1 acpi= [HW,ACPI,X86,ARM64]
2 Advanced Configuration and Power Interface
3 Format: { force | on | off | strict | noirq | rsdt |
5 force -- enable ACPI if default was off
6 on -- enable ACPI but allow fallback to DT [arm64]
7 off -- disable ACPI if default was on
8 noirq -- do not use ACPI for IRQ routing
9 strict -- Be less tolerant of platforms that are not
10 strictly ACPI specification compliant.
11 rsdt -- prefer RSDT over (default) XSDT
12 copy_dsdt -- copy DSDT to memory
13 For ARM64, ONLY "acpi=off", "acpi=on" or "acpi=force"
16 See also Documentation/power/runtime_pm.rst, pci=noacpi
18 acpi_apic_instance= [ACPI, IOAPIC]
20 2: use 2nd APIC table, if available
21 1,0: use 1st APIC table
24 acpi_backlight= [HW,ACPI]
25 { vendor | video | native | none }
26 If set to vendor, prefer vendor-specific driver
27 (e.g. thinkpad_acpi, sony_acpi, etc.) instead
28 of the ACPI video.ko driver.
29 If set to video, use the ACPI video.ko driver.
30 If set to native, use the device's native backlight mode.
31 If set to none, disable the ACPI backlight interface.
33 acpi_force_32bit_fadt_addr
34 force FADT to use 32 bit addresses rather than the
35 64 bit X_* addresses. Some firmware have broken 64
36 bit addresses for force ACPI ignore these and use
37 the older legacy 32 bit addresses.
39 acpica_no_return_repair [HW, ACPI]
40 Disable AML predefined validation mechanism
41 This mechanism can repair the evaluation result to make
42 the return objects more ACPI specification compliant.
43 This option is useful for developers to identify the
44 root cause of an AML interpreter issue when the issue
45 has something to do with the repair mechanism.
47 acpi.debug_layer= [HW,ACPI,ACPI_DEBUG]
48 acpi.debug_level= [HW,ACPI,ACPI_DEBUG]
50 CONFIG_ACPI_DEBUG must be enabled to produce any ACPI
51 debug output. Bits in debug_layer correspond to a
52 _COMPONENT in an ACPI source file, e.g.,
53 #define _COMPONENT ACPI_EVENTS
54 Bits in debug_level correspond to a level in
55 ACPI_DEBUG_PRINT statements, e.g.,
56 ACPI_DEBUG_PRINT((ACPI_DB_INFO, ...
57 The debug_level mask defaults to "info". See
58 Documentation/firmware-guide/acpi/debug.rst for more information about
59 debug layers and levels.
61 Enable processor driver info messages:
62 acpi.debug_layer=0x20000000
63 Enable AML "Debug" output, i.e., stores to the Debug
64 object while interpreting AML:
65 acpi.debug_layer=0xffffffff acpi.debug_level=0x2
66 Enable all messages related to ACPI hardware:
67 acpi.debug_layer=0x2 acpi.debug_level=0xffffffff
69 Some values produce so much output that the system is
70 unusable. The "log_buf_len" parameter may be useful
71 if you need to capture more output.
73 acpi_enforce_resources= [ACPI]
75 Check for resource conflicts between native drivers
76 and ACPI OperationRegions (SystemIO and SystemMemory
77 only). IO ports and memory declared in ACPI might be
78 used by the ACPI subsystem in arbitrary AML code and
79 can interfere with legacy drivers.
80 strict (default): access to resources claimed by ACPI
81 is denied; legacy drivers trying to access reserved
82 resources will fail to bind to device using them.
83 lax: access to resources claimed by ACPI is allowed;
84 legacy drivers trying to access reserved resources
85 will bind successfully but a warning message is logged.
86 no: ACPI OperationRegions are not marked as reserved,
87 no further checks are performed.
89 acpi_force_table_verification [HW,ACPI]
90 Enable table checksum verification during early stage.
91 By default, this is disabled due to x86 early mapping
94 acpi_irq_balance [HW,ACPI]
95 ACPI will balance active IRQs
98 acpi_irq_nobalance [HW,ACPI]
99 ACPI will not move active IRQs (default)
102 acpi_irq_isa= [HW,ACPI] If irq_balance, mark listed IRQs used by ISA
103 Format: <irq>,<irq>...
105 acpi_irq_pci= [HW,ACPI] If irq_balance, clear listed IRQs for
107 Format: <irq>,<irq>...
109 acpi_mask_gpe= [HW,ACPI]
110 Due to the existence of _Lxx/_Exx, some GPEs triggered
111 by unsupported hardware/firmware features can result in
112 GPE floodings that cannot be automatically disabled by
114 This facility can be used to prevent such uncontrolled
118 acpi_no_auto_serialize [HW,ACPI]
119 Disable auto-serialization of AML methods
120 AML control methods that contain the opcodes to create
121 named objects will be marked as "Serialized" by the
122 auto-serialization feature.
123 This feature is enabled by default.
124 This option allows to turn off the feature.
126 acpi_no_memhotplug [ACPI] Disable memory hotplug. Useful for kdump
129 acpi_no_static_ssdt [HW,ACPI]
130 Disable installation of static SSDTs at early boot time
131 By default, SSDTs contained in the RSDT/XSDT will be
132 installed automatically and they will appear under
133 /sys/firmware/acpi/tables.
134 This option turns off this feature.
135 Note that specifying this option does not affect
136 dynamic table installation which will install SSDT
137 tables to /sys/firmware/acpi/tables/dynamic.
139 acpi_no_watchdog [HW,ACPI,WDT]
140 Ignore the ACPI-based watchdog interface (WDAT) and let
141 a native driver control the watchdog device instead.
143 acpi_rsdp= [ACPI,EFI,KEXEC]
144 Pass the RSDP address to the kernel, mostly used
145 on machines running EFI runtime service to boot the
146 second kernel for kdump.
148 acpi_os_name= [HW,ACPI] Tell ACPI BIOS the name of the OS
149 Format: To spoof as Windows 98: ="Microsoft Windows"
151 acpi_rev_override [ACPI] Override the _REV object to return 5 (instead
152 of 2 which is mandated by ACPI 6) as the supported ACPI
153 specification revision (when using this switch, it may
154 be necessary to carry out a cold reboot _twice_ in a
155 row to make it take effect on the platform firmware).
157 acpi_osi= [HW,ACPI] Modify list of supported OS interface strings
158 acpi_osi="string1" # add string1
159 acpi_osi="!string2" # remove string2
160 acpi_osi=!* # remove all strings
161 acpi_osi=! # disable all built-in OS vendor
163 acpi_osi=!! # enable all built-in OS vendor
165 acpi_osi= # disable all strings
167 'acpi_osi=!' can be used in combination with single or
168 multiple 'acpi_osi="string1"' to support specific OS
169 vendor string(s). Note that such command can only
170 affect the default state of the OS vendor strings, thus
171 it cannot affect the default state of the feature group
172 strings and the current state of the OS vendor strings,
173 specifying it multiple times through kernel command line
174 is meaningless. This command is useful when one do not
175 care about the state of the feature group strings which
176 should be controlled by the OSPM.
178 1. 'acpi_osi=! acpi_osi="Windows 2000"' is equivalent
179 to 'acpi_osi="Windows 2000" acpi_osi=!', they all
180 can make '_OSI("Windows 2000")' TRUE.
182 'acpi_osi=' cannot be used in combination with other
183 'acpi_osi=' command lines, the _OSI method will not
184 exist in the ACPI namespace. NOTE that such command can
185 only affect the _OSI support state, thus specifying it
186 multiple times through kernel command line is also
189 1. 'acpi_osi=' can make 'CondRefOf(_OSI, Local1)'
192 'acpi_osi=!*' can be used in combination with single or
193 multiple 'acpi_osi="string1"' to support specific
194 string(s). Note that such command can affect the
195 current state of both the OS vendor strings and the
196 feature group strings, thus specifying it multiple times
197 through kernel command line is meaningful. But it may
198 still not able to affect the final state of a string if
199 there are quirks related to this string. This command
200 is useful when one want to control the state of the
201 feature group strings to debug BIOS issues related to
204 1. 'acpi_osi="Module Device" acpi_osi=!*' can make
205 '_OSI("Module Device")' FALSE.
206 2. 'acpi_osi=!* acpi_osi="Module Device"' can make
207 '_OSI("Module Device")' TRUE.
208 3. 'acpi_osi=! acpi_osi=!* acpi_osi="Windows 2000"' is
210 'acpi_osi=!* acpi_osi=! acpi_osi="Windows 2000"'
212 'acpi_osi=!* acpi_osi="Windows 2000" acpi_osi=!',
213 they all will make '_OSI("Windows 2000")' TRUE.
216 Override the pmtimer bug detection: force the kernel
217 to assume that this machine's pmtimer latches its value
218 and always returns good values.
220 acpi_sci= [HW,ACPI] ACPI System Control Interrupt trigger mode
221 Format: { level | edge | high | low }
223 acpi_skip_timer_override [HW,ACPI]
224 Recognize and ignore IRQ0/pin2 Interrupt Override.
225 For broken nForce2 BIOS resulting in XT-PIC timer.
227 acpi_sleep= [HW,ACPI] Sleep options
228 Format: { s3_bios, s3_mode, s3_beep, s4_nohwsig,
229 old_ordering, nonvs, sci_force_enable, nobl }
230 See Documentation/power/video.rst for information on
232 s3_beep is for debugging; it makes the PC's speaker beep
233 as soon as the kernel's real-mode entry point is called.
234 s4_nohwsig prevents ACPI hardware signature from being
235 used during resume from hibernation.
236 old_ordering causes the ACPI 1.0 ordering of the _PTS
237 control method, with respect to putting devices into
238 low power states, to be enforced (the ACPI 2.0 ordering
239 of _PTS is used by default).
240 nonvs prevents the kernel from saving/restoring the
241 ACPI NVS memory during suspend/hibernation and resume.
242 sci_force_enable causes the kernel to set SCI_EN directly
243 on resume from S1/S3 (which is against the ACPI spec,
244 but some broken systems don't work without it).
245 nobl causes the internal blacklist of systems known to
246 behave incorrectly in some ways with respect to system
247 suspend and resume to be ignored (use wisely).
249 acpi_use_timer_override [HW,ACPI]
250 Use timer override. For some broken Nvidia NF5 boards
251 that require a timer override, but don't have HPET
253 add_efi_memmap [EFI; X86] Include EFI memory map in
254 kernel's map of available physical RAM.
257 { off | try_unsupported }
258 off: disable AGP support
259 try_unsupported: try to drive unsupported chipsets
260 (may crash computer or cause data corruption)
263 See Documentation/sound/alsa-configuration.rst
266 Allow the default userspace alignment fault handler
267 behaviour to be specified. Bit 0 enables warnings,
268 bit 1 enables fixups, and bit 2 sends a segfault.
270 align_va_addr= [X86-64]
271 Align virtual addresses by clearing slice [14:12] when
272 allocating a VMA at process creation time. This option
273 gives you up to 3% performance improvement on AMD F15h
274 machines (where it is enabled by default) for a
275 CPU-intensive style benchmark, and it can vary highly in
276 a microbenchmark depending on workload and compiler.
278 32: only for 32-bit processes
279 64: only for 64-bit processes
280 on: enable for both 32- and 64-bit processes
281 off: disable for both 32- and 64-bit processes
283 alloc_snapshot [FTRACE]
284 Allocate the ftrace snapshot buffer on boot up when the
285 main buffer is allocated. This is handy if debugging
286 and you need to use tracing_snapshot() on boot up, and
287 do not want to use tracing_snapshot_alloc() as it needs
288 to be done where GFP_KERNEL allocations are allowed.
290 amd_iommu= [HW,X86-64]
291 Pass parameters to the AMD IOMMU driver in the system.
293 fullflush - enable flushing of IO/TLB entries when
294 they are unmapped. Otherwise they are
295 flushed before they will be reused, which
297 off - do not initialize any AMD IOMMU found in
299 force_isolation - Force device isolation for all
300 devices. The IOMMU driver is not
301 allowed anymore to lift isolation
302 requirements as needed. This option
303 does not override iommu=pt
305 amd_iommu_dump= [HW,X86-64]
306 Enable AMD IOMMU driver option to dump the ACPI table
307 for AMD IOMMU. With this option enabled, AMD IOMMU
308 driver will print ACPI tables for AMD IOMMU during
309 IOMMU initialization.
311 amd_iommu_intr= [HW,X86-64]
312 Specifies one of the following AMD IOMMU interrupt
314 legacy - Use legacy interrupt remapping mode.
315 vapic - Use virtual APIC mode, which allows IOMMU
316 to inject interrupts directly into guest.
317 This mode requires kvm-amd.avic=1.
318 (Default when IOMMU HW support is present.)
320 amijoy.map= [HW,JOY] Amiga joystick support
321 Map of devices attached to JOY0DAT and JOY1DAT
323 See also Documentation/input/joydev/joystick.rst
325 analog.map= [HW,JOY] Analog joystick and gamepad support
326 Specifies type or capabilities of an analog joystick
327 connected to one of 16 gameports
328 Format: <type1>,<type2>,..<type16>
331 Power management functions (SPARCstation-4/5 + deriv.)
333 Disable APC CPU standby support. SPARCstation-Fox does
334 not play well with APC CPU idle - disable it if you have
335 APC and your system crashes randomly.
337 apic= [APIC,X86] Advanced Programmable Interrupt Controller
338 Change the output verbosity while booting
339 Format: { quiet (default) | verbose | debug }
340 Change the amount of debugging information output
341 when initialising the APIC and IO-APIC components.
342 For X86-32, this can also be used to specify an APIC
344 Format: apic=driver_name
345 Examples: apic=bigsmp
347 apic_extnmi= [APIC,X86] External NMI delivery setting
348 Format: { bsp (default) | all | none }
349 bsp: External NMI is delivered only to CPU 0
350 all: External NMIs are broadcast to all CPUs as a
352 none: External NMI is masked for all CPUs. This is
353 useful so that a dump capture kernel won't be
357 See Documentation/networking/ipv6.rst.
359 show_lapic= [APIC,X86] Advanced Programmable Interrupt Controller
360 Limit apic dumping. The parameter defines the maximal
361 number of local apics being dumped. Also it is possible
362 to set it to "all" by meaning -- no limit here.
363 Format: { 1 (default) | 2 | ... | all }.
364 The parameter valid if only apic=debug or
365 apic=verbose is specified.
366 Example: apic=debug show_lapic=all
368 apm= [APM] Advanced Power Management
369 See header of arch/x86/kernel/apm_32.c.
371 arcrimi= [HW,NET] ARCnet - "RIM I" (entirely mem-mapped) cards
372 Format: <io>,<irq>,<nodeID>
374 arm64.nobti [ARM64] Unconditionally disable Branch Target
375 Identification support
377 arm64.nopauth [ARM64] Unconditionally disable Pointer Authentication
382 atarimouse= [HW,MOUSE] Atari Mouse
384 atkbd.extra= [HW] Enable extra LEDs and keys on IBM RapidAccess,
385 EzKey and similar keyboards
387 atkbd.reset= [HW] Reset keyboard during initialization
389 atkbd.set= [HW] Select keyboard code set
390 Format: <int> (2 = AT (default), 3 = PS/2)
392 atkbd.scroll= [HW] Enable scroll wheel on MS Office and similar
395 atkbd.softraw= [HW] Choose between synthetic and real raw mode
396 Format: <bool> (0 = real, 1 = synthetic (default))
398 atkbd.softrepeat= [HW]
399 Use software keyboard repeat
401 audit= [KNL] Enable the audit sub-system
402 Format: { "0" | "1" | "off" | "on" }
403 0 | off - kernel audit is disabled and can not be
404 enabled until the next reboot
405 unset - kernel audit is initialized but disabled and
406 will be fully enabled by the userspace auditd.
407 1 | on - kernel audit is initialized and partially
408 enabled, storing at most audit_backlog_limit
409 messages in RAM until it is fully enabled by the
413 audit_backlog_limit= [KNL] Set the audit queue size limit.
414 Format: <int> (must be >=0)
417 bau= [X86_UV] Enable the BAU on SGI UV. The default
418 behavior is to disable the BAU (i.e. bau=0).
419 Format: { "0" | "1" }
422 unset - Disable the BAU.
424 baycom_epp= [HW,AX25]
427 baycom_par= [HW,AX25] BayCom Parallel Port AX.25 Modem
429 See header of drivers/net/hamradio/baycom_par.c.
431 baycom_ser_fdx= [HW,AX25]
432 BayCom Serial Port AX.25 Modem (Full Duplex Mode)
433 Format: <io>,<irq>,<mode>[,<baud>]
434 See header of drivers/net/hamradio/baycom_ser_fdx.c.
436 baycom_ser_hdx= [HW,AX25]
437 BayCom Serial Port AX.25 Modem (Half Duplex Mode)
438 Format: <io>,<irq>,<mode>
439 See header of drivers/net/hamradio/baycom_ser_hdx.c.
441 blkdevparts= Manual partition parsing of block device(s) for
442 embedded devices based on command line input.
443 See Documentation/block/cmdline-partition.rst
445 boot_delay= Milliseconds to delay each printk during boot.
446 Values larger than 10 seconds (10000) are changed to
451 Extended command line options can be added to an initrd
452 and this will cause the kernel to look for it.
454 See Documentation/admin-guide/bootconfig.rst
457 Disable BERT OS support on buggy BIOSes.
459 bgrt_disable [ACPI][X86]
460 Disable BGRT to avoid flickering OEM logo.
462 bttv.card= [HW,V4L] bttv (bt848 + bt878 based grabber cards)
463 bttv.radio= Most important insmod options are available as
465 bttv.pll= See Documentation/admin-guide/media/bttv.rst
468 bulk_remove=off [PPC] This parameter disables the use of the pSeries
469 firmware feature for flushing multiple hpte entries
472 c101= [NET] Moxa C101 synchronous serial card
474 cachesize= [BUGS=X86-32] Override level 2 CPU cache size detection.
475 Sometimes CPU hardware bugs make them report the cache
476 size incorrectly. The kernel will attempt work arounds
477 to fix known problems, but for some CPUs it is not
478 possible to determine what the correct size should be.
479 This option provides an override for these situations.
482 [NET] Specifies amount of time (in seconds) that
483 the kernel should wait for a network carrier. By default
484 it waits 120 seconds.
486 ca_keys= [KEYS] This parameter identifies a specific key(s) on
487 the system trusted keyring to be used for certificate
489 format: { id:<keyid> | builtin }
491 cca= [MIPS] Override the kernel pages' cache coherency
492 algorithm. Accepted values range from 0 to 7
493 inclusive. See arch/mips/include/asm/pgtable-bits.h
494 for platform specific values (SB1, Loongson3 and
497 ccw_timeout_log [S390]
498 See Documentation/s390/common_io.rst for details.
500 cgroup_disable= [KNL] Disable a particular controller
501 Format: {name of the controller(s) to disable}
502 The effects of cgroup_disable=foo are:
503 - foo isn't auto-mounted if you mount all cgroups in
505 - foo isn't visible as an individually mountable
507 {Currently only "memory" controller deal with this and
508 cut the overhead, others just disable the usage. So
509 only cgroup_disable=memory is actually worthy}
511 cgroup_no_v1= [KNL] Disable cgroup controllers and named hierarchies in v1
512 Format: { { controller | "all" | "named" }
513 [,{ controller | "all" | "named" }...] }
514 Like cgroup_disable, but only applies to cgroup v1;
515 the blacklisted controllers remain available in cgroup2.
516 "all" blacklists all controllers and "named" disables
517 named mounts. Specifying both "all" and "named" disables
520 cgroup.memory= [KNL] Pass options to the cgroup memory controller.
522 nosocket -- Disable socket memory accounting.
523 nokmem -- Disable kernel memory accounting.
525 checkreqprot [SELINUX] Set initial checkreqprot flag value.
526 Format: { "0" | "1" }
527 See security/selinux/Kconfig help text.
528 0 -- check protection applied by kernel (includes
529 any implied execute protection).
530 1 -- check protection requested by application.
531 Default value is set via a kernel config option.
532 Value can be changed at runtime via
533 /sys/fs/selinux/checkreqprot.
534 Setting checkreqprot to 1 is deprecated.
537 See Documentation/s390/common_io.rst for details.
540 Prevents the clock framework from automatically gating
541 clocks that have not been explicitly enabled by a Linux
542 device driver but are enabled in hardware at reset or
543 by the bootloader/firmware. Note that this does not
544 force such clocks to be always-on nor does it reserve
545 those clocks in any way. This parameter is useful for
546 debug and development, but should not be needed on a
547 platform with proper driver support. For more
548 information, see Documentation/driver-api/clk.rst.
550 clock= [BUGS=X86-32, HW] gettimeofday clocksource override.
552 Forces specified clocksource (if available) to be used
553 when calculating gettimeofday(). If specified
554 clocksource is not available, it defaults to PIT.
555 Format: { pit | tsc | cyclone | pmtmr }
557 clocksource= Override the default clocksource
559 Override the default clocksource and use the clocksource
560 with the name specified.
561 Some clocksource names to choose from, depending on
563 [all] jiffies (this is the base, fallback clocksource)
565 [ARM] imx_timer1,OSTS,netx_timer,mpu_timer2,
566 pxa_timer,timer3,32k_counter,timer0_1
567 [X86-32] pit,hpet,tsc;
568 scx200_hrt on Geode; cyclone on IBM x440
576 clocksource.arm_arch_timer.evtstrm=
579 Enable/disable the eventstream feature of the ARM
580 architected timer so that code using WFE-based polling
581 loops can be debugged more effectively on production
584 clearcpuid=BITNUM[,BITNUM...] [X86]
585 Disable CPUID feature X for the kernel. See
586 arch/x86/include/asm/cpufeatures.h for the valid bit
587 numbers. Note the Linux specific bits are not necessarily
588 stable over kernel options, but the vendor specific
590 Also note that user programs calling CPUID directly
591 or using the feature without checking anything
592 will still see it. This just prevents it from
593 being used by the kernel or shown in /proc/cpuinfo.
594 Also note the kernel might malfunction if you disable
597 cma=nn[MG]@[start[MG][-end[MG]]]
599 Sets the size of kernel global memory area for
600 contiguous memory allocations and optionally the
601 placement constraint by the physical address range of
602 memory allocations. A value of 0 disables CMA
603 altogether. For more information, see
604 kernel/dma/contiguous.c
608 Sets the size of kernel per-numa memory area for
609 contiguous memory allocations. A value of 0 disables
610 per-numa CMA altogether. And If this option is not
611 specificed, the default value is 0.
612 With per-numa CMA enabled, DMA users on node nid will
613 first try to allocate buffer from the pernuma area
614 which is located in node nid, if the allocation fails,
615 they will fallback to the global default memory area.
617 cmo_free_hint= [PPC] Format: { yes | no }
618 Specify whether pages are marked as being inactive
619 when they are freed. This is used in CMO environments
620 to determine OS memory pressure for page stealing by
624 coherent_pool=nn[KMG] [ARM,KNL]
625 Sets the size of memory pool for coherent, atomic dma
626 allocations, by default set to 256K.
628 com20020= [HW,NET] ARCnet - COM20020 chipset
630 <io>[,<irq>[,<nodeID>[,<backplane>[,<ckp>[,<timeout>]]]]]
632 com90io= [HW,NET] ARCnet - COM90xx chipset (IO-mapped buffers)
636 ARCnet - COM90xx chipset (memory-mapped buffers)
637 Format: <io>[,<irq>[,<memstart>]]
639 condev= [HW,S390] console device
642 console= [KNL] Output console device and options.
644 tty<n> Use the virtual console device <n>.
648 Use the specified serial port. The options are of
649 the form "bbbbpnf", where "bbbb" is the baud rate,
650 "p" is parity ("n", "o", or "e"), "n" is number of
651 bits, and "f" is flow control ("r" for RTS or
652 omit it). Default is "9600n8".
654 See Documentation/admin-guide/serial-console.rst for more
656 Documentation/networking/netconsole.rst for an
659 uart[8250],io,<addr>[,options]
660 uart[8250],mmio,<addr>[,options]
661 uart[8250],mmio16,<addr>[,options]
662 uart[8250],mmio32,<addr>[,options]
663 uart[8250],0x<addr>[,options]
664 Start an early, polled-mode console on the 8250/16550
665 UART at the specified I/O port or MMIO address,
666 switching to the matching ttyS device later.
667 MMIO inter-register address stride is either 8-bit
668 (mmio), 16-bit (mmio16), or 32-bit (mmio32).
669 If none of [io|mmio|mmio16|mmio32], <addr> is assumed
670 to be equivalent to 'mmio'. 'options' are specified in
671 the same format described for ttyS above; if unspecified,
672 the h/w is not re-initialized.
674 hvc<n> Use the hypervisor console device <n>. This is for
675 both Xen and PowerPC hypervisors.
677 If the device connected to the port is not a TTY but a braille
678 device, prepend "brl," before the device type, for instance
680 For now, only VisioBraille is supported.
683 [KNL] Change console messages format
685 By default we print messages on consoles in
686 "[time stamp] text\n" format (time stamp may not be
687 printed, depending on CONFIG_PRINTK_TIME or
688 `printk_time' param).
690 Switch to syslog format: "<%u>[time stamp] text\n"
691 IOW, each message will have a facility and loglevel
692 prefix. The format is similar to one used by syslog()
693 syscall, or to executing "dmesg -S --raw" or to reading
696 consoleblank= [KNL] The console blank (screen saver) timeout in
697 seconds. A value of 0 disables the blank timer.
701 [KNL] Change the default value for
702 /proc/<pid>/coredump_filter.
703 See also Documentation/filesystems/proc.rst.
705 coresight_cpu_debug.enable
708 Enable/disable the CPU sampling based debugging.
709 0: default value, disable debugging
710 1: enable debugging at boot time
712 cpuidle.off=1 [CPU_IDLE]
713 disable the cpuidle sub-system
716 [CPU_IDLE] Name of the cpuidle governor to use.
718 cpufreq.off=1 [CPU_FREQ]
719 disable the cpufreq sub-system
721 cpufreq.default_governor=
722 [CPU_FREQ] Name of the default cpufreq governor or
723 policy to use. This governor must be registered in the
724 kernel before the cpufreq driver probes.
727 [X86] Delay for N microsec between assert and de-assert
728 of APIC INIT to start processors. This delay occurs
729 on every CPU online, such as boot, and resume from suspend.
732 cpcihp_generic= [HW,PCI] Generic port I/O CompactPCI driver
734 <first_slot>,<last_slot>,<port>,<enum_bit>[,<debug>]
736 crashkernel=size[KMG][@offset[KMG]]
737 [KNL] Using kexec, Linux can switch to a 'crash kernel'
738 upon panic. This parameter reserves the physical
739 memory region [offset, offset + size] for that kernel
740 image. If '@offset' is omitted, then a suitable offset
741 is selected automatically.
742 [KNL, X86-64] Select a region under 4G first, and
743 fall back to reserve region above 4G when '@offset'
744 hasn't been specified.
745 See Documentation/admin-guide/kdump/kdump.rst for further details.
747 crashkernel=range1:size1[,range2:size2,...][@offset]
748 [KNL] Same as above, but depends on the memory
749 in the running system. The syntax of range is
750 start-[end] where start and end are both
751 a memory unit (amount[KMG]). See also
752 Documentation/admin-guide/kdump/kdump.rst for an example.
754 crashkernel=size[KMG],high
755 [KNL, X86-64] range could be above 4G. Allow kernel
756 to allocate physical memory region from top, so could
757 be above 4G if system have more than 4G ram installed.
758 Otherwise memory region will be allocated below 4G, if
760 It will be ignored if crashkernel=X is specified.
761 crashkernel=size[KMG],low
762 [KNL, X86-64] range under 4G. When crashkernel=X,high
763 is passed, kernel could allocate physical memory region
764 above 4G, that cause second kernel crash on system
765 that require some amount of low memory, e.g. swiotlb
766 requires at least 64M+32K low memory, also enough extra
767 low memory is needed to make sure DMA buffers for 32-bit
768 devices won't run out. Kernel would try to allocate at
769 at least 256M below 4G automatically.
770 This one let user to specify own low range under 4G
771 for second kernel instead.
772 0: to disable low allocation.
773 It will be ignored when crashkernel=X,high is not used
774 or memory reserved is below 4G.
777 [KNL] Disable crypto self-tests
782 cs89x0_media= [HW,NET]
783 Format: { rj45 | aui | bnc }
786 See header of drivers/s390/block/dasd_devmap.c.
788 db9.dev[2|3]= [HW,JOY] Multisystem joystick support via parallel port
789 (one device per port)
790 Format: <port#>,<type>
791 See also Documentation/input/devices/joystick-parport.rst
793 ddebug_query= [KNL,DYNAMIC_DEBUG] Enable debug messages at early boot
795 Documentation/admin-guide/dynamic-debug-howto.rst for
796 details. Deprecated, see dyndbg.
798 debug [KNL] Enable kernel debugging (events log level).
801 [KNL] Enable printing [hashed] pointers early in the
802 boot sequence. If enabled, we use a weak hash instead
803 of siphash to hash pointers. Use this option if you are
804 seeing instances of '(___ptrval___)') and need to see a
805 value (hashed pointer) instead. Cryptographically
806 insecure, please do not use on production kernels.
809 [KNL] verbose locking self-tests
811 Print debugging info while doing the locking API
813 Bitmask for the various LOCKTYPE_ tests. Defaults to 0
814 (no extra messages), setting it to -1 (all bits set)
815 will print _a_lot_ more information - normally only
816 useful to lockdep developers.
818 debug_objects [KNL] Enable object debugging
821 [KNL] Disable object debugging
823 debug_guardpage_minorder=
824 [KNL] When CONFIG_DEBUG_PAGEALLOC is set, this
825 parameter allows control of the order of pages that will
826 be intentionally kept free (and hence protected) by the
827 buddy allocator. Bigger value increase the probability
828 of catching random memory corruption, but reduce the
829 amount of memory for normal system use. The maximum
830 possible value is MAX_ORDER/2. Setting this parameter
831 to 1 or 2 should be enough to identify most random
832 memory corruption problems caused by bugs in kernel or
833 driver code when a CPU writes to (or reads from) a
834 random memory location. Note that there exists a class
835 of memory corruptions problems caused by buggy H/W or
836 F/W or by drivers badly programing DMA (basically when
837 memory is written at bus level and the CPU MMU is
838 bypassed) which are not detectable by
839 CONFIG_DEBUG_PAGEALLOC, hence this option will not help
840 tracking down these problems.
843 [KNL] When CONFIG_DEBUG_PAGEALLOC is set, this parameter
844 enables the feature at boot time. By default, it is
845 disabled and the system will work mostly the same as a
846 kernel built without CONFIG_DEBUG_PAGEALLOC.
847 Note: to get most of debug_pagealloc error reports, it's
848 useful to also enable the page_owner functionality.
849 on: enable the feature
851 debugfs= [KNL] This parameter enables what is exposed to userspace
852 and debugfs internal clients.
853 Format: { on, no-mount, off }
854 on: All functions are enabled.
856 Filesystem is not registered but kernel clients can
857 access APIs and a crashkernel can be used to read
858 its content. There is nothing to mount.
859 off: Filesystem is not registered and clients
860 get a -EPERM as result when trying to register files
861 or directories within debugfs.
862 This is equivalent of the runtime functionality if
863 debugfs was not enabled in the kernel at all.
864 Default value is set in build-time with a kernel configuration.
866 debugpat [X86] Enable PAT debugging
868 decnet.addr= [HW,NET]
869 Format: <area>[,<node>]
870 See also Documentation/networking/decnet.rst.
873 [HW] The size of the default HugeTLB page. This is
874 the size represented by the legacy /proc/ hugepages
875 APIs. In addition, this is the default hugetlb size
876 used for shmget(), mmap() and mounting hugetlbfs
877 filesystems. If not specified, defaults to the
878 architecture's default huge page size. Huge page
879 sizes are architecture dependent. See also
880 Documentation/admin-guide/mm/hugetlbpage.rst.
883 deferred_probe_timeout=
884 [KNL] Debugging option to set a timeout in seconds for
885 deferred probe to give up waiting on dependencies to
886 probe. Only specific dependencies (subsystems or
887 drivers) that have opted in will be ignored. A timeout of 0
888 will timeout at the end of initcalls. This option will also
889 dump out devices still on the deferred probe list after
893 Format: { on | off | def_only | inf_only | always }
894 on: s390 zlib hardware support for compression on
895 level 1 and decompression (default)
896 off: No s390 zlib hardware support
897 def_only: s390 zlib hardware support for deflate
898 only (compression on level 1)
899 inf_only: s390 zlib hardware support for inflate
901 always: Same as 'on' but ignores the selected compression
902 level always using hardware support (used for debugging)
905 Set number of hash buckets for dentry cache.
907 disable_1tb_segments [PPC]
908 Disables the use of 1TB hash page table segments. This
909 causes the kernel to fall back to 256MB segments which
910 can be useful when debugging issues that require an SLB
914 Limits the number of kernel SLB entries, and flushes
915 them frequently to increase the rate of SLB faults
919 See Documentation/networking/ipv6.rst.
922 [KNL] Under CONFIG_HARDENED_USERCOPY, whether
923 hardening is enabled for this boot. Hardened
924 usercopy checking is used to protect the kernel
925 from reading or writing beyond known memory
926 allocation boundaries as a proactive defense
927 against bounds-checking flaws in the kernel's
928 copy_to_user()/copy_from_user() interface.
929 on Perform hardened usercopy checks (default).
930 off Disable hardened usercopy checks.
933 Disable RADIX MMU mode on POWER9
935 radix_hcall_invalidate=on [PPC/PSERIES]
936 Disable RADIX GTSE feature and use hcall for TLB
940 Disable TLBIE instruction. Currently does not work
941 with KVM, with HASH MMU, or with coherent accelerators.
943 disable_cpu_apicid= [X86,APIC,SMP]
945 The number of initial APIC ID for the
946 corresponding CPU to be disabled at boot,
947 mostly used for the kdump 2nd kernel to
948 disable BSP to wake up multiple CPUs without
949 causing system reset or hang due to sending
952 disable_ddw [PPC/PSERIES]
953 Disable Dynamic DMA Window support. Use this
954 to workaround buggy firmware.
957 See Documentation/networking/ipv6.rst.
959 disable_mtrr_cleanup [X86]
960 The kernel tries to adjust MTRR layout from continuous
961 to discrete, to make X server driver able to add WB
962 entry later. This parameter disables that.
964 disable_mtrr_trim [X86, Intel and AMD only]
965 By default the kernel will trim any uncacheable
966 memory out of your available memory pool based on
967 MTRR settings. This parameter disables that behavior,
968 possibly causing your machine to run very slowly.
970 disable_timer_pin_1 [X86]
971 Disable PIN 1 of APIC timer
972 Can be useful to work around chipset bugs.
974 dis_ucode_ldr [X86] Disable the microcode loader.
976 dma_debug=off If the kernel is compiled with DMA_API_DEBUG support,
977 this option disables the debugging code at boot.
979 dma_debug_entries=<number>
980 This option allows to tune the number of preallocated
981 entries for DMA-API debugging code. One entry is
982 required per DMA-API allocation. Use this if the
983 DMA-API debugging code disables itself because the
984 architectural default is too low.
986 dma_debug_driver=<driver_name>
987 With this option the DMA-API debugging driver
988 filter feature can be enabled at boot time. Just
989 pass the driver to filter for as the parameter.
990 The filter can be disabled or changed to another
991 driver later using sysfs.
993 driver_async_probe= [KNL]
994 List of driver names to be probed asynchronously.
995 Format: <driver_name1>,<driver_name2>...
997 drm.edid_firmware=[<connector>:]<file>[,[<connector>:]<file>]
998 Broken monitors, graphic adapters, KVMs and EDIDless
999 panels may send no or incorrect EDID data sets.
1000 This parameter allows to specify an EDID data sets
1001 in the /lib/firmware directory that are used instead.
1002 Generic built-in EDID data sets are used, if one of
1003 edid/1024x768.bin, edid/1280x1024.bin,
1004 edid/1680x1050.bin, or edid/1920x1080.bin is given
1005 and no file with the same name exists. Details and
1006 instructions how to build your own EDID data are
1007 available in Documentation/admin-guide/edid.rst. An EDID
1008 data set will only be used for a particular connector,
1009 if its name and a colon are prepended to the EDID
1010 name. Each connector may use a unique EDID data
1011 set by separating the files with a comma. An EDID
1012 data set with no connector name will be used for
1013 any connectors not explicitly specified.
1018 Format: {"off" | "known"}
1019 Control how the dt_cpu_ftrs device-tree binding is
1020 used for CPU feature discovery and setup (if it
1022 off: Do not use it, fall back to legacy cpu table.
1023 known: Do not pass through unknown features to guests
1024 or userspace, only those that the kernel is aware of.
1026 dump_apple_properties [X86]
1027 Dump name and content of EFI device properties on
1028 x86 Macs. Useful for driver authors to determine
1029 what data is available or for reverse-engineering.
1031 dyndbg[="val"] [KNL,DYNAMIC_DEBUG]
1032 <module>.dyndbg[="val"]
1033 Enable debug messages at boot time. See
1034 Documentation/admin-guide/dynamic-debug-howto.rst
1037 nopku [X86] Disable Memory Protection Keys CPU feature found
1040 <module>.async_probe [KNL]
1041 Enable asynchronous probe on this module.
1043 early_ioremap_debug [KNL]
1044 Enable debug messages in early_ioremap support. This
1045 is useful for tracking down temporary early mappings
1046 which are not unmapped.
1048 earlycon= [KNL] Output early console device and options.
1050 When used with no options, the early console is
1051 determined by stdout-path property in device tree's
1052 chosen node or the ACPI SPCR table if supported by
1055 cdns,<addr>[,options]
1056 Start an early, polled-mode console on a Cadence
1057 (xuartps) serial port at the specified address. Only
1058 supported option is baud rate. If baud rate is not
1059 specified, the serial port must already be setup and
1062 uart[8250],io,<addr>[,options]
1063 uart[8250],mmio,<addr>[,options]
1064 uart[8250],mmio32,<addr>[,options]
1065 uart[8250],mmio32be,<addr>[,options]
1066 uart[8250],0x<addr>[,options]
1067 Start an early, polled-mode console on the 8250/16550
1068 UART at the specified I/O port or MMIO address.
1069 MMIO inter-register address stride is either 8-bit
1070 (mmio) or 32-bit (mmio32 or mmio32be).
1071 If none of [io|mmio|mmio32|mmio32be], <addr> is assumed
1072 to be equivalent to 'mmio'. 'options' are specified
1073 in the same format described for "console=ttyS<n>"; if
1074 unspecified, the h/w is not initialized.
1078 Start an early, polled-mode console on a pl011 serial
1079 port at the specified address. The pl011 serial port
1080 must already be setup and configured. Options are not
1081 yet supported. If 'mmio32' is specified, then only
1082 the driver will use only 32-bit accessors to read/write
1083 the device registers.
1086 Start an early, polled-mode console on a meson serial
1087 port at the specified address. The serial port must
1088 already be setup and configured. Options are not yet
1092 Start an early, polled-mode console on an msm serial
1093 port at the specified address. The serial port
1094 must already be setup and configured. Options are not
1097 msm_serial_dm,<addr>
1098 Start an early, polled-mode console on an msm serial
1099 dm port at the specified address. The serial port
1100 must already be setup and configured. Options are not
1104 Start an early, polled-mode console on a serial port
1105 of an Actions Semi SoC, such as S500 or S900, at the
1106 specified address. The serial port must already be
1107 setup and configured. Options are not yet supported.
1110 Start an early, polled-mode console on a serial port
1111 of an RDA Micro SoC, such as RDA8810PL, at the
1112 specified address. The serial port must already be
1113 setup and configured. Options are not yet supported.
1116 Use RISC-V SBI (Supervisor Binary Interface) for early
1119 smh Use ARM semihosting calls for early console.
1127 Use early console provided by serial driver available
1128 on Samsung SoCs, requires selecting proper type and
1129 a correct base address of the selected UART port. The
1130 serial port must already be setup and configured.
1131 Options are not yet supported.
1134 Start an early, polled-mode console on a lantiq serial
1135 (lqasc) port at the specified address. The serial port
1136 must already be setup and configured. Options are not
1141 Use early console provided by Freescale LP UART driver
1142 found on Freescale Vybrid and QorIQ LS1021A processors.
1143 A valid base address must be provided, and the serial
1144 port must already be setup and configured.
1148 Start an early, polled-mode, output-only console on the
1149 Freescale i.MX UART at the specified address. The UART
1150 must already be setup and configured.
1153 Start an early, polled-mode console on the
1154 Armada 3700 serial port at the specified
1155 address. The serial port must already be setup
1156 and configured. Options are not yet supported.
1159 Start an early, polled-mode console on a Qualcomm
1160 Generic Interface (GENI) based serial port at the
1161 specified address. The serial port must already be
1162 setup and configured. Options are not yet supported.
1165 Start an early, unaccelerated console on the EFI
1166 memory mapped framebuffer (if available). On cache
1167 coherent non-x86 systems that use system memory for
1168 the framebuffer, pass the 'ram' option so that it is
1169 mapped with the correct attributes.
1172 Use early console provided by Freescale LINFlexD UART
1173 serial driver for NXP S32V234 SoCs. A valid base
1174 address must be provided, and the serial port must
1175 already be setup and configured.
1177 earlyprintk= [X86,SH,ARM,M68k,S390]
1181 earlyprintk=serial[,ttySn[,baudrate]]
1182 earlyprintk=serial[,0x...[,baudrate]]
1183 earlyprintk=ttySn[,baudrate]
1184 earlyprintk=dbgp[debugController#]
1185 earlyprintk=pciserial[,force],bus:device.function[,baudrate]
1186 earlyprintk=xdbc[xhciController#]
1188 earlyprintk is useful when the kernel crashes before
1189 the normal console is initialized. It is not enabled by
1190 default because it has some cosmetic problems.
1192 Append ",keep" to not disable it when the real console
1195 Only one of vga, efi, serial, or usb debug port can
1198 Currently only ttyS0 and ttyS1 may be specified by
1199 name. Other I/O ports may be explicitly specified
1200 on some architectures (x86 and arm at least) by
1201 replacing ttySn with an I/O port address, like this:
1202 earlyprintk=serial,0x1008,115200
1203 You can find the port for a given device in
1204 /proc/tty/driver/serial:
1205 2: uart:ST16650V2 port:00001008 irq:18 ...
1207 Interaction with the standard serial driver is not
1210 The VGA and EFI output is eventually overwritten by
1213 The xen output can only be used by Xen PV guests.
1215 The sclp output can only be used on s390.
1217 The optional "force" to "pciserial" enables use of a
1218 PCI device even when its classcode is not of the
1221 edac_report= [HW,EDAC] Control how to report EDAC event
1222 Format: {"on" | "off" | "force"}
1223 on: enable EDAC to report H/W event. May be overridden
1224 by other higher priority error reporting module.
1225 off: disable H/W event reporting through EDAC.
1226 force: enforce the use of EDAC to report H/W event.
1229 ekgdboc= [X86,KGDB] Allow early kernel console debugging
1232 This is designed to be used in conjunction with
1233 the boot argument: earlyprintk=vga
1235 This parameter works in place of the kgdboc parameter
1236 but can only be used if the backing tty is available
1237 very early in the boot process. For early debugging
1238 via a serial port see kgdboc_earlycon instead.
1241 Format: {"off" | "on" | "skip[mbr]"}
1244 Format: { "debug", "disable_early_pci_dma",
1245 "nochunk", "noruntime", "nosoftreserve",
1246 "novamap", "no_disable_early_pci_dma" }
1247 debug: enable misc debug output.
1248 disable_early_pci_dma: disable the busmaster bit on all
1249 PCI bridges while in the EFI boot stub.
1250 nochunk: disable reading files in "chunks" in the EFI
1251 boot stub, as chunking can cause problems with some
1252 firmware implementations.
1253 noruntime : disable EFI runtime services support
1254 nosoftreserve: The EFI_MEMORY_SP (Specific Purpose)
1255 attribute may cause the kernel to reserve the
1256 memory range for a memory mapping driver to
1257 claim. Specify efi=nosoftreserve to disable this
1258 reservation and treat the memory by its base type
1259 (i.e. EFI_CONVENTIONAL_MEMORY / "System RAM").
1260 novamap: do not call SetVirtualAddressMap().
1261 no_disable_early_pci_dma: Leave the busmaster bit set
1262 on all PCI bridges while in the EFI boot stub
1264 efi_no_storage_paranoia [EFI; X86]
1265 Using this parameter you can use more than 50% of
1266 your efi variable storage. Use this parameter only if
1267 you are really sure that your UEFI does sane gc and
1268 fulfills the spec otherwise your board may brick.
1270 efi_fake_mem= nn[KMG]@ss[KMG]:aa[,nn[KMG]@ss[KMG]:aa,..] [EFI; X86]
1271 Add arbitrary attribute to specific memory range by
1272 updating original EFI memory map.
1273 Region of memory which aa attribute is added to is
1276 If efi_fake_mem=2G@4G:0x10000,2G@0x10a0000000:0x10000
1277 is specified, EFI_MEMORY_MORE_RELIABLE(0x10000)
1278 attribute is added to range 0x100000000-0x180000000 and
1279 0x10a0000000-0x1120000000.
1281 If efi_fake_mem=8G@9G:0x40000 is specified, the
1282 EFI_MEMORY_SP(0x40000) attribute is added to
1283 range 0x240000000-0x43fffffff.
1285 Using this parameter you can do debugging of EFI memmap
1286 related features. For example, you can do debugging of
1287 Address Range Mirroring feature even if your box
1288 doesn't support it, or mark specific memory as
1291 efivar_ssdt= [EFI; X86] Name of an EFI variable that contains an SSDT
1292 that is to be dynamically loaded by Linux. If there are
1293 multiple variables with the same name but with different
1294 vendor GUIDs, all of them will be loaded. See
1295 Documentation/admin-guide/acpi/ssdt-overlays.rst for details.
1298 eisa_irq_edge= [PARISC,HW]
1299 See header of drivers/parisc/eisa.c.
1302 See comment before function elanfreq_setup() in
1303 arch/x86/kernel/cpu/cpufreq/elanfreq.c.
1305 elfcorehdr=[size[KMG]@]offset[KMG] [IA64,PPC,SH,X86,S390]
1306 Specifies physical address of start of kernel core
1307 image elf header and optionally the size. Generally
1308 kexec loader will pass this option to capture kernel.
1309 See Documentation/admin-guide/kdump/kdump.rst for details.
1311 enable_mtrr_cleanup [X86]
1312 The kernel tries to adjust MTRR layout from continuous
1313 to discrete, to make X server driver able to add WB
1314 entry later. This parameter enables that.
1316 enable_timer_pin_1 [X86]
1317 Enable PIN 1 of APIC timer
1318 Can be useful to work around chipset bugs
1319 (in particular on some ATI chipsets).
1320 The kernel tries to set a reasonable default.
1322 enforcing [SELINUX] Set initial enforcing status.
1324 See security/selinux/Kconfig help text.
1325 0 -- permissive (log only, no denials).
1326 1 -- enforcing (deny and log).
1328 Value can be changed at runtime via
1329 /sys/fs/selinux/enforce.
1332 Disable Error Record Serialization Table (ERST)
1335 ether= [HW,NET] Ethernet cards parameters
1336 This option is obsoleted by the "netdev=" option, which
1337 has equivalent usage. See its documentation for details.
1341 Permit 'security.evm' to be updated regardless of
1342 current integrity status.
1347 fail_make_request=[KNL]
1348 General fault injection mechanism.
1349 Format: <interval>,<probability>,<space>,<times>
1350 See also Documentation/fault-injection/.
1353 Format: { initns | none }
1354 See Documentation/admin-guide/sysctl/net.rst for
1355 fb_tunnels_only_for_init_ns
1358 See Documentation/admin-guide/blockdev/floppy.rst.
1360 force_pal_cache_flush
1361 [IA-64] Avoid check_sal_cache_flush which may hang on
1362 buggy SAL_CACHE_FLUSH implementations. Using this
1363 parameter will force ia64_sal_cache_flush to call
1364 ia64_pal_cache_flush instead of SAL_CACHE_FLUSH.
1367 Forcefully enable Physical Address Extension (PAE).
1368 Many Pentium M systems disable PAE but may have a
1369 functionally usable PAE implementation.
1370 Warning: use of this parameter will taint the kernel
1371 and may cause unknown problems.
1374 [FTRACE] will set and start the specified tracer
1375 as early as possible in order to facilitate early
1378 ftrace_dump_on_oops[=orig_cpu]
1379 [FTRACE] will dump the trace buffers on oops.
1380 If no parameter is passed, ftrace will dump
1381 buffers of all CPUs, but if you pass orig_cpu, it will
1382 dump only the buffer of the CPU that triggered the
1385 ftrace_filter=[function-list]
1386 [FTRACE] Limit the functions traced by the function
1387 tracer at boot up. function-list is a comma-separated
1388 list of functions. This list can be changed at run
1389 time by the set_ftrace_filter file in the debugfs
1392 ftrace_notrace=[function-list]
1393 [FTRACE] Do not trace the functions specified in
1394 function-list. This list can be changed at run time
1395 by the set_ftrace_notrace file in the debugfs
1398 ftrace_graph_filter=[function-list]
1399 [FTRACE] Limit the top level callers functions traced
1400 by the function graph tracer at boot up.
1401 function-list is a comma-separated list of functions
1402 that can be changed at run time by the
1403 set_graph_function file in the debugfs tracing directory.
1405 ftrace_graph_notrace=[function-list]
1406 [FTRACE] Do not trace from the functions specified in
1407 function-list. This list is a comma-separated list of
1408 functions that can be changed at run time by the
1409 set_graph_notrace file in the debugfs tracing directory.
1411 ftrace_graph_max_depth=<uint>
1412 [FTRACE] Used with the function graph tracer. This is
1413 the max depth it will trace into a function. This value
1414 can be changed at run time by the max_graph_depth file
1415 in the tracefs tracing directory. default: 0 (no limit)
1417 fw_devlink= [KNL] Create device links between consumer and supplier
1418 devices by scanning the firmware to infer the
1419 consumer/supplier relationships. This feature is
1420 especially useful when drivers are loaded as modules as
1421 it ensures proper ordering of tasks like device probing
1422 (suppliers first, then consumers), supplier boot state
1423 clean up (only after all consumers have probed),
1424 suspend/resume & runtime PM (consumers first, then
1426 Format: { off | permissive | on | rpm }
1427 off -- Don't create device links from firmware info.
1428 permissive -- Create device links from firmware info
1429 but use it only for ordering boot state clean
1430 up (sync_state() calls).
1431 on -- Create device links from firmware info and use it
1432 to enforce probe and suspend/resume ordering.
1433 rpm -- Like "on", but also use to order runtime PM.
1435 fw_devlink.strict=<bool>
1436 [KNL] Treat all inferred dependencies as mandatory
1437 dependencies. This only applies for fw_devlink=on|rpm.
1441 [HW,JOY] Multisystem joystick and NES/SNES/PSX pad
1442 support via parallel port (up to 5 devices per port)
1443 Format: <port#>,<pad1>,<pad2>,<pad3>,<pad4>,<pad5>
1444 See also Documentation/input/devices/joystick-parport.rst
1448 gart_fix_e820= [X86-64] disable the fix e820 for K8 GART
1452 gcov_persist= [GCOV] When non-zero (default), profiling data for
1453 kernel modules is saved and remains accessible via
1454 debugfs, even when the module is unloaded/reloaded.
1455 When zero, profiling data is discarded and associated
1456 debugfs files are removed at module unload time.
1458 goldfish [X86] Enable the goldfish android emulator platform.
1459 Don't use this when you are not running on the
1462 gpt [EFI] Forces disk with valid GPT signature but
1463 invalid Protective MBR to be treated as GPT. If the
1464 primary GPT is corrupted, it enables the backup/alternate
1465 GPT to be used instead.
1467 grcan.enable0= [HW] Configuration of physical interface 0. Determines
1468 the "Enable 0" bit of the configuration register.
1471 grcan.enable1= [HW] Configuration of physical interface 1. Determines
1472 the "Enable 0" bit of the configuration register.
1475 grcan.select= [HW] Select which physical interface to use.
1478 grcan.txsize= [HW] Sets the size of the tx buffer.
1479 Format: <unsigned int> such that (txsize & ~0x1fffc0) == 0.
1481 grcan.rxsize= [HW] Sets the size of the rx buffer.
1482 Format: <unsigned int> such that (rxsize & ~0x1fffc0) == 0.
1485 gpio-mockup.gpio_mockup_ranges
1486 [HW] Sets the ranges of gpiochip of for this device.
1487 Format: <start1>,<end1>,<start2>,<end2>...
1489 hardlockup_all_cpu_backtrace=
1490 [KNL] Should the hard-lockup detector generate
1491 backtraces on all cpus.
1494 hashdist= [KNL,NUMA] Large hashes allocated during boot
1495 are distributed across NUMA nodes. Defaults on
1496 for 64-bit NUMA, off otherwise.
1497 Format: 0 | 1 (for off | on)
1499 hcl= [IA-64] SGI's Hardware Graph compatibility layer
1501 hd= [EIDE] (E)IDE hard drive subsystem geometry
1502 Format: <cyl>,<head>,<sect>
1505 Disable Hardware Error Source Table (HEST) support;
1506 corresponding firmware-first mode error processing
1507 logic will be disabled.
1509 highmem=nn[KMG] [KNL,BOOT] forces the highmem zone to have an exact
1510 size of <nn>. This works even on boxes that have no
1511 highmem otherwise. This also works to reduce highmem
1512 size on bigger boxes.
1514 highres= [KNL] Enable/disable high resolution timer mode.
1515 Valid parameters: "on", "off"
1520 hpet= [X86-32,HPET] option to control HPET usage
1521 Format: { enable (default) | disable | force |
1523 disable: disable HPET and use PIT instead
1524 force: allow force enabled of undocumented chips (ICH4,
1526 verbose: show contents of HPET registers during setup
1528 hpet_mmap= [X86, HPET_MMAP] Allow userspace to mmap HPET
1529 registers. Default set by CONFIG_HPET_MMAP_DEFAULT.
1531 hugetlb_cma= [HW,CMA] The size of a CMA area used for allocation
1532 of gigantic hugepages.
1535 Reserve a CMA area of given size and allocate gigantic
1536 hugepages using the CMA allocator. If enabled, the
1537 boot-time allocation of gigantic hugepages is skipped.
1539 hugepages= [HW] Number of HugeTLB pages to allocate at boot.
1540 If this follows hugepagesz (below), it specifies
1541 the number of pages of hugepagesz to be allocated.
1542 If this is the first HugeTLB parameter on the command
1543 line, it specifies the number of pages to allocate for
1544 the default huge page size. See also
1545 Documentation/admin-guide/mm/hugetlbpage.rst.
1549 [HW] The size of the HugeTLB pages. This is used in
1550 conjunction with hugepages (above) to allocate huge
1551 pages of a specific size at boot. The pair
1552 hugepagesz=X hugepages=Y can be specified once for
1553 each supported huge page size. Huge page sizes are
1554 architecture dependent. See also
1555 Documentation/admin-guide/mm/hugetlbpage.rst.
1559 [KNL] Should the hung task detector generate panics.
1562 A value of 1 instructs the kernel to panic when a
1563 hung task is detected. The default value is controlled
1564 by the CONFIG_BOOTPARAM_HUNG_TASK_PANIC build-time
1565 option. The value selected by this boot parameter can
1566 be changed later by the kernel.hung_task_panic sysctl.
1568 hvc_iucv= [S390] Number of z/VM IUCV hypervisor console (HVC)
1569 terminal devices. Valid values: 0..8
1570 hvc_iucv_allow= [S390] Comma-separated list of z/VM user IDs.
1571 If specified, z/VM IUCV HVC accepts connections
1572 from listed z/VM user IDs only.
1574 hv_nopvspin [X86,HYPER_V] Disables the paravirt spinlock optimizations
1575 which allow the hypervisor to 'idle' the
1576 guest on lock contention.
1579 Do not unregister boot console at start. This is only
1580 useful for debugging when something happens in the window
1581 between unregistering the boot console and initializing
1584 i2c_bus= [HW] Override the default board specific I2C bus speed
1585 or register an additional I2C bus that is not
1586 registered from board initialization code.
1590 i8042.debug [HW] Toggle i8042 debug mode
1591 i8042.unmask_kbd_data
1592 [HW] Enable printing of interrupt data from the KBD port
1593 (disabled by default, and as a pre-condition
1594 requires that i8042.debug=1 be enabled)
1595 i8042.direct [HW] Put keyboard port into non-translated mode
1596 i8042.dumbkbd [HW] Pretend that controller can only read data from
1597 keyboard and cannot control its state
1598 (Don't attempt to blink the leds)
1599 i8042.noaux [HW] Don't check for auxiliary (== mouse) port
1600 i8042.nokbd [HW] Don't check/create keyboard port
1601 i8042.noloop [HW] Disable the AUX Loopback command while probing
1603 i8042.nomux [HW] Don't check presence of an active multiplexing
1605 i8042.nopnp [HW] Don't use ACPIPnP / PnPBIOS to discover KBD/AUX
1607 i8042.notimeout [HW] Ignore timeout condition signalled by controller
1608 i8042.reset [HW] Reset the controller during init, cleanup and
1609 suspend-to-ram transitions, only during s2r
1610 transitions, or never reset
1611 Format: { 1 | Y | y | 0 | N | n }
1612 1, Y, y: always reset controller
1613 0, N, n: don't ever reset controller
1614 Default: only on s2r transitions on x86; most other
1615 architectures force reset to be always executed
1616 i8042.unlock [HW] Unlock (ignore) the keylock
1617 i8042.kbdreset [HW] Reset device connected to KBD port
1621 i8k.ignore_dmi [HW] Continue probing hardware even if DMI data
1622 indicates that the driver is running on unsupported
1624 i8k.force [HW] Activate i8k driver even if SMM BIOS signature
1625 does not match list of supported models.
1627 [HW] Report power status in /proc/i8k
1628 (disabled by default)
1629 i8k.restricted [HW] Allow controlling fans only if SYS_ADMIN
1632 i915.invert_brightness=
1633 [DRM] Invert the sense of the variable that is used to
1634 set the brightness of the panel backlight. Normally a
1635 brightness value of 0 indicates backlight switched off,
1636 and the maximum of the brightness value sets the backlight
1637 to maximum brightness. If this parameter is set to 0
1638 (default) and the machine requires it, or this parameter
1639 is set to 1, a brightness value of 0 sets the backlight
1640 to maximum brightness, and the maximum of the brightness
1641 value switches the backlight off.
1642 -1 -- never invert brightness
1643 0 -- machine default
1644 1 -- force brightness inversion
1647 Format: <io>[,<membase>[,<icn_id>[,<icn_id2>]]]
1649 ide-core.nodma= [HW] (E)IDE subsystem
1650 Format: =0.0 to prevent dma on hda, =0.1 hdb =1.0 hdc
1651 .vlb_clock .pci_clock .noflush .nohpa .noprobe .nowerr
1652 .cdrom .chs .ignore_cable are additional options
1653 See Documentation/ide/ide.rst.
1655 ide-generic.probe-mask= [HW] (E)IDE subsystem
1657 Probe mask for legacy ISA IDE ports. Depending on
1658 platform up to 6 ports are supported, enabled by
1659 setting corresponding bits in the mask to 1. The
1660 default value is 0x0, which has a special meaning.
1661 On systems that have PCI, it triggers scanning the
1662 PCI bus for the first and the second port, which
1663 are then probed. On systems without PCI the value
1664 of 0x0 enables probing the two first ports as if it
1667 ide-pci-generic.all-generic-ide [HW] (E)IDE subsystem
1668 Claim all unknown PCI IDE storage controllers.
1671 Format: idle=poll, idle=halt, idle=nomwait
1672 Poll forces a polling idle loop that can slightly
1673 improve the performance of waking up a idle CPU, but
1674 will use a lot of power and make the system run hot.
1676 idle=halt: Halt is forced to be used for CPU idle.
1677 In such case C2/C3 won't be used again.
1678 idle=nomwait: Disable mwait for CPU C-states
1682 Allow force disabling of Shared Virtual Memory (SVA)
1683 support for the idxd driver. By default it is set to
1686 ieee754= [MIPS] Select IEEE Std 754 conformance mode
1687 Format: { strict | legacy | 2008 | relaxed }
1690 Choose which programs will be accepted for execution
1691 based on the IEEE 754 NaN encoding(s) supported by
1692 the FPU and the NaN encoding requested with the value
1693 of an ELF file header flag individually set by each
1694 binary. Hardware implementations are permitted to
1695 support either or both of the legacy and the 2008 NaN
1698 Available settings are as follows:
1699 strict accept binaries that request a NaN encoding
1700 supported by the FPU
1701 legacy only accept legacy-NaN binaries, if supported
1703 2008 only accept 2008-NaN binaries, if supported
1705 relaxed accept any binaries regardless of whether
1706 supported by the FPU
1708 The FPU emulator is always able to support both NaN
1709 encodings, so if no FPU hardware is present or it has
1710 been disabled with 'nofpu', then the settings of
1711 'legacy' and '2008' strap the emulator accordingly,
1712 'relaxed' straps the emulator for both legacy-NaN and
1713 2008-NaN, whereas 'strict' enables legacy-NaN only on
1714 legacy processors and both NaN encodings on MIPS32 or
1717 The setting for ABS.fmt/NEG.fmt instruction execution
1718 mode generally follows that for the NaN encoding,
1719 except where unsupported by hardware.
1721 ignore_loglevel [KNL]
1722 Ignore loglevel setting - this will print /all/
1723 kernel messages to the console. Useful for debugging.
1724 We also add it as printk module parameter, so users
1725 could change it dynamically, usually by
1726 /sys/module/printk/parameters/ignore_loglevel.
1729 Ignore RLIMIT_DATA setting for data mappings,
1730 print warning at first misuse. Can be changed via
1731 /sys/module/kernel/parameters/ignore_rlimit_data.
1733 ihash_entries= [KNL]
1734 Set number of hash buckets for inode cache.
1736 ima_appraise= [IMA] appraise integrity measurements
1737 Format: { "off" | "enforce" | "fix" | "log" }
1740 ima_appraise_tcb [IMA] Deprecated. Use ima_policy= instead.
1741 The builtin appraise policy appraises all files
1744 ima_canonical_fmt [IMA]
1745 Use the canonical format for the binary runtime
1746 measurements, instead of host native format.
1749 Format: { md5 | sha1 | rmd160 | sha256 | sha384
1753 The list of supported hash algorithms is defined
1754 in crypto/hash_info.h.
1757 The builtin policies to load during IMA setup.
1758 Format: "tcb | appraise_tcb | secure_boot |
1759 fail_securely | critical_data"
1761 The "tcb" policy measures all programs exec'd, files
1762 mmap'd for exec, and all files opened with the read
1763 mode bit set by either the effective uid (euid=0) or
1766 The "appraise_tcb" policy appraises the integrity of
1767 all files owned by root.
1769 The "secure_boot" policy appraises the integrity
1770 of files (eg. kexec kernel image, kernel modules,
1771 firmware, policy, etc) based on file signatures.
1773 The "fail_securely" policy forces file signature
1774 verification failure also on privileged mounted
1775 filesystems with the SB_I_UNVERIFIABLE_SIGNATURE
1778 The "critical_data" policy measures kernel integrity
1781 ima_tcb [IMA] Deprecated. Use ima_policy= instead.
1782 Load a policy which meets the needs of the Trusted
1783 Computing Base. This means IMA will measure all
1784 programs exec'd, files mmap'd for exec, and all files
1785 opened for read by uid=0.
1788 Select one of defined IMA measurements template formats.
1789 Formats: { "ima" | "ima-ng" | "ima-sig" }
1793 [IMA] Define a custom template format.
1794 Format: { "field1|...|fieldN" }
1796 ima.ahash_minsize= [IMA] Minimum file size for asynchronous hash usage
1797 Format: <min_file_size>
1798 Set the minimal file size for using asynchronous hash.
1799 If left unspecified, ahash usage is disabled.
1801 ahash performance varies for different data sizes on
1802 different crypto accelerators. This option can be used
1803 to achieve the best performance for a particular HW.
1805 ima.ahash_bufsize= [IMA] Asynchronous hash buffer size
1807 Set hashing buffer size. Default: 4k.
1809 ahash performance varies for different chunk sizes on
1810 different crypto accelerators. This option can be used
1811 to achieve best performance for particular HW.
1815 Run specified binary instead of /sbin/init as init
1818 initcall_debug [KNL] Trace initcalls as they are executed. Useful
1819 for working out where the kernel is dying during
1822 initcall_blacklist= [KNL] Do not execute a comma-separated list of
1823 initcall functions. Useful for debugging built-in
1824 modules and initcalls.
1826 initrd= [BOOT] Specify the location of the initial ramdisk
1828 initrdmem= [KNL] Specify a physical address and size from which to
1829 load the initrd. If an initrd is compiled in or
1830 specified in the bootparams, it takes priority over this
1832 Format: ss[KMG],nn[KMG]
1835 init_on_alloc= [MM] Fill newly allocated pages and heap objects with
1838 Default set by CONFIG_INIT_ON_ALLOC_DEFAULT_ON.
1840 init_on_free= [MM] Fill freed pages and heap objects with zeroes.
1842 Default set by CONFIG_INIT_ON_FREE_DEFAULT_ON.
1844 init_pkru= [X86] Specify the default memory protection keys rights
1845 register contents for all processes. 0x55555554 by
1846 default (disallow access to all but pkey 0). Can
1847 override in debugfs after boot.
1849 inport.irq= [HW] Inport (ATI XL and Microsoft) busmouse driver
1852 int_pln_enable [X86] Enable power limit notification interrupt
1854 integrity_audit=[IMA]
1855 Format: { "0" | "1" }
1856 0 -- basic integrity auditing messages. (Default)
1857 1 -- additional integrity auditing messages.
1859 intel_iommu= [DMAR] Intel IOMMU driver (DMAR) option
1861 Enable intel iommu driver.
1863 Disable intel iommu driver.
1864 igfx_off [Default Off]
1865 By default, gfx is mapped as normal device. If a gfx
1866 device has a dedicated DMAR unit, the DMAR unit is
1867 bypassed by not enabling DMAR with this option. In
1868 this case, gfx device will use physical address for
1871 With this option iommu will not optimize to look
1872 for io virtual address below 32-bit forcing dual
1873 address cycle on pci bus for cards supporting greater
1874 than 32-bit addressing. The default is to look
1875 for translation below 32-bit and if not available
1876 then look in the higher range.
1877 strict [Default Off]
1878 With this option on every unmap_single operation will
1879 result in a hardware IOTLB flush operation as opposed
1880 to batching them for performance.
1881 sp_off [Default Off]
1882 By default, super page will be supported if Intel IOMMU
1883 has the capability. With this option, super page will
1886 By default, scalable mode will be disabled even if the
1887 hardware advertises that it has support for the scalable
1888 mode translation. With this option set, scalable mode
1889 will be used on hardware which claims to support it.
1890 tboot_noforce [Default Off]
1891 Do not force the Intel IOMMU enabled under tboot.
1892 By default, tboot will force Intel IOMMU on, which
1893 could harm performance of some high-throughput
1894 devices like 40GBit network cards, even if identity
1896 Note that using this option lowers the security
1897 provided by tboot because it makes the system
1898 vulnerable to DMA attacks.
1900 intel_idle.max_cstate= [KNL,HW,ACPI,X86]
1901 0 disables intel_idle and fall back on acpi_idle.
1902 1 to 9 specify maximum depth of C-state.
1906 Do not enable intel_pstate as the default
1907 scaling driver for the supported processors
1909 Use intel_pstate as a scaling driver, but configure it
1910 to work with generic cpufreq governors (instead of
1911 enabling its internal governor). This mode cannot be
1912 used along with the hardware-managed P-states (HWP)
1915 Enable intel_pstate on systems that prohibit it by default
1916 in favor of acpi-cpufreq. Forcing the intel_pstate driver
1917 instead of acpi-cpufreq may disable platform features, such
1918 as thermal controls and power capping, that rely on ACPI
1919 P-States information being indicated to OSPM and therefore
1920 should be used with caution. This option does not work with
1921 processors that aren't supported by the intel_pstate driver
1922 or on platforms that use pcc-cpufreq instead of acpi-cpufreq.
1924 Do not enable hardware P state control (HWP)
1927 Only load intel_pstate on systems which support
1928 hardware P state control (HWP) if available.
1930 Enforce ACPI _PPC performance limits. If the Fixed ACPI
1931 Description Table, specifies preferred power management
1932 profile as "Enterprise Server" or "Performance Server",
1933 then this feature is turned on by default.
1935 Allow per-logical-CPU P-State performance control limits using
1936 cpufreq sysfs interface
1938 intremap= [X86-64, Intel-IOMMU]
1939 on enable Interrupt Remapping (default)
1940 off disable Interrupt Remapping
1941 nosid disable Source ID checking
1943 BIOS x2APIC opt-out request will be ignored
1944 nopost disable Interrupt Posting
1946 iomem= Disable strict checking of access to MMIO memory
1947 strict regions from userspace.
1962 nobypass [PPC/POWERNV]
1963 Disable IOMMU bypass, using IOMMU for PCI devices.
1965 iommu.strict= [ARM64] Configure TLB invalidation behaviour
1966 Format: { "0" | "1" }
1968 Request that DMA unmap operations use deferred
1969 invalidation of hardware TLBs, for increased
1970 throughput at the cost of reduced device isolation.
1971 Will fall back to strict mode if not supported by
1972 the relevant IOMMU driver.
1973 1 - Strict mode (default).
1974 DMA unmap operations invalidate IOMMU hardware TLBs
1978 [ARM64, X86] Configure DMA to bypass the IOMMU by default.
1979 Format: { "0" | "1" }
1980 0 - Use IOMMU translation for DMA.
1981 1 - Bypass the IOMMU for DMA.
1982 unset - Use value of CONFIG_IOMMU_DEFAULT_PASSTHROUGH.
1984 io7= [HW] IO7 for Marvel-based Alpha systems
1985 See comment before marvel_specify_io7 in
1986 arch/alpha/kernel/core_marvel.c.
1988 io_delay= [X86] I/O delay method
1990 Standard port 0x80 based delay
1992 Alternate port 0xed based delay (needed on some systems)
1994 Simple two microseconds delay
1999 See Documentation/admin-guide/nfs/nfsroot.rst.
2001 ipcmni_extend [KNL] Extend the maximum number of unique System V
2002 IPC identifiers from 32,768 to 16,777,216.
2004 irqaffinity= [SMP] Set the default irq affinity mask
2005 The argument is a cpu list, as described above.
2007 irqchip.gicv2_force_probe=
2010 Force the kernel to look for the second 4kB page
2011 of a GICv2 controller even if the memory range
2012 exposed by the device tree is too small.
2014 irqchip.gicv3_nolpi=
2016 Force the kernel to ignore the availability of
2017 LPIs (and by consequence ITSs). Intended for system
2018 that use the kernel as a bootloader, and thus want
2019 to let secondary kernels in charge of setting up
2022 irqchip.gicv3_pseudo_nmi= [ARM64]
2023 Enables support for pseudo-NMIs in the kernel. This
2024 requires the kernel to be built with
2025 CONFIG_ARM64_PSEUDO_NMI.
2028 When an interrupt is not handled search all handlers
2029 for it. Intended to get systems with badly broken
2033 When an interrupt is not handled search all handlers
2034 for it. Also check all handlers each timer
2035 interrupt. Intended to get systems with badly broken
2039 Format: <RDP>,<reset>,<pci_scan>,<verbosity>
2041 isolcpus= [KNL,SMP,ISOL] Isolate a given set of CPUs from disturbance.
2042 [Deprecated - use cpusets instead]
2043 Format: [flag-list,]<cpu-list>
2045 Specify one or more CPUs to isolate from disturbances
2046 specified in the flag list (default: domain):
2049 Disable the tick when a single task runs.
2051 A residual 1Hz tick is offloaded to workqueues, which you
2052 need to affine to housekeeping through the global
2053 workqueue's affinity configured via the
2054 /sys/devices/virtual/workqueue/cpumask sysfs file, or
2055 by using the 'domain' flag described below.
2057 NOTE: by default the global workqueue runs on all CPUs,
2058 so to protect individual CPUs the 'cpumask' file has to
2059 be configured manually after bootup.
2062 Isolate from the general SMP balancing and scheduling
2063 algorithms. Note that performing domain isolation this way
2064 is irreversible: it's not possible to bring back a CPU to
2065 the domains once isolated through isolcpus. It's strongly
2066 advised to use cpusets instead to disable scheduler load
2067 balancing through the "cpuset.sched_load_balance" file.
2068 It offers a much more flexible interface where CPUs can
2069 move in and out of an isolated set anytime.
2071 You can move a process onto or off an "isolated" CPU via
2072 the CPU affinity syscalls or cpuset.
2073 <cpu number> begins at 0 and the maximum value is
2074 "number of CPUs in system - 1".
2078 Isolate from being targeted by managed interrupts
2079 which have an interrupt mask containing isolated
2080 CPUs. The affinity of managed interrupts is
2081 handled by the kernel and cannot be changed via
2082 the /proc/irq/* interfaces.
2084 This isolation is best effort and only effective
2085 if the automatically assigned interrupt mask of a
2086 device queue contains isolated and housekeeping
2087 CPUs. If housekeeping CPUs are online then such
2088 interrupts are directed to the housekeeping CPU
2089 so that IO submitted on the housekeeping CPU
2090 cannot disturb the isolated CPU.
2092 If a queue's affinity mask contains only isolated
2093 CPUs then this parameter has no effect on the
2094 interrupt routing decision, though interrupts are
2095 only delivered when tasks running on those
2096 isolated CPUs submit IO. IO submitted on
2097 housekeeping CPUs has no influence on those
2100 The format of <cpu-list> is described above.
2104 ivrs_ioapic [HW,X86-64]
2105 Provide an override to the IOAPIC-ID<->DEVICE-ID
2106 mapping provided in the IVRS ACPI table. For
2107 example, to map IOAPIC-ID decimal 10 to
2108 PCI device 00:14.0 write the parameter as:
2109 ivrs_ioapic[10]=00:14.0
2111 ivrs_hpet [HW,X86-64]
2112 Provide an override to the HPET-ID<->DEVICE-ID
2113 mapping provided in the IVRS ACPI table. For
2114 example, to map HPET-ID decimal 0 to
2115 PCI device 00:14.0 write the parameter as:
2116 ivrs_hpet[0]=00:14.0
2118 ivrs_acpihid [HW,X86-64]
2119 Provide an override to the ACPI-HID:UID<->DEVICE-ID
2120 mapping provided in the IVRS ACPI table. For
2121 example, to map UART-HID:UID AMD0020:0 to
2122 PCI device 00:14.5 write the parameter as:
2123 ivrs_acpihid[00:14.5]=AMD0020:0
2125 js= [HW,JOY] Analog joystick
2126 See Documentation/input/joydev/joystick.rst.
2129 When CONFIG_RANDOMIZE_BASE is set, this disables
2130 kernel and module base offset ASLR (Address Space
2131 Layout Randomization).
2134 [KNL] Enforce KASAN (Kernel Address Sanitizer) to print
2135 report on every invalid memory access. Without this
2136 parameter KASAN will print report only for the first
2141 kernelcore= [KNL,X86,IA-64,PPC]
2142 Format: nn[KMGTPE] | nn% | "mirror"
2143 This parameter specifies the amount of memory usable by
2144 the kernel for non-movable allocations. The requested
2145 amount is spread evenly throughout all nodes in the
2146 system as ZONE_NORMAL. The remaining memory is used for
2147 movable memory in its own zone, ZONE_MOVABLE. In the
2148 event, a node is too small to have both ZONE_NORMAL and
2149 ZONE_MOVABLE, kernelcore memory will take priority and
2150 other nodes will have a larger ZONE_MOVABLE.
2152 ZONE_MOVABLE is used for the allocation of pages that
2153 may be reclaimed or moved by the page migration
2154 subsystem. Note that allocations like PTEs-from-HighMem
2155 still use the HighMem zone if it exists, and the Normal
2156 zone if it does not.
2158 It is possible to specify the exact amount of memory in
2159 the form of "nn[KMGTPE]", a percentage of total system
2160 memory in the form of "nn%", or "mirror". If "mirror"
2161 option is specified, mirrored (reliable) memory is used
2162 for non-movable allocations and remaining memory is used
2163 for Movable pages. "nn[KMGTPE]", "nn%", and "mirror"
2164 are exclusive, so you cannot specify multiple forms.
2166 kgdbdbgp= [KGDB,HW] kgdb over EHCI usb debug port.
2167 Format: <Controller#>[,poll interval]
2168 The controller # is the number of the ehci usb debug
2169 port as it is probed via PCI. The poll interval is
2170 optional and is the number seconds in between
2171 each poll cycle to the debug port in case you need
2172 the functionality for interrupting the kernel with
2173 gdb or control-c on the dbgp connection. When
2174 not using this parameter you use sysrq-g to break into
2175 the kernel debugger.
2177 kgdboc= [KGDB,HW] kgdb over consoles.
2178 Requires a tty driver that supports console polling,
2179 or a supported polling keyboard driver (non-usb).
2180 Serial only format: <serial_device>[,baud]
2181 keyboard only format: kbd
2182 keyboard and serial format: kbd,<serial_device>[,baud]
2183 Optional Kernel mode setting:
2184 kms, kbd format: kms,kbd
2185 kms, kbd and serial format: kms,kbd,<ser_dev>[,baud]
2187 kgdboc_earlycon= [KGDB,HW]
2188 If the boot console provides the ability to read
2189 characters and can work in polling mode, you can use
2190 this parameter to tell kgdb to use it as a backend
2191 until the normal console is registered. Intended to
2192 be used together with the kgdboc parameter which
2193 specifies the normal console to transition to.
2195 The name of the early console should be specified
2196 as the value of this parameter. Note that the name of
2197 the early console might be different than the tty
2198 name passed to kgdboc. It's OK to leave the value
2199 blank and the first boot console that implements
2200 read() will be picked.
2202 kgdbwait [KGDB] Stop kernel execution and enter the
2203 kernel debugger at the earliest opportunity.
2205 kmac= [MIPS] Korina ethernet MAC address.
2206 Configure the RouterBoard 532 series on-chip
2207 Ethernet adapter MAC address.
2209 kmemleak= [KNL] Boot-time kmemleak enable/disable
2210 Valid arguments: on, off
2212 Built with CONFIG_DEBUG_KMEMLEAK_DEFAULT_OFF=y,
2215 kprobe_event=[probe-list]
2216 [FTRACE] Add kprobe events and enable at boot time.
2217 The probe-list is a semicolon delimited list of probe
2218 definitions. Each definition is same as kprobe_events
2219 interface, but the parameters are comma delimited.
2220 For example, to add a kprobe event on vfs_read with
2221 arg1 and arg2, add to the command line;
2223 kprobe_event=p,vfs_read,$arg1,$arg2
2225 See also Documentation/trace/kprobetrace.rst "Kernel
2226 Boot Parameter" section.
2228 kpti= [ARM64] Control page table isolation of user
2229 and kernel address spaces.
2230 Default: enabled on cores which need mitigation.
2234 kvm.ignore_msrs=[KVM] Ignore guest accesses to unhandled MSRs.
2235 Default is 0 (don't ignore, but inject #GP)
2237 kvm.enable_vmware_backdoor=[KVM] Support VMware backdoor PV interface.
2238 Default is false (don't support).
2240 kvm.mmu_audit= [KVM] This is a R/W parameter which allows audit
2245 [KVM] Controls the software workaround for the
2246 X86_BUG_ITLB_MULTIHIT bug.
2247 force : Always deploy workaround.
2248 off : Never deploy workaround.
2249 auto : Deploy workaround based on the presence of
2250 X86_BUG_ITLB_MULTIHIT.
2254 If the software workaround is enabled for the host,
2255 guests do need not to enable it for nested guests.
2257 kvm.nx_huge_pages_recovery_ratio=
2258 [KVM] Controls how many 4KiB pages are periodically zapped
2259 back to huge pages. 0 disables the recovery, otherwise if
2260 the value is N KVM will zap 1/Nth of the 4KiB pages every
2261 minute. The default is 60.
2263 kvm-amd.nested= [KVM,AMD] Allow nested virtualization in KVM/SVM.
2264 Default is 1 (enabled)
2266 kvm-amd.npt= [KVM,AMD] Disable nested paging (virtualized MMU)
2268 Default is 1 (enabled) if in 64-bit or 32-bit PAE mode.
2271 [KVM,ARM] Select one of KVM/arm64's modes of operation.
2273 nvhe: Standard nVHE-based mode, without support for
2276 protected: nVHE-based mode with support for guests whose
2277 state is kept private from the host.
2278 Not valid if the kernel is running in EL2.
2280 Defaults to VHE/nVHE based on hardware support.
2282 kvm-arm.vgic_v3_group0_trap=
2283 [KVM,ARM] Trap guest accesses to GICv3 group-0
2286 kvm-arm.vgic_v3_group1_trap=
2287 [KVM,ARM] Trap guest accesses to GICv3 group-1
2290 kvm-arm.vgic_v3_common_trap=
2291 [KVM,ARM] Trap guest accesses to GICv3 common
2294 kvm-arm.vgic_v4_enable=
2295 [KVM,ARM] Allow use of GICv4 for direct injection of
2298 kvm_cma_resv_ratio=n [PPC]
2299 Reserves given percentage from system memory area for
2300 contiguous memory allocation for KVM hash pagetable
2302 By default it reserves 5% of total system memory.
2306 kvm-intel.ept= [KVM,Intel] Disable extended page tables
2307 (virtualized MMU) support on capable Intel chips.
2308 Default is 1 (enabled)
2310 kvm-intel.emulate_invalid_guest_state=
2311 [KVM,Intel] Enable emulation of invalid guest states
2312 Default is 0 (disabled)
2314 kvm-intel.flexpriority=
2315 [KVM,Intel] Disable FlexPriority feature (TPR shadow).
2316 Default is 1 (enabled)
2319 [KVM,Intel] Enable VMX nesting (nVMX).
2320 Default is 0 (disabled)
2322 kvm-intel.unrestricted_guest=
2323 [KVM,Intel] Disable unrestricted guest feature
2324 (virtualized real and unpaged mode) on capable
2325 Intel chips. Default is 1 (enabled)
2327 kvm-intel.vmentry_l1d_flush=[KVM,Intel] Mitigation for L1 Terminal Fault
2330 Valid arguments: never, cond, always
2332 always: L1D cache flush on every VMENTER.
2333 cond: Flush L1D on VMENTER only when the code between
2334 VMEXIT and VMENTER can leak host memory.
2335 never: Disables the mitigation
2337 Default is cond (do L1 cache flush in specific instances)
2339 kvm-intel.vpid= [KVM,Intel] Disable Virtual Processor Identification
2340 feature (tagged TLBs) on capable Intel chips.
2341 Default is 1 (enabled)
2343 l1tf= [X86] Control mitigation of the L1TF vulnerability on
2346 The kernel PTE inversion protection is unconditionally
2347 enabled and cannot be disabled.
2350 Provides all available mitigations for the
2351 L1TF vulnerability. Disables SMT and
2352 enables all mitigations in the
2353 hypervisors, i.e. unconditional L1D flush.
2355 SMT control and L1D flush control via the
2356 sysfs interface is still possible after
2357 boot. Hypervisors will issue a warning
2358 when the first VM is started in a
2359 potentially insecure configuration,
2360 i.e. SMT enabled or L1D flush disabled.
2363 Same as 'full', but disables SMT and L1D
2364 flush runtime control. Implies the
2365 'nosmt=force' command line option.
2366 (i.e. sysfs control of SMT is disabled.)
2369 Leaves SMT enabled and enables the default
2370 hypervisor mitigation, i.e. conditional
2373 SMT control and L1D flush control via the
2374 sysfs interface is still possible after
2375 boot. Hypervisors will issue a warning
2376 when the first VM is started in a
2377 potentially insecure configuration,
2378 i.e. SMT enabled or L1D flush disabled.
2382 Disables SMT and enables the default
2383 hypervisor mitigation.
2385 SMT control and L1D flush control via the
2386 sysfs interface is still possible after
2387 boot. Hypervisors will issue a warning
2388 when the first VM is started in a
2389 potentially insecure configuration,
2390 i.e. SMT enabled or L1D flush disabled.
2393 Same as 'flush', but hypervisors will not
2394 warn when a VM is started in a potentially
2395 insecure configuration.
2398 Disables hypervisor mitigations and doesn't
2400 It also drops the swap size and available
2401 RAM limit restriction on both hypervisor and
2406 For details see: Documentation/admin-guide/hw-vuln/l1tf.rst
2412 lapic [X86-32,APIC] Enable the local APIC even if BIOS
2415 lapic= [X86,APIC] Do not use TSC deadline
2416 value for LAPIC timer one-shot implementation. Default
2417 back to the programmable timer unit in the LAPIC.
2418 Format: notscdeadline
2420 lapic_timer_c2_ok [X86,APIC] trust the local apic timer
2423 libata.dma= [LIBATA] DMA control
2424 libata.dma=0 Disable all PATA and SATA DMA
2425 libata.dma=1 PATA and SATA Disk DMA only
2426 libata.dma=2 ATAPI (CDROM) DMA only
2427 libata.dma=4 Compact Flash DMA only
2428 Combinations also work, so libata.dma=3 enables DMA
2429 for disks and CDROMs, but not CFs.
2431 libata.ignore_hpa= [LIBATA] Ignore HPA limit
2432 libata.ignore_hpa=0 keep BIOS limits (default)
2433 libata.ignore_hpa=1 ignore limits, using full disk
2435 libata.noacpi [LIBATA] Disables use of ACPI in libata suspend/resume
2439 libata.force= [LIBATA] Force configurations. The format is comma-
2440 separated list of "[ID:]VAL" where ID is
2441 PORT[.DEVICE]. PORT and DEVICE are decimal numbers
2442 matching port, link or device. Basically, it matches
2443 the ATA ID string printed on console by libata. If
2444 the whole ID part is omitted, the last PORT and DEVICE
2445 values are used. If ID hasn't been specified yet, the
2446 configuration applies to all ports, links and devices.
2448 If only DEVICE is omitted, the parameter applies to
2449 the port and all links and devices behind it. DEVICE
2450 number of 0 either selects the first device or the
2451 first fan-out link behind PMP device. It does not
2452 select the host link. DEVICE number of 15 selects the
2453 host link and device attached to it.
2455 The VAL specifies the configuration to force. As long
2456 as there's no ambiguity shortcut notation is allowed.
2457 For example, both 1.5 and 1.5G would work for 1.5Gbps.
2458 The following configurations can be forced.
2460 * Cable type: 40c, 80c, short40c, unk, ign or sata.
2461 Any ID with matching PORT is used.
2463 * SATA link speed limit: 1.5Gbps or 3.0Gbps.
2465 * Transfer mode: pio[0-7], mwdma[0-4] and udma[0-7].
2466 udma[/][16,25,33,44,66,100,133] notation is also
2469 * [no]ncq: Turn on or off NCQ.
2471 * [no]ncqtrim: Turn off queued DSM TRIM.
2473 * nohrst, nosrst, norst: suppress hard, soft
2476 * rstonce: only attempt one reset during
2477 hot-unplug link recovery
2479 * dump_id: dump IDENTIFY data.
2481 * atapi_dmadir: Enable ATAPI DMADIR bridge support
2483 * disable: Disable this device.
2485 If there are multiple matching configurations changing
2486 the same attribute, the last one is used.
2488 memblock=debug [KNL] Enable memblock debug messages.
2490 load_ramdisk= [RAM] [Deprecated]
2492 lockd.nlm_grace_period=P [NFS] Assign grace period.
2495 lockd.nlm_tcpport=N [NFS] Assign TCP port.
2498 lockd.nlm_timeout=T [NFS] Assign timeout value.
2501 lockd.nlm_udpport=M [NFS] Assign UDP port.
2504 lockdown= [SECURITY]
2505 { integrity | confidentiality }
2506 Enable the kernel lockdown feature. If set to
2507 integrity, kernel features that allow userland to
2508 modify the running kernel are disabled. If set to
2509 confidentiality, kernel features that allow userland
2510 to extract confidential information from the kernel
2513 locktorture.nreaders_stress= [KNL]
2514 Set the number of locking read-acquisition kthreads.
2515 Defaults to being automatically set based on the
2516 number of online CPUs.
2518 locktorture.nwriters_stress= [KNL]
2519 Set the number of locking write-acquisition kthreads.
2521 locktorture.onoff_holdoff= [KNL]
2522 Set time (s) after boot for CPU-hotplug testing.
2524 locktorture.onoff_interval= [KNL]
2525 Set time (s) between CPU-hotplug operations, or
2526 zero to disable CPU-hotplug testing.
2528 locktorture.shuffle_interval= [KNL]
2529 Set task-shuffle interval (jiffies). Shuffling
2530 tasks allows some CPUs to go into dyntick-idle
2531 mode during the locktorture test.
2533 locktorture.shutdown_secs= [KNL]
2534 Set time (s) after boot system shutdown. This
2535 is useful for hands-off automated testing.
2537 locktorture.stat_interval= [KNL]
2538 Time (s) between statistics printk()s.
2540 locktorture.stutter= [KNL]
2541 Time (s) to stutter testing, for example,
2542 specifying five seconds causes the test to run for
2543 five seconds, wait for five seconds, and so on.
2544 This tests the locking primitive's ability to
2545 transition abruptly to and from idle.
2547 locktorture.torture_type= [KNL]
2548 Specify the locking implementation to test.
2550 locktorture.verbose= [KNL]
2551 Enable additional printk() statements.
2553 logibm.irq= [HW,MOUSE] Logitech Bus Mouse Driver
2556 loglevel= All Kernel Messages with a loglevel smaller than the
2557 console loglevel will be printed to the console. It can
2558 also be changed with klogd or other programs. The
2559 loglevels are defined as follows:
2561 0 (KERN_EMERG) system is unusable
2562 1 (KERN_ALERT) action must be taken immediately
2563 2 (KERN_CRIT) critical conditions
2564 3 (KERN_ERR) error conditions
2565 4 (KERN_WARNING) warning conditions
2566 5 (KERN_NOTICE) normal but significant condition
2567 6 (KERN_INFO) informational
2568 7 (KERN_DEBUG) debug-level messages
2570 log_buf_len=n[KMG] Sets the size of the printk ring buffer,
2571 in bytes. n must be a power of two and greater
2572 than the minimal size. The minimal size is defined
2573 by LOG_BUF_SHIFT kernel config parameter. There is
2574 also CONFIG_LOG_CPU_MAX_BUF_SHIFT config parameter
2575 that allows to increase the default size depending on
2576 the number of CPUs. See init/Kconfig for more details.
2578 logo.nologo [FB] Disables display of the built-in Linux logo.
2579 This may be used to provide more screen space for
2580 kernel log messages and is useful when debugging
2581 kernel boot problems.
2583 lp=0 [LP] Specify parallel ports to use, e.g,
2584 lp=port[,port...] lp=none,parport0 (lp0 not configured, lp1 uses
2585 lp=reset first parallel port). 'lp=0' disables the
2586 lp=auto printer driver. 'lp=reset' (which can be
2587 specified in addition to the ports) causes
2588 attached printers to be reset. Using
2589 lp=port1,port2,... specifies the parallel ports
2590 to associate lp devices with, starting with
2591 lp0. A port specification may be 'none' to skip
2592 that lp device, or a parport name such as
2593 'parport0'. Specifying 'lp=auto' instead of a
2594 port specification list means that device IDs
2595 from each port should be examined, to see if
2596 an IEEE 1284-compliant printer is attached; if
2597 so, the driver will manage that printer.
2598 See also header of drivers/char/lp.c.
2601 Sets loops_per_jiffy to given constant, thus avoiding
2602 time-consuming boot-time autodetection (up to 250 ms per
2603 CPU). 0 enables autodetection (default). To determine
2604 the correct value for your kernel, boot with normal
2605 autodetection and see what value is printed. Note that
2606 on SMP systems the preset will be applied to all CPUs,
2607 which is likely to cause problems if your CPUs need
2608 significantly divergent settings. An incorrect value
2609 will cause delays in the kernel to be wrong, leading to
2610 unpredictable I/O errors and other breakage. Although
2611 unlikely, in the extreme case this might damage your
2615 Format: <io>,<irq>,<dma>
2617 lsm.debug [SECURITY] Enable LSM initialization debugging output.
2620 [SECURITY] Choose order of LSM initialization. This
2621 overrides CONFIG_LSM, and the "security=" parameter.
2623 machvec= [IA-64] Force the use of a particular machine-vector
2624 (machvec) in a generic kernel.
2625 Example: machvec=hpzx1
2627 machtype= [Loongson] Share the same kernel image file between
2628 different yeeloong laptops.
2629 Example: machtype=lemote-yeeloong-2f-7inch
2631 max_addr=nn[KMG] [KNL,BOOT,ia64] All physical memory greater
2632 than or equal to this physical address is ignored.
2634 maxcpus= [SMP] Maximum number of processors that an SMP kernel
2635 will bring up during bootup. maxcpus=n : n >= 0 limits
2636 the kernel to bring up 'n' processors. Surely after
2637 bootup you can bring up the other plugged cpu by executing
2638 "echo 1 > /sys/devices/system/cpu/cpuX/online". So maxcpus
2639 only takes effect during system bootup.
2640 While n=0 is a special case, it is equivalent to "nosmp",
2641 which also disables the IO APIC.
2643 max_loop= [LOOP] The number of loop block devices that get
2644 (loop.max_loop) unconditionally pre-created at init time. The default
2645 number is configured by BLK_DEV_LOOP_MIN_COUNT. Instead
2646 of statically allocating a predefined number, loop
2647 devices can be requested on-demand with the
2648 /dev/loop-control interface.
2650 mce [X86-32] Machine Check Exception
2652 mce=option [X86-64] See Documentation/x86/x86_64/boot-options.rst
2654 md= [HW] RAID subsystems devices and level
2655 See Documentation/admin-guide/md.rst.
2658 Format: <first>,<last>
2659 Specifies range of consoles to be captured by the MDA.
2662 Control mitigation for the Micro-architectural Data
2663 Sampling (MDS) vulnerability.
2665 Certain CPUs are vulnerable to an exploit against CPU
2666 internal buffers which can forward information to a
2667 disclosure gadget under certain conditions.
2669 In vulnerable processors, the speculatively
2670 forwarded data can be used in a cache side channel
2671 attack, to access data to which the attacker does
2672 not have direct access.
2674 This parameter controls the MDS mitigation. The
2677 full - Enable MDS mitigation on vulnerable CPUs
2678 full,nosmt - Enable MDS mitigation and disable
2679 SMT on vulnerable CPUs
2680 off - Unconditionally disable MDS mitigation
2682 On TAA-affected machines, mds=off can be prevented by
2683 an active TAA mitigation as both vulnerabilities are
2684 mitigated with the same mechanism so in order to disable
2685 this mitigation, you need to specify tsx_async_abort=off
2688 Not specifying this option is equivalent to
2691 For details see: Documentation/admin-guide/hw-vuln/mds.rst
2693 mem=nn[KMG] [KNL,BOOT] Force usage of a specific amount of memory
2694 Amount of memory to be used in cases as follows:
2697 2 when the kernel is not able to see the whole system memory;
2698 3 memory that lies after 'mem=' boundary is excluded from
2699 the hypervisor, then assigned to KVM guests.
2701 [X86] Work as limiting max address. Use together
2702 with memmap= to avoid physical address space collisions.
2703 Without memmap= PCI devices could be placed at addresses
2704 belonging to unused RAM.
2706 Note that this only takes effects during boot time since
2707 in above case 3, memory may need be hot added after boot
2708 if system memory of hypervisor is not sufficient.
2710 mem=nopentium [BUGS=X86-32] Disable usage of 4MB pages for kernel
2714 [KNL,SH] Allow user to override the default size for
2715 per-device physically contiguous DMA buffers.
2717 memhp_default_state=online/offline
2718 [KNL] Set the initial state for the memory hotplug
2719 onlining policy. If not specified, the default value is
2720 set according to the
2721 CONFIG_MEMORY_HOTPLUG_DEFAULT_ONLINE kernel config
2723 See Documentation/admin-guide/mm/memory-hotplug.rst.
2725 memmap=exactmap [KNL,X86] Enable setting of an exact
2726 E820 memory map, as specified by the user.
2727 Such memmap=exactmap lines can be constructed based on
2728 BIOS output or other requirements. See the memmap=nn@ss
2731 memmap=nn[KMG]@ss[KMG]
2732 [KNL, X86, MIPS, XTENSA] Force usage of a specific region of memory.
2733 Region of memory to be used is from ss to ss+nn.
2734 If @ss[KMG] is omitted, it is equivalent to mem=nn[KMG],
2735 which limits max address to nn[KMG].
2736 Multiple different regions can be specified,
2739 memmap=100M@2G,100M#3G,1G!1024G
2741 memmap=nn[KMG]#ss[KMG]
2742 [KNL,ACPI] Mark specific memory as ACPI data.
2743 Region of memory to be marked is from ss to ss+nn.
2745 memmap=nn[KMG]$ss[KMG]
2746 [KNL,ACPI] Mark specific memory as reserved.
2747 Region of memory to be reserved is from ss to ss+nn.
2748 Example: Exclude memory from 0x18690000-0x1869ffff
2749 memmap=64K$0x18690000
2751 memmap=0x10000$0x18690000
2752 Some bootloaders may need an escape character before '$',
2753 like Grub2, otherwise '$' and the following number
2756 memmap=nn[KMG]!ss[KMG]
2757 [KNL,X86] Mark specific memory as protected.
2758 Region of memory to be used, from ss to ss+nn.
2759 The memory region may be marked as e820 type 12 (0xc)
2760 and is NVDIMM or ADR memory.
2762 memmap=<size>%<offset>-<oldtype>+<newtype>
2763 [KNL,ACPI] Convert memory within the specified region
2764 from <oldtype> to <newtype>. If "-<oldtype>" is left
2765 out, the whole region will be marked as <newtype>,
2766 even if previously unavailable. If "+<newtype>" is left
2767 out, matching memory will be removed. Types are
2768 specified as e820 types, e.g., 1 = RAM, 2 = reserved,
2769 3 = ACPI, 12 = PRAM.
2771 memory_corruption_check=0/1 [X86]
2772 Some BIOSes seem to corrupt the first 64k of
2773 memory when doing things like suspend/resume.
2774 Setting this option will scan the memory
2775 looking for corruption. Enabling this will
2776 both detect corruption and prevent the kernel
2777 from using the memory being corrupted.
2778 However, its intended as a diagnostic tool; if
2779 repeatable BIOS-originated corruption always
2780 affects the same memory, you can use memmap=
2781 to prevent the kernel from using that memory.
2783 memory_corruption_check_size=size [X86]
2784 By default it checks for corruption in the low
2785 64k, making this memory unavailable for normal
2786 use. Use this parameter to scan for
2787 corruption in more or less memory.
2789 memory_corruption_check_period=seconds [X86]
2790 By default it checks for corruption every 60
2791 seconds. Use this parameter to check at some
2792 other rate. 0 disables periodic checking.
2794 memtest= [KNL,X86,ARM,PPC] Enable memtest
2796 default : 0 <disable>
2797 Specifies the number of memtest passes to be
2798 performed. Each pass selects another test
2799 pattern from a given set of patterns. Memtest
2800 fills the memory with this pattern, validates
2801 memory contents and reserves bad memory
2802 regions that are detected.
2804 mem_encrypt= [X86-64] AMD Secure Memory Encryption (SME) control
2805 Valid arguments: on, off
2806 Default (depends on kernel configuration option):
2807 on (CONFIG_AMD_MEM_ENCRYPT_ACTIVE_BY_DEFAULT=y)
2808 off (CONFIG_AMD_MEM_ENCRYPT_ACTIVE_BY_DEFAULT=n)
2809 mem_encrypt=on: Activate SME
2810 mem_encrypt=off: Do not activate SME
2812 Refer to Documentation/virt/kvm/amd-memory-encryption.rst
2813 for details on when memory encryption can be activated.
2815 mem_sleep_default= [SUSPEND] Default system suspend mode:
2816 s2idle - Suspend-To-Idle
2817 shallow - Power-On Suspend or equivalent (if supported)
2818 deep - Suspend-To-RAM or equivalent (if supported)
2819 See Documentation/admin-guide/pm/sleep-states.rst.
2821 meye.*= [HW] Set MotionEye Camera parameters
2822 See Documentation/admin-guide/media/meye.rst.
2824 mfgpt_irq= [IA-32] Specify the IRQ to use for the
2825 Multi-Function General Purpose Timers on AMD Geode
2828 mfgptfix [X86-32] Fix MFGPT timers on AMD Geode platforms when
2829 the BIOS has incorrectly applied a workaround. TinyBIOS
2830 version 0.98 is known to be affected, 0.99 fixes the
2831 problem by letting the user disable the workaround.
2835 min_addr=nn[KMG] [KNL,BOOT,ia64] All physical memory below this
2836 physical address is ignored.
2838 mini2440= [ARM,HW,KNL]
2839 Format:[0..2][b][c][t]
2841 MINI2440 configuration specification:
2842 0 - The attached screen is the 3.5" TFT
2843 1 - The attached screen is the 7" TFT
2844 2 - The VGA Shield is attached (1024x768)
2845 Leaving out the screen size parameter will not load
2846 the TFT driver, and the framebuffer will be left
2848 b - Enable backlight. The TFT backlight pin will be
2849 linked to the kernel VESA blanking code and a GPIO
2850 LED. This parameter is not necessary when using the
2852 c - Enable the s3c camera interface.
2853 t - Reserved for enabling touchscreen support. The
2854 touchscreen support is not enabled in the mainstream
2855 kernel as of 2.6.30, a preliminary port can be found
2856 in the "bleeding edge" mini2440 support kernel at
2857 https://repo.or.cz/w/linux-2.6/mini2440.git
2860 [X86,PPC,S390,ARM64] Control optional mitigations for
2861 CPU vulnerabilities. This is a set of curated,
2862 arch-independent options, each of which is an
2863 aggregation of existing arch-specific options.
2866 Disable all optional CPU mitigations. This
2867 improves system performance, but it may also
2868 expose users to several CPU vulnerabilities.
2869 Equivalent to: nopti [X86,PPC]
2871 nospectre_v1 [X86,PPC]
2873 nospectre_v2 [X86,PPC,S390,ARM64]
2874 spectre_v2_user=off [X86]
2875 spec_store_bypass_disable=off [X86,PPC]
2876 ssbd=force-off [ARM64]
2879 tsx_async_abort=off [X86]
2880 kvm.nx_huge_pages=off [X86]
2881 no_entry_flush [PPC]
2882 no_uaccess_flush [PPC]
2885 This does not have any effect on
2886 kvm.nx_huge_pages when
2887 kvm.nx_huge_pages=force.
2890 Mitigate all CPU vulnerabilities, but leave SMT
2891 enabled, even if it's vulnerable. This is for
2892 users who don't want to be surprised by SMT
2893 getting disabled across kernel upgrades, or who
2894 have other ways of avoiding SMT-based attacks.
2895 Equivalent to: (default behavior)
2898 Mitigate all CPU vulnerabilities, disabling SMT
2899 if needed. This is for users who always want to
2900 be fully mitigated, even if it means losing SMT.
2901 Equivalent to: l1tf=flush,nosmt [X86]
2902 mds=full,nosmt [X86]
2903 tsx_async_abort=full,nosmt [X86]
2906 [KNL] When CONFIG_DEBUG_MEMORY_INIT is set, this
2907 parameter allows control of the logging verbosity for
2908 the additional memory initialisation checks. A value
2909 of 0 disables mminit logging and a level of 4 will
2910 log everything. Information is printed at KERN_DEBUG
2911 so loglevel=8 may also need to be specified.
2914 [KNL] When CONFIG_MODULE_SIG is set, this means that
2915 modules without (valid) signatures will fail to load.
2916 Note that if CONFIG_MODULE_SIG_FORCE is set, that
2917 is always true, so this option does nothing.
2919 module_blacklist= [KNL] Do not load a comma-separated list of
2920 modules. Useful for debugging problem modules.
2923 [MOUSE] Maximum time between finger touching and
2924 leaving touchpad surface for touch to be considered
2925 a tap and be reported as a left button click (for
2926 touchpads working in absolute mode only).
2928 mousedev.xres= [MOUSE] Horizontal screen resolution, used for devices
2929 reporting absolute coordinates, such as tablets
2930 mousedev.yres= [MOUSE] Vertical screen resolution, used for devices
2931 reporting absolute coordinates, such as tablets
2933 movablecore= [KNL,X86,IA-64,PPC]
2934 Format: nn[KMGTPE] | nn%
2935 This parameter is the complement to kernelcore=, it
2936 specifies the amount of memory used for migratable
2937 allocations. If both kernelcore and movablecore is
2938 specified, then kernelcore will be at *least* the
2939 specified value but may be more. If movablecore on its
2940 own is specified, the administrator must be careful
2941 that the amount of memory usable for all allocations
2944 movable_node [KNL] Boot-time switch to make hotplugable memory
2945 NUMA nodes to be movable. This means that the memory
2946 of such nodes will be usable only for movable
2947 allocations which rules out almost all kernel
2948 allocations. Use with caution!
2950 MTD_Partition= [MTD]
2951 Format: <name>,<region-number>,<size>,<offset>
2953 MTD_Region= [MTD] Format:
2954 <name>,<region-number>[,<base>,<size>,<buswidth>,<altbuswidth>]
2957 See drivers/mtd/parsers/cmdlinepart.c
2959 multitce=off [PPC] This parameter disables the use of the pSeries
2960 firmware feature for updating multiple TCE entries
2963 onenand.bdry= [HW,MTD] Flex-OneNAND Boundary Configuration
2965 Format: [die0_boundary][,die0_lock][,die1_boundary][,die1_lock]
2967 boundary - index of last SLC block on Flex-OneNAND.
2968 The remaining blocks are configured as MLC blocks.
2969 lock - Configure if Flex-OneNAND boundary should be locked.
2970 Once locked, the boundary cannot be changed.
2971 1 indicates lock status, 0 indicates unlock status.
2974 ARM/S3C2412 JIVE boot control
2976 See arch/arm/mach-s3c/mach-jive.c
2978 mtouchusb.raw_coordinates=
2979 [HW] Make the MicroTouch USB driver use raw coordinates
2980 ('y', default) or cooked coordinates ('n')
2982 mtrr_chunk_size=nn[KMG] [X86]
2983 used for mtrr cleanup. It is largest continuous chunk
2984 that could hold holes aka. UC entries.
2986 mtrr_gran_size=nn[KMG] [X86]
2987 Used for mtrr cleanup. It is granularity of mtrr block.
2989 Large value could prevent small alignment from
2992 mtrr_spare_reg_nr=n [X86]
2994 Range: 0,7 : spare reg number
2996 Used for mtrr cleanup. It is spare mtrr entries number.
2997 Set to 2 or more if your graphical card needs more.
2999 n2= [NET] SDL Inc. RISCom/N2 synchronous serial card
3001 netdev= [NET] Network devices parameters
3002 Format: <irq>,<io>,<mem_start>,<mem_end>,<name>
3003 Note that mem_start is often overloaded to mean
3004 something different and driver-specific.
3005 This usage is only documented in each driver source
3009 [NETFILTER] Enable connection tracking flow accounting
3010 0 to disable accounting
3011 1 to enable accounting
3014 nfsaddrs= [NFS] Deprecated. Use ip= instead.
3015 See Documentation/admin-guide/nfs/nfsroot.rst.
3017 nfsroot= [NFS] nfs root filesystem for disk-less boxes.
3018 See Documentation/admin-guide/nfs/nfsroot.rst.
3020 nfsrootdebug [NFS] enable nfsroot debugging messages.
3021 See Documentation/admin-guide/nfs/nfsroot.rst.
3023 nfs.callback_nr_threads=
3024 [NFSv4] set the total number of threads that the
3025 NFS client will assign to service NFSv4 callback
3028 nfs.callback_tcpport=
3029 [NFS] set the TCP port on which the NFSv4 callback
3030 channel should listen.
3033 [NFS] sets the pathname to the program which is used
3034 to update the NFS client cache entries.
3036 nfs.cache_getent_timeout=
3037 [NFS] sets the timeout after which an attempt to
3038 update a cache entry is deemed to have failed.
3040 nfs.idmap_cache_timeout=
3041 [NFS] set the maximum lifetime for idmapper cache
3045 [NFS] enable 64-bit inode numbers.
3046 If zero, the NFS client will fake up a 32-bit inode
3047 number for the readdir() and stat() syscalls instead
3048 of returning the full 64-bit number.
3049 The default is to return 64-bit inode numbers.
3051 nfs.max_session_cb_slots=
3052 [NFSv4.1] Sets the maximum number of session
3053 slots the client will assign to the callback
3054 channel. This determines the maximum number of
3055 callbacks the client will process in parallel for
3056 a particular server.
3058 nfs.max_session_slots=
3059 [NFSv4.1] Sets the maximum number of session slots
3060 the client will attempt to negotiate with the server.
3061 This limits the number of simultaneous RPC requests
3062 that the client can send to the NFSv4.1 server.
3063 Note that there is little point in setting this
3064 value higher than the max_tcp_slot_table_limit.
3066 nfs.nfs4_disable_idmapping=
3067 [NFSv4] When set to the default of '1', this option
3068 ensures that both the RPC level authentication
3069 scheme and the NFS level operations agree to use
3070 numeric uids/gids if the mount is using the
3071 'sec=sys' security flavour. In effect it is
3072 disabling idmapping, which can make migration from
3073 legacy NFSv2/v3 systems to NFSv4 easier.
3074 Servers that do not support this mode of operation
3075 will be autodetected by the client, and it will fall
3076 back to using the idmapper.
3077 To turn off this behaviour, set the value to '0'.
3079 [NFS4] Specify an additional fixed unique ident-
3080 ification string that NFSv4 clients can insert into
3081 their nfs_client_id4 string. This is typically a
3082 UUID that is generated at system install time.
3084 nfs.send_implementation_id =
3085 [NFSv4.1] Send client implementation identification
3086 information in exchange_id requests.
3087 If zero, no implementation identification information
3089 The default is to send the implementation identification
3092 nfs.recover_lost_locks =
3093 [NFSv4] Attempt to recover locks that were lost due
3094 to a lease timeout on the server. Please note that
3095 doing this risks data corruption, since there are
3096 no guarantees that the file will remain unchanged
3097 after the locks are lost.
3098 If you want to enable the kernel legacy behaviour of
3099 attempting to recover these locks, then set this
3101 The default parameter value of '0' causes the kernel
3102 not to attempt recovery of lost locks.
3104 nfs4.layoutstats_timer =
3105 [NFSv4.2] Change the rate at which the kernel sends
3106 layoutstats to the pNFS metadata server.
3108 Setting this to value to 0 causes the kernel to use
3109 whatever value is the default set by the layout
3110 driver. A non-zero value sets the minimum interval
3111 in seconds between layoutstats transmissions.
3113 nfsd.nfs4_disable_idmapping=
3114 [NFSv4] When set to the default of '1', the NFSv4
3115 server will return only numeric uids and gids to
3116 clients using auth_sys, and will accept numeric uids
3117 and gids from such clients. This is intended to ease
3118 migration from NFSv2/v3.
3120 nmi_backtrace.backtrace_idle [KNL]
3121 Dump stacks even of idle CPUs in response to an
3122 NMI stack-backtrace request.
3124 nmi_debug= [KNL,SH] Specify one or more actions to take
3125 when a NMI is triggered.
3126 Format: [state][,regs][,debounce][,die]
3128 nmi_watchdog= [KNL,BUGS=X86] Debugging features for SMP kernels
3129 Format: [panic,][nopanic,][num]
3131 0 - turn hardlockup detector in nmi_watchdog off
3132 1 - turn hardlockup detector in nmi_watchdog on
3133 When panic is specified, panic when an NMI watchdog
3134 timeout occurs (or 'nopanic' to not panic on an NMI
3135 watchdog, if CONFIG_BOOTPARAM_HARDLOCKUP_PANIC is set)
3136 To disable both hard and soft lockup detectors,
3137 please see 'nowatchdog'.
3138 This is useful when you use a panic=... timeout and
3139 need the box quickly up again.
3141 These settings can be accessed at runtime via
3142 the nmi_watchdog and hardlockup_panic sysctls.
3144 netpoll.carrier_timeout=
3145 [NET] Specifies amount of time (in seconds) that
3146 netpoll should wait for a carrier. By default netpoll
3149 no387 [BUGS=X86-32] Tells the kernel to use the 387 maths
3150 emulation library even if a 387 maths coprocessor
3153 no5lvl [X86-64] Disable 5-level paging mode. Forces
3154 kernel to use 4-level paging instead.
3156 nofsgsbase [X86] Disables FSGSBASE instructions.
3159 [HW] Never suspend the console
3160 Disable suspending of consoles during suspend and
3161 hibernate operations. Once disabled, debugging
3162 messages can reach various consoles while the rest
3163 of the system is being put to sleep (ie, while
3164 debugging driver suspend/resume hooks). This may
3165 not work reliably with all consoles, but is known
3166 to work with serial and VGA consoles.
3167 To facilitate more flexible debugging, we also add
3168 console_suspend, a printk module parameter to control
3169 it. Users could use console_suspend (usually
3170 /sys/module/printk/parameters/console_suspend) to
3171 turn on/off it dynamically.
3173 novmcoredd [KNL,KDUMP]
3174 Disable device dump. Device dump allows drivers to
3175 append dump data to vmcore so you can collect driver
3176 specified debug info. Drivers can append the data
3177 without any limit and this data is stored in memory,
3178 so this may cause significant memory stress. Disabling
3179 device dump can help save memory but the driver debug
3180 data will be no longer available. This parameter
3181 is only available when CONFIG_PROC_VMCORE_DEVICE_DUMP
3184 noaliencache [MM, NUMA, SLAB] Disables the allocation of alien
3185 caches in the slab allocator. Saves per-node memory,
3186 but will impact performance.
3190 noaltinstr [S390] Disables alternative instructions patching
3191 (CPU alternatives feature).
3193 noapic [SMP,APIC] Tells the kernel to not make use of any
3194 IOAPICs that may be present in the system.
3196 noautogroup Disable scheduler automatic task group creation.
3198 nobats [PPC] Do not use BATs for mapping kernel lowmem
3199 on "Classic" PPC cores.
3203 noclflush [BUGS=X86] Don't use the CLFLUSH instruction
3205 nodelayacct [KNL] Disable per-task delay accounting
3207 nodsp [SH] Disable hardware DSP at boot time.
3209 noefi Disable EFI runtime services support.
3211 no_entry_flush [PPC] Don't flush the L1-D cache when entering the kernel.
3216 On X86-32 available only on PAE configured kernels.
3217 noexec=on: enable non-executable mappings (default)
3218 noexec=off: disable non-executable mappings
3221 Disable SMAP (Supervisor Mode Access Prevention)
3222 even if it is supported by processor.
3225 Disable SMEP (Supervisor Mode Execution Prevention)
3226 even if it is supported by processor.
3229 This affects only 32-bit executables.
3230 noexec32=on: enable non-executable mappings (default)
3231 read doesn't imply executable mappings
3232 noexec32=off: disable non-executable mappings
3233 read implies executable mappings
3235 nofpu [MIPS,SH] Disable hardware FPU at boot time.
3237 nofxsr [BUGS=X86-32] Disables x86 floating point extended
3238 register save and restore. The kernel will only save
3239 legacy floating-point registers on task switch.
3241 nohugeiomap [KNL,X86,PPC,ARM64] Disable kernel huge I/O mappings.
3243 nosmt [KNL,S390] Disable symmetric multithreading (SMT).
3244 Equivalent to smt=1.
3246 [KNL,X86] Disable symmetric multithreading (SMT).
3247 nosmt=force: Force disable SMT, cannot be undone
3248 via the sysfs control file.
3250 nospectre_v1 [X86,PPC] Disable mitigations for Spectre Variant 1
3251 (bounds check bypass). With this option data leaks are
3252 possible in the system.
3254 nospectre_v2 [X86,PPC_FSL_BOOK3E,ARM64] Disable all mitigations for
3255 the Spectre variant 2 (indirect branch prediction)
3256 vulnerability. System may allow data leaks with this
3259 nospec_store_bypass_disable
3260 [HW] Disable all mitigations for the Speculative Store Bypass vulnerability
3263 [PPC] Don't flush the L1-D cache after accessing user data.
3265 noxsave [BUGS=X86] Disables x86 extended register state save
3266 and restore using xsave. The kernel will fallback to
3267 enabling legacy floating-point and sse state.
3269 noxsaveopt [X86] Disables xsaveopt used in saving x86 extended
3270 register states. The kernel will fall back to use
3271 xsave to save the states. By using this parameter,
3272 performance of saving the states is degraded because
3273 xsave doesn't support modified optimization while
3274 xsaveopt supports it on xsaveopt enabled systems.
3276 noxsaves [X86] Disables xsaves and xrstors used in saving and
3277 restoring x86 extended register state in compacted
3278 form of xsave area. The kernel will fall back to use
3279 xsaveopt and xrstor to save and restore the states
3280 in standard form of xsave area. By using this
3281 parameter, xsave area per process might occupy more
3282 memory on xsaves enabled systems.
3284 nohlt [ARM,ARM64,MICROBLAZE,SH] Forces the kernel to busy wait
3285 in do_idle() and not use the arch_cpu_idle()
3286 implementation; requires CONFIG_GENERIC_IDLE_POLL_SETUP
3287 to be effective. This is useful on platforms where the
3288 sleep(SH) or wfi(ARM,ARM64) instructions do not work
3289 correctly or when doing power measurements to evalute
3290 the impact of the sleep instructions. This is also
3291 useful when using JTAG debugger.
3293 no_file_caps Tells the kernel not to honor file capabilities. The
3294 only way then for a file to be executed with privilege
3295 is to be setuid root or executed by root.
3297 nohalt [IA-64] Tells the kernel not to use the power saving
3298 function PAL_HALT_LIGHT when idle. This increases
3299 power-consumption. On the positive side, it reduces
3300 interrupt wake-up latency, which may improve performance
3301 in certain environments such as networked servers or
3305 Force pointers printed to the console or buffers to be
3306 unhashed. By default, when a pointer is printed via %p
3307 format string, that pointer is "hashed", i.e. obscured
3308 by hashing the pointer value. This is a security feature
3309 that hides actual kernel addresses from unprivileged
3310 users, but it also makes debugging the kernel more
3311 difficult since unequal pointers can no longer be
3312 compared. However, if this command-line option is
3313 specified, then all normal pointers will have their true
3314 value printed. Pointers printed via %pK may still be
3315 hashed. This option should only be specified when
3316 debugging the kernel. Please do not use on production
3319 nohibernate [HIBERNATION] Disable hibernation and resume.
3321 nohz= [KNL] Boottime enable/disable dynamic ticks
3322 Valid arguments: on, off
3325 nohz_full= [KNL,BOOT,SMP,ISOL]
3326 The argument is a cpu list, as described above.
3327 In kernels built with CONFIG_NO_HZ_FULL=y, set
3328 the specified list of CPUs whose tick will be stopped
3329 whenever possible. The boot CPU will be forced outside
3330 the range to maintain the timekeeping. Any CPUs
3331 in this list will have their RCU callbacks offloaded,
3332 just as if they had also been called out in the
3333 rcu_nocbs= boot parameter.
3335 noiotrap [SH] Disables trapped I/O port accesses.
3337 noirqdebug [X86-32] Disables the code which attempts to detect and
3338 disable unhandled interrupt sources.
3340 no_timer_check [X86,APIC] Disables the code which tests for
3341 broken timer IRQ sources.
3343 noisapnp [ISAPNP] Disables ISA PnP code.
3345 noinitrd [RAM] Tells the kernel not to load any configured
3348 nointremap [X86-64, Intel-IOMMU] Do not enable interrupt
3350 [Deprecated - use intremap=off]
3354 noinvpcid [X86] Disable the INVPCID cpu feature.
3356 nojitter [IA-64] Disables jitter checking for ITC timers.
3358 no-kvmclock [X86,KVM] Disable paravirtualized KVM clock driver
3360 no-kvmapf [X86,KVM] Disable paravirtualized asynchronous page
3364 [X86,PV_OPS] Disable paravirtualized VMware scheduler
3365 clock and use the default one.
3367 no-steal-acc [X86,PV_OPS,ARM64] Disable paravirtualized steal time
3368 accounting. steal time is computed, but won't
3369 influence scheduler behaviour
3371 nolapic [X86-32,APIC] Do not enable or use the local APIC.
3373 nolapic_timer [X86-32,APIC] Do not use the local APIC timer.
3375 noltlbs [PPC] Do not use large page/tlb entries for kernel
3376 lowmem mapping on PPC40x and PPC8xx
3378 nomca [IA-64] Disable machine check abort handling
3380 nomce [X86-32] Disable Machine Check Exception
3382 nomfgpt [X86-32] Disable Multi-Function General Purpose
3383 Timer usage (for AMD Geode machines).
3385 nonmi_ipi [X86] Disable using NMI IPIs during panic/reboot to
3386 shutdown the other cpus. Instead use the REBOOT_VECTOR
3389 nomodule Disable module load
3391 nopat [X86] Disable PAT (page attribute table extension of
3392 pagetables) support.
3394 nopcid [X86-64] Disable the PCID cpu feature.
3396 norandmaps Don't use address space randomization. Equivalent to
3397 echo 0 > /proc/sys/kernel/randomize_va_space
3399 noreplace-smp [X86-32,SMP] Don't replace SMP instructions
3400 with UP alternatives
3402 nordrand [X86] Disable kernel use of the RDRAND and
3403 RDSEED instructions even if they are supported
3404 by the processor. RDRAND and RDSEED are still
3405 available to user space applications.
3407 noresume [SWSUSP] Disables resume and restores original swap
3410 no-scroll [VGA] Disables scrollback.
3411 This is required for the Braillex ib80-piezo Braille
3412 reader made by F.H. Papenmeier (Germany).
3416 nosep [BUGS=X86-32] Disables x86 SYSENTER/SYSEXIT support.
3418 nosgx [X86-64,SGX] Disables Intel SGX kernel support.
3420 nosmp [SMP] Tells an SMP kernel to act as a UP kernel,
3421 and disable the IO APIC. legacy for "maxcpus=0".
3423 nosoftlockup [KNL] Disable the soft-lockup detector.
3425 nosync [HW,M68K] Disables sync negotiation for all devices.
3427 nowatchdog [KNL] Disable both lockup detectors, i.e.
3428 soft-lockup and NMI watchdog (hard-lockup).
3432 nox2apic [X86-64,APIC] Do not enable x2APIC mode.
3434 cpu0_hotplug [X86] Turn on CPU0 hotplug feature when
3435 CONFIG_BOOTPARAM_HOTPLUG_CPU0 is off.
3436 Some features depend on CPU0. Known dependencies are:
3437 1. Resume from suspend/hibernate depends on CPU0.
3438 Suspend/hibernate will fail if CPU0 is offline and you
3439 need to online CPU0 before suspend/hibernate.
3440 2. PIC interrupts also depend on CPU0. CPU0 can't be
3441 removed if a PIC interrupt is detected.
3442 It's said poweroff/reboot may depend on CPU0 on some
3443 machines although I haven't seen such issues so far
3444 after CPU0 is offline on a few tested machines.
3445 If the dependencies are under your control, you can
3446 turn on cpu0_hotplug.
3448 nps_mtm_hs_ctr= [KNL,ARC]
3449 This parameter sets the maximum duration, in
3450 cycles, each HW thread of the CTOP can run
3451 without interruptions, before HW switches it.
3452 The actual maximum duration is 16 times this
3454 Format: integer between 1 and 255
3457 nptcg= [IA-64] Override max number of concurrent global TLB
3458 purges which is reported from either PAL_VM_SUMMARY or
3461 nr_cpus= [SMP] Maximum number of processors that an SMP kernel
3462 could support. nr_cpus=n : n >= 1 limits the kernel to
3463 support 'n' processors. It could be larger than the
3464 number of already plugged CPU during bootup, later in
3465 runtime you can physically add extra cpu until it reaches
3466 n. So during boot up some boot time memory for per-cpu
3467 variables need be pre-allocated for later physical cpu
3470 nr_uarts= [SERIAL] maximum number of UARTs to be registered.
3472 numa_balancing= [KNL,ARM64,PPC,RISCV,S390,X86] Enable or disable automatic
3474 Allowed values are enable and disable
3476 numa_zonelist_order= [KNL, BOOT] Select zonelist order for NUMA.
3477 'node', 'default' can be specified
3478 This can be set from sysctl after boot.
3479 See Documentation/admin-guide/sysctl/vm.rst for details.
3481 ohci1394_dma=early [HW] enable debugging via the ohci1394 driver.
3482 See Documentation/core-api/debugging-via-ohci1394.rst for more
3485 olpc_ec_timeout= [OLPC] ms delay when issuing EC commands
3486 Rather than timing out after 20 ms if an EC
3487 command is not properly ACKed, override the length
3488 of the timeout. We have interrupts disabled while
3489 waiting for the ACK, so if this is set too high
3490 interrupts *may* be lost!
3492 omap_mux= [OMAP] Override bootloader pin multiplexing.
3493 Format: <mux_mode0.mode_name=value>...
3494 For example, to override I2C bus2:
3495 omap_mux=i2c2_scl.i2c2_scl=0x100,i2c2_sda.i2c2_sda=0x100
3497 oops=panic Always panic on oopses. Default is to just kill the
3498 process, but there is a small probability of
3499 deadlocking the machine.
3500 This will also cause panics on machine check exceptions.
3501 Useful together with panic=30 to trigger a reboot.
3504 [KNL] Boolean flag to control whether the page allocator
3505 should randomize its free lists. The randomization may
3506 be automatically enabled if the kernel detects it is
3507 running on a platform with a direct-mapped memory-side
3508 cache, and this parameter can be used to
3509 override/disable that behavior. The state of the flag
3510 can be read from sysfs at:
3511 /sys/module/page_alloc/parameters/shuffle.
3513 page_owner= [KNL] Boot-time page_owner enabling option.
3514 Storage of the information about who allocated
3515 each page is disabled in default. With this switch,
3517 on: enable the feature
3519 page_poison= [KNL] Boot-time parameter changing the state of
3520 poisoning on the buddy allocator, available with
3521 CONFIG_PAGE_POISONING=y.
3522 off: turn off poisoning (default)
3523 on: turn on poisoning
3525 panic= [KNL] Kernel behaviour on panic: delay <timeout>
3526 timeout > 0: seconds before rebooting
3527 timeout = 0: wait forever
3528 timeout < 0: reboot immediately
3531 panic_print= Bitmask for printing system info when panic happens.
3532 User can chose combination of the following bits:
3533 bit 0: print all tasks info
3534 bit 1: print system memory info
3535 bit 2: print timer info
3536 bit 3: print locks info if CONFIG_LOCKDEP is on
3537 bit 4: print ftrace buffer
3538 bit 5: print all printk messages in buffer
3540 panic_on_taint= Bitmask for conditionally calling panic() in add_taint()
3541 Format: <hex>[,nousertaint]
3542 Hexadecimal bitmask representing the set of TAINT flags
3543 that will cause the kernel to panic when add_taint() is
3544 called with any of the flags in this set.
3545 The optional switch "nousertaint" can be utilized to
3546 prevent userspace forced crashes by writing to sysctl
3547 /proc/sys/kernel/tainted any flagset matching with the
3548 bitmask set on panic_on_taint.
3549 See Documentation/admin-guide/tainted-kernels.rst for
3550 extra details on the taint flags that users can pick
3551 to compose the bitmask to assign to panic_on_taint.
3553 panic_on_warn panic() instead of WARN(). Useful to cause kdump
3556 crash_kexec_post_notifiers
3557 Run kdump after running panic-notifiers and dumping
3558 kmsg. This only for the users who doubt kdump always
3559 succeeds in any situation.
3560 Note that this also increases risks of kdump failure,
3561 because some panic notifiers can make the crashed
3562 kernel more unstable.
3564 parkbd.port= [HW] Parallel port number the keyboard adapter is
3565 connected to, default is 0.
3567 parkbd.mode= [HW] Parallel port keyboard adapter mode of operation,
3568 0 for XT, 1 for AT (default is AT).
3571 parport= [HW,PPT] Specify parallel ports. 0 disables.
3572 Format: { 0 | auto | 0xBBB[,IRQ[,DMA]] }
3573 Use 'auto' to force the driver to use any
3574 IRQ/DMA settings detected (the default is to
3575 ignore detected IRQ/DMA settings because of
3576 possible conflicts). You can specify the base
3577 address, IRQ, and DMA settings; IRQ and DMA
3578 should be numbers, or 'auto' (for using detected
3579 settings on that particular port), or 'nofifo'
3580 (to avoid using a FIFO even if it is detected).
3581 Parallel ports are assigned in the order they
3582 are specified on the command line, starting
3585 parport_init_mode= [HW,PPT]
3586 Configure VIA parallel port to operate in
3587 a specific mode. This is necessary on Pegasos
3588 computer where firmware has no options for setting
3589 up parallel port mode and sets it to spp.
3590 Currently this function knows 686a and 8231 chips.
3591 Format: [spp|ps2|epp|ecp|ecpepp]
3594 Halt all CPUs after the first oops has been printed for
3595 the specified number of seconds. This is to be used if
3596 your oopses keep scrolling off the screen.
3601 See header of drivers/block/paride/pcd.c.
3602 See also Documentation/admin-guide/blockdev/paride.rst.
3604 pci=option[,option...] [PCI] various PCI subsystem options.
3606 Some options herein operate on a specific device
3607 or a set of devices (<pci_dev>). These are
3608 specified in one of the following formats:
3610 [<domain>:]<bus>:<dev>.<func>[/<dev>.<func>]*
3611 pci:<vendor>:<device>[:<subvendor>:<subdevice>]
3613 Note: the first format specifies a PCI
3614 bus/device/function address which may change
3615 if new hardware is inserted, if motherboard
3616 firmware changes, or due to changes caused
3617 by other kernel parameters. If the
3618 domain is left unspecified, it is
3619 taken to be zero. Optionally, a path
3620 to a device through multiple device/function
3621 addresses can be specified after the base
3622 address (this is more robust against
3623 renumbering issues). The second format
3624 selects devices using IDs from the
3625 configuration space which may match multiple
3626 devices in the system.
3628 earlydump dump PCI config space before the kernel
3630 off [X86] don't probe for the PCI bus
3631 bios [X86-32] force use of PCI BIOS, don't access
3632 the hardware directly. Use this if your machine
3633 has a non-standard PCI host bridge.
3634 nobios [X86-32] disallow use of PCI BIOS, only direct
3635 hardware access methods are allowed. Use this
3636 if you experience crashes upon bootup and you
3637 suspect they are caused by the BIOS.
3638 conf1 [X86] Force use of PCI Configuration Access
3639 Mechanism 1 (config address in IO port 0xCF8,
3640 data in IO port 0xCFC, both 32-bit).
3641 conf2 [X86] Force use of PCI Configuration Access
3642 Mechanism 2 (IO port 0xCF8 is an 8-bit port for
3643 the function, IO port 0xCFA, also 8-bit, sets
3644 bus number. The config space is then accessed
3645 through ports 0xC000-0xCFFF).
3646 See http://wiki.osdev.org/PCI for more info
3647 on the configuration access mechanisms.
3648 noaer [PCIE] If the PCIEAER kernel config parameter is
3649 enabled, this kernel boot option can be used to
3650 disable the use of PCIE advanced error reporting.
3651 nodomains [PCI] Disable support for multiple PCI
3652 root domains (aka PCI segments, in ACPI-speak).
3653 nommconf [X86] Disable use of MMCONFIG for PCI
3655 check_enable_amd_mmconf [X86] check for and enable
3656 properly configured MMIO access to PCI
3657 config space on AMD family 10h CPU
3658 nomsi [MSI] If the PCI_MSI kernel config parameter is
3659 enabled, this kernel boot option can be used to
3660 disable the use of MSI interrupts system-wide.
3661 noioapicquirk [APIC] Disable all boot interrupt quirks.
3662 Safety option to keep boot IRQs enabled. This
3663 should never be necessary.
3664 ioapicreroute [APIC] Enable rerouting of boot IRQs to the
3665 primary IO-APIC for bridges that cannot disable
3666 boot IRQs. This fixes a source of spurious IRQs
3667 when the system masks IRQs.
3668 noioapicreroute [APIC] Disable workaround that uses the
3669 boot IRQ equivalent of an IRQ that connects to
3670 a chipset where boot IRQs cannot be disabled.
3671 The opposite of ioapicreroute.
3672 biosirq [X86-32] Use PCI BIOS calls to get the interrupt
3673 routing table. These calls are known to be buggy
3674 on several machines and they hang the machine
3675 when used, but on other computers it's the only
3676 way to get the interrupt routing table. Try
3677 this option if the kernel is unable to allocate
3678 IRQs or discover secondary PCI buses on your
3680 rom [X86] Assign address space to expansion ROMs.
3681 Use with caution as certain devices share
3682 address decoders between ROMs and other
3684 norom [X86] Do not assign address space to
3685 expansion ROMs that do not already have
3686 BIOS assigned address ranges.
3687 nobar [X86] Do not assign address space to the
3688 BARs that weren't assigned by the BIOS.
3689 irqmask=0xMMMM [X86] Set a bit mask of IRQs allowed to be
3690 assigned automatically to PCI devices. You can
3691 make the kernel exclude IRQs of your ISA cards
3693 pirqaddr=0xAAAAA [X86] Specify the physical address
3694 of the PIRQ table (normally generated
3695 by the BIOS) if it is outside the
3696 F0000h-100000h range.
3697 lastbus=N [X86] Scan all buses thru bus #N. Can be
3698 useful if the kernel is unable to find your
3699 secondary buses and you want to tell it
3700 explicitly which ones they are.
3701 assign-busses [X86] Always assign all PCI bus
3702 numbers ourselves, overriding
3703 whatever the firmware may have done.
3704 usepirqmask [X86] Honor the possible IRQ mask stored
3705 in the BIOS $PIR table. This is needed on
3706 some systems with broken BIOSes, notably
3707 some HP Pavilion N5400 and Omnibook XE3
3708 notebooks. This will have no effect if ACPI
3709 IRQ routing is enabled.
3710 noacpi [X86] Do not use ACPI for IRQ routing
3711 or for PCI scanning.
3712 use_crs [X86] Use PCI host bridge window information
3713 from ACPI. On BIOSes from 2008 or later, this
3714 is enabled by default. If you need to use this,
3715 please report a bug.
3716 nocrs [X86] Ignore PCI host bridge windows from ACPI.
3717 If you need to use this, please report a bug.
3718 routeirq Do IRQ routing for all PCI devices.
3719 This is normally done in pci_enable_device(),
3720 so this option is a temporary workaround
3721 for broken drivers that don't call it.
3722 skip_isa_align [X86] do not align io start addr, so can
3723 handle more pci cards
3724 noearly [X86] Don't do any early type 1 scanning.
3725 This might help on some broken boards which
3726 machine check when some devices' config space
3727 is read. But various workarounds are disabled
3728 and some IOMMU drivers will not work.
3729 bfsort Sort PCI devices into breadth-first order.
3730 This sorting is done to get a device
3731 order compatible with older (<= 2.4) kernels.
3732 nobfsort Don't sort PCI devices into breadth-first order.
3733 pcie_bus_tune_off Disable PCIe MPS (Max Payload Size)
3734 tuning and use the BIOS-configured MPS defaults.
3735 pcie_bus_safe Set every device's MPS to the largest value
3736 supported by all devices below the root complex.
3737 pcie_bus_perf Set device MPS to the largest allowable MPS
3738 based on its parent bus. Also set MRRS (Max
3739 Read Request Size) to the largest supported
3740 value (no larger than the MPS that the device
3741 or bus can support) for best performance.
3742 pcie_bus_peer2peer Set every device's MPS to 128B, which
3743 every device is guaranteed to support. This
3744 configuration allows peer-to-peer DMA between
3745 any pair of devices, possibly at the cost of
3746 reduced performance. This also guarantees
3747 that hot-added devices will work.
3748 cbiosize=nn[KMG] The fixed amount of bus space which is
3749 reserved for the CardBus bridge's IO window.
3750 The default value is 256 bytes.
3751 cbmemsize=nn[KMG] The fixed amount of bus space which is
3752 reserved for the CardBus bridge's memory
3753 window. The default value is 64 megabytes.
3756 [<order of align>@]<pci_dev>[; ...]
3757 Specifies alignment and device to reassign
3758 aligned memory resources. How to
3759 specify the device is described above.
3760 If <order of align> is not specified,
3761 PAGE_SIZE is used as alignment.
3762 A PCI-PCI bridge can be specified if resource
3763 windows need to be expanded.
3764 To specify the alignment for several
3765 instances of a device, the PCI vendor,
3766 device, subvendor, and subdevice may be
3767 specified, e.g., 12@pci:8086:9c22:103c:198f
3768 for 4096-byte alignment.
3769 ecrc= Enable/disable PCIe ECRC (transaction layer
3770 end-to-end CRC checking).
3771 bios: Use BIOS/firmware settings. This is the
3775 hpiosize=nn[KMG] The fixed amount of bus space which is
3776 reserved for hotplug bridge's IO window.
3777 Default size is 256 bytes.
3778 hpmmiosize=nn[KMG] The fixed amount of bus space which is
3779 reserved for hotplug bridge's MMIO window.
3780 Default size is 2 megabytes.
3781 hpmmioprefsize=nn[KMG] The fixed amount of bus space which is
3782 reserved for hotplug bridge's MMIO_PREF window.
3783 Default size is 2 megabytes.
3784 hpmemsize=nn[KMG] The fixed amount of bus space which is
3785 reserved for hotplug bridge's MMIO and
3787 Default size is 2 megabytes.
3788 hpbussize=nn The minimum amount of additional bus numbers
3789 reserved for buses below a hotplug bridge.
3791 realloc= Enable/disable reallocating PCI bridge resources
3792 if allocations done by BIOS are too small to
3793 accommodate resources required by all child
3795 off: Turn realloc off
3797 realloc same as realloc=on
3798 noari do not use PCIe ARI.
3799 noats [PCIE, Intel-IOMMU, AMD-IOMMU]
3800 do not use PCIe ATS (and IOMMU device IOTLB).
3801 pcie_scan_all Scan all possible PCIe devices. Otherwise we
3802 only look for one device below a PCIe downstream
3804 big_root_window Try to add a big 64bit memory window to the PCIe
3805 root complex on AMD CPUs. Some GFX hardware
3806 can resize a BAR to allow access to all VRAM.
3807 Adding the window is slightly risky (it may
3808 conflict with unreported devices), so this
3810 disable_acs_redir=<pci_dev>[; ...]
3811 Specify one or more PCI devices (in the format
3812 specified above) separated by semicolons.
3813 Each device specified will have the PCI ACS
3814 redirect capabilities forced off which will
3815 allow P2P traffic between devices through
3816 bridges without forcing it upstream. Note:
3817 this removes isolation between devices and
3818 may put more devices in an IOMMU group.
3819 force_floating [S390] Force usage of floating interrupts.
3820 nomio [S390] Do not use MIO instructions.
3821 norid [S390] ignore the RID field and force use of
3822 one PCI domain per PCI function
3824 pcie_aspm= [PCIE] Forcibly enable or disable PCIe Active State Power
3827 force Enable ASPM even on devices that claim not to support it.
3828 WARNING: Forcing ASPM on may cause system lockups.
3830 pcie_ports= [PCIE] PCIe port services handling:
3831 native Use native PCIe services (PME, AER, DPC, PCIe hotplug)
3832 even if the platform doesn't give the OS permission to
3833 use them. This may cause conflicts if the platform
3834 also tries to use these services.
3835 dpc-native Use native PCIe service for DPC only. May
3836 cause conflicts if firmware uses AER or DPC.
3837 compat Disable native PCIe services (PME, AER, DPC, PCIe
3840 pcie_port_pm= [PCIE] PCIe port power management handling:
3841 off Disable power management of all PCIe ports
3842 force Forcibly enable power management of all PCIe ports
3844 pcie_pme= [PCIE,PM] Native PCIe PME signaling options:
3845 nomsi Do not use MSI for native PCIe PME signaling (this makes
3846 all PCIe root ports use INTx for all services).
3848 pcmv= [HW,PCMCIA] BadgePAD 4
3852 Keep all power-domains already enabled by bootloader on,
3853 even if no driver has claimed them. This is useful
3854 for debug and development, but should not be
3855 needed on a platform with proper driver support.
3858 See Documentation/admin-guide/blockdev/paride.rst.
3860 pdcchassis= [PARISC,HW] Disable/Enable PDC Chassis Status codes at
3863 See arch/parisc/kernel/pdc_chassis.c
3865 percpu_alloc= Select which percpu first chunk allocator to use.
3866 Currently supported values are "embed" and "page".
3867 Archs may support subset or none of the selections.
3868 See comments in mm/percpu.c for details on each
3869 allocator. This parameter is primarily for debugging
3870 and performance comparison.
3873 See Documentation/admin-guide/blockdev/paride.rst.
3876 See Documentation/admin-guide/blockdev/paride.rst.
3878 pirq= [SMP,APIC] Manual mp-table setup
3879 See Documentation/x86/i386/IO-APIC.rst.
3881 plip= [PPT,NET] Parallel port network link
3882 Format: { parport<nr> | timid | 0 }
3883 See also Documentation/admin-guide/parport.rst.
3885 pmtmr= [X86] Manual setup of pmtmr I/O Port.
3886 Override pmtimer IOPort with a hex value.
3889 pm_debug_messages [SUSPEND,KNL]
3890 Enable suspend/resume debug messages during boot up.
3893 Enable PNP debug messages (depends on the
3894 CONFIG_PNP_DEBUG_MESSAGES option). Change at run-time
3895 via /sys/module/pnp/parameters/debug. We always show
3896 current resource usage; turning this on also shows
3897 possible settings and some assignment information.
3903 { on | off | curr | res | no-curr | no-res }
3906 [ISAPNP] Exclude IRQs for the autoconfiguration
3909 [ISAPNP] Exclude DMAs for the autoconfiguration
3911 pnp_reserve_io= [ISAPNP] Exclude I/O ports for the autoconfiguration
3912 Ranges are in pairs (I/O port base and size).
3915 [ISAPNP] Exclude memory regions for the
3917 Ranges are in pairs (memory base and size).
3919 ports= [IP_VS_FTP] IPVS ftp helper module
3921 Up to 8 (IP_VS_APP_MAX_PORTS) ports
3923 Format: <port>,<port>....
3925 powersave=off [PPC] This option disables power saving features.
3926 It specifically disables cpuidle and sets the
3927 platform machine description specific power_save
3928 function to NULL. On Idle the CPU just reduces
3931 ppc_strict_facility_enable
3932 [PPC] This option catches any kernel floating point,
3933 Altivec, VSX and SPE outside of regions specifically
3934 allowed (eg kernel_enable_fpu()/kernel_disable_fpu()).
3935 There is some performance impact when enabling this.
3939 Disable Hardware Transactional Memory
3942 Select preemption mode if you have CONFIG_PREEMPT_DYNAMIC
3943 none - Limited to cond_resched() calls
3944 voluntary - Limited to cond_resched() and might_sleep() calls
3945 full - Any section that isn't explicitly preempt disabled
3946 can be preempted anytime.
3948 print-fatal-signals=
3949 [KNL] debug: print fatal signals
3951 If enabled, warn about various signal handling
3952 related application anomalies: too many signals,
3953 too many POSIX.1 timers, fatal signals causing a
3956 If you hit the warning due to signal overflow,
3957 you might want to try "ulimit -i unlimited".
3961 printk.always_kmsg_dump=
3962 Trigger kmsg_dump for cases other than kernel oops or
3964 Format: <bool> (1/Y/y=enable, 0/N/n=disable)
3967 printk.devkmsg={on,off,ratelimit}
3968 Control writing to /dev/kmsg.
3969 on - unlimited logging to /dev/kmsg from userspace
3970 off - logging to /dev/kmsg disabled
3971 ratelimit - ratelimit the logging
3974 printk.time= Show timing data prefixed to each printk message line
3975 Format: <bool> (1/Y/y=enable, 0/N/n=disable)
3977 processor.max_cstate= [HW,ACPI]
3978 Limit processor to maximum C-state
3979 max_cstate=9 overrides any DMI blacklist limit.
3981 processor.nocst [HW,ACPI]
3982 Ignore the _CST method to determine C-states,
3983 instead using the legacy FADT method
3985 profile= [KNL] Enable kernel profiling via /proc/profile
3986 Format: [<profiletype>,]<number>
3987 Param: <profiletype>: "schedule", "sleep", or "kvm"
3988 [defaults to kernel profiling]
3989 Param: "schedule" - profile schedule points.
3990 Param: "sleep" - profile D-state sleeping (millisecs).
3991 Requires CONFIG_SCHEDSTATS
3992 Param: "kvm" - profile VM exits.
3993 Param: <number> - step/bucket size as a power of 2 for
3994 statistical time based profiling.
3996 prompt_ramdisk= [RAM] [Deprecated]
3998 prot_virt= [S390] enable hosting protected virtual machines
3999 isolated from the hypervisor (if hardware supports
4003 psi= [KNL] Enable or disable pressure stall information
4007 psmouse.proto= [HW,MOUSE] Highest PS2 mouse protocol extension to
4008 probe for; one of (bare|imps|exps|lifebook|any).
4009 psmouse.rate= [HW,MOUSE] Set desired mouse report rate, in reports
4011 psmouse.resetafter= [HW,MOUSE]
4012 Try to reset the device after so many bad packets
4015 [HW,MOUSE] Set desired mouse resolution, in dpi.
4016 psmouse.smartscroll=
4017 [HW,MOUSE] Controls Logitech smartscroll autorepeat.
4018 0 = disabled, 1 = enabled (default).
4020 pstore.backend= Specify the name of the pstore backend to use
4023 See Documentation/admin-guide/blockdev/paride.rst.
4025 pti= [X86-64] Control Page Table Isolation of user and
4026 kernel address spaces. Disabling this feature
4027 removes hardening, but improves performance of
4028 system calls and interrupts.
4030 on - unconditionally enable
4031 off - unconditionally disable
4032 auto - kernel detects whether your CPU model is
4033 vulnerable to issues that PTI mitigates
4035 Not specifying this option is equivalent to pti=auto.
4038 Equivalent to pti=off
4041 [KNL] Number of legacy pty's. Overwrites compiled-in
4044 quiet [KNL] Disable most log messages
4049 See Documentation/admin-guide/md.rst.
4051 ramdisk_size= [RAM] Sizes of RAM disks in kilobytes
4052 See Documentation/admin-guide/blockdev/ramdisk.rst.
4054 ramdisk_start= [RAM] RAM disk image start address
4056 random.trust_cpu={on,off}
4057 [KNL] Enable or disable trusting the use of the
4058 CPU's random number generator (if available) to
4059 fully seed the kernel's CRNG. Default is controlled
4060 by CONFIG_RANDOM_TRUST_CPU.
4062 randomize_kstack_offset=
4063 [KNL] Enable or disable kernel stack offset
4064 randomization, which provides roughly 5 bits of
4065 entropy, frustrating memory corruption attacks
4066 that depend on stack address determinism or
4067 cross-syscall address exposures. This is only
4068 available on architectures that have defined
4069 CONFIG_HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET.
4070 Format: <bool> (1/Y/y=enable, 0/N/n=disable)
4071 Default is CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT.
4073 ras=option[,option,...] [KNL] RAS-specific options
4076 Disable the Correctable Errors Collector,
4077 see CONFIG_RAS_CEC help text.
4080 The argument is a cpu list, as described above,
4081 except that the string "all" can be used to
4082 specify every CPU on the system.
4084 In kernels built with CONFIG_RCU_NOCB_CPU=y, set
4085 the specified list of CPUs to be no-callback CPUs.
4086 Invocation of these CPUs' RCU callbacks will be
4087 offloaded to "rcuox/N" kthreads created for that
4088 purpose, where "x" is "p" for RCU-preempt, and
4089 "s" for RCU-sched, and "N" is the CPU number.
4090 This reduces OS jitter on the offloaded CPUs,
4091 which can be useful for HPC and real-time
4092 workloads. It can also improve energy efficiency
4093 for asymmetric multiprocessors.
4096 Rather than requiring that offloaded CPUs
4097 (specified by rcu_nocbs= above) explicitly
4098 awaken the corresponding "rcuoN" kthreads,
4099 make these kthreads poll for callbacks.
4100 This improves the real-time response for the
4101 offloaded CPUs by relieving them of the need to
4102 wake up the corresponding kthread, but degrades
4103 energy efficiency by requiring that the kthreads
4104 periodically wake up to do the polling.
4106 rcutree.blimit= [KNL]
4107 Set maximum number of finished RCU callbacks to
4108 process in one batch.
4110 rcutree.dump_tree= [KNL]
4111 Dump the structure of the rcu_node combining tree
4112 out at early boot. This is used for diagnostic
4113 purposes, to verify correct tree setup.
4115 rcutree.gp_cleanup_delay= [KNL]
4116 Set the number of jiffies to delay each step of
4117 RCU grace-period cleanup.
4119 rcutree.gp_init_delay= [KNL]
4120 Set the number of jiffies to delay each step of
4121 RCU grace-period initialization.
4123 rcutree.gp_preinit_delay= [KNL]
4124 Set the number of jiffies to delay each step of
4125 RCU grace-period pre-initialization, that is,
4126 the propagation of recent CPU-hotplug changes up
4127 the rcu_node combining tree.
4129 rcutree.use_softirq= [KNL]
4130 If set to zero, move all RCU_SOFTIRQ processing to
4131 per-CPU rcuc kthreads. Defaults to a non-zero
4132 value, meaning that RCU_SOFTIRQ is used by default.
4133 Specify rcutree.use_softirq=0 to use rcuc kthreads.
4135 But note that CONFIG_PREEMPT_RT=y kernels disable
4136 this kernel boot parameter, forcibly setting it
4139 rcutree.rcu_fanout_exact= [KNL]
4140 Disable autobalancing of the rcu_node combining
4141 tree. This is used by rcutorture, and might
4142 possibly be useful for architectures having high
4143 cache-to-cache transfer latencies.
4145 rcutree.rcu_fanout_leaf= [KNL]
4146 Change the number of CPUs assigned to each
4147 leaf rcu_node structure. Useful for very
4148 large systems, which will choose the value 64,
4149 and for NUMA systems with large remote-access
4150 latencies, which will choose a value aligned
4151 with the appropriate hardware boundaries.
4153 rcutree.rcu_min_cached_objs= [KNL]
4154 Minimum number of objects which are cached and
4155 maintained per one CPU. Object size is equal
4156 to PAGE_SIZE. The cache allows to reduce the
4157 pressure to page allocator, also it makes the
4158 whole algorithm to behave better in low memory
4161 rcutree.jiffies_till_first_fqs= [KNL]
4162 Set delay from grace-period initialization to
4163 first attempt to force quiescent states.
4164 Units are jiffies, minimum value is zero,
4165 and maximum value is HZ.
4167 rcutree.jiffies_till_next_fqs= [KNL]
4168 Set delay between subsequent attempts to force
4169 quiescent states. Units are jiffies, minimum
4170 value is one, and maximum value is HZ.
4172 rcutree.jiffies_till_sched_qs= [KNL]
4173 Set required age in jiffies for a
4174 given grace period before RCU starts
4175 soliciting quiescent-state help from
4176 rcu_note_context_switch() and cond_resched().
4177 If not specified, the kernel will calculate
4178 a value based on the most recent settings
4179 of rcutree.jiffies_till_first_fqs
4180 and rcutree.jiffies_till_next_fqs.
4181 This calculated value may be viewed in
4182 rcutree.jiffies_to_sched_qs. Any attempt to set
4183 rcutree.jiffies_to_sched_qs will be cheerfully
4186 rcutree.kthread_prio= [KNL,BOOT]
4187 Set the SCHED_FIFO priority of the RCU per-CPU
4188 kthreads (rcuc/N). This value is also used for
4189 the priority of the RCU boost threads (rcub/N)
4190 and for the RCU grace-period kthreads (rcu_bh,
4191 rcu_preempt, and rcu_sched). If RCU_BOOST is
4192 set, valid values are 1-99 and the default is 1
4193 (the least-favored priority). Otherwise, when
4194 RCU_BOOST is not set, valid values are 0-99 and
4195 the default is zero (non-realtime operation).
4197 rcutree.rcu_nocb_gp_stride= [KNL]
4198 Set the number of NOCB callback kthreads in
4199 each group, which defaults to the square root
4200 of the number of CPUs. Larger numbers reduce
4201 the wakeup overhead on the global grace-period
4202 kthread, but increases that same overhead on
4203 each group's NOCB grace-period kthread.
4205 rcutree.qhimark= [KNL]
4206 Set threshold of queued RCU callbacks beyond which
4207 batch limiting is disabled.
4209 rcutree.qlowmark= [KNL]
4210 Set threshold of queued RCU callbacks below which
4211 batch limiting is re-enabled.
4213 rcutree.qovld= [KNL]
4214 Set threshold of queued RCU callbacks beyond which
4215 RCU's force-quiescent-state scan will aggressively
4216 enlist help from cond_resched() and sched IPIs to
4217 help CPUs more quickly reach quiescent states.
4218 Set to less than zero to make this be set based
4219 on rcutree.qhimark at boot time and to zero to
4220 disable more aggressive help enlistment.
4222 rcutree.rcu_idle_gp_delay= [KNL]
4223 Set wakeup interval for idle CPUs that have
4224 RCU callbacks (RCU_FAST_NO_HZ=y).
4226 rcutree.rcu_kick_kthreads= [KNL]
4227 Cause the grace-period kthread to get an extra
4228 wake_up() if it sleeps three times longer than
4229 it should at force-quiescent-state time.
4230 This wake_up() will be accompanied by a
4231 WARN_ONCE() splat and an ftrace_dump().
4233 rcutree.rcu_unlock_delay= [KNL]
4234 In CONFIG_RCU_STRICT_GRACE_PERIOD=y kernels,
4235 this specifies an rcu_read_unlock()-time delay
4236 in microseconds. This defaults to zero.
4237 Larger delays increase the probability of
4238 catching RCU pointer leaks, that is, buggy use
4239 of RCU-protected pointers after the relevant
4240 rcu_read_unlock() has completed.
4242 rcutree.sysrq_rcu= [KNL]
4243 Commandeer a sysrq key to dump out Tree RCU's
4244 rcu_node tree with an eye towards determining
4245 why a new grace period has not yet started.
4247 rcuscale.gp_async= [KNL]
4248 Measure performance of asynchronous
4249 grace-period primitives such as call_rcu().
4251 rcuscale.gp_async_max= [KNL]
4252 Specify the maximum number of outstanding
4253 callbacks per writer thread. When a writer
4254 thread exceeds this limit, it invokes the
4255 corresponding flavor of rcu_barrier() to allow
4256 previously posted callbacks to drain.
4258 rcuscale.gp_exp= [KNL]
4259 Measure performance of expedited synchronous
4260 grace-period primitives.
4262 rcuscale.holdoff= [KNL]
4263 Set test-start holdoff period. The purpose of
4264 this parameter is to delay the start of the
4265 test until boot completes in order to avoid
4268 rcuscale.kfree_rcu_test= [KNL]
4269 Set to measure performance of kfree_rcu() flooding.
4271 rcuscale.kfree_nthreads= [KNL]
4272 The number of threads running loops of kfree_rcu().
4274 rcuscale.kfree_alloc_num= [KNL]
4275 Number of allocations and frees done in an iteration.
4277 rcuscale.kfree_loops= [KNL]
4278 Number of loops doing rcuscale.kfree_alloc_num number
4279 of allocations and frees.
4281 rcuscale.nreaders= [KNL]
4282 Set number of RCU readers. The value -1 selects
4283 N, where N is the number of CPUs. A value
4284 "n" less than -1 selects N-n+1, where N is again
4285 the number of CPUs. For example, -2 selects N
4286 (the number of CPUs), -3 selects N+1, and so on.
4287 A value of "n" less than or equal to -N selects
4290 rcuscale.nwriters= [KNL]
4291 Set number of RCU writers. The values operate
4292 the same as for rcuscale.nreaders.
4293 N, where N is the number of CPUs
4295 rcuscale.perf_type= [KNL]
4296 Specify the RCU implementation to test.
4298 rcuscale.shutdown= [KNL]
4299 Shut the system down after performance tests
4300 complete. This is useful for hands-off automated
4303 rcuscale.verbose= [KNL]
4304 Enable additional printk() statements.
4306 rcuscale.writer_holdoff= [KNL]
4307 Write-side holdoff between grace periods,
4308 in microseconds. The default of zero says
4311 rcutorture.fqs_duration= [KNL]
4312 Set duration of force_quiescent_state bursts
4315 rcutorture.fqs_holdoff= [KNL]
4316 Set holdoff time within force_quiescent_state bursts
4319 rcutorture.fqs_stutter= [KNL]
4320 Set wait time between force_quiescent_state bursts
4323 rcutorture.fwd_progress= [KNL]
4324 Enable RCU grace-period forward-progress testing
4325 for the types of RCU supporting this notion.
4327 rcutorture.fwd_progress_div= [KNL]
4328 Specify the fraction of a CPU-stall-warning
4329 period to do tight-loop forward-progress testing.
4331 rcutorture.fwd_progress_holdoff= [KNL]
4332 Number of seconds to wait between successive
4333 forward-progress tests.
4335 rcutorture.fwd_progress_need_resched= [KNL]
4336 Enclose cond_resched() calls within checks for
4337 need_resched() during tight-loop forward-progress
4340 rcutorture.gp_cond= [KNL]
4341 Use conditional/asynchronous update-side
4342 primitives, if available.
4344 rcutorture.gp_exp= [KNL]
4345 Use expedited update-side primitives, if available.
4347 rcutorture.gp_normal= [KNL]
4348 Use normal (non-expedited) asynchronous
4349 update-side primitives, if available.
4351 rcutorture.gp_sync= [KNL]
4352 Use normal (non-expedited) synchronous
4353 update-side primitives, if available. If all
4354 of rcutorture.gp_cond=, rcutorture.gp_exp=,
4355 rcutorture.gp_normal=, and rcutorture.gp_sync=
4356 are zero, rcutorture acts as if is interpreted
4357 they are all non-zero.
4359 rcutorture.irqreader= [KNL]
4360 Run RCU readers from irq handlers, or, more
4361 accurately, from a timer handler. Not all RCU
4362 flavors take kindly to this sort of thing.
4364 rcutorture.leakpointer= [KNL]
4365 Leak an RCU-protected pointer out of the reader.
4366 This can of course result in splats, and is
4367 intended to test the ability of things like
4368 CONFIG_RCU_STRICT_GRACE_PERIOD=y to detect
4371 rcutorture.n_barrier_cbs= [KNL]
4372 Set callbacks/threads for rcu_barrier() testing.
4374 rcutorture.nfakewriters= [KNL]
4375 Set number of concurrent RCU writers. These just
4376 stress RCU, they don't participate in the actual
4377 test, hence the "fake".
4379 rcutorture.nocbs_nthreads= [KNL]
4380 Set number of RCU callback-offload togglers.
4381 Zero (the default) disables toggling.
4383 rcutorture.nocbs_toggle= [KNL]
4384 Set the delay in milliseconds between successive
4385 callback-offload toggling attempts.
4387 rcutorture.nreaders= [KNL]
4388 Set number of RCU readers. The value -1 selects
4389 N-1, where N is the number of CPUs. A value
4390 "n" less than -1 selects N-n-2, where N is again
4391 the number of CPUs. For example, -2 selects N
4392 (the number of CPUs), -3 selects N+1, and so on.
4394 rcutorture.object_debug= [KNL]
4395 Enable debug-object double-call_rcu() testing.
4397 rcutorture.onoff_holdoff= [KNL]
4398 Set time (s) after boot for CPU-hotplug testing.
4400 rcutorture.onoff_interval= [KNL]
4401 Set time (jiffies) between CPU-hotplug operations,
4402 or zero to disable CPU-hotplug testing.
4404 rcutorture.read_exit= [KNL]
4405 Set the number of read-then-exit kthreads used
4406 to test the interaction of RCU updaters and
4407 task-exit processing.
4409 rcutorture.read_exit_burst= [KNL]
4410 The number of times in a given read-then-exit
4411 episode that a set of read-then-exit kthreads
4414 rcutorture.read_exit_delay= [KNL]
4415 The delay, in seconds, between successive
4416 read-then-exit testing episodes.
4418 rcutorture.shuffle_interval= [KNL]
4419 Set task-shuffle interval (s). Shuffling tasks
4420 allows some CPUs to go into dyntick-idle mode
4421 during the rcutorture test.
4423 rcutorture.shutdown_secs= [KNL]
4424 Set time (s) after boot system shutdown. This
4425 is useful for hands-off automated testing.
4427 rcutorture.stall_cpu= [KNL]
4428 Duration of CPU stall (s) to test RCU CPU stall
4429 warnings, zero to disable.
4431 rcutorture.stall_cpu_block= [KNL]
4432 Sleep while stalling if set. This will result
4433 in warnings from preemptible RCU in addition
4434 to any other stall-related activity.
4436 rcutorture.stall_cpu_holdoff= [KNL]
4437 Time to wait (s) after boot before inducing stall.
4439 rcutorture.stall_cpu_irqsoff= [KNL]
4440 Disable interrupts while stalling if set.
4442 rcutorture.stall_gp_kthread= [KNL]
4443 Duration (s) of forced sleep within RCU
4444 grace-period kthread to test RCU CPU stall
4445 warnings, zero to disable. If both stall_cpu
4446 and stall_gp_kthread are specified, the
4447 kthread is starved first, then the CPU.
4449 rcutorture.stat_interval= [KNL]
4450 Time (s) between statistics printk()s.
4452 rcutorture.stutter= [KNL]
4453 Time (s) to stutter testing, for example, specifying
4454 five seconds causes the test to run for five seconds,
4455 wait for five seconds, and so on. This tests RCU's
4456 ability to transition abruptly to and from idle.
4458 rcutorture.test_boost= [KNL]
4459 Test RCU priority boosting? 0=no, 1=maybe, 2=yes.
4460 "Maybe" means test if the RCU implementation
4461 under test support RCU priority boosting.
4463 rcutorture.test_boost_duration= [KNL]
4464 Duration (s) of each individual boost test.
4466 rcutorture.test_boost_interval= [KNL]
4467 Interval (s) between each boost test.
4469 rcutorture.test_no_idle_hz= [KNL]
4470 Test RCU's dyntick-idle handling. See also the
4471 rcutorture.shuffle_interval parameter.
4473 rcutorture.torture_type= [KNL]
4474 Specify the RCU implementation to test.
4476 rcutorture.verbose= [KNL]
4477 Enable additional printk() statements.
4479 rcupdate.rcu_cpu_stall_ftrace_dump= [KNL]
4480 Dump ftrace buffer after reporting RCU CPU
4483 rcupdate.rcu_cpu_stall_suppress= [KNL]
4484 Suppress RCU CPU stall warning messages.
4486 rcupdate.rcu_cpu_stall_suppress_at_boot= [KNL]
4487 Suppress RCU CPU stall warning messages and
4488 rcutorture writer stall warnings that occur
4489 during early boot, that is, during the time
4490 before the init task is spawned.
4492 rcupdate.rcu_cpu_stall_timeout= [KNL]
4493 Set timeout for RCU CPU stall warning messages.
4495 rcupdate.rcu_expedited= [KNL]
4496 Use expedited grace-period primitives, for
4497 example, synchronize_rcu_expedited() instead
4498 of synchronize_rcu(). This reduces latency,
4499 but can increase CPU utilization, degrade
4500 real-time latency, and degrade energy efficiency.
4501 No effect on CONFIG_TINY_RCU kernels.
4503 rcupdate.rcu_normal= [KNL]
4504 Use only normal grace-period primitives,
4505 for example, synchronize_rcu() instead of
4506 synchronize_rcu_expedited(). This improves
4507 real-time latency, CPU utilization, and
4508 energy efficiency, but can expose users to
4509 increased grace-period latency. This parameter
4510 overrides rcupdate.rcu_expedited. No effect on
4511 CONFIG_TINY_RCU kernels.
4513 rcupdate.rcu_normal_after_boot= [KNL]
4514 Once boot has completed (that is, after
4515 rcu_end_inkernel_boot() has been invoked), use
4516 only normal grace-period primitives. No effect
4517 on CONFIG_TINY_RCU kernels.
4519 But note that CONFIG_PREEMPT_RT=y kernels enables
4520 this kernel boot parameter, forcibly setting
4521 it to the value one, that is, converting any
4522 post-boot attempt at an expedited RCU grace
4523 period to instead use normal non-expedited
4524 grace-period processing.
4526 rcupdate.rcu_task_ipi_delay= [KNL]
4527 Set time in jiffies during which RCU tasks will
4528 avoid sending IPIs, starting with the beginning
4529 of a given grace period. Setting a large
4530 number avoids disturbing real-time workloads,
4531 but lengthens grace periods.
4533 rcupdate.rcu_task_stall_timeout= [KNL]
4534 Set timeout in jiffies for RCU task stall warning
4535 messages. Disable with a value less than or equal
4538 rcupdate.rcu_self_test= [KNL]
4539 Run the RCU early boot self tests
4543 Run specified binary instead of /init from the ramdisk,
4544 used for early userspace startup. See initrd.
4547 force - Override the decision by the kernel to hide the
4548 advertisement of RDRAND support (this affects
4549 certain AMD processors because of buggy BIOS
4550 support, specifically around the suspend/resume
4554 Turn on/off individual RDT features. List is:
4555 cmt, mbmtotal, mbmlocal, l3cat, l3cdp, l2cat, l2cdp,
4557 E.g. to turn on cmt and turn off mba use:
4561 Format (x86 or x86_64):
4562 [w[arm] | c[old] | h[ard] | s[oft] | g[pio]] \
4564 [[,]b[ios] | a[cpi] | k[bd] | t[riple] | e[fi] | p[ci]] \
4566 Where reboot_mode is one of warm (soft) or cold (hard) or gpio
4567 (prefix with 'panic_' to set mode for panic
4569 reboot_type is one of bios, acpi, kbd, triple, efi, or pci,
4570 reboot_force is either force or not specified,
4571 reboot_cpu is s[mp]#### with #### being the processor
4572 to be used for rebooting.
4574 refscale.holdoff= [KNL]
4575 Set test-start holdoff period. The purpose of
4576 this parameter is to delay the start of the
4577 test until boot completes in order to avoid
4580 refscale.loops= [KNL]
4581 Set the number of loops over the synchronization
4582 primitive under test. Increasing this number
4583 reduces noise due to loop start/end overhead,
4584 but the default has already reduced the per-pass
4585 noise to a handful of picoseconds on ca. 2020
4588 refscale.nreaders= [KNL]
4589 Set number of readers. The default value of -1
4590 selects N, where N is roughly 75% of the number
4591 of CPUs. A value of zero is an interesting choice.
4593 refscale.nruns= [KNL]
4594 Set number of runs, each of which is dumped onto
4597 refscale.readdelay= [KNL]
4598 Set the read-side critical-section duration,
4599 measured in microseconds.
4601 refscale.scale_type= [KNL]
4602 Specify the read-protection implementation to test.
4604 refscale.shutdown= [KNL]
4605 Shut down the system at the end of the performance
4606 test. This defaults to 1 (shut it down) when
4607 refscale is built into the kernel and to 0 (leave
4608 it running) when refscale is built as a module.
4610 refscale.verbose= [KNL]
4611 Enable additional printk() statements.
4613 refscale.verbose_batched= [KNL]
4614 Batch the additional printk() statements. If zero
4615 (the default) or negative, print everything. Otherwise,
4616 print every Nth verbose statement, where N is the value
4620 [KNL, SMP] Set scheduler's default relax_domain_level.
4621 See Documentation/admin-guide/cgroup-v1/cpusets.rst.
4623 reserve= [KNL,BUGS] Force kernel to ignore I/O ports or memory
4624 Format: <base1>,<size1>[,<base2>,<size2>,...]
4625 Reserve I/O ports or memory so the kernel won't use
4626 them. If <base> is less than 0x10000, the region
4627 is assumed to be I/O ports; otherwise it is memory.
4629 reservetop= [X86-32]
4631 Reserves a hole at the top of the kernel virtual
4636 Set the amount of memory to reserve for BIOS at
4637 the bottom of the address space.
4639 reset_devices [KNL] Force drivers to reset the underlying device
4640 during initialization.
4643 Specify the partition device for software suspend
4645 {/dev/<dev> | PARTUUID=<uuid> | <int>:<int> | <hex>}
4647 resume_offset= [SWSUSP]
4648 Specify the offset from the beginning of the partition
4649 given by "resume=" at which the swap header is located,
4650 in <PAGE_SIZE> units (needed only for swap files).
4651 See Documentation/power/swsusp-and-swap-files.rst
4653 resumedelay= [HIBERNATION] Delay (in seconds) to pause before attempting to
4654 read the resume files
4656 resumewait [HIBERNATION] Wait (indefinitely) for resume device to show up.
4657 Useful for devices that are detected asynchronously
4658 (e.g. USB and MMC devices).
4660 hibernate= [HIBERNATION]
4661 noresume Don't check if there's a hibernation image
4662 present during boot.
4663 nocompress Don't compress/decompress hibernation images.
4664 no Disable hibernation and resume.
4665 protect_image Turn on image protection during restoration
4666 (that will set all pages holding image data
4667 during restoration read-only).
4669 retain_initrd [RAM] Keep initrd memory after extraction
4671 rfkill.default_state=
4672 0 "airplane mode". All wifi, bluetooth, wimax, gps, fm,
4673 etc. communication is blocked by default.
4676 rfkill.master_switch_mode=
4677 0 The "airplane mode" button does nothing.
4678 1 The "airplane mode" button toggles between everything
4679 blocked and the previous configuration.
4680 2 The "airplane mode" button toggles between everything
4681 blocked and everything unblocked.
4683 rhash_entries= [KNL,NET]
4684 Set number of hash buckets for route cache
4687 [KNL] Disable ring 3 MONITOR/MWAIT feature on supported
4690 ro [KNL] Mount root device read-only on boot
4693 on Mark read-only kernel memory as read-only (default).
4694 off Leave read-only kernel memory writable for debugging.
4697 Enable the uart passthrough on the designated usb port
4698 on Rockchip SoCs. When active, the signals of the
4699 debug-uart get routed to the D+ and D- pins of the usb
4700 port and the regular usb controller gets disabled.
4702 root= [KNL] Root filesystem
4703 See name_to_dev_t comment in init/do_mounts.c.
4705 rootdelay= [KNL] Delay (in seconds) to pause before attempting to
4706 mount the root filesystem
4708 rootflags= [KNL] Set root filesystem mount option string
4710 rootfstype= [KNL] Set root filesystem type
4712 rootwait [KNL] Wait (indefinitely) for root device to show up.
4713 Useful for devices that are detected asynchronously
4714 (e.g. USB and MMC devices).
4716 rproc_mem=nn[KMG][@address]
4717 [KNL,ARM,CMA] Remoteproc physical memory block.
4718 Memory area to be used by remote processor image,
4721 rw [KNL] Mount root device read-write on boot
4723 S [KNL] Run init in single mode
4725 s390_iommu= [HW,S390]
4726 Set s390 IOTLB flushing mode
4728 With strict flushing every unmap operation will result in
4729 an IOTLB flush. Default is lazy flushing before reuse,
4733 See drivers/net/irda/sa1100_ir.c.
4735 sbni= [NET] Granch SBNI12 leased line adapter
4737 sched_debug [KNL] Enables verbose scheduler debug messages.
4739 schedstats= [KNL,X86] Enable or disable scheduled statistics.
4740 Allowed values are enable and disable. This feature
4741 incurs a small amount of overhead in the scheduler
4742 but is useful for debugging and performance tuning.
4744 sched_thermal_decay_shift=
4745 [KNL, SMP] Set a decay shift for scheduler thermal
4746 pressure signal. Thermal pressure signal follows the
4747 default decay period of other scheduler pelt
4748 signals(usually 32 ms but configurable). Setting
4749 sched_thermal_decay_shift will left shift the decay
4750 period for the thermal pressure signal by the shift
4752 i.e. with the default pelt decay period of 32 ms
4753 sched_thermal_decay_shift thermal pressure decay pr
4757 Format: integer between 0 and 10
4760 scftorture.holdoff= [KNL]
4761 Number of seconds to hold off before starting
4762 test. Defaults to zero for module insertion and
4763 to 10 seconds for built-in smp_call_function()
4766 scftorture.longwait= [KNL]
4767 Request ridiculously long waits randomly selected
4768 up to the chosen limit in seconds. Zero (the
4769 default) disables this feature. Please note
4770 that requesting even small non-zero numbers of
4771 seconds can result in RCU CPU stall warnings,
4772 softlockup complaints, and so on.
4774 scftorture.nthreads= [KNL]
4775 Number of kthreads to spawn to invoke the
4776 smp_call_function() family of functions.
4777 The default of -1 specifies a number of kthreads
4778 equal to the number of CPUs.
4780 scftorture.onoff_holdoff= [KNL]
4781 Number seconds to wait after the start of the
4782 test before initiating CPU-hotplug operations.
4784 scftorture.onoff_interval= [KNL]
4785 Number seconds to wait between successive
4786 CPU-hotplug operations. Specifying zero (which
4787 is the default) disables CPU-hotplug operations.
4789 scftorture.shutdown_secs= [KNL]
4790 The number of seconds following the start of the
4791 test after which to shut down the system. The
4792 default of zero avoids shutting down the system.
4793 Non-zero values are useful for automated tests.
4795 scftorture.stat_interval= [KNL]
4796 The number of seconds between outputting the
4797 current test statistics to the console. A value
4798 of zero disables statistics output.
4800 scftorture.stutter_cpus= [KNL]
4801 The number of jiffies to wait between each change
4802 to the set of CPUs under test.
4804 scftorture.use_cpus_read_lock= [KNL]
4805 Use use_cpus_read_lock() instead of the default
4806 preempt_disable() to disable CPU hotplug
4807 while invoking one of the smp_call_function*()
4810 scftorture.verbose= [KNL]
4811 Enable additional printk() statements.
4813 scftorture.weight_single= [KNL]
4814 The probability weighting to use for the
4815 smp_call_function_single() function with a zero
4816 "wait" parameter. A value of -1 selects the
4817 default if all other weights are -1. However,
4818 if at least one weight has some other value, a
4819 value of -1 will instead select a weight of zero.
4821 scftorture.weight_single_wait= [KNL]
4822 The probability weighting to use for the
4823 smp_call_function_single() function with a
4824 non-zero "wait" parameter. See weight_single.
4826 scftorture.weight_many= [KNL]
4827 The probability weighting to use for the
4828 smp_call_function_many() function with a zero
4829 "wait" parameter. See weight_single.
4830 Note well that setting a high probability for
4831 this weighting can place serious IPI load
4834 scftorture.weight_many_wait= [KNL]
4835 The probability weighting to use for the
4836 smp_call_function_many() function with a
4837 non-zero "wait" parameter. See weight_single
4840 scftorture.weight_all= [KNL]
4841 The probability weighting to use for the
4842 smp_call_function_all() function with a zero
4843 "wait" parameter. See weight_single and
4846 scftorture.weight_all_wait= [KNL]
4847 The probability weighting to use for the
4848 smp_call_function_all() function with a
4849 non-zero "wait" parameter. See weight_single
4852 skew_tick= [KNL] Offset the periodic timer tick per cpu to mitigate
4853 xtime_lock contention on larger systems, and/or RCU lock
4854 contention on all systems with CONFIG_MAXSMP set.
4855 Format: { "0" | "1" }
4856 0 -- disable. (may be 1 via CONFIG_CMDLINE="skew_tick=1"
4858 Note: increases power consumption, thus should only be
4859 enabled if running jitter sensitive (HPC/RT) workloads.
4861 security= [SECURITY] Choose a legacy "major" security module to
4862 enable at boot. This has been deprecated by the
4865 selinux= [SELINUX] Disable or enable SELinux at boot time.
4866 Format: { "0" | "1" }
4867 See security/selinux/Kconfig help text.
4872 apparmor= [APPARMOR] Disable or enable AppArmor at boot time
4873 Format: { "0" | "1" }
4874 See security/apparmor/Kconfig help text
4877 Default value is set via kernel config option.
4879 serialnumber [BUGS=X86-32]
4882 Maximal number of shapers.
4890 Disable merging of slabs with similar size. May be
4891 necessary if there is some reason to distinguish
4892 allocs to different slabs, especially in hardened
4893 environments where the risk of heap overflows and
4894 layout control by attackers can usually be
4895 frustrated by disabling merging. This will reduce
4896 most of the exposure of a heap attack to a single
4897 cache (risks via metadata attacks are mostly
4898 unchanged). Debug options disable merging on their
4900 For more information see Documentation/vm/slub.rst.
4902 slab_max_order= [MM, SLAB]
4903 Determines the maximum allowed order for slabs.
4904 A high setting may cause OOMs due to memory
4905 fragmentation. Defaults to 1 for systems with
4906 more than 32MB of RAM, 0 otherwise.
4908 slub_debug[=options[,slabs][;[options[,slabs]]...] [MM, SLUB]
4909 Enabling slub_debug allows one to determine the
4910 culprit if slab objects become corrupted. Enabling
4911 slub_debug can create guard zones around objects and
4912 may poison objects when not in use. Also tracks the
4913 last alloc / free. For more information see
4914 Documentation/vm/slub.rst.
4916 slub_max_order= [MM, SLUB]
4917 Determines the maximum allowed order for slabs.
4918 A high setting may cause OOMs due to memory
4919 fragmentation. For more information see
4920 Documentation/vm/slub.rst.
4922 slub_min_objects= [MM, SLUB]
4923 The minimum number of objects per slab. SLUB will
4924 increase the slab order up to slub_max_order to
4925 generate a sufficiently large slab able to contain
4926 the number of objects indicated. The higher the number
4927 of objects the smaller the overhead of tracking slabs
4928 and the less frequently locks need to be acquired.
4929 For more information see Documentation/vm/slub.rst.
4931 slub_min_order= [MM, SLUB]
4932 Determines the minimum page order for slabs. Must be
4933 lower than slub_max_order.
4934 For more information see Documentation/vm/slub.rst.
4936 slub_nomerge [MM, SLUB]
4937 Same with slab_nomerge. This is supported for legacy.
4938 See slab_nomerge for more information.
4941 Format: <io1>[,<io2>[,...,<io8>]]
4943 smsc-ircc2.nopnp [HW] Don't use PNP to discover SMC devices
4944 smsc-ircc2.ircc_cfg= [HW] Device configuration I/O port
4945 smsc-ircc2.ircc_sir= [HW] SIR base I/O port
4946 smsc-ircc2.ircc_fir= [HW] FIR base I/O port
4947 smsc-ircc2.ircc_irq= [HW] IRQ line
4948 smsc-ircc2.ircc_dma= [HW] DMA channel
4949 smsc-ircc2.ircc_transceiver= [HW] Transceiver type:
4950 0: Toshiba Satellite 1800 (GP data pin select)
4951 1: Fast pin select (default)
4954 smt [KNL,S390] Set the maximum number of threads (logical
4955 CPUs) to use per physical CPU on systems capable of
4956 symmetric multithreading (SMT). Will be capped to the
4957 actual hardware limit.
4959 Default: -1 (no limit)
4962 [KNL] Should the soft-lockup detector generate panics.
4965 A value of 1 instructs the soft-lockup detector
4966 to panic the machine when a soft-lockup occurs. It is
4967 also controlled by the kernel.softlockup_panic sysctl
4968 and CONFIG_BOOTPARAM_SOFTLOCKUP_PANIC, which is the
4969 respective build-time switch to that functionality.
4971 softlockup_all_cpu_backtrace=
4972 [KNL] Should the soft-lockup detector generate
4973 backtraces on all cpus.
4976 sonypi.*= [HW] Sony Programmable I/O Control Device driver
4977 See Documentation/admin-guide/laptops/sonypi.rst
4979 spectre_v2= [X86] Control mitigation of Spectre variant 2
4980 (indirect branch speculation) vulnerability.
4981 The default operation protects the kernel from
4984 on - unconditionally enable, implies
4986 off - unconditionally disable, implies
4988 auto - kernel detects whether your CPU model is
4991 Selecting 'on' will, and 'auto' may, choose a
4992 mitigation method at run time according to the
4993 CPU, the available microcode, the setting of the
4994 CONFIG_RETPOLINE configuration option, and the
4995 compiler with which the kernel was built.
4997 Selecting 'on' will also enable the mitigation
4998 against user space to user space task attacks.
5000 Selecting 'off' will disable both the kernel and
5001 the user space protections.
5003 Specific mitigations can also be selected manually:
5005 retpoline - replace indirect branches
5006 retpoline,generic - google's original retpoline
5007 retpoline,amd - AMD-specific minimal thunk
5009 Not specifying this option is equivalent to
5013 [X86] Control mitigation of Spectre variant 2
5014 (indirect branch speculation) vulnerability between
5017 on - Unconditionally enable mitigations. Is
5018 enforced by spectre_v2=on
5020 off - Unconditionally disable mitigations. Is
5021 enforced by spectre_v2=off
5023 prctl - Indirect branch speculation is enabled,
5024 but mitigation can be enabled via prctl
5025 per thread. The mitigation control state
5026 is inherited on fork.
5029 - Like "prctl" above, but only STIBP is
5030 controlled per thread. IBPB is issued
5031 always when switching between different user
5035 - Same as "prctl" above, but all seccomp
5036 threads will enable the mitigation unless
5037 they explicitly opt out.
5040 - Like "seccomp" above, but only STIBP is
5041 controlled per thread. IBPB is issued
5042 always when switching between different
5043 user space processes.
5045 auto - Kernel selects the mitigation depending on
5046 the available CPU features and vulnerability.
5049 If CONFIG_SECCOMP=y then "seccomp", otherwise "prctl"
5051 Not specifying this option is equivalent to
5052 spectre_v2_user=auto.
5054 spec_store_bypass_disable=
5055 [HW] Control Speculative Store Bypass (SSB) Disable mitigation
5056 (Speculative Store Bypass vulnerability)
5058 Certain CPUs are vulnerable to an exploit against a
5059 a common industry wide performance optimization known
5060 as "Speculative Store Bypass" in which recent stores
5061 to the same memory location may not be observed by
5062 later loads during speculative execution. The idea
5063 is that such stores are unlikely and that they can
5064 be detected prior to instruction retirement at the
5065 end of a particular speculation execution window.
5067 In vulnerable processors, the speculatively forwarded
5068 store can be used in a cache side channel attack, for
5069 example to read memory to which the attacker does not
5070 directly have access (e.g. inside sandboxed code).
5072 This parameter controls whether the Speculative Store
5073 Bypass optimization is used.
5075 On x86 the options are:
5077 on - Unconditionally disable Speculative Store Bypass
5078 off - Unconditionally enable Speculative Store Bypass
5079 auto - Kernel detects whether the CPU model contains an
5080 implementation of Speculative Store Bypass and
5081 picks the most appropriate mitigation. If the
5082 CPU is not vulnerable, "off" is selected. If the
5083 CPU is vulnerable the default mitigation is
5084 architecture and Kconfig dependent. See below.
5085 prctl - Control Speculative Store Bypass per thread
5086 via prctl. Speculative Store Bypass is enabled
5087 for a process by default. The state of the control
5088 is inherited on fork.
5089 seccomp - Same as "prctl" above, but all seccomp threads
5090 will disable SSB unless they explicitly opt out.
5092 Default mitigations:
5093 X86: If CONFIG_SECCOMP=y "seccomp", otherwise "prctl"
5095 On powerpc the options are:
5097 on,auto - On Power8 and Power9 insert a store-forwarding
5098 barrier on kernel entry and exit. On Power7
5099 perform a software flush on kernel entry and
5103 Not specifying this option is equivalent to
5104 spec_store_bypass_disable=auto.
5106 spia_io_base= [HW,MTD]
5112 [X86] Enable split lock detection or bus lock detection
5114 When enabled (and if hardware support is present), atomic
5115 instructions that access data across cache line
5116 boundaries will result in an alignment check exception
5117 for split lock detection or a debug exception for
5122 warn - the kernel will emit rate-limited warnings
5123 about applications triggering the #AC
5124 exception or the #DB exception. This mode is
5125 the default on CPUs that support split lock
5126 detection or bus lock detection. Default
5127 behavior is by #AC if both features are
5128 enabled in hardware.
5130 fatal - the kernel will send SIGBUS to applications
5131 that trigger the #AC exception or the #DB
5132 exception. Default behavior is by #AC if
5133 both features are enabled in hardware.
5135 If an #AC exception is hit in the kernel or in
5136 firmware (i.e. not while executing in user mode)
5137 the kernel will oops in either "warn" or "fatal"
5140 #DB exception for bus lock is triggered only when
5144 Control the Special Register Buffer Data Sampling
5147 Certain CPUs are vulnerable to an MDS-like
5148 exploit which can leak bits from the random
5151 By default, this issue is mitigated by
5152 microcode. However, the microcode fix can cause
5153 the RDRAND and RDSEED instructions to become
5154 much slower. Among other effects, this will
5155 result in reduced throughput from /dev/urandom.
5157 The microcode mitigation can be disabled with
5158 the following option:
5160 off: Disable mitigation and remove
5161 performance impact to RDRAND and RDSEED
5163 srcutree.counter_wrap_check [KNL]
5164 Specifies how frequently to check for
5165 grace-period sequence counter wrap for the
5166 srcu_data structure's ->srcu_gp_seq_needed field.
5167 The greater the number of bits set in this kernel
5168 parameter, the less frequently counter wrap will
5169 be checked for. Note that the bottom two bits
5172 srcutree.exp_holdoff [KNL]
5173 Specifies how many nanoseconds must elapse
5174 since the end of the last SRCU grace period for
5175 a given srcu_struct until the next normal SRCU
5176 grace period will be considered for automatic
5177 expediting. Set to zero to disable automatic
5181 Speculative Store Bypass Disable control
5183 On CPUs that are vulnerable to the Speculative
5184 Store Bypass vulnerability and offer a
5185 firmware based mitigation, this parameter
5186 indicates how the mitigation should be used:
5188 force-on: Unconditionally enable mitigation for
5189 for both kernel and userspace
5190 force-off: Unconditionally disable mitigation for
5191 for both kernel and userspace
5192 kernel: Always enable mitigation in the
5193 kernel, and offer a prctl interface
5194 to allow userspace to register its
5195 interest in being mitigated too.
5197 stack_guard_gap= [MM]
5198 override the default stack gap protection. The value
5199 is in page units and it defines how many pages prior
5200 to (for stacks growing down) resp. after (for stacks
5201 growing up) the main stack are reserved for no other
5202 mapping. Default value is 256 pages.
5204 stack_depot_disable= [KNL]
5205 Setting this to true through kernel command line will
5206 disable the stack depot thereby saving the static memory
5207 consumed by the stack hash table. By default this is set
5211 Enabled the stack tracer on boot up.
5213 stacktrace_filter=[function-list]
5214 [FTRACE] Limit the functions that the stack tracer
5215 will trace at boot up. function-list is a comma-separated
5216 list of functions. This list can be changed at run
5217 time by the stack_trace_filter file in the debugfs
5218 tracing directory. Note, this enables stack tracing
5219 and the stacktrace above is not needed.
5223 Set the STI (builtin display/keyboard on the HP-PARISC
5224 machines) console (graphic card) which should be used
5225 as the initial boot-console.
5226 See also comment in drivers/video/console/sticore.c.
5229 See comment in drivers/video/console/sticore.c.
5232 Format: bpp:<bpp1>[:<bpp2>[:<bpp3>...]]
5234 sunrpc.min_resvport=
5235 sunrpc.max_resvport=
5237 SunRPC servers often require that client requests
5238 originate from a privileged port (i.e. a port in the
5239 range 0 < portnr < 1024).
5240 An administrator who wishes to reserve some of these
5241 ports for other uses may adjust the range that the
5242 kernel's sunrpc client considers to be privileged
5243 using these two parameters to set the minimum and
5244 maximum port values.
5246 sunrpc.svc_rpc_per_connection_limit=
5248 Limit the number of requests that the server will
5249 process in parallel from a single connection.
5250 The default value is 0 (no limit).
5254 Control how the NFS server code allocates CPUs to
5255 service thread pools. Depending on how many NICs
5256 you have and where their interrupts are bound, this
5257 option will affect which CPUs will do NFS serving.
5258 Note: this parameter cannot be changed while the
5259 NFS server is running.
5261 auto the server chooses an appropriate mode
5262 automatically using heuristics
5263 global a single global pool contains all CPUs
5264 percpu one pool for each CPU
5265 pernode one pool for each NUMA node (equivalent
5266 to global on non-NUMA machines)
5268 sunrpc.tcp_slot_table_entries=
5269 sunrpc.udp_slot_table_entries=
5271 Sets the upper limit on the number of simultaneous
5272 RPC calls that can be sent from the client to a
5273 server. Increasing these values may allow you to
5274 improve throughput, but will also increase the
5275 amount of memory reserved for use by the client.
5277 suspend.pm_test_delay=
5279 Sets the number of seconds to remain in a suspend test
5280 mode before resuming the system (see
5281 /sys/power/pm_test). Only available when CONFIG_PM_DEBUG
5282 is set. Default value is 5.
5285 Format: { on | off | y | n | 1 | 0 }
5286 This parameter controls use of the Protected
5287 Execution Facility on pSeries.
5290 [KNL] Enable accounting of swap in memory resource
5291 controller if no parameter or 1 is given or disable
5292 it if 0 is given (See Documentation/admin-guide/cgroup-v1/memory.rst)
5294 swiotlb= [ARM,IA-64,PPC,MIPS,X86]
5295 Format: { <int> | force | noforce }
5296 <int> -- Number of I/O TLB slabs
5297 force -- force using of bounce buffers even if they
5298 wouldn't be automatically used by the kernel
5299 noforce -- Never use bounce buffers (for debugging)
5304 Set a sysctl parameter, right before loading the init
5305 process, as if the value was written to the respective
5306 /proc/sys/... file. Both '.' and '/' are recognized as
5307 separators. Unrecognized parameters and invalid values
5308 are reported in the kernel log. Sysctls registered
5309 later by a loaded module cannot be set this way.
5310 Example: sysctl.vm.swappiness=40
5312 sysfs.deprecated=0|1 [KNL]
5313 Enable/disable old style sysfs layout for old udev
5314 on older distributions. When this option is enabled
5315 very new udev will not work anymore. When this option
5316 is disabled (or CONFIG_SYSFS_DEPRECATED not compiled)
5317 in older udev will not work anymore.
5318 Default depends on CONFIG_SYSFS_DEPRECATED_V2 set in
5319 the kernel configuration.
5321 sysrq_always_enabled
5323 Ignore sysrq setting - this boot parameter will
5324 neutralize any effect of /proc/sys/kernel/sysrq.
5325 Useful for debugging.
5327 tcpmhash_entries= [KNL,NET]
5328 Set the number of tcp_metrics_hash slots.
5329 Default value is 8192 or 16384 depending on total
5330 ram pages. This is used to specify the TCP metrics
5331 cache size. See Documentation/networking/ip-sysctl.rst
5332 "tcp_no_metrics_save" section for more details.
5336 test_suspend= [SUSPEND][,N]
5337 Specify "mem" (for Suspend-to-RAM) or "standby" (for
5338 standby suspend) or "freeze" (for suspend type freeze)
5339 as the system sleep state during system startup with
5340 the optional capability to repeat N number of times.
5341 The system is woken from this state using a
5342 wakeup-capable RTC alarm.
5344 thash_entries= [KNL,NET]
5345 Set number of hash buckets for TCP connection
5347 thermal.act= [HW,ACPI]
5348 -1: disable all active trip points in all thermal zones
5349 <degrees C>: override all lowest active trip points
5351 thermal.crt= [HW,ACPI]
5352 -1: disable all critical trip points in all thermal zones
5353 <degrees C>: override all critical trip points
5355 thermal.nocrt= [HW,ACPI]
5356 Set to disable actions on ACPI thermal zone
5357 critical and hot trip points.
5359 thermal.off= [HW,ACPI]
5360 1: disable ACPI thermal control
5362 thermal.psv= [HW,ACPI]
5363 -1: disable all passive trip points
5364 <degrees C>: override all passive trip points to this
5367 thermal.tzp= [HW,ACPI]
5368 Specify global default ACPI thermal zone polling rate
5369 <deci-seconds>: poll all this frequency
5370 0: no polling (default)
5373 Force threading of all interrupt handlers except those
5374 marked explicitly IRQF_NO_THREAD.
5378 Specify if the kernel should make use of the cpu
5379 topology information if the hardware supports this.
5380 The scheduler will make use of this information and
5381 e.g. base its process migration decisions on it.
5384 topology_updates= [KNL, PPC, NUMA]
5386 Specify if the kernel should ignore (off)
5387 topology updates sent by the hypervisor to this
5390 torture.disable_onoff_at_boot= [KNL]
5391 Prevent the CPU-hotplug component of torturing
5392 until after init has spawned.
5394 torture.ftrace_dump_at_shutdown= [KNL]
5395 Dump the ftrace buffer at torture-test shutdown,
5396 even if there were no errors. This can be a
5397 very costly operation when many torture tests
5398 are running concurrently, especially on systems
5399 with rotating-rust storage.
5401 torture.verbose_sleep_frequency= [KNL]
5402 Specifies how many verbose printk()s should be
5403 emitted between each sleep. The default of zero
5404 disables verbose-printk() sleeping.
5406 torture.verbose_sleep_duration= [KNL]
5407 Duration of each verbose-printk() sleep in jiffies.
5411 tpm_suspend_pcr=[HW,TPM]
5412 Format: integer pcr id
5413 Specify that at suspend time, the tpm driver
5414 should extend the specified pcr with zeros,
5415 as a workaround for some chips which fail to
5416 flush the last written pcr on TPM_SaveState.
5417 This will guarantee that all the other pcrs
5420 trace_buf_size=nn[KMG]
5421 [FTRACE] will set tracing buffer size on each cpu.
5423 trace_event=[event-list]
5424 [FTRACE] Set and start specified trace events in order
5425 to facilitate early boot debugging. The event-list is a
5426 comma-separated list of trace events to enable. See
5427 also Documentation/trace/events.rst
5429 trace_options=[option-list]
5430 [FTRACE] Enable or disable tracer options at boot.
5431 The option-list is a comma delimited list of options
5432 that can be enabled or disabled just as if you were
5433 to echo the option name into
5435 /sys/kernel/debug/tracing/trace_options
5437 For example, to enable stacktrace option (to dump the
5438 stack trace of each event), add to the command line:
5440 trace_options=stacktrace
5442 See also Documentation/trace/ftrace.rst "trace options"
5446 Have the tracepoints sent to printk as well as the
5447 tracing ring buffer. This is useful for early boot up
5448 where the system hangs or reboots and does not give the
5449 option for reading the tracing buffer or performing a
5450 ftrace_dump_on_oops.
5452 To turn off having tracepoints sent to printk,
5453 echo 0 > /proc/sys/kernel/tracepoint_printk
5454 Note, echoing 1 into this file without the
5455 tracepoint_printk kernel cmdline option has no effect.
5459 Having tracepoints sent to printk() and activating high
5460 frequency tracepoints such as irq or sched, can cause
5461 the system to live lock.
5464 [FTRACE] enable this option to disable tracing when a
5465 warning is hit. This turns off "tracing_on". Tracing can
5466 be enabled again by echoing '1' into the "tracing_on"
5467 file located in /sys/kernel/debug/tracing/
5469 This option is useful, as it disables the trace before
5470 the WARNING dump is called, which prevents the trace to
5471 be filled with content caused by the warning output.
5473 This option can also be set at run time via the sysctl
5474 option: kernel/traceoff_on_warning
5476 transparent_hugepage=
5478 Format: [always|madvise|never]
5479 Can be used to control the default behavior of the system
5480 with respect to transparent hugepages.
5481 See Documentation/admin-guide/mm/transhuge.rst
5484 trusted.source= [KEYS]
5486 This parameter identifies the trust source as a backend
5487 for trusted keys implementation. Supported trust
5491 If not specified then it defaults to iterating through
5492 the trust source list starting with TPM and assigns the
5493 first trust source as a backend which is initialized
5494 successfully during iteration.
5496 tsc= Disable clocksource stability checks for TSC.
5498 [x86] reliable: mark tsc clocksource as reliable, this
5499 disables clocksource verification at runtime, as well
5500 as the stability checks done at bootup. Used to enable
5501 high-resolution timer mode on older hardware, and in
5502 virtualized environment.
5503 [x86] noirqtime: Do not use TSC to do irq accounting.
5504 Used to run time disable IRQ_TIME_ACCOUNTING on any
5505 platforms where RDTSC is slow and this accounting
5507 [x86] unstable: mark the TSC clocksource as unstable, this
5508 marks the TSC unconditionally unstable at bootup and
5509 avoids any further wobbles once the TSC watchdog notices.
5510 [x86] nowatchdog: disable clocksource watchdog. Used
5511 in situations with strict latency requirements (where
5512 interruptions from clocksource watchdog are not
5515 tsc_early_khz= [X86] Skip early TSC calibration and use the given
5516 value instead. Useful when the early TSC frequency discovery
5517 procedure is not reliable, such as on overclocked systems
5518 with CPUID.16h support and partial CPUID.15h support.
5519 Format: <unsigned int>
5521 tsx= [X86] Control Transactional Synchronization
5522 Extensions (TSX) feature in Intel processors that
5523 support TSX control.
5525 This parameter controls the TSX feature. The options are:
5527 on - Enable TSX on the system. Although there are
5528 mitigations for all known security vulnerabilities,
5529 TSX has been known to be an accelerator for
5530 several previous speculation-related CVEs, and
5531 so there may be unknown security risks associated
5532 with leaving it enabled.
5534 off - Disable TSX on the system. (Note that this
5535 option takes effect only on newer CPUs which are
5536 not vulnerable to MDS, i.e., have
5537 MSR_IA32_ARCH_CAPABILITIES.MDS_NO=1 and which get
5538 the new IA32_TSX_CTRL MSR through a microcode
5539 update. This new MSR allows for the reliable
5540 deactivation of the TSX functionality.)
5542 auto - Disable TSX if X86_BUG_TAA is present,
5543 otherwise enable TSX on the system.
5545 Not specifying this option is equivalent to tsx=off.
5547 See Documentation/admin-guide/hw-vuln/tsx_async_abort.rst
5550 tsx_async_abort= [X86,INTEL] Control mitigation for the TSX Async
5551 Abort (TAA) vulnerability.
5553 Similar to Micro-architectural Data Sampling (MDS)
5554 certain CPUs that support Transactional
5555 Synchronization Extensions (TSX) are vulnerable to an
5556 exploit against CPU internal buffers which can forward
5557 information to a disclosure gadget under certain
5560 In vulnerable processors, the speculatively forwarded
5561 data can be used in a cache side channel attack, to
5562 access data to which the attacker does not have direct
5565 This parameter controls the TAA mitigation. The
5568 full - Enable TAA mitigation on vulnerable CPUs
5571 full,nosmt - Enable TAA mitigation and disable SMT on
5572 vulnerable CPUs. If TSX is disabled, SMT
5573 is not disabled because CPU is not
5574 vulnerable to cross-thread TAA attacks.
5575 off - Unconditionally disable TAA mitigation
5577 On MDS-affected machines, tsx_async_abort=off can be
5578 prevented by an active MDS mitigation as both vulnerabilities
5579 are mitigated with the same mechanism so in order to disable
5580 this mitigation, you need to specify mds=off too.
5582 Not specifying this option is equivalent to
5583 tsx_async_abort=full. On CPUs which are MDS affected
5584 and deploy MDS mitigation, TAA mitigation is not
5585 required and doesn't provide any additional
5589 Documentation/admin-guide/hw-vuln/tsx_async_abort.rst
5591 turbografx.map[2|3]= [HW,JOY]
5592 TurboGraFX parallel port interface
5594 <port#>,<js1>,<js2>,<js3>,<js4>,<js5>,<js6>,<js7>
5595 See also Documentation/input/devices/joystick-parport.rst
5597 udbg-immortal [PPC] When debugging early kernel crashes that
5598 happen after console_init() and before a proper
5599 console driver takes over, this boot options might
5600 help "seeing" what's going on.
5602 uhash_entries= [KNL,NET]
5603 Set number of hash buckets for UDP/UDP-Lite connections
5606 [USB] Ignore overcurrent events (default N).
5607 Some badly-designed motherboards generate lots of
5608 bogus events, for ports that aren't wired to
5609 anything. Set this parameter to avoid log spamming.
5610 Note that genuine overcurrent events won't be
5614 [X86] Cause panic on unknown NMI.
5616 usbcore.authorized_default=
5617 [USB] Default USB device authorization:
5618 (default -1 = authorized except for wireless USB,
5619 0 = not authorized, 1 = authorized, 2 = authorized
5620 if device connected to internal port)
5622 usbcore.autosuspend=
5623 [USB] The autosuspend time delay (in seconds) used
5624 for newly-detected USB devices (default 2). This
5625 is the time required before an idle device will be
5626 autosuspended. Devices for which the delay is set
5627 to a negative value won't be autosuspended at all.
5629 usbcore.usbfs_snoop=
5630 [USB] Set to log all usbfs traffic (default 0 = off).
5632 usbcore.usbfs_snoop_max=
5633 [USB] Maximum number of bytes to snoop in each URB
5636 usbcore.blinkenlights=
5637 [USB] Set to cycle leds on hubs (default 0 = off).
5639 usbcore.old_scheme_first=
5640 [USB] Start with the old device initialization
5641 scheme (default 0 = off).
5643 usbcore.usbfs_memory_mb=
5644 [USB] Memory limit (in MB) for buffers allocated by
5645 usbfs (default = 16, 0 = max = 2047).
5647 usbcore.use_both_schemes=
5648 [USB] Try the other device initialization scheme
5649 if the first one fails (default 1 = enabled).
5651 usbcore.initial_descriptor_timeout=
5652 [USB] Specifies timeout for the initial 64-byte
5653 USB_REQ_GET_DESCRIPTOR request in milliseconds
5654 (default 5000 = 5.0 seconds).
5656 usbcore.nousb [USB] Disable the USB subsystem
5659 [USB] A list of quirk entries to augment the built-in
5660 usb core quirk list. List entries are separated by
5661 commas. Each entry has the form
5662 VendorID:ProductID:Flags. The IDs are 4-digit hex
5663 numbers and Flags is a set of letters. Each letter
5664 will change the built-in quirk; setting it if it is
5665 clear and clearing it if it is set. The letters have
5666 the following meanings:
5667 a = USB_QUIRK_STRING_FETCH_255 (string
5668 descriptors must not be fetched using
5670 b = USB_QUIRK_RESET_RESUME (device can't resume
5671 correctly so reset it instead);
5672 c = USB_QUIRK_NO_SET_INTF (device can't handle
5673 Set-Interface requests);
5674 d = USB_QUIRK_CONFIG_INTF_STRINGS (device can't
5675 handle its Configuration or Interface
5677 e = USB_QUIRK_RESET (device can't be reset
5678 (e.g morph devices), don't use reset);
5679 f = USB_QUIRK_HONOR_BNUMINTERFACES (device has
5680 more interface descriptions than the
5681 bNumInterfaces count, and can't handle
5682 talking to these interfaces);
5683 g = USB_QUIRK_DELAY_INIT (device needs a pause
5684 during initialization, after we read
5685 the device descriptor);
5686 h = USB_QUIRK_LINEAR_UFRAME_INTR_BINTERVAL (For
5687 high speed and super speed interrupt
5688 endpoints, the USB 2.0 and USB 3.0 spec
5689 require the interval in microframes (1
5690 microframe = 125 microseconds) to be
5691 calculated as interval = 2 ^
5693 Devices with this quirk report their
5694 bInterval as the result of this
5695 calculation instead of the exponent
5696 variable used in the calculation);
5697 i = USB_QUIRK_DEVICE_QUALIFIER (device can't
5698 handle device_qualifier descriptor
5700 j = USB_QUIRK_IGNORE_REMOTE_WAKEUP (device
5701 generates spurious wakeup, ignore
5702 remote wakeup capability);
5703 k = USB_QUIRK_NO_LPM (device can't handle Link
5705 l = USB_QUIRK_LINEAR_FRAME_INTR_BINTERVAL
5706 (Device reports its bInterval as linear
5707 frames instead of the USB 2.0
5709 m = USB_QUIRK_DISCONNECT_SUSPEND (Device needs
5710 to be disconnected before suspend to
5711 prevent spurious wakeup);
5712 n = USB_QUIRK_DELAY_CTRL_MSG (Device needs a
5713 pause after every control message);
5714 o = USB_QUIRK_HUB_SLOW_RESET (Hub needs extra
5715 delay after resetting its port);
5716 Example: quirks=0781:5580:bk,0a5c:5834:gij
5719 [USBHID] The interval which mice are to be polled at.
5722 [USBHID] The interval which joysticks are to be polled at.
5725 [USBHID] The interval which keyboards are to be polled at.
5727 usb-storage.delay_use=
5728 [UMS] The delay in seconds before a new device is
5729 scanned for Logical Units (default 1).
5732 [UMS] A list of quirks entries to supplement or
5733 override the built-in unusual_devs list. List
5734 entries are separated by commas. Each entry has
5735 the form VID:PID:Flags where VID and PID are Vendor
5736 and Product ID values (4-digit hex numbers) and
5737 Flags is a set of characters, each corresponding
5738 to a common usb-storage quirk flag as follows:
5739 a = SANE_SENSE (collect more than 18 bytes
5740 of sense data, not on uas);
5741 b = BAD_SENSE (don't collect more than 18
5742 bytes of sense data, not on uas);
5743 c = FIX_CAPACITY (decrease the reported
5744 device capacity by one sector);
5745 d = NO_READ_DISC_INFO (don't use
5746 READ_DISC_INFO command, not on uas);
5747 e = NO_READ_CAPACITY_16 (don't use
5748 READ_CAPACITY_16 command);
5749 f = NO_REPORT_OPCODES (don't use report opcodes
5751 g = MAX_SECTORS_240 (don't transfer more than
5752 240 sectors at a time, uas only);
5753 h = CAPACITY_HEURISTICS (decrease the
5754 reported device capacity by one
5755 sector if the number is odd);
5756 i = IGNORE_DEVICE (don't bind to this
5758 j = NO_REPORT_LUNS (don't use report luns
5760 k = NO_SAME (do not use WRITE_SAME, uas only)
5761 l = NOT_LOCKABLE (don't try to lock and
5762 unlock ejectable media, not on uas);
5763 m = MAX_SECTORS_64 (don't transfer more
5764 than 64 sectors = 32 KB at a time,
5766 n = INITIAL_READ10 (force a retry of the
5767 initial READ(10) command, not on uas);
5768 o = CAPACITY_OK (accept the capacity
5769 reported by the device, not on uas);
5770 p = WRITE_CACHE (the device cache is ON
5771 by default, not on uas);
5772 r = IGNORE_RESIDUE (the device reports
5773 bogus residue values, not on uas);
5774 s = SINGLE_LUN (the device has only one
5776 t = NO_ATA_1X (don't allow ATA(12) and ATA(16)
5777 commands, uas only);
5778 u = IGNORE_UAS (don't bind to the uas driver);
5779 w = NO_WP_DETECT (don't test whether the
5780 medium is write-protected).
5781 y = ALWAYS_SYNC (issue a SYNCHRONIZE_CACHE
5782 even if the device claims no cache,
5784 Example: quirks=0419:aaf5:rl,0421:0433:rc
5786 user_debug= [KNL,ARM]
5788 See arch/arm/Kconfig.debug help text.
5789 1 - undefined instruction events
5791 4 - invalid data aborts
5794 Example: user_debug=31
5797 [X86] Flags controlling user PTE allocations.
5799 nohigh = do not allocate PTE pages in
5800 HIGHMEM regardless of setting
5804 On X86_32, this is an alias for vdso32=. Otherwise:
5806 vdso=1: enable VDSO (the default)
5807 vdso=0: disable VDSO mapping
5809 vdso32= [X86] Control the 32-bit vDSO
5810 vdso32=1: enable 32-bit VDSO
5811 vdso32=0 or vdso32=2: disable 32-bit VDSO
5813 See the help text for CONFIG_COMPAT_VDSO for more
5814 details. If CONFIG_COMPAT_VDSO is set, the default is
5815 vdso32=0; otherwise, the default is vdso32=1.
5817 For compatibility with older kernels, vdso32=2 is an
5820 Try vdso32=0 if you encounter an error that says:
5821 dl_main: Assertion `(void *) ph->p_vaddr == _rtld_local._dl_sysinfo_dso' failed!
5824 vector=percpu: enable percpu vector domain
5826 video= [FB] Frame buffer configuration
5827 See Documentation/fb/modedb.rst.
5829 video.brightness_switch_enabled= [0,1]
5830 If set to 1, on receiving an ACPI notify event
5831 generated by hotkey, video driver will adjust brightness
5832 level and then send out the event to user space through
5833 the allocated input device; If set to 0, video driver
5834 will only send out the event without touching backlight
5839 [VMMIO] Memory mapped virtio (platform) device.
5841 <size>@<baseaddr>:<irq>[:<id>]
5843 <size> := size (can use standard suffixes
5845 <baseaddr> := physical base address
5846 <irq> := interrupt number (as passed to
5848 <id> := (optional) platform device id
5850 virtio_mmio.device=1K@0x100b0000:48:7
5852 Can be used multiple times for multiple devices.
5854 vga= [BOOT,X86-32] Select a particular video mode
5855 See Documentation/x86/boot.rst and
5856 Documentation/admin-guide/svga.rst.
5857 Use vga=ask for menu.
5858 This is actually a boot loader parameter; the value is
5859 passed to the kernel using a special protocol.
5861 vm_debug[=options] [KNL] Available with CONFIG_DEBUG_VM=y.
5862 May slow down system boot speed, especially when
5863 enabled on systems with a large amount of memory.
5864 All options are enabled by default, and this
5865 interface is meant to allow for selectively
5866 enabling or disabling specific virtual memory
5869 Available options are:
5870 P Enable page structure init time poisoning
5871 - Disable all of the above options
5873 vmalloc=nn[KMG] [KNL,BOOT] Forces the vmalloc area to have an exact
5874 size of <nn>. This can be used to increase the
5875 minimum size (128MB on x86). It can also be used to
5876 decrease the size and leave more room for directly
5879 vmcp_cma=nn[MG] [KNL,S390]
5880 Sets the memory size reserved for contiguous memory
5881 allocations for the vmcp device driver.
5883 vmhalt= [KNL,S390] Perform z/VM CP command after system halt.
5886 vmpanic= [KNL,S390] Perform z/VM CP command after kernel panic.
5889 vmpoff= [KNL,S390] Perform z/VM CP command after power off.
5893 Controls the behavior of vsyscalls (i.e. calls to
5894 fixed addresses of 0xffffffffff600x00 from legacy
5895 code). Most statically-linked binaries and older
5896 versions of glibc use these calls. Because these
5897 functions are at fixed addresses, they make nice
5898 targets for exploits that can control RIP.
5900 emulate [default] Vsyscalls turn into traps and are
5901 emulated reasonably safely. The vsyscall
5904 xonly Vsyscalls turn into traps and are
5905 emulated reasonably safely. The vsyscall
5906 page is not readable.
5908 none Vsyscalls don't work at all. This makes
5909 them quite hard to use for exploits but
5910 might break your system.
5912 vt.color= [VT] Default text color.
5913 Format: 0xYX, X = foreground, Y = background.
5914 Default: 0x07 = light gray on black.
5916 vt.cur_default= [VT] Default cursor shape.
5917 Format: 0xCCBBAA, where AA, BB, and CC are the same as
5918 the parameters of the <Esc>[?A;B;Cc escape sequence;
5919 see VGA-softcursor.txt. Default: 2 = underline.
5921 vt.default_blu= [VT]
5922 Format: <blue0>,<blue1>,<blue2>,...,<blue15>
5923 Change the default blue palette of the console.
5924 This is a 16-member array composed of values
5927 vt.default_grn= [VT]
5928 Format: <green0>,<green1>,<green2>,...,<green15>
5929 Change the default green palette of the console.
5930 This is a 16-member array composed of values
5933 vt.default_red= [VT]
5934 Format: <red0>,<red1>,<red2>,...,<red15>
5935 Change the default red palette of the console.
5936 This is a 16-member array composed of values
5942 Set system-wide default UTF-8 mode for all tty's.
5943 Default is 1, i.e. UTF-8 mode is enabled for all
5944 newly opened terminals.
5946 vt.global_cursor_default=
5949 Set system-wide default for whether a cursor
5950 is shown on new VTs. Default is -1,
5951 i.e. cursors will be created by default unless
5952 overridden by individual drivers. 0 will hide
5953 cursors, 1 will display them.
5955 vt.italic= [VT] Default color for italic text; 0-15.
5958 vt.underline= [VT] Default color for underlined text; 0-15.
5961 watchdog timers [HW,WDT] For information on watchdog timers,
5962 see Documentation/watchdog/watchdog-parameters.rst
5963 or other driver-specific files in the
5964 Documentation/watchdog/ directory.
5968 Set the hard lockup detector stall duration
5969 threshold in seconds. The soft lockup detector
5970 threshold is set to twice the value. A value of 0
5971 disables both lockup detectors. Default is 10
5974 workqueue.watchdog_thresh=
5975 If CONFIG_WQ_WATCHDOG is configured, workqueue can
5976 warn stall conditions and dump internal state to
5977 help debugging. 0 disables workqueue stall
5978 detection; otherwise, it's the stall threshold
5979 duration in seconds. The default value is 30 and
5980 it can be updated at runtime by writing to the
5981 corresponding sysfs file.
5983 workqueue.disable_numa
5984 By default, all work items queued to unbound
5985 workqueues are affine to the NUMA nodes they're
5986 issued on, which results in better behavior in
5987 general. If NUMA affinity needs to be disabled for
5988 whatever reason, this option can be used. Note
5989 that this also can be controlled per-workqueue for
5990 workqueues visible under /sys/bus/workqueue/.
5992 workqueue.power_efficient
5993 Per-cpu workqueues are generally preferred because
5994 they show better performance thanks to cache
5995 locality; unfortunately, per-cpu workqueues tend to
5996 be more power hungry than unbound workqueues.
5998 Enabling this makes the per-cpu workqueues which
5999 were observed to contribute significantly to power
6000 consumption unbound, leading to measurably lower
6001 power usage at the cost of small performance
6004 The default value of this parameter is determined by
6005 the config option CONFIG_WQ_POWER_EFFICIENT_DEFAULT.
6007 workqueue.debug_force_rr_cpu
6008 Workqueue used to implicitly guarantee that work
6009 items queued without explicit CPU specified are put
6010 on the local CPU. This guarantee is no longer true
6011 and while local CPU is still preferred work items
6012 may be put on foreign CPUs. This debug option
6013 forces round-robin CPU selection to flush out
6014 usages which depend on the now broken guarantee.
6015 When enabled, memory and cache locality will be
6018 x2apic_phys [X86-64,APIC] Use x2apic physical mode instead of
6019 default x2apic cluster mode on platforms
6022 xen_512gb_limit [KNL,X86-64,XEN]
6023 Restricts the kernel running paravirtualized under Xen
6024 to use only up to 512 GB of RAM. The reason to do so is
6025 crash analysis tools and Xen tools for doing domain
6026 save/restore/migration must be enabled to handle larger
6029 xen_emul_unplug= [HW,X86,XEN]
6030 Unplug Xen emulated devices
6031 Format: [unplug0,][unplug1]
6032 ide-disks -- unplug primary master IDE devices
6033 aux-ide-disks -- unplug non-primary-master IDE devices
6034 nics -- unplug network devices
6035 all -- unplug all emulated devices (NICs and IDE disks)
6036 unnecessary -- unplugging emulated devices is
6037 unnecessary even if the host did not respond to
6039 never -- do not unplug even if version check succeeds
6041 xen_legacy_crash [X86,XEN]
6042 Crash from Xen panic notifier, without executing late
6043 panic() code such as dumping handler.
6045 xen_nopvspin [X86,XEN]
6046 Disables the qspinlock slowpath using Xen PV optimizations.
6047 This parameter is obsoleted by "nopvspin" parameter, which
6048 has equivalent effect for XEN platform.
6051 Disables the PV optimizations forcing the HVM guest to
6052 run as generic HVM guest with no PV drivers.
6053 This option is obsoleted by the "nopv" option, which
6054 has equivalent effect for XEN platform.
6056 xen_no_vector_callback
6057 [KNL,X86,XEN] Disable the vector callback for Xen
6058 event channel interrupts.
6060 xen_scrub_pages= [XEN]
6061 Boolean option to control scrubbing pages before giving them back
6062 to Xen, for use by other domains. Can be also changed at runtime
6063 with /sys/devices/system/xen_memory/xen_memory0/scrub_pages.
6064 Default value controlled with CONFIG_XEN_SCRUB_PAGES_DEFAULT.
6066 xen_timer_slop= [X86-64,XEN]
6067 Set the timer slop (in nanoseconds) for the virtual Xen
6068 timers (default is 100000). This adjusts the minimum
6069 delta of virtualized Xen timers, where lower values
6070 improve timer resolution at the expense of processing
6071 more timer interrupts.
6073 xen.event_eoi_delay= [XEN]
6074 How long to delay EOI handling in case of event
6075 storms (jiffies). Default is 10.
6077 xen.event_loop_timeout= [XEN]
6078 After which time (jiffies) the event handling loop
6079 should start to delay EOI handling. Default is 2.
6081 xen.fifo_events= [XEN]
6082 Boolean parameter to disable using fifo event handling
6083 even if available. Normally fifo event handling is
6084 preferred over the 2-level event handling, as it is
6085 fairer and the number of possible event channels is
6086 much higher. Default is on (use fifo events).
6088 nopv= [X86,XEN,KVM,HYPER_V,VMWARE]
6089 Disables the PV optimizations forcing the guest to run
6090 as generic guest with no PV drivers. Currently support
6091 XEN HVM, KVM, HYPER_V and VMWARE guest.
6093 nopvspin [X86,XEN,KVM]
6094 Disables the qspinlock slow path using PV optimizations
6095 which allow the hypervisor to 'idle' the guest on lock
6098 xirc2ps_cs= [NET,PCMCIA]
6100 <irq>,<irq_mask>,<io>,<full_duplex>,<do_sound>,<lockup_hack>[,<irq2>[,<irq3>[,<irq4>]]]
6103 By default on POWER9 and above, the kernel will
6104 natively use the XIVE interrupt controller. This option
6105 allows the fallback firmware mode to be used:
6107 off Fallback to firmware control of XIVE interrupt
6108 controller on both pseries and powernv
6109 platforms. Only useful on POWER9 and above.
6111 xhci-hcd.quirks [USB,KNL]
6112 A hex value specifying bitmask with supplemental xhci
6113 host controller quirks. Meaning of each bit can be
6114 consulted in header drivers/usb/host/xhci.h.
6117 Format: { early | on | rw | ro | off }
6118 Controls if xmon debugger is enabled. Default is off.
6119 Passing only "xmon" is equivalent to "xmon=early".
6120 early Call xmon as early as possible on boot; xmon
6121 debugger is called from setup_arch().
6122 on xmon debugger hooks will be installed so xmon
6123 is only called on a kernel crash. Default mode,
6124 i.e. either "ro" or "rw" mode, is controlled
6125 with CONFIG_XMON_DEFAULT_RO_MODE.
6126 rw xmon debugger hooks will be installed so xmon
6127 is called only on a kernel crash, mode is write,
6128 meaning SPR registers, memory and, other data
6129 can be written using xmon commands.
6130 ro same as "rw" option above but SPR registers,
6131 memory, and other data can't be written using
6133 off xmon is disabled.