How to configure the Linux kernel/arch/um

Howto configure the Linux kernel / arch / um


 * UML uses the generic IRQ sugsystem


 * Option: GENERIC_HARDIRQS
 * Kernel Versions: 2.6.15.6 ...
 * (on/off)
 * default y


 * Option: UML
 * Kernel Versions: 2.6.15.6 ...
 * (on/off)
 * default y


 * Option: MMU
 * Kernel Versions: 2.6.15.6 ...
 * (on/off)
 * default y

"Linux/Usermode Kernel Configuration"


 * Option: ISA
 * Kernel Versions: 2.6.15.6 ...
 * (on/off)


 * Option: SBUS
 * Kernel Versions: 2.6.15.6 ...
 * (on/off)


 * Option: PCI
 * Kernel Versions: 2.6.15.6 ...
 * (on/off)


 * Option: UID16
 * Kernel Versions: 2.6.15.6 ...
 * (on/off)
 * default y


 * Option: GENERIC_CALIBRATE_DELAY
 * Kernel Versions: 2.6.15.6 ...
 * (on/off)
 * default y


 * Used in kernel/irq/manage.c and include/linux/irq.h


 * Option: IRQ_RELEASE_METHOD
 * Kernel Versions: 2.6.15.6 ...
 * (on/off)
 * default y

UML-specific options
option controls whether tracing thread support is compiled UML. Normally, this should be set to Y. If you intend to only skas mode (and the host has the skas patch applied to it), it is OK to say N here.
 * Option: MODE_TT
 * Kernel Versions: 2.6.15.6 ...
 * (on/off) Tracing thread support
 * default y

CONFIG_MODE_TT is disabled, then this option gives you the ability force a static link of UML. Normally, if only skas mode is built to UML, it will be linked as a shared binary. This is inconvenient use in a chroot jail. So, if you intend to run UML inside a, and you disable CONFIG_MODE_TT, you probably want to say Y.
 * Option: STATIC_LINK
 * Kernel Versions: 2.6.15.6 ...
 * (on/off) Force a static link
 * default n
 * depends on !MODE_TT

is needed when the host on which you run has a 2G/2G memory , instead of the customary 3G/1G.
 * Option: HOST_2G_2G
 * Kernel Versions: 2.6.15.6 ...
 * (on/off) 2G/2G host address space split
 * default n
 * depends on MODE_TT

that to enable such a host

patches.
 * Option: uration, which makes sense only in some cases, you need special
 * Kernel Versions: 2.6.15.6 ...

, if you do not know what to do here, say 'N'.

"Kernel address space size (in .5G units)"
 * Option: KERNEL_HALF_GIGS
 * Kernel Versions: 2.6.15.6 ...
 * default "1"
 * depends on MODE_TT
 * This determines the amount of address space that UML will allocate for      its own, measured in half Gigabyte units.  The default is 1.       Change this only if you need to boot UML with an unusually large amount       of physical memory.

option controls whether skas (separate kernel address space) is compiled in. If you have applied the skas patch to the , then you certainly want to say Y here (and consider saying N CONFIG_MODE_TT). Otherwise, it is safe to say Y. Disabling this will shrink the UML binary slightly.
 * Option: MODE_SKAS
 * Kernel Versions: 2.6.15.6 ...
 * (on/off) Separate Kernel Address Space support
 * default y


 * Option: Howto configure the Linux kernel/arch/um.arch"


 * Option: Howto configure the Linux kernel/mm


 * Option: LD_SCRIPT_STATIC
 * Kernel Versions: 2.6.15.6 ...
 * (on/off)
 * default y
 * depends on MODE_TT || STATIC_LINK


 * Option: LD_SCRIPT_DYN
 * Kernel Versions: 2.6.15.6 ...
 * (on/off)
 * default y
 * depends on !LD_SCRIPT_STATIC

you really know what you are doing, you should say Y here. reason is that some programs need kernel networking support even running on a stand-alone machine that isn't connected to any computer. If you are upgrading from an older kernel, you consider updating your networking tools too because changes the kernel and the tools often go hand in hand. The tools are in the package net-tools, the location and version number which are given in .
 * Option: NET
 * Kernel Versions: 2.6.15.6 ...
 * (on/off) Networking support

a general introduction to Linux networking, it is highly to read the NET-HOWTO, available from .


 * Option: Howto configure the Linux kernel/fs.binfmt"


 * Option: HOSTFS
 * Kernel Versions: 2.6.15.6 ...
 * (on/off/module) Host filesystem
 * While the User-Mode Linux port uses its own root file system for      booting and normal file access, this module lets the UML user       access files stored on the host.  It does not require any       network connection between the Host and UML.  An example use of       this might be:
 * mount none /tmp/fromhost -t hostfs -o /tmp/umlshare
 * where /tmp/fromhost is an empty directory inside UML and      /tmp/umlshare is a directory on the host with files the UML user       wishes to access.
 * For more information, see      .
 * If you'd like to be able to work with files stored on the host,      say Y or M here; otherwise say N.

(HoneyPot ProcFS) is a filesystem which allows UML /proc to be overridden, removed, or fabricated from the host. purpose is to allow a UML to appear to be a physical machine removing or changing anything in /proc which gives away the of a UML.
 * Option: HPPFS
 * Kernel Versions: 2.6.15.6 ...
 * (on/off/module) HoneyPot ProcFS (EXPERIMENTAL)

 for more information.

only need this if you are setting up a UML honeypot. Otherwise, is safe to say 'N' here.

you are actively using it, please report any problems, since it's fixed. In this moment, it is experimental on 2.6 (it works on 2.4).


 * Option: MCONSOLE
 * Kernel Versions: 2.6.15.6 ...
 * (on/off) Management console
 * default y
 * The user mode linux management console is a low-level interface to      the kernel, somewhat like the i386 SysRq interface.  Since there is       a full-blown operating system running under every user mode linux       instance, there is much greater flexibility possible than with the       SysRq mechanism.
 * If you answer 'Y' to this option, to use this feature, you need the      mconsole client (called uml_mconsole) which is present in CVS in       2.4.5-9um and later (path /tools/mconsole), and is also in the       distribution RPM package in 2.4.6 and later.
 * It is safe to say 'Y' here.

you say Y here, you will have some control over the system even the system crashes for example during kernel debugging (e.g., you be able to flush the buffer cache to disk, reboot the system or dump some status information). A key for each of the requests is provided.
 * Option: MAGIC_SYSRQ
 * Kernel Versions: 2.6.15.6 ...
 * (on/off) Magic SysRq key
 * depends on MCONSOLE

is the feature normally accomplished by pressing a key holding SysRq (Alt+PrintScreen).

UML, this is accomplished by sending a "sysrq" command with , followed by the letter for the requested command.

keys are documented in . Don't say Y you really know what this hack does.


 * Option: SMP
 * Kernel Versions: 2.6.15.6 ...
 * (on/off) Symmetric multi-processing support (EXPERIMENTAL)
 * default n
 * SMP_BROKEN is for x86_64.

option enables UML SMP support. is NOT related to having a real SMP box. Not directly, at least.
 * depends on MODE_TT && EXPERIMENTAL && (!SMP_BROKEN || (BROKEN && SMP_BROKEN))

implements virtual SMP by allowing as many processes to run on the host as there are virtual processors configured.

, if the host is a uniprocessor, those processes will , but, inside UML, will appear to be running simultaneously. the host is a multiprocessor, then UML processes may run , depending on the host scheduler.

, however, is supported only in TT mode. So, if you use the SKAS on your host, switching to TT mode and enabling SMP usually gives worse performances. , since the support for SMP has been under-developed, there could some bugs being exposed by enabling SMP.

you don't know what to do, say N.

"Maximum number of CPUs (2-32)" 2 32
 * Option: NR_CPUS
 * Kernel Versions: 2.6.15.6 ...
 * depends on SMP
 * default "32"

"Nesting level"
 * Option: NEST_LEVEL
 * Kernel Versions: 2.6.15.6 ...
 * default "0"
 * This is set to the number of layers of UMLs that this UML will be run      in.  Normally, this is zero, meaning that it will run directly on the       host.  Setting it to one will build a UML that can run inside a UML       that is running on the host.  Generally, if you intend this UML to run       inside another UML, set CONFIG_NEST_LEVEL to one more than the host       UML.
 * Note that if the hosting UML has its CONFIG_KERNEL_HALF_GIGS set to      greater than one, then the guest UML should have its CONFIG_NEST_LEVEL       set to the host's CONFIG_NEST_LEVEL + CONFIG_KERNEL_HALF_GIGS.       Only change this if you are running nested UMLs.


 * Option: HIGHMEM
 * Kernel Versions: 2.6.15.6 ...
 * (on/off) Highmem support
 * depends on !64BIT

"Kernel stack size order" option determines the size of UML kernel stacks. They will 1 << order pages. The default is OK unless you're running Valgrind UML, in which case, set this to 3.
 * Option: KERNEL_STACK_ORDER
 * Kernel Versions: 2.6.15.6 ...
 * default 2

option makes UML time deltas match wall clock deltas. This should be enabled. The exception would be if you are debugging with and spend long times with UML stopped at a breakpoint. In this , when UML is restarted, it will call the timer enough times to make for the time spent at the breakpoint. This could result in a lag. If this is a problem, then disable this option.
 * Option: UML_REAL_TIME_CLOCK
 * Kernel Versions: 2.6.15.6 ...
 * (on/off) Real-time Clock
 * default y


 * Option: Howto configure the Linux kernel/init


 * Option: Howto configure the Linux kernel/net


 * Option: Howto configure the Linux kernel/drivers/base


 * Option: Howto configure the Linux kernel/arch/um.char"


 * Option: Howto configure the Linux kernel/drivers/block


 * Option: NETDEVICES
 * Kernel Versions: 2.6.15.6 ...
 * (on/off)
 * default NET


 * Option: Howto configure the Linux kernel/arch/um.net"


 * Option: Howto configure the Linux kernel/drivers/net


 * Option: Howto configure the Linux kernel/drivers/connector


 * Option: Howto configure the Linux kernel/fs


 * Option: Howto configure the Linux kernel/security


 * Option: Howto configure the Linux kernel/crypto


 * Option: Howto configure the Linux kernel/lib

SCSI support

 * depends on BROKEN


 * Option: SCSI
 * Kernel Versions: 2.6.15.6 ...
 * (on/off/module) SCSI support


 * This gives us free_dma, which scsi.c wants.


 * Option: GENERIC_ISA_DMA
 * Kernel Versions: 2.6.15.6 ...
 * (on/off)
 * depends on SCSI
 * default y


 * Option: Howto configure the Linux kernel/arch/um.scsi"


 * Option: Howto configure the Linux kernel/drivers/md

BROKEN


 * Option: Howto configure the Linux kernel/drivers/mtd


 * This is just to shut up some Kconfig warnings, so no prompt.


 * Option: INPUT
 * Kernel Versions: 2.6.15.6 ...
 * (on/off)
 * default n


 * Option: Howto configure the Linux kernel/arch/um.debug"

Linux Kernel Configuration