man.bsd.lv manual page server

KERNEL ENVIRONMENT VARIABLES AND LOCAL VARIABLES

The loader actually has two different kinds of variables. These are kernel environment variables, which are visible to the kernel when it is started, and a separate space of local variables used by loader, which are not available to the kernel.

Both local variable and kernel environment variable of the same name are changed by assignment (variable=value).

Kernel environment variables can be set and unset through the set and unset builtins, and can have their values examined through the show and ifset builtins. Variables in command arguments or value in assignments (“$VAR” and “${VAR}”) refers to kernel environment variables.

Local variables can be unset with lunset and lunsetif builtin, and can have their values examined through the local builtin.

Notice that these environment variables are not inherited by any shell after the system has been booted, but can be examined by kenv(1).

Note that a variable can have two instances with differnet values: both a local variable instance and a kernel environment variable instance can exist for the same name and with different values. This can cause confusion and is seldom done on purpose.

A few variables are set automatically by loader. Others can affect the behavior of either loader or the kernel at boot. Some options may require a value, while others define behavior just by being set. Both types of variables are described below.

acpi_load

Used for handling automatic loading of the acpi(4) module. To disable automatic loading of the ACPI module use:

lunset acpi_load

set hint.acpi.0.disabled=1

autoboot_delay

Number of seconds autoboot and menu will wait before booting. Default value is 10 seconds.

If set to “NO”, no autoboot will be automatically attempted after processing dloader.rc, though explicit autoboot's will be processed normally, defaulting to 10 seconds delay.

boot_askname

Instructs the kernel to prompt the user for the name of the root device when the kernel is booted.

boot_ddb

Instructs the kernel to start in the DDB debugger, rather than proceeding to initialize when booted.

boot_gdb

Selects gdb-remote mode for the kernel debugger by default.

boot_single

Prevents the kernel from initiating a multi-user startup; instead single-user mode will be entered when the kernel has finished device probing.

boot_verbose

Setting this variable causes extra debugging information to be printed by the kernel during and after the boot phase.

bootfile

List of semicolon-separated search path for bootable kernels. The default is “kernel”.

console

Defines the current console.

currdev

Selects the default device. Syntax for devices is odd.

default_kernel

Selects default kernel loaded by menu command. Defaults to kernel.

dumpdev

The name of a device where the kernel can save a crash dump in case of a panic. This automatically sets the kern.dumpdev sysctl(3) MIB variable.

ehci_load

Used for handling automatic loading of the ehci(4) module. To disable automatic loading of the EHCI module use:

lunset ehci_load

set hint.ehci.0.disabled=1

xhci_load

Used for handling automatic loading of the xhci(4) module. To disable automatic loading of the XHCI module use:

lunset xhci_load

set hint.xhci.0.disabled=1

init_chroot

Directory init(8) will chroot() to on startup. By setting this variable DragonFly can be run from a subdirectory of the root file system.

init_path

Sets the list of binaries which the kernel will try to run as the initial process. The first matching binary is used. The default list is “/sbin/init:/sbin/oinit:/sbin/init.bak”.

kernel_options

Set kernel boot flags. See also boot command.

LINES

Define the number of lines on the screen, to be used by the pager.

local_modules

Setting this variable to “YES” causes /boot/modules.local to be included after modules_path directories list and passed to kernel for kldload(8).

module_path

Sets the list of directories which will be searched for modules named in a load command or implicitly required by a dependency. The default value for this variable is “;modules”, which first searches the current working directory and then modules.

num_ide_disks

Sets the number of IDE disks as a workaround for some problems in finding the root disk at boot. This has been deprecated in favor of root_disk_unit.

prompt

Value of loader's prompt. Defaults to “OK”. Kernel environment variables can be used in prompt by including “${VAR}”. E.g. (note that ‘$’ must be escaped with ‘\’ to be included in prompt):

set prompt="\${currdev} OK"

root_disk_unit

If the code which detects the disk unit number for the root disk is confused, e.g. by a mix of SCSI and IDE disks, or IDE disks with gaps in the sequence (e.g. no primary slave), the unit number can be forced by setting this variable.

See also vfs.root.mountfrom variable.

rootdev

By default the value of currdev is used to set the root file system when the kernel is booted. This can be overridden by setting rootdev explicitly.

See also vfs.root.mountfrom variable.

Other variables are used to override kernel tunable parameters. The following loader tunables are available:

hw.ioapic_enable

Control use of I/O APIC. Set to 1 to enable, 0 to disable. Default is 1.

hw.irq.X.dest

Set irqX's destination to the given CPUID, which starts from 0. If the specified value is larger than the last CPUID, then the first CPUID will be used. This variable should not be used if I/O APIC use is disabled.

hw.physmem

Limit the amount of physical memory the system will use. By default the size is in bytes, but the k, K, m, M, g and G suffixes are also accepted and indicate kilobytes, megabytes and gigabytes respectively. An invalid suffix will result in the variable being ignored by the kernel.

hw.usb.hack_defer_exploration

The USB keyboard will sometimes not attach properly unless you set this variable to 0.

kern.emergency_intr_enable

Setting this to 1 enables emergency interrupt polling. All interrupt handlers are executed periodically. This mode is very expensive and should only be used to get a system accessible when interrupt routing is otherwise broken. It is primarily used by kernel developers to debug new systems.

kern.emergency_intr_freq

Set the polling rate for the emergency interrupt polling code. The default is 10 (hz) to dissuade casual use. If you are doing real work with emergency interrupt polling mode enabled, it is recommended that you use a frequency between 100hz and 300hz.

kern.maxusers

Set the size of a number of statically allocated system tables; see tuning(7) for a description of how to select an appropriate value for this tunable. When set, this tunable replaces the value declared in the kernel compile-time configuration file.

kern.ipc.nmbclusters

Set the number of mbuf clusters to be allocated. The value cannot be set below the default determined when the kernel was compiled. Modifies NMBCLUSTERS.

kern.mmxopt

Toggles the mmx optimizations for the bcopy/copyin/copyout routines

kern.user_scheduler

Default userland scheduler (usched). If set, values can be “bsd4” or “dfly”. Default is “dfly”.

kern.vm.kmem.size

Sets the size of kernel memory (bytes). This overrides the value determined when the kernel was compiled.

kern.maxswzone

Limits the amount of KVM to be used to hold swap meta information, which directly governs the maximum amount of swap the system can support. This value is specified in bytes of KVA space and defaults to around 70MBytes. Care should be taken to not reduce this value such that the actual amount of configured swap exceeds ½ the kernel-supported swap. The default 70MB allows the kernel to support a maximum of (approximately) 14GB of configured swap. Only mess around with this parameter if you need to greatly extend the KVM reservation for other resources such as the buffer cache or NMBCLUSTERS. Modifies VM_SWZONE_SIZE_MAX.

kern.maxbcache

Limits the amount of KVM reserved for use by the buffer cache, specified in bytes. The default maximum is 200MB on 32-bit and unspecified on 64-bit. This parameter is used to prevent the buffer cache from eating too much KVM in large-memory machine configurations. Only mess around with this parameter if you need to greatly extend the KVM reservation for other resources such as the swap zone or NMBCLUSTERS. Note that the NBUF parameter will override this limit. Modifies VM_BCACHE_SIZE_MAX.

machdep.disable_mtrrs

Disable the use of i686 MTRRs (x86 only).

net.inet.tcp.tcbhashsize

Overrides the compile-time set value of TCBHASHSIZE or the preset default of 512. Must be a power of 2.

vfs.root.wakedelay

Specify an additional delay (default is 2 seconds if unspecified) before trying to mount root.

vfs.root.mountfrom

Specify root file system. A semicolon separated list of file systems to try as the kernel root file system. File system format is file system type and disk store, separated by colon. This variable needs to be set when using a boot-only partition, which is typically mounted on root file system as /boot.

One file system example:

hammer:da8s1a

One file system HAMMER2 multi volume example:

hammer2:da8s1a:da9s1a

Several file systems, boot list, example:

ufs:da0s1a;hammer2:ad1s1d

Each file system in the list will be tried in the order specified until the mount succeeds. If all fail, the ‘mountroot>’ prompt is displayed for manual entry.

You may not specify devtab labels here but you can specify paths available to devfs(5) such as:

hammer:serno/L41JYE0G.s1d

vfs.root.realroot

Root file system and extra options for initrd. See initrd(7).

nfsroot.iosize

Override the default NFS netbooted root mount io block size. The default is 8192 which works with most servers. suggested values are 8192, 16384, or 32768.

nfsroot.rahead

Override the default read-ahead used by netbooted root mounts. The default is 4 which is reasonable. Suggested values are 1 through 8.

machdep.hack_efifb_probe_early

Setting this to 1 works around an issue that occurs on some recent systems where there is no system console when booting via UEFI. See bug report #3167.