Documentation/x86/x86_64/boot-options.txt

   1 AMD64 specific boot options
   2
   3 There are many others (usually documented in driver documentation), but
   4 only the AMD64 specific ones are listed here.
   5
   6 Machine check
   7
   8    Please see Documentation/x86/x86_64/machinecheck for sysfs runtime tunables.
   9
  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.
  16    mce=nobootlog
  17                 Disable boot machine check logging.
  18    mce=tolerancelevel[,monarchtimeout] (number,number)
  19                 tolerance levels:
  20                 0: always panic on uncorrected errors, log corrected errors
  21                 1: panic or SIGBUS on uncorrected errors, log corrected errors
  22                 2: SIGBUS or log uncorrected errors, log corrected errors
  23                 3: never panic or SIGBUS, log all errors (for testing only)
  24                 Default is 1
  25                 Can be also set using sysfs which is preferable.
  26                 monarchtimeout:
  27                 Sets the time in us to wait for other CPUs on machine checks. 0
  28                 to disable.
  29
  30    nomce (for compatibility with i386): same as mce=off
  31
  32    Everything else is in sysfs now.
  33
  34 APICs
  35
  36    apic          Use IO-APIC. Default
  37
  38    noapic        Don't use the IO-APIC.
  39
  40    disableapic   Don't use the local APIC
  41
  42    nolapic       Don't use the local APIC (alias for i386 compatibility)
  43
  44    pirq=...      See Documentation/x86/i386/IO-APIC.txt
  45
  46    noapictimer   Don't set up the APIC timer
  47
  48    no_timer_check Don't check the IO-APIC timer. This can work around
  49                  problems with incorrect timer initialization on some boards.
  50
  51    apicmaintimer Run time keeping from the local APIC timer instead
  52                  of using the PIT/HPET interrupt for this. This is useful
  53                  when the PIT/HPET interrupts are unreliable.
  54
  55    noapicmaintimer  Don't do time keeping using the APIC timer.
  56                  Useful when this option was auto selected, but doesn't work.
  57
  58    apicpmtimer
  59                  Do APIC timer calibration using the pmtimer. Implies
  60                  apicmaintimer. Useful when your PIT timer is totally
  61                  broken.
  62
  63 Early Console
  64
  65    syntax: earlyprintk=vga
  66            earlyprintk=serial[,ttySn[,baudrate]]
  67
  68    The early console is useful when the kernel crashes before the
  69    normal console is initialized. It is not enabled by
  70    default because it has some cosmetic problems.
  71    Append ,keep to not disable it when the real console takes over.
  72    Only vga or serial at a time, not both.
  73    Currently only ttyS0 and ttyS1 are supported.
  74    Interaction with the standard serial driver is not very good.
  75    The VGA output is eventually overwritten by the real console.
  76
  77 Timing
  78
  79   notsc
  80   Don't use the CPU time stamp counter to read the wall time.
  81   This can be used to work around timing problems on multiprocessor systems
  82   with not properly synchronized CPUs.
  83
  84   report_lost_ticks
  85   Report when timer interrupts are lost because some code turned off
  86   interrupts for too long.
  87
  88   nohpet
  89   Don't use the HPET timer.
  90
  91 Idle loop
  92
  93   idle=poll
  94   Don't do power saving in the idle loop using HLT, but poll for rescheduling
  95   event. This will make the CPUs eat a lot more power, but may be useful
  96   to get slightly better performance in multiprocessor benchmarks. It also
  97   makes some profiling using performance counters more accurate.
  98   Please note that on systems with MONITOR/MWAIT support (like Intel EM64T
  99   CPUs) this option has no performance advantage over the normal idle loop.
 100   It may also interact badly with hyperthreading.
 101
 102 Rebooting
 103
 104    reboot=b[ios] | t[riple] | k[bd] | a[cpi] | e[fi] [, [w]arm | [c]old]
 105    bios   Use the CPU reboot vector for warm reset
 106    warm   Don't set the cold reboot flag
 107    cold   Set the cold reboot flag
 108    triple Force a triple fault (init)
 109    kbd    Use the keyboard controller. cold reset (default)
 110    acpi   Use the ACPI RESET_REG in the FADT. If ACPI is not configured or the
 111           ACPI reset does not work, the reboot path attempts the reset using
 112           the keyboard controller.
 113    efi    Use efi reset_system runtime service. If EFI is not configured or the
 114           EFI reset does not work, the reboot path attempts the reset using
 115           the keyboard controller.
 116
 117    Using warm reset will be much faster especially on big memory
 118    systems because the BIOS will not go through the memory check.
 119    Disadvantage is that not all hardware will be completely reinitialized
 120    on reboot so there may be boot problems on some systems.
 121
 122    reboot=force
 123
 124    Don't stop other CPUs on reboot. This can make reboot more reliable
 125    in some cases.
 126
 127 Non Executable Mappings
 128
 129   noexec=on|off
 130
 131   on      Enable(default)
 132   off     Disable
 133
 134 SMP
 135
 136   additional_cpus=NUM Allow NUM more CPUs for hotplug
 137                  (defaults are specified by the BIOS, see Documentation/x86/x86_64/cpu-hotplug-spec)
 138
 139 NUMA
 140
 141   numa=off      Only set up a single NUMA node spanning all memory.
 142
 143   numa=noacpi   Don't parse the SRAT table for NUMA setup
 144
 145   numa=fake=CMDLINE
 146                 If a number, fakes CMDLINE nodes and ignores NUMA setup of the
 147                 actual machine.  Otherwise, system memory is configured
 148                 depending on the sizes and coefficients listed.  For example:
 149                         numa=fake=2*512,1024,4*256,*128
 150                 gives two 512M nodes, a 1024M node, four 256M nodes, and the
 151                 rest split into 128M chunks.  If the last character of CMDLINE
 152                 is a *, the remaining memory is divided up equally among its
 153                 coefficient:
 154                         numa=fake=2*512,2*
 155                 gives two 512M nodes and the rest split into two nodes.
 156                 Otherwise, the remaining system RAM is allocated to an
 157                 additional node.
 158
 159   numa=hotadd=percent
 160                 Only allow hotadd memory to preallocate page structures upto
 161                 percent of already available memory.
 162                 numa=hotadd=0 will disable hotadd memory.
 163
 164 ACPI
 165
 166   acpi=off      Don't enable ACPI
 167   acpi=ht       Use ACPI boot table parsing, but don't enable ACPI
 168                 interpreter
 169   acpi=force    Force ACPI on (currently not needed)
 170
 171   acpi=strict   Disable out of spec ACPI workarounds.
 172
 173   acpi_sci={edge,level,high,low}  Set up ACPI SCI interrupt.
 174
 175   acpi=noirq    Don't route interrupts
 176
 177 PCI
 178
 179   pci=off       Don't use PCI
 180   pci=conf1     Use conf1 access.
 181   pci=conf2     Use conf2 access.
 182   pci=rom       Assign ROMs.
 183   pci=assign-busses    Assign busses
 184   pci=irqmask=MASK             Set PCI interrupt mask to MASK
 185   pci=lastbus=NUMBER           Scan upto NUMBER busses, no matter what the mptable says.
 186   pci=noacpi            Don't use ACPI to set up PCI interrupt routing.
 187
 188 IOMMU (input/output memory management unit)
 189
 190  Currently four x86-64 PCI-DMA mapping implementations exist:
 191
 192    1. <arch/x86_64/kernel/pci-nommu.c>: use no hardware/software IOMMU at all
 193       (e.g. because you have < 3 GB memory).
 194       Kernel boot message: "PCI-DMA: Disabling IOMMU"
 195
 196    2. <arch/x86_64/kernel/pci-gart.c>: AMD GART based hardware IOMMU.
 197       Kernel boot message: "PCI-DMA: using GART IOMMU"
 198
 199    3. <arch/x86_64/kernel/pci-swiotlb.c> : Software IOMMU implementation. Used
 200       e.g. if there is no hardware IOMMU in the system and it is need because
 201       you have >3GB memory or told the kernel to us it (iommu=soft))
 202       Kernel boot message: "PCI-DMA: Using software bounce buffering
 203       for IO (SWIOTLB)"
 204
 205    4. <arch/x86_64/pci-calgary.c> : IBM Calgary hardware IOMMU. Used in IBM
 206       pSeries and xSeries servers. This hardware IOMMU supports DMA address
 207       mapping with memory protection, etc.
 208       Kernel boot message: "PCI-DMA: Using Calgary IOMMU"
 209
 210  iommu=[<size>][,noagp][,off][,force][,noforce][,leak[=<nr_of_leak_pages>]
 211         [,memaper[=<order>]][,merge][,forcesac][,fullflush][,nomerge]
 212         [,noaperture][,calgary]
 213
 214   General iommu options:
 215     off                Don't initialize and use any kind of IOMMU.
 216     noforce            Don't force hardware IOMMU usage when it is not needed.
 217                        (default).
 218     force              Force the use of the hardware IOMMU even when it is
 219                        not actually needed (e.g. because < 3 GB memory).
 220     soft               Use software bounce buffering (SWIOTLB) (default for
 221                        Intel machines). This can be used to prevent the usage
 222                        of an available hardware IOMMU.
 223
 224   iommu options only relevant to the AMD GART hardware IOMMU:
 225     <size>             Set the size of the remapping area in bytes.
 226     allowed            Overwrite iommu off workarounds for specific chipsets.
 227     fullflush          Flush IOMMU on each allocation (default).
 228     nofullflush        Don't use IOMMU fullflush.
 229     leak               Turn on simple iommu leak tracing (only when
 230                        CONFIG_IOMMU_LEAK is on). Default number of leak pages
 231                        is 20.
 232     memaper[=<order>]  Allocate an own aperture over RAM with size 32MB<<order.
 233                        (default: order=1, i.e. 64MB)
 234     merge              Do scatter-gather (SG) merging. Implies "force"
 235                        (experimental).
 236     nomerge            Don't do scatter-gather (SG) merging.
 237     noaperture         Ask the IOMMU not to touch the aperture for AGP.
 238     forcesac           Force single-address cycle (SAC) mode for masks <40bits
 239                        (experimental).
 240     noagp              Don't initialize the AGP driver and use full aperture.
 241     allowdac           Allow double-address cycle (DAC) mode, i.e. DMA >4GB.
 242                        DAC is used with 32-bit PCI to push a 64-bit address in
 243                        two cycles. When off all DMA over >4GB is forced through
 244                        an IOMMU or software bounce buffering.
 245     nodac              Forbid DAC mode, i.e. DMA >4GB.
 246     panic              Always panic when IOMMU overflows.
 247     calgary            Use the Calgary IOMMU if it is available
 248
 249   iommu options only relevant to the software bounce buffering (SWIOTLB) IOMMU
 250   implementation:
 251     swiotlb=<pages>[,force]
 252     <pages>            Prereserve that many 128K pages for the software IO
 253                        bounce buffering.
 254     force              Force all IO through the software TLB.
 255
 256   Settings for the IBM Calgary hardware IOMMU currently found in IBM
 257   pSeries and xSeries machines:
 258
 259     calgary=[64k,128k,256k,512k,1M,2M,4M,8M]
 260     calgary=[translate_empty_slots]
 261     calgary=[disable=<PCI bus number>]
 262     panic              Always panic when IOMMU overflows
 263
 264     64k,...,8M - Set the size of each PCI slot's translation table
 265     when using the Calgary IOMMU. This is the size of the translation
 266     table itself in main memory. The smallest table, 64k, covers an IO
 267     space of 32MB; the largest, 8MB table, can cover an IO space of
 268     4GB. Normally the kernel will make the right choice by itself.
 269
 270     translate_empty_slots - Enable translation even on slots that have
 271     no devices attached to them, in case a device will be hotplugged
 272     in the future.
 273
 274     disable=<PCI bus number> - Disable translation on a given PHB. For
 275     example, the built-in graphics adapter resides on the first bridge
 276     (PCI bus number 0); if translation (isolation) is enabled on this
 277     bridge, X servers that access the hardware directly from user
 278     space might stop working. Use this option if you have devices that
 279     are accessed from userspace directly on some PCI host bridge.
 280
 281 Debugging
 282
 283   oops=panic    Always panic on oopses. Default is to just kill the process,
 284                 but there is a small probability of deadlocking the machine.
 285                 This will also cause panics on machine check exceptions.
 286                 Useful together with panic=30 to trigger a reboot.
 287
 288   kstack=N      Print N words from the kernel stack in oops dumps.
 289
 290   pagefaulttrace  Dump all page faults. Only useful for extreme debugging
 291                 and will create a lot of output.
 292
 293   call_trace=[old|both|newfallback|new]
 294                 old: use old inexact backtracer
 295                 new: use new exact dwarf2 unwinder
 296                 both: print entries from both
 297                 newfallback: use new unwinder but fall back to old if it gets
 298                         stuck (default)
 299
 300 Miscellaneous
 301
 302         nogbpages
 303                 Do not use GB pages for kernel direct mappings.
 304         gbpages
 305                 Use GB pages for kernel direct mappings.