1 AMD64 specific boot options
3 There are many others (usually documented in driver documentation), but
4 only the AMD64 specific ones are listed here.
8 Please see Documentation/x86/x86_64/machinecheck for sysfs runtime tunables.
10 mce=off disable machine check
11 mce=bootlog Enable logging of machine checks left over from booting.
12 Disabled by default on AMD because some BIOS leave bogus ones.
13 If your BIOS doesn't do that it's a good idea to enable though
14 to make sure you log even machine check events that result
15 in a reboot. On Intel systems it is enabled by default.
17 Disable boot machine check logging.
18 mce=tolerancelevel (number)
19 0: always panic on uncorrected errors, log corrected errors
20 1: panic or SIGBUS on uncorrected errors, log corrected errors
21 2: SIGBUS or log uncorrected errors, log corrected errors
22 3: never panic or SIGBUS, log all errors (for testing only)
24 Can be also set using sysfs which is preferable.
26 nomce (for compatibility with i386): same as mce=off
28 Everything else is in sysfs now.
32 apic Use IO-APIC. Default
34 noapic Don't use the IO-APIC.
36 disableapic Don't use the local APIC
38 nolapic Don't use the local APIC (alias for i386 compatibility)
40 pirq=... See Documentation/x86/i386/IO-APIC.txt
42 noapictimer Don't set up the APIC timer
44 no_timer_check Don't check the IO-APIC timer. This can work around
45 problems with incorrect timer initialization on some boards.
47 apicmaintimer Run time keeping from the local APIC timer instead
48 of using the PIT/HPET interrupt for this. This is useful
49 when the PIT/HPET interrupts are unreliable.
51 noapicmaintimer Don't do time keeping using the APIC timer.
52 Useful when this option was auto selected, but doesn't work.
55 Do APIC timer calibration using the pmtimer. Implies
56 apicmaintimer. Useful when your PIT timer is totally
61 syntax: earlyprintk=vga
62 earlyprintk=serial[,ttySn[,baudrate]]
64 The early console is useful when the kernel crashes before the
65 normal console is initialized. It is not enabled by
66 default because it has some cosmetic problems.
67 Append ,keep to not disable it when the real console takes over.
68 Only vga or serial at a time, not both.
69 Currently only ttyS0 and ttyS1 are supported.
70 Interaction with the standard serial driver is not very good.
71 The VGA output is eventually overwritten by the real console.
76 Don't use the CPU time stamp counter to read the wall time.
77 This can be used to work around timing problems on multiprocessor systems
78 with not properly synchronized CPUs.
81 Report when timer interrupts are lost because some code turned off
82 interrupts for too long.
85 Don't use the HPET timer.
90 Don't do power saving in the idle loop using HLT, but poll for rescheduling
91 event. This will make the CPUs eat a lot more power, but may be useful
92 to get slightly better performance in multiprocessor benchmarks. It also
93 makes some profiling using performance counters more accurate.
94 Please note that on systems with MONITOR/MWAIT support (like Intel EM64T
95 CPUs) this option has no performance advantage over the normal idle loop.
96 It may also interact badly with hyperthreading.
100 reboot=b[ios] | t[riple] | k[bd] | a[cpi] | e[fi] [, [w]arm | [c]old]
101 bios Use the CPU reboot vector for warm reset
102 warm Don't set the cold reboot flag
103 cold Set the cold reboot flag
104 triple Force a triple fault (init)
105 kbd Use the keyboard controller. cold reset (default)
106 acpi Use the ACPI RESET_REG in the FADT. If ACPI is not configured or the
107 ACPI reset does not work, the reboot path attempts the reset using
108 the keyboard controller.
109 efi Use efi reset_system runtime service. If EFI is not configured or the
110 EFI reset does not work, the reboot path attempts the reset using
111 the keyboard controller.
113 Using warm reset will be much faster especially on big memory
114 systems because the BIOS will not go through the memory check.
115 Disadvantage is that not all hardware will be completely reinitialized
116 on reboot so there may be boot problems on some systems.
120 Don't stop other CPUs on reboot. This can make reboot more reliable
123 Non Executable Mappings
132 additional_cpus=NUM Allow NUM more CPUs for hotplug
133 (defaults are specified by the BIOS, see Documentation/x86/x86_64/cpu-hotplug-spec)
137 numa=off Only set up a single NUMA node spanning all memory.
139 numa=noacpi Don't parse the SRAT table for NUMA setup
142 If a number, fakes CMDLINE nodes and ignores NUMA setup of the
143 actual machine. Otherwise, system memory is configured
144 depending on the sizes and coefficients listed. For example:
145 numa=fake=2*512,1024,4*256,*128
146 gives two 512M nodes, a 1024M node, four 256M nodes, and the
147 rest split into 128M chunks. If the last character of CMDLINE
148 is a *, the remaining memory is divided up equally among its
151 gives two 512M nodes and the rest split into two nodes.
152 Otherwise, the remaining system RAM is allocated to an
156 Only allow hotadd memory to preallocate page structures upto
157 percent of already available memory.
158 numa=hotadd=0 will disable hotadd memory.
162 acpi=off Don't enable ACPI
163 acpi=ht Use ACPI boot table parsing, but don't enable ACPI
165 acpi=force Force ACPI on (currently not needed)
167 acpi=strict Disable out of spec ACPI workarounds.
169 acpi_sci={edge,level,high,low} Set up ACPI SCI interrupt.
171 acpi=noirq Don't route interrupts
175 pci=off Don't use PCI
176 pci=conf1 Use conf1 access.
177 pci=conf2 Use conf2 access.
179 pci=assign-busses Assign busses
180 pci=irqmask=MASK Set PCI interrupt mask to MASK
181 pci=lastbus=NUMBER Scan upto NUMBER busses, no matter what the mptable says.
182 pci=noacpi Don't use ACPI to set up PCI interrupt routing.
184 IOMMU (input/output memory management unit)
186 Currently four x86-64 PCI-DMA mapping implementations exist:
188 1. <arch/x86_64/kernel/pci-nommu.c>: use no hardware/software IOMMU at all
189 (e.g. because you have < 3 GB memory).
190 Kernel boot message: "PCI-DMA: Disabling IOMMU"
192 2. <arch/x86_64/kernel/pci-gart.c>: AMD GART based hardware IOMMU.
193 Kernel boot message: "PCI-DMA: using GART IOMMU"
195 3. <arch/x86_64/kernel/pci-swiotlb.c> : Software IOMMU implementation. Used
196 e.g. if there is no hardware IOMMU in the system and it is need because
197 you have >3GB memory or told the kernel to us it (iommu=soft))
198 Kernel boot message: "PCI-DMA: Using software bounce buffering
201 4. <arch/x86_64/pci-calgary.c> : IBM Calgary hardware IOMMU. Used in IBM
202 pSeries and xSeries servers. This hardware IOMMU supports DMA address
203 mapping with memory protection, etc.
204 Kernel boot message: "PCI-DMA: Using Calgary IOMMU"
206 iommu=[<size>][,noagp][,off][,force][,noforce][,leak[=<nr_of_leak_pages>]
207 [,memaper[=<order>]][,merge][,forcesac][,fullflush][,nomerge]
208 [,noaperture][,calgary]
210 General iommu options:
211 off Don't initialize and use any kind of IOMMU.
212 noforce Don't force hardware IOMMU usage when it is not needed.
214 force Force the use of the hardware IOMMU even when it is
215 not actually needed (e.g. because < 3 GB memory).
216 soft Use software bounce buffering (SWIOTLB) (default for
217 Intel machines). This can be used to prevent the usage
218 of an available hardware IOMMU.
220 iommu options only relevant to the AMD GART hardware IOMMU:
221 <size> Set the size of the remapping area in bytes.
222 allowed Overwrite iommu off workarounds for specific chipsets.
223 fullflush Flush IOMMU on each allocation (default).
224 nofullflush Don't use IOMMU fullflush.
225 leak Turn on simple iommu leak tracing (only when
226 CONFIG_IOMMU_LEAK is on). Default number of leak pages
228 memaper[=<order>] Allocate an own aperture over RAM with size 32MB<<order.
229 (default: order=1, i.e. 64MB)
230 merge Do scatter-gather (SG) merging. Implies "force"
232 nomerge Don't do scatter-gather (SG) merging.
233 noaperture Ask the IOMMU not to touch the aperture for AGP.
234 forcesac Force single-address cycle (SAC) mode for masks <40bits
236 noagp Don't initialize the AGP driver and use full aperture.
237 allowdac Allow double-address cycle (DAC) mode, i.e. DMA >4GB.
238 DAC is used with 32-bit PCI to push a 64-bit address in
239 two cycles. When off all DMA over >4GB is forced through
240 an IOMMU or software bounce buffering.
241 nodac Forbid DAC mode, i.e. DMA >4GB.
242 panic Always panic when IOMMU overflows.
243 calgary Use the Calgary IOMMU if it is available
245 iommu options only relevant to the software bounce buffering (SWIOTLB) IOMMU
247 swiotlb=<pages>[,force]
248 <pages> Prereserve that many 128K pages for the software IO
250 force Force all IO through the software TLB.
252 Settings for the IBM Calgary hardware IOMMU currently found in IBM
253 pSeries and xSeries machines:
255 calgary=[64k,128k,256k,512k,1M,2M,4M,8M]
256 calgary=[translate_empty_slots]
257 calgary=[disable=<PCI bus number>]
258 panic Always panic when IOMMU overflows
260 64k,...,8M - Set the size of each PCI slot's translation table
261 when using the Calgary IOMMU. This is the size of the translation
262 table itself in main memory. The smallest table, 64k, covers an IO
263 space of 32MB; the largest, 8MB table, can cover an IO space of
264 4GB. Normally the kernel will make the right choice by itself.
266 translate_empty_slots - Enable translation even on slots that have
267 no devices attached to them, in case a device will be hotplugged
270 disable=<PCI bus number> - Disable translation on a given PHB. For
271 example, the built-in graphics adapter resides on the first bridge
272 (PCI bus number 0); if translation (isolation) is enabled on this
273 bridge, X servers that access the hardware directly from user
274 space might stop working. Use this option if you have devices that
275 are accessed from userspace directly on some PCI host bridge.
279 oops=panic Always panic on oopses. Default is to just kill the process,
280 but there is a small probability of deadlocking the machine.
281 This will also cause panics on machine check exceptions.
282 Useful together with panic=30 to trigger a reboot.
284 kstack=N Print N words from the kernel stack in oops dumps.
286 pagefaulttrace Dump all page faults. Only useful for extreme debugging
287 and will create a lot of output.
289 call_trace=[old|both|newfallback|new]
290 old: use old inexact backtracer
291 new: use new exact dwarf2 unwinder
292 both: print entries from both
293 newfallback: use new unwinder but fall back to old if it gets
299 Do not use GB pages for kernel direct mappings.
301 Use GB pages for kernel direct mappings.