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]
27 If set to vendor, prefer vendor specific driver
28 (e.g. thinkpad_acpi, sony_acpi, etc.) instead
29 of the ACPI video.ko driver.
31 acpi_force_32bit_fadt_addr
32 force FADT to use 32 bit addresses rather than the
33 64 bit X_* addresses. Some firmware have broken 64
34 bit addresses for force ACPI ignore these and use
35 the older legacy 32 bit addresses.
37 acpica_no_return_repair [HW, ACPI]
38 Disable AML predefined validation mechanism
39 This mechanism can repair the evaluation result to make
40 the return objects more ACPI specification compliant.
41 This option is useful for developers to identify the
42 root cause of an AML interpreter issue when the issue
43 has something to do with the repair mechanism.
45 acpi.debug_layer= [HW,ACPI,ACPI_DEBUG]
46 acpi.debug_level= [HW,ACPI,ACPI_DEBUG]
48 CONFIG_ACPI_DEBUG must be enabled to produce any ACPI
49 debug output. Bits in debug_layer correspond to a
50 _COMPONENT in an ACPI source file, e.g.,
51 #define _COMPONENT ACPI_PCI_COMPONENT
52 Bits in debug_level correspond to a level in
53 ACPI_DEBUG_PRINT statements, e.g.,
54 ACPI_DEBUG_PRINT((ACPI_DB_INFO, ...
55 The debug_level mask defaults to "info". See
56 Documentation/firmware-guide/acpi/debug.rst for more information about
57 debug layers and levels.
59 Enable processor driver info messages:
60 acpi.debug_layer=0x20000000
61 Enable PCI/PCI interrupt routing info messages:
62 acpi.debug_layer=0x400000
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_rsdp= [ACPI,EFI,KEXEC]
140 Pass the RSDP address to the kernel, mostly used
141 on machines running EFI runtime service to boot the
142 second kernel for kdump.
144 acpi_os_name= [HW,ACPI] Tell ACPI BIOS the name of the OS
145 Format: To spoof as Windows 98: ="Microsoft Windows"
147 acpi_rev_override [ACPI] Override the _REV object to return 5 (instead
148 of 2 which is mandated by ACPI 6) as the supported ACPI
149 specification revision (when using this switch, it may
150 be necessary to carry out a cold reboot _twice_ in a
151 row to make it take effect on the platform firmware).
153 acpi_osi= [HW,ACPI] Modify list of supported OS interface strings
154 acpi_osi="string1" # add string1
155 acpi_osi="!string2" # remove string2
156 acpi_osi=!* # remove all strings
157 acpi_osi=! # disable all built-in OS vendor
159 acpi_osi=!! # enable all built-in OS vendor
161 acpi_osi= # disable all strings
163 'acpi_osi=!' can be used in combination with single or
164 multiple 'acpi_osi="string1"' to support specific OS
165 vendor string(s). Note that such command can only
166 affect the default state of the OS vendor strings, thus
167 it cannot affect the default state of the feature group
168 strings and the current state of the OS vendor strings,
169 specifying it multiple times through kernel command line
170 is meaningless. This command is useful when one do not
171 care about the state of the feature group strings which
172 should be controlled by the OSPM.
174 1. 'acpi_osi=! acpi_osi="Windows 2000"' is equivalent
175 to 'acpi_osi="Windows 2000" acpi_osi=!', they all
176 can make '_OSI("Windows 2000")' TRUE.
178 'acpi_osi=' cannot be used in combination with other
179 'acpi_osi=' command lines, the _OSI method will not
180 exist in the ACPI namespace. NOTE that such command can
181 only affect the _OSI support state, thus specifying it
182 multiple times through kernel command line is also
185 1. 'acpi_osi=' can make 'CondRefOf(_OSI, Local1)'
188 'acpi_osi=!*' can be used in combination with single or
189 multiple 'acpi_osi="string1"' to support specific
190 string(s). Note that such command can affect the
191 current state of both the OS vendor strings and the
192 feature group strings, thus specifying it multiple times
193 through kernel command line is meaningful. But it may
194 still not able to affect the final state of a string if
195 there are quirks related to this string. This command
196 is useful when one want to control the state of the
197 feature group strings to debug BIOS issues related to
200 1. 'acpi_osi="Module Device" acpi_osi=!*' can make
201 '_OSI("Module Device")' FALSE.
202 2. 'acpi_osi=!* acpi_osi="Module Device"' can make
203 '_OSI("Module Device")' TRUE.
204 3. 'acpi_osi=! acpi_osi=!* acpi_osi="Windows 2000"' is
206 'acpi_osi=!* acpi_osi=! acpi_osi="Windows 2000"'
208 'acpi_osi=!* acpi_osi="Windows 2000" acpi_osi=!',
209 they all will make '_OSI("Windows 2000")' TRUE.
212 Override the pmtimer bug detection: force the kernel
213 to assume that this machine's pmtimer latches its value
214 and always returns good values.
216 acpi_sci= [HW,ACPI] ACPI System Control Interrupt trigger mode
217 Format: { level | edge | high | low }
219 acpi_skip_timer_override [HW,ACPI]
220 Recognize and ignore IRQ0/pin2 Interrupt Override.
221 For broken nForce2 BIOS resulting in XT-PIC timer.
223 acpi_sleep= [HW,ACPI] Sleep options
224 Format: { s3_bios, s3_mode, s3_beep, s4_nohwsig,
225 old_ordering, nonvs, sci_force_enable, nobl }
226 See Documentation/power/video.rst for information on
228 s3_beep is for debugging; it makes the PC's speaker beep
229 as soon as the kernel's real-mode entry point is called.
230 s4_nohwsig prevents ACPI hardware signature from being
231 used during resume from hibernation.
232 old_ordering causes the ACPI 1.0 ordering of the _PTS
233 control method, with respect to putting devices into
234 low power states, to be enforced (the ACPI 2.0 ordering
235 of _PTS is used by default).
236 nonvs prevents the kernel from saving/restoring the
237 ACPI NVS memory during suspend/hibernation and resume.
238 sci_force_enable causes the kernel to set SCI_EN directly
239 on resume from S1/S3 (which is against the ACPI spec,
240 but some broken systems don't work without it).
241 nobl causes the internal blacklist of systems known to
242 behave incorrectly in some ways with respect to system
243 suspend and resume to be ignored (use wisely).
245 acpi_use_timer_override [HW,ACPI]
246 Use timer override. For some broken Nvidia NF5 boards
247 that require a timer override, but don't have HPET
249 add_efi_memmap [EFI; X86] Include EFI memory map in
250 kernel's map of available physical RAM.
253 { off | try_unsupported }
254 off: disable AGP support
255 try_unsupported: try to drive unsupported chipsets
256 (may crash computer or cause data corruption)
259 See Documentation/sound/alsa-configuration.rst
262 Allow the default userspace alignment fault handler
263 behaviour to be specified. Bit 0 enables warnings,
264 bit 1 enables fixups, and bit 2 sends a segfault.
266 align_va_addr= [X86-64]
267 Align virtual addresses by clearing slice [14:12] when
268 allocating a VMA at process creation time. This option
269 gives you up to 3% performance improvement on AMD F15h
270 machines (where it is enabled by default) for a
271 CPU-intensive style benchmark, and it can vary highly in
272 a microbenchmark depending on workload and compiler.
274 32: only for 32-bit processes
275 64: only for 64-bit processes
276 on: enable for both 32- and 64-bit processes
277 off: disable for both 32- and 64-bit processes
279 alloc_snapshot [FTRACE]
280 Allocate the ftrace snapshot buffer on boot up when the
281 main buffer is allocated. This is handy if debugging
282 and you need to use tracing_snapshot() on boot up, and
283 do not want to use tracing_snapshot_alloc() as it needs
284 to be done where GFP_KERNEL allocations are allowed.
286 amd_iommu= [HW,X86-64]
287 Pass parameters to the AMD IOMMU driver in the system.
289 fullflush - enable flushing of IO/TLB entries when
290 they are unmapped. Otherwise they are
291 flushed before they will be reused, which
293 off - do not initialize any AMD IOMMU found in
295 force_isolation - Force device isolation for all
296 devices. The IOMMU driver is not
297 allowed anymore to lift isolation
298 requirements as needed. This option
299 does not override iommu=pt
301 amd_iommu_dump= [HW,X86-64]
302 Enable AMD IOMMU driver option to dump the ACPI table
303 for AMD IOMMU. With this option enabled, AMD IOMMU
304 driver will print ACPI tables for AMD IOMMU during
305 IOMMU initialization.
307 amd_iommu_intr= [HW,X86-64]
308 Specifies one of the following AMD IOMMU interrupt
310 legacy - Use legacy interrupt remapping mode.
311 vapic - Use virtual APIC mode, which allows IOMMU
312 to inject interrupts directly into guest.
313 This mode requires kvm-amd.avic=1.
314 (Default when IOMMU HW support is present.)
316 amijoy.map= [HW,JOY] Amiga joystick support
317 Map of devices attached to JOY0DAT and JOY1DAT
319 See also Documentation/input/joydev/joystick.rst
321 analog.map= [HW,JOY] Analog joystick and gamepad support
322 Specifies type or capabilities of an analog joystick
323 connected to one of 16 gameports
324 Format: <type1>,<type2>,..<type16>
327 Power management functions (SPARCstation-4/5 + deriv.)
329 Disable APC CPU standby support. SPARCstation-Fox does
330 not play well with APC CPU idle - disable it if you have
331 APC and your system crashes randomly.
333 apic= [APIC,X86] Advanced Programmable Interrupt Controller
334 Change the output verbosity while booting
335 Format: { quiet (default) | verbose | debug }
336 Change the amount of debugging information output
337 when initialising the APIC and IO-APIC components.
338 For X86-32, this can also be used to specify an APIC
340 Format: apic=driver_name
341 Examples: apic=bigsmp
343 apic_extnmi= [APIC,X86] External NMI delivery setting
344 Format: { bsp (default) | all | none }
345 bsp: External NMI is delivered only to CPU 0
346 all: External NMIs are broadcast to all CPUs as a
348 none: External NMI is masked for all CPUs. This is
349 useful so that a dump capture kernel won't be
353 See Documentation/networking/ipv6.txt.
355 show_lapic= [APIC,X86] Advanced Programmable Interrupt Controller
356 Limit apic dumping. The parameter defines the maximal
357 number of local apics being dumped. Also it is possible
358 to set it to "all" by meaning -- no limit here.
359 Format: { 1 (default) | 2 | ... | all }.
360 The parameter valid if only apic=debug or
361 apic=verbose is specified.
362 Example: apic=debug show_lapic=all
364 apm= [APM] Advanced Power Management
365 See header of arch/x86/kernel/apm_32.c.
367 arcrimi= [HW,NET] ARCnet - "RIM I" (entirely mem-mapped) cards
368 Format: <io>,<irq>,<nodeID>
372 atarimouse= [HW,MOUSE] Atari Mouse
374 atkbd.extra= [HW] Enable extra LEDs and keys on IBM RapidAccess,
375 EzKey and similar keyboards
377 atkbd.reset= [HW] Reset keyboard during initialization
379 atkbd.set= [HW] Select keyboard code set
380 Format: <int> (2 = AT (default), 3 = PS/2)
382 atkbd.scroll= [HW] Enable scroll wheel on MS Office and similar
385 atkbd.softraw= [HW] Choose between synthetic and real raw mode
386 Format: <bool> (0 = real, 1 = synthetic (default))
388 atkbd.softrepeat= [HW]
389 Use software keyboard repeat
391 audit= [KNL] Enable the audit sub-system
392 Format: { "0" | "1" | "off" | "on" }
393 0 | off - kernel audit is disabled and can not be
394 enabled until the next reboot
395 unset - kernel audit is initialized but disabled and
396 will be fully enabled by the userspace auditd.
397 1 | on - kernel audit is initialized and partially
398 enabled, storing at most audit_backlog_limit
399 messages in RAM until it is fully enabled by the
403 audit_backlog_limit= [KNL] Set the audit queue size limit.
404 Format: <int> (must be >=0)
407 bau= [X86_UV] Enable the BAU on SGI UV. The default
408 behavior is to disable the BAU (i.e. bau=0).
409 Format: { "0" | "1" }
412 unset - Disable the BAU.
414 baycom_epp= [HW,AX25]
417 baycom_par= [HW,AX25] BayCom Parallel Port AX.25 Modem
419 See header of drivers/net/hamradio/baycom_par.c.
421 baycom_ser_fdx= [HW,AX25]
422 BayCom Serial Port AX.25 Modem (Full Duplex Mode)
423 Format: <io>,<irq>,<mode>[,<baud>]
424 See header of drivers/net/hamradio/baycom_ser_fdx.c.
426 baycom_ser_hdx= [HW,AX25]
427 BayCom Serial Port AX.25 Modem (Half Duplex Mode)
428 Format: <io>,<irq>,<mode>
429 See header of drivers/net/hamradio/baycom_ser_hdx.c.
431 blkdevparts= Manual partition parsing of block device(s) for
432 embedded devices based on command line input.
433 See Documentation/block/cmdline-partition.rst
435 boot_delay= Milliseconds to delay each printk during boot.
436 Values larger than 10 seconds (10000) are changed to
440 bootmem_debug [KNL] Enable bootmem allocator debug messages.
443 Disable BERT OS support on buggy BIOSes.
445 bttv.card= [HW,V4L] bttv (bt848 + bt878 based grabber cards)
446 bttv.radio= Most important insmod options are available as
448 bttv.pll= See Documentation/media/v4l-drivers/bttv.rst
451 bulk_remove=off [PPC] This parameter disables the use of the pSeries
452 firmware feature for flushing multiple hpte entries
455 c101= [NET] Moxa C101 synchronous serial card
457 cachesize= [BUGS=X86-32] Override level 2 CPU cache size detection.
458 Sometimes CPU hardware bugs make them report the cache
459 size incorrectly. The kernel will attempt work arounds
460 to fix known problems, but for some CPUs it is not
461 possible to determine what the correct size should be.
462 This option provides an override for these situations.
465 [NET] Specifies amount of time (in seconds) that
466 the kernel should wait for a network carrier. By default
467 it waits 120 seconds.
469 ca_keys= [KEYS] This parameter identifies a specific key(s) on
470 the system trusted keyring to be used for certificate
472 format: { id:<keyid> | builtin }
474 cca= [MIPS] Override the kernel pages' cache coherency
475 algorithm. Accepted values range from 0 to 7
476 inclusive. See arch/mips/include/asm/pgtable-bits.h
477 for platform specific values (SB1, Loongson3 and
480 ccw_timeout_log [S390]
481 See Documentation/s390/common_io.rst for details.
483 cgroup_disable= [KNL] Disable a particular controller
484 Format: {name of the controller(s) to disable}
485 The effects of cgroup_disable=foo are:
486 - foo isn't auto-mounted if you mount all cgroups in
488 - foo isn't visible as an individually mountable
490 {Currently only "memory" controller deal with this and
491 cut the overhead, others just disable the usage. So
492 only cgroup_disable=memory is actually worthy}
494 cgroup_no_v1= [KNL] Disable cgroup controllers and named hierarchies in v1
495 Format: { { controller | "all" | "named" }
496 [,{ controller | "all" | "named" }...] }
497 Like cgroup_disable, but only applies to cgroup v1;
498 the blacklisted controllers remain available in cgroup2.
499 "all" blacklists all controllers and "named" disables
500 named mounts. Specifying both "all" and "named" disables
503 cgroup.memory= [KNL] Pass options to the cgroup memory controller.
505 nosocket -- Disable socket memory accounting.
506 nokmem -- Disable kernel memory accounting.
508 checkreqprot [SELINUX] Set initial checkreqprot flag value.
509 Format: { "0" | "1" }
510 See security/selinux/Kconfig help text.
511 0 -- check protection applied by kernel (includes
512 any implied execute protection).
513 1 -- check protection requested by application.
514 Default value is set via a kernel config option.
515 Value can be changed at runtime via
516 /selinux/checkreqprot.
519 See Documentation/s390/common_io.rst for details.
522 Prevents the clock framework from automatically gating
523 clocks that have not been explicitly enabled by a Linux
524 device driver but are enabled in hardware at reset or
525 by the bootloader/firmware. Note that this does not
526 force such clocks to be always-on nor does it reserve
527 those clocks in any way. This parameter is useful for
528 debug and development, but should not be needed on a
529 platform with proper driver support. For more
530 information, see Documentation/driver-api/clk.rst.
532 clock= [BUGS=X86-32, HW] gettimeofday clocksource override.
534 Forces specified clocksource (if available) to be used
535 when calculating gettimeofday(). If specified
536 clocksource is not available, it defaults to PIT.
537 Format: { pit | tsc | cyclone | pmtmr }
539 clocksource= Override the default clocksource
541 Override the default clocksource and use the clocksource
542 with the name specified.
543 Some clocksource names to choose from, depending on
545 [all] jiffies (this is the base, fallback clocksource)
547 [ARM] imx_timer1,OSTS,netx_timer,mpu_timer2,
548 pxa_timer,timer3,32k_counter,timer0_1
549 [X86-32] pit,hpet,tsc;
550 scx200_hrt on Geode; cyclone on IBM x440
558 clocksource.arm_arch_timer.evtstrm=
561 Enable/disable the eventstream feature of the ARM
562 architected timer so that code using WFE-based polling
563 loops can be debugged more effectively on production
566 clearcpuid=BITNUM [X86]
567 Disable CPUID feature X for the kernel. See
568 arch/x86/include/asm/cpufeatures.h for the valid bit
569 numbers. Note the Linux specific bits are not necessarily
570 stable over kernel options, but the vendor specific
572 Also note that user programs calling CPUID directly
573 or using the feature without checking anything
574 will still see it. This just prevents it from
575 being used by the kernel or shown in /proc/cpuinfo.
576 Also note the kernel might malfunction if you disable
579 cma=nn[MG]@[start[MG][-end[MG]]]
581 Sets the size of kernel global memory area for
582 contiguous memory allocations and optionally the
583 placement constraint by the physical address range of
584 memory allocations. A value of 0 disables CMA
585 altogether. For more information, see
586 include/linux/dma-contiguous.h
588 cmo_free_hint= [PPC] Format: { yes | no }
589 Specify whether pages are marked as being inactive
590 when they are freed. This is used in CMO environments
591 to determine OS memory pressure for page stealing by
595 coherent_pool=nn[KMG] [ARM,KNL]
596 Sets the size of memory pool for coherent, atomic dma
597 allocations, by default set to 256K.
599 com20020= [HW,NET] ARCnet - COM20020 chipset
601 <io>[,<irq>[,<nodeID>[,<backplane>[,<ckp>[,<timeout>]]]]]
603 com90io= [HW,NET] ARCnet - COM90xx chipset (IO-mapped buffers)
607 ARCnet - COM90xx chipset (memory-mapped buffers)
608 Format: <io>[,<irq>[,<memstart>]]
610 condev= [HW,S390] console device
613 console= [KNL] Output console device and options.
615 tty<n> Use the virtual console device <n>.
619 Use the specified serial port. The options are of
620 the form "bbbbpnf", where "bbbb" is the baud rate,
621 "p" is parity ("n", "o", or "e"), "n" is number of
622 bits, and "f" is flow control ("r" for RTS or
623 omit it). Default is "9600n8".
625 See Documentation/admin-guide/serial-console.rst for more
627 Documentation/networking/netconsole.txt for an
630 uart[8250],io,<addr>[,options]
631 uart[8250],mmio,<addr>[,options]
632 uart[8250],mmio16,<addr>[,options]
633 uart[8250],mmio32,<addr>[,options]
634 uart[8250],0x<addr>[,options]
635 Start an early, polled-mode console on the 8250/16550
636 UART at the specified I/O port or MMIO address,
637 switching to the matching ttyS device later.
638 MMIO inter-register address stride is either 8-bit
639 (mmio), 16-bit (mmio16), or 32-bit (mmio32).
640 If none of [io|mmio|mmio16|mmio32], <addr> is assumed
641 to be equivalent to 'mmio'. 'options' are specified in
642 the same format described for ttyS above; if unspecified,
643 the h/w is not re-initialized.
645 hvc<n> Use the hypervisor console device <n>. This is for
646 both Xen and PowerPC hypervisors.
648 If the device connected to the port is not a TTY but a braille
649 device, prepend "brl," before the device type, for instance
651 For now, only VisioBraille is supported.
654 [KNL] Change console messages format
656 By default we print messages on consoles in
657 "[time stamp] text\n" format (time stamp may not be
658 printed, depending on CONFIG_PRINTK_TIME or
659 `printk_time' param).
661 Switch to syslog format: "<%u>[time stamp] text\n"
662 IOW, each message will have a facility and loglevel
663 prefix. The format is similar to one used by syslog()
664 syscall, or to executing "dmesg -S --raw" or to reading
667 consoleblank= [KNL] The console blank (screen saver) timeout in
668 seconds. A value of 0 disables the blank timer.
672 [KNL] Change the default value for
673 /proc/<pid>/coredump_filter.
674 See also Documentation/filesystems/proc.txt.
676 coresight_cpu_debug.enable
679 Enable/disable the CPU sampling based debugging.
680 0: default value, disable debugging
681 1: enable debugging at boot time
683 cpuidle.off=1 [CPU_IDLE]
684 disable the cpuidle sub-system
687 [CPU_IDLE] Name of the cpuidle governor to use.
689 cpufreq.off=1 [CPU_FREQ]
690 disable the cpufreq sub-system
693 [X86] Delay for N microsec between assert and de-assert
694 of APIC INIT to start processors. This delay occurs
695 on every CPU online, such as boot, and resume from suspend.
698 cpcihp_generic= [HW,PCI] Generic port I/O CompactPCI driver
700 <first_slot>,<last_slot>,<port>,<enum_bit>[,<debug>]
702 crashkernel=size[KMG][@offset[KMG]]
703 [KNL] Using kexec, Linux can switch to a 'crash kernel'
704 upon panic. This parameter reserves the physical
705 memory region [offset, offset + size] for that kernel
706 image. If '@offset' is omitted, then a suitable offset
707 is selected automatically.
708 [KNL, x86_64] select a region under 4G first, and
709 fall back to reserve region above 4G when '@offset'
710 hasn't been specified.
711 See Documentation/admin-guide/kdump/kdump.rst for further details.
713 crashkernel=range1:size1[,range2:size2,...][@offset]
714 [KNL] Same as above, but depends on the memory
715 in the running system. The syntax of range is
716 start-[end] where start and end are both
717 a memory unit (amount[KMG]). See also
718 Documentation/admin-guide/kdump/kdump.rst for an example.
720 crashkernel=size[KMG],high
721 [KNL, x86_64] range could be above 4G. Allow kernel
722 to allocate physical memory region from top, so could
723 be above 4G if system have more than 4G ram installed.
724 Otherwise memory region will be allocated below 4G, if
726 It will be ignored if crashkernel=X is specified.
727 crashkernel=size[KMG],low
728 [KNL, x86_64] range under 4G. When crashkernel=X,high
729 is passed, kernel could allocate physical memory region
730 above 4G, that cause second kernel crash on system
731 that require some amount of low memory, e.g. swiotlb
732 requires at least 64M+32K low memory, also enough extra
733 low memory is needed to make sure DMA buffers for 32-bit
734 devices won't run out. Kernel would try to allocate at
735 at least 256M below 4G automatically.
736 This one let user to specify own low range under 4G
737 for second kernel instead.
738 0: to disable low allocation.
739 It will be ignored when crashkernel=X,high is not used
740 or memory reserved is below 4G.
743 [KNL] Disable crypto self-tests
748 cs89x0_media= [HW,NET]
749 Format: { rj45 | aui | bnc }
752 See header of drivers/s390/block/dasd_devmap.c.
754 db9.dev[2|3]= [HW,JOY] Multisystem joystick support via parallel port
755 (one device per port)
756 Format: <port#>,<type>
757 See also Documentation/input/devices/joystick-parport.rst
759 ddebug_query= [KNL,DYNAMIC_DEBUG] Enable debug messages at early boot
761 Documentation/admin-guide/dynamic-debug-howto.rst for
762 details. Deprecated, see dyndbg.
764 debug [KNL] Enable kernel debugging (events log level).
767 [KNL] Enable printing [hashed] pointers early in the
768 boot sequence. If enabled, we use a weak hash instead
769 of siphash to hash pointers. Use this option if you are
770 seeing instances of '(___ptrval___)') and need to see a
771 value (hashed pointer) instead. Cryptographically
772 insecure, please do not use on production kernels.
775 [KNL] verbose self-tests
777 Print debugging info while doing the locking API
779 We default to 0 (no extra messages), setting it to
780 1 will print _a lot_ more information - normally
781 only useful to kernel developers.
783 debug_objects [KNL] Enable object debugging
786 [KNL] Disable object debugging
788 debug_guardpage_minorder=
789 [KNL] When CONFIG_DEBUG_PAGEALLOC is set, this
790 parameter allows control of the order of pages that will
791 be intentionally kept free (and hence protected) by the
792 buddy allocator. Bigger value increase the probability
793 of catching random memory corruption, but reduce the
794 amount of memory for normal system use. The maximum
795 possible value is MAX_ORDER/2. Setting this parameter
796 to 1 or 2 should be enough to identify most random
797 memory corruption problems caused by bugs in kernel or
798 driver code when a CPU writes to (or reads from) a
799 random memory location. Note that there exists a class
800 of memory corruptions problems caused by buggy H/W or
801 F/W or by drivers badly programing DMA (basically when
802 memory is written at bus level and the CPU MMU is
803 bypassed) which are not detectable by
804 CONFIG_DEBUG_PAGEALLOC, hence this option will not help
805 tracking down these problems.
808 [KNL] When CONFIG_DEBUG_PAGEALLOC is set, this parameter
809 enables the feature at boot time. By default, it is
810 disabled and the system will work mostly the same as a
811 kernel built without CONFIG_DEBUG_PAGEALLOC.
812 on: enable the feature
814 debugpat [X86] Enable PAT debugging
816 decnet.addr= [HW,NET]
817 Format: <area>[,<node>]
818 See also Documentation/networking/decnet.txt.
821 [same as hugepagesz=] The size of the default
822 HugeTLB page size. This is the size represented by
823 the legacy /proc/ hugepages APIs, used for SHM, and
824 default size when mounting hugetlbfs filesystems.
825 Defaults to the default architecture's huge page size
828 deferred_probe_timeout=
829 [KNL] Debugging option to set a timeout in seconds for
830 deferred probe to give up waiting on dependencies to
831 probe. Only specific dependencies (subsystems or
832 drivers) that have opted in will be ignored. A timeout of 0
833 will timeout at the end of initcalls. This option will also
834 dump out devices still on the deferred probe list after
838 Set number of hash buckets for dentry cache.
840 disable_1tb_segments [PPC]
841 Disables the use of 1TB hash page table segments. This
842 causes the kernel to fall back to 256MB segments which
843 can be useful when debugging issues that require an SLB
847 See Documentation/networking/ipv6.txt.
850 [KNL] Under CONFIG_HARDENED_USERCOPY, whether
851 hardening is enabled for this boot. Hardened
852 usercopy checking is used to protect the kernel
853 from reading or writing beyond known memory
854 allocation boundaries as a proactive defense
855 against bounds-checking flaws in the kernel's
856 copy_to_user()/copy_from_user() interface.
857 on Perform hardened usercopy checks (default).
858 off Disable hardened usercopy checks.
861 Disable RADIX MMU mode on POWER9
863 disable_cpu_apicid= [X86,APIC,SMP]
865 The number of initial APIC ID for the
866 corresponding CPU to be disabled at boot,
867 mostly used for the kdump 2nd kernel to
868 disable BSP to wake up multiple CPUs without
869 causing system reset or hang due to sending
872 perf_v4_pmi= [X86,INTEL]
874 Disable Intel PMU counter freezing feature.
875 The feature only exists starting from
876 Arch Perfmon v4 (Skylake and newer).
878 disable_ddw [PPC/PSERIES]
879 Disable Dynamic DMA Window support. Use this if
880 to workaround buggy firmware.
883 See Documentation/networking/ipv6.txt.
885 disable_mtrr_cleanup [X86]
886 The kernel tries to adjust MTRR layout from continuous
887 to discrete, to make X server driver able to add WB
888 entry later. This parameter disables that.
890 disable_mtrr_trim [X86, Intel and AMD only]
891 By default the kernel will trim any uncacheable
892 memory out of your available memory pool based on
893 MTRR settings. This parameter disables that behavior,
894 possibly causing your machine to run very slowly.
896 disable_timer_pin_1 [X86]
897 Disable PIN 1 of APIC timer
898 Can be useful to work around chipset bugs.
900 dis_ucode_ldr [X86] Disable the microcode loader.
902 dma_debug=off If the kernel is compiled with DMA_API_DEBUG support,
903 this option disables the debugging code at boot.
905 dma_debug_entries=<number>
906 This option allows to tune the number of preallocated
907 entries for DMA-API debugging code. One entry is
908 required per DMA-API allocation. Use this if the
909 DMA-API debugging code disables itself because the
910 architectural default is too low.
912 dma_debug_driver=<driver_name>
913 With this option the DMA-API debugging driver
914 filter feature can be enabled at boot time. Just
915 pass the driver to filter for as the parameter.
916 The filter can be disabled or changed to another
917 driver later using sysfs.
919 driver_async_probe= [KNL]
920 List of driver names to be probed asynchronously.
921 Format: <driver_name1>,<driver_name2>...
923 drm.edid_firmware=[<connector>:]<file>[,[<connector>:]<file>]
924 Broken monitors, graphic adapters, KVMs and EDIDless
925 panels may send no or incorrect EDID data sets.
926 This parameter allows to specify an EDID data sets
927 in the /lib/firmware directory that are used instead.
928 Generic built-in EDID data sets are used, if one of
929 edid/1024x768.bin, edid/1280x1024.bin,
930 edid/1680x1050.bin, or edid/1920x1080.bin is given
931 and no file with the same name exists. Details and
932 instructions how to build your own EDID data are
933 available in Documentation/driver-api/edid.rst. An EDID
934 data set will only be used for a particular connector,
935 if its name and a colon are prepended to the EDID
936 name. Each connector may use a unique EDID data
937 set by separating the files with a comma. An EDID
938 data set with no connector name will be used for
939 any connectors not explicitly specified.
944 Format: {"off" | "known"}
945 Control how the dt_cpu_ftrs device-tree binding is
946 used for CPU feature discovery and setup (if it
948 off: Do not use it, fall back to legacy cpu table.
949 known: Do not pass through unknown features to guests
950 or userspace, only those that the kernel is aware of.
952 dump_apple_properties [X86]
953 Dump name and content of EFI device properties on
954 x86 Macs. Useful for driver authors to determine
955 what data is available or for reverse-engineering.
957 dyndbg[="val"] [KNL,DYNAMIC_DEBUG]
958 module.dyndbg[="val"]
959 Enable debug messages at boot time. See
960 Documentation/admin-guide/dynamic-debug-howto.rst
963 nompx [X86] Disables Intel Memory Protection Extensions.
964 See Documentation/x86/intel_mpx.rst for more
965 information about the feature.
967 nopku [X86] Disable Memory Protection Keys CPU feature found
970 module.async_probe [KNL]
971 Enable asynchronous probe on this module.
973 early_ioremap_debug [KNL]
974 Enable debug messages in early_ioremap support. This
975 is useful for tracking down temporary early mappings
976 which are not unmapped.
978 earlycon= [KNL] Output early console device and options.
980 [ARM64] The early console is determined by the
981 stdout-path property in device tree's chosen node,
982 or determined by the ACPI SPCR table.
984 [X86] When used with no options the early console is
985 determined by the ACPI SPCR table.
987 cdns,<addr>[,options]
988 Start an early, polled-mode console on a Cadence
989 (xuartps) serial port at the specified address. Only
990 supported option is baud rate. If baud rate is not
991 specified, the serial port must already be setup and
994 uart[8250],io,<addr>[,options]
995 uart[8250],mmio,<addr>[,options]
996 uart[8250],mmio32,<addr>[,options]
997 uart[8250],mmio32be,<addr>[,options]
998 uart[8250],0x<addr>[,options]
999 Start an early, polled-mode console on the 8250/16550
1000 UART at the specified I/O port or MMIO address.
1001 MMIO inter-register address stride is either 8-bit
1002 (mmio) or 32-bit (mmio32 or mmio32be).
1003 If none of [io|mmio|mmio32|mmio32be], <addr> is assumed
1004 to be equivalent to 'mmio'. 'options' are specified
1005 in the same format described for "console=ttyS<n>"; if
1006 unspecified, the h/w is not initialized.
1010 Start an early, polled-mode console on a pl011 serial
1011 port at the specified address. The pl011 serial port
1012 must already be setup and configured. Options are not
1013 yet supported. If 'mmio32' is specified, then only
1014 the driver will use only 32-bit accessors to read/write
1015 the device registers.
1018 Start an early, polled-mode console on a meson serial
1019 port at the specified address. The serial port must
1020 already be setup and configured. Options are not yet
1024 Start an early, polled-mode console on an msm serial
1025 port at the specified address. The serial port
1026 must already be setup and configured. Options are not
1029 msm_serial_dm,<addr>
1030 Start an early, polled-mode console on an msm serial
1031 dm port at the specified address. The serial port
1032 must already be setup and configured. Options are not
1036 Start an early, polled-mode console on a serial port
1037 of an Actions Semi SoC, such as S500 or S900, at the
1038 specified address. The serial port must already be
1039 setup and configured. Options are not yet supported.
1042 Start an early, polled-mode console on a serial port
1043 of an RDA Micro SoC, such as RDA8810PL, at the
1044 specified address. The serial port must already be
1045 setup and configured. Options are not yet supported.
1047 smh Use ARM semihosting calls for early console.
1055 Use early console provided by serial driver available
1056 on Samsung SoCs, requires selecting proper type and
1057 a correct base address of the selected UART port. The
1058 serial port must already be setup and configured.
1059 Options are not yet supported.
1062 Start an early, polled-mode console on a lantiq serial
1063 (lqasc) port at the specified address. The serial port
1064 must already be setup and configured. Options are not
1069 Use early console provided by Freescale LP UART driver
1070 found on Freescale Vybrid and QorIQ LS1021A processors.
1071 A valid base address must be provided, and the serial
1072 port must already be setup and configured.
1075 Start an early, polled-mode console on the
1076 Armada 3700 serial port at the specified
1077 address. The serial port must already be setup
1078 and configured. Options are not yet supported.
1081 Start an early, polled-mode console on a Qualcomm
1082 Generic Interface (GENI) based serial port at the
1083 specified address. The serial port must already be
1084 setup and configured. Options are not yet supported.
1087 Start an early, unaccelerated console on the EFI
1088 memory mapped framebuffer (if available). On cache
1089 coherent non-x86 systems that use system memory for
1090 the framebuffer, pass the 'ram' option so that it is
1091 mapped with the correct attributes.
1093 earlyprintk= [X86,SH,ARM,M68k,S390]
1097 earlyprintk=serial[,ttySn[,baudrate]]
1098 earlyprintk=serial[,0x...[,baudrate]]
1099 earlyprintk=ttySn[,baudrate]
1100 earlyprintk=dbgp[debugController#]
1101 earlyprintk=pciserial[,force],bus:device.function[,baudrate]
1102 earlyprintk=xdbc[xhciController#]
1104 earlyprintk is useful when the kernel crashes before
1105 the normal console is initialized. It is not enabled by
1106 default because it has some cosmetic problems.
1108 Append ",keep" to not disable it when the real console
1111 Only one of vga, efi, serial, or usb debug port can
1114 Currently only ttyS0 and ttyS1 may be specified by
1115 name. Other I/O ports may be explicitly specified
1116 on some architectures (x86 and arm at least) by
1117 replacing ttySn with an I/O port address, like this:
1118 earlyprintk=serial,0x1008,115200
1119 You can find the port for a given device in
1120 /proc/tty/driver/serial:
1121 2: uart:ST16650V2 port:00001008 irq:18 ...
1123 Interaction with the standard serial driver is not
1126 The VGA and EFI output is eventually overwritten by
1129 The xen output can only be used by Xen PV guests.
1131 The sclp output can only be used on s390.
1133 The optional "force" to "pciserial" enables use of a
1134 PCI device even when its classcode is not of the
1137 edac_report= [HW,EDAC] Control how to report EDAC event
1138 Format: {"on" | "off" | "force"}
1139 on: enable EDAC to report H/W event. May be overridden
1140 by other higher priority error reporting module.
1141 off: disable H/W event reporting through EDAC.
1142 force: enforce the use of EDAC to report H/W event.
1145 ekgdboc= [X86,KGDB] Allow early kernel console debugging
1148 This is designed to be used in conjunction with
1149 the boot argument: earlyprintk=vga
1152 Format: {"off" | "on" | "skip[mbr]"}
1155 Format: { "old_map", "nochunk", "noruntime", "debug" }
1156 old_map [X86-64]: switch to the old ioremap-based EFI
1157 runtime services mapping. 32-bit still uses this one by
1159 nochunk: disable reading files in "chunks" in the EFI
1160 boot stub, as chunking can cause problems with some
1161 firmware implementations.
1162 noruntime : disable EFI runtime services support
1163 debug: enable misc debug output
1165 efi_no_storage_paranoia [EFI; X86]
1166 Using this parameter you can use more than 50% of
1167 your efi variable storage. Use this parameter only if
1168 you are really sure that your UEFI does sane gc and
1169 fulfills the spec otherwise your board may brick.
1171 efi_fake_mem= nn[KMG]@ss[KMG]:aa[,nn[KMG]@ss[KMG]:aa,..] [EFI; X86]
1172 Add arbitrary attribute to specific memory range by
1173 updating original EFI memory map.
1174 Region of memory which aa attribute is added to is
1176 If efi_fake_mem=2G@4G:0x10000,2G@0x10a0000000:0x10000
1177 is specified, EFI_MEMORY_MORE_RELIABLE(0x10000)
1178 attribute is added to range 0x100000000-0x180000000 and
1179 0x10a0000000-0x1120000000.
1181 Using this parameter you can do debugging of EFI memmap
1182 related feature. For example, you can do debugging of
1183 Address Range Mirroring feature even if your box
1186 efivar_ssdt= [EFI; X86] Name of an EFI variable that contains an SSDT
1187 that is to be dynamically loaded by Linux. If there are
1188 multiple variables with the same name but with different
1189 vendor GUIDs, all of them will be loaded. See
1190 Documentation/admin-guide/acpi/ssdt-overlays.rst for details.
1193 eisa_irq_edge= [PARISC,HW]
1194 See header of drivers/parisc/eisa.c.
1197 See comment before function elanfreq_setup() in
1198 arch/x86/kernel/cpu/cpufreq/elanfreq.c.
1201 Format: { "mq-deadline" | "kyber" | "bfq" }
1202 See Documentation/block/deadline-iosched.rst,
1203 Documentation/block/kyber-iosched.rst and
1204 Documentation/block/bfq-iosched.rst for details.
1206 elfcorehdr=[size[KMG]@]offset[KMG] [IA64,PPC,SH,X86,S390]
1207 Specifies physical address of start of kernel core
1208 image elf header and optionally the size. Generally
1209 kexec loader will pass this option to capture kernel.
1210 See Documentation/admin-guide/kdump/kdump.rst for details.
1212 enable_mtrr_cleanup [X86]
1213 The kernel tries to adjust MTRR layout from continuous
1214 to discrete, to make X server driver able to add WB
1215 entry later. This parameter enables that.
1217 enable_timer_pin_1 [X86]
1218 Enable PIN 1 of APIC timer
1219 Can be useful to work around chipset bugs
1220 (in particular on some ATI chipsets).
1221 The kernel tries to set a reasonable default.
1223 enforcing [SELINUX] Set initial enforcing status.
1225 See security/selinux/Kconfig help text.
1226 0 -- permissive (log only, no denials).
1227 1 -- enforcing (deny and log).
1229 Value can be changed at runtime via /selinux/enforce.
1232 Disable Error Record Serialization Table (ERST)
1235 ether= [HW,NET] Ethernet cards parameters
1236 This option is obsoleted by the "netdev=" option, which
1237 has equivalent usage. See its documentation for details.
1241 Permit 'security.evm' to be updated regardless of
1242 current integrity status.
1246 fail_make_request=[KNL]
1247 General fault injection mechanism.
1248 Format: <interval>,<probability>,<space>,<times>
1249 See also Documentation/fault-injection/.
1252 See Documentation/admin-guide/blockdev/floppy.rst.
1254 force_pal_cache_flush
1255 [IA-64] Avoid check_sal_cache_flush which may hang on
1256 buggy SAL_CACHE_FLUSH implementations. Using this
1257 parameter will force ia64_sal_cache_flush to call
1258 ia64_pal_cache_flush instead of SAL_CACHE_FLUSH.
1261 Forcefully enable Physical Address Extension (PAE).
1262 Many Pentium M systems disable PAE but may have a
1263 functionally usable PAE implementation.
1264 Warning: use of this parameter will taint the kernel
1265 and may cause unknown problems.
1268 [FTRACE] will set and start the specified tracer
1269 as early as possible in order to facilitate early
1272 ftrace_dump_on_oops[=orig_cpu]
1273 [FTRACE] will dump the trace buffers on oops.
1274 If no parameter is passed, ftrace will dump
1275 buffers of all CPUs, but if you pass orig_cpu, it will
1276 dump only the buffer of the CPU that triggered the
1279 ftrace_filter=[function-list]
1280 [FTRACE] Limit the functions traced by the function
1281 tracer at boot up. function-list is a comma separated
1282 list of functions. This list can be changed at run
1283 time by the set_ftrace_filter file in the debugfs
1286 ftrace_notrace=[function-list]
1287 [FTRACE] Do not trace the functions specified in
1288 function-list. This list can be changed at run time
1289 by the set_ftrace_notrace file in the debugfs
1292 ftrace_graph_filter=[function-list]
1293 [FTRACE] Limit the top level callers functions traced
1294 by the function graph tracer at boot up.
1295 function-list is a comma separated list of functions
1296 that can be changed at run time by the
1297 set_graph_function file in the debugfs tracing directory.
1299 ftrace_graph_notrace=[function-list]
1300 [FTRACE] Do not trace from the functions specified in
1301 function-list. This list is a comma separated list of
1302 functions that can be changed at run time by the
1303 set_graph_notrace file in the debugfs tracing directory.
1305 ftrace_graph_max_depth=<uint>
1306 [FTRACE] Used with the function graph tracer. This is
1307 the max depth it will trace into a function. This value
1308 can be changed at run time by the max_graph_depth file
1309 in the tracefs tracing directory. default: 0 (no limit)
1312 [HW,JOY] Multisystem joystick and NES/SNES/PSX pad
1313 support via parallel port (up to 5 devices per port)
1314 Format: <port#>,<pad1>,<pad2>,<pad3>,<pad4>,<pad5>
1315 See also Documentation/input/devices/joystick-parport.rst
1319 gart_fix_e820= [X86_64] disable the fix e820 for K8 GART
1323 gcov_persist= [GCOV] When non-zero (default), profiling data for
1324 kernel modules is saved and remains accessible via
1325 debugfs, even when the module is unloaded/reloaded.
1326 When zero, profiling data is discarded and associated
1327 debugfs files are removed at module unload time.
1329 goldfish [X86] Enable the goldfish android emulator platform.
1330 Don't use this when you are not running on the
1333 gpt [EFI] Forces disk with valid GPT signature but
1334 invalid Protective MBR to be treated as GPT. If the
1335 primary GPT is corrupted, it enables the backup/alternate
1336 GPT to be used instead.
1338 grcan.enable0= [HW] Configuration of physical interface 0. Determines
1339 the "Enable 0" bit of the configuration register.
1342 grcan.enable1= [HW] Configuration of physical interface 1. Determines
1343 the "Enable 0" bit of the configuration register.
1346 grcan.select= [HW] Select which physical interface to use.
1349 grcan.txsize= [HW] Sets the size of the tx buffer.
1350 Format: <unsigned int> such that (txsize & ~0x1fffc0) == 0.
1352 grcan.rxsize= [HW] Sets the size of the rx buffer.
1353 Format: <unsigned int> such that (rxsize & ~0x1fffc0) == 0.
1356 gpio-mockup.gpio_mockup_ranges
1357 [HW] Sets the ranges of gpiochip of for this device.
1358 Format: <start1>,<end1>,<start2>,<end2>...
1360 hardlockup_all_cpu_backtrace=
1361 [KNL] Should the hard-lockup detector generate
1362 backtraces on all cpus.
1365 hashdist= [KNL,NUMA] Large hashes allocated during boot
1366 are distributed across NUMA nodes. Defaults on
1367 for 64-bit NUMA, off otherwise.
1368 Format: 0 | 1 (for off | on)
1370 hcl= [IA-64] SGI's Hardware Graph compatibility layer
1372 hd= [EIDE] (E)IDE hard drive subsystem geometry
1373 Format: <cyl>,<head>,<sect>
1376 Disable Hardware Error Source Table (HEST) support;
1377 corresponding firmware-first mode error processing
1378 logic will be disabled.
1380 highmem=nn[KMG] [KNL,BOOT] forces the highmem zone to have an exact
1381 size of <nn>. This works even on boxes that have no
1382 highmem otherwise. This also works to reduce highmem
1383 size on bigger boxes.
1385 highres= [KNL] Enable/disable high resolution timer mode.
1386 Valid parameters: "on", "off"
1391 hpet= [X86-32,HPET] option to control HPET usage
1392 Format: { enable (default) | disable | force |
1394 disable: disable HPET and use PIT instead
1395 force: allow force enabled of undocumented chips (ICH4,
1397 verbose: show contents of HPET registers during setup
1399 hpet_mmap= [X86, HPET_MMAP] Allow userspace to mmap HPET
1400 registers. Default set by CONFIG_HPET_MMAP_DEFAULT.
1402 hugepages= [HW,X86-32,IA-64] HugeTLB pages to allocate at boot.
1403 hugepagesz= [HW,IA-64,PPC,X86-64] The size of the HugeTLB pages.
1404 On x86-64 and powerpc, this option can be specified
1405 multiple times interleaved with hugepages= to reserve
1406 huge pages of different sizes. Valid pages sizes on
1407 x86-64 are 2M (when the CPU supports "pse") and 1G
1408 (when the CPU supports the "pdpe1gb" cpuinfo flag).
1411 [KNL] Should the hung task detector generate panics.
1414 A nonzero value instructs the kernel to panic when a
1415 hung task is detected. The default value is controlled
1416 by the CONFIG_BOOTPARAM_HUNG_TASK_PANIC build-time
1417 option. The value selected by this boot parameter can
1418 be changed later by the kernel.hung_task_panic sysctl.
1420 hvc_iucv= [S390] Number of z/VM IUCV hypervisor console (HVC)
1421 terminal devices. Valid values: 0..8
1422 hvc_iucv_allow= [S390] Comma-separated list of z/VM user IDs.
1423 If specified, z/VM IUCV HVC accepts connections
1424 from listed z/VM user IDs only.
1426 hv_nopvspin [X86,HYPER_V] Disables the paravirt spinlock optimizations
1427 which allow the hypervisor to 'idle' the
1428 guest on lock contention.
1431 Do not unregister boot console at start. This is only
1432 useful for debugging when something happens in the window
1433 between unregistering the boot console and initializing
1436 i2c_bus= [HW] Override the default board specific I2C bus speed
1437 or register an additional I2C bus that is not
1438 registered from board initialization code.
1442 i8042.debug [HW] Toggle i8042 debug mode
1443 i8042.unmask_kbd_data
1444 [HW] Enable printing of interrupt data from the KBD port
1445 (disabled by default, and as a pre-condition
1446 requires that i8042.debug=1 be enabled)
1447 i8042.direct [HW] Put keyboard port into non-translated mode
1448 i8042.dumbkbd [HW] Pretend that controller can only read data from
1449 keyboard and cannot control its state
1450 (Don't attempt to blink the leds)
1451 i8042.noaux [HW] Don't check for auxiliary (== mouse) port
1452 i8042.nokbd [HW] Don't check/create keyboard port
1453 i8042.noloop [HW] Disable the AUX Loopback command while probing
1455 i8042.nomux [HW] Don't check presence of an active multiplexing
1457 i8042.nopnp [HW] Don't use ACPIPnP / PnPBIOS to discover KBD/AUX
1459 i8042.notimeout [HW] Ignore timeout condition signalled by controller
1460 i8042.reset [HW] Reset the controller during init, cleanup and
1461 suspend-to-ram transitions, only during s2r
1462 transitions, or never reset
1463 Format: { 1 | Y | y | 0 | N | n }
1464 1, Y, y: always reset controller
1465 0, N, n: don't ever reset controller
1466 Default: only on s2r transitions on x86; most other
1467 architectures force reset to be always executed
1468 i8042.unlock [HW] Unlock (ignore) the keylock
1469 i8042.kbdreset [HW] Reset device connected to KBD port
1473 i8k.ignore_dmi [HW] Continue probing hardware even if DMI data
1474 indicates that the driver is running on unsupported
1476 i8k.force [HW] Activate i8k driver even if SMM BIOS signature
1477 does not match list of supported models.
1479 [HW] Report power status in /proc/i8k
1480 (disabled by default)
1481 i8k.restricted [HW] Allow controlling fans only if SYS_ADMIN
1484 i915.invert_brightness=
1485 [DRM] Invert the sense of the variable that is used to
1486 set the brightness of the panel backlight. Normally a
1487 brightness value of 0 indicates backlight switched off,
1488 and the maximum of the brightness value sets the backlight
1489 to maximum brightness. If this parameter is set to 0
1490 (default) and the machine requires it, or this parameter
1491 is set to 1, a brightness value of 0 sets the backlight
1492 to maximum brightness, and the maximum of the brightness
1493 value switches the backlight off.
1494 -1 -- never invert brightness
1495 0 -- machine default
1496 1 -- force brightness inversion
1499 Format: <io>[,<membase>[,<icn_id>[,<icn_id2>]]]
1501 ide-core.nodma= [HW] (E)IDE subsystem
1502 Format: =0.0 to prevent dma on hda, =0.1 hdb =1.0 hdc
1503 .vlb_clock .pci_clock .noflush .nohpa .noprobe .nowerr
1504 .cdrom .chs .ignore_cable are additional options
1505 See Documentation/ide/ide.rst.
1507 ide-generic.probe-mask= [HW] (E)IDE subsystem
1509 Probe mask for legacy ISA IDE ports. Depending on
1510 platform up to 6 ports are supported, enabled by
1511 setting corresponding bits in the mask to 1. The
1512 default value is 0x0, which has a special meaning.
1513 On systems that have PCI, it triggers scanning the
1514 PCI bus for the first and the second port, which
1515 are then probed. On systems without PCI the value
1516 of 0x0 enables probing the two first ports as if it
1519 ide-pci-generic.all-generic-ide [HW] (E)IDE subsystem
1520 Claim all unknown PCI IDE storage controllers.
1523 Format: idle=poll, idle=halt, idle=nomwait
1524 Poll forces a polling idle loop that can slightly
1525 improve the performance of waking up a idle CPU, but
1526 will use a lot of power and make the system run hot.
1528 idle=halt: Halt is forced to be used for CPU idle.
1529 In such case C2/C3 won't be used again.
1530 idle=nomwait: Disable mwait for CPU C-states
1532 ieee754= [MIPS] Select IEEE Std 754 conformance mode
1533 Format: { strict | legacy | 2008 | relaxed }
1536 Choose which programs will be accepted for execution
1537 based on the IEEE 754 NaN encoding(s) supported by
1538 the FPU and the NaN encoding requested with the value
1539 of an ELF file header flag individually set by each
1540 binary. Hardware implementations are permitted to
1541 support either or both of the legacy and the 2008 NaN
1544 Available settings are as follows:
1545 strict accept binaries that request a NaN encoding
1546 supported by the FPU
1547 legacy only accept legacy-NaN binaries, if supported
1549 2008 only accept 2008-NaN binaries, if supported
1551 relaxed accept any binaries regardless of whether
1552 supported by the FPU
1554 The FPU emulator is always able to support both NaN
1555 encodings, so if no FPU hardware is present or it has
1556 been disabled with 'nofpu', then the settings of
1557 'legacy' and '2008' strap the emulator accordingly,
1558 'relaxed' straps the emulator for both legacy-NaN and
1559 2008-NaN, whereas 'strict' enables legacy-NaN only on
1560 legacy processors and both NaN encodings on MIPS32 or
1563 The setting for ABS.fmt/NEG.fmt instruction execution
1564 mode generally follows that for the NaN encoding,
1565 except where unsupported by hardware.
1567 ignore_loglevel [KNL]
1568 Ignore loglevel setting - this will print /all/
1569 kernel messages to the console. Useful for debugging.
1570 We also add it as printk module parameter, so users
1571 could change it dynamically, usually by
1572 /sys/module/printk/parameters/ignore_loglevel.
1575 Ignore RLIMIT_DATA setting for data mappings,
1576 print warning at first misuse. Can be changed via
1577 /sys/module/kernel/parameters/ignore_rlimit_data.
1579 ihash_entries= [KNL]
1580 Set number of hash buckets for inode cache.
1582 ima_appraise= [IMA] appraise integrity measurements
1583 Format: { "off" | "enforce" | "fix" | "log" }
1586 ima_appraise_tcb [IMA] Deprecated. Use ima_policy= instead.
1587 The builtin appraise policy appraises all files
1590 ima_canonical_fmt [IMA]
1591 Use the canonical format for the binary runtime
1592 measurements, instead of host native format.
1595 Format: { md5 | sha1 | rmd160 | sha256 | sha384
1599 The list of supported hash algorithms is defined
1600 in crypto/hash_info.h.
1603 The builtin policies to load during IMA setup.
1604 Format: "tcb | appraise_tcb | secure_boot |
1607 The "tcb" policy measures all programs exec'd, files
1608 mmap'd for exec, and all files opened with the read
1609 mode bit set by either the effective uid (euid=0) or
1612 The "appraise_tcb" policy appraises the integrity of
1613 all files owned by root.
1615 The "secure_boot" policy appraises the integrity
1616 of files (eg. kexec kernel image, kernel modules,
1617 firmware, policy, etc) based on file signatures.
1619 The "fail_securely" policy forces file signature
1620 verification failure also on privileged mounted
1621 filesystems with the SB_I_UNVERIFIABLE_SIGNATURE
1624 ima_tcb [IMA] Deprecated. Use ima_policy= instead.
1625 Load a policy which meets the needs of the Trusted
1626 Computing Base. This means IMA will measure all
1627 programs exec'd, files mmap'd for exec, and all files
1628 opened for read by uid=0.
1631 Select one of defined IMA measurements template formats.
1632 Formats: { "ima" | "ima-ng" | "ima-sig" }
1636 [IMA] Define a custom template format.
1637 Format: { "field1|...|fieldN" }
1639 ima.ahash_minsize= [IMA] Minimum file size for asynchronous hash usage
1640 Format: <min_file_size>
1641 Set the minimal file size for using asynchronous hash.
1642 If left unspecified, ahash usage is disabled.
1644 ahash performance varies for different data sizes on
1645 different crypto accelerators. This option can be used
1646 to achieve the best performance for a particular HW.
1648 ima.ahash_bufsize= [IMA] Asynchronous hash buffer size
1650 Set hashing buffer size. Default: 4k.
1652 ahash performance varies for different chunk sizes on
1653 different crypto accelerators. This option can be used
1654 to achieve best performance for particular HW.
1658 Run specified binary instead of /sbin/init as init
1661 initcall_debug [KNL] Trace initcalls as they are executed. Useful
1662 for working out where the kernel is dying during
1665 initcall_blacklist= [KNL] Do not execute a comma-separated list of
1666 initcall functions. Useful for debugging built-in
1667 modules and initcalls.
1669 initrd= [BOOT] Specify the location of the initial ramdisk
1671 init_on_alloc= [MM] Fill newly allocated pages and heap objects with
1674 Default set by CONFIG_INIT_ON_ALLOC_DEFAULT_ON.
1676 init_on_free= [MM] Fill freed pages and heap objects with zeroes.
1678 Default set by CONFIG_INIT_ON_FREE_DEFAULT_ON.
1680 init_pkru= [x86] Specify the default memory protection keys rights
1681 register contents for all processes. 0x55555554 by
1682 default (disallow access to all but pkey 0). Can
1683 override in debugfs after boot.
1685 inport.irq= [HW] Inport (ATI XL and Microsoft) busmouse driver
1688 int_pln_enable [x86] Enable power limit notification interrupt
1690 integrity_audit=[IMA]
1691 Format: { "0" | "1" }
1692 0 -- basic integrity auditing messages. (Default)
1693 1 -- additional integrity auditing messages.
1695 intel_iommu= [DMAR] Intel IOMMU driver (DMAR) option
1697 Enable intel iommu driver.
1699 Disable intel iommu driver.
1700 igfx_off [Default Off]
1701 By default, gfx is mapped as normal device. If a gfx
1702 device has a dedicated DMAR unit, the DMAR unit is
1703 bypassed by not enabling DMAR with this option. In
1704 this case, gfx device will use physical address for
1707 With this option iommu will not optimize to look
1708 for io virtual address below 32-bit forcing dual
1709 address cycle on pci bus for cards supporting greater
1710 than 32-bit addressing. The default is to look
1711 for translation below 32-bit and if not available
1712 then look in the higher range.
1713 strict [Default Off]
1714 With this option on every unmap_single operation will
1715 result in a hardware IOTLB flush operation as opposed
1716 to batching them for performance.
1717 sp_off [Default Off]
1718 By default, super page will be supported if Intel IOMMU
1719 has the capability. With this option, super page will
1722 By default, scalable mode will be disabled even if the
1723 hardware advertises that it has support for the scalable
1724 mode translation. With this option set, scalable mode
1725 will be used on hardware which claims to support it.
1726 tboot_noforce [Default Off]
1727 Do not force the Intel IOMMU enabled under tboot.
1728 By default, tboot will force Intel IOMMU on, which
1729 could harm performance of some high-throughput
1730 devices like 40GBit network cards, even if identity
1732 Note that using this option lowers the security
1733 provided by tboot because it makes the system
1734 vulnerable to DMA attacks.
1736 intel_idle.max_cstate= [KNL,HW,ACPI,X86]
1737 0 disables intel_idle and fall back on acpi_idle.
1738 1 to 9 specify maximum depth of C-state.
1742 Do not enable intel_pstate as the default
1743 scaling driver for the supported processors
1745 Use intel_pstate as a scaling driver, but configure it
1746 to work with generic cpufreq governors (instead of
1747 enabling its internal governor). This mode cannot be
1748 used along with the hardware-managed P-states (HWP)
1751 Enable intel_pstate on systems that prohibit it by default
1752 in favor of acpi-cpufreq. Forcing the intel_pstate driver
1753 instead of acpi-cpufreq may disable platform features, such
1754 as thermal controls and power capping, that rely on ACPI
1755 P-States information being indicated to OSPM and therefore
1756 should be used with caution. This option does not work with
1757 processors that aren't supported by the intel_pstate driver
1758 or on platforms that use pcc-cpufreq instead of acpi-cpufreq.
1760 Do not enable hardware P state control (HWP)
1763 Only load intel_pstate on systems which support
1764 hardware P state control (HWP) if available.
1766 Enforce ACPI _PPC performance limits. If the Fixed ACPI
1767 Description Table, specifies preferred power management
1768 profile as "Enterprise Server" or "Performance Server",
1769 then this feature is turned on by default.
1771 Allow per-logical-CPU P-State performance control limits using
1772 cpufreq sysfs interface
1774 intremap= [X86-64, Intel-IOMMU]
1775 on enable Interrupt Remapping (default)
1776 off disable Interrupt Remapping
1777 nosid disable Source ID checking
1779 BIOS x2APIC opt-out request will be ignored
1780 nopost disable Interrupt Posting
1782 iomem= Disable strict checking of access to MMIO memory
1783 strict regions from userspace.
1798 nobypass [PPC/POWERNV]
1799 Disable IOMMU bypass, using IOMMU for PCI devices.
1801 iommu.strict= [ARM64] Configure TLB invalidation behaviour
1802 Format: { "0" | "1" }
1804 Request that DMA unmap operations use deferred
1805 invalidation of hardware TLBs, for increased
1806 throughput at the cost of reduced device isolation.
1807 Will fall back to strict mode if not supported by
1808 the relevant IOMMU driver.
1809 1 - Strict mode (default).
1810 DMA unmap operations invalidate IOMMU hardware TLBs
1814 [ARM64] Configure DMA to bypass the IOMMU by default.
1815 Format: { "0" | "1" }
1816 0 - Use IOMMU translation for DMA.
1817 1 - Bypass the IOMMU for DMA.
1818 unset - Use value of CONFIG_IOMMU_DEFAULT_PASSTHROUGH.
1820 io7= [HW] IO7 for Marvel based alpha systems
1821 See comment before marvel_specify_io7 in
1822 arch/alpha/kernel/core_marvel.c.
1824 io_delay= [X86] I/O delay method
1826 Standard port 0x80 based delay
1828 Alternate port 0xed based delay (needed on some systems)
1830 Simple two microseconds delay
1835 See Documentation/filesystems/nfs/nfsroot.txt.
1837 ipcmni_extend [KNL] Extend the maximum number of unique System V
1838 IPC identifiers from 32,768 to 16,777,216.
1840 irqaffinity= [SMP] Set the default irq affinity mask
1841 The argument is a cpu list, as described above.
1843 irqchip.gicv2_force_probe=
1846 Force the kernel to look for the second 4kB page
1847 of a GICv2 controller even if the memory range
1848 exposed by the device tree is too small.
1850 irqchip.gicv3_nolpi=
1852 Force the kernel to ignore the availability of
1853 LPIs (and by consequence ITSs). Intended for system
1854 that use the kernel as a bootloader, and thus want
1855 to let secondary kernels in charge of setting up
1858 irqchip.gicv3_pseudo_nmi= [ARM64]
1859 Enables support for pseudo-NMIs in the kernel. This
1860 requires the kernel to be built with
1861 CONFIG_ARM64_PSEUDO_NMI.
1864 When an interrupt is not handled search all handlers
1865 for it. Intended to get systems with badly broken
1869 When an interrupt is not handled search all handlers
1870 for it. Also check all handlers each timer
1871 interrupt. Intended to get systems with badly broken
1875 Format: <RDP>,<reset>,<pci_scan>,<verbosity>
1877 isolcpus= [KNL,SMP,ISOL] Isolate a given set of CPUs from disturbance.
1878 [Deprecated - use cpusets instead]
1879 Format: [flag-list,]<cpu-list>
1881 Specify one or more CPUs to isolate from disturbances
1882 specified in the flag list (default: domain):
1885 Disable the tick when a single task runs.
1887 A residual 1Hz tick is offloaded to workqueues, which you
1888 need to affine to housekeeping through the global
1889 workqueue's affinity configured via the
1890 /sys/devices/virtual/workqueue/cpumask sysfs file, or
1891 by using the 'domain' flag described below.
1893 NOTE: by default the global workqueue runs on all CPUs,
1894 so to protect individual CPUs the 'cpumask' file has to
1895 be configured manually after bootup.
1898 Isolate from the general SMP balancing and scheduling
1899 algorithms. Note that performing domain isolation this way
1900 is irreversible: it's not possible to bring back a CPU to
1901 the domains once isolated through isolcpus. It's strongly
1902 advised to use cpusets instead to disable scheduler load
1903 balancing through the "cpuset.sched_load_balance" file.
1904 It offers a much more flexible interface where CPUs can
1905 move in and out of an isolated set anytime.
1907 You can move a process onto or off an "isolated" CPU via
1908 the CPU affinity syscalls or cpuset.
1909 <cpu number> begins at 0 and the maximum value is
1910 "number of CPUs in system - 1".
1912 The format of <cpu-list> is described above.
1918 ivrs_ioapic [HW,X86_64]
1919 Provide an override to the IOAPIC-ID<->DEVICE-ID
1920 mapping provided in the IVRS ACPI table. For
1921 example, to map IOAPIC-ID decimal 10 to
1922 PCI device 00:14.0 write the parameter as:
1923 ivrs_ioapic[10]=00:14.0
1925 ivrs_hpet [HW,X86_64]
1926 Provide an override to the HPET-ID<->DEVICE-ID
1927 mapping provided in the IVRS ACPI table. For
1928 example, to map HPET-ID decimal 0 to
1929 PCI device 00:14.0 write the parameter as:
1930 ivrs_hpet[0]=00:14.0
1932 ivrs_acpihid [HW,X86_64]
1933 Provide an override to the ACPI-HID:UID<->DEVICE-ID
1934 mapping provided in the IVRS ACPI table. For
1935 example, to map UART-HID:UID AMD0020:0 to
1936 PCI device 00:14.5 write the parameter as:
1937 ivrs_acpihid[00:14.5]=AMD0020:0
1939 js= [HW,JOY] Analog joystick
1940 See Documentation/input/joydev/joystick.rst.
1943 When CONFIG_RANDOMIZE_BASE is set, this disables
1944 kernel and module base offset ASLR (Address Space
1945 Layout Randomization).
1948 [KNL] Enforce KASAN (Kernel Address Sanitizer) to print
1949 report on every invalid memory access. Without this
1950 parameter KASAN will print report only for the first
1955 kernelcore= [KNL,X86,IA-64,PPC]
1956 Format: nn[KMGTPE] | nn% | "mirror"
1957 This parameter specifies the amount of memory usable by
1958 the kernel for non-movable allocations. The requested
1959 amount is spread evenly throughout all nodes in the
1960 system as ZONE_NORMAL. The remaining memory is used for
1961 movable memory in its own zone, ZONE_MOVABLE. In the
1962 event, a node is too small to have both ZONE_NORMAL and
1963 ZONE_MOVABLE, kernelcore memory will take priority and
1964 other nodes will have a larger ZONE_MOVABLE.
1966 ZONE_MOVABLE is used for the allocation of pages that
1967 may be reclaimed or moved by the page migration
1968 subsystem. Note that allocations like PTEs-from-HighMem
1969 still use the HighMem zone if it exists, and the Normal
1970 zone if it does not.
1972 It is possible to specify the exact amount of memory in
1973 the form of "nn[KMGTPE]", a percentage of total system
1974 memory in the form of "nn%", or "mirror". If "mirror"
1975 option is specified, mirrored (reliable) memory is used
1976 for non-movable allocations and remaining memory is used
1977 for Movable pages. "nn[KMGTPE]", "nn%", and "mirror"
1978 are exclusive, so you cannot specify multiple forms.
1980 kgdbdbgp= [KGDB,HW] kgdb over EHCI usb debug port.
1981 Format: <Controller#>[,poll interval]
1982 The controller # is the number of the ehci usb debug
1983 port as it is probed via PCI. The poll interval is
1984 optional and is the number seconds in between
1985 each poll cycle to the debug port in case you need
1986 the functionality for interrupting the kernel with
1987 gdb or control-c on the dbgp connection. When
1988 not using this parameter you use sysrq-g to break into
1989 the kernel debugger.
1991 kgdboc= [KGDB,HW] kgdb over consoles.
1992 Requires a tty driver that supports console polling,
1993 or a supported polling keyboard driver (non-usb).
1994 Serial only format: <serial_device>[,baud]
1995 keyboard only format: kbd
1996 keyboard and serial format: kbd,<serial_device>[,baud]
1997 Optional Kernel mode setting:
1998 kms, kbd format: kms,kbd
1999 kms, kbd and serial format: kms,kbd,<ser_dev>[,baud]
2001 kgdbwait [KGDB] Stop kernel execution and enter the
2002 kernel debugger at the earliest opportunity.
2004 kmac= [MIPS] korina ethernet MAC address.
2005 Configure the RouterBoard 532 series on-chip
2006 Ethernet adapter MAC address.
2008 kmemleak= [KNL] Boot-time kmemleak enable/disable
2009 Valid arguments: on, off
2011 Built with CONFIG_DEBUG_KMEMLEAK_DEFAULT_OFF=y,
2014 kprobe_event=[probe-list]
2015 [FTRACE] Add kprobe events and enable at boot time.
2016 The probe-list is a semicolon delimited list of probe
2017 definitions. Each definition is same as kprobe_events
2018 interface, but the parameters are comma delimited.
2019 For example, to add a kprobe event on vfs_read with
2020 arg1 and arg2, add to the command line;
2022 kprobe_event=p,vfs_read,$arg1,$arg2
2024 See also Documentation/trace/kprobetrace.rst "Kernel
2025 Boot Parameter" section.
2027 kpti= [ARM64] Control page table isolation of user
2028 and kernel address spaces.
2029 Default: enabled on cores which need mitigation.
2033 kvm.ignore_msrs=[KVM] Ignore guest accesses to unhandled MSRs.
2034 Default is 0 (don't ignore, but inject #GP)
2036 kvm.enable_vmware_backdoor=[KVM] Support VMware backdoor PV interface.
2037 Default is false (don't support).
2039 kvm.mmu_audit= [KVM] This is a R/W parameter which allows audit
2044 [KVM] Controls the software workaround for the
2045 X86_BUG_ITLB_MULTIHIT bug.
2046 force : Always deploy workaround.
2047 off : Never deploy workaround.
2048 auto : Deploy workaround based on the presence of
2049 X86_BUG_ITLB_MULTIHIT.
2053 If the software workaround is enabled for the host,
2054 guests do need not to enable it for nested guests.
2056 kvm.nx_huge_pages_recovery_ratio=
2057 [KVM] Controls how many 4KiB pages are periodically zapped
2058 back to huge pages. 0 disables the recovery, otherwise if
2059 the value is N KVM will zap 1/Nth of the 4KiB pages every
2060 minute. The default is 60.
2062 kvm-amd.nested= [KVM,AMD] Allow nested virtualization in KVM/SVM.
2063 Default is 1 (enabled)
2065 kvm-amd.npt= [KVM,AMD] Disable nested paging (virtualized MMU)
2067 Default is 1 (enabled) if in 64-bit or 32-bit PAE mode.
2069 kvm-arm.vgic_v3_group0_trap=
2070 [KVM,ARM] Trap guest accesses to GICv3 group-0
2073 kvm-arm.vgic_v3_group1_trap=
2074 [KVM,ARM] Trap guest accesses to GICv3 group-1
2077 kvm-arm.vgic_v3_common_trap=
2078 [KVM,ARM] Trap guest accesses to GICv3 common
2081 kvm-arm.vgic_v4_enable=
2082 [KVM,ARM] Allow use of GICv4 for direct injection of
2085 kvm-intel.ept= [KVM,Intel] Disable extended page tables
2086 (virtualized MMU) support on capable Intel chips.
2087 Default is 1 (enabled)
2089 kvm-intel.emulate_invalid_guest_state=
2090 [KVM,Intel] Enable emulation of invalid guest states
2091 Default is 0 (disabled)
2093 kvm-intel.flexpriority=
2094 [KVM,Intel] Disable FlexPriority feature (TPR shadow).
2095 Default is 1 (enabled)
2098 [KVM,Intel] Enable VMX nesting (nVMX).
2099 Default is 0 (disabled)
2101 kvm-intel.unrestricted_guest=
2102 [KVM,Intel] Disable unrestricted guest feature
2103 (virtualized real and unpaged mode) on capable
2104 Intel chips. Default is 1 (enabled)
2106 kvm-intel.vmentry_l1d_flush=[KVM,Intel] Mitigation for L1 Terminal Fault
2109 Valid arguments: never, cond, always
2111 always: L1D cache flush on every VMENTER.
2112 cond: Flush L1D on VMENTER only when the code between
2113 VMEXIT and VMENTER can leak host memory.
2114 never: Disables the mitigation
2116 Default is cond (do L1 cache flush in specific instances)
2118 kvm-intel.vpid= [KVM,Intel] Disable Virtual Processor Identification
2119 feature (tagged TLBs) on capable Intel chips.
2120 Default is 1 (enabled)
2122 l1tf= [X86] Control mitigation of the L1TF vulnerability on
2125 The kernel PTE inversion protection is unconditionally
2126 enabled and cannot be disabled.
2129 Provides all available mitigations for the
2130 L1TF vulnerability. Disables SMT and
2131 enables all mitigations in the
2132 hypervisors, i.e. unconditional L1D flush.
2134 SMT control and L1D flush control via the
2135 sysfs interface is still possible after
2136 boot. Hypervisors will issue a warning
2137 when the first VM is started in a
2138 potentially insecure configuration,
2139 i.e. SMT enabled or L1D flush disabled.
2142 Same as 'full', but disables SMT and L1D
2143 flush runtime control. Implies the
2144 'nosmt=force' command line option.
2145 (i.e. sysfs control of SMT is disabled.)
2148 Leaves SMT enabled and enables the default
2149 hypervisor mitigation, i.e. conditional
2152 SMT control and L1D flush control via the
2153 sysfs interface is still possible after
2154 boot. Hypervisors will issue a warning
2155 when the first VM is started in a
2156 potentially insecure configuration,
2157 i.e. SMT enabled or L1D flush disabled.
2161 Disables SMT and enables the default
2162 hypervisor mitigation.
2164 SMT control and L1D flush control via the
2165 sysfs interface is still possible after
2166 boot. Hypervisors will issue a warning
2167 when the first VM is started in a
2168 potentially insecure configuration,
2169 i.e. SMT enabled or L1D flush disabled.
2172 Same as 'flush', but hypervisors will not
2173 warn when a VM is started in a potentially
2174 insecure configuration.
2177 Disables hypervisor mitigations and doesn't
2179 It also drops the swap size and available
2180 RAM limit restriction on both hypervisor and
2185 For details see: Documentation/admin-guide/hw-vuln/l1tf.rst
2191 lapic [X86-32,APIC] Enable the local APIC even if BIOS
2194 lapic= [x86,APIC] "notscdeadline" Do not use TSC deadline
2195 value for LAPIC timer one-shot implementation. Default
2196 back to the programmable timer unit in the LAPIC.
2198 lapic_timer_c2_ok [X86,APIC] trust the local apic timer
2201 libata.dma= [LIBATA] DMA control
2202 libata.dma=0 Disable all PATA and SATA DMA
2203 libata.dma=1 PATA and SATA Disk DMA only
2204 libata.dma=2 ATAPI (CDROM) DMA only
2205 libata.dma=4 Compact Flash DMA only
2206 Combinations also work, so libata.dma=3 enables DMA
2207 for disks and CDROMs, but not CFs.
2209 libata.ignore_hpa= [LIBATA] Ignore HPA limit
2210 libata.ignore_hpa=0 keep BIOS limits (default)
2211 libata.ignore_hpa=1 ignore limits, using full disk
2213 libata.noacpi [LIBATA] Disables use of ACPI in libata suspend/resume
2217 libata.force= [LIBATA] Force configurations. The format is comma
2218 separated list of "[ID:]VAL" where ID is
2219 PORT[.DEVICE]. PORT and DEVICE are decimal numbers
2220 matching port, link or device. Basically, it matches
2221 the ATA ID string printed on console by libata. If
2222 the whole ID part is omitted, the last PORT and DEVICE
2223 values are used. If ID hasn't been specified yet, the
2224 configuration applies to all ports, links and devices.
2226 If only DEVICE is omitted, the parameter applies to
2227 the port and all links and devices behind it. DEVICE
2228 number of 0 either selects the first device or the
2229 first fan-out link behind PMP device. It does not
2230 select the host link. DEVICE number of 15 selects the
2231 host link and device attached to it.
2233 The VAL specifies the configuration to force. As long
2234 as there's no ambiguity shortcut notation is allowed.
2235 For example, both 1.5 and 1.5G would work for 1.5Gbps.
2236 The following configurations can be forced.
2238 * Cable type: 40c, 80c, short40c, unk, ign or sata.
2239 Any ID with matching PORT is used.
2241 * SATA link speed limit: 1.5Gbps or 3.0Gbps.
2243 * Transfer mode: pio[0-7], mwdma[0-4] and udma[0-7].
2244 udma[/][16,25,33,44,66,100,133] notation is also
2247 * [no]ncq: Turn on or off NCQ.
2249 * [no]ncqtrim: Turn off queued DSM TRIM.
2251 * nohrst, nosrst, norst: suppress hard, soft
2254 * rstonce: only attempt one reset during
2255 hot-unplug link recovery
2257 * dump_id: dump IDENTIFY data.
2259 * atapi_dmadir: Enable ATAPI DMADIR bridge support
2261 * disable: Disable this device.
2263 If there are multiple matching configurations changing
2264 the same attribute, the last one is used.
2266 memblock=debug [KNL] Enable memblock debug messages.
2268 load_ramdisk= [RAM] List of ramdisks to load from floppy
2269 See Documentation/admin-guide/blockdev/ramdisk.rst.
2271 lockd.nlm_grace_period=P [NFS] Assign grace period.
2274 lockd.nlm_tcpport=N [NFS] Assign TCP port.
2277 lockd.nlm_timeout=T [NFS] Assign timeout value.
2280 lockd.nlm_udpport=M [NFS] Assign UDP port.
2283 locktorture.nreaders_stress= [KNL]
2284 Set the number of locking read-acquisition kthreads.
2285 Defaults to being automatically set based on the
2286 number of online CPUs.
2288 locktorture.nwriters_stress= [KNL]
2289 Set the number of locking write-acquisition kthreads.
2291 locktorture.onoff_holdoff= [KNL]
2292 Set time (s) after boot for CPU-hotplug testing.
2294 locktorture.onoff_interval= [KNL]
2295 Set time (s) between CPU-hotplug operations, or
2296 zero to disable CPU-hotplug testing.
2298 locktorture.shuffle_interval= [KNL]
2299 Set task-shuffle interval (jiffies). Shuffling
2300 tasks allows some CPUs to go into dyntick-idle
2301 mode during the locktorture test.
2303 locktorture.shutdown_secs= [KNL]
2304 Set time (s) after boot system shutdown. This
2305 is useful for hands-off automated testing.
2307 locktorture.stat_interval= [KNL]
2308 Time (s) between statistics printk()s.
2310 locktorture.stutter= [KNL]
2311 Time (s) to stutter testing, for example,
2312 specifying five seconds causes the test to run for
2313 five seconds, wait for five seconds, and so on.
2314 This tests the locking primitive's ability to
2315 transition abruptly to and from idle.
2317 locktorture.torture_type= [KNL]
2318 Specify the locking implementation to test.
2320 locktorture.verbose= [KNL]
2321 Enable additional printk() statements.
2323 logibm.irq= [HW,MOUSE] Logitech Bus Mouse Driver
2326 loglevel= All Kernel Messages with a loglevel smaller than the
2327 console loglevel will be printed to the console. It can
2328 also be changed with klogd or other programs. The
2329 loglevels are defined as follows:
2331 0 (KERN_EMERG) system is unusable
2332 1 (KERN_ALERT) action must be taken immediately
2333 2 (KERN_CRIT) critical conditions
2334 3 (KERN_ERR) error conditions
2335 4 (KERN_WARNING) warning conditions
2336 5 (KERN_NOTICE) normal but significant condition
2337 6 (KERN_INFO) informational
2338 7 (KERN_DEBUG) debug-level messages
2340 log_buf_len=n[KMG] Sets the size of the printk ring buffer,
2341 in bytes. n must be a power of two and greater
2342 than the minimal size. The minimal size is defined
2343 by LOG_BUF_SHIFT kernel config parameter. There is
2344 also CONFIG_LOG_CPU_MAX_BUF_SHIFT config parameter
2345 that allows to increase the default size depending on
2346 the number of CPUs. See init/Kconfig for more details.
2348 logo.nologo [FB] Disables display of the built-in Linux logo.
2349 This may be used to provide more screen space for
2350 kernel log messages and is useful when debugging
2351 kernel boot problems.
2353 lp=0 [LP] Specify parallel ports to use, e.g,
2354 lp=port[,port...] lp=none,parport0 (lp0 not configured, lp1 uses
2355 lp=reset first parallel port). 'lp=0' disables the
2356 lp=auto printer driver. 'lp=reset' (which can be
2357 specified in addition to the ports) causes
2358 attached printers to be reset. Using
2359 lp=port1,port2,... specifies the parallel ports
2360 to associate lp devices with, starting with
2361 lp0. A port specification may be 'none' to skip
2362 that lp device, or a parport name such as
2363 'parport0'. Specifying 'lp=auto' instead of a
2364 port specification list means that device IDs
2365 from each port should be examined, to see if
2366 an IEEE 1284-compliant printer is attached; if
2367 so, the driver will manage that printer.
2368 See also header of drivers/char/lp.c.
2371 Sets loops_per_jiffy to given constant, thus avoiding
2372 time-consuming boot-time autodetection (up to 250 ms per
2373 CPU). 0 enables autodetection (default). To determine
2374 the correct value for your kernel, boot with normal
2375 autodetection and see what value is printed. Note that
2376 on SMP systems the preset will be applied to all CPUs,
2377 which is likely to cause problems if your CPUs need
2378 significantly divergent settings. An incorrect value
2379 will cause delays in the kernel to be wrong, leading to
2380 unpredictable I/O errors and other breakage. Although
2381 unlikely, in the extreme case this might damage your
2385 Format: <io>,<irq>,<dma>
2387 lsm.debug [SECURITY] Enable LSM initialization debugging output.
2390 [SECURITY] Choose order of LSM initialization. This
2391 overrides CONFIG_LSM, and the "security=" parameter.
2393 machvec= [IA-64] Force the use of a particular machine-vector
2394 (machvec) in a generic kernel.
2395 Example: machvec=hpzx1_swiotlb
2397 machtype= [Loongson] Share the same kernel image file between different
2399 Example: machtype=lemote-yeeloong-2f-7inch
2401 max_addr=nn[KMG] [KNL,BOOT,ia64] All physical memory greater
2402 than or equal to this physical address is ignored.
2404 maxcpus= [SMP] Maximum number of processors that an SMP kernel
2405 will bring up during bootup. maxcpus=n : n >= 0 limits
2406 the kernel to bring up 'n' processors. Surely after
2407 bootup you can bring up the other plugged cpu by executing
2408 "echo 1 > /sys/devices/system/cpu/cpuX/online". So maxcpus
2409 only takes effect during system bootup.
2410 While n=0 is a special case, it is equivalent to "nosmp",
2411 which also disables the IO APIC.
2413 max_loop= [LOOP] The number of loop block devices that get
2414 (loop.max_loop) unconditionally pre-created at init time. The default
2415 number is configured by BLK_DEV_LOOP_MIN_COUNT. Instead
2416 of statically allocating a predefined number, loop
2417 devices can be requested on-demand with the
2418 /dev/loop-control interface.
2420 mce [X86-32] Machine Check Exception
2422 mce=option [X86-64] See Documentation/x86/x86_64/boot-options.rst
2424 md= [HW] RAID subsystems devices and level
2425 See Documentation/admin-guide/md.rst.
2428 Format: <first>,<last>
2429 Specifies range of consoles to be captured by the MDA.
2432 Control mitigation for the Micro-architectural Data
2433 Sampling (MDS) vulnerability.
2435 Certain CPUs are vulnerable to an exploit against CPU
2436 internal buffers which can forward information to a
2437 disclosure gadget under certain conditions.
2439 In vulnerable processors, the speculatively
2440 forwarded data can be used in a cache side channel
2441 attack, to access data to which the attacker does
2442 not have direct access.
2444 This parameter controls the MDS mitigation. The
2447 full - Enable MDS mitigation on vulnerable CPUs
2448 full,nosmt - Enable MDS mitigation and disable
2449 SMT on vulnerable CPUs
2450 off - Unconditionally disable MDS mitigation
2452 Not specifying this option is equivalent to
2455 For details see: Documentation/admin-guide/hw-vuln/mds.rst
2457 mem=nn[KMG] [KNL,BOOT] Force usage of a specific amount of memory
2458 Amount of memory to be used when the kernel is not able
2459 to see the whole system memory or for test.
2460 [X86] Work as limiting max address. Use together
2461 with memmap= to avoid physical address space collisions.
2462 Without memmap= PCI devices could be placed at addresses
2463 belonging to unused RAM.
2465 mem=nopentium [BUGS=X86-32] Disable usage of 4MB pages for kernel
2469 [KNL,SH] Allow user to override the default size for
2470 per-device physically contiguous DMA buffers.
2472 memhp_default_state=online/offline
2473 [KNL] Set the initial state for the memory hotplug
2474 onlining policy. If not specified, the default value is
2475 set according to the
2476 CONFIG_MEMORY_HOTPLUG_DEFAULT_ONLINE kernel config
2478 See Documentation/admin-guide/mm/memory-hotplug.rst.
2480 memmap=exactmap [KNL,X86] Enable setting of an exact
2481 E820 memory map, as specified by the user.
2482 Such memmap=exactmap lines can be constructed based on
2483 BIOS output or other requirements. See the memmap=nn@ss
2486 memmap=nn[KMG]@ss[KMG]
2487 [KNL] Force usage of a specific region of memory.
2488 Region of memory to be used is from ss to ss+nn.
2489 If @ss[KMG] is omitted, it is equivalent to mem=nn[KMG],
2490 which limits max address to nn[KMG].
2491 Multiple different regions can be specified,
2494 memmap=100M@2G,100M#3G,1G!1024G
2496 memmap=nn[KMG]#ss[KMG]
2497 [KNL,ACPI] Mark specific memory as ACPI data.
2498 Region of memory to be marked is from ss to ss+nn.
2500 memmap=nn[KMG]$ss[KMG]
2501 [KNL,ACPI] Mark specific memory as reserved.
2502 Region of memory to be reserved is from ss to ss+nn.
2503 Example: Exclude memory from 0x18690000-0x1869ffff
2504 memmap=64K$0x18690000
2506 memmap=0x10000$0x18690000
2507 Some bootloaders may need an escape character before '$',
2508 like Grub2, otherwise '$' and the following number
2511 memmap=nn[KMG]!ss[KMG]
2512 [KNL,X86] Mark specific memory as protected.
2513 Region of memory to be used, from ss to ss+nn.
2514 The memory region may be marked as e820 type 12 (0xc)
2515 and is NVDIMM or ADR memory.
2517 memmap=<size>%<offset>-<oldtype>+<newtype>
2518 [KNL,ACPI] Convert memory within the specified region
2519 from <oldtype> to <newtype>. If "-<oldtype>" is left
2520 out, the whole region will be marked as <newtype>,
2521 even if previously unavailable. If "+<newtype>" is left
2522 out, matching memory will be removed. Types are
2523 specified as e820 types, e.g., 1 = RAM, 2 = reserved,
2524 3 = ACPI, 12 = PRAM.
2526 memory_corruption_check=0/1 [X86]
2527 Some BIOSes seem to corrupt the first 64k of
2528 memory when doing things like suspend/resume.
2529 Setting this option will scan the memory
2530 looking for corruption. Enabling this will
2531 both detect corruption and prevent the kernel
2532 from using the memory being corrupted.
2533 However, its intended as a diagnostic tool; if
2534 repeatable BIOS-originated corruption always
2535 affects the same memory, you can use memmap=
2536 to prevent the kernel from using that memory.
2538 memory_corruption_check_size=size [X86]
2539 By default it checks for corruption in the low
2540 64k, making this memory unavailable for normal
2541 use. Use this parameter to scan for
2542 corruption in more or less memory.
2544 memory_corruption_check_period=seconds [X86]
2545 By default it checks for corruption every 60
2546 seconds. Use this parameter to check at some
2547 other rate. 0 disables periodic checking.
2549 memtest= [KNL,X86,ARM,PPC] Enable memtest
2551 default : 0 <disable>
2552 Specifies the number of memtest passes to be
2553 performed. Each pass selects another test
2554 pattern from a given set of patterns. Memtest
2555 fills the memory with this pattern, validates
2556 memory contents and reserves bad memory
2557 regions that are detected.
2559 mem_encrypt= [X86-64] AMD Secure Memory Encryption (SME) control
2560 Valid arguments: on, off
2561 Default (depends on kernel configuration option):
2562 on (CONFIG_AMD_MEM_ENCRYPT_ACTIVE_BY_DEFAULT=y)
2563 off (CONFIG_AMD_MEM_ENCRYPT_ACTIVE_BY_DEFAULT=n)
2564 mem_encrypt=on: Activate SME
2565 mem_encrypt=off: Do not activate SME
2567 Refer to Documentation/virt/kvm/amd-memory-encryption.rst
2568 for details on when memory encryption can be activated.
2570 mem_sleep_default= [SUSPEND] Default system suspend mode:
2571 s2idle - Suspend-To-Idle
2572 shallow - Power-On Suspend or equivalent (if supported)
2573 deep - Suspend-To-RAM or equivalent (if supported)
2574 See Documentation/admin-guide/pm/sleep-states.rst.
2576 meye.*= [HW] Set MotionEye Camera parameters
2577 See Documentation/media/v4l-drivers/meye.rst.
2579 mfgpt_irq= [IA-32] Specify the IRQ to use for the
2580 Multi-Function General Purpose Timers on AMD Geode
2583 mfgptfix [X86-32] Fix MFGPT timers on AMD Geode platforms when
2584 the BIOS has incorrectly applied a workaround. TinyBIOS
2585 version 0.98 is known to be affected, 0.99 fixes the
2586 problem by letting the user disable the workaround.
2590 min_addr=nn[KMG] [KNL,BOOT,ia64] All physical memory below this
2591 physical address is ignored.
2593 mini2440= [ARM,HW,KNL]
2594 Format:[0..2][b][c][t]
2596 MINI2440 configuration specification:
2597 0 - The attached screen is the 3.5" TFT
2598 1 - The attached screen is the 7" TFT
2599 2 - The VGA Shield is attached (1024x768)
2600 Leaving out the screen size parameter will not load
2601 the TFT driver, and the framebuffer will be left
2603 b - Enable backlight. The TFT backlight pin will be
2604 linked to the kernel VESA blanking code and a GPIO
2605 LED. This parameter is not necessary when using the
2607 c - Enable the s3c camera interface.
2608 t - Reserved for enabling touchscreen support. The
2609 touchscreen support is not enabled in the mainstream
2610 kernel as of 2.6.30, a preliminary port can be found
2611 in the "bleeding edge" mini2440 support kernel at
2612 http://repo.or.cz/w/linux-2.6/mini2440.git
2615 [X86,PPC,S390,ARM64] Control optional mitigations for
2616 CPU vulnerabilities. This is a set of curated,
2617 arch-independent options, each of which is an
2618 aggregation of existing arch-specific options.
2621 Disable all optional CPU mitigations. This
2622 improves system performance, but it may also
2623 expose users to several CPU vulnerabilities.
2624 Equivalent to: nopti [X86,PPC]
2626 nospectre_v1 [X86,PPC]
2628 nospectre_v2 [X86,PPC,S390,ARM64]
2629 spectre_v2_user=off [X86]
2630 spec_store_bypass_disable=off [X86,PPC]
2631 ssbd=force-off [ARM64]
2634 tsx_async_abort=off [X86]
2635 kvm.nx_huge_pages=off [X86]
2638 This does not have any effect on
2639 kvm.nx_huge_pages when
2640 kvm.nx_huge_pages=force.
2643 Mitigate all CPU vulnerabilities, but leave SMT
2644 enabled, even if it's vulnerable. This is for
2645 users who don't want to be surprised by SMT
2646 getting disabled across kernel upgrades, or who
2647 have other ways of avoiding SMT-based attacks.
2648 Equivalent to: (default behavior)
2651 Mitigate all CPU vulnerabilities, disabling SMT
2652 if needed. This is for users who always want to
2653 be fully mitigated, even if it means losing SMT.
2654 Equivalent to: l1tf=flush,nosmt [X86]
2655 mds=full,nosmt [X86]
2656 tsx_async_abort=full,nosmt [X86]
2659 [KNL] When CONFIG_DEBUG_MEMORY_INIT is set, this
2660 parameter allows control of the logging verbosity for
2661 the additional memory initialisation checks. A value
2662 of 0 disables mminit logging and a level of 4 will
2663 log everything. Information is printed at KERN_DEBUG
2664 so loglevel=8 may also need to be specified.
2667 [KNL] When CONFIG_MODULE_SIG is set, this means that
2668 modules without (valid) signatures will fail to load.
2669 Note that if CONFIG_MODULE_SIG_FORCE is set, that
2670 is always true, so this option does nothing.
2672 module_blacklist= [KNL] Do not load a comma-separated list of
2673 modules. Useful for debugging problem modules.
2676 [MOUSE] Maximum time between finger touching and
2677 leaving touchpad surface for touch to be considered
2678 a tap and be reported as a left button click (for
2679 touchpads working in absolute mode only).
2681 mousedev.xres= [MOUSE] Horizontal screen resolution, used for devices
2682 reporting absolute coordinates, such as tablets
2683 mousedev.yres= [MOUSE] Vertical screen resolution, used for devices
2684 reporting absolute coordinates, such as tablets
2686 movablecore= [KNL,X86,IA-64,PPC]
2687 Format: nn[KMGTPE] | nn%
2688 This parameter is the complement to kernelcore=, it
2689 specifies the amount of memory used for migratable
2690 allocations. If both kernelcore and movablecore is
2691 specified, then kernelcore will be at *least* the
2692 specified value but may be more. If movablecore on its
2693 own is specified, the administrator must be careful
2694 that the amount of memory usable for all allocations
2697 movable_node [KNL] Boot-time switch to make hotplugable memory
2698 NUMA nodes to be movable. This means that the memory
2699 of such nodes will be usable only for movable
2700 allocations which rules out almost all kernel
2701 allocations. Use with caution!
2703 MTD_Partition= [MTD]
2704 Format: <name>,<region-number>,<size>,<offset>
2706 MTD_Region= [MTD] Format:
2707 <name>,<region-number>[,<base>,<size>,<buswidth>,<altbuswidth>]
2710 See drivers/mtd/cmdlinepart.c.
2712 multitce=off [PPC] This parameter disables the use of the pSeries
2713 firmware feature for updating multiple TCE entries
2716 onenand.bdry= [HW,MTD] Flex-OneNAND Boundary Configuration
2718 Format: [die0_boundary][,die0_lock][,die1_boundary][,die1_lock]
2720 boundary - index of last SLC block on Flex-OneNAND.
2721 The remaining blocks are configured as MLC blocks.
2722 lock - Configure if Flex-OneNAND boundary should be locked.
2723 Once locked, the boundary cannot be changed.
2724 1 indicates lock status, 0 indicates unlock status.
2727 ARM/S3C2412 JIVE boot control
2729 See arch/arm/mach-s3c2412/mach-jive.c
2731 mtouchusb.raw_coordinates=
2732 [HW] Make the MicroTouch USB driver use raw coordinates
2733 ('y', default) or cooked coordinates ('n')
2735 mtrr_chunk_size=nn[KMG] [X86]
2736 used for mtrr cleanup. It is largest continuous chunk
2737 that could hold holes aka. UC entries.
2739 mtrr_gran_size=nn[KMG] [X86]
2740 Used for mtrr cleanup. It is granularity of mtrr block.
2742 Large value could prevent small alignment from
2745 mtrr_spare_reg_nr=n [X86]
2747 Range: 0,7 : spare reg number
2749 Used for mtrr cleanup. It is spare mtrr entries number.
2750 Set to 2 or more if your graphical card needs more.
2752 n2= [NET] SDL Inc. RISCom/N2 synchronous serial card
2754 netdev= [NET] Network devices parameters
2755 Format: <irq>,<io>,<mem_start>,<mem_end>,<name>
2756 Note that mem_start is often overloaded to mean
2757 something different and driver-specific.
2758 This usage is only documented in each driver source
2762 [NETFILTER] Enable connection tracking flow accounting
2763 0 to disable accounting
2764 1 to enable accounting
2767 nfsaddrs= [NFS] Deprecated. Use ip= instead.
2768 See Documentation/filesystems/nfs/nfsroot.txt.
2770 nfsroot= [NFS] nfs root filesystem for disk-less boxes.
2771 See Documentation/filesystems/nfs/nfsroot.txt.
2773 nfsrootdebug [NFS] enable nfsroot debugging messages.
2774 See Documentation/filesystems/nfs/nfsroot.txt.
2776 nfs.callback_nr_threads=
2777 [NFSv4] set the total number of threads that the
2778 NFS client will assign to service NFSv4 callback
2781 nfs.callback_tcpport=
2782 [NFS] set the TCP port on which the NFSv4 callback
2783 channel should listen.
2786 [NFS] sets the pathname to the program which is used
2787 to update the NFS client cache entries.
2789 nfs.cache_getent_timeout=
2790 [NFS] sets the timeout after which an attempt to
2791 update a cache entry is deemed to have failed.
2793 nfs.idmap_cache_timeout=
2794 [NFS] set the maximum lifetime for idmapper cache
2798 [NFS] enable 64-bit inode numbers.
2799 If zero, the NFS client will fake up a 32-bit inode
2800 number for the readdir() and stat() syscalls instead
2801 of returning the full 64-bit number.
2802 The default is to return 64-bit inode numbers.
2804 nfs.max_session_cb_slots=
2805 [NFSv4.1] Sets the maximum number of session
2806 slots the client will assign to the callback
2807 channel. This determines the maximum number of
2808 callbacks the client will process in parallel for
2809 a particular server.
2811 nfs.max_session_slots=
2812 [NFSv4.1] Sets the maximum number of session slots
2813 the client will attempt to negotiate with the server.
2814 This limits the number of simultaneous RPC requests
2815 that the client can send to the NFSv4.1 server.
2816 Note that there is little point in setting this
2817 value higher than the max_tcp_slot_table_limit.
2819 nfs.nfs4_disable_idmapping=
2820 [NFSv4] When set to the default of '1', this option
2821 ensures that both the RPC level authentication
2822 scheme and the NFS level operations agree to use
2823 numeric uids/gids if the mount is using the
2824 'sec=sys' security flavour. In effect it is
2825 disabling idmapping, which can make migration from
2826 legacy NFSv2/v3 systems to NFSv4 easier.
2827 Servers that do not support this mode of operation
2828 will be autodetected by the client, and it will fall
2829 back to using the idmapper.
2830 To turn off this behaviour, set the value to '0'.
2832 [NFS4] Specify an additional fixed unique ident-
2833 ification string that NFSv4 clients can insert into
2834 their nfs_client_id4 string. This is typically a
2835 UUID that is generated at system install time.
2837 nfs.send_implementation_id =
2838 [NFSv4.1] Send client implementation identification
2839 information in exchange_id requests.
2840 If zero, no implementation identification information
2842 The default is to send the implementation identification
2845 nfs.recover_lost_locks =
2846 [NFSv4] Attempt to recover locks that were lost due
2847 to a lease timeout on the server. Please note that
2848 doing this risks data corruption, since there are
2849 no guarantees that the file will remain unchanged
2850 after the locks are lost.
2851 If you want to enable the kernel legacy behaviour of
2852 attempting to recover these locks, then set this
2854 The default parameter value of '0' causes the kernel
2855 not to attempt recovery of lost locks.
2857 nfs4.layoutstats_timer =
2858 [NFSv4.2] Change the rate at which the kernel sends
2859 layoutstats to the pNFS metadata server.
2861 Setting this to value to 0 causes the kernel to use
2862 whatever value is the default set by the layout
2863 driver. A non-zero value sets the minimum interval
2864 in seconds between layoutstats transmissions.
2866 nfsd.nfs4_disable_idmapping=
2867 [NFSv4] When set to the default of '1', the NFSv4
2868 server will return only numeric uids and gids to
2869 clients using auth_sys, and will accept numeric uids
2870 and gids from such clients. This is intended to ease
2871 migration from NFSv2/v3.
2873 nmi_debug= [KNL,SH] Specify one or more actions to take
2874 when a NMI is triggered.
2875 Format: [state][,regs][,debounce][,die]
2877 nmi_watchdog= [KNL,BUGS=X86] Debugging features for SMP kernels
2878 Format: [panic,][nopanic,][num]
2880 0 - turn hardlockup detector in nmi_watchdog off
2881 1 - turn hardlockup detector in nmi_watchdog on
2882 When panic is specified, panic when an NMI watchdog
2883 timeout occurs (or 'nopanic' to not panic on an NMI
2884 watchdog, if CONFIG_BOOTPARAM_HARDLOCKUP_PANIC is set)
2885 To disable both hard and soft lockup detectors,
2886 please see 'nowatchdog'.
2887 This is useful when you use a panic=... timeout and
2888 need the box quickly up again.
2890 These settings can be accessed at runtime via
2891 the nmi_watchdog and hardlockup_panic sysctls.
2893 netpoll.carrier_timeout=
2894 [NET] Specifies amount of time (in seconds) that
2895 netpoll should wait for a carrier. By default netpoll
2898 no387 [BUGS=X86-32] Tells the kernel to use the 387 maths
2899 emulation library even if a 387 maths coprocessor
2902 no5lvl [X86-64] Disable 5-level paging mode. Forces
2903 kernel to use 4-level paging instead.
2906 [HW] Never suspend the console
2907 Disable suspending of consoles during suspend and
2908 hibernate operations. Once disabled, debugging
2909 messages can reach various consoles while the rest
2910 of the system is being put to sleep (ie, while
2911 debugging driver suspend/resume hooks). This may
2912 not work reliably with all consoles, but is known
2913 to work with serial and VGA consoles.
2914 To facilitate more flexible debugging, we also add
2915 console_suspend, a printk module parameter to control
2916 it. Users could use console_suspend (usually
2917 /sys/module/printk/parameters/console_suspend) to
2918 turn on/off it dynamically.
2920 novmcoredd [KNL,KDUMP]
2921 Disable device dump. Device dump allows drivers to
2922 append dump data to vmcore so you can collect driver
2923 specified debug info. Drivers can append the data
2924 without any limit and this data is stored in memory,
2925 so this may cause significant memory stress. Disabling
2926 device dump can help save memory but the driver debug
2927 data will be no longer available. This parameter
2928 is only available when CONFIG_PROC_VMCORE_DEVICE_DUMP
2931 noaliencache [MM, NUMA, SLAB] Disables the allocation of alien
2932 caches in the slab allocator. Saves per-node memory,
2933 but will impact performance.
2937 noaltinstr [S390] Disables alternative instructions patching
2938 (CPU alternatives feature).
2940 noapic [SMP,APIC] Tells the kernel to not make use of any
2941 IOAPICs that may be present in the system.
2943 noautogroup Disable scheduler automatic task group creation.
2945 nobats [PPC] Do not use BATs for mapping kernel lowmem
2946 on "Classic" PPC cores.
2950 noclflush [BUGS=X86] Don't use the CLFLUSH instruction
2952 nodelayacct [KNL] Disable per-task delay accounting
2954 nodsp [SH] Disable hardware DSP at boot time.
2956 noefi Disable EFI runtime services support.
2961 On X86-32 available only on PAE configured kernels.
2962 noexec=on: enable non-executable mappings (default)
2963 noexec=off: disable non-executable mappings
2966 Disable SMAP (Supervisor Mode Access Prevention)
2967 even if it is supported by processor.
2970 Disable SMEP (Supervisor Mode Execution Prevention)
2971 even if it is supported by processor.
2974 This affects only 32-bit executables.
2975 noexec32=on: enable non-executable mappings (default)
2976 read doesn't imply executable mappings
2977 noexec32=off: disable non-executable mappings
2978 read implies executable mappings
2980 nofpu [MIPS,SH] Disable hardware FPU at boot time.
2982 nofxsr [BUGS=X86-32] Disables x86 floating point extended
2983 register save and restore. The kernel will only save
2984 legacy floating-point registers on task switch.
2986 nohugeiomap [KNL,x86,PPC] Disable kernel huge I/O mappings.
2988 nosmt [KNL,S390] Disable symmetric multithreading (SMT).
2989 Equivalent to smt=1.
2991 [KNL,x86] Disable symmetric multithreading (SMT).
2992 nosmt=force: Force disable SMT, cannot be undone
2993 via the sysfs control file.
2995 nospectre_v1 [X86,PPC] Disable mitigations for Spectre Variant 1
2996 (bounds check bypass). With this option data leaks are
2997 possible in the system.
2999 nospectre_v2 [X86,PPC_FSL_BOOK3E,ARM64] Disable all mitigations for
3000 the Spectre variant 2 (indirect branch prediction)
3001 vulnerability. System may allow data leaks with this
3004 nospec_store_bypass_disable
3005 [HW] Disable all mitigations for the Speculative Store Bypass vulnerability
3007 noxsave [BUGS=X86] Disables x86 extended register state save
3008 and restore using xsave. The kernel will fallback to
3009 enabling legacy floating-point and sse state.
3011 noxsaveopt [X86] Disables xsaveopt used in saving x86 extended
3012 register states. The kernel will fall back to use
3013 xsave to save the states. By using this parameter,
3014 performance of saving the states is degraded because
3015 xsave doesn't support modified optimization while
3016 xsaveopt supports it on xsaveopt enabled systems.
3018 noxsaves [X86] Disables xsaves and xrstors used in saving and
3019 restoring x86 extended register state in compacted
3020 form of xsave area. The kernel will fall back to use
3021 xsaveopt and xrstor to save and restore the states
3022 in standard form of xsave area. By using this
3023 parameter, xsave area per process might occupy more
3024 memory on xsaves enabled systems.
3026 nohlt [BUGS=ARM,SH] Tells the kernel that the sleep(SH) or
3027 wfi(ARM) instruction doesn't work correctly and not to
3028 use it. This is also useful when using JTAG debugger.
3030 no_file_caps Tells the kernel not to honor file capabilities. The
3031 only way then for a file to be executed with privilege
3032 is to be setuid root or executed by root.
3034 nohalt [IA-64] Tells the kernel not to use the power saving
3035 function PAL_HALT_LIGHT when idle. This increases
3036 power-consumption. On the positive side, it reduces
3037 interrupt wake-up latency, which may improve performance
3038 in certain environments such as networked servers or
3041 nohibernate [HIBERNATION] Disable hibernation and resume.
3043 nohz= [KNL] Boottime enable/disable dynamic ticks
3044 Valid arguments: on, off
3047 nohz_full= [KNL,BOOT,SMP,ISOL]
3048 The argument is a cpu list, as described above.
3049 In kernels built with CONFIG_NO_HZ_FULL=y, set
3050 the specified list of CPUs whose tick will be stopped
3051 whenever possible. The boot CPU will be forced outside
3052 the range to maintain the timekeeping. Any CPUs
3053 in this list will have their RCU callbacks offloaded,
3054 just as if they had also been called out in the
3055 rcu_nocbs= boot parameter.
3057 noiotrap [SH] Disables trapped I/O port accesses.
3059 noirqdebug [X86-32] Disables the code which attempts to detect and
3060 disable unhandled interrupt sources.
3062 no_timer_check [X86,APIC] Disables the code which tests for
3063 broken timer IRQ sources.
3065 noisapnp [ISAPNP] Disables ISA PnP code.
3067 noinitrd [RAM] Tells the kernel not to load any configured
3070 nointremap [X86-64, Intel-IOMMU] Do not enable interrupt
3072 [Deprecated - use intremap=off]
3076 noinvpcid [X86] Disable the INVPCID cpu feature.
3078 nojitter [IA-64] Disables jitter checking for ITC timers.
3080 no-kvmclock [X86,KVM] Disable paravirtualized KVM clock driver
3082 no-kvmapf [X86,KVM] Disable paravirtualized asynchronous page
3086 [X86,PV_OPS] Disable paravirtualized VMware scheduler
3087 clock and use the default one.
3089 no-steal-acc [X86,KVM] Disable paravirtualized steal time accounting.
3090 steal time is computed, but won't influence scheduler
3093 nolapic [X86-32,APIC] Do not enable or use the local APIC.
3095 nolapic_timer [X86-32,APIC] Do not use the local APIC timer.
3097 noltlbs [PPC] Do not use large page/tlb entries for kernel
3098 lowmem mapping on PPC40x and PPC8xx
3100 nomca [IA-64] Disable machine check abort handling
3102 nomce [X86-32] Disable Machine Check Exception
3104 nomfgpt [X86-32] Disable Multi-Function General Purpose
3105 Timer usage (for AMD Geode machines).
3107 nonmi_ipi [X86] Disable using NMI IPIs during panic/reboot to
3108 shutdown the other cpus. Instead use the REBOOT_VECTOR
3111 nomodule Disable module load
3113 nopat [X86] Disable PAT (page attribute table extension of
3114 pagetables) support.
3116 nopcid [X86-64] Disable the PCID cpu feature.
3118 norandmaps Don't use address space randomization. Equivalent to
3119 echo 0 > /proc/sys/kernel/randomize_va_space
3121 noreplace-smp [X86-32,SMP] Don't replace SMP instructions
3122 with UP alternatives
3124 nordrand [X86] Disable kernel use of the RDRAND and
3125 RDSEED instructions even if they are supported
3126 by the processor. RDRAND and RDSEED are still
3127 available to user space applications.
3129 noresume [SWSUSP] Disables resume and restores original swap
3132 no-scroll [VGA] Disables scrollback.
3133 This is required for the Braillex ib80-piezo Braille
3134 reader made by F.H. Papenmeier (Germany).
3138 nosep [BUGS=X86-32] Disables x86 SYSENTER/SYSEXIT support.
3140 nosmp [SMP] Tells an SMP kernel to act as a UP kernel,
3141 and disable the IO APIC. legacy for "maxcpus=0".
3143 nosoftlockup [KNL] Disable the soft-lockup detector.
3145 nosync [HW,M68K] Disables sync negotiation for all devices.
3147 nowatchdog [KNL] Disable both lockup detectors, i.e.
3148 soft-lockup and NMI watchdog (hard-lockup).
3152 nox2apic [X86-64,APIC] Do not enable x2APIC mode.
3154 cpu0_hotplug [X86] Turn on CPU0 hotplug feature when
3155 CONFIG_BOOTPARAM_HOTPLUG_CPU0 is off.
3156 Some features depend on CPU0. Known dependencies are:
3157 1. Resume from suspend/hibernate depends on CPU0.
3158 Suspend/hibernate will fail if CPU0 is offline and you
3159 need to online CPU0 before suspend/hibernate.
3160 2. PIC interrupts also depend on CPU0. CPU0 can't be
3161 removed if a PIC interrupt is detected.
3162 It's said poweroff/reboot may depend on CPU0 on some
3163 machines although I haven't seen such issues so far
3164 after CPU0 is offline on a few tested machines.
3165 If the dependencies are under your control, you can
3166 turn on cpu0_hotplug.
3168 nps_mtm_hs_ctr= [KNL,ARC]
3169 This parameter sets the maximum duration, in
3170 cycles, each HW thread of the CTOP can run
3171 without interruptions, before HW switches it.
3172 The actual maximum duration is 16 times this
3174 Format: integer between 1 and 255
3177 nptcg= [IA-64] Override max number of concurrent global TLB
3178 purges which is reported from either PAL_VM_SUMMARY or
3181 nr_cpus= [SMP] Maximum number of processors that an SMP kernel
3182 could support. nr_cpus=n : n >= 1 limits the kernel to
3183 support 'n' processors. It could be larger than the
3184 number of already plugged CPU during bootup, later in
3185 runtime you can physically add extra cpu until it reaches
3186 n. So during boot up some boot time memory for per-cpu
3187 variables need be pre-allocated for later physical cpu
3190 nr_uarts= [SERIAL] maximum number of UARTs to be registered.
3192 numa_balancing= [KNL,X86] Enable or disable automatic NUMA balancing.
3193 Allowed values are enable and disable
3195 numa_zonelist_order= [KNL, BOOT] Select zonelist order for NUMA.
3196 'node', 'default' can be specified
3197 This can be set from sysctl after boot.
3198 See Documentation/admin-guide/sysctl/vm.rst for details.
3200 ohci1394_dma=early [HW] enable debugging via the ohci1394 driver.
3201 See Documentation/debugging-via-ohci1394.txt for more
3204 olpc_ec_timeout= [OLPC] ms delay when issuing EC commands
3205 Rather than timing out after 20 ms if an EC
3206 command is not properly ACKed, override the length
3207 of the timeout. We have interrupts disabled while
3208 waiting for the ACK, so if this is set too high
3209 interrupts *may* be lost!
3211 omap_mux= [OMAP] Override bootloader pin multiplexing.
3212 Format: <mux_mode0.mode_name=value>...
3213 For example, to override I2C bus2:
3214 omap_mux=i2c2_scl.i2c2_scl=0x100,i2c2_sda.i2c2_sda=0x100
3216 oprofile.timer= [HW]
3217 Use timer interrupt instead of performance counters
3219 oprofile.cpu_type= Force an oprofile cpu type
3220 This might be useful if you have an older oprofile
3221 userland or if you want common events.
3222 Format: { arch_perfmon }
3223 arch_perfmon: [X86] Force use of architectural
3224 perfmon on Intel CPUs instead of the
3225 CPU specific event set.
3226 timer: [X86] Force use of architectural NMI
3227 timer mode (see also oprofile.timer
3228 for generic hr timer mode)
3230 oops=panic Always panic on oopses. Default is to just kill the
3231 process, but there is a small probability of
3232 deadlocking the machine.
3233 This will also cause panics on machine check exceptions.
3234 Useful together with panic=30 to trigger a reboot.
3237 [KNL] Boolean flag to control whether the page allocator
3238 should randomize its free lists. The randomization may
3239 be automatically enabled if the kernel detects it is
3240 running on a platform with a direct-mapped memory-side
3241 cache, and this parameter can be used to
3242 override/disable that behavior. The state of the flag
3243 can be read from sysfs at:
3244 /sys/module/page_alloc/parameters/shuffle.
3246 page_owner= [KNL] Boot-time page_owner enabling option.
3247 Storage of the information about who allocated
3248 each page is disabled in default. With this switch,
3250 on: enable the feature
3252 page_poison= [KNL] Boot-time parameter changing the state of
3253 poisoning on the buddy allocator, available with
3254 CONFIG_PAGE_POISONING=y.
3255 off: turn off poisoning (default)
3256 on: turn on poisoning
3258 panic= [KNL] Kernel behaviour on panic: delay <timeout>
3259 timeout > 0: seconds before rebooting
3260 timeout = 0: wait forever
3261 timeout < 0: reboot immediately
3264 panic_print= Bitmask for printing system info when panic happens.
3265 User can chose combination of the following bits:
3266 bit 0: print all tasks info
3267 bit 1: print system memory info
3268 bit 2: print timer info
3269 bit 3: print locks info if CONFIG_LOCKDEP is on
3270 bit 4: print ftrace buffer
3271 bit 5: print all printk messages in buffer
3273 panic_on_warn panic() instead of WARN(). Useful to cause kdump
3276 crash_kexec_post_notifiers
3277 Run kdump after running panic-notifiers and dumping
3278 kmsg. This only for the users who doubt kdump always
3279 succeeds in any situation.
3280 Note that this also increases risks of kdump failure,
3281 because some panic notifiers can make the crashed
3282 kernel more unstable.
3284 parkbd.port= [HW] Parallel port number the keyboard adapter is
3285 connected to, default is 0.
3287 parkbd.mode= [HW] Parallel port keyboard adapter mode of operation,
3288 0 for XT, 1 for AT (default is AT).
3291 parport= [HW,PPT] Specify parallel ports. 0 disables.
3292 Format: { 0 | auto | 0xBBB[,IRQ[,DMA]] }
3293 Use 'auto' to force the driver to use any
3294 IRQ/DMA settings detected (the default is to
3295 ignore detected IRQ/DMA settings because of
3296 possible conflicts). You can specify the base
3297 address, IRQ, and DMA settings; IRQ and DMA
3298 should be numbers, or 'auto' (for using detected
3299 settings on that particular port), or 'nofifo'
3300 (to avoid using a FIFO even if it is detected).
3301 Parallel ports are assigned in the order they
3302 are specified on the command line, starting
3305 parport_init_mode= [HW,PPT]
3306 Configure VIA parallel port to operate in
3307 a specific mode. This is necessary on Pegasos
3308 computer where firmware has no options for setting
3309 up parallel port mode and sets it to spp.
3310 Currently this function knows 686a and 8231 chips.
3311 Format: [spp|ps2|epp|ecp|ecpepp]
3314 Halt all CPUs after the first oops has been printed for
3315 the specified number of seconds. This is to be used if
3316 your oopses keep scrolling off the screen.
3321 See header of drivers/block/paride/pcd.c.
3322 See also Documentation/admin-guide/blockdev/paride.rst.
3324 pci=option[,option...] [PCI] various PCI subsystem options.
3326 Some options herein operate on a specific device
3327 or a set of devices (<pci_dev>). These are
3328 specified in one of the following formats:
3330 [<domain>:]<bus>:<dev>.<func>[/<dev>.<func>]*
3331 pci:<vendor>:<device>[:<subvendor>:<subdevice>]
3333 Note: the first format specifies a PCI
3334 bus/device/function address which may change
3335 if new hardware is inserted, if motherboard
3336 firmware changes, or due to changes caused
3337 by other kernel parameters. If the
3338 domain is left unspecified, it is
3339 taken to be zero. Optionally, a path
3340 to a device through multiple device/function
3341 addresses can be specified after the base
3342 address (this is more robust against
3343 renumbering issues). The second format
3344 selects devices using IDs from the
3345 configuration space which may match multiple
3346 devices in the system.
3348 earlydump dump PCI config space before the kernel
3350 off [X86] don't probe for the PCI bus
3351 bios [X86-32] force use of PCI BIOS, don't access
3352 the hardware directly. Use this if your machine
3353 has a non-standard PCI host bridge.
3354 nobios [X86-32] disallow use of PCI BIOS, only direct
3355 hardware access methods are allowed. Use this
3356 if you experience crashes upon bootup and you
3357 suspect they are caused by the BIOS.
3358 conf1 [X86] Force use of PCI Configuration Access
3359 Mechanism 1 (config address in IO port 0xCF8,
3360 data in IO port 0xCFC, both 32-bit).
3361 conf2 [X86] Force use of PCI Configuration Access
3362 Mechanism 2 (IO port 0xCF8 is an 8-bit port for
3363 the function, IO port 0xCFA, also 8-bit, sets
3364 bus number. The config space is then accessed
3365 through ports 0xC000-0xCFFF).
3366 See http://wiki.osdev.org/PCI for more info
3367 on the configuration access mechanisms.
3368 noaer [PCIE] If the PCIEAER kernel config parameter is
3369 enabled, this kernel boot option can be used to
3370 disable the use of PCIE advanced error reporting.
3371 nodomains [PCI] Disable support for multiple PCI
3372 root domains (aka PCI segments, in ACPI-speak).
3373 nommconf [X86] Disable use of MMCONFIG for PCI
3375 check_enable_amd_mmconf [X86] check for and enable
3376 properly configured MMIO access to PCI
3377 config space on AMD family 10h CPU
3378 nomsi [MSI] If the PCI_MSI kernel config parameter is
3379 enabled, this kernel boot option can be used to
3380 disable the use of MSI interrupts system-wide.
3381 noioapicquirk [APIC] Disable all boot interrupt quirks.
3382 Safety option to keep boot IRQs enabled. This
3383 should never be necessary.
3384 ioapicreroute [APIC] Enable rerouting of boot IRQs to the
3385 primary IO-APIC for bridges that cannot disable
3386 boot IRQs. This fixes a source of spurious IRQs
3387 when the system masks IRQs.
3388 noioapicreroute [APIC] Disable workaround that uses the
3389 boot IRQ equivalent of an IRQ that connects to
3390 a chipset where boot IRQs cannot be disabled.
3391 The opposite of ioapicreroute.
3392 biosirq [X86-32] Use PCI BIOS calls to get the interrupt
3393 routing table. These calls are known to be buggy
3394 on several machines and they hang the machine
3395 when used, but on other computers it's the only
3396 way to get the interrupt routing table. Try
3397 this option if the kernel is unable to allocate
3398 IRQs or discover secondary PCI buses on your
3400 rom [X86] Assign address space to expansion ROMs.
3401 Use with caution as certain devices share
3402 address decoders between ROMs and other
3404 norom [X86] Do not assign address space to
3405 expansion ROMs that do not already have
3406 BIOS assigned address ranges.
3407 nobar [X86] Do not assign address space to the
3408 BARs that weren't assigned by the BIOS.
3409 irqmask=0xMMMM [X86] Set a bit mask of IRQs allowed to be
3410 assigned automatically to PCI devices. You can
3411 make the kernel exclude IRQs of your ISA cards
3413 pirqaddr=0xAAAAA [X86] Specify the physical address
3414 of the PIRQ table (normally generated
3415 by the BIOS) if it is outside the
3416 F0000h-100000h range.
3417 lastbus=N [X86] Scan all buses thru bus #N. Can be
3418 useful if the kernel is unable to find your
3419 secondary buses and you want to tell it
3420 explicitly which ones they are.
3421 assign-busses [X86] Always assign all PCI bus
3422 numbers ourselves, overriding
3423 whatever the firmware may have done.
3424 usepirqmask [X86] Honor the possible IRQ mask stored
3425 in the BIOS $PIR table. This is needed on
3426 some systems with broken BIOSes, notably
3427 some HP Pavilion N5400 and Omnibook XE3
3428 notebooks. This will have no effect if ACPI
3429 IRQ routing is enabled.
3430 noacpi [X86] Do not use ACPI for IRQ routing
3431 or for PCI scanning.
3432 use_crs [X86] Use PCI host bridge window information
3433 from ACPI. On BIOSes from 2008 or later, this
3434 is enabled by default. If you need to use this,
3435 please report a bug.
3436 nocrs [X86] Ignore PCI host bridge windows from ACPI.
3437 If you need to use this, please report a bug.
3438 routeirq Do IRQ routing for all PCI devices.
3439 This is normally done in pci_enable_device(),
3440 so this option is a temporary workaround
3441 for broken drivers that don't call it.
3442 skip_isa_align [X86] do not align io start addr, so can
3443 handle more pci cards
3444 noearly [X86] Don't do any early type 1 scanning.
3445 This might help on some broken boards which
3446 machine check when some devices' config space
3447 is read. But various workarounds are disabled
3448 and some IOMMU drivers will not work.
3449 bfsort Sort PCI devices into breadth-first order.
3450 This sorting is done to get a device
3451 order compatible with older (<= 2.4) kernels.
3452 nobfsort Don't sort PCI devices into breadth-first order.
3453 pcie_bus_tune_off Disable PCIe MPS (Max Payload Size)
3454 tuning and use the BIOS-configured MPS defaults.
3455 pcie_bus_safe Set every device's MPS to the largest value
3456 supported by all devices below the root complex.
3457 pcie_bus_perf Set device MPS to the largest allowable MPS
3458 based on its parent bus. Also set MRRS (Max
3459 Read Request Size) to the largest supported
3460 value (no larger than the MPS that the device
3461 or bus can support) for best performance.
3462 pcie_bus_peer2peer Set every device's MPS to 128B, which
3463 every device is guaranteed to support. This
3464 configuration allows peer-to-peer DMA between
3465 any pair of devices, possibly at the cost of
3466 reduced performance. This also guarantees
3467 that hot-added devices will work.
3468 cbiosize=nn[KMG] The fixed amount of bus space which is
3469 reserved for the CardBus bridge's IO window.
3470 The default value is 256 bytes.
3471 cbmemsize=nn[KMG] The fixed amount of bus space which is
3472 reserved for the CardBus bridge's memory
3473 window. The default value is 64 megabytes.
3476 [<order of align>@]<pci_dev>[; ...]
3477 Specifies alignment and device to reassign
3478 aligned memory resources. How to
3479 specify the device is described above.
3480 If <order of align> is not specified,
3481 PAGE_SIZE is used as alignment.
3482 PCI-PCI bridge can be specified, if resource
3483 windows need to be expanded.
3484 To specify the alignment for several
3485 instances of a device, the PCI vendor,
3486 device, subvendor, and subdevice may be
3487 specified, e.g., 4096@pci:8086:9c22:103c:198f
3488 ecrc= Enable/disable PCIe ECRC (transaction layer
3489 end-to-end CRC checking).
3490 bios: Use BIOS/firmware settings. This is the
3494 hpiosize=nn[KMG] The fixed amount of bus space which is
3495 reserved for hotplug bridge's IO window.
3496 Default size is 256 bytes.
3497 hpmemsize=nn[KMG] The fixed amount of bus space which is
3498 reserved for hotplug bridge's memory window.
3499 Default size is 2 megabytes.
3500 hpbussize=nn The minimum amount of additional bus numbers
3501 reserved for buses below a hotplug bridge.
3503 realloc= Enable/disable reallocating PCI bridge resources
3504 if allocations done by BIOS are too small to
3505 accommodate resources required by all child
3507 off: Turn realloc off
3509 realloc same as realloc=on
3510 noari do not use PCIe ARI.
3511 noats [PCIE, Intel-IOMMU, AMD-IOMMU]
3512 do not use PCIe ATS (and IOMMU device IOTLB).
3513 pcie_scan_all Scan all possible PCIe devices. Otherwise we
3514 only look for one device below a PCIe downstream
3516 big_root_window Try to add a big 64bit memory window to the PCIe
3517 root complex on AMD CPUs. Some GFX hardware
3518 can resize a BAR to allow access to all VRAM.
3519 Adding the window is slightly risky (it may
3520 conflict with unreported devices), so this
3522 disable_acs_redir=<pci_dev>[; ...]
3523 Specify one or more PCI devices (in the format
3524 specified above) separated by semicolons.
3525 Each device specified will have the PCI ACS
3526 redirect capabilities forced off which will
3527 allow P2P traffic between devices through
3528 bridges without forcing it upstream. Note:
3529 this removes isolation between devices and
3530 may put more devices in an IOMMU group.
3531 force_floating [S390] Force usage of floating interrupts.
3532 nomio [S390] Do not use MIO instructions.
3534 pcie_aspm= [PCIE] Forcibly enable or disable PCIe Active State Power
3537 force Enable ASPM even on devices that claim not to support it.
3538 WARNING: Forcing ASPM on may cause system lockups.
3540 pcie_ports= [PCIE] PCIe port services handling:
3541 native Use native PCIe services (PME, AER, DPC, PCIe hotplug)
3542 even if the platform doesn't give the OS permission to
3543 use them. This may cause conflicts if the platform
3544 also tries to use these services.
3545 compat Disable native PCIe services (PME, AER, DPC, PCIe
3548 pcie_port_pm= [PCIE] PCIe port power management handling:
3549 off Disable power management of all PCIe ports
3550 force Forcibly enable power management of all PCIe ports
3552 pcie_pme= [PCIE,PM] Native PCIe PME signaling options:
3553 nomsi Do not use MSI for native PCIe PME signaling (this makes
3554 all PCIe root ports use INTx for all services).
3556 pcmv= [HW,PCMCIA] BadgePAD 4
3560 Keep all power-domains already enabled by bootloader on,
3561 even if no driver has claimed them. This is useful
3562 for debug and development, but should not be
3563 needed on a platform with proper driver support.
3566 See Documentation/admin-guide/blockdev/paride.rst.
3568 pdcchassis= [PARISC,HW] Disable/Enable PDC Chassis Status codes at
3571 See arch/parisc/kernel/pdc_chassis.c
3573 percpu_alloc= Select which percpu first chunk allocator to use.
3574 Currently supported values are "embed" and "page".
3575 Archs may support subset or none of the selections.
3576 See comments in mm/percpu.c for details on each
3577 allocator. This parameter is primarily for debugging
3578 and performance comparison.
3581 See Documentation/admin-guide/blockdev/paride.rst.
3584 See Documentation/admin-guide/blockdev/paride.rst.
3586 pirq= [SMP,APIC] Manual mp-table setup
3587 See Documentation/x86/i386/IO-APIC.rst.
3589 plip= [PPT,NET] Parallel port network link
3590 Format: { parport<nr> | timid | 0 }
3591 See also Documentation/admin-guide/parport.rst.
3593 pmtmr= [X86] Manual setup of pmtmr I/O Port.
3594 Override pmtimer IOPort with a hex value.
3598 Enable PNP debug messages (depends on the
3599 CONFIG_PNP_DEBUG_MESSAGES option). Change at run-time
3600 via /sys/module/pnp/parameters/debug. We always show
3601 current resource usage; turning this on also shows
3602 possible settings and some assignment information.
3608 { on | off | curr | res | no-curr | no-res }
3611 [ISAPNP] Exclude IRQs for the autoconfiguration
3614 [ISAPNP] Exclude DMAs for the autoconfiguration
3616 pnp_reserve_io= [ISAPNP] Exclude I/O ports for the autoconfiguration
3617 Ranges are in pairs (I/O port base and size).
3620 [ISAPNP] Exclude memory regions for the
3622 Ranges are in pairs (memory base and size).
3624 ports= [IP_VS_FTP] IPVS ftp helper module
3626 Up to 8 (IP_VS_APP_MAX_PORTS) ports
3628 Format: <port>,<port>....
3630 powersave=off [PPC] This option disables power saving features.
3631 It specifically disables cpuidle and sets the
3632 platform machine description specific power_save
3633 function to NULL. On Idle the CPU just reduces
3636 ppc_strict_facility_enable
3637 [PPC] This option catches any kernel floating point,
3638 Altivec, VSX and SPE outside of regions specifically
3639 allowed (eg kernel_enable_fpu()/kernel_disable_fpu()).
3640 There is some performance impact when enabling this.
3644 Disable Hardware Transactional Memory
3646 print-fatal-signals=
3647 [KNL] debug: print fatal signals
3649 If enabled, warn about various signal handling
3650 related application anomalies: too many signals,
3651 too many POSIX.1 timers, fatal signals causing a
3654 If you hit the warning due to signal overflow,
3655 you might want to try "ulimit -i unlimited".
3659 printk.always_kmsg_dump=
3660 Trigger kmsg_dump for cases other than kernel oops or
3662 Format: <bool> (1/Y/y=enable, 0/N/n=disable)
3665 printk.devkmsg={on,off,ratelimit}
3666 Control writing to /dev/kmsg.
3667 on - unlimited logging to /dev/kmsg from userspace
3668 off - logging to /dev/kmsg disabled
3669 ratelimit - ratelimit the logging
3672 printk.time= Show timing data prefixed to each printk message line
3673 Format: <bool> (1/Y/y=enable, 0/N/n=disable)
3675 processor.max_cstate= [HW,ACPI]
3676 Limit processor to maximum C-state
3677 max_cstate=9 overrides any DMI blacklist limit.
3679 processor.nocst [HW,ACPI]
3680 Ignore the _CST method to determine C-states,
3681 instead using the legacy FADT method
3683 profile= [KNL] Enable kernel profiling via /proc/profile
3684 Format: [<profiletype>,]<number>
3685 Param: <profiletype>: "schedule", "sleep", or "kvm"
3686 [defaults to kernel profiling]
3687 Param: "schedule" - profile schedule points.
3688 Param: "sleep" - profile D-state sleeping (millisecs).
3689 Requires CONFIG_SCHEDSTATS
3690 Param: "kvm" - profile VM exits.
3691 Param: <number> - step/bucket size as a power of 2 for
3692 statistical time based profiling.
3694 prompt_ramdisk= [RAM] List of RAM disks to prompt for floppy disk
3696 See Documentation/admin-guide/blockdev/ramdisk.rst.
3698 psi= [KNL] Enable or disable pressure stall information
3702 psmouse.proto= [HW,MOUSE] Highest PS2 mouse protocol extension to
3703 probe for; one of (bare|imps|exps|lifebook|any).
3704 psmouse.rate= [HW,MOUSE] Set desired mouse report rate, in reports
3706 psmouse.resetafter= [HW,MOUSE]
3707 Try to reset the device after so many bad packets
3710 [HW,MOUSE] Set desired mouse resolution, in dpi.
3711 psmouse.smartscroll=
3712 [HW,MOUSE] Controls Logitech smartscroll autorepeat.
3713 0 = disabled, 1 = enabled (default).
3715 pstore.backend= Specify the name of the pstore backend to use
3718 See Documentation/admin-guide/blockdev/paride.rst.
3720 pti= [X86_64] Control Page Table Isolation of user and
3721 kernel address spaces. Disabling this feature
3722 removes hardening, but improves performance of
3723 system calls and interrupts.
3725 on - unconditionally enable
3726 off - unconditionally disable
3727 auto - kernel detects whether your CPU model is
3728 vulnerable to issues that PTI mitigates
3730 Not specifying this option is equivalent to pti=auto.
3733 Equivalent to pti=off
3736 [KNL] Number of legacy pty's. Overwrites compiled-in
3739 quiet [KNL] Disable most log messages
3744 See Documentation/admin-guide/md.rst.
3746 ramdisk_size= [RAM] Sizes of RAM disks in kilobytes
3747 See Documentation/admin-guide/blockdev/ramdisk.rst.
3749 random.trust_cpu={on,off}
3750 [KNL] Enable or disable trusting the use of the
3751 CPU's random number generator (if available) to
3752 fully seed the kernel's CRNG. Default is controlled
3753 by CONFIG_RANDOM_TRUST_CPU.
3755 ras=option[,option,...] [KNL] RAS-specific options
3758 Disable the Correctable Errors Collector,
3759 see CONFIG_RAS_CEC help text.
3762 The argument is a cpu list, as described above,
3763 except that the string "all" can be used to
3764 specify every CPU on the system.
3766 In kernels built with CONFIG_RCU_NOCB_CPU=y, set
3767 the specified list of CPUs to be no-callback CPUs.
3768 Invocation of these CPUs' RCU callbacks will be
3769 offloaded to "rcuox/N" kthreads created for that
3770 purpose, where "x" is "p" for RCU-preempt, and
3771 "s" for RCU-sched, and "N" is the CPU number.
3772 This reduces OS jitter on the offloaded CPUs,
3773 which can be useful for HPC and real-time
3774 workloads. It can also improve energy efficiency
3775 for asymmetric multiprocessors.
3778 Rather than requiring that offloaded CPUs
3779 (specified by rcu_nocbs= above) explicitly
3780 awaken the corresponding "rcuoN" kthreads,
3781 make these kthreads poll for callbacks.
3782 This improves the real-time response for the
3783 offloaded CPUs by relieving them of the need to
3784 wake up the corresponding kthread, but degrades
3785 energy efficiency by requiring that the kthreads
3786 periodically wake up to do the polling.
3788 rcutree.blimit= [KNL]
3789 Set maximum number of finished RCU callbacks to
3790 process in one batch.
3792 rcutree.dump_tree= [KNL]
3793 Dump the structure of the rcu_node combining tree
3794 out at early boot. This is used for diagnostic
3795 purposes, to verify correct tree setup.
3797 rcutree.gp_cleanup_delay= [KNL]
3798 Set the number of jiffies to delay each step of
3799 RCU grace-period cleanup.
3801 rcutree.gp_init_delay= [KNL]
3802 Set the number of jiffies to delay each step of
3803 RCU grace-period initialization.
3805 rcutree.gp_preinit_delay= [KNL]
3806 Set the number of jiffies to delay each step of
3807 RCU grace-period pre-initialization, that is,
3808 the propagation of recent CPU-hotplug changes up
3809 the rcu_node combining tree.
3811 rcutree.use_softirq= [KNL]
3812 If set to zero, move all RCU_SOFTIRQ processing to
3813 per-CPU rcuc kthreads. Defaults to a non-zero
3814 value, meaning that RCU_SOFTIRQ is used by default.
3815 Specify rcutree.use_softirq=0 to use rcuc kthreads.
3817 rcutree.rcu_fanout_exact= [KNL]
3818 Disable autobalancing of the rcu_node combining
3819 tree. This is used by rcutorture, and might
3820 possibly be useful for architectures having high
3821 cache-to-cache transfer latencies.
3823 rcutree.rcu_fanout_leaf= [KNL]
3824 Change the number of CPUs assigned to each
3825 leaf rcu_node structure. Useful for very
3826 large systems, which will choose the value 64,
3827 and for NUMA systems with large remote-access
3828 latencies, which will choose a value aligned
3829 with the appropriate hardware boundaries.
3831 rcutree.jiffies_till_first_fqs= [KNL]
3832 Set delay from grace-period initialization to
3833 first attempt to force quiescent states.
3834 Units are jiffies, minimum value is zero,
3835 and maximum value is HZ.
3837 rcutree.jiffies_till_next_fqs= [KNL]
3838 Set delay between subsequent attempts to force
3839 quiescent states. Units are jiffies, minimum
3840 value is one, and maximum value is HZ.
3842 rcutree.jiffies_till_sched_qs= [KNL]
3843 Set required age in jiffies for a
3844 given grace period before RCU starts
3845 soliciting quiescent-state help from
3846 rcu_note_context_switch() and cond_resched().
3847 If not specified, the kernel will calculate
3848 a value based on the most recent settings
3849 of rcutree.jiffies_till_first_fqs
3850 and rcutree.jiffies_till_next_fqs.
3851 This calculated value may be viewed in
3852 rcutree.jiffies_to_sched_qs. Any attempt to set
3853 rcutree.jiffies_to_sched_qs will be cheerfully
3856 rcutree.kthread_prio= [KNL,BOOT]
3857 Set the SCHED_FIFO priority of the RCU per-CPU
3858 kthreads (rcuc/N). This value is also used for
3859 the priority of the RCU boost threads (rcub/N)
3860 and for the RCU grace-period kthreads (rcu_bh,
3861 rcu_preempt, and rcu_sched). If RCU_BOOST is
3862 set, valid values are 1-99 and the default is 1
3863 (the least-favored priority). Otherwise, when
3864 RCU_BOOST is not set, valid values are 0-99 and
3865 the default is zero (non-realtime operation).
3867 rcutree.rcu_nocb_leader_stride= [KNL]
3868 Set the number of NOCB kthread groups, which
3869 defaults to the square root of the number of
3870 CPUs. Larger numbers reduces the wakeup overhead
3871 on the per-CPU grace-period kthreads, but increases
3872 that same overhead on each group's leader.
3874 rcutree.qhimark= [KNL]
3875 Set threshold of queued RCU callbacks beyond which
3876 batch limiting is disabled.
3878 rcutree.qlowmark= [KNL]
3879 Set threshold of queued RCU callbacks below which
3880 batch limiting is re-enabled.
3882 rcutree.rcu_idle_gp_delay= [KNL]
3883 Set wakeup interval for idle CPUs that have
3884 RCU callbacks (RCU_FAST_NO_HZ=y).
3886 rcutree.rcu_idle_lazy_gp_delay= [KNL]
3887 Set wakeup interval for idle CPUs that have
3888 only "lazy" RCU callbacks (RCU_FAST_NO_HZ=y).
3889 Lazy RCU callbacks are those which RCU can
3890 prove do nothing more than free memory.
3892 rcutree.rcu_kick_kthreads= [KNL]
3893 Cause the grace-period kthread to get an extra
3894 wake_up() if it sleeps three times longer than
3895 it should at force-quiescent-state time.
3896 This wake_up() will be accompanied by a
3897 WARN_ONCE() splat and an ftrace_dump().
3899 rcutree.sysrq_rcu= [KNL]
3900 Commandeer a sysrq key to dump out Tree RCU's
3901 rcu_node tree with an eye towards determining
3902 why a new grace period has not yet started.
3904 rcuperf.gp_async= [KNL]
3905 Measure performance of asynchronous
3906 grace-period primitives such as call_rcu().
3908 rcuperf.gp_async_max= [KNL]
3909 Specify the maximum number of outstanding
3910 callbacks per writer thread. When a writer
3911 thread exceeds this limit, it invokes the
3912 corresponding flavor of rcu_barrier() to allow
3913 previously posted callbacks to drain.
3915 rcuperf.gp_exp= [KNL]
3916 Measure performance of expedited synchronous
3917 grace-period primitives.
3919 rcuperf.holdoff= [KNL]
3920 Set test-start holdoff period. The purpose of
3921 this parameter is to delay the start of the
3922 test until boot completes in order to avoid
3925 rcuperf.nreaders= [KNL]
3926 Set number of RCU readers. The value -1 selects
3927 N, where N is the number of CPUs. A value
3928 "n" less than -1 selects N-n+1, where N is again
3929 the number of CPUs. For example, -2 selects N
3930 (the number of CPUs), -3 selects N+1, and so on.
3931 A value of "n" less than or equal to -N selects
3934 rcuperf.nwriters= [KNL]
3935 Set number of RCU writers. The values operate
3936 the same as for rcuperf.nreaders.
3937 N, where N is the number of CPUs
3939 rcuperf.perf_type= [KNL]
3940 Specify the RCU implementation to test.
3942 rcuperf.shutdown= [KNL]
3943 Shut the system down after performance tests
3944 complete. This is useful for hands-off automated
3947 rcuperf.verbose= [KNL]
3948 Enable additional printk() statements.
3950 rcuperf.writer_holdoff= [KNL]
3951 Write-side holdoff between grace periods,
3952 in microseconds. The default of zero says
3955 rcutorture.fqs_duration= [KNL]
3956 Set duration of force_quiescent_state bursts
3959 rcutorture.fqs_holdoff= [KNL]
3960 Set holdoff time within force_quiescent_state bursts
3963 rcutorture.fqs_stutter= [KNL]
3964 Set wait time between force_quiescent_state bursts
3967 rcutorture.fwd_progress= [KNL]
3968 Enable RCU grace-period forward-progress testing
3969 for the types of RCU supporting this notion.
3971 rcutorture.fwd_progress_div= [KNL]
3972 Specify the fraction of a CPU-stall-warning
3973 period to do tight-loop forward-progress testing.
3975 rcutorture.fwd_progress_holdoff= [KNL]
3976 Number of seconds to wait between successive
3977 forward-progress tests.
3979 rcutorture.fwd_progress_need_resched= [KNL]
3980 Enclose cond_resched() calls within checks for
3981 need_resched() during tight-loop forward-progress
3984 rcutorture.gp_cond= [KNL]
3985 Use conditional/asynchronous update-side
3986 primitives, if available.
3988 rcutorture.gp_exp= [KNL]
3989 Use expedited update-side primitives, if available.
3991 rcutorture.gp_normal= [KNL]
3992 Use normal (non-expedited) asynchronous
3993 update-side primitives, if available.
3995 rcutorture.gp_sync= [KNL]
3996 Use normal (non-expedited) synchronous
3997 update-side primitives, if available. If all
3998 of rcutorture.gp_cond=, rcutorture.gp_exp=,
3999 rcutorture.gp_normal=, and rcutorture.gp_sync=
4000 are zero, rcutorture acts as if is interpreted
4001 they are all non-zero.
4003 rcutorture.n_barrier_cbs= [KNL]
4004 Set callbacks/threads for rcu_barrier() testing.
4006 rcutorture.nfakewriters= [KNL]
4007 Set number of concurrent RCU writers. These just
4008 stress RCU, they don't participate in the actual
4009 test, hence the "fake".
4011 rcutorture.nreaders= [KNL]
4012 Set number of RCU readers. The value -1 selects
4013 N-1, where N is the number of CPUs. A value
4014 "n" less than -1 selects N-n-2, where N is again
4015 the number of CPUs. For example, -2 selects N
4016 (the number of CPUs), -3 selects N+1, and so on.
4018 rcutorture.object_debug= [KNL]
4019 Enable debug-object double-call_rcu() testing.
4021 rcutorture.onoff_holdoff= [KNL]
4022 Set time (s) after boot for CPU-hotplug testing.
4024 rcutorture.onoff_interval= [KNL]
4025 Set time (jiffies) between CPU-hotplug operations,
4026 or zero to disable CPU-hotplug testing.
4028 rcutorture.shuffle_interval= [KNL]
4029 Set task-shuffle interval (s). Shuffling tasks
4030 allows some CPUs to go into dyntick-idle mode
4031 during the rcutorture test.
4033 rcutorture.shutdown_secs= [KNL]
4034 Set time (s) after boot system shutdown. This
4035 is useful for hands-off automated testing.
4037 rcutorture.stall_cpu= [KNL]
4038 Duration of CPU stall (s) to test RCU CPU stall
4039 warnings, zero to disable.
4041 rcutorture.stall_cpu_holdoff= [KNL]
4042 Time to wait (s) after boot before inducing stall.
4044 rcutorture.stall_cpu_irqsoff= [KNL]
4045 Disable interrupts while stalling if set.
4047 rcutorture.stat_interval= [KNL]
4048 Time (s) between statistics printk()s.
4050 rcutorture.stutter= [KNL]
4051 Time (s) to stutter testing, for example, specifying
4052 five seconds causes the test to run for five seconds,
4053 wait for five seconds, and so on. This tests RCU's
4054 ability to transition abruptly to and from idle.
4056 rcutorture.test_boost= [KNL]
4057 Test RCU priority boosting? 0=no, 1=maybe, 2=yes.
4058 "Maybe" means test if the RCU implementation
4059 under test support RCU priority boosting.
4061 rcutorture.test_boost_duration= [KNL]
4062 Duration (s) of each individual boost test.
4064 rcutorture.test_boost_interval= [KNL]
4065 Interval (s) between each boost test.
4067 rcutorture.test_no_idle_hz= [KNL]
4068 Test RCU's dyntick-idle handling. See also the
4069 rcutorture.shuffle_interval parameter.
4071 rcutorture.torture_type= [KNL]
4072 Specify the RCU implementation to test.
4074 rcutorture.verbose= [KNL]
4075 Enable additional printk() statements.
4077 rcupdate.rcu_cpu_stall_suppress= [KNL]
4078 Suppress RCU CPU stall warning messages.
4080 rcupdate.rcu_cpu_stall_timeout= [KNL]
4081 Set timeout for RCU CPU stall warning messages.
4083 rcupdate.rcu_expedited= [KNL]
4084 Use expedited grace-period primitives, for
4085 example, synchronize_rcu_expedited() instead
4086 of synchronize_rcu(). This reduces latency,
4087 but can increase CPU utilization, degrade
4088 real-time latency, and degrade energy efficiency.
4089 No effect on CONFIG_TINY_RCU kernels.
4091 rcupdate.rcu_normal= [KNL]
4092 Use only normal grace-period primitives,
4093 for example, synchronize_rcu() instead of
4094 synchronize_rcu_expedited(). This improves
4095 real-time latency, CPU utilization, and
4096 energy efficiency, but can expose users to
4097 increased grace-period latency. This parameter
4098 overrides rcupdate.rcu_expedited. No effect on
4099 CONFIG_TINY_RCU kernels.
4101 rcupdate.rcu_normal_after_boot= [KNL]
4102 Once boot has completed (that is, after
4103 rcu_end_inkernel_boot() has been invoked), use
4104 only normal grace-period primitives. No effect
4105 on CONFIG_TINY_RCU kernels.
4107 rcupdate.rcu_task_stall_timeout= [KNL]
4108 Set timeout in jiffies for RCU task stall warning
4109 messages. Disable with a value less than or equal
4112 rcupdate.rcu_self_test= [KNL]
4113 Run the RCU early boot self tests
4117 Run specified binary instead of /init from the ramdisk,
4118 used for early userspace startup. See initrd.
4121 force - Override the decision by the kernel to hide the
4122 advertisement of RDRAND support (this affects
4123 certain AMD processors because of buggy BIOS
4124 support, specifically around the suspend/resume
4128 Turn on/off individual RDT features. List is:
4129 cmt, mbmtotal, mbmlocal, l3cat, l3cdp, l2cat, l2cdp,
4131 E.g. to turn on cmt and turn off mba use:
4135 Format (x86 or x86_64):
4136 [w[arm] | c[old] | h[ard] | s[oft] | g[pio]] \
4138 [[,]b[ios] | a[cpi] | k[bd] | t[riple] | e[fi] | p[ci]] \
4140 Where reboot_mode is one of warm (soft) or cold (hard) or gpio
4141 (prefix with 'panic_' to set mode for panic
4143 reboot_type is one of bios, acpi, kbd, triple, efi, or pci,
4144 reboot_force is either force or not specified,
4145 reboot_cpu is s[mp]#### with #### being the processor
4146 to be used for rebooting.
4149 [KNL, SMP] Set scheduler's default relax_domain_level.
4150 See Documentation/admin-guide/cgroup-v1/cpusets.rst.
4152 reserve= [KNL,BUGS] Force kernel to ignore I/O ports or memory
4153 Format: <base1>,<size1>[,<base2>,<size2>,...]
4154 Reserve I/O ports or memory so the kernel won't use
4155 them. If <base> is less than 0x10000, the region
4156 is assumed to be I/O ports; otherwise it is memory.
4158 reservetop= [X86-32]
4160 Reserves a hole at the top of the kernel virtual
4165 Set the amount of memory to reserve for BIOS at
4166 the bottom of the address space.
4168 reset_devices [KNL] Force drivers to reset the underlying device
4169 during initialization.
4172 Specify the partition device for software suspend
4174 {/dev/<dev> | PARTUUID=<uuid> | <int>:<int> | <hex>}
4176 resume_offset= [SWSUSP]
4177 Specify the offset from the beginning of the partition
4178 given by "resume=" at which the swap header is located,
4179 in <PAGE_SIZE> units (needed only for swap files).
4180 See Documentation/power/swsusp-and-swap-files.rst
4182 resumedelay= [HIBERNATION] Delay (in seconds) to pause before attempting to
4183 read the resume files
4185 resumewait [HIBERNATION] Wait (indefinitely) for resume device to show up.
4186 Useful for devices that are detected asynchronously
4187 (e.g. USB and MMC devices).
4189 hibernate= [HIBERNATION]
4190 noresume Don't check if there's a hibernation image
4191 present during boot.
4192 nocompress Don't compress/decompress hibernation images.
4193 no Disable hibernation and resume.
4194 protect_image Turn on image protection during restoration
4195 (that will set all pages holding image data
4196 during restoration read-only).
4198 retain_initrd [RAM] Keep initrd memory after extraction
4200 rfkill.default_state=
4201 0 "airplane mode". All wifi, bluetooth, wimax, gps, fm,
4202 etc. communication is blocked by default.
4205 rfkill.master_switch_mode=
4206 0 The "airplane mode" button does nothing.
4207 1 The "airplane mode" button toggles between everything
4208 blocked and the previous configuration.
4209 2 The "airplane mode" button toggles between everything
4210 blocked and everything unblocked.
4212 rhash_entries= [KNL,NET]
4213 Set number of hash buckets for route cache
4216 [KNL] Disable ring 3 MONITOR/MWAIT feature on supported
4219 ro [KNL] Mount root device read-only on boot
4222 on Mark read-only kernel memory as read-only (default).
4223 off Leave read-only kernel memory writable for debugging.
4226 Enable the uart passthrough on the designated usb port
4227 on Rockchip SoCs. When active, the signals of the
4228 debug-uart get routed to the D+ and D- pins of the usb
4229 port and the regular usb controller gets disabled.
4231 root= [KNL] Root filesystem
4232 See name_to_dev_t comment in init/do_mounts.c.
4234 rootdelay= [KNL] Delay (in seconds) to pause before attempting to
4235 mount the root filesystem
4237 rootflags= [KNL] Set root filesystem mount option string
4239 rootfstype= [KNL] Set root filesystem type
4241 rootwait [KNL] Wait (indefinitely) for root device to show up.
4242 Useful for devices that are detected asynchronously
4243 (e.g. USB and MMC devices).
4245 rproc_mem=nn[KMG][@address]
4246 [KNL,ARM,CMA] Remoteproc physical memory block.
4247 Memory area to be used by remote processor image,
4250 rw [KNL] Mount root device read-write on boot
4252 S [KNL] Run init in single mode
4254 s390_iommu= [HW,S390]
4255 Set s390 IOTLB flushing mode
4257 With strict flushing every unmap operation will result in
4258 an IOTLB flush. Default is lazy flushing before reuse,
4262 See drivers/net/irda/sa1100_ir.c.
4264 sbni= [NET] Granch SBNI12 leased line adapter
4266 sched_debug [KNL] Enables verbose scheduler debug messages.
4268 schedstats= [KNL,X86] Enable or disable scheduled statistics.
4269 Allowed values are enable and disable. This feature
4270 incurs a small amount of overhead in the scheduler
4271 but is useful for debugging and performance tuning.
4273 skew_tick= [KNL] Offset the periodic timer tick per cpu to mitigate
4274 xtime_lock contention on larger systems, and/or RCU lock
4275 contention on all systems with CONFIG_MAXSMP set.
4276 Format: { "0" | "1" }
4277 0 -- disable. (may be 1 via CONFIG_CMDLINE="skew_tick=1"
4279 Note: increases power consumption, thus should only be
4280 enabled if running jitter sensitive (HPC/RT) workloads.
4282 security= [SECURITY] Choose a legacy "major" security module to
4283 enable at boot. This has been deprecated by the
4286 selinux= [SELINUX] Disable or enable SELinux at boot time.
4287 Format: { "0" | "1" }
4288 See security/selinux/Kconfig help text.
4291 Default value is set via kernel config option.
4292 If enabled at boot time, /selinux/disable can be used
4293 later to disable prior to initial policy load.
4295 apparmor= [APPARMOR] Disable or enable AppArmor at boot time
4296 Format: { "0" | "1" }
4297 See security/apparmor/Kconfig help text
4300 Default value is set via kernel config option.
4302 serialnumber [BUGS=X86-32]
4305 Maximal number of shapers.
4313 Disable merging of slabs with similar size. May be
4314 necessary if there is some reason to distinguish
4315 allocs to different slabs, especially in hardened
4316 environments where the risk of heap overflows and
4317 layout control by attackers can usually be
4318 frustrated by disabling merging. This will reduce
4319 most of the exposure of a heap attack to a single
4320 cache (risks via metadata attacks are mostly
4321 unchanged). Debug options disable merging on their
4323 For more information see Documentation/vm/slub.rst.
4325 slab_max_order= [MM, SLAB]
4326 Determines the maximum allowed order for slabs.
4327 A high setting may cause OOMs due to memory
4328 fragmentation. Defaults to 1 for systems with
4329 more than 32MB of RAM, 0 otherwise.
4331 slub_debug[=options[,slabs]] [MM, SLUB]
4332 Enabling slub_debug allows one to determine the
4333 culprit if slab objects become corrupted. Enabling
4334 slub_debug can create guard zones around objects and
4335 may poison objects when not in use. Also tracks the
4336 last alloc / free. For more information see
4337 Documentation/vm/slub.rst.
4339 slub_memcg_sysfs= [MM, SLUB]
4340 Determines whether to enable sysfs directories for
4341 memory cgroup sub-caches. 1 to enable, 0 to disable.
4342 The default is determined by CONFIG_SLUB_MEMCG_SYSFS_ON.
4343 Enabling this can lead to a very high number of debug
4344 directories and files being created under
4347 slub_max_order= [MM, SLUB]
4348 Determines the maximum allowed order for slabs.
4349 A high setting may cause OOMs due to memory
4350 fragmentation. For more information see
4351 Documentation/vm/slub.rst.
4353 slub_min_objects= [MM, SLUB]
4354 The minimum number of objects per slab. SLUB will
4355 increase the slab order up to slub_max_order to
4356 generate a sufficiently large slab able to contain
4357 the number of objects indicated. The higher the number
4358 of objects the smaller the overhead of tracking slabs
4359 and the less frequently locks need to be acquired.
4360 For more information see Documentation/vm/slub.rst.
4362 slub_min_order= [MM, SLUB]
4363 Determines the minimum page order for slabs. Must be
4364 lower than slub_max_order.
4365 For more information see Documentation/vm/slub.rst.
4367 slub_nomerge [MM, SLUB]
4368 Same with slab_nomerge. This is supported for legacy.
4369 See slab_nomerge for more information.
4372 Format: <io1>[,<io2>[,...,<io8>]]
4374 smsc-ircc2.nopnp [HW] Don't use PNP to discover SMC devices
4375 smsc-ircc2.ircc_cfg= [HW] Device configuration I/O port
4376 smsc-ircc2.ircc_sir= [HW] SIR base I/O port
4377 smsc-ircc2.ircc_fir= [HW] FIR base I/O port
4378 smsc-ircc2.ircc_irq= [HW] IRQ line
4379 smsc-ircc2.ircc_dma= [HW] DMA channel
4380 smsc-ircc2.ircc_transceiver= [HW] Transceiver type:
4381 0: Toshiba Satellite 1800 (GP data pin select)
4382 1: Fast pin select (default)
4385 smt [KNL,S390] Set the maximum number of threads (logical
4386 CPUs) to use per physical CPU on systems capable of
4387 symmetric multithreading (SMT). Will be capped to the
4388 actual hardware limit.
4390 Default: -1 (no limit)
4393 [KNL] Should the soft-lockup detector generate panics.
4396 A nonzero value instructs the soft-lockup detector
4397 to panic the machine when a soft-lockup occurs. This
4398 is also controlled by CONFIG_BOOTPARAM_SOFTLOCKUP_PANIC
4399 which is the respective build-time switch to that
4402 softlockup_all_cpu_backtrace=
4403 [KNL] Should the soft-lockup detector generate
4404 backtraces on all cpus.
4407 sonypi.*= [HW] Sony Programmable I/O Control Device driver
4408 See Documentation/admin-guide/laptops/sonypi.rst
4410 spectre_v2= [X86] Control mitigation of Spectre variant 2
4411 (indirect branch speculation) vulnerability.
4412 The default operation protects the kernel from
4415 on - unconditionally enable, implies
4417 off - unconditionally disable, implies
4419 auto - kernel detects whether your CPU model is
4422 Selecting 'on' will, and 'auto' may, choose a
4423 mitigation method at run time according to the
4424 CPU, the available microcode, the setting of the
4425 CONFIG_RETPOLINE configuration option, and the
4426 compiler with which the kernel was built.
4428 Selecting 'on' will also enable the mitigation
4429 against user space to user space task attacks.
4431 Selecting 'off' will disable both the kernel and
4432 the user space protections.
4434 Specific mitigations can also be selected manually:
4436 retpoline - replace indirect branches
4437 retpoline,generic - google's original retpoline
4438 retpoline,amd - AMD-specific minimal thunk
4440 Not specifying this option is equivalent to
4444 [X86] Control mitigation of Spectre variant 2
4445 (indirect branch speculation) vulnerability between
4448 on - Unconditionally enable mitigations. Is
4449 enforced by spectre_v2=on
4451 off - Unconditionally disable mitigations. Is
4452 enforced by spectre_v2=off
4454 prctl - Indirect branch speculation is enabled,
4455 but mitigation can be enabled via prctl
4456 per thread. The mitigation control state
4457 is inherited on fork.
4460 - Like "prctl" above, but only STIBP is
4461 controlled per thread. IBPB is issued
4462 always when switching between different user
4466 - Same as "prctl" above, but all seccomp
4467 threads will enable the mitigation unless
4468 they explicitly opt out.
4471 - Like "seccomp" above, but only STIBP is
4472 controlled per thread. IBPB is issued
4473 always when switching between different
4474 user space processes.
4476 auto - Kernel selects the mitigation depending on
4477 the available CPU features and vulnerability.
4480 If CONFIG_SECCOMP=y then "seccomp", otherwise "prctl"
4482 Not specifying this option is equivalent to
4483 spectre_v2_user=auto.
4485 spec_store_bypass_disable=
4486 [HW] Control Speculative Store Bypass (SSB) Disable mitigation
4487 (Speculative Store Bypass vulnerability)
4489 Certain CPUs are vulnerable to an exploit against a
4490 a common industry wide performance optimization known
4491 as "Speculative Store Bypass" in which recent stores
4492 to the same memory location may not be observed by
4493 later loads during speculative execution. The idea
4494 is that such stores are unlikely and that they can
4495 be detected prior to instruction retirement at the
4496 end of a particular speculation execution window.
4498 In vulnerable processors, the speculatively forwarded
4499 store can be used in a cache side channel attack, for
4500 example to read memory to which the attacker does not
4501 directly have access (e.g. inside sandboxed code).
4503 This parameter controls whether the Speculative Store
4504 Bypass optimization is used.
4506 On x86 the options are:
4508 on - Unconditionally disable Speculative Store Bypass
4509 off - Unconditionally enable Speculative Store Bypass
4510 auto - Kernel detects whether the CPU model contains an
4511 implementation of Speculative Store Bypass and
4512 picks the most appropriate mitigation. If the
4513 CPU is not vulnerable, "off" is selected. If the
4514 CPU is vulnerable the default mitigation is
4515 architecture and Kconfig dependent. See below.
4516 prctl - Control Speculative Store Bypass per thread
4517 via prctl. Speculative Store Bypass is enabled
4518 for a process by default. The state of the control
4519 is inherited on fork.
4520 seccomp - Same as "prctl" above, but all seccomp threads
4521 will disable SSB unless they explicitly opt out.
4523 Default mitigations:
4524 X86: If CONFIG_SECCOMP=y "seccomp", otherwise "prctl"
4526 On powerpc the options are:
4528 on,auto - On Power8 and Power9 insert a store-forwarding
4529 barrier on kernel entry and exit. On Power7
4530 perform a software flush on kernel entry and
4534 Not specifying this option is equivalent to
4535 spec_store_bypass_disable=auto.
4537 spia_io_base= [HW,MTD]
4542 srcutree.counter_wrap_check [KNL]
4543 Specifies how frequently to check for
4544 grace-period sequence counter wrap for the
4545 srcu_data structure's ->srcu_gp_seq_needed field.
4546 The greater the number of bits set in this kernel
4547 parameter, the less frequently counter wrap will
4548 be checked for. Note that the bottom two bits
4551 srcutree.exp_holdoff [KNL]
4552 Specifies how many nanoseconds must elapse
4553 since the end of the last SRCU grace period for
4554 a given srcu_struct until the next normal SRCU
4555 grace period will be considered for automatic
4556 expediting. Set to zero to disable automatic
4560 Speculative Store Bypass Disable control
4562 On CPUs that are vulnerable to the Speculative
4563 Store Bypass vulnerability and offer a
4564 firmware based mitigation, this parameter
4565 indicates how the mitigation should be used:
4567 force-on: Unconditionally enable mitigation for
4568 for both kernel and userspace
4569 force-off: Unconditionally disable mitigation for
4570 for both kernel and userspace
4571 kernel: Always enable mitigation in the
4572 kernel, and offer a prctl interface
4573 to allow userspace to register its
4574 interest in being mitigated too.
4576 stack_guard_gap= [MM]
4577 override the default stack gap protection. The value
4578 is in page units and it defines how many pages prior
4579 to (for stacks growing down) resp. after (for stacks
4580 growing up) the main stack are reserved for no other
4581 mapping. Default value is 256 pages.
4584 Enabled the stack tracer on boot up.
4586 stacktrace_filter=[function-list]
4587 [FTRACE] Limit the functions that the stack tracer
4588 will trace at boot up. function-list is a comma separated
4589 list of functions. This list can be changed at run
4590 time by the stack_trace_filter file in the debugfs
4591 tracing directory. Note, this enables stack tracing
4592 and the stacktrace above is not needed.
4596 Set the STI (builtin display/keyboard on the HP-PARISC
4597 machines) console (graphic card) which should be used
4598 as the initial boot-console.
4599 See also comment in drivers/video/console/sticore.c.
4602 See comment in drivers/video/console/sticore.c.
4605 Format: bpp:<bpp1>[:<bpp2>[:<bpp3>...]]
4607 sunrpc.min_resvport=
4608 sunrpc.max_resvport=
4610 SunRPC servers often require that client requests
4611 originate from a privileged port (i.e. a port in the
4612 range 0 < portnr < 1024).
4613 An administrator who wishes to reserve some of these
4614 ports for other uses may adjust the range that the
4615 kernel's sunrpc client considers to be privileged
4616 using these two parameters to set the minimum and
4617 maximum port values.
4619 sunrpc.svc_rpc_per_connection_limit=
4621 Limit the number of requests that the server will
4622 process in parallel from a single connection.
4623 The default value is 0 (no limit).
4627 Control how the NFS server code allocates CPUs to
4628 service thread pools. Depending on how many NICs
4629 you have and where their interrupts are bound, this
4630 option will affect which CPUs will do NFS serving.
4631 Note: this parameter cannot be changed while the
4632 NFS server is running.
4634 auto the server chooses an appropriate mode
4635 automatically using heuristics
4636 global a single global pool contains all CPUs
4637 percpu one pool for each CPU
4638 pernode one pool for each NUMA node (equivalent
4639 to global on non-NUMA machines)
4641 sunrpc.tcp_slot_table_entries=
4642 sunrpc.udp_slot_table_entries=
4644 Sets the upper limit on the number of simultaneous
4645 RPC calls that can be sent from the client to a
4646 server. Increasing these values may allow you to
4647 improve throughput, but will also increase the
4648 amount of memory reserved for use by the client.
4650 suspend.pm_test_delay=
4652 Sets the number of seconds to remain in a suspend test
4653 mode before resuming the system (see
4654 /sys/power/pm_test). Only available when CONFIG_PM_DEBUG
4655 is set. Default value is 5.
4658 [KNL] Enable accounting of swap in memory resource
4659 controller if no parameter or 1 is given or disable
4660 it if 0 is given (See Documentation/admin-guide/cgroup-v1/memory.rst)
4662 swiotlb= [ARM,IA-64,PPC,MIPS,X86]
4663 Format: { <int> | force | noforce }
4664 <int> -- Number of I/O TLB slabs
4665 force -- force using of bounce buffers even if they
4666 wouldn't be automatically used by the kernel
4667 noforce -- Never use bounce buffers (for debugging)
4671 sysfs.deprecated=0|1 [KNL]
4672 Enable/disable old style sysfs layout for old udev
4673 on older distributions. When this option is enabled
4674 very new udev will not work anymore. When this option
4675 is disabled (or CONFIG_SYSFS_DEPRECATED not compiled)
4676 in older udev will not work anymore.
4677 Default depends on CONFIG_SYSFS_DEPRECATED_V2 set in
4678 the kernel configuration.
4680 sysrq_always_enabled
4682 Ignore sysrq setting - this boot parameter will
4683 neutralize any effect of /proc/sys/kernel/sysrq.
4684 Useful for debugging.
4686 tcpmhash_entries= [KNL,NET]
4687 Set the number of tcp_metrics_hash slots.
4688 Default value is 8192 or 16384 depending on total
4689 ram pages. This is used to specify the TCP metrics
4690 cache size. See Documentation/networking/ip-sysctl.txt
4691 "tcp_no_metrics_save" section for more details.
4695 test_suspend= [SUSPEND][,N]
4696 Specify "mem" (for Suspend-to-RAM) or "standby" (for
4697 standby suspend) or "freeze" (for suspend type freeze)
4698 as the system sleep state during system startup with
4699 the optional capability to repeat N number of times.
4700 The system is woken from this state using a
4701 wakeup-capable RTC alarm.
4703 thash_entries= [KNL,NET]
4704 Set number of hash buckets for TCP connection
4706 thermal.act= [HW,ACPI]
4707 -1: disable all active trip points in all thermal zones
4708 <degrees C>: override all lowest active trip points
4710 thermal.crt= [HW,ACPI]
4711 -1: disable all critical trip points in all thermal zones
4712 <degrees C>: override all critical trip points
4714 thermal.nocrt= [HW,ACPI]
4715 Set to disable actions on ACPI thermal zone
4716 critical and hot trip points.
4718 thermal.off= [HW,ACPI]
4719 1: disable ACPI thermal control
4721 thermal.psv= [HW,ACPI]
4722 -1: disable all passive trip points
4723 <degrees C>: override all passive trip points to this
4726 thermal.tzp= [HW,ACPI]
4727 Specify global default ACPI thermal zone polling rate
4728 <deci-seconds>: poll all this frequency
4729 0: no polling (default)
4732 Force threading of all interrupt handlers except those
4733 marked explicitly IRQF_NO_THREAD.
4737 Specify if the kernel should make use of the cpu
4738 topology information if the hardware supports this.
4739 The scheduler will make use of this information and
4740 e.g. base its process migration decisions on it.
4743 topology_updates= [KNL, PPC, NUMA]
4745 Specify if the kernel should ignore (off)
4746 topology updates sent by the hypervisor to this
4751 tpm_suspend_pcr=[HW,TPM]
4752 Format: integer pcr id
4753 Specify that at suspend time, the tpm driver
4754 should extend the specified pcr with zeros,
4755 as a workaround for some chips which fail to
4756 flush the last written pcr on TPM_SaveState.
4757 This will guarantee that all the other pcrs
4760 trace_buf_size=nn[KMG]
4761 [FTRACE] will set tracing buffer size on each cpu.
4763 trace_event=[event-list]
4764 [FTRACE] Set and start specified trace events in order
4765 to facilitate early boot debugging. The event-list is a
4766 comma separated list of trace events to enable. See
4767 also Documentation/trace/events.rst
4769 trace_options=[option-list]
4770 [FTRACE] Enable or disable tracer options at boot.
4771 The option-list is a comma delimited list of options
4772 that can be enabled or disabled just as if you were
4773 to echo the option name into
4775 /sys/kernel/debug/tracing/trace_options
4777 For example, to enable stacktrace option (to dump the
4778 stack trace of each event), add to the command line:
4780 trace_options=stacktrace
4782 See also Documentation/trace/ftrace.rst "trace options"
4786 Have the tracepoints sent to printk as well as the
4787 tracing ring buffer. This is useful for early boot up
4788 where the system hangs or reboots and does not give the
4789 option for reading the tracing buffer or performing a
4790 ftrace_dump_on_oops.
4792 To turn off having tracepoints sent to printk,
4793 echo 0 > /proc/sys/kernel/tracepoint_printk
4794 Note, echoing 1 into this file without the
4795 tracepoint_printk kernel cmdline option has no effect.
4799 Having tracepoints sent to printk() and activating high
4800 frequency tracepoints such as irq or sched, can cause
4801 the system to live lock.
4804 [FTRACE] enable this option to disable tracing when a
4805 warning is hit. This turns off "tracing_on". Tracing can
4806 be enabled again by echoing '1' into the "tracing_on"
4807 file located in /sys/kernel/debug/tracing/
4809 This option is useful, as it disables the trace before
4810 the WARNING dump is called, which prevents the trace to
4811 be filled with content caused by the warning output.
4813 This option can also be set at run time via the sysctl
4814 option: kernel/traceoff_on_warning
4816 transparent_hugepage=
4818 Format: [always|madvise|never]
4819 Can be used to control the default behavior of the system
4820 with respect to transparent hugepages.
4821 See Documentation/admin-guide/mm/transhuge.rst
4824 tsc= Disable clocksource stability checks for TSC.
4826 [x86] reliable: mark tsc clocksource as reliable, this
4827 disables clocksource verification at runtime, as well
4828 as the stability checks done at bootup. Used to enable
4829 high-resolution timer mode on older hardware, and in
4830 virtualized environment.
4831 [x86] noirqtime: Do not use TSC to do irq accounting.
4832 Used to run time disable IRQ_TIME_ACCOUNTING on any
4833 platforms where RDTSC is slow and this accounting
4835 [x86] unstable: mark the TSC clocksource as unstable, this
4836 marks the TSC unconditionally unstable at bootup and
4837 avoids any further wobbles once the TSC watchdog notices.
4838 [x86] nowatchdog: disable clocksource watchdog. Used
4839 in situations with strict latency requirements (where
4840 interruptions from clocksource watchdog are not
4843 tsx= [X86] Control Transactional Synchronization
4844 Extensions (TSX) feature in Intel processors that
4845 support TSX control.
4847 This parameter controls the TSX feature. The options are:
4849 on - Enable TSX on the system. Although there are
4850 mitigations for all known security vulnerabilities,
4851 TSX has been known to be an accelerator for
4852 several previous speculation-related CVEs, and
4853 so there may be unknown security risks associated
4854 with leaving it enabled.
4856 off - Disable TSX on the system. (Note that this
4857 option takes effect only on newer CPUs which are
4858 not vulnerable to MDS, i.e., have
4859 MSR_IA32_ARCH_CAPABILITIES.MDS_NO=1 and which get
4860 the new IA32_TSX_CTRL MSR through a microcode
4861 update. This new MSR allows for the reliable
4862 deactivation of the TSX functionality.)
4864 auto - Disable TSX if X86_BUG_TAA is present,
4865 otherwise enable TSX on the system.
4867 Not specifying this option is equivalent to tsx=off.
4869 See Documentation/admin-guide/hw-vuln/tsx_async_abort.rst
4872 tsx_async_abort= [X86,INTEL] Control mitigation for the TSX Async
4873 Abort (TAA) vulnerability.
4875 Similar to Micro-architectural Data Sampling (MDS)
4876 certain CPUs that support Transactional
4877 Synchronization Extensions (TSX) are vulnerable to an
4878 exploit against CPU internal buffers which can forward
4879 information to a disclosure gadget under certain
4882 In vulnerable processors, the speculatively forwarded
4883 data can be used in a cache side channel attack, to
4884 access data to which the attacker does not have direct
4887 This parameter controls the TAA mitigation. The
4890 full - Enable TAA mitigation on vulnerable CPUs
4893 full,nosmt - Enable TAA mitigation and disable SMT on
4894 vulnerable CPUs. If TSX is disabled, SMT
4895 is not disabled because CPU is not
4896 vulnerable to cross-thread TAA attacks.
4897 off - Unconditionally disable TAA mitigation
4899 Not specifying this option is equivalent to
4900 tsx_async_abort=full. On CPUs which are MDS affected
4901 and deploy MDS mitigation, TAA mitigation is not
4902 required and doesn't provide any additional
4906 Documentation/admin-guide/hw-vuln/tsx_async_abort.rst
4908 turbografx.map[2|3]= [HW,JOY]
4909 TurboGraFX parallel port interface
4911 <port#>,<js1>,<js2>,<js3>,<js4>,<js5>,<js6>,<js7>
4912 See also Documentation/input/devices/joystick-parport.rst
4914 udbg-immortal [PPC] When debugging early kernel crashes that
4915 happen after console_init() and before a proper
4916 console driver takes over, this boot options might
4917 help "seeing" what's going on.
4919 uhash_entries= [KNL,NET]
4920 Set number of hash buckets for UDP/UDP-Lite connections
4923 [USB] Ignore overcurrent events (default N).
4924 Some badly-designed motherboards generate lots of
4925 bogus events, for ports that aren't wired to
4926 anything. Set this parameter to avoid log spamming.
4927 Note that genuine overcurrent events won't be
4931 [X86] Cause panic on unknown NMI.
4933 usbcore.authorized_default=
4934 [USB] Default USB device authorization:
4935 (default -1 = authorized except for wireless USB,
4936 0 = not authorized, 1 = authorized, 2 = authorized
4937 if device connected to internal port)
4939 usbcore.autosuspend=
4940 [USB] The autosuspend time delay (in seconds) used
4941 for newly-detected USB devices (default 2). This
4942 is the time required before an idle device will be
4943 autosuspended. Devices for which the delay is set
4944 to a negative value won't be autosuspended at all.
4946 usbcore.usbfs_snoop=
4947 [USB] Set to log all usbfs traffic (default 0 = off).
4949 usbcore.usbfs_snoop_max=
4950 [USB] Maximum number of bytes to snoop in each URB
4953 usbcore.blinkenlights=
4954 [USB] Set to cycle leds on hubs (default 0 = off).
4956 usbcore.old_scheme_first=
4957 [USB] Start with the old device initialization
4958 scheme, applies only to low and full-speed devices
4961 usbcore.usbfs_memory_mb=
4962 [USB] Memory limit (in MB) for buffers allocated by
4963 usbfs (default = 16, 0 = max = 2047).
4965 usbcore.use_both_schemes=
4966 [USB] Try the other device initialization scheme
4967 if the first one fails (default 1 = enabled).
4969 usbcore.initial_descriptor_timeout=
4970 [USB] Specifies timeout for the initial 64-byte
4971 USB_REQ_GET_DESCRIPTOR request in milliseconds
4972 (default 5000 = 5.0 seconds).
4974 usbcore.nousb [USB] Disable the USB subsystem
4977 [USB] A list of quirk entries to augment the built-in
4978 usb core quirk list. List entries are separated by
4979 commas. Each entry has the form
4980 VendorID:ProductID:Flags. The IDs are 4-digit hex
4981 numbers and Flags is a set of letters. Each letter
4982 will change the built-in quirk; setting it if it is
4983 clear and clearing it if it is set. The letters have
4984 the following meanings:
4985 a = USB_QUIRK_STRING_FETCH_255 (string
4986 descriptors must not be fetched using
4988 b = USB_QUIRK_RESET_RESUME (device can't resume
4989 correctly so reset it instead);
4990 c = USB_QUIRK_NO_SET_INTF (device can't handle
4991 Set-Interface requests);
4992 d = USB_QUIRK_CONFIG_INTF_STRINGS (device can't
4993 handle its Configuration or Interface
4995 e = USB_QUIRK_RESET (device can't be reset
4996 (e.g morph devices), don't use reset);
4997 f = USB_QUIRK_HONOR_BNUMINTERFACES (device has
4998 more interface descriptions than the
4999 bNumInterfaces count, and can't handle
5000 talking to these interfaces);
5001 g = USB_QUIRK_DELAY_INIT (device needs a pause
5002 during initialization, after we read
5003 the device descriptor);
5004 h = USB_QUIRK_LINEAR_UFRAME_INTR_BINTERVAL (For
5005 high speed and super speed interrupt
5006 endpoints, the USB 2.0 and USB 3.0 spec
5007 require the interval in microframes (1
5008 microframe = 125 microseconds) to be
5009 calculated as interval = 2 ^
5011 Devices with this quirk report their
5012 bInterval as the result of this
5013 calculation instead of the exponent
5014 variable used in the calculation);
5015 i = USB_QUIRK_DEVICE_QUALIFIER (device can't
5016 handle device_qualifier descriptor
5018 j = USB_QUIRK_IGNORE_REMOTE_WAKEUP (device
5019 generates spurious wakeup, ignore
5020 remote wakeup capability);
5021 k = USB_QUIRK_NO_LPM (device can't handle Link
5023 l = USB_QUIRK_LINEAR_FRAME_INTR_BINTERVAL
5024 (Device reports its bInterval as linear
5025 frames instead of the USB 2.0
5027 m = USB_QUIRK_DISCONNECT_SUSPEND (Device needs
5028 to be disconnected before suspend to
5029 prevent spurious wakeup);
5030 n = USB_QUIRK_DELAY_CTRL_MSG (Device needs a
5031 pause after every control message);
5032 o = USB_QUIRK_HUB_SLOW_RESET (Hub needs extra
5033 delay after resetting its port);
5034 Example: quirks=0781:5580:bk,0a5c:5834:gij
5037 [USBHID] The interval which mice are to be polled at.
5040 [USBHID] The interval which joysticks are to be polled at.
5043 [USBHID] The interval which keyboards are to be polled at.
5045 usb-storage.delay_use=
5046 [UMS] The delay in seconds before a new device is
5047 scanned for Logical Units (default 1).
5050 [UMS] A list of quirks entries to supplement or
5051 override the built-in unusual_devs list. List
5052 entries are separated by commas. Each entry has
5053 the form VID:PID:Flags where VID and PID are Vendor
5054 and Product ID values (4-digit hex numbers) and
5055 Flags is a set of characters, each corresponding
5056 to a common usb-storage quirk flag as follows:
5057 a = SANE_SENSE (collect more than 18 bytes
5059 b = BAD_SENSE (don't collect more than 18
5060 bytes of sense data);
5061 c = FIX_CAPACITY (decrease the reported
5062 device capacity by one sector);
5063 d = NO_READ_DISC_INFO (don't use
5064 READ_DISC_INFO command);
5065 e = NO_READ_CAPACITY_16 (don't use
5066 READ_CAPACITY_16 command);
5067 f = NO_REPORT_OPCODES (don't use report opcodes
5069 g = MAX_SECTORS_240 (don't transfer more than
5070 240 sectors at a time, uas only);
5071 h = CAPACITY_HEURISTICS (decrease the
5072 reported device capacity by one
5073 sector if the number is odd);
5074 i = IGNORE_DEVICE (don't bind to this
5076 j = NO_REPORT_LUNS (don't use report luns
5078 l = NOT_LOCKABLE (don't try to lock and
5079 unlock ejectable media);
5080 m = MAX_SECTORS_64 (don't transfer more
5081 than 64 sectors = 32 KB at a time);
5082 n = INITIAL_READ10 (force a retry of the
5083 initial READ(10) command);
5084 o = CAPACITY_OK (accept the capacity
5085 reported by the device);
5086 p = WRITE_CACHE (the device cache is ON
5088 r = IGNORE_RESIDUE (the device reports
5089 bogus residue values);
5090 s = SINGLE_LUN (the device has only one
5092 t = NO_ATA_1X (don't allow ATA(12) and ATA(16)
5093 commands, uas only);
5094 u = IGNORE_UAS (don't bind to the uas driver);
5095 w = NO_WP_DETECT (don't test whether the
5096 medium is write-protected).
5097 y = ALWAYS_SYNC (issue a SYNCHRONIZE_CACHE
5098 even if the device claims no cache)
5099 Example: quirks=0419:aaf5:rl,0421:0433:rc
5101 user_debug= [KNL,ARM]
5103 See arch/arm/Kconfig.debug help text.
5104 1 - undefined instruction events
5106 4 - invalid data aborts
5109 Example: user_debug=31
5112 [X86] Flags controlling user PTE allocations.
5114 nohigh = do not allocate PTE pages in
5115 HIGHMEM regardless of setting
5119 On X86_32, this is an alias for vdso32=. Otherwise:
5121 vdso=1: enable VDSO (the default)
5122 vdso=0: disable VDSO mapping
5124 vdso32= [X86] Control the 32-bit vDSO
5125 vdso32=1: enable 32-bit VDSO
5126 vdso32=0 or vdso32=2: disable 32-bit VDSO
5128 See the help text for CONFIG_COMPAT_VDSO for more
5129 details. If CONFIG_COMPAT_VDSO is set, the default is
5130 vdso32=0; otherwise, the default is vdso32=1.
5132 For compatibility with older kernels, vdso32=2 is an
5135 Try vdso32=0 if you encounter an error that says:
5136 dl_main: Assertion `(void *) ph->p_vaddr == _rtld_local._dl_sysinfo_dso' failed!
5139 vector=percpu: enable percpu vector domain
5141 video= [FB] Frame buffer configuration
5142 See Documentation/fb/modedb.rst.
5144 video.brightness_switch_enabled= [0,1]
5145 If set to 1, on receiving an ACPI notify event
5146 generated by hotkey, video driver will adjust brightness
5147 level and then send out the event to user space through
5148 the allocated input device; If set to 0, video driver
5149 will only send out the event without touching backlight
5154 [VMMIO] Memory mapped virtio (platform) device.
5156 <size>@<baseaddr>:<irq>[:<id>]
5158 <size> := size (can use standard suffixes
5160 <baseaddr> := physical base address
5161 <irq> := interrupt number (as passed to
5163 <id> := (optional) platform device id
5165 virtio_mmio.device=1K@0x100b0000:48:7
5167 Can be used multiple times for multiple devices.
5169 vga= [BOOT,X86-32] Select a particular video mode
5170 See Documentation/x86/boot.rst and
5171 Documentation/admin-guide/svga.rst.
5172 Use vga=ask for menu.
5173 This is actually a boot loader parameter; the value is
5174 passed to the kernel using a special protocol.
5176 vm_debug[=options] [KNL] Available with CONFIG_DEBUG_VM=y.
5177 May slow down system boot speed, especially when
5178 enabled on systems with a large amount of memory.
5179 All options are enabled by default, and this
5180 interface is meant to allow for selectively
5181 enabling or disabling specific virtual memory
5184 Available options are:
5185 P Enable page structure init time poisoning
5186 - Disable all of the above options
5188 vmalloc=nn[KMG] [KNL,BOOT] Forces the vmalloc area to have an exact
5189 size of <nn>. This can be used to increase the
5190 minimum size (128MB on x86). It can also be used to
5191 decrease the size and leave more room for directly
5194 vmcp_cma=nn[MG] [KNL,S390]
5195 Sets the memory size reserved for contiguous memory
5196 allocations for the vmcp device driver.
5198 vmhalt= [KNL,S390] Perform z/VM CP command after system halt.
5201 vmpanic= [KNL,S390] Perform z/VM CP command after kernel panic.
5204 vmpoff= [KNL,S390] Perform z/VM CP command after power off.
5208 Controls the behavior of vsyscalls (i.e. calls to
5209 fixed addresses of 0xffffffffff600x00 from legacy
5210 code). Most statically-linked binaries and older
5211 versions of glibc use these calls. Because these
5212 functions are at fixed addresses, they make nice
5213 targets for exploits that can control RIP.
5215 emulate [default] Vsyscalls turn into traps and are
5216 emulated reasonably safely. The vsyscall
5219 xonly Vsyscalls turn into traps and are
5220 emulated reasonably safely. The vsyscall
5221 page is not readable.
5223 none Vsyscalls don't work at all. This makes
5224 them quite hard to use for exploits but
5225 might break your system.
5227 vt.color= [VT] Default text color.
5228 Format: 0xYX, X = foreground, Y = background.
5229 Default: 0x07 = light gray on black.
5231 vt.cur_default= [VT] Default cursor shape.
5232 Format: 0xCCBBAA, where AA, BB, and CC are the same as
5233 the parameters of the <Esc>[?A;B;Cc escape sequence;
5234 see VGA-softcursor.txt. Default: 2 = underline.
5236 vt.default_blu= [VT]
5237 Format: <blue0>,<blue1>,<blue2>,...,<blue15>
5238 Change the default blue palette of the console.
5239 This is a 16-member array composed of values
5242 vt.default_grn= [VT]
5243 Format: <green0>,<green1>,<green2>,...,<green15>
5244 Change the default green palette of the console.
5245 This is a 16-member array composed of values
5248 vt.default_red= [VT]
5249 Format: <red0>,<red1>,<red2>,...,<red15>
5250 Change the default red palette of the console.
5251 This is a 16-member array composed of values
5257 Set system-wide default UTF-8 mode for all tty's.
5258 Default is 1, i.e. UTF-8 mode is enabled for all
5259 newly opened terminals.
5261 vt.global_cursor_default=
5264 Set system-wide default for whether a cursor
5265 is shown on new VTs. Default is -1,
5266 i.e. cursors will be created by default unless
5267 overridden by individual drivers. 0 will hide
5268 cursors, 1 will display them.
5270 vt.italic= [VT] Default color for italic text; 0-15.
5273 vt.underline= [VT] Default color for underlined text; 0-15.
5276 watchdog timers [HW,WDT] For information on watchdog timers,
5277 see Documentation/watchdog/watchdog-parameters.rst
5278 or other driver-specific files in the
5279 Documentation/watchdog/ directory.
5283 Set the hard lockup detector stall duration
5284 threshold in seconds. The soft lockup detector
5285 threshold is set to twice the value. A value of 0
5286 disables both lockup detectors. Default is 10
5289 workqueue.watchdog_thresh=
5290 If CONFIG_WQ_WATCHDOG is configured, workqueue can
5291 warn stall conditions and dump internal state to
5292 help debugging. 0 disables workqueue stall
5293 detection; otherwise, it's the stall threshold
5294 duration in seconds. The default value is 30 and
5295 it can be updated at runtime by writing to the
5296 corresponding sysfs file.
5298 workqueue.disable_numa
5299 By default, all work items queued to unbound
5300 workqueues are affine to the NUMA nodes they're
5301 issued on, which results in better behavior in
5302 general. If NUMA affinity needs to be disabled for
5303 whatever reason, this option can be used. Note
5304 that this also can be controlled per-workqueue for
5305 workqueues visible under /sys/bus/workqueue/.
5307 workqueue.power_efficient
5308 Per-cpu workqueues are generally preferred because
5309 they show better performance thanks to cache
5310 locality; unfortunately, per-cpu workqueues tend to
5311 be more power hungry than unbound workqueues.
5313 Enabling this makes the per-cpu workqueues which
5314 were observed to contribute significantly to power
5315 consumption unbound, leading to measurably lower
5316 power usage at the cost of small performance
5319 The default value of this parameter is determined by
5320 the config option CONFIG_WQ_POWER_EFFICIENT_DEFAULT.
5322 workqueue.debug_force_rr_cpu
5323 Workqueue used to implicitly guarantee that work
5324 items queued without explicit CPU specified are put
5325 on the local CPU. This guarantee is no longer true
5326 and while local CPU is still preferred work items
5327 may be put on foreign CPUs. This debug option
5328 forces round-robin CPU selection to flush out
5329 usages which depend on the now broken guarantee.
5330 When enabled, memory and cache locality will be
5333 x2apic_phys [X86-64,APIC] Use x2apic physical mode instead of
5334 default x2apic cluster mode on platforms
5337 x86_intel_mid_timer= [X86-32,APBT]
5338 Choose timer option for x86 Intel MID platform.
5339 Two valid options are apbt timer only and lapic timer
5340 plus one apbt timer for broadcast timer.
5341 x86_intel_mid_timer=apbt_only | lapic_and_apbt
5343 xen_512gb_limit [KNL,X86-64,XEN]
5344 Restricts the kernel running paravirtualized under Xen
5345 to use only up to 512 GB of RAM. The reason to do so is
5346 crash analysis tools and Xen tools for doing domain
5347 save/restore/migration must be enabled to handle larger
5350 xen_emul_unplug= [HW,X86,XEN]
5351 Unplug Xen emulated devices
5352 Format: [unplug0,][unplug1]
5353 ide-disks -- unplug primary master IDE devices
5354 aux-ide-disks -- unplug non-primary-master IDE devices
5355 nics -- unplug network devices
5356 all -- unplug all emulated devices (NICs and IDE disks)
5357 unnecessary -- unplugging emulated devices is
5358 unnecessary even if the host did not respond to
5360 never -- do not unplug even if version check succeeds
5362 xen_legacy_crash [X86,XEN]
5363 Crash from Xen panic notifier, without executing late
5364 panic() code such as dumping handler.
5366 xen_nopvspin [X86,XEN]
5367 Disables the ticketlock slowpath using Xen PV
5371 Disables the PV optimizations forcing the HVM guest to
5372 run as generic HVM guest with no PV drivers.
5373 This option is obsoleted by the "nopv" option, which
5374 has equivalent effect for XEN platform.
5376 xen_scrub_pages= [XEN]
5377 Boolean option to control scrubbing pages before giving them back
5378 to Xen, for use by other domains. Can be also changed at runtime
5379 with /sys/devices/system/xen_memory/xen_memory0/scrub_pages.
5380 Default value controlled with CONFIG_XEN_SCRUB_PAGES_DEFAULT.
5382 xen_timer_slop= [X86-64,XEN]
5383 Set the timer slop (in nanoseconds) for the virtual Xen
5384 timers (default is 100000). This adjusts the minimum
5385 delta of virtualized Xen timers, where lower values
5386 improve timer resolution at the expense of processing
5387 more timer interrupts.
5389 nopv= [X86,XEN,KVM,HYPER_V,VMWARE]
5390 Disables the PV optimizations forcing the guest to run
5391 as generic guest with no PV drivers. Currently support
5392 XEN HVM, KVM, HYPER_V and VMWARE guest.
5394 xirc2ps_cs= [NET,PCMCIA]
5396 <irq>,<irq_mask>,<io>,<full_duplex>,<do_sound>,<lockup_hack>[,<irq2>[,<irq3>[,<irq4>]]]
5399 By default on POWER9 and above, the kernel will
5400 natively use the XIVE interrupt controller. This option
5401 allows the fallback firmware mode to be used:
5403 off Fallback to firmware control of XIVE interrupt
5404 controller on both pseries and powernv
5405 platforms. Only useful on POWER9 and above.
5407 xhci-hcd.quirks [USB,KNL]
5408 A hex value specifying bitmask with supplemental xhci
5409 host controller quirks. Meaning of each bit can be
5410 consulted in header drivers/usb/host/xhci.h.