; ...]
4476 Specifies alignment and device to reassign
4477 aligned memory resources. How to
4478 specify the device is described above.
4479 If <order of align> is not specified,
4480 PAGE_SIZE is used as alignment.
4481 A PCI-PCI bridge can be specified if resource
4482 windows need to be expanded.
4483 To specify the alignment for several
4484 instances of a device, the PCI vendor,
4485 device, subvendor, and subdevice may be
4486 specified, e.g., 12@pci:8086:9c22:103c:198f
4487 for 4096-byte alignment.
4488 ecrc= Enable/disable PCIe ECRC (transaction layer
4489 end-to-end CRC checking). Only effective if
4490 OS has native AER control (either granted by
4491 ACPI _OSC or forced via "pcie_ports=native")
4492 bios: Use BIOS/firmware settings. This is the
4496 hpiosize=nn[KMG] The fixed amount of bus space which is
4497 reserved for hotplug bridge's IO window.
4498 Default size is 256 bytes.
4499 hpmmiosize=nn[KMG] The fixed amount of bus space which is
4500 reserved for hotplug bridge's MMIO window.
4501 Default size is 2 megabytes.
4502 hpmmioprefsize=nn[KMG] The fixed amount of bus space which is
4503 reserved for hotplug bridge's MMIO_PREF window.
4504 Default size is 2 megabytes.
4505 hpmemsize=nn[KMG] The fixed amount of bus space which is
4506 reserved for hotplug bridge's MMIO and
4508 Default size is 2 megabytes.
4509 hpbussize=nn The minimum amount of additional bus numbers
4510 reserved for buses below a hotplug bridge.
4512 realloc= Enable/disable reallocating PCI bridge resources
4513 if allocations done by BIOS are too small to
4514 accommodate resources required by all child
4516 off: Turn realloc off
4518 realloc same as realloc=on
4519 noari do not use PCIe ARI.
4520 noats [PCIE, Intel-IOMMU, AMD-IOMMU]
4521 do not use PCIe ATS (and IOMMU device IOTLB).
4522 pcie_scan_all Scan all possible PCIe devices. Otherwise we
4523 only look for one device below a PCIe downstream
4525 big_root_window Try to add a big 64bit memory window to the PCIe
4526 root complex on AMD CPUs. Some GFX hardware
4527 can resize a BAR to allow access to all VRAM.
4528 Adding the window is slightly risky (it may
4529 conflict with unreported devices), so this
4531 disable_acs_redir=<pci_dev>[; ...]
4532 Specify one or more PCI devices (in the format
4533 specified above) separated by semicolons.
4534 Each device specified will have the PCI ACS
4535 redirect capabilities forced off which will
4536 allow P2P traffic between devices through
4537 bridges without forcing it upstream. Note:
4538 this removes isolation between devices and
4539 may put more devices in an IOMMU group.
4540 force_floating [S390] Force usage of floating interrupts.
4541 nomio [S390] Do not use MIO instructions.
4542 norid [S390] ignore the RID field and force use of
4543 one PCI domain per PCI function
4545 pcie_aspm= [PCIE] Forcibly enable or disable PCIe Active State Power
4548 force Enable ASPM even on devices that claim not to support it.
4549 WARNING: Forcing ASPM on may cause system lockups.
4551 pcie_ports= [PCIE] PCIe port services handling:
4552 native Use native PCIe services (PME, AER, DPC, PCIe hotplug)
4553 even if the platform doesn't give the OS permission to
4554 use them. This may cause conflicts if the platform
4555 also tries to use these services.
4556 dpc-native Use native PCIe service for DPC only. May
4557 cause conflicts if firmware uses AER or DPC.
4558 compat Disable native PCIe services (PME, AER, DPC, PCIe
4561 pcie_port_pm= [PCIE] PCIe port power management handling:
4562 off Disable power management of all PCIe ports
4563 force Forcibly enable power management of all PCIe ports
4565 pcie_pme= [PCIE,PM] Native PCIe PME signaling options:
4566 nomsi Do not use MSI for native PCIe PME signaling (this makes
4567 all PCIe root ports use INTx for all services).
4569 pcmv= [HW,PCMCIA] BadgePAD 4
4573 Keep all power-domains already enabled by bootloader on,
4574 even if no driver has claimed them. This is useful
4575 for debug and development, but should not be
4576 needed on a platform with proper driver support.
4578 pdcchassis= [PARISC,HW] Disable/Enable PDC Chassis Status codes at
4581 See arch/parisc/kernel/pdc_chassis.c
4583 percpu_alloc= Select which percpu first chunk allocator to use.
4584 Currently supported values are "embed" and "page".
4585 Archs may support subset or none of the selections.
4586 See comments in mm/percpu.c for details on each
4587 allocator. This parameter is primarily for debugging
4588 and performance comparison.
4590 pirq= [SMP,APIC] Manual mp-table setup
4591 See Documentation/arch/x86/i386/IO-APIC.rst.
4593 plip= [PPT,NET] Parallel port network link
4594 Format: { parport<nr> | timid | 0 }
4595 See also Documentation/admin-guide/parport.rst.
4597 pmtmr= [X86] Manual setup of pmtmr I/O Port.
4598 Override pmtimer IOPort with a hex value.
4601 pmu_override= [PPC] Override the PMU.
4602 This option takes over the PMU facility, so it is no
4603 longer usable by perf. Setting this option starts the
4604 PMU counters by setting MMCR0 to 0 (the FC bit is
4605 cleared). If a number is given, then MMCR1 is set to
4606 that number, otherwise (e.g., 'pmu_override=on'), MMCR1
4609 pm_debug_messages [SUSPEND,KNL]
4610 Enable suspend/resume debug messages during boot up.
4613 Enable PNP debug messages (depends on the
4614 CONFIG_PNP_DEBUG_MESSAGES option). Change at run-time
4615 via /sys/module/pnp/parameters/debug. We always show
4616 current resource usage; turning this on also shows
4617 possible settings and some assignment information.
4623 { on | off | curr | res | no-curr | no-res }
4626 [ISAPNP] Exclude IRQs for the autoconfiguration
4629 [ISAPNP] Exclude DMAs for the autoconfiguration
4631 pnp_reserve_io= [ISAPNP] Exclude I/O ports for the autoconfiguration
4632 Ranges are in pairs (I/O port base and size).
4635 [ISAPNP] Exclude memory regions for the
4637 Ranges are in pairs (memory base and size).
4639 ports= [IP_VS_FTP] IPVS ftp helper module
4641 Up to 8 (IP_VS_APP_MAX_PORTS) ports
4643 Format: <port>,<port>....
4645 powersave=off [PPC] This option disables power saving features.
4646 It specifically disables cpuidle and sets the
4647 platform machine description specific power_save
4648 function to NULL. On Idle the CPU just reduces
4651 ppc_strict_facility_enable
4652 [PPC] This option catches any kernel floating point,
4653 Altivec, VSX and SPE outside of regions specifically
4654 allowed (eg kernel_enable_fpu()/kernel_disable_fpu()).
4655 There is some performance impact when enabling this.
4659 Disable Hardware Transactional Memory
4662 Select preemption mode if you have CONFIG_PREEMPT_DYNAMIC
4663 none - Limited to cond_resched() calls
4664 voluntary - Limited to cond_resched() and might_sleep() calls
4665 full - Any section that isn't explicitly preempt disabled
4666 can be preempted anytime.
4668 print-fatal-signals=
4669 [KNL] debug: print fatal signals
4671 If enabled, warn about various signal handling
4672 related application anomalies: too many signals,
4673 too many POSIX.1 timers, fatal signals causing a
4676 If you hit the warning due to signal overflow,
4677 you might want to try "ulimit -i unlimited".
4681 printk.always_kmsg_dump=
4682 Trigger kmsg_dump for cases other than kernel oops or
4684 Format: <bool> (1/Y/y=enable, 0/N/n=disable)
4687 printk.console_no_auto_verbose=
4688 Disable console loglevel raise on oops, panic
4689 or lockdep-detected issues (only if lock debug is on).
4690 With an exception to setups with low baudrate on
4691 serial console, keeping this 0 is a good choice
4692 in order to provide more debug information.
4694 default: 0 (auto_verbose is enabled)
4696 printk.devkmsg={on,off,ratelimit}
4697 Control writing to /dev/kmsg.
4698 on - unlimited logging to /dev/kmsg from userspace
4699 off - logging to /dev/kmsg disabled
4700 ratelimit - ratelimit the logging
4703 printk.time= Show timing data prefixed to each printk message line
4704 Format: <bool> (1/Y/y=enable, 0/N/n=disable)
4706 processor.max_cstate= [HW,ACPI]
4707 Limit processor to maximum C-state
4708 max_cstate=9 overrides any DMI blacklist limit.
4710 processor.nocst [HW,ACPI]
4711 Ignore the _CST method to determine C-states,
4712 instead using the legacy FADT method
4714 profile= [KNL] Enable kernel profiling via /proc/profile
4715 Format: [<profiletype>,]<number>
4716 Param: <profiletype>: "schedule", "sleep", or "kvm"
4717 [defaults to kernel profiling]
4718 Param: "schedule" - profile schedule points.
4719 Param: "sleep" - profile D-state sleeping (millisecs).
4720 Requires CONFIG_SCHEDSTATS
4721 Param: "kvm" - profile VM exits.
4722 Param: <number> - step/bucket size as a power of 2 for
4723 statistical time based profiling.
4725 prompt_ramdisk= [RAM] [Deprecated]
4727 prot_virt= [S390] enable hosting protected virtual machines
4728 isolated from the hypervisor (if hardware supports
4732 psi= [KNL] Enable or disable pressure stall information
4736 psmouse.proto= [HW,MOUSE] Highest PS2 mouse protocol extension to
4737 probe for; one of (bare|imps|exps|lifebook|any).
4738 psmouse.rate= [HW,MOUSE] Set desired mouse report rate, in reports
4740 psmouse.resetafter= [HW,MOUSE]
4741 Try to reset the device after so many bad packets
4744 [HW,MOUSE] Set desired mouse resolution, in dpi.
4745 psmouse.smartscroll=
4746 [HW,MOUSE] Controls Logitech smartscroll autorepeat.
4747 0 = disabled, 1 = enabled (default).
4749 pstore.backend= Specify the name of the pstore backend to use
4751 pti= [X86-64] Control Page Table Isolation of user and
4752 kernel address spaces. Disabling this feature
4753 removes hardening, but improves performance of
4754 system calls and interrupts.
4756 on - unconditionally enable
4757 off - unconditionally disable
4758 auto - kernel detects whether your CPU model is
4759 vulnerable to issues that PTI mitigates
4761 Not specifying this option is equivalent to pti=auto.
4764 [KNL] Number of legacy pty's. Overwrites compiled-in
4767 quiet [KNL] Disable most log messages
4771 radix_hcall_invalidate=on [PPC/PSERIES]
4772 Disable RADIX GTSE feature and use hcall for TLB
4776 See Documentation/admin-guide/md.rst.
4778 ramdisk_size= [RAM] Sizes of RAM disks in kilobytes
4779 See Documentation/admin-guide/blockdev/ramdisk.rst.
4781 ramdisk_start= [RAM] RAM disk image start address
4783 random.trust_cpu=off
4784 [KNL] Disable trusting the use of the CPU's
4785 random number generator (if available) to
4786 initialize the kernel's RNG.
4788 random.trust_bootloader=off
4789 [KNL] Disable trusting the use of the a seed
4790 passed by the bootloader (if available) to
4791 initialize the kernel's RNG.
4793 randomize_kstack_offset=
4794 [KNL] Enable or disable kernel stack offset
4795 randomization, which provides roughly 5 bits of
4796 entropy, frustrating memory corruption attacks
4797 that depend on stack address determinism or
4798 cross-syscall address exposures. This is only
4799 available on architectures that have defined
4800 CONFIG_HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET.
4801 Format: <bool> (1/Y/y=enable, 0/N/n=disable)
4802 Default is CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT.
4804 ras=option[,option,...] [KNL] RAS-specific options
4807 Disable the Correctable Errors Collector,
4808 see CONFIG_RAS_CEC help text.
4810 rcu_nocbs[=cpu-list]
4811 [KNL] The optional argument is a cpu list,
4814 In kernels built with CONFIG_RCU_NOCB_CPU=y,
4815 enable the no-callback CPU mode, which prevents
4816 such CPUs' callbacks from being invoked in
4817 softirq context. Invocation of such CPUs' RCU
4818 callbacks will instead be offloaded to "rcuox/N"
4819 kthreads created for that purpose, where "x" is
4820 "p" for RCU-preempt, "s" for RCU-sched, and "g"
4821 for the kthreads that mediate grace periods; and
4822 "N" is the CPU number. This reduces OS jitter on
4823 the offloaded CPUs, which can be useful for HPC
4824 and real-time workloads. It can also improve
4825 energy efficiency for asymmetric multiprocessors.
4827 If a cpulist is passed as an argument, the specified
4828 list of CPUs is set to no-callback mode from boot.
4830 Otherwise, if the '=' sign and the cpulist
4831 arguments are omitted, no CPU will be set to
4832 no-callback mode from boot but the mode may be
4833 toggled at runtime via cpusets.
4835 Note that this argument takes precedence over
4836 the CONFIG_RCU_NOCB_CPU_DEFAULT_ALL option.
4839 Rather than requiring that offloaded CPUs
4840 (specified by rcu_nocbs= above) explicitly
4841 awaken the corresponding "rcuoN" kthreads,
4842 make these kthreads poll for callbacks.
4843 This improves the real-time response for the
4844 offloaded CPUs by relieving them of the need to
4845 wake up the corresponding kthread, but degrades
4846 energy efficiency by requiring that the kthreads
4847 periodically wake up to do the polling.
4849 rcutree.blimit= [KNL]
4850 Set maximum number of finished RCU callbacks to
4851 process in one batch.
4853 rcutree.do_rcu_barrier= [KNL]
4854 Request a call to rcu_barrier(). This is
4855 throttled so that userspace tests can safely
4856 hammer on the sysfs variable if they so choose.
4857 If triggered before the RCU grace-period machinery
4858 is fully active, this will error out with EAGAIN.
4860 rcutree.dump_tree= [KNL]
4861 Dump the structure of the rcu_node combining tree
4862 out at early boot. This is used for diagnostic
4863 purposes, to verify correct tree setup.
4865 rcutree.gp_cleanup_delay= [KNL]
4866 Set the number of jiffies to delay each step of
4867 RCU grace-period cleanup.
4869 rcutree.gp_init_delay= [KNL]
4870 Set the number of jiffies to delay each step of
4871 RCU grace-period initialization.
4873 rcutree.gp_preinit_delay= [KNL]
4874 Set the number of jiffies to delay each step of
4875 RCU grace-period pre-initialization, that is,
4876 the propagation of recent CPU-hotplug changes up
4877 the rcu_node combining tree.
4879 rcutree.jiffies_till_first_fqs= [KNL]
4880 Set delay from grace-period initialization to
4881 first attempt to force quiescent states.
4882 Units are jiffies, minimum value is zero,
4883 and maximum value is HZ.
4885 rcutree.jiffies_till_next_fqs= [KNL]
4886 Set delay between subsequent attempts to force
4887 quiescent states. Units are jiffies, minimum
4888 value is one, and maximum value is HZ.
4890 rcutree.jiffies_till_sched_qs= [KNL]
4891 Set required age in jiffies for a
4892 given grace period before RCU starts
4893 soliciting quiescent-state help from
4894 rcu_note_context_switch() and cond_resched().
4895 If not specified, the kernel will calculate
4896 a value based on the most recent settings
4897 of rcutree.jiffies_till_first_fqs
4898 and rcutree.jiffies_till_next_fqs.
4899 This calculated value may be viewed in
4900 rcutree.jiffies_to_sched_qs. Any attempt to set
4901 rcutree.jiffies_to_sched_qs will be cheerfully
4904 rcutree.kthread_prio= [KNL,BOOT]
4905 Set the SCHED_FIFO priority of the RCU per-CPU
4906 kthreads (rcuc/N). This value is also used for
4907 the priority of the RCU boost threads (rcub/N)
4908 and for the RCU grace-period kthreads (rcu_bh,
4909 rcu_preempt, and rcu_sched). If RCU_BOOST is
4910 set, valid values are 1-99 and the default is 1
4911 (the least-favored priority). Otherwise, when
4912 RCU_BOOST is not set, valid values are 0-99 and
4913 the default is zero (non-realtime operation).
4914 When RCU_NOCB_CPU is set, also adjust the
4915 priority of NOCB callback kthreads.
4917 rcutree.nocb_nobypass_lim_per_jiffy= [KNL]
4918 On callback-offloaded (rcu_nocbs) CPUs,
4919 RCU reduces the lock contention that would
4920 otherwise be caused by callback floods through
4921 use of the ->nocb_bypass list. However, in the
4922 common non-flooded case, RCU queues directly to
4923 the main ->cblist in order to avoid the extra
4924 overhead of the ->nocb_bypass list and its lock.
4925 But if there are too many callbacks queued during
4926 a single jiffy, RCU pre-queues the callbacks into
4927 the ->nocb_bypass queue. The definition of "too
4928 many" is supplied by this kernel boot parameter.
4930 rcutree.qhimark= [KNL]
4931 Set threshold of queued RCU callbacks beyond which
4932 batch limiting is disabled.
4934 rcutree.qlowmark= [KNL]
4935 Set threshold of queued RCU callbacks below which
4936 batch limiting is re-enabled.
4938 rcutree.qovld= [KNL]
4939 Set threshold of queued RCU callbacks beyond which
4940 RCU's force-quiescent-state scan will aggressively
4941 enlist help from cond_resched() and sched IPIs to
4942 help CPUs more quickly reach quiescent states.
4943 Set to less than zero to make this be set based
4944 on rcutree.qhimark at boot time and to zero to
4945 disable more aggressive help enlistment.
4947 rcutree.rcu_delay_page_cache_fill_msec= [KNL]
4948 Set the page-cache refill delay (in milliseconds)
4949 in response to low-memory conditions. The range
4950 of permitted values is in the range 0:100000.
4952 rcutree.rcu_divisor= [KNL]
4953 Set the shift-right count to use to compute
4954 the callback-invocation batch limit bl from
4955 the number of callbacks queued on this CPU.
4956 The result will be bounded below by the value of
4957 the rcutree.blimit kernel parameter. Every bl
4958 callbacks, the softirq handler will exit in
4959 order to allow the CPU to do other work.
4961 Please note that this callback-invocation batch
4962 limit applies only to non-offloaded callback
4963 invocation. Offloaded callbacks are instead
4964 invoked in the context of an rcuoc kthread, which
4965 scheduler will preempt as it does any other task.
4967 rcutree.rcu_fanout_exact= [KNL]
4968 Disable autobalancing of the rcu_node combining
4969 tree. This is used by rcutorture, and might
4970 possibly be useful for architectures having high
4971 cache-to-cache transfer latencies.
4973 rcutree.rcu_fanout_leaf= [KNL]
4974 Change the number of CPUs assigned to each
4975 leaf rcu_node structure. Useful for very
4976 large systems, which will choose the value 64,
4977 and for NUMA systems with large remote-access
4978 latencies, which will choose a value aligned
4979 with the appropriate hardware boundaries.
4981 rcutree.rcu_min_cached_objs= [KNL]
4982 Minimum number of objects which are cached and
4983 maintained per one CPU. Object size is equal
4984 to PAGE_SIZE. The cache allows to reduce the
4985 pressure to page allocator, also it makes the
4986 whole algorithm to behave better in low memory
4989 rcutree.rcu_nocb_gp_stride= [KNL]
4990 Set the number of NOCB callback kthreads in
4991 each group, which defaults to the square root
4992 of the number of CPUs. Larger numbers reduce
4993 the wakeup overhead on the global grace-period
4994 kthread, but increases that same overhead on
4995 each group's NOCB grace-period kthread.
4997 rcutree.rcu_kick_kthreads= [KNL]
4998 Cause the grace-period kthread to get an extra
4999 wake_up() if it sleeps three times longer than
5000 it should at force-quiescent-state time.
5001 This wake_up() will be accompanied by a
5002 WARN_ONCE() splat and an ftrace_dump().
5004 rcutree.rcu_resched_ns= [KNL]
5005 Limit the time spend invoking a batch of RCU
5006 callbacks to the specified number of nanoseconds.
5007 By default, this limit is checked only once
5008 every 32 callbacks in order to limit the pain
5009 inflicted by local_clock() overhead.
5011 rcutree.rcu_unlock_delay= [KNL]
5012 In CONFIG_RCU_STRICT_GRACE_PERIOD=y kernels,
5013 this specifies an rcu_read_unlock()-time delay
5014 in microseconds. This defaults to zero.
5015 Larger delays increase the probability of
5016 catching RCU pointer leaks, that is, buggy use
5017 of RCU-protected pointers after the relevant
5018 rcu_read_unlock() has completed.
5020 rcutree.sysrq_rcu= [KNL]
5021 Commandeer a sysrq key to dump out Tree RCU's
5022 rcu_node tree with an eye towards determining
5023 why a new grace period has not yet started.
5025 rcutree.use_softirq= [KNL]
5026 If set to zero, move all RCU_SOFTIRQ processing to
5027 per-CPU rcuc kthreads. Defaults to a non-zero
5028 value, meaning that RCU_SOFTIRQ is used by default.
5029 Specify rcutree.use_softirq=0 to use rcuc kthreads.
5031 But note that CONFIG_PREEMPT_RT=y kernels disable
5032 this kernel boot parameter, forcibly setting it
5035 rcuscale.gp_async= [KNL]
5036 Measure performance of asynchronous
5037 grace-period primitives such as call_rcu().
5039 rcuscale.gp_async_max= [KNL]
5040 Specify the maximum number of outstanding
5041 callbacks per writer thread. When a writer
5042 thread exceeds this limit, it invokes the
5043 corresponding flavor of rcu_barrier() to allow
5044 previously posted callbacks to drain.
5046 rcuscale.gp_exp= [KNL]
5047 Measure performance of expedited synchronous
5048 grace-period primitives.
5050 rcuscale.holdoff= [KNL]
5051 Set test-start holdoff period. The purpose of
5052 this parameter is to delay the start of the
5053 test until boot completes in order to avoid
5056 rcuscale.kfree_by_call_rcu= [KNL]
5057 In kernels built with CONFIG_RCU_LAZY=y, test
5058 call_rcu() instead of kfree_rcu().
5060 rcuscale.kfree_mult= [KNL]
5061 Instead of allocating an object of size kfree_obj,
5062 allocate one of kfree_mult * sizeof(kfree_obj).
5065 rcuscale.kfree_rcu_test= [KNL]
5066 Set to measure performance of kfree_rcu() flooding.
5068 rcuscale.kfree_rcu_test_double= [KNL]
5069 Test the double-argument variant of kfree_rcu().
5070 If this parameter has the same value as
5071 rcuscale.kfree_rcu_test_single, both the single-
5072 and double-argument variants are tested.
5074 rcuscale.kfree_rcu_test_single= [KNL]
5075 Test the single-argument variant of kfree_rcu().
5076 If this parameter has the same value as
5077 rcuscale.kfree_rcu_test_double, both the single-
5078 and double-argument variants are tested.
5080 rcuscale.kfree_nthreads= [KNL]
5081 The number of threads running loops of kfree_rcu().
5083 rcuscale.kfree_alloc_num= [KNL]
5084 Number of allocations and frees done in an iteration.
5086 rcuscale.kfree_loops= [KNL]
5087 Number of loops doing rcuscale.kfree_alloc_num number
5088 of allocations and frees.
5090 rcuscale.minruntime= [KNL]
5091 Set the minimum test run time in seconds. This
5092 does not affect the data-collection interval,
5093 but instead allows better measurement of things
5094 like CPU consumption.
5096 rcuscale.nreaders= [KNL]
5097 Set number of RCU readers. The value -1 selects
5098 N, where N is the number of CPUs. A value
5099 "n" less than -1 selects N-n+1, where N is again
5100 the number of CPUs. For example, -2 selects N
5101 (the number of CPUs), -3 selects N+1, and so on.
5102 A value of "n" less than or equal to -N selects
5105 rcuscale.nwriters= [KNL]
5106 Set number of RCU writers. The values operate
5107 the same as for rcuscale.nreaders.
5108 N, where N is the number of CPUs
5110 rcuscale.scale_type= [KNL]
5111 Specify the RCU implementation to test.
5113 rcuscale.shutdown= [KNL]
5114 Shut the system down after performance tests
5115 complete. This is useful for hands-off automated
5118 rcuscale.verbose= [KNL]
5119 Enable additional printk() statements.
5121 rcuscale.writer_holdoff= [KNL]
5122 Write-side holdoff between grace periods,
5123 in microseconds. The default of zero says
5126 rcuscale.writer_holdoff_jiffies= [KNL]
5127 Additional write-side holdoff between grace
5128 periods, but in jiffies. The default of zero
5131 rcutorture.fqs_duration= [KNL]
5132 Set duration of force_quiescent_state bursts
5135 rcutorture.fqs_holdoff= [KNL]
5136 Set holdoff time within force_quiescent_state bursts
5139 rcutorture.fqs_stutter= [KNL]
5140 Set wait time between force_quiescent_state bursts
5143 rcutorture.fwd_progress= [KNL]
5144 Specifies the number of kthreads to be used
5145 for RCU grace-period forward-progress testing
5146 for the types of RCU supporting this notion.
5147 Defaults to 1 kthread, values less than zero or
5148 greater than the number of CPUs cause the number
5151 rcutorture.fwd_progress_div= [KNL]
5152 Specify the fraction of a CPU-stall-warning
5153 period to do tight-loop forward-progress testing.
5155 rcutorture.fwd_progress_holdoff= [KNL]
5156 Number of seconds to wait between successive
5157 forward-progress tests.
5159 rcutorture.fwd_progress_need_resched= [KNL]
5160 Enclose cond_resched() calls within checks for
5161 need_resched() during tight-loop forward-progress
5164 rcutorture.gp_cond= [KNL]
5165 Use conditional/asynchronous update-side
5166 primitives, if available.
5168 rcutorture.gp_exp= [KNL]
5169 Use expedited update-side primitives, if available.
5171 rcutorture.gp_normal= [KNL]
5172 Use normal (non-expedited) asynchronous
5173 update-side primitives, if available.
5175 rcutorture.gp_sync= [KNL]
5176 Use normal (non-expedited) synchronous
5177 update-side primitives, if available. If all
5178 of rcutorture.gp_cond=, rcutorture.gp_exp=,
5179 rcutorture.gp_normal=, and rcutorture.gp_sync=
5180 are zero, rcutorture acts as if is interpreted
5181 they are all non-zero.
5183 rcutorture.irqreader= [KNL]
5184 Run RCU readers from irq handlers, or, more
5185 accurately, from a timer handler. Not all RCU
5186 flavors take kindly to this sort of thing.
5188 rcutorture.leakpointer= [KNL]
5189 Leak an RCU-protected pointer out of the reader.
5190 This can of course result in splats, and is
5191 intended to test the ability of things like
5192 CONFIG_RCU_STRICT_GRACE_PERIOD=y to detect
5195 rcutorture.n_barrier_cbs= [KNL]
5196 Set callbacks/threads for rcu_barrier() testing.
5198 rcutorture.nfakewriters= [KNL]
5199 Set number of concurrent RCU writers. These just
5200 stress RCU, they don't participate in the actual
5201 test, hence the "fake".
5203 rcutorture.nocbs_nthreads= [KNL]
5204 Set number of RCU callback-offload togglers.
5205 Zero (the default) disables toggling.
5207 rcutorture.nocbs_toggle= [KNL]
5208 Set the delay in milliseconds between successive
5209 callback-offload toggling attempts.
5211 rcutorture.nreaders= [KNL]
5212 Set number of RCU readers. The value -1 selects
5213 N-1, where N is the number of CPUs. A value
5214 "n" less than -1 selects N-n-2, where N is again
5215 the number of CPUs. For example, -2 selects N
5216 (the number of CPUs), -3 selects N+1, and so on.
5218 rcutorture.object_debug= [KNL]
5219 Enable debug-object double-call_rcu() testing.
5221 rcutorture.onoff_holdoff= [KNL]
5222 Set time (s) after boot for CPU-hotplug testing.
5224 rcutorture.onoff_interval= [KNL]
5225 Set time (jiffies) between CPU-hotplug operations,
5226 or zero to disable CPU-hotplug testing.
5228 rcutorture.read_exit= [KNL]
5229 Set the number of read-then-exit kthreads used
5230 to test the interaction of RCU updaters and
5231 task-exit processing.
5233 rcutorture.read_exit_burst= [KNL]
5234 The number of times in a given read-then-exit
5235 episode that a set of read-then-exit kthreads
5238 rcutorture.read_exit_delay= [KNL]
5239 The delay, in seconds, between successive
5240 read-then-exit testing episodes.
5242 rcutorture.shuffle_interval= [KNL]
5243 Set task-shuffle interval (s). Shuffling tasks
5244 allows some CPUs to go into dyntick-idle mode
5245 during the rcutorture test.
5247 rcutorture.shutdown_secs= [KNL]
5248 Set time (s) after boot system shutdown. This
5249 is useful for hands-off automated testing.
5251 rcutorture.stall_cpu= [KNL]
5252 Duration of CPU stall (s) to test RCU CPU stall
5253 warnings, zero to disable.
5255 rcutorture.stall_cpu_block= [KNL]
5256 Sleep while stalling if set. This will result
5257 in warnings from preemptible RCU in addition to
5258 any other stall-related activity. Note that
5259 in kernels built with CONFIG_PREEMPTION=n and
5260 CONFIG_PREEMPT_COUNT=y, this parameter will
5261 cause the CPU to pass through a quiescent state.
5262 Given CONFIG_PREEMPTION=n, this will suppress
5263 RCU CPU stall warnings, but will instead result
5264 in scheduling-while-atomic splats.
5266 Use of this module parameter results in splats.
5269 rcutorture.stall_cpu_holdoff= [KNL]
5270 Time to wait (s) after boot before inducing stall.
5272 rcutorture.stall_cpu_irqsoff= [KNL]
5273 Disable interrupts while stalling if set.
5275 rcutorture.stall_gp_kthread= [KNL]
5276 Duration (s) of forced sleep within RCU
5277 grace-period kthread to test RCU CPU stall
5278 warnings, zero to disable. If both stall_cpu
5279 and stall_gp_kthread are specified, the
5280 kthread is starved first, then the CPU.
5282 rcutorture.stat_interval= [KNL]
5283 Time (s) between statistics printk()s.
5285 rcutorture.stutter= [KNL]
5286 Time (s) to stutter testing, for example, specifying
5287 five seconds causes the test to run for five seconds,
5288 wait for five seconds, and so on. This tests RCU's
5289 ability to transition abruptly to and from idle.
5291 rcutorture.test_boost= [KNL]
5292 Test RCU priority boosting? 0=no, 1=maybe, 2=yes.
5293 "Maybe" means test if the RCU implementation
5294 under test support RCU priority boosting.
5296 rcutorture.test_boost_duration= [KNL]
5297 Duration (s) of each individual boost test.
5299 rcutorture.test_boost_interval= [KNL]
5300 Interval (s) between each boost test.
5302 rcutorture.test_no_idle_hz= [KNL]
5303 Test RCU's dyntick-idle handling. See also the
5304 rcutorture.shuffle_interval parameter.
5306 rcutorture.torture_type= [KNL]
5307 Specify the RCU implementation to test.
5309 rcutorture.verbose= [KNL]
5310 Enable additional printk() statements.
5312 rcupdate.rcu_cpu_stall_ftrace_dump= [KNL]
5313 Dump ftrace buffer after reporting RCU CPU
5316 rcupdate.rcu_cpu_stall_notifiers= [KNL]
5317 Provide RCU CPU stall notifiers, but see the
5318 warnings in the RCU_CPU_STALL_NOTIFIER Kconfig
5319 option's help text. TL;DR: You almost certainly
5320 do not want rcupdate.rcu_cpu_stall_notifiers.
5322 rcupdate.rcu_cpu_stall_suppress= [KNL]
5323 Suppress RCU CPU stall warning messages.
5325 rcupdate.rcu_cpu_stall_suppress_at_boot= [KNL]
5326 Suppress RCU CPU stall warning messages and
5327 rcutorture writer stall warnings that occur
5328 during early boot, that is, during the time
5329 before the init task is spawned.
5331 rcupdate.rcu_cpu_stall_timeout= [KNL]
5332 Set timeout for RCU CPU stall warning messages.
5333 The value is in seconds and the maximum allowed
5334 value is 300 seconds.
5336 rcupdate.rcu_exp_cpu_stall_timeout= [KNL]
5337 Set timeout for expedited RCU CPU stall warning
5338 messages. The value is in milliseconds
5339 and the maximum allowed value is 21000
5340 milliseconds. Please note that this value is
5341 adjusted to an arch timer tick resolution.
5342 Setting this to zero causes the value from
5343 rcupdate.rcu_cpu_stall_timeout to be used (after
5344 conversion from seconds to milliseconds).
5346 rcupdate.rcu_cpu_stall_cputime= [KNL]
5347 Provide statistics on the cputime and count of
5348 interrupts and tasks during the sampling period. For
5349 multiple continuous RCU stalls, all sampling periods
5350 begin at half of the first RCU stall timeout.
5352 rcupdate.rcu_exp_stall_task_details= [KNL]
5353 Print stack dumps of any tasks blocking the
5354 current expedited RCU grace period during an
5355 expedited RCU CPU stall warning.
5357 rcupdate.rcu_expedited= [KNL]
5358 Use expedited grace-period primitives, for
5359 example, synchronize_rcu_expedited() instead
5360 of synchronize_rcu(). This reduces latency,
5361 but can increase CPU utilization, degrade
5362 real-time latency, and degrade energy efficiency.
5363 No effect on CONFIG_TINY_RCU kernels.
5365 rcupdate.rcu_normal= [KNL]
5366 Use only normal grace-period primitives,
5367 for example, synchronize_rcu() instead of
5368 synchronize_rcu_expedited(). This improves
5369 real-time latency, CPU utilization, and
5370 energy efficiency, but can expose users to
5371 increased grace-period latency. This parameter
5372 overrides rcupdate.rcu_expedited. No effect on
5373 CONFIG_TINY_RCU kernels.
5375 rcupdate.rcu_normal_after_boot= [KNL]
5376 Once boot has completed (that is, after
5377 rcu_end_inkernel_boot() has been invoked), use
5378 only normal grace-period primitives. No effect
5379 on CONFIG_TINY_RCU kernels.
5381 But note that CONFIG_PREEMPT_RT=y kernels enables
5382 this kernel boot parameter, forcibly setting
5383 it to the value one, that is, converting any
5384 post-boot attempt at an expedited RCU grace
5385 period to instead use normal non-expedited
5386 grace-period processing.
5388 rcupdate.rcu_task_collapse_lim= [KNL]
5389 Set the maximum number of callbacks present
5390 at the beginning of a grace period that allows
5391 the RCU Tasks flavors to collapse back to using
5392 a single callback queue. This switching only
5393 occurs when rcupdate.rcu_task_enqueue_lim is
5394 set to the default value of -1.
5396 rcupdate.rcu_task_contend_lim= [KNL]
5397 Set the minimum number of callback-queuing-time
5398 lock-contention events per jiffy required to
5399 cause the RCU Tasks flavors to switch to per-CPU
5400 callback queuing. This switching only occurs
5401 when rcupdate.rcu_task_enqueue_lim is set to
5402 the default value of -1.
5404 rcupdate.rcu_task_enqueue_lim= [KNL]
5405 Set the number of callback queues to use for the
5406 RCU Tasks family of RCU flavors. The default
5407 of -1 allows this to be automatically (and
5408 dynamically) adjusted. This parameter is intended
5411 rcupdate.rcu_task_ipi_delay= [KNL]
5412 Set time in jiffies during which RCU tasks will
5413 avoid sending IPIs, starting with the beginning
5414 of a given grace period. Setting a large
5415 number avoids disturbing real-time workloads,
5416 but lengthens grace periods.
5418 rcupdate.rcu_task_lazy_lim= [KNL]
5419 Number of callbacks on a given CPU that will
5420 cancel laziness on that CPU. Use -1 to disable
5421 cancellation of laziness, but be advised that
5422 doing so increases the danger of OOM due to
5425 rcupdate.rcu_task_stall_info= [KNL]
5426 Set initial timeout in jiffies for RCU task stall
5427 informational messages, which give some indication
5428 of the problem for those not patient enough to
5429 wait for ten minutes. Informational messages are
5430 only printed prior to the stall-warning message
5431 for a given grace period. Disable with a value
5432 less than or equal to zero. Defaults to ten
5433 seconds. A change in value does not take effect
5434 until the beginning of the next grace period.
5436 rcupdate.rcu_task_stall_info_mult= [KNL]
5437 Multiplier for time interval between successive
5438 RCU task stall informational messages for a given
5439 RCU tasks grace period. This value is clamped
5440 to one through ten, inclusive. It defaults to
5441 the value three, so that the first informational
5442 message is printed 10 seconds into the grace
5443 period, the second at 40 seconds, the third at
5444 160 seconds, and then the stall warning at 600
5445 seconds would prevent a fourth at 640 seconds.
5447 rcupdate.rcu_task_stall_timeout= [KNL]
5448 Set timeout in jiffies for RCU task stall
5449 warning messages. Disable with a value less
5450 than or equal to zero. Defaults to ten minutes.
5451 A change in value does not take effect until
5452 the beginning of the next grace period.
5454 rcupdate.rcu_tasks_lazy_ms= [KNL]
5455 Set timeout in milliseconds RCU Tasks asynchronous
5456 callback batching for call_rcu_tasks().
5457 A negative value will take the default. A value
5458 of zero will disable batching. Batching is
5459 always disabled for synchronize_rcu_tasks().
5461 rcupdate.rcu_tasks_rude_lazy_ms= [KNL]
5462 Set timeout in milliseconds RCU Tasks
5463 Rude asynchronous callback batching for
5464 call_rcu_tasks_rude(). A negative value
5465 will take the default. A value of zero will
5466 disable batching. Batching is always disabled
5467 for synchronize_rcu_tasks_rude().
5469 rcupdate.rcu_tasks_trace_lazy_ms= [KNL]
5470 Set timeout in milliseconds RCU Tasks
5471 Trace asynchronous callback batching for
5472 call_rcu_tasks_trace(). A negative value
5473 will take the default. A value of zero will
5474 disable batching. Batching is always disabled
5475 for synchronize_rcu_tasks_trace().
5477 rcupdate.rcu_self_test= [KNL]
5478 Run the RCU early boot self tests
5482 Run specified binary instead of /init from the ramdisk,
5483 used for early userspace startup. See initrd.
5486 force - Override the decision by the kernel to hide the
5487 advertisement of RDRAND support (this affects
5488 certain AMD processors because of buggy BIOS
5489 support, specifically around the suspend/resume
5493 Turn on/off individual RDT features. List is:
5494 cmt, mbmtotal, mbmlocal, l3cat, l3cdp, l2cat, l2cdp,
5496 E.g. to turn on cmt and turn off mba use:
5500 Format (x86 or x86_64):
5501 [w[arm] | c[old] | h[ard] | s[oft] | g[pio]] | d[efault] \
5503 [[,]b[ios] | a[cpi] | k[bd] | t[riple] | e[fi] | p[ci]] \
5505 Where reboot_mode is one of warm (soft) or cold (hard) or gpio
5506 (prefix with 'panic_' to set mode for panic
5508 reboot_type is one of bios, acpi, kbd, triple, efi, or pci,
5509 reboot_force is either force or not specified,
5510 reboot_cpu is s[mp]#### with #### being the processor
5511 to be used for rebooting.
5513 refscale.holdoff= [KNL]
5514 Set test-start holdoff period. The purpose of
5515 this parameter is to delay the start of the
5516 test until boot completes in order to avoid
5519 refscale.lookup_instances= [KNL]
5520 Number of data elements to use for the forms of
5521 SLAB_TYPESAFE_BY_RCU testing. A negative number
5522 is negated and multiplied by nr_cpu_ids, while
5523 zero specifies nr_cpu_ids.
5525 refscale.loops= [KNL]
5526 Set the number of loops over the synchronization
5527 primitive under test. Increasing this number
5528 reduces noise due to loop start/end overhead,
5529 but the default has already reduced the per-pass
5530 noise to a handful of picoseconds on ca. 2020
5533 refscale.nreaders= [KNL]
5534 Set number of readers. The default value of -1
5535 selects N, where N is roughly 75% of the number
5536 of CPUs. A value of zero is an interesting choice.
5538 refscale.nruns= [KNL]
5539 Set number of runs, each of which is dumped onto
5542 refscale.readdelay= [KNL]
5543 Set the read-side critical-section duration,
5544 measured in microseconds.
5546 refscale.scale_type= [KNL]
5547 Specify the read-protection implementation to test.
5549 refscale.shutdown= [KNL]
5550 Shut down the system at the end of the performance
5551 test. This defaults to 1 (shut it down) when
5552 refscale is built into the kernel and to 0 (leave
5553 it running) when refscale is built as a module.
5555 refscale.verbose= [KNL]
5556 Enable additional printk() statements.
5558 refscale.verbose_batched= [KNL]
5559 Batch the additional printk() statements. If zero
5560 (the default) or negative, print everything. Otherwise,
5561 print every Nth verbose statement, where N is the value
5564 regulator_ignore_unused
5566 Prevents regulator framework from disabling regulators
5567 that are unused, due no driver claiming them. This may
5568 be useful for debug and development, but should not be
5569 needed on a platform with proper driver support.
5572 [KNL, SMP] Set scheduler's default relax_domain_level.
5573 See Documentation/admin-guide/cgroup-v1/cpusets.rst.
5575 reserve= [KNL,BUGS] Force kernel to ignore I/O ports or memory
5576 Format: <base1>,<size1>[,<base2>,<size2>,...]
5577 Reserve I/O ports or memory so the kernel won't use
5578 them. If <base> is less than 0x10000, the region
5579 is assumed to be I/O ports; otherwise it is memory.
5581 reservetop= [X86-32]
5583 Reserves a hole at the top of the kernel virtual
5586 reset_devices [KNL] Force drivers to reset the underlying device
5587 during initialization.
5590 Specify the partition device for software suspend
5592 {/dev/<dev> | PARTUUID=<uuid> | <int>:<int> | <hex>}
5594 resume_offset= [SWSUSP]
5595 Specify the offset from the beginning of the partition
5596 given by "resume=" at which the swap header is located,
5597 in <PAGE_SIZE> units (needed only for swap files).
5598 See Documentation/power/swsusp-and-swap-files.rst
5600 resumedelay= [HIBERNATION] Delay (in seconds) to pause before attempting to
5601 read the resume files
5603 resumewait [HIBERNATION] Wait (indefinitely) for resume device to show up.
5604 Useful for devices that are detected asynchronously
5605 (e.g. USB and MMC devices).
5607 retain_initrd [RAM] Keep initrd memory after extraction. After boot, it will
5608 be accessible via /sys/firmware/initrd.
5610 retbleed= [X86] Control mitigation of RETBleed (Arbitrary
5611 Speculative Code Execution with Return Instructions)
5614 AMD-based UNRET and IBPB mitigations alone do not stop
5615 sibling threads from influencing the predictions of other
5616 sibling threads. For that reason, STIBP is used on pro-
5617 cessors that support it, and mitigate SMT on processors
5621 auto - automatically select a migitation
5622 auto,nosmt - automatically select a mitigation,
5623 disabling SMT if necessary for
5624 the full mitigation (only on Zen1
5625 and older without STIBP).
5626 ibpb - On AMD, mitigate short speculation
5627 windows on basic block boundaries too.
5628 Safe, highest perf impact. It also
5629 enables STIBP if present. Not suitable
5631 ibpb,nosmt - Like "ibpb" above but will disable SMT
5632 when STIBP is not available. This is
5633 the alternative for systems which do not
5635 unret - Force enable untrained return thunks,
5636 only effective on AMD f15h-f17h based
5638 unret,nosmt - Like unret, but will disable SMT when STIBP
5639 is not available. This is the alternative for
5640 systems which do not have STIBP.
5642 Selecting 'auto' will choose a mitigation method at run
5643 time according to the CPU.
5645 Not specifying this option is equivalent to retbleed=auto.
5647 rfkill.default_state=
5648 0 "airplane mode". All wifi, bluetooth, wimax, gps, fm,
5649 etc. communication is blocked by default.
5652 rfkill.master_switch_mode=
5653 0 The "airplane mode" button does nothing.
5654 1 The "airplane mode" button toggles between everything
5655 blocked and the previous configuration.
5656 2 The "airplane mode" button toggles between everything
5657 blocked and everything unblocked.
5659 rhash_entries= [KNL,NET]
5660 Set number of hash buckets for route cache
5663 [KNL] Disable ring 3 MONITOR/MWAIT feature on supported
5666 riscv_isa_fallback [RISCV]
5667 When CONFIG_RISCV_ISA_FALLBACK is not enabled, permit
5668 falling back to detecting extension support by parsing
5669 "riscv,isa" property on devicetree systems when the
5670 replacement properties are not found. See the Kconfig
5671 entry for RISCV_ISA_FALLBACK.
5673 ro [KNL] Mount root device read-only on boot
5676 on Mark read-only kernel memory as read-only (default).
5677 off Leave read-only kernel memory writable for debugging.
5678 full Mark read-only kernel memory and aliases as read-only
5682 Enable the uart passthrough on the designated usb port
5683 on Rockchip SoCs. When active, the signals of the
5684 debug-uart get routed to the D+ and D- pins of the usb
5685 port and the regular usb controller gets disabled.
5687 root= [KNL] Root filesystem
5688 Usually this a a block device specifier of some kind,
5689 see the early_lookup_bdev comment in
5690 block/early-lookup.c for details.
5691 Alternatively this can be "ram" for the legacy initial
5692 ramdisk, "nfs" and "cifs" for root on a network file
5693 system, or "mtd" and "ubi" for mounting from raw flash.
5695 rootdelay= [KNL] Delay (in seconds) to pause before attempting to
5696 mount the root filesystem
5698 rootflags= [KNL] Set root filesystem mount option string
5700 rootfstype= [KNL] Set root filesystem type
5702 rootwait [KNL] Wait (indefinitely) for root device to show up.
5703 Useful for devices that are detected asynchronously
5704 (e.g. USB and MMC devices).
5706 rootwait= [KNL] Maximum time (in seconds) to wait for root device
5707 to show up before attempting to mount the root
5710 rproc_mem=nn[KMG][@address]
5711 [KNL,ARM,CMA] Remoteproc physical memory block.
5712 Memory area to be used by remote processor image,
5715 rw [KNL] Mount root device read-write on boot
5717 S [KNL] Run init in single mode
5719 s390_iommu= [HW,S390]
5720 Set s390 IOTLB flushing mode
5722 With strict flushing every unmap operation will result
5723 in an IOTLB flush. Default is lazy flushing before
5724 reuse, which is faster. Deprecated, equivalent to
5727 s390_iommu_aperture= [KNL,S390]
5728 Specifies the size of the per device DMA address space
5729 accessible through the DMA and IOMMU APIs as a decimal
5730 factor of the size of main memory.
5731 The default is 1 meaning that one can concurrently use
5732 as many DMA addresses as physical memory is installed,
5733 if supported by hardware, and thus map all of memory
5734 once. With a value of 2 one can map all of memory twice
5735 and so on. As a special case a factor of 0 imposes no
5736 restrictions other than those given by hardware at the
5737 cost of significant additional memory use for tables.
5740 See drivers/net/irda/sa1100_ir.c.
5742 sched_verbose [KNL] Enables verbose scheduler debug messages.
5744 schedstats= [KNL,X86] Enable or disable scheduled statistics.
5745 Allowed values are enable and disable. This feature
5746 incurs a small amount of overhead in the scheduler
5747 but is useful for debugging and performance tuning.
5749 sched_thermal_decay_shift=
5750 [KNL, SMP] Set a decay shift for scheduler thermal
5751 pressure signal. Thermal pressure signal follows the
5752 default decay period of other scheduler pelt
5753 signals(usually 32 ms but configurable). Setting
5754 sched_thermal_decay_shift will left shift the decay
5755 period for the thermal pressure signal by the shift
5757 i.e. with the default pelt decay period of 32 ms
5758 sched_thermal_decay_shift thermal pressure decay pr
5762 Format: integer between 0 and 10
5765 scftorture.holdoff= [KNL]
5766 Number of seconds to hold off before starting
5767 test. Defaults to zero for module insertion and
5768 to 10 seconds for built-in smp_call_function()
5771 scftorture.longwait= [KNL]
5772 Request ridiculously long waits randomly selected
5773 up to the chosen limit in seconds. Zero (the
5774 default) disables this feature. Please note
5775 that requesting even small non-zero numbers of
5776 seconds can result in RCU CPU stall warnings,
5777 softlockup complaints, and so on.
5779 scftorture.nthreads= [KNL]
5780 Number of kthreads to spawn to invoke the
5781 smp_call_function() family of functions.
5782 The default of -1 specifies a number of kthreads
5783 equal to the number of CPUs.
5785 scftorture.onoff_holdoff= [KNL]
5786 Number seconds to wait after the start of the
5787 test before initiating CPU-hotplug operations.
5789 scftorture.onoff_interval= [KNL]
5790 Number seconds to wait between successive
5791 CPU-hotplug operations. Specifying zero (which
5792 is the default) disables CPU-hotplug operations.
5794 scftorture.shutdown_secs= [KNL]
5795 The number of seconds following the start of the
5796 test after which to shut down the system. The
5797 default of zero avoids shutting down the system.
5798 Non-zero values are useful for automated tests.
5800 scftorture.stat_interval= [KNL]
5801 The number of seconds between outputting the
5802 current test statistics to the console. A value
5803 of zero disables statistics output.
5805 scftorture.stutter_cpus= [KNL]
5806 The number of jiffies to wait between each change
5807 to the set of CPUs under test.
5809 scftorture.use_cpus_read_lock= [KNL]
5810 Use use_cpus_read_lock() instead of the default
5811 preempt_disable() to disable CPU hotplug
5812 while invoking one of the smp_call_function*()
5815 scftorture.verbose= [KNL]
5816 Enable additional printk() statements.
5818 scftorture.weight_single= [KNL]
5819 The probability weighting to use for the
5820 smp_call_function_single() function with a zero
5821 "wait" parameter. A value of -1 selects the
5822 default if all other weights are -1. However,
5823 if at least one weight has some other value, a
5824 value of -1 will instead select a weight of zero.
5826 scftorture.weight_single_wait= [KNL]
5827 The probability weighting to use for the
5828 smp_call_function_single() function with a
5829 non-zero "wait" parameter. See weight_single.
5831 scftorture.weight_many= [KNL]
5832 The probability weighting to use for the
5833 smp_call_function_many() function with a zero
5834 "wait" parameter. See weight_single.
5835 Note well that setting a high probability for
5836 this weighting can place serious IPI load
5839 scftorture.weight_many_wait= [KNL]
5840 The probability weighting to use for the
5841 smp_call_function_many() function with a
5842 non-zero "wait" parameter. See weight_single
5845 scftorture.weight_all= [KNL]
5846 The probability weighting to use for the
5847 smp_call_function_all() function with a zero
5848 "wait" parameter. See weight_single and
5851 scftorture.weight_all_wait= [KNL]
5852 The probability weighting to use for the
5853 smp_call_function_all() function with a
5854 non-zero "wait" parameter. See weight_single
5857 skew_tick= [KNL] Offset the periodic timer tick per cpu to mitigate
5858 xtime_lock contention on larger systems, and/or RCU lock
5859 contention on all systems with CONFIG_MAXSMP set.
5860 Format: { "0" | "1" }
5861 0 -- disable. (may be 1 via CONFIG_CMDLINE="skew_tick=1"
5863 Note: increases power consumption, thus should only be
5864 enabled if running jitter sensitive (HPC/RT) workloads.
5866 security= [SECURITY] Choose a legacy "major" security module to
5867 enable at boot. This has been deprecated by the
5870 selinux= [SELINUX] Disable or enable SELinux at boot time.
5871 Format: { "0" | "1" }
5872 See security/selinux/Kconfig help text.
5877 serialnumber [BUGS=X86-32]
5879 sev=option[,option...] [X86-64] See Documentation/arch/x86/x86_64/boot-options.rst
5882 Maximal number of shapers.
5884 show_lapic= [APIC,X86] Advanced Programmable Interrupt Controller
5885 Limit apic dumping. The parameter defines the maximal
5886 number of local apics being dumped. Also it is possible
5887 to set it to "all" by meaning -- no limit here.
5888 Format: { 1 (default) | 2 | ... | all }.
5889 The parameter valid if only apic=debug or
5890 apic=verbose is specified.
5891 Example: apic=debug show_lapic=all
5899 Enable merging of slabs with similar size when the
5900 kernel is built without CONFIG_SLAB_MERGE_DEFAULT.
5903 Disable merging of slabs with similar size. May be
5904 necessary if there is some reason to distinguish
5905 allocs to different slabs, especially in hardened
5906 environments where the risk of heap overflows and
5907 layout control by attackers can usually be
5908 frustrated by disabling merging. This will reduce
5909 most of the exposure of a heap attack to a single
5910 cache (risks via metadata attacks are mostly
5911 unchanged). Debug options disable merging on their
5913 For more information see Documentation/mm/slub.rst.
5915 slab_max_order= [MM, SLAB]
5916 Determines the maximum allowed order for slabs.
5917 A high setting may cause OOMs due to memory
5918 fragmentation. Defaults to 1 for systems with
5919 more than 32MB of RAM, 0 otherwise.
5921 slub_debug[=options[,slabs][;[options[,slabs]]...] [MM, SLUB]
5922 Enabling slub_debug allows one to determine the
5923 culprit if slab objects become corrupted. Enabling
5924 slub_debug can create guard zones around objects and
5925 may poison objects when not in use. Also tracks the
5926 last alloc / free. For more information see
5927 Documentation/mm/slub.rst.
5929 slub_max_order= [MM, SLUB]
5930 Determines the maximum allowed order for slabs.
5931 A high setting may cause OOMs due to memory
5932 fragmentation. For more information see
5933 Documentation/mm/slub.rst.
5935 slub_min_objects= [MM, SLUB]
5936 The minimum number of objects per slab. SLUB will
5937 increase the slab order up to slub_max_order to
5938 generate a sufficiently large slab able to contain
5939 the number of objects indicated. The higher the number
5940 of objects the smaller the overhead of tracking slabs
5941 and the less frequently locks need to be acquired.
5942 For more information see Documentation/mm/slub.rst.
5944 slub_min_order= [MM, SLUB]
5945 Determines the minimum page order for slabs. Must be
5946 lower than slub_max_order.
5947 For more information see Documentation/mm/slub.rst.
5949 slub_merge [MM, SLUB]
5950 Same with slab_merge.
5952 slub_nomerge [MM, SLUB]
5953 Same with slab_nomerge. This is supported for legacy.
5954 See slab_nomerge for more information.
5957 Format: <io1>[,<io2>[,...,<io8>]]
5959 smp.csd_lock_timeout= [KNL]
5960 Specify the period of time in milliseconds
5961 that smp_call_function() and friends will wait
5962 for a CPU to release the CSD lock. This is
5963 useful when diagnosing bugs involving CPUs
5964 disabling interrupts for extended periods
5965 of time. Defaults to 5,000 milliseconds, and
5966 setting a value of zero disables this feature.
5967 This feature may be more efficiently disabled
5968 using the csdlock_debug- kernel parameter.
5970 smp.panic_on_ipistall= [KNL]
5971 If a csd_lock_timeout extends for more than
5972 the specified number of milliseconds, panic the
5973 system. By default, let CSD-lock acquisition
5974 take as long as they take. Specifying 300,000
5975 for this value provides a 5-minute timeout.
5977 smsc-ircc2.nopnp [HW] Don't use PNP to discover SMC devices
5978 smsc-ircc2.ircc_cfg= [HW] Device configuration I/O port
5979 smsc-ircc2.ircc_sir= [HW] SIR base I/O port
5980 smsc-ircc2.ircc_fir= [HW] FIR base I/O port
5981 smsc-ircc2.ircc_irq= [HW] IRQ line
5982 smsc-ircc2.ircc_dma= [HW] DMA channel
5983 smsc-ircc2.ircc_transceiver= [HW] Transceiver type:
5984 0: Toshiba Satellite 1800 (GP data pin select)
5985 1: Fast pin select (default)
5988 smt= [KNL,MIPS,S390] Set the maximum number of threads (logical
5989 CPUs) to use per physical CPU on systems capable of
5990 symmetric multithreading (SMT). Will be capped to the
5991 actual hardware limit.
5993 Default: -1 (no limit)
5996 [KNL] Should the soft-lockup detector generate panics.
5999 A value of 1 instructs the soft-lockup detector
6000 to panic the machine when a soft-lockup occurs. It is
6001 also controlled by the kernel.softlockup_panic sysctl
6002 and CONFIG_BOOTPARAM_SOFTLOCKUP_PANIC, which is the
6003 respective build-time switch to that functionality.
6005 softlockup_all_cpu_backtrace=
6006 [KNL] Should the soft-lockup detector generate
6007 backtraces on all cpus.
6010 sonypi.*= [HW] Sony Programmable I/O Control Device driver
6011 See Documentation/admin-guide/laptops/sonypi.rst
6013 spectre_v2= [X86] Control mitigation of Spectre variant 2
6014 (indirect branch speculation) vulnerability.
6015 The default operation protects the kernel from
6018 on - unconditionally enable, implies
6020 off - unconditionally disable, implies
6022 auto - kernel detects whether your CPU model is
6025 Selecting 'on' will, and 'auto' may, choose a
6026 mitigation method at run time according to the
6027 CPU, the available microcode, the setting of the
6028 CONFIG_RETPOLINE configuration option, and the
6029 compiler with which the kernel was built.
6031 Selecting 'on' will also enable the mitigation
6032 against user space to user space task attacks.
6034 Selecting 'off' will disable both the kernel and
6035 the user space protections.
6037 Specific mitigations can also be selected manually:
6039 retpoline - replace indirect branches
6040 retpoline,generic - Retpolines
6041 retpoline,lfence - LFENCE; indirect branch
6042 retpoline,amd - alias for retpoline,lfence
6043 eibrs - Enhanced/Auto IBRS
6044 eibrs,retpoline - Enhanced/Auto IBRS + Retpolines
6045 eibrs,lfence - Enhanced/Auto IBRS + LFENCE
6046 ibrs - use IBRS to protect kernel
6048 Not specifying this option is equivalent to
6052 [X86] Control mitigation of Spectre variant 2
6053 (indirect branch speculation) vulnerability between
6056 on - Unconditionally enable mitigations. Is
6057 enforced by spectre_v2=on
6059 off - Unconditionally disable mitigations. Is
6060 enforced by spectre_v2=off
6062 prctl - Indirect branch speculation is enabled,
6063 but mitigation can be enabled via prctl
6064 per thread. The mitigation control state
6065 is inherited on fork.
6068 - Like "prctl" above, but only STIBP is
6069 controlled per thread. IBPB is issued
6070 always when switching between different user
6074 - Same as "prctl" above, but all seccomp
6075 threads will enable the mitigation unless
6076 they explicitly opt out.
6079 - Like "seccomp" above, but only STIBP is
6080 controlled per thread. IBPB is issued
6081 always when switching between different
6082 user space processes.
6084 auto - Kernel selects the mitigation depending on
6085 the available CPU features and vulnerability.
6087 Default mitigation: "prctl"
6089 Not specifying this option is equivalent to
6090 spectre_v2_user=auto.
6092 spec_rstack_overflow=
6093 [X86] Control RAS overflow mitigation on AMD Zen CPUs
6095 off - Disable mitigation
6096 microcode - Enable microcode mitigation only
6097 safe-ret - Enable sw-only safe RET mitigation (default)
6098 ibpb - Enable mitigation by issuing IBPB on
6100 ibpb-vmexit - Issue IBPB only on VMEXIT
6101 (cloud-specific mitigation)
6103 spec_store_bypass_disable=
6104 [HW] Control Speculative Store Bypass (SSB) Disable mitigation
6105 (Speculative Store Bypass vulnerability)
6107 Certain CPUs are vulnerable to an exploit against a
6108 a common industry wide performance optimization known
6109 as "Speculative Store Bypass" in which recent stores
6110 to the same memory location may not be observed by
6111 later loads during speculative execution. The idea
6112 is that such stores are unlikely and that they can
6113 be detected prior to instruction retirement at the
6114 end of a particular speculation execution window.
6116 In vulnerable processors, the speculatively forwarded
6117 store can be used in a cache side channel attack, for
6118 example to read memory to which the attacker does not
6119 directly have access (e.g. inside sandboxed code).
6121 This parameter controls whether the Speculative Store
6122 Bypass optimization is used.
6124 On x86 the options are:
6126 on - Unconditionally disable Speculative Store Bypass
6127 off - Unconditionally enable Speculative Store Bypass
6128 auto - Kernel detects whether the CPU model contains an
6129 implementation of Speculative Store Bypass and
6130 picks the most appropriate mitigation. If the
6131 CPU is not vulnerable, "off" is selected. If the
6132 CPU is vulnerable the default mitigation is
6133 architecture and Kconfig dependent. See below.
6134 prctl - Control Speculative Store Bypass per thread
6135 via prctl. Speculative Store Bypass is enabled
6136 for a process by default. The state of the control
6137 is inherited on fork.
6138 seccomp - Same as "prctl" above, but all seccomp threads
6139 will disable SSB unless they explicitly opt out.
6141 Default mitigations:
6144 On powerpc the options are:
6146 on,auto - On Power8 and Power9 insert a store-forwarding
6147 barrier on kernel entry and exit. On Power7
6148 perform a software flush on kernel entry and
6152 Not specifying this option is equivalent to
6153 spec_store_bypass_disable=auto.
6155 spia_io_base= [HW,MTD]
6161 [X86] Enable split lock detection or bus lock detection
6163 When enabled (and if hardware support is present), atomic
6164 instructions that access data across cache line
6165 boundaries will result in an alignment check exception
6166 for split lock detection or a debug exception for
6171 warn - the kernel will emit rate-limited warnings
6172 about applications triggering the #AC
6173 exception or the #DB exception. This mode is
6174 the default on CPUs that support split lock
6175 detection or bus lock detection. Default
6176 behavior is by #AC if both features are
6177 enabled in hardware.
6179 fatal - the kernel will send SIGBUS to applications
6180 that trigger the #AC exception or the #DB
6181 exception. Default behavior is by #AC if
6182 both features are enabled in hardware.
6185 Set system wide rate limit to N bus locks
6186 per second for bus lock detection.
6189 N/A for split lock detection.
6192 If an #AC exception is hit in the kernel or in
6193 firmware (i.e. not while executing in user mode)
6194 the kernel will oops in either "warn" or "fatal"
6197 #DB exception for bus lock is triggered only when
6201 Control the Special Register Buffer Data Sampling
6204 Certain CPUs are vulnerable to an MDS-like
6205 exploit which can leak bits from the random
6208 By default, this issue is mitigated by
6209 microcode. However, the microcode fix can cause
6210 the RDRAND and RDSEED instructions to become
6211 much slower. Among other effects, this will
6212 result in reduced throughput from /dev/urandom.
6214 The microcode mitigation can be disabled with
6215 the following option:
6217 off: Disable mitigation and remove
6218 performance impact to RDRAND and RDSEED
6220 srcutree.big_cpu_lim [KNL]
6221 Specifies the number of CPUs constituting a
6222 large system, such that srcu_struct structures
6223 should immediately allocate an srcu_node array.
6224 This kernel-boot parameter defaults to 128,
6225 but takes effect only when the low-order four
6226 bits of srcutree.convert_to_big is equal to 3
6229 srcutree.convert_to_big [KNL]
6230 Specifies under what conditions an SRCU tree
6231 srcu_struct structure will be converted to big
6232 form, that is, with an rcu_node tree:
6235 1: At init_srcu_struct() time.
6236 2: When rcutorture decides to.
6237 3: Decide at boot time (default).
6238 0x1X: Above plus if high contention.
6240 Either way, the srcu_node tree will be sized based
6241 on the actual runtime number of CPUs (nr_cpu_ids)
6242 instead of the compile-time CONFIG_NR_CPUS.
6244 srcutree.counter_wrap_check [KNL]
6245 Specifies how frequently to check for
6246 grace-period sequence counter wrap for the
6247 srcu_data structure's ->srcu_gp_seq_needed field.
6248 The greater the number of bits set in this kernel
6249 parameter, the less frequently counter wrap will
6250 be checked for. Note that the bottom two bits
6253 srcutree.exp_holdoff [KNL]
6254 Specifies how many nanoseconds must elapse
6255 since the end of the last SRCU grace period for
6256 a given srcu_struct until the next normal SRCU
6257 grace period will be considered for automatic
6258 expediting. Set to zero to disable automatic
6261 srcutree.srcu_max_nodelay [KNL]
6262 Specifies the number of no-delay instances
6263 per jiffy for which the SRCU grace period
6264 worker thread will be rescheduled with zero
6265 delay. Beyond this limit, worker thread will
6266 be rescheduled with a sleep delay of one jiffy.
6268 srcutree.srcu_max_nodelay_phase [KNL]
6269 Specifies the per-grace-period phase, number of
6270 non-sleeping polls of readers. Beyond this limit,
6271 grace period worker thread will be rescheduled
6272 with a sleep delay of one jiffy, between each
6273 rescan of the readers, for a grace period phase.
6275 srcutree.srcu_retry_check_delay [KNL]
6276 Specifies number of microseconds of non-sleeping
6277 delay between each non-sleeping poll of readers.
6279 srcutree.small_contention_lim [KNL]
6280 Specifies the number of update-side contention
6281 events per jiffy will be tolerated before
6282 initiating a conversion of an srcu_struct
6283 structure to big form. Note that the value of
6284 srcutree.convert_to_big must have the 0x10 bit
6285 set for contention-based conversions to occur.
6288 Speculative Store Bypass Disable control
6290 On CPUs that are vulnerable to the Speculative
6291 Store Bypass vulnerability and offer a
6292 firmware based mitigation, this parameter
6293 indicates how the mitigation should be used:
6295 force-on: Unconditionally enable mitigation for
6296 for both kernel and userspace
6297 force-off: Unconditionally disable mitigation for
6298 for both kernel and userspace
6299 kernel: Always enable mitigation in the
6300 kernel, and offer a prctl interface
6301 to allow userspace to register its
6302 interest in being mitigated too.
6304 stack_guard_gap= [MM]
6305 override the default stack gap protection. The value
6306 is in page units and it defines how many pages prior
6307 to (for stacks growing down) resp. after (for stacks
6308 growing up) the main stack are reserved for no other
6309 mapping. Default value is 256 pages.
6311 stack_depot_disable= [KNL]
6312 Setting this to true through kernel command line will
6313 disable the stack depot thereby saving the static memory
6314 consumed by the stack hash table. By default this is set
6318 Enabled the stack tracer on boot up.
6320 stacktrace_filter=[function-list]
6321 [FTRACE] Limit the functions that the stack tracer
6322 will trace at boot up. function-list is a comma-separated
6323 list of functions. This list can be changed at run
6324 time by the stack_trace_filter file in the debugfs
6325 tracing directory. Note, this enables stack tracing
6326 and the stacktrace above is not needed.
6330 Set the STI (builtin display/keyboard on the HP-PARISC
6331 machines) console (graphic card) which should be used
6332 as the initial boot-console.
6333 See also comment in drivers/video/console/sticore.c.
6336 See comment in drivers/video/console/sticore.c.
6339 Format: bpp:<bpp1>[:<bpp2>[:<bpp3>...]]
6344 Enable or disable strict sigaltstack size checks
6345 against the required signal frame size which
6346 depends on the supported FPU features. This can
6347 be used to filter out binaries which have
6348 not yet been made aware of AT_MINSIGSTKSZ.
6351 Limits the number of kernel HPT entries in the hash
6352 page table to increase the rate of hash page table
6353 faults on kernel addresses.
6356 Limits the number of kernel SLB entries, and flushes
6357 them frequently to increase the rate of SLB faults
6358 on kernel addresses.
6360 sunrpc.min_resvport=
6361 sunrpc.max_resvport=
6363 SunRPC servers often require that client requests
6364 originate from a privileged port (i.e. a port in the
6365 range 0 < portnr < 1024).
6366 An administrator who wishes to reserve some of these
6367 ports for other uses may adjust the range that the
6368 kernel's sunrpc client considers to be privileged
6369 using these two parameters to set the minimum and
6370 maximum port values.
6372 sunrpc.svc_rpc_per_connection_limit=
6374 Limit the number of requests that the server will
6375 process in parallel from a single connection.
6376 The default value is 0 (no limit).
6380 Control how the NFS server code allocates CPUs to
6381 service thread pools. Depending on how many NICs
6382 you have and where their interrupts are bound, this
6383 option will affect which CPUs will do NFS serving.
6384 Note: this parameter cannot be changed while the
6385 NFS server is running.
6387 auto the server chooses an appropriate mode
6388 automatically using heuristics
6389 global a single global pool contains all CPUs
6390 percpu one pool for each CPU
6391 pernode one pool for each NUMA node (equivalent
6392 to global on non-NUMA machines)
6394 sunrpc.tcp_slot_table_entries=
6395 sunrpc.udp_slot_table_entries=
6397 Sets the upper limit on the number of simultaneous
6398 RPC calls that can be sent from the client to a
6399 server. Increasing these values may allow you to
6400 improve throughput, but will also increase the
6401 amount of memory reserved for use by the client.
6403 suspend.pm_test_delay=
6405 Sets the number of seconds to remain in a suspend test
6406 mode before resuming the system (see
6407 /sys/power/pm_test). Only available when CONFIG_PM_DEBUG
6408 is set. Default value is 5.
6411 Format: { on | off | y | n | 1 | 0 }
6412 This parameter controls use of the Protected
6413 Execution Facility on pSeries.
6415 swiotlb= [ARM,IA-64,PPC,MIPS,X86]
6416 Format: { <int> [,<int>] | force | noforce }
6417 <int> -- Number of I/O TLB slabs
6418 <int> -- Second integer after comma. Number of swiotlb
6419 areas with their own lock. Will be rounded up
6421 force -- force using of bounce buffers even if they
6422 wouldn't be automatically used by the kernel
6423 noforce -- Never use bounce buffers (for debugging)
6428 Set a sysctl parameter, right before loading the init
6429 process, as if the value was written to the respective
6430 /proc/sys/... file. Both '.' and '/' are recognized as
6431 separators. Unrecognized parameters and invalid values
6432 are reported in the kernel log. Sysctls registered
6433 later by a loaded module cannot be set this way.
6434 Example: sysctl.vm.swappiness=40
6436 sysrq_always_enabled
6438 Ignore sysrq setting - this boot parameter will
6439 neutralize any effect of /proc/sys/kernel/sysrq.
6440 Useful for debugging.
6442 tcpmhash_entries= [KNL,NET]
6443 Set the number of tcp_metrics_hash slots.
6444 Default value is 8192 or 16384 depending on total
6445 ram pages. This is used to specify the TCP metrics
6446 cache size. See Documentation/networking/ip-sysctl.rst
6447 "tcp_no_metrics_save" section for more details.
6451 test_suspend= [SUSPEND]
6452 Format: { "mem" | "standby" | "freeze" }[,N]
6453 Specify "mem" (for Suspend-to-RAM) or "standby" (for
6454 standby suspend) or "freeze" (for suspend type freeze)
6455 as the system sleep state during system startup with
6456 the optional capability to repeat N number of times.
6457 The system is woken from this state using a
6458 wakeup-capable RTC alarm.
6460 thash_entries= [KNL,NET]
6461 Set number of hash buckets for TCP connection
6463 thermal.act= [HW,ACPI]
6464 -1: disable all active trip points in all thermal zones
6465 <degrees C>: override all lowest active trip points
6467 thermal.crt= [HW,ACPI]
6468 -1: disable all critical trip points in all thermal zones
6469 <degrees C>: override all critical trip points
6471 thermal.off= [HW,ACPI]
6472 1: disable ACPI thermal control
6474 thermal.psv= [HW,ACPI]
6475 -1: disable all passive trip points
6476 <degrees C>: override all passive trip points to this
6479 thermal.tzp= [HW,ACPI]
6480 Specify global default ACPI thermal zone polling rate
6481 <deci-seconds>: poll all this frequency
6482 0: no polling (default)
6485 Force threading of all interrupt handlers except those
6486 marked explicitly IRQF_NO_THREAD.
6490 Specify if the kernel should make use of the cpu
6491 topology information if the hardware supports this.
6492 The scheduler will make use of this information and
6493 e.g. base its process migration decisions on it.
6496 topology_updates= [KNL, PPC, NUMA]
6498 Specify if the kernel should ignore (off)
6499 topology updates sent by the hypervisor to this
6502 torture.disable_onoff_at_boot= [KNL]
6503 Prevent the CPU-hotplug component of torturing
6504 until after init has spawned.
6506 torture.ftrace_dump_at_shutdown= [KNL]
6507 Dump the ftrace buffer at torture-test shutdown,
6508 even if there were no errors. This can be a
6509 very costly operation when many torture tests
6510 are running concurrently, especially on systems
6511 with rotating-rust storage.
6513 torture.verbose_sleep_frequency= [KNL]
6514 Specifies how many verbose printk()s should be
6515 emitted between each sleep. The default of zero
6516 disables verbose-printk() sleeping.
6518 torture.verbose_sleep_duration= [KNL]
6519 Duration of each verbose-printk() sleep in jiffies.
6523 tpm_suspend_pcr=[HW,TPM]
6524 Format: integer pcr id
6525 Specify that at suspend time, the tpm driver
6526 should extend the specified pcr with zeros,
6527 as a workaround for some chips which fail to
6528 flush the last written pcr on TPM_SaveState.
6529 This will guarantee that all the other pcrs
6532 tpm_tis.interrupts= [HW,TPM]
6533 Enable interrupts for the MMIO based physical layer
6534 for the FIFO interface. By default it is set to false
6535 (0). For more information about TPM hardware interfaces
6536 defined by Trusted Computing Group (TCG) see
6537 https://trustedcomputinggroup.org/resource/pc-client-platform-tpm-profile-ptp-specification/
6540 Have the tracepoints sent to printk as well as the
6541 tracing ring buffer. This is useful for early boot up
6542 where the system hangs or reboots and does not give the
6543 option for reading the tracing buffer or performing a
6544 ftrace_dump_on_oops.
6546 To turn off having tracepoints sent to printk,
6547 echo 0 > /proc/sys/kernel/tracepoint_printk
6548 Note, echoing 1 into this file without the
6549 tracepoint_printk kernel cmdline option has no effect.
6551 The tp_printk_stop_on_boot (see below) can also be used
6552 to stop the printing of events to console at
6557 Having tracepoints sent to printk() and activating high
6558 frequency tracepoints such as irq or sched, can cause
6559 the system to live lock.
6561 tp_printk_stop_on_boot [FTRACE]
6562 When tp_printk (above) is set, it can cause a lot of noise
6563 on the console. It may be useful to only include the
6564 printing of events during boot up, as user space may
6565 make the system inoperable.
6567 This command line option will stop the printing of events
6568 to console at the late_initcall_sync() time frame.
6570 trace_buf_size=nn[KMG]
6571 [FTRACE] will set tracing buffer size on each cpu.
6573 trace_clock= [FTRACE] Set the clock used for tracing events
6575 local - Use the per CPU time stamp counter
6576 (converted into nanoseconds). Fast, but
6577 depending on the architecture, may not be
6578 in sync between CPUs.
6579 global - Event time stamps are synchronize across
6580 CPUs. May be slower than the local clock,
6581 but better for some race conditions.
6582 counter - Simple counting of events (1, 2, ..)
6583 note, some counts may be skipped due to the
6584 infrastructure grabbing the clock more than
6586 uptime - Use jiffies as the time stamp.
6587 perf - Use the same clock that perf uses.
6588 mono - Use ktime_get_mono_fast_ns() for time stamps.
6589 mono_raw - Use ktime_get_raw_fast_ns() for time
6591 boot - Use ktime_get_boot_fast_ns() for time stamps.
6592 Architectures may add more clocks. See
6593 Documentation/trace/ftrace.rst for more details.
6595 trace_event=[event-list]
6596 [FTRACE] Set and start specified trace events in order
6597 to facilitate early boot debugging. The event-list is a
6598 comma-separated list of trace events to enable. See
6599 also Documentation/trace/events.rst
6601 trace_instance=[instance-info]
6602 [FTRACE] Create a ring buffer instance early in boot up.
6603 This will be listed in:
6605 /sys/kernel/tracing/instances
6607 Events can be enabled at the time the instance is created
6610 trace_instance=<name>,<system1>:<event1>,<system2>:<event2>
6612 Note, the "<system*>:" portion is optional if the event is
6615 trace_instance=foo,sched:sched_switch,irq_handler_entry,initcall
6617 will enable the "sched_switch" event (note, the "sched:" is optional, and
6618 the same thing would happen if it was left off). The irq_handler_entry
6619 event, and all events under the "initcall" system.
6621 trace_options=[option-list]
6622 [FTRACE] Enable or disable tracer options at boot.
6623 The option-list is a comma delimited list of options
6624 that can be enabled or disabled just as if you were
6625 to echo the option name into
6627 /sys/kernel/tracing/trace_options
6629 For example, to enable stacktrace option (to dump the
6630 stack trace of each event), add to the command line:
6632 trace_options=stacktrace
6634 See also Documentation/trace/ftrace.rst "trace options"
6637 trace_trigger=[trigger-list]
6638 [FTRACE] Add a event trigger on specific events.
6639 Set a trigger on top of a specific event, with an optional
6642 The format is is "trace_trigger=<event>.<trigger>[ if <filter>],..."
6643 Where more than one trigger may be specified that are comma deliminated.
6647 trace_trigger="sched_switch.stacktrace if prev_state == 2"
6649 The above will enable the "stacktrace" trigger on the "sched_switch"
6650 event but only trigger it if the "prev_state" of the "sched_switch"
6651 event is "2" (TASK_UNINTERUPTIBLE).
6653 See also "Event triggers" in Documentation/trace/events.rst
6657 [FTRACE] enable this option to disable tracing when a
6658 warning is hit. This turns off "tracing_on". Tracing can
6659 be enabled again by echoing '1' into the "tracing_on"
6660 file located in /sys/kernel/tracing/
6662 This option is useful, as it disables the trace before
6663 the WARNING dump is called, which prevents the trace to
6664 be filled with content caused by the warning output.
6666 This option can also be set at run time via the sysctl
6667 option: kernel/traceoff_on_warning
6669 transparent_hugepage=
6671 Format: [always|madvise|never]
6672 Can be used to control the default behavior of the system
6673 with respect to transparent hugepages.
6674 See Documentation/admin-guide/mm/transhuge.rst
6677 trusted.source= [KEYS]
6679 This parameter identifies the trust source as a backend
6680 for trusted keys implementation. Supported trust
6685 If not specified then it defaults to iterating through
6686 the trust source list starting with TPM and assigns the
6687 first trust source as a backend which is initialized
6688 successfully during iteration.
6692 The RNG used to generate key material for trusted keys.
6695 - the same value as trusted.source: "tpm" or "tee"
6697 If not specified, "default" is used. In this case,
6698 the RNG's choice is left to each individual trust source.
6700 tsc= Disable clocksource stability checks for TSC.
6702 [x86] reliable: mark tsc clocksource as reliable, this
6703 disables clocksource verification at runtime, as well
6704 as the stability checks done at bootup. Used to enable
6705 high-resolution timer mode on older hardware, and in
6706 virtualized environment.
6707 [x86] noirqtime: Do not use TSC to do irq accounting.
6708 Used to run time disable IRQ_TIME_ACCOUNTING on any
6709 platforms where RDTSC is slow and this accounting
6711 [x86] unstable: mark the TSC clocksource as unstable, this
6712 marks the TSC unconditionally unstable at bootup and
6713 avoids any further wobbles once the TSC watchdog notices.
6714 [x86] nowatchdog: disable clocksource watchdog. Used
6715 in situations with strict latency requirements (where
6716 interruptions from clocksource watchdog are not
6718 [x86] recalibrate: force recalibration against a HW timer
6719 (HPET or PM timer) on systems whose TSC frequency was
6720 obtained from HW or FW using either an MSR or CPUID(0x15).
6721 Warn if the difference is more than 500 ppm.
6722 [x86] watchdog: Use TSC as the watchdog clocksource with
6723 which to check other HW timers (HPET or PM timer), but
6724 only on systems where TSC has been deemed trustworthy.
6725 This will be suppressed by an earlier tsc=nowatchdog and
6726 can be overridden by a later tsc=nowatchdog. A console
6727 message will flag any such suppression or overriding.
6729 tsc_early_khz= [X86] Skip early TSC calibration and use the given
6730 value instead. Useful when the early TSC frequency discovery
6731 procedure is not reliable, such as on overclocked systems
6732 with CPUID.16h support and partial CPUID.15h support.
6733 Format: <unsigned int>
6735 tsx= [X86] Control Transactional Synchronization
6736 Extensions (TSX) feature in Intel processors that
6737 support TSX control.
6739 This parameter controls the TSX feature. The options are:
6741 on - Enable TSX on the system. Although there are
6742 mitigations for all known security vulnerabilities,
6743 TSX has been known to be an accelerator for
6744 several previous speculation-related CVEs, and
6745 so there may be unknown security risks associated
6746 with leaving it enabled.
6748 off - Disable TSX on the system. (Note that this
6749 option takes effect only on newer CPUs which are
6750 not vulnerable to MDS, i.e., have
6751 MSR_IA32_ARCH_CAPABILITIES.MDS_NO=1 and which get
6752 the new IA32_TSX_CTRL MSR through a microcode
6753 update. This new MSR allows for the reliable
6754 deactivation of the TSX functionality.)
6756 auto - Disable TSX if X86_BUG_TAA is present,
6757 otherwise enable TSX on the system.
6759 Not specifying this option is equivalent to tsx=off.
6761 See Documentation/admin-guide/hw-vuln/tsx_async_abort.rst
6764 tsx_async_abort= [X86,INTEL] Control mitigation for the TSX Async
6765 Abort (TAA) vulnerability.
6767 Similar to Micro-architectural Data Sampling (MDS)
6768 certain CPUs that support Transactional
6769 Synchronization Extensions (TSX) are vulnerable to an
6770 exploit against CPU internal buffers which can forward
6771 information to a disclosure gadget under certain
6774 In vulnerable processors, the speculatively forwarded
6775 data can be used in a cache side channel attack, to
6776 access data to which the attacker does not have direct
6779 This parameter controls the TAA mitigation. The
6782 full - Enable TAA mitigation on vulnerable CPUs
6785 full,nosmt - Enable TAA mitigation and disable SMT on
6786 vulnerable CPUs. If TSX is disabled, SMT
6787 is not disabled because CPU is not
6788 vulnerable to cross-thread TAA attacks.
6789 off - Unconditionally disable TAA mitigation
6791 On MDS-affected machines, tsx_async_abort=off can be
6792 prevented by an active MDS mitigation as both vulnerabilities
6793 are mitigated with the same mechanism so in order to disable
6794 this mitigation, you need to specify mds=off too.
6796 Not specifying this option is equivalent to
6797 tsx_async_abort=full. On CPUs which are MDS affected
6798 and deploy MDS mitigation, TAA mitigation is not
6799 required and doesn't provide any additional
6803 Documentation/admin-guide/hw-vuln/tsx_async_abort.rst
6805 turbografx.map[2|3]= [HW,JOY]
6806 TurboGraFX parallel port interface
6808 <port#>,<js1>,<js2>,<js3>,<js4>,<js5>,<js6>,<js7>
6809 See also Documentation/input/devices/joystick-parport.rst
6811 udbg-immortal [PPC] When debugging early kernel crashes that
6812 happen after console_init() and before a proper
6813 console driver takes over, this boot options might
6814 help "seeing" what's going on.
6816 uhash_entries= [KNL,NET]
6817 Set number of hash buckets for UDP/UDP-Lite connections
6820 [USB] Ignore overcurrent events (default N).
6821 Some badly-designed motherboards generate lots of
6822 bogus events, for ports that aren't wired to
6823 anything. Set this parameter to avoid log spamming.
6824 Note that genuine overcurrent events won't be
6828 [X86] Cause panic on unknown NMI.
6830 unwind_debug [X86-64]
6831 Enable unwinder debug output. This can be
6832 useful for debugging certain unwinder error
6833 conditions, including corrupt stacks and
6834 bad/missing unwinder metadata.
6836 usbcore.authorized_default=
6837 [USB] Default USB device authorization:
6838 (default -1 = authorized (same as 1),
6839 0 = not authorized, 1 = authorized, 2 = authorized
6840 if device connected to internal port)
6842 usbcore.autosuspend=
6843 [USB] The autosuspend time delay (in seconds) used
6844 for newly-detected USB devices (default 2). This
6845 is the time required before an idle device will be
6846 autosuspended. Devices for which the delay is set
6847 to a negative value won't be autosuspended at all.
6849 usbcore.usbfs_snoop=
6850 [USB] Set to log all usbfs traffic (default 0 = off).
6852 usbcore.usbfs_snoop_max=
6853 [USB] Maximum number of bytes to snoop in each URB
6856 usbcore.blinkenlights=
6857 [USB] Set to cycle leds on hubs (default 0 = off).
6859 usbcore.old_scheme_first=
6860 [USB] Start with the old device initialization
6861 scheme (default 0 = off).
6863 usbcore.usbfs_memory_mb=
6864 [USB] Memory limit (in MB) for buffers allocated by
6865 usbfs (default = 16, 0 = max = 2047).
6867 usbcore.use_both_schemes=
6868 [USB] Try the other device initialization scheme
6869 if the first one fails (default 1 = enabled).
6871 usbcore.initial_descriptor_timeout=
6872 [USB] Specifies timeout for the initial 64-byte
6873 USB_REQ_GET_DESCRIPTOR request in milliseconds
6874 (default 5000 = 5.0 seconds).
6876 usbcore.nousb [USB] Disable the USB subsystem
6879 [USB] A list of quirk entries to augment the built-in
6880 usb core quirk list. List entries are separated by
6881 commas. Each entry has the form
6882 VendorID:ProductID:Flags. The IDs are 4-digit hex
6883 numbers and Flags is a set of letters. Each letter
6884 will change the built-in quirk; setting it if it is
6885 clear and clearing it if it is set. The letters have
6886 the following meanings:
6887 a = USB_QUIRK_STRING_FETCH_255 (string
6888 descriptors must not be fetched using
6890 b = USB_QUIRK_RESET_RESUME (device can't resume
6891 correctly so reset it instead);
6892 c = USB_QUIRK_NO_SET_INTF (device can't handle
6893 Set-Interface requests);
6894 d = USB_QUIRK_CONFIG_INTF_STRINGS (device can't
6895 handle its Configuration or Interface
6897 e = USB_QUIRK_RESET (device can't be reset
6898 (e.g morph devices), don't use reset);
6899 f = USB_QUIRK_HONOR_BNUMINTERFACES (device has
6900 more interface descriptions than the
6901 bNumInterfaces count, and can't handle
6902 talking to these interfaces);
6903 g = USB_QUIRK_DELAY_INIT (device needs a pause
6904 during initialization, after we read
6905 the device descriptor);
6906 h = USB_QUIRK_LINEAR_UFRAME_INTR_BINTERVAL (For
6907 high speed and super speed interrupt
6908 endpoints, the USB 2.0 and USB 3.0 spec
6909 require the interval in microframes (1
6910 microframe = 125 microseconds) to be
6911 calculated as interval = 2 ^
6913 Devices with this quirk report their
6914 bInterval as the result of this
6915 calculation instead of the exponent
6916 variable used in the calculation);
6917 i = USB_QUIRK_DEVICE_QUALIFIER (device can't
6918 handle device_qualifier descriptor
6920 j = USB_QUIRK_IGNORE_REMOTE_WAKEUP (device
6921 generates spurious wakeup, ignore
6922 remote wakeup capability);
6923 k = USB_QUIRK_NO_LPM (device can't handle Link
6925 l = USB_QUIRK_LINEAR_FRAME_INTR_BINTERVAL
6926 (Device reports its bInterval as linear
6927 frames instead of the USB 2.0
6929 m = USB_QUIRK_DISCONNECT_SUSPEND (Device needs
6930 to be disconnected before suspend to
6931 prevent spurious wakeup);
6932 n = USB_QUIRK_DELAY_CTRL_MSG (Device needs a
6933 pause after every control message);
6934 o = USB_QUIRK_HUB_SLOW_RESET (Hub needs extra
6935 delay after resetting its port);
6936 p = USB_QUIRK_SHORT_SET_ADDRESS_REQ_TIMEOUT
6937 (Reduce timeout of the SET_ADDRESS
6938 request from 5000 ms to 500 ms);
6939 Example: quirks=0781:5580:bk,0a5c:5834:gij
6942 [USBHID] The interval which mice are to be polled at.
6945 [USBHID] The interval which joysticks are to be polled at.
6948 [USBHID] The interval which keyboards are to be polled at.
6950 usb-storage.delay_use=
6951 [UMS] The delay in seconds before a new device is
6952 scanned for Logical Units (default 1).
6955 [UMS] A list of quirks entries to supplement or
6956 override the built-in unusual_devs list. List
6957 entries are separated by commas. Each entry has
6958 the form VID:PID:Flags where VID and PID are Vendor
6959 and Product ID values (4-digit hex numbers) and
6960 Flags is a set of characters, each corresponding
6961 to a common usb-storage quirk flag as follows:
6962 a = SANE_SENSE (collect more than 18 bytes
6963 of sense data, not on uas);
6964 b = BAD_SENSE (don't collect more than 18
6965 bytes of sense data, not on uas);
6966 c = FIX_CAPACITY (decrease the reported
6967 device capacity by one sector);
6968 d = NO_READ_DISC_INFO (don't use
6969 READ_DISC_INFO command, not on uas);
6970 e = NO_READ_CAPACITY_16 (don't use
6971 READ_CAPACITY_16 command);
6972 f = NO_REPORT_OPCODES (don't use report opcodes
6974 g = MAX_SECTORS_240 (don't transfer more than
6975 240 sectors at a time, uas only);
6976 h = CAPACITY_HEURISTICS (decrease the
6977 reported device capacity by one
6978 sector if the number is odd);
6979 i = IGNORE_DEVICE (don't bind to this
6981 j = NO_REPORT_LUNS (don't use report luns
6983 k = NO_SAME (do not use WRITE_SAME, uas only)
6984 l = NOT_LOCKABLE (don't try to lock and
6985 unlock ejectable media, not on uas);
6986 m = MAX_SECTORS_64 (don't transfer more
6987 than 64 sectors = 32 KB at a time,
6989 n = INITIAL_READ10 (force a retry of the
6990 initial READ(10) command, not on uas);
6991 o = CAPACITY_OK (accept the capacity
6992 reported by the device, not on uas);
6993 p = WRITE_CACHE (the device cache is ON
6994 by default, not on uas);
6995 r = IGNORE_RESIDUE (the device reports
6996 bogus residue values, not on uas);
6997 s = SINGLE_LUN (the device has only one
6999 t = NO_ATA_1X (don't allow ATA(12) and ATA(16)
7000 commands, uas only);
7001 u = IGNORE_UAS (don't bind to the uas driver);
7002 w = NO_WP_DETECT (don't test whether the
7003 medium is write-protected).
7004 y = ALWAYS_SYNC (issue a SYNCHRONIZE_CACHE
7005 even if the device claims no cache,
7007 Example: quirks=0419:aaf5:rl,0421:0433:rc
7009 user_debug= [KNL,ARM]
7011 See arch/arm/Kconfig.debug help text.
7012 1 - undefined instruction events
7014 4 - invalid data aborts
7017 Example: user_debug=31
7020 [X86] Flags controlling user PTE allocations.
7022 nohigh = do not allocate PTE pages in
7023 HIGHMEM regardless of setting
7026 vdso= [X86,SH,SPARC]
7027 On X86_32, this is an alias for vdso32=. Otherwise:
7029 vdso=1: enable VDSO (the default)
7030 vdso=0: disable VDSO mapping
7032 vdso32= [X86] Control the 32-bit vDSO
7033 vdso32=1: enable 32-bit VDSO
7034 vdso32=0 or vdso32=2: disable 32-bit VDSO
7036 See the help text for CONFIG_COMPAT_VDSO for more
7037 details. If CONFIG_COMPAT_VDSO is set, the default is
7038 vdso32=0; otherwise, the default is vdso32=1.
7040 For compatibility with older kernels, vdso32=2 is an
7043 Try vdso32=0 if you encounter an error that says:
7044 dl_main: Assertion `(void *) ph->p_vaddr == _rtld_local._dl_sysinfo_dso' failed!
7047 vector=percpu: enable percpu vector domain
7049 video= [FB] Frame buffer configuration
7050 See Documentation/fb/modedb.rst.
7052 video.brightness_switch_enabled= [ACPI]
7054 If set to 1, on receiving an ACPI notify event
7055 generated by hotkey, video driver will adjust brightness
7056 level and then send out the event to user space through
7057 the allocated input device. If set to 0, video driver
7058 will only send out the event without touching backlight
7063 [VMMIO] Memory mapped virtio (platform) device.
7065 <size>@<baseaddr>:<irq>[:<id>]
7067 <size> := size (can use standard suffixes
7069 <baseaddr> := physical base address
7070 <irq> := interrupt number (as passed to
7072 <id> := (optional) platform device id
7074 virtio_mmio.device=1K@0x100b0000:48:7
7076 Can be used multiple times for multiple devices.
7078 vga= [BOOT,X86-32] Select a particular video mode
7079 See Documentation/arch/x86/boot.rst and
7080 Documentation/admin-guide/svga.rst.
7081 Use vga=ask for menu.
7082 This is actually a boot loader parameter; the value is
7083 passed to the kernel using a special protocol.
7085 vm_debug[=options] [KNL] Available with CONFIG_DEBUG_VM=y.
7086 May slow down system boot speed, especially when
7087 enabled on systems with a large amount of memory.
7088 All options are enabled by default, and this
7089 interface is meant to allow for selectively
7090 enabling or disabling specific virtual memory
7093 Available options are:
7094 P Enable page structure init time poisoning
7095 - Disable all of the above options
7097 vmalloc=nn[KMG] [KNL,BOOT] Forces the vmalloc area to have an exact
7098 size of <nn>. This can be used to increase the
7099 minimum size (128MB on x86). It can also be used to
7100 decrease the size and leave more room for directly
7103 vmcp_cma=nn[MG] [KNL,S390]
7104 Sets the memory size reserved for contiguous memory
7105 allocations for the vmcp device driver.
7107 vmhalt= [KNL,S390] Perform z/VM CP command after system halt.
7110 vmpanic= [KNL,S390] Perform z/VM CP command after kernel panic.
7113 vmpoff= [KNL,S390] Perform z/VM CP command after power off.
7117 Controls the behavior of vsyscalls (i.e. calls to
7118 fixed addresses of 0xffffffffff600x00 from legacy
7119 code). Most statically-linked binaries and older
7120 versions of glibc use these calls. Because these
7121 functions are at fixed addresses, they make nice
7122 targets for exploits that can control RIP.
7124 emulate Vsyscalls turn into traps and are emulated
7125 reasonably safely. The vsyscall page is
7128 xonly [default] Vsyscalls turn into traps and are
7129 emulated reasonably safely. The vsyscall
7130 page is not readable.
7132 none Vsyscalls don't work at all. This makes
7133 them quite hard to use for exploits but
7134 might break your system.
7136 vt.color= [VT] Default text color.
7137 Format: 0xYX, X = foreground, Y = background.
7138 Default: 0x07 = light gray on black.
7140 vt.cur_default= [VT] Default cursor shape.
7141 Format: 0xCCBBAA, where AA, BB, and CC are the same as
7142 the parameters of the <Esc>[?A;B;Cc escape sequence;
7143 see VGA-softcursor.txt. Default: 2 = underline.
7145 vt.default_blu= [VT]
7146 Format: <blue0>,<blue1>,<blue2>,...,<blue15>
7147 Change the default blue palette of the console.
7148 This is a 16-member array composed of values
7151 vt.default_grn= [VT]
7152 Format: <green0>,<green1>,<green2>,...,<green15>
7153 Change the default green palette of the console.
7154 This is a 16-member array composed of values
7157 vt.default_red= [VT]
7158 Format: <red0>,<red1>,<red2>,...,<red15>
7159 Change the default red palette of the console.
7160 This is a 16-member array composed of values
7166 Set system-wide default UTF-8 mode for all tty's.
7167 Default is 1, i.e. UTF-8 mode is enabled for all
7168 newly opened terminals.
7170 vt.global_cursor_default=
7173 Set system-wide default for whether a cursor
7174 is shown on new VTs. Default is -1,
7175 i.e. cursors will be created by default unless
7176 overridden by individual drivers. 0 will hide
7177 cursors, 1 will display them.
7179 vt.italic= [VT] Default color for italic text; 0-15.
7182 vt.underline= [VT] Default color for underlined text; 0-15.
7185 watchdog timers [HW,WDT] For information on watchdog timers,
7186 see Documentation/watchdog/watchdog-parameters.rst
7187 or other driver-specific files in the
7188 Documentation/watchdog/ directory.
7192 Set the hard lockup detector stall duration
7193 threshold in seconds. The soft lockup detector
7194 threshold is set to twice the value. A value of 0
7195 disables both lockup detectors. Default is 10
7198 workqueue.unbound_cpus=
7199 [KNL,SMP] Specify to constrain one or some CPUs
7200 to use in unbound workqueues.
7202 By default, all online CPUs are available for
7205 workqueue.watchdog_thresh=
7206 If CONFIG_WQ_WATCHDOG is configured, workqueue can
7207 warn stall conditions and dump internal state to
7208 help debugging. 0 disables workqueue stall
7209 detection; otherwise, it's the stall threshold
7210 duration in seconds. The default value is 30 and
7211 it can be updated at runtime by writing to the
7212 corresponding sysfs file.
7214 workqueue.cpu_intensive_thresh_us=
7215 Per-cpu work items which run for longer than this
7216 threshold are automatically considered CPU intensive
7217 and excluded from concurrency management to prevent
7218 them from noticeably delaying other per-cpu work
7219 items. Default is 10000 (10ms).
7221 If CONFIG_WQ_CPU_INTENSIVE_REPORT is set, the kernel
7222 will report the work functions which violate this
7223 threshold repeatedly. They are likely good
7224 candidates for using WQ_UNBOUND workqueues instead.
7226 workqueue.power_efficient
7227 Per-cpu workqueues are generally preferred because
7228 they show better performance thanks to cache
7229 locality; unfortunately, per-cpu workqueues tend to
7230 be more power hungry than unbound workqueues.
7232 Enabling this makes the per-cpu workqueues which
7233 were observed to contribute significantly to power
7234 consumption unbound, leading to measurably lower
7235 power usage at the cost of small performance
7238 The default value of this parameter is determined by
7239 the config option CONFIG_WQ_POWER_EFFICIENT_DEFAULT.
7241 workqueue.default_affinity_scope=
7242 Select the default affinity scope to use for unbound
7243 workqueues. Can be one of "cpu", "smt", "cache",
7244 "numa" and "system". Default is "cache". For more
7245 information, see the Affinity Scopes section in
7246 Documentation/core-api/workqueue.rst.
7248 This can be changed after boot by writing to the
7249 matching /sys/module/workqueue/parameters file. All
7250 workqueues with the "default" affinity scope will be
7251 updated accordignly.
7253 workqueue.debug_force_rr_cpu
7254 Workqueue used to implicitly guarantee that work
7255 items queued without explicit CPU specified are put
7256 on the local CPU. This guarantee is no longer true
7257 and while local CPU is still preferred work items
7258 may be put on foreign CPUs. This debug option
7259 forces round-robin CPU selection to flush out
7260 usages which depend on the now broken guarantee.
7261 When enabled, memory and cache locality will be
7264 writecombine= [LOONGARCH] Control the MAT (Memory Access Type) of
7267 on - Enable writecombine, use WUC for ioremap_wc()
7268 off - Disable writecombine, use SUC for ioremap_wc()
7270 x2apic_phys [X86-64,APIC] Use x2apic physical mode instead of
7271 default x2apic cluster mode on platforms
7274 xen_512gb_limit [KNL,X86-64,XEN]
7275 Restricts the kernel running paravirtualized under Xen
7276 to use only up to 512 GB of RAM. The reason to do so is
7277 crash analysis tools and Xen tools for doing domain
7278 save/restore/migration must be enabled to handle larger
7281 xen_emul_unplug= [HW,X86,XEN]
7282 Unplug Xen emulated devices
7283 Format: [unplug0,][unplug1]
7284 ide-disks -- unplug primary master IDE devices
7285 aux-ide-disks -- unplug non-primary-master IDE devices
7286 nics -- unplug network devices
7287 all -- unplug all emulated devices (NICs and IDE disks)
7288 unnecessary -- unplugging emulated devices is
7289 unnecessary even if the host did not respond to
7291 never -- do not unplug even if version check succeeds
7293 xen_legacy_crash [X86,XEN]
7294 Crash from Xen panic notifier, without executing late
7295 panic() code such as dumping handler.
7297 xen_msr_safe= [X86,XEN]
7299 Select whether to always use non-faulting (safe) MSR
7300 access functions when running as Xen PV guest. The
7301 default value is controlled by CONFIG_XEN_PV_MSR_SAFE.
7303 xen_nopvspin [X86,XEN]
7304 Disables the qspinlock slowpath using Xen PV optimizations.
7305 This parameter is obsoleted by "nopvspin" parameter, which
7306 has equivalent effect for XEN platform.
7309 Disables the PV optimizations forcing the HVM guest to
7310 run as generic HVM guest with no PV drivers.
7311 This option is obsoleted by the "nopv" option, which
7312 has equivalent effect for XEN platform.
7314 xen_no_vector_callback
7315 [KNL,X86,XEN] Disable the vector callback for Xen
7316 event channel interrupts.
7318 xen_scrub_pages= [XEN]
7319 Boolean option to control scrubbing pages before giving them back
7320 to Xen, for use by other domains. Can be also changed at runtime
7321 with /sys/devices/system/xen_memory/xen_memory0/scrub_pages.
7322 Default value controlled with CONFIG_XEN_SCRUB_PAGES_DEFAULT.
7324 xen_timer_slop= [X86-64,XEN]
7325 Set the timer slop (in nanoseconds) for the virtual Xen
7326 timers (default is 100000). This adjusts the minimum
7327 delta of virtualized Xen timers, where lower values
7328 improve timer resolution at the expense of processing
7329 more timer interrupts.
7331 xen.balloon_boot_timeout= [XEN]
7332 The time (in seconds) to wait before giving up to boot
7333 in case initial ballooning fails to free enough memory.
7334 Applies only when running as HVM or PVH guest and
7335 started with less memory configured than allowed at
7336 max. Default is 180.
7338 xen.event_eoi_delay= [XEN]
7339 How long to delay EOI handling in case of event
7340 storms (jiffies). Default is 10.
7342 xen.event_loop_timeout= [XEN]
7343 After which time (jiffies) the event handling loop
7344 should start to delay EOI handling. Default is 2.
7346 xen.fifo_events= [XEN]
7347 Boolean parameter to disable using fifo event handling
7348 even if available. Normally fifo event handling is
7349 preferred over the 2-level event handling, as it is
7350 fairer and the number of possible event channels is
7351 much higher. Default is on (use fifo events).
7353 xirc2ps_cs= [NET,PCMCIA]
7355 <irq>,<irq_mask>,<io>,<full_duplex>,<do_sound>,<lockup_hack>[,<irq2>[,<irq3>[,<irq4>]]]
7358 By default on POWER9 and above, the kernel will
7359 natively use the XIVE interrupt controller. This option
7360 allows the fallback firmware mode to be used:
7362 off Fallback to firmware control of XIVE interrupt
7363 controller on both pseries and powernv
7364 platforms. Only useful on POWER9 and above.
7366 xive.store-eoi=off [PPC]
7367 By default on POWER10 and above, the kernel will use
7368 stores for EOI handling when the XIVE interrupt mode
7369 is active. This option allows the XIVE driver to use
7370 loads instead, as on POWER9.
7372 xhci-hcd.quirks [USB,KNL]
7373 A hex value specifying bitmask with supplemental xhci
7374 host controller quirks. Meaning of each bit can be
7375 consulted in header drivers/usb/host/xhci.h.
7378 Format: { early | on | rw | ro | off }
7379 Controls if xmon debugger is enabled. Default is off.
7380 Passing only "xmon" is equivalent to "xmon=early".
7381 early Call xmon as early as possible on boot; xmon
7382 debugger is called from setup_arch().
7383 on xmon debugger hooks will be installed so xmon
7384 is only called on a kernel crash. Default mode,
7385 i.e. either "ro" or "rw" mode, is controlled
7386 with CONFIG_XMON_DEFAULT_RO_MODE.
7387 rw xmon debugger hooks will be installed so xmon
7388 is called only on a kernel crash, mode is write,
7389 meaning SPR registers, memory and, other data
7390 can be written using xmon commands.
7391 ro same as "rw" option above but SPR registers,
7392 memory, and other data can't be written using
7394 off xmon is disabled.