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,EARLY]
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 nocmcff -- Disable firmware first mode for corrected
25 errors. This disables parsing the HEST CMC error
26 source to check if firmware has set the FF flag. This
27 may result in duplicate corrected error reports.
28 nospcr -- disable console in ACPI SPCR table as
29 default _serial_ console on ARM64
30 For ARM64, ONLY "acpi=off", "acpi=on", "acpi=force" or
31 "acpi=nospcr" are available
32 For RISCV64, ONLY "acpi=off", "acpi=on" or "acpi=force"
35 See also Documentation/power/runtime_pm.rst, pci=noacpi
37 acpi_apic_instance= [ACPI,IOAPIC,EARLY]
39 2: use 2nd APIC table, if available
40 1,0: use 1st APIC table
43 acpi_backlight= [HW,ACPI]
44 { vendor | video | native | none }
45 If set to vendor, prefer vendor-specific driver
46 (e.g. thinkpad_acpi, sony_acpi, etc.) instead
47 of the ACPI video.ko driver.
48 If set to video, use the ACPI video.ko driver.
49 If set to native, use the device's native backlight mode.
50 If set to none, disable the ACPI backlight interface.
52 acpi_force_32bit_fadt_addr [ACPI,EARLY]
53 force FADT to use 32 bit addresses rather than the
54 64 bit X_* addresses. Some firmware have broken 64
55 bit addresses for force ACPI ignore these and use
56 the older legacy 32 bit addresses.
58 acpica_no_return_repair [HW, ACPI]
59 Disable AML predefined validation mechanism
60 This mechanism can repair the evaluation result to make
61 the return objects more ACPI specification compliant.
62 This option is useful for developers to identify the
63 root cause of an AML interpreter issue when the issue
64 has something to do with the repair mechanism.
66 acpi.debug_layer= [HW,ACPI,ACPI_DEBUG]
67 acpi.debug_level= [HW,ACPI,ACPI_DEBUG]
69 CONFIG_ACPI_DEBUG must be enabled to produce any ACPI
70 debug output. Bits in debug_layer correspond to a
71 _COMPONENT in an ACPI source file, e.g.,
72 #define _COMPONENT ACPI_EVENTS
73 Bits in debug_level correspond to a level in
74 ACPI_DEBUG_PRINT statements, e.g.,
75 ACPI_DEBUG_PRINT((ACPI_DB_INFO, ...
76 The debug_level mask defaults to "info". See
77 Documentation/firmware-guide/acpi/debug.rst for more information about
78 debug layers and levels.
80 Enable processor driver info messages:
81 acpi.debug_layer=0x20000000
82 Enable AML "Debug" output, i.e., stores to the Debug
83 object while interpreting AML:
84 acpi.debug_layer=0xffffffff acpi.debug_level=0x2
85 Enable all messages related to ACPI hardware:
86 acpi.debug_layer=0x2 acpi.debug_level=0xffffffff
88 Some values produce so much output that the system is
89 unusable. The "log_buf_len" parameter may be useful
90 if you need to capture more output.
92 acpi_enforce_resources= [ACPI]
94 Check for resource conflicts between native drivers
95 and ACPI OperationRegions (SystemIO and SystemMemory
96 only). IO ports and memory declared in ACPI might be
97 used by the ACPI subsystem in arbitrary AML code and
98 can interfere with legacy drivers.
99 strict (default): access to resources claimed by ACPI
100 is denied; legacy drivers trying to access reserved
101 resources will fail to bind to device using them.
102 lax: access to resources claimed by ACPI is allowed;
103 legacy drivers trying to access reserved resources
104 will bind successfully but a warning message is logged.
105 no: ACPI OperationRegions are not marked as reserved,
106 no further checks are performed.
108 acpi_force_table_verification [HW,ACPI,EARLY]
109 Enable table checksum verification during early stage.
110 By default, this is disabled due to x86 early mapping
113 acpi_irq_balance [HW,ACPI]
114 ACPI will balance active IRQs
117 acpi_irq_nobalance [HW,ACPI]
118 ACPI will not move active IRQs (default)
121 acpi_irq_isa= [HW,ACPI] If irq_balance, mark listed IRQs used by ISA
122 Format: <irq>,<irq>...
124 acpi_irq_pci= [HW,ACPI] If irq_balance, clear listed IRQs for
126 Format: <irq>,<irq>...
128 acpi_mask_gpe= [HW,ACPI]
129 Due to the existence of _Lxx/_Exx, some GPEs triggered
130 by unsupported hardware/firmware features can result in
131 GPE floodings that cannot be automatically disabled by
133 This facility can be used to prevent such uncontrolled
135 Format: <byte> or <bitmap-list>
137 acpi_no_auto_serialize [HW,ACPI]
138 Disable auto-serialization of AML methods
139 AML control methods that contain the opcodes to create
140 named objects will be marked as "Serialized" by the
141 auto-serialization feature.
142 This feature is enabled by default.
143 This option allows to turn off the feature.
145 acpi_no_memhotplug [ACPI] Disable memory hotplug. Useful for kdump
148 acpi_no_static_ssdt [HW,ACPI,EARLY]
149 Disable installation of static SSDTs at early boot time
150 By default, SSDTs contained in the RSDT/XSDT will be
151 installed automatically and they will appear under
152 /sys/firmware/acpi/tables.
153 This option turns off this feature.
154 Note that specifying this option does not affect
155 dynamic table installation which will install SSDT
156 tables to /sys/firmware/acpi/tables/dynamic.
158 acpi_no_watchdog [HW,ACPI,WDT]
159 Ignore the ACPI-based watchdog interface (WDAT) and let
160 a native driver control the watchdog device instead.
162 acpi_rsdp= [ACPI,EFI,KEXEC,EARLY]
163 Pass the RSDP address to the kernel, mostly used
164 on machines running EFI runtime service to boot the
165 second kernel for kdump.
167 acpi_os_name= [HW,ACPI] Tell ACPI BIOS the name of the OS
168 Format: To spoof as Windows 98: ="Microsoft Windows"
170 acpi_rev_override [ACPI] Override the _REV object to return 5 (instead
171 of 2 which is mandated by ACPI 6) as the supported ACPI
172 specification revision (when using this switch, it may
173 be necessary to carry out a cold reboot _twice_ in a
174 row to make it take effect on the platform firmware).
176 acpi_osi= [HW,ACPI] Modify list of supported OS interface strings
177 acpi_osi="string1" # add string1
178 acpi_osi="!string2" # remove string2
179 acpi_osi=!* # remove all strings
180 acpi_osi=! # disable all built-in OS vendor
182 acpi_osi=!! # enable all built-in OS vendor
184 acpi_osi= # disable all strings
186 'acpi_osi=!' can be used in combination with single or
187 multiple 'acpi_osi="string1"' to support specific OS
188 vendor string(s). Note that such command can only
189 affect the default state of the OS vendor strings, thus
190 it cannot affect the default state of the feature group
191 strings and the current state of the OS vendor strings,
192 specifying it multiple times through kernel command line
193 is meaningless. This command is useful when one do not
194 care about the state of the feature group strings which
195 should be controlled by the OSPM.
197 1. 'acpi_osi=! acpi_osi="Windows 2000"' is equivalent
198 to 'acpi_osi="Windows 2000" acpi_osi=!', they all
199 can make '_OSI("Windows 2000")' TRUE.
201 'acpi_osi=' cannot be used in combination with other
202 'acpi_osi=' command lines, the _OSI method will not
203 exist in the ACPI namespace. NOTE that such command can
204 only affect the _OSI support state, thus specifying it
205 multiple times through kernel command line is also
208 1. 'acpi_osi=' can make 'CondRefOf(_OSI, Local1)'
211 'acpi_osi=!*' can be used in combination with single or
212 multiple 'acpi_osi="string1"' to support specific
213 string(s). Note that such command can affect the
214 current state of both the OS vendor strings and the
215 feature group strings, thus specifying it multiple times
216 through kernel command line is meaningful. But it may
217 still not able to affect the final state of a string if
218 there are quirks related to this string. This command
219 is useful when one want to control the state of the
220 feature group strings to debug BIOS issues related to
223 1. 'acpi_osi="Module Device" acpi_osi=!*' can make
224 '_OSI("Module Device")' FALSE.
225 2. 'acpi_osi=!* acpi_osi="Module Device"' can make
226 '_OSI("Module Device")' TRUE.
227 3. 'acpi_osi=! acpi_osi=!* acpi_osi="Windows 2000"' is
229 'acpi_osi=!* acpi_osi=! acpi_osi="Windows 2000"'
231 'acpi_osi=!* acpi_osi="Windows 2000" acpi_osi=!',
232 they all will make '_OSI("Windows 2000")' TRUE.
235 Override the pmtimer bug detection: force the kernel
236 to assume that this machine's pmtimer latches its value
237 and always returns good values.
239 acpi_sci= [HW,ACPI,EARLY] ACPI System Control Interrupt trigger mode
240 Format: { level | edge | high | low }
242 acpi_skip_timer_override [HW,ACPI,EARLY]
243 Recognize and ignore IRQ0/pin2 Interrupt Override.
244 For broken nForce2 BIOS resulting in XT-PIC timer.
246 acpi_sleep= [HW,ACPI] Sleep options
247 Format: { s3_bios, s3_mode, s3_beep, s4_hwsig,
248 s4_nohwsig, old_ordering, nonvs,
249 sci_force_enable, nobl }
250 See Documentation/power/video.rst for information on
252 s3_beep is for debugging; it makes the PC's speaker beep
253 as soon as the kernel's real-mode entry point is called.
254 s4_hwsig causes the kernel to check the ACPI hardware
255 signature during resume from hibernation, and gracefully
256 refuse to resume if it has changed. This complies with
257 the ACPI specification but not with reality, since
258 Windows does not do this and many laptops do change it
259 on docking. So the default behaviour is to allow resume
260 and simply warn when the signature changes, unless the
261 s4_hwsig option is enabled.
262 s4_nohwsig prevents ACPI hardware signature from being
263 used (or even warned about) during resume.
264 old_ordering causes the ACPI 1.0 ordering of the _PTS
265 control method, with respect to putting devices into
266 low power states, to be enforced (the ACPI 2.0 ordering
267 of _PTS is used by default).
268 nonvs prevents the kernel from saving/restoring the
269 ACPI NVS memory during suspend/hibernation and resume.
270 sci_force_enable causes the kernel to set SCI_EN directly
271 on resume from S1/S3 (which is against the ACPI spec,
272 but some broken systems don't work without it).
273 nobl causes the internal blacklist of systems known to
274 behave incorrectly in some ways with respect to system
275 suspend and resume to be ignored (use wisely).
277 acpi_use_timer_override [HW,ACPI,EARLY]
278 Use timer override. For some broken Nvidia NF5 boards
279 that require a timer override, but don't have HPET
281 add_efi_memmap [EFI,X86,EARLY] Include EFI memory map in
282 kernel's map of available physical RAM.
285 { off | try_unsupported }
286 off: disable AGP support
287 try_unsupported: try to drive unsupported chipsets
288 (may crash computer or cause data corruption)
291 See Documentation/sound/alsa-configuration.rst
294 Allow the default userspace alignment fault handler
295 behaviour to be specified. Bit 0 enables warnings,
296 bit 1 enables fixups, and bit 2 sends a segfault.
298 align_va_addr= [X86-64]
299 Align virtual addresses by clearing slice [14:12] when
300 allocating a VMA at process creation time. This option
301 gives you up to 3% performance improvement on AMD F15h
302 machines (where it is enabled by default) for a
303 CPU-intensive style benchmark, and it can vary highly in
304 a microbenchmark depending on workload and compiler.
306 32: only for 32-bit processes
307 64: only for 64-bit processes
308 on: enable for both 32- and 64-bit processes
309 off: disable for both 32- and 64-bit processes
311 alloc_snapshot [FTRACE]
312 Allocate the ftrace snapshot buffer on boot up when the
313 main buffer is allocated. This is handy if debugging
314 and you need to use tracing_snapshot() on boot up, and
315 do not want to use tracing_snapshot_alloc() as it needs
316 to be done where GFP_KERNEL allocations are allowed.
318 allow_mismatched_32bit_el0 [ARM64,EARLY]
319 Allow execve() of 32-bit applications and setting of the
320 PER_LINUX32 personality on systems where only a strict
321 subset of the CPUs support 32-bit EL0. When this
322 parameter is present, the set of CPUs supporting 32-bit
323 EL0 is indicated by /sys/devices/system/cpu/aarch32_el0
324 and hot-unplug operations may be restricted.
326 See Documentation/arch/arm64/asymmetric-32bit.rst for more
329 amd_iommu= [HW,X86-64]
330 Pass parameters to the AMD IOMMU driver in the system.
332 fullflush - Deprecated, equivalent to iommu.strict=1
333 off - do not initialize any AMD IOMMU found in
335 force_isolation - Force device isolation for all
336 devices. The IOMMU driver is not
337 allowed anymore to lift isolation
338 requirements as needed. This option
339 does not override iommu=pt
340 force_enable - Force enable the IOMMU on platforms known
341 to be buggy with IOMMU enabled. Use this
343 pgtbl_v1 - Use v1 page table for DMA-API (Default).
344 pgtbl_v2 - Use v2 page table for DMA-API.
345 irtcachedis - Disable Interrupt Remapping Table (IRT) caching.
346 nohugepages - Limit page-sizes used for v1 page-tables
348 v2_pgsizes_only - Limit page-sizes used for v1 page-tables
352 amd_iommu_dump= [HW,X86-64]
353 Enable AMD IOMMU driver option to dump the ACPI table
354 for AMD IOMMU. With this option enabled, AMD IOMMU
355 driver will print ACPI tables for AMD IOMMU during
356 IOMMU initialization.
358 amd_iommu_intr= [HW,X86-64]
359 Specifies one of the following AMD IOMMU interrupt
361 legacy - Use legacy interrupt remapping mode.
362 vapic - Use virtual APIC mode, which allows IOMMU
363 to inject interrupts directly into guest.
364 This mode requires kvm-amd.avic=1.
365 (Default when IOMMU HW support is present.)
367 amd_pstate= [X86,EARLY]
369 Do not enable amd_pstate as the default
370 scaling driver for the supported processors
372 Use amd_pstate with passive mode as a scaling driver.
373 In this mode autonomous selection is disabled.
374 Driver requests a desired performance level and platform
375 tries to match the same performance level if it is
376 satisfied by guaranteed performance level.
378 Use amd_pstate_epp driver instance as the scaling driver,
379 driver provides a hint to the hardware if software wants
380 to bias toward performance (0x0) or energy efficiency (0xff)
381 to the CPPC firmware. then CPPC power algorithm will
382 calculate the runtime workload and adjust the realtime cores
385 Activate guided autonomous mode. Driver requests minimum and
386 maximum performance level and the platform autonomously
387 selects a performance level in this range and appropriate
388 to the current workload.
393 Disable amd-pstate preferred core.
395 amijoy.map= [HW,JOY] Amiga joystick support
396 Map of devices attached to JOY0DAT and JOY1DAT
398 See also Documentation/input/joydev/joystick.rst
400 analog.map= [HW,JOY] Analog joystick and gamepad support
401 Specifies type or capabilities of an analog joystick
402 connected to one of 16 gameports
403 Format: <type1>,<type2>,..<type16>
406 Power management functions (SPARCstation-4/5 + deriv.)
408 Disable APC CPU standby support. SPARCstation-Fox does
409 not play well with APC CPU idle - disable it if you have
410 APC and your system crashes randomly.
412 apic [APIC,X86-64] Use IO-APIC. Default.
414 apic= [APIC,X86,EARLY] Advanced Programmable Interrupt Controller
415 Change the output verbosity while booting
416 Format: { quiet (default) | verbose | debug }
417 Change the amount of debugging information output
418 when initialising the APIC and IO-APIC components.
420 apic_extnmi= [APIC,X86,EARLY] External NMI delivery setting
421 Format: { bsp (default) | all | none }
422 bsp: External NMI is delivered only to CPU 0
423 all: External NMIs are broadcast to all CPUs as a
425 none: External NMI is masked for all CPUs. This is
426 useful so that a dump capture kernel won't be
429 apicpmtimer Do APIC timer calibration using the pmtimer. Implies
430 apicmaintimer. Useful when your PIT timer is totally
434 See Documentation/networking/ipv6.rst.
436 apm= [APM] Advanced Power Management
437 See header of arch/x86/kernel/apm_32.c.
439 apparmor= [APPARMOR] Disable or enable AppArmor at boot time
440 Format: { "0" | "1" }
441 See security/apparmor/Kconfig help text
444 Default value is set via kernel config option.
446 arcrimi= [HW,NET] ARCnet - "RIM I" (entirely mem-mapped) cards
447 Format: <io>,<irq>,<nodeID>
449 arm64.no32bit_el0 [ARM64] Unconditionally disable the execution of
452 arm64.nobti [ARM64] Unconditionally disable Branch Target
453 Identification support
455 arm64.nogcs [ARM64] Unconditionally disable Guarded Control Stack
458 arm64.nomops [ARM64] Unconditionally disable Memory Copy and Memory
459 Set instructions support
461 arm64.nompam [ARM64] Unconditionally disable Memory Partitioning And
464 arm64.nomte [ARM64] Unconditionally disable Memory Tagging Extension
467 arm64.nopauth [ARM64] Unconditionally disable Pointer Authentication
470 arm64.nosme [ARM64] Unconditionally disable Scalable Matrix
473 arm64.nosve [ARM64] Unconditionally disable Scalable Vector
478 atarimouse= [HW,MOUSE] Atari Mouse
480 atkbd.extra= [HW] Enable extra LEDs and keys on IBM RapidAccess,
481 EzKey and similar keyboards
483 atkbd.reset= [HW] Reset keyboard during initialization
485 atkbd.set= [HW] Select keyboard code set
486 Format: <int> (2 = AT (default), 3 = PS/2)
488 atkbd.scroll= [HW] Enable scroll wheel on MS Office and similar
491 atkbd.softraw= [HW] Choose between synthetic and real raw mode
492 Format: <bool> (0 = real, 1 = synthetic (default))
494 atkbd.softrepeat= [HW]
495 Use software keyboard repeat
497 audit= [KNL] Enable the audit sub-system
498 Format: { "0" | "1" | "off" | "on" }
499 0 | off - kernel audit is disabled and can not be
500 enabled until the next reboot
501 unset - kernel audit is initialized but disabled and
502 will be fully enabled by the userspace auditd.
503 1 | on - kernel audit is initialized and partially
504 enabled, storing at most audit_backlog_limit
505 messages in RAM until it is fully enabled by the
509 audit_backlog_limit= [KNL] Set the audit queue size limit.
510 Format: <int> (must be >=0)
513 bau= [X86_UV] Enable the BAU on SGI UV. The default
514 behavior is to disable the BAU (i.e. bau=0).
515 Format: { "0" | "1" }
518 unset - Disable the BAU.
520 baycom_epp= [HW,AX25]
523 baycom_par= [HW,AX25] BayCom Parallel Port AX.25 Modem
525 See header of drivers/net/hamradio/baycom_par.c.
527 baycom_ser_fdx= [HW,AX25]
528 BayCom Serial Port AX.25 Modem (Full Duplex Mode)
529 Format: <io>,<irq>,<mode>[,<baud>]
530 See header of drivers/net/hamradio/baycom_ser_fdx.c.
532 baycom_ser_hdx= [HW,AX25]
533 BayCom Serial Port AX.25 Modem (Half Duplex Mode)
534 Format: <io>,<irq>,<mode>
535 See header of drivers/net/hamradio/baycom_ser_hdx.c.
537 bdev_allow_write_mounted=
539 Control the ability to open a mounted block device
540 for writing, i.e., allow / disallow writes that bypass
541 the FS. This was implemented as a means to prevent
542 fuzzers from crashing the kernel by overwriting the
543 metadata underneath a mounted FS without its awareness.
544 This also prevents destructive formatting of mounted
545 filesystems by naive storage tooling that don't use
546 O_EXCL. Default is Y and can be changed through the
547 Kconfig option CONFIG_BLK_DEV_WRITE_MOUNTED.
550 Disable BERT OS support on buggy BIOSes.
552 bgrt_disable [ACPI,X86,EARLY]
553 Disable BGRT to avoid flickering OEM logo.
555 blkdevparts= Manual partition parsing of block device(s) for
556 embedded devices based on command line input.
557 See Documentation/block/cmdline-partition.rst
559 boot_delay= [KNL,EARLY]
560 Milliseconds to delay each printk during boot.
561 Only works if CONFIG_BOOT_PRINTK_DELAY is enabled,
562 and you may also have to specify "lpj=". Boot_delay
563 values larger than 10 seconds (10000) are assumed
564 erroneous and ignored.
567 bootconfig [KNL,EARLY]
568 Extended command line options can be added to an initrd
569 and this will cause the kernel to look for it.
571 See Documentation/admin-guide/bootconfig.rst
573 bttv.card= [HW,V4L] bttv (bt848 + bt878 based grabber cards)
574 bttv.radio= Most important insmod options are available as
576 bttv.pll= See Documentation/admin-guide/media/bttv.rst
579 bulk_remove=off [PPC] This parameter disables the use of the pSeries
580 firmware feature for flushing multiple hpte entries
583 c101= [NET] Moxa C101 synchronous serial card
585 cachesize= [BUGS=X86-32] Override level 2 CPU cache size detection.
586 Sometimes CPU hardware bugs make them report the cache
587 size incorrectly. The kernel will attempt work arounds
588 to fix known problems, but for some CPUs it is not
589 possible to determine what the correct size should be.
590 This option provides an override for these situations.
593 [NET] Specifies amount of time (in seconds) that
594 the kernel should wait for a network carrier. By default
595 it waits 120 seconds.
597 ca_keys= [KEYS] This parameter identifies a specific key(s) on
598 the system trusted keyring to be used for certificate
600 format: { id:<keyid> | builtin }
602 cca= [MIPS,EARLY] Override the kernel pages' cache coherency
603 algorithm. Accepted values range from 0 to 7
604 inclusive. See arch/mips/include/asm/pgtable-bits.h
605 for platform specific values (SB1, Loongson3 and
608 ccw_timeout_log [S390]
609 See Documentation/arch/s390/common_io.rst for details.
611 cgroup_disable= [KNL] Disable a particular controller or optional feature
612 Format: {name of the controller(s) or feature(s) to disable}
613 The effects of cgroup_disable=foo are:
614 - foo isn't auto-mounted if you mount all cgroups in
616 - foo isn't visible as an individually mountable
618 - if foo is an optional feature then the feature is
619 disabled and corresponding cgroup files are not
621 {Currently only "memory" controller deal with this and
622 cut the overhead, others just disable the usage. So
623 only cgroup_disable=memory is actually worthy}
624 Specifying "pressure" disables per-cgroup pressure
625 stall information accounting feature
627 cgroup_no_v1= [KNL] Disable cgroup controllers and named hierarchies in v1
628 Format: { { controller | "all" | "named" }
629 [,{ controller | "all" | "named" }...] }
630 Like cgroup_disable, but only applies to cgroup v1;
631 the blacklisted controllers remain available in cgroup2.
632 "all" blacklists all controllers and "named" disables
633 named mounts. Specifying both "all" and "named" disables
636 cgroup_favordynmods= [KNL] Enable or Disable favordynmods.
637 Format: { "true" | "false" }
638 Defaults to the value of CONFIG_CGROUP_FAVOR_DYNMODS.
640 cgroup.memory= [KNL] Pass options to the cgroup memory controller.
642 nosocket -- Disable socket memory accounting.
643 nokmem -- Disable kernel memory accounting.
644 nobpf -- Disable BPF memory accounting.
646 checkreqprot= [SELINUX] Set initial checkreqprot flag value.
647 Format: { "0" | "1" }
648 See security/selinux/Kconfig help text.
649 0 -- check protection applied by kernel (includes
650 any implied execute protection).
651 1 -- check protection requested by application.
652 Default value is set via a kernel config option.
653 Value can be changed at runtime via
654 /sys/fs/selinux/checkreqprot.
655 Setting checkreqprot to 1 is deprecated.
658 See Documentation/arch/s390/common_io.rst for details.
660 clearcpuid=X[,X...] [X86]
661 Disable CPUID feature X for the kernel. See
662 arch/x86/include/asm/cpufeatures.h for the valid bit
663 numbers X. Note the Linux-specific bits are not necessarily
664 stable over kernel options, but the vendor-specific
666 X can also be a string as appearing in the flags: line
667 in /proc/cpuinfo which does not have the above
668 instability issue. However, not all features have names
670 Note that using this option will taint your kernel.
671 Also note that user programs calling CPUID directly
672 or using the feature without checking anything
673 will still see it. This just prevents it from
674 being used by the kernel or shown in /proc/cpuinfo.
675 Also note the kernel might malfunction if you disable
680 Prevents the clock framework from automatically gating
681 clocks that have not been explicitly enabled by a Linux
682 device driver but are enabled in hardware at reset or
683 by the bootloader/firmware. Note that this does not
684 force such clocks to be always-on nor does it reserve
685 those clocks in any way. This parameter is useful for
686 debug and development, but should not be needed on a
687 platform with proper driver support. For more
688 information, see Documentation/driver-api/clk.rst.
690 clock= [BUGS=X86-32, HW] gettimeofday clocksource override.
692 Forces specified clocksource (if available) to be used
693 when calculating gettimeofday(). If specified
694 clocksource is not available, it defaults to PIT.
695 Format: { pit | tsc | cyclone | pmtmr }
697 clocksource= Override the default clocksource
699 Override the default clocksource and use the clocksource
700 with the name specified.
701 Some clocksource names to choose from, depending on
703 [all] jiffies (this is the base, fallback clocksource)
705 [ARM] imx_timer1,OSTS,netx_timer,mpu_timer2,
706 pxa_timer,timer3,32k_counter,timer0_1
707 [X86-32] pit,hpet,tsc;
708 scx200_hrt on Geode; cyclone on IBM x440
716 clocksource.arm_arch_timer.evtstrm=
719 Enable/disable the eventstream feature of the ARM
720 architected timer so that code using WFE-based polling
721 loops can be debugged more effectively on production
724 clocksource.verify_n_cpus= [KNL]
725 Limit the number of CPUs checked for clocksources
726 marked with CLOCK_SOURCE_VERIFY_PERCPU that
727 are marked unstable due to excessive skew.
728 A negative value says to check all CPUs, while
729 zero says not to check any. Values larger than
730 nr_cpu_ids are silently truncated to nr_cpu_ids.
731 The actual CPUs are chosen randomly, with
732 no replacement if the same CPU is chosen twice.
734 clocksource-wdtest.holdoff= [KNL]
735 Set the time in seconds that the clocksource
736 watchdog test waits before commencing its tests.
737 Defaults to zero when built as a module and to
738 10 seconds when built into the kernel.
740 cma=nn[MG]@[start[MG][-end[MG]]]
742 Sets the size of kernel global memory area for
743 contiguous memory allocations and optionally the
744 placement constraint by the physical address range of
745 memory allocations. A value of 0 disables CMA
746 altogether. For more information, see
747 kernel/dma/contiguous.c
751 Sets the size of kernel per-numa memory area for
752 contiguous memory allocations. A value of 0 disables
753 per-numa CMA altogether. And If this option is not
754 specified, the default value is 0.
755 With per-numa CMA enabled, DMA users on node nid will
756 first try to allocate buffer from the pernuma area
757 which is located in node nid, if the allocation fails,
758 they will fallback to the global default memory area.
760 numa_cma=<node>:nn[MG][,<node>:nn[MG]]
762 Sets the size of kernel numa memory area for
763 contiguous memory allocations. It will reserve CMA
764 area for the specified node.
766 With numa CMA enabled, DMA users on node nid will
767 first try to allocate buffer from the numa area
768 which is located in node nid, if the allocation fails,
769 they will fallback to the global default memory area.
771 cmo_free_hint= [PPC] Format: { yes | no }
772 Specify whether pages are marked as being inactive
773 when they are freed. This is used in CMO environments
774 to determine OS memory pressure for page stealing by
778 coherent_pool=nn[KMG] [ARM,KNL,EARLY]
779 Sets the size of memory pool for coherent, atomic dma
780 allocations, by default set to 256K.
782 com20020= [HW,NET] ARCnet - COM20020 chipset
784 <io>[,<irq>[,<nodeID>[,<backplane>[,<ckp>[,<timeout>]]]]]
786 com90io= [HW,NET] ARCnet - COM90xx chipset (IO-mapped buffers)
790 ARCnet - COM90xx chipset (memory-mapped buffers)
791 Format: <io>[,<irq>[,<memstart>]]
793 condev= [HW,S390] console device
796 con3215_drop= [S390,EARLY] 3215 console drop mode.
798 When set to true, drop data on the 3215 console when
799 the console buffer is full. In this case the
800 operator using a 3270 terminal emulator (for example
801 x3270) does not have to enter the clear key for the
802 console output to advance and the kernel to continue.
803 This leads to a much faster boot time when a 3270
804 terminal emulator is active. If no 3270 terminal
805 emulator is used, this parameter has no effect.
807 console= [KNL] Output console device and options.
809 tty<n> Use the virtual console device <n>.
813 Use the specified serial port. The options are of
814 the form "bbbbpnf", where "bbbb" is the baud rate,
815 "p" is parity ("n", "o", or "e"), "n" is number of
816 bits, and "f" is flow control ("r" for RTS or
817 omit it). Default is "9600n8".
819 See Documentation/admin-guide/serial-console.rst for more
821 Documentation/networking/netconsole.rst for an
824 <DEVNAME>:<n>.<n>[,options]
825 Use the specified serial port on the serial core bus.
826 The addressing uses DEVNAME of the physical serial port
827 device, followed by the serial core controller instance,
828 and the serial port instance. The options are the same
829 as documented for the ttyS addressing above.
831 The mapping of the serial ports to the tty instances
834 $ ls -d /sys/bus/serial-base/devices/*:*.*/tty/*
835 /sys/bus/serial-base/devices/00:04:0.0/tty/ttyS0
837 In the above example, the console can be addressed with
838 console=00:04:0.0. Note that a console addressed this
839 way will only get added when the related device driver
840 is ready. The use of an earlycon parameter in addition to
841 the console may be desired for console output early on.
843 uart[8250],io,<addr>[,options]
844 uart[8250],mmio,<addr>[,options]
845 uart[8250],mmio16,<addr>[,options]
846 uart[8250],mmio32,<addr>[,options]
847 uart[8250],0x<addr>[,options]
848 Start an early, polled-mode console on the 8250/16550
849 UART at the specified I/O port or MMIO address,
850 switching to the matching ttyS device later.
851 MMIO inter-register address stride is either 8-bit
852 (mmio), 16-bit (mmio16), or 32-bit (mmio32).
853 If none of [io|mmio|mmio16|mmio32], <addr> is assumed
854 to be equivalent to 'mmio'. 'options' are specified in
855 the same format described for ttyS above; if unspecified,
856 the h/w is not re-initialized.
858 hvc<n> Use the hypervisor console device <n>. This is for
859 both Xen and PowerPC hypervisors.
862 Use to disable console output, i.e., to have kernel
863 console messages discarded.
864 This must be the only console= parameter used on the
867 If the device connected to the port is not a TTY but a braille
868 device, prepend "brl," before the device type, for instance
870 For now, only VisioBraille is supported.
873 [KNL] Change console messages format
875 By default we print messages on consoles in
876 "[time stamp] text\n" format (time stamp may not be
877 printed, depending on CONFIG_PRINTK_TIME or
878 `printk_time' param).
880 Switch to syslog format: "<%u>[time stamp] text\n"
881 IOW, each message will have a facility and loglevel
882 prefix. The format is similar to one used by syslog()
883 syscall, or to executing "dmesg -S --raw" or to reading
886 consoleblank= [KNL] The console blank (screen saver) timeout in
887 seconds. A value of 0 disables the blank timer.
891 [KNL] Change the default value for
892 /proc/<pid>/coredump_filter.
893 See also Documentation/filesystems/proc.rst.
895 coresight_cpu_debug.enable
898 Enable/disable the CPU sampling based debugging.
899 0: default value, disable debugging
900 1: enable debugging at boot time
902 cpcihp_generic= [HW,PCI] Generic port I/O CompactPCI driver
904 <first_slot>,<last_slot>,<port>,<enum_bit>[,<debug>]
906 cpuidle.off=1 [CPU_IDLE]
907 disable the cpuidle sub-system
910 [CPU_IDLE] Name of the cpuidle governor to use.
912 cpufreq.off=1 [CPU_FREQ]
913 disable the cpufreq sub-system
915 cpufreq.default_governor=
916 [CPU_FREQ] Name of the default cpufreq governor or
917 policy to use. This governor must be registered in the
918 kernel before the cpufreq driver probes.
921 [X86,EARLY] Delay for N microsec between assert and de-assert
922 of APIC INIT to start processors. This delay occurs
923 on every CPU online, such as boot, and resume from suspend.
927 [SMP] Enable/disable parallel bringup of secondary CPUs
929 Default is enabled if CONFIG_HOTPLUG_PARALLEL=y. Otherwise
930 the parameter has no effect.
932 crash_kexec_post_notifiers
933 Only jump to kdump kernel after running the panic
934 notifiers and dumping kmsg. This option increases
935 the risks of a kdump failure, since some panic
936 notifiers can make the crashed kernel more unstable.
937 In configurations where kdump may not be reliable,
938 running the panic notifiers could allow collecting
939 more data on dmesg, like stack traces from other CPUS
940 or extra data dumped by panic_print. Note that some
941 configurations enable this option unconditionally,
942 like Hyper-V, PowerPC (fadump) and AMD SEV-SNP.
944 crashkernel=size[KMG][@offset[KMG]]
945 [KNL,EARLY] Using kexec, Linux can switch to a 'crash kernel'
946 upon panic. This parameter reserves the physical
947 memory region [offset, offset + size] for that kernel
948 image. If '@offset' is omitted, then a suitable offset
949 is selected automatically.
950 [KNL, X86-64, ARM64, RISCV, LoongArch] Select a region
951 under 4G first, and fall back to reserve region above
952 4G when '@offset' hasn't been specified.
953 See Documentation/admin-guide/kdump/kdump.rst for further details.
955 crashkernel=range1:size1[,range2:size2,...][@offset]
956 [KNL] Same as above, but depends on the memory
957 in the running system. The syntax of range is
958 start-[end] where start and end are both
959 a memory unit (amount[KMG]). See also
960 Documentation/admin-guide/kdump/kdump.rst for an example.
962 crashkernel=size[KMG],high
963 [KNL, X86-64, ARM64, RISCV, LoongArch] range could be
965 Allow kernel to allocate physical memory region from top,
966 so could be above 4G if system have more than 4G ram
967 installed. Otherwise memory region will be allocated
968 below 4G, if available.
969 It will be ignored if crashkernel=X is specified.
970 crashkernel=size[KMG],low
971 [KNL, X86-64, ARM64, RISCV, LoongArch] range under 4G.
972 When crashkernel=X,high is passed, kernel could allocate
973 physical memory region above 4G, that cause second kernel
974 crash on system that require some amount of low memory,
975 e.g. swiotlb requires at least 64M+32K low memory, also
976 enough extra low memory is needed to make sure DMA buffers
977 for 32-bit devices won't run out. Kernel would try to allocate
978 default size of memory below 4G automatically. The default
979 size is platform dependent.
980 --> x86: max(swiotlb_size_or_default() + 8MiB, 256MiB)
983 --> loongarch: 128MiB
984 This one lets the user specify own low range under 4G
985 for second kernel instead.
986 0: to disable low allocation.
987 It will be ignored when crashkernel=X,high is not used
988 or memory reserved is below 4G.
991 [KNL] Disable crypto self-tests
996 cs89x0_media= [HW,NET]
997 Format: { rj45 | aui | bnc }
999 csdlock_debug= [KNL] Enable or disable debug add-ons of cross-CPU
1000 function call handling. When switched on,
1001 additional debug data is printed to the console
1002 in case a hanging CPU is detected, and that
1003 CPU is pinged again in order to try to resolve
1004 the hang situation. The default value of this
1005 option depends on the CSD_LOCK_WAIT_DEBUG_DEFAULT
1009 See header of drivers/s390/block/dasd_devmap.c.
1011 db9.dev[2|3]= [HW,JOY] Multisystem joystick support via parallel port
1012 (one device per port)
1013 Format: <port#>,<type>
1014 See also Documentation/input/devices/joystick-parport.rst
1016 debug [KNL,EARLY] Enable kernel debugging (events log level).
1018 debug_boot_weak_hash
1019 [KNL,EARLY] Enable printing [hashed] pointers early in the
1020 boot sequence. If enabled, we use a weak hash instead
1021 of siphash to hash pointers. Use this option if you are
1022 seeing instances of '(___ptrval___)') and need to see a
1023 value (hashed pointer) instead. Cryptographically
1024 insecure, please do not use on production kernels.
1026 debug_locks_verbose=
1027 [KNL] verbose locking self-tests
1029 Print debugging info while doing the locking API
1031 Bitmask for the various LOCKTYPE_ tests. Defaults to 0
1032 (no extra messages), setting it to -1 (all bits set)
1033 will print _a_lot_ more information - normally only
1034 useful to lockdep developers.
1036 debug_objects [KNL,EARLY] Enable object debugging
1038 debug_guardpage_minorder=
1039 [KNL,EARLY] When CONFIG_DEBUG_PAGEALLOC is set, this
1040 parameter allows control of the order of pages that will
1041 be intentionally kept free (and hence protected) by the
1042 buddy allocator. Bigger value increase the probability
1043 of catching random memory corruption, but reduce the
1044 amount of memory for normal system use. The maximum
1045 possible value is MAX_PAGE_ORDER/2. Setting this
1046 parameter to 1 or 2 should be enough to identify most
1047 random memory corruption problems caused by bugs in
1048 kernel or driver code when a CPU writes to (or reads
1049 from) a random memory location. Note that there exists
1050 a class of memory corruptions problems caused by buggy
1051 H/W or F/W or by drivers badly programming DMA
1052 (basically when memory is written at bus level and the
1053 CPU MMU is bypassed) which are not detectable by
1054 CONFIG_DEBUG_PAGEALLOC, hence this option will not
1055 help tracking down these problems.
1058 [KNL,EARLY] When CONFIG_DEBUG_PAGEALLOC is set, this parameter
1059 enables the feature at boot time. By default, it is
1060 disabled and the system will work mostly the same as a
1061 kernel built without CONFIG_DEBUG_PAGEALLOC.
1062 Note: to get most of debug_pagealloc error reports, it's
1063 useful to also enable the page_owner functionality.
1064 on: enable the feature
1066 debugfs= [KNL,EARLY] This parameter enables what is exposed to
1067 userspace and debugfs internal clients.
1068 Format: { on, no-mount, off }
1069 on: All functions are enabled.
1071 Filesystem is not registered but kernel clients can
1072 access APIs and a crashkernel can be used to read
1073 its content. There is nothing to mount.
1074 off: Filesystem is not registered and clients
1075 get a -EPERM as result when trying to register files
1076 or directories within debugfs.
1077 This is equivalent of the runtime functionality if
1078 debugfs was not enabled in the kernel at all.
1079 Default value is set in build-time with a kernel configuration.
1081 debugpat [X86] Enable PAT debugging
1084 [HW] The size of the default HugeTLB page. This is
1085 the size represented by the legacy /proc/ hugepages
1086 APIs. In addition, this is the default hugetlb size
1087 used for shmget(), mmap() and mounting hugetlbfs
1088 filesystems. If not specified, defaults to the
1089 architecture's default huge page size. Huge page
1090 sizes are architecture dependent. See also
1091 Documentation/admin-guide/mm/hugetlbpage.rst.
1094 deferred_probe_timeout=
1095 [KNL] Debugging option to set a timeout in seconds for
1096 deferred probe to give up waiting on dependencies to
1097 probe. Only specific dependencies (subsystems or
1098 drivers) that have opted in will be ignored. A timeout
1099 of 0 will timeout at the end of initcalls. If the time
1100 out hasn't expired, it'll be restarted by each
1101 successful driver registration. This option will also
1102 dump out devices still on the deferred probe list after
1105 delayacct [KNL] Enable per-task delay accounting
1107 dell_smm_hwmon.ignore_dmi=
1108 [HW] Continue probing hardware even if DMI data
1109 indicates that the driver is running on unsupported
1112 dell_smm_hwmon.force=
1113 [HW] Activate driver even if SMM BIOS signature does
1114 not match list of supported models and enable otherwise
1115 blacklisted features.
1117 dell_smm_hwmon.power_status=
1118 [HW] Report power status in /proc/i8k
1119 (disabled by default).
1121 dell_smm_hwmon.restricted=
1122 [HW] Allow controlling fans only if SYS_ADMIN
1125 dell_smm_hwmon.fan_mult=
1126 [HW] Factor to multiply fan speed with.
1128 dell_smm_hwmon.fan_max=
1129 [HW] Maximum configurable fan speed.
1132 Format: { on | off | def_only | inf_only | always }
1133 on: s390 zlib hardware support for compression on
1134 level 1 and decompression (default)
1135 off: No s390 zlib hardware support
1136 def_only: s390 zlib hardware support for deflate
1137 only (compression on level 1)
1138 inf_only: s390 zlib hardware support for inflate
1139 only (decompression)
1140 always: Same as 'on' but ignores the selected compression
1141 level always using hardware support (used for debugging)
1143 dhash_entries= [KNL]
1144 Set number of hash buckets for dentry cache.
1146 disable_1tb_segments [PPC,EARLY]
1147 Disables the use of 1TB hash page table segments. This
1148 causes the kernel to fall back to 256MB segments which
1149 can be useful when debugging issues that require an SLB
1153 See Documentation/networking/ipv6.rst.
1155 disable_radix [PPC,EARLY]
1156 Disable RADIX MMU mode on POWER9
1159 Disable TLBIE instruction. Currently does not work
1160 with KVM, with HASH MMU, or with coherent accelerators.
1162 disable_ddw [PPC/PSERIES,EARLY]
1163 Disable Dynamic DMA Window support. Use this
1164 to workaround buggy firmware.
1166 disable_ipv6= [IPV6]
1167 See Documentation/networking/ipv6.rst.
1169 disable_mtrr_cleanup [X86,EARLY]
1170 The kernel tries to adjust MTRR layout from continuous
1171 to discrete, to make X server driver able to add WB
1172 entry later. This parameter disables that.
1174 disable_mtrr_trim [X86, Intel and AMD only,EARLY]
1175 By default the kernel will trim any uncacheable
1176 memory out of your available memory pool based on
1177 MTRR settings. This parameter disables that behavior,
1178 possibly causing your machine to run very slowly.
1180 disable_timer_pin_1 [X86,EARLY]
1181 Disable PIN 1 of APIC timer
1182 Can be useful to work around chipset bugs.
1184 dis_ucode_ldr [X86] Disable the microcode loader.
1186 dma_debug=off If the kernel is compiled with DMA_API_DEBUG support,
1187 this option disables the debugging code at boot.
1189 dma_debug_entries=<number>
1190 This option allows to tune the number of preallocated
1191 entries for DMA-API debugging code. One entry is
1192 required per DMA-API allocation. Use this if the
1193 DMA-API debugging code disables itself because the
1194 architectural default is too low.
1196 dma_debug_driver=<driver_name>
1197 With this option the DMA-API debugging driver
1198 filter feature can be enabled at boot time. Just
1199 pass the driver to filter for as the parameter.
1200 The filter can be disabled or changed to another
1201 driver later using sysfs.
1203 reg_file_data_sampling=
1204 [X86] Controls mitigation for Register File Data
1205 Sampling (RFDS) vulnerability. RFDS is a CPU
1206 vulnerability which may allow userspace to infer
1207 kernel data values previously stored in floating point
1208 registers, vector registers, or integer registers.
1209 RFDS only affects Intel Atom processors.
1211 on: Turns ON the mitigation.
1212 off: Turns OFF the mitigation.
1214 This parameter overrides the compile time default set
1215 by CONFIG_MITIGATION_RFDS. Mitigation cannot be
1216 disabled when other VERW based mitigations (like MDS)
1217 are enabled. In order to disable RFDS mitigation all
1218 VERW based mitigations need to be disabled.
1221 Documentation/admin-guide/hw-vuln/reg-file-data-sampling.rst
1223 driver_async_probe= [KNL]
1224 List of driver names to be probed asynchronously. *
1225 matches with all driver names. If * is specified, the
1226 rest of the listed driver names are those that will NOT
1228 Format: <driver_name1>,<driver_name2>...
1230 drm.edid_firmware=[<connector>:]<file>[,[<connector>:]<file>]
1231 Broken monitors, graphic adapters, KVMs and EDIDless
1232 panels may send no or incorrect EDID data sets.
1233 This parameter allows to specify an EDID data sets
1234 in the /lib/firmware directory that are used instead.
1235 An EDID data set will only be used for a particular
1236 connector, if its name and a colon are prepended to
1237 the EDID name. Each connector may use a unique EDID
1238 data set by separating the files with a comma. An EDID
1239 data set with no connector name will be used for
1240 any connectors not explicitly specified.
1244 dt_cpu_ftrs= [PPC,EARLY]
1245 Format: {"off" | "known"}
1246 Control how the dt_cpu_ftrs device-tree binding is
1247 used for CPU feature discovery and setup (if it
1249 off: Do not use it, fall back to legacy cpu table.
1250 known: Do not pass through unknown features to guests
1251 or userspace, only those that the kernel is aware of.
1253 dump_apple_properties [X86]
1254 Dump name and content of EFI device properties on
1255 x86 Macs. Useful for driver authors to determine
1256 what data is available or for reverse-engineering.
1258 dyndbg[="val"] [KNL,DYNAMIC_DEBUG]
1259 <module>.dyndbg[="val"]
1260 Enable debug messages at boot time. See
1261 Documentation/admin-guide/dynamic-debug-howto.rst
1264 early_ioremap_debug [KNL,EARLY]
1265 Enable debug messages in early_ioremap support. This
1266 is useful for tracking down temporary early mappings
1267 which are not unmapped.
1269 earlycon= [KNL,EARLY] Output early console device and options.
1271 When used with no options, the early console is
1272 determined by stdout-path property in device tree's
1273 chosen node or the ACPI SPCR table if supported by
1276 cdns,<addr>[,options]
1277 Start an early, polled-mode console on a Cadence
1278 (xuartps) serial port at the specified address. Only
1279 supported option is baud rate. If baud rate is not
1280 specified, the serial port must already be setup and
1283 uart[8250],io,<addr>[,options[,uartclk]]
1284 uart[8250],mmio,<addr>[,options[,uartclk]]
1285 uart[8250],mmio32,<addr>[,options[,uartclk]]
1286 uart[8250],mmio32be,<addr>[,options[,uartclk]]
1287 uart[8250],0x<addr>[,options]
1288 Start an early, polled-mode console on the 8250/16550
1289 UART at the specified I/O port or MMIO address.
1290 MMIO inter-register address stride is either 8-bit
1291 (mmio) or 32-bit (mmio32 or mmio32be).
1292 If none of [io|mmio|mmio32|mmio32be], <addr> is assumed
1293 to be equivalent to 'mmio'. 'options' are specified
1294 in the same format described for "console=ttyS<n>"; if
1295 unspecified, the h/w is not initialized. 'uartclk' is
1296 the uart clock frequency; if unspecified, it is set
1297 to 'BASE_BAUD' * 16.
1301 Start an early, polled-mode console on a pl011 serial
1302 port at the specified address. The pl011 serial port
1303 must already be setup and configured. Options are not
1304 yet supported. If 'mmio32' is specified, then only
1305 the driver will use only 32-bit accessors to read/write
1306 the device registers.
1309 Start an early console on a litex serial port at the
1310 specified address. The serial port must already be
1311 setup and configured. Options are not yet supported.
1314 Start an early, polled-mode console on a meson serial
1315 port at the specified address. The serial port must
1316 already be setup and configured. Options are not yet
1320 Start an early, polled-mode console on an msm serial
1321 port at the specified address. The serial port
1322 must already be setup and configured. Options are not
1325 msm_serial_dm,<addr>
1326 Start an early, polled-mode console on an msm serial
1327 dm port at the specified address. The serial port
1328 must already be setup and configured. Options are not
1332 Start an early, polled-mode console on a serial port
1333 of an Actions Semi SoC, such as S500 or S900, at the
1334 specified address. The serial port must already be
1335 setup and configured. Options are not yet supported.
1338 Start an early, polled-mode console on a serial port
1339 of an RDA Micro SoC, such as RDA8810PL, at the
1340 specified address. The serial port must already be
1341 setup and configured. Options are not yet supported.
1344 Use RISC-V SBI (Supervisor Binary Interface) for early
1347 smh Use ARM semihosting calls for early console.
1355 Use early console provided by serial driver available
1356 on Samsung SoCs, requires selecting proper type and
1357 a correct base address of the selected UART port. The
1358 serial port must already be setup and configured.
1359 Options are not yet supported.
1362 Start an early, polled-mode console on a lantiq serial
1363 (lqasc) port at the specified address. The serial port
1364 must already be setup and configured. Options are not
1369 Use early console provided by Freescale LP UART driver
1370 found on Freescale Vybrid and QorIQ LS1021A processors.
1371 A valid base address must be provided, and the serial
1372 port must already be setup and configured.
1376 Start an early, polled-mode, output-only console on the
1377 Freescale i.MX UART at the specified address. The UART
1378 must already be setup and configured.
1381 Start an early, polled-mode console on the
1382 Armada 3700 serial port at the specified
1383 address. The serial port must already be setup
1384 and configured. Options are not yet supported.
1387 Start an early, polled-mode console on a Qualcomm
1388 Generic Interface (GENI) based serial port at the
1389 specified address. The serial port must already be
1390 setup and configured. Options are not yet supported.
1393 Start an early, unaccelerated console on the EFI
1394 memory mapped framebuffer (if available). On cache
1395 coherent non-x86 systems that use system memory for
1396 the framebuffer, pass the 'ram' option so that it is
1397 mapped with the correct attributes.
1400 Use early console provided by Freescale LINFlexD UART
1401 serial driver for NXP S32V234 SoCs. A valid base
1402 address must be provided, and the serial port must
1403 already be setup and configured.
1405 earlyprintk= [X86,SH,ARM,M68k,S390,UM,EARLY]
1409 earlyprintk=serial[,ttySn[,baudrate]]
1410 earlyprintk=serial[,0x...[,baudrate]]
1411 earlyprintk=ttySn[,baudrate]
1412 earlyprintk=dbgp[debugController#]
1413 earlyprintk=mmio32,membase[,{nocfg|baudrate}]
1414 earlyprintk=pciserial[,force],bus:device.function[,{nocfg|baudrate}]
1415 earlyprintk=xdbc[xhciController#]
1418 earlyprintk is useful when the kernel crashes before
1419 the normal console is initialized. It is not enabled by
1420 default because it has some cosmetic problems.
1422 Use "nocfg" to skip UART configuration, assume
1423 BIOS/firmware has configured UART correctly.
1425 Append ",keep" to not disable it when the real console
1428 Only one of vga, serial, or usb debug port can
1431 Currently only ttyS0 and ttyS1 may be specified by
1432 name. Other I/O ports may be explicitly specified
1433 on some architectures (x86 and arm at least) by
1434 replacing ttySn with an I/O port address, like this:
1435 earlyprintk=serial,0x1008,115200
1436 You can find the port for a given device in
1437 /proc/tty/driver/serial:
1438 2: uart:ST16650V2 port:00001008 irq:18 ...
1440 Interaction with the standard serial driver is not
1443 The VGA output is eventually overwritten by
1446 The xen option can only be used in Xen domains.
1448 The sclp output can only be used on s390.
1450 The bios output can only be used on SuperH.
1452 The optional "force" to "pciserial" enables use of a
1453 PCI device even when its classcode is not of the
1456 edac_report= [HW,EDAC] Control how to report EDAC event
1457 Format: {"on" | "off" | "force"}
1458 on: enable EDAC to report H/W event. May be overridden
1459 by other higher priority error reporting module.
1460 off: disable H/W event reporting through EDAC.
1461 force: enforce the use of EDAC to report H/W event.
1465 Format: {"off" | "on" | "skip[mbr]"}
1468 Format: { "debug", "disable_early_pci_dma",
1469 "nochunk", "noruntime", "nosoftreserve",
1470 "novamap", "no_disable_early_pci_dma" }
1471 debug: enable misc debug output.
1472 disable_early_pci_dma: disable the busmaster bit on all
1473 PCI bridges while in the EFI boot stub.
1474 nochunk: disable reading files in "chunks" in the EFI
1475 boot stub, as chunking can cause problems with some
1476 firmware implementations.
1477 noruntime : disable EFI runtime services support
1478 nosoftreserve: The EFI_MEMORY_SP (Specific Purpose)
1479 attribute may cause the kernel to reserve the
1480 memory range for a memory mapping driver to
1481 claim. Specify efi=nosoftreserve to disable this
1482 reservation and treat the memory by its base type
1483 (i.e. EFI_CONVENTIONAL_MEMORY / "System RAM").
1484 novamap: do not call SetVirtualAddressMap().
1485 no_disable_early_pci_dma: Leave the busmaster bit set
1486 on all PCI bridges while in the EFI boot stub
1488 efi_no_storage_paranoia [EFI,X86,EARLY]
1489 Using this parameter you can use more than 50% of
1490 your efi variable storage. Use this parameter only if
1491 you are really sure that your UEFI does sane gc and
1492 fulfills the spec otherwise your board may brick.
1494 efivar_ssdt= [EFI; X86] Name of an EFI variable that contains an SSDT
1495 that is to be dynamically loaded by Linux. If there are
1496 multiple variables with the same name but with different
1497 vendor GUIDs, all of them will be loaded. See
1498 Documentation/admin-guide/acpi/ssdt-overlays.rst for details.
1501 eisa_irq_edge= [PARISC,HW]
1502 See header of drivers/parisc/eisa.c.
1504 ekgdboc= [X86,KGDB,EARLY] Allow early kernel console debugging
1507 This is designed to be used in conjunction with
1508 the boot argument: earlyprintk=vga
1510 This parameter works in place of the kgdboc parameter
1511 but can only be used if the backing tty is available
1512 very early in the boot process. For early debugging
1513 via a serial port see kgdboc_earlycon instead.
1516 See comment before function elanfreq_setup() in
1517 arch/x86/kernel/cpu/cpufreq/elanfreq.c.
1519 elfcorehdr=[size[KMG]@]offset[KMG] [PPC,SH,X86,S390,EARLY]
1520 Specifies physical address of start of kernel core
1521 image elf header and optionally the size. Generally
1522 kexec loader will pass this option to capture kernel.
1523 See Documentation/admin-guide/kdump/kdump.rst for details.
1525 enable_mtrr_cleanup [X86,EARLY]
1526 The kernel tries to adjust MTRR layout from continuous
1527 to discrete, to make X server driver able to add WB
1528 entry later. This parameter enables that.
1530 enable_timer_pin_1 [X86]
1531 Enable PIN 1 of APIC timer
1532 Can be useful to work around chipset bugs
1533 (in particular on some ATI chipsets).
1534 The kernel tries to set a reasonable default.
1536 enforcing= [SELINUX] Set initial enforcing status.
1538 See security/selinux/Kconfig help text.
1539 0 -- permissive (log only, no denials).
1540 1 -- enforcing (deny and log).
1542 Value can be changed at runtime via
1543 /sys/fs/selinux/enforce.
1546 Disable Error Record Serialization Table (ERST)
1549 ether= [HW,NET] Ethernet cards parameters
1550 This option is obsoleted by the "netdev=" option, which
1551 has equivalent usage. See its documentation for details.
1555 Permit 'security.evm' to be updated regardless of
1556 current integrity status.
1558 early_page_ext [KNL,EARLY] Enforces page_ext initialization to earlier
1559 stages so cover more early boot allocations.
1560 Please note that as side effect some optimizations
1561 might be disabled to achieve that (e.g. parallelized
1562 memory initialization is disabled) so the boot process
1563 might take longer, especially on systems with a lot of
1564 memory. Available with CONFIG_PAGE_EXTENSION=y.
1570 fail_make_request=[KNL]
1571 General fault injection mechanism.
1572 Format: <interval>,<probability>,<space>,<times>
1573 See also Documentation/fault-injection/.
1576 Format: { initns | none }
1577 See Documentation/admin-guide/sysctl/net.rst for
1578 fb_tunnels_only_for_init_ns
1581 See Documentation/admin-guide/blockdev/floppy.rst.
1584 Forcefully enable Physical Address Extension (PAE).
1585 Many Pentium M systems disable PAE but may have a
1586 functionally usable PAE implementation.
1587 Warning: use of this parameter will taint the kernel
1588 and may cause unknown problems.
1591 Enable/disable Flexible Return and Event Delivery.
1592 Format: { on | off }
1593 on: enable FRED when it's present.
1594 off: disable FRED, the default setting.
1597 [FTRACE] will set and start the specified tracer
1598 as early as possible in order to facilitate early
1601 ftrace_boot_snapshot
1602 [FTRACE] On boot up, a snapshot will be taken of the
1603 ftrace ring buffer that can be read at:
1604 /sys/kernel/tracing/snapshot.
1605 This is useful if you need tracing information from kernel
1606 boot up that is likely to be overridden by user space
1607 start up functionality.
1609 Optionally, the snapshot can also be defined for a tracing
1610 instance that was created by the trace_instance= command
1613 trace_instance=foo,sched_switch ftrace_boot_snapshot=foo
1615 The above will cause the "foo" tracing instance to trigger
1616 a snapshot at the end of boot up.
1618 ftrace_dump_on_oops[=2(orig_cpu) | =<instance>][,<instance> |
1619 ,<instance>=2(orig_cpu)]
1620 [FTRACE] will dump the trace buffers on oops.
1621 If no parameter is passed, ftrace will dump global
1622 buffers of all CPUs, if you pass 2 or orig_cpu, it
1623 will dump only the buffer of the CPU that triggered
1624 the oops, or the specific instance will be dumped if
1625 its name is passed. Multiple instance dump is also
1626 supported, and instances are separated by commas. Each
1627 instance supports only dump on CPU that triggered the
1628 oops by passing 2 or orig_cpu to it.
1630 ftrace_dump_on_oops=foo=orig_cpu
1632 The above will dump only the buffer of "foo" instance
1633 on CPU that triggered the oops.
1635 ftrace_dump_on_oops,foo,bar=orig_cpu
1637 The above will dump global buffer on all CPUs, the
1638 buffer of "foo" instance on all CPUs and the buffer
1639 of "bar" instance on CPU that triggered the oops.
1641 ftrace_filter=[function-list]
1642 [FTRACE] Limit the functions traced by the function
1643 tracer at boot up. function-list is a comma-separated
1644 list of functions. This list can be changed at run
1645 time by the set_ftrace_filter file in the debugfs
1648 ftrace_notrace=[function-list]
1649 [FTRACE] Do not trace the functions specified in
1650 function-list. This list can be changed at run time
1651 by the set_ftrace_notrace file in the debugfs
1654 ftrace_graph_filter=[function-list]
1655 [FTRACE] Limit the top level callers functions traced
1656 by the function graph tracer at boot up.
1657 function-list is a comma-separated list of functions
1658 that can be changed at run time by the
1659 set_graph_function file in the debugfs tracing directory.
1661 ftrace_graph_notrace=[function-list]
1662 [FTRACE] Do not trace from the functions specified in
1663 function-list. This list is a comma-separated list of
1664 functions that can be changed at run time by the
1665 set_graph_notrace file in the debugfs tracing directory.
1667 ftrace_graph_max_depth=<uint>
1668 [FTRACE] Used with the function graph tracer. This is
1669 the max depth it will trace into a function. This value
1670 can be changed at run time by the max_graph_depth file
1671 in the tracefs tracing directory. default: 0 (no limit)
1673 fw_devlink= [KNL,EARLY] Create device links between consumer and supplier
1674 devices by scanning the firmware to infer the
1675 consumer/supplier relationships. This feature is
1676 especially useful when drivers are loaded as modules as
1677 it ensures proper ordering of tasks like device probing
1678 (suppliers first, then consumers), supplier boot state
1679 clean up (only after all consumers have probed),
1680 suspend/resume & runtime PM (consumers first, then
1682 Format: { off | permissive | on | rpm }
1683 off -- Don't create device links from firmware info.
1684 permissive -- Create device links from firmware info
1685 but use it only for ordering boot state clean
1686 up (sync_state() calls).
1687 on -- Create device links from firmware info and use it
1688 to enforce probe and suspend/resume ordering.
1689 rpm -- Like "on", but also use to order runtime PM.
1691 fw_devlink.strict=<bool>
1692 [KNL,EARLY] Treat all inferred dependencies as mandatory
1693 dependencies. This only applies for fw_devlink=on|rpm.
1696 fw_devlink.sync_state =
1697 [KNL,EARLY] When all devices that could probe have finished
1698 probing, this parameter controls what to do with
1699 devices that haven't yet received their sync_state()
1701 Format: { strict | timeout }
1702 strict -- Default. Continue waiting on consumers to
1704 timeout -- Give up waiting on consumers and call
1705 sync_state() on any devices that haven't yet
1706 received their sync_state() calls after
1707 deferred_probe_timeout has expired or by
1708 late_initcall() if !CONFIG_MODULES.
1711 [HW,JOY] Multisystem joystick and NES/SNES/PSX pad
1712 support via parallel port (up to 5 devices per port)
1713 Format: <port#>,<pad1>,<pad2>,<pad3>,<pad4>,<pad5>
1714 See also Documentation/input/devices/joystick-parport.rst
1718 gart_fix_e820= [X86-64,EARLY] disable the fix e820 for K8 GART
1722 gather_data_sampling=
1723 [X86,INTEL,EARLY] Control the Gather Data Sampling (GDS)
1726 Gather Data Sampling is a hardware vulnerability which
1727 allows unprivileged speculative access to data which was
1728 previously stored in vector registers.
1730 This issue is mitigated by default in updated microcode.
1731 The mitigation may have a performance impact but can be
1732 disabled. On systems without the microcode mitigation
1733 disabling AVX serves as a mitigation.
1735 force: Disable AVX to mitigate systems without
1736 microcode mitigation. No effect if the microcode
1737 mitigation is present. Known to cause crashes in
1738 userspace with buggy AVX enumeration.
1740 off: Disable GDS mitigation.
1742 gbpages [X86] Use GB pages for kernel direct mappings.
1744 gcov_persist= [GCOV] When non-zero (default), profiling data for
1745 kernel modules is saved and remains accessible via
1746 debugfs, even when the module is unloaded/reloaded.
1747 When zero, profiling data is discarded and associated
1748 debugfs files are removed at module unload time.
1750 goldfish [X86] Enable the goldfish android emulator platform.
1751 Don't use this when you are not running on the
1754 gpio-mockup.gpio_mockup_ranges
1755 [HW] Sets the ranges of gpiochip of for this device.
1756 Format: <start1>,<end1>,<start2>,<end2>...
1757 gpio-mockup.gpio_mockup_named_lines
1758 [HW] Let the driver know GPIO lines should be named.
1760 gpt [EFI] Forces disk with valid GPT signature but
1761 invalid Protective MBR to be treated as GPT. If the
1762 primary GPT is corrupted, it enables the backup/alternate
1763 GPT to be used instead.
1765 grcan.enable0= [HW] Configuration of physical interface 0. Determines
1766 the "Enable 0" bit of the configuration register.
1769 grcan.enable1= [HW] Configuration of physical interface 1. Determines
1770 the "Enable 0" bit of the configuration register.
1773 grcan.select= [HW] Select which physical interface to use.
1776 grcan.txsize= [HW] Sets the size of the tx buffer.
1777 Format: <unsigned int> such that (txsize & ~0x1fffc0) == 0.
1779 grcan.rxsize= [HW] Sets the size of the rx buffer.
1780 Format: <unsigned int> such that (rxsize & ~0x1fffc0) == 0.
1784 [KNL] Under CONFIG_HARDENED_USERCOPY, whether
1785 hardening is enabled for this boot. Hardened
1786 usercopy checking is used to protect the kernel
1787 from reading or writing beyond known memory
1788 allocation boundaries as a proactive defense
1789 against bounds-checking flaws in the kernel's
1790 copy_to_user()/copy_from_user() interface.
1791 The default is determined by
1792 CONFIG_HARDENED_USERCOPY_DEFAULT_ON.
1793 on Perform hardened usercopy checks.
1794 off Disable hardened usercopy checks.
1796 hardlockup_all_cpu_backtrace=
1797 [KNL] Should the hard-lockup detector generate
1798 backtraces on all cpus.
1801 hashdist= [KNL,NUMA] Large hashes allocated during boot
1802 are distributed across NUMA nodes. Defaults on
1803 for 64-bit NUMA, off otherwise.
1804 Format: 0 | 1 (for off | on)
1806 hd= [EIDE] (E)IDE hard drive subsystem geometry
1807 Format: <cyl>,<head>,<sect>
1810 Disable Hardware Error Source Table (HEST) support;
1811 corresponding firmware-first mode error processing
1812 logic will be disabled.
1814 hibernate= [HIBERNATION]
1815 noresume Don't check if there's a hibernation image
1816 present during boot.
1817 nocompress Don't compress/decompress hibernation images.
1818 no Disable hibernation and resume.
1819 protect_image Turn on image protection during restoration
1820 (that will set all pages holding image data
1821 during restoration read-only).
1823 hibernate.compressor= [HIBERNATION] Compression algorithm to be
1824 used with hibernation.
1825 Format: { lzo | lz4 }
1828 lzo: Select LZO compression algorithm to
1829 compress/decompress hibernation image.
1831 lz4: Select LZ4 compression algorithm to
1832 compress/decompress hibernation image.
1834 hibernate.pm_test_delay=
1836 Sets the number of seconds to remain in a hibernation test
1837 mode before resuming the system (see
1838 /sys/power/pm_test). Only available when CONFIG_PM_DEBUG
1839 is set. Default value is 5.
1841 highmem=nn[KMG] [KNL,BOOT,EARLY] forces the highmem zone to have an exact
1842 size of <nn>. This works even on boxes that have no
1843 highmem otherwise. This also works to reduce highmem
1844 size on bigger boxes.
1846 highres= [KNL] Enable/disable high resolution timer mode.
1847 Valid parameters: "on", "off"
1852 hostname= [KNL,EARLY] Set the hostname (aka UTS nodename).
1854 This allows setting the system's hostname during early
1855 startup. This sets the name returned by gethostname.
1856 Using this parameter to set the hostname makes it
1857 possible to ensure the hostname is correctly set before
1858 any userspace processes run, avoiding the possibility
1859 that a process may call gethostname before the hostname
1860 has been explicitly set, resulting in the calling
1861 process getting an incorrect result. The string must
1862 not exceed the maximum allowed hostname length (usually
1863 64 characters) and will be truncated otherwise.
1865 hpet= [X86-32,HPET] option to control HPET usage
1866 Format: { enable (default) | disable | force |
1868 disable: disable HPET and use PIT instead
1869 force: allow force enabled of undocumented chips (ICH4,
1871 verbose: show contents of HPET registers during setup
1873 hpet_mmap= [X86, HPET_MMAP] Allow userspace to mmap HPET
1874 registers. Default set by CONFIG_HPET_MMAP_DEFAULT.
1876 hugepages= [HW,EARLY] Number of HugeTLB pages to allocate at boot.
1877 If this follows hugepagesz (below), it specifies
1878 the number of pages of hugepagesz to be allocated.
1879 If this is the first HugeTLB parameter on the command
1880 line, it specifies the number of pages to allocate for
1881 the default huge page size. If using node format, the
1882 number of pages to allocate per-node can be specified.
1883 See also Documentation/admin-guide/mm/hugetlbpage.rst.
1884 Format: <integer> or (node format)
1885 <node>:<integer>[,<node>:<integer>]
1888 [HW,EARLY] The size of the HugeTLB pages. This is
1889 used in conjunction with hugepages (above) to
1890 allocate huge pages of a specific size at boot. The
1891 pair hugepagesz=X hugepages=Y can be specified once
1892 for each supported huge page size. Huge page sizes
1893 are architecture dependent. See also
1894 Documentation/admin-guide/mm/hugetlbpage.rst.
1897 hugepage_alloc_threads=
1898 [HW] The number of threads that should be used to
1899 allocate hugepages during boot. This option can be
1900 used to improve system bootup time when allocating
1901 a large amount of huge pages.
1902 The default value is 25% of the available hardware threads.
1904 Note that this parameter only applies to non-gigantic huge pages.
1906 hugetlb_cma= [HW,CMA,EARLY] The size of a CMA area used for allocation
1907 of gigantic hugepages. Or using node format, the size
1908 of a CMA area per node can be specified.
1909 Format: nn[KMGTPE] or (node format)
1910 <node>:nn[KMGTPE][,<node>:nn[KMGTPE]]
1912 Reserve a CMA area of given size and allocate gigantic
1913 hugepages using the CMA allocator. If enabled, the
1914 boot-time allocation of gigantic hugepages is skipped.
1917 [HW,CMA,EARLY] When allocating new HugeTLB pages, only
1918 try to allocate from the CMA areas.
1920 This option does nothing if hugetlb_cma= is not also
1923 hugetlb_free_vmemmap=
1924 [KNL] Requires CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP
1926 Control if HugeTLB Vmemmap Optimization (HVO) is enabled.
1927 Allows heavy hugetlb users to free up some more
1928 memory (7 * PAGE_SIZE for each 2MB hugetlb page).
1929 Format: { on | off (default) }
1934 Built with CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP_DEFAULT_ON=y,
1937 Note that the vmemmap pages may be allocated from the added
1938 memory block itself when memory_hotplug.memmap_on_memory is
1939 enabled, those vmemmap pages cannot be optimized even if this
1940 feature is enabled. Other vmemmap pages not allocated from
1941 the added memory block itself do not be affected.
1944 [KNL] Should the hung task detector generate panics.
1947 A value of 1 instructs the kernel to panic when a
1948 hung task is detected. The default value is controlled
1949 by the CONFIG_BOOTPARAM_HUNG_TASK_PANIC build-time
1950 option. The value selected by this boot parameter can
1951 be changed later by the kernel.hung_task_panic sysctl.
1953 hvc_iucv= [S390] Number of z/VM IUCV hypervisor console (HVC)
1954 terminal devices. Valid values: 0..8
1955 hvc_iucv_allow= [S390] Comma-separated list of z/VM user IDs.
1956 If specified, z/VM IUCV HVC accepts connections
1957 from listed z/VM user IDs only.
1959 hv_nopvspin [X86,HYPER_V,EARLY]
1960 Disables the paravirt spinlock optimizations
1961 which allow the hypervisor to 'idle' the guest
1965 Format: reboot | shutdown
1967 Hardware protection action taken on critical events like
1968 overtemperature or imminent voltage loss.
1970 i2c_bus= [HW] Override the default board specific I2C bus speed
1971 or register an additional I2C bus that is not
1972 registered from board initialization code.
1976 i2c_touchscreen_props= [HW,ACPI,X86]
1977 Set device-properties for ACPI-enumerated I2C-attached
1978 touchscreen, to e.g. fix coordinates of upside-down
1979 mounted touchscreens. If you need this option please
1980 submit a drivers/platform/x86/touchscreen_dmi.c patch
1981 adding a DMI quirk for this.
1984 <ACPI_HW_ID>:<prop_name>=<val>[:prop_name=val][:...]
1985 Where <val> is one of:
1986 Omit "=<val>" entirely Set a boolean device-property
1987 Unsigned number Set a u32 device-property
1988 Anything else Set a string device-property
1990 Examples (split over multiple lines):
1991 i2c_touchscreen_props=GDIX1001:touchscreen-inverted-x:
1992 touchscreen-inverted-y
1994 i2c_touchscreen_props=MSSL1680:touchscreen-size-x=1920:
1995 touchscreen-size-y=1080:touchscreen-inverted-y:
1996 firmware-name=gsl1680-vendor-model.fw:silead,home-button
1998 i8042.debug [HW] Toggle i8042 debug mode
1999 i8042.unmask_kbd_data
2000 [HW] Enable printing of interrupt data from the KBD port
2001 (disabled by default, and as a pre-condition
2002 requires that i8042.debug=1 be enabled)
2003 i8042.direct [HW] Put keyboard port into non-translated mode
2004 i8042.dumbkbd [HW] Pretend that controller can only read data from
2005 keyboard and cannot control its state
2006 (Don't attempt to blink the leds)
2007 i8042.noaux [HW] Don't check for auxiliary (== mouse) port
2008 i8042.nokbd [HW] Don't check/create keyboard port
2009 i8042.noloop [HW] Disable the AUX Loopback command while probing
2011 i8042.nomux [HW] Don't check presence of an active multiplexing
2013 i8042.nopnp [HW] Don't use ACPIPnP / PnPBIOS to discover KBD/AUX
2015 i8042.notimeout [HW] Ignore timeout condition signalled by controller
2016 i8042.reset [HW] Reset the controller during init, cleanup and
2017 suspend-to-ram transitions, only during s2r
2018 transitions, or never reset
2019 Format: { 1 | Y | y | 0 | N | n }
2020 1, Y, y: always reset controller
2021 0, N, n: don't ever reset controller
2022 Default: only on s2r transitions on x86; most other
2023 architectures force reset to be always executed
2024 i8042.unlock [HW] Unlock (ignore) the keylock
2025 i8042.kbdreset [HW] Reset device connected to KBD port
2027 [HW] Allow deferred probing upon i8042 probe errors
2031 i915.invert_brightness=
2032 [DRM] Invert the sense of the variable that is used to
2033 set the brightness of the panel backlight. Normally a
2034 brightness value of 0 indicates backlight switched off,
2035 and the maximum of the brightness value sets the backlight
2036 to maximum brightness. If this parameter is set to 0
2037 (default) and the machine requires it, or this parameter
2038 is set to 1, a brightness value of 0 sets the backlight
2039 to maximum brightness, and the maximum of the brightness
2040 value switches the backlight off.
2041 -1 -- never invert brightness
2042 0 -- machine default
2043 1 -- force brightness inversion
2045 ia32_emulation= [X86-64]
2047 When true, allows loading 32-bit programs and executing 32-bit
2048 syscalls, essentially overriding IA32_EMULATION_DEFAULT_DISABLED at
2049 boot time. When false, unconditionally disables IA32 emulation.
2052 Format: <io>[,<membase>[,<icn_id>[,<icn_id2>]]]
2056 Format: idle=poll, idle=halt, idle=nomwait
2058 idle=poll: Don't do power saving in the idle loop
2059 using HLT, but poll for rescheduling event. This will
2060 make the CPUs eat a lot more power, but may be useful
2061 to get slightly better performance in multiprocessor
2062 benchmarks. It also makes some profiling using
2063 performance counters more accurate. Please note that
2064 on systems with MONITOR/MWAIT support (like Intel
2065 EM64T CPUs) this option has no performance advantage
2066 over the normal idle loop. It may also interact badly
2067 with hyperthreading.
2069 idle=halt: Halt is forced to be used for CPU idle.
2070 In such case C2/C3 won't be used again.
2072 idle=nomwait: Disable mwait for CPU C-states
2076 Allow force disabling of Shared Virtual Memory (SVA)
2077 support for the idxd driver. By default it is set to
2080 idxd.tc_override= [HW]
2082 Allow override of default traffic class configuration
2083 for the device. By default it is set to false (0).
2085 ieee754= [MIPS] Select IEEE Std 754 conformance mode
2086 Format: { strict | legacy | 2008 | relaxed | emulated }
2089 Choose which programs will be accepted for execution
2090 based on the IEEE 754 NaN encoding(s) supported by
2091 the FPU and the NaN encoding requested with the value
2092 of an ELF file header flag individually set by each
2093 binary. Hardware implementations are permitted to
2094 support either or both of the legacy and the 2008 NaN
2097 Available settings are as follows:
2098 strict accept binaries that request a NaN encoding
2099 supported by the FPU
2100 legacy only accept legacy-NaN binaries, if supported
2102 2008 only accept 2008-NaN binaries, if supported
2104 relaxed accept any binaries regardless of whether
2105 supported by the FPU
2106 emulated accept any binaries but enable FPU emulator
2107 if binary mode is unsupported by the FPU.
2109 The FPU emulator is always able to support both NaN
2110 encodings, so if no FPU hardware is present or it has
2111 been disabled with 'nofpu', then the settings of
2112 'legacy' and '2008' strap the emulator accordingly,
2113 'relaxed' straps the emulator for both legacy-NaN and
2114 2008-NaN, whereas 'strict' enables legacy-NaN only on
2115 legacy processors and both NaN encodings on MIPS32 or
2118 The setting for ABS.fmt/NEG.fmt instruction execution
2119 mode generally follows that for the NaN encoding,
2120 except where unsupported by hardware.
2122 ignore_loglevel [KNL,EARLY]
2123 Ignore loglevel setting - this will print /all/
2124 kernel messages to the console. Useful for debugging.
2125 We also add it as printk module parameter, so users
2126 could change it dynamically, usually by
2127 /sys/module/printk/parameters/ignore_loglevel.
2130 Ignore RLIMIT_DATA setting for data mappings,
2131 print warning at first misuse. Can be changed via
2132 /sys/module/kernel/parameters/ignore_rlimit_data.
2134 ihash_entries= [KNL]
2135 Set number of hash buckets for inode cache.
2137 ima_appraise= [IMA] appraise integrity measurements
2138 Format: { "off" | "enforce" | "fix" | "log" }
2141 ima_appraise_tcb [IMA] Deprecated. Use ima_policy= instead.
2142 The builtin appraise policy appraises all files
2145 ima_canonical_fmt [IMA]
2146 Use the canonical format for the binary runtime
2147 measurements, instead of host native format.
2150 Format: { md5 | sha1 | rmd160 | sha256 | sha384
2154 The list of supported hash algorithms is defined
2155 in crypto/hash_info.h.
2158 The builtin policies to load during IMA setup.
2159 Format: "tcb | appraise_tcb | secure_boot |
2160 fail_securely | critical_data"
2162 The "tcb" policy measures all programs exec'd, files
2163 mmap'd for exec, and all files opened with the read
2164 mode bit set by either the effective uid (euid=0) or
2167 The "appraise_tcb" policy appraises the integrity of
2168 all files owned by root.
2170 The "secure_boot" policy appraises the integrity
2171 of files (eg. kexec kernel image, kernel modules,
2172 firmware, policy, etc) based on file signatures.
2174 The "fail_securely" policy forces file signature
2175 verification failure also on privileged mounted
2176 filesystems with the SB_I_UNVERIFIABLE_SIGNATURE
2179 The "critical_data" policy measures kernel integrity
2182 ima_tcb [IMA] Deprecated. Use ima_policy= instead.
2183 Load a policy which meets the needs of the Trusted
2184 Computing Base. This means IMA will measure all
2185 programs exec'd, files mmap'd for exec, and all files
2186 opened for read by uid=0.
2189 Select one of defined IMA measurements template formats.
2190 Formats: { "ima" | "ima-ng" | "ima-ngv2" | "ima-sig" |
2195 [IMA] Define a custom template format.
2196 Format: { "field1|...|fieldN" }
2198 ima.ahash_minsize= [IMA] Minimum file size for asynchronous hash usage
2199 Format: <min_file_size>
2200 Set the minimal file size for using asynchronous hash.
2201 If left unspecified, ahash usage is disabled.
2203 ahash performance varies for different data sizes on
2204 different crypto accelerators. This option can be used
2205 to achieve the best performance for a particular HW.
2207 ima.ahash_bufsize= [IMA] Asynchronous hash buffer size
2209 Set hashing buffer size. Default: 4k.
2211 ahash performance varies for different chunk sizes on
2212 different crypto accelerators. This option can be used
2213 to achieve best performance for particular HW.
2215 indirect_target_selection= [X86,Intel] Mitigation control for Indirect
2216 Target Selection(ITS) bug in Intel CPUs. Updated
2217 microcode is also required for a fix in IBPB.
2219 on: Enable mitigation (default).
2220 off: Disable mitigation.
2221 force: Force the ITS bug and deploy default
2223 vmexit: Only deploy mitigation if CPU is affected by
2224 guest/host isolation part of ITS.
2225 stuff: Deploy RSB-fill mitigation when retpoline is
2226 also deployed. Otherwise, deploy the default
2230 Documentation/admin-guide/hw-vuln/indirect-target-selection.rst
2234 Run specified binary instead of /sbin/init as init
2237 initcall_debug [KNL] Trace initcalls as they are executed. Useful
2238 for working out where the kernel is dying during
2241 initcall_blacklist= [KNL] Do not execute a comma-separated list of
2242 initcall functions. Useful for debugging built-in
2243 modules and initcalls.
2245 initramfs_async= [KNL]
2248 This parameter controls whether the initramfs
2249 image is unpacked asynchronously, concurrently
2250 with devices being probed and
2251 initialized. This should normally just work,
2252 but as a debugging aid, one can get the
2253 historical behaviour of the initramfs
2254 unpacking being completed before device_ and
2257 initrd= [BOOT,EARLY] Specify the location of the initial ramdisk
2259 initrdmem= [KNL,EARLY] Specify a physical address and size from which to
2260 load the initrd. If an initrd is compiled in or
2261 specified in the bootparams, it takes priority over this
2263 Format: ss[KMG],nn[KMG]
2266 init_on_alloc= [MM,EARLY] Fill newly allocated pages and heap objects with
2269 Default set by CONFIG_INIT_ON_ALLOC_DEFAULT_ON.
2271 init_on_free= [MM,EARLY] Fill freed pages and heap objects with zeroes.
2273 Default set by CONFIG_INIT_ON_FREE_DEFAULT_ON.
2275 init_pkru= [X86] Specify the default memory protection keys rights
2276 register contents for all processes. 0x55555554 by
2277 default (disallow access to all but pkey 0). Can
2278 override in debugfs after boot.
2280 inport.irq= [HW] Inport (ATI XL and Microsoft) busmouse driver
2283 int_pln_enable [X86] Enable power limit notification interrupt
2285 integrity_audit=[IMA]
2286 Format: { "0" | "1" }
2287 0 -- basic integrity auditing messages. (Default)
2288 1 -- additional integrity auditing messages.
2290 intel_iommu= [DMAR] Intel IOMMU driver (DMAR) option
2292 Enable intel iommu driver.
2294 Disable intel iommu driver.
2295 igfx_off [Default Off]
2296 By default, gfx is mapped as normal device. If a gfx
2297 device has a dedicated DMAR unit, the DMAR unit is
2298 bypassed by not enabling DMAR with this option. In
2299 this case, gfx device will use physical address for
2301 strict [Default Off]
2302 Deprecated, equivalent to iommu.strict=1.
2303 sp_off [Default Off]
2304 By default, super page will be supported if Intel IOMMU
2305 has the capability. With this option, super page will
2308 Enable the Intel IOMMU scalable mode if the hardware
2309 advertises that it has support for the scalable mode
2312 Disallow use of the Intel IOMMU scalable mode.
2313 tboot_noforce [Default Off]
2314 Do not force the Intel IOMMU enabled under tboot.
2315 By default, tboot will force Intel IOMMU on, which
2316 could harm performance of some high-throughput
2317 devices like 40GBit network cards, even if identity
2319 Note that using this option lowers the security
2320 provided by tboot because it makes the system
2321 vulnerable to DMA attacks.
2323 intel_idle.max_cstate= [KNL,HW,ACPI,X86]
2324 0 disables intel_idle and fall back on acpi_idle.
2325 1 to 9 specify maximum depth of C-state.
2327 intel_pstate= [X86,EARLY]
2329 Do not enable intel_pstate as the default
2330 scaling driver for the supported processors
2332 Use intel_pstate driver to bypass the scaling
2333 governors layer of cpufreq and provides it own
2334 algorithms for p-state selection. There are two
2335 P-state selection algorithms provided by
2336 intel_pstate in the active mode: powersave and
2337 performance. The way they both operate depends
2338 on whether or not the hardware managed P-states
2339 (HWP) feature has been enabled in the processor
2340 and possibly on the processor model.
2342 Use intel_pstate as a scaling driver, but configure it
2343 to work with generic cpufreq governors (instead of
2344 enabling its internal governor). This mode cannot be
2345 used along with the hardware-managed P-states (HWP)
2348 Enable intel_pstate on systems that prohibit it by default
2349 in favor of acpi-cpufreq. Forcing the intel_pstate driver
2350 instead of acpi-cpufreq may disable platform features, such
2351 as thermal controls and power capping, that rely on ACPI
2352 P-States information being indicated to OSPM and therefore
2353 should be used with caution. This option does not work with
2354 processors that aren't supported by the intel_pstate driver
2355 or on platforms that use pcc-cpufreq instead of acpi-cpufreq.
2357 Do not enable hardware P state control (HWP)
2360 Only load intel_pstate on systems which support
2361 hardware P state control (HWP) if available.
2363 Enforce ACPI _PPC performance limits. If the Fixed ACPI
2364 Description Table, specifies preferred power management
2365 profile as "Enterprise Server" or "Performance Server",
2366 then this feature is turned on by default.
2368 Allow per-logical-CPU P-State performance control limits using
2369 cpufreq sysfs interface
2371 Do not enable capacity-aware scheduling (CAS) on
2374 intremap= [X86-64,Intel-IOMMU,EARLY]
2375 on enable Interrupt Remapping (default)
2376 off disable Interrupt Remapping
2377 nosid disable Source ID checking
2379 BIOS x2APIC opt-out request will be ignored
2380 nopost disable Interrupt Posting
2382 enable MSIs delivered as posted interrupts
2384 iomem= Disable strict checking of access to MMIO memory
2385 strict regions from userspace.
2391 Don't initialize and use any kind of IOMMU.
2394 Force the use of the hardware IOMMU even when
2395 it is not actually needed (e.g. because < 3 GB
2399 Don't force hardware IOMMU usage when it is not
2409 Use software bounce buffering (SWIOTLB) (default for
2410 Intel machines). This can be used to prevent the usage
2411 of an available hardware IOMMU.
2419 Disable IOMMU bypass, using IOMMU for PCI devices.
2422 AMD Gart HW IOMMU-specific options:
2425 Set the size of the remapping area in bytes.
2428 Overwrite iommu off workarounds for specific chipsets
2431 Flush IOMMU on each allocation (default).
2434 Don't use IOMMU fullflush.
2437 Allocate an own aperture over RAM with size
2438 32MB<<order. (default: order=1, i.e. 64MB)
2441 Do scatter-gather (SG) merging. Implies "force"
2445 Don't do scatter-gather (SG) merging.
2448 Ask the IOMMU not to touch the aperture for AGP.
2451 Don't initialize the AGP driver and use full aperture.
2454 Always panic when IOMMU overflows.
2456 iommu.forcedac= [ARM64,X86,EARLY] Control IOVA allocation for PCI devices.
2457 Format: { "0" | "1" }
2458 0 - Try to allocate a 32-bit DMA address first, before
2459 falling back to the full range if needed.
2460 1 - Allocate directly from the full usable range,
2461 forcing Dual Address Cycle for PCI cards supporting
2462 greater than 32-bit addressing.
2464 iommu.strict= [ARM64,X86,S390,EARLY] Configure TLB invalidation behaviour
2465 Format: { "0" | "1" }
2467 Request that DMA unmap operations use deferred
2468 invalidation of hardware TLBs, for increased
2469 throughput at the cost of reduced device isolation.
2470 Will fall back to strict mode if not supported by
2471 the relevant IOMMU driver.
2473 DMA unmap operations invalidate IOMMU hardware TLBs
2475 unset - Use value of CONFIG_IOMMU_DEFAULT_DMA_{LAZY,STRICT}.
2476 Note: on x86, strict mode specified via one of the
2477 legacy driver-specific options takes precedence.
2480 [ARM64,X86,EARLY] Configure DMA to bypass the IOMMU by default.
2481 Format: { "0" | "1" }
2482 0 - Use IOMMU translation for DMA.
2483 1 - Bypass the IOMMU for DMA.
2484 unset - Use value of CONFIG_IOMMU_DEFAULT_PASSTHROUGH.
2486 io7= [HW] IO7 for Marvel-based Alpha systems
2487 See comment before marvel_specify_io7 in
2488 arch/alpha/kernel/core_marvel.c.
2490 io_delay= [X86,EARLY] I/O delay method
2492 Standard port 0x80 based delay
2494 Alternate port 0xed based delay (needed on some systems)
2496 Simple two microseconds delay
2501 See Documentation/admin-guide/nfs/nfsroot.rst.
2503 ipcmni_extend [KNL,EARLY] Extend the maximum number of unique System V
2504 IPC identifiers from 32,768 to 16,777,216.
2508 Determine whether IPE starts in permissive (0) or
2509 enforce (1) mode. The default is enforce.
2514 Start IPE with success auditing enabled, emitting
2515 an audit event when a binary is allowed. The default
2518 irqaffinity= [SMP] Set the default irq affinity mask
2519 The argument is a cpu list, as described above.
2521 irqchip.gicv2_force_probe=
2524 Force the kernel to look for the second 4kB page
2525 of a GICv2 controller even if the memory range
2526 exposed by the device tree is too small.
2528 irqchip.gicv3_nolpi=
2530 Force the kernel to ignore the availability of
2531 LPIs (and by consequence ITSs). Intended for system
2532 that use the kernel as a bootloader, and thus want
2533 to let secondary kernels in charge of setting up
2536 irqchip.gicv3_pseudo_nmi= [ARM64,EARLY]
2537 Enables support for pseudo-NMIs in the kernel. This
2538 requires the kernel to be built with
2539 CONFIG_ARM64_PSEUDO_NMI.
2542 When an interrupt is not handled search all handlers
2543 for it. Intended to get systems with badly broken
2547 When an interrupt is not handled search all handlers
2548 for it. Also check all handlers each timer
2549 interrupt. Intended to get systems with badly broken
2553 Format: <RDP>,<reset>,<pci_scan>,<verbosity>
2555 isolcpus= [KNL,SMP,ISOL] Isolate a given set of CPUs from disturbance.
2556 [Deprecated - use cpusets instead]
2557 Format: [flag-list,]<cpu-list>
2559 Specify one or more CPUs to isolate from disturbances
2560 specified in the flag list (default: domain):
2563 Disable the tick when a single task runs as well as
2564 disabling other kernel noises like having RCU callbacks
2565 offloaded. This is equivalent to the nohz_full parameter.
2567 A residual 1Hz tick is offloaded to workqueues, which you
2568 need to affine to housekeeping through the global
2569 workqueue's affinity configured via the
2570 /sys/devices/virtual/workqueue/cpumask sysfs file, or
2571 by using the 'domain' flag described below.
2573 NOTE: by default the global workqueue runs on all CPUs,
2574 so to protect individual CPUs the 'cpumask' file has to
2575 be configured manually after bootup.
2578 Isolate from the general SMP balancing and scheduling
2579 algorithms. Note that performing domain isolation this way
2580 is irreversible: it's not possible to bring back a CPU to
2581 the domains once isolated through isolcpus. It's strongly
2582 advised to use cpusets instead to disable scheduler load
2583 balancing through the "cpuset.sched_load_balance" file.
2584 It offers a much more flexible interface where CPUs can
2585 move in and out of an isolated set anytime.
2587 You can move a process onto or off an "isolated" CPU via
2588 the CPU affinity syscalls or cpuset.
2589 <cpu number> begins at 0 and the maximum value is
2590 "number of CPUs in system - 1".
2594 Isolate from being targeted by managed interrupts
2595 which have an interrupt mask containing isolated
2596 CPUs. The affinity of managed interrupts is
2597 handled by the kernel and cannot be changed via
2598 the /proc/irq/* interfaces.
2600 This isolation is best effort and only effective
2601 if the automatically assigned interrupt mask of a
2602 device queue contains isolated and housekeeping
2603 CPUs. If housekeeping CPUs are online then such
2604 interrupts are directed to the housekeeping CPU
2605 so that IO submitted on the housekeeping CPU
2606 cannot disturb the isolated CPU.
2608 If a queue's affinity mask contains only isolated
2609 CPUs then this parameter has no effect on the
2610 interrupt routing decision, though interrupts are
2611 only delivered when tasks running on those
2612 isolated CPUs submit IO. IO submitted on
2613 housekeeping CPUs has no influence on those
2616 The format of <cpu-list> is described above.
2620 ivrs_ioapic [HW,X86-64]
2621 Provide an override to the IOAPIC-ID<->DEVICE-ID
2622 mapping provided in the IVRS ACPI table.
2623 By default, PCI segment is 0, and can be omitted.
2625 For example, to map IOAPIC-ID decimal 10 to
2626 PCI segment 0x1 and PCI device 00:14.0,
2627 write the parameter as:
2628 ivrs_ioapic=10@0001:00:14.0
2631 * To map IOAPIC-ID decimal 10 to PCI device 00:14.0
2632 write the parameter as:
2633 ivrs_ioapic[10]=00:14.0
2634 * To map IOAPIC-ID decimal 10 to PCI segment 0x1 and
2635 PCI device 00:14.0 write the parameter as:
2636 ivrs_ioapic[10]=0001:00:14.0
2638 ivrs_hpet [HW,X86-64]
2639 Provide an override to the HPET-ID<->DEVICE-ID
2640 mapping provided in the IVRS ACPI table.
2641 By default, PCI segment is 0, and can be omitted.
2643 For example, to map HPET-ID decimal 10 to
2644 PCI segment 0x1 and PCI device 00:14.0,
2645 write the parameter as:
2646 ivrs_hpet=10@0001:00:14.0
2649 * To map HPET-ID decimal 0 to PCI device 00:14.0
2650 write the parameter as:
2651 ivrs_hpet[0]=00:14.0
2652 * To map HPET-ID decimal 10 to PCI segment 0x1 and
2653 PCI device 00:14.0 write the parameter as:
2654 ivrs_ioapic[10]=0001:00:14.0
2656 ivrs_acpihid [HW,X86-64]
2657 Provide an override to the ACPI-HID:UID<->DEVICE-ID
2658 mapping provided in the IVRS ACPI table.
2659 By default, PCI segment is 0, and can be omitted.
2661 For example, to map UART-HID:UID AMD0020:0 to
2662 PCI segment 0x1 and PCI device ID 00:14.5,
2663 write the parameter as:
2664 ivrs_acpihid=AMD0020:0@0001:00:14.5
2667 * To map UART-HID:UID AMD0020:0 to PCI segment is 0,
2668 PCI device ID 00:14.5, write the parameter as:
2669 ivrs_acpihid[00:14.5]=AMD0020:0
2670 * To map UART-HID:UID AMD0020:0 to PCI segment 0x1 and
2671 PCI device ID 00:14.5, write the parameter as:
2672 ivrs_acpihid[0001:00:14.5]=AMD0020:0
2674 js= [HW,JOY] Analog joystick
2675 See Documentation/input/joydev/joystick.rst.
2678 [KNL] Enforce KASAN (Kernel Address Sanitizer) to print
2679 report on every invalid memory access. Without this
2680 parameter KASAN will print report only for the first
2683 keep_bootcon [KNL,EARLY]
2684 Do not unregister boot console at start. This is only
2685 useful for debugging when something happens in the window
2686 between unregistering the boot console and initializing
2689 keepinitrd [HW,ARM] See retain_initrd.
2691 kernelcore= [KNL,X86,PPC,EARLY]
2692 Format: nn[KMGTPE] | nn% | "mirror"
2693 This parameter specifies the amount of memory usable by
2694 the kernel for non-movable allocations. The requested
2695 amount is spread evenly throughout all nodes in the
2696 system as ZONE_NORMAL. The remaining memory is used for
2697 movable memory in its own zone, ZONE_MOVABLE. In the
2698 event, a node is too small to have both ZONE_NORMAL and
2699 ZONE_MOVABLE, kernelcore memory will take priority and
2700 other nodes will have a larger ZONE_MOVABLE.
2702 ZONE_MOVABLE is used for the allocation of pages that
2703 may be reclaimed or moved by the page migration
2704 subsystem. Note that allocations like PTEs-from-HighMem
2705 still use the HighMem zone if it exists, and the Normal
2706 zone if it does not.
2708 It is possible to specify the exact amount of memory in
2709 the form of "nn[KMGTPE]", a percentage of total system
2710 memory in the form of "nn%", or "mirror". If "mirror"
2711 option is specified, mirrored (reliable) memory is used
2712 for non-movable allocations and remaining memory is used
2713 for Movable pages. "nn[KMGTPE]", "nn%", and "mirror"
2714 are exclusive, so you cannot specify multiple forms.
2716 kgdbdbgp= [KGDB,HW,EARLY] kgdb over EHCI usb debug port.
2717 Format: <Controller#>[,poll interval]
2718 The controller # is the number of the ehci usb debug
2719 port as it is probed via PCI. The poll interval is
2720 optional and is the number seconds in between
2721 each poll cycle to the debug port in case you need
2722 the functionality for interrupting the kernel with
2723 gdb or control-c on the dbgp connection. When
2724 not using this parameter you use sysrq-g to break into
2725 the kernel debugger.
2727 kgdboc= [KGDB,HW] kgdb over consoles.
2728 Requires a tty driver that supports console polling,
2729 or a supported polling keyboard driver (non-usb).
2730 Serial only format: <serial_device>[,baud]
2731 keyboard only format: kbd
2732 keyboard and serial format: kbd,<serial_device>[,baud]
2733 Optional Kernel mode setting:
2734 kms, kbd format: kms,kbd
2735 kms, kbd and serial format: kms,kbd,<ser_dev>[,baud]
2737 kgdboc_earlycon= [KGDB,HW,EARLY]
2738 If the boot console provides the ability to read
2739 characters and can work in polling mode, you can use
2740 this parameter to tell kgdb to use it as a backend
2741 until the normal console is registered. Intended to
2742 be used together with the kgdboc parameter which
2743 specifies the normal console to transition to.
2745 The name of the early console should be specified
2746 as the value of this parameter. Note that the name of
2747 the early console might be different than the tty
2748 name passed to kgdboc. It's OK to leave the value
2749 blank and the first boot console that implements
2750 read() will be picked.
2752 kgdbwait [KGDB,EARLY] Stop kernel execution and enter the
2753 kernel debugger at the earliest opportunity.
2756 Format: { "0" | "1" | "off" | "on" | "y" | "n" }
2757 Enables or disables Kexec HandOver.
2758 "0" | "off" | "n" - kexec handover is disabled
2759 "1" | "on" | "y" - kexec handover is enabled
2761 kho_scratch= [KEXEC,EARLY]
2762 Format: ll[KMG],mm[KMG],nn[KMG] | nn%
2763 Defines the size of the KHO scratch region. The KHO
2764 scratch regions are physically contiguous memory
2765 ranges that can only be used for non-kernel
2766 allocations. That way, even when memory is heavily
2767 fragmented with handed over memory, the kexeced
2768 kernel will always have enough contiguous ranges to
2771 It is possible to specify the exact amount of
2772 memory in the form of "ll[KMG],mm[KMG],nn[KMG]"
2773 where the first parameter defines the size of a low
2774 memory scratch area, the second parameter defines
2775 the size of a global scratch area and the third
2776 parameter defines the size of additional per-node
2777 scratch areas. The form "nn%" defines scale factor
2778 (in percents) of memory that was used during boot.
2780 kmac= [MIPS] Korina ethernet MAC address.
2781 Configure the RouterBoard 532 series on-chip
2782 Ethernet adapter MAC address.
2784 kmemleak= [KNL,EARLY] Boot-time kmemleak enable/disable
2785 Valid arguments: on, off
2787 Built with CONFIG_DEBUG_KMEMLEAK_DEFAULT_OFF=y,
2790 kprobe_event=[probe-list]
2791 [FTRACE] Add kprobe events and enable at boot time.
2792 The probe-list is a semicolon delimited list of probe
2793 definitions. Each definition is same as kprobe_events
2794 interface, but the parameters are comma delimited.
2795 For example, to add a kprobe event on vfs_read with
2796 arg1 and arg2, add to the command line;
2798 kprobe_event=p,vfs_read,$arg1,$arg2
2800 See also Documentation/trace/kprobetrace.rst "Kernel
2801 Boot Parameter" section.
2803 kpti= [ARM64,EARLY] Control page table isolation of
2804 user and kernel address spaces.
2805 Default: enabled on cores which need mitigation.
2809 kunit.enable= [KUNIT] Enable executing KUnit tests. Requires
2810 CONFIG_KUNIT to be set to be fully enabled. The
2811 default value can be overridden via
2812 KUNIT_DEFAULT_ENABLED.
2813 Default is 1 (enabled)
2815 kvm.ignore_msrs=[KVM] Ignore guest accesses to unhandled MSRs.
2816 Default is 0 (don't ignore, but inject #GP)
2818 kvm.eager_page_split=
2819 [KVM,X86] Controls whether or not KVM will try to
2820 proactively split all huge pages during dirty logging.
2821 Eager page splitting reduces interruptions to vCPU
2822 execution by eliminating the write-protection faults
2823 and MMU lock contention that would otherwise be
2824 required to split huge pages lazily.
2826 VM workloads that rarely perform writes or that write
2827 only to a small region of VM memory may benefit from
2828 disabling eager page splitting to allow huge pages to
2829 still be used for reads.
2831 The behavior of eager page splitting depends on whether
2832 KVM_DIRTY_LOG_INITIALLY_SET is enabled or disabled. If
2833 disabled, all huge pages in a memslot will be eagerly
2834 split when dirty logging is enabled on that memslot. If
2835 enabled, eager page splitting will be performed during
2836 the KVM_CLEAR_DIRTY ioctl, and only for the pages being
2839 Eager page splitting is only supported when kvm.tdp_mmu=Y.
2843 kvm.enable_virt_at_load=[KVM,ARM64,LOONGARCH,MIPS,RISCV,X86]
2844 If enabled, KVM will enable virtualization in hardware
2845 when KVM is loaded, and disable virtualization when KVM
2846 is unloaded (if KVM is built as a module).
2848 If disabled, KVM will dynamically enable and disable
2849 virtualization on-demand when creating and destroying
2850 VMs, i.e. on the 0=>1 and 1=>0 transitions of the
2853 Enabling virtualization at module load avoids potential
2854 latency for creation of the 0=>1 VM, as KVM serializes
2855 virtualization enabling across all online CPUs. The
2856 "cost" of enabling virtualization when KVM is loaded,
2857 is that doing so may interfere with using out-of-tree
2858 hypervisors that want to "own" virtualization hardware.
2860 kvm.enable_vmware_backdoor=[KVM] Support VMware backdoor PV interface.
2861 Default is false (don't support).
2864 [KVM] Controls the software workaround for the
2865 X86_BUG_ITLB_MULTIHIT bug.
2866 force : Always deploy workaround.
2867 off : Never deploy workaround.
2868 auto : Deploy workaround based on the presence of
2869 X86_BUG_ITLB_MULTIHIT.
2873 If the software workaround is enabled for the host,
2874 guests do need not to enable it for nested guests.
2876 kvm.nx_huge_pages_recovery_ratio=
2877 [KVM] Controls how many 4KiB pages are periodically zapped
2878 back to huge pages. 0 disables the recovery, otherwise if
2879 the value is N KVM will zap 1/Nth of the 4KiB pages every
2880 period (see below). The default is 60.
2882 kvm.nx_huge_pages_recovery_period_ms=
2883 [KVM] Controls the time period at which KVM zaps 4KiB pages
2884 back to huge pages. If the value is a non-zero N, KVM will
2885 zap a portion (see ratio above) of the pages every N msecs.
2886 If the value is 0 (the default), KVM will pick a period based
2887 on the ratio, such that a page is zapped after 1 hour on average.
2889 kvm-amd.nested= [KVM,AMD] Control nested virtualization feature in
2890 KVM/SVM. Default is 1 (enabled).
2892 kvm-amd.npt= [KVM,AMD] Control KVM's use of Nested Page Tables,
2893 a.k.a. Two-Dimensional Page Tables. Default is 1
2894 (enabled). Disable by KVM if hardware lacks support
2898 [KVM,ARM,EARLY] Select one of KVM/arm64's modes of
2901 none: Forcefully disable KVM.
2903 nvhe: Standard nVHE-based mode, without support for
2906 protected: Mode with support for guests whose state is
2907 kept private from the host, using VHE or
2908 nVHE depending on HW support.
2910 nested: VHE-based mode with support for nested
2911 virtualization. Requires at least ARMv8.4
2912 hardware (with FEAT_NV2).
2914 Defaults to VHE/nVHE based on hardware support. Setting
2915 mode to "protected" will disable kexec and hibernation
2916 for the host. To force nVHE on VHE hardware, add
2917 "arm64_sw.hvhe=0 id_aa64mmfr1.vh=0" to the
2919 "nested" is experimental and should be used with
2922 kvm-arm.vgic_v3_group0_trap=
2923 [KVM,ARM,EARLY] Trap guest accesses to GICv3 group-0
2926 kvm-arm.vgic_v3_group1_trap=
2927 [KVM,ARM,EARLY] Trap guest accesses to GICv3 group-1
2930 kvm-arm.vgic_v3_common_trap=
2931 [KVM,ARM,EARLY] Trap guest accesses to GICv3 common
2934 kvm-arm.vgic_v4_enable=
2935 [KVM,ARM,EARLY] Allow use of GICv4 for direct
2938 kvm-arm.wfe_trap_policy=
2939 [KVM,ARM] Control when to set WFE instruction trap for
2940 KVM VMs. Traps are allowed but not guaranteed by the
2943 trap: set WFE instruction trap
2945 notrap: clear WFE instruction trap
2947 kvm-arm.wfi_trap_policy=
2948 [KVM,ARM] Control when to set WFI instruction trap for
2949 KVM VMs. Traps are allowed but not guaranteed by the
2952 trap: set WFI instruction trap
2954 notrap: clear WFI instruction trap
2956 kvm_cma_resv_ratio=n [PPC,EARLY]
2957 Reserves given percentage from system memory area for
2958 contiguous memory allocation for KVM hash pagetable
2960 By default it reserves 5% of total system memory.
2964 kvm-intel.ept= [KVM,Intel] Control KVM's use of Extended Page Tables,
2965 a.k.a. Two-Dimensional Page Tables. Default is 1
2966 (enabled). Disable by KVM if hardware lacks support
2969 kvm-intel.emulate_invalid_guest_state=
2970 [KVM,Intel] Control whether to emulate invalid guest
2971 state. Ignored if kvm-intel.enable_unrestricted_guest=1,
2972 as guest state is never invalid for unrestricted
2973 guests. This param doesn't apply to nested guests (L2),
2974 as KVM never emulates invalid L2 guest state.
2975 Default is 1 (enabled).
2977 kvm-intel.flexpriority=
2978 [KVM,Intel] Control KVM's use of FlexPriority feature
2979 (TPR shadow). Default is 1 (enabled). Disable by KVM if
2980 hardware lacks support for it.
2983 [KVM,Intel] Control nested virtualization feature in
2984 KVM/VMX. Default is 1 (enabled).
2986 kvm-intel.unrestricted_guest=
2987 [KVM,Intel] Control KVM's use of unrestricted guest
2988 feature (virtualized real and unpaged mode). Default
2989 is 1 (enabled). Disable by KVM if EPT is disabled or
2990 hardware lacks support for it.
2992 kvm-intel.vmentry_l1d_flush=[KVM,Intel] Mitigation for L1 Terminal Fault
2995 Valid arguments: never, cond, always
2997 always: L1D cache flush on every VMENTER.
2998 cond: Flush L1D on VMENTER only when the code between
2999 VMEXIT and VMENTER can leak host memory.
3000 never: Disables the mitigation
3002 Default is cond (do L1 cache flush in specific instances)
3004 kvm-intel.vpid= [KVM,Intel] Control KVM's use of Virtual Processor
3005 Identification feature (tagged TLBs). Default is 1
3006 (enabled). Disable by KVM if hardware lacks support
3009 l1d_flush= [X86,INTEL,EARLY]
3010 Control mitigation for L1D based snooping vulnerability.
3012 Certain CPUs are vulnerable to an exploit against CPU
3013 internal buffers which can forward information to a
3014 disclosure gadget under certain conditions.
3016 In vulnerable processors, the speculatively
3017 forwarded data can be used in a cache side channel
3018 attack, to access data to which the attacker does
3019 not have direct access.
3021 This parameter controls the mitigation. The
3024 on - enable the interface for the mitigation
3026 l1tf= [X86,EARLY] Control mitigation of the L1TF vulnerability on
3029 The kernel PTE inversion protection is unconditionally
3030 enabled and cannot be disabled.
3033 Provides all available mitigations for the
3034 L1TF vulnerability. Disables SMT and
3035 enables all mitigations in the
3036 hypervisors, i.e. unconditional L1D flush.
3038 SMT control and L1D flush control via the
3039 sysfs interface is still possible after
3040 boot. Hypervisors will issue a warning
3041 when the first VM is started in a
3042 potentially insecure configuration,
3043 i.e. SMT enabled or L1D flush disabled.
3046 Same as 'full', but disables SMT and L1D
3047 flush runtime control. Implies the
3048 'nosmt=force' command line option.
3049 (i.e. sysfs control of SMT is disabled.)
3052 Leaves SMT enabled and enables the default
3053 hypervisor mitigation, i.e. conditional
3056 SMT control and L1D flush control via the
3057 sysfs interface is still possible after
3058 boot. Hypervisors will issue a warning
3059 when the first VM is started in a
3060 potentially insecure configuration,
3061 i.e. SMT enabled or L1D flush disabled.
3065 Disables SMT and enables the default
3066 hypervisor mitigation.
3068 SMT control and L1D flush control via the
3069 sysfs interface is still possible after
3070 boot. Hypervisors will issue a warning
3071 when the first VM is started in a
3072 potentially insecure configuration,
3073 i.e. SMT enabled or L1D flush disabled.
3076 Same as 'flush', but hypervisors will not
3077 warn when a VM is started in a potentially
3078 insecure configuration.
3081 Disables hypervisor mitigations and doesn't
3083 It also drops the swap size and available
3084 RAM limit restriction on both hypervisor and
3089 For details see: Documentation/admin-guide/hw-vuln/l1tf.rst
3095 lapic [X86-32,APIC,EARLY] Enable the local APIC even if BIOS
3098 lapic= [X86,APIC] Do not use TSC deadline
3099 value for LAPIC timer one-shot implementation. Default
3100 back to the programmable timer unit in the LAPIC.
3101 Format: notscdeadline
3103 lapic_timer_c2_ok [X86,APIC,EARLY] trust the local apic timer
3106 libata.dma= [LIBATA] DMA control
3107 libata.dma=0 Disable all PATA and SATA DMA
3108 libata.dma=1 PATA and SATA Disk DMA only
3109 libata.dma=2 ATAPI (CDROM) DMA only
3110 libata.dma=4 Compact Flash DMA only
3111 Combinations also work, so libata.dma=3 enables DMA
3112 for disks and CDROMs, but not CFs.
3114 libata.ignore_hpa= [LIBATA] Ignore HPA limit
3115 libata.ignore_hpa=0 keep BIOS limits (default)
3116 libata.ignore_hpa=1 ignore limits, using full disk
3118 libata.noacpi [LIBATA] Disables use of ACPI in libata suspend/resume
3122 libata.force= [LIBATA] Force configurations. The format is a comma-
3123 separated list of "[ID:]VAL" where ID is PORT[.DEVICE].
3124 PORT and DEVICE are decimal numbers matching port, link
3125 or device. Basically, it matches the ATA ID string
3126 printed on console by libata. If the whole ID part is
3127 omitted, the last PORT and DEVICE values are used. If
3128 ID hasn't been specified yet, the configuration applies
3129 to all ports, links and devices.
3131 If only DEVICE is omitted, the parameter applies to
3132 the port and all links and devices behind it. DEVICE
3133 number of 0 either selects the first device or the
3134 first fan-out link behind PMP device. It does not
3135 select the host link. DEVICE number of 15 selects the
3136 host link and device attached to it.
3138 The VAL specifies the configuration to force. As long
3139 as there is no ambiguity, shortcut notation is allowed.
3140 For example, both 1.5 and 1.5G would work for 1.5Gbps.
3141 The following configurations can be forced.
3143 * Cable type: 40c, 80c, short40c, unk, ign or sata.
3144 Any ID with matching PORT is used.
3146 * SATA link speed limit: 1.5Gbps or 3.0Gbps.
3148 * Transfer mode: pio[0-7], mwdma[0-4] and udma[0-7].
3149 udma[/][16,25,33,44,66,100,133] notation is also
3152 * nohrst, nosrst, norst: suppress hard, soft and both
3155 * rstonce: only attempt one reset during hot-unplug
3158 * [no]dbdelay: Enable or disable the extra 200ms delay
3159 before debouncing a link PHY and device presence
3162 * [no]ncq: Turn on or off NCQ.
3164 * [no]ncqtrim: Enable or disable queued DSM TRIM.
3166 * [no]ncqati: Enable or disable NCQ trim on ATI chipset.
3168 * [no]trim: Enable or disable (unqueued) TRIM.
3170 * trim_zero: Indicate that TRIM command zeroes data.
3172 * max_trim_128m: Set 128M maximum trim size limit.
3174 * [no]dma: Turn on or off DMA transfers.
3176 * atapi_dmadir: Enable ATAPI DMADIR bridge support.
3178 * atapi_mod16_dma: Enable the use of ATAPI DMA for
3179 commands that are not a multiple of 16 bytes.
3181 * [no]dmalog: Enable or disable the use of the
3182 READ LOG DMA EXT command to access logs.
3184 * [no]iddevlog: Enable or disable access to the
3185 identify device data log.
3187 * [no]logdir: Enable or disable access to the general
3188 purpose log directory.
3190 * max_sec_128: Set transfer size limit to 128 sectors.
3192 * max_sec_1024: Set or clear transfer size limit to
3195 * max_sec_lba48: Set or clear transfer size limit to
3198 * external: Mark port as external (hotplug-capable).
3200 * [no]lpm: Enable or disable link power management.
3202 * [no]setxfer: Indicate if transfer speed mode setting
3205 * [no]fua: Disable or enable FUA (Force Unit Access)
3206 support for devices supporting this feature.
3208 * dump_id: Dump IDENTIFY data.
3210 * disable: Disable this device.
3212 If there are multiple matching configurations changing
3213 the same attribute, the last one is used.
3215 load_ramdisk= [RAM] [Deprecated]
3217 lockd.nlm_grace_period=P [NFS] Assign grace period.
3220 lockd.nlm_tcpport=N [NFS] Assign TCP port.
3223 lockd.nlm_timeout=T [NFS] Assign timeout value.
3226 lockd.nlm_udpport=M [NFS] Assign UDP port.
3229 lockdown= [SECURITY,EARLY]
3230 { integrity | confidentiality }
3231 Enable the kernel lockdown feature. If set to
3232 integrity, kernel features that allow userland to
3233 modify the running kernel are disabled. If set to
3234 confidentiality, kernel features that allow userland
3235 to extract confidential information from the kernel
3238 locktorture.acq_writer_lim= [KNL]
3239 Set the time limit in jiffies for a lock
3240 acquisition. Acquisitions exceeding this limit
3241 will result in a splat once they do complete.
3243 locktorture.bind_readers= [KNL]
3244 Specify the list of CPUs to which the readers are
3247 locktorture.bind_writers= [KNL]
3248 Specify the list of CPUs to which the writers are
3251 locktorture.call_rcu_chains= [KNL]
3252 Specify the number of self-propagating call_rcu()
3253 chains to set up. These are used to ensure that
3254 there is a high probability of an RCU grace period
3255 in progress at any given time. Defaults to 0,
3256 which disables these call_rcu() chains.
3258 locktorture.long_hold= [KNL]
3259 Specify the duration in milliseconds for the
3260 occasional long-duration lock hold time. Defaults
3261 to 100 milliseconds. Select 0 to disable.
3263 locktorture.nested_locks= [KNL]
3264 Specify the maximum lock nesting depth that
3265 locktorture is to exercise, up to a limit of 8
3266 (MAX_NESTED_LOCKS). Specify zero to disable.
3267 Note that this parameter is ineffective on types
3268 of locks that do not support nested acquisition.
3270 locktorture.nreaders_stress= [KNL]
3271 Set the number of locking read-acquisition kthreads.
3272 Defaults to being automatically set based on the
3273 number of online CPUs.
3275 locktorture.nwriters_stress= [KNL]
3276 Set the number of locking write-acquisition kthreads.
3278 locktorture.onoff_holdoff= [KNL]
3279 Set time (s) after boot for CPU-hotplug testing.
3281 locktorture.onoff_interval= [KNL]
3282 Set time (s) between CPU-hotplug operations, or
3283 zero to disable CPU-hotplug testing.
3285 locktorture.rt_boost= [KNL]
3286 Do periodic testing of real-time lock priority
3287 boosting. Select 0 to disable, 1 to boost
3288 only rt_mutex, and 2 to boost unconditionally.
3289 Defaults to 2, which might seem to be an
3290 odd choice, but which should be harmless for
3291 non-real-time spinlocks, due to their disabling
3292 of preemption. Note that non-realtime mutexes
3295 locktorture.rt_boost_factor= [KNL]
3296 Number that determines how often and for how
3297 long priority boosting is exercised. This is
3298 scaled down by the number of writers, so that the
3299 number of boosts per unit time remains roughly
3300 constant as the number of writers increases.
3301 On the other hand, the duration of each boost
3302 increases with the number of writers.
3304 locktorture.shuffle_interval= [KNL]
3305 Set task-shuffle interval (jiffies). Shuffling
3306 tasks allows some CPUs to go into dyntick-idle
3307 mode during the locktorture test.
3309 locktorture.shutdown_secs= [KNL]
3310 Set time (s) after boot system shutdown. This
3311 is useful for hands-off automated testing.
3313 locktorture.stat_interval= [KNL]
3314 Time (s) between statistics printk()s.
3316 locktorture.stutter= [KNL]
3317 Time (s) to stutter testing, for example,
3318 specifying five seconds causes the test to run for
3319 five seconds, wait for five seconds, and so on.
3320 This tests the locking primitive's ability to
3321 transition abruptly to and from idle.
3323 locktorture.torture_type= [KNL]
3324 Specify the locking implementation to test.
3326 locktorture.verbose= [KNL]
3327 Enable additional printk() statements.
3329 locktorture.writer_fifo= [KNL]
3330 Run the write-side locktorture kthreads at
3331 sched_set_fifo() real-time priority.
3333 logibm.irq= [HW,MOUSE] Logitech Bus Mouse Driver
3336 loglevel= [KNL,EARLY]
3337 All Kernel Messages with a loglevel smaller than the
3338 console loglevel will be printed to the console. It can
3339 also be changed with klogd or other programs. The
3340 loglevels are defined as follows:
3342 0 (KERN_EMERG) system is unusable
3343 1 (KERN_ALERT) action must be taken immediately
3344 2 (KERN_CRIT) critical conditions
3345 3 (KERN_ERR) error conditions
3346 4 (KERN_WARNING) warning conditions
3347 5 (KERN_NOTICE) normal but significant condition
3348 6 (KERN_INFO) informational
3349 7 (KERN_DEBUG) debug-level messages
3351 log_buf_len=n[KMG] [KNL,EARLY]
3352 Sets the size of the printk ring buffer, in bytes.
3353 n must be a power of two and greater than the
3354 minimal size. The minimal size is defined by
3355 LOG_BUF_SHIFT kernel config parameter. There
3356 is also CONFIG_LOG_CPU_MAX_BUF_SHIFT config
3357 parameter that allows to increase the default size
3358 depending on the number of CPUs. See init/Kconfig
3361 logo.nologo [FB] Disables display of the built-in Linux logo.
3362 This may be used to provide more screen space for
3363 kernel log messages and is useful when debugging
3364 kernel boot problems.
3366 lp=0 [LP] Specify parallel ports to use, e.g,
3367 lp=port[,port...] lp=none,parport0 (lp0 not configured, lp1 uses
3368 lp=reset first parallel port). 'lp=0' disables the
3369 lp=auto printer driver. 'lp=reset' (which can be
3370 specified in addition to the ports) causes
3371 attached printers to be reset. Using
3372 lp=port1,port2,... specifies the parallel ports
3373 to associate lp devices with, starting with
3374 lp0. A port specification may be 'none' to skip
3375 that lp device, or a parport name such as
3376 'parport0'. Specifying 'lp=auto' instead of a
3377 port specification list means that device IDs
3378 from each port should be examined, to see if
3379 an IEEE 1284-compliant printer is attached; if
3380 so, the driver will manage that printer.
3381 See also header of drivers/char/lp.c.
3384 Sets loops_per_jiffy to given constant, thus avoiding
3385 time-consuming boot-time autodetection (up to 250 ms per
3386 CPU). 0 enables autodetection (default). To determine
3387 the correct value for your kernel, boot with normal
3388 autodetection and see what value is printed. Note that
3389 on SMP systems the preset will be applied to all CPUs,
3390 which is likely to cause problems if your CPUs need
3391 significantly divergent settings. An incorrect value
3392 will cause delays in the kernel to be wrong, leading to
3393 unpredictable I/O errors and other breakage. Although
3394 unlikely, in the extreme case this might damage your
3397 lsm.debug [SECURITY] Enable LSM initialization debugging output.
3400 [SECURITY] Choose order of LSM initialization. This
3401 overrides CONFIG_LSM, and the "security=" parameter.
3403 machtype= [Loongson] Share the same kernel image file between
3404 different yeeloong laptops.
3405 Example: machtype=lemote-yeeloong-2f-7inch
3407 maxcpus= [SMP,EARLY] Maximum number of processors that an SMP kernel
3408 will bring up during bootup. maxcpus=n : n >= 0 limits
3409 the kernel to bring up 'n' processors. Surely after
3410 bootup you can bring up the other plugged cpu by executing
3411 "echo 1 > /sys/devices/system/cpu/cpuX/online". So maxcpus
3412 only takes effect during system bootup.
3413 While n=0 is a special case, it is equivalent to "nosmp",
3414 which also disables the IO APIC.
3416 max_loop= [LOOP] The number of loop block devices that get
3417 (loop.max_loop) unconditionally pre-created at init time. The default
3418 number is configured by BLK_DEV_LOOP_MIN_COUNT. Instead
3419 of statically allocating a predefined number, loop
3420 devices can be requested on-demand with the
3421 /dev/loop-control interface.
3425 Please see Documentation/arch/x86/x86_64/machinecheck.rst for sysfs runtime tunables.
3428 disable machine check
3431 disable CMCI(Corrected Machine Check Interrupt) that
3432 Intel processor supports. Usually this disablement is
3433 not recommended, but it might be handy if your
3434 hardware is misbehaving.
3436 Note that you'll get more problems without CMCI than
3437 with due to the shared banks, i.e. you might get
3438 duplicated error logs.
3441 don't make logs for corrected errors. All events
3442 reported as corrected are silently cleared by OS. This
3443 option will be useful if you have no interest in any
3444 of corrected errors.
3447 disable features for corrected errors, e.g.
3448 polling timer and CMCI. All events reported as
3449 corrected are not cleared by OS and remained in its
3452 Usually this disablement is not recommended, however
3453 if there is an agent checking/clearing corrected
3454 errors (e.g. BIOS or hardware monitoring
3455 applications), conflicting with OS's error handling,
3456 and you cannot deactivate the agent, then this option
3460 do not opt-in to Local MCE delivery. Use legacy method
3464 enable logging of machine checks left over from
3465 booting. Disabled by default on AMD Fam10h and older
3466 because some BIOS leave bogus ones.
3468 If your BIOS doesn't do that it's a good idea to
3469 enable though to make sure you log even machine check
3470 events that result in a reboot. On Intel systems it is
3474 disable boot machine check logging.
3476 monarchtimeout (number)
3477 sets the time in us to wait for other CPUs on machine
3478 checks. 0 to disable.
3481 don't overwrite the bios-set CMCI threshold. This boot
3482 option prevents Linux from overwriting the CMCI
3483 threshold set by the bios. Without this option, Linux
3484 always sets the CMCI threshold to 1. Enabling this may
3485 make memory predictive failure analysis less effective
3486 if the bios sets thresholds for memory errors since we
3487 will not see details for all errors.
3490 force-enable recoverable machine check code paths
3492 Everything else is in sysfs now.
3495 md= [HW] RAID subsystems devices and level
3496 See Documentation/admin-guide/md.rst.
3499 Format: <first>,<last>
3500 Specifies range of consoles to be captured by the MDA.
3502 mds= [X86,INTEL,EARLY]
3503 Control mitigation for the Micro-architectural Data
3504 Sampling (MDS) vulnerability.
3506 Certain CPUs are vulnerable to an exploit against CPU
3507 internal buffers which can forward information to a
3508 disclosure gadget under certain conditions.
3510 In vulnerable processors, the speculatively
3511 forwarded data can be used in a cache side channel
3512 attack, to access data to which the attacker does
3513 not have direct access.
3515 This parameter controls the MDS mitigation. The
3518 full - Enable MDS mitigation on vulnerable CPUs
3519 full,nosmt - Enable MDS mitigation and disable
3520 SMT on vulnerable CPUs
3521 off - Unconditionally disable MDS mitigation
3523 On TAA-affected machines, mds=off can be prevented by
3524 an active TAA mitigation as both vulnerabilities are
3525 mitigated with the same mechanism so in order to disable
3526 this mitigation, you need to specify tsx_async_abort=off
3529 Not specifying this option is equivalent to
3532 For details see: Documentation/admin-guide/hw-vuln/mds.rst
3534 mem=nn[KMG] [HEXAGON,EARLY] Set the memory size.
3535 Must be specified, otherwise memory size will be 0.
3537 mem=nn[KMG] [KNL,BOOT,EARLY] Force usage of a specific amount
3538 of memory Amount of memory to be used in cases
3542 2 when the kernel is not able to see the whole system memory;
3543 3 memory that lies after 'mem=' boundary is excluded from
3544 the hypervisor, then assigned to KVM guests.
3545 4 to limit the memory available for kdump kernel.
3547 [ARC,MICROBLAZE] - the limit applies only to low memory,
3548 high memory is not affected.
3550 [ARM64] - only limits memory covered by the linear
3551 mapping. The NOMAP regions are not affected.
3553 [X86] Work as limiting max address. Use together
3554 with memmap= to avoid physical address space collisions.
3555 Without memmap= PCI devices could be placed at addresses
3556 belonging to unused RAM.
3558 Note that this only takes effects during boot time since
3559 in above case 3, memory may need be hot added after boot
3560 if system memory of hypervisor is not sufficient.
3563 [ARM,MIPS,EARLY] - override the memory layout
3564 reported by firmware.
3565 Define a memory region of size nn[KMG] starting at
3567 Multiple different regions can be specified with
3568 multiple mem= parameters on the command line.
3570 mem=nopentium [BUGS=X86-32] Disable usage of 4MB pages for kernel
3573 memblock=debug [KNL,EARLY] Enable memblock debug messages.
3576 [KNL,SH] Allow user to override the default size for
3577 per-device physically contiguous DMA buffers.
3579 memhp_default_state=online/offline/online_kernel/online_movable
3580 [KNL] Set the initial state for the memory hotplug
3581 onlining policy. If not specified, the default value is
3582 set according to the
3583 CONFIG_MHP_DEFAULT_ONLINE_TYPE kernel config
3585 See Documentation/admin-guide/mm/memory-hotplug.rst.
3587 memmap=exactmap [KNL,X86,EARLY] Enable setting of an exact
3588 E820 memory map, as specified by the user.
3589 Such memmap=exactmap lines can be constructed based on
3590 BIOS output or other requirements. See the memmap=nn@ss
3593 memmap=nn[KMG]@ss[KMG]
3594 [KNL, X86,MIPS,XTENSA,EARLY] Force usage of a specific region of memory.
3595 Region of memory to be used is from ss to ss+nn.
3596 If @ss[KMG] is omitted, it is equivalent to mem=nn[KMG],
3597 which limits max address to nn[KMG].
3598 Multiple different regions can be specified,
3601 memmap=100M@2G,100M#3G,1G!1024G
3603 memmap=nn[KMG]#ss[KMG]
3604 [KNL,ACPI,EARLY] Mark specific memory as ACPI data.
3605 Region of memory to be marked is from ss to ss+nn.
3607 memmap=nn[KMG]$ss[KMG]
3608 [KNL,ACPI,EARLY] Mark specific memory as reserved.
3609 Region of memory to be reserved is from ss to ss+nn.
3610 Example: Exclude memory from 0x18690000-0x1869ffff
3611 memmap=64K$0x18690000
3613 memmap=0x10000$0x18690000
3614 Some bootloaders may need an escape character before '$',
3615 like Grub2, otherwise '$' and the following number
3618 memmap=nn[KMG]!ss[KMG,EARLY]
3619 [KNL,X86] Mark specific memory as protected.
3620 Region of memory to be used, from ss to ss+nn.
3621 The memory region may be marked as e820 type 12 (0xc)
3622 and is NVDIMM or ADR memory.
3624 memmap=<size>%<offset>-<oldtype>+<newtype>
3625 [KNL,ACPI,EARLY] Convert memory within the specified region
3626 from <oldtype> to <newtype>. If "-<oldtype>" is left
3627 out, the whole region will be marked as <newtype>,
3628 even if previously unavailable. If "+<newtype>" is left
3629 out, matching memory will be removed. Types are
3630 specified as e820 types, e.g., 1 = RAM, 2 = reserved,
3631 3 = ACPI, 12 = PRAM.
3633 memory_corruption_check=0/1 [X86,EARLY]
3634 Some BIOSes seem to corrupt the first 64k of
3635 memory when doing things like suspend/resume.
3636 Setting this option will scan the memory
3637 looking for corruption. Enabling this will
3638 both detect corruption and prevent the kernel
3639 from using the memory being corrupted.
3640 However, its intended as a diagnostic tool; if
3641 repeatable BIOS-originated corruption always
3642 affects the same memory, you can use memmap=
3643 to prevent the kernel from using that memory.
3645 memory_corruption_check_size=size [X86,EARLY]
3646 By default it checks for corruption in the low
3647 64k, making this memory unavailable for normal
3648 use. Use this parameter to scan for
3649 corruption in more or less memory.
3651 memory_corruption_check_period=seconds [X86,EARLY]
3652 By default it checks for corruption every 60
3653 seconds. Use this parameter to check at some
3654 other rate. 0 disables periodic checking.
3656 memory_hotplug.memmap_on_memory
3657 [KNL,X86,ARM] Boolean flag to enable this feature.
3658 Format: {on | off (default)}
3659 When enabled, runtime hotplugged memory will
3660 allocate its internal metadata (struct pages,
3661 those vmemmap pages cannot be optimized even
3662 if hugetlb_free_vmemmap is enabled) from the
3663 hotadded memory which will allow to hotadd a
3664 lot of memory without requiring additional
3666 This feature is disabled by default because it
3667 has some implication on large (e.g. GB)
3668 allocations in some configurations (e.g. small
3670 The state of the flag can be read in
3671 /sys/module/memory_hotplug/parameters/memmap_on_memory.
3672 Note that even when enabled, there are a few cases where
3673 the feature is not effective.
3675 memtest= [KNL,X86,ARM,M68K,PPC,RISCV,EARLY] Enable memtest
3677 default : 0 <disable>
3678 Specifies the number of memtest passes to be
3679 performed. Each pass selects another test
3680 pattern from a given set of patterns. Memtest
3681 fills the memory with this pattern, validates
3682 memory contents and reserves bad memory
3683 regions that are detected.
3685 mem_encrypt= [X86-64] AMD Secure Memory Encryption (SME) control
3686 Valid arguments: on, off
3688 mem_encrypt=on: Activate SME
3689 mem_encrypt=off: Do not activate SME
3691 Refer to Documentation/virt/kvm/x86/amd-memory-encryption.rst
3692 for details on when memory encryption can be activated.
3694 mem_sleep_default= [SUSPEND] Default system suspend mode:
3695 s2idle - Suspend-To-Idle
3696 shallow - Power-On Suspend or equivalent (if supported)
3697 deep - Suspend-To-RAM or equivalent (if supported)
3698 See Documentation/admin-guide/pm/sleep-states.rst.
3700 mfgptfix [X86-32] Fix MFGPT timers on AMD Geode platforms when
3701 the BIOS has incorrectly applied a workaround. TinyBIOS
3702 version 0.98 is known to be affected, 0.99 fixes the
3703 problem by letting the user disable the workaround.
3707 microcode.force_minrev= [X86]
3709 Enable or disable the microcode minimal revision
3710 enforcement for the runtime microcode loader.
3712 mini2440= [ARM,HW,KNL]
3713 Format:[0..2][b][c][t]
3715 MINI2440 configuration specification:
3716 0 - The attached screen is the 3.5" TFT
3717 1 - The attached screen is the 7" TFT
3718 2 - The VGA Shield is attached (1024x768)
3719 Leaving out the screen size parameter will not load
3720 the TFT driver, and the framebuffer will be left
3722 b - Enable backlight. The TFT backlight pin will be
3723 linked to the kernel VESA blanking code and a GPIO
3724 LED. This parameter is not necessary when using the
3726 c - Enable the s3c camera interface.
3727 t - Reserved for enabling touchscreen support. The
3728 touchscreen support is not enabled in the mainstream
3729 kernel as of 2.6.30, a preliminary port can be found
3730 in the "bleeding edge" mini2440 support kernel at
3731 https://repo.or.cz/w/linux-2.6/mini2440.git
3734 [X86,PPC,S390,ARM64,EARLY] Control optional mitigations for
3735 CPU vulnerabilities. This is a set of curated,
3736 arch-independent options, each of which is an
3737 aggregation of existing arch-specific options.
3739 Note, "mitigations" is supported if and only if the
3740 kernel was built with CPU_MITIGATIONS=y.
3743 Disable all optional CPU mitigations. This
3744 improves system performance, but it may also
3745 expose users to several CPU vulnerabilities.
3746 Equivalent to: if nokaslr then kpti=0 [ARM64]
3747 gather_data_sampling=off [X86]
3748 indirect_target_selection=off [X86]
3749 kvm.nx_huge_pages=off [X86]
3752 mmio_stale_data=off [X86]
3753 no_entry_flush [PPC]
3754 no_uaccess_flush [PPC]
3757 nospectre_bhb [ARM64]
3758 nospectre_v1 [X86,PPC]
3759 nospectre_v2 [X86,PPC,S390,ARM64]
3760 reg_file_data_sampling=off [X86]
3762 spec_rstack_overflow=off [X86]
3763 spec_store_bypass_disable=off [X86,PPC]
3764 spectre_bhi=off [X86]
3765 spectre_v2_user=off [X86]
3766 srbds=off [X86,INTEL]
3767 ssbd=force-off [ARM64]
3768 tsx_async_abort=off [X86]
3771 This does not have any effect on
3772 kvm.nx_huge_pages when
3773 kvm.nx_huge_pages=force.
3776 Mitigate all CPU vulnerabilities, but leave SMT
3777 enabled, even if it's vulnerable. This is for
3778 users who don't want to be surprised by SMT
3779 getting disabled across kernel upgrades, or who
3780 have other ways of avoiding SMT-based attacks.
3781 Equivalent to: (default behavior)
3784 Mitigate all CPU vulnerabilities, disabling SMT
3785 if needed. This is for users who always want to
3786 be fully mitigated, even if it means losing SMT.
3787 Equivalent to: l1tf=flush,nosmt [X86]
3788 mds=full,nosmt [X86]
3789 tsx_async_abort=full,nosmt [X86]
3790 mmio_stale_data=full,nosmt [X86]
3791 retbleed=auto,nosmt [X86]
3794 [KNL,EARLY] When CONFIG_DEBUG_MEMORY_INIT is set, this
3795 parameter allows control of the logging verbosity for
3796 the additional memory initialisation checks. A value
3797 of 0 disables mminit logging and a level of 4 will
3798 log everything. Information is printed at KERN_DEBUG
3799 so loglevel=8 may also need to be specified.
3802 [X86,INTEL,EARLY] Control mitigation for the Processor
3803 MMIO Stale Data vulnerabilities.
3805 Processor MMIO Stale Data is a class of
3806 vulnerabilities that may expose data after an MMIO
3807 operation. Exposed data could originate or end in
3808 the same CPU buffers as affected by MDS and TAA.
3809 Therefore, similar to MDS and TAA, the mitigation
3810 is to clear the affected CPU buffers.
3812 This parameter controls the mitigation. The
3815 full - Enable mitigation on vulnerable CPUs
3817 full,nosmt - Enable mitigation and disable SMT on
3820 off - Unconditionally disable mitigation
3822 On MDS or TAA affected machines,
3823 mmio_stale_data=off can be prevented by an active
3824 MDS or TAA mitigation as these vulnerabilities are
3825 mitigated with the same mechanism so in order to
3826 disable this mitigation, you need to specify
3827 mds=off and tsx_async_abort=off too.
3829 Not specifying this option is equivalent to
3830 mmio_stale_data=full.
3833 Documentation/admin-guide/hw-vuln/processor_mmio_stale_data.rst
3835 <module>.async_probe[=<bool>] [KNL]
3836 If no <bool> value is specified or if the value
3837 specified is not a valid <bool>, enable asynchronous
3838 probe on this module. Otherwise, enable/disable
3839 asynchronous probe on this module as indicated by the
3840 <bool> value. See also: module.async_probe
3842 module.async_probe=<bool>
3843 [KNL] When set to true, modules will use async probing
3844 by default. To enable/disable async probing for a
3845 specific module, use the module specific control that
3846 is documented under <module>.async_probe. When both
3847 module.async_probe and <module>.async_probe are
3848 specified, <module>.async_probe takes precedence for
3849 the specific module.
3851 module.enable_dups_trace
3852 [KNL] When CONFIG_MODULE_DEBUG_AUTOLOAD_DUPS is set,
3853 this means that duplicate request_module() calls will
3854 trigger a WARN_ON() instead of a pr_warn(). Note that
3855 if MODULE_DEBUG_AUTOLOAD_DUPS_TRACE is set, WARN_ON()s
3856 will always be issued and this option does nothing.
3858 [KNL] When CONFIG_MODULE_SIG is set, this means that
3859 modules without (valid) signatures will fail to load.
3860 Note that if CONFIG_MODULE_SIG_FORCE is set, that
3861 is always true, so this option does nothing.
3863 module_blacklist= [KNL] Do not load a comma-separated list of
3864 modules. Useful for debugging problem modules.
3867 [MOUSE] Maximum time between finger touching and
3868 leaving touchpad surface for touch to be considered
3869 a tap and be reported as a left button click (for
3870 touchpads working in absolute mode only).
3872 mousedev.xres= [MOUSE] Horizontal screen resolution, used for devices
3873 reporting absolute coordinates, such as tablets
3874 mousedev.yres= [MOUSE] Vertical screen resolution, used for devices
3875 reporting absolute coordinates, such as tablets
3877 movablecore= [KNL,X86,PPC,EARLY]
3878 Format: nn[KMGTPE] | nn%
3879 This parameter is the complement to kernelcore=, it
3880 specifies the amount of memory used for migratable
3881 allocations. If both kernelcore and movablecore is
3882 specified, then kernelcore will be at *least* the
3883 specified value but may be more. If movablecore on its
3884 own is specified, the administrator must be careful
3885 that the amount of memory usable for all allocations
3888 movable_node [KNL,EARLY] Boot-time switch to make hotplugable memory
3889 NUMA nodes to be movable. This means that the memory
3890 of such nodes will be usable only for movable
3891 allocations which rules out almost all kernel
3892 allocations. Use with caution!
3894 MTD_Partition= [MTD]
3895 Format: <name>,<region-number>,<size>,<offset>
3897 MTD_Region= [MTD] Format:
3898 <name>,<region-number>[,<base>,<size>,<buswidth>,<altbuswidth>]
3901 See drivers/mtd/parsers/cmdlinepart.c
3903 mtouchusb.raw_coordinates=
3904 [HW] Make the MicroTouch USB driver use raw coordinates
3905 ('y', default) or cooked coordinates ('n')
3907 mtrr=debug [X86,EARLY]
3908 Enable printing debug information related to MTRR
3909 registers at boot time.
3911 mtrr_chunk_size=nn[KMG,X86,EARLY]
3912 used for mtrr cleanup. It is largest continuous chunk
3913 that could hold holes aka. UC entries.
3915 mtrr_gran_size=nn[KMG,X86,EARLY]
3916 Used for mtrr cleanup. It is granularity of mtrr block.
3918 Large value could prevent small alignment from
3921 mtrr_spare_reg_nr=n [X86,EARLY]
3923 Range: 0,7 : spare reg number
3925 Used for mtrr cleanup. It is spare mtrr entries number.
3926 Set to 2 or more if your graphical card needs more.
3928 multitce=off [PPC] This parameter disables the use of the pSeries
3929 firmware feature for updating multiple TCE entries
3932 n2= [NET] SDL Inc. RISCom/N2 synchronous serial card
3934 netdev= [NET] Network devices parameters
3935 Format: <irq>,<io>,<mem_start>,<mem_end>,<name>
3936 Note that mem_start is often overloaded to mean
3937 something different and driver-specific.
3938 This usage is only documented in each driver source
3941 netpoll.carrier_timeout=
3942 [NET] Specifies amount of time (in seconds) that
3943 netpoll should wait for a carrier. By default netpoll
3947 [NETFILTER] Enable connection tracking flow accounting
3948 0 to disable accounting
3949 1 to enable accounting
3953 [NFS] sets the pathname to the program which is used
3954 to update the NFS client cache entries.
3956 nfs.cache_getent_timeout=
3957 [NFS] sets the timeout after which an attempt to
3958 update a cache entry is deemed to have failed.
3960 nfs.callback_nr_threads=
3961 [NFSv4] set the total number of threads that the
3962 NFS client will assign to service NFSv4 callback
3965 nfs.callback_tcpport=
3966 [NFS] set the TCP port on which the NFSv4 callback
3967 channel should listen.
3970 [NFS] specifies the number of times the NFSv4 client
3971 retries the request before returning an EAGAIN error,
3972 after a reply of NFS4ERR_DELAY from the server.
3973 Only applies if the softerr mount option is enabled,
3974 and the specified value is >= 0.
3977 [NFS] enable 64-bit inode numbers.
3978 If zero, the NFS client will fake up a 32-bit inode
3979 number for the readdir() and stat() syscalls instead
3980 of returning the full 64-bit number.
3981 The default is to return 64-bit inode numbers.
3983 nfs.idmap_cache_timeout=
3984 [NFS] set the maximum lifetime for idmapper cache
3987 nfs.max_session_cb_slots=
3988 [NFSv4.1] Sets the maximum number of session
3989 slots the client will assign to the callback
3990 channel. This determines the maximum number of
3991 callbacks the client will process in parallel for
3992 a particular server.
3994 nfs.max_session_slots=
3995 [NFSv4.1] Sets the maximum number of session slots
3996 the client will attempt to negotiate with the server.
3997 This limits the number of simultaneous RPC requests
3998 that the client can send to the NFSv4.1 server.
3999 Note that there is little point in setting this
4000 value higher than the max_tcp_slot_table_limit.
4002 nfs.nfs4_disable_idmapping=
4003 [NFSv4] When set to the default of '1', this option
4004 ensures that both the RPC level authentication
4005 scheme and the NFS level operations agree to use
4006 numeric uids/gids if the mount is using the
4007 'sec=sys' security flavour. In effect it is
4008 disabling idmapping, which can make migration from
4009 legacy NFSv2/v3 systems to NFSv4 easier.
4010 Servers that do not support this mode of operation
4011 will be autodetected by the client, and it will fall
4012 back to using the idmapper.
4013 To turn off this behaviour, set the value to '0'.
4016 [NFS4] Specify an additional fixed unique ident-
4017 ification string that NFSv4 clients can insert into
4018 their nfs_client_id4 string. This is typically a
4019 UUID that is generated at system install time.
4021 nfs.recover_lost_locks=
4022 [NFSv4] Attempt to recover locks that were lost due
4023 to a lease timeout on the server. Please note that
4024 doing this risks data corruption, since there are
4025 no guarantees that the file will remain unchanged
4026 after the locks are lost.
4027 If you want to enable the kernel legacy behaviour of
4028 attempting to recover these locks, then set this
4030 The default parameter value of '0' causes the kernel
4031 not to attempt recovery of lost locks.
4033 nfs.send_implementation_id=
4034 [NFSv4.1] Send client implementation identification
4035 information in exchange_id requests.
4036 If zero, no implementation identification information
4038 The default is to send the implementation identification
4041 nfs4.layoutstats_timer=
4042 [NFSv4.2] Change the rate at which the kernel sends
4043 layoutstats to the pNFS metadata server.
4045 Setting this to value to 0 causes the kernel to use
4046 whatever value is the default set by the layout
4047 driver. A non-zero value sets the minimum interval
4048 in seconds between layoutstats transmissions.
4050 nfsd.inter_copy_offload_enable=
4051 [NFSv4.2] When set to 1, the server will support
4052 server-to-server copies for which this server is
4053 the destination of the copy.
4055 nfsd.nfs4_disable_idmapping=
4056 [NFSv4] When set to the default of '1', the NFSv4
4057 server will return only numeric uids and gids to
4058 clients using auth_sys, and will accept numeric uids
4059 and gids from such clients. This is intended to ease
4060 migration from NFSv2/v3.
4062 nfsd.nfsd4_ssc_umount_timeout=
4063 [NFSv4.2] When used as the destination of a
4064 server-to-server copy, knfsd temporarily mounts
4065 the source server. It caches the mount in case
4066 it will be needed again, and discards it if not
4067 used for the number of milliseconds specified by
4070 nfsaddrs= [NFS] Deprecated. Use ip= instead.
4071 See Documentation/admin-guide/nfs/nfsroot.rst.
4073 nfsroot= [NFS] nfs root filesystem for disk-less boxes.
4074 See Documentation/admin-guide/nfs/nfsroot.rst.
4076 nfsrootdebug [NFS] enable nfsroot debugging messages.
4077 See Documentation/admin-guide/nfs/nfsroot.rst.
4079 nmi_backtrace.backtrace_idle [KNL]
4080 Dump stacks even of idle CPUs in response to an
4081 NMI stack-backtrace request.
4083 nmi_debug= [KNL,SH] Specify one or more actions to take
4084 when a NMI is triggered.
4085 Format: [state][,regs][,debounce][,die]
4087 nmi_watchdog= [KNL,BUGS=X86] Debugging features for SMP kernels
4088 Format: [panic,][nopanic,][rNNN,][num]
4090 0 - turn hardlockup detector in nmi_watchdog off
4091 1 - turn hardlockup detector in nmi_watchdog on
4092 rNNN - configure the watchdog with raw perf event 0xNNN
4094 When panic is specified, panic when an NMI watchdog
4095 timeout occurs (or 'nopanic' to not panic on an NMI
4096 watchdog, if CONFIG_BOOTPARAM_HARDLOCKUP_PANIC is set)
4097 To disable both hard and soft lockup detectors,
4098 please see 'nowatchdog'.
4099 This is useful when you use a panic=... timeout and
4100 need the box quickly up again.
4102 These settings can be accessed at runtime via
4103 the nmi_watchdog and hardlockup_panic sysctls.
4105 no387 [BUGS=X86-32] Tells the kernel to use the 387 maths
4106 emulation library even if a 387 maths coprocessor
4109 no4lvl [RISCV,EARLY] Disable 4-level and 5-level paging modes.
4110 Forces kernel to use 3-level paging instead.
4112 no5lvl [X86-64,RISCV,EARLY] Disable 5-level paging mode. Forces
4113 kernel to use 4-level paging instead.
4117 noapic [SMP,APIC,EARLY] Tells the kernel to not make use of any
4118 IOAPICs that may be present in the system.
4120 noapictimer [APIC,X86] Don't set up the APIC timer
4122 noautogroup Disable scheduler automatic task group creation.
4127 [HW] Never suspend the console
4128 Disable suspending of consoles during suspend and
4129 hibernate operations. Once disabled, debugging
4130 messages can reach various consoles while the rest
4131 of the system is being put to sleep (ie, while
4132 debugging driver suspend/resume hooks). This may
4133 not work reliably with all consoles, but is known
4134 to work with serial and VGA consoles.
4135 To facilitate more flexible debugging, we also add
4136 console_suspend, a printk module parameter to control
4137 it. Users could use console_suspend (usually
4138 /sys/module/printk/parameters/console_suspend) to
4139 turn on/off it dynamically.
4142 [KNL,EARLY] Disable object debugging
4144 nodsp [SH] Disable hardware DSP at boot time.
4146 noefi [EFI,EARLY] Disable EFI runtime services support.
4148 no_entry_flush [PPC,EARLY] Don't flush the L1-D cache when entering the kernel.
4151 This affects only 32-bit executables.
4152 noexec32=on: enable non-executable mappings (default)
4153 read doesn't imply executable mappings
4154 noexec32=off: disable non-executable mappings
4155 read implies executable mappings
4157 no_file_caps Tells the kernel not to honor file capabilities. The
4158 only way then for a file to be executed with privilege
4159 is to be setuid root or executed by root.
4161 nofpu [MIPS,SH] Disable hardware FPU at boot time.
4163 nofsgsbase [X86] Disables FSGSBASE instructions.
4165 nofxsr [BUGS=X86-32] Disables x86 floating point extended
4166 register save and restore. The kernel will only save
4167 legacy floating-point registers on task switch.
4169 nogbpages [X86] Do not use GB pages for kernel direct mappings.
4173 Force pointers printed to the console or buffers to be
4174 unhashed. By default, when a pointer is printed via %p
4175 format string, that pointer is "hashed", i.e. obscured
4176 by hashing the pointer value. This is a security feature
4177 that hides actual kernel addresses from unprivileged
4178 users, but it also makes debugging the kernel more
4179 difficult since unequal pointers can no longer be
4180 compared. However, if this command-line option is
4181 specified, then all normal pointers will have their true
4182 value printed. This option should only be specified when
4183 debugging the kernel. Please do not use on production
4186 nohibernate [HIBERNATION] Disable hibernation and resume.
4188 nohlt [ARM,ARM64,MICROBLAZE,MIPS,PPC,RISCV,SH] Forces the kernel to
4189 busy wait in do_idle() and not use the arch_cpu_idle()
4190 implementation; requires CONFIG_GENERIC_IDLE_POLL_SETUP
4191 to be effective. This is useful on platforms where the
4192 sleep(SH) or wfi(ARM,ARM64) instructions do not work
4193 correctly or when doing power measurements to evaluate
4194 the impact of the sleep instructions. This is also
4195 useful when using JTAG debugger.
4197 nohpet [X86] Don't use the HPET timer.
4199 nohugeiomap [KNL,X86,PPC,ARM64,EARLY] Disable kernel huge I/O mappings.
4201 nohugevmalloc [KNL,X86,PPC,ARM64,EARLY] Disable kernel huge vmalloc mappings.
4203 nohz= [KNL] Boottime enable/disable dynamic ticks
4204 Valid arguments: on, off
4207 nohz_full= [KNL,BOOT,SMP,ISOL]
4208 The argument is a cpu list, as described above.
4209 In kernels built with CONFIG_NO_HZ_FULL=y, set
4210 the specified list of CPUs whose tick will be stopped
4211 whenever possible. The boot CPU will be forced outside
4212 the range to maintain the timekeeping. Any CPUs
4213 in this list will have their RCU callbacks offloaded,
4214 just as if they had also been called out in the
4215 rcu_nocbs= boot parameter.
4217 Note that this argument takes precedence over
4218 the CONFIG_RCU_NOCB_CPU_DEFAULT_ALL option.
4220 noinitrd [RAM] Tells the kernel not to load any configured
4223 nointremap [X86-64,Intel-IOMMU,EARLY] Do not enable interrupt
4225 [Deprecated - use intremap=off]
4227 noinvpcid [X86,EARLY] Disable the INVPCID cpu feature.
4229 noiotrap [SH] Disables trapped I/O port accesses.
4231 noirqdebug [X86-32] Disables the code which attempts to detect and
4232 disable unhandled interrupt sources.
4234 noisapnp [ISAPNP] Disables ISA PnP code.
4237 When CONFIG_RANDOMIZE_BASE is set, this disables
4238 kernel and module base offset ASLR (Address Space
4239 Layout Randomization).
4241 no-kvmapf [X86,KVM,EARLY] Disable paravirtualized asynchronous page
4244 no-kvmclock [X86,KVM,EARLY] Disable paravirtualized KVM clock driver
4246 nolapic [X86-32,APIC,EARLY] Do not enable or use the local APIC.
4248 nolapic_timer [X86-32,APIC,EARLY] Do not use the local APIC timer.
4250 nomce [X86-32] Disable Machine Check Exception
4252 nomfgpt [X86-32] Disable Multi-Function General Purpose
4253 Timer usage (for AMD Geode machines).
4255 nomodeset Disable kernel modesetting. Most systems' firmware
4256 sets up a display mode and provides framebuffer memory
4257 for output. With nomodeset, DRM and fbdev drivers will
4258 not load if they could possibly displace the pre-
4259 initialized output. Only the system framebuffer will
4260 be available for use. The respective drivers will not
4261 perform display-mode changes or accelerated rendering.
4263 Useful as error fallback, or for testing and debugging.
4265 nomodule Disable module load
4267 nonmi_ipi [X86] Disable using NMI IPIs during panic/reboot to
4268 shutdown the other cpus. Instead use the REBOOT_VECTOR
4271 nopat [X86,EARLY] Disable PAT (page attribute table extension of
4272 pagetables) support.
4274 nopcid [X86-64,EARLY] Disable the PCID cpu feature.
4276 nopku [X86] Disable Memory Protection Keys CPU feature found
4279 nopti [X86-64,EARLY]
4280 Equivalent to pti=off
4282 nopv= [X86,XEN,KVM,HYPER_V,VMWARE,EARLY]
4283 Disables the PV optimizations forcing the guest to run
4284 as generic guest with no PV drivers. Currently support
4285 XEN HVM, KVM, HYPER_V and VMWARE guest.
4287 nopvspin [X86,XEN,KVM,EARLY]
4288 Disables the qspinlock slow path using PV optimizations
4289 which allow the hypervisor to 'idle' the guest on lock
4292 norandmaps Don't use address space randomization. Equivalent to
4293 echo 0 > /proc/sys/kernel/randomize_va_space
4295 noreplace-smp [X86-32,SMP] Don't replace SMP instructions
4296 with UP alternatives
4298 noresume [SWSUSP] Disables resume and restores original swap
4301 no-scroll [VGA] Disables scrollback.
4302 This is required for the Braillex ib80-piezo Braille
4303 reader made by F.H. Papenmeier (Germany).
4305 nosgx [X86-64,SGX,EARLY] Disables Intel SGX kernel support.
4308 Disable SMAP (Supervisor Mode Access Prevention)
4309 even if it is supported by processor.
4311 nosmep [PPC64s,EARLY]
4312 Disable SMEP (Supervisor Mode Execution Prevention)
4313 even if it is supported by processor.
4315 nosmp [SMP,EARLY] Tells an SMP kernel to act as a UP kernel,
4316 and disable the IO APIC. legacy for "maxcpus=0".
4318 nosmt [KNL,MIPS,PPC,EARLY] Disable symmetric multithreading (SMT).
4319 Equivalent to smt=1.
4321 [KNL,X86,PPC,S390] Disable symmetric multithreading (SMT).
4322 nosmt=force: Force disable SMT, cannot be undone
4323 via the sysfs control file.
4325 nosoftlockup [KNL] Disable the soft-lockup detector.
4327 nospec_store_bypass_disable
4328 [HW,EARLY] Disable all mitigations for the Speculative
4329 Store Bypass vulnerability
4331 nospectre_bhb [ARM64,EARLY] Disable all mitigations for Spectre-BHB (branch
4332 history injection) vulnerability. System may allow data leaks
4335 nospectre_v1 [X86,PPC,EARLY] Disable mitigations for Spectre Variant 1
4336 (bounds check bypass). With this option data leaks are
4337 possible in the system.
4339 nospectre_v2 [X86,PPC_E500,ARM64,EARLY] Disable all mitigations
4340 for the Spectre variant 2 (indirect branch
4341 prediction) vulnerability. System may allow data
4342 leaks with this option.
4344 no-steal-acc [X86,PV_OPS,ARM64,PPC/PSERIES,RISCV,LOONGARCH,EARLY]
4345 Disable paravirtualized steal time accounting. steal time
4346 is computed, but won't influence scheduler behaviour
4348 nosync [HW,M68K] Disables sync negotiation for all devices.
4350 no_timer_check [X86,APIC] Disables the code which tests for broken
4351 timer IRQ sources, i.e., the IO-APIC timer. This can
4352 work around problems with incorrect timer
4353 initialization on some boards.
4356 [PPC,EARLY] Don't flush the L1-D cache after accessing user data.
4358 novmcoredd [KNL,KDUMP]
4359 Disable device dump. Device dump allows drivers to
4360 append dump data to vmcore so you can collect driver
4361 specified debug info. Drivers can append the data
4362 without any limit and this data is stored in memory,
4363 so this may cause significant memory stress. Disabling
4364 device dump can help save memory but the driver debug
4365 data will be no longer available. This parameter
4366 is only available when CONFIG_PROC_VMCORE_DEVICE_DUMP
4370 [X86,PV_OPS,EARLY] Disable paravirtualized VMware
4371 scheduler clock and use the default one.
4373 nowatchdog [KNL] Disable both lockup detectors, i.e.
4374 soft-lockup and NMI watchdog (hard-lockup).
4378 nox2apic [X86-64,APIC,EARLY] Do not enable x2APIC mode.
4380 NOTE: this parameter will be ignored on systems with the
4381 LEGACY_XAPIC_DISABLED bit set in the
4382 IA32_XAPIC_DISABLE_STATUS MSR.
4384 noxsave [BUGS=X86] Disables x86 extended register state save
4385 and restore using xsave. The kernel will fallback to
4386 enabling legacy floating-point and sse state.
4388 noxsaveopt [X86] Disables xsaveopt used in saving x86 extended
4389 register states. The kernel will fall back to use
4390 xsave to save the states. By using this parameter,
4391 performance of saving the states is degraded because
4392 xsave doesn't support modified optimization while
4393 xsaveopt supports it on xsaveopt enabled systems.
4395 noxsaves [X86] Disables xsaves and xrstors used in saving and
4396 restoring x86 extended register state in compacted
4397 form of xsave area. The kernel will fall back to use
4398 xsaveopt and xrstor to save and restore the states
4399 in standard form of xsave area. By using this
4400 parameter, xsave area per process might occupy more
4401 memory on xsaves enabled systems.
4403 nr_cpus= [SMP,EARLY] Maximum number of processors that an SMP kernel
4404 could support. nr_cpus=n : n >= 1 limits the kernel to
4405 support 'n' processors. It could be larger than the
4406 number of already plugged CPU during bootup, later in
4407 runtime you can physically add extra cpu until it reaches
4408 n. So during boot up some boot time memory for per-cpu
4409 variables need be pre-allocated for later physical cpu
4412 nr_uarts= [SERIAL] maximum number of UARTs to be registered.
4414 numa=off [KNL, ARM64, PPC, RISCV, SPARC, X86, EARLY]
4415 Disable NUMA, Only set up a single NUMA node
4416 spanning all memory.
4418 numa=fake=<size>[MG]
4419 [KNL, ARM64, RISCV, X86, EARLY]
4420 If given as a memory unit, fills all system RAM with
4421 nodes of size interleaved over physical nodes.
4424 [KNL, ARM64, RISCV, X86, EARLY]
4425 If given as an integer, fills all system RAM with N
4426 fake nodes interleaved over physical nodes.
4429 [KNL, ARM64, RISCV, X86, EARLY]
4430 If given as an integer followed by 'U', it will
4431 divide each physical node into N emulated nodes.
4433 numa=noacpi [X86] Don't parse the SRAT table for NUMA setup
4435 numa=nohmat [X86] Don't parse the HMAT table for NUMA setup, or
4436 soft-reserved memory partitioning.
4438 numa_balancing= [KNL,ARM64,PPC,RISCV,S390,X86] Enable or disable automatic
4440 Allowed values are enable and disable
4442 numa_zonelist_order= [KNL, BOOT] Select zonelist order for NUMA.
4443 'node', 'default' can be specified
4444 This can be set from sysctl after boot.
4445 See Documentation/admin-guide/sysctl/vm.rst for details.
4447 ohci1394_dma=early [HW,EARLY] enable debugging via the ohci1394 driver.
4448 See Documentation/core-api/debugging-via-ohci1394.rst for more
4451 olpc_ec_timeout= [OLPC] ms delay when issuing EC commands
4452 Rather than timing out after 20 ms if an EC
4453 command is not properly ACKed, override the length
4454 of the timeout. We have interrupts disabled while
4455 waiting for the ACK, so if this is set too high
4456 interrupts *may* be lost!
4458 omap_mux= [OMAP] Override bootloader pin multiplexing.
4459 Format: <mux_mode0.mode_name=value>...
4460 For example, to override I2C bus2:
4461 omap_mux=i2c2_scl.i2c2_scl=0x100,i2c2_sda.i2c2_sda=0x100
4463 onenand.bdry= [HW,MTD] Flex-OneNAND Boundary Configuration
4465 Format: [die0_boundary][,die0_lock][,die1_boundary][,die1_lock]
4467 boundary - index of last SLC block on Flex-OneNAND.
4468 The remaining blocks are configured as MLC blocks.
4469 lock - Configure if Flex-OneNAND boundary should be locked.
4470 Once locked, the boundary cannot be changed.
4471 1 indicates lock status, 0 indicates unlock status.
4473 oops=panic [KNL,EARLY]
4474 Always panic on oopses. Default is to just kill the
4475 process, but there is a small probability of
4476 deadlocking the machine.
4477 This will also cause panics on machine check exceptions.
4478 Useful together with panic=30 to trigger a reboot.
4481 [KNL] Boolean flag to control whether the page allocator
4482 should randomize its free lists. This parameter can be
4483 used to enable/disable page randomization. The state of
4484 the flag can be read from sysfs at:
4485 /sys/module/page_alloc/parameters/shuffle.
4486 This parameter is only available if CONFIG_SHUFFLE_PAGE_ALLOCATOR=y.
4488 page_owner= [KNL,EARLY] Boot-time page_owner enabling option.
4489 Storage of the information about who allocated
4490 each page is disabled in default. With this switch,
4492 on: enable the feature
4494 page_poison= [KNL,EARLY] Boot-time parameter changing the state of
4495 poisoning on the buddy allocator, available with
4496 CONFIG_PAGE_POISONING=y.
4497 off: turn off poisoning (default)
4498 on: turn on poisoning
4500 page_reporting.page_reporting_order=
4501 [KNL] Minimal page reporting order
4503 Adjust the minimal page reporting order. The page
4504 reporting is disabled when it exceeds MAX_PAGE_ORDER.
4506 panic= [KNL] Kernel behaviour on panic: delay <timeout>
4507 timeout > 0: seconds before rebooting
4508 timeout = 0: wait forever
4509 timeout < 0: reboot immediately
4512 panic_on_taint= [KNL,EARLY]
4513 Bitmask for conditionally calling panic() in add_taint()
4514 Format: <hex>[,nousertaint]
4515 Hexadecimal bitmask representing the set of TAINT flags
4516 that will cause the kernel to panic when add_taint() is
4517 called with any of the flags in this set.
4518 The optional switch "nousertaint" can be utilized to
4519 prevent userspace forced crashes by writing to sysctl
4520 /proc/sys/kernel/tainted any flagset matching with the
4521 bitmask set on panic_on_taint.
4522 See Documentation/admin-guide/tainted-kernels.rst for
4523 extra details on the taint flags that users can pick
4524 to compose the bitmask to assign to panic_on_taint.
4526 panic_on_warn=1 panic() instead of WARN(). Useful to cause kdump
4529 panic_print= Bitmask for printing system info when panic happens.
4530 User can chose combination of the following bits:
4531 bit 0: print all tasks info
4532 bit 1: print system memory info
4533 bit 2: print timer info
4534 bit 3: print locks info if CONFIG_LOCKDEP is on
4535 bit 4: print ftrace buffer
4536 bit 5: print all printk messages in buffer
4537 bit 6: print all CPUs backtrace (if available in the arch)
4538 bit 7: print only tasks in uninterruptible (blocked) state
4539 *Be aware* that this option may print a _lot_ of lines,
4540 so there are risks of losing older messages in the log.
4541 Use this option carefully, maybe worth to setup a
4542 bigger log buffer with "log_buf_len" along with this.
4544 parkbd.port= [HW] Parallel port number the keyboard adapter is
4545 connected to, default is 0.
4547 parkbd.mode= [HW] Parallel port keyboard adapter mode of operation,
4548 0 for XT, 1 for AT (default is AT).
4551 parport= [HW,PPT] Specify parallel ports. 0 disables.
4552 Format: { 0 | auto | 0xBBB[,IRQ[,DMA]] }
4553 Use 'auto' to force the driver to use any
4554 IRQ/DMA settings detected (the default is to
4555 ignore detected IRQ/DMA settings because of
4556 possible conflicts). You can specify the base
4557 address, IRQ, and DMA settings; IRQ and DMA
4558 should be numbers, or 'auto' (for using detected
4559 settings on that particular port), or 'nofifo'
4560 (to avoid using a FIFO even if it is detected).
4561 Parallel ports are assigned in the order they
4562 are specified on the command line, starting
4565 parport_init_mode= [HW,PPT]
4566 Configure VIA parallel port to operate in
4567 a specific mode. This is necessary on Pegasos
4568 computer where firmware has no options for setting
4569 up parallel port mode and sets it to spp.
4570 Currently this function knows 686a and 8231 chips.
4571 Format: [spp|ps2|epp|ecp|ecpepp]
4573 pata_legacy.all= [HW,LIBATA]
4575 Set to non-zero to probe primary and secondary ISA
4576 port ranges on PCI systems where no PCI PATA device
4577 has been found at either range. Disabled by default.
4579 pata_legacy.autospeed= [HW,LIBATA]
4581 Set to non-zero if a chip is present that snoops speed
4582 changes. Disabled by default.
4584 pata_legacy.ht6560a= [HW,LIBATA]
4586 Set to 1, 2, or 3 for HT 6560A on the primary channel,
4587 the secondary channel, or both channels respectively.
4588 Disabled by default.
4590 pata_legacy.ht6560b= [HW,LIBATA]
4592 Set to 1, 2, or 3 for HT 6560B on the primary channel,
4593 the secondary channel, or both channels respectively.
4594 Disabled by default.
4596 pata_legacy.iordy_mask= [HW,LIBATA]
4598 IORDY enable mask. Set individual bits to allow IORDY
4599 for the respective channel. Bit 0 is for the first
4600 legacy channel handled by this driver, bit 1 is for
4601 the second channel, and so on. The sequence will often
4602 correspond to the primary legacy channel, the secondary
4603 legacy channel, and so on, but the handling of a PCI
4604 bus and the use of other driver options may interfere
4605 with the sequence. By default IORDY is allowed across
4608 pata_legacy.opti82c46x= [HW,LIBATA]
4610 Set to 1, 2, or 3 for Opti 82c611A on the primary
4611 channel, the secondary channel, or both channels
4612 respectively. Disabled by default.
4614 pata_legacy.opti82c611a= [HW,LIBATA]
4616 Set to 1, 2, or 3 for Opti 82c465MV on the primary
4617 channel, the secondary channel, or both channels
4618 respectively. Disabled by default.
4620 pata_legacy.pio_mask= [HW,LIBATA]
4622 PIO mode mask for autospeed devices. Set individual
4623 bits to allow the use of the respective PIO modes.
4624 Bit 0 is for mode 0, bit 1 is for mode 1, and so on.
4625 All modes allowed by default.
4627 pata_legacy.probe_all= [HW,LIBATA]
4629 Set to non-zero to probe tertiary and further ISA
4630 port ranges on PCI systems. Disabled by default.
4632 pata_legacy.probe_mask= [HW,LIBATA]
4634 Probe mask for legacy ISA PATA ports. Depending on
4635 platform configuration and the use of other driver
4636 options up to 6 legacy ports are supported: 0x1f0,
4637 0x170, 0x1e8, 0x168, 0x1e0, 0x160, however probing
4638 of individual ports can be disabled by setting the
4639 corresponding bits in the mask to 1. Bit 0 is for
4640 the first port in the list above (0x1f0), and so on.
4641 By default all supported ports are probed.
4643 pata_legacy.qdi= [HW,LIBATA]
4645 Set to non-zero to probe QDI controllers. By default
4646 set to 1 if CONFIG_PATA_QDI_MODULE, 0 otherwise.
4648 pata_legacy.winbond= [HW,LIBATA]
4650 Set to non-zero to probe Winbond controllers. Use
4651 the standard I/O port (0x130) if 1, otherwise the
4652 value given is the I/O port to use (typically 0x1b0).
4653 By default set to 1 if CONFIG_PATA_WINBOND_VLB_MODULE,
4656 pata_platform.pio_mask= [HW,LIBATA]
4658 Supported PIO mode mask. Set individual bits to allow
4659 the use of the respective PIO modes. Bit 0 is for
4660 mode 0, bit 1 is for mode 1, and so on. Mode 0 only
4664 Halt all CPUs after the first oops has been printed for
4665 the specified number of seconds. This is to be used if
4666 your oopses keep scrolling off the screen.
4670 pci=option[,option...] [PCI,EARLY] various PCI subsystem options.
4672 Some options herein operate on a specific device
4673 or a set of devices (<pci_dev>). These are
4674 specified in one of the following formats:
4676 [<domain>:]<bus>:<dev>.<func>[/<dev>.<func>]*
4677 pci:<vendor>:<device>[:<subvendor>:<subdevice>]
4679 Note: the first format specifies a PCI
4680 bus/device/function address which may change
4681 if new hardware is inserted, if motherboard
4682 firmware changes, or due to changes caused
4683 by other kernel parameters. If the
4684 domain is left unspecified, it is
4685 taken to be zero. Optionally, a path
4686 to a device through multiple device/function
4687 addresses can be specified after the base
4688 address (this is more robust against
4689 renumbering issues). The second format
4690 selects devices using IDs from the
4691 configuration space which may match multiple
4692 devices in the system.
4694 earlydump dump PCI config space before the kernel
4696 off [X86] don't probe for the PCI bus
4697 bios [X86-32] force use of PCI BIOS, don't access
4698 the hardware directly. Use this if your machine
4699 has a non-standard PCI host bridge.
4700 nobios [X86-32] disallow use of PCI BIOS, only direct
4701 hardware access methods are allowed. Use this
4702 if you experience crashes upon bootup and you
4703 suspect they are caused by the BIOS.
4704 conf1 [X86] Force use of PCI Configuration Access
4705 Mechanism 1 (config address in IO port 0xCF8,
4706 data in IO port 0xCFC, both 32-bit).
4707 conf2 [X86] Force use of PCI Configuration Access
4708 Mechanism 2 (IO port 0xCF8 is an 8-bit port for
4709 the function, IO port 0xCFA, also 8-bit, sets
4710 bus number. The config space is then accessed
4711 through ports 0xC000-0xCFFF).
4712 See http://wiki.osdev.org/PCI for more info
4713 on the configuration access mechanisms.
4714 noaer [PCIE] If the PCIEAER kernel config parameter is
4715 enabled, this kernel boot option can be used to
4716 disable the use of PCIE advanced error reporting.
4717 nodomains [PCI] Disable support for multiple PCI
4718 root domains (aka PCI segments, in ACPI-speak).
4719 nommconf [X86] Disable use of MMCONFIG for PCI
4721 check_enable_amd_mmconf [X86] check for and enable
4722 properly configured MMIO access to PCI
4723 config space on AMD family 10h CPU
4724 nomsi [MSI] If the PCI_MSI kernel config parameter is
4725 enabled, this kernel boot option can be used to
4726 disable the use of MSI interrupts system-wide.
4727 noioapicquirk [APIC] Disable all boot interrupt quirks.
4728 Safety option to keep boot IRQs enabled. This
4729 should never be necessary.
4730 ioapicreroute [APIC] Enable rerouting of boot IRQs to the
4731 primary IO-APIC for bridges that cannot disable
4732 boot IRQs. This fixes a source of spurious IRQs
4733 when the system masks IRQs.
4734 noioapicreroute [APIC] Disable workaround that uses the
4735 boot IRQ equivalent of an IRQ that connects to
4736 a chipset where boot IRQs cannot be disabled.
4737 The opposite of ioapicreroute.
4738 biosirq [X86-32] Use PCI BIOS calls to get the interrupt
4739 routing table. These calls are known to be buggy
4740 on several machines and they hang the machine
4741 when used, but on other computers it's the only
4742 way to get the interrupt routing table. Try
4743 this option if the kernel is unable to allocate
4744 IRQs or discover secondary PCI buses on your
4746 rom [X86] Assign address space to expansion ROMs.
4747 Use with caution as certain devices share
4748 address decoders between ROMs and other
4750 norom [X86] Do not assign address space to
4751 expansion ROMs that do not already have
4752 BIOS assigned address ranges.
4753 nobar [X86] Do not assign address space to the
4754 BARs that weren't assigned by the BIOS.
4755 irqmask=0xMMMM [X86] Set a bit mask of IRQs allowed to be
4756 assigned automatically to PCI devices. You can
4757 make the kernel exclude IRQs of your ISA cards
4759 pirqaddr=0xAAAAA [X86] Specify the physical address
4760 of the PIRQ table (normally generated
4761 by the BIOS) if it is outside the
4762 F0000h-100000h range.
4763 lastbus=N [X86] Scan all buses thru bus #N. Can be
4764 useful if the kernel is unable to find your
4765 secondary buses and you want to tell it
4766 explicitly which ones they are.
4767 assign-busses [X86] Always assign all PCI bus
4768 numbers ourselves, overriding
4769 whatever the firmware may have done.
4770 usepirqmask [X86] Honor the possible IRQ mask stored
4771 in the BIOS $PIR table. This is needed on
4772 some systems with broken BIOSes, notably
4773 some HP Pavilion N5400 and Omnibook XE3
4774 notebooks. This will have no effect if ACPI
4775 IRQ routing is enabled.
4776 noacpi [X86] Do not use ACPI for IRQ routing
4777 or for PCI scanning.
4778 use_crs [X86] Use PCI host bridge window information
4779 from ACPI. On BIOSes from 2008 or later, this
4780 is enabled by default. If you need to use this,
4781 please report a bug.
4782 nocrs [X86] Ignore PCI host bridge windows from ACPI.
4783 If you need to use this, please report a bug.
4784 use_e820 [X86] Use E820 reservations to exclude parts of
4785 PCI host bridge windows. This is a workaround
4786 for BIOS defects in host bridge _CRS methods.
4787 If you need to use this, please report a bug to
4788 <linux-pci@vger.kernel.org>.
4789 no_e820 [X86] Ignore E820 reservations for PCI host
4790 bridge windows. This is the default on modern
4791 hardware. If you need to use this, please report
4792 a bug to <linux-pci@vger.kernel.org>.
4793 routeirq Do IRQ routing for all PCI devices.
4794 This is normally done in pci_enable_device(),
4795 so this option is a temporary workaround
4796 for broken drivers that don't call it.
4797 skip_isa_align [X86] do not align io start addr, so can
4798 handle more pci cards
4799 noearly [X86] Don't do any early type 1 scanning.
4800 This might help on some broken boards which
4801 machine check when some devices' config space
4802 is read. But various workarounds are disabled
4803 and some IOMMU drivers will not work.
4804 bfsort Sort PCI devices into breadth-first order.
4805 This sorting is done to get a device
4806 order compatible with older (<= 2.4) kernels.
4807 nobfsort Don't sort PCI devices into breadth-first order.
4808 pcie_bus_tune_off Disable PCIe MPS (Max Payload Size)
4809 tuning and use the BIOS-configured MPS defaults.
4810 pcie_bus_safe Set every device's MPS to the largest value
4811 supported by all devices below the root complex.
4812 pcie_bus_perf Set device MPS to the largest allowable MPS
4813 based on its parent bus. Also set MRRS (Max
4814 Read Request Size) to the largest supported
4815 value (no larger than the MPS that the device
4816 or bus can support) for best performance.
4817 pcie_bus_peer2peer Set every device's MPS to 128B, which
4818 every device is guaranteed to support. This
4819 configuration allows peer-to-peer DMA between
4820 any pair of devices, possibly at the cost of
4821 reduced performance. This also guarantees
4822 that hot-added devices will work.
4823 cbiosize=nn[KMG] The fixed amount of bus space which is
4824 reserved for the CardBus bridge's IO window.
4825 The default value is 256 bytes.
4826 cbmemsize=nn[KMG] The fixed amount of bus space which is
4827 reserved for the CardBus bridge's memory
4828 window. The default value is 64 megabytes.
4831 [<order of align>@]<pci_dev>[; ...]
4832 Specifies alignment and device to reassign
4833 aligned memory resources. How to
4834 specify the device is described above.
4835 If <order of align> is not specified,
4836 PAGE_SIZE is used as alignment.
4837 A PCI-PCI bridge can be specified if resource
4838 windows need to be expanded.
4839 To specify the alignment for several
4840 instances of a device, the PCI vendor,
4841 device, subvendor, and subdevice may be
4842 specified, e.g., 12@pci:8086:9c22:103c:198f
4843 for 4096-byte alignment.
4844 ecrc= Enable/disable PCIe ECRC (transaction layer
4845 end-to-end CRC checking). Only effective if
4846 OS has native AER control (either granted by
4847 ACPI _OSC or forced via "pcie_ports=native")
4848 bios: Use BIOS/firmware settings. This is the
4852 hpiosize=nn[KMG] The fixed amount of bus space which is
4853 reserved for hotplug bridge's IO window.
4854 Default size is 256 bytes.
4855 hpmmiosize=nn[KMG] The fixed amount of bus space which is
4856 reserved for hotplug bridge's MMIO window.
4857 Default size is 2 megabytes.
4858 hpmmioprefsize=nn[KMG] The fixed amount of bus space which is
4859 reserved for hotplug bridge's MMIO_PREF window.
4860 Default size is 2 megabytes.
4861 hpmemsize=nn[KMG] The fixed amount of bus space which is
4862 reserved for hotplug bridge's MMIO and
4864 Default size is 2 megabytes.
4865 hpbussize=nn The minimum amount of additional bus numbers
4866 reserved for buses below a hotplug bridge.
4868 realloc= Enable/disable reallocating PCI bridge resources
4869 if allocations done by BIOS are too small to
4870 accommodate resources required by all child
4872 off: Turn realloc off
4874 realloc same as realloc=on
4875 noari do not use PCIe ARI.
4876 noats [PCIE, Intel-IOMMU, AMD-IOMMU]
4877 do not use PCIe ATS (and IOMMU device IOTLB).
4878 pcie_scan_all Scan all possible PCIe devices. Otherwise we
4879 only look for one device below a PCIe downstream
4881 big_root_window Try to add a big 64bit memory window to the PCIe
4882 root complex on AMD CPUs. Some GFX hardware
4883 can resize a BAR to allow access to all VRAM.
4884 Adding the window is slightly risky (it may
4885 conflict with unreported devices), so this
4887 disable_acs_redir=<pci_dev>[; ...]
4888 Specify one or more PCI devices (in the format
4889 specified above) separated by semicolons.
4890 Each device specified will have the PCI ACS
4891 redirect capabilities forced off which will
4892 allow P2P traffic between devices through
4893 bridges without forcing it upstream. Note:
4894 this removes isolation between devices and
4895 may put more devices in an IOMMU group.
4898 <ACS flags>@<pci_dev>[; ...]
4899 Specify one or more PCI devices (in the format
4900 specified above) optionally prepended with flags
4901 and separated by semicolons. The respective
4902 capabilities will be enabled, disabled or
4903 unchanged based on what is specified in
4906 ACS Flags is defined as follows:
4907 bit-0 : ACS Source Validation
4908 bit-1 : ACS Translation Blocking
4909 bit-2 : ACS P2P Request Redirect
4910 bit-3 : ACS P2P Completion Redirect
4911 bit-4 : ACS Upstream Forwarding
4912 bit-5 : ACS P2P Egress Control
4913 bit-6 : ACS Direct Translated P2P
4914 Each bit can be marked as:
4915 '0' – force disabled
4916 '1' – force enabled
4919 pci=config_acs=10x@pci:0:0
4920 would configure all devices that support
4921 ACS to enable P2P Request Redirect, disable
4922 Translation Blocking, and leave Source
4923 Validation unchanged from whatever power-up
4924 or firmware set it to.
4926 Note: this may remove isolation between devices
4927 and may put more devices in an IOMMU group.
4928 force_floating [S390] Force usage of floating interrupts.
4929 nomio [S390] Do not use MIO instructions.
4930 norid [S390] ignore the RID field and force use of
4931 one PCI domain per PCI function
4932 notph [PCIE] If the PCIE_TPH kernel config parameter
4933 is enabled, this kernel boot option can be used
4934 to disable PCIe TLP Processing Hints support
4937 pcie_aspm= [PCIE] Forcibly enable or ignore PCIe Active State Power
4939 off Don't touch ASPM configuration at all. Leave any
4940 configuration done by firmware unchanged.
4941 force Enable ASPM even on devices that claim not to support it.
4942 WARNING: Forcing ASPM on may cause system lockups.
4944 pcie_ports= [PCIE] PCIe port services handling:
4945 native Use native PCIe services (PME, AER, DPC, PCIe hotplug)
4946 even if the platform doesn't give the OS permission to
4947 use them. This may cause conflicts if the platform
4948 also tries to use these services.
4949 dpc-native Use native PCIe service for DPC only. May
4950 cause conflicts if firmware uses AER or DPC.
4951 compat Disable native PCIe services (PME, AER, DPC, PCIe
4954 pcie_port_pm= [PCIE] PCIe port power management handling:
4955 off Disable power management of all PCIe ports
4956 force Forcibly enable power management of all PCIe ports
4958 pcie_pme= [PCIE,PM] Native PCIe PME signaling options:
4959 nomsi Do not use MSI for native PCIe PME signaling (this makes
4960 all PCIe root ports use INTx for all services).
4962 pcmv= [HW,PCMCIA] BadgePAD 4
4966 Keep all power-domains already enabled by bootloader on,
4967 even if no driver has claimed them. This is useful
4968 for debug and development, but should not be
4969 needed on a platform with proper driver support.
4971 pdcchassis= [PARISC,HW] Disable/Enable PDC Chassis Status codes at
4974 See arch/parisc/kernel/pdc_chassis.c
4976 percpu_alloc= [MM,EARLY]
4977 Select which percpu first chunk allocator to use.
4978 Currently supported values are "embed" and "page".
4979 Archs may support subset or none of the selections.
4980 See comments in mm/percpu.c for details on each
4981 allocator. This parameter is primarily for debugging
4982 and performance comparison.
4984 pirq= [SMP,APIC] Manual mp-table setup
4985 See Documentation/arch/x86/i386/IO-APIC.rst.
4987 plip= [PPT,NET] Parallel port network link
4988 Format: { parport<nr> | timid | 0 }
4989 See also Documentation/admin-guide/parport.rst.
4991 pmtmr= [X86] Manual setup of pmtmr I/O Port.
4992 Override pmtimer IOPort with a hex value.
4995 pmu_override= [PPC] Override the PMU.
4996 This option takes over the PMU facility, so it is no
4997 longer usable by perf. Setting this option starts the
4998 PMU counters by setting MMCR0 to 0 (the FC bit is
4999 cleared). If a number is given, then MMCR1 is set to
5000 that number, otherwise (e.g., 'pmu_override=on'), MMCR1
5003 pm_debug_messages [SUSPEND,KNL]
5004 Enable suspend/resume debug messages during boot up.
5007 Enable PNP debug messages (depends on the
5008 CONFIG_PNP_DEBUG_MESSAGES option). Change at run-time
5009 via /sys/module/pnp/parameters/debug. We always show
5010 current resource usage; turning this on also shows
5011 possible settings and some assignment information.
5017 { on | off | curr | res | no-curr | no-res }
5020 [ISAPNP] Exclude IRQs for the autoconfiguration
5023 [ISAPNP] Exclude DMAs for the autoconfiguration
5025 pnp_reserve_io= [ISAPNP] Exclude I/O ports for the autoconfiguration
5026 Ranges are in pairs (I/O port base and size).
5029 [ISAPNP] Exclude memory regions for the
5031 Ranges are in pairs (memory base and size).
5033 ports= [IP_VS_FTP] IPVS ftp helper module
5035 Up to 8 (IP_VS_APP_MAX_PORTS) ports
5037 Format: <port>,<port>....
5039 possible_cpus= [SMP,S390,X86]
5040 Format: <unsigned int>
5041 Set the number of possible CPUs, overriding the
5042 regular discovery mechanisms (such as ACPI/FW, etc).
5044 powersave=off [PPC] This option disables power saving features.
5045 It specifically disables cpuidle and sets the
5046 platform machine description specific power_save
5047 function to NULL. On Idle the CPU just reduces
5050 ppc_strict_facility_enable
5051 [PPC,ENABLE] This option catches any kernel floating point,
5052 Altivec, VSX and SPE outside of regions specifically
5053 allowed (eg kernel_enable_fpu()/kernel_disable_fpu()).
5054 There is some performance impact when enabling this.
5058 Disable Hardware Transactional Memory
5061 Select preemption mode if you have CONFIG_PREEMPT_DYNAMIC
5062 none - Limited to cond_resched() calls
5063 voluntary - Limited to cond_resched() and might_sleep() calls
5064 full - Any section that isn't explicitly preempt disabled
5065 can be preempted anytime. Tasks will also yield
5066 contended spinlocks (if the critical section isn't
5067 explicitly preempt disabled beyond the lock itself).
5068 lazy - Scheduler controlled. Similar to full but instead
5069 of preempting the task immediately, the task gets
5070 one HZ tick time to yield itself before the
5071 preemption will be forced. One preemption is when the
5072 task returns to user space.
5074 print-fatal-signals=
5075 [KNL] debug: print fatal signals
5077 If enabled, warn about various signal handling
5078 related application anomalies: too many signals,
5079 too many POSIX.1 timers, fatal signals causing a
5082 If you hit the warning due to signal overflow,
5083 you might want to try "ulimit -i unlimited".
5087 printk.always_kmsg_dump=
5088 Trigger kmsg_dump for cases other than kernel oops or
5090 Format: <bool> (1/Y/y=enable, 0/N/n=disable)
5093 printk.console_no_auto_verbose=
5094 Disable console loglevel raise on oops, panic
5095 or lockdep-detected issues (only if lock debug is on).
5096 With an exception to setups with low baudrate on
5097 serial console, keeping this 0 is a good choice
5098 in order to provide more debug information.
5100 default: 0 (auto_verbose is enabled)
5102 printk.debug_non_panic_cpus=
5103 Allows storing messages from non-panic CPUs into
5104 the printk log buffer during panic(). They are
5105 flushed to consoles by the panic-CPU on
5106 a best-effort basis.
5107 Format: <bool> (1/Y/y=enable, 0/N/n=disable)
5110 printk.devkmsg={on,off,ratelimit}
5111 Control writing to /dev/kmsg.
5112 on - unlimited logging to /dev/kmsg from userspace
5113 off - logging to /dev/kmsg disabled
5114 ratelimit - ratelimit the logging
5117 printk.time= Show timing data prefixed to each printk message line
5118 Format: <bool> (1/Y/y=enable, 0/N/n=disable)
5120 proc_mem.force_override= [KNL]
5121 Format: {always | ptrace | never}
5122 Traditionally /proc/pid/mem allows memory permissions to be
5123 overridden without restrictions. This option may be set to
5124 restrict that. Can be one of:
5125 - 'always': traditional behavior always allows mem overrides.
5126 - 'ptrace': only allow mem overrides for active ptracers.
5127 - 'never': never allow mem overrides.
5128 If not specified, default is the CONFIG_PROC_MEM_* choice.
5130 processor.max_cstate= [HW,ACPI]
5131 Limit processor to maximum C-state
5132 max_cstate=9 overrides any DMI blacklist limit.
5134 processor.nocst [HW,ACPI]
5135 Ignore the _CST method to determine C-states,
5136 instead using the legacy FADT method
5138 profile= [KNL] Enable kernel profiling via /proc/profile
5139 Format: [<profiletype>,]<number>
5140 Param: <profiletype>: "schedule" or "kvm"
5141 [defaults to kernel profiling]
5142 Param: "schedule" - profile schedule points.
5143 Param: "kvm" - profile VM exits.
5144 Param: <number> - step/bucket size as a power of 2 for
5145 statistical time based profiling.
5147 prompt_ramdisk= [RAM] [Deprecated]
5149 prot_virt= [S390] enable hosting protected virtual machines
5150 isolated from the hypervisor (if hardware supports
5151 that). If enabled, the default kernel base address
5152 might be overridden even when Kernel Address Space
5153 Layout Randomization is disabled.
5156 psi= [KNL] Enable or disable pressure stall information
5160 psmouse.proto= [HW,MOUSE] Highest PS2 mouse protocol extension to
5161 probe for; one of (bare|imps|exps|lifebook|any).
5162 psmouse.rate= [HW,MOUSE] Set desired mouse report rate, in reports
5164 psmouse.resetafter= [HW,MOUSE]
5165 Try to reset the device after so many bad packets
5168 [HW,MOUSE] Set desired mouse resolution, in dpi.
5169 psmouse.smartscroll=
5170 [HW,MOUSE] Controls Logitech smartscroll autorepeat.
5171 0 = disabled, 1 = enabled (default).
5173 pstore.backend= Specify the name of the pstore backend to use
5175 pti= [X86-64] Control Page Table Isolation of user and
5176 kernel address spaces. Disabling this feature
5177 removes hardening, but improves performance of
5178 system calls and interrupts.
5180 on - unconditionally enable
5181 off - unconditionally disable
5182 auto - kernel detects whether your CPU model is
5183 vulnerable to issues that PTI mitigates
5185 Not specifying this option is equivalent to pti=auto.
5188 [KNL] Number of legacy pty's. Overwrites compiled-in
5191 quiet [KNL,EARLY] Disable most log messages
5195 radix_hcall_invalidate=on [PPC/PSERIES]
5196 Disable RADIX GTSE feature and use hcall for TLB
5200 See Documentation/admin-guide/md.rst.
5202 ramdisk_size= [RAM] Sizes of RAM disks in kilobytes
5203 See Documentation/admin-guide/blockdev/ramdisk.rst.
5205 ramdisk_start= [RAM] RAM disk image start address
5207 random.trust_cpu=off
5208 [KNL,EARLY] Disable trusting the use of the CPU's
5209 random number generator (if available) to
5210 initialize the kernel's RNG.
5212 random.trust_bootloader=off
5213 [KNL,EARLY] Disable trusting the use of the a seed
5214 passed by the bootloader (if available) to
5215 initialize the kernel's RNG.
5217 randomize_kstack_offset=
5218 [KNL,EARLY] Enable or disable kernel stack offset
5219 randomization, which provides roughly 5 bits of
5220 entropy, frustrating memory corruption attacks
5221 that depend on stack address determinism or
5222 cross-syscall address exposures. This is only
5223 available on architectures that have defined
5224 CONFIG_HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET.
5225 Format: <bool> (1/Y/y=enable, 0/N/n=disable)
5226 Default is CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT.
5228 ras=option[,option,...] [KNL] RAS-specific options
5231 Disable the Correctable Errors Collector,
5232 see CONFIG_RAS_CEC help text.
5234 rcu_nocbs[=cpu-list]
5235 [KNL] The optional argument is a cpu list,
5238 In kernels built with CONFIG_RCU_NOCB_CPU=y,
5239 enable the no-callback CPU mode, which prevents
5240 such CPUs' callbacks from being invoked in
5241 softirq context. Invocation of such CPUs' RCU
5242 callbacks will instead be offloaded to "rcuox/N"
5243 kthreads created for that purpose, where "x" is
5244 "p" for RCU-preempt, "s" for RCU-sched, and "g"
5245 for the kthreads that mediate grace periods; and
5246 "N" is the CPU number. This reduces OS jitter on
5247 the offloaded CPUs, which can be useful for HPC
5248 and real-time workloads. It can also improve
5249 energy efficiency for asymmetric multiprocessors.
5251 If a cpulist is passed as an argument, the specified
5252 list of CPUs is set to no-callback mode from boot.
5254 Otherwise, if the '=' sign and the cpulist
5255 arguments are omitted, no CPU will be set to
5256 no-callback mode from boot but the mode may be
5257 toggled at runtime via cpusets.
5259 Note that this argument takes precedence over
5260 the CONFIG_RCU_NOCB_CPU_DEFAULT_ALL option.
5263 Rather than requiring that offloaded CPUs
5264 (specified by rcu_nocbs= above) explicitly
5265 awaken the corresponding "rcuoN" kthreads,
5266 make these kthreads poll for callbacks.
5267 This improves the real-time response for the
5268 offloaded CPUs by relieving them of the need to
5269 wake up the corresponding kthread, but degrades
5270 energy efficiency by requiring that the kthreads
5271 periodically wake up to do the polling.
5273 rcutree.blimit= [KNL]
5274 Set maximum number of finished RCU callbacks to
5275 process in one batch.
5277 rcutree.csd_lock_suppress_rcu_stall= [KNL]
5278 Do only a one-line RCU CPU stall warning when
5279 there is an ongoing too-long CSD-lock wait.
5281 rcutree.do_rcu_barrier= [KNL]
5282 Request a call to rcu_barrier(). This is
5283 throttled so that userspace tests can safely
5284 hammer on the sysfs variable if they so choose.
5285 If triggered before the RCU grace-period machinery
5286 is fully active, this will error out with EAGAIN.
5288 rcutree.dump_tree= [KNL]
5289 Dump the structure of the rcu_node combining tree
5290 out at early boot. This is used for diagnostic
5291 purposes, to verify correct tree setup.
5293 rcutree.gp_cleanup_delay= [KNL]
5294 Set the number of jiffies to delay each step of
5295 RCU grace-period cleanup.
5297 rcutree.gp_init_delay= [KNL]
5298 Set the number of jiffies to delay each step of
5299 RCU grace-period initialization.
5301 rcutree.gp_preinit_delay= [KNL]
5302 Set the number of jiffies to delay each step of
5303 RCU grace-period pre-initialization, that is,
5304 the propagation of recent CPU-hotplug changes up
5305 the rcu_node combining tree.
5307 rcutree.jiffies_till_first_fqs= [KNL]
5308 Set delay from grace-period initialization to
5309 first attempt to force quiescent states.
5310 Units are jiffies, minimum value is zero,
5311 and maximum value is HZ.
5313 rcutree.jiffies_till_next_fqs= [KNL]
5314 Set delay between subsequent attempts to force
5315 quiescent states. Units are jiffies, minimum
5316 value is one, and maximum value is HZ.
5318 rcutree.jiffies_till_sched_qs= [KNL]
5319 Set required age in jiffies for a
5320 given grace period before RCU starts
5321 soliciting quiescent-state help from
5322 rcu_note_context_switch() and cond_resched().
5323 If not specified, the kernel will calculate
5324 a value based on the most recent settings
5325 of rcutree.jiffies_till_first_fqs
5326 and rcutree.jiffies_till_next_fqs.
5327 This calculated value may be viewed in
5328 rcutree.jiffies_to_sched_qs. Any attempt to set
5329 rcutree.jiffies_to_sched_qs will be cheerfully
5332 rcutree.kthread_prio= [KNL,BOOT]
5333 Set the SCHED_FIFO priority of the RCU per-CPU
5334 kthreads (rcuc/N). This value is also used for
5335 the priority of the RCU boost threads (rcub/N)
5336 and for the RCU grace-period kthreads (rcu_bh,
5337 rcu_preempt, and rcu_sched). If RCU_BOOST is
5338 set, valid values are 1-99 and the default is 1
5339 (the least-favored priority). Otherwise, when
5340 RCU_BOOST is not set, valid values are 0-99 and
5341 the default is zero (non-realtime operation).
5342 When RCU_NOCB_CPU is set, also adjust the
5343 priority of NOCB callback kthreads.
5345 rcutree.nocb_nobypass_lim_per_jiffy= [KNL]
5346 On callback-offloaded (rcu_nocbs) CPUs,
5347 RCU reduces the lock contention that would
5348 otherwise be caused by callback floods through
5349 use of the ->nocb_bypass list. However, in the
5350 common non-flooded case, RCU queues directly to
5351 the main ->cblist in order to avoid the extra
5352 overhead of the ->nocb_bypass list and its lock.
5353 But if there are too many callbacks queued during
5354 a single jiffy, RCU pre-queues the callbacks into
5355 the ->nocb_bypass queue. The definition of "too
5356 many" is supplied by this kernel boot parameter.
5358 rcutree.nohz_full_patience_delay= [KNL]
5359 On callback-offloaded (rcu_nocbs) CPUs, avoid
5360 disturbing RCU unless the grace period has
5361 reached the specified age in milliseconds.
5362 Defaults to zero. Large values will be capped
5363 at five seconds. All values will be rounded down
5364 to the nearest value representable by jiffies.
5366 rcutree.qhimark= [KNL]
5367 Set threshold of queued RCU callbacks beyond which
5368 batch limiting is disabled.
5370 rcutree.qlowmark= [KNL]
5371 Set threshold of queued RCU callbacks below which
5372 batch limiting is re-enabled.
5374 rcutree.qovld= [KNL]
5375 Set threshold of queued RCU callbacks beyond which
5376 RCU's force-quiescent-state scan will aggressively
5377 enlist help from cond_resched() and sched IPIs to
5378 help CPUs more quickly reach quiescent states.
5379 Set to less than zero to make this be set based
5380 on rcutree.qhimark at boot time and to zero to
5381 disable more aggressive help enlistment.
5383 rcutree.rcu_delay_page_cache_fill_msec= [KNL]
5384 Set the page-cache refill delay (in milliseconds)
5385 in response to low-memory conditions. The range
5386 of permitted values is in the range 0:100000.
5388 rcutree.rcu_divisor= [KNL]
5389 Set the shift-right count to use to compute
5390 the callback-invocation batch limit bl from
5391 the number of callbacks queued on this CPU.
5392 The result will be bounded below by the value of
5393 the rcutree.blimit kernel parameter. Every bl
5394 callbacks, the softirq handler will exit in
5395 order to allow the CPU to do other work.
5397 Please note that this callback-invocation batch
5398 limit applies only to non-offloaded callback
5399 invocation. Offloaded callbacks are instead
5400 invoked in the context of an rcuoc kthread, which
5401 scheduler will preempt as it does any other task.
5403 rcutree.rcu_fanout_exact= [KNL]
5404 Disable autobalancing of the rcu_node combining
5405 tree. This is used by rcutorture, and might
5406 possibly be useful for architectures having high
5407 cache-to-cache transfer latencies.
5409 rcutree.rcu_fanout_leaf= [KNL]
5410 Change the number of CPUs assigned to each
5411 leaf rcu_node structure. Useful for very
5412 large systems, which will choose the value 64,
5413 and for NUMA systems with large remote-access
5414 latencies, which will choose a value aligned
5415 with the appropriate hardware boundaries.
5417 rcutree.rcu_min_cached_objs= [KNL]
5418 Minimum number of objects which are cached and
5419 maintained per one CPU. Object size is equal
5420 to PAGE_SIZE. The cache allows to reduce the
5421 pressure to page allocator, also it makes the
5422 whole algorithm to behave better in low memory
5425 rcutree.rcu_nocb_gp_stride= [KNL]
5426 Set the number of NOCB callback kthreads in
5427 each group, which defaults to the square root
5428 of the number of CPUs. Larger numbers reduce
5429 the wakeup overhead on the global grace-period
5430 kthread, but increases that same overhead on
5431 each group's NOCB grace-period kthread.
5433 rcutree.rcu_kick_kthreads= [KNL]
5434 Cause the grace-period kthread to get an extra
5435 wake_up() if it sleeps three times longer than
5436 it should at force-quiescent-state time.
5437 This wake_up() will be accompanied by a
5438 WARN_ONCE() splat and an ftrace_dump().
5440 rcutree.rcu_resched_ns= [KNL]
5441 Limit the time spend invoking a batch of RCU
5442 callbacks to the specified number of nanoseconds.
5443 By default, this limit is checked only once
5444 every 32 callbacks in order to limit the pain
5445 inflicted by local_clock() overhead.
5447 rcutree.rcu_unlock_delay= [KNL]
5448 In CONFIG_RCU_STRICT_GRACE_PERIOD=y kernels,
5449 this specifies an rcu_read_unlock()-time delay
5450 in microseconds. This defaults to zero.
5451 Larger delays increase the probability of
5452 catching RCU pointer leaks, that is, buggy use
5453 of RCU-protected pointers after the relevant
5454 rcu_read_unlock() has completed.
5456 rcutree.sysrq_rcu= [KNL]
5457 Commandeer a sysrq key to dump out Tree RCU's
5458 rcu_node tree with an eye towards determining
5459 why a new grace period has not yet started.
5461 rcutree.use_softirq= [KNL]
5462 If set to zero, move all RCU_SOFTIRQ processing to
5463 per-CPU rcuc kthreads. Defaults to a non-zero
5464 value, meaning that RCU_SOFTIRQ is used by default.
5465 Specify rcutree.use_softirq=0 to use rcuc kthreads.
5467 But note that CONFIG_PREEMPT_RT=y kernels disable
5468 this kernel boot parameter, forcibly setting it
5471 rcutree.enable_rcu_lazy= [KNL]
5472 To save power, batch RCU callbacks and flush after
5473 delay, memory pressure or callback list growing too
5476 rcutree.rcu_normal_wake_from_gp= [KNL]
5477 Reduces a latency of synchronize_rcu() call. This approach
5478 maintains its own track of synchronize_rcu() callers, so it
5479 does not interact with regular callbacks because it does not
5480 use a call_rcu[_hurry]() path. Please note, this is for a
5481 normal grace period.
5485 echo 1 > /sys/module/rcutree/parameters/rcu_normal_wake_from_gp
5486 or pass a boot parameter "rcutree.rcu_normal_wake_from_gp=1"
5490 rcuscale.gp_async= [KNL]
5491 Measure performance of asynchronous
5492 grace-period primitives such as call_rcu().
5494 rcuscale.gp_async_max= [KNL]
5495 Specify the maximum number of outstanding
5496 callbacks per writer thread. When a writer
5497 thread exceeds this limit, it invokes the
5498 corresponding flavor of rcu_barrier() to allow
5499 previously posted callbacks to drain.
5501 rcuscale.gp_exp= [KNL]
5502 Measure performance of expedited synchronous
5503 grace-period primitives.
5505 rcuscale.holdoff= [KNL]
5506 Set test-start holdoff period. The purpose of
5507 this parameter is to delay the start of the
5508 test until boot completes in order to avoid
5511 rcuscale.kfree_by_call_rcu= [KNL]
5512 In kernels built with CONFIG_RCU_LAZY=y, test
5513 call_rcu() instead of kfree_rcu().
5515 rcuscale.kfree_mult= [KNL]
5516 Instead of allocating an object of size kfree_obj,
5517 allocate one of kfree_mult * sizeof(kfree_obj).
5520 rcuscale.kfree_rcu_test= [KNL]
5521 Set to measure performance of kfree_rcu() flooding.
5523 rcuscale.kfree_rcu_test_double= [KNL]
5524 Test the double-argument variant of kfree_rcu().
5525 If this parameter has the same value as
5526 rcuscale.kfree_rcu_test_single, both the single-
5527 and double-argument variants are tested.
5529 rcuscale.kfree_rcu_test_single= [KNL]
5530 Test the single-argument variant of kfree_rcu().
5531 If this parameter has the same value as
5532 rcuscale.kfree_rcu_test_double, both the single-
5533 and double-argument variants are tested.
5535 rcuscale.kfree_nthreads= [KNL]
5536 The number of threads running loops of kfree_rcu().
5538 rcuscale.kfree_alloc_num= [KNL]
5539 Number of allocations and frees done in an iteration.
5541 rcuscale.kfree_loops= [KNL]
5542 Number of loops doing rcuscale.kfree_alloc_num number
5543 of allocations and frees.
5545 rcuscale.minruntime= [KNL]
5546 Set the minimum test run time in seconds. This
5547 does not affect the data-collection interval,
5548 but instead allows better measurement of things
5549 like CPU consumption.
5551 rcuscale.nreaders= [KNL]
5552 Set number of RCU readers. The value -1 selects
5553 N, where N is the number of CPUs. A value
5554 "n" less than -1 selects N-n+1, where N is again
5555 the number of CPUs. For example, -2 selects N
5556 (the number of CPUs), -3 selects N+1, and so on.
5557 A value of "n" less than or equal to -N selects
5560 rcuscale.nwriters= [KNL]
5561 Set number of RCU writers. The values operate
5562 the same as for rcuscale.nreaders.
5563 N, where N is the number of CPUs
5565 rcuscale.scale_type= [KNL]
5566 Specify the RCU implementation to test.
5568 rcuscale.shutdown= [KNL]
5569 Shut the system down after performance tests
5570 complete. This is useful for hands-off automated
5573 rcuscale.verbose= [KNL]
5574 Enable additional printk() statements.
5576 rcuscale.writer_holdoff= [KNL]
5577 Write-side holdoff between grace periods,
5578 in microseconds. The default of zero says
5581 rcuscale.writer_holdoff_jiffies= [KNL]
5582 Additional write-side holdoff between grace
5583 periods, but in jiffies. The default of zero
5586 rcutorture.fqs_duration= [KNL]
5587 Set duration of force_quiescent_state bursts
5590 rcutorture.fqs_holdoff= [KNL]
5591 Set holdoff time within force_quiescent_state bursts
5594 rcutorture.fqs_stutter= [KNL]
5595 Set wait time between force_quiescent_state bursts
5598 rcutorture.fwd_progress= [KNL]
5599 Specifies the number of kthreads to be used
5600 for RCU grace-period forward-progress testing
5601 for the types of RCU supporting this notion.
5602 Defaults to 1 kthread, values less than zero or
5603 greater than the number of CPUs cause the number
5606 rcutorture.fwd_progress_div= [KNL]
5607 Specify the fraction of a CPU-stall-warning
5608 period to do tight-loop forward-progress testing.
5610 rcutorture.fwd_progress_holdoff= [KNL]
5611 Number of seconds to wait between successive
5612 forward-progress tests.
5614 rcutorture.fwd_progress_need_resched= [KNL]
5615 Enclose cond_resched() calls within checks for
5616 need_resched() during tight-loop forward-progress
5619 rcutorture.gp_cond= [KNL]
5620 Use conditional/asynchronous update-side
5621 normal-grace-period primitives, if available.
5623 rcutorture.gp_cond_exp= [KNL]
5624 Use conditional/asynchronous update-side
5625 expedited-grace-period primitives, if available.
5627 rcutorture.gp_cond_full= [KNL]
5628 Use conditional/asynchronous update-side
5629 normal-grace-period primitives that also take
5630 concurrent expedited grace periods into account,
5633 rcutorture.gp_cond_exp_full= [KNL]
5634 Use conditional/asynchronous update-side
5635 expedited-grace-period primitives that also take
5636 concurrent normal grace periods into account,
5639 rcutorture.gp_cond_wi= [KNL]
5640 Nominal wait interval for normal conditional
5641 grace periods (specified by rcutorture's
5642 gp_cond and gp_cond_full module parameters),
5643 in microseconds. The actual wait interval will
5644 be randomly selected to nanosecond granularity up
5645 to this wait interval. Defaults to 16 jiffies,
5646 for example, 16,000 microseconds on a system
5649 rcutorture.gp_cond_wi_exp= [KNL]
5650 Nominal wait interval for expedited conditional
5651 grace periods (specified by rcutorture's
5652 gp_cond_exp and gp_cond_exp_full module
5653 parameters), in microseconds. The actual wait
5654 interval will be randomly selected to nanosecond
5655 granularity up to this wait interval. Defaults to
5658 rcutorture.gp_exp= [KNL]
5659 Use expedited update-side primitives, if available.
5661 rcutorture.gp_normal= [KNL]
5662 Use normal (non-expedited) asynchronous
5663 update-side primitives, if available.
5665 rcutorture.gp_poll= [KNL]
5666 Use polled update-side normal-grace-period
5667 primitives, if available.
5669 rcutorture.gp_poll_exp= [KNL]
5670 Use polled update-side expedited-grace-period
5671 primitives, if available.
5673 rcutorture.gp_poll_full= [KNL]
5674 Use polled update-side normal-grace-period
5675 primitives that also take concurrent expedited
5676 grace periods into account, if available.
5678 rcutorture.gp_poll_exp_full= [KNL]
5679 Use polled update-side expedited-grace-period
5680 primitives that also take concurrent normal
5681 grace periods into account, if available.
5683 rcutorture.gp_poll_wi= [KNL]
5684 Nominal wait interval for normal conditional
5685 grace periods (specified by rcutorture's
5686 gp_poll and gp_poll_full module parameters),
5687 in microseconds. The actual wait interval will
5688 be randomly selected to nanosecond granularity up
5689 to this wait interval. Defaults to 16 jiffies,
5690 for example, 16,000 microseconds on a system
5693 rcutorture.gp_poll_wi_exp= [KNL]
5694 Nominal wait interval for expedited conditional
5695 grace periods (specified by rcutorture's
5696 gp_poll_exp and gp_poll_exp_full module
5697 parameters), in microseconds. The actual wait
5698 interval will be randomly selected to nanosecond
5699 granularity up to this wait interval. Defaults to
5702 rcutorture.gp_sync= [KNL]
5703 Use normal (non-expedited) synchronous
5704 update-side primitives, if available. If all
5705 of rcutorture.gp_cond=, rcutorture.gp_exp=,
5706 rcutorture.gp_normal=, and rcutorture.gp_sync=
5707 are zero, rcutorture acts as if is interpreted
5708 they are all non-zero.
5710 rcutorture.gpwrap_lag= [KNL]
5711 Enable grace-period wrap lag testing. Setting
5712 to false prevents the gpwrap lag test from
5713 running. Default is true.
5715 rcutorture.gpwrap_lag_gps= [KNL]
5716 Set the value for grace-period wrap lag during
5717 active lag testing periods. This controls how many
5718 grace periods differences we tolerate between
5719 rdp and rnp's gp_seq before setting overflow flag.
5720 The default is always set to 8.
5722 rcutorture.gpwrap_lag_cycle_mins= [KNL]
5723 Set the total cycle duration for gpwrap lag
5724 testing in minutes. This is the total time for
5725 one complete cycle of active and inactive
5726 testing periods. Default is 30 minutes.
5728 rcutorture.gpwrap_lag_active_mins= [KNL]
5729 Set the duration for which gpwrap lag is active
5730 within each cycle, in minutes. During this time,
5731 the grace-period wrap lag will be set to the
5732 value specified by gpwrap_lag_gps. Default is
5735 rcutorture.irqreader= [KNL]
5736 Run RCU readers from irq handlers, or, more
5737 accurately, from a timer handler. Not all RCU
5738 flavors take kindly to this sort of thing.
5740 rcutorture.leakpointer= [KNL]
5741 Leak an RCU-protected pointer out of the reader.
5742 This can of course result in splats, and is
5743 intended to test the ability of things like
5744 CONFIG_RCU_STRICT_GRACE_PERIOD=y to detect
5747 rcutorture.n_barrier_cbs= [KNL]
5748 Set callbacks/threads for rcu_barrier() testing.
5750 rcutorture.nfakewriters= [KNL]
5751 Set number of concurrent RCU writers. These just
5752 stress RCU, they don't participate in the actual
5753 test, hence the "fake".
5755 rcutorture.nocbs_nthreads= [KNL]
5756 Set number of RCU callback-offload togglers.
5757 Zero (the default) disables toggling.
5759 rcutorture.nocbs_toggle= [KNL]
5760 Set the delay in milliseconds between successive
5761 callback-offload toggling attempts.
5763 rcutorture.nreaders= [KNL]
5764 Set number of RCU readers. The value -1 selects
5765 N-1, where N is the number of CPUs. A value
5766 "n" less than -1 selects N-n-2, where N is again
5767 the number of CPUs. For example, -2 selects N
5768 (the number of CPUs), -3 selects N+1, and so on.
5770 rcutorture.object_debug= [KNL]
5771 Enable debug-object double-call_rcu() testing.
5773 rcutorture.onoff_holdoff= [KNL]
5774 Set time (s) after boot for CPU-hotplug testing.
5776 rcutorture.onoff_interval= [KNL]
5777 Set time (jiffies) between CPU-hotplug operations,
5778 or zero to disable CPU-hotplug testing.
5780 rcutorture.preempt_duration= [KNL]
5781 Set duration (in milliseconds) of preemptions
5782 by a high-priority FIFO real-time task. Set to
5783 zero (the default) to disable. The CPUs to
5784 preempt are selected randomly from the set that
5785 are online at a given point in time. Races with
5786 CPUs going offline are ignored, with that attempt
5787 at preemption skipped.
5789 rcutorture.preempt_interval= [KNL]
5790 Set interval (in milliseconds, defaulting to one
5791 second) between preemptions by a high-priority
5792 FIFO real-time task. This delay is mediated
5793 by an hrtimer and is further fuzzed to avoid
5794 inadvertent synchronizations.
5796 rcutorture.read_exit_burst= [KNL]
5797 The number of times in a given read-then-exit
5798 episode that a set of read-then-exit kthreads
5801 rcutorture.read_exit_delay= [KNL]
5802 The delay, in seconds, between successive
5803 read-then-exit testing episodes.
5805 rcutorture.reader_flavor= [KNL]
5806 A bit mask indicating which readers to use.
5807 If there is more than one bit set, the readers
5808 are entered from low-order bit up, and are
5809 exited in the opposite order. For SRCU, the
5810 0x1 bit is normal readers, 0x2 NMI-safe readers,
5811 and 0x4 light-weight readers.
5813 rcutorture.shuffle_interval= [KNL]
5814 Set task-shuffle interval (s). Shuffling tasks
5815 allows some CPUs to go into dyntick-idle mode
5816 during the rcutorture test.
5818 rcutorture.shutdown_secs= [KNL]
5819 Set time (s) after boot system shutdown. This
5820 is useful for hands-off automated testing.
5822 rcutorture.stall_cpu= [KNL]
5823 Duration of CPU stall (s) to test RCU CPU stall
5824 warnings, zero to disable.
5826 rcutorture.stall_cpu_block= [KNL]
5827 Sleep while stalling if set. This will result
5828 in warnings from preemptible RCU in addition to
5829 any other stall-related activity. Note that
5830 in kernels built with CONFIG_PREEMPTION=n and
5831 CONFIG_PREEMPT_COUNT=y, this parameter will
5832 cause the CPU to pass through a quiescent state.
5833 Given CONFIG_PREEMPTION=n, this will suppress
5834 RCU CPU stall warnings, but will instead result
5835 in scheduling-while-atomic splats.
5837 Use of this module parameter results in splats.
5840 rcutorture.stall_cpu_holdoff= [KNL]
5841 Time to wait (s) after boot before inducing stall.
5843 rcutorture.stall_cpu_irqsoff= [KNL]
5844 Disable interrupts while stalling if set, but only
5845 on the first stall in the set.
5847 rcutorture.stall_cpu_repeat= [KNL]
5848 Number of times to repeat the stall sequence,
5849 so that rcutorture.stall_cpu_repeat=3 will result
5850 in four stall sequences.
5852 rcutorture.stall_gp_kthread= [KNL]
5853 Duration (s) of forced sleep within RCU
5854 grace-period kthread to test RCU CPU stall
5855 warnings, zero to disable. If both stall_cpu
5856 and stall_gp_kthread are specified, the
5857 kthread is starved first, then the CPU.
5859 rcutorture.stat_interval= [KNL]
5860 Time (s) between statistics printk()s.
5862 rcutorture.stutter= [KNL]
5863 Time (s) to stutter testing, for example, specifying
5864 five seconds causes the test to run for five seconds,
5865 wait for five seconds, and so on. This tests RCU's
5866 ability to transition abruptly to and from idle.
5868 rcutorture.test_boost= [KNL]
5869 Test RCU priority boosting? 0=no, 1=maybe, 2=yes.
5870 "Maybe" means test if the RCU implementation
5871 under test support RCU priority boosting.
5873 rcutorture.test_boost_duration= [KNL]
5874 Duration (s) of each individual boost test.
5876 rcutorture.test_boost_holdoff= [KNL]
5877 Holdoff time (s) from start of test to the start
5878 of RCU priority-boost testing. Defaults to zero,
5879 that is, no holdoff.
5881 rcutorture.test_boost_interval= [KNL]
5882 Interval (s) between each boost test.
5884 rcutorture.test_no_idle_hz= [KNL]
5885 Test RCU's dyntick-idle handling. See also the
5886 rcutorture.shuffle_interval parameter.
5888 rcutorture.torture_type= [KNL]
5889 Specify the RCU implementation to test.
5891 rcutorture.verbose= [KNL]
5892 Enable additional printk() statements.
5894 rcupdate.rcu_cpu_stall_ftrace_dump= [KNL]
5895 Dump ftrace buffer after reporting RCU CPU
5898 rcupdate.rcu_cpu_stall_notifiers= [KNL]
5899 Provide RCU CPU stall notifiers, but see the
5900 warnings in the RCU_CPU_STALL_NOTIFIER Kconfig
5901 option's help text. TL;DR: You almost certainly
5902 do not want rcupdate.rcu_cpu_stall_notifiers.
5904 rcupdate.rcu_cpu_stall_suppress= [KNL]
5905 Suppress RCU CPU stall warning messages.
5907 rcupdate.rcu_cpu_stall_suppress_at_boot= [KNL]
5908 Suppress RCU CPU stall warning messages and
5909 rcutorture writer stall warnings that occur
5910 during early boot, that is, during the time
5911 before the init task is spawned.
5913 rcupdate.rcu_cpu_stall_timeout= [KNL]
5914 Set timeout for RCU CPU stall warning messages.
5915 The value is in seconds and the maximum allowed
5916 value is 300 seconds.
5918 rcupdate.rcu_exp_cpu_stall_timeout= [KNL]
5919 Set timeout for expedited RCU CPU stall warning
5920 messages. The value is in milliseconds
5921 and the maximum allowed value is 21000
5922 milliseconds. Please note that this value is
5923 adjusted to an arch timer tick resolution.
5924 Setting this to zero causes the value from
5925 rcupdate.rcu_cpu_stall_timeout to be used (after
5926 conversion from seconds to milliseconds).
5928 rcupdate.rcu_cpu_stall_cputime= [KNL]
5929 Provide statistics on the cputime and count of
5930 interrupts and tasks during the sampling period. For
5931 multiple continuous RCU stalls, all sampling periods
5932 begin at half of the first RCU stall timeout.
5934 rcupdate.rcu_exp_stall_task_details= [KNL]
5935 Print stack dumps of any tasks blocking the
5936 current expedited RCU grace period during an
5937 expedited RCU CPU stall warning.
5939 rcupdate.rcu_expedited= [KNL]
5940 Use expedited grace-period primitives, for
5941 example, synchronize_rcu_expedited() instead
5942 of synchronize_rcu(). This reduces latency,
5943 but can increase CPU utilization, degrade
5944 real-time latency, and degrade energy efficiency.
5945 No effect on CONFIG_TINY_RCU kernels.
5947 rcupdate.rcu_normal= [KNL]
5948 Use only normal grace-period primitives,
5949 for example, synchronize_rcu() instead of
5950 synchronize_rcu_expedited(). This improves
5951 real-time latency, CPU utilization, and
5952 energy efficiency, but can expose users to
5953 increased grace-period latency. This parameter
5954 overrides rcupdate.rcu_expedited. No effect on
5955 CONFIG_TINY_RCU kernels.
5957 rcupdate.rcu_normal_after_boot= [KNL]
5958 Once boot has completed (that is, after
5959 rcu_end_inkernel_boot() has been invoked), use
5960 only normal grace-period primitives. No effect
5961 on CONFIG_TINY_RCU kernels.
5963 But note that CONFIG_PREEMPT_RT=y kernels enables
5964 this kernel boot parameter, forcibly setting
5965 it to the value one, that is, converting any
5966 post-boot attempt at an expedited RCU grace
5967 period to instead use normal non-expedited
5968 grace-period processing.
5970 rcupdate.rcu_task_collapse_lim= [KNL]
5971 Set the maximum number of callbacks present
5972 at the beginning of a grace period that allows
5973 the RCU Tasks flavors to collapse back to using
5974 a single callback queue. This switching only
5975 occurs when rcupdate.rcu_task_enqueue_lim is
5976 set to the default value of -1.
5978 rcupdate.rcu_task_contend_lim= [KNL]
5979 Set the minimum number of callback-queuing-time
5980 lock-contention events per jiffy required to
5981 cause the RCU Tasks flavors to switch to per-CPU
5982 callback queuing. This switching only occurs
5983 when rcupdate.rcu_task_enqueue_lim is set to
5984 the default value of -1.
5986 rcupdate.rcu_task_enqueue_lim= [KNL]
5987 Set the number of callback queues to use for the
5988 RCU Tasks family of RCU flavors. The default
5989 of -1 allows this to be automatically (and
5990 dynamically) adjusted. This parameter is intended
5993 rcupdate.rcu_task_ipi_delay= [KNL]
5994 Set time in jiffies during which RCU tasks will
5995 avoid sending IPIs, starting with the beginning
5996 of a given grace period. Setting a large
5997 number avoids disturbing real-time workloads,
5998 but lengthens grace periods.
6000 rcupdate.rcu_task_lazy_lim= [KNL]
6001 Number of callbacks on a given CPU that will
6002 cancel laziness on that CPU. Use -1 to disable
6003 cancellation of laziness, but be advised that
6004 doing so increases the danger of OOM due to
6007 rcupdate.rcu_task_stall_info= [KNL]
6008 Set initial timeout in jiffies for RCU task stall
6009 informational messages, which give some indication
6010 of the problem for those not patient enough to
6011 wait for ten minutes. Informational messages are
6012 only printed prior to the stall-warning message
6013 for a given grace period. Disable with a value
6014 less than or equal to zero. Defaults to ten
6015 seconds. A change in value does not take effect
6016 until the beginning of the next grace period.
6018 rcupdate.rcu_task_stall_info_mult= [KNL]
6019 Multiplier for time interval between successive
6020 RCU task stall informational messages for a given
6021 RCU tasks grace period. This value is clamped
6022 to one through ten, inclusive. It defaults to
6023 the value three, so that the first informational
6024 message is printed 10 seconds into the grace
6025 period, the second at 40 seconds, the third at
6026 160 seconds, and then the stall warning at 600
6027 seconds would prevent a fourth at 640 seconds.
6029 rcupdate.rcu_task_stall_timeout= [KNL]
6030 Set timeout in jiffies for RCU task stall
6031 warning messages. Disable with a value less
6032 than or equal to zero. Defaults to ten minutes.
6033 A change in value does not take effect until
6034 the beginning of the next grace period.
6036 rcupdate.rcu_tasks_lazy_ms= [KNL]
6037 Set timeout in milliseconds RCU Tasks asynchronous
6038 callback batching for call_rcu_tasks().
6039 A negative value will take the default. A value
6040 of zero will disable batching. Batching is
6041 always disabled for synchronize_rcu_tasks().
6043 rcupdate.rcu_tasks_trace_lazy_ms= [KNL]
6044 Set timeout in milliseconds RCU Tasks
6045 Trace asynchronous callback batching for
6046 call_rcu_tasks_trace(). A negative value
6047 will take the default. A value of zero will
6048 disable batching. Batching is always disabled
6049 for synchronize_rcu_tasks_trace().
6051 rcupdate.rcu_self_test= [KNL]
6052 Run the RCU early boot self tests
6056 Run specified binary instead of /init from the ramdisk,
6057 used for early userspace startup. See initrd.
6060 force - Override the decision by the kernel to hide the
6061 advertisement of RDRAND support (this affects
6062 certain AMD processors because of buggy BIOS
6063 support, specifically around the suspend/resume
6067 Turn on/off individual RDT features. List is:
6068 cmt, mbmtotal, mbmlocal, l3cat, l3cdp, l2cat, l2cdp,
6070 E.g. to turn on cmt and turn off mba use:
6074 Format (x86 or x86_64):
6075 [w[arm] | c[old] | h[ard] | s[oft] | g[pio]] | d[efault] \
6077 [[,]b[ios] | a[cpi] | k[bd] | t[riple] | e[fi] | p[ci]] \
6079 Where reboot_mode is one of warm (soft) or cold (hard) or gpio
6080 (prefix with 'panic_' to set mode for panic
6082 reboot_type is one of bios, acpi, kbd, triple, efi, or pci,
6083 reboot_force is either force or not specified,
6084 reboot_cpu is s[mp]#### with #### being the processor
6085 to be used for rebooting.
6088 Use the ACPI RESET_REG in the FADT. If ACPI is not
6089 configured or the ACPI reset does not work, the reboot
6090 path attempts the reset using the keyboard controller.
6093 Use the CPU reboot vector for warm reset
6096 Set the cold reboot flag
6099 There are some built-in platform specific "quirks"
6100 - you may see: "reboot: <name> series board detected.
6101 Selecting <type> for reboots." In the case where you
6102 think the quirk is in error (e.g. you have newer BIOS,
6103 or newer board) using this option will ignore the
6104 built-in quirk table, and use the generic default
6108 Use efi reset_system runtime service. If EFI is not
6109 configured or the EFI reset does not work, the reboot
6110 path attempts the reset using the keyboard controller.
6113 Don't stop other CPUs on reboot. This can make reboot
6114 more reliable in some cases.
6117 Use the keyboard controller. cold reset (default)
6120 Use a write to the PCI config space register 0xcf9 to
6124 Force a triple fault (init)
6127 Don't set the cold reboot flag
6129 Using warm reset will be much faster especially on big
6130 memory systems because the BIOS will not go through
6131 the memory check. Disadvantage is that not all
6132 hardware will be completely reinitialized on reboot so
6133 there may be boot problems on some systems.
6136 refscale.holdoff= [KNL]
6137 Set test-start holdoff period. The purpose of
6138 this parameter is to delay the start of the
6139 test until boot completes in order to avoid
6142 refscale.lookup_instances= [KNL]
6143 Number of data elements to use for the forms of
6144 SLAB_TYPESAFE_BY_RCU testing. A negative number
6145 is negated and multiplied by nr_cpu_ids, while
6146 zero specifies nr_cpu_ids.
6148 refscale.loops= [KNL]
6149 Set the number of loops over the synchronization
6150 primitive under test. Increasing this number
6151 reduces noise due to loop start/end overhead,
6152 but the default has already reduced the per-pass
6153 noise to a handful of picoseconds on ca. 2020
6156 refscale.nreaders= [KNL]
6157 Set number of readers. The default value of -1
6158 selects N, where N is roughly 75% of the number
6159 of CPUs. A value of zero is an interesting choice.
6161 refscale.nruns= [KNL]
6162 Set number of runs, each of which is dumped onto
6165 refscale.readdelay= [KNL]
6166 Set the read-side critical-section duration,
6167 measured in microseconds.
6169 refscale.scale_type= [KNL]
6170 Specify the read-protection implementation to test.
6172 refscale.shutdown= [KNL]
6173 Shut down the system at the end of the performance
6174 test. This defaults to 1 (shut it down) when
6175 refscale is built into the kernel and to 0 (leave
6176 it running) when refscale is built as a module.
6178 refscale.verbose= [KNL]
6179 Enable additional printk() statements.
6181 refscale.verbose_batched= [KNL]
6182 Batch the additional printk() statements. If zero
6183 (the default) or negative, print everything. Otherwise,
6184 print every Nth verbose statement, where N is the value
6187 regulator_ignore_unused
6189 Prevents regulator framework from disabling regulators
6190 that are unused, due no driver claiming them. This may
6191 be useful for debug and development, but should not be
6192 needed on a platform with proper driver support.
6195 [KNL, SMP] Set scheduler's default relax_domain_level.
6196 See Documentation/admin-guide/cgroup-v1/cpusets.rst.
6198 reserve= [KNL,BUGS] Force kernel to ignore I/O ports or memory
6199 Format: <base1>,<size1>[,<base2>,<size2>,...]
6200 Reserve I/O ports or memory so the kernel won't use
6201 them. If <base> is less than 0x10000, the region
6202 is assumed to be I/O ports; otherwise it is memory.
6205 Format: nn[KMG]:<align>:<label>
6206 Reserve physical memory and label it with a name that
6207 other subsystems can use to access it. This is typically
6208 used for systems that do not wipe the RAM, and this command
6209 line will try to reserve the same physical memory on
6210 soft reboots. Note, it is not guaranteed to be the same
6211 location. For example, if anything about the system changes
6212 or if booting a different kernel. It can also fail if KASLR
6213 places the kernel at the location of where the RAM reservation
6214 was from a previous boot, the new reservation will be at a
6216 Any subsystem using this feature must add a way to verify
6217 that the contents of the physical memory is from a previous
6218 boot, as there may be cases where the memory will not be
6219 located at the same location.
6221 The format is size:align:label for example, to request
6222 12 megabytes of 4096 alignment for ramoops:
6224 reserve_mem=12M:4096:oops ramoops.mem_name=oops
6226 reservetop= [X86-32,EARLY]
6228 Reserves a hole at the top of the kernel virtual
6231 reset_devices [KNL] Force drivers to reset the underlying device
6232 during initialization.
6235 Specify the partition device for software suspend
6237 {/dev/<dev> | PARTUUID=<uuid> | <int>:<int> | <hex>}
6239 resume_offset= [SWSUSP]
6240 Specify the offset from the beginning of the partition
6241 given by "resume=" at which the swap header is located,
6242 in <PAGE_SIZE> units (needed only for swap files).
6243 See Documentation/power/swsusp-and-swap-files.rst
6245 resumedelay= [HIBERNATION] Delay (in seconds) to pause before attempting to
6246 read the resume files
6248 resumewait [HIBERNATION] Wait (indefinitely) for resume device to show up.
6249 Useful for devices that are detected asynchronously
6250 (e.g. USB and MMC devices).
6252 retain_initrd [RAM] Keep initrd memory after extraction. After boot, it will
6253 be accessible via /sys/firmware/initrd.
6255 retbleed= [X86] Control mitigation of RETBleed (Arbitrary
6256 Speculative Code Execution with Return Instructions)
6259 AMD-based UNRET and IBPB mitigations alone do not stop
6260 sibling threads from influencing the predictions of other
6261 sibling threads. For that reason, STIBP is used on pro-
6262 cessors that support it, and mitigate SMT on processors
6266 auto - automatically select a migitation
6267 auto,nosmt - automatically select a mitigation,
6268 disabling SMT if necessary for
6269 the full mitigation (only on Zen1
6270 and older without STIBP).
6271 ibpb - On AMD, mitigate short speculation
6272 windows on basic block boundaries too.
6273 Safe, highest perf impact. It also
6274 enables STIBP if present. Not suitable
6276 ibpb,nosmt - Like "ibpb" above but will disable SMT
6277 when STIBP is not available. This is
6278 the alternative for systems which do not
6280 unret - Force enable untrained return thunks,
6281 only effective on AMD f15h-f17h based
6283 unret,nosmt - Like unret, but will disable SMT when STIBP
6284 is not available. This is the alternative for
6285 systems which do not have STIBP.
6287 Selecting 'auto' will choose a mitigation method at run
6288 time according to the CPU.
6290 Not specifying this option is equivalent to retbleed=auto.
6292 rfkill.default_state=
6293 0 "airplane mode". All wifi, bluetooth, wimax, gps, fm,
6294 etc. communication is blocked by default.
6297 rfkill.master_switch_mode=
6298 0 The "airplane mode" button does nothing.
6299 1 The "airplane mode" button toggles between everything
6300 blocked and the previous configuration.
6301 2 The "airplane mode" button toggles between everything
6302 blocked and everything unblocked.
6305 [KNL] Disable ring 3 MONITOR/MWAIT feature on supported
6308 riscv_isa_fallback [RISCV,EARLY]
6309 When CONFIG_RISCV_ISA_FALLBACK is not enabled, permit
6310 falling back to detecting extension support by parsing
6311 "riscv,isa" property on devicetree systems when the
6312 replacement properties are not found. See the Kconfig
6313 entry for RISCV_ISA_FALLBACK.
6315 ro [KNL] Mount root device read-only on boot
6318 on Mark read-only kernel memory as read-only (default).
6319 off Leave read-only kernel memory writable for debugging.
6320 full Mark read-only kernel memory and aliases as read-only
6325 Enable the uart passthrough on the designated usb port
6326 on Rockchip SoCs. When active, the signals of the
6327 debug-uart get routed to the D+ and D- pins of the usb
6328 port and the regular usb controller gets disabled.
6330 root= [KNL] Root filesystem
6331 Usually this is a block device specifier of some kind,
6332 see the early_lookup_bdev comment in
6333 block/early-lookup.c for details.
6334 Alternatively this can be "ram" for the legacy initial
6335 ramdisk, "nfs" and "cifs" for root on a network file
6336 system, or "mtd" and "ubi" for mounting from raw flash.
6338 rootdelay= [KNL] Delay (in seconds) to pause before attempting to
6339 mount the root filesystem
6341 rootflags= [KNL] Set root filesystem mount option string
6343 rootfstype= [KNL] Set root filesystem type
6345 rootwait [KNL] Wait (indefinitely) for root device to show up.
6346 Useful for devices that are detected asynchronously
6347 (e.g. USB and MMC devices).
6349 rootwait= [KNL] Maximum time (in seconds) to wait for root device
6350 to show up before attempting to mount the root
6353 rproc_mem=nn[KMG][@address]
6354 [KNL,ARM,CMA] Remoteproc physical memory block.
6355 Memory area to be used by remote processor image,
6358 rt_group_sched= [KNL] Enable or disable SCHED_RR/FIFO group scheduling
6359 when CONFIG_RT_GROUP_SCHED=y. Defaults to
6360 !CONFIG_RT_GROUP_SCHED_DEFAULT_DISABLED.
6363 rw [KNL] Mount root device read-write on boot
6365 S [KNL] Run init in single mode
6367 s390_iommu= [HW,S390]
6368 Set s390 IOTLB flushing mode
6370 With strict flushing every unmap operation will result
6371 in an IOTLB flush. Default is lazy flushing before
6372 reuse, which is faster. Deprecated, equivalent to
6375 s390_iommu_aperture= [KNL,S390]
6376 Specifies the size of the per device DMA address space
6377 accessible through the DMA and IOMMU APIs as a decimal
6378 factor of the size of main memory.
6379 The default is 1 meaning that one can concurrently use
6380 as many DMA addresses as physical memory is installed,
6381 if supported by hardware, and thus map all of memory
6382 once. With a value of 2 one can map all of memory twice
6383 and so on. As a special case a factor of 0 imposes no
6384 restrictions other than those given by hardware at the
6385 cost of significant additional memory use for tables.
6388 See drivers/net/irda/sa1100_ir.c.
6390 sched_verbose [KNL,EARLY] Enables verbose scheduler debug messages.
6392 schedstats= [KNL,X86] Enable or disable scheduled statistics.
6393 Allowed values are enable and disable. This feature
6394 incurs a small amount of overhead in the scheduler
6395 but is useful for debugging and performance tuning.
6397 sched_thermal_decay_shift=
6399 [KNL, SMP] Set a decay shift for scheduler thermal
6400 pressure signal. Thermal pressure signal follows the
6401 default decay period of other scheduler pelt
6402 signals(usually 32 ms but configurable). Setting
6403 sched_thermal_decay_shift will left shift the decay
6404 period for the thermal pressure signal by the shift
6406 i.e. with the default pelt decay period of 32 ms
6407 sched_thermal_decay_shift thermal pressure decay pr
6411 Format: integer between 0 and 10
6414 scftorture.holdoff= [KNL]
6415 Number of seconds to hold off before starting
6416 test. Defaults to zero for module insertion and
6417 to 10 seconds for built-in smp_call_function()
6420 scftorture.longwait= [KNL]
6421 Request ridiculously long waits randomly selected
6422 up to the chosen limit in seconds. Zero (the
6423 default) disables this feature. Please note
6424 that requesting even small non-zero numbers of
6425 seconds can result in RCU CPU stall warnings,
6426 softlockup complaints, and so on.
6428 scftorture.nthreads= [KNL]
6429 Number of kthreads to spawn to invoke the
6430 smp_call_function() family of functions.
6431 The default of -1 specifies a number of kthreads
6432 equal to the number of CPUs.
6434 scftorture.onoff_holdoff= [KNL]
6435 Number seconds to wait after the start of the
6436 test before initiating CPU-hotplug operations.
6438 scftorture.onoff_interval= [KNL]
6439 Number seconds to wait between successive
6440 CPU-hotplug operations. Specifying zero (which
6441 is the default) disables CPU-hotplug operations.
6443 scftorture.shutdown_secs= [KNL]
6444 The number of seconds following the start of the
6445 test after which to shut down the system. The
6446 default of zero avoids shutting down the system.
6447 Non-zero values are useful for automated tests.
6449 scftorture.stat_interval= [KNL]
6450 The number of seconds between outputting the
6451 current test statistics to the console. A value
6452 of zero disables statistics output.
6454 scftorture.stutter_cpus= [KNL]
6455 The number of jiffies to wait between each change
6456 to the set of CPUs under test.
6458 scftorture.use_cpus_read_lock= [KNL]
6459 Use use_cpus_read_lock() instead of the default
6460 preempt_disable() to disable CPU hotplug
6461 while invoking one of the smp_call_function*()
6464 scftorture.verbose= [KNL]
6465 Enable additional printk() statements.
6467 scftorture.weight_single= [KNL]
6468 The probability weighting to use for the
6469 smp_call_function_single() function with a zero
6470 "wait" parameter. A value of -1 selects the
6471 default if all other weights are -1. However,
6472 if at least one weight has some other value, a
6473 value of -1 will instead select a weight of zero.
6475 scftorture.weight_single_wait= [KNL]
6476 The probability weighting to use for the
6477 smp_call_function_single() function with a
6478 non-zero "wait" parameter. See weight_single.
6480 scftorture.weight_many= [KNL]
6481 The probability weighting to use for the
6482 smp_call_function_many() function with a zero
6483 "wait" parameter. See weight_single.
6484 Note well that setting a high probability for
6485 this weighting can place serious IPI load
6488 scftorture.weight_many_wait= [KNL]
6489 The probability weighting to use for the
6490 smp_call_function_many() function with a
6491 non-zero "wait" parameter. See weight_single
6494 scftorture.weight_all= [KNL]
6495 The probability weighting to use for the
6496 smp_call_function_all() function with a zero
6497 "wait" parameter. See weight_single and
6500 scftorture.weight_all_wait= [KNL]
6501 The probability weighting to use for the
6502 smp_call_function_all() function with a
6503 non-zero "wait" parameter. See weight_single
6506 sdw_mclk_divider=[SDW]
6507 Specify the MCLK divider for Intel SoundWire buses in
6508 case the BIOS does not provide the clock rate properly.
6510 skew_tick= [KNL,EARLY] Offset the periodic timer tick per cpu to mitigate
6511 xtime_lock contention on larger systems, and/or RCU lock
6512 contention on all systems with CONFIG_MAXSMP set.
6513 Format: { "0" | "1" }
6514 0 -- disable. (may be 1 via CONFIG_CMDLINE="skew_tick=1"
6516 Note: increases power consumption, thus should only be
6517 enabled if running jitter sensitive (HPC/RT) workloads.
6519 security= [SECURITY] Choose a legacy "major" security module to
6520 enable at boot. This has been deprecated by the
6523 selinux= [SELINUX] Disable or enable SELinux at boot time.
6524 Format: { "0" | "1" }
6525 See security/selinux/Kconfig help text.
6530 serialnumber [BUGS=X86-32]
6532 sev=option[,option...] [X86-64]
6535 Enable debug messages.
6538 Do not enable SEV-SNP (applies to host/hypervisor
6539 only). Setting 'nosnp' avoids the RMP check overhead
6540 in memory accesses when users do not want to run
6544 Maximal number of shapers.
6546 show_lapic= [APIC,X86] Advanced Programmable Interrupt Controller
6547 Limit apic dumping. The parameter defines the maximal
6548 number of local apics being dumped. Also it is possible
6549 to set it to "all" by meaning -- no limit here.
6550 Format: { 1 (default) | 2 | ... | all }.
6551 The parameter valid if only apic=debug or
6552 apic=verbose is specified.
6553 Example: apic=debug show_lapic=all
6555 slab_debug[=options[,slabs][;[options[,slabs]]...] [MM]
6556 Enabling slab_debug allows one to determine the
6557 culprit if slab objects become corrupted. Enabling
6558 slab_debug can create guard zones around objects and
6559 may poison objects when not in use. Also tracks the
6560 last alloc / free. For more information see
6561 Documentation/mm/slub.rst.
6562 (slub_debug legacy name also accepted for now)
6564 slab_max_order= [MM]
6565 Determines the maximum allowed order for slabs.
6566 A high setting may cause OOMs due to memory
6567 fragmentation. For more information see
6568 Documentation/mm/slub.rst.
6569 (slub_max_order legacy name also accepted for now)
6572 Enable merging of slabs with similar size when the
6573 kernel is built without CONFIG_SLAB_MERGE_DEFAULT.
6574 (slub_merge legacy name also accepted for now)
6576 slab_min_objects= [MM]
6577 The minimum number of objects per slab. SLUB will
6578 increase the slab order up to slab_max_order to
6579 generate a sufficiently large slab able to contain
6580 the number of objects indicated. The higher the number
6581 of objects the smaller the overhead of tracking slabs
6582 and the less frequently locks need to be acquired.
6583 For more information see Documentation/mm/slub.rst.
6584 (slub_min_objects legacy name also accepted for now)
6586 slab_min_order= [MM]
6587 Determines the minimum page order for slabs. Must be
6588 lower or equal to slab_max_order. For more information see
6589 Documentation/mm/slub.rst.
6590 (slub_min_order legacy name also accepted for now)
6593 Disable merging of slabs with similar size. May be
6594 necessary if there is some reason to distinguish
6595 allocs to different slabs, especially in hardened
6596 environments where the risk of heap overflows and
6597 layout control by attackers can usually be
6598 frustrated by disabling merging. This will reduce
6599 most of the exposure of a heap attack to a single
6600 cache (risks via metadata attacks are mostly
6601 unchanged). Debug options disable merging on their
6603 For more information see Documentation/mm/slub.rst.
6604 (slub_nomerge legacy name also accepted for now)
6606 slab_strict_numa [MM]
6607 Support memory policies on a per object level
6608 in the slab allocator. The default is for memory
6609 policies to be applied at the folio level when
6610 a new folio is needed or a partial folio is
6611 retrieved from the lists. Increases overhead
6612 in the slab fastpaths but gains more accurate
6613 NUMA kernel object placement which helps with slow
6614 interconnects in NUMA systems.
6619 Format: <io1>[,<io2>[,...,<io8>]]
6621 smp.csd_lock_timeout= [KNL]
6622 Specify the period of time in milliseconds
6623 that smp_call_function() and friends will wait
6624 for a CPU to release the CSD lock. This is
6625 useful when diagnosing bugs involving CPUs
6626 disabling interrupts for extended periods
6627 of time. Defaults to 5,000 milliseconds, and
6628 setting a value of zero disables this feature.
6629 This feature may be more efficiently disabled
6630 using the csdlock_debug- kernel parameter.
6632 smp.panic_on_ipistall= [KNL]
6633 If a csd_lock_timeout extends for more than
6634 the specified number of milliseconds, panic the
6635 system. By default, let CSD-lock acquisition
6636 take as long as they take. Specifying 300,000
6637 for this value provides a 5-minute timeout.
6639 smsc-ircc2.nopnp [HW] Don't use PNP to discover SMC devices
6640 smsc-ircc2.ircc_cfg= [HW] Device configuration I/O port
6641 smsc-ircc2.ircc_sir= [HW] SIR base I/O port
6642 smsc-ircc2.ircc_fir= [HW] FIR base I/O port
6643 smsc-ircc2.ircc_irq= [HW] IRQ line
6644 smsc-ircc2.ircc_dma= [HW] DMA channel
6645 smsc-ircc2.ircc_transceiver= [HW] Transceiver type:
6646 0: Toshiba Satellite 1800 (GP data pin select)
6647 1: Fast pin select (default)
6650 smt= [KNL,MIPS,S390,EARLY] Set the maximum number of threads
6651 (logical CPUs) to use per physical CPU on systems
6652 capable of symmetric multithreading (SMT). Will
6653 be capped to the actual hardware limit.
6655 Default: -1 (no limit)
6658 [KNL] Should the soft-lockup detector generate panics.
6661 A value of 1 instructs the soft-lockup detector
6662 to panic the machine when a soft-lockup occurs. It is
6663 also controlled by the kernel.softlockup_panic sysctl
6664 and CONFIG_BOOTPARAM_SOFTLOCKUP_PANIC, which is the
6665 respective build-time switch to that functionality.
6667 softlockup_all_cpu_backtrace=
6668 [KNL] Should the soft-lockup detector generate
6669 backtraces on all cpus.
6672 sonypi.*= [HW] Sony Programmable I/O Control Device driver
6673 See Documentation/admin-guide/laptops/sonypi.rst
6675 spectre_bhi= [X86] Control mitigation of Branch History Injection
6676 (BHI) vulnerability. This setting affects the
6677 deployment of the HW BHI control and the SW BHB
6680 on - (default) Enable the HW or SW mitigation as
6681 needed. This protects the kernel from
6682 both syscalls and VMs.
6683 vmexit - On systems which don't have the HW mitigation
6684 available, enable the SW mitigation on vmexit
6685 ONLY. On such systems, the host kernel is
6686 protected from VM-originated BHI attacks, but
6687 may still be vulnerable to syscall attacks.
6688 off - Disable the mitigation.
6690 spectre_v2= [X86,EARLY] Control mitigation of Spectre variant 2
6691 (indirect branch speculation) vulnerability.
6692 The default operation protects the kernel from
6695 on - unconditionally enable, implies
6697 off - unconditionally disable, implies
6699 auto - kernel detects whether your CPU model is
6702 Selecting 'on' will, and 'auto' may, choose a
6703 mitigation method at run time according to the
6704 CPU, the available microcode, the setting of the
6705 CONFIG_MITIGATION_RETPOLINE configuration option,
6706 and the compiler with which the kernel was built.
6708 Selecting 'on' will also enable the mitigation
6709 against user space to user space task attacks.
6710 Selecting specific mitigation does not force enable
6713 Selecting 'off' will disable both the kernel and
6714 the user space protections.
6716 Specific mitigations can also be selected manually:
6718 retpoline - replace indirect branches
6719 retpoline,generic - Retpolines
6720 retpoline,lfence - LFENCE; indirect branch
6721 retpoline,amd - alias for retpoline,lfence
6722 eibrs - Enhanced/Auto IBRS
6723 eibrs,retpoline - Enhanced/Auto IBRS + Retpolines
6724 eibrs,lfence - Enhanced/Auto IBRS + LFENCE
6725 ibrs - use IBRS to protect kernel
6727 Not specifying this option is equivalent to
6731 [X86] Control mitigation of Spectre variant 2
6732 (indirect branch speculation) vulnerability between
6735 on - Unconditionally enable mitigations. Is
6736 enforced by spectre_v2=on
6738 off - Unconditionally disable mitigations. Is
6739 enforced by spectre_v2=off
6741 prctl - Indirect branch speculation is enabled,
6742 but mitigation can be enabled via prctl
6743 per thread. The mitigation control state
6744 is inherited on fork.
6747 - Like "prctl" above, but only STIBP is
6748 controlled per thread. IBPB is issued
6749 always when switching between different user
6753 - Same as "prctl" above, but all seccomp
6754 threads will enable the mitigation unless
6755 they explicitly opt out.
6758 - Like "seccomp" above, but only STIBP is
6759 controlled per thread. IBPB is issued
6760 always when switching between different
6761 user space processes.
6763 auto - Kernel selects the mitigation depending on
6764 the available CPU features and vulnerability.
6766 Default mitigation: "prctl"
6768 Not specifying this option is equivalent to
6769 spectre_v2_user=auto.
6771 spec_rstack_overflow=
6772 [X86,EARLY] Control RAS overflow mitigation on AMD Zen CPUs
6774 off - Disable mitigation
6775 microcode - Enable microcode mitigation only
6776 safe-ret - Enable sw-only safe RET mitigation (default)
6777 ibpb - Enable mitigation by issuing IBPB on
6779 ibpb-vmexit - Issue IBPB only on VMEXIT
6780 (cloud-specific mitigation)
6782 spec_store_bypass_disable=
6783 [HW,EARLY] Control Speculative Store Bypass (SSB) Disable mitigation
6784 (Speculative Store Bypass vulnerability)
6786 Certain CPUs are vulnerable to an exploit against a
6787 a common industry wide performance optimization known
6788 as "Speculative Store Bypass" in which recent stores
6789 to the same memory location may not be observed by
6790 later loads during speculative execution. The idea
6791 is that such stores are unlikely and that they can
6792 be detected prior to instruction retirement at the
6793 end of a particular speculation execution window.
6795 In vulnerable processors, the speculatively forwarded
6796 store can be used in a cache side channel attack, for
6797 example to read memory to which the attacker does not
6798 directly have access (e.g. inside sandboxed code).
6800 This parameter controls whether the Speculative Store
6801 Bypass optimization is used.
6803 On x86 the options are:
6805 on - Unconditionally disable Speculative Store Bypass
6806 off - Unconditionally enable Speculative Store Bypass
6807 auto - Kernel detects whether the CPU model contains an
6808 implementation of Speculative Store Bypass and
6809 picks the most appropriate mitigation. If the
6810 CPU is not vulnerable, "off" is selected. If the
6811 CPU is vulnerable the default mitigation is
6812 architecture and Kconfig dependent. See below.
6813 prctl - Control Speculative Store Bypass per thread
6814 via prctl. Speculative Store Bypass is enabled
6815 for a process by default. The state of the control
6816 is inherited on fork.
6817 seccomp - Same as "prctl" above, but all seccomp threads
6818 will disable SSB unless they explicitly opt out.
6820 Default mitigations:
6823 On powerpc the options are:
6825 on,auto - On Power8 and Power9 insert a store-forwarding
6826 barrier on kernel entry and exit. On Power7
6827 perform a software flush on kernel entry and
6831 Not specifying this option is equivalent to
6832 spec_store_bypass_disable=auto.
6835 [X86] Enable split lock detection or bus lock detection
6837 When enabled (and if hardware support is present), atomic
6838 instructions that access data across cache line
6839 boundaries will result in an alignment check exception
6840 for split lock detection or a debug exception for
6845 warn - the kernel will emit rate-limited warnings
6846 about applications triggering the #AC
6847 exception or the #DB exception. This mode is
6848 the default on CPUs that support split lock
6849 detection or bus lock detection. Default
6850 behavior is by #AC if both features are
6851 enabled in hardware.
6853 fatal - the kernel will send SIGBUS to applications
6854 that trigger the #AC exception or the #DB
6855 exception. Default behavior is by #AC if
6856 both features are enabled in hardware.
6859 Set system wide rate limit to N bus locks
6860 per second for bus lock detection.
6863 N/A for split lock detection.
6866 If an #AC exception is hit in the kernel or in
6867 firmware (i.e. not while executing in user mode)
6868 the kernel will oops in either "warn" or "fatal"
6871 #DB exception for bus lock is triggered only when
6874 srbds= [X86,INTEL,EARLY]
6875 Control the Special Register Buffer Data Sampling
6878 Certain CPUs are vulnerable to an MDS-like
6879 exploit which can leak bits from the random
6882 By default, this issue is mitigated by
6883 microcode. However, the microcode fix can cause
6884 the RDRAND and RDSEED instructions to become
6885 much slower. Among other effects, this will
6886 result in reduced throughput from /dev/urandom.
6888 The microcode mitigation can be disabled with
6889 the following option:
6891 off: Disable mitigation and remove
6892 performance impact to RDRAND and RDSEED
6894 srcutree.big_cpu_lim [KNL]
6895 Specifies the number of CPUs constituting a
6896 large system, such that srcu_struct structures
6897 should immediately allocate an srcu_node array.
6898 This kernel-boot parameter defaults to 128,
6899 but takes effect only when the low-order four
6900 bits of srcutree.convert_to_big is equal to 3
6903 srcutree.convert_to_big [KNL]
6904 Specifies under what conditions an SRCU tree
6905 srcu_struct structure will be converted to big
6906 form, that is, with an rcu_node tree:
6909 1: At init_srcu_struct() time.
6910 2: When rcutorture decides to.
6911 3: Decide at boot time (default).
6912 0x1X: Above plus if high contention.
6914 Either way, the srcu_node tree will be sized based
6915 on the actual runtime number of CPUs (nr_cpu_ids)
6916 instead of the compile-time CONFIG_NR_CPUS.
6918 srcutree.counter_wrap_check [KNL]
6919 Specifies how frequently to check for
6920 grace-period sequence counter wrap for the
6921 srcu_data structure's ->srcu_gp_seq_needed field.
6922 The greater the number of bits set in this kernel
6923 parameter, the less frequently counter wrap will
6924 be checked for. Note that the bottom two bits
6927 srcutree.exp_holdoff [KNL]
6928 Specifies how many nanoseconds must elapse
6929 since the end of the last SRCU grace period for
6930 a given srcu_struct until the next normal SRCU
6931 grace period will be considered for automatic
6932 expediting. Set to zero to disable automatic
6935 srcutree.srcu_max_nodelay [KNL]
6936 Specifies the number of no-delay instances
6937 per jiffy for which the SRCU grace period
6938 worker thread will be rescheduled with zero
6939 delay. Beyond this limit, worker thread will
6940 be rescheduled with a sleep delay of one jiffy.
6942 srcutree.srcu_max_nodelay_phase [KNL]
6943 Specifies the per-grace-period phase, number of
6944 non-sleeping polls of readers. Beyond this limit,
6945 grace period worker thread will be rescheduled
6946 with a sleep delay of one jiffy, between each
6947 rescan of the readers, for a grace period phase.
6949 srcutree.srcu_retry_check_delay [KNL]
6950 Specifies number of microseconds of non-sleeping
6951 delay between each non-sleeping poll of readers.
6953 srcutree.small_contention_lim [KNL]
6954 Specifies the number of update-side contention
6955 events per jiffy will be tolerated before
6956 initiating a conversion of an srcu_struct
6957 structure to big form. Note that the value of
6958 srcutree.convert_to_big must have the 0x10 bit
6959 set for contention-based conversions to occur.
6961 ssbd= [ARM64,HW,EARLY]
6962 Speculative Store Bypass Disable control
6964 On CPUs that are vulnerable to the Speculative
6965 Store Bypass vulnerability and offer a
6966 firmware based mitigation, this parameter
6967 indicates how the mitigation should be used:
6969 force-on: Unconditionally enable mitigation for
6970 for both kernel and userspace
6971 force-off: Unconditionally disable mitigation for
6972 for both kernel and userspace
6973 kernel: Always enable mitigation in the
6974 kernel, and offer a prctl interface
6975 to allow userspace to register its
6976 interest in being mitigated too.
6978 stack_guard_gap= [MM]
6979 override the default stack gap protection. The value
6980 is in page units and it defines how many pages prior
6981 to (for stacks growing down) resp. after (for stacks
6982 growing up) the main stack are reserved for no other
6983 mapping. Default value is 256 pages.
6985 stack_depot_disable= [KNL,EARLY]
6986 Setting this to true through kernel command line will
6987 disable the stack depot thereby saving the static memory
6988 consumed by the stack hash table. By default this is set
6992 Enabled the stack tracer on boot up.
6994 stacktrace_filter=[function-list]
6995 [FTRACE] Limit the functions that the stack tracer
6996 will trace at boot up. function-list is a comma-separated
6997 list of functions. This list can be changed at run
6998 time by the stack_trace_filter file in the debugfs
6999 tracing directory. Note, this enables stack tracing
7000 and the stacktrace above is not needed.
7004 Set the STI (builtin display/keyboard on the HP-PARISC
7005 machines) console (graphic card) which should be used
7006 as the initial boot-console.
7007 See also comment in drivers/video/console/sticore.c.
7010 See comment in drivers/video/console/sticore.c.
7013 Format: bpp:<bpp1>[:<bpp2>[:<bpp3>...]]
7018 Enable or disable strict sigaltstack size checks
7019 against the required signal frame size which
7020 depends on the supported FPU features. This can
7021 be used to filter out binaries which have
7022 not yet been made aware of AT_MINSIGSTKSZ.
7024 stress_hpt [PPC,EARLY]
7025 Limits the number of kernel HPT entries in the hash
7026 page table to increase the rate of hash page table
7027 faults on kernel addresses.
7029 stress_slb [PPC,EARLY]
7030 Limits the number of kernel SLB entries, and flushes
7031 them frequently to increase the rate of SLB faults
7032 on kernel addresses.
7034 sunrpc.min_resvport=
7035 sunrpc.max_resvport=
7037 SunRPC servers often require that client requests
7038 originate from a privileged port (i.e. a port in the
7039 range 0 < portnr < 1024).
7040 An administrator who wishes to reserve some of these
7041 ports for other uses may adjust the range that the
7042 kernel's sunrpc client considers to be privileged
7043 using these two parameters to set the minimum and
7044 maximum port values.
7046 sunrpc.svc_rpc_per_connection_limit=
7048 Limit the number of requests that the server will
7049 process in parallel from a single connection.
7050 The default value is 0 (no limit).
7054 Control how the NFS server code allocates CPUs to
7055 service thread pools. Depending on how many NICs
7056 you have and where their interrupts are bound, this
7057 option will affect which CPUs will do NFS serving.
7058 Note: this parameter cannot be changed while the
7059 NFS server is running.
7061 auto the server chooses an appropriate mode
7062 automatically using heuristics
7063 global a single global pool contains all CPUs
7064 percpu one pool for each CPU
7065 pernode one pool for each NUMA node (equivalent
7066 to global on non-NUMA machines)
7068 sunrpc.tcp_slot_table_entries=
7069 sunrpc.udp_slot_table_entries=
7071 Sets the upper limit on the number of simultaneous
7072 RPC calls that can be sent from the client to a
7073 server. Increasing these values may allow you to
7074 improve throughput, but will also increase the
7075 amount of memory reserved for use by the client.
7077 suspend.pm_test_delay=
7079 Sets the number of seconds to remain in a suspend test
7080 mode before resuming the system (see
7081 /sys/power/pm_test). Only available when CONFIG_PM_DEBUG
7082 is set. Default value is 5.
7085 Format: { on | off | y | n | 1 | 0 }
7086 This parameter controls use of the Protected
7087 Execution Facility on pSeries.
7089 swiotlb= [ARM,PPC,MIPS,X86,S390,EARLY]
7090 Format: { <int> [,<int>] | force | noforce }
7091 <int> -- Number of I/O TLB slabs
7092 <int> -- Second integer after comma. Number of swiotlb
7093 areas with their own lock. Will be rounded up
7095 force -- force using of bounce buffers even if they
7096 wouldn't be automatically used by the kernel
7097 noforce -- Never use bounce buffers (for debugging)
7099 switches= [HW,M68k,EARLY]
7102 Set a sysctl parameter, right before loading the init
7103 process, as if the value was written to the respective
7104 /proc/sys/... file. Both '.' and '/' are recognized as
7105 separators. Unrecognized parameters and invalid values
7106 are reported in the kernel log. Sysctls registered
7107 later by a loaded module cannot be set this way.
7108 Example: sysctl.vm.swappiness=40
7110 sysrq_always_enabled
7112 Ignore sysrq setting - this boot parameter will
7113 neutralize any effect of /proc/sys/kernel/sysrq.
7114 Useful for debugging.
7116 tcpmhash_entries= [KNL,NET]
7117 Set the number of tcp_metrics_hash slots.
7118 Default value is 8192 or 16384 depending on total
7119 ram pages. This is used to specify the TCP metrics
7120 cache size. See Documentation/networking/ip-sysctl.rst
7121 "tcp_no_metrics_save" section for more details.
7125 test_suspend= [SUSPEND]
7126 Format: { "mem" | "standby" | "freeze" }[,N]
7127 Specify "mem" (for Suspend-to-RAM) or "standby" (for
7128 standby suspend) or "freeze" (for suspend type freeze)
7129 as the system sleep state during system startup with
7130 the optional capability to repeat N number of times.
7131 The system is woken from this state using a
7132 wakeup-capable RTC alarm.
7134 thash_entries= [KNL,NET]
7135 Set number of hash buckets for TCP connection
7137 thermal.act= [HW,ACPI]
7138 -1: disable all active trip points in all thermal zones
7139 <degrees C>: override all lowest active trip points
7141 thermal.crt= [HW,ACPI]
7142 -1: disable all critical trip points in all thermal zones
7143 <degrees C>: override all critical trip points
7145 thermal.off= [HW,ACPI]
7146 1: disable ACPI thermal control
7148 thermal.psv= [HW,ACPI]
7149 -1: disable all passive trip points
7150 <degrees C>: override all passive trip points to this
7153 thermal.tzp= [HW,ACPI]
7154 Specify global default ACPI thermal zone polling rate
7155 <deci-seconds>: poll all this frequency
7156 0: no polling (default)
7159 Format: <size>[KMG],<size>[KMG]:<state>;<size>[KMG]-<size>[KMG]:<state>
7160 state is one of "always", "madvise", "never" or "inherit".
7161 Control the default behavior of the system with respect
7162 to anonymous transparent hugepages.
7163 Can be used multiple times for multiple anon THP sizes.
7164 See Documentation/admin-guide/mm/transhuge.rst for more
7167 threadirqs [KNL,EARLY]
7168 Force threading of all interrupt handlers except those
7169 marked explicitly IRQF_NO_THREAD.
7172 Format: <size>[KMG],<size>[KMG]:<policy>;<size>[KMG]-<size>[KMG]:<policy>
7173 Control the default policy of each hugepage size for the
7174 internal shmem mount. <policy> is one of policies available
7175 for the shmem mount ("always", "inherit", "never", "within_size",
7177 It can be used multiple times for multiple shmem THP sizes.
7178 See Documentation/admin-guide/mm/transhuge.rst for more
7181 topology= [S390,EARLY]
7183 Specify if the kernel should make use of the cpu
7184 topology information if the hardware supports this.
7185 The scheduler will make use of this information and
7186 e.g. base its process migration decisions on it.
7189 torture.disable_onoff_at_boot= [KNL]
7190 Prevent the CPU-hotplug component of torturing
7191 until after init has spawned.
7193 torture.ftrace_dump_at_shutdown= [KNL]
7194 Dump the ftrace buffer at torture-test shutdown,
7195 even if there were no errors. This can be a
7196 very costly operation when many torture tests
7197 are running concurrently, especially on systems
7198 with rotating-rust storage.
7200 torture.verbose_sleep_frequency= [KNL]
7201 Specifies how many verbose printk()s should be
7202 emitted between each sleep. The default of zero
7203 disables verbose-printk() sleeping.
7205 torture.verbose_sleep_duration= [KNL]
7206 Duration of each verbose-printk() sleep in jiffies.
7208 tpm.disable_pcr_integrity= [HW,TPM]
7209 Do not protect PCR registers from unintended physical
7210 access, or interposers in the bus by the means of
7211 having an integrity protected session wrapped around
7212 TPM2_PCR_Extend command. Consider this in a situation
7213 where TPM is heavily utilized by IMA, thus protection
7214 causing a major performance hit, and the space where
7215 machines are deployed is by other means guarded.
7217 tpm_suspend_pcr=[HW,TPM]
7218 Format: integer pcr id
7219 Specify that at suspend time, the tpm driver
7220 should extend the specified pcr with zeros,
7221 as a workaround for some chips which fail to
7222 flush the last written pcr on TPM_SaveState.
7223 This will guarantee that all the other pcrs
7226 tpm_tis.interrupts= [HW,TPM]
7227 Enable interrupts for the MMIO based physical layer
7228 for the FIFO interface. By default it is set to false
7229 (0). For more information about TPM hardware interfaces
7230 defined by Trusted Computing Group (TCG) see
7231 https://trustedcomputinggroup.org/resource/pc-client-platform-tpm-profile-ptp-specification/
7234 Have the tracepoints sent to printk as well as the
7235 tracing ring buffer. This is useful for early boot up
7236 where the system hangs or reboots and does not give the
7237 option for reading the tracing buffer or performing a
7238 ftrace_dump_on_oops.
7240 To turn off having tracepoints sent to printk,
7241 echo 0 > /proc/sys/kernel/tracepoint_printk
7242 Note, echoing 1 into this file without the
7243 tp_printk kernel cmdline option has no effect.
7245 The tp_printk_stop_on_boot (see below) can also be used
7246 to stop the printing of events to console at
7251 Having tracepoints sent to printk() and activating high
7252 frequency tracepoints such as irq or sched, can cause
7253 the system to live lock.
7255 tp_printk_stop_on_boot [FTRACE]
7256 When tp_printk (above) is set, it can cause a lot of noise
7257 on the console. It may be useful to only include the
7258 printing of events during boot up, as user space may
7259 make the system inoperable.
7261 This command line option will stop the printing of events
7262 to console at the late_initcall_sync() time frame.
7264 trace_buf_size=nn[KMG]
7265 [FTRACE] will set tracing buffer size on each cpu.
7267 trace_clock= [FTRACE] Set the clock used for tracing events
7269 local - Use the per CPU time stamp counter
7270 (converted into nanoseconds). Fast, but
7271 depending on the architecture, may not be
7272 in sync between CPUs.
7273 global - Event time stamps are synchronize across
7274 CPUs. May be slower than the local clock,
7275 but better for some race conditions.
7276 counter - Simple counting of events (1, 2, ..)
7277 note, some counts may be skipped due to the
7278 infrastructure grabbing the clock more than
7280 uptime - Use jiffies as the time stamp.
7281 perf - Use the same clock that perf uses.
7282 mono - Use ktime_get_mono_fast_ns() for time stamps.
7283 mono_raw - Use ktime_get_raw_fast_ns() for time
7285 boot - Use ktime_get_boot_fast_ns() for time stamps.
7286 Architectures may add more clocks. See
7287 Documentation/trace/ftrace.rst for more details.
7289 trace_event=[event-list]
7290 [FTRACE] Set and start specified trace events in order
7291 to facilitate early boot debugging. The event-list is a
7292 comma-separated list of trace events to enable. See
7293 also Documentation/trace/events.rst
7295 To enable modules, use :mod: keyword:
7297 trace_event=:mod:<module>
7299 The value before :mod: will only enable specific events
7300 that are part of the module. See the above mentioned
7301 document for more information.
7303 trace_instance=[instance-info]
7304 [FTRACE] Create a ring buffer instance early in boot up.
7305 This will be listed in:
7307 /sys/kernel/tracing/instances
7309 Events can be enabled at the time the instance is created
7312 trace_instance=<name>,<system1>:<event1>,<system2>:<event2>
7314 Note, the "<system*>:" portion is optional if the event is
7317 trace_instance=foo,sched:sched_switch,irq_handler_entry,initcall
7319 will enable the "sched_switch" event (note, the "sched:" is optional, and
7320 the same thing would happen if it was left off). The irq_handler_entry
7321 event, and all events under the "initcall" system.
7323 Flags can be added to the instance to modify its behavior when it is
7324 created. The flags are separated by '^'.
7326 The available flags are:
7328 traceoff - Have the tracing instance tracing disabled after it is created.
7329 traceprintk - Have trace_printk() write into this trace instance
7330 (note, "printk" and "trace_printk" can also be used)
7332 trace_instance=foo^traceoff^traceprintk,sched,irq
7334 The flags must come before the defined events.
7336 If memory has been reserved (see memmap for x86), the instance
7337 can use that memory:
7339 memmap=12M$0x284500000 trace_instance=boot_map@0x284500000:12M
7341 The above will create a "boot_map" instance that uses the physical
7342 memory at 0x284500000 that is 12Megs. The per CPU buffers of that
7343 instance will be split up accordingly.
7345 Alternatively, the memory can be reserved by the reserve_mem option:
7347 reserve_mem=12M:4096:trace trace_instance=boot_map@trace
7349 This will reserve 12 megabytes at boot up with a 4096 byte alignment
7350 and place the ring buffer in this memory. Note that due to KASLR, the
7351 memory may not be the same location each time, which will not preserve
7354 Also note that the layout of the ring buffer data may change between
7355 kernel versions where the validator will fail and reset the ring buffer
7356 if the layout is not the same as the previous kernel.
7358 If the ring buffer is used for persistent bootups and has events enabled,
7359 it is recommend to disable tracing so that events from a previous boot do not
7360 mix with events of the current boot (unless you are debugging a random crash
7363 reserve_mem=12M:4096:trace trace_instance=boot_map^traceoff^traceprintk@trace,sched,irq
7365 Note, saving the trace buffer across reboots does require that the system
7366 is set up to not wipe memory. For instance, CONFIG_RESET_ATTACK_MITIGATION
7367 can force a memory reset on boot which will clear any trace that was stored.
7368 This is just one of many ways that can clear memory. Make sure your system
7369 keeps the content of memory across reboots before relying on this option.
7371 NB: Both the mapped address and size must be page aligned for the architecture.
7373 See also Documentation/trace/debugging.rst
7376 trace_options=[option-list]
7377 [FTRACE] Enable or disable tracer options at boot.
7378 The option-list is a comma delimited list of options
7379 that can be enabled or disabled just as if you were
7380 to echo the option name into
7382 /sys/kernel/tracing/trace_options
7384 For example, to enable stacktrace option (to dump the
7385 stack trace of each event), add to the command line:
7387 trace_options=stacktrace
7389 See also Documentation/trace/ftrace.rst "trace options"
7392 trace_trigger=[trigger-list]
7393 [FTRACE] Add a event trigger on specific events.
7394 Set a trigger on top of a specific event, with an optional
7397 The format is is "trace_trigger=<event>.<trigger>[ if <filter>],..."
7398 Where more than one trigger may be specified that are comma deliminated.
7402 trace_trigger="sched_switch.stacktrace if prev_state == 2"
7404 The above will enable the "stacktrace" trigger on the "sched_switch"
7405 event but only trigger it if the "prev_state" of the "sched_switch"
7406 event is "2" (TASK_UNINTERUPTIBLE).
7408 See also "Event triggers" in Documentation/trace/events.rst
7412 [FTRACE] Sometimes tracing is used to debug issues
7413 during the boot process. Since the trace buffer has a
7414 limited amount of storage, it may be prudent to
7415 disable tracing after the boot is finished, otherwise
7416 the critical information may be overwritten. With this
7417 option, the main tracing buffer will be turned off at
7418 the end of the boot process.
7421 [FTRACE] enable this option to disable tracing when a
7422 warning is hit. This turns off "tracing_on". Tracing can
7423 be enabled again by echoing '1' into the "tracing_on"
7424 file located in /sys/kernel/tracing/
7426 This option is useful, as it disables the trace before
7427 the WARNING dump is called, which prevents the trace to
7428 be filled with content caused by the warning output.
7430 This option can also be set at run time via the sysctl
7431 option: kernel/traceoff_on_warning
7433 transparent_hugepage=
7435 Format: [always|madvise|never]
7436 Can be used to control the default behavior of the system
7437 with respect to transparent hugepages.
7438 See Documentation/admin-guide/mm/transhuge.rst
7441 transparent_hugepage_shmem= [KNL]
7442 Format: [always|within_size|advise|never|deny|force]
7443 Can be used to control the hugepage allocation policy for
7444 the internal shmem mount.
7445 See Documentation/admin-guide/mm/transhuge.rst
7448 transparent_hugepage_tmpfs= [KNL]
7449 Format: [always|within_size|advise|never]
7450 Can be used to control the default hugepage allocation policy
7451 for the tmpfs mount.
7452 See Documentation/admin-guide/mm/transhuge.rst
7455 trusted.source= [KEYS]
7457 This parameter identifies the trust source as a backend
7458 for trusted keys implementation. Supported trust
7464 If not specified then it defaults to iterating through
7465 the trust source list starting with TPM and assigns the
7466 first trust source as a backend which is initialized
7467 successfully during iteration.
7471 The RNG used to generate key material for trusted keys.
7474 - the same value as trusted.source: "tpm" or "tee"
7476 If not specified, "default" is used. In this case,
7477 the RNG's choice is left to each individual trust source.
7479 trusted.dcp_use_otp_key
7480 This is intended to be used in combination with
7481 trusted.source=dcp and will select the DCP OTP key
7482 instead of the DCP UNIQUE key blob encryption.
7484 trusted.dcp_skip_zk_test
7485 This is intended to be used in combination with
7486 trusted.source=dcp and will disable the check if the
7487 blob key is all zeros. This is helpful for situations where
7488 having this key zero'ed is acceptable. E.g. in testing
7491 tsc= Disable clocksource stability checks for TSC.
7493 [x86] reliable: mark tsc clocksource as reliable, this
7494 disables clocksource verification at runtime, as well
7495 as the stability checks done at bootup. Used to enable
7496 high-resolution timer mode on older hardware, and in
7497 virtualized environment.
7498 [x86] noirqtime: Do not use TSC to do irq accounting.
7499 Used to run time disable IRQ_TIME_ACCOUNTING on any
7500 platforms where RDTSC is slow and this accounting
7502 [x86] unstable: mark the TSC clocksource as unstable, this
7503 marks the TSC unconditionally unstable at bootup and
7504 avoids any further wobbles once the TSC watchdog notices.
7505 [x86] nowatchdog: disable clocksource watchdog. Used
7506 in situations with strict latency requirements (where
7507 interruptions from clocksource watchdog are not
7509 [x86] recalibrate: force recalibration against a HW timer
7510 (HPET or PM timer) on systems whose TSC frequency was
7511 obtained from HW or FW using either an MSR or CPUID(0x15).
7512 Warn if the difference is more than 500 ppm.
7513 [x86] watchdog: Use TSC as the watchdog clocksource with
7514 which to check other HW timers (HPET or PM timer), but
7515 only on systems where TSC has been deemed trustworthy.
7516 This will be suppressed by an earlier tsc=nowatchdog and
7517 can be overridden by a later tsc=nowatchdog. A console
7518 message will flag any such suppression or overriding.
7520 tsc_early_khz= [X86,EARLY] Skip early TSC calibration and use the given
7521 value instead. Useful when the early TSC frequency discovery
7522 procedure is not reliable, such as on overclocked systems
7523 with CPUID.16h support and partial CPUID.15h support.
7524 Format: <unsigned int>
7526 tsx= [X86] Control Transactional Synchronization
7527 Extensions (TSX) feature in Intel processors that
7528 support TSX control.
7530 This parameter controls the TSX feature. The options are:
7532 on - Enable TSX on the system. Although there are
7533 mitigations for all known security vulnerabilities,
7534 TSX has been known to be an accelerator for
7535 several previous speculation-related CVEs, and
7536 so there may be unknown security risks associated
7537 with leaving it enabled.
7539 off - Disable TSX on the system. (Note that this
7540 option takes effect only on newer CPUs which are
7541 not vulnerable to MDS, i.e., have
7542 MSR_IA32_ARCH_CAPABILITIES.MDS_NO=1 and which get
7543 the new IA32_TSX_CTRL MSR through a microcode
7544 update. This new MSR allows for the reliable
7545 deactivation of the TSX functionality.)
7547 auto - Disable TSX if X86_BUG_TAA is present,
7548 otherwise enable TSX on the system.
7550 Not specifying this option is equivalent to tsx=off.
7552 See Documentation/admin-guide/hw-vuln/tsx_async_abort.rst
7555 tsx_async_abort= [X86,INTEL,EARLY] Control mitigation for the TSX Async
7556 Abort (TAA) vulnerability.
7558 Similar to Micro-architectural Data Sampling (MDS)
7559 certain CPUs that support Transactional
7560 Synchronization Extensions (TSX) are vulnerable to an
7561 exploit against CPU internal buffers which can forward
7562 information to a disclosure gadget under certain
7565 In vulnerable processors, the speculatively forwarded
7566 data can be used in a cache side channel attack, to
7567 access data to which the attacker does not have direct
7570 This parameter controls the TAA mitigation. The
7573 full - Enable TAA mitigation on vulnerable CPUs
7576 full,nosmt - Enable TAA mitigation and disable SMT on
7577 vulnerable CPUs. If TSX is disabled, SMT
7578 is not disabled because CPU is not
7579 vulnerable to cross-thread TAA attacks.
7580 off - Unconditionally disable TAA mitigation
7582 On MDS-affected machines, tsx_async_abort=off can be
7583 prevented by an active MDS mitigation as both vulnerabilities
7584 are mitigated with the same mechanism so in order to disable
7585 this mitigation, you need to specify mds=off too.
7587 Not specifying this option is equivalent to
7588 tsx_async_abort=full. On CPUs which are MDS affected
7589 and deploy MDS mitigation, TAA mitigation is not
7590 required and doesn't provide any additional
7594 Documentation/admin-guide/hw-vuln/tsx_async_abort.rst
7596 turbografx.map[2|3]= [HW,JOY]
7597 TurboGraFX parallel port interface
7599 <port#>,<js1>,<js2>,<js3>,<js4>,<js5>,<js6>,<js7>
7600 See also Documentation/input/devices/joystick-parport.rst
7602 udbg-immortal [PPC] When debugging early kernel crashes that
7603 happen after console_init() and before a proper
7604 console driver takes over, this boot options might
7605 help "seeing" what's going on.
7607 uhash_entries= [KNL,NET]
7608 Set number of hash buckets for UDP/UDP-Lite connections
7611 [USB] Ignore overcurrent events (default N).
7612 Some badly-designed motherboards generate lots of
7613 bogus events, for ports that aren't wired to
7614 anything. Set this parameter to avoid log spamming.
7615 Note that genuine overcurrent events won't be
7618 unaligned_scalar_speed=
7620 Format: {slow | fast | unsupported}
7621 Allow skipping scalar unaligned access speed tests. This
7622 is useful for testing alternative code paths and to skip
7623 the tests in environments where they run too slowly. All
7624 CPUs must have the same scalar unaligned access speed.
7626 unaligned_vector_speed=
7628 Format: {slow | fast | unsupported}
7629 Allow skipping vector unaligned access speed tests. This
7630 is useful for testing alternative code paths and to skip
7631 the tests in environments where they run too slowly. All
7632 CPUs must have the same vector unaligned access speed.
7635 [X86] Cause panic on unknown NMI.
7637 unwind_debug [X86-64,EARLY]
7638 Enable unwinder debug output. This can be
7639 useful for debugging certain unwinder error
7640 conditions, including corrupt stacks and
7641 bad/missing unwinder metadata.
7643 usbcore.authorized_default=
7644 [USB] Default USB device authorization:
7645 (default -1 = authorized (same as 1),
7646 0 = not authorized, 1 = authorized, 2 = authorized
7647 if device connected to internal port)
7649 usbcore.autosuspend=
7650 [USB] The autosuspend time delay (in seconds) used
7651 for newly-detected USB devices (default 2). This
7652 is the time required before an idle device will be
7653 autosuspended. Devices for which the delay is set
7654 to a negative value won't be autosuspended at all.
7656 usbcore.usbfs_snoop=
7657 [USB] Set to log all usbfs traffic (default 0 = off).
7659 usbcore.usbfs_snoop_max=
7660 [USB] Maximum number of bytes to snoop in each URB
7663 usbcore.blinkenlights=
7664 [USB] Set to cycle leds on hubs (default 0 = off).
7666 usbcore.old_scheme_first=
7667 [USB] Start with the old device initialization
7668 scheme (default 0 = off).
7670 usbcore.usbfs_memory_mb=
7671 [USB] Memory limit (in MB) for buffers allocated by
7672 usbfs (default = 16, 0 = max = 2047).
7674 usbcore.use_both_schemes=
7675 [USB] Try the other device initialization scheme
7676 if the first one fails (default 1 = enabled).
7678 usbcore.initial_descriptor_timeout=
7679 [USB] Specifies timeout for the initial 64-byte
7680 USB_REQ_GET_DESCRIPTOR request in milliseconds
7681 (default 5000 = 5.0 seconds).
7683 usbcore.nousb [USB] Disable the USB subsystem
7686 [USB] A list of quirk entries to augment the built-in
7687 usb core quirk list. List entries are separated by
7688 commas. Each entry has the form
7689 VendorID:ProductID:Flags. The IDs are 4-digit hex
7690 numbers and Flags is a set of letters. Each letter
7691 will change the built-in quirk; setting it if it is
7692 clear and clearing it if it is set. The letters have
7693 the following meanings:
7694 a = USB_QUIRK_STRING_FETCH_255 (string
7695 descriptors must not be fetched using
7697 b = USB_QUIRK_RESET_RESUME (device can't resume
7698 correctly so reset it instead);
7699 c = USB_QUIRK_NO_SET_INTF (device can't handle
7700 Set-Interface requests);
7701 d = USB_QUIRK_CONFIG_INTF_STRINGS (device can't
7702 handle its Configuration or Interface
7704 e = USB_QUIRK_RESET (device can't be reset
7705 (e.g morph devices), don't use reset);
7706 f = USB_QUIRK_HONOR_BNUMINTERFACES (device has
7707 more interface descriptions than the
7708 bNumInterfaces count, and can't handle
7709 talking to these interfaces);
7710 g = USB_QUIRK_DELAY_INIT (device needs a pause
7711 during initialization, after we read
7712 the device descriptor);
7713 h = USB_QUIRK_LINEAR_UFRAME_INTR_BINTERVAL (For
7714 high speed and super speed interrupt
7715 endpoints, the USB 2.0 and USB 3.0 spec
7716 require the interval in microframes (1
7717 microframe = 125 microseconds) to be
7718 calculated as interval = 2 ^
7720 Devices with this quirk report their
7721 bInterval as the result of this
7722 calculation instead of the exponent
7723 variable used in the calculation);
7724 i = USB_QUIRK_DEVICE_QUALIFIER (device can't
7725 handle device_qualifier descriptor
7727 j = USB_QUIRK_IGNORE_REMOTE_WAKEUP (device
7728 generates spurious wakeup, ignore
7729 remote wakeup capability);
7730 k = USB_QUIRK_NO_LPM (device can't handle Link
7732 l = USB_QUIRK_LINEAR_FRAME_INTR_BINTERVAL
7733 (Device reports its bInterval as linear
7734 frames instead of the USB 2.0
7736 m = USB_QUIRK_DISCONNECT_SUSPEND (Device needs
7737 to be disconnected before suspend to
7738 prevent spurious wakeup);
7739 n = USB_QUIRK_DELAY_CTRL_MSG (Device needs a
7740 pause after every control message);
7741 o = USB_QUIRK_HUB_SLOW_RESET (Hub needs extra
7742 delay after resetting its port);
7743 p = USB_QUIRK_SHORT_SET_ADDRESS_REQ_TIMEOUT
7744 (Reduce timeout of the SET_ADDRESS
7745 request from 5000 ms to 500 ms);
7746 Example: quirks=0781:5580:bk,0a5c:5834:gij
7749 [USBHID] The interval which mice are to be polled at.
7752 [USBHID] The interval which joysticks are to be polled at.
7755 [USBHID] The interval which keyboards are to be polled at.
7757 usb-storage.delay_use=
7758 [UMS] The delay in seconds before a new device is
7759 scanned for Logical Units (default 1).
7760 Optionally the delay in milliseconds if the value has
7762 Example: delay_use=2567ms
7765 [UMS] A list of quirks entries to supplement or
7766 override the built-in unusual_devs list. List
7767 entries are separated by commas. Each entry has
7768 the form VID:PID:Flags where VID and PID are Vendor
7769 and Product ID values (4-digit hex numbers) and
7770 Flags is a set of characters, each corresponding
7771 to a common usb-storage quirk flag as follows:
7772 a = SANE_SENSE (collect more than 18 bytes
7773 of sense data, not on uas);
7774 b = BAD_SENSE (don't collect more than 18
7775 bytes of sense data, not on uas);
7776 c = FIX_CAPACITY (decrease the reported
7777 device capacity by one sector);
7778 d = NO_READ_DISC_INFO (don't use
7779 READ_DISC_INFO command, not on uas);
7780 e = NO_READ_CAPACITY_16 (don't use
7781 READ_CAPACITY_16 command);
7782 f = NO_REPORT_OPCODES (don't use report opcodes
7784 g = MAX_SECTORS_240 (don't transfer more than
7785 240 sectors at a time, uas only);
7786 h = CAPACITY_HEURISTICS (decrease the
7787 reported device capacity by one
7788 sector if the number is odd);
7789 i = IGNORE_DEVICE (don't bind to this
7791 j = NO_REPORT_LUNS (don't use report luns
7793 k = NO_SAME (do not use WRITE_SAME, uas only)
7794 l = NOT_LOCKABLE (don't try to lock and
7795 unlock ejectable media, not on uas);
7796 m = MAX_SECTORS_64 (don't transfer more
7797 than 64 sectors = 32 KB at a time,
7799 n = INITIAL_READ10 (force a retry of the
7800 initial READ(10) command, not on uas);
7801 o = CAPACITY_OK (accept the capacity
7802 reported by the device, not on uas);
7803 p = WRITE_CACHE (the device cache is ON
7804 by default, not on uas);
7805 r = IGNORE_RESIDUE (the device reports
7806 bogus residue values, not on uas);
7807 s = SINGLE_LUN (the device has only one
7809 t = NO_ATA_1X (don't allow ATA(12) and ATA(16)
7810 commands, uas only);
7811 u = IGNORE_UAS (don't bind to the uas driver);
7812 w = NO_WP_DETECT (don't test whether the
7813 medium is write-protected).
7814 y = ALWAYS_SYNC (issue a SYNCHRONIZE_CACHE
7815 even if the device claims no cache,
7817 Example: quirks=0419:aaf5:rl,0421:0433:rc
7819 user_debug= [KNL,ARM]
7821 See arch/arm/Kconfig.debug help text.
7822 1 - undefined instruction events
7824 4 - invalid data aborts
7827 Example: user_debug=31
7829 vdso= [X86,SH,SPARC]
7830 On X86_32, this is an alias for vdso32=. Otherwise:
7832 vdso=1: enable VDSO (the default)
7833 vdso=0: disable VDSO mapping
7835 vdso32= [X86] Control the 32-bit vDSO
7836 vdso32=1: enable 32-bit VDSO
7837 vdso32=0 or vdso32=2: disable 32-bit VDSO
7839 See the help text for CONFIG_COMPAT_VDSO for more
7840 details. If CONFIG_COMPAT_VDSO is set, the default is
7841 vdso32=0; otherwise, the default is vdso32=1.
7843 For compatibility with older kernels, vdso32=2 is an
7846 Try vdso32=0 if you encounter an error that says:
7847 dl_main: Assertion `(void *) ph->p_vaddr == _rtld_local._dl_sysinfo_dso' failed!
7849 video= [FB,EARLY] Frame buffer configuration
7850 See Documentation/fb/modedb.rst.
7852 video.brightness_switch_enabled= [ACPI]
7854 If set to 1, on receiving an ACPI notify event
7855 generated by hotkey, video driver will adjust brightness
7856 level and then send out the event to user space through
7857 the allocated input device. If set to 0, video driver
7858 will only send out the event without touching backlight
7863 [VMMIO] Memory mapped virtio (platform) device.
7865 <size>@<baseaddr>:<irq>[:<id>]
7867 <size> := size (can use standard suffixes
7869 <baseaddr> := physical base address
7870 <irq> := interrupt number (as passed to
7872 <id> := (optional) platform device id
7874 virtio_mmio.device=1K@0x100b0000:48:7
7876 Can be used multiple times for multiple devices.
7878 vga= [BOOT,X86-32] Select a particular video mode
7879 See Documentation/arch/x86/boot.rst and
7880 Documentation/admin-guide/svga.rst.
7881 Use vga=ask for menu.
7882 This is actually a boot loader parameter; the value is
7883 passed to the kernel using a special protocol.
7885 vm_debug[=options] [KNL] Available with CONFIG_DEBUG_VM=y.
7886 May slow down system boot speed, especially when
7887 enabled on systems with a large amount of memory.
7888 All options are enabled by default, and this
7889 interface is meant to allow for selectively
7890 enabling or disabling specific virtual memory
7893 Available options are:
7894 P Enable page structure init time poisoning
7895 - Disable all of the above options
7897 vmalloc=nn[KMG] [KNL,BOOT,EARLY] Forces the vmalloc area to have an
7898 exact size of <nn>. This can be used to increase
7899 the minimum size (128MB on x86, arm32 platforms).
7900 It can also be used to decrease the size and leave more room
7901 for directly mapped kernel RAM. Note that this parameter does
7902 not exist on many other platforms (including arm64, alpha,
7903 loongarch, arc, csky, hexagon, microblaze, mips, nios2, openrisc,
7904 parisc, m64k, powerpc, riscv, sh, um, xtensa, s390, sparc).
7906 vmcp_cma=nn[MG] [KNL,S390,EARLY]
7907 Sets the memory size reserved for contiguous memory
7908 allocations for the vmcp device driver.
7910 vmhalt= [KNL,S390] Perform z/VM CP command after system halt.
7913 vmpanic= [KNL,S390] Perform z/VM CP command after kernel panic.
7916 vmpoff= [KNL,S390] Perform z/VM CP command after power off.
7919 vsyscall= [X86-64,EARLY]
7920 Controls the behavior of vsyscalls (i.e. calls to
7921 fixed addresses of 0xffffffffff600x00 from legacy
7922 code). Most statically-linked binaries and older
7923 versions of glibc use these calls. Because these
7924 functions are at fixed addresses, they make nice
7925 targets for exploits that can control RIP.
7927 emulate Vsyscalls turn into traps and are emulated
7928 reasonably safely. The vsyscall page is
7931 xonly [default] Vsyscalls turn into traps and are
7932 emulated reasonably safely. The vsyscall
7933 page is not readable.
7935 none Vsyscalls don't work at all. This makes
7936 them quite hard to use for exploits but
7937 might break your system.
7939 vt.color= [VT] Default text color.
7940 Format: 0xYX, X = foreground, Y = background.
7941 Default: 0x07 = light gray on black.
7943 vt.cur_default= [VT] Default cursor shape.
7944 Format: 0xCCBBAA, where AA, BB, and CC are the same as
7945 the parameters of the <Esc>[?A;B;Cc escape sequence;
7946 see vga-softcursor.rst. Default: 2 = underline.
7948 vt.default_blu= [VT]
7949 Format: <blue0>,<blue1>,<blue2>,...,<blue15>
7950 Change the default blue palette of the console.
7951 This is a 16-member array composed of values
7954 vt.default_grn= [VT]
7955 Format: <green0>,<green1>,<green2>,...,<green15>
7956 Change the default green palette of the console.
7957 This is a 16-member array composed of values
7960 vt.default_red= [VT]
7961 Format: <red0>,<red1>,<red2>,...,<red15>
7962 Change the default red palette of the console.
7963 This is a 16-member array composed of values
7969 Set system-wide default UTF-8 mode for all tty's.
7970 Default is 1, i.e. UTF-8 mode is enabled for all
7971 newly opened terminals.
7973 vt.global_cursor_default=
7976 Set system-wide default for whether a cursor
7977 is shown on new VTs. Default is -1,
7978 i.e. cursors will be created by default unless
7979 overridden by individual drivers. 0 will hide
7980 cursors, 1 will display them.
7982 vt.italic= [VT] Default color for italic text; 0-15.
7985 vt.underline= [VT] Default color for underlined text; 0-15.
7988 watchdog timers [HW,WDT] For information on watchdog timers,
7989 see Documentation/watchdog/watchdog-parameters.rst
7990 or other driver-specific files in the
7991 Documentation/watchdog/ directory.
7995 Set the hard lockup detector stall duration
7996 threshold in seconds. The soft lockup detector
7997 threshold is set to twice the value. A value of 0
7998 disables both lockup detectors. Default is 10
8001 workqueue.unbound_cpus=
8002 [KNL,SMP] Specify to constrain one or some CPUs
8003 to use in unbound workqueues.
8005 By default, all online CPUs are available for
8008 workqueue.watchdog_thresh=
8009 If CONFIG_WQ_WATCHDOG is configured, workqueue can
8010 warn stall conditions and dump internal state to
8011 help debugging. 0 disables workqueue stall
8012 detection; otherwise, it's the stall threshold
8013 duration in seconds. The default value is 30 and
8014 it can be updated at runtime by writing to the
8015 corresponding sysfs file.
8017 workqueue.panic_on_stall=<uint>
8018 Panic when workqueue stall is detected by
8019 CONFIG_WQ_WATCHDOG. It sets the number times of the
8020 stall to trigger panic.
8022 The default is 0, which disables the panic on stall.
8024 workqueue.cpu_intensive_thresh_us=
8025 Per-cpu work items which run for longer than this
8026 threshold are automatically considered CPU intensive
8027 and excluded from concurrency management to prevent
8028 them from noticeably delaying other per-cpu work
8029 items. Default is 10000 (10ms).
8031 If CONFIG_WQ_CPU_INTENSIVE_REPORT is set, the kernel
8032 will report the work functions which violate this
8033 threshold repeatedly. They are likely good
8034 candidates for using WQ_UNBOUND workqueues instead.
8036 workqueue.cpu_intensive_warning_thresh=<uint>
8037 If CONFIG_WQ_CPU_INTENSIVE_REPORT is set, the kernel
8038 will report the work functions which violate the
8039 intensive_threshold_us repeatedly. In order to prevent
8040 spurious warnings, start printing only after a work
8041 function has violated this threshold number of times.
8043 The default is 4 times. 0 disables the warning.
8045 workqueue.power_efficient
8046 Per-cpu workqueues are generally preferred because
8047 they show better performance thanks to cache
8048 locality; unfortunately, per-cpu workqueues tend to
8049 be more power hungry than unbound workqueues.
8051 Enabling this makes the per-cpu workqueues which
8052 were observed to contribute significantly to power
8053 consumption unbound, leading to measurably lower
8054 power usage at the cost of small performance
8057 The default value of this parameter is determined by
8058 the config option CONFIG_WQ_POWER_EFFICIENT_DEFAULT.
8060 workqueue.default_affinity_scope=
8061 Select the default affinity scope to use for unbound
8062 workqueues. Can be one of "cpu", "smt", "cache",
8063 "numa" and "system". Default is "cache". For more
8064 information, see the Affinity Scopes section in
8065 Documentation/core-api/workqueue.rst.
8067 This can be changed after boot by writing to the
8068 matching /sys/module/workqueue/parameters file. All
8069 workqueues with the "default" affinity scope will be
8070 updated accordingly.
8072 workqueue.debug_force_rr_cpu
8073 Workqueue used to implicitly guarantee that work
8074 items queued without explicit CPU specified are put
8075 on the local CPU. This guarantee is no longer true
8076 and while local CPU is still preferred work items
8077 may be put on foreign CPUs. This debug option
8078 forces round-robin CPU selection to flush out
8079 usages which depend on the now broken guarantee.
8080 When enabled, memory and cache locality will be
8083 writecombine= [LOONGARCH,EARLY] Control the MAT (Memory Access
8084 Type) of ioremap_wc().
8086 on - Enable writecombine, use WUC for ioremap_wc()
8087 off - Disable writecombine, use SUC for ioremap_wc()
8089 x2apic_phys [X86-64,APIC,EARLY] Use x2apic physical mode instead of
8090 default x2apic cluster mode on platforms
8093 xen_512gb_limit [KNL,X86-64,XEN]
8094 Restricts the kernel running paravirtualized under Xen
8095 to use only up to 512 GB of RAM. The reason to do so is
8096 crash analysis tools and Xen tools for doing domain
8097 save/restore/migration must be enabled to handle larger
8100 xen_emul_unplug= [HW,X86,XEN,EARLY]
8101 Unplug Xen emulated devices
8102 Format: [unplug0,][unplug1]
8103 ide-disks -- unplug primary master IDE devices
8104 aux-ide-disks -- unplug non-primary-master IDE devices
8105 nics -- unplug network devices
8106 all -- unplug all emulated devices (NICs and IDE disks)
8107 unnecessary -- unplugging emulated devices is
8108 unnecessary even if the host did not respond to
8110 never -- do not unplug even if version check succeeds
8112 xen_legacy_crash [X86,XEN,EARLY]
8113 Crash from Xen panic notifier, without executing late
8114 panic() code such as dumping handler.
8116 xen_mc_debug [X86,XEN,EARLY]
8117 Enable multicall debugging when running as a Xen PV guest.
8118 Enabling this feature will reduce performance a little
8119 bit, so it should only be enabled for obtaining extended
8120 debug data in case of multicall errors.
8122 xen_msr_safe= [X86,XEN,EARLY]
8124 Select whether to always use non-faulting (safe) MSR
8125 access functions when running as Xen PV guest. The
8126 default value is controlled by CONFIG_XEN_PV_MSR_SAFE.
8129 Disables the PV optimizations forcing the HVM guest to
8130 run as generic HVM guest with no PV drivers.
8131 This option is obsoleted by the "nopv" option, which
8132 has equivalent effect for XEN platform.
8134 xen_no_vector_callback
8135 [KNL,X86,XEN,EARLY] Disable the vector callback for Xen
8136 event channel interrupts.
8138 xen_scrub_pages= [XEN]
8139 Boolean option to control scrubbing pages before giving them back
8140 to Xen, for use by other domains. Can be also changed at runtime
8141 with /sys/devices/system/xen_memory/xen_memory0/scrub_pages.
8142 Default value controlled with CONFIG_XEN_SCRUB_PAGES_DEFAULT.
8144 xen_timer_slop= [X86-64,XEN,EARLY]
8145 Set the timer slop (in nanoseconds) for the virtual Xen
8146 timers (default is 100000). This adjusts the minimum
8147 delta of virtualized Xen timers, where lower values
8148 improve timer resolution at the expense of processing
8149 more timer interrupts.
8151 xen.balloon_boot_timeout= [XEN]
8152 The time (in seconds) to wait before giving up to boot
8153 in case initial ballooning fails to free enough memory.
8154 Applies only when running as HVM or PVH guest and
8155 started with less memory configured than allowed at
8156 max. Default is 180.
8158 xen.event_eoi_delay= [XEN]
8159 How long to delay EOI handling in case of event
8160 storms (jiffies). Default is 10.
8162 xen.event_loop_timeout= [XEN]
8163 After which time (jiffies) the event handling loop
8164 should start to delay EOI handling. Default is 2.
8166 xen.fifo_events= [XEN]
8167 Boolean parameter to disable using fifo event handling
8168 even if available. Normally fifo event handling is
8169 preferred over the 2-level event handling, as it is
8170 fairer and the number of possible event channels is
8171 much higher. Default is on (use fifo events).
8173 xirc2ps_cs= [NET,PCMCIA]
8175 <irq>,<irq_mask>,<io>,<full_duplex>,<do_sound>,<lockup_hack>[,<irq2>[,<irq3>[,<irq4>]]]
8178 By default on POWER9 and above, the kernel will
8179 natively use the XIVE interrupt controller. This option
8180 allows the fallback firmware mode to be used:
8182 off Fallback to firmware control of XIVE interrupt
8183 controller on both pseries and powernv
8184 platforms. Only useful on POWER9 and above.
8186 xive.store-eoi=off [PPC]
8187 By default on POWER10 and above, the kernel will use
8188 stores for EOI handling when the XIVE interrupt mode
8189 is active. This option allows the XIVE driver to use
8190 loads instead, as on POWER9.
8192 xhci-hcd.quirks [USB,KNL]
8193 A hex value specifying bitmask with supplemental xhci
8194 host controller quirks. Meaning of each bit can be
8195 consulted in header drivers/usb/host/xhci.h.
8198 Format: { early | on | rw | ro | off }
8199 Controls if xmon debugger is enabled. Default is off.
8200 Passing only "xmon" is equivalent to "xmon=early".
8201 early Call xmon as early as possible on boot; xmon
8202 debugger is called from setup_arch().
8203 on xmon debugger hooks will be installed so xmon
8204 is only called on a kernel crash. Default mode,
8205 i.e. either "ro" or "rw" mode, is controlled
8206 with CONFIG_XMON_DEFAULT_RO_MODE.
8207 rw xmon debugger hooks will be installed so xmon
8208 is called only on a kernel crash, mode is write,
8209 meaning SPR registers, memory and, other data
8210 can be written using xmon commands.
8211 ro same as "rw" option above but SPR registers,
8212 memory, and other data can't be written using
8214 off xmon is disabled.