1 acpi= [HW,ACPI,X86,ARM64]
2 Advanced Configuration and Power Interface
3 Format: { force | on | off | strict | noirq | rsdt |
5 force -- enable ACPI if default was off
6 on -- enable ACPI but allow fallback to DT [arm64]
7 off -- disable ACPI if default was on
8 noirq -- do not use ACPI for IRQ routing
9 strict -- Be less tolerant of platforms that are not
10 strictly ACPI specification compliant.
11 rsdt -- prefer RSDT over (default) XSDT
12 copy_dsdt -- copy DSDT to memory
13 For ARM64, ONLY "acpi=off", "acpi=on" or "acpi=force"
16 See also Documentation/power/runtime_pm.rst, pci=noacpi
18 acpi_apic_instance= [ACPI, IOAPIC]
20 2: use 2nd APIC table, if available
21 1,0: use 1st APIC table
24 acpi_backlight= [HW,ACPI]
25 { vendor | video | native | none }
26 If set to vendor, prefer vendor-specific driver
27 (e.g. thinkpad_acpi, sony_acpi, etc.) instead
28 of the ACPI video.ko driver.
29 If set to video, use the ACPI video.ko driver.
30 If set to native, use the device's native backlight mode.
31 If set to none, disable the ACPI backlight interface.
33 acpi_force_32bit_fadt_addr
34 force FADT to use 32 bit addresses rather than the
35 64 bit X_* addresses. Some firmware have broken 64
36 bit addresses for force ACPI ignore these and use
37 the older legacy 32 bit addresses.
39 acpica_no_return_repair [HW, ACPI]
40 Disable AML predefined validation mechanism
41 This mechanism can repair the evaluation result to make
42 the return objects more ACPI specification compliant.
43 This option is useful for developers to identify the
44 root cause of an AML interpreter issue when the issue
45 has something to do with the repair mechanism.
47 acpi.debug_layer= [HW,ACPI,ACPI_DEBUG]
48 acpi.debug_level= [HW,ACPI,ACPI_DEBUG]
50 CONFIG_ACPI_DEBUG must be enabled to produce any ACPI
51 debug output. Bits in debug_layer correspond to a
52 _COMPONENT in an ACPI source file, e.g.,
53 #define _COMPONENT ACPI_EVENTS
54 Bits in debug_level correspond to a level in
55 ACPI_DEBUG_PRINT statements, e.g.,
56 ACPI_DEBUG_PRINT((ACPI_DB_INFO, ...
57 The debug_level mask defaults to "info". See
58 Documentation/firmware-guide/acpi/debug.rst for more information about
59 debug layers and levels.
61 Enable processor driver info messages:
62 acpi.debug_layer=0x20000000
63 Enable AML "Debug" output, i.e., stores to the Debug
64 object while interpreting AML:
65 acpi.debug_layer=0xffffffff acpi.debug_level=0x2
66 Enable all messages related to ACPI hardware:
67 acpi.debug_layer=0x2 acpi.debug_level=0xffffffff
69 Some values produce so much output that the system is
70 unusable. The "log_buf_len" parameter may be useful
71 if you need to capture more output.
73 acpi_enforce_resources= [ACPI]
75 Check for resource conflicts between native drivers
76 and ACPI OperationRegions (SystemIO and SystemMemory
77 only). IO ports and memory declared in ACPI might be
78 used by the ACPI subsystem in arbitrary AML code and
79 can interfere with legacy drivers.
80 strict (default): access to resources claimed by ACPI
81 is denied; legacy drivers trying to access reserved
82 resources will fail to bind to device using them.
83 lax: access to resources claimed by ACPI is allowed;
84 legacy drivers trying to access reserved resources
85 will bind successfully but a warning message is logged.
86 no: ACPI OperationRegions are not marked as reserved,
87 no further checks are performed.
89 acpi_force_table_verification [HW,ACPI]
90 Enable table checksum verification during early stage.
91 By default, this is disabled due to x86 early mapping
94 acpi_irq_balance [HW,ACPI]
95 ACPI will balance active IRQs
98 acpi_irq_nobalance [HW,ACPI]
99 ACPI will not move active IRQs (default)
102 acpi_irq_isa= [HW,ACPI] If irq_balance, mark listed IRQs used by ISA
103 Format: <irq>,<irq>...
105 acpi_irq_pci= [HW,ACPI] If irq_balance, clear listed IRQs for
107 Format: <irq>,<irq>...
109 acpi_mask_gpe= [HW,ACPI]
110 Due to the existence of _Lxx/_Exx, some GPEs triggered
111 by unsupported hardware/firmware features can result in
112 GPE floodings that cannot be automatically disabled by
114 This facility can be used to prevent such uncontrolled
116 Format: <byte> or <bitmap-list>
118 acpi_no_auto_serialize [HW,ACPI]
119 Disable auto-serialization of AML methods
120 AML control methods that contain the opcodes to create
121 named objects will be marked as "Serialized" by the
122 auto-serialization feature.
123 This feature is enabled by default.
124 This option allows to turn off the feature.
126 acpi_no_memhotplug [ACPI] Disable memory hotplug. Useful for kdump
129 acpi_no_static_ssdt [HW,ACPI]
130 Disable installation of static SSDTs at early boot time
131 By default, SSDTs contained in the RSDT/XSDT will be
132 installed automatically and they will appear under
133 /sys/firmware/acpi/tables.
134 This option turns off this feature.
135 Note that specifying this option does not affect
136 dynamic table installation which will install SSDT
137 tables to /sys/firmware/acpi/tables/dynamic.
139 acpi_no_watchdog [HW,ACPI,WDT]
140 Ignore the ACPI-based watchdog interface (WDAT) and let
141 a native driver control the watchdog device instead.
143 acpi_rsdp= [ACPI,EFI,KEXEC]
144 Pass the RSDP address to the kernel, mostly used
145 on machines running EFI runtime service to boot the
146 second kernel for kdump.
148 acpi_os_name= [HW,ACPI] Tell ACPI BIOS the name of the OS
149 Format: To spoof as Windows 98: ="Microsoft Windows"
151 acpi_rev_override [ACPI] Override the _REV object to return 5 (instead
152 of 2 which is mandated by ACPI 6) as the supported ACPI
153 specification revision (when using this switch, it may
154 be necessary to carry out a cold reboot _twice_ in a
155 row to make it take effect on the platform firmware).
157 acpi_osi= [HW,ACPI] Modify list of supported OS interface strings
158 acpi_osi="string1" # add string1
159 acpi_osi="!string2" # remove string2
160 acpi_osi=!* # remove all strings
161 acpi_osi=! # disable all built-in OS vendor
163 acpi_osi=!! # enable all built-in OS vendor
165 acpi_osi= # disable all strings
167 'acpi_osi=!' can be used in combination with single or
168 multiple 'acpi_osi="string1"' to support specific OS
169 vendor string(s). Note that such command can only
170 affect the default state of the OS vendor strings, thus
171 it cannot affect the default state of the feature group
172 strings and the current state of the OS vendor strings,
173 specifying it multiple times through kernel command line
174 is meaningless. This command is useful when one do not
175 care about the state of the feature group strings which
176 should be controlled by the OSPM.
178 1. 'acpi_osi=! acpi_osi="Windows 2000"' is equivalent
179 to 'acpi_osi="Windows 2000" acpi_osi=!', they all
180 can make '_OSI("Windows 2000")' TRUE.
182 'acpi_osi=' cannot be used in combination with other
183 'acpi_osi=' command lines, the _OSI method will not
184 exist in the ACPI namespace. NOTE that such command can
185 only affect the _OSI support state, thus specifying it
186 multiple times through kernel command line is also
189 1. 'acpi_osi=' can make 'CondRefOf(_OSI, Local1)'
192 'acpi_osi=!*' can be used in combination with single or
193 multiple 'acpi_osi="string1"' to support specific
194 string(s). Note that such command can affect the
195 current state of both the OS vendor strings and the
196 feature group strings, thus specifying it multiple times
197 through kernel command line is meaningful. But it may
198 still not able to affect the final state of a string if
199 there are quirks related to this string. This command
200 is useful when one want to control the state of the
201 feature group strings to debug BIOS issues related to
204 1. 'acpi_osi="Module Device" acpi_osi=!*' can make
205 '_OSI("Module Device")' FALSE.
206 2. 'acpi_osi=!* acpi_osi="Module Device"' can make
207 '_OSI("Module Device")' TRUE.
208 3. 'acpi_osi=! acpi_osi=!* acpi_osi="Windows 2000"' is
210 'acpi_osi=!* acpi_osi=! acpi_osi="Windows 2000"'
212 'acpi_osi=!* acpi_osi="Windows 2000" acpi_osi=!',
213 they all will make '_OSI("Windows 2000")' TRUE.
216 Override the pmtimer bug detection: force the kernel
217 to assume that this machine's pmtimer latches its value
218 and always returns good values.
220 acpi_sci= [HW,ACPI] ACPI System Control Interrupt trigger mode
221 Format: { level | edge | high | low }
223 acpi_skip_timer_override [HW,ACPI]
224 Recognize and ignore IRQ0/pin2 Interrupt Override.
225 For broken nForce2 BIOS resulting in XT-PIC timer.
227 acpi_sleep= [HW,ACPI] Sleep options
228 Format: { s3_bios, s3_mode, s3_beep, s4_hwsig,
229 s4_nohwsig, old_ordering, nonvs,
230 sci_force_enable, nobl }
231 See Documentation/power/video.rst for information on
233 s3_beep is for debugging; it makes the PC's speaker beep
234 as soon as the kernel's real-mode entry point is called.
235 s4_hwsig causes the kernel to check the ACPI hardware
236 signature during resume from hibernation, and gracefully
237 refuse to resume if it has changed. This complies with
238 the ACPI specification but not with reality, since
239 Windows does not do this and many laptops do change it
240 on docking. So the default behaviour is to allow resume
241 and simply warn when the signature changes, unless the
242 s4_hwsig option is enabled.
243 s4_nohwsig prevents ACPI hardware signature from being
244 used (or even warned about) during resume.
245 old_ordering causes the ACPI 1.0 ordering of the _PTS
246 control method, with respect to putting devices into
247 low power states, to be enforced (the ACPI 2.0 ordering
248 of _PTS is used by default).
249 nonvs prevents the kernel from saving/restoring the
250 ACPI NVS memory during suspend/hibernation and resume.
251 sci_force_enable causes the kernel to set SCI_EN directly
252 on resume from S1/S3 (which is against the ACPI spec,
253 but some broken systems don't work without it).
254 nobl causes the internal blacklist of systems known to
255 behave incorrectly in some ways with respect to system
256 suspend and resume to be ignored (use wisely).
258 acpi_use_timer_override [HW,ACPI]
259 Use timer override. For some broken Nvidia NF5 boards
260 that require a timer override, but don't have HPET
262 add_efi_memmap [EFI; X86] Include EFI memory map in
263 kernel's map of available physical RAM.
266 { off | try_unsupported }
267 off: disable AGP support
268 try_unsupported: try to drive unsupported chipsets
269 (may crash computer or cause data corruption)
272 See Documentation/sound/alsa-configuration.rst
275 Allow the default userspace alignment fault handler
276 behaviour to be specified. Bit 0 enables warnings,
277 bit 1 enables fixups, and bit 2 sends a segfault.
279 align_va_addr= [X86-64]
280 Align virtual addresses by clearing slice [14:12] when
281 allocating a VMA at process creation time. This option
282 gives you up to 3% performance improvement on AMD F15h
283 machines (where it is enabled by default) for a
284 CPU-intensive style benchmark, and it can vary highly in
285 a microbenchmark depending on workload and compiler.
287 32: only for 32-bit processes
288 64: only for 64-bit processes
289 on: enable for both 32- and 64-bit processes
290 off: disable for both 32- and 64-bit processes
292 alloc_snapshot [FTRACE]
293 Allocate the ftrace snapshot buffer on boot up when the
294 main buffer is allocated. This is handy if debugging
295 and you need to use tracing_snapshot() on boot up, and
296 do not want to use tracing_snapshot_alloc() as it needs
297 to be done where GFP_KERNEL allocations are allowed.
299 allow_mismatched_32bit_el0 [ARM64]
300 Allow execve() of 32-bit applications and setting of the
301 PER_LINUX32 personality on systems where only a strict
302 subset of the CPUs support 32-bit EL0. When this
303 parameter is present, the set of CPUs supporting 32-bit
304 EL0 is indicated by /sys/devices/system/cpu/aarch32_el0
305 and hot-unplug operations may be restricted.
307 See Documentation/arm64/asymmetric-32bit.rst for more
310 amd_iommu= [HW,X86-64]
311 Pass parameters to the AMD IOMMU driver in the system.
313 fullflush - Deprecated, equivalent to iommu.strict=1
314 off - do not initialize any AMD IOMMU found in
316 force_isolation - Force device isolation for all
317 devices. The IOMMU driver is not
318 allowed anymore to lift isolation
319 requirements as needed. This option
320 does not override iommu=pt
321 force_enable - Force enable the IOMMU on platforms known
322 to be buggy with IOMMU enabled. Use this
324 pgtbl_v1 - Use v1 page table for DMA-API (Default).
325 pgtbl_v2 - Use v2 page table for DMA-API.
327 amd_iommu_dump= [HW,X86-64]
328 Enable AMD IOMMU driver option to dump the ACPI table
329 for AMD IOMMU. With this option enabled, AMD IOMMU
330 driver will print ACPI tables for AMD IOMMU during
331 IOMMU initialization.
333 amd_iommu_intr= [HW,X86-64]
334 Specifies one of the following AMD IOMMU interrupt
336 legacy - Use legacy interrupt remapping mode.
337 vapic - Use virtual APIC mode, which allows IOMMU
338 to inject interrupts directly into guest.
339 This mode requires kvm-amd.avic=1.
340 (Default when IOMMU HW support is present.)
342 amijoy.map= [HW,JOY] Amiga joystick support
343 Map of devices attached to JOY0DAT and JOY1DAT
345 See also Documentation/input/joydev/joystick.rst
347 analog.map= [HW,JOY] Analog joystick and gamepad support
348 Specifies type or capabilities of an analog joystick
349 connected to one of 16 gameports
350 Format: <type1>,<type2>,..<type16>
353 Power management functions (SPARCstation-4/5 + deriv.)
355 Disable APC CPU standby support. SPARCstation-Fox does
356 not play well with APC CPU idle - disable it if you have
357 APC and your system crashes randomly.
359 apic= [APIC,X86] Advanced Programmable Interrupt Controller
360 Change the output verbosity while booting
361 Format: { quiet (default) | verbose | debug }
362 Change the amount of debugging information output
363 when initialising the APIC and IO-APIC components.
364 For X86-32, this can also be used to specify an APIC
366 Format: apic=driver_name
367 Examples: apic=bigsmp
369 apic_extnmi= [APIC,X86] External NMI delivery setting
370 Format: { bsp (default) | all | none }
371 bsp: External NMI is delivered only to CPU 0
372 all: External NMIs are broadcast to all CPUs as a
374 none: External NMI is masked for all CPUs. This is
375 useful so that a dump capture kernel won't be
379 See Documentation/networking/ipv6.rst.
381 apm= [APM] Advanced Power Management
382 See header of arch/x86/kernel/apm_32.c.
384 apparmor= [APPARMOR] Disable or enable AppArmor at boot time
385 Format: { "0" | "1" }
386 See security/apparmor/Kconfig help text
389 Default value is set via kernel config option.
391 arcrimi= [HW,NET] ARCnet - "RIM I" (entirely mem-mapped) cards
392 Format: <io>,<irq>,<nodeID>
394 arm64.nobti [ARM64] Unconditionally disable Branch Target
395 Identification support
397 arm64.nopauth [ARM64] Unconditionally disable Pointer Authentication
400 arm64.nomte [ARM64] Unconditionally disable Memory Tagging Extension
403 arm64.nosve [ARM64] Unconditionally disable Scalable Vector
406 arm64.nosme [ARM64] Unconditionally disable Scalable Matrix
411 atarimouse= [HW,MOUSE] Atari Mouse
413 atkbd.extra= [HW] Enable extra LEDs and keys on IBM RapidAccess,
414 EzKey and similar keyboards
416 atkbd.reset= [HW] Reset keyboard during initialization
418 atkbd.set= [HW] Select keyboard code set
419 Format: <int> (2 = AT (default), 3 = PS/2)
421 atkbd.scroll= [HW] Enable scroll wheel on MS Office and similar
424 atkbd.softraw= [HW] Choose between synthetic and real raw mode
425 Format: <bool> (0 = real, 1 = synthetic (default))
427 atkbd.softrepeat= [HW]
428 Use software keyboard repeat
430 audit= [KNL] Enable the audit sub-system
431 Format: { "0" | "1" | "off" | "on" }
432 0 | off - kernel audit is disabled and can not be
433 enabled until the next reboot
434 unset - kernel audit is initialized but disabled and
435 will be fully enabled by the userspace auditd.
436 1 | on - kernel audit is initialized and partially
437 enabled, storing at most audit_backlog_limit
438 messages in RAM until it is fully enabled by the
442 audit_backlog_limit= [KNL] Set the audit queue size limit.
443 Format: <int> (must be >=0)
446 bau= [X86_UV] Enable the BAU on SGI UV. The default
447 behavior is to disable the BAU (i.e. bau=0).
448 Format: { "0" | "1" }
451 unset - Disable the BAU.
453 baycom_epp= [HW,AX25]
456 baycom_par= [HW,AX25] BayCom Parallel Port AX.25 Modem
458 See header of drivers/net/hamradio/baycom_par.c.
460 baycom_ser_fdx= [HW,AX25]
461 BayCom Serial Port AX.25 Modem (Full Duplex Mode)
462 Format: <io>,<irq>,<mode>[,<baud>]
463 See header of drivers/net/hamradio/baycom_ser_fdx.c.
465 baycom_ser_hdx= [HW,AX25]
466 BayCom Serial Port AX.25 Modem (Half Duplex Mode)
467 Format: <io>,<irq>,<mode>
468 See header of drivers/net/hamradio/baycom_ser_hdx.c.
471 Disable BERT OS support on buggy BIOSes.
473 bgrt_disable [ACPI][X86]
474 Disable BGRT to avoid flickering OEM logo.
476 blkdevparts= Manual partition parsing of block device(s) for
477 embedded devices based on command line input.
478 See Documentation/block/cmdline-partition.rst
480 boot_delay= Milliseconds to delay each printk during boot.
481 Only works if CONFIG_BOOT_PRINTK_DELAY is enabled,
482 and you may also have to specify "lpj=". Boot_delay
483 values larger than 10 seconds (10000) are assumed
484 erroneous and ignored.
488 Extended command line options can be added to an initrd
489 and this will cause the kernel to look for it.
491 See Documentation/admin-guide/bootconfig.rst
493 bttv.card= [HW,V4L] bttv (bt848 + bt878 based grabber cards)
494 bttv.radio= Most important insmod options are available as
496 bttv.pll= See Documentation/admin-guide/media/bttv.rst
499 bulk_remove=off [PPC] This parameter disables the use of the pSeries
500 firmware feature for flushing multiple hpte entries
503 c101= [NET] Moxa C101 synchronous serial card
505 cachesize= [BUGS=X86-32] Override level 2 CPU cache size detection.
506 Sometimes CPU hardware bugs make them report the cache
507 size incorrectly. The kernel will attempt work arounds
508 to fix known problems, but for some CPUs it is not
509 possible to determine what the correct size should be.
510 This option provides an override for these situations.
513 [NET] Specifies amount of time (in seconds) that
514 the kernel should wait for a network carrier. By default
515 it waits 120 seconds.
517 ca_keys= [KEYS] This parameter identifies a specific key(s) on
518 the system trusted keyring to be used for certificate
520 format: { id:<keyid> | builtin }
522 cca= [MIPS] Override the kernel pages' cache coherency
523 algorithm. Accepted values range from 0 to 7
524 inclusive. See arch/mips/include/asm/pgtable-bits.h
525 for platform specific values (SB1, Loongson3 and
528 ccw_timeout_log [S390]
529 See Documentation/s390/common_io.rst for details.
531 cgroup_disable= [KNL] Disable a particular controller or optional feature
532 Format: {name of the controller(s) or feature(s) to disable}
533 The effects of cgroup_disable=foo are:
534 - foo isn't auto-mounted if you mount all cgroups in
536 - foo isn't visible as an individually mountable
538 - if foo is an optional feature then the feature is
539 disabled and corresponding cgroup files are not
541 {Currently only "memory" controller deal with this and
542 cut the overhead, others just disable the usage. So
543 only cgroup_disable=memory is actually worthy}
544 Specifying "pressure" disables per-cgroup pressure
545 stall information accounting feature
547 cgroup_no_v1= [KNL] Disable cgroup controllers and named hierarchies in v1
548 Format: { { controller | "all" | "named" }
549 [,{ controller | "all" | "named" }...] }
550 Like cgroup_disable, but only applies to cgroup v1;
551 the blacklisted controllers remain available in cgroup2.
552 "all" blacklists all controllers and "named" disables
553 named mounts. Specifying both "all" and "named" disables
556 cgroup.memory= [KNL] Pass options to the cgroup memory controller.
558 nosocket -- Disable socket memory accounting.
559 nokmem -- Disable kernel memory accounting.
560 nobpf -- Disable BPF memory accounting.
562 checkreqprot= [SELINUX] Set initial checkreqprot flag value.
563 Format: { "0" | "1" }
564 See security/selinux/Kconfig help text.
565 0 -- check protection applied by kernel (includes
566 any implied execute protection).
567 1 -- check protection requested by application.
568 Default value is set via a kernel config option.
569 Value can be changed at runtime via
570 /sys/fs/selinux/checkreqprot.
571 Setting checkreqprot to 1 is deprecated.
574 See Documentation/s390/common_io.rst for details.
576 clearcpuid=X[,X...] [X86]
577 Disable CPUID feature X for the kernel. See
578 arch/x86/include/asm/cpufeatures.h for the valid bit
579 numbers X. Note the Linux-specific bits are not necessarily
580 stable over kernel options, but the vendor-specific
582 X can also be a string as appearing in the flags: line
583 in /proc/cpuinfo which does not have the above
584 instability issue. However, not all features have names
586 Note that using this option will taint your kernel.
587 Also note that user programs calling CPUID directly
588 or using the feature without checking anything
589 will still see it. This just prevents it from
590 being used by the kernel or shown in /proc/cpuinfo.
591 Also note the kernel might malfunction if you disable
596 Prevents the clock framework from automatically gating
597 clocks that have not been explicitly enabled by a Linux
598 device driver but are enabled in hardware at reset or
599 by the bootloader/firmware. Note that this does not
600 force such clocks to be always-on nor does it reserve
601 those clocks in any way. This parameter is useful for
602 debug and development, but should not be needed on a
603 platform with proper driver support. For more
604 information, see Documentation/driver-api/clk.rst.
606 clock= [BUGS=X86-32, HW] gettimeofday clocksource override.
608 Forces specified clocksource (if available) to be used
609 when calculating gettimeofday(). If specified
610 clocksource is not available, it defaults to PIT.
611 Format: { pit | tsc | cyclone | pmtmr }
613 clocksource= Override the default clocksource
615 Override the default clocksource and use the clocksource
616 with the name specified.
617 Some clocksource names to choose from, depending on
619 [all] jiffies (this is the base, fallback clocksource)
621 [ARM] imx_timer1,OSTS,netx_timer,mpu_timer2,
622 pxa_timer,timer3,32k_counter,timer0_1
623 [X86-32] pit,hpet,tsc;
624 scx200_hrt on Geode; cyclone on IBM x440
632 clocksource.arm_arch_timer.evtstrm=
635 Enable/disable the eventstream feature of the ARM
636 architected timer so that code using WFE-based polling
637 loops can be debugged more effectively on production
640 clocksource.max_cswd_read_retries= [KNL]
641 Number of clocksource_watchdog() retries due to
642 external delays before the clock will be marked
643 unstable. Defaults to two retries, that is,
644 three attempts to read the clock under test.
646 clocksource.verify_n_cpus= [KNL]
647 Limit the number of CPUs checked for clocksources
648 marked with CLOCK_SOURCE_VERIFY_PERCPU that
649 are marked unstable due to excessive skew.
650 A negative value says to check all CPUs, while
651 zero says not to check any. Values larger than
652 nr_cpu_ids are silently truncated to nr_cpu_ids.
653 The actual CPUs are chosen randomly, with
654 no replacement if the same CPU is chosen twice.
656 clocksource-wdtest.holdoff= [KNL]
657 Set the time in seconds that the clocksource
658 watchdog test waits before commencing its tests.
659 Defaults to zero when built as a module and to
660 10 seconds when built into the kernel.
662 cma=nn[MG]@[start[MG][-end[MG]]]
664 Sets the size of kernel global memory area for
665 contiguous memory allocations and optionally the
666 placement constraint by the physical address range of
667 memory allocations. A value of 0 disables CMA
668 altogether. For more information, see
669 kernel/dma/contiguous.c
673 Sets the size of kernel per-numa memory area for
674 contiguous memory allocations. A value of 0 disables
675 per-numa CMA altogether. And If this option is not
676 specified, the default value is 0.
677 With per-numa CMA enabled, DMA users on node nid will
678 first try to allocate buffer from the pernuma area
679 which is located in node nid, if the allocation fails,
680 they will fallback to the global default memory area.
682 cmo_free_hint= [PPC] Format: { yes | no }
683 Specify whether pages are marked as being inactive
684 when they are freed. This is used in CMO environments
685 to determine OS memory pressure for page stealing by
689 coherent_pool=nn[KMG] [ARM,KNL]
690 Sets the size of memory pool for coherent, atomic dma
691 allocations, by default set to 256K.
693 com20020= [HW,NET] ARCnet - COM20020 chipset
695 <io>[,<irq>[,<nodeID>[,<backplane>[,<ckp>[,<timeout>]]]]]
697 com90io= [HW,NET] ARCnet - COM90xx chipset (IO-mapped buffers)
701 ARCnet - COM90xx chipset (memory-mapped buffers)
702 Format: <io>[,<irq>[,<memstart>]]
704 condev= [HW,S390] console device
707 con3215_drop= [S390] 3215 console drop mode.
709 When set to true, drop data on the 3215 console when
710 the console buffer is full. In this case the
711 operator using a 3270 terminal emulator (for example
712 x3270) does not have to enter the clear key for the
713 console output to advance and the kernel to continue.
714 This leads to a much faster boot time when a 3270
715 terminal emulator is active. If no 3270 terminal
716 emulator is used, this parameter has no effect.
718 console= [KNL] Output console device and options.
720 tty<n> Use the virtual console device <n>.
724 Use the specified serial port. The options are of
725 the form "bbbbpnf", where "bbbb" is the baud rate,
726 "p" is parity ("n", "o", or "e"), "n" is number of
727 bits, and "f" is flow control ("r" for RTS or
728 omit it). Default is "9600n8".
730 See Documentation/admin-guide/serial-console.rst for more
732 Documentation/networking/netconsole.rst for an
735 uart[8250],io,<addr>[,options]
736 uart[8250],mmio,<addr>[,options]
737 uart[8250],mmio16,<addr>[,options]
738 uart[8250],mmio32,<addr>[,options]
739 uart[8250],0x<addr>[,options]
740 Start an early, polled-mode console on the 8250/16550
741 UART at the specified I/O port or MMIO address,
742 switching to the matching ttyS device later.
743 MMIO inter-register address stride is either 8-bit
744 (mmio), 16-bit (mmio16), or 32-bit (mmio32).
745 If none of [io|mmio|mmio16|mmio32], <addr> is assumed
746 to be equivalent to 'mmio'. 'options' are specified in
747 the same format described for ttyS above; if unspecified,
748 the h/w is not re-initialized.
750 hvc<n> Use the hypervisor console device <n>. This is for
751 both Xen and PowerPC hypervisors.
754 Use to disable console output, i.e., to have kernel
755 console messages discarded.
756 This must be the only console= parameter used on the
759 If the device connected to the port is not a TTY but a braille
760 device, prepend "brl," before the device type, for instance
762 For now, only VisioBraille is supported.
765 [KNL] Change console messages format
767 By default we print messages on consoles in
768 "[time stamp] text\n" format (time stamp may not be
769 printed, depending on CONFIG_PRINTK_TIME or
770 `printk_time' param).
772 Switch to syslog format: "<%u>[time stamp] text\n"
773 IOW, each message will have a facility and loglevel
774 prefix. The format is similar to one used by syslog()
775 syscall, or to executing "dmesg -S --raw" or to reading
778 consoleblank= [KNL] The console blank (screen saver) timeout in
779 seconds. A value of 0 disables the blank timer.
783 [KNL] Change the default value for
784 /proc/<pid>/coredump_filter.
785 See also Documentation/filesystems/proc.rst.
787 coresight_cpu_debug.enable
790 Enable/disable the CPU sampling based debugging.
791 0: default value, disable debugging
792 1: enable debugging at boot time
794 cpcihp_generic= [HW,PCI] Generic port I/O CompactPCI driver
796 <first_slot>,<last_slot>,<port>,<enum_bit>[,<debug>]
798 cpu0_hotplug [X86] Turn on CPU0 hotplug feature when
799 CONFIG_BOOTPARAM_HOTPLUG_CPU0 is off.
800 Some features depend on CPU0. Known dependencies are:
801 1. Resume from suspend/hibernate depends on CPU0.
802 Suspend/hibernate will fail if CPU0 is offline and you
803 need to online CPU0 before suspend/hibernate.
804 2. PIC interrupts also depend on CPU0. CPU0 can't be
805 removed if a PIC interrupt is detected.
806 It's said poweroff/reboot may depend on CPU0 on some
807 machines although I haven't seen such issues so far
808 after CPU0 is offline on a few tested machines.
809 If the dependencies are under your control, you can
810 turn on cpu0_hotplug.
812 cpuidle.off=1 [CPU_IDLE]
813 disable the cpuidle sub-system
816 [CPU_IDLE] Name of the cpuidle governor to use.
818 cpufreq.off=1 [CPU_FREQ]
819 disable the cpufreq sub-system
821 cpufreq.default_governor=
822 [CPU_FREQ] Name of the default cpufreq governor or
823 policy to use. This governor must be registered in the
824 kernel before the cpufreq driver probes.
827 [X86] Delay for N microsec between assert and de-assert
828 of APIC INIT to start processors. This delay occurs
829 on every CPU online, such as boot, and resume from suspend.
832 crash_kexec_post_notifiers
833 Run kdump after running panic-notifiers and dumping
834 kmsg. This only for the users who doubt kdump always
835 succeeds in any situation.
836 Note that this also increases risks of kdump failure,
837 because some panic notifiers can make the crashed
838 kernel more unstable.
840 crashkernel=size[KMG][@offset[KMG]]
841 [KNL] Using kexec, Linux can switch to a 'crash kernel'
842 upon panic. This parameter reserves the physical
843 memory region [offset, offset + size] for that kernel
844 image. If '@offset' is omitted, then a suitable offset
845 is selected automatically.
846 [KNL, X86-64, ARM64] Select a region under 4G first, and
847 fall back to reserve region above 4G when '@offset'
848 hasn't been specified.
849 See Documentation/admin-guide/kdump/kdump.rst for further details.
851 crashkernel=range1:size1[,range2:size2,...][@offset]
852 [KNL] Same as above, but depends on the memory
853 in the running system. The syntax of range is
854 start-[end] where start and end are both
855 a memory unit (amount[KMG]). See also
856 Documentation/admin-guide/kdump/kdump.rst for an example.
858 crashkernel=size[KMG],high
859 [KNL, X86-64, ARM64] range could be above 4G. Allow kernel
860 to allocate physical memory region from top, so could
861 be above 4G if system have more than 4G ram installed.
862 Otherwise memory region will be allocated below 4G, if
864 It will be ignored if crashkernel=X is specified.
865 crashkernel=size[KMG],low
866 [KNL, X86-64, ARM64] range under 4G. When crashkernel=X,high
867 is passed, kernel could allocate physical memory region
868 above 4G, that cause second kernel crash on system
869 that require some amount of low memory, e.g. swiotlb
870 requires at least 64M+32K low memory, also enough extra
871 low memory is needed to make sure DMA buffers for 32-bit
872 devices won't run out. Kernel would try to allocate
873 default size of memory below 4G automatically. The default
874 size is platform dependent.
875 --> x86: max(swiotlb_size_or_default() + 8MiB, 256MiB)
877 This one lets the user specify own low range under 4G
878 for second kernel instead.
879 0: to disable low allocation.
880 It will be ignored when crashkernel=X,high is not used
881 or memory reserved is below 4G.
884 [KNL] Disable crypto self-tests
889 cs89x0_media= [HW,NET]
890 Format: { rj45 | aui | bnc }
892 csdlock_debug= [KNL] Enable debug add-ons of cross-CPU function call
893 handling. When switched on, additional debug data is
894 printed to the console in case a hanging CPU is
895 detected, and that CPU is pinged again in order to try
896 to resolve the hang situation.
897 0: disable csdlock debugging (default)
898 1: enable basic csdlock debugging (minor impact)
899 ext: enable extended csdlock debugging (more impact,
903 See header of drivers/s390/block/dasd_devmap.c.
905 db9.dev[2|3]= [HW,JOY] Multisystem joystick support via parallel port
906 (one device per port)
907 Format: <port#>,<type>
908 See also Documentation/input/devices/joystick-parport.rst
910 debug [KNL] Enable kernel debugging (events log level).
913 [KNL] Enable printing [hashed] pointers early in the
914 boot sequence. If enabled, we use a weak hash instead
915 of siphash to hash pointers. Use this option if you are
916 seeing instances of '(___ptrval___)') and need to see a
917 value (hashed pointer) instead. Cryptographically
918 insecure, please do not use on production kernels.
921 [KNL] verbose locking self-tests
923 Print debugging info while doing the locking API
925 Bitmask for the various LOCKTYPE_ tests. Defaults to 0
926 (no extra messages), setting it to -1 (all bits set)
927 will print _a_lot_ more information - normally only
928 useful to lockdep developers.
930 debug_objects [KNL] Enable object debugging
933 [KNL] Disable object debugging
935 debug_guardpage_minorder=
936 [KNL] When CONFIG_DEBUG_PAGEALLOC is set, this
937 parameter allows control of the order of pages that will
938 be intentionally kept free (and hence protected) by the
939 buddy allocator. Bigger value increase the probability
940 of catching random memory corruption, but reduce the
941 amount of memory for normal system use. The maximum
942 possible value is MAX_ORDER/2. Setting this parameter
943 to 1 or 2 should be enough to identify most random
944 memory corruption problems caused by bugs in kernel or
945 driver code when a CPU writes to (or reads from) a
946 random memory location. Note that there exists a class
947 of memory corruptions problems caused by buggy H/W or
948 F/W or by drivers badly programming DMA (basically when
949 memory is written at bus level and the CPU MMU is
950 bypassed) which are not detectable by
951 CONFIG_DEBUG_PAGEALLOC, hence this option will not help
952 tracking down these problems.
955 [KNL] When CONFIG_DEBUG_PAGEALLOC is set, this parameter
956 enables the feature at boot time. By default, it is
957 disabled and the system will work mostly the same as a
958 kernel built without CONFIG_DEBUG_PAGEALLOC.
959 Note: to get most of debug_pagealloc error reports, it's
960 useful to also enable the page_owner functionality.
961 on: enable the feature
963 debugfs= [KNL] This parameter enables what is exposed to userspace
964 and debugfs internal clients.
965 Format: { on, no-mount, off }
966 on: All functions are enabled.
968 Filesystem is not registered but kernel clients can
969 access APIs and a crashkernel can be used to read
970 its content. There is nothing to mount.
971 off: Filesystem is not registered and clients
972 get a -EPERM as result when trying to register files
973 or directories within debugfs.
974 This is equivalent of the runtime functionality if
975 debugfs was not enabled in the kernel at all.
976 Default value is set in build-time with a kernel configuration.
978 debugpat [X86] Enable PAT debugging
981 [HW] The size of the default HugeTLB page. This is
982 the size represented by the legacy /proc/ hugepages
983 APIs. In addition, this is the default hugetlb size
984 used for shmget(), mmap() and mounting hugetlbfs
985 filesystems. If not specified, defaults to the
986 architecture's default huge page size. Huge page
987 sizes are architecture dependent. See also
988 Documentation/admin-guide/mm/hugetlbpage.rst.
991 deferred_probe_timeout=
992 [KNL] Debugging option to set a timeout in seconds for
993 deferred probe to give up waiting on dependencies to
994 probe. Only specific dependencies (subsystems or
995 drivers) that have opted in will be ignored. A timeout
996 of 0 will timeout at the end of initcalls. If the time
997 out hasn't expired, it'll be restarted by each
998 successful driver registration. This option will also
999 dump out devices still on the deferred probe list after
1002 delayacct [KNL] Enable per-task delay accounting
1004 dell_smm_hwmon.ignore_dmi=
1005 [HW] Continue probing hardware even if DMI data
1006 indicates that the driver is running on unsupported
1009 dell_smm_hwmon.force=
1010 [HW] Activate driver even if SMM BIOS signature does
1011 not match list of supported models and enable otherwise
1012 blacklisted features.
1014 dell_smm_hwmon.power_status=
1015 [HW] Report power status in /proc/i8k
1016 (disabled by default).
1018 dell_smm_hwmon.restricted=
1019 [HW] Allow controlling fans only if SYS_ADMIN
1022 dell_smm_hwmon.fan_mult=
1023 [HW] Factor to multiply fan speed with.
1025 dell_smm_hwmon.fan_max=
1026 [HW] Maximum configurable fan speed.
1029 Format: { on | off | def_only | inf_only | always }
1030 on: s390 zlib hardware support for compression on
1031 level 1 and decompression (default)
1032 off: No s390 zlib hardware support
1033 def_only: s390 zlib hardware support for deflate
1034 only (compression on level 1)
1035 inf_only: s390 zlib hardware support for inflate
1036 only (decompression)
1037 always: Same as 'on' but ignores the selected compression
1038 level always using hardware support (used for debugging)
1040 dhash_entries= [KNL]
1041 Set number of hash buckets for dentry cache.
1043 disable_1tb_segments [PPC]
1044 Disables the use of 1TB hash page table segments. This
1045 causes the kernel to fall back to 256MB segments which
1046 can be useful when debugging issues that require an SLB
1050 See Documentation/networking/ipv6.rst.
1053 Disable RADIX MMU mode on POWER9
1056 Disable TLBIE instruction. Currently does not work
1057 with KVM, with HASH MMU, or with coherent accelerators.
1059 disable_cpu_apicid= [X86,APIC,SMP]
1061 The number of initial APIC ID for the
1062 corresponding CPU to be disabled at boot,
1063 mostly used for the kdump 2nd kernel to
1064 disable BSP to wake up multiple CPUs without
1065 causing system reset or hang due to sending
1066 INIT from AP to BSP.
1068 disable_ddw [PPC/PSERIES]
1069 Disable Dynamic DMA Window support. Use this
1070 to workaround buggy firmware.
1072 disable_ipv6= [IPV6]
1073 See Documentation/networking/ipv6.rst.
1075 disable_mtrr_cleanup [X86]
1076 The kernel tries to adjust MTRR layout from continuous
1077 to discrete, to make X server driver able to add WB
1078 entry later. This parameter disables that.
1080 disable_mtrr_trim [X86, Intel and AMD only]
1081 By default the kernel will trim any uncacheable
1082 memory out of your available memory pool based on
1083 MTRR settings. This parameter disables that behavior,
1084 possibly causing your machine to run very slowly.
1086 disable_timer_pin_1 [X86]
1087 Disable PIN 1 of APIC timer
1088 Can be useful to work around chipset bugs.
1090 dis_ucode_ldr [X86] Disable the microcode loader.
1092 dma_debug=off If the kernel is compiled with DMA_API_DEBUG support,
1093 this option disables the debugging code at boot.
1095 dma_debug_entries=<number>
1096 This option allows to tune the number of preallocated
1097 entries for DMA-API debugging code. One entry is
1098 required per DMA-API allocation. Use this if the
1099 DMA-API debugging code disables itself because the
1100 architectural default is too low.
1102 dma_debug_driver=<driver_name>
1103 With this option the DMA-API debugging driver
1104 filter feature can be enabled at boot time. Just
1105 pass the driver to filter for as the parameter.
1106 The filter can be disabled or changed to another
1107 driver later using sysfs.
1109 driver_async_probe= [KNL]
1110 List of driver names to be probed asynchronously. *
1111 matches with all driver names. If * is specified, the
1112 rest of the listed driver names are those that will NOT
1114 Format: <driver_name1>,<driver_name2>...
1116 drm.edid_firmware=[<connector>:]<file>[,[<connector>:]<file>]
1117 Broken monitors, graphic adapters, KVMs and EDIDless
1118 panels may send no or incorrect EDID data sets.
1119 This parameter allows to specify an EDID data sets
1120 in the /lib/firmware directory that are used instead.
1121 Generic built-in EDID data sets are used, if one of
1122 edid/1024x768.bin, edid/1280x1024.bin,
1123 edid/1680x1050.bin, or edid/1920x1080.bin is given
1124 and no file with the same name exists. Details and
1125 instructions how to build your own EDID data are
1126 available in Documentation/admin-guide/edid.rst. An EDID
1127 data set will only be used for a particular connector,
1128 if its name and a colon are prepended to the EDID
1129 name. Each connector may use a unique EDID data
1130 set by separating the files with a comma. An EDID
1131 data set with no connector name will be used for
1132 any connectors not explicitly specified.
1137 Format: {"off" | "known"}
1138 Control how the dt_cpu_ftrs device-tree binding is
1139 used for CPU feature discovery and setup (if it
1141 off: Do not use it, fall back to legacy cpu table.
1142 known: Do not pass through unknown features to guests
1143 or userspace, only those that the kernel is aware of.
1145 dump_apple_properties [X86]
1146 Dump name and content of EFI device properties on
1147 x86 Macs. Useful for driver authors to determine
1148 what data is available or for reverse-engineering.
1150 dyndbg[="val"] [KNL,DYNAMIC_DEBUG]
1151 <module>.dyndbg[="val"]
1152 Enable debug messages at boot time. See
1153 Documentation/admin-guide/dynamic-debug-howto.rst
1156 early_ioremap_debug [KNL]
1157 Enable debug messages in early_ioremap support. This
1158 is useful for tracking down temporary early mappings
1159 which are not unmapped.
1161 earlycon= [KNL] Output early console device and options.
1163 When used with no options, the early console is
1164 determined by stdout-path property in device tree's
1165 chosen node or the ACPI SPCR table if supported by
1168 cdns,<addr>[,options]
1169 Start an early, polled-mode console on a Cadence
1170 (xuartps) serial port at the specified address. Only
1171 supported option is baud rate. If baud rate is not
1172 specified, the serial port must already be setup and
1175 uart[8250],io,<addr>[,options[,uartclk]]
1176 uart[8250],mmio,<addr>[,options[,uartclk]]
1177 uart[8250],mmio32,<addr>[,options[,uartclk]]
1178 uart[8250],mmio32be,<addr>[,options[,uartclk]]
1179 uart[8250],0x<addr>[,options]
1180 Start an early, polled-mode console on the 8250/16550
1181 UART at the specified I/O port or MMIO address.
1182 MMIO inter-register address stride is either 8-bit
1183 (mmio) or 32-bit (mmio32 or mmio32be).
1184 If none of [io|mmio|mmio32|mmio32be], <addr> is assumed
1185 to be equivalent to 'mmio'. 'options' are specified
1186 in the same format described for "console=ttyS<n>"; if
1187 unspecified, the h/w is not initialized. 'uartclk' is
1188 the uart clock frequency; if unspecified, it is set
1189 to 'BASE_BAUD' * 16.
1193 Start an early, polled-mode console on a pl011 serial
1194 port at the specified address. The pl011 serial port
1195 must already be setup and configured. Options are not
1196 yet supported. If 'mmio32' is specified, then only
1197 the driver will use only 32-bit accessors to read/write
1198 the device registers.
1201 Start an early console on a litex serial port at the
1202 specified address. The serial port must already be
1203 setup and configured. Options are not yet supported.
1206 Start an early, polled-mode console on a meson serial
1207 port at the specified address. The serial port must
1208 already be setup and configured. Options are not yet
1212 Start an early, polled-mode console on an msm serial
1213 port at the specified address. The serial port
1214 must already be setup and configured. Options are not
1217 msm_serial_dm,<addr>
1218 Start an early, polled-mode console on an msm serial
1219 dm port at the specified address. The serial port
1220 must already be setup and configured. Options are not
1224 Start an early, polled-mode console on a serial port
1225 of an Actions Semi SoC, such as S500 or S900, at the
1226 specified address. The serial port must already be
1227 setup and configured. Options are not yet supported.
1230 Start an early, polled-mode console on a serial port
1231 of an RDA Micro SoC, such as RDA8810PL, at the
1232 specified address. The serial port must already be
1233 setup and configured. Options are not yet supported.
1236 Use RISC-V SBI (Supervisor Binary Interface) for early
1239 smh Use ARM semihosting calls for early console.
1247 Use early console provided by serial driver available
1248 on Samsung SoCs, requires selecting proper type and
1249 a correct base address of the selected UART port. The
1250 serial port must already be setup and configured.
1251 Options are not yet supported.
1254 Start an early, polled-mode console on a lantiq serial
1255 (lqasc) port at the specified address. The serial port
1256 must already be setup and configured. Options are not
1261 Use early console provided by Freescale LP UART driver
1262 found on Freescale Vybrid and QorIQ LS1021A processors.
1263 A valid base address must be provided, and the serial
1264 port must already be setup and configured.
1268 Start an early, polled-mode, output-only console on the
1269 Freescale i.MX UART at the specified address. The UART
1270 must already be setup and configured.
1273 Start an early, polled-mode console on the
1274 Armada 3700 serial port at the specified
1275 address. The serial port must already be setup
1276 and configured. Options are not yet supported.
1279 Start an early, polled-mode console on a Qualcomm
1280 Generic Interface (GENI) based serial port at the
1281 specified address. The serial port must already be
1282 setup and configured. Options are not yet supported.
1285 Start an early, unaccelerated console on the EFI
1286 memory mapped framebuffer (if available). On cache
1287 coherent non-x86 systems that use system memory for
1288 the framebuffer, pass the 'ram' option so that it is
1289 mapped with the correct attributes.
1292 Use early console provided by Freescale LINFlexD UART
1293 serial driver for NXP S32V234 SoCs. A valid base
1294 address must be provided, and the serial port must
1295 already be setup and configured.
1297 earlyprintk= [X86,SH,ARM,M68k,S390]
1301 earlyprintk=serial[,ttySn[,baudrate]]
1302 earlyprintk=serial[,0x...[,baudrate]]
1303 earlyprintk=ttySn[,baudrate]
1304 earlyprintk=dbgp[debugController#]
1305 earlyprintk=pciserial[,force],bus:device.function[,baudrate]
1306 earlyprintk=xdbc[xhciController#]
1308 earlyprintk is useful when the kernel crashes before
1309 the normal console is initialized. It is not enabled by
1310 default because it has some cosmetic problems.
1312 Append ",keep" to not disable it when the real console
1315 Only one of vga, serial, or usb debug port can
1318 Currently only ttyS0 and ttyS1 may be specified by
1319 name. Other I/O ports may be explicitly specified
1320 on some architectures (x86 and arm at least) by
1321 replacing ttySn with an I/O port address, like this:
1322 earlyprintk=serial,0x1008,115200
1323 You can find the port for a given device in
1324 /proc/tty/driver/serial:
1325 2: uart:ST16650V2 port:00001008 irq:18 ...
1327 Interaction with the standard serial driver is not
1330 The VGA output is eventually overwritten by
1333 The xen option can only be used in Xen domains.
1335 The sclp output can only be used on s390.
1337 The optional "force" to "pciserial" enables use of a
1338 PCI device even when its classcode is not of the
1341 edac_report= [HW,EDAC] Control how to report EDAC event
1342 Format: {"on" | "off" | "force"}
1343 on: enable EDAC to report H/W event. May be overridden
1344 by other higher priority error reporting module.
1345 off: disable H/W event reporting through EDAC.
1346 force: enforce the use of EDAC to report H/W event.
1350 Format: {"off" | "on" | "skip[mbr]"}
1353 Format: { "debug", "disable_early_pci_dma",
1354 "nochunk", "noruntime", "nosoftreserve",
1355 "novamap", "no_disable_early_pci_dma" }
1356 debug: enable misc debug output.
1357 disable_early_pci_dma: disable the busmaster bit on all
1358 PCI bridges while in the EFI boot stub.
1359 nochunk: disable reading files in "chunks" in the EFI
1360 boot stub, as chunking can cause problems with some
1361 firmware implementations.
1362 noruntime : disable EFI runtime services support
1363 nosoftreserve: The EFI_MEMORY_SP (Specific Purpose)
1364 attribute may cause the kernel to reserve the
1365 memory range for a memory mapping driver to
1366 claim. Specify efi=nosoftreserve to disable this
1367 reservation and treat the memory by its base type
1368 (i.e. EFI_CONVENTIONAL_MEMORY / "System RAM").
1369 novamap: do not call SetVirtualAddressMap().
1370 no_disable_early_pci_dma: Leave the busmaster bit set
1371 on all PCI bridges while in the EFI boot stub
1373 efi_no_storage_paranoia [EFI; X86]
1374 Using this parameter you can use more than 50% of
1375 your efi variable storage. Use this parameter only if
1376 you are really sure that your UEFI does sane gc and
1377 fulfills the spec otherwise your board may brick.
1379 efi_fake_mem= nn[KMG]@ss[KMG]:aa[,nn[KMG]@ss[KMG]:aa,..] [EFI; X86]
1380 Add arbitrary attribute to specific memory range by
1381 updating original EFI memory map.
1382 Region of memory which aa attribute is added to is
1385 If efi_fake_mem=2G@4G:0x10000,2G@0x10a0000000:0x10000
1386 is specified, EFI_MEMORY_MORE_RELIABLE(0x10000)
1387 attribute is added to range 0x100000000-0x180000000 and
1388 0x10a0000000-0x1120000000.
1390 If efi_fake_mem=8G@9G:0x40000 is specified, the
1391 EFI_MEMORY_SP(0x40000) attribute is added to
1392 range 0x240000000-0x43fffffff.
1394 Using this parameter you can do debugging of EFI memmap
1395 related features. For example, you can do debugging of
1396 Address Range Mirroring feature even if your box
1397 doesn't support it, or mark specific memory as
1400 efivar_ssdt= [EFI; X86] Name of an EFI variable that contains an SSDT
1401 that is to be dynamically loaded by Linux. If there are
1402 multiple variables with the same name but with different
1403 vendor GUIDs, all of them will be loaded. See
1404 Documentation/admin-guide/acpi/ssdt-overlays.rst for details.
1407 eisa_irq_edge= [PARISC,HW]
1408 See header of drivers/parisc/eisa.c.
1410 ekgdboc= [X86,KGDB] Allow early kernel console debugging
1413 This is designed to be used in conjunction with
1414 the boot argument: earlyprintk=vga
1416 This parameter works in place of the kgdboc parameter
1417 but can only be used if the backing tty is available
1418 very early in the boot process. For early debugging
1419 via a serial port see kgdboc_earlycon instead.
1422 See comment before function elanfreq_setup() in
1423 arch/x86/kernel/cpu/cpufreq/elanfreq.c.
1425 elfcorehdr=[size[KMG]@]offset[KMG] [IA64,PPC,SH,X86,S390]
1426 Specifies physical address of start of kernel core
1427 image elf header and optionally the size. Generally
1428 kexec loader will pass this option to capture kernel.
1429 See Documentation/admin-guide/kdump/kdump.rst for details.
1431 enable_mtrr_cleanup [X86]
1432 The kernel tries to adjust MTRR layout from continuous
1433 to discrete, to make X server driver able to add WB
1434 entry later. This parameter enables that.
1436 enable_timer_pin_1 [X86]
1437 Enable PIN 1 of APIC timer
1438 Can be useful to work around chipset bugs
1439 (in particular on some ATI chipsets).
1440 The kernel tries to set a reasonable default.
1442 enforcing= [SELINUX] Set initial enforcing status.
1444 See security/selinux/Kconfig help text.
1445 0 -- permissive (log only, no denials).
1446 1 -- enforcing (deny and log).
1448 Value can be changed at runtime via
1449 /sys/fs/selinux/enforce.
1452 Disable Error Record Serialization Table (ERST)
1455 ether= [HW,NET] Ethernet cards parameters
1456 This option is obsoleted by the "netdev=" option, which
1457 has equivalent usage. See its documentation for details.
1461 Permit 'security.evm' to be updated regardless of
1462 current integrity status.
1464 early_page_ext [KNL] Enforces page_ext initialization to earlier
1465 stages so cover more early boot allocations.
1466 Please note that as side effect some optimizations
1467 might be disabled to achieve that (e.g. parallelized
1468 memory initialization is disabled) so the boot process
1469 might take longer, especially on systems with a lot of
1470 memory. Available with CONFIG_PAGE_EXTENSION=y.
1475 fail_make_request=[KNL]
1476 General fault injection mechanism.
1477 Format: <interval>,<probability>,<space>,<times>
1478 See also Documentation/fault-injection/.
1481 Format: { initns | none }
1482 See Documentation/admin-guide/sysctl/net.rst for
1483 fb_tunnels_only_for_init_ns
1486 See Documentation/admin-guide/blockdev/floppy.rst.
1488 force_pal_cache_flush
1489 [IA-64] Avoid check_sal_cache_flush which may hang on
1490 buggy SAL_CACHE_FLUSH implementations. Using this
1491 parameter will force ia64_sal_cache_flush to call
1492 ia64_pal_cache_flush instead of SAL_CACHE_FLUSH.
1495 Forcefully enable Physical Address Extension (PAE).
1496 Many Pentium M systems disable PAE but may have a
1497 functionally usable PAE implementation.
1498 Warning: use of this parameter will taint the kernel
1499 and may cause unknown problems.
1502 [FTRACE] will set and start the specified tracer
1503 as early as possible in order to facilitate early
1506 ftrace_boot_snapshot
1507 [FTRACE] On boot up, a snapshot will be taken of the
1508 ftrace ring buffer that can be read at:
1509 /sys/kernel/tracing/snapshot.
1510 This is useful if you need tracing information from kernel
1511 boot up that is likely to be overridden by user space
1512 start up functionality.
1514 Optionally, the snapshot can also be defined for a tracing
1515 instance that was created by the trace_instance= command
1518 trace_instance=foo,sched_switch ftrace_boot_snapshot=foo
1520 The above will cause the "foo" tracing instance to trigger
1521 a snapshot at the end of boot up.
1523 ftrace_dump_on_oops[=orig_cpu]
1524 [FTRACE] will dump the trace buffers on oops.
1525 If no parameter is passed, ftrace will dump
1526 buffers of all CPUs, but if you pass orig_cpu, it will
1527 dump only the buffer of the CPU that triggered the
1530 ftrace_filter=[function-list]
1531 [FTRACE] Limit the functions traced by the function
1532 tracer at boot up. function-list is a comma-separated
1533 list of functions. This list can be changed at run
1534 time by the set_ftrace_filter file in the debugfs
1537 ftrace_notrace=[function-list]
1538 [FTRACE] Do not trace the functions specified in
1539 function-list. This list can be changed at run time
1540 by the set_ftrace_notrace file in the debugfs
1543 ftrace_graph_filter=[function-list]
1544 [FTRACE] Limit the top level callers functions traced
1545 by the function graph tracer at boot up.
1546 function-list is a comma-separated list of functions
1547 that can be changed at run time by the
1548 set_graph_function file in the debugfs tracing directory.
1550 ftrace_graph_notrace=[function-list]
1551 [FTRACE] Do not trace from the functions specified in
1552 function-list. This list is a comma-separated list of
1553 functions that can be changed at run time by the
1554 set_graph_notrace file in the debugfs tracing directory.
1556 ftrace_graph_max_depth=<uint>
1557 [FTRACE] Used with the function graph tracer. This is
1558 the max depth it will trace into a function. This value
1559 can be changed at run time by the max_graph_depth file
1560 in the tracefs tracing directory. default: 0 (no limit)
1562 fw_devlink= [KNL] Create device links between consumer and supplier
1563 devices by scanning the firmware to infer the
1564 consumer/supplier relationships. This feature is
1565 especially useful when drivers are loaded as modules as
1566 it ensures proper ordering of tasks like device probing
1567 (suppliers first, then consumers), supplier boot state
1568 clean up (only after all consumers have probed),
1569 suspend/resume & runtime PM (consumers first, then
1571 Format: { off | permissive | on | rpm }
1572 off -- Don't create device links from firmware info.
1573 permissive -- Create device links from firmware info
1574 but use it only for ordering boot state clean
1575 up (sync_state() calls).
1576 on -- Create device links from firmware info and use it
1577 to enforce probe and suspend/resume ordering.
1578 rpm -- Like "on", but also use to order runtime PM.
1580 fw_devlink.strict=<bool>
1581 [KNL] Treat all inferred dependencies as mandatory
1582 dependencies. This only applies for fw_devlink=on|rpm.
1586 [HW,JOY] Multisystem joystick and NES/SNES/PSX pad
1587 support via parallel port (up to 5 devices per port)
1588 Format: <port#>,<pad1>,<pad2>,<pad3>,<pad4>,<pad5>
1589 See also Documentation/input/devices/joystick-parport.rst
1593 gart_fix_e820= [X86-64] disable the fix e820 for K8 GART
1597 gcov_persist= [GCOV] When non-zero (default), profiling data for
1598 kernel modules is saved and remains accessible via
1599 debugfs, even when the module is unloaded/reloaded.
1600 When zero, profiling data is discarded and associated
1601 debugfs files are removed at module unload time.
1603 goldfish [X86] Enable the goldfish android emulator platform.
1604 Don't use this when you are not running on the
1607 gpio-mockup.gpio_mockup_ranges
1608 [HW] Sets the ranges of gpiochip of for this device.
1609 Format: <start1>,<end1>,<start2>,<end2>...
1610 gpio-mockup.gpio_mockup_named_lines
1611 [HW] Let the driver know GPIO lines should be named.
1613 gpt [EFI] Forces disk with valid GPT signature but
1614 invalid Protective MBR to be treated as GPT. If the
1615 primary GPT is corrupted, it enables the backup/alternate
1616 GPT to be used instead.
1618 grcan.enable0= [HW] Configuration of physical interface 0. Determines
1619 the "Enable 0" bit of the configuration register.
1622 grcan.enable1= [HW] Configuration of physical interface 1. Determines
1623 the "Enable 0" bit of the configuration register.
1626 grcan.select= [HW] Select which physical interface to use.
1629 grcan.txsize= [HW] Sets the size of the tx buffer.
1630 Format: <unsigned int> such that (txsize & ~0x1fffc0) == 0.
1632 grcan.rxsize= [HW] Sets the size of the rx buffer.
1633 Format: <unsigned int> such that (rxsize & ~0x1fffc0) == 0.
1637 [KNL] Under CONFIG_HARDENED_USERCOPY, whether
1638 hardening is enabled for this boot. Hardened
1639 usercopy checking is used to protect the kernel
1640 from reading or writing beyond known memory
1641 allocation boundaries as a proactive defense
1642 against bounds-checking flaws in the kernel's
1643 copy_to_user()/copy_from_user() interface.
1644 on Perform hardened usercopy checks (default).
1645 off Disable hardened usercopy checks.
1647 hardlockup_all_cpu_backtrace=
1648 [KNL] Should the hard-lockup detector generate
1649 backtraces on all cpus.
1652 hashdist= [KNL,NUMA] Large hashes allocated during boot
1653 are distributed across NUMA nodes. Defaults on
1654 for 64-bit NUMA, off otherwise.
1655 Format: 0 | 1 (for off | on)
1657 hcl= [IA-64] SGI's Hardware Graph compatibility layer
1659 hd= [EIDE] (E)IDE hard drive subsystem geometry
1660 Format: <cyl>,<head>,<sect>
1663 Disable Hardware Error Source Table (HEST) support;
1664 corresponding firmware-first mode error processing
1665 logic will be disabled.
1667 hibernate= [HIBERNATION]
1668 noresume Don't check if there's a hibernation image
1669 present during boot.
1670 nocompress Don't compress/decompress hibernation images.
1671 no Disable hibernation and resume.
1672 protect_image Turn on image protection during restoration
1673 (that will set all pages holding image data
1674 during restoration read-only).
1676 highmem=nn[KMG] [KNL,BOOT] forces the highmem zone to have an exact
1677 size of <nn>. This works even on boxes that have no
1678 highmem otherwise. This also works to reduce highmem
1679 size on bigger boxes.
1681 highres= [KNL] Enable/disable high resolution timer mode.
1682 Valid parameters: "on", "off"
1687 hostname= [KNL] Set the hostname (aka UTS nodename).
1689 This allows setting the system's hostname during early
1690 startup. This sets the name returned by gethostname.
1691 Using this parameter to set the hostname makes it
1692 possible to ensure the hostname is correctly set before
1693 any userspace processes run, avoiding the possibility
1694 that a process may call gethostname before the hostname
1695 has been explicitly set, resulting in the calling
1696 process getting an incorrect result. The string must
1697 not exceed the maximum allowed hostname length (usually
1698 64 characters) and will be truncated otherwise.
1700 hpet= [X86-32,HPET] option to control HPET usage
1701 Format: { enable (default) | disable | force |
1703 disable: disable HPET and use PIT instead
1704 force: allow force enabled of undocumented chips (ICH4,
1706 verbose: show contents of HPET registers during setup
1708 hpet_mmap= [X86, HPET_MMAP] Allow userspace to mmap HPET
1709 registers. Default set by CONFIG_HPET_MMAP_DEFAULT.
1711 hugepages= [HW] Number of HugeTLB pages to allocate at boot.
1712 If this follows hugepagesz (below), it specifies
1713 the number of pages of hugepagesz to be allocated.
1714 If this is the first HugeTLB parameter on the command
1715 line, it specifies the number of pages to allocate for
1716 the default huge page size. If using node format, the
1717 number of pages to allocate per-node can be specified.
1718 See also Documentation/admin-guide/mm/hugetlbpage.rst.
1719 Format: <integer> or (node format)
1720 <node>:<integer>[,<node>:<integer>]
1723 [HW] The size of the HugeTLB pages. This is used in
1724 conjunction with hugepages (above) to allocate huge
1725 pages of a specific size at boot. The pair
1726 hugepagesz=X hugepages=Y can be specified once for
1727 each supported huge page size. Huge page sizes are
1728 architecture dependent. See also
1729 Documentation/admin-guide/mm/hugetlbpage.rst.
1732 hugetlb_cma= [HW,CMA] The size of a CMA area used for allocation
1733 of gigantic hugepages. Or using node format, the size
1734 of a CMA area per node can be specified.
1735 Format: nn[KMGTPE] or (node format)
1736 <node>:nn[KMGTPE][,<node>:nn[KMGTPE]]
1738 Reserve a CMA area of given size and allocate gigantic
1739 hugepages using the CMA allocator. If enabled, the
1740 boot-time allocation of gigantic hugepages is skipped.
1742 hugetlb_free_vmemmap=
1743 [KNL] Requires CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP
1745 Control if HugeTLB Vmemmap Optimization (HVO) is enabled.
1746 Allows heavy hugetlb users to free up some more
1747 memory (7 * PAGE_SIZE for each 2MB hugetlb page).
1748 Format: { on | off (default) }
1753 Built with CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP_DEFAULT_ON=y,
1756 Note that the vmemmap pages may be allocated from the added
1757 memory block itself when memory_hotplug.memmap_on_memory is
1758 enabled, those vmemmap pages cannot be optimized even if this
1759 feature is enabled. Other vmemmap pages not allocated from
1760 the added memory block itself do not be affected.
1763 [KNL] Should the hung task detector generate panics.
1766 A value of 1 instructs the kernel to panic when a
1767 hung task is detected. The default value is controlled
1768 by the CONFIG_BOOTPARAM_HUNG_TASK_PANIC build-time
1769 option. The value selected by this boot parameter can
1770 be changed later by the kernel.hung_task_panic sysctl.
1772 hvc_iucv= [S390] Number of z/VM IUCV hypervisor console (HVC)
1773 terminal devices. Valid values: 0..8
1774 hvc_iucv_allow= [S390] Comma-separated list of z/VM user IDs.
1775 If specified, z/VM IUCV HVC accepts connections
1776 from listed z/VM user IDs only.
1778 hv_nopvspin [X86,HYPER_V] Disables the paravirt spinlock optimizations
1779 which allow the hypervisor to 'idle' the
1780 guest on lock contention.
1782 i2c_bus= [HW] Override the default board specific I2C bus speed
1783 or register an additional I2C bus that is not
1784 registered from board initialization code.
1788 i8042.debug [HW] Toggle i8042 debug mode
1789 i8042.unmask_kbd_data
1790 [HW] Enable printing of interrupt data from the KBD port
1791 (disabled by default, and as a pre-condition
1792 requires that i8042.debug=1 be enabled)
1793 i8042.direct [HW] Put keyboard port into non-translated mode
1794 i8042.dumbkbd [HW] Pretend that controller can only read data from
1795 keyboard and cannot control its state
1796 (Don't attempt to blink the leds)
1797 i8042.noaux [HW] Don't check for auxiliary (== mouse) port
1798 i8042.nokbd [HW] Don't check/create keyboard port
1799 i8042.noloop [HW] Disable the AUX Loopback command while probing
1801 i8042.nomux [HW] Don't check presence of an active multiplexing
1803 i8042.nopnp [HW] Don't use ACPIPnP / PnPBIOS to discover KBD/AUX
1805 i8042.notimeout [HW] Ignore timeout condition signalled by controller
1806 i8042.reset [HW] Reset the controller during init, cleanup and
1807 suspend-to-ram transitions, only during s2r
1808 transitions, or never reset
1809 Format: { 1 | Y | y | 0 | N | n }
1810 1, Y, y: always reset controller
1811 0, N, n: don't ever reset controller
1812 Default: only on s2r transitions on x86; most other
1813 architectures force reset to be always executed
1814 i8042.unlock [HW] Unlock (ignore) the keylock
1815 i8042.kbdreset [HW] Reset device connected to KBD port
1817 [HW] Allow deferred probing upon i8042 probe errors
1821 i915.invert_brightness=
1822 [DRM] Invert the sense of the variable that is used to
1823 set the brightness of the panel backlight. Normally a
1824 brightness value of 0 indicates backlight switched off,
1825 and the maximum of the brightness value sets the backlight
1826 to maximum brightness. If this parameter is set to 0
1827 (default) and the machine requires it, or this parameter
1828 is set to 1, a brightness value of 0 sets the backlight
1829 to maximum brightness, and the maximum of the brightness
1830 value switches the backlight off.
1831 -1 -- never invert brightness
1832 0 -- machine default
1833 1 -- force brightness inversion
1836 Format: <io>[,<membase>[,<icn_id>[,<icn_id2>]]]
1840 Format: idle=poll, idle=halt, idle=nomwait
1841 Poll forces a polling idle loop that can slightly
1842 improve the performance of waking up a idle CPU, but
1843 will use a lot of power and make the system run hot.
1845 idle=halt: Halt is forced to be used for CPU idle.
1846 In such case C2/C3 won't be used again.
1847 idle=nomwait: Disable mwait for CPU C-states
1851 Allow force disabling of Shared Virtual Memory (SVA)
1852 support for the idxd driver. By default it is set to
1855 idxd.tc_override= [HW]
1857 Allow override of default traffic class configuration
1858 for the device. By default it is set to false (0).
1860 ieee754= [MIPS] Select IEEE Std 754 conformance mode
1861 Format: { strict | legacy | 2008 | relaxed }
1864 Choose which programs will be accepted for execution
1865 based on the IEEE 754 NaN encoding(s) supported by
1866 the FPU and the NaN encoding requested with the value
1867 of an ELF file header flag individually set by each
1868 binary. Hardware implementations are permitted to
1869 support either or both of the legacy and the 2008 NaN
1872 Available settings are as follows:
1873 strict accept binaries that request a NaN encoding
1874 supported by the FPU
1875 legacy only accept legacy-NaN binaries, if supported
1877 2008 only accept 2008-NaN binaries, if supported
1879 relaxed accept any binaries regardless of whether
1880 supported by the FPU
1882 The FPU emulator is always able to support both NaN
1883 encodings, so if no FPU hardware is present or it has
1884 been disabled with 'nofpu', then the settings of
1885 'legacy' and '2008' strap the emulator accordingly,
1886 'relaxed' straps the emulator for both legacy-NaN and
1887 2008-NaN, whereas 'strict' enables legacy-NaN only on
1888 legacy processors and both NaN encodings on MIPS32 or
1891 The setting for ABS.fmt/NEG.fmt instruction execution
1892 mode generally follows that for the NaN encoding,
1893 except where unsupported by hardware.
1895 ignore_loglevel [KNL]
1896 Ignore loglevel setting - this will print /all/
1897 kernel messages to the console. Useful for debugging.
1898 We also add it as printk module parameter, so users
1899 could change it dynamically, usually by
1900 /sys/module/printk/parameters/ignore_loglevel.
1903 Ignore RLIMIT_DATA setting for data mappings,
1904 print warning at first misuse. Can be changed via
1905 /sys/module/kernel/parameters/ignore_rlimit_data.
1907 ihash_entries= [KNL]
1908 Set number of hash buckets for inode cache.
1910 ima_appraise= [IMA] appraise integrity measurements
1911 Format: { "off" | "enforce" | "fix" | "log" }
1914 ima_appraise_tcb [IMA] Deprecated. Use ima_policy= instead.
1915 The builtin appraise policy appraises all files
1918 ima_canonical_fmt [IMA]
1919 Use the canonical format for the binary runtime
1920 measurements, instead of host native format.
1923 Format: { md5 | sha1 | rmd160 | sha256 | sha384
1927 The list of supported hash algorithms is defined
1928 in crypto/hash_info.h.
1931 The builtin policies to load during IMA setup.
1932 Format: "tcb | appraise_tcb | secure_boot |
1933 fail_securely | critical_data"
1935 The "tcb" policy measures all programs exec'd, files
1936 mmap'd for exec, and all files opened with the read
1937 mode bit set by either the effective uid (euid=0) or
1940 The "appraise_tcb" policy appraises the integrity of
1941 all files owned by root.
1943 The "secure_boot" policy appraises the integrity
1944 of files (eg. kexec kernel image, kernel modules,
1945 firmware, policy, etc) based on file signatures.
1947 The "fail_securely" policy forces file signature
1948 verification failure also on privileged mounted
1949 filesystems with the SB_I_UNVERIFIABLE_SIGNATURE
1952 The "critical_data" policy measures kernel integrity
1955 ima_tcb [IMA] Deprecated. Use ima_policy= instead.
1956 Load a policy which meets the needs of the Trusted
1957 Computing Base. This means IMA will measure all
1958 programs exec'd, files mmap'd for exec, and all files
1959 opened for read by uid=0.
1962 Select one of defined IMA measurements template formats.
1963 Formats: { "ima" | "ima-ng" | "ima-ngv2" | "ima-sig" |
1968 [IMA] Define a custom template format.
1969 Format: { "field1|...|fieldN" }
1971 ima.ahash_minsize= [IMA] Minimum file size for asynchronous hash usage
1972 Format: <min_file_size>
1973 Set the minimal file size for using asynchronous hash.
1974 If left unspecified, ahash usage is disabled.
1976 ahash performance varies for different data sizes on
1977 different crypto accelerators. This option can be used
1978 to achieve the best performance for a particular HW.
1980 ima.ahash_bufsize= [IMA] Asynchronous hash buffer size
1982 Set hashing buffer size. Default: 4k.
1984 ahash performance varies for different chunk sizes on
1985 different crypto accelerators. This option can be used
1986 to achieve best performance for particular HW.
1990 Run specified binary instead of /sbin/init as init
1993 initcall_debug [KNL] Trace initcalls as they are executed. Useful
1994 for working out where the kernel is dying during
1997 initcall_blacklist= [KNL] Do not execute a comma-separated list of
1998 initcall functions. Useful for debugging built-in
1999 modules and initcalls.
2001 initramfs_async= [KNL]
2004 This parameter controls whether the initramfs
2005 image is unpacked asynchronously, concurrently
2006 with devices being probed and
2007 initialized. This should normally just work,
2008 but as a debugging aid, one can get the
2009 historical behaviour of the initramfs
2010 unpacking being completed before device_ and
2013 initrd= [BOOT] Specify the location of the initial ramdisk
2015 initrdmem= [KNL] Specify a physical address and size from which to
2016 load the initrd. If an initrd is compiled in or
2017 specified in the bootparams, it takes priority over this
2019 Format: ss[KMG],nn[KMG]
2022 init_on_alloc= [MM] Fill newly allocated pages and heap objects with
2025 Default set by CONFIG_INIT_ON_ALLOC_DEFAULT_ON.
2027 init_on_free= [MM] Fill freed pages and heap objects with zeroes.
2029 Default set by CONFIG_INIT_ON_FREE_DEFAULT_ON.
2031 init_pkru= [X86] Specify the default memory protection keys rights
2032 register contents for all processes. 0x55555554 by
2033 default (disallow access to all but pkey 0). Can
2034 override in debugfs after boot.
2036 inport.irq= [HW] Inport (ATI XL and Microsoft) busmouse driver
2039 int_pln_enable [X86] Enable power limit notification interrupt
2041 integrity_audit=[IMA]
2042 Format: { "0" | "1" }
2043 0 -- basic integrity auditing messages. (Default)
2044 1 -- additional integrity auditing messages.
2046 intel_iommu= [DMAR] Intel IOMMU driver (DMAR) option
2048 Enable intel iommu driver.
2050 Disable intel iommu driver.
2051 igfx_off [Default Off]
2052 By default, gfx is mapped as normal device. If a gfx
2053 device has a dedicated DMAR unit, the DMAR unit is
2054 bypassed by not enabling DMAR with this option. In
2055 this case, gfx device will use physical address for
2057 strict [Default Off]
2058 Deprecated, equivalent to iommu.strict=1.
2059 sp_off [Default Off]
2060 By default, super page will be supported if Intel IOMMU
2061 has the capability. With this option, super page will
2064 Enable the Intel IOMMU scalable mode if the hardware
2065 advertises that it has support for the scalable mode
2068 Disallow use of the Intel IOMMU scalable mode.
2069 tboot_noforce [Default Off]
2070 Do not force the Intel IOMMU enabled under tboot.
2071 By default, tboot will force Intel IOMMU on, which
2072 could harm performance of some high-throughput
2073 devices like 40GBit network cards, even if identity
2075 Note that using this option lowers the security
2076 provided by tboot because it makes the system
2077 vulnerable to DMA attacks.
2079 intel_idle.max_cstate= [KNL,HW,ACPI,X86]
2080 0 disables intel_idle and fall back on acpi_idle.
2081 1 to 9 specify maximum depth of C-state.
2085 Do not enable intel_pstate as the default
2086 scaling driver for the supported processors
2088 Use intel_pstate as a scaling driver, but configure it
2089 to work with generic cpufreq governors (instead of
2090 enabling its internal governor). This mode cannot be
2091 used along with the hardware-managed P-states (HWP)
2094 Enable intel_pstate on systems that prohibit it by default
2095 in favor of acpi-cpufreq. Forcing the intel_pstate driver
2096 instead of acpi-cpufreq may disable platform features, such
2097 as thermal controls and power capping, that rely on ACPI
2098 P-States information being indicated to OSPM and therefore
2099 should be used with caution. This option does not work with
2100 processors that aren't supported by the intel_pstate driver
2101 or on platforms that use pcc-cpufreq instead of acpi-cpufreq.
2103 Do not enable hardware P state control (HWP)
2106 Only load intel_pstate on systems which support
2107 hardware P state control (HWP) if available.
2109 Enforce ACPI _PPC performance limits. If the Fixed ACPI
2110 Description Table, specifies preferred power management
2111 profile as "Enterprise Server" or "Performance Server",
2112 then this feature is turned on by default.
2114 Allow per-logical-CPU P-State performance control limits using
2115 cpufreq sysfs interface
2117 intremap= [X86-64, Intel-IOMMU]
2118 on enable Interrupt Remapping (default)
2119 off disable Interrupt Remapping
2120 nosid disable Source ID checking
2122 BIOS x2APIC opt-out request will be ignored
2123 nopost disable Interrupt Posting
2125 iomem= Disable strict checking of access to MMIO memory
2126 strict regions from userspace.
2141 nobypass [PPC/POWERNV]
2142 Disable IOMMU bypass, using IOMMU for PCI devices.
2144 iommu.forcedac= [ARM64, X86] Control IOVA allocation for PCI devices.
2145 Format: { "0" | "1" }
2146 0 - Try to allocate a 32-bit DMA address first, before
2147 falling back to the full range if needed.
2148 1 - Allocate directly from the full usable range,
2149 forcing Dual Address Cycle for PCI cards supporting
2150 greater than 32-bit addressing.
2152 iommu.strict= [ARM64, X86] Configure TLB invalidation behaviour
2153 Format: { "0" | "1" }
2155 Request that DMA unmap operations use deferred
2156 invalidation of hardware TLBs, for increased
2157 throughput at the cost of reduced device isolation.
2158 Will fall back to strict mode if not supported by
2159 the relevant IOMMU driver.
2161 DMA unmap operations invalidate IOMMU hardware TLBs
2163 unset - Use value of CONFIG_IOMMU_DEFAULT_DMA_{LAZY,STRICT}.
2164 Note: on x86, strict mode specified via one of the
2165 legacy driver-specific options takes precedence.
2168 [ARM64, X86] Configure DMA to bypass the IOMMU by default.
2169 Format: { "0" | "1" }
2170 0 - Use IOMMU translation for DMA.
2171 1 - Bypass the IOMMU for DMA.
2172 unset - Use value of CONFIG_IOMMU_DEFAULT_PASSTHROUGH.
2174 io7= [HW] IO7 for Marvel-based Alpha systems
2175 See comment before marvel_specify_io7 in
2176 arch/alpha/kernel/core_marvel.c.
2178 io_delay= [X86] I/O delay method
2180 Standard port 0x80 based delay
2182 Alternate port 0xed based delay (needed on some systems)
2184 Simple two microseconds delay
2189 See Documentation/admin-guide/nfs/nfsroot.rst.
2191 ipcmni_extend [KNL] Extend the maximum number of unique System V
2192 IPC identifiers from 32,768 to 16,777,216.
2194 irqaffinity= [SMP] Set the default irq affinity mask
2195 The argument is a cpu list, as described above.
2197 irqchip.gicv2_force_probe=
2200 Force the kernel to look for the second 4kB page
2201 of a GICv2 controller even if the memory range
2202 exposed by the device tree is too small.
2204 irqchip.gicv3_nolpi=
2206 Force the kernel to ignore the availability of
2207 LPIs (and by consequence ITSs). Intended for system
2208 that use the kernel as a bootloader, and thus want
2209 to let secondary kernels in charge of setting up
2212 irqchip.gicv3_pseudo_nmi= [ARM64]
2213 Enables support for pseudo-NMIs in the kernel. This
2214 requires the kernel to be built with
2215 CONFIG_ARM64_PSEUDO_NMI.
2218 When an interrupt is not handled search all handlers
2219 for it. Intended to get systems with badly broken
2223 When an interrupt is not handled search all handlers
2224 for it. Also check all handlers each timer
2225 interrupt. Intended to get systems with badly broken
2229 Format: <RDP>,<reset>,<pci_scan>,<verbosity>
2231 isolcpus= [KNL,SMP,ISOL] Isolate a given set of CPUs from disturbance.
2232 [Deprecated - use cpusets instead]
2233 Format: [flag-list,]<cpu-list>
2235 Specify one or more CPUs to isolate from disturbances
2236 specified in the flag list (default: domain):
2239 Disable the tick when a single task runs.
2241 A residual 1Hz tick is offloaded to workqueues, which you
2242 need to affine to housekeeping through the global
2243 workqueue's affinity configured via the
2244 /sys/devices/virtual/workqueue/cpumask sysfs file, or
2245 by using the 'domain' flag described below.
2247 NOTE: by default the global workqueue runs on all CPUs,
2248 so to protect individual CPUs the 'cpumask' file has to
2249 be configured manually after bootup.
2252 Isolate from the general SMP balancing and scheduling
2253 algorithms. Note that performing domain isolation this way
2254 is irreversible: it's not possible to bring back a CPU to
2255 the domains once isolated through isolcpus. It's strongly
2256 advised to use cpusets instead to disable scheduler load
2257 balancing through the "cpuset.sched_load_balance" file.
2258 It offers a much more flexible interface where CPUs can
2259 move in and out of an isolated set anytime.
2261 You can move a process onto or off an "isolated" CPU via
2262 the CPU affinity syscalls or cpuset.
2263 <cpu number> begins at 0 and the maximum value is
2264 "number of CPUs in system - 1".
2268 Isolate from being targeted by managed interrupts
2269 which have an interrupt mask containing isolated
2270 CPUs. The affinity of managed interrupts is
2271 handled by the kernel and cannot be changed via
2272 the /proc/irq/* interfaces.
2274 This isolation is best effort and only effective
2275 if the automatically assigned interrupt mask of a
2276 device queue contains isolated and housekeeping
2277 CPUs. If housekeeping CPUs are online then such
2278 interrupts are directed to the housekeeping CPU
2279 so that IO submitted on the housekeeping CPU
2280 cannot disturb the isolated CPU.
2282 If a queue's affinity mask contains only isolated
2283 CPUs then this parameter has no effect on the
2284 interrupt routing decision, though interrupts are
2285 only delivered when tasks running on those
2286 isolated CPUs submit IO. IO submitted on
2287 housekeeping CPUs has no influence on those
2290 The format of <cpu-list> is described above.
2294 ivrs_ioapic [HW,X86-64]
2295 Provide an override to the IOAPIC-ID<->DEVICE-ID
2296 mapping provided in the IVRS ACPI table.
2297 By default, PCI segment is 0, and can be omitted.
2299 For example, to map IOAPIC-ID decimal 10 to
2300 PCI segment 0x1 and PCI device 00:14.0,
2301 write the parameter as:
2302 ivrs_ioapic=10@0001:00:14.0
2305 * To map IOAPIC-ID decimal 10 to PCI device 00:14.0
2306 write the parameter as:
2307 ivrs_ioapic[10]=00:14.0
2308 * To map IOAPIC-ID decimal 10 to PCI segment 0x1 and
2309 PCI device 00:14.0 write the parameter as:
2310 ivrs_ioapic[10]=0001:00:14.0
2312 ivrs_hpet [HW,X86-64]
2313 Provide an override to the HPET-ID<->DEVICE-ID
2314 mapping provided in the IVRS ACPI table.
2315 By default, PCI segment is 0, and can be omitted.
2317 For example, to map HPET-ID decimal 10 to
2318 PCI segment 0x1 and PCI device 00:14.0,
2319 write the parameter as:
2320 ivrs_hpet=10@0001:00:14.0
2323 * To map HPET-ID decimal 0 to PCI device 00:14.0
2324 write the parameter as:
2325 ivrs_hpet[0]=00:14.0
2326 * To map HPET-ID decimal 10 to PCI segment 0x1 and
2327 PCI device 00:14.0 write the parameter as:
2328 ivrs_ioapic[10]=0001:00:14.0
2330 ivrs_acpihid [HW,X86-64]
2331 Provide an override to the ACPI-HID:UID<->DEVICE-ID
2332 mapping provided in the IVRS ACPI table.
2333 By default, PCI segment is 0, and can be omitted.
2335 For example, to map UART-HID:UID AMD0020:0 to
2336 PCI segment 0x1 and PCI device ID 00:14.5,
2337 write the parameter as:
2338 ivrs_acpihid=AMD0020:0@0001:00:14.5
2341 * To map UART-HID:UID AMD0020:0 to PCI segment is 0,
2342 PCI device ID 00:14.5, write the parameter as:
2343 ivrs_acpihid[00:14.5]=AMD0020:0
2344 * To map UART-HID:UID AMD0020:0 to PCI segment 0x1 and
2345 PCI device ID 00:14.5, write the parameter as:
2346 ivrs_acpihid[0001:00:14.5]=AMD0020:0
2348 js= [HW,JOY] Analog joystick
2349 See Documentation/input/joydev/joystick.rst.
2352 [KNL] Enforce KASAN (Kernel Address Sanitizer) to print
2353 report on every invalid memory access. Without this
2354 parameter KASAN will print report only for the first
2358 Do not unregister boot console at start. This is only
2359 useful for debugging when something happens in the window
2360 between unregistering the boot console and initializing
2365 kernelcore= [KNL,X86,IA-64,PPC]
2366 Format: nn[KMGTPE] | nn% | "mirror"
2367 This parameter specifies the amount of memory usable by
2368 the kernel for non-movable allocations. The requested
2369 amount is spread evenly throughout all nodes in the
2370 system as ZONE_NORMAL. The remaining memory is used for
2371 movable memory in its own zone, ZONE_MOVABLE. In the
2372 event, a node is too small to have both ZONE_NORMAL and
2373 ZONE_MOVABLE, kernelcore memory will take priority and
2374 other nodes will have a larger ZONE_MOVABLE.
2376 ZONE_MOVABLE is used for the allocation of pages that
2377 may be reclaimed or moved by the page migration
2378 subsystem. Note that allocations like PTEs-from-HighMem
2379 still use the HighMem zone if it exists, and the Normal
2380 zone if it does not.
2382 It is possible to specify the exact amount of memory in
2383 the form of "nn[KMGTPE]", a percentage of total system
2384 memory in the form of "nn%", or "mirror". If "mirror"
2385 option is specified, mirrored (reliable) memory is used
2386 for non-movable allocations and remaining memory is used
2387 for Movable pages. "nn[KMGTPE]", "nn%", and "mirror"
2388 are exclusive, so you cannot specify multiple forms.
2390 kgdbdbgp= [KGDB,HW] kgdb over EHCI usb debug port.
2391 Format: <Controller#>[,poll interval]
2392 The controller # is the number of the ehci usb debug
2393 port as it is probed via PCI. The poll interval is
2394 optional and is the number seconds in between
2395 each poll cycle to the debug port in case you need
2396 the functionality for interrupting the kernel with
2397 gdb or control-c on the dbgp connection. When
2398 not using this parameter you use sysrq-g to break into
2399 the kernel debugger.
2401 kgdboc= [KGDB,HW] kgdb over consoles.
2402 Requires a tty driver that supports console polling,
2403 or a supported polling keyboard driver (non-usb).
2404 Serial only format: <serial_device>[,baud]
2405 keyboard only format: kbd
2406 keyboard and serial format: kbd,<serial_device>[,baud]
2407 Optional Kernel mode setting:
2408 kms, kbd format: kms,kbd
2409 kms, kbd and serial format: kms,kbd,<ser_dev>[,baud]
2411 kgdboc_earlycon= [KGDB,HW]
2412 If the boot console provides the ability to read
2413 characters and can work in polling mode, you can use
2414 this parameter to tell kgdb to use it as a backend
2415 until the normal console is registered. Intended to
2416 be used together with the kgdboc parameter which
2417 specifies the normal console to transition to.
2419 The name of the early console should be specified
2420 as the value of this parameter. Note that the name of
2421 the early console might be different than the tty
2422 name passed to kgdboc. It's OK to leave the value
2423 blank and the first boot console that implements
2424 read() will be picked.
2426 kgdbwait [KGDB] Stop kernel execution and enter the
2427 kernel debugger at the earliest opportunity.
2429 kmac= [MIPS] Korina ethernet MAC address.
2430 Configure the RouterBoard 532 series on-chip
2431 Ethernet adapter MAC address.
2433 kmemleak= [KNL] Boot-time kmemleak enable/disable
2434 Valid arguments: on, off
2436 Built with CONFIG_DEBUG_KMEMLEAK_DEFAULT_OFF=y,
2439 kprobe_event=[probe-list]
2440 [FTRACE] Add kprobe events and enable at boot time.
2441 The probe-list is a semicolon delimited list of probe
2442 definitions. Each definition is same as kprobe_events
2443 interface, but the parameters are comma delimited.
2444 For example, to add a kprobe event on vfs_read with
2445 arg1 and arg2, add to the command line;
2447 kprobe_event=p,vfs_read,$arg1,$arg2
2449 See also Documentation/trace/kprobetrace.rst "Kernel
2450 Boot Parameter" section.
2452 kpti= [ARM64] Control page table isolation of user
2453 and kernel address spaces.
2454 Default: enabled on cores which need mitigation.
2458 kunit.enable= [KUNIT] Enable executing KUnit tests. Requires
2459 CONFIG_KUNIT to be set to be fully enabled. The
2460 default value can be overridden via
2461 KUNIT_DEFAULT_ENABLED.
2462 Default is 1 (enabled)
2464 kvm.ignore_msrs=[KVM] Ignore guest accesses to unhandled MSRs.
2465 Default is 0 (don't ignore, but inject #GP)
2467 kvm.eager_page_split=
2468 [KVM,X86] Controls whether or not KVM will try to
2469 proactively split all huge pages during dirty logging.
2470 Eager page splitting reduces interruptions to vCPU
2471 execution by eliminating the write-protection faults
2472 and MMU lock contention that would otherwise be
2473 required to split huge pages lazily.
2475 VM workloads that rarely perform writes or that write
2476 only to a small region of VM memory may benefit from
2477 disabling eager page splitting to allow huge pages to
2478 still be used for reads.
2480 The behavior of eager page splitting depends on whether
2481 KVM_DIRTY_LOG_INITIALLY_SET is enabled or disabled. If
2482 disabled, all huge pages in a memslot will be eagerly
2483 split when dirty logging is enabled on that memslot. If
2484 enabled, eager page splitting will be performed during
2485 the KVM_CLEAR_DIRTY ioctl, and only for the pages being
2488 Eager page splitting is only supported when kvm.tdp_mmu=Y.
2492 kvm.enable_vmware_backdoor=[KVM] Support VMware backdoor PV interface.
2493 Default is false (don't support).
2496 [KVM] Controls the software workaround for the
2497 X86_BUG_ITLB_MULTIHIT bug.
2498 force : Always deploy workaround.
2499 off : Never deploy workaround.
2500 auto : Deploy workaround based on the presence of
2501 X86_BUG_ITLB_MULTIHIT.
2505 If the software workaround is enabled for the host,
2506 guests do need not to enable it for nested guests.
2508 kvm.nx_huge_pages_recovery_ratio=
2509 [KVM] Controls how many 4KiB pages are periodically zapped
2510 back to huge pages. 0 disables the recovery, otherwise if
2511 the value is N KVM will zap 1/Nth of the 4KiB pages every
2512 period (see below). The default is 60.
2514 kvm.nx_huge_pages_recovery_period_ms=
2515 [KVM] Controls the time period at which KVM zaps 4KiB pages
2516 back to huge pages. If the value is a non-zero N, KVM will
2517 zap a portion (see ratio above) of the pages every N msecs.
2518 If the value is 0 (the default), KVM will pick a period based
2519 on the ratio, such that a page is zapped after 1 hour on average.
2521 kvm-amd.nested= [KVM,AMD] Allow nested virtualization in KVM/SVM.
2522 Default is 1 (enabled)
2524 kvm-amd.npt= [KVM,AMD] Disable nested paging (virtualized MMU)
2526 Default is 1 (enabled) if in 64-bit or 32-bit PAE mode.
2529 [KVM,ARM] Select one of KVM/arm64's modes of operation.
2531 none: Forcefully disable KVM.
2533 nvhe: Standard nVHE-based mode, without support for
2536 protected: nVHE-based mode with support for guests whose
2537 state is kept private from the host.
2539 nested: VHE-based mode with support for nested
2540 virtualization. Requires at least ARMv8.3
2543 Defaults to VHE/nVHE based on hardware support. Setting
2544 mode to "protected" will disable kexec and hibernation
2545 for the host. "nested" is experimental and should be
2546 used with extreme caution.
2548 kvm-arm.vgic_v3_group0_trap=
2549 [KVM,ARM] Trap guest accesses to GICv3 group-0
2552 kvm-arm.vgic_v3_group1_trap=
2553 [KVM,ARM] Trap guest accesses to GICv3 group-1
2556 kvm-arm.vgic_v3_common_trap=
2557 [KVM,ARM] Trap guest accesses to GICv3 common
2560 kvm-arm.vgic_v4_enable=
2561 [KVM,ARM] Allow use of GICv4 for direct injection of
2564 kvm_cma_resv_ratio=n [PPC]
2565 Reserves given percentage from system memory area for
2566 contiguous memory allocation for KVM hash pagetable
2568 By default it reserves 5% of total system memory.
2572 kvm-intel.ept= [KVM,Intel] Disable extended page tables
2573 (virtualized MMU) support on capable Intel chips.
2574 Default is 1 (enabled)
2576 kvm-intel.emulate_invalid_guest_state=
2577 [KVM,Intel] Disable emulation of invalid guest state.
2578 Ignored if kvm-intel.enable_unrestricted_guest=1, as
2579 guest state is never invalid for unrestricted guests.
2580 This param doesn't apply to nested guests (L2), as KVM
2581 never emulates invalid L2 guest state.
2582 Default is 1 (enabled)
2584 kvm-intel.flexpriority=
2585 [KVM,Intel] Disable FlexPriority feature (TPR shadow).
2586 Default is 1 (enabled)
2589 [KVM,Intel] Enable VMX nesting (nVMX).
2590 Default is 0 (disabled)
2592 kvm-intel.unrestricted_guest=
2593 [KVM,Intel] Disable unrestricted guest feature
2594 (virtualized real and unpaged mode) on capable
2595 Intel chips. Default is 1 (enabled)
2597 kvm-intel.vmentry_l1d_flush=[KVM,Intel] Mitigation for L1 Terminal Fault
2600 Valid arguments: never, cond, always
2602 always: L1D cache flush on every VMENTER.
2603 cond: Flush L1D on VMENTER only when the code between
2604 VMEXIT and VMENTER can leak host memory.
2605 never: Disables the mitigation
2607 Default is cond (do L1 cache flush in specific instances)
2609 kvm-intel.vpid= [KVM,Intel] Disable Virtual Processor Identification
2610 feature (tagged TLBs) on capable Intel chips.
2611 Default is 1 (enabled)
2613 l1d_flush= [X86,INTEL]
2614 Control mitigation for L1D based snooping vulnerability.
2616 Certain CPUs are vulnerable to an exploit against CPU
2617 internal buffers which can forward information to a
2618 disclosure gadget under certain conditions.
2620 In vulnerable processors, the speculatively
2621 forwarded data can be used in a cache side channel
2622 attack, to access data to which the attacker does
2623 not have direct access.
2625 This parameter controls the mitigation. The
2628 on - enable the interface for the mitigation
2630 l1tf= [X86] Control mitigation of the L1TF vulnerability on
2633 The kernel PTE inversion protection is unconditionally
2634 enabled and cannot be disabled.
2637 Provides all available mitigations for the
2638 L1TF vulnerability. Disables SMT and
2639 enables all mitigations in the
2640 hypervisors, i.e. unconditional L1D flush.
2642 SMT control and L1D flush control via the
2643 sysfs interface is still possible after
2644 boot. Hypervisors will issue a warning
2645 when the first VM is started in a
2646 potentially insecure configuration,
2647 i.e. SMT enabled or L1D flush disabled.
2650 Same as 'full', but disables SMT and L1D
2651 flush runtime control. Implies the
2652 'nosmt=force' command line option.
2653 (i.e. sysfs control of SMT is disabled.)
2656 Leaves SMT enabled and enables the default
2657 hypervisor mitigation, i.e. conditional
2660 SMT control and L1D flush control via the
2661 sysfs interface is still possible after
2662 boot. Hypervisors will issue a warning
2663 when the first VM is started in a
2664 potentially insecure configuration,
2665 i.e. SMT enabled or L1D flush disabled.
2669 Disables SMT and enables the default
2670 hypervisor mitigation.
2672 SMT control and L1D flush control via the
2673 sysfs interface is still possible after
2674 boot. Hypervisors will issue a warning
2675 when the first VM is started in a
2676 potentially insecure configuration,
2677 i.e. SMT enabled or L1D flush disabled.
2680 Same as 'flush', but hypervisors will not
2681 warn when a VM is started in a potentially
2682 insecure configuration.
2685 Disables hypervisor mitigations and doesn't
2687 It also drops the swap size and available
2688 RAM limit restriction on both hypervisor and
2693 For details see: Documentation/admin-guide/hw-vuln/l1tf.rst
2699 lapic [X86-32,APIC] Enable the local APIC even if BIOS
2702 lapic= [X86,APIC] Do not use TSC deadline
2703 value for LAPIC timer one-shot implementation. Default
2704 back to the programmable timer unit in the LAPIC.
2705 Format: notscdeadline
2707 lapic_timer_c2_ok [X86,APIC] trust the local apic timer
2710 libata.dma= [LIBATA] DMA control
2711 libata.dma=0 Disable all PATA and SATA DMA
2712 libata.dma=1 PATA and SATA Disk DMA only
2713 libata.dma=2 ATAPI (CDROM) DMA only
2714 libata.dma=4 Compact Flash DMA only
2715 Combinations also work, so libata.dma=3 enables DMA
2716 for disks and CDROMs, but not CFs.
2718 libata.ignore_hpa= [LIBATA] Ignore HPA limit
2719 libata.ignore_hpa=0 keep BIOS limits (default)
2720 libata.ignore_hpa=1 ignore limits, using full disk
2722 libata.noacpi [LIBATA] Disables use of ACPI in libata suspend/resume
2726 libata.force= [LIBATA] Force configurations. The format is a comma-
2727 separated list of "[ID:]VAL" where ID is PORT[.DEVICE].
2728 PORT and DEVICE are decimal numbers matching port, link
2729 or device. Basically, it matches the ATA ID string
2730 printed on console by libata. If the whole ID part is
2731 omitted, the last PORT and DEVICE values are used. If
2732 ID hasn't been specified yet, the configuration applies
2733 to all ports, links and devices.
2735 If only DEVICE is omitted, the parameter applies to
2736 the port and all links and devices behind it. DEVICE
2737 number of 0 either selects the first device or the
2738 first fan-out link behind PMP device. It does not
2739 select the host link. DEVICE number of 15 selects the
2740 host link and device attached to it.
2742 The VAL specifies the configuration to force. As long
2743 as there is no ambiguity, shortcut notation is allowed.
2744 For example, both 1.5 and 1.5G would work for 1.5Gbps.
2745 The following configurations can be forced.
2747 * Cable type: 40c, 80c, short40c, unk, ign or sata.
2748 Any ID with matching PORT is used.
2750 * SATA link speed limit: 1.5Gbps or 3.0Gbps.
2752 * Transfer mode: pio[0-7], mwdma[0-4] and udma[0-7].
2753 udma[/][16,25,33,44,66,100,133] notation is also
2756 * nohrst, nosrst, norst: suppress hard, soft and both
2759 * rstonce: only attempt one reset during hot-unplug
2762 * [no]dbdelay: Enable or disable the extra 200ms delay
2763 before debouncing a link PHY and device presence
2766 * [no]ncq: Turn on or off NCQ.
2768 * [no]ncqtrim: Enable or disable queued DSM TRIM.
2770 * [no]ncqati: Enable or disable NCQ trim on ATI chipset.
2772 * [no]trim: Enable or disable (unqueued) TRIM.
2774 * trim_zero: Indicate that TRIM command zeroes data.
2776 * max_trim_128m: Set 128M maximum trim size limit.
2778 * [no]dma: Turn on or off DMA transfers.
2780 * atapi_dmadir: Enable ATAPI DMADIR bridge support.
2782 * atapi_mod16_dma: Enable the use of ATAPI DMA for
2783 commands that are not a multiple of 16 bytes.
2785 * [no]dmalog: Enable or disable the use of the
2786 READ LOG DMA EXT command to access logs.
2788 * [no]iddevlog: Enable or disable access to the
2789 identify device data log.
2791 * [no]logdir: Enable or disable access to the general
2792 purpose log directory.
2794 * max_sec_128: Set transfer size limit to 128 sectors.
2796 * max_sec_1024: Set or clear transfer size limit to
2799 * max_sec_lba48: Set or clear transfer size limit to
2802 * [no]lpm: Enable or disable link power management.
2804 * [no]setxfer: Indicate if transfer speed mode setting
2807 * [no]fua: Disable or enable FUA (Force Unit Access)
2808 support for devices supporting this feature.
2810 * dump_id: Dump IDENTIFY data.
2812 * disable: Disable this device.
2814 If there are multiple matching configurations changing
2815 the same attribute, the last one is used.
2817 load_ramdisk= [RAM] [Deprecated]
2819 lockd.nlm_grace_period=P [NFS] Assign grace period.
2822 lockd.nlm_tcpport=N [NFS] Assign TCP port.
2825 lockd.nlm_timeout=T [NFS] Assign timeout value.
2828 lockd.nlm_udpport=M [NFS] Assign UDP port.
2831 lockdown= [SECURITY]
2832 { integrity | confidentiality }
2833 Enable the kernel lockdown feature. If set to
2834 integrity, kernel features that allow userland to
2835 modify the running kernel are disabled. If set to
2836 confidentiality, kernel features that allow userland
2837 to extract confidential information from the kernel
2840 locktorture.nreaders_stress= [KNL]
2841 Set the number of locking read-acquisition kthreads.
2842 Defaults to being automatically set based on the
2843 number of online CPUs.
2845 locktorture.nwriters_stress= [KNL]
2846 Set the number of locking write-acquisition kthreads.
2848 locktorture.onoff_holdoff= [KNL]
2849 Set time (s) after boot for CPU-hotplug testing.
2851 locktorture.onoff_interval= [KNL]
2852 Set time (s) between CPU-hotplug operations, or
2853 zero to disable CPU-hotplug testing.
2855 locktorture.shuffle_interval= [KNL]
2856 Set task-shuffle interval (jiffies). Shuffling
2857 tasks allows some CPUs to go into dyntick-idle
2858 mode during the locktorture test.
2860 locktorture.shutdown_secs= [KNL]
2861 Set time (s) after boot system shutdown. This
2862 is useful for hands-off automated testing.
2864 locktorture.stat_interval= [KNL]
2865 Time (s) between statistics printk()s.
2867 locktorture.stutter= [KNL]
2868 Time (s) to stutter testing, for example,
2869 specifying five seconds causes the test to run for
2870 five seconds, wait for five seconds, and so on.
2871 This tests the locking primitive's ability to
2872 transition abruptly to and from idle.
2874 locktorture.torture_type= [KNL]
2875 Specify the locking implementation to test.
2877 locktorture.verbose= [KNL]
2878 Enable additional printk() statements.
2880 logibm.irq= [HW,MOUSE] Logitech Bus Mouse Driver
2883 loglevel= All Kernel Messages with a loglevel smaller than the
2884 console loglevel will be printed to the console. It can
2885 also be changed with klogd or other programs. The
2886 loglevels are defined as follows:
2888 0 (KERN_EMERG) system is unusable
2889 1 (KERN_ALERT) action must be taken immediately
2890 2 (KERN_CRIT) critical conditions
2891 3 (KERN_ERR) error conditions
2892 4 (KERN_WARNING) warning conditions
2893 5 (KERN_NOTICE) normal but significant condition
2894 6 (KERN_INFO) informational
2895 7 (KERN_DEBUG) debug-level messages
2897 log_buf_len=n[KMG] Sets the size of the printk ring buffer,
2898 in bytes. n must be a power of two and greater
2899 than the minimal size. The minimal size is defined
2900 by LOG_BUF_SHIFT kernel config parameter. There is
2901 also CONFIG_LOG_CPU_MAX_BUF_SHIFT config parameter
2902 that allows to increase the default size depending on
2903 the number of CPUs. See init/Kconfig for more details.
2905 logo.nologo [FB] Disables display of the built-in Linux logo.
2906 This may be used to provide more screen space for
2907 kernel log messages and is useful when debugging
2908 kernel boot problems.
2910 lp=0 [LP] Specify parallel ports to use, e.g,
2911 lp=port[,port...] lp=none,parport0 (lp0 not configured, lp1 uses
2912 lp=reset first parallel port). 'lp=0' disables the
2913 lp=auto printer driver. 'lp=reset' (which can be
2914 specified in addition to the ports) causes
2915 attached printers to be reset. Using
2916 lp=port1,port2,... specifies the parallel ports
2917 to associate lp devices with, starting with
2918 lp0. A port specification may be 'none' to skip
2919 that lp device, or a parport name such as
2920 'parport0'. Specifying 'lp=auto' instead of a
2921 port specification list means that device IDs
2922 from each port should be examined, to see if
2923 an IEEE 1284-compliant printer is attached; if
2924 so, the driver will manage that printer.
2925 See also header of drivers/char/lp.c.
2928 Sets loops_per_jiffy to given constant, thus avoiding
2929 time-consuming boot-time autodetection (up to 250 ms per
2930 CPU). 0 enables autodetection (default). To determine
2931 the correct value for your kernel, boot with normal
2932 autodetection and see what value is printed. Note that
2933 on SMP systems the preset will be applied to all CPUs,
2934 which is likely to cause problems if your CPUs need
2935 significantly divergent settings. An incorrect value
2936 will cause delays in the kernel to be wrong, leading to
2937 unpredictable I/O errors and other breakage. Although
2938 unlikely, in the extreme case this might damage your
2942 Format: <io>,<irq>,<dma>
2944 lsm.debug [SECURITY] Enable LSM initialization debugging output.
2947 [SECURITY] Choose order of LSM initialization. This
2948 overrides CONFIG_LSM, and the "security=" parameter.
2950 machvec= [IA-64] Force the use of a particular machine-vector
2951 (machvec) in a generic kernel.
2952 Example: machvec=hpzx1
2954 machtype= [Loongson] Share the same kernel image file between
2955 different yeeloong laptops.
2956 Example: machtype=lemote-yeeloong-2f-7inch
2958 max_addr=nn[KMG] [KNL,BOOT,IA-64] All physical memory greater
2959 than or equal to this physical address is ignored.
2961 maxcpus= [SMP] Maximum number of processors that an SMP kernel
2962 will bring up during bootup. maxcpus=n : n >= 0 limits
2963 the kernel to bring up 'n' processors. Surely after
2964 bootup you can bring up the other plugged cpu by executing
2965 "echo 1 > /sys/devices/system/cpu/cpuX/online". So maxcpus
2966 only takes effect during system bootup.
2967 While n=0 is a special case, it is equivalent to "nosmp",
2968 which also disables the IO APIC.
2970 max_loop= [LOOP] The number of loop block devices that get
2971 (loop.max_loop) unconditionally pre-created at init time. The default
2972 number is configured by BLK_DEV_LOOP_MIN_COUNT. Instead
2973 of statically allocating a predefined number, loop
2974 devices can be requested on-demand with the
2975 /dev/loop-control interface.
2977 mce [X86-32] Machine Check Exception
2979 mce=option [X86-64] See Documentation/x86/x86_64/boot-options.rst
2981 md= [HW] RAID subsystems devices and level
2982 See Documentation/admin-guide/md.rst.
2985 Format: <first>,<last>
2986 Specifies range of consoles to be captured by the MDA.
2989 Control mitigation for the Micro-architectural Data
2990 Sampling (MDS) vulnerability.
2992 Certain CPUs are vulnerable to an exploit against CPU
2993 internal buffers which can forward information to a
2994 disclosure gadget under certain conditions.
2996 In vulnerable processors, the speculatively
2997 forwarded data can be used in a cache side channel
2998 attack, to access data to which the attacker does
2999 not have direct access.
3001 This parameter controls the MDS mitigation. The
3004 full - Enable MDS mitigation on vulnerable CPUs
3005 full,nosmt - Enable MDS mitigation and disable
3006 SMT on vulnerable CPUs
3007 off - Unconditionally disable MDS mitigation
3009 On TAA-affected machines, mds=off can be prevented by
3010 an active TAA mitigation as both vulnerabilities are
3011 mitigated with the same mechanism so in order to disable
3012 this mitigation, you need to specify tsx_async_abort=off
3015 Not specifying this option is equivalent to
3018 For details see: Documentation/admin-guide/hw-vuln/mds.rst
3020 mem=nn[KMG] [HEXAGON] Set the memory size.
3021 Must be specified, otherwise memory size will be 0.
3023 mem=nn[KMG] [KNL,BOOT] Force usage of a specific amount of memory
3024 Amount of memory to be used in cases as follows:
3027 2 when the kernel is not able to see the whole system memory;
3028 3 memory that lies after 'mem=' boundary is excluded from
3029 the hypervisor, then assigned to KVM guests.
3030 4 to limit the memory available for kdump kernel.
3032 [ARC,MICROBLAZE] - the limit applies only to low memory,
3033 high memory is not affected.
3035 [ARM64] - only limits memory covered by the linear
3036 mapping. The NOMAP regions are not affected.
3038 [X86] Work as limiting max address. Use together
3039 with memmap= to avoid physical address space collisions.
3040 Without memmap= PCI devices could be placed at addresses
3041 belonging to unused RAM.
3043 Note that this only takes effects during boot time since
3044 in above case 3, memory may need be hot added after boot
3045 if system memory of hypervisor is not sufficient.
3048 [ARM,MIPS] - override the memory layout reported by
3050 Define a memory region of size nn[KMG] starting at
3052 Multiple different regions can be specified with
3053 multiple mem= parameters on the command line.
3055 mem=nopentium [BUGS=X86-32] Disable usage of 4MB pages for kernel
3058 memblock=debug [KNL] Enable memblock debug messages.
3061 [KNL,SH] Allow user to override the default size for
3062 per-device physically contiguous DMA buffers.
3064 memhp_default_state=online/offline
3065 [KNL] Set the initial state for the memory hotplug
3066 onlining policy. If not specified, the default value is
3067 set according to the
3068 CONFIG_MEMORY_HOTPLUG_DEFAULT_ONLINE kernel config
3070 See Documentation/admin-guide/mm/memory-hotplug.rst.
3072 memmap=exactmap [KNL,X86] Enable setting of an exact
3073 E820 memory map, as specified by the user.
3074 Such memmap=exactmap lines can be constructed based on
3075 BIOS output or other requirements. See the memmap=nn@ss
3078 memmap=nn[KMG]@ss[KMG]
3079 [KNL, X86, MIPS, XTENSA] Force usage of a specific region of memory.
3080 Region of memory to be used is from ss to ss+nn.
3081 If @ss[KMG] is omitted, it is equivalent to mem=nn[KMG],
3082 which limits max address to nn[KMG].
3083 Multiple different regions can be specified,
3086 memmap=100M@2G,100M#3G,1G!1024G
3088 memmap=nn[KMG]#ss[KMG]
3089 [KNL,ACPI] Mark specific memory as ACPI data.
3090 Region of memory to be marked is from ss to ss+nn.
3092 memmap=nn[KMG]$ss[KMG]
3093 [KNL,ACPI] Mark specific memory as reserved.
3094 Region of memory to be reserved is from ss to ss+nn.
3095 Example: Exclude memory from 0x18690000-0x1869ffff
3096 memmap=64K$0x18690000
3098 memmap=0x10000$0x18690000
3099 Some bootloaders may need an escape character before '$',
3100 like Grub2, otherwise '$' and the following number
3103 memmap=nn[KMG]!ss[KMG]
3104 [KNL,X86] Mark specific memory as protected.
3105 Region of memory to be used, from ss to ss+nn.
3106 The memory region may be marked as e820 type 12 (0xc)
3107 and is NVDIMM or ADR memory.
3109 memmap=<size>%<offset>-<oldtype>+<newtype>
3110 [KNL,ACPI] Convert memory within the specified region
3111 from <oldtype> to <newtype>. If "-<oldtype>" is left
3112 out, the whole region will be marked as <newtype>,
3113 even if previously unavailable. If "+<newtype>" is left
3114 out, matching memory will be removed. Types are
3115 specified as e820 types, e.g., 1 = RAM, 2 = reserved,
3116 3 = ACPI, 12 = PRAM.
3118 memory_corruption_check=0/1 [X86]
3119 Some BIOSes seem to corrupt the first 64k of
3120 memory when doing things like suspend/resume.
3121 Setting this option will scan the memory
3122 looking for corruption. Enabling this will
3123 both detect corruption and prevent the kernel
3124 from using the memory being corrupted.
3125 However, its intended as a diagnostic tool; if
3126 repeatable BIOS-originated corruption always
3127 affects the same memory, you can use memmap=
3128 to prevent the kernel from using that memory.
3130 memory_corruption_check_size=size [X86]
3131 By default it checks for corruption in the low
3132 64k, making this memory unavailable for normal
3133 use. Use this parameter to scan for
3134 corruption in more or less memory.
3136 memory_corruption_check_period=seconds [X86]
3137 By default it checks for corruption every 60
3138 seconds. Use this parameter to check at some
3139 other rate. 0 disables periodic checking.
3141 memory_hotplug.memmap_on_memory
3142 [KNL,X86,ARM] Boolean flag to enable this feature.
3143 Format: {on | off (default)}
3144 When enabled, runtime hotplugged memory will
3145 allocate its internal metadata (struct pages,
3146 those vmemmap pages cannot be optimized even
3147 if hugetlb_free_vmemmap is enabled) from the
3148 hotadded memory which will allow to hotadd a
3149 lot of memory without requiring additional
3151 This feature is disabled by default because it
3152 has some implication on large (e.g. GB)
3153 allocations in some configurations (e.g. small
3155 The state of the flag can be read in
3156 /sys/module/memory_hotplug/parameters/memmap_on_memory.
3157 Note that even when enabled, there are a few cases where
3158 the feature is not effective.
3160 memtest= [KNL,X86,ARM,M68K,PPC,RISCV] Enable memtest
3162 default : 0 <disable>
3163 Specifies the number of memtest passes to be
3164 performed. Each pass selects another test
3165 pattern from a given set of patterns. Memtest
3166 fills the memory with this pattern, validates
3167 memory contents and reserves bad memory
3168 regions that are detected.
3170 mem_encrypt= [X86-64] AMD Secure Memory Encryption (SME) control
3171 Valid arguments: on, off
3172 Default (depends on kernel configuration option):
3173 on (CONFIG_AMD_MEM_ENCRYPT_ACTIVE_BY_DEFAULT=y)
3174 off (CONFIG_AMD_MEM_ENCRYPT_ACTIVE_BY_DEFAULT=n)
3175 mem_encrypt=on: Activate SME
3176 mem_encrypt=off: Do not activate SME
3178 Refer to Documentation/virt/kvm/x86/amd-memory-encryption.rst
3179 for details on when memory encryption can be activated.
3181 mem_sleep_default= [SUSPEND] Default system suspend mode:
3182 s2idle - Suspend-To-Idle
3183 shallow - Power-On Suspend or equivalent (if supported)
3184 deep - Suspend-To-RAM or equivalent (if supported)
3185 See Documentation/admin-guide/pm/sleep-states.rst.
3187 meye.*= [HW] Set MotionEye Camera parameters
3188 See Documentation/admin-guide/media/meye.rst.
3190 mfgpt_irq= [IA-32] Specify the IRQ to use for the
3191 Multi-Function General Purpose Timers on AMD Geode
3194 mfgptfix [X86-32] Fix MFGPT timers on AMD Geode platforms when
3195 the BIOS has incorrectly applied a workaround. TinyBIOS
3196 version 0.98 is known to be affected, 0.99 fixes the
3197 problem by letting the user disable the workaround.
3201 min_addr=nn[KMG] [KNL,BOOT,IA-64] All physical memory below this
3202 physical address is ignored.
3204 mini2440= [ARM,HW,KNL]
3205 Format:[0..2][b][c][t]
3207 MINI2440 configuration specification:
3208 0 - The attached screen is the 3.5" TFT
3209 1 - The attached screen is the 7" TFT
3210 2 - The VGA Shield is attached (1024x768)
3211 Leaving out the screen size parameter will not load
3212 the TFT driver, and the framebuffer will be left
3214 b - Enable backlight. The TFT backlight pin will be
3215 linked to the kernel VESA blanking code and a GPIO
3216 LED. This parameter is not necessary when using the
3218 c - Enable the s3c camera interface.
3219 t - Reserved for enabling touchscreen support. The
3220 touchscreen support is not enabled in the mainstream
3221 kernel as of 2.6.30, a preliminary port can be found
3222 in the "bleeding edge" mini2440 support kernel at
3223 https://repo.or.cz/w/linux-2.6/mini2440.git
3226 [X86,PPC,S390,ARM64] Control optional mitigations for
3227 CPU vulnerabilities. This is a set of curated,
3228 arch-independent options, each of which is an
3229 aggregation of existing arch-specific options.
3232 Disable all optional CPU mitigations. This
3233 improves system performance, but it may also
3234 expose users to several CPU vulnerabilities.
3235 Equivalent to: nopti [X86,PPC]
3236 if nokaslr then kpti=0 [ARM64]
3237 nospectre_v1 [X86,PPC]
3239 nospectre_v2 [X86,PPC,S390,ARM64]
3240 spectre_v2_user=off [X86]
3241 spec_store_bypass_disable=off [X86,PPC]
3242 ssbd=force-off [ARM64]
3243 nospectre_bhb [ARM64]
3246 tsx_async_abort=off [X86]
3247 kvm.nx_huge_pages=off [X86]
3248 srbds=off [X86,INTEL]
3249 no_entry_flush [PPC]
3250 no_uaccess_flush [PPC]
3251 mmio_stale_data=off [X86]
3255 This does not have any effect on
3256 kvm.nx_huge_pages when
3257 kvm.nx_huge_pages=force.
3260 Mitigate all CPU vulnerabilities, but leave SMT
3261 enabled, even if it's vulnerable. This is for
3262 users who don't want to be surprised by SMT
3263 getting disabled across kernel upgrades, or who
3264 have other ways of avoiding SMT-based attacks.
3265 Equivalent to: (default behavior)
3268 Mitigate all CPU vulnerabilities, disabling SMT
3269 if needed. This is for users who always want to
3270 be fully mitigated, even if it means losing SMT.
3271 Equivalent to: l1tf=flush,nosmt [X86]
3272 mds=full,nosmt [X86]
3273 tsx_async_abort=full,nosmt [X86]
3274 mmio_stale_data=full,nosmt [X86]
3275 retbleed=auto,nosmt [X86]
3278 [KNL] When CONFIG_DEBUG_MEMORY_INIT is set, this
3279 parameter allows control of the logging verbosity for
3280 the additional memory initialisation checks. A value
3281 of 0 disables mminit logging and a level of 4 will
3282 log everything. Information is printed at KERN_DEBUG
3283 so loglevel=8 may also need to be specified.
3286 [X86,INTEL] Control mitigation for the Processor
3287 MMIO Stale Data vulnerabilities.
3289 Processor MMIO Stale Data is a class of
3290 vulnerabilities that may expose data after an MMIO
3291 operation. Exposed data could originate or end in
3292 the same CPU buffers as affected by MDS and TAA.
3293 Therefore, similar to MDS and TAA, the mitigation
3294 is to clear the affected CPU buffers.
3296 This parameter controls the mitigation. The
3299 full - Enable mitigation on vulnerable CPUs
3301 full,nosmt - Enable mitigation and disable SMT on
3304 off - Unconditionally disable mitigation
3306 On MDS or TAA affected machines,
3307 mmio_stale_data=off can be prevented by an active
3308 MDS or TAA mitigation as these vulnerabilities are
3309 mitigated with the same mechanism so in order to
3310 disable this mitigation, you need to specify
3311 mds=off and tsx_async_abort=off too.
3313 Not specifying this option is equivalent to
3314 mmio_stale_data=full.
3317 Documentation/admin-guide/hw-vuln/processor_mmio_stale_data.rst
3319 <module>.async_probe[=<bool>] [KNL]
3320 If no <bool> value is specified or if the value
3321 specified is not a valid <bool>, enable asynchronous
3322 probe on this module. Otherwise, enable/disable
3323 asynchronous probe on this module as indicated by the
3324 <bool> value. See also: module.async_probe
3326 module.async_probe=<bool>
3327 [KNL] When set to true, modules will use async probing
3328 by default. To enable/disable async probing for a
3329 specific module, use the module specific control that
3330 is documented under <module>.async_probe. When both
3331 module.async_probe and <module>.async_probe are
3332 specified, <module>.async_probe takes precedence for
3333 the specific module.
3336 [KNL] When CONFIG_MODULE_SIG is set, this means that
3337 modules without (valid) signatures will fail to load.
3338 Note that if CONFIG_MODULE_SIG_FORCE is set, that
3339 is always true, so this option does nothing.
3341 module_blacklist= [KNL] Do not load a comma-separated list of
3342 modules. Useful for debugging problem modules.
3345 [MOUSE] Maximum time between finger touching and
3346 leaving touchpad surface for touch to be considered
3347 a tap and be reported as a left button click (for
3348 touchpads working in absolute mode only).
3350 mousedev.xres= [MOUSE] Horizontal screen resolution, used for devices
3351 reporting absolute coordinates, such as tablets
3352 mousedev.yres= [MOUSE] Vertical screen resolution, used for devices
3353 reporting absolute coordinates, such as tablets
3355 movablecore= [KNL,X86,IA-64,PPC]
3356 Format: nn[KMGTPE] | nn%
3357 This parameter is the complement to kernelcore=, it
3358 specifies the amount of memory used for migratable
3359 allocations. If both kernelcore and movablecore is
3360 specified, then kernelcore will be at *least* the
3361 specified value but may be more. If movablecore on its
3362 own is specified, the administrator must be careful
3363 that the amount of memory usable for all allocations
3366 movable_node [KNL] Boot-time switch to make hotplugable memory
3367 NUMA nodes to be movable. This means that the memory
3368 of such nodes will be usable only for movable
3369 allocations which rules out almost all kernel
3370 allocations. Use with caution!
3372 MTD_Partition= [MTD]
3373 Format: <name>,<region-number>,<size>,<offset>
3375 MTD_Region= [MTD] Format:
3376 <name>,<region-number>[,<base>,<size>,<buswidth>,<altbuswidth>]
3379 See drivers/mtd/parsers/cmdlinepart.c
3382 ARM/S3C2412 JIVE boot control
3384 See arch/arm/mach-s3c/mach-jive.c
3386 mtouchusb.raw_coordinates=
3387 [HW] Make the MicroTouch USB driver use raw coordinates
3388 ('y', default) or cooked coordinates ('n')
3390 mtrr_chunk_size=nn[KMG] [X86]
3391 used for mtrr cleanup. It is largest continuous chunk
3392 that could hold holes aka. UC entries.
3394 mtrr_gran_size=nn[KMG] [X86]
3395 Used for mtrr cleanup. It is granularity of mtrr block.
3397 Large value could prevent small alignment from
3400 mtrr_spare_reg_nr=n [X86]
3402 Range: 0,7 : spare reg number
3404 Used for mtrr cleanup. It is spare mtrr entries number.
3405 Set to 2 or more if your graphical card needs more.
3407 multitce=off [PPC] This parameter disables the use of the pSeries
3408 firmware feature for updating multiple TCE entries
3411 n2= [NET] SDL Inc. RISCom/N2 synchronous serial card
3413 netdev= [NET] Network devices parameters
3414 Format: <irq>,<io>,<mem_start>,<mem_end>,<name>
3415 Note that mem_start is often overloaded to mean
3416 something different and driver-specific.
3417 This usage is only documented in each driver source
3420 netpoll.carrier_timeout=
3421 [NET] Specifies amount of time (in seconds) that
3422 netpoll should wait for a carrier. By default netpoll
3426 [NETFILTER] Enable connection tracking flow accounting
3427 0 to disable accounting
3428 1 to enable accounting
3431 nfsaddrs= [NFS] Deprecated. Use ip= instead.
3432 See Documentation/admin-guide/nfs/nfsroot.rst.
3434 nfsroot= [NFS] nfs root filesystem for disk-less boxes.
3435 See Documentation/admin-guide/nfs/nfsroot.rst.
3437 nfsrootdebug [NFS] enable nfsroot debugging messages.
3438 See Documentation/admin-guide/nfs/nfsroot.rst.
3440 nfs.callback_nr_threads=
3441 [NFSv4] set the total number of threads that the
3442 NFS client will assign to service NFSv4 callback
3445 nfs.callback_tcpport=
3446 [NFS] set the TCP port on which the NFSv4 callback
3447 channel should listen.
3450 [NFS] sets the pathname to the program which is used
3451 to update the NFS client cache entries.
3453 nfs.cache_getent_timeout=
3454 [NFS] sets the timeout after which an attempt to
3455 update a cache entry is deemed to have failed.
3457 nfs.idmap_cache_timeout=
3458 [NFS] set the maximum lifetime for idmapper cache
3462 [NFS] enable 64-bit inode numbers.
3463 If zero, the NFS client will fake up a 32-bit inode
3464 number for the readdir() and stat() syscalls instead
3465 of returning the full 64-bit number.
3466 The default is to return 64-bit inode numbers.
3468 nfs.max_session_cb_slots=
3469 [NFSv4.1] Sets the maximum number of session
3470 slots the client will assign to the callback
3471 channel. This determines the maximum number of
3472 callbacks the client will process in parallel for
3473 a particular server.
3475 nfs.max_session_slots=
3476 [NFSv4.1] Sets the maximum number of session slots
3477 the client will attempt to negotiate with the server.
3478 This limits the number of simultaneous RPC requests
3479 that the client can send to the NFSv4.1 server.
3480 Note that there is little point in setting this
3481 value higher than the max_tcp_slot_table_limit.
3483 nfs.nfs4_disable_idmapping=
3484 [NFSv4] When set to the default of '1', this option
3485 ensures that both the RPC level authentication
3486 scheme and the NFS level operations agree to use
3487 numeric uids/gids if the mount is using the
3488 'sec=sys' security flavour. In effect it is
3489 disabling idmapping, which can make migration from
3490 legacy NFSv2/v3 systems to NFSv4 easier.
3491 Servers that do not support this mode of operation
3492 will be autodetected by the client, and it will fall
3493 back to using the idmapper.
3494 To turn off this behaviour, set the value to '0'.
3496 [NFS4] Specify an additional fixed unique ident-
3497 ification string that NFSv4 clients can insert into
3498 their nfs_client_id4 string. This is typically a
3499 UUID that is generated at system install time.
3501 nfs.send_implementation_id =
3502 [NFSv4.1] Send client implementation identification
3503 information in exchange_id requests.
3504 If zero, no implementation identification information
3506 The default is to send the implementation identification
3509 nfs.recover_lost_locks =
3510 [NFSv4] Attempt to recover locks that were lost due
3511 to a lease timeout on the server. Please note that
3512 doing this risks data corruption, since there are
3513 no guarantees that the file will remain unchanged
3514 after the locks are lost.
3515 If you want to enable the kernel legacy behaviour of
3516 attempting to recover these locks, then set this
3518 The default parameter value of '0' causes the kernel
3519 not to attempt recovery of lost locks.
3521 nfs4.layoutstats_timer =
3522 [NFSv4.2] Change the rate at which the kernel sends
3523 layoutstats to the pNFS metadata server.
3525 Setting this to value to 0 causes the kernel to use
3526 whatever value is the default set by the layout
3527 driver. A non-zero value sets the minimum interval
3528 in seconds between layoutstats transmissions.
3530 nfsd.inter_copy_offload_enable =
3531 [NFSv4.2] When set to 1, the server will support
3532 server-to-server copies for which this server is
3533 the destination of the copy.
3535 nfsd.nfsd4_ssc_umount_timeout =
3536 [NFSv4.2] When used as the destination of a
3537 server-to-server copy, knfsd temporarily mounts
3538 the source server. It caches the mount in case
3539 it will be needed again, and discards it if not
3540 used for the number of milliseconds specified by
3543 nfsd.nfs4_disable_idmapping=
3544 [NFSv4] When set to the default of '1', the NFSv4
3545 server will return only numeric uids and gids to
3546 clients using auth_sys, and will accept numeric uids
3547 and gids from such clients. This is intended to ease
3548 migration from NFSv2/v3.
3551 nmi_backtrace.backtrace_idle [KNL]
3552 Dump stacks even of idle CPUs in response to an
3553 NMI stack-backtrace request.
3555 nmi_debug= [KNL,SH] Specify one or more actions to take
3556 when a NMI is triggered.
3557 Format: [state][,regs][,debounce][,die]
3559 nmi_watchdog= [KNL,BUGS=X86] Debugging features for SMP kernels
3560 Format: [panic,][nopanic,][num]
3562 0 - turn hardlockup detector in nmi_watchdog off
3563 1 - turn hardlockup detector in nmi_watchdog on
3564 When panic is specified, panic when an NMI watchdog
3565 timeout occurs (or 'nopanic' to not panic on an NMI
3566 watchdog, if CONFIG_BOOTPARAM_HARDLOCKUP_PANIC is set)
3567 To disable both hard and soft lockup detectors,
3568 please see 'nowatchdog'.
3569 This is useful when you use a panic=... timeout and
3570 need the box quickly up again.
3572 These settings can be accessed at runtime via
3573 the nmi_watchdog and hardlockup_panic sysctls.
3575 no387 [BUGS=X86-32] Tells the kernel to use the 387 maths
3576 emulation library even if a 387 maths coprocessor
3579 no5lvl [X86-64] Disable 5-level paging mode. Forces
3580 kernel to use 4-level paging instead.
3582 nofsgsbase [X86] Disables FSGSBASE instructions.
3585 [HW] Never suspend the console
3586 Disable suspending of consoles during suspend and
3587 hibernate operations. Once disabled, debugging
3588 messages can reach various consoles while the rest
3589 of the system is being put to sleep (ie, while
3590 debugging driver suspend/resume hooks). This may
3591 not work reliably with all consoles, but is known
3592 to work with serial and VGA consoles.
3593 To facilitate more flexible debugging, we also add
3594 console_suspend, a printk module parameter to control
3595 it. Users could use console_suspend (usually
3596 /sys/module/printk/parameters/console_suspend) to
3597 turn on/off it dynamically.
3599 novmcoredd [KNL,KDUMP]
3600 Disable device dump. Device dump allows drivers to
3601 append dump data to vmcore so you can collect driver
3602 specified debug info. Drivers can append the data
3603 without any limit and this data is stored in memory,
3604 so this may cause significant memory stress. Disabling
3605 device dump can help save memory but the driver debug
3606 data will be no longer available. This parameter
3607 is only available when CONFIG_PROC_VMCORE_DEVICE_DUMP
3610 noaliencache [MM, NUMA, SLAB] Disables the allocation of alien
3611 caches in the slab allocator. Saves per-node memory,
3612 but will impact performance.
3616 noaltinstr [S390] Disables alternative instructions patching
3617 (CPU alternatives feature).
3619 noapic [SMP,APIC] Tells the kernel to not make use of any
3620 IOAPICs that may be present in the system.
3622 noautogroup Disable scheduler automatic task group creation.
3626 nodsp [SH] Disable hardware DSP at boot time.
3628 noefi Disable EFI runtime services support.
3630 no_entry_flush [PPC] Don't flush the L1-D cache when entering the kernel.
3635 Disable SMAP (Supervisor Mode Access Prevention)
3636 even if it is supported by processor.
3639 Disable SMEP (Supervisor Mode Execution Prevention)
3640 even if it is supported by processor.
3643 This affects only 32-bit executables.
3644 noexec32=on: enable non-executable mappings (default)
3645 read doesn't imply executable mappings
3646 noexec32=off: disable non-executable mappings
3647 read implies executable mappings
3649 nofpu [MIPS,SH] Disable hardware FPU at boot time.
3651 nofxsr [BUGS=X86-32] Disables x86 floating point extended
3652 register save and restore. The kernel will only save
3653 legacy floating-point registers on task switch.
3655 nohugeiomap [KNL,X86,PPC,ARM64] Disable kernel huge I/O mappings.
3657 nohugevmalloc [KNL,X86,PPC,ARM64] Disable kernel huge vmalloc mappings.
3659 nosmt [KNL,S390] Disable symmetric multithreading (SMT).
3660 Equivalent to smt=1.
3662 [KNL,X86] Disable symmetric multithreading (SMT).
3663 nosmt=force: Force disable SMT, cannot be undone
3664 via the sysfs control file.
3666 nospectre_v1 [X86,PPC] Disable mitigations for Spectre Variant 1
3667 (bounds check bypass). With this option data leaks are
3668 possible in the system.
3670 nospectre_v2 [X86,PPC_E500,ARM64] Disable all mitigations for
3671 the Spectre variant 2 (indirect branch prediction)
3672 vulnerability. System may allow data leaks with this
3675 nospectre_bhb [ARM64] Disable all mitigations for Spectre-BHB (branch
3676 history injection) vulnerability. System may allow data leaks
3679 nospec_store_bypass_disable
3680 [HW] Disable all mitigations for the Speculative Store Bypass vulnerability
3683 [PPC] Don't flush the L1-D cache after accessing user data.
3685 noxsave [BUGS=X86] Disables x86 extended register state save
3686 and restore using xsave. The kernel will fallback to
3687 enabling legacy floating-point and sse state.
3689 noxsaveopt [X86] Disables xsaveopt used in saving x86 extended
3690 register states. The kernel will fall back to use
3691 xsave to save the states. By using this parameter,
3692 performance of saving the states is degraded because
3693 xsave doesn't support modified optimization while
3694 xsaveopt supports it on xsaveopt enabled systems.
3696 noxsaves [X86] Disables xsaves and xrstors used in saving and
3697 restoring x86 extended register state in compacted
3698 form of xsave area. The kernel will fall back to use
3699 xsaveopt and xrstor to save and restore the states
3700 in standard form of xsave area. By using this
3701 parameter, xsave area per process might occupy more
3702 memory on xsaves enabled systems.
3704 nohlt [ARM,ARM64,MICROBLAZE,SH] Forces the kernel to busy wait
3705 in do_idle() and not use the arch_cpu_idle()
3706 implementation; requires CONFIG_GENERIC_IDLE_POLL_SETUP
3707 to be effective. This is useful on platforms where the
3708 sleep(SH) or wfi(ARM,ARM64) instructions do not work
3709 correctly or when doing power measurements to evaluate
3710 the impact of the sleep instructions. This is also
3711 useful when using JTAG debugger.
3713 no_file_caps Tells the kernel not to honor file capabilities. The
3714 only way then for a file to be executed with privilege
3715 is to be setuid root or executed by root.
3717 nohalt [IA-64] Tells the kernel not to use the power saving
3718 function PAL_HALT_LIGHT when idle. This increases
3719 power-consumption. On the positive side, it reduces
3720 interrupt wake-up latency, which may improve performance
3721 in certain environments such as networked servers or
3725 Force pointers printed to the console or buffers to be
3726 unhashed. By default, when a pointer is printed via %p
3727 format string, that pointer is "hashed", i.e. obscured
3728 by hashing the pointer value. This is a security feature
3729 that hides actual kernel addresses from unprivileged
3730 users, but it also makes debugging the kernel more
3731 difficult since unequal pointers can no longer be
3732 compared. However, if this command-line option is
3733 specified, then all normal pointers will have their true
3734 value printed. This option should only be specified when
3735 debugging the kernel. Please do not use on production
3738 nohibernate [HIBERNATION] Disable hibernation and resume.
3740 nohz= [KNL] Boottime enable/disable dynamic ticks
3741 Valid arguments: on, off
3744 nohz_full= [KNL,BOOT,SMP,ISOL]
3745 The argument is a cpu list, as described above.
3746 In kernels built with CONFIG_NO_HZ_FULL=y, set
3747 the specified list of CPUs whose tick will be stopped
3748 whenever possible. The boot CPU will be forced outside
3749 the range to maintain the timekeeping. Any CPUs
3750 in this list will have their RCU callbacks offloaded,
3751 just as if they had also been called out in the
3752 rcu_nocbs= boot parameter.
3754 Note that this argument takes precedence over
3755 the CONFIG_RCU_NOCB_CPU_DEFAULT_ALL option.
3757 noiotrap [SH] Disables trapped I/O port accesses.
3759 noirqdebug [X86-32] Disables the code which attempts to detect and
3760 disable unhandled interrupt sources.
3762 no_timer_check [X86,APIC] Disables the code which tests for
3763 broken timer IRQ sources.
3765 noisapnp [ISAPNP] Disables ISA PnP code.
3767 noinitrd [RAM] Tells the kernel not to load any configured
3770 nointremap [X86-64, Intel-IOMMU] Do not enable interrupt
3772 [Deprecated - use intremap=off]
3776 noinvpcid [X86] Disable the INVPCID cpu feature.
3778 nojitter [IA-64] Disables jitter checking for ITC timers.
3781 When CONFIG_RANDOMIZE_BASE is set, this disables
3782 kernel and module base offset ASLR (Address Space
3783 Layout Randomization).
3785 no-kvmclock [X86,KVM] Disable paravirtualized KVM clock driver
3787 no-kvmapf [X86,KVM] Disable paravirtualized asynchronous page
3791 [X86,PV_OPS] Disable paravirtualized VMware scheduler
3792 clock and use the default one.
3794 no-steal-acc [X86,PV_OPS,ARM64,PPC/PSERIES] Disable paravirtualized
3795 steal time accounting. steal time is computed, but
3796 won't influence scheduler behaviour
3798 nolapic [X86-32,APIC] Do not enable or use the local APIC.
3800 nolapic_timer [X86-32,APIC] Do not use the local APIC timer.
3802 nomca [IA-64] Disable machine check abort handling
3804 nomce [X86-32] Disable Machine Check Exception
3806 nomfgpt [X86-32] Disable Multi-Function General Purpose
3807 Timer usage (for AMD Geode machines).
3809 nonmi_ipi [X86] Disable using NMI IPIs during panic/reboot to
3810 shutdown the other cpus. Instead use the REBOOT_VECTOR
3813 nomodeset Disable kernel modesetting. Most systems' firmware
3814 sets up a display mode and provides framebuffer memory
3815 for output. With nomodeset, DRM and fbdev drivers will
3816 not load if they could possibly displace the pre-
3817 initialized output. Only the system framebuffer will
3818 be available for use. The respective drivers will not
3819 perform display-mode changes or accelerated rendering.
3821 Useful as error fallback, or for testing and debugging.
3823 nomodule Disable module load
3825 nopat [X86] Disable PAT (page attribute table extension of
3826 pagetables) support.
3828 nopcid [X86-64] Disable the PCID cpu feature.
3830 nopku [X86] Disable Memory Protection Keys CPU feature found
3833 nopv= [X86,XEN,KVM,HYPER_V,VMWARE]
3834 Disables the PV optimizations forcing the guest to run
3835 as generic guest with no PV drivers. Currently support
3836 XEN HVM, KVM, HYPER_V and VMWARE guest.
3838 nopvspin [X86,XEN,KVM]
3839 Disables the qspinlock slow path using PV optimizations
3840 which allow the hypervisor to 'idle' the guest on lock
3843 norandmaps Don't use address space randomization. Equivalent to
3844 echo 0 > /proc/sys/kernel/randomize_va_space
3846 noreplace-smp [X86-32,SMP] Don't replace SMP instructions
3847 with UP alternatives
3849 noresume [SWSUSP] Disables resume and restores original swap
3852 no-scroll [VGA] Disables scrollback.
3853 This is required for the Braillex ib80-piezo Braille
3854 reader made by F.H. Papenmeier (Germany).
3858 nosgx [X86-64,SGX] Disables Intel SGX kernel support.
3860 nosmp [SMP] Tells an SMP kernel to act as a UP kernel,
3861 and disable the IO APIC. legacy for "maxcpus=0".
3863 nosoftlockup [KNL] Disable the soft-lockup detector.
3865 nosync [HW,M68K] Disables sync negotiation for all devices.
3867 nowatchdog [KNL] Disable both lockup detectors, i.e.
3868 soft-lockup and NMI watchdog (hard-lockup).
3872 nox2apic [X86-64,APIC] Do not enable x2APIC mode.
3874 NOTE: this parameter will be ignored on systems with the
3875 LEGACY_XAPIC_DISABLED bit set in the
3876 IA32_XAPIC_DISABLE_STATUS MSR.
3878 nps_mtm_hs_ctr= [KNL,ARC]
3879 This parameter sets the maximum duration, in
3880 cycles, each HW thread of the CTOP can run
3881 without interruptions, before HW switches it.
3882 The actual maximum duration is 16 times this
3884 Format: integer between 1 and 255
3887 nptcg= [IA-64] Override max number of concurrent global TLB
3888 purges which is reported from either PAL_VM_SUMMARY or
3891 nr_cpus= [SMP] Maximum number of processors that an SMP kernel
3892 could support. nr_cpus=n : n >= 1 limits the kernel to
3893 support 'n' processors. It could be larger than the
3894 number of already plugged CPU during bootup, later in
3895 runtime you can physically add extra cpu until it reaches
3896 n. So during boot up some boot time memory for per-cpu
3897 variables need be pre-allocated for later physical cpu
3900 nr_uarts= [SERIAL] maximum number of UARTs to be registered.
3902 numa=off [KNL, ARM64, PPC, RISCV, SPARC, X86] Disable NUMA, Only
3903 set up a single NUMA node spanning all memory.
3905 numa_balancing= [KNL,ARM64,PPC,RISCV,S390,X86] Enable or disable automatic
3907 Allowed values are enable and disable
3909 numa_zonelist_order= [KNL, BOOT] Select zonelist order for NUMA.
3910 'node', 'default' can be specified
3911 This can be set from sysctl after boot.
3912 See Documentation/admin-guide/sysctl/vm.rst for details.
3914 ohci1394_dma=early [HW] enable debugging via the ohci1394 driver.
3915 See Documentation/core-api/debugging-via-ohci1394.rst for more
3918 olpc_ec_timeout= [OLPC] ms delay when issuing EC commands
3919 Rather than timing out after 20 ms if an EC
3920 command is not properly ACKed, override the length
3921 of the timeout. We have interrupts disabled while
3922 waiting for the ACK, so if this is set too high
3923 interrupts *may* be lost!
3925 omap_mux= [OMAP] Override bootloader pin multiplexing.
3926 Format: <mux_mode0.mode_name=value>...
3927 For example, to override I2C bus2:
3928 omap_mux=i2c2_scl.i2c2_scl=0x100,i2c2_sda.i2c2_sda=0x100
3930 onenand.bdry= [HW,MTD] Flex-OneNAND Boundary Configuration
3932 Format: [die0_boundary][,die0_lock][,die1_boundary][,die1_lock]
3934 boundary - index of last SLC block on Flex-OneNAND.
3935 The remaining blocks are configured as MLC blocks.
3936 lock - Configure if Flex-OneNAND boundary should be locked.
3937 Once locked, the boundary cannot be changed.
3938 1 indicates lock status, 0 indicates unlock status.
3940 oops=panic Always panic on oopses. Default is to just kill the
3941 process, but there is a small probability of
3942 deadlocking the machine.
3943 This will also cause panics on machine check exceptions.
3944 Useful together with panic=30 to trigger a reboot.
3947 [KNL] Boolean flag to control whether the page allocator
3948 should randomize its free lists. The randomization may
3949 be automatically enabled if the kernel detects it is
3950 running on a platform with a direct-mapped memory-side
3951 cache, and this parameter can be used to
3952 override/disable that behavior. The state of the flag
3953 can be read from sysfs at:
3954 /sys/module/page_alloc/parameters/shuffle.
3956 page_owner= [KNL] Boot-time page_owner enabling option.
3957 Storage of the information about who allocated
3958 each page is disabled in default. With this switch,
3960 on: enable the feature
3962 page_poison= [KNL] Boot-time parameter changing the state of
3963 poisoning on the buddy allocator, available with
3964 CONFIG_PAGE_POISONING=y.
3965 off: turn off poisoning (default)
3966 on: turn on poisoning
3968 page_reporting.page_reporting_order=
3969 [KNL] Minimal page reporting order
3971 Adjust the minimal page reporting order. The page
3972 reporting is disabled when it exceeds (MAX_ORDER-1).
3974 panic= [KNL] Kernel behaviour on panic: delay <timeout>
3975 timeout > 0: seconds before rebooting
3976 timeout = 0: wait forever
3977 timeout < 0: reboot immediately
3980 panic_print= Bitmask for printing system info when panic happens.
3981 User can chose combination of the following bits:
3982 bit 0: print all tasks info
3983 bit 1: print system memory info
3984 bit 2: print timer info
3985 bit 3: print locks info if CONFIG_LOCKDEP is on
3986 bit 4: print ftrace buffer
3987 bit 5: print all printk messages in buffer
3988 bit 6: print all CPUs backtrace (if available in the arch)
3989 *Be aware* that this option may print a _lot_ of lines,
3990 so there are risks of losing older messages in the log.
3991 Use this option carefully, maybe worth to setup a
3992 bigger log buffer with "log_buf_len" along with this.
3994 panic_on_taint= Bitmask for conditionally calling panic() in add_taint()
3995 Format: <hex>[,nousertaint]
3996 Hexadecimal bitmask representing the set of TAINT flags
3997 that will cause the kernel to panic when add_taint() is
3998 called with any of the flags in this set.
3999 The optional switch "nousertaint" can be utilized to
4000 prevent userspace forced crashes by writing to sysctl
4001 /proc/sys/kernel/tainted any flagset matching with the
4002 bitmask set on panic_on_taint.
4003 See Documentation/admin-guide/tainted-kernels.rst for
4004 extra details on the taint flags that users can pick
4005 to compose the bitmask to assign to panic_on_taint.
4007 panic_on_warn panic() instead of WARN(). Useful to cause kdump
4010 parkbd.port= [HW] Parallel port number the keyboard adapter is
4011 connected to, default is 0.
4013 parkbd.mode= [HW] Parallel port keyboard adapter mode of operation,
4014 0 for XT, 1 for AT (default is AT).
4017 parport= [HW,PPT] Specify parallel ports. 0 disables.
4018 Format: { 0 | auto | 0xBBB[,IRQ[,DMA]] }
4019 Use 'auto' to force the driver to use any
4020 IRQ/DMA settings detected (the default is to
4021 ignore detected IRQ/DMA settings because of
4022 possible conflicts). You can specify the base
4023 address, IRQ, and DMA settings; IRQ and DMA
4024 should be numbers, or 'auto' (for using detected
4025 settings on that particular port), or 'nofifo'
4026 (to avoid using a FIFO even if it is detected).
4027 Parallel ports are assigned in the order they
4028 are specified on the command line, starting
4031 parport_init_mode= [HW,PPT]
4032 Configure VIA parallel port to operate in
4033 a specific mode. This is necessary on Pegasos
4034 computer where firmware has no options for setting
4035 up parallel port mode and sets it to spp.
4036 Currently this function knows 686a and 8231 chips.
4037 Format: [spp|ps2|epp|ecp|ecpepp]
4039 pata_legacy.all= [HW,LIBATA]
4041 Set to non-zero to probe primary and secondary ISA
4042 port ranges on PCI systems where no PCI PATA device
4043 has been found at either range. Disabled by default.
4045 pata_legacy.autospeed= [HW,LIBATA]
4047 Set to non-zero if a chip is present that snoops speed
4048 changes. Disabled by default.
4050 pata_legacy.ht6560a= [HW,LIBATA]
4052 Set to 1, 2, or 3 for HT 6560A on the primary channel,
4053 the secondary channel, or both channels respectively.
4054 Disabled by default.
4056 pata_legacy.ht6560b= [HW,LIBATA]
4058 Set to 1, 2, or 3 for HT 6560B on the primary channel,
4059 the secondary channel, or both channels respectively.
4060 Disabled by default.
4062 pata_legacy.iordy_mask= [HW,LIBATA]
4064 IORDY enable mask. Set individual bits to allow IORDY
4065 for the respective channel. Bit 0 is for the first
4066 legacy channel handled by this driver, bit 1 is for
4067 the second channel, and so on. The sequence will often
4068 correspond to the primary legacy channel, the secondary
4069 legacy channel, and so on, but the handling of a PCI
4070 bus and the use of other driver options may interfere
4071 with the sequence. By default IORDY is allowed across
4074 pata_legacy.opti82c46x= [HW,LIBATA]
4076 Set to 1, 2, or 3 for Opti 82c611A on the primary
4077 channel, the secondary channel, or both channels
4078 respectively. Disabled by default.
4080 pata_legacy.opti82c611a= [HW,LIBATA]
4082 Set to 1, 2, or 3 for Opti 82c465MV on the primary
4083 channel, the secondary channel, or both channels
4084 respectively. Disabled by default.
4086 pata_legacy.pio_mask= [HW,LIBATA]
4088 PIO mode mask for autospeed devices. Set individual
4089 bits to allow the use of the respective PIO modes.
4090 Bit 0 is for mode 0, bit 1 is for mode 1, and so on.
4091 All modes allowed by default.
4093 pata_legacy.probe_all= [HW,LIBATA]
4095 Set to non-zero to probe tertiary and further ISA
4096 port ranges on PCI systems. Disabled by default.
4098 pata_legacy.probe_mask= [HW,LIBATA]
4100 Probe mask for legacy ISA PATA ports. Depending on
4101 platform configuration and the use of other driver
4102 options up to 6 legacy ports are supported: 0x1f0,
4103 0x170, 0x1e8, 0x168, 0x1e0, 0x160, however probing
4104 of individual ports can be disabled by setting the
4105 corresponding bits in the mask to 1. Bit 0 is for
4106 the first port in the list above (0x1f0), and so on.
4107 By default all supported ports are probed.
4109 pata_legacy.qdi= [HW,LIBATA]
4111 Set to non-zero to probe QDI controllers. By default
4112 set to 1 if CONFIG_PATA_QDI_MODULE, 0 otherwise.
4114 pata_legacy.winbond= [HW,LIBATA]
4116 Set to non-zero to probe Winbond controllers. Use
4117 the standard I/O port (0x130) if 1, otherwise the
4118 value given is the I/O port to use (typically 0x1b0).
4119 By default set to 1 if CONFIG_PATA_WINBOND_VLB_MODULE,
4122 pata_platform.pio_mask= [HW,LIBATA]
4124 Supported PIO mode mask. Set individual bits to allow
4125 the use of the respective PIO modes. Bit 0 is for
4126 mode 0, bit 1 is for mode 1, and so on. Mode 0 only
4130 Halt all CPUs after the first oops has been printed for
4131 the specified number of seconds. This is to be used if
4132 your oopses keep scrolling off the screen.
4136 pci=option[,option...] [PCI] various PCI subsystem options.
4138 Some options herein operate on a specific device
4139 or a set of devices (<pci_dev>). These are
4140 specified in one of the following formats:
4142 [<domain>:]<bus>:<dev>.<func>[/<dev>.<func>]*
4143 pci:<vendor>:<device>[:<subvendor>:<subdevice>]
4145 Note: the first format specifies a PCI
4146 bus/device/function address which may change
4147 if new hardware is inserted, if motherboard
4148 firmware changes, or due to changes caused
4149 by other kernel parameters. If the
4150 domain is left unspecified, it is
4151 taken to be zero. Optionally, a path
4152 to a device through multiple device/function
4153 addresses can be specified after the base
4154 address (this is more robust against
4155 renumbering issues). The second format
4156 selects devices using IDs from the
4157 configuration space which may match multiple
4158 devices in the system.
4160 earlydump dump PCI config space before the kernel
4162 off [X86] don't probe for the PCI bus
4163 bios [X86-32] force use of PCI BIOS, don't access
4164 the hardware directly. Use this if your machine
4165 has a non-standard PCI host bridge.
4166 nobios [X86-32] disallow use of PCI BIOS, only direct
4167 hardware access methods are allowed. Use this
4168 if you experience crashes upon bootup and you
4169 suspect they are caused by the BIOS.
4170 conf1 [X86] Force use of PCI Configuration Access
4171 Mechanism 1 (config address in IO port 0xCF8,
4172 data in IO port 0xCFC, both 32-bit).
4173 conf2 [X86] Force use of PCI Configuration Access
4174 Mechanism 2 (IO port 0xCF8 is an 8-bit port for
4175 the function, IO port 0xCFA, also 8-bit, sets
4176 bus number. The config space is then accessed
4177 through ports 0xC000-0xCFFF).
4178 See http://wiki.osdev.org/PCI for more info
4179 on the configuration access mechanisms.
4180 noaer [PCIE] If the PCIEAER kernel config parameter is
4181 enabled, this kernel boot option can be used to
4182 disable the use of PCIE advanced error reporting.
4183 nodomains [PCI] Disable support for multiple PCI
4184 root domains (aka PCI segments, in ACPI-speak).
4185 nommconf [X86] Disable use of MMCONFIG for PCI
4187 check_enable_amd_mmconf [X86] check for and enable
4188 properly configured MMIO access to PCI
4189 config space on AMD family 10h CPU
4190 nomsi [MSI] If the PCI_MSI kernel config parameter is
4191 enabled, this kernel boot option can be used to
4192 disable the use of MSI interrupts system-wide.
4193 noioapicquirk [APIC] Disable all boot interrupt quirks.
4194 Safety option to keep boot IRQs enabled. This
4195 should never be necessary.
4196 ioapicreroute [APIC] Enable rerouting of boot IRQs to the
4197 primary IO-APIC for bridges that cannot disable
4198 boot IRQs. This fixes a source of spurious IRQs
4199 when the system masks IRQs.
4200 noioapicreroute [APIC] Disable workaround that uses the
4201 boot IRQ equivalent of an IRQ that connects to
4202 a chipset where boot IRQs cannot be disabled.
4203 The opposite of ioapicreroute.
4204 biosirq [X86-32] Use PCI BIOS calls to get the interrupt
4205 routing table. These calls are known to be buggy
4206 on several machines and they hang the machine
4207 when used, but on other computers it's the only
4208 way to get the interrupt routing table. Try
4209 this option if the kernel is unable to allocate
4210 IRQs or discover secondary PCI buses on your
4212 rom [X86] Assign address space to expansion ROMs.
4213 Use with caution as certain devices share
4214 address decoders between ROMs and other
4216 norom [X86] Do not assign address space to
4217 expansion ROMs that do not already have
4218 BIOS assigned address ranges.
4219 nobar [X86] Do not assign address space to the
4220 BARs that weren't assigned by the BIOS.
4221 irqmask=0xMMMM [X86] Set a bit mask of IRQs allowed to be
4222 assigned automatically to PCI devices. You can
4223 make the kernel exclude IRQs of your ISA cards
4225 pirqaddr=0xAAAAA [X86] Specify the physical address
4226 of the PIRQ table (normally generated
4227 by the BIOS) if it is outside the
4228 F0000h-100000h range.
4229 lastbus=N [X86] Scan all buses thru bus #N. Can be
4230 useful if the kernel is unable to find your
4231 secondary buses and you want to tell it
4232 explicitly which ones they are.
4233 assign-busses [X86] Always assign all PCI bus
4234 numbers ourselves, overriding
4235 whatever the firmware may have done.
4236 usepirqmask [X86] Honor the possible IRQ mask stored
4237 in the BIOS $PIR table. This is needed on
4238 some systems with broken BIOSes, notably
4239 some HP Pavilion N5400 and Omnibook XE3
4240 notebooks. This will have no effect if ACPI
4241 IRQ routing is enabled.
4242 noacpi [X86] Do not use ACPI for IRQ routing
4243 or for PCI scanning.
4244 use_crs [X86] Use PCI host bridge window information
4245 from ACPI. On BIOSes from 2008 or later, this
4246 is enabled by default. If you need to use this,
4247 please report a bug.
4248 nocrs [X86] Ignore PCI host bridge windows from ACPI.
4249 If you need to use this, please report a bug.
4250 use_e820 [X86] Use E820 reservations to exclude parts of
4251 PCI host bridge windows. This is a workaround
4252 for BIOS defects in host bridge _CRS methods.
4253 If you need to use this, please report a bug to
4254 <linux-pci@vger.kernel.org>.
4255 no_e820 [X86] Ignore E820 reservations for PCI host
4256 bridge windows. This is the default on modern
4257 hardware. If you need to use this, please report
4258 a bug to <linux-pci@vger.kernel.org>.
4259 routeirq Do IRQ routing for all PCI devices.
4260 This is normally done in pci_enable_device(),
4261 so this option is a temporary workaround
4262 for broken drivers that don't call it.
4263 skip_isa_align [X86] do not align io start addr, so can
4264 handle more pci cards
4265 noearly [X86] Don't do any early type 1 scanning.
4266 This might help on some broken boards which
4267 machine check when some devices' config space
4268 is read. But various workarounds are disabled
4269 and some IOMMU drivers will not work.
4270 bfsort Sort PCI devices into breadth-first order.
4271 This sorting is done to get a device
4272 order compatible with older (<= 2.4) kernels.
4273 nobfsort Don't sort PCI devices into breadth-first order.
4274 pcie_bus_tune_off Disable PCIe MPS (Max Payload Size)
4275 tuning and use the BIOS-configured MPS defaults.
4276 pcie_bus_safe Set every device's MPS to the largest value
4277 supported by all devices below the root complex.
4278 pcie_bus_perf Set device MPS to the largest allowable MPS
4279 based on its parent bus. Also set MRRS (Max
4280 Read Request Size) to the largest supported
4281 value (no larger than the MPS that the device
4282 or bus can support) for best performance.
4283 pcie_bus_peer2peer Set every device's MPS to 128B, which
4284 every device is guaranteed to support. This
4285 configuration allows peer-to-peer DMA between
4286 any pair of devices, possibly at the cost of
4287 reduced performance. This also guarantees
4288 that hot-added devices will work.
4289 cbiosize=nn[KMG] The fixed amount of bus space which is
4290 reserved for the CardBus bridge's IO window.
4291 The default value is 256 bytes.
4292 cbmemsize=nn[KMG] The fixed amount of bus space which is
4293 reserved for the CardBus bridge's memory
4294 window. The default value is 64 megabytes.
4297 [<order of align>@]<pci_dev>[; ...]
4298 Specifies alignment and device to reassign
4299 aligned memory resources. How to
4300 specify the device is described above.
4301 If <order of align> is not specified,
4302 PAGE_SIZE is used as alignment.
4303 A PCI-PCI bridge can be specified if resource
4304 windows need to be expanded.
4305 To specify the alignment for several
4306 instances of a device, the PCI vendor,
4307 device, subvendor, and subdevice may be
4308 specified, e.g., 12@pci:8086:9c22:103c:198f
4309 for 4096-byte alignment.
4310 ecrc= Enable/disable PCIe ECRC (transaction layer
4311 end-to-end CRC checking). Only effective if
4312 OS has native AER control (either granted by
4313 ACPI _OSC or forced via "pcie_ports=native")
4314 bios: Use BIOS/firmware settings. This is the
4318 hpiosize=nn[KMG] The fixed amount of bus space which is
4319 reserved for hotplug bridge's IO window.
4320 Default size is 256 bytes.
4321 hpmmiosize=nn[KMG] The fixed amount of bus space which is
4322 reserved for hotplug bridge's MMIO window.
4323 Default size is 2 megabytes.
4324 hpmmioprefsize=nn[KMG] The fixed amount of bus space which is
4325 reserved for hotplug bridge's MMIO_PREF window.
4326 Default size is 2 megabytes.
4327 hpmemsize=nn[KMG] The fixed amount of bus space which is
4328 reserved for hotplug bridge's MMIO and
4330 Default size is 2 megabytes.
4331 hpbussize=nn The minimum amount of additional bus numbers
4332 reserved for buses below a hotplug bridge.
4334 realloc= Enable/disable reallocating PCI bridge resources
4335 if allocations done by BIOS are too small to
4336 accommodate resources required by all child
4338 off: Turn realloc off
4340 realloc same as realloc=on
4341 noari do not use PCIe ARI.
4342 noats [PCIE, Intel-IOMMU, AMD-IOMMU]
4343 do not use PCIe ATS (and IOMMU device IOTLB).
4344 pcie_scan_all Scan all possible PCIe devices. Otherwise we
4345 only look for one device below a PCIe downstream
4347 big_root_window Try to add a big 64bit memory window to the PCIe
4348 root complex on AMD CPUs. Some GFX hardware
4349 can resize a BAR to allow access to all VRAM.
4350 Adding the window is slightly risky (it may
4351 conflict with unreported devices), so this
4353 disable_acs_redir=<pci_dev>[; ...]
4354 Specify one or more PCI devices (in the format
4355 specified above) separated by semicolons.
4356 Each device specified will have the PCI ACS
4357 redirect capabilities forced off which will
4358 allow P2P traffic between devices through
4359 bridges without forcing it upstream. Note:
4360 this removes isolation between devices and
4361 may put more devices in an IOMMU group.
4362 force_floating [S390] Force usage of floating interrupts.
4363 nomio [S390] Do not use MIO instructions.
4364 norid [S390] ignore the RID field and force use of
4365 one PCI domain per PCI function
4367 pcie_aspm= [PCIE] Forcibly enable or disable PCIe Active State Power
4370 force Enable ASPM even on devices that claim not to support it.
4371 WARNING: Forcing ASPM on may cause system lockups.
4373 pcie_ports= [PCIE] PCIe port services handling:
4374 native Use native PCIe services (PME, AER, DPC, PCIe hotplug)
4375 even if the platform doesn't give the OS permission to
4376 use them. This may cause conflicts if the platform
4377 also tries to use these services.
4378 dpc-native Use native PCIe service for DPC only. May
4379 cause conflicts if firmware uses AER or DPC.
4380 compat Disable native PCIe services (PME, AER, DPC, PCIe
4383 pcie_port_pm= [PCIE] PCIe port power management handling:
4384 off Disable power management of all PCIe ports
4385 force Forcibly enable power management of all PCIe ports
4387 pcie_pme= [PCIE,PM] Native PCIe PME signaling options:
4388 nomsi Do not use MSI for native PCIe PME signaling (this makes
4389 all PCIe root ports use INTx for all services).
4391 pcmv= [HW,PCMCIA] BadgePAD 4
4395 Keep all power-domains already enabled by bootloader on,
4396 even if no driver has claimed them. This is useful
4397 for debug and development, but should not be
4398 needed on a platform with proper driver support.
4400 pdcchassis= [PARISC,HW] Disable/Enable PDC Chassis Status codes at
4403 See arch/parisc/kernel/pdc_chassis.c
4405 percpu_alloc= Select which percpu first chunk allocator to use.
4406 Currently supported values are "embed" and "page".
4407 Archs may support subset or none of the selections.
4408 See comments in mm/percpu.c for details on each
4409 allocator. This parameter is primarily for debugging
4410 and performance comparison.
4412 pirq= [SMP,APIC] Manual mp-table setup
4413 See Documentation/x86/i386/IO-APIC.rst.
4415 plip= [PPT,NET] Parallel port network link
4416 Format: { parport<nr> | timid | 0 }
4417 See also Documentation/admin-guide/parport.rst.
4419 pmtmr= [X86] Manual setup of pmtmr I/O Port.
4420 Override pmtimer IOPort with a hex value.
4423 pmu_override= [PPC] Override the PMU.
4424 This option takes over the PMU facility, so it is no
4425 longer usable by perf. Setting this option starts the
4426 PMU counters by setting MMCR0 to 0 (the FC bit is
4427 cleared). If a number is given, then MMCR1 is set to
4428 that number, otherwise (e.g., 'pmu_override=on'), MMCR1
4431 pm_debug_messages [SUSPEND,KNL]
4432 Enable suspend/resume debug messages during boot up.
4435 Enable PNP debug messages (depends on the
4436 CONFIG_PNP_DEBUG_MESSAGES option). Change at run-time
4437 via /sys/module/pnp/parameters/debug. We always show
4438 current resource usage; turning this on also shows
4439 possible settings and some assignment information.
4445 { on | off | curr | res | no-curr | no-res }
4448 [ISAPNP] Exclude IRQs for the autoconfiguration
4451 [ISAPNP] Exclude DMAs for the autoconfiguration
4453 pnp_reserve_io= [ISAPNP] Exclude I/O ports for the autoconfiguration
4454 Ranges are in pairs (I/O port base and size).
4457 [ISAPNP] Exclude memory regions for the
4459 Ranges are in pairs (memory base and size).
4461 ports= [IP_VS_FTP] IPVS ftp helper module
4463 Up to 8 (IP_VS_APP_MAX_PORTS) ports
4465 Format: <port>,<port>....
4467 powersave=off [PPC] This option disables power saving features.
4468 It specifically disables cpuidle and sets the
4469 platform machine description specific power_save
4470 function to NULL. On Idle the CPU just reduces
4473 ppc_strict_facility_enable
4474 [PPC] This option catches any kernel floating point,
4475 Altivec, VSX and SPE outside of regions specifically
4476 allowed (eg kernel_enable_fpu()/kernel_disable_fpu()).
4477 There is some performance impact when enabling this.
4481 Disable Hardware Transactional Memory
4484 Select preemption mode if you have CONFIG_PREEMPT_DYNAMIC
4485 none - Limited to cond_resched() calls
4486 voluntary - Limited to cond_resched() and might_sleep() calls
4487 full - Any section that isn't explicitly preempt disabled
4488 can be preempted anytime.
4490 print-fatal-signals=
4491 [KNL] debug: print fatal signals
4493 If enabled, warn about various signal handling
4494 related application anomalies: too many signals,
4495 too many POSIX.1 timers, fatal signals causing a
4498 If you hit the warning due to signal overflow,
4499 you might want to try "ulimit -i unlimited".
4503 printk.always_kmsg_dump=
4504 Trigger kmsg_dump for cases other than kernel oops or
4506 Format: <bool> (1/Y/y=enable, 0/N/n=disable)
4509 printk.console_no_auto_verbose=
4510 Disable console loglevel raise on oops, panic
4511 or lockdep-detected issues (only if lock debug is on).
4512 With an exception to setups with low baudrate on
4513 serial console, keeping this 0 is a good choice
4514 in order to provide more debug information.
4516 default: 0 (auto_verbose is enabled)
4518 printk.devkmsg={on,off,ratelimit}
4519 Control writing to /dev/kmsg.
4520 on - unlimited logging to /dev/kmsg from userspace
4521 off - logging to /dev/kmsg disabled
4522 ratelimit - ratelimit the logging
4525 printk.time= Show timing data prefixed to each printk message line
4526 Format: <bool> (1/Y/y=enable, 0/N/n=disable)
4528 processor.max_cstate= [HW,ACPI]
4529 Limit processor to maximum C-state
4530 max_cstate=9 overrides any DMI blacklist limit.
4532 processor.nocst [HW,ACPI]
4533 Ignore the _CST method to determine C-states,
4534 instead using the legacy FADT method
4536 profile= [KNL] Enable kernel profiling via /proc/profile
4537 Format: [<profiletype>,]<number>
4538 Param: <profiletype>: "schedule", "sleep", or "kvm"
4539 [defaults to kernel profiling]
4540 Param: "schedule" - profile schedule points.
4541 Param: "sleep" - profile D-state sleeping (millisecs).
4542 Requires CONFIG_SCHEDSTATS
4543 Param: "kvm" - profile VM exits.
4544 Param: <number> - step/bucket size as a power of 2 for
4545 statistical time based profiling.
4547 prompt_ramdisk= [RAM] [Deprecated]
4549 prot_virt= [S390] enable hosting protected virtual machines
4550 isolated from the hypervisor (if hardware supports
4554 psi= [KNL] Enable or disable pressure stall information
4558 psmouse.proto= [HW,MOUSE] Highest PS2 mouse protocol extension to
4559 probe for; one of (bare|imps|exps|lifebook|any).
4560 psmouse.rate= [HW,MOUSE] Set desired mouse report rate, in reports
4562 psmouse.resetafter= [HW,MOUSE]
4563 Try to reset the device after so many bad packets
4566 [HW,MOUSE] Set desired mouse resolution, in dpi.
4567 psmouse.smartscroll=
4568 [HW,MOUSE] Controls Logitech smartscroll autorepeat.
4569 0 = disabled, 1 = enabled (default).
4571 pstore.backend= Specify the name of the pstore backend to use
4573 pti= [X86-64] Control Page Table Isolation of user and
4574 kernel address spaces. Disabling this feature
4575 removes hardening, but improves performance of
4576 system calls and interrupts.
4578 on - unconditionally enable
4579 off - unconditionally disable
4580 auto - kernel detects whether your CPU model is
4581 vulnerable to issues that PTI mitigates
4583 Not specifying this option is equivalent to pti=auto.
4586 Equivalent to pti=off
4589 [KNL] Number of legacy pty's. Overwrites compiled-in
4592 quiet [KNL] Disable most log messages
4596 radix_hcall_invalidate=on [PPC/PSERIES]
4597 Disable RADIX GTSE feature and use hcall for TLB
4601 See Documentation/admin-guide/md.rst.
4603 ramdisk_size= [RAM] Sizes of RAM disks in kilobytes
4604 See Documentation/admin-guide/blockdev/ramdisk.rst.
4606 ramdisk_start= [RAM] RAM disk image start address
4608 random.trust_cpu=off
4609 [KNL] Disable trusting the use of the CPU's
4610 random number generator (if available) to
4611 initialize the kernel's RNG.
4613 random.trust_bootloader=off
4614 [KNL] Disable trusting the use of the a seed
4615 passed by the bootloader (if available) to
4616 initialize the kernel's RNG.
4618 randomize_kstack_offset=
4619 [KNL] Enable or disable kernel stack offset
4620 randomization, which provides roughly 5 bits of
4621 entropy, frustrating memory corruption attacks
4622 that depend on stack address determinism or
4623 cross-syscall address exposures. This is only
4624 available on architectures that have defined
4625 CONFIG_HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET.
4626 Format: <bool> (1/Y/y=enable, 0/N/n=disable)
4627 Default is CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT.
4629 ras=option[,option,...] [KNL] RAS-specific options
4632 Disable the Correctable Errors Collector,
4633 see CONFIG_RAS_CEC help text.
4635 rcu_nocbs[=cpu-list]
4636 [KNL] The optional argument is a cpu list,
4639 In kernels built with CONFIG_RCU_NOCB_CPU=y,
4640 enable the no-callback CPU mode, which prevents
4641 such CPUs' callbacks from being invoked in
4642 softirq context. Invocation of such CPUs' RCU
4643 callbacks will instead be offloaded to "rcuox/N"
4644 kthreads created for that purpose, where "x" is
4645 "p" for RCU-preempt, "s" for RCU-sched, and "g"
4646 for the kthreads that mediate grace periods; and
4647 "N" is the CPU number. This reduces OS jitter on
4648 the offloaded CPUs, which can be useful for HPC
4649 and real-time workloads. It can also improve
4650 energy efficiency for asymmetric multiprocessors.
4652 If a cpulist is passed as an argument, the specified
4653 list of CPUs is set to no-callback mode from boot.
4655 Otherwise, if the '=' sign and the cpulist
4656 arguments are omitted, no CPU will be set to
4657 no-callback mode from boot but the mode may be
4658 toggled at runtime via cpusets.
4660 Note that this argument takes precedence over
4661 the CONFIG_RCU_NOCB_CPU_DEFAULT_ALL option.
4664 Rather than requiring that offloaded CPUs
4665 (specified by rcu_nocbs= above) explicitly
4666 awaken the corresponding "rcuoN" kthreads,
4667 make these kthreads poll for callbacks.
4668 This improves the real-time response for the
4669 offloaded CPUs by relieving them of the need to
4670 wake up the corresponding kthread, but degrades
4671 energy efficiency by requiring that the kthreads
4672 periodically wake up to do the polling.
4674 rcutree.blimit= [KNL]
4675 Set maximum number of finished RCU callbacks to
4676 process in one batch.
4678 rcutree.dump_tree= [KNL]
4679 Dump the structure of the rcu_node combining tree
4680 out at early boot. This is used for diagnostic
4681 purposes, to verify correct tree setup.
4683 rcutree.gp_cleanup_delay= [KNL]
4684 Set the number of jiffies to delay each step of
4685 RCU grace-period cleanup.
4687 rcutree.gp_init_delay= [KNL]
4688 Set the number of jiffies to delay each step of
4689 RCU grace-period initialization.
4691 rcutree.gp_preinit_delay= [KNL]
4692 Set the number of jiffies to delay each step of
4693 RCU grace-period pre-initialization, that is,
4694 the propagation of recent CPU-hotplug changes up
4695 the rcu_node combining tree.
4697 rcutree.use_softirq= [KNL]
4698 If set to zero, move all RCU_SOFTIRQ processing to
4699 per-CPU rcuc kthreads. Defaults to a non-zero
4700 value, meaning that RCU_SOFTIRQ is used by default.
4701 Specify rcutree.use_softirq=0 to use rcuc kthreads.
4703 But note that CONFIG_PREEMPT_RT=y kernels disable
4704 this kernel boot parameter, forcibly setting it
4707 rcutree.rcu_fanout_exact= [KNL]
4708 Disable autobalancing of the rcu_node combining
4709 tree. This is used by rcutorture, and might
4710 possibly be useful for architectures having high
4711 cache-to-cache transfer latencies.
4713 rcutree.rcu_fanout_leaf= [KNL]
4714 Change the number of CPUs assigned to each
4715 leaf rcu_node structure. Useful for very
4716 large systems, which will choose the value 64,
4717 and for NUMA systems with large remote-access
4718 latencies, which will choose a value aligned
4719 with the appropriate hardware boundaries.
4721 rcutree.rcu_min_cached_objs= [KNL]
4722 Minimum number of objects which are cached and
4723 maintained per one CPU. Object size is equal
4724 to PAGE_SIZE. The cache allows to reduce the
4725 pressure to page allocator, also it makes the
4726 whole algorithm to behave better in low memory
4729 rcutree.rcu_delay_page_cache_fill_msec= [KNL]
4730 Set the page-cache refill delay (in milliseconds)
4731 in response to low-memory conditions. The range
4732 of permitted values is in the range 0:100000.
4734 rcutree.jiffies_till_first_fqs= [KNL]
4735 Set delay from grace-period initialization to
4736 first attempt to force quiescent states.
4737 Units are jiffies, minimum value is zero,
4738 and maximum value is HZ.
4740 rcutree.jiffies_till_next_fqs= [KNL]
4741 Set delay between subsequent attempts to force
4742 quiescent states. Units are jiffies, minimum
4743 value is one, and maximum value is HZ.
4745 rcutree.jiffies_till_sched_qs= [KNL]
4746 Set required age in jiffies for a
4747 given grace period before RCU starts
4748 soliciting quiescent-state help from
4749 rcu_note_context_switch() and cond_resched().
4750 If not specified, the kernel will calculate
4751 a value based on the most recent settings
4752 of rcutree.jiffies_till_first_fqs
4753 and rcutree.jiffies_till_next_fqs.
4754 This calculated value may be viewed in
4755 rcutree.jiffies_to_sched_qs. Any attempt to set
4756 rcutree.jiffies_to_sched_qs will be cheerfully
4759 rcutree.kthread_prio= [KNL,BOOT]
4760 Set the SCHED_FIFO priority of the RCU per-CPU
4761 kthreads (rcuc/N). This value is also used for
4762 the priority of the RCU boost threads (rcub/N)
4763 and for the RCU grace-period kthreads (rcu_bh,
4764 rcu_preempt, and rcu_sched). If RCU_BOOST is
4765 set, valid values are 1-99 and the default is 1
4766 (the least-favored priority). Otherwise, when
4767 RCU_BOOST is not set, valid values are 0-99 and
4768 the default is zero (non-realtime operation).
4769 When RCU_NOCB_CPU is set, also adjust the
4770 priority of NOCB callback kthreads.
4772 rcutree.rcu_divisor= [KNL]
4773 Set the shift-right count to use to compute
4774 the callback-invocation batch limit bl from
4775 the number of callbacks queued on this CPU.
4776 The result will be bounded below by the value of
4777 the rcutree.blimit kernel parameter. Every bl
4778 callbacks, the softirq handler will exit in
4779 order to allow the CPU to do other work.
4781 Please note that this callback-invocation batch
4782 limit applies only to non-offloaded callback
4783 invocation. Offloaded callbacks are instead
4784 invoked in the context of an rcuoc kthread, which
4785 scheduler will preempt as it does any other task.
4787 rcutree.nocb_nobypass_lim_per_jiffy= [KNL]
4788 On callback-offloaded (rcu_nocbs) CPUs,
4789 RCU reduces the lock contention that would
4790 otherwise be caused by callback floods through
4791 use of the ->nocb_bypass list. However, in the
4792 common non-flooded case, RCU queues directly to
4793 the main ->cblist in order to avoid the extra
4794 overhead of the ->nocb_bypass list and its lock.
4795 But if there are too many callbacks queued during
4796 a single jiffy, RCU pre-queues the callbacks into
4797 the ->nocb_bypass queue. The definition of "too
4798 many" is supplied by this kernel boot parameter.
4800 rcutree.rcu_nocb_gp_stride= [KNL]
4801 Set the number of NOCB callback kthreads in
4802 each group, which defaults to the square root
4803 of the number of CPUs. Larger numbers reduce
4804 the wakeup overhead on the global grace-period
4805 kthread, but increases that same overhead on
4806 each group's NOCB grace-period kthread.
4808 rcutree.qhimark= [KNL]
4809 Set threshold of queued RCU callbacks beyond which
4810 batch limiting is disabled.
4812 rcutree.qlowmark= [KNL]
4813 Set threshold of queued RCU callbacks below which
4814 batch limiting is re-enabled.
4816 rcutree.qovld= [KNL]
4817 Set threshold of queued RCU callbacks beyond which
4818 RCU's force-quiescent-state scan will aggressively
4819 enlist help from cond_resched() and sched IPIs to
4820 help CPUs more quickly reach quiescent states.
4821 Set to less than zero to make this be set based
4822 on rcutree.qhimark at boot time and to zero to
4823 disable more aggressive help enlistment.
4825 rcutree.rcu_kick_kthreads= [KNL]
4826 Cause the grace-period kthread to get an extra
4827 wake_up() if it sleeps three times longer than
4828 it should at force-quiescent-state time.
4829 This wake_up() will be accompanied by a
4830 WARN_ONCE() splat and an ftrace_dump().
4832 rcutree.rcu_unlock_delay= [KNL]
4833 In CONFIG_RCU_STRICT_GRACE_PERIOD=y kernels,
4834 this specifies an rcu_read_unlock()-time delay
4835 in microseconds. This defaults to zero.
4836 Larger delays increase the probability of
4837 catching RCU pointer leaks, that is, buggy use
4838 of RCU-protected pointers after the relevant
4839 rcu_read_unlock() has completed.
4841 rcutree.sysrq_rcu= [KNL]
4842 Commandeer a sysrq key to dump out Tree RCU's
4843 rcu_node tree with an eye towards determining
4844 why a new grace period has not yet started.
4846 rcuscale.gp_async= [KNL]
4847 Measure performance of asynchronous
4848 grace-period primitives such as call_rcu().
4850 rcuscale.gp_async_max= [KNL]
4851 Specify the maximum number of outstanding
4852 callbacks per writer thread. When a writer
4853 thread exceeds this limit, it invokes the
4854 corresponding flavor of rcu_barrier() to allow
4855 previously posted callbacks to drain.
4857 rcuscale.gp_exp= [KNL]
4858 Measure performance of expedited synchronous
4859 grace-period primitives.
4861 rcuscale.holdoff= [KNL]
4862 Set test-start holdoff period. The purpose of
4863 this parameter is to delay the start of the
4864 test until boot completes in order to avoid
4867 rcuscale.kfree_rcu_test= [KNL]
4868 Set to measure performance of kfree_rcu() flooding.
4870 rcuscale.kfree_rcu_test_double= [KNL]
4871 Test the double-argument variant of kfree_rcu().
4872 If this parameter has the same value as
4873 rcuscale.kfree_rcu_test_single, both the single-
4874 and double-argument variants are tested.
4876 rcuscale.kfree_rcu_test_single= [KNL]
4877 Test the single-argument variant of kfree_rcu().
4878 If this parameter has the same value as
4879 rcuscale.kfree_rcu_test_double, both the single-
4880 and double-argument variants are tested.
4882 rcuscale.kfree_nthreads= [KNL]
4883 The number of threads running loops of kfree_rcu().
4885 rcuscale.kfree_alloc_num= [KNL]
4886 Number of allocations and frees done in an iteration.
4888 rcuscale.kfree_loops= [KNL]
4889 Number of loops doing rcuscale.kfree_alloc_num number
4890 of allocations and frees.
4892 rcuscale.nreaders= [KNL]
4893 Set number of RCU readers. The value -1 selects
4894 N, where N is the number of CPUs. A value
4895 "n" less than -1 selects N-n+1, where N is again
4896 the number of CPUs. For example, -2 selects N
4897 (the number of CPUs), -3 selects N+1, and so on.
4898 A value of "n" less than or equal to -N selects
4901 rcuscale.nwriters= [KNL]
4902 Set number of RCU writers. The values operate
4903 the same as for rcuscale.nreaders.
4904 N, where N is the number of CPUs
4906 rcuscale.perf_type= [KNL]
4907 Specify the RCU implementation to test.
4909 rcuscale.shutdown= [KNL]
4910 Shut the system down after performance tests
4911 complete. This is useful for hands-off automated
4914 rcuscale.verbose= [KNL]
4915 Enable additional printk() statements.
4917 rcuscale.writer_holdoff= [KNL]
4918 Write-side holdoff between grace periods,
4919 in microseconds. The default of zero says
4922 rcutorture.fqs_duration= [KNL]
4923 Set duration of force_quiescent_state bursts
4926 rcutorture.fqs_holdoff= [KNL]
4927 Set holdoff time within force_quiescent_state bursts
4930 rcutorture.fqs_stutter= [KNL]
4931 Set wait time between force_quiescent_state bursts
4934 rcutorture.fwd_progress= [KNL]
4935 Specifies the number of kthreads to be used
4936 for RCU grace-period forward-progress testing
4937 for the types of RCU supporting this notion.
4938 Defaults to 1 kthread, values less than zero or
4939 greater than the number of CPUs cause the number
4942 rcutorture.fwd_progress_div= [KNL]
4943 Specify the fraction of a CPU-stall-warning
4944 period to do tight-loop forward-progress testing.
4946 rcutorture.fwd_progress_holdoff= [KNL]
4947 Number of seconds to wait between successive
4948 forward-progress tests.
4950 rcutorture.fwd_progress_need_resched= [KNL]
4951 Enclose cond_resched() calls within checks for
4952 need_resched() during tight-loop forward-progress
4955 rcutorture.gp_cond= [KNL]
4956 Use conditional/asynchronous update-side
4957 primitives, if available.
4959 rcutorture.gp_exp= [KNL]
4960 Use expedited update-side primitives, if available.
4962 rcutorture.gp_normal= [KNL]
4963 Use normal (non-expedited) asynchronous
4964 update-side primitives, if available.
4966 rcutorture.gp_sync= [KNL]
4967 Use normal (non-expedited) synchronous
4968 update-side primitives, if available. If all
4969 of rcutorture.gp_cond=, rcutorture.gp_exp=,
4970 rcutorture.gp_normal=, and rcutorture.gp_sync=
4971 are zero, rcutorture acts as if is interpreted
4972 they are all non-zero.
4974 rcutorture.irqreader= [KNL]
4975 Run RCU readers from irq handlers, or, more
4976 accurately, from a timer handler. Not all RCU
4977 flavors take kindly to this sort of thing.
4979 rcutorture.leakpointer= [KNL]
4980 Leak an RCU-protected pointer out of the reader.
4981 This can of course result in splats, and is
4982 intended to test the ability of things like
4983 CONFIG_RCU_STRICT_GRACE_PERIOD=y to detect
4986 rcutorture.n_barrier_cbs= [KNL]
4987 Set callbacks/threads for rcu_barrier() testing.
4989 rcutorture.nfakewriters= [KNL]
4990 Set number of concurrent RCU writers. These just
4991 stress RCU, they don't participate in the actual
4992 test, hence the "fake".
4994 rcutorture.nocbs_nthreads= [KNL]
4995 Set number of RCU callback-offload togglers.
4996 Zero (the default) disables toggling.
4998 rcutorture.nocbs_toggle= [KNL]
4999 Set the delay in milliseconds between successive
5000 callback-offload toggling attempts.
5002 rcutorture.nreaders= [KNL]
5003 Set number of RCU readers. The value -1 selects
5004 N-1, where N is the number of CPUs. A value
5005 "n" less than -1 selects N-n-2, where N is again
5006 the number of CPUs. For example, -2 selects N
5007 (the number of CPUs), -3 selects N+1, and so on.
5009 rcutorture.object_debug= [KNL]
5010 Enable debug-object double-call_rcu() testing.
5012 rcutorture.onoff_holdoff= [KNL]
5013 Set time (s) after boot for CPU-hotplug testing.
5015 rcutorture.onoff_interval= [KNL]
5016 Set time (jiffies) between CPU-hotplug operations,
5017 or zero to disable CPU-hotplug testing.
5019 rcutorture.read_exit= [KNL]
5020 Set the number of read-then-exit kthreads used
5021 to test the interaction of RCU updaters and
5022 task-exit processing.
5024 rcutorture.read_exit_burst= [KNL]
5025 The number of times in a given read-then-exit
5026 episode that a set of read-then-exit kthreads
5029 rcutorture.read_exit_delay= [KNL]
5030 The delay, in seconds, between successive
5031 read-then-exit testing episodes.
5033 rcutorture.shuffle_interval= [KNL]
5034 Set task-shuffle interval (s). Shuffling tasks
5035 allows some CPUs to go into dyntick-idle mode
5036 during the rcutorture test.
5038 rcutorture.shutdown_secs= [KNL]
5039 Set time (s) after boot system shutdown. This
5040 is useful for hands-off automated testing.
5042 rcutorture.stall_cpu= [KNL]
5043 Duration of CPU stall (s) to test RCU CPU stall
5044 warnings, zero to disable.
5046 rcutorture.stall_cpu_block= [KNL]
5047 Sleep while stalling if set. This will result
5048 in warnings from preemptible RCU in addition
5049 to any other stall-related activity.
5051 rcutorture.stall_cpu_holdoff= [KNL]
5052 Time to wait (s) after boot before inducing stall.
5054 rcutorture.stall_cpu_irqsoff= [KNL]
5055 Disable interrupts while stalling if set.
5057 rcutorture.stall_gp_kthread= [KNL]
5058 Duration (s) of forced sleep within RCU
5059 grace-period kthread to test RCU CPU stall
5060 warnings, zero to disable. If both stall_cpu
5061 and stall_gp_kthread are specified, the
5062 kthread is starved first, then the CPU.
5064 rcutorture.stat_interval= [KNL]
5065 Time (s) between statistics printk()s.
5067 rcutorture.stutter= [KNL]
5068 Time (s) to stutter testing, for example, specifying
5069 five seconds causes the test to run for five seconds,
5070 wait for five seconds, and so on. This tests RCU's
5071 ability to transition abruptly to and from idle.
5073 rcutorture.test_boost= [KNL]
5074 Test RCU priority boosting? 0=no, 1=maybe, 2=yes.
5075 "Maybe" means test if the RCU implementation
5076 under test support RCU priority boosting.
5078 rcutorture.test_boost_duration= [KNL]
5079 Duration (s) of each individual boost test.
5081 rcutorture.test_boost_interval= [KNL]
5082 Interval (s) between each boost test.
5084 rcutorture.test_no_idle_hz= [KNL]
5085 Test RCU's dyntick-idle handling. See also the
5086 rcutorture.shuffle_interval parameter.
5088 rcutorture.torture_type= [KNL]
5089 Specify the RCU implementation to test.
5091 rcutorture.verbose= [KNL]
5092 Enable additional printk() statements.
5094 rcupdate.rcu_cpu_stall_ftrace_dump= [KNL]
5095 Dump ftrace buffer after reporting RCU CPU
5098 rcupdate.rcu_cpu_stall_suppress= [KNL]
5099 Suppress RCU CPU stall warning messages.
5101 rcupdate.rcu_cpu_stall_suppress_at_boot= [KNL]
5102 Suppress RCU CPU stall warning messages and
5103 rcutorture writer stall warnings that occur
5104 during early boot, that is, during the time
5105 before the init task is spawned.
5107 rcupdate.rcu_cpu_stall_timeout= [KNL]
5108 Set timeout for RCU CPU stall warning messages.
5109 The value is in seconds and the maximum allowed
5110 value is 300 seconds.
5112 rcupdate.rcu_exp_cpu_stall_timeout= [KNL]
5113 Set timeout for expedited RCU CPU stall warning
5114 messages. The value is in milliseconds
5115 and the maximum allowed value is 21000
5116 milliseconds. Please note that this value is
5117 adjusted to an arch timer tick resolution.
5118 Setting this to zero causes the value from
5119 rcupdate.rcu_cpu_stall_timeout to be used (after
5120 conversion from seconds to milliseconds).
5122 rcupdate.rcu_cpu_stall_cputime= [KNL]
5123 Provide statistics on the cputime and count of
5124 interrupts and tasks during the sampling period. For
5125 multiple continuous RCU stalls, all sampling periods
5126 begin at half of the first RCU stall timeout.
5128 rcupdate.rcu_exp_stall_task_details= [KNL]
5129 Print stack dumps of any tasks blocking the
5130 current expedited RCU grace period during an
5131 expedited RCU CPU stall warning.
5133 rcupdate.rcu_expedited= [KNL]
5134 Use expedited grace-period primitives, for
5135 example, synchronize_rcu_expedited() instead
5136 of synchronize_rcu(). This reduces latency,
5137 but can increase CPU utilization, degrade
5138 real-time latency, and degrade energy efficiency.
5139 No effect on CONFIG_TINY_RCU kernels.
5141 rcupdate.rcu_normal= [KNL]
5142 Use only normal grace-period primitives,
5143 for example, synchronize_rcu() instead of
5144 synchronize_rcu_expedited(). This improves
5145 real-time latency, CPU utilization, and
5146 energy efficiency, but can expose users to
5147 increased grace-period latency. This parameter
5148 overrides rcupdate.rcu_expedited. No effect on
5149 CONFIG_TINY_RCU kernels.
5151 rcupdate.rcu_normal_after_boot= [KNL]
5152 Once boot has completed (that is, after
5153 rcu_end_inkernel_boot() has been invoked), use
5154 only normal grace-period primitives. No effect
5155 on CONFIG_TINY_RCU kernels.
5157 But note that CONFIG_PREEMPT_RT=y kernels enables
5158 this kernel boot parameter, forcibly setting
5159 it to the value one, that is, converting any
5160 post-boot attempt at an expedited RCU grace
5161 period to instead use normal non-expedited
5162 grace-period processing.
5164 rcupdate.rcu_task_collapse_lim= [KNL]
5165 Set the maximum number of callbacks present
5166 at the beginning of a grace period that allows
5167 the RCU Tasks flavors to collapse back to using
5168 a single callback queue. This switching only
5169 occurs when rcupdate.rcu_task_enqueue_lim is
5170 set to the default value of -1.
5172 rcupdate.rcu_task_contend_lim= [KNL]
5173 Set the minimum number of callback-queuing-time
5174 lock-contention events per jiffy required to
5175 cause the RCU Tasks flavors to switch to per-CPU
5176 callback queuing. This switching only occurs
5177 when rcupdate.rcu_task_enqueue_lim is set to
5178 the default value of -1.
5180 rcupdate.rcu_task_enqueue_lim= [KNL]
5181 Set the number of callback queues to use for the
5182 RCU Tasks family of RCU flavors. The default
5183 of -1 allows this to be automatically (and
5184 dynamically) adjusted. This parameter is intended
5187 rcupdate.rcu_task_ipi_delay= [KNL]
5188 Set time in jiffies during which RCU tasks will
5189 avoid sending IPIs, starting with the beginning
5190 of a given grace period. Setting a large
5191 number avoids disturbing real-time workloads,
5192 but lengthens grace periods.
5194 rcupdate.rcu_task_stall_info= [KNL]
5195 Set initial timeout in jiffies for RCU task stall
5196 informational messages, which give some indication
5197 of the problem for those not patient enough to
5198 wait for ten minutes. Informational messages are
5199 only printed prior to the stall-warning message
5200 for a given grace period. Disable with a value
5201 less than or equal to zero. Defaults to ten
5202 seconds. A change in value does not take effect
5203 until the beginning of the next grace period.
5205 rcupdate.rcu_task_stall_info_mult= [KNL]
5206 Multiplier for time interval between successive
5207 RCU task stall informational messages for a given
5208 RCU tasks grace period. This value is clamped
5209 to one through ten, inclusive. It defaults to
5210 the value three, so that the first informational
5211 message is printed 10 seconds into the grace
5212 period, the second at 40 seconds, the third at
5213 160 seconds, and then the stall warning at 600
5214 seconds would prevent a fourth at 640 seconds.
5216 rcupdate.rcu_task_stall_timeout= [KNL]
5217 Set timeout in jiffies for RCU task stall
5218 warning messages. Disable with a value less
5219 than or equal to zero. Defaults to ten minutes.
5220 A change in value does not take effect until
5221 the beginning of the next grace period.
5223 rcupdate.rcu_self_test= [KNL]
5224 Run the RCU early boot self tests
5228 Run specified binary instead of /init from the ramdisk,
5229 used for early userspace startup. See initrd.
5232 force - Override the decision by the kernel to hide the
5233 advertisement of RDRAND support (this affects
5234 certain AMD processors because of buggy BIOS
5235 support, specifically around the suspend/resume
5239 Turn on/off individual RDT features. List is:
5240 cmt, mbmtotal, mbmlocal, l3cat, l3cdp, l2cat, l2cdp,
5242 E.g. to turn on cmt and turn off mba use:
5246 Format (x86 or x86_64):
5247 [w[arm] | c[old] | h[ard] | s[oft] | g[pio]] | d[efault] \
5249 [[,]b[ios] | a[cpi] | k[bd] | t[riple] | e[fi] | p[ci]] \
5251 Where reboot_mode is one of warm (soft) or cold (hard) or gpio
5252 (prefix with 'panic_' to set mode for panic
5254 reboot_type is one of bios, acpi, kbd, triple, efi, or pci,
5255 reboot_force is either force or not specified,
5256 reboot_cpu is s[mp]#### with #### being the processor
5257 to be used for rebooting.
5259 refscale.holdoff= [KNL]
5260 Set test-start holdoff period. The purpose of
5261 this parameter is to delay the start of the
5262 test until boot completes in order to avoid
5265 refscale.loops= [KNL]
5266 Set the number of loops over the synchronization
5267 primitive under test. Increasing this number
5268 reduces noise due to loop start/end overhead,
5269 but the default has already reduced the per-pass
5270 noise to a handful of picoseconds on ca. 2020
5273 refscale.nreaders= [KNL]
5274 Set number of readers. The default value of -1
5275 selects N, where N is roughly 75% of the number
5276 of CPUs. A value of zero is an interesting choice.
5278 refscale.nruns= [KNL]
5279 Set number of runs, each of which is dumped onto
5282 refscale.readdelay= [KNL]
5283 Set the read-side critical-section duration,
5284 measured in microseconds.
5286 refscale.scale_type= [KNL]
5287 Specify the read-protection implementation to test.
5289 refscale.shutdown= [KNL]
5290 Shut down the system at the end of the performance
5291 test. This defaults to 1 (shut it down) when
5292 refscale is built into the kernel and to 0 (leave
5293 it running) when refscale is built as a module.
5295 refscale.verbose= [KNL]
5296 Enable additional printk() statements.
5298 refscale.verbose_batched= [KNL]
5299 Batch the additional printk() statements. If zero
5300 (the default) or negative, print everything. Otherwise,
5301 print every Nth verbose statement, where N is the value
5305 [KNL, SMP] Set scheduler's default relax_domain_level.
5306 See Documentation/admin-guide/cgroup-v1/cpusets.rst.
5308 reserve= [KNL,BUGS] Force kernel to ignore I/O ports or memory
5309 Format: <base1>,<size1>[,<base2>,<size2>,...]
5310 Reserve I/O ports or memory so the kernel won't use
5311 them. If <base> is less than 0x10000, the region
5312 is assumed to be I/O ports; otherwise it is memory.
5314 reservetop= [X86-32]
5316 Reserves a hole at the top of the kernel virtual
5319 reset_devices [KNL] Force drivers to reset the underlying device
5320 during initialization.
5323 Specify the partition device for software suspend
5325 {/dev/<dev> | PARTUUID=<uuid> | <int>:<int> | <hex>}
5327 resume_offset= [SWSUSP]
5328 Specify the offset from the beginning of the partition
5329 given by "resume=" at which the swap header is located,
5330 in <PAGE_SIZE> units (needed only for swap files).
5331 See Documentation/power/swsusp-and-swap-files.rst
5333 resumedelay= [HIBERNATION] Delay (in seconds) to pause before attempting to
5334 read the resume files
5336 resumewait [HIBERNATION] Wait (indefinitely) for resume device to show up.
5337 Useful for devices that are detected asynchronously
5338 (e.g. USB and MMC devices).
5340 retain_initrd [RAM] Keep initrd memory after extraction
5342 retbleed= [X86] Control mitigation of RETBleed (Arbitrary
5343 Speculative Code Execution with Return Instructions)
5346 AMD-based UNRET and IBPB mitigations alone do not stop
5347 sibling threads from influencing the predictions of other
5348 sibling threads. For that reason, STIBP is used on pro-
5349 cessors that support it, and mitigate SMT on processors
5353 auto - automatically select a migitation
5354 auto,nosmt - automatically select a mitigation,
5355 disabling SMT if necessary for
5356 the full mitigation (only on Zen1
5357 and older without STIBP).
5358 ibpb - On AMD, mitigate short speculation
5359 windows on basic block boundaries too.
5360 Safe, highest perf impact. It also
5361 enables STIBP if present. Not suitable
5363 ibpb,nosmt - Like "ibpb" above but will disable SMT
5364 when STIBP is not available. This is
5365 the alternative for systems which do not
5367 unret - Force enable untrained return thunks,
5368 only effective on AMD f15h-f17h based
5370 unret,nosmt - Like unret, but will disable SMT when STIBP
5371 is not available. This is the alternative for
5372 systems which do not have STIBP.
5374 Selecting 'auto' will choose a mitigation method at run
5375 time according to the CPU.
5377 Not specifying this option is equivalent to retbleed=auto.
5379 rfkill.default_state=
5380 0 "airplane mode". All wifi, bluetooth, wimax, gps, fm,
5381 etc. communication is blocked by default.
5384 rfkill.master_switch_mode=
5385 0 The "airplane mode" button does nothing.
5386 1 The "airplane mode" button toggles between everything
5387 blocked and the previous configuration.
5388 2 The "airplane mode" button toggles between everything
5389 blocked and everything unblocked.
5391 rhash_entries= [KNL,NET]
5392 Set number of hash buckets for route cache
5395 [KNL] Disable ring 3 MONITOR/MWAIT feature on supported
5398 ro [KNL] Mount root device read-only on boot
5401 on Mark read-only kernel memory as read-only (default).
5402 off Leave read-only kernel memory writable for debugging.
5403 full Mark read-only kernel memory and aliases as read-only
5407 Enable the uart passthrough on the designated usb port
5408 on Rockchip SoCs. When active, the signals of the
5409 debug-uart get routed to the D+ and D- pins of the usb
5410 port and the regular usb controller gets disabled.
5412 root= [KNL] Root filesystem
5413 See name_to_dev_t comment in init/do_mounts.c.
5415 rootdelay= [KNL] Delay (in seconds) to pause before attempting to
5416 mount the root filesystem
5418 rootflags= [KNL] Set root filesystem mount option string
5420 rootfstype= [KNL] Set root filesystem type
5422 rootwait [KNL] Wait (indefinitely) for root device to show up.
5423 Useful for devices that are detected asynchronously
5424 (e.g. USB and MMC devices).
5426 rproc_mem=nn[KMG][@address]
5427 [KNL,ARM,CMA] Remoteproc physical memory block.
5428 Memory area to be used by remote processor image,
5431 rw [KNL] Mount root device read-write on boot
5433 S [KNL] Run init in single mode
5435 s390_iommu= [HW,S390]
5436 Set s390 IOTLB flushing mode
5438 With strict flushing every unmap operation will result in
5439 an IOTLB flush. Default is lazy flushing before reuse,
5442 s390_iommu_aperture= [KNL,S390]
5443 Specifies the size of the per device DMA address space
5444 accessible through the DMA and IOMMU APIs as a decimal
5445 factor of the size of main memory.
5446 The default is 1 meaning that one can concurrently use
5447 as many DMA addresses as physical memory is installed,
5448 if supported by hardware, and thus map all of memory
5449 once. With a value of 2 one can map all of memory twice
5450 and so on. As a special case a factor of 0 imposes no
5451 restrictions other than those given by hardware at the
5452 cost of significant additional memory use for tables.
5455 See drivers/net/irda/sa1100_ir.c.
5457 sched_verbose [KNL] Enables verbose scheduler debug messages.
5459 schedstats= [KNL,X86] Enable or disable scheduled statistics.
5460 Allowed values are enable and disable. This feature
5461 incurs a small amount of overhead in the scheduler
5462 but is useful for debugging and performance tuning.
5464 sched_thermal_decay_shift=
5465 [KNL, SMP] Set a decay shift for scheduler thermal
5466 pressure signal. Thermal pressure signal follows the
5467 default decay period of other scheduler pelt
5468 signals(usually 32 ms but configurable). Setting
5469 sched_thermal_decay_shift will left shift the decay
5470 period for the thermal pressure signal by the shift
5472 i.e. with the default pelt decay period of 32 ms
5473 sched_thermal_decay_shift thermal pressure decay pr
5477 Format: integer between 0 and 10
5480 scftorture.holdoff= [KNL]
5481 Number of seconds to hold off before starting
5482 test. Defaults to zero for module insertion and
5483 to 10 seconds for built-in smp_call_function()
5486 scftorture.longwait= [KNL]
5487 Request ridiculously long waits randomly selected
5488 up to the chosen limit in seconds. Zero (the
5489 default) disables this feature. Please note
5490 that requesting even small non-zero numbers of
5491 seconds can result in RCU CPU stall warnings,
5492 softlockup complaints, and so on.
5494 scftorture.nthreads= [KNL]
5495 Number of kthreads to spawn to invoke the
5496 smp_call_function() family of functions.
5497 The default of -1 specifies a number of kthreads
5498 equal to the number of CPUs.
5500 scftorture.onoff_holdoff= [KNL]
5501 Number seconds to wait after the start of the
5502 test before initiating CPU-hotplug operations.
5504 scftorture.onoff_interval= [KNL]
5505 Number seconds to wait between successive
5506 CPU-hotplug operations. Specifying zero (which
5507 is the default) disables CPU-hotplug operations.
5509 scftorture.shutdown_secs= [KNL]
5510 The number of seconds following the start of the
5511 test after which to shut down the system. The
5512 default of zero avoids shutting down the system.
5513 Non-zero values are useful for automated tests.
5515 scftorture.stat_interval= [KNL]
5516 The number of seconds between outputting the
5517 current test statistics to the console. A value
5518 of zero disables statistics output.
5520 scftorture.stutter_cpus= [KNL]
5521 The number of jiffies to wait between each change
5522 to the set of CPUs under test.
5524 scftorture.use_cpus_read_lock= [KNL]
5525 Use use_cpus_read_lock() instead of the default
5526 preempt_disable() to disable CPU hotplug
5527 while invoking one of the smp_call_function*()
5530 scftorture.verbose= [KNL]
5531 Enable additional printk() statements.
5533 scftorture.weight_single= [KNL]
5534 The probability weighting to use for the
5535 smp_call_function_single() function with a zero
5536 "wait" parameter. A value of -1 selects the
5537 default if all other weights are -1. However,
5538 if at least one weight has some other value, a
5539 value of -1 will instead select a weight of zero.
5541 scftorture.weight_single_wait= [KNL]
5542 The probability weighting to use for the
5543 smp_call_function_single() function with a
5544 non-zero "wait" parameter. See weight_single.
5546 scftorture.weight_many= [KNL]
5547 The probability weighting to use for the
5548 smp_call_function_many() function with a zero
5549 "wait" parameter. See weight_single.
5550 Note well that setting a high probability for
5551 this weighting can place serious IPI load
5554 scftorture.weight_many_wait= [KNL]
5555 The probability weighting to use for the
5556 smp_call_function_many() function with a
5557 non-zero "wait" parameter. See weight_single
5560 scftorture.weight_all= [KNL]
5561 The probability weighting to use for the
5562 smp_call_function_all() function with a zero
5563 "wait" parameter. See weight_single and
5566 scftorture.weight_all_wait= [KNL]
5567 The probability weighting to use for the
5568 smp_call_function_all() function with a
5569 non-zero "wait" parameter. See weight_single
5572 skew_tick= [KNL] Offset the periodic timer tick per cpu to mitigate
5573 xtime_lock contention on larger systems, and/or RCU lock
5574 contention on all systems with CONFIG_MAXSMP set.
5575 Format: { "0" | "1" }
5576 0 -- disable. (may be 1 via CONFIG_CMDLINE="skew_tick=1"
5578 Note: increases power consumption, thus should only be
5579 enabled if running jitter sensitive (HPC/RT) workloads.
5581 security= [SECURITY] Choose a legacy "major" security module to
5582 enable at boot. This has been deprecated by the
5585 selinux= [SELINUX] Disable or enable SELinux at boot time.
5586 Format: { "0" | "1" }
5587 See security/selinux/Kconfig help text.
5592 serialnumber [BUGS=X86-32]
5594 sev=option[,option...] [X86-64] See Documentation/x86/x86_64/boot-options.rst
5597 Maximal number of shapers.
5599 show_lapic= [APIC,X86] Advanced Programmable Interrupt Controller
5600 Limit apic dumping. The parameter defines the maximal
5601 number of local apics being dumped. Also it is possible
5602 to set it to "all" by meaning -- no limit here.
5603 Format: { 1 (default) | 2 | ... | all }.
5604 The parameter valid if only apic=debug or
5605 apic=verbose is specified.
5606 Example: apic=debug show_lapic=all
5614 Enable merging of slabs with similar size when the
5615 kernel is built without CONFIG_SLAB_MERGE_DEFAULT.
5618 Disable merging of slabs with similar size. May be
5619 necessary if there is some reason to distinguish
5620 allocs to different slabs, especially in hardened
5621 environments where the risk of heap overflows and
5622 layout control by attackers can usually be
5623 frustrated by disabling merging. This will reduce
5624 most of the exposure of a heap attack to a single
5625 cache (risks via metadata attacks are mostly
5626 unchanged). Debug options disable merging on their
5628 For more information see Documentation/mm/slub.rst.
5630 slab_max_order= [MM, SLAB]
5631 Determines the maximum allowed order for slabs.
5632 A high setting may cause OOMs due to memory
5633 fragmentation. Defaults to 1 for systems with
5634 more than 32MB of RAM, 0 otherwise.
5636 slub_debug[=options[,slabs][;[options[,slabs]]...] [MM, SLUB]
5637 Enabling slub_debug allows one to determine the
5638 culprit if slab objects become corrupted. Enabling
5639 slub_debug can create guard zones around objects and
5640 may poison objects when not in use. Also tracks the
5641 last alloc / free. For more information see
5642 Documentation/mm/slub.rst.
5644 slub_max_order= [MM, SLUB]
5645 Determines the maximum allowed order for slabs.
5646 A high setting may cause OOMs due to memory
5647 fragmentation. For more information see
5648 Documentation/mm/slub.rst.
5650 slub_min_objects= [MM, SLUB]
5651 The minimum number of objects per slab. SLUB will
5652 increase the slab order up to slub_max_order to
5653 generate a sufficiently large slab able to contain
5654 the number of objects indicated. The higher the number
5655 of objects the smaller the overhead of tracking slabs
5656 and the less frequently locks need to be acquired.
5657 For more information see Documentation/mm/slub.rst.
5659 slub_min_order= [MM, SLUB]
5660 Determines the minimum page order for slabs. Must be
5661 lower than slub_max_order.
5662 For more information see Documentation/mm/slub.rst.
5664 slub_merge [MM, SLUB]
5665 Same with slab_merge.
5667 slub_nomerge [MM, SLUB]
5668 Same with slab_nomerge. This is supported for legacy.
5669 See slab_nomerge for more information.
5672 Format: <io1>[,<io2>[,...,<io8>]]
5674 smp.csd_lock_timeout= [KNL]
5675 Specify the period of time in milliseconds
5676 that smp_call_function() and friends will wait
5677 for a CPU to release the CSD lock. This is
5678 useful when diagnosing bugs involving CPUs
5679 disabling interrupts for extended periods
5680 of time. Defaults to 5,000 milliseconds, and
5681 setting a value of zero disables this feature.
5682 This feature may be more efficiently disabled
5683 using the csdlock_debug- kernel parameter.
5685 smsc-ircc2.nopnp [HW] Don't use PNP to discover SMC devices
5686 smsc-ircc2.ircc_cfg= [HW] Device configuration I/O port
5687 smsc-ircc2.ircc_sir= [HW] SIR base I/O port
5688 smsc-ircc2.ircc_fir= [HW] FIR base I/O port
5689 smsc-ircc2.ircc_irq= [HW] IRQ line
5690 smsc-ircc2.ircc_dma= [HW] DMA channel
5691 smsc-ircc2.ircc_transceiver= [HW] Transceiver type:
5692 0: Toshiba Satellite 1800 (GP data pin select)
5693 1: Fast pin select (default)
5696 smt= [KNL,S390] Set the maximum number of threads (logical
5697 CPUs) to use per physical CPU on systems capable of
5698 symmetric multithreading (SMT). Will be capped to the
5699 actual hardware limit.
5701 Default: -1 (no limit)
5704 [KNL] Should the soft-lockup detector generate panics.
5707 A value of 1 instructs the soft-lockup detector
5708 to panic the machine when a soft-lockup occurs. It is
5709 also controlled by the kernel.softlockup_panic sysctl
5710 and CONFIG_BOOTPARAM_SOFTLOCKUP_PANIC, which is the
5711 respective build-time switch to that functionality.
5713 softlockup_all_cpu_backtrace=
5714 [KNL] Should the soft-lockup detector generate
5715 backtraces on all cpus.
5718 sonypi.*= [HW] Sony Programmable I/O Control Device driver
5719 See Documentation/admin-guide/laptops/sonypi.rst
5721 spectre_v2= [X86] Control mitigation of Spectre variant 2
5722 (indirect branch speculation) vulnerability.
5723 The default operation protects the kernel from
5726 on - unconditionally enable, implies
5728 off - unconditionally disable, implies
5730 auto - kernel detects whether your CPU model is
5733 Selecting 'on' will, and 'auto' may, choose a
5734 mitigation method at run time according to the
5735 CPU, the available microcode, the setting of the
5736 CONFIG_RETPOLINE configuration option, and the
5737 compiler with which the kernel was built.
5739 Selecting 'on' will also enable the mitigation
5740 against user space to user space task attacks.
5742 Selecting 'off' will disable both the kernel and
5743 the user space protections.
5745 Specific mitigations can also be selected manually:
5747 retpoline - replace indirect branches
5748 retpoline,generic - Retpolines
5749 retpoline,lfence - LFENCE; indirect branch
5750 retpoline,amd - alias for retpoline,lfence
5751 eibrs - Enhanced/Auto IBRS
5752 eibrs,retpoline - Enhanced/Auto IBRS + Retpolines
5753 eibrs,lfence - Enhanced/Auto IBRS + LFENCE
5754 ibrs - use IBRS to protect kernel
5756 Not specifying this option is equivalent to
5760 [X86] Control mitigation of Spectre variant 2
5761 (indirect branch speculation) vulnerability between
5764 on - Unconditionally enable mitigations. Is
5765 enforced by spectre_v2=on
5767 off - Unconditionally disable mitigations. Is
5768 enforced by spectre_v2=off
5770 prctl - Indirect branch speculation is enabled,
5771 but mitigation can be enabled via prctl
5772 per thread. The mitigation control state
5773 is inherited on fork.
5776 - Like "prctl" above, but only STIBP is
5777 controlled per thread. IBPB is issued
5778 always when switching between different user
5782 - Same as "prctl" above, but all seccomp
5783 threads will enable the mitigation unless
5784 they explicitly opt out.
5787 - Like "seccomp" above, but only STIBP is
5788 controlled per thread. IBPB is issued
5789 always when switching between different
5790 user space processes.
5792 auto - Kernel selects the mitigation depending on
5793 the available CPU features and vulnerability.
5795 Default mitigation: "prctl"
5797 Not specifying this option is equivalent to
5798 spectre_v2_user=auto.
5800 spec_store_bypass_disable=
5801 [HW] Control Speculative Store Bypass (SSB) Disable mitigation
5802 (Speculative Store Bypass vulnerability)
5804 Certain CPUs are vulnerable to an exploit against a
5805 a common industry wide performance optimization known
5806 as "Speculative Store Bypass" in which recent stores
5807 to the same memory location may not be observed by
5808 later loads during speculative execution. The idea
5809 is that such stores are unlikely and that they can
5810 be detected prior to instruction retirement at the
5811 end of a particular speculation execution window.
5813 In vulnerable processors, the speculatively forwarded
5814 store can be used in a cache side channel attack, for
5815 example to read memory to which the attacker does not
5816 directly have access (e.g. inside sandboxed code).
5818 This parameter controls whether the Speculative Store
5819 Bypass optimization is used.
5821 On x86 the options are:
5823 on - Unconditionally disable Speculative Store Bypass
5824 off - Unconditionally enable Speculative Store Bypass
5825 auto - Kernel detects whether the CPU model contains an
5826 implementation of Speculative Store Bypass and
5827 picks the most appropriate mitigation. If the
5828 CPU is not vulnerable, "off" is selected. If the
5829 CPU is vulnerable the default mitigation is
5830 architecture and Kconfig dependent. See below.
5831 prctl - Control Speculative Store Bypass per thread
5832 via prctl. Speculative Store Bypass is enabled
5833 for a process by default. The state of the control
5834 is inherited on fork.
5835 seccomp - Same as "prctl" above, but all seccomp threads
5836 will disable SSB unless they explicitly opt out.
5838 Default mitigations:
5841 On powerpc the options are:
5843 on,auto - On Power8 and Power9 insert a store-forwarding
5844 barrier on kernel entry and exit. On Power7
5845 perform a software flush on kernel entry and
5849 Not specifying this option is equivalent to
5850 spec_store_bypass_disable=auto.
5852 spia_io_base= [HW,MTD]
5858 [X86] Enable split lock detection or bus lock detection
5860 When enabled (and if hardware support is present), atomic
5861 instructions that access data across cache line
5862 boundaries will result in an alignment check exception
5863 for split lock detection or a debug exception for
5868 warn - the kernel will emit rate-limited warnings
5869 about applications triggering the #AC
5870 exception or the #DB exception. This mode is
5871 the default on CPUs that support split lock
5872 detection or bus lock detection. Default
5873 behavior is by #AC if both features are
5874 enabled in hardware.
5876 fatal - the kernel will send SIGBUS to applications
5877 that trigger the #AC exception or the #DB
5878 exception. Default behavior is by #AC if
5879 both features are enabled in hardware.
5882 Set system wide rate limit to N bus locks
5883 per second for bus lock detection.
5886 N/A for split lock detection.
5889 If an #AC exception is hit in the kernel or in
5890 firmware (i.e. not while executing in user mode)
5891 the kernel will oops in either "warn" or "fatal"
5894 #DB exception for bus lock is triggered only when
5898 Control the Special Register Buffer Data Sampling
5901 Certain CPUs are vulnerable to an MDS-like
5902 exploit which can leak bits from the random
5905 By default, this issue is mitigated by
5906 microcode. However, the microcode fix can cause
5907 the RDRAND and RDSEED instructions to become
5908 much slower. Among other effects, this will
5909 result in reduced throughput from /dev/urandom.
5911 The microcode mitigation can be disabled with
5912 the following option:
5914 off: Disable mitigation and remove
5915 performance impact to RDRAND and RDSEED
5917 srcutree.big_cpu_lim [KNL]
5918 Specifies the number of CPUs constituting a
5919 large system, such that srcu_struct structures
5920 should immediately allocate an srcu_node array.
5921 This kernel-boot parameter defaults to 128,
5922 but takes effect only when the low-order four
5923 bits of srcutree.convert_to_big is equal to 3
5926 srcutree.convert_to_big [KNL]
5927 Specifies under what conditions an SRCU tree
5928 srcu_struct structure will be converted to big
5929 form, that is, with an rcu_node tree:
5932 1: At init_srcu_struct() time.
5933 2: When rcutorture decides to.
5934 3: Decide at boot time (default).
5935 0x1X: Above plus if high contention.
5937 Either way, the srcu_node tree will be sized based
5938 on the actual runtime number of CPUs (nr_cpu_ids)
5939 instead of the compile-time CONFIG_NR_CPUS.
5941 srcutree.counter_wrap_check [KNL]
5942 Specifies how frequently to check for
5943 grace-period sequence counter wrap for the
5944 srcu_data structure's ->srcu_gp_seq_needed field.
5945 The greater the number of bits set in this kernel
5946 parameter, the less frequently counter wrap will
5947 be checked for. Note that the bottom two bits
5950 srcutree.exp_holdoff [KNL]
5951 Specifies how many nanoseconds must elapse
5952 since the end of the last SRCU grace period for
5953 a given srcu_struct until the next normal SRCU
5954 grace period will be considered for automatic
5955 expediting. Set to zero to disable automatic
5958 srcutree.srcu_max_nodelay [KNL]
5959 Specifies the number of no-delay instances
5960 per jiffy for which the SRCU grace period
5961 worker thread will be rescheduled with zero
5962 delay. Beyond this limit, worker thread will
5963 be rescheduled with a sleep delay of one jiffy.
5965 srcutree.srcu_max_nodelay_phase [KNL]
5966 Specifies the per-grace-period phase, number of
5967 non-sleeping polls of readers. Beyond this limit,
5968 grace period worker thread will be rescheduled
5969 with a sleep delay of one jiffy, between each
5970 rescan of the readers, for a grace period phase.
5972 srcutree.srcu_retry_check_delay [KNL]
5973 Specifies number of microseconds of non-sleeping
5974 delay between each non-sleeping poll of readers.
5976 srcutree.small_contention_lim [KNL]
5977 Specifies the number of update-side contention
5978 events per jiffy will be tolerated before
5979 initiating a conversion of an srcu_struct
5980 structure to big form. Note that the value of
5981 srcutree.convert_to_big must have the 0x10 bit
5982 set for contention-based conversions to occur.
5985 Speculative Store Bypass Disable control
5987 On CPUs that are vulnerable to the Speculative
5988 Store Bypass vulnerability and offer a
5989 firmware based mitigation, this parameter
5990 indicates how the mitigation should be used:
5992 force-on: Unconditionally enable mitigation for
5993 for both kernel and userspace
5994 force-off: Unconditionally disable mitigation for
5995 for both kernel and userspace
5996 kernel: Always enable mitigation in the
5997 kernel, and offer a prctl interface
5998 to allow userspace to register its
5999 interest in being mitigated too.
6001 stack_guard_gap= [MM]
6002 override the default stack gap protection. The value
6003 is in page units and it defines how many pages prior
6004 to (for stacks growing down) resp. after (for stacks
6005 growing up) the main stack are reserved for no other
6006 mapping. Default value is 256 pages.
6008 stack_depot_disable= [KNL]
6009 Setting this to true through kernel command line will
6010 disable the stack depot thereby saving the static memory
6011 consumed by the stack hash table. By default this is set
6015 Enabled the stack tracer on boot up.
6017 stacktrace_filter=[function-list]
6018 [FTRACE] Limit the functions that the stack tracer
6019 will trace at boot up. function-list is a comma-separated
6020 list of functions. This list can be changed at run
6021 time by the stack_trace_filter file in the debugfs
6022 tracing directory. Note, this enables stack tracing
6023 and the stacktrace above is not needed.
6027 Set the STI (builtin display/keyboard on the HP-PARISC
6028 machines) console (graphic card) which should be used
6029 as the initial boot-console.
6030 See also comment in drivers/video/console/sticore.c.
6033 See comment in drivers/video/console/sticore.c.
6036 Format: bpp:<bpp1>[:<bpp2>[:<bpp3>...]]
6041 Enable or disable strict sigaltstack size checks
6042 against the required signal frame size which
6043 depends on the supported FPU features. This can
6044 be used to filter out binaries which have
6045 not yet been made aware of AT_MINSIGSTKSZ.
6048 Limits the number of kernel HPT entries in the hash
6049 page table to increase the rate of hash page table
6050 faults on kernel addresses.
6053 Limits the number of kernel SLB entries, and flushes
6054 them frequently to increase the rate of SLB faults
6055 on kernel addresses.
6057 sunrpc.min_resvport=
6058 sunrpc.max_resvport=
6060 SunRPC servers often require that client requests
6061 originate from a privileged port (i.e. a port in the
6062 range 0 < portnr < 1024).
6063 An administrator who wishes to reserve some of these
6064 ports for other uses may adjust the range that the
6065 kernel's sunrpc client considers to be privileged
6066 using these two parameters to set the minimum and
6067 maximum port values.
6069 sunrpc.svc_rpc_per_connection_limit=
6071 Limit the number of requests that the server will
6072 process in parallel from a single connection.
6073 The default value is 0 (no limit).
6077 Control how the NFS server code allocates CPUs to
6078 service thread pools. Depending on how many NICs
6079 you have and where their interrupts are bound, this
6080 option will affect which CPUs will do NFS serving.
6081 Note: this parameter cannot be changed while the
6082 NFS server is running.
6084 auto the server chooses an appropriate mode
6085 automatically using heuristics
6086 global a single global pool contains all CPUs
6087 percpu one pool for each CPU
6088 pernode one pool for each NUMA node (equivalent
6089 to global on non-NUMA machines)
6091 sunrpc.tcp_slot_table_entries=
6092 sunrpc.udp_slot_table_entries=
6094 Sets the upper limit on the number of simultaneous
6095 RPC calls that can be sent from the client to a
6096 server. Increasing these values may allow you to
6097 improve throughput, but will also increase the
6098 amount of memory reserved for use by the client.
6100 suspend.pm_test_delay=
6102 Sets the number of seconds to remain in a suspend test
6103 mode before resuming the system (see
6104 /sys/power/pm_test). Only available when CONFIG_PM_DEBUG
6105 is set. Default value is 5.
6108 Format: { on | off | y | n | 1 | 0 }
6109 This parameter controls use of the Protected
6110 Execution Facility on pSeries.
6112 swiotlb= [ARM,IA-64,PPC,MIPS,X86]
6113 Format: { <int> [,<int>] | force | noforce }
6114 <int> -- Number of I/O TLB slabs
6115 <int> -- Second integer after comma. Number of swiotlb
6116 areas with their own lock. Will be rounded up
6118 force -- force using of bounce buffers even if they
6119 wouldn't be automatically used by the kernel
6120 noforce -- Never use bounce buffers (for debugging)
6125 Set a sysctl parameter, right before loading the init
6126 process, as if the value was written to the respective
6127 /proc/sys/... file. Both '.' and '/' are recognized as
6128 separators. Unrecognized parameters and invalid values
6129 are reported in the kernel log. Sysctls registered
6130 later by a loaded module cannot be set this way.
6131 Example: sysctl.vm.swappiness=40
6133 sysfs.deprecated=0|1 [KNL]
6134 Enable/disable old style sysfs layout for old udev
6135 on older distributions. When this option is enabled
6136 very new udev will not work anymore. When this option
6137 is disabled (or CONFIG_SYSFS_DEPRECATED not compiled)
6138 in older udev will not work anymore.
6139 Default depends on CONFIG_SYSFS_DEPRECATED_V2 set in
6140 the kernel configuration.
6142 sysrq_always_enabled
6144 Ignore sysrq setting - this boot parameter will
6145 neutralize any effect of /proc/sys/kernel/sysrq.
6146 Useful for debugging.
6148 tcpmhash_entries= [KNL,NET]
6149 Set the number of tcp_metrics_hash slots.
6150 Default value is 8192 or 16384 depending on total
6151 ram pages. This is used to specify the TCP metrics
6152 cache size. See Documentation/networking/ip-sysctl.rst
6153 "tcp_no_metrics_save" section for more details.
6157 test_suspend= [SUSPEND]
6158 Format: { "mem" | "standby" | "freeze" }[,N]
6159 Specify "mem" (for Suspend-to-RAM) or "standby" (for
6160 standby suspend) or "freeze" (for suspend type freeze)
6161 as the system sleep state during system startup with
6162 the optional capability to repeat N number of times.
6163 The system is woken from this state using a
6164 wakeup-capable RTC alarm.
6166 thash_entries= [KNL,NET]
6167 Set number of hash buckets for TCP connection
6169 thermal.act= [HW,ACPI]
6170 -1: disable all active trip points in all thermal zones
6171 <degrees C>: override all lowest active trip points
6173 thermal.crt= [HW,ACPI]
6174 -1: disable all critical trip points in all thermal zones
6175 <degrees C>: override all critical trip points
6177 thermal.nocrt= [HW,ACPI]
6178 Set to disable actions on ACPI thermal zone
6179 critical and hot trip points.
6181 thermal.off= [HW,ACPI]
6182 1: disable ACPI thermal control
6184 thermal.psv= [HW,ACPI]
6185 -1: disable all passive trip points
6186 <degrees C>: override all passive trip points to this
6189 thermal.tzp= [HW,ACPI]
6190 Specify global default ACPI thermal zone polling rate
6191 <deci-seconds>: poll all this frequency
6192 0: no polling (default)
6195 Force threading of all interrupt handlers except those
6196 marked explicitly IRQF_NO_THREAD.
6200 Specify if the kernel should make use of the cpu
6201 topology information if the hardware supports this.
6202 The scheduler will make use of this information and
6203 e.g. base its process migration decisions on it.
6206 topology_updates= [KNL, PPC, NUMA]
6208 Specify if the kernel should ignore (off)
6209 topology updates sent by the hypervisor to this
6212 torture.disable_onoff_at_boot= [KNL]
6213 Prevent the CPU-hotplug component of torturing
6214 until after init has spawned.
6216 torture.ftrace_dump_at_shutdown= [KNL]
6217 Dump the ftrace buffer at torture-test shutdown,
6218 even if there were no errors. This can be a
6219 very costly operation when many torture tests
6220 are running concurrently, especially on systems
6221 with rotating-rust storage.
6223 torture.verbose_sleep_frequency= [KNL]
6224 Specifies how many verbose printk()s should be
6225 emitted between each sleep. The default of zero
6226 disables verbose-printk() sleeping.
6228 torture.verbose_sleep_duration= [KNL]
6229 Duration of each verbose-printk() sleep in jiffies.
6233 tpm_suspend_pcr=[HW,TPM]
6234 Format: integer pcr id
6235 Specify that at suspend time, the tpm driver
6236 should extend the specified pcr with zeros,
6237 as a workaround for some chips which fail to
6238 flush the last written pcr on TPM_SaveState.
6239 This will guarantee that all the other pcrs
6243 Have the tracepoints sent to printk as well as the
6244 tracing ring buffer. This is useful for early boot up
6245 where the system hangs or reboots and does not give the
6246 option for reading the tracing buffer or performing a
6247 ftrace_dump_on_oops.
6249 To turn off having tracepoints sent to printk,
6250 echo 0 > /proc/sys/kernel/tracepoint_printk
6251 Note, echoing 1 into this file without the
6252 tracepoint_printk kernel cmdline option has no effect.
6254 The tp_printk_stop_on_boot (see below) can also be used
6255 to stop the printing of events to console at
6260 Having tracepoints sent to printk() and activating high
6261 frequency tracepoints such as irq or sched, can cause
6262 the system to live lock.
6264 tp_printk_stop_on_boot [FTRACE]
6265 When tp_printk (above) is set, it can cause a lot of noise
6266 on the console. It may be useful to only include the
6267 printing of events during boot up, as user space may
6268 make the system inoperable.
6270 This command line option will stop the printing of events
6271 to console at the late_initcall_sync() time frame.
6273 trace_buf_size=nn[KMG]
6274 [FTRACE] will set tracing buffer size on each cpu.
6276 trace_clock= [FTRACE] Set the clock used for tracing events
6278 local - Use the per CPU time stamp counter
6279 (converted into nanoseconds). Fast, but
6280 depending on the architecture, may not be
6281 in sync between CPUs.
6282 global - Event time stamps are synchronize across
6283 CPUs. May be slower than the local clock,
6284 but better for some race conditions.
6285 counter - Simple counting of events (1, 2, ..)
6286 note, some counts may be skipped due to the
6287 infrastructure grabbing the clock more than
6289 uptime - Use jiffies as the time stamp.
6290 perf - Use the same clock that perf uses.
6291 mono - Use ktime_get_mono_fast_ns() for time stamps.
6292 mono_raw - Use ktime_get_raw_fast_ns() for time
6294 boot - Use ktime_get_boot_fast_ns() for time stamps.
6295 Architectures may add more clocks. See
6296 Documentation/trace/ftrace.rst for more details.
6298 trace_event=[event-list]
6299 [FTRACE] Set and start specified trace events in order
6300 to facilitate early boot debugging. The event-list is a
6301 comma-separated list of trace events to enable. See
6302 also Documentation/trace/events.rst
6304 trace_instance=[instance-info]
6305 [FTRACE] Create a ring buffer instance early in boot up.
6306 This will be listed in:
6308 /sys/kernel/tracing/instances
6310 Events can be enabled at the time the instance is created
6313 trace_instance=<name>,<system1>:<event1>,<system2>:<event2>
6315 Note, the "<system*>:" portion is optional if the event is
6318 trace_instance=foo,sched:sched_switch,irq_handler_entry,initcall
6320 will enable the "sched_switch" event (note, the "sched:" is optional, and
6321 the same thing would happen if it was left off). The irq_handler_entry
6322 event, and all events under the "initcall" system.
6324 trace_options=[option-list]
6325 [FTRACE] Enable or disable tracer options at boot.
6326 The option-list is a comma delimited list of options
6327 that can be enabled or disabled just as if you were
6328 to echo the option name into
6330 /sys/kernel/tracing/trace_options
6332 For example, to enable stacktrace option (to dump the
6333 stack trace of each event), add to the command line:
6335 trace_options=stacktrace
6337 See also Documentation/trace/ftrace.rst "trace options"
6340 trace_trigger=[trigger-list]
6341 [FTRACE] Add a event trigger on specific events.
6342 Set a trigger on top of a specific event, with an optional
6345 The format is is "trace_trigger=<event>.<trigger>[ if <filter>],..."
6346 Where more than one trigger may be specified that are comma deliminated.
6350 trace_trigger="sched_switch.stacktrace if prev_state == 2"
6352 The above will enable the "stacktrace" trigger on the "sched_switch"
6353 event but only trigger it if the "prev_state" of the "sched_switch"
6354 event is "2" (TASK_UNINTERUPTIBLE).
6356 See also "Event triggers" in Documentation/trace/events.rst
6360 [FTRACE] enable this option to disable tracing when a
6361 warning is hit. This turns off "tracing_on". Tracing can
6362 be enabled again by echoing '1' into the "tracing_on"
6363 file located in /sys/kernel/tracing/
6365 This option is useful, as it disables the trace before
6366 the WARNING dump is called, which prevents the trace to
6367 be filled with content caused by the warning output.
6369 This option can also be set at run time via the sysctl
6370 option: kernel/traceoff_on_warning
6372 transparent_hugepage=
6374 Format: [always|madvise|never]
6375 Can be used to control the default behavior of the system
6376 with respect to transparent hugepages.
6377 See Documentation/admin-guide/mm/transhuge.rst
6380 trusted.source= [KEYS]
6382 This parameter identifies the trust source as a backend
6383 for trusted keys implementation. Supported trust
6388 If not specified then it defaults to iterating through
6389 the trust source list starting with TPM and assigns the
6390 first trust source as a backend which is initialized
6391 successfully during iteration.
6395 The RNG used to generate key material for trusted keys.
6398 - the same value as trusted.source: "tpm" or "tee"
6400 If not specified, "default" is used. In this case,
6401 the RNG's choice is left to each individual trust source.
6403 tsc= Disable clocksource stability checks for TSC.
6405 [x86] reliable: mark tsc clocksource as reliable, this
6406 disables clocksource verification at runtime, as well
6407 as the stability checks done at bootup. Used to enable
6408 high-resolution timer mode on older hardware, and in
6409 virtualized environment.
6410 [x86] noirqtime: Do not use TSC to do irq accounting.
6411 Used to run time disable IRQ_TIME_ACCOUNTING on any
6412 platforms where RDTSC is slow and this accounting
6414 [x86] unstable: mark the TSC clocksource as unstable, this
6415 marks the TSC unconditionally unstable at bootup and
6416 avoids any further wobbles once the TSC watchdog notices.
6417 [x86] nowatchdog: disable clocksource watchdog. Used
6418 in situations with strict latency requirements (where
6419 interruptions from clocksource watchdog are not
6421 [x86] recalibrate: force recalibration against a HW timer
6422 (HPET or PM timer) on systems whose TSC frequency was
6423 obtained from HW or FW using either an MSR or CPUID(0x15).
6424 Warn if the difference is more than 500 ppm.
6425 [x86] watchdog: Use TSC as the watchdog clocksource with
6426 which to check other HW timers (HPET or PM timer), but
6427 only on systems where TSC has been deemed trustworthy.
6428 This will be suppressed by an earlier tsc=nowatchdog and
6429 can be overridden by a later tsc=nowatchdog. A console
6430 message will flag any such suppression or overriding.
6432 tsc_early_khz= [X86] Skip early TSC calibration and use the given
6433 value instead. Useful when the early TSC frequency discovery
6434 procedure is not reliable, such as on overclocked systems
6435 with CPUID.16h support and partial CPUID.15h support.
6436 Format: <unsigned int>
6438 tsx= [X86] Control Transactional Synchronization
6439 Extensions (TSX) feature in Intel processors that
6440 support TSX control.
6442 This parameter controls the TSX feature. The options are:
6444 on - Enable TSX on the system. Although there are
6445 mitigations for all known security vulnerabilities,
6446 TSX has been known to be an accelerator for
6447 several previous speculation-related CVEs, and
6448 so there may be unknown security risks associated
6449 with leaving it enabled.
6451 off - Disable TSX on the system. (Note that this
6452 option takes effect only on newer CPUs which are
6453 not vulnerable to MDS, i.e., have
6454 MSR_IA32_ARCH_CAPABILITIES.MDS_NO=1 and which get
6455 the new IA32_TSX_CTRL MSR through a microcode
6456 update. This new MSR allows for the reliable
6457 deactivation of the TSX functionality.)
6459 auto - Disable TSX if X86_BUG_TAA is present,
6460 otherwise enable TSX on the system.
6462 Not specifying this option is equivalent to tsx=off.
6464 See Documentation/admin-guide/hw-vuln/tsx_async_abort.rst
6467 tsx_async_abort= [X86,INTEL] Control mitigation for the TSX Async
6468 Abort (TAA) vulnerability.
6470 Similar to Micro-architectural Data Sampling (MDS)
6471 certain CPUs that support Transactional
6472 Synchronization Extensions (TSX) are vulnerable to an
6473 exploit against CPU internal buffers which can forward
6474 information to a disclosure gadget under certain
6477 In vulnerable processors, the speculatively forwarded
6478 data can be used in a cache side channel attack, to
6479 access data to which the attacker does not have direct
6482 This parameter controls the TAA mitigation. The
6485 full - Enable TAA mitigation on vulnerable CPUs
6488 full,nosmt - Enable TAA mitigation and disable SMT on
6489 vulnerable CPUs. If TSX is disabled, SMT
6490 is not disabled because CPU is not
6491 vulnerable to cross-thread TAA attacks.
6492 off - Unconditionally disable TAA mitigation
6494 On MDS-affected machines, tsx_async_abort=off can be
6495 prevented by an active MDS mitigation as both vulnerabilities
6496 are mitigated with the same mechanism so in order to disable
6497 this mitigation, you need to specify mds=off too.
6499 Not specifying this option is equivalent to
6500 tsx_async_abort=full. On CPUs which are MDS affected
6501 and deploy MDS mitigation, TAA mitigation is not
6502 required and doesn't provide any additional
6506 Documentation/admin-guide/hw-vuln/tsx_async_abort.rst
6508 turbografx.map[2|3]= [HW,JOY]
6509 TurboGraFX parallel port interface
6511 <port#>,<js1>,<js2>,<js3>,<js4>,<js5>,<js6>,<js7>
6512 See also Documentation/input/devices/joystick-parport.rst
6514 udbg-immortal [PPC] When debugging early kernel crashes that
6515 happen after console_init() and before a proper
6516 console driver takes over, this boot options might
6517 help "seeing" what's going on.
6519 uhash_entries= [KNL,NET]
6520 Set number of hash buckets for UDP/UDP-Lite connections
6523 [USB] Ignore overcurrent events (default N).
6524 Some badly-designed motherboards generate lots of
6525 bogus events, for ports that aren't wired to
6526 anything. Set this parameter to avoid log spamming.
6527 Note that genuine overcurrent events won't be
6531 [X86] Cause panic on unknown NMI.
6533 usbcore.authorized_default=
6534 [USB] Default USB device authorization:
6535 (default -1 = authorized except for wireless USB,
6536 0 = not authorized, 1 = authorized, 2 = authorized
6537 if device connected to internal port)
6539 usbcore.autosuspend=
6540 [USB] The autosuspend time delay (in seconds) used
6541 for newly-detected USB devices (default 2). This
6542 is the time required before an idle device will be
6543 autosuspended. Devices for which the delay is set
6544 to a negative value won't be autosuspended at all.
6546 usbcore.usbfs_snoop=
6547 [USB] Set to log all usbfs traffic (default 0 = off).
6549 usbcore.usbfs_snoop_max=
6550 [USB] Maximum number of bytes to snoop in each URB
6553 usbcore.blinkenlights=
6554 [USB] Set to cycle leds on hubs (default 0 = off).
6556 usbcore.old_scheme_first=
6557 [USB] Start with the old device initialization
6558 scheme (default 0 = off).
6560 usbcore.usbfs_memory_mb=
6561 [USB] Memory limit (in MB) for buffers allocated by
6562 usbfs (default = 16, 0 = max = 2047).
6564 usbcore.use_both_schemes=
6565 [USB] Try the other device initialization scheme
6566 if the first one fails (default 1 = enabled).
6568 usbcore.initial_descriptor_timeout=
6569 [USB] Specifies timeout for the initial 64-byte
6570 USB_REQ_GET_DESCRIPTOR request in milliseconds
6571 (default 5000 = 5.0 seconds).
6573 usbcore.nousb [USB] Disable the USB subsystem
6576 [USB] A list of quirk entries to augment the built-in
6577 usb core quirk list. List entries are separated by
6578 commas. Each entry has the form
6579 VendorID:ProductID:Flags. The IDs are 4-digit hex
6580 numbers and Flags is a set of letters. Each letter
6581 will change the built-in quirk; setting it if it is
6582 clear and clearing it if it is set. The letters have
6583 the following meanings:
6584 a = USB_QUIRK_STRING_FETCH_255 (string
6585 descriptors must not be fetched using
6587 b = USB_QUIRK_RESET_RESUME (device can't resume
6588 correctly so reset it instead);
6589 c = USB_QUIRK_NO_SET_INTF (device can't handle
6590 Set-Interface requests);
6591 d = USB_QUIRK_CONFIG_INTF_STRINGS (device can't
6592 handle its Configuration or Interface
6594 e = USB_QUIRK_RESET (device can't be reset
6595 (e.g morph devices), don't use reset);
6596 f = USB_QUIRK_HONOR_BNUMINTERFACES (device has
6597 more interface descriptions than the
6598 bNumInterfaces count, and can't handle
6599 talking to these interfaces);
6600 g = USB_QUIRK_DELAY_INIT (device needs a pause
6601 during initialization, after we read
6602 the device descriptor);
6603 h = USB_QUIRK_LINEAR_UFRAME_INTR_BINTERVAL (For
6604 high speed and super speed interrupt
6605 endpoints, the USB 2.0 and USB 3.0 spec
6606 require the interval in microframes (1
6607 microframe = 125 microseconds) to be
6608 calculated as interval = 2 ^
6610 Devices with this quirk report their
6611 bInterval as the result of this
6612 calculation instead of the exponent
6613 variable used in the calculation);
6614 i = USB_QUIRK_DEVICE_QUALIFIER (device can't
6615 handle device_qualifier descriptor
6617 j = USB_QUIRK_IGNORE_REMOTE_WAKEUP (device
6618 generates spurious wakeup, ignore
6619 remote wakeup capability);
6620 k = USB_QUIRK_NO_LPM (device can't handle Link
6622 l = USB_QUIRK_LINEAR_FRAME_INTR_BINTERVAL
6623 (Device reports its bInterval as linear
6624 frames instead of the USB 2.0
6626 m = USB_QUIRK_DISCONNECT_SUSPEND (Device needs
6627 to be disconnected before suspend to
6628 prevent spurious wakeup);
6629 n = USB_QUIRK_DELAY_CTRL_MSG (Device needs a
6630 pause after every control message);
6631 o = USB_QUIRK_HUB_SLOW_RESET (Hub needs extra
6632 delay after resetting its port);
6633 Example: quirks=0781:5580:bk,0a5c:5834:gij
6636 [USBHID] The interval which mice are to be polled at.
6639 [USBHID] The interval which joysticks are to be polled at.
6642 [USBHID] The interval which keyboards are to be polled at.
6644 usb-storage.delay_use=
6645 [UMS] The delay in seconds before a new device is
6646 scanned for Logical Units (default 1).
6649 [UMS] A list of quirks entries to supplement or
6650 override the built-in unusual_devs list. List
6651 entries are separated by commas. Each entry has
6652 the form VID:PID:Flags where VID and PID are Vendor
6653 and Product ID values (4-digit hex numbers) and
6654 Flags is a set of characters, each corresponding
6655 to a common usb-storage quirk flag as follows:
6656 a = SANE_SENSE (collect more than 18 bytes
6657 of sense data, not on uas);
6658 b = BAD_SENSE (don't collect more than 18
6659 bytes of sense data, not on uas);
6660 c = FIX_CAPACITY (decrease the reported
6661 device capacity by one sector);
6662 d = NO_READ_DISC_INFO (don't use
6663 READ_DISC_INFO command, not on uas);
6664 e = NO_READ_CAPACITY_16 (don't use
6665 READ_CAPACITY_16 command);
6666 f = NO_REPORT_OPCODES (don't use report opcodes
6668 g = MAX_SECTORS_240 (don't transfer more than
6669 240 sectors at a time, uas only);
6670 h = CAPACITY_HEURISTICS (decrease the
6671 reported device capacity by one
6672 sector if the number is odd);
6673 i = IGNORE_DEVICE (don't bind to this
6675 j = NO_REPORT_LUNS (don't use report luns
6677 k = NO_SAME (do not use WRITE_SAME, uas only)
6678 l = NOT_LOCKABLE (don't try to lock and
6679 unlock ejectable media, not on uas);
6680 m = MAX_SECTORS_64 (don't transfer more
6681 than 64 sectors = 32 KB at a time,
6683 n = INITIAL_READ10 (force a retry of the
6684 initial READ(10) command, not on uas);
6685 o = CAPACITY_OK (accept the capacity
6686 reported by the device, not on uas);
6687 p = WRITE_CACHE (the device cache is ON
6688 by default, not on uas);
6689 r = IGNORE_RESIDUE (the device reports
6690 bogus residue values, not on uas);
6691 s = SINGLE_LUN (the device has only one
6693 t = NO_ATA_1X (don't allow ATA(12) and ATA(16)
6694 commands, uas only);
6695 u = IGNORE_UAS (don't bind to the uas driver);
6696 w = NO_WP_DETECT (don't test whether the
6697 medium is write-protected).
6698 y = ALWAYS_SYNC (issue a SYNCHRONIZE_CACHE
6699 even if the device claims no cache,
6701 Example: quirks=0419:aaf5:rl,0421:0433:rc
6703 user_debug= [KNL,ARM]
6705 See arch/arm/Kconfig.debug help text.
6706 1 - undefined instruction events
6708 4 - invalid data aborts
6711 Example: user_debug=31
6714 [X86] Flags controlling user PTE allocations.
6716 nohigh = do not allocate PTE pages in
6717 HIGHMEM regardless of setting
6720 vdso= [X86,SH,SPARC]
6721 On X86_32, this is an alias for vdso32=. Otherwise:
6723 vdso=1: enable VDSO (the default)
6724 vdso=0: disable VDSO mapping
6726 vdso32= [X86] Control the 32-bit vDSO
6727 vdso32=1: enable 32-bit VDSO
6728 vdso32=0 or vdso32=2: disable 32-bit VDSO
6730 See the help text for CONFIG_COMPAT_VDSO for more
6731 details. If CONFIG_COMPAT_VDSO is set, the default is
6732 vdso32=0; otherwise, the default is vdso32=1.
6734 For compatibility with older kernels, vdso32=2 is an
6737 Try vdso32=0 if you encounter an error that says:
6738 dl_main: Assertion `(void *) ph->p_vaddr == _rtld_local._dl_sysinfo_dso' failed!
6741 vector=percpu: enable percpu vector domain
6743 video= [FB] Frame buffer configuration
6744 See Documentation/fb/modedb.rst.
6746 video.brightness_switch_enabled= [ACPI]
6748 If set to 1, on receiving an ACPI notify event
6749 generated by hotkey, video driver will adjust brightness
6750 level and then send out the event to user space through
6751 the allocated input device. If set to 0, video driver
6752 will only send out the event without touching backlight
6757 [VMMIO] Memory mapped virtio (platform) device.
6759 <size>@<baseaddr>:<irq>[:<id>]
6761 <size> := size (can use standard suffixes
6763 <baseaddr> := physical base address
6764 <irq> := interrupt number (as passed to
6766 <id> := (optional) platform device id
6768 virtio_mmio.device=1K@0x100b0000:48:7
6770 Can be used multiple times for multiple devices.
6772 vga= [BOOT,X86-32] Select a particular video mode
6773 See Documentation/x86/boot.rst and
6774 Documentation/admin-guide/svga.rst.
6775 Use vga=ask for menu.
6776 This is actually a boot loader parameter; the value is
6777 passed to the kernel using a special protocol.
6779 vm_debug[=options] [KNL] Available with CONFIG_DEBUG_VM=y.
6780 May slow down system boot speed, especially when
6781 enabled on systems with a large amount of memory.
6782 All options are enabled by default, and this
6783 interface is meant to allow for selectively
6784 enabling or disabling specific virtual memory
6787 Available options are:
6788 P Enable page structure init time poisoning
6789 - Disable all of the above options
6791 vmalloc=nn[KMG] [KNL,BOOT] Forces the vmalloc area to have an exact
6792 size of <nn>. This can be used to increase the
6793 minimum size (128MB on x86). It can also be used to
6794 decrease the size and leave more room for directly
6797 vmcp_cma=nn[MG] [KNL,S390]
6798 Sets the memory size reserved for contiguous memory
6799 allocations for the vmcp device driver.
6801 vmhalt= [KNL,S390] Perform z/VM CP command after system halt.
6804 vmpanic= [KNL,S390] Perform z/VM CP command after kernel panic.
6807 vmpoff= [KNL,S390] Perform z/VM CP command after power off.
6811 Controls the behavior of vsyscalls (i.e. calls to
6812 fixed addresses of 0xffffffffff600x00 from legacy
6813 code). Most statically-linked binaries and older
6814 versions of glibc use these calls. Because these
6815 functions are at fixed addresses, they make nice
6816 targets for exploits that can control RIP.
6818 emulate Vsyscalls turn into traps and are emulated
6819 reasonably safely. The vsyscall page is
6822 xonly [default] Vsyscalls turn into traps and are
6823 emulated reasonably safely. The vsyscall
6824 page is not readable.
6826 none Vsyscalls don't work at all. This makes
6827 them quite hard to use for exploits but
6828 might break your system.
6830 vt.color= [VT] Default text color.
6831 Format: 0xYX, X = foreground, Y = background.
6832 Default: 0x07 = light gray on black.
6834 vt.cur_default= [VT] Default cursor shape.
6835 Format: 0xCCBBAA, where AA, BB, and CC are the same as
6836 the parameters of the <Esc>[?A;B;Cc escape sequence;
6837 see VGA-softcursor.txt. Default: 2 = underline.
6839 vt.default_blu= [VT]
6840 Format: <blue0>,<blue1>,<blue2>,...,<blue15>
6841 Change the default blue palette of the console.
6842 This is a 16-member array composed of values
6845 vt.default_grn= [VT]
6846 Format: <green0>,<green1>,<green2>,...,<green15>
6847 Change the default green palette of the console.
6848 This is a 16-member array composed of values
6851 vt.default_red= [VT]
6852 Format: <red0>,<red1>,<red2>,...,<red15>
6853 Change the default red palette of the console.
6854 This is a 16-member array composed of values
6860 Set system-wide default UTF-8 mode for all tty's.
6861 Default is 1, i.e. UTF-8 mode is enabled for all
6862 newly opened terminals.
6864 vt.global_cursor_default=
6867 Set system-wide default for whether a cursor
6868 is shown on new VTs. Default is -1,
6869 i.e. cursors will be created by default unless
6870 overridden by individual drivers. 0 will hide
6871 cursors, 1 will display them.
6873 vt.italic= [VT] Default color for italic text; 0-15.
6876 vt.underline= [VT] Default color for underlined text; 0-15.
6879 watchdog timers [HW,WDT] For information on watchdog timers,
6880 see Documentation/watchdog/watchdog-parameters.rst
6881 or other driver-specific files in the
6882 Documentation/watchdog/ directory.
6886 Set the hard lockup detector stall duration
6887 threshold in seconds. The soft lockup detector
6888 threshold is set to twice the value. A value of 0
6889 disables both lockup detectors. Default is 10
6892 workqueue.watchdog_thresh=
6893 If CONFIG_WQ_WATCHDOG is configured, workqueue can
6894 warn stall conditions and dump internal state to
6895 help debugging. 0 disables workqueue stall
6896 detection; otherwise, it's the stall threshold
6897 duration in seconds. The default value is 30 and
6898 it can be updated at runtime by writing to the
6899 corresponding sysfs file.
6901 workqueue.disable_numa
6902 By default, all work items queued to unbound
6903 workqueues are affine to the NUMA nodes they're
6904 issued on, which results in better behavior in
6905 general. If NUMA affinity needs to be disabled for
6906 whatever reason, this option can be used. Note
6907 that this also can be controlled per-workqueue for
6908 workqueues visible under /sys/bus/workqueue/.
6910 workqueue.power_efficient
6911 Per-cpu workqueues are generally preferred because
6912 they show better performance thanks to cache
6913 locality; unfortunately, per-cpu workqueues tend to
6914 be more power hungry than unbound workqueues.
6916 Enabling this makes the per-cpu workqueues which
6917 were observed to contribute significantly to power
6918 consumption unbound, leading to measurably lower
6919 power usage at the cost of small performance
6922 The default value of this parameter is determined by
6923 the config option CONFIG_WQ_POWER_EFFICIENT_DEFAULT.
6925 workqueue.debug_force_rr_cpu
6926 Workqueue used to implicitly guarantee that work
6927 items queued without explicit CPU specified are put
6928 on the local CPU. This guarantee is no longer true
6929 and while local CPU is still preferred work items
6930 may be put on foreign CPUs. This debug option
6931 forces round-robin CPU selection to flush out
6932 usages which depend on the now broken guarantee.
6933 When enabled, memory and cache locality will be
6936 writecombine= [LOONGARCH] Control the MAT (Memory Access Type) of
6939 on - Enable writecombine, use WUC for ioremap_wc()
6940 off - Disable writecombine, use SUC for ioremap_wc()
6942 x2apic_phys [X86-64,APIC] Use x2apic physical mode instead of
6943 default x2apic cluster mode on platforms
6946 xen_512gb_limit [KNL,X86-64,XEN]
6947 Restricts the kernel running paravirtualized under Xen
6948 to use only up to 512 GB of RAM. The reason to do so is
6949 crash analysis tools and Xen tools for doing domain
6950 save/restore/migration must be enabled to handle larger
6953 xen_emul_unplug= [HW,X86,XEN]
6954 Unplug Xen emulated devices
6955 Format: [unplug0,][unplug1]
6956 ide-disks -- unplug primary master IDE devices
6957 aux-ide-disks -- unplug non-primary-master IDE devices
6958 nics -- unplug network devices
6959 all -- unplug all emulated devices (NICs and IDE disks)
6960 unnecessary -- unplugging emulated devices is
6961 unnecessary even if the host did not respond to
6963 never -- do not unplug even if version check succeeds
6965 xen_legacy_crash [X86,XEN]
6966 Crash from Xen panic notifier, without executing late
6967 panic() code such as dumping handler.
6969 xen_msr_safe= [X86,XEN]
6971 Select whether to always use non-faulting (safe) MSR
6972 access functions when running as Xen PV guest. The
6973 default value is controlled by CONFIG_XEN_PV_MSR_SAFE.
6975 xen_nopvspin [X86,XEN]
6976 Disables the qspinlock slowpath using Xen PV optimizations.
6977 This parameter is obsoleted by "nopvspin" parameter, which
6978 has equivalent effect for XEN platform.
6981 Disables the PV optimizations forcing the HVM guest to
6982 run as generic HVM guest with no PV drivers.
6983 This option is obsoleted by the "nopv" option, which
6984 has equivalent effect for XEN platform.
6986 xen_no_vector_callback
6987 [KNL,X86,XEN] Disable the vector callback for Xen
6988 event channel interrupts.
6990 xen_scrub_pages= [XEN]
6991 Boolean option to control scrubbing pages before giving them back
6992 to Xen, for use by other domains. Can be also changed at runtime
6993 with /sys/devices/system/xen_memory/xen_memory0/scrub_pages.
6994 Default value controlled with CONFIG_XEN_SCRUB_PAGES_DEFAULT.
6996 xen_timer_slop= [X86-64,XEN]
6997 Set the timer slop (in nanoseconds) for the virtual Xen
6998 timers (default is 100000). This adjusts the minimum
6999 delta of virtualized Xen timers, where lower values
7000 improve timer resolution at the expense of processing
7001 more timer interrupts.
7003 xen.balloon_boot_timeout= [XEN]
7004 The time (in seconds) to wait before giving up to boot
7005 in case initial ballooning fails to free enough memory.
7006 Applies only when running as HVM or PVH guest and
7007 started with less memory configured than allowed at
7008 max. Default is 180.
7010 xen.event_eoi_delay= [XEN]
7011 How long to delay EOI handling in case of event
7012 storms (jiffies). Default is 10.
7014 xen.event_loop_timeout= [XEN]
7015 After which time (jiffies) the event handling loop
7016 should start to delay EOI handling. Default is 2.
7018 xen.fifo_events= [XEN]
7019 Boolean parameter to disable using fifo event handling
7020 even if available. Normally fifo event handling is
7021 preferred over the 2-level event handling, as it is
7022 fairer and the number of possible event channels is
7023 much higher. Default is on (use fifo events).
7025 xirc2ps_cs= [NET,PCMCIA]
7027 <irq>,<irq_mask>,<io>,<full_duplex>,<do_sound>,<lockup_hack>[,<irq2>[,<irq3>[,<irq4>]]]
7030 By default on POWER9 and above, the kernel will
7031 natively use the XIVE interrupt controller. This option
7032 allows the fallback firmware mode to be used:
7034 off Fallback to firmware control of XIVE interrupt
7035 controller on both pseries and powernv
7036 platforms. Only useful on POWER9 and above.
7038 xive.store-eoi=off [PPC]
7039 By default on POWER10 and above, the kernel will use
7040 stores for EOI handling when the XIVE interrupt mode
7041 is active. This option allows the XIVE driver to use
7042 loads instead, as on POWER9.
7044 xhci-hcd.quirks [USB,KNL]
7045 A hex value specifying bitmask with supplemental xhci
7046 host controller quirks. Meaning of each bit can be
7047 consulted in header drivers/usb/host/xhci.h.
7050 Format: { early | on | rw | ro | off }
7051 Controls if xmon debugger is enabled. Default is off.
7052 Passing only "xmon" is equivalent to "xmon=early".
7053 early Call xmon as early as possible on boot; xmon
7054 debugger is called from setup_arch().
7055 on xmon debugger hooks will be installed so xmon
7056 is only called on a kernel crash. Default mode,
7057 i.e. either "ro" or "rw" mode, is controlled
7058 with CONFIG_XMON_DEFAULT_RO_MODE.
7059 rw xmon debugger hooks will be installed so xmon
7060 is called only on a kernel crash, mode is write,
7061 meaning SPR registers, memory and, other data
7062 can be written using xmon commands.
7063 ro same as "rw" option above but SPR registers,
7064 memory, and other data can't be written using
7066 off xmon is disabled.
7070 Do not enable amd_pstate as the default
7071 scaling driver for the supported processors
7073 Use amd_pstate as a scaling driver, driver requests a
7074 desired performance on this abstract scale and the power
7075 management firmware translates the requests into actual
7076 hardware states (core frequency, data fabric and memory
7079 Use amd_pstate_epp driver instance as the scaling driver,
7080 driver provides a hint to the hardware if software wants
7081 to bias toward performance (0x0) or energy efficiency (0xff)
7082 to the CPPC firmware. then CPPC power algorithm will
7083 calculate the runtime workload and adjust the realtime cores