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_PCI_COMPONENT
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 PCI/PCI interrupt routing info messages:
64 acpi.debug_layer=0x400000
65 Enable AML "Debug" output, i.e., stores to the Debug
66 object while interpreting AML:
67 acpi.debug_layer=0xffffffff acpi.debug_level=0x2
68 Enable all messages related to ACPI hardware:
69 acpi.debug_layer=0x2 acpi.debug_level=0xffffffff
71 Some values produce so much output that the system is
72 unusable. The "log_buf_len" parameter may be useful
73 if you need to capture more output.
75 acpi_enforce_resources= [ACPI]
77 Check for resource conflicts between native drivers
78 and ACPI OperationRegions (SystemIO and SystemMemory
79 only). IO ports and memory declared in ACPI might be
80 used by the ACPI subsystem in arbitrary AML code and
81 can interfere with legacy drivers.
82 strict (default): access to resources claimed by ACPI
83 is denied; legacy drivers trying to access reserved
84 resources will fail to bind to device using them.
85 lax: access to resources claimed by ACPI is allowed;
86 legacy drivers trying to access reserved resources
87 will bind successfully but a warning message is logged.
88 no: ACPI OperationRegions are not marked as reserved,
89 no further checks are performed.
91 acpi_force_table_verification [HW,ACPI]
92 Enable table checksum verification during early stage.
93 By default, this is disabled due to x86 early mapping
96 acpi_irq_balance [HW,ACPI]
97 ACPI will balance active IRQs
100 acpi_irq_nobalance [HW,ACPI]
101 ACPI will not move active IRQs (default)
104 acpi_irq_isa= [HW,ACPI] If irq_balance, mark listed IRQs used by ISA
105 Format: <irq>,<irq>...
107 acpi_irq_pci= [HW,ACPI] If irq_balance, clear listed IRQs for
109 Format: <irq>,<irq>...
111 acpi_mask_gpe= [HW,ACPI]
112 Due to the existence of _Lxx/_Exx, some GPEs triggered
113 by unsupported hardware/firmware features can result in
114 GPE floodings that cannot be automatically disabled by
116 This facility can be used to prevent such uncontrolled
120 acpi_no_auto_serialize [HW,ACPI]
121 Disable auto-serialization of AML methods
122 AML control methods that contain the opcodes to create
123 named objects will be marked as "Serialized" by the
124 auto-serialization feature.
125 This feature is enabled by default.
126 This option allows to turn off the feature.
128 acpi_no_memhotplug [ACPI] Disable memory hotplug. Useful for kdump
131 acpi_no_static_ssdt [HW,ACPI]
132 Disable installation of static SSDTs at early boot time
133 By default, SSDTs contained in the RSDT/XSDT will be
134 installed automatically and they will appear under
135 /sys/firmware/acpi/tables.
136 This option turns off this feature.
137 Note that specifying this option does not affect
138 dynamic table installation which will install SSDT
139 tables to /sys/firmware/acpi/tables/dynamic.
141 acpi_no_watchdog [HW,ACPI,WDT]
142 Ignore the ACPI-based watchdog interface (WDAT) and let
143 a native driver control the watchdog device instead.
145 acpi_rsdp= [ACPI,EFI,KEXEC]
146 Pass the RSDP address to the kernel, mostly used
147 on machines running EFI runtime service to boot the
148 second kernel for kdump.
150 acpi_os_name= [HW,ACPI] Tell ACPI BIOS the name of the OS
151 Format: To spoof as Windows 98: ="Microsoft Windows"
153 acpi_rev_override [ACPI] Override the _REV object to return 5 (instead
154 of 2 which is mandated by ACPI 6) as the supported ACPI
155 specification revision (when using this switch, it may
156 be necessary to carry out a cold reboot _twice_ in a
157 row to make it take effect on the platform firmware).
159 acpi_osi= [HW,ACPI] Modify list of supported OS interface strings
160 acpi_osi="string1" # add string1
161 acpi_osi="!string2" # remove string2
162 acpi_osi=!* # remove all strings
163 acpi_osi=! # disable all built-in OS vendor
165 acpi_osi=!! # enable all built-in OS vendor
167 acpi_osi= # disable all strings
169 'acpi_osi=!' can be used in combination with single or
170 multiple 'acpi_osi="string1"' to support specific OS
171 vendor string(s). Note that such command can only
172 affect the default state of the OS vendor strings, thus
173 it cannot affect the default state of the feature group
174 strings and the current state of the OS vendor strings,
175 specifying it multiple times through kernel command line
176 is meaningless. This command is useful when one do not
177 care about the state of the feature group strings which
178 should be controlled by the OSPM.
180 1. 'acpi_osi=! acpi_osi="Windows 2000"' is equivalent
181 to 'acpi_osi="Windows 2000" acpi_osi=!', they all
182 can make '_OSI("Windows 2000")' TRUE.
184 'acpi_osi=' cannot be used in combination with other
185 'acpi_osi=' command lines, the _OSI method will not
186 exist in the ACPI namespace. NOTE that such command can
187 only affect the _OSI support state, thus specifying it
188 multiple times through kernel command line is also
191 1. 'acpi_osi=' can make 'CondRefOf(_OSI, Local1)'
194 'acpi_osi=!*' can be used in combination with single or
195 multiple 'acpi_osi="string1"' to support specific
196 string(s). Note that such command can affect the
197 current state of both the OS vendor strings and the
198 feature group strings, thus specifying it multiple times
199 through kernel command line is meaningful. But it may
200 still not able to affect the final state of a string if
201 there are quirks related to this string. This command
202 is useful when one want to control the state of the
203 feature group strings to debug BIOS issues related to
206 1. 'acpi_osi="Module Device" acpi_osi=!*' can make
207 '_OSI("Module Device")' FALSE.
208 2. 'acpi_osi=!* acpi_osi="Module Device"' can make
209 '_OSI("Module Device")' TRUE.
210 3. 'acpi_osi=! acpi_osi=!* acpi_osi="Windows 2000"' is
212 'acpi_osi=!* acpi_osi=! acpi_osi="Windows 2000"'
214 'acpi_osi=!* acpi_osi="Windows 2000" acpi_osi=!',
215 they all will make '_OSI("Windows 2000")' TRUE.
218 Override the pmtimer bug detection: force the kernel
219 to assume that this machine's pmtimer latches its value
220 and always returns good values.
222 acpi_sci= [HW,ACPI] ACPI System Control Interrupt trigger mode
223 Format: { level | edge | high | low }
225 acpi_skip_timer_override [HW,ACPI]
226 Recognize and ignore IRQ0/pin2 Interrupt Override.
227 For broken nForce2 BIOS resulting in XT-PIC timer.
229 acpi_sleep= [HW,ACPI] Sleep options
230 Format: { s3_bios, s3_mode, s3_beep, s4_nohwsig,
231 old_ordering, nonvs, sci_force_enable, nobl }
232 See Documentation/power/video.rst for information on
234 s3_beep is for debugging; it makes the PC's speaker beep
235 as soon as the kernel's real-mode entry point is called.
236 s4_nohwsig prevents ACPI hardware signature from being
237 used during resume from hibernation.
238 old_ordering causes the ACPI 1.0 ordering of the _PTS
239 control method, with respect to putting devices into
240 low power states, to be enforced (the ACPI 2.0 ordering
241 of _PTS is used by default).
242 nonvs prevents the kernel from saving/restoring the
243 ACPI NVS memory during suspend/hibernation and resume.
244 sci_force_enable causes the kernel to set SCI_EN directly
245 on resume from S1/S3 (which is against the ACPI spec,
246 but some broken systems don't work without it).
247 nobl causes the internal blacklist of systems known to
248 behave incorrectly in some ways with respect to system
249 suspend and resume to be ignored (use wisely).
251 acpi_use_timer_override [HW,ACPI]
252 Use timer override. For some broken Nvidia NF5 boards
253 that require a timer override, but don't have HPET
255 add_efi_memmap [EFI; X86] Include EFI memory map in
256 kernel's map of available physical RAM.
259 { off | try_unsupported }
260 off: disable AGP support
261 try_unsupported: try to drive unsupported chipsets
262 (may crash computer or cause data corruption)
265 See Documentation/sound/alsa-configuration.rst
268 Allow the default userspace alignment fault handler
269 behaviour to be specified. Bit 0 enables warnings,
270 bit 1 enables fixups, and bit 2 sends a segfault.
272 align_va_addr= [X86-64]
273 Align virtual addresses by clearing slice [14:12] when
274 allocating a VMA at process creation time. This option
275 gives you up to 3% performance improvement on AMD F15h
276 machines (where it is enabled by default) for a
277 CPU-intensive style benchmark, and it can vary highly in
278 a microbenchmark depending on workload and compiler.
280 32: only for 32-bit processes
281 64: only for 64-bit processes
282 on: enable for both 32- and 64-bit processes
283 off: disable for both 32- and 64-bit processes
285 alloc_snapshot [FTRACE]
286 Allocate the ftrace snapshot buffer on boot up when the
287 main buffer is allocated. This is handy if debugging
288 and you need to use tracing_snapshot() on boot up, and
289 do not want to use tracing_snapshot_alloc() as it needs
290 to be done where GFP_KERNEL allocations are allowed.
292 amd_iommu= [HW,X86-64]
293 Pass parameters to the AMD IOMMU driver in the system.
295 fullflush - enable flushing of IO/TLB entries when
296 they are unmapped. Otherwise they are
297 flushed before they will be reused, which
299 off - do not initialize any AMD IOMMU found in
301 force_isolation - Force device isolation for all
302 devices. The IOMMU driver is not
303 allowed anymore to lift isolation
304 requirements as needed. This option
305 does not override iommu=pt
307 amd_iommu_dump= [HW,X86-64]
308 Enable AMD IOMMU driver option to dump the ACPI table
309 for AMD IOMMU. With this option enabled, AMD IOMMU
310 driver will print ACPI tables for AMD IOMMU during
311 IOMMU initialization.
313 amd_iommu_intr= [HW,X86-64]
314 Specifies one of the following AMD IOMMU interrupt
316 legacy - Use legacy interrupt remapping mode.
317 vapic - Use virtual APIC mode, which allows IOMMU
318 to inject interrupts directly into guest.
319 This mode requires kvm-amd.avic=1.
320 (Default when IOMMU HW support is present.)
322 amijoy.map= [HW,JOY] Amiga joystick support
323 Map of devices attached to JOY0DAT and JOY1DAT
325 See also Documentation/input/joydev/joystick.rst
327 analog.map= [HW,JOY] Analog joystick and gamepad support
328 Specifies type or capabilities of an analog joystick
329 connected to one of 16 gameports
330 Format: <type1>,<type2>,..<type16>
333 Power management functions (SPARCstation-4/5 + deriv.)
335 Disable APC CPU standby support. SPARCstation-Fox does
336 not play well with APC CPU idle - disable it if you have
337 APC and your system crashes randomly.
339 apic= [APIC,X86] Advanced Programmable Interrupt Controller
340 Change the output verbosity while booting
341 Format: { quiet (default) | verbose | debug }
342 Change the amount of debugging information output
343 when initialising the APIC and IO-APIC components.
344 For X86-32, this can also be used to specify an APIC
346 Format: apic=driver_name
347 Examples: apic=bigsmp
349 apic_extnmi= [APIC,X86] External NMI delivery setting
350 Format: { bsp (default) | all | none }
351 bsp: External NMI is delivered only to CPU 0
352 all: External NMIs are broadcast to all CPUs as a
354 none: External NMI is masked for all CPUs. This is
355 useful so that a dump capture kernel won't be
359 See Documentation/networking/ipv6.rst.
361 show_lapic= [APIC,X86] Advanced Programmable Interrupt Controller
362 Limit apic dumping. The parameter defines the maximal
363 number of local apics being dumped. Also it is possible
364 to set it to "all" by meaning -- no limit here.
365 Format: { 1 (default) | 2 | ... | all }.
366 The parameter valid if only apic=debug or
367 apic=verbose is specified.
368 Example: apic=debug show_lapic=all
370 apm= [APM] Advanced Power Management
371 See header of arch/x86/kernel/apm_32.c.
373 arcrimi= [HW,NET] ARCnet - "RIM I" (entirely mem-mapped) cards
374 Format: <io>,<irq>,<nodeID>
378 atarimouse= [HW,MOUSE] Atari Mouse
380 atkbd.extra= [HW] Enable extra LEDs and keys on IBM RapidAccess,
381 EzKey and similar keyboards
383 atkbd.reset= [HW] Reset keyboard during initialization
385 atkbd.set= [HW] Select keyboard code set
386 Format: <int> (2 = AT (default), 3 = PS/2)
388 atkbd.scroll= [HW] Enable scroll wheel on MS Office and similar
391 atkbd.softraw= [HW] Choose between synthetic and real raw mode
392 Format: <bool> (0 = real, 1 = synthetic (default))
394 atkbd.softrepeat= [HW]
395 Use software keyboard repeat
397 audit= [KNL] Enable the audit sub-system
398 Format: { "0" | "1" | "off" | "on" }
399 0 | off - kernel audit is disabled and can not be
400 enabled until the next reboot
401 unset - kernel audit is initialized but disabled and
402 will be fully enabled by the userspace auditd.
403 1 | on - kernel audit is initialized and partially
404 enabled, storing at most audit_backlog_limit
405 messages in RAM until it is fully enabled by the
409 audit_backlog_limit= [KNL] Set the audit queue size limit.
410 Format: <int> (must be >=0)
413 bau= [X86_UV] Enable the BAU on SGI UV. The default
414 behavior is to disable the BAU (i.e. bau=0).
415 Format: { "0" | "1" }
418 unset - Disable the BAU.
420 baycom_epp= [HW,AX25]
423 baycom_par= [HW,AX25] BayCom Parallel Port AX.25 Modem
425 See header of drivers/net/hamradio/baycom_par.c.
427 baycom_ser_fdx= [HW,AX25]
428 BayCom Serial Port AX.25 Modem (Full Duplex Mode)
429 Format: <io>,<irq>,<mode>[,<baud>]
430 See header of drivers/net/hamradio/baycom_ser_fdx.c.
432 baycom_ser_hdx= [HW,AX25]
433 BayCom Serial Port AX.25 Modem (Half Duplex Mode)
434 Format: <io>,<irq>,<mode>
435 See header of drivers/net/hamradio/baycom_ser_hdx.c.
437 blkdevparts= Manual partition parsing of block device(s) for
438 embedded devices based on command line input.
439 See Documentation/block/cmdline-partition.rst
441 boot_delay= Milliseconds to delay each printk during boot.
442 Values larger than 10 seconds (10000) are changed to
447 Extended command line options can be added to an initrd
448 and this will cause the kernel to look for it.
450 See Documentation/admin-guide/bootconfig.rst
453 Disable BERT OS support on buggy BIOSes.
455 bgrt_disable [ACPI][X86]
456 Disable BGRT to avoid flickering OEM logo.
458 bttv.card= [HW,V4L] bttv (bt848 + bt878 based grabber cards)
459 bttv.radio= Most important insmod options are available as
461 bttv.pll= See Documentation/admin-guide/media/bttv.rst
464 bulk_remove=off [PPC] This parameter disables the use of the pSeries
465 firmware feature for flushing multiple hpte entries
468 c101= [NET] Moxa C101 synchronous serial card
470 cachesize= [BUGS=X86-32] Override level 2 CPU cache size detection.
471 Sometimes CPU hardware bugs make them report the cache
472 size incorrectly. The kernel will attempt work arounds
473 to fix known problems, but for some CPUs it is not
474 possible to determine what the correct size should be.
475 This option provides an override for these situations.
478 [NET] Specifies amount of time (in seconds) that
479 the kernel should wait for a network carrier. By default
480 it waits 120 seconds.
482 ca_keys= [KEYS] This parameter identifies a specific key(s) on
483 the system trusted keyring to be used for certificate
485 format: { id:<keyid> | builtin }
487 cca= [MIPS] Override the kernel pages' cache coherency
488 algorithm. Accepted values range from 0 to 7
489 inclusive. See arch/mips/include/asm/pgtable-bits.h
490 for platform specific values (SB1, Loongson3 and
493 ccw_timeout_log [S390]
494 See Documentation/s390/common_io.rst for details.
496 cgroup_disable= [KNL] Disable a particular controller
497 Format: {name of the controller(s) to disable}
498 The effects of cgroup_disable=foo are:
499 - foo isn't auto-mounted if you mount all cgroups in
501 - foo isn't visible as an individually mountable
503 {Currently only "memory" controller deal with this and
504 cut the overhead, others just disable the usage. So
505 only cgroup_disable=memory is actually worthy}
507 cgroup_no_v1= [KNL] Disable cgroup controllers and named hierarchies in v1
508 Format: { { controller | "all" | "named" }
509 [,{ controller | "all" | "named" }...] }
510 Like cgroup_disable, but only applies to cgroup v1;
511 the blacklisted controllers remain available in cgroup2.
512 "all" blacklists all controllers and "named" disables
513 named mounts. Specifying both "all" and "named" disables
516 cgroup.memory= [KNL] Pass options to the cgroup memory controller.
518 nosocket -- Disable socket memory accounting.
519 nokmem -- Disable kernel memory accounting.
521 checkreqprot [SELINUX] Set initial checkreqprot flag value.
522 Format: { "0" | "1" }
523 See security/selinux/Kconfig help text.
524 0 -- check protection applied by kernel (includes
525 any implied execute protection).
526 1 -- check protection requested by application.
527 Default value is set via a kernel config option.
528 Value can be changed at runtime via
529 /sys/fs/selinux/checkreqprot.
530 Setting checkreqprot to 1 is deprecated.
533 See Documentation/s390/common_io.rst for details.
536 Prevents the clock framework from automatically gating
537 clocks that have not been explicitly enabled by a Linux
538 device driver but are enabled in hardware at reset or
539 by the bootloader/firmware. Note that this does not
540 force such clocks to be always-on nor does it reserve
541 those clocks in any way. This parameter is useful for
542 debug and development, but should not be needed on a
543 platform with proper driver support. For more
544 information, see Documentation/driver-api/clk.rst.
546 clock= [BUGS=X86-32, HW] gettimeofday clocksource override.
548 Forces specified clocksource (if available) to be used
549 when calculating gettimeofday(). If specified
550 clocksource is not available, it defaults to PIT.
551 Format: { pit | tsc | cyclone | pmtmr }
553 clocksource= Override the default clocksource
555 Override the default clocksource and use the clocksource
556 with the name specified.
557 Some clocksource names to choose from, depending on
559 [all] jiffies (this is the base, fallback clocksource)
561 [ARM] imx_timer1,OSTS,netx_timer,mpu_timer2,
562 pxa_timer,timer3,32k_counter,timer0_1
563 [X86-32] pit,hpet,tsc;
564 scx200_hrt on Geode; cyclone on IBM x440
572 clocksource.arm_arch_timer.evtstrm=
575 Enable/disable the eventstream feature of the ARM
576 architected timer so that code using WFE-based polling
577 loops can be debugged more effectively on production
580 clearcpuid=BITNUM[,BITNUM...] [X86]
581 Disable CPUID feature X for the kernel. See
582 arch/x86/include/asm/cpufeatures.h for the valid bit
583 numbers. Note the Linux specific bits are not necessarily
584 stable over kernel options, but the vendor specific
586 Also note that user programs calling CPUID directly
587 or using the feature without checking anything
588 will still see it. This just prevents it from
589 being used by the kernel or shown in /proc/cpuinfo.
590 Also note the kernel might malfunction if you disable
593 cma=nn[MG]@[start[MG][-end[MG]]]
595 Sets the size of kernel global memory area for
596 contiguous memory allocations and optionally the
597 placement constraint by the physical address range of
598 memory allocations. A value of 0 disables CMA
599 altogether. For more information, see
600 kernel/dma/contiguous.c
604 Sets the size of kernel per-numa memory area for
605 contiguous memory allocations. A value of 0 disables
606 per-numa CMA altogether. And If this option is not
607 specificed, the default value is 0.
608 With per-numa CMA enabled, DMA users on node nid will
609 first try to allocate buffer from the pernuma area
610 which is located in node nid, if the allocation fails,
611 they will fallback to the global default memory area.
613 cmo_free_hint= [PPC] Format: { yes | no }
614 Specify whether pages are marked as being inactive
615 when they are freed. This is used in CMO environments
616 to determine OS memory pressure for page stealing by
620 coherent_pool=nn[KMG] [ARM,KNL]
621 Sets the size of memory pool for coherent, atomic dma
622 allocations, by default set to 256K.
624 com20020= [HW,NET] ARCnet - COM20020 chipset
626 <io>[,<irq>[,<nodeID>[,<backplane>[,<ckp>[,<timeout>]]]]]
628 com90io= [HW,NET] ARCnet - COM90xx chipset (IO-mapped buffers)
632 ARCnet - COM90xx chipset (memory-mapped buffers)
633 Format: <io>[,<irq>[,<memstart>]]
635 condev= [HW,S390] console device
638 console= [KNL] Output console device and options.
640 tty<n> Use the virtual console device <n>.
644 Use the specified serial port. The options are of
645 the form "bbbbpnf", where "bbbb" is the baud rate,
646 "p" is parity ("n", "o", or "e"), "n" is number of
647 bits, and "f" is flow control ("r" for RTS or
648 omit it). Default is "9600n8".
650 See Documentation/admin-guide/serial-console.rst for more
652 Documentation/networking/netconsole.rst for an
655 uart[8250],io,<addr>[,options]
656 uart[8250],mmio,<addr>[,options]
657 uart[8250],mmio16,<addr>[,options]
658 uart[8250],mmio32,<addr>[,options]
659 uart[8250],0x<addr>[,options]
660 Start an early, polled-mode console on the 8250/16550
661 UART at the specified I/O port or MMIO address,
662 switching to the matching ttyS device later.
663 MMIO inter-register address stride is either 8-bit
664 (mmio), 16-bit (mmio16), or 32-bit (mmio32).
665 If none of [io|mmio|mmio16|mmio32], <addr> is assumed
666 to be equivalent to 'mmio'. 'options' are specified in
667 the same format described for ttyS above; if unspecified,
668 the h/w is not re-initialized.
670 hvc<n> Use the hypervisor console device <n>. This is for
671 both Xen and PowerPC hypervisors.
673 If the device connected to the port is not a TTY but a braille
674 device, prepend "brl," before the device type, for instance
676 For now, only VisioBraille is supported.
679 [KNL] Change console messages format
681 By default we print messages on consoles in
682 "[time stamp] text\n" format (time stamp may not be
683 printed, depending on CONFIG_PRINTK_TIME or
684 `printk_time' param).
686 Switch to syslog format: "<%u>[time stamp] text\n"
687 IOW, each message will have a facility and loglevel
688 prefix. The format is similar to one used by syslog()
689 syscall, or to executing "dmesg -S --raw" or to reading
692 consoleblank= [KNL] The console blank (screen saver) timeout in
693 seconds. A value of 0 disables the blank timer.
697 [KNL] Change the default value for
698 /proc/<pid>/coredump_filter.
699 See also Documentation/filesystems/proc.rst.
701 coresight_cpu_debug.enable
704 Enable/disable the CPU sampling based debugging.
705 0: default value, disable debugging
706 1: enable debugging at boot time
708 cpuidle.off=1 [CPU_IDLE]
709 disable the cpuidle sub-system
712 [CPU_IDLE] Name of the cpuidle governor to use.
714 cpufreq.off=1 [CPU_FREQ]
715 disable the cpufreq sub-system
717 cpufreq.default_governor=
718 [CPU_FREQ] Name of the default cpufreq governor or
719 policy to use. This governor must be registered in the
720 kernel before the cpufreq driver probes.
723 [X86] Delay for N microsec between assert and de-assert
724 of APIC INIT to start processors. This delay occurs
725 on every CPU online, such as boot, and resume from suspend.
728 cpcihp_generic= [HW,PCI] Generic port I/O CompactPCI driver
730 <first_slot>,<last_slot>,<port>,<enum_bit>[,<debug>]
732 crashkernel=size[KMG][@offset[KMG]]
733 [KNL] Using kexec, Linux can switch to a 'crash kernel'
734 upon panic. This parameter reserves the physical
735 memory region [offset, offset + size] for that kernel
736 image. If '@offset' is omitted, then a suitable offset
737 is selected automatically.
738 [KNL, X86-64] Select a region under 4G first, and
739 fall back to reserve region above 4G when '@offset'
740 hasn't been specified.
741 See Documentation/admin-guide/kdump/kdump.rst for further details.
743 crashkernel=range1:size1[,range2:size2,...][@offset]
744 [KNL] Same as above, but depends on the memory
745 in the running system. The syntax of range is
746 start-[end] where start and end are both
747 a memory unit (amount[KMG]). See also
748 Documentation/admin-guide/kdump/kdump.rst for an example.
750 crashkernel=size[KMG],high
751 [KNL, X86-64] range could be above 4G. Allow kernel
752 to allocate physical memory region from top, so could
753 be above 4G if system have more than 4G ram installed.
754 Otherwise memory region will be allocated below 4G, if
756 It will be ignored if crashkernel=X is specified.
757 crashkernel=size[KMG],low
758 [KNL, X86-64] range under 4G. When crashkernel=X,high
759 is passed, kernel could allocate physical memory region
760 above 4G, that cause second kernel crash on system
761 that require some amount of low memory, e.g. swiotlb
762 requires at least 64M+32K low memory, also enough extra
763 low memory is needed to make sure DMA buffers for 32-bit
764 devices won't run out. Kernel would try to allocate at
765 at least 256M below 4G automatically.
766 This one let user to specify own low range under 4G
767 for second kernel instead.
768 0: to disable low allocation.
769 It will be ignored when crashkernel=X,high is not used
770 or memory reserved is below 4G.
773 [KNL] Disable crypto self-tests
778 cs89x0_media= [HW,NET]
779 Format: { rj45 | aui | bnc }
782 See header of drivers/s390/block/dasd_devmap.c.
784 db9.dev[2|3]= [HW,JOY] Multisystem joystick support via parallel port
785 (one device per port)
786 Format: <port#>,<type>
787 See also Documentation/input/devices/joystick-parport.rst
789 ddebug_query= [KNL,DYNAMIC_DEBUG] Enable debug messages at early boot
791 Documentation/admin-guide/dynamic-debug-howto.rst for
792 details. Deprecated, see dyndbg.
794 debug [KNL] Enable kernel debugging (events log level).
797 [KNL] Enable printing [hashed] pointers early in the
798 boot sequence. If enabled, we use a weak hash instead
799 of siphash to hash pointers. Use this option if you are
800 seeing instances of '(___ptrval___)') and need to see a
801 value (hashed pointer) instead. Cryptographically
802 insecure, please do not use on production kernels.
805 [KNL] verbose self-tests
807 Print debugging info while doing the locking API
809 We default to 0 (no extra messages), setting it to
810 1 will print _a lot_ more information - normally
811 only useful to kernel developers.
813 debug_objects [KNL] Enable object debugging
816 [KNL] Disable object debugging
818 debug_guardpage_minorder=
819 [KNL] When CONFIG_DEBUG_PAGEALLOC is set, this
820 parameter allows control of the order of pages that will
821 be intentionally kept free (and hence protected) by the
822 buddy allocator. Bigger value increase the probability
823 of catching random memory corruption, but reduce the
824 amount of memory for normal system use. The maximum
825 possible value is MAX_ORDER/2. Setting this parameter
826 to 1 or 2 should be enough to identify most random
827 memory corruption problems caused by bugs in kernel or
828 driver code when a CPU writes to (or reads from) a
829 random memory location. Note that there exists a class
830 of memory corruptions problems caused by buggy H/W or
831 F/W or by drivers badly programing DMA (basically when
832 memory is written at bus level and the CPU MMU is
833 bypassed) which are not detectable by
834 CONFIG_DEBUG_PAGEALLOC, hence this option will not help
835 tracking down these problems.
838 [KNL] When CONFIG_DEBUG_PAGEALLOC is set, this parameter
839 enables the feature at boot time. By default, it is
840 disabled and the system will work mostly the same as a
841 kernel built without CONFIG_DEBUG_PAGEALLOC.
842 Note: to get most of debug_pagealloc error reports, it's
843 useful to also enable the page_owner functionality.
844 on: enable the feature
846 debugfs= [KNL] This parameter enables what is exposed to userspace
847 and debugfs internal clients.
848 Format: { on, no-mount, off }
849 on: All functions are enabled.
851 Filesystem is not registered but kernel clients can
852 access APIs and a crashkernel can be used to read
853 its content. There is nothing to mount.
854 off: Filesystem is not registered and clients
855 get a -EPERM as result when trying to register files
856 or directories within debugfs.
857 This is equivalent of the runtime functionality if
858 debugfs was not enabled in the kernel at all.
859 Default value is set in build-time with a kernel configuration.
861 debugpat [X86] Enable PAT debugging
863 decnet.addr= [HW,NET]
864 Format: <area>[,<node>]
865 See also Documentation/networking/decnet.rst.
868 [HW] The size of the default HugeTLB page. This is
869 the size represented by the legacy /proc/ hugepages
870 APIs. In addition, this is the default hugetlb size
871 used for shmget(), mmap() and mounting hugetlbfs
872 filesystems. If not specified, defaults to the
873 architecture's default huge page size. Huge page
874 sizes are architecture dependent. See also
875 Documentation/admin-guide/mm/hugetlbpage.rst.
878 deferred_probe_timeout=
879 [KNL] Debugging option to set a timeout in seconds for
880 deferred probe to give up waiting on dependencies to
881 probe. Only specific dependencies (subsystems or
882 drivers) that have opted in will be ignored. A timeout of 0
883 will timeout at the end of initcalls. This option will also
884 dump out devices still on the deferred probe list after
888 Format: { on | off | def_only | inf_only | always }
889 on: s390 zlib hardware support for compression on
890 level 1 and decompression (default)
891 off: No s390 zlib hardware support
892 def_only: s390 zlib hardware support for deflate
893 only (compression on level 1)
894 inf_only: s390 zlib hardware support for inflate
896 always: Same as 'on' but ignores the selected compression
897 level always using hardware support (used for debugging)
900 Set number of hash buckets for dentry cache.
902 disable_1tb_segments [PPC]
903 Disables the use of 1TB hash page table segments. This
904 causes the kernel to fall back to 256MB segments which
905 can be useful when debugging issues that require an SLB
909 Limits the number of kernel SLB entries, and flushes
910 them frequently to increase the rate of SLB faults
914 See Documentation/networking/ipv6.rst.
917 [KNL] Under CONFIG_HARDENED_USERCOPY, whether
918 hardening is enabled for this boot. Hardened
919 usercopy checking is used to protect the kernel
920 from reading or writing beyond known memory
921 allocation boundaries as a proactive defense
922 against bounds-checking flaws in the kernel's
923 copy_to_user()/copy_from_user() interface.
924 on Perform hardened usercopy checks (default).
925 off Disable hardened usercopy checks.
928 Disable RADIX MMU mode on POWER9
930 radix_hcall_invalidate=on [PPC/PSERIES]
931 Disable RADIX GTSE feature and use hcall for TLB
935 Disable TLBIE instruction. Currently does not work
936 with KVM, with HASH MMU, or with coherent accelerators.
938 disable_cpu_apicid= [X86,APIC,SMP]
940 The number of initial APIC ID for the
941 corresponding CPU to be disabled at boot,
942 mostly used for the kdump 2nd kernel to
943 disable BSP to wake up multiple CPUs without
944 causing system reset or hang due to sending
947 perf_v4_pmi= [X86,INTEL]
949 Disable Intel PMU counter freezing feature.
950 The feature only exists starting from
951 Arch Perfmon v4 (Skylake and newer).
953 disable_ddw [PPC/PSERIES]
954 Disable Dynamic DMA Window support. Use this
955 to workaround buggy firmware.
958 See Documentation/networking/ipv6.rst.
960 disable_mtrr_cleanup [X86]
961 The kernel tries to adjust MTRR layout from continuous
962 to discrete, to make X server driver able to add WB
963 entry later. This parameter disables that.
965 disable_mtrr_trim [X86, Intel and AMD only]
966 By default the kernel will trim any uncacheable
967 memory out of your available memory pool based on
968 MTRR settings. This parameter disables that behavior,
969 possibly causing your machine to run very slowly.
971 disable_timer_pin_1 [X86]
972 Disable PIN 1 of APIC timer
973 Can be useful to work around chipset bugs.
975 dis_ucode_ldr [X86] Disable the microcode loader.
977 dma_debug=off If the kernel is compiled with DMA_API_DEBUG support,
978 this option disables the debugging code at boot.
980 dma_debug_entries=<number>
981 This option allows to tune the number of preallocated
982 entries for DMA-API debugging code. One entry is
983 required per DMA-API allocation. Use this if the
984 DMA-API debugging code disables itself because the
985 architectural default is too low.
987 dma_debug_driver=<driver_name>
988 With this option the DMA-API debugging driver
989 filter feature can be enabled at boot time. Just
990 pass the driver to filter for as the parameter.
991 The filter can be disabled or changed to another
992 driver later using sysfs.
994 driver_async_probe= [KNL]
995 List of driver names to be probed asynchronously.
996 Format: <driver_name1>,<driver_name2>...
998 drm.edid_firmware=[<connector>:]<file>[,[<connector>:]<file>]
999 Broken monitors, graphic adapters, KVMs and EDIDless
1000 panels may send no or incorrect EDID data sets.
1001 This parameter allows to specify an EDID data sets
1002 in the /lib/firmware directory that are used instead.
1003 Generic built-in EDID data sets are used, if one of
1004 edid/1024x768.bin, edid/1280x1024.bin,
1005 edid/1680x1050.bin, or edid/1920x1080.bin is given
1006 and no file with the same name exists. Details and
1007 instructions how to build your own EDID data are
1008 available in Documentation/admin-guide/edid.rst. An EDID
1009 data set will only be used for a particular connector,
1010 if its name and a colon are prepended to the EDID
1011 name. Each connector may use a unique EDID data
1012 set by separating the files with a comma. An EDID
1013 data set with no connector name will be used for
1014 any connectors not explicitly specified.
1019 Format: {"off" | "known"}
1020 Control how the dt_cpu_ftrs device-tree binding is
1021 used for CPU feature discovery and setup (if it
1023 off: Do not use it, fall back to legacy cpu table.
1024 known: Do not pass through unknown features to guests
1025 or userspace, only those that the kernel is aware of.
1027 dump_apple_properties [X86]
1028 Dump name and content of EFI device properties on
1029 x86 Macs. Useful for driver authors to determine
1030 what data is available or for reverse-engineering.
1032 dyndbg[="val"] [KNL,DYNAMIC_DEBUG]
1033 <module>.dyndbg[="val"]
1034 Enable debug messages at boot time. See
1035 Documentation/admin-guide/dynamic-debug-howto.rst
1038 nopku [X86] Disable Memory Protection Keys CPU feature found
1041 <module>.async_probe [KNL]
1042 Enable asynchronous probe on this module.
1044 early_ioremap_debug [KNL]
1045 Enable debug messages in early_ioremap support. This
1046 is useful for tracking down temporary early mappings
1047 which are not unmapped.
1049 earlycon= [KNL] Output early console device and options.
1051 When used with no options, the early console is
1052 determined by stdout-path property in device tree's
1053 chosen node or the ACPI SPCR table if supported by
1056 cdns,<addr>[,options]
1057 Start an early, polled-mode console on a Cadence
1058 (xuartps) serial port at the specified address. Only
1059 supported option is baud rate. If baud rate is not
1060 specified, the serial port must already be setup and
1063 uart[8250],io,<addr>[,options]
1064 uart[8250],mmio,<addr>[,options]
1065 uart[8250],mmio32,<addr>[,options]
1066 uart[8250],mmio32be,<addr>[,options]
1067 uart[8250],0x<addr>[,options]
1068 Start an early, polled-mode console on the 8250/16550
1069 UART at the specified I/O port or MMIO address.
1070 MMIO inter-register address stride is either 8-bit
1071 (mmio) or 32-bit (mmio32 or mmio32be).
1072 If none of [io|mmio|mmio32|mmio32be], <addr> is assumed
1073 to be equivalent to 'mmio'. 'options' are specified
1074 in the same format described for "console=ttyS<n>"; if
1075 unspecified, the h/w is not initialized.
1079 Start an early, polled-mode console on a pl011 serial
1080 port at the specified address. The pl011 serial port
1081 must already be setup and configured. Options are not
1082 yet supported. If 'mmio32' is specified, then only
1083 the driver will use only 32-bit accessors to read/write
1084 the device registers.
1087 Start an early, polled-mode console on a meson serial
1088 port at the specified address. The serial port must
1089 already be setup and configured. Options are not yet
1093 Start an early, polled-mode console on an msm serial
1094 port at the specified address. The serial port
1095 must already be setup and configured. Options are not
1098 msm_serial_dm,<addr>
1099 Start an early, polled-mode console on an msm serial
1100 dm port at the specified address. The serial port
1101 must already be setup and configured. Options are not
1105 Start an early, polled-mode console on a serial port
1106 of an Actions Semi SoC, such as S500 or S900, at the
1107 specified address. The serial port must already be
1108 setup and configured. Options are not yet supported.
1111 Start an early, polled-mode console on a serial port
1112 of an RDA Micro SoC, such as RDA8810PL, at the
1113 specified address. The serial port must already be
1114 setup and configured. Options are not yet supported.
1117 Use RISC-V SBI (Supervisor Binary Interface) for early
1120 smh Use ARM semihosting calls for early console.
1128 Use early console provided by serial driver available
1129 on Samsung SoCs, requires selecting proper type and
1130 a correct base address of the selected UART port. The
1131 serial port must already be setup and configured.
1132 Options are not yet supported.
1135 Start an early, polled-mode console on a lantiq serial
1136 (lqasc) port at the specified address. The serial port
1137 must already be setup and configured. Options are not
1142 Use early console provided by Freescale LP UART driver
1143 found on Freescale Vybrid and QorIQ LS1021A processors.
1144 A valid base address must be provided, and the serial
1145 port must already be setup and configured.
1149 Start an early, polled-mode, output-only console on the
1150 Freescale i.MX UART at the specified address. The UART
1151 must already be setup and configured.
1154 Start an early, polled-mode console on the
1155 Armada 3700 serial port at the specified
1156 address. The serial port must already be setup
1157 and configured. Options are not yet supported.
1160 Start an early, polled-mode console on a Qualcomm
1161 Generic Interface (GENI) based serial port at the
1162 specified address. The serial port must already be
1163 setup and configured. Options are not yet supported.
1166 Start an early, unaccelerated console on the EFI
1167 memory mapped framebuffer (if available). On cache
1168 coherent non-x86 systems that use system memory for
1169 the framebuffer, pass the 'ram' option so that it is
1170 mapped with the correct attributes.
1173 Use early console provided by Freescale LINFlexD UART
1174 serial driver for NXP S32V234 SoCs. A valid base
1175 address must be provided, and the serial port must
1176 already be setup and configured.
1178 earlyprintk= [X86,SH,ARM,M68k,S390]
1182 earlyprintk=serial[,ttySn[,baudrate]]
1183 earlyprintk=serial[,0x...[,baudrate]]
1184 earlyprintk=ttySn[,baudrate]
1185 earlyprintk=dbgp[debugController#]
1186 earlyprintk=pciserial[,force],bus:device.function[,baudrate]
1187 earlyprintk=xdbc[xhciController#]
1189 earlyprintk is useful when the kernel crashes before
1190 the normal console is initialized. It is not enabled by
1191 default because it has some cosmetic problems.
1193 Append ",keep" to not disable it when the real console
1196 Only one of vga, efi, serial, or usb debug port can
1199 Currently only ttyS0 and ttyS1 may be specified by
1200 name. Other I/O ports may be explicitly specified
1201 on some architectures (x86 and arm at least) by
1202 replacing ttySn with an I/O port address, like this:
1203 earlyprintk=serial,0x1008,115200
1204 You can find the port for a given device in
1205 /proc/tty/driver/serial:
1206 2: uart:ST16650V2 port:00001008 irq:18 ...
1208 Interaction with the standard serial driver is not
1211 The VGA and EFI output is eventually overwritten by
1214 The xen output can only be used by Xen PV guests.
1216 The sclp output can only be used on s390.
1218 The optional "force" to "pciserial" enables use of a
1219 PCI device even when its classcode is not of the
1222 edac_report= [HW,EDAC] Control how to report EDAC event
1223 Format: {"on" | "off" | "force"}
1224 on: enable EDAC to report H/W event. May be overridden
1225 by other higher priority error reporting module.
1226 off: disable H/W event reporting through EDAC.
1227 force: enforce the use of EDAC to report H/W event.
1230 ekgdboc= [X86,KGDB] Allow early kernel console debugging
1233 This is designed to be used in conjunction with
1234 the boot argument: earlyprintk=vga
1236 This parameter works in place of the kgdboc parameter
1237 but can only be used if the backing tty is available
1238 very early in the boot process. For early debugging
1239 via a serial port see kgdboc_earlycon instead.
1242 Format: {"off" | "on" | "skip[mbr]"}
1245 Format: { "debug", "disable_early_pci_dma",
1246 "nochunk", "noruntime", "nosoftreserve",
1247 "novamap", "no_disable_early_pci_dma" }
1248 debug: enable misc debug output.
1249 disable_early_pci_dma: disable the busmaster bit on all
1250 PCI bridges while in the EFI boot stub.
1251 nochunk: disable reading files in "chunks" in the EFI
1252 boot stub, as chunking can cause problems with some
1253 firmware implementations.
1254 noruntime : disable EFI runtime services support
1255 nosoftreserve: The EFI_MEMORY_SP (Specific Purpose)
1256 attribute may cause the kernel to reserve the
1257 memory range for a memory mapping driver to
1258 claim. Specify efi=nosoftreserve to disable this
1259 reservation and treat the memory by its base type
1260 (i.e. EFI_CONVENTIONAL_MEMORY / "System RAM").
1261 novamap: do not call SetVirtualAddressMap().
1262 no_disable_early_pci_dma: Leave the busmaster bit set
1263 on all PCI bridges while in the EFI boot stub
1265 efi_no_storage_paranoia [EFI; X86]
1266 Using this parameter you can use more than 50% of
1267 your efi variable storage. Use this parameter only if
1268 you are really sure that your UEFI does sane gc and
1269 fulfills the spec otherwise your board may brick.
1271 efi_fake_mem= nn[KMG]@ss[KMG]:aa[,nn[KMG]@ss[KMG]:aa,..] [EFI; X86]
1272 Add arbitrary attribute to specific memory range by
1273 updating original EFI memory map.
1274 Region of memory which aa attribute is added to is
1277 If efi_fake_mem=2G@4G:0x10000,2G@0x10a0000000:0x10000
1278 is specified, EFI_MEMORY_MORE_RELIABLE(0x10000)
1279 attribute is added to range 0x100000000-0x180000000 and
1280 0x10a0000000-0x1120000000.
1282 If efi_fake_mem=8G@9G:0x40000 is specified, the
1283 EFI_MEMORY_SP(0x40000) attribute is added to
1284 range 0x240000000-0x43fffffff.
1286 Using this parameter you can do debugging of EFI memmap
1287 related features. For example, you can do debugging of
1288 Address Range Mirroring feature even if your box
1289 doesn't support it, or mark specific memory as
1292 efivar_ssdt= [EFI; X86] Name of an EFI variable that contains an SSDT
1293 that is to be dynamically loaded by Linux. If there are
1294 multiple variables with the same name but with different
1295 vendor GUIDs, all of them will be loaded. See
1296 Documentation/admin-guide/acpi/ssdt-overlays.rst for details.
1299 eisa_irq_edge= [PARISC,HW]
1300 See header of drivers/parisc/eisa.c.
1303 See comment before function elanfreq_setup() in
1304 arch/x86/kernel/cpu/cpufreq/elanfreq.c.
1306 elfcorehdr=[size[KMG]@]offset[KMG] [IA64,PPC,SH,X86,S390]
1307 Specifies physical address of start of kernel core
1308 image elf header and optionally the size. Generally
1309 kexec loader will pass this option to capture kernel.
1310 See Documentation/admin-guide/kdump/kdump.rst for details.
1312 enable_mtrr_cleanup [X86]
1313 The kernel tries to adjust MTRR layout from continuous
1314 to discrete, to make X server driver able to add WB
1315 entry later. This parameter enables that.
1317 enable_timer_pin_1 [X86]
1318 Enable PIN 1 of APIC timer
1319 Can be useful to work around chipset bugs
1320 (in particular on some ATI chipsets).
1321 The kernel tries to set a reasonable default.
1323 enforcing [SELINUX] Set initial enforcing status.
1325 See security/selinux/Kconfig help text.
1326 0 -- permissive (log only, no denials).
1327 1 -- enforcing (deny and log).
1329 Value can be changed at runtime via
1330 /sys/fs/selinux/enforce.
1333 Disable Error Record Serialization Table (ERST)
1336 ether= [HW,NET] Ethernet cards parameters
1337 This option is obsoleted by the "netdev=" option, which
1338 has equivalent usage. See its documentation for details.
1342 Permit 'security.evm' to be updated regardless of
1343 current integrity status.
1348 fail_make_request=[KNL]
1349 General fault injection mechanism.
1350 Format: <interval>,<probability>,<space>,<times>
1351 See also Documentation/fault-injection/.
1354 Format: { initns | none }
1355 See Documentation/admin-guide/sysctl/net.rst for
1356 fb_tunnels_only_for_init_ns
1359 See Documentation/admin-guide/blockdev/floppy.rst.
1361 force_pal_cache_flush
1362 [IA-64] Avoid check_sal_cache_flush which may hang on
1363 buggy SAL_CACHE_FLUSH implementations. Using this
1364 parameter will force ia64_sal_cache_flush to call
1365 ia64_pal_cache_flush instead of SAL_CACHE_FLUSH.
1368 Forcefully enable Physical Address Extension (PAE).
1369 Many Pentium M systems disable PAE but may have a
1370 functionally usable PAE implementation.
1371 Warning: use of this parameter will taint the kernel
1372 and may cause unknown problems.
1375 [FTRACE] will set and start the specified tracer
1376 as early as possible in order to facilitate early
1379 ftrace_dump_on_oops[=orig_cpu]
1380 [FTRACE] will dump the trace buffers on oops.
1381 If no parameter is passed, ftrace will dump
1382 buffers of all CPUs, but if you pass orig_cpu, it will
1383 dump only the buffer of the CPU that triggered the
1386 ftrace_filter=[function-list]
1387 [FTRACE] Limit the functions traced by the function
1388 tracer at boot up. function-list is a comma-separated
1389 list of functions. This list can be changed at run
1390 time by the set_ftrace_filter file in the debugfs
1393 ftrace_notrace=[function-list]
1394 [FTRACE] Do not trace the functions specified in
1395 function-list. This list can be changed at run time
1396 by the set_ftrace_notrace file in the debugfs
1399 ftrace_graph_filter=[function-list]
1400 [FTRACE] Limit the top level callers functions traced
1401 by the function graph tracer at boot up.
1402 function-list is a comma-separated list of functions
1403 that can be changed at run time by the
1404 set_graph_function file in the debugfs tracing directory.
1406 ftrace_graph_notrace=[function-list]
1407 [FTRACE] Do not trace from the functions specified in
1408 function-list. This list is a comma-separated list of
1409 functions that can be changed at run time by the
1410 set_graph_notrace file in the debugfs tracing directory.
1412 ftrace_graph_max_depth=<uint>
1413 [FTRACE] Used with the function graph tracer. This is
1414 the max depth it will trace into a function. This value
1415 can be changed at run time by the max_graph_depth file
1416 in the tracefs tracing directory. default: 0 (no limit)
1418 fw_devlink= [KNL] Create device links between consumer and supplier
1419 devices by scanning the firmware to infer the
1420 consumer/supplier relationships. This feature is
1421 especially useful when drivers are loaded as modules as
1422 it ensures proper ordering of tasks like device probing
1423 (suppliers first, then consumers), supplier boot state
1424 clean up (only after all consumers have probed),
1425 suspend/resume & runtime PM (consumers first, then
1427 Format: { off | permissive | on | rpm }
1428 off -- Don't create device links from firmware info.
1429 permissive -- Create device links from firmware info
1430 but use it only for ordering boot state clean
1431 up (sync_state() calls).
1432 on -- Create device links from firmware info and use it
1433 to enforce probe and suspend/resume ordering.
1434 rpm -- Like "on", but also use to order runtime PM.
1437 [HW,JOY] Multisystem joystick and NES/SNES/PSX pad
1438 support via parallel port (up to 5 devices per port)
1439 Format: <port#>,<pad1>,<pad2>,<pad3>,<pad4>,<pad5>
1440 See also Documentation/input/devices/joystick-parport.rst
1444 gart_fix_e820= [X86-64] disable the fix e820 for K8 GART
1448 gcov_persist= [GCOV] When non-zero (default), profiling data for
1449 kernel modules is saved and remains accessible via
1450 debugfs, even when the module is unloaded/reloaded.
1451 When zero, profiling data is discarded and associated
1452 debugfs files are removed at module unload time.
1454 goldfish [X86] Enable the goldfish android emulator platform.
1455 Don't use this when you are not running on the
1458 gpt [EFI] Forces disk with valid GPT signature but
1459 invalid Protective MBR to be treated as GPT. If the
1460 primary GPT is corrupted, it enables the backup/alternate
1461 GPT to be used instead.
1463 grcan.enable0= [HW] Configuration of physical interface 0. Determines
1464 the "Enable 0" bit of the configuration register.
1467 grcan.enable1= [HW] Configuration of physical interface 1. Determines
1468 the "Enable 0" bit of the configuration register.
1471 grcan.select= [HW] Select which physical interface to use.
1474 grcan.txsize= [HW] Sets the size of the tx buffer.
1475 Format: <unsigned int> such that (txsize & ~0x1fffc0) == 0.
1477 grcan.rxsize= [HW] Sets the size of the rx buffer.
1478 Format: <unsigned int> such that (rxsize & ~0x1fffc0) == 0.
1481 gpio-mockup.gpio_mockup_ranges
1482 [HW] Sets the ranges of gpiochip of for this device.
1483 Format: <start1>,<end1>,<start2>,<end2>...
1485 hardlockup_all_cpu_backtrace=
1486 [KNL] Should the hard-lockup detector generate
1487 backtraces on all cpus.
1490 hashdist= [KNL,NUMA] Large hashes allocated during boot
1491 are distributed across NUMA nodes. Defaults on
1492 for 64-bit NUMA, off otherwise.
1493 Format: 0 | 1 (for off | on)
1495 hcl= [IA-64] SGI's Hardware Graph compatibility layer
1497 hd= [EIDE] (E)IDE hard drive subsystem geometry
1498 Format: <cyl>,<head>,<sect>
1501 Disable Hardware Error Source Table (HEST) support;
1502 corresponding firmware-first mode error processing
1503 logic will be disabled.
1505 highmem=nn[KMG] [KNL,BOOT] forces the highmem zone to have an exact
1506 size of <nn>. This works even on boxes that have no
1507 highmem otherwise. This also works to reduce highmem
1508 size on bigger boxes.
1510 highres= [KNL] Enable/disable high resolution timer mode.
1511 Valid parameters: "on", "off"
1516 hpet= [X86-32,HPET] option to control HPET usage
1517 Format: { enable (default) | disable | force |
1519 disable: disable HPET and use PIT instead
1520 force: allow force enabled of undocumented chips (ICH4,
1522 verbose: show contents of HPET registers during setup
1524 hpet_mmap= [X86, HPET_MMAP] Allow userspace to mmap HPET
1525 registers. Default set by CONFIG_HPET_MMAP_DEFAULT.
1527 hugetlb_cma= [HW] The size of a cma area used for allocation
1528 of gigantic hugepages.
1531 Reserve a cma area of given size and allocate gigantic
1532 hugepages using the cma allocator. If enabled, the
1533 boot-time allocation of gigantic hugepages is skipped.
1535 hugepages= [HW] Number of HugeTLB pages to allocate at boot.
1536 If this follows hugepagesz (below), it specifies
1537 the number of pages of hugepagesz to be allocated.
1538 If this is the first HugeTLB parameter on the command
1539 line, it specifies the number of pages to allocate for
1540 the default huge page size. See also
1541 Documentation/admin-guide/mm/hugetlbpage.rst.
1545 [HW] The size of the HugeTLB pages. This is used in
1546 conjunction with hugepages (above) to allocate huge
1547 pages of a specific size at boot. The pair
1548 hugepagesz=X hugepages=Y can be specified once for
1549 each supported huge page size. Huge page sizes are
1550 architecture dependent. See also
1551 Documentation/admin-guide/mm/hugetlbpage.rst.
1555 [KNL] Should the hung task detector generate panics.
1558 A value of 1 instructs the kernel to panic when a
1559 hung task is detected. The default value is controlled
1560 by the CONFIG_BOOTPARAM_HUNG_TASK_PANIC build-time
1561 option. The value selected by this boot parameter can
1562 be changed later by the kernel.hung_task_panic sysctl.
1564 hvc_iucv= [S390] Number of z/VM IUCV hypervisor console (HVC)
1565 terminal devices. Valid values: 0..8
1566 hvc_iucv_allow= [S390] Comma-separated list of z/VM user IDs.
1567 If specified, z/VM IUCV HVC accepts connections
1568 from listed z/VM user IDs only.
1570 hv_nopvspin [X86,HYPER_V] Disables the paravirt spinlock optimizations
1571 which allow the hypervisor to 'idle' the
1572 guest on lock contention.
1575 Do not unregister boot console at start. This is only
1576 useful for debugging when something happens in the window
1577 between unregistering the boot console and initializing
1580 i2c_bus= [HW] Override the default board specific I2C bus speed
1581 or register an additional I2C bus that is not
1582 registered from board initialization code.
1586 i8042.debug [HW] Toggle i8042 debug mode
1587 i8042.unmask_kbd_data
1588 [HW] Enable printing of interrupt data from the KBD port
1589 (disabled by default, and as a pre-condition
1590 requires that i8042.debug=1 be enabled)
1591 i8042.direct [HW] Put keyboard port into non-translated mode
1592 i8042.dumbkbd [HW] Pretend that controller can only read data from
1593 keyboard and cannot control its state
1594 (Don't attempt to blink the leds)
1595 i8042.noaux [HW] Don't check for auxiliary (== mouse) port
1596 i8042.nokbd [HW] Don't check/create keyboard port
1597 i8042.noloop [HW] Disable the AUX Loopback command while probing
1599 i8042.nomux [HW] Don't check presence of an active multiplexing
1601 i8042.nopnp [HW] Don't use ACPIPnP / PnPBIOS to discover KBD/AUX
1603 i8042.notimeout [HW] Ignore timeout condition signalled by controller
1604 i8042.reset [HW] Reset the controller during init, cleanup and
1605 suspend-to-ram transitions, only during s2r
1606 transitions, or never reset
1607 Format: { 1 | Y | y | 0 | N | n }
1608 1, Y, y: always reset controller
1609 0, N, n: don't ever reset controller
1610 Default: only on s2r transitions on x86; most other
1611 architectures force reset to be always executed
1612 i8042.unlock [HW] Unlock (ignore) the keylock
1613 i8042.kbdreset [HW] Reset device connected to KBD port
1617 i8k.ignore_dmi [HW] Continue probing hardware even if DMI data
1618 indicates that the driver is running on unsupported
1620 i8k.force [HW] Activate i8k driver even if SMM BIOS signature
1621 does not match list of supported models.
1623 [HW] Report power status in /proc/i8k
1624 (disabled by default)
1625 i8k.restricted [HW] Allow controlling fans only if SYS_ADMIN
1628 i915.invert_brightness=
1629 [DRM] Invert the sense of the variable that is used to
1630 set the brightness of the panel backlight. Normally a
1631 brightness value of 0 indicates backlight switched off,
1632 and the maximum of the brightness value sets the backlight
1633 to maximum brightness. If this parameter is set to 0
1634 (default) and the machine requires it, or this parameter
1635 is set to 1, a brightness value of 0 sets the backlight
1636 to maximum brightness, and the maximum of the brightness
1637 value switches the backlight off.
1638 -1 -- never invert brightness
1639 0 -- machine default
1640 1 -- force brightness inversion
1643 Format: <io>[,<membase>[,<icn_id>[,<icn_id2>]]]
1645 ide-core.nodma= [HW] (E)IDE subsystem
1646 Format: =0.0 to prevent dma on hda, =0.1 hdb =1.0 hdc
1647 .vlb_clock .pci_clock .noflush .nohpa .noprobe .nowerr
1648 .cdrom .chs .ignore_cable are additional options
1649 See Documentation/ide/ide.rst.
1651 ide-generic.probe-mask= [HW] (E)IDE subsystem
1653 Probe mask for legacy ISA IDE ports. Depending on
1654 platform up to 6 ports are supported, enabled by
1655 setting corresponding bits in the mask to 1. The
1656 default value is 0x0, which has a special meaning.
1657 On systems that have PCI, it triggers scanning the
1658 PCI bus for the first and the second port, which
1659 are then probed. On systems without PCI the value
1660 of 0x0 enables probing the two first ports as if it
1663 ide-pci-generic.all-generic-ide [HW] (E)IDE subsystem
1664 Claim all unknown PCI IDE storage controllers.
1667 Format: idle=poll, idle=halt, idle=nomwait
1668 Poll forces a polling idle loop that can slightly
1669 improve the performance of waking up a idle CPU, but
1670 will use a lot of power and make the system run hot.
1672 idle=halt: Halt is forced to be used for CPU idle.
1673 In such case C2/C3 won't be used again.
1674 idle=nomwait: Disable mwait for CPU C-states
1676 ieee754= [MIPS] Select IEEE Std 754 conformance mode
1677 Format: { strict | legacy | 2008 | relaxed }
1680 Choose which programs will be accepted for execution
1681 based on the IEEE 754 NaN encoding(s) supported by
1682 the FPU and the NaN encoding requested with the value
1683 of an ELF file header flag individually set by each
1684 binary. Hardware implementations are permitted to
1685 support either or both of the legacy and the 2008 NaN
1688 Available settings are as follows:
1689 strict accept binaries that request a NaN encoding
1690 supported by the FPU
1691 legacy only accept legacy-NaN binaries, if supported
1693 2008 only accept 2008-NaN binaries, if supported
1695 relaxed accept any binaries regardless of whether
1696 supported by the FPU
1698 The FPU emulator is always able to support both NaN
1699 encodings, so if no FPU hardware is present or it has
1700 been disabled with 'nofpu', then the settings of
1701 'legacy' and '2008' strap the emulator accordingly,
1702 'relaxed' straps the emulator for both legacy-NaN and
1703 2008-NaN, whereas 'strict' enables legacy-NaN only on
1704 legacy processors and both NaN encodings on MIPS32 or
1707 The setting for ABS.fmt/NEG.fmt instruction execution
1708 mode generally follows that for the NaN encoding,
1709 except where unsupported by hardware.
1711 ignore_loglevel [KNL]
1712 Ignore loglevel setting - this will print /all/
1713 kernel messages to the console. Useful for debugging.
1714 We also add it as printk module parameter, so users
1715 could change it dynamically, usually by
1716 /sys/module/printk/parameters/ignore_loglevel.
1719 Ignore RLIMIT_DATA setting for data mappings,
1720 print warning at first misuse. Can be changed via
1721 /sys/module/kernel/parameters/ignore_rlimit_data.
1723 ihash_entries= [KNL]
1724 Set number of hash buckets for inode cache.
1726 ima_appraise= [IMA] appraise integrity measurements
1727 Format: { "off" | "enforce" | "fix" | "log" }
1730 ima_appraise_tcb [IMA] Deprecated. Use ima_policy= instead.
1731 The builtin appraise policy appraises all files
1734 ima_canonical_fmt [IMA]
1735 Use the canonical format for the binary runtime
1736 measurements, instead of host native format.
1739 Format: { md5 | sha1 | rmd160 | sha256 | sha384
1743 The list of supported hash algorithms is defined
1744 in crypto/hash_info.h.
1747 The builtin policies to load during IMA setup.
1748 Format: "tcb | appraise_tcb | secure_boot |
1751 The "tcb" policy measures all programs exec'd, files
1752 mmap'd for exec, and all files opened with the read
1753 mode bit set by either the effective uid (euid=0) or
1756 The "appraise_tcb" policy appraises the integrity of
1757 all files owned by root.
1759 The "secure_boot" policy appraises the integrity
1760 of files (eg. kexec kernel image, kernel modules,
1761 firmware, policy, etc) based on file signatures.
1763 The "fail_securely" policy forces file signature
1764 verification failure also on privileged mounted
1765 filesystems with the SB_I_UNVERIFIABLE_SIGNATURE
1768 ima_tcb [IMA] Deprecated. Use ima_policy= instead.
1769 Load a policy which meets the needs of the Trusted
1770 Computing Base. This means IMA will measure all
1771 programs exec'd, files mmap'd for exec, and all files
1772 opened for read by uid=0.
1775 Select one of defined IMA measurements template formats.
1776 Formats: { "ima" | "ima-ng" | "ima-sig" }
1780 [IMA] Define a custom template format.
1781 Format: { "field1|...|fieldN" }
1783 ima.ahash_minsize= [IMA] Minimum file size for asynchronous hash usage
1784 Format: <min_file_size>
1785 Set the minimal file size for using asynchronous hash.
1786 If left unspecified, ahash usage is disabled.
1788 ahash performance varies for different data sizes on
1789 different crypto accelerators. This option can be used
1790 to achieve the best performance for a particular HW.
1792 ima.ahash_bufsize= [IMA] Asynchronous hash buffer size
1794 Set hashing buffer size. Default: 4k.
1796 ahash performance varies for different chunk sizes on
1797 different crypto accelerators. This option can be used
1798 to achieve best performance for particular HW.
1802 Run specified binary instead of /sbin/init as init
1805 initcall_debug [KNL] Trace initcalls as they are executed. Useful
1806 for working out where the kernel is dying during
1809 initcall_blacklist= [KNL] Do not execute a comma-separated list of
1810 initcall functions. Useful for debugging built-in
1811 modules and initcalls.
1813 initrd= [BOOT] Specify the location of the initial ramdisk
1815 initrdmem= [KNL] Specify a physical address and size from which to
1816 load the initrd. If an initrd is compiled in or
1817 specified in the bootparams, it takes priority over this
1819 Format: ss[KMG],nn[KMG]
1822 init_on_alloc= [MM] Fill newly allocated pages and heap objects with
1825 Default set by CONFIG_INIT_ON_ALLOC_DEFAULT_ON.
1827 init_on_free= [MM] Fill freed pages and heap objects with zeroes.
1829 Default set by CONFIG_INIT_ON_FREE_DEFAULT_ON.
1831 init_pkru= [X86] Specify the default memory protection keys rights
1832 register contents for all processes. 0x55555554 by
1833 default (disallow access to all but pkey 0). Can
1834 override in debugfs after boot.
1836 inport.irq= [HW] Inport (ATI XL and Microsoft) busmouse driver
1839 int_pln_enable [X86] Enable power limit notification interrupt
1841 integrity_audit=[IMA]
1842 Format: { "0" | "1" }
1843 0 -- basic integrity auditing messages. (Default)
1844 1 -- additional integrity auditing messages.
1846 intel_iommu= [DMAR] Intel IOMMU driver (DMAR) option
1848 Enable intel iommu driver.
1850 Disable intel iommu driver.
1851 igfx_off [Default Off]
1852 By default, gfx is mapped as normal device. If a gfx
1853 device has a dedicated DMAR unit, the DMAR unit is
1854 bypassed by not enabling DMAR with this option. In
1855 this case, gfx device will use physical address for
1858 With this option iommu will not optimize to look
1859 for io virtual address below 32-bit forcing dual
1860 address cycle on pci bus for cards supporting greater
1861 than 32-bit addressing. The default is to look
1862 for translation below 32-bit and if not available
1863 then look in the higher range.
1864 strict [Default Off]
1865 With this option on every unmap_single operation will
1866 result in a hardware IOTLB flush operation as opposed
1867 to batching them for performance.
1868 sp_off [Default Off]
1869 By default, super page will be supported if Intel IOMMU
1870 has the capability. With this option, super page will
1873 By default, scalable mode will be disabled even if the
1874 hardware advertises that it has support for the scalable
1875 mode translation. With this option set, scalable mode
1876 will be used on hardware which claims to support it.
1877 tboot_noforce [Default Off]
1878 Do not force the Intel IOMMU enabled under tboot.
1879 By default, tboot will force Intel IOMMU on, which
1880 could harm performance of some high-throughput
1881 devices like 40GBit network cards, even if identity
1883 Note that using this option lowers the security
1884 provided by tboot because it makes the system
1885 vulnerable to DMA attacks.
1887 intel_idle.max_cstate= [KNL,HW,ACPI,X86]
1888 0 disables intel_idle and fall back on acpi_idle.
1889 1 to 9 specify maximum depth of C-state.
1893 Do not enable intel_pstate as the default
1894 scaling driver for the supported processors
1896 Use intel_pstate as a scaling driver, but configure it
1897 to work with generic cpufreq governors (instead of
1898 enabling its internal governor). This mode cannot be
1899 used along with the hardware-managed P-states (HWP)
1902 Enable intel_pstate on systems that prohibit it by default
1903 in favor of acpi-cpufreq. Forcing the intel_pstate driver
1904 instead of acpi-cpufreq may disable platform features, such
1905 as thermal controls and power capping, that rely on ACPI
1906 P-States information being indicated to OSPM and therefore
1907 should be used with caution. This option does not work with
1908 processors that aren't supported by the intel_pstate driver
1909 or on platforms that use pcc-cpufreq instead of acpi-cpufreq.
1911 Do not enable hardware P state control (HWP)
1914 Only load intel_pstate on systems which support
1915 hardware P state control (HWP) if available.
1917 Enforce ACPI _PPC performance limits. If the Fixed ACPI
1918 Description Table, specifies preferred power management
1919 profile as "Enterprise Server" or "Performance Server",
1920 then this feature is turned on by default.
1922 Allow per-logical-CPU P-State performance control limits using
1923 cpufreq sysfs interface
1925 intremap= [X86-64, Intel-IOMMU]
1926 on enable Interrupt Remapping (default)
1927 off disable Interrupt Remapping
1928 nosid disable Source ID checking
1930 BIOS x2APIC opt-out request will be ignored
1931 nopost disable Interrupt Posting
1933 iomem= Disable strict checking of access to MMIO memory
1934 strict regions from userspace.
1949 nobypass [PPC/POWERNV]
1950 Disable IOMMU bypass, using IOMMU for PCI devices.
1952 iommu.strict= [ARM64] Configure TLB invalidation behaviour
1953 Format: { "0" | "1" }
1955 Request that DMA unmap operations use deferred
1956 invalidation of hardware TLBs, for increased
1957 throughput at the cost of reduced device isolation.
1958 Will fall back to strict mode if not supported by
1959 the relevant IOMMU driver.
1960 1 - Strict mode (default).
1961 DMA unmap operations invalidate IOMMU hardware TLBs
1965 [ARM64, X86] Configure DMA to bypass the IOMMU by default.
1966 Format: { "0" | "1" }
1967 0 - Use IOMMU translation for DMA.
1968 1 - Bypass the IOMMU for DMA.
1969 unset - Use value of CONFIG_IOMMU_DEFAULT_PASSTHROUGH.
1971 io7= [HW] IO7 for Marvel-based Alpha systems
1972 See comment before marvel_specify_io7 in
1973 arch/alpha/kernel/core_marvel.c.
1975 io_delay= [X86] I/O delay method
1977 Standard port 0x80 based delay
1979 Alternate port 0xed based delay (needed on some systems)
1981 Simple two microseconds delay
1986 See Documentation/admin-guide/nfs/nfsroot.rst.
1988 ipcmni_extend [KNL] Extend the maximum number of unique System V
1989 IPC identifiers from 32,768 to 16,777,216.
1991 irqaffinity= [SMP] Set the default irq affinity mask
1992 The argument is a cpu list, as described above.
1994 irqchip.gicv2_force_probe=
1997 Force the kernel to look for the second 4kB page
1998 of a GICv2 controller even if the memory range
1999 exposed by the device tree is too small.
2001 irqchip.gicv3_nolpi=
2003 Force the kernel to ignore the availability of
2004 LPIs (and by consequence ITSs). Intended for system
2005 that use the kernel as a bootloader, and thus want
2006 to let secondary kernels in charge of setting up
2009 irqchip.gicv3_pseudo_nmi= [ARM64]
2010 Enables support for pseudo-NMIs in the kernel. This
2011 requires the kernel to be built with
2012 CONFIG_ARM64_PSEUDO_NMI.
2015 When an interrupt is not handled search all handlers
2016 for it. Intended to get systems with badly broken
2020 When an interrupt is not handled search all handlers
2021 for it. Also check all handlers each timer
2022 interrupt. Intended to get systems with badly broken
2026 Format: <RDP>,<reset>,<pci_scan>,<verbosity>
2028 isolcpus= [KNL,SMP,ISOL] Isolate a given set of CPUs from disturbance.
2029 [Deprecated - use cpusets instead]
2030 Format: [flag-list,]<cpu-list>
2032 Specify one or more CPUs to isolate from disturbances
2033 specified in the flag list (default: domain):
2036 Disable the tick when a single task runs.
2038 A residual 1Hz tick is offloaded to workqueues, which you
2039 need to affine to housekeeping through the global
2040 workqueue's affinity configured via the
2041 /sys/devices/virtual/workqueue/cpumask sysfs file, or
2042 by using the 'domain' flag described below.
2044 NOTE: by default the global workqueue runs on all CPUs,
2045 so to protect individual CPUs the 'cpumask' file has to
2046 be configured manually after bootup.
2049 Isolate from the general SMP balancing and scheduling
2050 algorithms. Note that performing domain isolation this way
2051 is irreversible: it's not possible to bring back a CPU to
2052 the domains once isolated through isolcpus. It's strongly
2053 advised to use cpusets instead to disable scheduler load
2054 balancing through the "cpuset.sched_load_balance" file.
2055 It offers a much more flexible interface where CPUs can
2056 move in and out of an isolated set anytime.
2058 You can move a process onto or off an "isolated" CPU via
2059 the CPU affinity syscalls or cpuset.
2060 <cpu number> begins at 0 and the maximum value is
2061 "number of CPUs in system - 1".
2065 Isolate from being targeted by managed interrupts
2066 which have an interrupt mask containing isolated
2067 CPUs. The affinity of managed interrupts is
2068 handled by the kernel and cannot be changed via
2069 the /proc/irq/* interfaces.
2071 This isolation is best effort and only effective
2072 if the automatically assigned interrupt mask of a
2073 device queue contains isolated and housekeeping
2074 CPUs. If housekeeping CPUs are online then such
2075 interrupts are directed to the housekeeping CPU
2076 so that IO submitted on the housekeeping CPU
2077 cannot disturb the isolated CPU.
2079 If a queue's affinity mask contains only isolated
2080 CPUs then this parameter has no effect on the
2081 interrupt routing decision, though interrupts are
2082 only delivered when tasks running on those
2083 isolated CPUs submit IO. IO submitted on
2084 housekeeping CPUs has no influence on those
2087 The format of <cpu-list> is described above.
2091 ivrs_ioapic [HW,X86-64]
2092 Provide an override to the IOAPIC-ID<->DEVICE-ID
2093 mapping provided in the IVRS ACPI table. For
2094 example, to map IOAPIC-ID decimal 10 to
2095 PCI device 00:14.0 write the parameter as:
2096 ivrs_ioapic[10]=00:14.0
2098 ivrs_hpet [HW,X86-64]
2099 Provide an override to the HPET-ID<->DEVICE-ID
2100 mapping provided in the IVRS ACPI table. For
2101 example, to map HPET-ID decimal 0 to
2102 PCI device 00:14.0 write the parameter as:
2103 ivrs_hpet[0]=00:14.0
2105 ivrs_acpihid [HW,X86-64]
2106 Provide an override to the ACPI-HID:UID<->DEVICE-ID
2107 mapping provided in the IVRS ACPI table. For
2108 example, to map UART-HID:UID AMD0020:0 to
2109 PCI device 00:14.5 write the parameter as:
2110 ivrs_acpihid[00:14.5]=AMD0020:0
2112 js= [HW,JOY] Analog joystick
2113 See Documentation/input/joydev/joystick.rst.
2116 When CONFIG_RANDOMIZE_BASE is set, this disables
2117 kernel and module base offset ASLR (Address Space
2118 Layout Randomization).
2121 [KNL] Enforce KASAN (Kernel Address Sanitizer) to print
2122 report on every invalid memory access. Without this
2123 parameter KASAN will print report only for the first
2128 kernelcore= [KNL,X86,IA-64,PPC]
2129 Format: nn[KMGTPE] | nn% | "mirror"
2130 This parameter specifies the amount of memory usable by
2131 the kernel for non-movable allocations. The requested
2132 amount is spread evenly throughout all nodes in the
2133 system as ZONE_NORMAL. The remaining memory is used for
2134 movable memory in its own zone, ZONE_MOVABLE. In the
2135 event, a node is too small to have both ZONE_NORMAL and
2136 ZONE_MOVABLE, kernelcore memory will take priority and
2137 other nodes will have a larger ZONE_MOVABLE.
2139 ZONE_MOVABLE is used for the allocation of pages that
2140 may be reclaimed or moved by the page migration
2141 subsystem. Note that allocations like PTEs-from-HighMem
2142 still use the HighMem zone if it exists, and the Normal
2143 zone if it does not.
2145 It is possible to specify the exact amount of memory in
2146 the form of "nn[KMGTPE]", a percentage of total system
2147 memory in the form of "nn%", or "mirror". If "mirror"
2148 option is specified, mirrored (reliable) memory is used
2149 for non-movable allocations and remaining memory is used
2150 for Movable pages. "nn[KMGTPE]", "nn%", and "mirror"
2151 are exclusive, so you cannot specify multiple forms.
2153 kgdbdbgp= [KGDB,HW] kgdb over EHCI usb debug port.
2154 Format: <Controller#>[,poll interval]
2155 The controller # is the number of the ehci usb debug
2156 port as it is probed via PCI. The poll interval is
2157 optional and is the number seconds in between
2158 each poll cycle to the debug port in case you need
2159 the functionality for interrupting the kernel with
2160 gdb or control-c on the dbgp connection. When
2161 not using this parameter you use sysrq-g to break into
2162 the kernel debugger.
2164 kgdboc= [KGDB,HW] kgdb over consoles.
2165 Requires a tty driver that supports console polling,
2166 or a supported polling keyboard driver (non-usb).
2167 Serial only format: <serial_device>[,baud]
2168 keyboard only format: kbd
2169 keyboard and serial format: kbd,<serial_device>[,baud]
2170 Optional Kernel mode setting:
2171 kms, kbd format: kms,kbd
2172 kms, kbd and serial format: kms,kbd,<ser_dev>[,baud]
2174 kgdboc_earlycon= [KGDB,HW]
2175 If the boot console provides the ability to read
2176 characters and can work in polling mode, you can use
2177 this parameter to tell kgdb to use it as a backend
2178 until the normal console is registered. Intended to
2179 be used together with the kgdboc parameter which
2180 specifies the normal console to transition to.
2182 The name of the early console should be specified
2183 as the value of this parameter. Note that the name of
2184 the early console might be different than the tty
2185 name passed to kgdboc. It's OK to leave the value
2186 blank and the first boot console that implements
2187 read() will be picked.
2189 kgdbwait [KGDB] Stop kernel execution and enter the
2190 kernel debugger at the earliest opportunity.
2192 kmac= [MIPS] Korina ethernet MAC address.
2193 Configure the RouterBoard 532 series on-chip
2194 Ethernet adapter MAC address.
2196 kmemleak= [KNL] Boot-time kmemleak enable/disable
2197 Valid arguments: on, off
2199 Built with CONFIG_DEBUG_KMEMLEAK_DEFAULT_OFF=y,
2202 kprobe_event=[probe-list]
2203 [FTRACE] Add kprobe events and enable at boot time.
2204 The probe-list is a semicolon delimited list of probe
2205 definitions. Each definition is same as kprobe_events
2206 interface, but the parameters are comma delimited.
2207 For example, to add a kprobe event on vfs_read with
2208 arg1 and arg2, add to the command line;
2210 kprobe_event=p,vfs_read,$arg1,$arg2
2212 See also Documentation/trace/kprobetrace.rst "Kernel
2213 Boot Parameter" section.
2215 kpti= [ARM64] Control page table isolation of user
2216 and kernel address spaces.
2217 Default: enabled on cores which need mitigation.
2221 kvm.ignore_msrs=[KVM] Ignore guest accesses to unhandled MSRs.
2222 Default is 0 (don't ignore, but inject #GP)
2224 kvm.enable_vmware_backdoor=[KVM] Support VMware backdoor PV interface.
2225 Default is false (don't support).
2227 kvm.mmu_audit= [KVM] This is a R/W parameter which allows audit
2232 [KVM] Controls the software workaround for the
2233 X86_BUG_ITLB_MULTIHIT bug.
2234 force : Always deploy workaround.
2235 off : Never deploy workaround.
2236 auto : Deploy workaround based on the presence of
2237 X86_BUG_ITLB_MULTIHIT.
2241 If the software workaround is enabled for the host,
2242 guests do need not to enable it for nested guests.
2244 kvm.nx_huge_pages_recovery_ratio=
2245 [KVM] Controls how many 4KiB pages are periodically zapped
2246 back to huge pages. 0 disables the recovery, otherwise if
2247 the value is N KVM will zap 1/Nth of the 4KiB pages every
2248 minute. The default is 60.
2250 kvm-amd.nested= [KVM,AMD] Allow nested virtualization in KVM/SVM.
2251 Default is 1 (enabled)
2253 kvm-amd.npt= [KVM,AMD] Disable nested paging (virtualized MMU)
2255 Default is 1 (enabled) if in 64-bit or 32-bit PAE mode.
2258 [KVM,ARM] Select one of KVM/arm64's modes of operation.
2260 protected: nVHE-based mode with support for guests whose
2261 state is kept private from the host.
2262 Not valid if the kernel is running in EL2.
2264 Defaults to VHE/nVHE based on hardware support and
2265 the value of CONFIG_ARM64_VHE.
2267 kvm-arm.vgic_v3_group0_trap=
2268 [KVM,ARM] Trap guest accesses to GICv3 group-0
2271 kvm-arm.vgic_v3_group1_trap=
2272 [KVM,ARM] Trap guest accesses to GICv3 group-1
2275 kvm-arm.vgic_v3_common_trap=
2276 [KVM,ARM] Trap guest accesses to GICv3 common
2279 kvm-arm.vgic_v4_enable=
2280 [KVM,ARM] Allow use of GICv4 for direct injection of
2283 kvm_cma_resv_ratio=n [PPC]
2284 Reserves given percentage from system memory area for
2285 contiguous memory allocation for KVM hash pagetable
2287 By default it reserves 5% of total system memory.
2291 kvm-intel.ept= [KVM,Intel] Disable extended page tables
2292 (virtualized MMU) support on capable Intel chips.
2293 Default is 1 (enabled)
2295 kvm-intel.emulate_invalid_guest_state=
2296 [KVM,Intel] Enable emulation of invalid guest states
2297 Default is 0 (disabled)
2299 kvm-intel.flexpriority=
2300 [KVM,Intel] Disable FlexPriority feature (TPR shadow).
2301 Default is 1 (enabled)
2304 [KVM,Intel] Enable VMX nesting (nVMX).
2305 Default is 0 (disabled)
2307 kvm-intel.unrestricted_guest=
2308 [KVM,Intel] Disable unrestricted guest feature
2309 (virtualized real and unpaged mode) on capable
2310 Intel chips. Default is 1 (enabled)
2312 kvm-intel.vmentry_l1d_flush=[KVM,Intel] Mitigation for L1 Terminal Fault
2315 Valid arguments: never, cond, always
2317 always: L1D cache flush on every VMENTER.
2318 cond: Flush L1D on VMENTER only when the code between
2319 VMEXIT and VMENTER can leak host memory.
2320 never: Disables the mitigation
2322 Default is cond (do L1 cache flush in specific instances)
2324 kvm-intel.vpid= [KVM,Intel] Disable Virtual Processor Identification
2325 feature (tagged TLBs) on capable Intel chips.
2326 Default is 1 (enabled)
2328 l1tf= [X86] Control mitigation of the L1TF vulnerability on
2331 The kernel PTE inversion protection is unconditionally
2332 enabled and cannot be disabled.
2335 Provides all available mitigations for the
2336 L1TF vulnerability. Disables SMT and
2337 enables all mitigations in the
2338 hypervisors, i.e. unconditional L1D flush.
2340 SMT control and L1D flush control via the
2341 sysfs interface is still possible after
2342 boot. Hypervisors will issue a warning
2343 when the first VM is started in a
2344 potentially insecure configuration,
2345 i.e. SMT enabled or L1D flush disabled.
2348 Same as 'full', but disables SMT and L1D
2349 flush runtime control. Implies the
2350 'nosmt=force' command line option.
2351 (i.e. sysfs control of SMT is disabled.)
2354 Leaves SMT enabled and enables the default
2355 hypervisor mitigation, i.e. conditional
2358 SMT control and L1D flush control via the
2359 sysfs interface is still possible after
2360 boot. Hypervisors will issue a warning
2361 when the first VM is started in a
2362 potentially insecure configuration,
2363 i.e. SMT enabled or L1D flush disabled.
2367 Disables SMT and enables the default
2368 hypervisor mitigation.
2370 SMT control and L1D flush control via the
2371 sysfs interface is still possible after
2372 boot. Hypervisors will issue a warning
2373 when the first VM is started in a
2374 potentially insecure configuration,
2375 i.e. SMT enabled or L1D flush disabled.
2378 Same as 'flush', but hypervisors will not
2379 warn when a VM is started in a potentially
2380 insecure configuration.
2383 Disables hypervisor mitigations and doesn't
2385 It also drops the swap size and available
2386 RAM limit restriction on both hypervisor and
2391 For details see: Documentation/admin-guide/hw-vuln/l1tf.rst
2397 lapic [X86-32,APIC] Enable the local APIC even if BIOS
2400 lapic= [X86,APIC] Do not use TSC deadline
2401 value for LAPIC timer one-shot implementation. Default
2402 back to the programmable timer unit in the LAPIC.
2403 Format: notscdeadline
2405 lapic_timer_c2_ok [X86,APIC] trust the local apic timer
2408 libata.dma= [LIBATA] DMA control
2409 libata.dma=0 Disable all PATA and SATA DMA
2410 libata.dma=1 PATA and SATA Disk DMA only
2411 libata.dma=2 ATAPI (CDROM) DMA only
2412 libata.dma=4 Compact Flash DMA only
2413 Combinations also work, so libata.dma=3 enables DMA
2414 for disks and CDROMs, but not CFs.
2416 libata.ignore_hpa= [LIBATA] Ignore HPA limit
2417 libata.ignore_hpa=0 keep BIOS limits (default)
2418 libata.ignore_hpa=1 ignore limits, using full disk
2420 libata.noacpi [LIBATA] Disables use of ACPI in libata suspend/resume
2424 libata.force= [LIBATA] Force configurations. The format is comma-
2425 separated list of "[ID:]VAL" where ID is
2426 PORT[.DEVICE]. PORT and DEVICE are decimal numbers
2427 matching port, link or device. Basically, it matches
2428 the ATA ID string printed on console by libata. If
2429 the whole ID part is omitted, the last PORT and DEVICE
2430 values are used. If ID hasn't been specified yet, the
2431 configuration applies to all ports, links and devices.
2433 If only DEVICE is omitted, the parameter applies to
2434 the port and all links and devices behind it. DEVICE
2435 number of 0 either selects the first device or the
2436 first fan-out link behind PMP device. It does not
2437 select the host link. DEVICE number of 15 selects the
2438 host link and device attached to it.
2440 The VAL specifies the configuration to force. As long
2441 as there's no ambiguity shortcut notation is allowed.
2442 For example, both 1.5 and 1.5G would work for 1.5Gbps.
2443 The following configurations can be forced.
2445 * Cable type: 40c, 80c, short40c, unk, ign or sata.
2446 Any ID with matching PORT is used.
2448 * SATA link speed limit: 1.5Gbps or 3.0Gbps.
2450 * Transfer mode: pio[0-7], mwdma[0-4] and udma[0-7].
2451 udma[/][16,25,33,44,66,100,133] notation is also
2454 * [no]ncq: Turn on or off NCQ.
2456 * [no]ncqtrim: Turn off queued DSM TRIM.
2458 * nohrst, nosrst, norst: suppress hard, soft
2461 * rstonce: only attempt one reset during
2462 hot-unplug link recovery
2464 * dump_id: dump IDENTIFY data.
2466 * atapi_dmadir: Enable ATAPI DMADIR bridge support
2468 * disable: Disable this device.
2470 If there are multiple matching configurations changing
2471 the same attribute, the last one is used.
2473 memblock=debug [KNL] Enable memblock debug messages.
2475 load_ramdisk= [RAM] [Deprecated]
2477 lockd.nlm_grace_period=P [NFS] Assign grace period.
2480 lockd.nlm_tcpport=N [NFS] Assign TCP port.
2483 lockd.nlm_timeout=T [NFS] Assign timeout value.
2486 lockd.nlm_udpport=M [NFS] Assign UDP port.
2489 lockdown= [SECURITY]
2490 { integrity | confidentiality }
2491 Enable the kernel lockdown feature. If set to
2492 integrity, kernel features that allow userland to
2493 modify the running kernel are disabled. If set to
2494 confidentiality, kernel features that allow userland
2495 to extract confidential information from the kernel
2498 locktorture.nreaders_stress= [KNL]
2499 Set the number of locking read-acquisition kthreads.
2500 Defaults to being automatically set based on the
2501 number of online CPUs.
2503 locktorture.nwriters_stress= [KNL]
2504 Set the number of locking write-acquisition kthreads.
2506 locktorture.onoff_holdoff= [KNL]
2507 Set time (s) after boot for CPU-hotplug testing.
2509 locktorture.onoff_interval= [KNL]
2510 Set time (s) between CPU-hotplug operations, or
2511 zero to disable CPU-hotplug testing.
2513 locktorture.shuffle_interval= [KNL]
2514 Set task-shuffle interval (jiffies). Shuffling
2515 tasks allows some CPUs to go into dyntick-idle
2516 mode during the locktorture test.
2518 locktorture.shutdown_secs= [KNL]
2519 Set time (s) after boot system shutdown. This
2520 is useful for hands-off automated testing.
2522 locktorture.stat_interval= [KNL]
2523 Time (s) between statistics printk()s.
2525 locktorture.stutter= [KNL]
2526 Time (s) to stutter testing, for example,
2527 specifying five seconds causes the test to run for
2528 five seconds, wait for five seconds, and so on.
2529 This tests the locking primitive's ability to
2530 transition abruptly to and from idle.
2532 locktorture.torture_type= [KNL]
2533 Specify the locking implementation to test.
2535 locktorture.verbose= [KNL]
2536 Enable additional printk() statements.
2538 logibm.irq= [HW,MOUSE] Logitech Bus Mouse Driver
2541 loglevel= All Kernel Messages with a loglevel smaller than the
2542 console loglevel will be printed to the console. It can
2543 also be changed with klogd or other programs. The
2544 loglevels are defined as follows:
2546 0 (KERN_EMERG) system is unusable
2547 1 (KERN_ALERT) action must be taken immediately
2548 2 (KERN_CRIT) critical conditions
2549 3 (KERN_ERR) error conditions
2550 4 (KERN_WARNING) warning conditions
2551 5 (KERN_NOTICE) normal but significant condition
2552 6 (KERN_INFO) informational
2553 7 (KERN_DEBUG) debug-level messages
2555 log_buf_len=n[KMG] Sets the size of the printk ring buffer,
2556 in bytes. n must be a power of two and greater
2557 than the minimal size. The minimal size is defined
2558 by LOG_BUF_SHIFT kernel config parameter. There is
2559 also CONFIG_LOG_CPU_MAX_BUF_SHIFT config parameter
2560 that allows to increase the default size depending on
2561 the number of CPUs. See init/Kconfig for more details.
2563 logo.nologo [FB] Disables display of the built-in Linux logo.
2564 This may be used to provide more screen space for
2565 kernel log messages and is useful when debugging
2566 kernel boot problems.
2568 lp=0 [LP] Specify parallel ports to use, e.g,
2569 lp=port[,port...] lp=none,parport0 (lp0 not configured, lp1 uses
2570 lp=reset first parallel port). 'lp=0' disables the
2571 lp=auto printer driver. 'lp=reset' (which can be
2572 specified in addition to the ports) causes
2573 attached printers to be reset. Using
2574 lp=port1,port2,... specifies the parallel ports
2575 to associate lp devices with, starting with
2576 lp0. A port specification may be 'none' to skip
2577 that lp device, or a parport name such as
2578 'parport0'. Specifying 'lp=auto' instead of a
2579 port specification list means that device IDs
2580 from each port should be examined, to see if
2581 an IEEE 1284-compliant printer is attached; if
2582 so, the driver will manage that printer.
2583 See also header of drivers/char/lp.c.
2586 Sets loops_per_jiffy to given constant, thus avoiding
2587 time-consuming boot-time autodetection (up to 250 ms per
2588 CPU). 0 enables autodetection (default). To determine
2589 the correct value for your kernel, boot with normal
2590 autodetection and see what value is printed. Note that
2591 on SMP systems the preset will be applied to all CPUs,
2592 which is likely to cause problems if your CPUs need
2593 significantly divergent settings. An incorrect value
2594 will cause delays in the kernel to be wrong, leading to
2595 unpredictable I/O errors and other breakage. Although
2596 unlikely, in the extreme case this might damage your
2600 Format: <io>,<irq>,<dma>
2602 lsm.debug [SECURITY] Enable LSM initialization debugging output.
2605 [SECURITY] Choose order of LSM initialization. This
2606 overrides CONFIG_LSM, and the "security=" parameter.
2608 machvec= [IA-64] Force the use of a particular machine-vector
2609 (machvec) in a generic kernel.
2610 Example: machvec=hpzx1
2612 machtype= [Loongson] Share the same kernel image file between
2613 different yeeloong laptops.
2614 Example: machtype=lemote-yeeloong-2f-7inch
2616 max_addr=nn[KMG] [KNL,BOOT,ia64] All physical memory greater
2617 than or equal to this physical address is ignored.
2619 maxcpus= [SMP] Maximum number of processors that an SMP kernel
2620 will bring up during bootup. maxcpus=n : n >= 0 limits
2621 the kernel to bring up 'n' processors. Surely after
2622 bootup you can bring up the other plugged cpu by executing
2623 "echo 1 > /sys/devices/system/cpu/cpuX/online". So maxcpus
2624 only takes effect during system bootup.
2625 While n=0 is a special case, it is equivalent to "nosmp",
2626 which also disables the IO APIC.
2628 max_loop= [LOOP] The number of loop block devices that get
2629 (loop.max_loop) unconditionally pre-created at init time. The default
2630 number is configured by BLK_DEV_LOOP_MIN_COUNT. Instead
2631 of statically allocating a predefined number, loop
2632 devices can be requested on-demand with the
2633 /dev/loop-control interface.
2635 mce [X86-32] Machine Check Exception
2637 mce=option [X86-64] See Documentation/x86/x86_64/boot-options.rst
2639 md= [HW] RAID subsystems devices and level
2640 See Documentation/admin-guide/md.rst.
2643 Format: <first>,<last>
2644 Specifies range of consoles to be captured by the MDA.
2647 Control mitigation for the Micro-architectural Data
2648 Sampling (MDS) vulnerability.
2650 Certain CPUs are vulnerable to an exploit against CPU
2651 internal buffers which can forward information to a
2652 disclosure gadget under certain conditions.
2654 In vulnerable processors, the speculatively
2655 forwarded data can be used in a cache side channel
2656 attack, to access data to which the attacker does
2657 not have direct access.
2659 This parameter controls the MDS mitigation. The
2662 full - Enable MDS mitigation on vulnerable CPUs
2663 full,nosmt - Enable MDS mitigation and disable
2664 SMT on vulnerable CPUs
2665 off - Unconditionally disable MDS mitigation
2667 On TAA-affected machines, mds=off can be prevented by
2668 an active TAA mitigation as both vulnerabilities are
2669 mitigated with the same mechanism so in order to disable
2670 this mitigation, you need to specify tsx_async_abort=off
2673 Not specifying this option is equivalent to
2676 For details see: Documentation/admin-guide/hw-vuln/mds.rst
2678 mem=nn[KMG] [KNL,BOOT] Force usage of a specific amount of memory
2679 Amount of memory to be used in cases as follows:
2682 2 when the kernel is not able to see the whole system memory;
2683 3 memory that lies after 'mem=' boundary is excluded from
2684 the hypervisor, then assigned to KVM guests.
2686 [X86] Work as limiting max address. Use together
2687 with memmap= to avoid physical address space collisions.
2688 Without memmap= PCI devices could be placed at addresses
2689 belonging to unused RAM.
2691 Note that this only takes effects during boot time since
2692 in above case 3, memory may need be hot added after boot
2693 if system memory of hypervisor is not sufficient.
2695 mem=nopentium [BUGS=X86-32] Disable usage of 4MB pages for kernel
2699 [KNL,SH] Allow user to override the default size for
2700 per-device physically contiguous DMA buffers.
2702 memhp_default_state=online/offline
2703 [KNL] Set the initial state for the memory hotplug
2704 onlining policy. If not specified, the default value is
2705 set according to the
2706 CONFIG_MEMORY_HOTPLUG_DEFAULT_ONLINE kernel config
2708 See Documentation/admin-guide/mm/memory-hotplug.rst.
2710 memmap=exactmap [KNL,X86] Enable setting of an exact
2711 E820 memory map, as specified by the user.
2712 Such memmap=exactmap lines can be constructed based on
2713 BIOS output or other requirements. See the memmap=nn@ss
2716 memmap=nn[KMG]@ss[KMG]
2717 [KNL, X86, MIPS, XTENSA] Force usage of a specific region of memory.
2718 Region of memory to be used is from ss to ss+nn.
2719 If @ss[KMG] is omitted, it is equivalent to mem=nn[KMG],
2720 which limits max address to nn[KMG].
2721 Multiple different regions can be specified,
2724 memmap=100M@2G,100M#3G,1G!1024G
2726 memmap=nn[KMG]#ss[KMG]
2727 [KNL,ACPI] Mark specific memory as ACPI data.
2728 Region of memory to be marked is from ss to ss+nn.
2730 memmap=nn[KMG]$ss[KMG]
2731 [KNL,ACPI] Mark specific memory as reserved.
2732 Region of memory to be reserved is from ss to ss+nn.
2733 Example: Exclude memory from 0x18690000-0x1869ffff
2734 memmap=64K$0x18690000
2736 memmap=0x10000$0x18690000
2737 Some bootloaders may need an escape character before '$',
2738 like Grub2, otherwise '$' and the following number
2741 memmap=nn[KMG]!ss[KMG]
2742 [KNL,X86] Mark specific memory as protected.
2743 Region of memory to be used, from ss to ss+nn.
2744 The memory region may be marked as e820 type 12 (0xc)
2745 and is NVDIMM or ADR memory.
2747 memmap=<size>%<offset>-<oldtype>+<newtype>
2748 [KNL,ACPI] Convert memory within the specified region
2749 from <oldtype> to <newtype>. If "-<oldtype>" is left
2750 out, the whole region will be marked as <newtype>,
2751 even if previously unavailable. If "+<newtype>" is left
2752 out, matching memory will be removed. Types are
2753 specified as e820 types, e.g., 1 = RAM, 2 = reserved,
2754 3 = ACPI, 12 = PRAM.
2756 memory_corruption_check=0/1 [X86]
2757 Some BIOSes seem to corrupt the first 64k of
2758 memory when doing things like suspend/resume.
2759 Setting this option will scan the memory
2760 looking for corruption. Enabling this will
2761 both detect corruption and prevent the kernel
2762 from using the memory being corrupted.
2763 However, its intended as a diagnostic tool; if
2764 repeatable BIOS-originated corruption always
2765 affects the same memory, you can use memmap=
2766 to prevent the kernel from using that memory.
2768 memory_corruption_check_size=size [X86]
2769 By default it checks for corruption in the low
2770 64k, making this memory unavailable for normal
2771 use. Use this parameter to scan for
2772 corruption in more or less memory.
2774 memory_corruption_check_period=seconds [X86]
2775 By default it checks for corruption every 60
2776 seconds. Use this parameter to check at some
2777 other rate. 0 disables periodic checking.
2779 memtest= [KNL,X86,ARM,PPC] Enable memtest
2781 default : 0 <disable>
2782 Specifies the number of memtest passes to be
2783 performed. Each pass selects another test
2784 pattern from a given set of patterns. Memtest
2785 fills the memory with this pattern, validates
2786 memory contents and reserves bad memory
2787 regions that are detected.
2789 mem_encrypt= [X86-64] AMD Secure Memory Encryption (SME) control
2790 Valid arguments: on, off
2791 Default (depends on kernel configuration option):
2792 on (CONFIG_AMD_MEM_ENCRYPT_ACTIVE_BY_DEFAULT=y)
2793 off (CONFIG_AMD_MEM_ENCRYPT_ACTIVE_BY_DEFAULT=n)
2794 mem_encrypt=on: Activate SME
2795 mem_encrypt=off: Do not activate SME
2797 Refer to Documentation/virt/kvm/amd-memory-encryption.rst
2798 for details on when memory encryption can be activated.
2800 mem_sleep_default= [SUSPEND] Default system suspend mode:
2801 s2idle - Suspend-To-Idle
2802 shallow - Power-On Suspend or equivalent (if supported)
2803 deep - Suspend-To-RAM or equivalent (if supported)
2804 See Documentation/admin-guide/pm/sleep-states.rst.
2806 meye.*= [HW] Set MotionEye Camera parameters
2807 See Documentation/admin-guide/media/meye.rst.
2809 mfgpt_irq= [IA-32] Specify the IRQ to use for the
2810 Multi-Function General Purpose Timers on AMD Geode
2813 mfgptfix [X86-32] Fix MFGPT timers on AMD Geode platforms when
2814 the BIOS has incorrectly applied a workaround. TinyBIOS
2815 version 0.98 is known to be affected, 0.99 fixes the
2816 problem by letting the user disable the workaround.
2820 min_addr=nn[KMG] [KNL,BOOT,ia64] All physical memory below this
2821 physical address is ignored.
2823 mini2440= [ARM,HW,KNL]
2824 Format:[0..2][b][c][t]
2826 MINI2440 configuration specification:
2827 0 - The attached screen is the 3.5" TFT
2828 1 - The attached screen is the 7" TFT
2829 2 - The VGA Shield is attached (1024x768)
2830 Leaving out the screen size parameter will not load
2831 the TFT driver, and the framebuffer will be left
2833 b - Enable backlight. The TFT backlight pin will be
2834 linked to the kernel VESA blanking code and a GPIO
2835 LED. This parameter is not necessary when using the
2837 c - Enable the s3c camera interface.
2838 t - Reserved for enabling touchscreen support. The
2839 touchscreen support is not enabled in the mainstream
2840 kernel as of 2.6.30, a preliminary port can be found
2841 in the "bleeding edge" mini2440 support kernel at
2842 https://repo.or.cz/w/linux-2.6/mini2440.git
2845 [X86,PPC,S390,ARM64] Control optional mitigations for
2846 CPU vulnerabilities. This is a set of curated,
2847 arch-independent options, each of which is an
2848 aggregation of existing arch-specific options.
2851 Disable all optional CPU mitigations. This
2852 improves system performance, but it may also
2853 expose users to several CPU vulnerabilities.
2854 Equivalent to: nopti [X86,PPC]
2856 nospectre_v1 [X86,PPC]
2858 nospectre_v2 [X86,PPC,S390,ARM64]
2859 spectre_v2_user=off [X86]
2860 spec_store_bypass_disable=off [X86,PPC]
2861 ssbd=force-off [ARM64]
2864 tsx_async_abort=off [X86]
2865 kvm.nx_huge_pages=off [X86]
2866 no_entry_flush [PPC]
2867 no_uaccess_flush [PPC]
2870 This does not have any effect on
2871 kvm.nx_huge_pages when
2872 kvm.nx_huge_pages=force.
2875 Mitigate all CPU vulnerabilities, but leave SMT
2876 enabled, even if it's vulnerable. This is for
2877 users who don't want to be surprised by SMT
2878 getting disabled across kernel upgrades, or who
2879 have other ways of avoiding SMT-based attacks.
2880 Equivalent to: (default behavior)
2883 Mitigate all CPU vulnerabilities, disabling SMT
2884 if needed. This is for users who always want to
2885 be fully mitigated, even if it means losing SMT.
2886 Equivalent to: l1tf=flush,nosmt [X86]
2887 mds=full,nosmt [X86]
2888 tsx_async_abort=full,nosmt [X86]
2891 [KNL] When CONFIG_DEBUG_MEMORY_INIT is set, this
2892 parameter allows control of the logging verbosity for
2893 the additional memory initialisation checks. A value
2894 of 0 disables mminit logging and a level of 4 will
2895 log everything. Information is printed at KERN_DEBUG
2896 so loglevel=8 may also need to be specified.
2899 [KNL] When CONFIG_MODULE_SIG is set, this means that
2900 modules without (valid) signatures will fail to load.
2901 Note that if CONFIG_MODULE_SIG_FORCE is set, that
2902 is always true, so this option does nothing.
2904 module_blacklist= [KNL] Do not load a comma-separated list of
2905 modules. Useful for debugging problem modules.
2908 [MOUSE] Maximum time between finger touching and
2909 leaving touchpad surface for touch to be considered
2910 a tap and be reported as a left button click (for
2911 touchpads working in absolute mode only).
2913 mousedev.xres= [MOUSE] Horizontal screen resolution, used for devices
2914 reporting absolute coordinates, such as tablets
2915 mousedev.yres= [MOUSE] Vertical screen resolution, used for devices
2916 reporting absolute coordinates, such as tablets
2918 movablecore= [KNL,X86,IA-64,PPC]
2919 Format: nn[KMGTPE] | nn%
2920 This parameter is the complement to kernelcore=, it
2921 specifies the amount of memory used for migratable
2922 allocations. If both kernelcore and movablecore is
2923 specified, then kernelcore will be at *least* the
2924 specified value but may be more. If movablecore on its
2925 own is specified, the administrator must be careful
2926 that the amount of memory usable for all allocations
2929 movable_node [KNL] Boot-time switch to make hotplugable memory
2930 NUMA nodes to be movable. This means that the memory
2931 of such nodes will be usable only for movable
2932 allocations which rules out almost all kernel
2933 allocations. Use with caution!
2935 MTD_Partition= [MTD]
2936 Format: <name>,<region-number>,<size>,<offset>
2938 MTD_Region= [MTD] Format:
2939 <name>,<region-number>[,<base>,<size>,<buswidth>,<altbuswidth>]
2942 See drivers/mtd/parsers/cmdlinepart.c
2944 multitce=off [PPC] This parameter disables the use of the pSeries
2945 firmware feature for updating multiple TCE entries
2948 onenand.bdry= [HW,MTD] Flex-OneNAND Boundary Configuration
2950 Format: [die0_boundary][,die0_lock][,die1_boundary][,die1_lock]
2952 boundary - index of last SLC block on Flex-OneNAND.
2953 The remaining blocks are configured as MLC blocks.
2954 lock - Configure if Flex-OneNAND boundary should be locked.
2955 Once locked, the boundary cannot be changed.
2956 1 indicates lock status, 0 indicates unlock status.
2959 ARM/S3C2412 JIVE boot control
2961 See arch/arm/mach-s3c/mach-jive.c
2963 mtouchusb.raw_coordinates=
2964 [HW] Make the MicroTouch USB driver use raw coordinates
2965 ('y', default) or cooked coordinates ('n')
2967 mtrr_chunk_size=nn[KMG] [X86]
2968 used for mtrr cleanup. It is largest continuous chunk
2969 that could hold holes aka. UC entries.
2971 mtrr_gran_size=nn[KMG] [X86]
2972 Used for mtrr cleanup. It is granularity of mtrr block.
2974 Large value could prevent small alignment from
2977 mtrr_spare_reg_nr=n [X86]
2979 Range: 0,7 : spare reg number
2981 Used for mtrr cleanup. It is spare mtrr entries number.
2982 Set to 2 or more if your graphical card needs more.
2984 n2= [NET] SDL Inc. RISCom/N2 synchronous serial card
2986 netdev= [NET] Network devices parameters
2987 Format: <irq>,<io>,<mem_start>,<mem_end>,<name>
2988 Note that mem_start is often overloaded to mean
2989 something different and driver-specific.
2990 This usage is only documented in each driver source
2994 [NETFILTER] Enable connection tracking flow accounting
2995 0 to disable accounting
2996 1 to enable accounting
2999 nfsaddrs= [NFS] Deprecated. Use ip= instead.
3000 See Documentation/admin-guide/nfs/nfsroot.rst.
3002 nfsroot= [NFS] nfs root filesystem for disk-less boxes.
3003 See Documentation/admin-guide/nfs/nfsroot.rst.
3005 nfsrootdebug [NFS] enable nfsroot debugging messages.
3006 See Documentation/admin-guide/nfs/nfsroot.rst.
3008 nfs.callback_nr_threads=
3009 [NFSv4] set the total number of threads that the
3010 NFS client will assign to service NFSv4 callback
3013 nfs.callback_tcpport=
3014 [NFS] set the TCP port on which the NFSv4 callback
3015 channel should listen.
3018 [NFS] sets the pathname to the program which is used
3019 to update the NFS client cache entries.
3021 nfs.cache_getent_timeout=
3022 [NFS] sets the timeout after which an attempt to
3023 update a cache entry is deemed to have failed.
3025 nfs.idmap_cache_timeout=
3026 [NFS] set the maximum lifetime for idmapper cache
3030 [NFS] enable 64-bit inode numbers.
3031 If zero, the NFS client will fake up a 32-bit inode
3032 number for the readdir() and stat() syscalls instead
3033 of returning the full 64-bit number.
3034 The default is to return 64-bit inode numbers.
3036 nfs.max_session_cb_slots=
3037 [NFSv4.1] Sets the maximum number of session
3038 slots the client will assign to the callback
3039 channel. This determines the maximum number of
3040 callbacks the client will process in parallel for
3041 a particular server.
3043 nfs.max_session_slots=
3044 [NFSv4.1] Sets the maximum number of session slots
3045 the client will attempt to negotiate with the server.
3046 This limits the number of simultaneous RPC requests
3047 that the client can send to the NFSv4.1 server.
3048 Note that there is little point in setting this
3049 value higher than the max_tcp_slot_table_limit.
3051 nfs.nfs4_disable_idmapping=
3052 [NFSv4] When set to the default of '1', this option
3053 ensures that both the RPC level authentication
3054 scheme and the NFS level operations agree to use
3055 numeric uids/gids if the mount is using the
3056 'sec=sys' security flavour. In effect it is
3057 disabling idmapping, which can make migration from
3058 legacy NFSv2/v3 systems to NFSv4 easier.
3059 Servers that do not support this mode of operation
3060 will be autodetected by the client, and it will fall
3061 back to using the idmapper.
3062 To turn off this behaviour, set the value to '0'.
3064 [NFS4] Specify an additional fixed unique ident-
3065 ification string that NFSv4 clients can insert into
3066 their nfs_client_id4 string. This is typically a
3067 UUID that is generated at system install time.
3069 nfs.send_implementation_id =
3070 [NFSv4.1] Send client implementation identification
3071 information in exchange_id requests.
3072 If zero, no implementation identification information
3074 The default is to send the implementation identification
3077 nfs.recover_lost_locks =
3078 [NFSv4] Attempt to recover locks that were lost due
3079 to a lease timeout on the server. Please note that
3080 doing this risks data corruption, since there are
3081 no guarantees that the file will remain unchanged
3082 after the locks are lost.
3083 If you want to enable the kernel legacy behaviour of
3084 attempting to recover these locks, then set this
3086 The default parameter value of '0' causes the kernel
3087 not to attempt recovery of lost locks.
3089 nfs4.layoutstats_timer =
3090 [NFSv4.2] Change the rate at which the kernel sends
3091 layoutstats to the pNFS metadata server.
3093 Setting this to value to 0 causes the kernel to use
3094 whatever value is the default set by the layout
3095 driver. A non-zero value sets the minimum interval
3096 in seconds between layoutstats transmissions.
3098 nfsd.nfs4_disable_idmapping=
3099 [NFSv4] When set to the default of '1', the NFSv4
3100 server will return only numeric uids and gids to
3101 clients using auth_sys, and will accept numeric uids
3102 and gids from such clients. This is intended to ease
3103 migration from NFSv2/v3.
3105 nmi_backtrace.backtrace_idle [KNL]
3106 Dump stacks even of idle CPUs in response to an
3107 NMI stack-backtrace request.
3109 nmi_debug= [KNL,SH] Specify one or more actions to take
3110 when a NMI is triggered.
3111 Format: [state][,regs][,debounce][,die]
3113 nmi_watchdog= [KNL,BUGS=X86] Debugging features for SMP kernels
3114 Format: [panic,][nopanic,][num]
3116 0 - turn hardlockup detector in nmi_watchdog off
3117 1 - turn hardlockup detector in nmi_watchdog on
3118 When panic is specified, panic when an NMI watchdog
3119 timeout occurs (or 'nopanic' to not panic on an NMI
3120 watchdog, if CONFIG_BOOTPARAM_HARDLOCKUP_PANIC is set)
3121 To disable both hard and soft lockup detectors,
3122 please see 'nowatchdog'.
3123 This is useful when you use a panic=... timeout and
3124 need the box quickly up again.
3126 These settings can be accessed at runtime via
3127 the nmi_watchdog and hardlockup_panic sysctls.
3129 netpoll.carrier_timeout=
3130 [NET] Specifies amount of time (in seconds) that
3131 netpoll should wait for a carrier. By default netpoll
3134 no387 [BUGS=X86-32] Tells the kernel to use the 387 maths
3135 emulation library even if a 387 maths coprocessor
3138 no5lvl [X86-64] Disable 5-level paging mode. Forces
3139 kernel to use 4-level paging instead.
3141 nofsgsbase [X86] Disables FSGSBASE instructions.
3144 [HW] Never suspend the console
3145 Disable suspending of consoles during suspend and
3146 hibernate operations. Once disabled, debugging
3147 messages can reach various consoles while the rest
3148 of the system is being put to sleep (ie, while
3149 debugging driver suspend/resume hooks). This may
3150 not work reliably with all consoles, but is known
3151 to work with serial and VGA consoles.
3152 To facilitate more flexible debugging, we also add
3153 console_suspend, a printk module parameter to control
3154 it. Users could use console_suspend (usually
3155 /sys/module/printk/parameters/console_suspend) to
3156 turn on/off it dynamically.
3158 novmcoredd [KNL,KDUMP]
3159 Disable device dump. Device dump allows drivers to
3160 append dump data to vmcore so you can collect driver
3161 specified debug info. Drivers can append the data
3162 without any limit and this data is stored in memory,
3163 so this may cause significant memory stress. Disabling
3164 device dump can help save memory but the driver debug
3165 data will be no longer available. This parameter
3166 is only available when CONFIG_PROC_VMCORE_DEVICE_DUMP
3169 noaliencache [MM, NUMA, SLAB] Disables the allocation of alien
3170 caches in the slab allocator. Saves per-node memory,
3171 but will impact performance.
3175 noaltinstr [S390] Disables alternative instructions patching
3176 (CPU alternatives feature).
3178 noapic [SMP,APIC] Tells the kernel to not make use of any
3179 IOAPICs that may be present in the system.
3181 noautogroup Disable scheduler automatic task group creation.
3183 nobats [PPC] Do not use BATs for mapping kernel lowmem
3184 on "Classic" PPC cores.
3188 noclflush [BUGS=X86] Don't use the CLFLUSH instruction
3190 nodelayacct [KNL] Disable per-task delay accounting
3192 nodsp [SH] Disable hardware DSP at boot time.
3194 noefi Disable EFI runtime services support.
3196 no_entry_flush [PPC] Don't flush the L1-D cache when entering the kernel.
3201 On X86-32 available only on PAE configured kernels.
3202 noexec=on: enable non-executable mappings (default)
3203 noexec=off: disable non-executable mappings
3206 Disable SMAP (Supervisor Mode Access Prevention)
3207 even if it is supported by processor.
3210 Disable SMEP (Supervisor Mode Execution Prevention)
3211 even if it is supported by processor.
3214 This affects only 32-bit executables.
3215 noexec32=on: enable non-executable mappings (default)
3216 read doesn't imply executable mappings
3217 noexec32=off: disable non-executable mappings
3218 read implies executable mappings
3220 nofpu [MIPS,SH] Disable hardware FPU at boot time.
3222 nofxsr [BUGS=X86-32] Disables x86 floating point extended
3223 register save and restore. The kernel will only save
3224 legacy floating-point registers on task switch.
3226 nohugeiomap [KNL,X86,PPC,ARM64] Disable kernel huge I/O mappings.
3228 nosmt [KNL,S390] Disable symmetric multithreading (SMT).
3229 Equivalent to smt=1.
3231 [KNL,X86] Disable symmetric multithreading (SMT).
3232 nosmt=force: Force disable SMT, cannot be undone
3233 via the sysfs control file.
3235 nospectre_v1 [X86,PPC] Disable mitigations for Spectre Variant 1
3236 (bounds check bypass). With this option data leaks are
3237 possible in the system.
3239 nospectre_v2 [X86,PPC_FSL_BOOK3E,ARM64] Disable all mitigations for
3240 the Spectre variant 2 (indirect branch prediction)
3241 vulnerability. System may allow data leaks with this
3244 nospec_store_bypass_disable
3245 [HW] Disable all mitigations for the Speculative Store Bypass vulnerability
3248 [PPC] Don't flush the L1-D cache after accessing user data.
3250 noxsave [BUGS=X86] Disables x86 extended register state save
3251 and restore using xsave. The kernel will fallback to
3252 enabling legacy floating-point and sse state.
3254 noxsaveopt [X86] Disables xsaveopt used in saving x86 extended
3255 register states. The kernel will fall back to use
3256 xsave to save the states. By using this parameter,
3257 performance of saving the states is degraded because
3258 xsave doesn't support modified optimization while
3259 xsaveopt supports it on xsaveopt enabled systems.
3261 noxsaves [X86] Disables xsaves and xrstors used in saving and
3262 restoring x86 extended register state in compacted
3263 form of xsave area. The kernel will fall back to use
3264 xsaveopt and xrstor to save and restore the states
3265 in standard form of xsave area. By using this
3266 parameter, xsave area per process might occupy more
3267 memory on xsaves enabled systems.
3269 nohlt [BUGS=ARM,SH] Tells the kernel that the sleep(SH) or
3270 wfi(ARM) instruction doesn't work correctly and not to
3271 use it. This is also useful when using JTAG debugger.
3273 no_file_caps Tells the kernel not to honor file capabilities. The
3274 only way then for a file to be executed with privilege
3275 is to be setuid root or executed by root.
3277 nohalt [IA-64] Tells the kernel not to use the power saving
3278 function PAL_HALT_LIGHT when idle. This increases
3279 power-consumption. On the positive side, it reduces
3280 interrupt wake-up latency, which may improve performance
3281 in certain environments such as networked servers or
3284 nohibernate [HIBERNATION] Disable hibernation and resume.
3286 nohz= [KNL] Boottime enable/disable dynamic ticks
3287 Valid arguments: on, off
3290 nohz_full= [KNL,BOOT,SMP,ISOL]
3291 The argument is a cpu list, as described above.
3292 In kernels built with CONFIG_NO_HZ_FULL=y, set
3293 the specified list of CPUs whose tick will be stopped
3294 whenever possible. The boot CPU will be forced outside
3295 the range to maintain the timekeeping. Any CPUs
3296 in this list will have their RCU callbacks offloaded,
3297 just as if they had also been called out in the
3298 rcu_nocbs= boot parameter.
3300 noiotrap [SH] Disables trapped I/O port accesses.
3302 noirqdebug [X86-32] Disables the code which attempts to detect and
3303 disable unhandled interrupt sources.
3305 no_timer_check [X86,APIC] Disables the code which tests for
3306 broken timer IRQ sources.
3308 noisapnp [ISAPNP] Disables ISA PnP code.
3310 noinitrd [RAM] Tells the kernel not to load any configured
3313 nointremap [X86-64, Intel-IOMMU] Do not enable interrupt
3315 [Deprecated - use intremap=off]
3319 noinvpcid [X86] Disable the INVPCID cpu feature.
3321 nojitter [IA-64] Disables jitter checking for ITC timers.
3323 no-kvmclock [X86,KVM] Disable paravirtualized KVM clock driver
3325 no-kvmapf [X86,KVM] Disable paravirtualized asynchronous page
3329 [X86,PV_OPS] Disable paravirtualized VMware scheduler
3330 clock and use the default one.
3332 no-steal-acc [X86,PV_OPS,ARM64] Disable paravirtualized steal time
3333 accounting. steal time is computed, but won't
3334 influence scheduler behaviour
3336 nolapic [X86-32,APIC] Do not enable or use the local APIC.
3338 nolapic_timer [X86-32,APIC] Do not use the local APIC timer.
3340 noltlbs [PPC] Do not use large page/tlb entries for kernel
3341 lowmem mapping on PPC40x and PPC8xx
3343 nomca [IA-64] Disable machine check abort handling
3345 nomce [X86-32] Disable Machine Check Exception
3347 nomfgpt [X86-32] Disable Multi-Function General Purpose
3348 Timer usage (for AMD Geode machines).
3350 nonmi_ipi [X86] Disable using NMI IPIs during panic/reboot to
3351 shutdown the other cpus. Instead use the REBOOT_VECTOR
3354 nomodule Disable module load
3356 nopat [X86] Disable PAT (page attribute table extension of
3357 pagetables) support.
3359 nopcid [X86-64] Disable the PCID cpu feature.
3361 norandmaps Don't use address space randomization. Equivalent to
3362 echo 0 > /proc/sys/kernel/randomize_va_space
3364 noreplace-smp [X86-32,SMP] Don't replace SMP instructions
3365 with UP alternatives
3367 nordrand [X86] Disable kernel use of the RDRAND and
3368 RDSEED instructions even if they are supported
3369 by the processor. RDRAND and RDSEED are still
3370 available to user space applications.
3372 noresume [SWSUSP] Disables resume and restores original swap
3375 no-scroll [VGA] Disables scrollback.
3376 This is required for the Braillex ib80-piezo Braille
3377 reader made by F.H. Papenmeier (Germany).
3381 nosep [BUGS=X86-32] Disables x86 SYSENTER/SYSEXIT support.
3383 nosgx [X86-64,SGX] Disables Intel SGX kernel support.
3385 nosmp [SMP] Tells an SMP kernel to act as a UP kernel,
3386 and disable the IO APIC. legacy for "maxcpus=0".
3388 nosoftlockup [KNL] Disable the soft-lockup detector.
3390 nosync [HW,M68K] Disables sync negotiation for all devices.
3392 nowatchdog [KNL] Disable both lockup detectors, i.e.
3393 soft-lockup and NMI watchdog (hard-lockup).
3397 nox2apic [X86-64,APIC] Do not enable x2APIC mode.
3399 cpu0_hotplug [X86] Turn on CPU0 hotplug feature when
3400 CONFIG_BOOTPARAM_HOTPLUG_CPU0 is off.
3401 Some features depend on CPU0. Known dependencies are:
3402 1. Resume from suspend/hibernate depends on CPU0.
3403 Suspend/hibernate will fail if CPU0 is offline and you
3404 need to online CPU0 before suspend/hibernate.
3405 2. PIC interrupts also depend on CPU0. CPU0 can't be
3406 removed if a PIC interrupt is detected.
3407 It's said poweroff/reboot may depend on CPU0 on some
3408 machines although I haven't seen such issues so far
3409 after CPU0 is offline on a few tested machines.
3410 If the dependencies are under your control, you can
3411 turn on cpu0_hotplug.
3413 nps_mtm_hs_ctr= [KNL,ARC]
3414 This parameter sets the maximum duration, in
3415 cycles, each HW thread of the CTOP can run
3416 without interruptions, before HW switches it.
3417 The actual maximum duration is 16 times this
3419 Format: integer between 1 and 255
3422 nptcg= [IA-64] Override max number of concurrent global TLB
3423 purges which is reported from either PAL_VM_SUMMARY or
3426 nr_cpus= [SMP] Maximum number of processors that an SMP kernel
3427 could support. nr_cpus=n : n >= 1 limits the kernel to
3428 support 'n' processors. It could be larger than the
3429 number of already plugged CPU during bootup, later in
3430 runtime you can physically add extra cpu until it reaches
3431 n. So during boot up some boot time memory for per-cpu
3432 variables need be pre-allocated for later physical cpu
3435 nr_uarts= [SERIAL] maximum number of UARTs to be registered.
3437 numa_balancing= [KNL,X86] Enable or disable automatic NUMA balancing.
3438 Allowed values are enable and disable
3440 numa_zonelist_order= [KNL, BOOT] Select zonelist order for NUMA.
3441 'node', 'default' can be specified
3442 This can be set from sysctl after boot.
3443 See Documentation/admin-guide/sysctl/vm.rst for details.
3445 ohci1394_dma=early [HW] enable debugging via the ohci1394 driver.
3446 See Documentation/core-api/debugging-via-ohci1394.rst for more
3449 olpc_ec_timeout= [OLPC] ms delay when issuing EC commands
3450 Rather than timing out after 20 ms if an EC
3451 command is not properly ACKed, override the length
3452 of the timeout. We have interrupts disabled while
3453 waiting for the ACK, so if this is set too high
3454 interrupts *may* be lost!
3456 omap_mux= [OMAP] Override bootloader pin multiplexing.
3457 Format: <mux_mode0.mode_name=value>...
3458 For example, to override I2C bus2:
3459 omap_mux=i2c2_scl.i2c2_scl=0x100,i2c2_sda.i2c2_sda=0x100
3461 oops=panic Always panic on oopses. Default is to just kill the
3462 process, but there is a small probability of
3463 deadlocking the machine.
3464 This will also cause panics on machine check exceptions.
3465 Useful together with panic=30 to trigger a reboot.
3468 [KNL] Boolean flag to control whether the page allocator
3469 should randomize its free lists. The randomization may
3470 be automatically enabled if the kernel detects it is
3471 running on a platform with a direct-mapped memory-side
3472 cache, and this parameter can be used to
3473 override/disable that behavior. The state of the flag
3474 can be read from sysfs at:
3475 /sys/module/page_alloc/parameters/shuffle.
3477 page_owner= [KNL] Boot-time page_owner enabling option.
3478 Storage of the information about who allocated
3479 each page is disabled in default. With this switch,
3481 on: enable the feature
3483 page_poison= [KNL] Boot-time parameter changing the state of
3484 poisoning on the buddy allocator, available with
3485 CONFIG_PAGE_POISONING=y.
3486 off: turn off poisoning (default)
3487 on: turn on poisoning
3489 panic= [KNL] Kernel behaviour on panic: delay <timeout>
3490 timeout > 0: seconds before rebooting
3491 timeout = 0: wait forever
3492 timeout < 0: reboot immediately
3495 panic_print= Bitmask for printing system info when panic happens.
3496 User can chose combination of the following bits:
3497 bit 0: print all tasks info
3498 bit 1: print system memory info
3499 bit 2: print timer info
3500 bit 3: print locks info if CONFIG_LOCKDEP is on
3501 bit 4: print ftrace buffer
3502 bit 5: print all printk messages in buffer
3504 panic_on_taint= Bitmask for conditionally calling panic() in add_taint()
3505 Format: <hex>[,nousertaint]
3506 Hexadecimal bitmask representing the set of TAINT flags
3507 that will cause the kernel to panic when add_taint() is
3508 called with any of the flags in this set.
3509 The optional switch "nousertaint" can be utilized to
3510 prevent userspace forced crashes by writing to sysctl
3511 /proc/sys/kernel/tainted any flagset matching with the
3512 bitmask set on panic_on_taint.
3513 See Documentation/admin-guide/tainted-kernels.rst for
3514 extra details on the taint flags that users can pick
3515 to compose the bitmask to assign to panic_on_taint.
3517 panic_on_warn panic() instead of WARN(). Useful to cause kdump
3520 crash_kexec_post_notifiers
3521 Run kdump after running panic-notifiers and dumping
3522 kmsg. This only for the users who doubt kdump always
3523 succeeds in any situation.
3524 Note that this also increases risks of kdump failure,
3525 because some panic notifiers can make the crashed
3526 kernel more unstable.
3528 parkbd.port= [HW] Parallel port number the keyboard adapter is
3529 connected to, default is 0.
3531 parkbd.mode= [HW] Parallel port keyboard adapter mode of operation,
3532 0 for XT, 1 for AT (default is AT).
3535 parport= [HW,PPT] Specify parallel ports. 0 disables.
3536 Format: { 0 | auto | 0xBBB[,IRQ[,DMA]] }
3537 Use 'auto' to force the driver to use any
3538 IRQ/DMA settings detected (the default is to
3539 ignore detected IRQ/DMA settings because of
3540 possible conflicts). You can specify the base
3541 address, IRQ, and DMA settings; IRQ and DMA
3542 should be numbers, or 'auto' (for using detected
3543 settings on that particular port), or 'nofifo'
3544 (to avoid using a FIFO even if it is detected).
3545 Parallel ports are assigned in the order they
3546 are specified on the command line, starting
3549 parport_init_mode= [HW,PPT]
3550 Configure VIA parallel port to operate in
3551 a specific mode. This is necessary on Pegasos
3552 computer where firmware has no options for setting
3553 up parallel port mode and sets it to spp.
3554 Currently this function knows 686a and 8231 chips.
3555 Format: [spp|ps2|epp|ecp|ecpepp]
3558 Halt all CPUs after the first oops has been printed for
3559 the specified number of seconds. This is to be used if
3560 your oopses keep scrolling off the screen.
3565 See header of drivers/block/paride/pcd.c.
3566 See also Documentation/admin-guide/blockdev/paride.rst.
3568 pci=option[,option...] [PCI] various PCI subsystem options.
3570 Some options herein operate on a specific device
3571 or a set of devices (<pci_dev>). These are
3572 specified in one of the following formats:
3574 [<domain>:]<bus>:<dev>.<func>[/<dev>.<func>]*
3575 pci:<vendor>:<device>[:<subvendor>:<subdevice>]
3577 Note: the first format specifies a PCI
3578 bus/device/function address which may change
3579 if new hardware is inserted, if motherboard
3580 firmware changes, or due to changes caused
3581 by other kernel parameters. If the
3582 domain is left unspecified, it is
3583 taken to be zero. Optionally, a path
3584 to a device through multiple device/function
3585 addresses can be specified after the base
3586 address (this is more robust against
3587 renumbering issues). The second format
3588 selects devices using IDs from the
3589 configuration space which may match multiple
3590 devices in the system.
3592 earlydump dump PCI config space before the kernel
3594 off [X86] don't probe for the PCI bus
3595 bios [X86-32] force use of PCI BIOS, don't access
3596 the hardware directly. Use this if your machine
3597 has a non-standard PCI host bridge.
3598 nobios [X86-32] disallow use of PCI BIOS, only direct
3599 hardware access methods are allowed. Use this
3600 if you experience crashes upon bootup and you
3601 suspect they are caused by the BIOS.
3602 conf1 [X86] Force use of PCI Configuration Access
3603 Mechanism 1 (config address in IO port 0xCF8,
3604 data in IO port 0xCFC, both 32-bit).
3605 conf2 [X86] Force use of PCI Configuration Access
3606 Mechanism 2 (IO port 0xCF8 is an 8-bit port for
3607 the function, IO port 0xCFA, also 8-bit, sets
3608 bus number. The config space is then accessed
3609 through ports 0xC000-0xCFFF).
3610 See http://wiki.osdev.org/PCI for more info
3611 on the configuration access mechanisms.
3612 noaer [PCIE] If the PCIEAER kernel config parameter is
3613 enabled, this kernel boot option can be used to
3614 disable the use of PCIE advanced error reporting.
3615 nodomains [PCI] Disable support for multiple PCI
3616 root domains (aka PCI segments, in ACPI-speak).
3617 nommconf [X86] Disable use of MMCONFIG for PCI
3619 check_enable_amd_mmconf [X86] check for and enable
3620 properly configured MMIO access to PCI
3621 config space on AMD family 10h CPU
3622 nomsi [MSI] If the PCI_MSI kernel config parameter is
3623 enabled, this kernel boot option can be used to
3624 disable the use of MSI interrupts system-wide.
3625 noioapicquirk [APIC] Disable all boot interrupt quirks.
3626 Safety option to keep boot IRQs enabled. This
3627 should never be necessary.
3628 ioapicreroute [APIC] Enable rerouting of boot IRQs to the
3629 primary IO-APIC for bridges that cannot disable
3630 boot IRQs. This fixes a source of spurious IRQs
3631 when the system masks IRQs.
3632 noioapicreroute [APIC] Disable workaround that uses the
3633 boot IRQ equivalent of an IRQ that connects to
3634 a chipset where boot IRQs cannot be disabled.
3635 The opposite of ioapicreroute.
3636 biosirq [X86-32] Use PCI BIOS calls to get the interrupt
3637 routing table. These calls are known to be buggy
3638 on several machines and they hang the machine
3639 when used, but on other computers it's the only
3640 way to get the interrupt routing table. Try
3641 this option if the kernel is unable to allocate
3642 IRQs or discover secondary PCI buses on your
3644 rom [X86] Assign address space to expansion ROMs.
3645 Use with caution as certain devices share
3646 address decoders between ROMs and other
3648 norom [X86] Do not assign address space to
3649 expansion ROMs that do not already have
3650 BIOS assigned address ranges.
3651 nobar [X86] Do not assign address space to the
3652 BARs that weren't assigned by the BIOS.
3653 irqmask=0xMMMM [X86] Set a bit mask of IRQs allowed to be
3654 assigned automatically to PCI devices. You can
3655 make the kernel exclude IRQs of your ISA cards
3657 pirqaddr=0xAAAAA [X86] Specify the physical address
3658 of the PIRQ table (normally generated
3659 by the BIOS) if it is outside the
3660 F0000h-100000h range.
3661 lastbus=N [X86] Scan all buses thru bus #N. Can be
3662 useful if the kernel is unable to find your
3663 secondary buses and you want to tell it
3664 explicitly which ones they are.
3665 assign-busses [X86] Always assign all PCI bus
3666 numbers ourselves, overriding
3667 whatever the firmware may have done.
3668 usepirqmask [X86] Honor the possible IRQ mask stored
3669 in the BIOS $PIR table. This is needed on
3670 some systems with broken BIOSes, notably
3671 some HP Pavilion N5400 and Omnibook XE3
3672 notebooks. This will have no effect if ACPI
3673 IRQ routing is enabled.
3674 noacpi [X86] Do not use ACPI for IRQ routing
3675 or for PCI scanning.
3676 use_crs [X86] Use PCI host bridge window information
3677 from ACPI. On BIOSes from 2008 or later, this
3678 is enabled by default. If you need to use this,
3679 please report a bug.
3680 nocrs [X86] Ignore PCI host bridge windows from ACPI.
3681 If you need to use this, please report a bug.
3682 routeirq Do IRQ routing for all PCI devices.
3683 This is normally done in pci_enable_device(),
3684 so this option is a temporary workaround
3685 for broken drivers that don't call it.
3686 skip_isa_align [X86] do not align io start addr, so can
3687 handle more pci cards
3688 noearly [X86] Don't do any early type 1 scanning.
3689 This might help on some broken boards which
3690 machine check when some devices' config space
3691 is read. But various workarounds are disabled
3692 and some IOMMU drivers will not work.
3693 bfsort Sort PCI devices into breadth-first order.
3694 This sorting is done to get a device
3695 order compatible with older (<= 2.4) kernels.
3696 nobfsort Don't sort PCI devices into breadth-first order.
3697 pcie_bus_tune_off Disable PCIe MPS (Max Payload Size)
3698 tuning and use the BIOS-configured MPS defaults.
3699 pcie_bus_safe Set every device's MPS to the largest value
3700 supported by all devices below the root complex.
3701 pcie_bus_perf Set device MPS to the largest allowable MPS
3702 based on its parent bus. Also set MRRS (Max
3703 Read Request Size) to the largest supported
3704 value (no larger than the MPS that the device
3705 or bus can support) for best performance.
3706 pcie_bus_peer2peer Set every device's MPS to 128B, which
3707 every device is guaranteed to support. This
3708 configuration allows peer-to-peer DMA between
3709 any pair of devices, possibly at the cost of
3710 reduced performance. This also guarantees
3711 that hot-added devices will work.
3712 cbiosize=nn[KMG] The fixed amount of bus space which is
3713 reserved for the CardBus bridge's IO window.
3714 The default value is 256 bytes.
3715 cbmemsize=nn[KMG] The fixed amount of bus space which is
3716 reserved for the CardBus bridge's memory
3717 window. The default value is 64 megabytes.
3720 [<order of align>@]<pci_dev>[; ...]
3721 Specifies alignment and device to reassign
3722 aligned memory resources. How to
3723 specify the device is described above.
3724 If <order of align> is not specified,
3725 PAGE_SIZE is used as alignment.
3726 A PCI-PCI bridge can be specified if resource
3727 windows need to be expanded.
3728 To specify the alignment for several
3729 instances of a device, the PCI vendor,
3730 device, subvendor, and subdevice may be
3731 specified, e.g., 12@pci:8086:9c22:103c:198f
3732 for 4096-byte alignment.
3733 ecrc= Enable/disable PCIe ECRC (transaction layer
3734 end-to-end CRC checking).
3735 bios: Use BIOS/firmware settings. This is the
3739 hpiosize=nn[KMG] The fixed amount of bus space which is
3740 reserved for hotplug bridge's IO window.
3741 Default size is 256 bytes.
3742 hpmmiosize=nn[KMG] The fixed amount of bus space which is
3743 reserved for hotplug bridge's MMIO window.
3744 Default size is 2 megabytes.
3745 hpmmioprefsize=nn[KMG] The fixed amount of bus space which is
3746 reserved for hotplug bridge's MMIO_PREF window.
3747 Default size is 2 megabytes.
3748 hpmemsize=nn[KMG] The fixed amount of bus space which is
3749 reserved for hotplug bridge's MMIO and
3751 Default size is 2 megabytes.
3752 hpbussize=nn The minimum amount of additional bus numbers
3753 reserved for buses below a hotplug bridge.
3755 realloc= Enable/disable reallocating PCI bridge resources
3756 if allocations done by BIOS are too small to
3757 accommodate resources required by all child
3759 off: Turn realloc off
3761 realloc same as realloc=on
3762 noari do not use PCIe ARI.
3763 noats [PCIE, Intel-IOMMU, AMD-IOMMU]
3764 do not use PCIe ATS (and IOMMU device IOTLB).
3765 pcie_scan_all Scan all possible PCIe devices. Otherwise we
3766 only look for one device below a PCIe downstream
3768 big_root_window Try to add a big 64bit memory window to the PCIe
3769 root complex on AMD CPUs. Some GFX hardware
3770 can resize a BAR to allow access to all VRAM.
3771 Adding the window is slightly risky (it may
3772 conflict with unreported devices), so this
3774 disable_acs_redir=<pci_dev>[; ...]
3775 Specify one or more PCI devices (in the format
3776 specified above) separated by semicolons.
3777 Each device specified will have the PCI ACS
3778 redirect capabilities forced off which will
3779 allow P2P traffic between devices through
3780 bridges without forcing it upstream. Note:
3781 this removes isolation between devices and
3782 may put more devices in an IOMMU group.
3783 force_floating [S390] Force usage of floating interrupts.
3784 nomio [S390] Do not use MIO instructions.
3785 norid [S390] ignore the RID field and force use of
3786 one PCI domain per PCI function
3788 pcie_aspm= [PCIE] Forcibly enable or disable PCIe Active State Power
3791 force Enable ASPM even on devices that claim not to support it.
3792 WARNING: Forcing ASPM on may cause system lockups.
3794 pcie_ports= [PCIE] PCIe port services handling:
3795 native Use native PCIe services (PME, AER, DPC, PCIe hotplug)
3796 even if the platform doesn't give the OS permission to
3797 use them. This may cause conflicts if the platform
3798 also tries to use these services.
3799 dpc-native Use native PCIe service for DPC only. May
3800 cause conflicts if firmware uses AER or DPC.
3801 compat Disable native PCIe services (PME, AER, DPC, PCIe
3804 pcie_port_pm= [PCIE] PCIe port power management handling:
3805 off Disable power management of all PCIe ports
3806 force Forcibly enable power management of all PCIe ports
3808 pcie_pme= [PCIE,PM] Native PCIe PME signaling options:
3809 nomsi Do not use MSI for native PCIe PME signaling (this makes
3810 all PCIe root ports use INTx for all services).
3812 pcmv= [HW,PCMCIA] BadgePAD 4
3816 Keep all power-domains already enabled by bootloader on,
3817 even if no driver has claimed them. This is useful
3818 for debug and development, but should not be
3819 needed on a platform with proper driver support.
3822 See Documentation/admin-guide/blockdev/paride.rst.
3824 pdcchassis= [PARISC,HW] Disable/Enable PDC Chassis Status codes at
3827 See arch/parisc/kernel/pdc_chassis.c
3829 percpu_alloc= Select which percpu first chunk allocator to use.
3830 Currently supported values are "embed" and "page".
3831 Archs may support subset or none of the selections.
3832 See comments in mm/percpu.c for details on each
3833 allocator. This parameter is primarily for debugging
3834 and performance comparison.
3837 See Documentation/admin-guide/blockdev/paride.rst.
3840 See Documentation/admin-guide/blockdev/paride.rst.
3842 pirq= [SMP,APIC] Manual mp-table setup
3843 See Documentation/x86/i386/IO-APIC.rst.
3845 plip= [PPT,NET] Parallel port network link
3846 Format: { parport<nr> | timid | 0 }
3847 See also Documentation/admin-guide/parport.rst.
3849 pmtmr= [X86] Manual setup of pmtmr I/O Port.
3850 Override pmtimer IOPort with a hex value.
3853 pm_debug_messages [SUSPEND,KNL]
3854 Enable suspend/resume debug messages during boot up.
3857 Enable PNP debug messages (depends on the
3858 CONFIG_PNP_DEBUG_MESSAGES option). Change at run-time
3859 via /sys/module/pnp/parameters/debug. We always show
3860 current resource usage; turning this on also shows
3861 possible settings and some assignment information.
3867 { on | off | curr | res | no-curr | no-res }
3870 [ISAPNP] Exclude IRQs for the autoconfiguration
3873 [ISAPNP] Exclude DMAs for the autoconfiguration
3875 pnp_reserve_io= [ISAPNP] Exclude I/O ports for the autoconfiguration
3876 Ranges are in pairs (I/O port base and size).
3879 [ISAPNP] Exclude memory regions for the
3881 Ranges are in pairs (memory base and size).
3883 ports= [IP_VS_FTP] IPVS ftp helper module
3885 Up to 8 (IP_VS_APP_MAX_PORTS) ports
3887 Format: <port>,<port>....
3889 powersave=off [PPC] This option disables power saving features.
3890 It specifically disables cpuidle and sets the
3891 platform machine description specific power_save
3892 function to NULL. On Idle the CPU just reduces
3895 ppc_strict_facility_enable
3896 [PPC] This option catches any kernel floating point,
3897 Altivec, VSX and SPE outside of regions specifically
3898 allowed (eg kernel_enable_fpu()/kernel_disable_fpu()).
3899 There is some performance impact when enabling this.
3903 Disable Hardware Transactional Memory
3905 print-fatal-signals=
3906 [KNL] debug: print fatal signals
3908 If enabled, warn about various signal handling
3909 related application anomalies: too many signals,
3910 too many POSIX.1 timers, fatal signals causing a
3913 If you hit the warning due to signal overflow,
3914 you might want to try "ulimit -i unlimited".
3918 printk.always_kmsg_dump=
3919 Trigger kmsg_dump for cases other than kernel oops or
3921 Format: <bool> (1/Y/y=enable, 0/N/n=disable)
3924 printk.devkmsg={on,off,ratelimit}
3925 Control writing to /dev/kmsg.
3926 on - unlimited logging to /dev/kmsg from userspace
3927 off - logging to /dev/kmsg disabled
3928 ratelimit - ratelimit the logging
3931 printk.time= Show timing data prefixed to each printk message line
3932 Format: <bool> (1/Y/y=enable, 0/N/n=disable)
3934 processor.max_cstate= [HW,ACPI]
3935 Limit processor to maximum C-state
3936 max_cstate=9 overrides any DMI blacklist limit.
3938 processor.nocst [HW,ACPI]
3939 Ignore the _CST method to determine C-states,
3940 instead using the legacy FADT method
3942 profile= [KNL] Enable kernel profiling via /proc/profile
3943 Format: [<profiletype>,]<number>
3944 Param: <profiletype>: "schedule", "sleep", or "kvm"
3945 [defaults to kernel profiling]
3946 Param: "schedule" - profile schedule points.
3947 Param: "sleep" - profile D-state sleeping (millisecs).
3948 Requires CONFIG_SCHEDSTATS
3949 Param: "kvm" - profile VM exits.
3950 Param: <number> - step/bucket size as a power of 2 for
3951 statistical time based profiling.
3953 prompt_ramdisk= [RAM] [Deprecated]
3955 prot_virt= [S390] enable hosting protected virtual machines
3956 isolated from the hypervisor (if hardware supports
3960 psi= [KNL] Enable or disable pressure stall information
3964 psmouse.proto= [HW,MOUSE] Highest PS2 mouse protocol extension to
3965 probe for; one of (bare|imps|exps|lifebook|any).
3966 psmouse.rate= [HW,MOUSE] Set desired mouse report rate, in reports
3968 psmouse.resetafter= [HW,MOUSE]
3969 Try to reset the device after so many bad packets
3972 [HW,MOUSE] Set desired mouse resolution, in dpi.
3973 psmouse.smartscroll=
3974 [HW,MOUSE] Controls Logitech smartscroll autorepeat.
3975 0 = disabled, 1 = enabled (default).
3977 pstore.backend= Specify the name of the pstore backend to use
3980 See Documentation/admin-guide/blockdev/paride.rst.
3982 pti= [X86-64] Control Page Table Isolation of user and
3983 kernel address spaces. Disabling this feature
3984 removes hardening, but improves performance of
3985 system calls and interrupts.
3987 on - unconditionally enable
3988 off - unconditionally disable
3989 auto - kernel detects whether your CPU model is
3990 vulnerable to issues that PTI mitigates
3992 Not specifying this option is equivalent to pti=auto.
3995 Equivalent to pti=off
3998 [KNL] Number of legacy pty's. Overwrites compiled-in
4001 quiet [KNL] Disable most log messages
4006 See Documentation/admin-guide/md.rst.
4008 ramdisk_size= [RAM] Sizes of RAM disks in kilobytes
4009 See Documentation/admin-guide/blockdev/ramdisk.rst.
4011 ramdisk_start= [RAM] RAM disk image start address
4013 random.trust_cpu={on,off}
4014 [KNL] Enable or disable trusting the use of the
4015 CPU's random number generator (if available) to
4016 fully seed the kernel's CRNG. Default is controlled
4017 by CONFIG_RANDOM_TRUST_CPU.
4019 ras=option[,option,...] [KNL] RAS-specific options
4022 Disable the Correctable Errors Collector,
4023 see CONFIG_RAS_CEC help text.
4026 The argument is a cpu list, as described above,
4027 except that the string "all" can be used to
4028 specify every CPU on the system.
4030 In kernels built with CONFIG_RCU_NOCB_CPU=y, set
4031 the specified list of CPUs to be no-callback CPUs.
4032 Invocation of these CPUs' RCU callbacks will be
4033 offloaded to "rcuox/N" kthreads created for that
4034 purpose, where "x" is "p" for RCU-preempt, and
4035 "s" for RCU-sched, and "N" is the CPU number.
4036 This reduces OS jitter on the offloaded CPUs,
4037 which can be useful for HPC and real-time
4038 workloads. It can also improve energy efficiency
4039 for asymmetric multiprocessors.
4042 Rather than requiring that offloaded CPUs
4043 (specified by rcu_nocbs= above) explicitly
4044 awaken the corresponding "rcuoN" kthreads,
4045 make these kthreads poll for callbacks.
4046 This improves the real-time response for the
4047 offloaded CPUs by relieving them of the need to
4048 wake up the corresponding kthread, but degrades
4049 energy efficiency by requiring that the kthreads
4050 periodically wake up to do the polling.
4052 rcutree.blimit= [KNL]
4053 Set maximum number of finished RCU callbacks to
4054 process in one batch.
4056 rcutree.dump_tree= [KNL]
4057 Dump the structure of the rcu_node combining tree
4058 out at early boot. This is used for diagnostic
4059 purposes, to verify correct tree setup.
4061 rcutree.gp_cleanup_delay= [KNL]
4062 Set the number of jiffies to delay each step of
4063 RCU grace-period cleanup.
4065 rcutree.gp_init_delay= [KNL]
4066 Set the number of jiffies to delay each step of
4067 RCU grace-period initialization.
4069 rcutree.gp_preinit_delay= [KNL]
4070 Set the number of jiffies to delay each step of
4071 RCU grace-period pre-initialization, that is,
4072 the propagation of recent CPU-hotplug changes up
4073 the rcu_node combining tree.
4075 rcutree.use_softirq= [KNL]
4076 If set to zero, move all RCU_SOFTIRQ processing to
4077 per-CPU rcuc kthreads. Defaults to a non-zero
4078 value, meaning that RCU_SOFTIRQ is used by default.
4079 Specify rcutree.use_softirq=0 to use rcuc kthreads.
4081 But note that CONFIG_PREEMPT_RT=y kernels disable
4082 this kernel boot parameter, forcibly setting it
4085 rcutree.rcu_fanout_exact= [KNL]
4086 Disable autobalancing of the rcu_node combining
4087 tree. This is used by rcutorture, and might
4088 possibly be useful for architectures having high
4089 cache-to-cache transfer latencies.
4091 rcutree.rcu_fanout_leaf= [KNL]
4092 Change the number of CPUs assigned to each
4093 leaf rcu_node structure. Useful for very
4094 large systems, which will choose the value 64,
4095 and for NUMA systems with large remote-access
4096 latencies, which will choose a value aligned
4097 with the appropriate hardware boundaries.
4099 rcutree.rcu_min_cached_objs= [KNL]
4100 Minimum number of objects which are cached and
4101 maintained per one CPU. Object size is equal
4102 to PAGE_SIZE. The cache allows to reduce the
4103 pressure to page allocator, also it makes the
4104 whole algorithm to behave better in low memory
4107 rcutree.jiffies_till_first_fqs= [KNL]
4108 Set delay from grace-period initialization to
4109 first attempt to force quiescent states.
4110 Units are jiffies, minimum value is zero,
4111 and maximum value is HZ.
4113 rcutree.jiffies_till_next_fqs= [KNL]
4114 Set delay between subsequent attempts to force
4115 quiescent states. Units are jiffies, minimum
4116 value is one, and maximum value is HZ.
4118 rcutree.jiffies_till_sched_qs= [KNL]
4119 Set required age in jiffies for a
4120 given grace period before RCU starts
4121 soliciting quiescent-state help from
4122 rcu_note_context_switch() and cond_resched().
4123 If not specified, the kernel will calculate
4124 a value based on the most recent settings
4125 of rcutree.jiffies_till_first_fqs
4126 and rcutree.jiffies_till_next_fqs.
4127 This calculated value may be viewed in
4128 rcutree.jiffies_to_sched_qs. Any attempt to set
4129 rcutree.jiffies_to_sched_qs will be cheerfully
4132 rcutree.kthread_prio= [KNL,BOOT]
4133 Set the SCHED_FIFO priority of the RCU per-CPU
4134 kthreads (rcuc/N). This value is also used for
4135 the priority of the RCU boost threads (rcub/N)
4136 and for the RCU grace-period kthreads (rcu_bh,
4137 rcu_preempt, and rcu_sched). If RCU_BOOST is
4138 set, valid values are 1-99 and the default is 1
4139 (the least-favored priority). Otherwise, when
4140 RCU_BOOST is not set, valid values are 0-99 and
4141 the default is zero (non-realtime operation).
4143 rcutree.rcu_nocb_gp_stride= [KNL]
4144 Set the number of NOCB callback kthreads in
4145 each group, which defaults to the square root
4146 of the number of CPUs. Larger numbers reduce
4147 the wakeup overhead on the global grace-period
4148 kthread, but increases that same overhead on
4149 each group's NOCB grace-period kthread.
4151 rcutree.qhimark= [KNL]
4152 Set threshold of queued RCU callbacks beyond which
4153 batch limiting is disabled.
4155 rcutree.qlowmark= [KNL]
4156 Set threshold of queued RCU callbacks below which
4157 batch limiting is re-enabled.
4159 rcutree.qovld= [KNL]
4160 Set threshold of queued RCU callbacks beyond which
4161 RCU's force-quiescent-state scan will aggressively
4162 enlist help from cond_resched() and sched IPIs to
4163 help CPUs more quickly reach quiescent states.
4164 Set to less than zero to make this be set based
4165 on rcutree.qhimark at boot time and to zero to
4166 disable more aggressive help enlistment.
4168 rcutree.rcu_idle_gp_delay= [KNL]
4169 Set wakeup interval for idle CPUs that have
4170 RCU callbacks (RCU_FAST_NO_HZ=y).
4172 rcutree.rcu_kick_kthreads= [KNL]
4173 Cause the grace-period kthread to get an extra
4174 wake_up() if it sleeps three times longer than
4175 it should at force-quiescent-state time.
4176 This wake_up() will be accompanied by a
4177 WARN_ONCE() splat and an ftrace_dump().
4179 rcutree.rcu_unlock_delay= [KNL]
4180 In CONFIG_RCU_STRICT_GRACE_PERIOD=y kernels,
4181 this specifies an rcu_read_unlock()-time delay
4182 in microseconds. This defaults to zero.
4183 Larger delays increase the probability of
4184 catching RCU pointer leaks, that is, buggy use
4185 of RCU-protected pointers after the relevant
4186 rcu_read_unlock() has completed.
4188 rcutree.sysrq_rcu= [KNL]
4189 Commandeer a sysrq key to dump out Tree RCU's
4190 rcu_node tree with an eye towards determining
4191 why a new grace period has not yet started.
4193 rcuscale.gp_async= [KNL]
4194 Measure performance of asynchronous
4195 grace-period primitives such as call_rcu().
4197 rcuscale.gp_async_max= [KNL]
4198 Specify the maximum number of outstanding
4199 callbacks per writer thread. When a writer
4200 thread exceeds this limit, it invokes the
4201 corresponding flavor of rcu_barrier() to allow
4202 previously posted callbacks to drain.
4204 rcuscale.gp_exp= [KNL]
4205 Measure performance of expedited synchronous
4206 grace-period primitives.
4208 rcuscale.holdoff= [KNL]
4209 Set test-start holdoff period. The purpose of
4210 this parameter is to delay the start of the
4211 test until boot completes in order to avoid
4214 rcuscale.kfree_rcu_test= [KNL]
4215 Set to measure performance of kfree_rcu() flooding.
4217 rcuscale.kfree_nthreads= [KNL]
4218 The number of threads running loops of kfree_rcu().
4220 rcuscale.kfree_alloc_num= [KNL]
4221 Number of allocations and frees done in an iteration.
4223 rcuscale.kfree_loops= [KNL]
4224 Number of loops doing rcuscale.kfree_alloc_num number
4225 of allocations and frees.
4227 rcuscale.nreaders= [KNL]
4228 Set number of RCU readers. The value -1 selects
4229 N, where N is the number of CPUs. A value
4230 "n" less than -1 selects N-n+1, where N is again
4231 the number of CPUs. For example, -2 selects N
4232 (the number of CPUs), -3 selects N+1, and so on.
4233 A value of "n" less than or equal to -N selects
4236 rcuscale.nwriters= [KNL]
4237 Set number of RCU writers. The values operate
4238 the same as for rcuscale.nreaders.
4239 N, where N is the number of CPUs
4241 rcuscale.perf_type= [KNL]
4242 Specify the RCU implementation to test.
4244 rcuscale.shutdown= [KNL]
4245 Shut the system down after performance tests
4246 complete. This is useful for hands-off automated
4249 rcuscale.verbose= [KNL]
4250 Enable additional printk() statements.
4252 rcuscale.writer_holdoff= [KNL]
4253 Write-side holdoff between grace periods,
4254 in microseconds. The default of zero says
4257 rcutorture.fqs_duration= [KNL]
4258 Set duration of force_quiescent_state bursts
4261 rcutorture.fqs_holdoff= [KNL]
4262 Set holdoff time within force_quiescent_state bursts
4265 rcutorture.fqs_stutter= [KNL]
4266 Set wait time between force_quiescent_state bursts
4269 rcutorture.fwd_progress= [KNL]
4270 Enable RCU grace-period forward-progress testing
4271 for the types of RCU supporting this notion.
4273 rcutorture.fwd_progress_div= [KNL]
4274 Specify the fraction of a CPU-stall-warning
4275 period to do tight-loop forward-progress testing.
4277 rcutorture.fwd_progress_holdoff= [KNL]
4278 Number of seconds to wait between successive
4279 forward-progress tests.
4281 rcutorture.fwd_progress_need_resched= [KNL]
4282 Enclose cond_resched() calls within checks for
4283 need_resched() during tight-loop forward-progress
4286 rcutorture.gp_cond= [KNL]
4287 Use conditional/asynchronous update-side
4288 primitives, if available.
4290 rcutorture.gp_exp= [KNL]
4291 Use expedited update-side primitives, if available.
4293 rcutorture.gp_normal= [KNL]
4294 Use normal (non-expedited) asynchronous
4295 update-side primitives, if available.
4297 rcutorture.gp_sync= [KNL]
4298 Use normal (non-expedited) synchronous
4299 update-side primitives, if available. If all
4300 of rcutorture.gp_cond=, rcutorture.gp_exp=,
4301 rcutorture.gp_normal=, and rcutorture.gp_sync=
4302 are zero, rcutorture acts as if is interpreted
4303 they are all non-zero.
4305 rcutorture.irqreader= [KNL]
4306 Run RCU readers from irq handlers, or, more
4307 accurately, from a timer handler. Not all RCU
4308 flavors take kindly to this sort of thing.
4310 rcutorture.leakpointer= [KNL]
4311 Leak an RCU-protected pointer out of the reader.
4312 This can of course result in splats, and is
4313 intended to test the ability of things like
4314 CONFIG_RCU_STRICT_GRACE_PERIOD=y to detect
4317 rcutorture.n_barrier_cbs= [KNL]
4318 Set callbacks/threads for rcu_barrier() testing.
4320 rcutorture.nfakewriters= [KNL]
4321 Set number of concurrent RCU writers. These just
4322 stress RCU, they don't participate in the actual
4323 test, hence the "fake".
4325 rcutorture.nocbs_nthreads= [KNL]
4326 Set number of RCU callback-offload togglers.
4327 Zero (the default) disables toggling.
4329 rcutorture.nocbs_toggle= [KNL]
4330 Set the delay in milliseconds between successive
4331 callback-offload toggling attempts.
4333 rcutorture.nreaders= [KNL]
4334 Set number of RCU readers. The value -1 selects
4335 N-1, where N is the number of CPUs. A value
4336 "n" less than -1 selects N-n-2, where N is again
4337 the number of CPUs. For example, -2 selects N
4338 (the number of CPUs), -3 selects N+1, and so on.
4340 rcutorture.object_debug= [KNL]
4341 Enable debug-object double-call_rcu() testing.
4343 rcutorture.onoff_holdoff= [KNL]
4344 Set time (s) after boot for CPU-hotplug testing.
4346 rcutorture.onoff_interval= [KNL]
4347 Set time (jiffies) between CPU-hotplug operations,
4348 or zero to disable CPU-hotplug testing.
4350 rcutorture.read_exit= [KNL]
4351 Set the number of read-then-exit kthreads used
4352 to test the interaction of RCU updaters and
4353 task-exit processing.
4355 rcutorture.read_exit_burst= [KNL]
4356 The number of times in a given read-then-exit
4357 episode that a set of read-then-exit kthreads
4360 rcutorture.read_exit_delay= [KNL]
4361 The delay, in seconds, between successive
4362 read-then-exit testing episodes.
4364 rcutorture.shuffle_interval= [KNL]
4365 Set task-shuffle interval (s). Shuffling tasks
4366 allows some CPUs to go into dyntick-idle mode
4367 during the rcutorture test.
4369 rcutorture.shutdown_secs= [KNL]
4370 Set time (s) after boot system shutdown. This
4371 is useful for hands-off automated testing.
4373 rcutorture.stall_cpu= [KNL]
4374 Duration of CPU stall (s) to test RCU CPU stall
4375 warnings, zero to disable.
4377 rcutorture.stall_cpu_block= [KNL]
4378 Sleep while stalling if set. This will result
4379 in warnings from preemptible RCU in addition
4380 to any other stall-related activity.
4382 rcutorture.stall_cpu_holdoff= [KNL]
4383 Time to wait (s) after boot before inducing stall.
4385 rcutorture.stall_cpu_irqsoff= [KNL]
4386 Disable interrupts while stalling if set.
4388 rcutorture.stall_gp_kthread= [KNL]
4389 Duration (s) of forced sleep within RCU
4390 grace-period kthread to test RCU CPU stall
4391 warnings, zero to disable. If both stall_cpu
4392 and stall_gp_kthread are specified, the
4393 kthread is starved first, then the CPU.
4395 rcutorture.stat_interval= [KNL]
4396 Time (s) between statistics printk()s.
4398 rcutorture.stutter= [KNL]
4399 Time (s) to stutter testing, for example, specifying
4400 five seconds causes the test to run for five seconds,
4401 wait for five seconds, and so on. This tests RCU's
4402 ability to transition abruptly to and from idle.
4404 rcutorture.test_boost= [KNL]
4405 Test RCU priority boosting? 0=no, 1=maybe, 2=yes.
4406 "Maybe" means test if the RCU implementation
4407 under test support RCU priority boosting.
4409 rcutorture.test_boost_duration= [KNL]
4410 Duration (s) of each individual boost test.
4412 rcutorture.test_boost_interval= [KNL]
4413 Interval (s) between each boost test.
4415 rcutorture.test_no_idle_hz= [KNL]
4416 Test RCU's dyntick-idle handling. See also the
4417 rcutorture.shuffle_interval parameter.
4419 rcutorture.torture_type= [KNL]
4420 Specify the RCU implementation to test.
4422 rcutorture.verbose= [KNL]
4423 Enable additional printk() statements.
4425 rcupdate.rcu_cpu_stall_ftrace_dump= [KNL]
4426 Dump ftrace buffer after reporting RCU CPU
4429 rcupdate.rcu_cpu_stall_suppress= [KNL]
4430 Suppress RCU CPU stall warning messages.
4432 rcupdate.rcu_cpu_stall_suppress_at_boot= [KNL]
4433 Suppress RCU CPU stall warning messages and
4434 rcutorture writer stall warnings that occur
4435 during early boot, that is, during the time
4436 before the init task is spawned.
4438 rcupdate.rcu_cpu_stall_timeout= [KNL]
4439 Set timeout for RCU CPU stall warning messages.
4441 rcupdate.rcu_expedited= [KNL]
4442 Use expedited grace-period primitives, for
4443 example, synchronize_rcu_expedited() instead
4444 of synchronize_rcu(). This reduces latency,
4445 but can increase CPU utilization, degrade
4446 real-time latency, and degrade energy efficiency.
4447 No effect on CONFIG_TINY_RCU kernels.
4449 rcupdate.rcu_normal= [KNL]
4450 Use only normal grace-period primitives,
4451 for example, synchronize_rcu() instead of
4452 synchronize_rcu_expedited(). This improves
4453 real-time latency, CPU utilization, and
4454 energy efficiency, but can expose users to
4455 increased grace-period latency. This parameter
4456 overrides rcupdate.rcu_expedited. No effect on
4457 CONFIG_TINY_RCU kernels.
4459 rcupdate.rcu_normal_after_boot= [KNL]
4460 Once boot has completed (that is, after
4461 rcu_end_inkernel_boot() has been invoked), use
4462 only normal grace-period primitives. No effect
4463 on CONFIG_TINY_RCU kernels.
4465 But note that CONFIG_PREEMPT_RT=y kernels enables
4466 this kernel boot parameter, forcibly setting
4467 it to the value one, that is, converting any
4468 post-boot attempt at an expedited RCU grace
4469 period to instead use normal non-expedited
4470 grace-period processing.
4472 rcupdate.rcu_task_ipi_delay= [KNL]
4473 Set time in jiffies during which RCU tasks will
4474 avoid sending IPIs, starting with the beginning
4475 of a given grace period. Setting a large
4476 number avoids disturbing real-time workloads,
4477 but lengthens grace periods.
4479 rcupdate.rcu_task_stall_timeout= [KNL]
4480 Set timeout in jiffies for RCU task stall warning
4481 messages. Disable with a value less than or equal
4484 rcupdate.rcu_self_test= [KNL]
4485 Run the RCU early boot self tests
4489 Run specified binary instead of /init from the ramdisk,
4490 used for early userspace startup. See initrd.
4493 force - Override the decision by the kernel to hide the
4494 advertisement of RDRAND support (this affects
4495 certain AMD processors because of buggy BIOS
4496 support, specifically around the suspend/resume
4500 Turn on/off individual RDT features. List is:
4501 cmt, mbmtotal, mbmlocal, l3cat, l3cdp, l2cat, l2cdp,
4503 E.g. to turn on cmt and turn off mba use:
4507 Format (x86 or x86_64):
4508 [w[arm] | c[old] | h[ard] | s[oft] | g[pio]] \
4510 [[,]b[ios] | a[cpi] | k[bd] | t[riple] | e[fi] | p[ci]] \
4512 Where reboot_mode is one of warm (soft) or cold (hard) or gpio
4513 (prefix with 'panic_' to set mode for panic
4515 reboot_type is one of bios, acpi, kbd, triple, efi, or pci,
4516 reboot_force is either force or not specified,
4517 reboot_cpu is s[mp]#### with #### being the processor
4518 to be used for rebooting.
4520 refscale.holdoff= [KNL]
4521 Set test-start holdoff period. The purpose of
4522 this parameter is to delay the start of the
4523 test until boot completes in order to avoid
4526 refscale.loops= [KNL]
4527 Set the number of loops over the synchronization
4528 primitive under test. Increasing this number
4529 reduces noise due to loop start/end overhead,
4530 but the default has already reduced the per-pass
4531 noise to a handful of picoseconds on ca. 2020
4534 refscale.nreaders= [KNL]
4535 Set number of readers. The default value of -1
4536 selects N, where N is roughly 75% of the number
4537 of CPUs. A value of zero is an interesting choice.
4539 refscale.nruns= [KNL]
4540 Set number of runs, each of which is dumped onto
4543 refscale.readdelay= [KNL]
4544 Set the read-side critical-section duration,
4545 measured in microseconds.
4547 refscale.scale_type= [KNL]
4548 Specify the read-protection implementation to test.
4550 refscale.shutdown= [KNL]
4551 Shut down the system at the end of the performance
4552 test. This defaults to 1 (shut it down) when
4553 refscale is built into the kernel and to 0 (leave
4554 it running) when refscale is built as a module.
4556 refscale.verbose= [KNL]
4557 Enable additional printk() statements.
4559 refscale.verbose_batched= [KNL]
4560 Batch the additional printk() statements. If zero
4561 (the default) or negative, print everything. Otherwise,
4562 print every Nth verbose statement, where N is the value
4566 [KNL, SMP] Set scheduler's default relax_domain_level.
4567 See Documentation/admin-guide/cgroup-v1/cpusets.rst.
4569 reserve= [KNL,BUGS] Force kernel to ignore I/O ports or memory
4570 Format: <base1>,<size1>[,<base2>,<size2>,...]
4571 Reserve I/O ports or memory so the kernel won't use
4572 them. If <base> is less than 0x10000, the region
4573 is assumed to be I/O ports; otherwise it is memory.
4575 reservetop= [X86-32]
4577 Reserves a hole at the top of the kernel virtual
4582 Set the amount of memory to reserve for BIOS at
4583 the bottom of the address space.
4585 reset_devices [KNL] Force drivers to reset the underlying device
4586 during initialization.
4589 Specify the partition device for software suspend
4591 {/dev/<dev> | PARTUUID=<uuid> | <int>:<int> | <hex>}
4593 resume_offset= [SWSUSP]
4594 Specify the offset from the beginning of the partition
4595 given by "resume=" at which the swap header is located,
4596 in <PAGE_SIZE> units (needed only for swap files).
4597 See Documentation/power/swsusp-and-swap-files.rst
4599 resumedelay= [HIBERNATION] Delay (in seconds) to pause before attempting to
4600 read the resume files
4602 resumewait [HIBERNATION] Wait (indefinitely) for resume device to show up.
4603 Useful for devices that are detected asynchronously
4604 (e.g. USB and MMC devices).
4606 hibernate= [HIBERNATION]
4607 noresume Don't check if there's a hibernation image
4608 present during boot.
4609 nocompress Don't compress/decompress hibernation images.
4610 no Disable hibernation and resume.
4611 protect_image Turn on image protection during restoration
4612 (that will set all pages holding image data
4613 during restoration read-only).
4615 retain_initrd [RAM] Keep initrd memory after extraction
4617 rfkill.default_state=
4618 0 "airplane mode". All wifi, bluetooth, wimax, gps, fm,
4619 etc. communication is blocked by default.
4622 rfkill.master_switch_mode=
4623 0 The "airplane mode" button does nothing.
4624 1 The "airplane mode" button toggles between everything
4625 blocked and the previous configuration.
4626 2 The "airplane mode" button toggles between everything
4627 blocked and everything unblocked.
4629 rhash_entries= [KNL,NET]
4630 Set number of hash buckets for route cache
4633 [KNL] Disable ring 3 MONITOR/MWAIT feature on supported
4636 ro [KNL] Mount root device read-only on boot
4639 on Mark read-only kernel memory as read-only (default).
4640 off Leave read-only kernel memory writable for debugging.
4643 Enable the uart passthrough on the designated usb port
4644 on Rockchip SoCs. When active, the signals of the
4645 debug-uart get routed to the D+ and D- pins of the usb
4646 port and the regular usb controller gets disabled.
4648 root= [KNL] Root filesystem
4649 See name_to_dev_t comment in init/do_mounts.c.
4651 rootdelay= [KNL] Delay (in seconds) to pause before attempting to
4652 mount the root filesystem
4654 rootflags= [KNL] Set root filesystem mount option string
4656 rootfstype= [KNL] Set root filesystem type
4658 rootwait [KNL] Wait (indefinitely) for root device to show up.
4659 Useful for devices that are detected asynchronously
4660 (e.g. USB and MMC devices).
4662 rproc_mem=nn[KMG][@address]
4663 [KNL,ARM,CMA] Remoteproc physical memory block.
4664 Memory area to be used by remote processor image,
4667 rw [KNL] Mount root device read-write on boot
4669 S [KNL] Run init in single mode
4671 s390_iommu= [HW,S390]
4672 Set s390 IOTLB flushing mode
4674 With strict flushing every unmap operation will result in
4675 an IOTLB flush. Default is lazy flushing before reuse,
4679 See drivers/net/irda/sa1100_ir.c.
4681 sbni= [NET] Granch SBNI12 leased line adapter
4683 sched_debug [KNL] Enables verbose scheduler debug messages.
4685 schedstats= [KNL,X86] Enable or disable scheduled statistics.
4686 Allowed values are enable and disable. This feature
4687 incurs a small amount of overhead in the scheduler
4688 but is useful for debugging and performance tuning.
4690 sched_thermal_decay_shift=
4691 [KNL, SMP] Set a decay shift for scheduler thermal
4692 pressure signal. Thermal pressure signal follows the
4693 default decay period of other scheduler pelt
4694 signals(usually 32 ms but configurable). Setting
4695 sched_thermal_decay_shift will left shift the decay
4696 period for the thermal pressure signal by the shift
4698 i.e. with the default pelt decay period of 32 ms
4699 sched_thermal_decay_shift thermal pressure decay pr
4703 Format: integer between 0 and 10
4706 scftorture.holdoff= [KNL]
4707 Number of seconds to hold off before starting
4708 test. Defaults to zero for module insertion and
4709 to 10 seconds for built-in smp_call_function()
4712 scftorture.longwait= [KNL]
4713 Request ridiculously long waits randomly selected
4714 up to the chosen limit in seconds. Zero (the
4715 default) disables this feature. Please note
4716 that requesting even small non-zero numbers of
4717 seconds can result in RCU CPU stall warnings,
4718 softlockup complaints, and so on.
4720 scftorture.nthreads= [KNL]
4721 Number of kthreads to spawn to invoke the
4722 smp_call_function() family of functions.
4723 The default of -1 specifies a number of kthreads
4724 equal to the number of CPUs.
4726 scftorture.onoff_holdoff= [KNL]
4727 Number seconds to wait after the start of the
4728 test before initiating CPU-hotplug operations.
4730 scftorture.onoff_interval= [KNL]
4731 Number seconds to wait between successive
4732 CPU-hotplug operations. Specifying zero (which
4733 is the default) disables CPU-hotplug operations.
4735 scftorture.shutdown_secs= [KNL]
4736 The number of seconds following the start of the
4737 test after which to shut down the system. The
4738 default of zero avoids shutting down the system.
4739 Non-zero values are useful for automated tests.
4741 scftorture.stat_interval= [KNL]
4742 The number of seconds between outputting the
4743 current test statistics to the console. A value
4744 of zero disables statistics output.
4746 scftorture.stutter_cpus= [KNL]
4747 The number of jiffies to wait between each change
4748 to the set of CPUs under test.
4750 scftorture.use_cpus_read_lock= [KNL]
4751 Use use_cpus_read_lock() instead of the default
4752 preempt_disable() to disable CPU hotplug
4753 while invoking one of the smp_call_function*()
4756 scftorture.verbose= [KNL]
4757 Enable additional printk() statements.
4759 scftorture.weight_single= [KNL]
4760 The probability weighting to use for the
4761 smp_call_function_single() function with a zero
4762 "wait" parameter. A value of -1 selects the
4763 default if all other weights are -1. However,
4764 if at least one weight has some other value, a
4765 value of -1 will instead select a weight of zero.
4767 scftorture.weight_single_wait= [KNL]
4768 The probability weighting to use for the
4769 smp_call_function_single() function with a
4770 non-zero "wait" parameter. See weight_single.
4772 scftorture.weight_many= [KNL]
4773 The probability weighting to use for the
4774 smp_call_function_many() function with a zero
4775 "wait" parameter. See weight_single.
4776 Note well that setting a high probability for
4777 this weighting can place serious IPI load
4780 scftorture.weight_many_wait= [KNL]
4781 The probability weighting to use for the
4782 smp_call_function_many() function with a
4783 non-zero "wait" parameter. See weight_single
4786 scftorture.weight_all= [KNL]
4787 The probability weighting to use for the
4788 smp_call_function_all() function with a zero
4789 "wait" parameter. See weight_single and
4792 scftorture.weight_all_wait= [KNL]
4793 The probability weighting to use for the
4794 smp_call_function_all() function with a
4795 non-zero "wait" parameter. See weight_single
4798 skew_tick= [KNL] Offset the periodic timer tick per cpu to mitigate
4799 xtime_lock contention on larger systems, and/or RCU lock
4800 contention on all systems with CONFIG_MAXSMP set.
4801 Format: { "0" | "1" }
4802 0 -- disable. (may be 1 via CONFIG_CMDLINE="skew_tick=1"
4804 Note: increases power consumption, thus should only be
4805 enabled if running jitter sensitive (HPC/RT) workloads.
4807 security= [SECURITY] Choose a legacy "major" security module to
4808 enable at boot. This has been deprecated by the
4811 selinux= [SELINUX] Disable or enable SELinux at boot time.
4812 Format: { "0" | "1" }
4813 See security/selinux/Kconfig help text.
4818 apparmor= [APPARMOR] Disable or enable AppArmor at boot time
4819 Format: { "0" | "1" }
4820 See security/apparmor/Kconfig help text
4823 Default value is set via kernel config option.
4825 serialnumber [BUGS=X86-32]
4828 Maximal number of shapers.
4836 Disable merging of slabs with similar size. May be
4837 necessary if there is some reason to distinguish
4838 allocs to different slabs, especially in hardened
4839 environments where the risk of heap overflows and
4840 layout control by attackers can usually be
4841 frustrated by disabling merging. This will reduce
4842 most of the exposure of a heap attack to a single
4843 cache (risks via metadata attacks are mostly
4844 unchanged). Debug options disable merging on their
4846 For more information see Documentation/vm/slub.rst.
4848 slab_max_order= [MM, SLAB]
4849 Determines the maximum allowed order for slabs.
4850 A high setting may cause OOMs due to memory
4851 fragmentation. Defaults to 1 for systems with
4852 more than 32MB of RAM, 0 otherwise.
4854 slub_debug[=options[,slabs][;[options[,slabs]]...] [MM, SLUB]
4855 Enabling slub_debug allows one to determine the
4856 culprit if slab objects become corrupted. Enabling
4857 slub_debug can create guard zones around objects and
4858 may poison objects when not in use. Also tracks the
4859 last alloc / free. For more information see
4860 Documentation/vm/slub.rst.
4862 slub_memcg_sysfs= [MM, SLUB]
4863 Determines whether to enable sysfs directories for
4864 memory cgroup sub-caches. 1 to enable, 0 to disable.
4865 The default is determined by CONFIG_SLUB_MEMCG_SYSFS_ON.
4866 Enabling this can lead to a very high number of debug
4867 directories and files being created under
4870 slub_max_order= [MM, SLUB]
4871 Determines the maximum allowed order for slabs.
4872 A high setting may cause OOMs due to memory
4873 fragmentation. For more information see
4874 Documentation/vm/slub.rst.
4876 slub_min_objects= [MM, SLUB]
4877 The minimum number of objects per slab. SLUB will
4878 increase the slab order up to slub_max_order to
4879 generate a sufficiently large slab able to contain
4880 the number of objects indicated. The higher the number
4881 of objects the smaller the overhead of tracking slabs
4882 and the less frequently locks need to be acquired.
4883 For more information see Documentation/vm/slub.rst.
4885 slub_min_order= [MM, SLUB]
4886 Determines the minimum page order for slabs. Must be
4887 lower than slub_max_order.
4888 For more information see Documentation/vm/slub.rst.
4890 slub_nomerge [MM, SLUB]
4891 Same with slab_nomerge. This is supported for legacy.
4892 See slab_nomerge for more information.
4895 Format: <io1>[,<io2>[,...,<io8>]]
4897 smsc-ircc2.nopnp [HW] Don't use PNP to discover SMC devices
4898 smsc-ircc2.ircc_cfg= [HW] Device configuration I/O port
4899 smsc-ircc2.ircc_sir= [HW] SIR base I/O port
4900 smsc-ircc2.ircc_fir= [HW] FIR base I/O port
4901 smsc-ircc2.ircc_irq= [HW] IRQ line
4902 smsc-ircc2.ircc_dma= [HW] DMA channel
4903 smsc-ircc2.ircc_transceiver= [HW] Transceiver type:
4904 0: Toshiba Satellite 1800 (GP data pin select)
4905 1: Fast pin select (default)
4908 smt [KNL,S390] Set the maximum number of threads (logical
4909 CPUs) to use per physical CPU on systems capable of
4910 symmetric multithreading (SMT). Will be capped to the
4911 actual hardware limit.
4913 Default: -1 (no limit)
4916 [KNL] Should the soft-lockup detector generate panics.
4919 A value of 1 instructs the soft-lockup detector
4920 to panic the machine when a soft-lockup occurs. It is
4921 also controlled by the kernel.softlockup_panic sysctl
4922 and CONFIG_BOOTPARAM_SOFTLOCKUP_PANIC, which is the
4923 respective build-time switch to that functionality.
4925 softlockup_all_cpu_backtrace=
4926 [KNL] Should the soft-lockup detector generate
4927 backtraces on all cpus.
4930 sonypi.*= [HW] Sony Programmable I/O Control Device driver
4931 See Documentation/admin-guide/laptops/sonypi.rst
4933 spectre_v2= [X86] Control mitigation of Spectre variant 2
4934 (indirect branch speculation) vulnerability.
4935 The default operation protects the kernel from
4938 on - unconditionally enable, implies
4940 off - unconditionally disable, implies
4942 auto - kernel detects whether your CPU model is
4945 Selecting 'on' will, and 'auto' may, choose a
4946 mitigation method at run time according to the
4947 CPU, the available microcode, the setting of the
4948 CONFIG_RETPOLINE configuration option, and the
4949 compiler with which the kernel was built.
4951 Selecting 'on' will also enable the mitigation
4952 against user space to user space task attacks.
4954 Selecting 'off' will disable both the kernel and
4955 the user space protections.
4957 Specific mitigations can also be selected manually:
4959 retpoline - replace indirect branches
4960 retpoline,generic - google's original retpoline
4961 retpoline,amd - AMD-specific minimal thunk
4963 Not specifying this option is equivalent to
4967 [X86] Control mitigation of Spectre variant 2
4968 (indirect branch speculation) vulnerability between
4971 on - Unconditionally enable mitigations. Is
4972 enforced by spectre_v2=on
4974 off - Unconditionally disable mitigations. Is
4975 enforced by spectre_v2=off
4977 prctl - Indirect branch speculation is enabled,
4978 but mitigation can be enabled via prctl
4979 per thread. The mitigation control state
4980 is inherited on fork.
4983 - Like "prctl" above, but only STIBP is
4984 controlled per thread. IBPB is issued
4985 always when switching between different user
4989 - Same as "prctl" above, but all seccomp
4990 threads will enable the mitigation unless
4991 they explicitly opt out.
4994 - Like "seccomp" above, but only STIBP is
4995 controlled per thread. IBPB is issued
4996 always when switching between different
4997 user space processes.
4999 auto - Kernel selects the mitigation depending on
5000 the available CPU features and vulnerability.
5003 If CONFIG_SECCOMP=y then "seccomp", otherwise "prctl"
5005 Not specifying this option is equivalent to
5006 spectre_v2_user=auto.
5008 spec_store_bypass_disable=
5009 [HW] Control Speculative Store Bypass (SSB) Disable mitigation
5010 (Speculative Store Bypass vulnerability)
5012 Certain CPUs are vulnerable to an exploit against a
5013 a common industry wide performance optimization known
5014 as "Speculative Store Bypass" in which recent stores
5015 to the same memory location may not be observed by
5016 later loads during speculative execution. The idea
5017 is that such stores are unlikely and that they can
5018 be detected prior to instruction retirement at the
5019 end of a particular speculation execution window.
5021 In vulnerable processors, the speculatively forwarded
5022 store can be used in a cache side channel attack, for
5023 example to read memory to which the attacker does not
5024 directly have access (e.g. inside sandboxed code).
5026 This parameter controls whether the Speculative Store
5027 Bypass optimization is used.
5029 On x86 the options are:
5031 on - Unconditionally disable Speculative Store Bypass
5032 off - Unconditionally enable Speculative Store Bypass
5033 auto - Kernel detects whether the CPU model contains an
5034 implementation of Speculative Store Bypass and
5035 picks the most appropriate mitigation. If the
5036 CPU is not vulnerable, "off" is selected. If the
5037 CPU is vulnerable the default mitigation is
5038 architecture and Kconfig dependent. See below.
5039 prctl - Control Speculative Store Bypass per thread
5040 via prctl. Speculative Store Bypass is enabled
5041 for a process by default. The state of the control
5042 is inherited on fork.
5043 seccomp - Same as "prctl" above, but all seccomp threads
5044 will disable SSB unless they explicitly opt out.
5046 Default mitigations:
5047 X86: If CONFIG_SECCOMP=y "seccomp", otherwise "prctl"
5049 On powerpc the options are:
5051 on,auto - On Power8 and Power9 insert a store-forwarding
5052 barrier on kernel entry and exit. On Power7
5053 perform a software flush on kernel entry and
5057 Not specifying this option is equivalent to
5058 spec_store_bypass_disable=auto.
5060 spia_io_base= [HW,MTD]
5066 [X86] Enable split lock detection
5068 When enabled (and if hardware support is present), atomic
5069 instructions that access data across cache line
5070 boundaries will result in an alignment check exception.
5074 warn - the kernel will emit rate limited warnings
5075 about applications triggering the #AC
5076 exception. This mode is the default on CPUs
5077 that supports split lock detection.
5079 fatal - the kernel will send SIGBUS to applications
5080 that trigger the #AC exception.
5082 If an #AC exception is hit in the kernel or in
5083 firmware (i.e. not while executing in user mode)
5084 the kernel will oops in either "warn" or "fatal"
5088 Control the Special Register Buffer Data Sampling
5091 Certain CPUs are vulnerable to an MDS-like
5092 exploit which can leak bits from the random
5095 By default, this issue is mitigated by
5096 microcode. However, the microcode fix can cause
5097 the RDRAND and RDSEED instructions to become
5098 much slower. Among other effects, this will
5099 result in reduced throughput from /dev/urandom.
5101 The microcode mitigation can be disabled with
5102 the following option:
5104 off: Disable mitigation and remove
5105 performance impact to RDRAND and RDSEED
5107 srcutree.counter_wrap_check [KNL]
5108 Specifies how frequently to check for
5109 grace-period sequence counter wrap for the
5110 srcu_data structure's ->srcu_gp_seq_needed field.
5111 The greater the number of bits set in this kernel
5112 parameter, the less frequently counter wrap will
5113 be checked for. Note that the bottom two bits
5116 srcutree.exp_holdoff [KNL]
5117 Specifies how many nanoseconds must elapse
5118 since the end of the last SRCU grace period for
5119 a given srcu_struct until the next normal SRCU
5120 grace period will be considered for automatic
5121 expediting. Set to zero to disable automatic
5125 Speculative Store Bypass Disable control
5127 On CPUs that are vulnerable to the Speculative
5128 Store Bypass vulnerability and offer a
5129 firmware based mitigation, this parameter
5130 indicates how the mitigation should be used:
5132 force-on: Unconditionally enable mitigation for
5133 for both kernel and userspace
5134 force-off: Unconditionally disable mitigation for
5135 for both kernel and userspace
5136 kernel: Always enable mitigation in the
5137 kernel, and offer a prctl interface
5138 to allow userspace to register its
5139 interest in being mitigated too.
5141 stack_guard_gap= [MM]
5142 override the default stack gap protection. The value
5143 is in page units and it defines how many pages prior
5144 to (for stacks growing down) resp. after (for stacks
5145 growing up) the main stack are reserved for no other
5146 mapping. Default value is 256 pages.
5149 Enabled the stack tracer on boot up.
5151 stacktrace_filter=[function-list]
5152 [FTRACE] Limit the functions that the stack tracer
5153 will trace at boot up. function-list is a comma-separated
5154 list of functions. This list can be changed at run
5155 time by the stack_trace_filter file in the debugfs
5156 tracing directory. Note, this enables stack tracing
5157 and the stacktrace above is not needed.
5161 Set the STI (builtin display/keyboard on the HP-PARISC
5162 machines) console (graphic card) which should be used
5163 as the initial boot-console.
5164 See also comment in drivers/video/console/sticore.c.
5167 See comment in drivers/video/console/sticore.c.
5170 Format: bpp:<bpp1>[:<bpp2>[:<bpp3>...]]
5172 sunrpc.min_resvport=
5173 sunrpc.max_resvport=
5175 SunRPC servers often require that client requests
5176 originate from a privileged port (i.e. a port in the
5177 range 0 < portnr < 1024).
5178 An administrator who wishes to reserve some of these
5179 ports for other uses may adjust the range that the
5180 kernel's sunrpc client considers to be privileged
5181 using these two parameters to set the minimum and
5182 maximum port values.
5184 sunrpc.svc_rpc_per_connection_limit=
5186 Limit the number of requests that the server will
5187 process in parallel from a single connection.
5188 The default value is 0 (no limit).
5192 Control how the NFS server code allocates CPUs to
5193 service thread pools. Depending on how many NICs
5194 you have and where their interrupts are bound, this
5195 option will affect which CPUs will do NFS serving.
5196 Note: this parameter cannot be changed while the
5197 NFS server is running.
5199 auto the server chooses an appropriate mode
5200 automatically using heuristics
5201 global a single global pool contains all CPUs
5202 percpu one pool for each CPU
5203 pernode one pool for each NUMA node (equivalent
5204 to global on non-NUMA machines)
5206 sunrpc.tcp_slot_table_entries=
5207 sunrpc.udp_slot_table_entries=
5209 Sets the upper limit on the number of simultaneous
5210 RPC calls that can be sent from the client to a
5211 server. Increasing these values may allow you to
5212 improve throughput, but will also increase the
5213 amount of memory reserved for use by the client.
5215 suspend.pm_test_delay=
5217 Sets the number of seconds to remain in a suspend test
5218 mode before resuming the system (see
5219 /sys/power/pm_test). Only available when CONFIG_PM_DEBUG
5220 is set. Default value is 5.
5223 Format: { on | off | y | n | 1 | 0 }
5224 This parameter controls use of the Protected
5225 Execution Facility on pSeries.
5228 [KNL] Enable accounting of swap in memory resource
5229 controller if no parameter or 1 is given or disable
5230 it if 0 is given (See Documentation/admin-guide/cgroup-v1/memory.rst)
5232 swiotlb= [ARM,IA-64,PPC,MIPS,X86]
5233 Format: { <int> | force | noforce }
5234 <int> -- Number of I/O TLB slabs
5235 force -- force using of bounce buffers even if they
5236 wouldn't be automatically used by the kernel
5237 noforce -- Never use bounce buffers (for debugging)
5242 Set a sysctl parameter, right before loading the init
5243 process, as if the value was written to the respective
5244 /proc/sys/... file. Both '.' and '/' are recognized as
5245 separators. Unrecognized parameters and invalid values
5246 are reported in the kernel log. Sysctls registered
5247 later by a loaded module cannot be set this way.
5248 Example: sysctl.vm.swappiness=40
5250 sysfs.deprecated=0|1 [KNL]
5251 Enable/disable old style sysfs layout for old udev
5252 on older distributions. When this option is enabled
5253 very new udev will not work anymore. When this option
5254 is disabled (or CONFIG_SYSFS_DEPRECATED not compiled)
5255 in older udev will not work anymore.
5256 Default depends on CONFIG_SYSFS_DEPRECATED_V2 set in
5257 the kernel configuration.
5259 sysrq_always_enabled
5261 Ignore sysrq setting - this boot parameter will
5262 neutralize any effect of /proc/sys/kernel/sysrq.
5263 Useful for debugging.
5265 tcpmhash_entries= [KNL,NET]
5266 Set the number of tcp_metrics_hash slots.
5267 Default value is 8192 or 16384 depending on total
5268 ram pages. This is used to specify the TCP metrics
5269 cache size. See Documentation/networking/ip-sysctl.rst
5270 "tcp_no_metrics_save" section for more details.
5274 test_suspend= [SUSPEND][,N]
5275 Specify "mem" (for Suspend-to-RAM) or "standby" (for
5276 standby suspend) or "freeze" (for suspend type freeze)
5277 as the system sleep state during system startup with
5278 the optional capability to repeat N number of times.
5279 The system is woken from this state using a
5280 wakeup-capable RTC alarm.
5282 thash_entries= [KNL,NET]
5283 Set number of hash buckets for TCP connection
5285 thermal.act= [HW,ACPI]
5286 -1: disable all active trip points in all thermal zones
5287 <degrees C>: override all lowest active trip points
5289 thermal.crt= [HW,ACPI]
5290 -1: disable all critical trip points in all thermal zones
5291 <degrees C>: override all critical trip points
5293 thermal.nocrt= [HW,ACPI]
5294 Set to disable actions on ACPI thermal zone
5295 critical and hot trip points.
5297 thermal.off= [HW,ACPI]
5298 1: disable ACPI thermal control
5300 thermal.psv= [HW,ACPI]
5301 -1: disable all passive trip points
5302 <degrees C>: override all passive trip points to this
5305 thermal.tzp= [HW,ACPI]
5306 Specify global default ACPI thermal zone polling rate
5307 <deci-seconds>: poll all this frequency
5308 0: no polling (default)
5311 Force threading of all interrupt handlers except those
5312 marked explicitly IRQF_NO_THREAD.
5316 Specify if the kernel should make use of the cpu
5317 topology information if the hardware supports this.
5318 The scheduler will make use of this information and
5319 e.g. base its process migration decisions on it.
5322 topology_updates= [KNL, PPC, NUMA]
5324 Specify if the kernel should ignore (off)
5325 topology updates sent by the hypervisor to this
5328 torture.disable_onoff_at_boot= [KNL]
5329 Prevent the CPU-hotplug component of torturing
5330 until after init has spawned.
5332 torture.ftrace_dump_at_shutdown= [KNL]
5333 Dump the ftrace buffer at torture-test shutdown,
5334 even if there were no errors. This can be a
5335 very costly operation when many torture tests
5336 are running concurrently, especially on systems
5337 with rotating-rust storage.
5339 torture.verbose_sleep_frequency= [KNL]
5340 Specifies how many verbose printk()s should be
5341 emitted between each sleep. The default of zero
5342 disables verbose-printk() sleeping.
5344 torture.verbose_sleep_duration= [KNL]
5345 Duration of each verbose-printk() sleep in jiffies.
5349 tpm_suspend_pcr=[HW,TPM]
5350 Format: integer pcr id
5351 Specify that at suspend time, the tpm driver
5352 should extend the specified pcr with zeros,
5353 as a workaround for some chips which fail to
5354 flush the last written pcr on TPM_SaveState.
5355 This will guarantee that all the other pcrs
5358 trace_buf_size=nn[KMG]
5359 [FTRACE] will set tracing buffer size on each cpu.
5361 trace_event=[event-list]
5362 [FTRACE] Set and start specified trace events in order
5363 to facilitate early boot debugging. The event-list is a
5364 comma-separated list of trace events to enable. See
5365 also Documentation/trace/events.rst
5367 trace_options=[option-list]
5368 [FTRACE] Enable or disable tracer options at boot.
5369 The option-list is a comma delimited list of options
5370 that can be enabled or disabled just as if you were
5371 to echo the option name into
5373 /sys/kernel/debug/tracing/trace_options
5375 For example, to enable stacktrace option (to dump the
5376 stack trace of each event), add to the command line:
5378 trace_options=stacktrace
5380 See also Documentation/trace/ftrace.rst "trace options"
5384 Have the tracepoints sent to printk as well as the
5385 tracing ring buffer. This is useful for early boot up
5386 where the system hangs or reboots and does not give the
5387 option for reading the tracing buffer or performing a
5388 ftrace_dump_on_oops.
5390 To turn off having tracepoints sent to printk,
5391 echo 0 > /proc/sys/kernel/tracepoint_printk
5392 Note, echoing 1 into this file without the
5393 tracepoint_printk kernel cmdline option has no effect.
5397 Having tracepoints sent to printk() and activating high
5398 frequency tracepoints such as irq or sched, can cause
5399 the system to live lock.
5402 [FTRACE] enable this option to disable tracing when a
5403 warning is hit. This turns off "tracing_on". Tracing can
5404 be enabled again by echoing '1' into the "tracing_on"
5405 file located in /sys/kernel/debug/tracing/
5407 This option is useful, as it disables the trace before
5408 the WARNING dump is called, which prevents the trace to
5409 be filled with content caused by the warning output.
5411 This option can also be set at run time via the sysctl
5412 option: kernel/traceoff_on_warning
5414 transparent_hugepage=
5416 Format: [always|madvise|never]
5417 Can be used to control the default behavior of the system
5418 with respect to transparent hugepages.
5419 See Documentation/admin-guide/mm/transhuge.rst
5422 tsc= Disable clocksource stability checks for TSC.
5424 [x86] reliable: mark tsc clocksource as reliable, this
5425 disables clocksource verification at runtime, as well
5426 as the stability checks done at bootup. Used to enable
5427 high-resolution timer mode on older hardware, and in
5428 virtualized environment.
5429 [x86] noirqtime: Do not use TSC to do irq accounting.
5430 Used to run time disable IRQ_TIME_ACCOUNTING on any
5431 platforms where RDTSC is slow and this accounting
5433 [x86] unstable: mark the TSC clocksource as unstable, this
5434 marks the TSC unconditionally unstable at bootup and
5435 avoids any further wobbles once the TSC watchdog notices.
5436 [x86] nowatchdog: disable clocksource watchdog. Used
5437 in situations with strict latency requirements (where
5438 interruptions from clocksource watchdog are not
5441 tsc_early_khz= [X86] Skip early TSC calibration and use the given
5442 value instead. Useful when the early TSC frequency discovery
5443 procedure is not reliable, such as on overclocked systems
5444 with CPUID.16h support and partial CPUID.15h support.
5445 Format: <unsigned int>
5447 tsx= [X86] Control Transactional Synchronization
5448 Extensions (TSX) feature in Intel processors that
5449 support TSX control.
5451 This parameter controls the TSX feature. The options are:
5453 on - Enable TSX on the system. Although there are
5454 mitigations for all known security vulnerabilities,
5455 TSX has been known to be an accelerator for
5456 several previous speculation-related CVEs, and
5457 so there may be unknown security risks associated
5458 with leaving it enabled.
5460 off - Disable TSX on the system. (Note that this
5461 option takes effect only on newer CPUs which are
5462 not vulnerable to MDS, i.e., have
5463 MSR_IA32_ARCH_CAPABILITIES.MDS_NO=1 and which get
5464 the new IA32_TSX_CTRL MSR through a microcode
5465 update. This new MSR allows for the reliable
5466 deactivation of the TSX functionality.)
5468 auto - Disable TSX if X86_BUG_TAA is present,
5469 otherwise enable TSX on the system.
5471 Not specifying this option is equivalent to tsx=off.
5473 See Documentation/admin-guide/hw-vuln/tsx_async_abort.rst
5476 tsx_async_abort= [X86,INTEL] Control mitigation for the TSX Async
5477 Abort (TAA) vulnerability.
5479 Similar to Micro-architectural Data Sampling (MDS)
5480 certain CPUs that support Transactional
5481 Synchronization Extensions (TSX) are vulnerable to an
5482 exploit against CPU internal buffers which can forward
5483 information to a disclosure gadget under certain
5486 In vulnerable processors, the speculatively forwarded
5487 data can be used in a cache side channel attack, to
5488 access data to which the attacker does not have direct
5491 This parameter controls the TAA mitigation. The
5494 full - Enable TAA mitigation on vulnerable CPUs
5497 full,nosmt - Enable TAA mitigation and disable SMT on
5498 vulnerable CPUs. If TSX is disabled, SMT
5499 is not disabled because CPU is not
5500 vulnerable to cross-thread TAA attacks.
5501 off - Unconditionally disable TAA mitigation
5503 On MDS-affected machines, tsx_async_abort=off can be
5504 prevented by an active MDS mitigation as both vulnerabilities
5505 are mitigated with the same mechanism so in order to disable
5506 this mitigation, you need to specify mds=off too.
5508 Not specifying this option is equivalent to
5509 tsx_async_abort=full. On CPUs which are MDS affected
5510 and deploy MDS mitigation, TAA mitigation is not
5511 required and doesn't provide any additional
5515 Documentation/admin-guide/hw-vuln/tsx_async_abort.rst
5517 turbografx.map[2|3]= [HW,JOY]
5518 TurboGraFX parallel port interface
5520 <port#>,<js1>,<js2>,<js3>,<js4>,<js5>,<js6>,<js7>
5521 See also Documentation/input/devices/joystick-parport.rst
5523 udbg-immortal [PPC] When debugging early kernel crashes that
5524 happen after console_init() and before a proper
5525 console driver takes over, this boot options might
5526 help "seeing" what's going on.
5528 uhash_entries= [KNL,NET]
5529 Set number of hash buckets for UDP/UDP-Lite connections
5532 [USB] Ignore overcurrent events (default N).
5533 Some badly-designed motherboards generate lots of
5534 bogus events, for ports that aren't wired to
5535 anything. Set this parameter to avoid log spamming.
5536 Note that genuine overcurrent events won't be
5540 [X86] Cause panic on unknown NMI.
5542 usbcore.authorized_default=
5543 [USB] Default USB device authorization:
5544 (default -1 = authorized except for wireless USB,
5545 0 = not authorized, 1 = authorized, 2 = authorized
5546 if device connected to internal port)
5548 usbcore.autosuspend=
5549 [USB] The autosuspend time delay (in seconds) used
5550 for newly-detected USB devices (default 2). This
5551 is the time required before an idle device will be
5552 autosuspended. Devices for which the delay is set
5553 to a negative value won't be autosuspended at all.
5555 usbcore.usbfs_snoop=
5556 [USB] Set to log all usbfs traffic (default 0 = off).
5558 usbcore.usbfs_snoop_max=
5559 [USB] Maximum number of bytes to snoop in each URB
5562 usbcore.blinkenlights=
5563 [USB] Set to cycle leds on hubs (default 0 = off).
5565 usbcore.old_scheme_first=
5566 [USB] Start with the old device initialization
5567 scheme (default 0 = off).
5569 usbcore.usbfs_memory_mb=
5570 [USB] Memory limit (in MB) for buffers allocated by
5571 usbfs (default = 16, 0 = max = 2047).
5573 usbcore.use_both_schemes=
5574 [USB] Try the other device initialization scheme
5575 if the first one fails (default 1 = enabled).
5577 usbcore.initial_descriptor_timeout=
5578 [USB] Specifies timeout for the initial 64-byte
5579 USB_REQ_GET_DESCRIPTOR request in milliseconds
5580 (default 5000 = 5.0 seconds).
5582 usbcore.nousb [USB] Disable the USB subsystem
5585 [USB] A list of quirk entries to augment the built-in
5586 usb core quirk list. List entries are separated by
5587 commas. Each entry has the form
5588 VendorID:ProductID:Flags. The IDs are 4-digit hex
5589 numbers and Flags is a set of letters. Each letter
5590 will change the built-in quirk; setting it if it is
5591 clear and clearing it if it is set. The letters have
5592 the following meanings:
5593 a = USB_QUIRK_STRING_FETCH_255 (string
5594 descriptors must not be fetched using
5596 b = USB_QUIRK_RESET_RESUME (device can't resume
5597 correctly so reset it instead);
5598 c = USB_QUIRK_NO_SET_INTF (device can't handle
5599 Set-Interface requests);
5600 d = USB_QUIRK_CONFIG_INTF_STRINGS (device can't
5601 handle its Configuration or Interface
5603 e = USB_QUIRK_RESET (device can't be reset
5604 (e.g morph devices), don't use reset);
5605 f = USB_QUIRK_HONOR_BNUMINTERFACES (device has
5606 more interface descriptions than the
5607 bNumInterfaces count, and can't handle
5608 talking to these interfaces);
5609 g = USB_QUIRK_DELAY_INIT (device needs a pause
5610 during initialization, after we read
5611 the device descriptor);
5612 h = USB_QUIRK_LINEAR_UFRAME_INTR_BINTERVAL (For
5613 high speed and super speed interrupt
5614 endpoints, the USB 2.0 and USB 3.0 spec
5615 require the interval in microframes (1
5616 microframe = 125 microseconds) to be
5617 calculated as interval = 2 ^
5619 Devices with this quirk report their
5620 bInterval as the result of this
5621 calculation instead of the exponent
5622 variable used in the calculation);
5623 i = USB_QUIRK_DEVICE_QUALIFIER (device can't
5624 handle device_qualifier descriptor
5626 j = USB_QUIRK_IGNORE_REMOTE_WAKEUP (device
5627 generates spurious wakeup, ignore
5628 remote wakeup capability);
5629 k = USB_QUIRK_NO_LPM (device can't handle Link
5631 l = USB_QUIRK_LINEAR_FRAME_INTR_BINTERVAL
5632 (Device reports its bInterval as linear
5633 frames instead of the USB 2.0
5635 m = USB_QUIRK_DISCONNECT_SUSPEND (Device needs
5636 to be disconnected before suspend to
5637 prevent spurious wakeup);
5638 n = USB_QUIRK_DELAY_CTRL_MSG (Device needs a
5639 pause after every control message);
5640 o = USB_QUIRK_HUB_SLOW_RESET (Hub needs extra
5641 delay after resetting its port);
5642 Example: quirks=0781:5580:bk,0a5c:5834:gij
5645 [USBHID] The interval which mice are to be polled at.
5648 [USBHID] The interval which joysticks are to be polled at.
5651 [USBHID] The interval which keyboards are to be polled at.
5653 usb-storage.delay_use=
5654 [UMS] The delay in seconds before a new device is
5655 scanned for Logical Units (default 1).
5658 [UMS] A list of quirks entries to supplement or
5659 override the built-in unusual_devs list. List
5660 entries are separated by commas. Each entry has
5661 the form VID:PID:Flags where VID and PID are Vendor
5662 and Product ID values (4-digit hex numbers) and
5663 Flags is a set of characters, each corresponding
5664 to a common usb-storage quirk flag as follows:
5665 a = SANE_SENSE (collect more than 18 bytes
5666 of sense data, not on uas);
5667 b = BAD_SENSE (don't collect more than 18
5668 bytes of sense data, not on uas);
5669 c = FIX_CAPACITY (decrease the reported
5670 device capacity by one sector);
5671 d = NO_READ_DISC_INFO (don't use
5672 READ_DISC_INFO command, not on uas);
5673 e = NO_READ_CAPACITY_16 (don't use
5674 READ_CAPACITY_16 command);
5675 f = NO_REPORT_OPCODES (don't use report opcodes
5677 g = MAX_SECTORS_240 (don't transfer more than
5678 240 sectors at a time, uas only);
5679 h = CAPACITY_HEURISTICS (decrease the
5680 reported device capacity by one
5681 sector if the number is odd);
5682 i = IGNORE_DEVICE (don't bind to this
5684 j = NO_REPORT_LUNS (don't use report luns
5686 k = NO_SAME (do not use WRITE_SAME, uas only)
5687 l = NOT_LOCKABLE (don't try to lock and
5688 unlock ejectable media, not on uas);
5689 m = MAX_SECTORS_64 (don't transfer more
5690 than 64 sectors = 32 KB at a time,
5692 n = INITIAL_READ10 (force a retry of the
5693 initial READ(10) command, not on uas);
5694 o = CAPACITY_OK (accept the capacity
5695 reported by the device, not on uas);
5696 p = WRITE_CACHE (the device cache is ON
5697 by default, not on uas);
5698 r = IGNORE_RESIDUE (the device reports
5699 bogus residue values, not on uas);
5700 s = SINGLE_LUN (the device has only one
5702 t = NO_ATA_1X (don't allow ATA(12) and ATA(16)
5703 commands, uas only);
5704 u = IGNORE_UAS (don't bind to the uas driver);
5705 w = NO_WP_DETECT (don't test whether the
5706 medium is write-protected).
5707 y = ALWAYS_SYNC (issue a SYNCHRONIZE_CACHE
5708 even if the device claims no cache,
5710 Example: quirks=0419:aaf5:rl,0421:0433:rc
5712 user_debug= [KNL,ARM]
5714 See arch/arm/Kconfig.debug help text.
5715 1 - undefined instruction events
5717 4 - invalid data aborts
5720 Example: user_debug=31
5723 [X86] Flags controlling user PTE allocations.
5725 nohigh = do not allocate PTE pages in
5726 HIGHMEM regardless of setting
5730 On X86_32, this is an alias for vdso32=. Otherwise:
5732 vdso=1: enable VDSO (the default)
5733 vdso=0: disable VDSO mapping
5735 vdso32= [X86] Control the 32-bit vDSO
5736 vdso32=1: enable 32-bit VDSO
5737 vdso32=0 or vdso32=2: disable 32-bit VDSO
5739 See the help text for CONFIG_COMPAT_VDSO for more
5740 details. If CONFIG_COMPAT_VDSO is set, the default is
5741 vdso32=0; otherwise, the default is vdso32=1.
5743 For compatibility with older kernels, vdso32=2 is an
5746 Try vdso32=0 if you encounter an error that says:
5747 dl_main: Assertion `(void *) ph->p_vaddr == _rtld_local._dl_sysinfo_dso' failed!
5750 vector=percpu: enable percpu vector domain
5752 video= [FB] Frame buffer configuration
5753 See Documentation/fb/modedb.rst.
5755 video.brightness_switch_enabled= [0,1]
5756 If set to 1, on receiving an ACPI notify event
5757 generated by hotkey, video driver will adjust brightness
5758 level and then send out the event to user space through
5759 the allocated input device; If set to 0, video driver
5760 will only send out the event without touching backlight
5765 [VMMIO] Memory mapped virtio (platform) device.
5767 <size>@<baseaddr>:<irq>[:<id>]
5769 <size> := size (can use standard suffixes
5771 <baseaddr> := physical base address
5772 <irq> := interrupt number (as passed to
5774 <id> := (optional) platform device id
5776 virtio_mmio.device=1K@0x100b0000:48:7
5778 Can be used multiple times for multiple devices.
5780 vga= [BOOT,X86-32] Select a particular video mode
5781 See Documentation/x86/boot.rst and
5782 Documentation/admin-guide/svga.rst.
5783 Use vga=ask for menu.
5784 This is actually a boot loader parameter; the value is
5785 passed to the kernel using a special protocol.
5787 vm_debug[=options] [KNL] Available with CONFIG_DEBUG_VM=y.
5788 May slow down system boot speed, especially when
5789 enabled on systems with a large amount of memory.
5790 All options are enabled by default, and this
5791 interface is meant to allow for selectively
5792 enabling or disabling specific virtual memory
5795 Available options are:
5796 P Enable page structure init time poisoning
5797 - Disable all of the above options
5799 vmalloc=nn[KMG] [KNL,BOOT] Forces the vmalloc area to have an exact
5800 size of <nn>. This can be used to increase the
5801 minimum size (128MB on x86). It can also be used to
5802 decrease the size and leave more room for directly
5805 vmcp_cma=nn[MG] [KNL,S390]
5806 Sets the memory size reserved for contiguous memory
5807 allocations for the vmcp device driver.
5809 vmhalt= [KNL,S390] Perform z/VM CP command after system halt.
5812 vmpanic= [KNL,S390] Perform z/VM CP command after kernel panic.
5815 vmpoff= [KNL,S390] Perform z/VM CP command after power off.
5819 Controls the behavior of vsyscalls (i.e. calls to
5820 fixed addresses of 0xffffffffff600x00 from legacy
5821 code). Most statically-linked binaries and older
5822 versions of glibc use these calls. Because these
5823 functions are at fixed addresses, they make nice
5824 targets for exploits that can control RIP.
5826 emulate [default] Vsyscalls turn into traps and are
5827 emulated reasonably safely. The vsyscall
5830 xonly Vsyscalls turn into traps and are
5831 emulated reasonably safely. The vsyscall
5832 page is not readable.
5834 none Vsyscalls don't work at all. This makes
5835 them quite hard to use for exploits but
5836 might break your system.
5838 vt.color= [VT] Default text color.
5839 Format: 0xYX, X = foreground, Y = background.
5840 Default: 0x07 = light gray on black.
5842 vt.cur_default= [VT] Default cursor shape.
5843 Format: 0xCCBBAA, where AA, BB, and CC are the same as
5844 the parameters of the <Esc>[?A;B;Cc escape sequence;
5845 see VGA-softcursor.txt. Default: 2 = underline.
5847 vt.default_blu= [VT]
5848 Format: <blue0>,<blue1>,<blue2>,...,<blue15>
5849 Change the default blue palette of the console.
5850 This is a 16-member array composed of values
5853 vt.default_grn= [VT]
5854 Format: <green0>,<green1>,<green2>,...,<green15>
5855 Change the default green palette of the console.
5856 This is a 16-member array composed of values
5859 vt.default_red= [VT]
5860 Format: <red0>,<red1>,<red2>,...,<red15>
5861 Change the default red palette of the console.
5862 This is a 16-member array composed of values
5868 Set system-wide default UTF-8 mode for all tty's.
5869 Default is 1, i.e. UTF-8 mode is enabled for all
5870 newly opened terminals.
5872 vt.global_cursor_default=
5875 Set system-wide default for whether a cursor
5876 is shown on new VTs. Default is -1,
5877 i.e. cursors will be created by default unless
5878 overridden by individual drivers. 0 will hide
5879 cursors, 1 will display them.
5881 vt.italic= [VT] Default color for italic text; 0-15.
5884 vt.underline= [VT] Default color for underlined text; 0-15.
5887 watchdog timers [HW,WDT] For information on watchdog timers,
5888 see Documentation/watchdog/watchdog-parameters.rst
5889 or other driver-specific files in the
5890 Documentation/watchdog/ directory.
5894 Set the hard lockup detector stall duration
5895 threshold in seconds. The soft lockup detector
5896 threshold is set to twice the value. A value of 0
5897 disables both lockup detectors. Default is 10
5900 workqueue.watchdog_thresh=
5901 If CONFIG_WQ_WATCHDOG is configured, workqueue can
5902 warn stall conditions and dump internal state to
5903 help debugging. 0 disables workqueue stall
5904 detection; otherwise, it's the stall threshold
5905 duration in seconds. The default value is 30 and
5906 it can be updated at runtime by writing to the
5907 corresponding sysfs file.
5909 workqueue.disable_numa
5910 By default, all work items queued to unbound
5911 workqueues are affine to the NUMA nodes they're
5912 issued on, which results in better behavior in
5913 general. If NUMA affinity needs to be disabled for
5914 whatever reason, this option can be used. Note
5915 that this also can be controlled per-workqueue for
5916 workqueues visible under /sys/bus/workqueue/.
5918 workqueue.power_efficient
5919 Per-cpu workqueues are generally preferred because
5920 they show better performance thanks to cache
5921 locality; unfortunately, per-cpu workqueues tend to
5922 be more power hungry than unbound workqueues.
5924 Enabling this makes the per-cpu workqueues which
5925 were observed to contribute significantly to power
5926 consumption unbound, leading to measurably lower
5927 power usage at the cost of small performance
5930 The default value of this parameter is determined by
5931 the config option CONFIG_WQ_POWER_EFFICIENT_DEFAULT.
5933 workqueue.debug_force_rr_cpu
5934 Workqueue used to implicitly guarantee that work
5935 items queued without explicit CPU specified are put
5936 on the local CPU. This guarantee is no longer true
5937 and while local CPU is still preferred work items
5938 may be put on foreign CPUs. This debug option
5939 forces round-robin CPU selection to flush out
5940 usages which depend on the now broken guarantee.
5941 When enabled, memory and cache locality will be
5944 x2apic_phys [X86-64,APIC] Use x2apic physical mode instead of
5945 default x2apic cluster mode on platforms
5948 x86_intel_mid_timer= [X86-32,APBT]
5949 Choose timer option for x86 Intel MID platform.
5950 Two valid options are apbt timer only and lapic timer
5951 plus one apbt timer for broadcast timer.
5952 x86_intel_mid_timer=apbt_only | lapic_and_apbt
5954 xen_512gb_limit [KNL,X86-64,XEN]
5955 Restricts the kernel running paravirtualized under Xen
5956 to use only up to 512 GB of RAM. The reason to do so is
5957 crash analysis tools and Xen tools for doing domain
5958 save/restore/migration must be enabled to handle larger
5961 xen_emul_unplug= [HW,X86,XEN]
5962 Unplug Xen emulated devices
5963 Format: [unplug0,][unplug1]
5964 ide-disks -- unplug primary master IDE devices
5965 aux-ide-disks -- unplug non-primary-master IDE devices
5966 nics -- unplug network devices
5967 all -- unplug all emulated devices (NICs and IDE disks)
5968 unnecessary -- unplugging emulated devices is
5969 unnecessary even if the host did not respond to
5971 never -- do not unplug even if version check succeeds
5973 xen_legacy_crash [X86,XEN]
5974 Crash from Xen panic notifier, without executing late
5975 panic() code such as dumping handler.
5977 xen_nopvspin [X86,XEN]
5978 Disables the qspinlock slowpath using Xen PV optimizations.
5979 This parameter is obsoleted by "nopvspin" parameter, which
5980 has equivalent effect for XEN platform.
5983 Disables the PV optimizations forcing the HVM guest to
5984 run as generic HVM guest with no PV drivers.
5985 This option is obsoleted by the "nopv" option, which
5986 has equivalent effect for XEN platform.
5988 xen_no_vector_callback
5989 [KNL,X86,XEN] Disable the vector callback for Xen
5990 event channel interrupts.
5992 xen_scrub_pages= [XEN]
5993 Boolean option to control scrubbing pages before giving them back
5994 to Xen, for use by other domains. Can be also changed at runtime
5995 with /sys/devices/system/xen_memory/xen_memory0/scrub_pages.
5996 Default value controlled with CONFIG_XEN_SCRUB_PAGES_DEFAULT.
5998 xen_timer_slop= [X86-64,XEN]
5999 Set the timer slop (in nanoseconds) for the virtual Xen
6000 timers (default is 100000). This adjusts the minimum
6001 delta of virtualized Xen timers, where lower values
6002 improve timer resolution at the expense of processing
6003 more timer interrupts.
6005 xen.event_eoi_delay= [XEN]
6006 How long to delay EOI handling in case of event
6007 storms (jiffies). Default is 10.
6009 xen.event_loop_timeout= [XEN]
6010 After which time (jiffies) the event handling loop
6011 should start to delay EOI handling. Default is 2.
6013 xen.fifo_events= [XEN]
6014 Boolean parameter to disable using fifo event handling
6015 even if available. Normally fifo event handling is
6016 preferred over the 2-level event handling, as it is
6017 fairer and the number of possible event channels is
6018 much higher. Default is on (use fifo events).
6020 nopv= [X86,XEN,KVM,HYPER_V,VMWARE]
6021 Disables the PV optimizations forcing the guest to run
6022 as generic guest with no PV drivers. Currently support
6023 XEN HVM, KVM, HYPER_V and VMWARE guest.
6025 nopvspin [X86,XEN,KVM]
6026 Disables the qspinlock slow path using PV optimizations
6027 which allow the hypervisor to 'idle' the guest on lock
6030 xirc2ps_cs= [NET,PCMCIA]
6032 <irq>,<irq_mask>,<io>,<full_duplex>,<do_sound>,<lockup_hack>[,<irq2>[,<irq3>[,<irq4>]]]
6035 By default on POWER9 and above, the kernel will
6036 natively use the XIVE interrupt controller. This option
6037 allows the fallback firmware mode to be used:
6039 off Fallback to firmware control of XIVE interrupt
6040 controller on both pseries and powernv
6041 platforms. Only useful on POWER9 and above.
6043 xhci-hcd.quirks [USB,KNL]
6044 A hex value specifying bitmask with supplemental xhci
6045 host controller quirks. Meaning of each bit can be
6046 consulted in header drivers/usb/host/xhci.h.
6049 Format: { early | on | rw | ro | off }
6050 Controls if xmon debugger is enabled. Default is off.
6051 Passing only "xmon" is equivalent to "xmon=early".
6052 early Call xmon as early as possible on boot; xmon
6053 debugger is called from setup_arch().
6054 on xmon debugger hooks will be installed so xmon
6055 is only called on a kernel crash. Default mode,
6056 i.e. either "ro" or "rw" mode, is controlled
6057 with CONFIG_XMON_DEFAULT_RO_MODE.
6058 rw xmon debugger hooks will be installed so xmon
6059 is called only on a kernel crash, mode is write,
6060 meaning SPR registers, memory and, other data
6061 can be written using xmon commands.
6062 ro same as "rw" option above but SPR registers,
6063 memory, and other data can't be written using
6065 off xmon is disabled.