[Previous] [Contents] [Index] [Next]


Network I/O support

Note: You must be root to start this manager.


io-net [-avms] [-i instance] [-c 0] [-d driver [driver_options]] ...
       -p protocol [protocol_options] 
       [-t threads] ... &


Disable the automatic loading of converter modules.
-c 0
Disable caching of packet buffers. This should only be needed as a debugging facility and should be used in conjunction with the "cache=0" option to the tcpip (big) stack.
-d driver [driver_options]
Load the specified driver and pass it the given driver_options. For information on the drivers and their syntax and options, see the devn-* entries:
Driver Shared object
artesyn devn-artesyn.so
cpci-mcp750 devn-cpci-mcp750.so
crys8900 devn-crys8900.so
dm9102 devn-dm9102.so
eepro devn-eepro.so
el509 devn-el509.so
el589 devn-el589.so
el900 devn-el900.so
epic devn-epic.so
fd devn-fd.so
marvell devn-gt6260.so
intel devn-i82544.so
klsi devn-klsi.so
lance devn-lance.so
ne2000 devn-ne2000.so
ne2000-403 devn-ne2000-403.so
ns83815 devn-ns83815.so
orinoco devn-orinoco.so
pcnet devn-pcnet.so
pegasus devn-pegasus.so
ppc800-ads devn-ppc800-ads.so
ppc800-cllf devn-ppc800-cllf.so
ppc800-mbx devn-ppc800-mbx.so
ppc800-rpxlite devn-ppc800-rpxlite.so
ppc860t_mii devn-ppc860_mii.so
ppc8260t devn-ppc8260.so
prism devn-prism.so
rtl devn-rtl.so
sis9 devn-sis9.so
smc9000 devn-smc9000.so
speedo devn-speedo.so
tulip devn-tulip.so
tulip-p5064 devn-tulip-p5064.so
rhine devn-via-rhine.so
wd devn-wd.so

The -d option isn't necessary for driverless interfaces such as pppd.

-i instance
Instance number: useful when running multiple io-nets.
Allow multicast devctl() functions on /dev/io-net/ entries.
-p protocol [protocol_options]
Load the specified protocol and pass it the given protocol_options. The protocols include:
Protocol Shared object
pppmgr npm-pppmgr.so
pppoe npm-pppoe.so
qnet npm-qnet.so
tcpip npm-tcpip.so
ttcpip npm-ttcpip.so

"Static" configuration: no mounting or unmounting of modules is performed after the command line is processed.

-t threads
Maximum number of threads available in the shared pool exported by io-net to the modules that registered with it. Default is 100.
Verbose output.


The io-net manager provides support for dynamically loaded networking modules. You can load drivers and protocols when you start io-net, by using the -d and -p command-line options.

Note: You can start more than one driver or protocol by using multiple -d and -p command-line options.

Once io-net has started, you can dynamically load and unload modules by using the mount and umount commands. For example, this command:

io-net -dne2000 -pttcpip if=en0:10.163 &

is the same as this sequence:

io-net &
mount -T io-net devn-ne2000.so
mount -T io-net -o if=en0:10.163 npm-ttcpip.so

The io-net manager uses the LD_LIBRARY_PATH environment variable when searching for the shared objects. If it's not set, or the shared object in question isn't in one of its directories, or you want to override the default, specify the full path in the mount command.

To unload a module, use a command like this:

umount /dev/io-net/en0

By default, io-net populates the /dev/io-net portion of the namespace and handles mount requests of type io-net. This may cause confusion when multiple io-nets are running, as their portion of the namespace overlap. The -i option can be used to append an instance number to the above strings. See the examples below:

mount -T io-net /lib/dll/devn-speedo.so
ls /dev/io-net
io-net -i1
mount -T io-net1 -o prefix=/alt /lib/dll/npm-tcpip.so
ls /dev/io-net1


Start io-net, load the NE-2000 Ethernet driver and tiny TCP/IP protocol module, and autoload the IP-to-Ethernet-converter module (ncm-ip_en.so).

io-net -dne2000 -pttcpip

This sample mkifs boot script starts io-net with an NE 2000 Ethernet adapter, the tiny TCP/IP stack, and a PPP manager:

[uid=* gid=*]

[virtual=x86/bios +compress] boot = {
    startup-bios -v 
    PATH=/proc/boot LD_LIBRARY_PATH=/proc/boot:/dev/shmem procnto -vvv

[+script] startup.script = {
    [+session] devc-con -e -n8 &
    waitfor /dev/con1
    reopen /dev/con1

    [+session] devc-ser8250 &
    waitfor /dev/ser2 &

    io-net -dne2000 -pttcpip if=en0:10.163: -p pppmgr
    waitfor /dev/socket

    fs-nfs2 -u 10.25:/usr/nto/x86 / -u 10.25:/ /q &
    waitfor /bin

    reopen /dev/con1
    [+session] TZ=est05edt04,M4.1.0/2,M10.5.0/2 PATH=/bin:/usr/local/bin \
        sh &

    /bin/pipe &
    /bin/devc-pty &

[type=link] /usr/lib/ldqnx.so.2 = /proc/boot/libc.so
[type=link] /tmp = /dev/shmem
[type=link] /bin/login = /bin/esh
[type=link] /dev/console = /dev/con1
[type=link] /dev/log = /dev/null
[type=link] /proc/dumper = /dev/null




Mount two Tulip-based Netgear FA310TX cards, one with PCI index 1 that's connected to the Internet, and one with PCI index 0 that's connected to a local network, setting the default route:

io-net -dtulip chipset=168,media=3,pci=0x1 \
       -dtulip chipset=168,media=5,pci=0x0 \
       -pttcpip if=en0:,\

See also:

mount, pppd, umount

"Network drivers (devn-*)" and "Network protocol modules (npm-*)" in the Utilities Summary

Using Qnet for Transparent Distributed Processing and TCP/IP Networking in the Neutrino User's Guide

[Previous] [Contents] [Index] [Next]