AUTOCONFIG(8) System Manager's Manual AUTOCONFIG(8)
autoconfig - configure the running system to the hardware
autoconfig [-i ifile] [-n nfile] [-k kfile] [-v] [-d] [-c]
Autoconfig is called by init(8) to configure the currently running system. Init checks the exit status of autoconfig to determine if the
configuration was successful. Autoconfig reads the device table /etc/dtab for a list of devices which may be on the system. It first ver-
ifies that the kernel has an attach routine for each device (and therefore has a device handler) and that the kernel has a probe routine.
It then checks each of these devices to see if it is present, and if it is, attempts to make it interrupt (if possible) to verify that the
interrupt vector is correct. The interrupt vector is checked to see that it has not previously been used. An interrupt through any of the
device's consecutive vectors is sufficient.
Devices which use programmable vectors (MSCP and TMSCP) are permitted to have a value of 0 in the dtab vector field. This special value
tells autoconfig to call the kernel's get next available vector routine and assign that to the device. For programmable vector devices if
the dtab vector field is non 0 then the value specified in the dtab file is used. In both cases the driver is called at its xxVec() rou-
tine with the vector being assigned to the device.
If the address and vector are correct, it then attaches the device by passing the address and unit number to the kernel's attach routine
and setting up the interrupt vector according to the interrupt handlers and priority listed in the device table. If the unit number is
given as a '?' in the device table, it will be assigned the next available unit number if the device exists. If the device is not present
or the vector is incorrect, and if the unit number was specified (not a '?'), then the kernel is notified that that unit is not present,
preventing accesses to a nonexistent device address.
There are only a few flags which are mostly useful for debugging but for completeness, here they are.
-i ifile Use ifile instead of /etc/dtab as the device table.
-n nfile Use nfile instead of /unix for finding the namelist of the currently running kernel.
-k kfile The file kfile should be used instead of /dev/kmem to alter and read kernel memory.
-v Verbose output, indicates reason for rejecting any device in the device table. Normally only attached devices are reported.
-c Report error messages for devices skipped because of problems with their interrupt vectors.
-d Turn on debugging mode. Shows many gory details of autoconfig's internal processing.
Devices of the same type must be listed with ascending unit numbers or with wildcards.
Disks that could be root devices must have their addresses and vectors initialized in the kernel; the kernel uses a root attach entry in
the block device switch to allow disk drivers to do any probes necessary before autoconfiguration.
Must be run only by init(8). There is a flag set in the kernel that autoconfig has already run, running autoconfig a second time results
in the error:
"namelist doesn't match running kernel."
Autoconfig attempts to open /dev/kmem for write. If the kernel is in securelevel 1 or higher the open of /dev/kmem will fail.
/etc/dtab device table
ucall(2), nlist(3), dtab(5)
3rd Berkeley Distribution December 30, 1992 AUTOCONFIG(8)