Hello folks,

If I read you correctly, at a cold boot, the HD does not have time to
initialize before Linux boots from the Bios.  On reset, the HD remains hot,
while the board resets and the boot up process works as advertised.

If this is correct, you need to change the values on the RC network that you
will find on the CPU's reset pin to one with a longer time constant.  The
two ways are to increase the resistance or increase the capacitance.  In
order to change the resistance to a higher value, you would need to replace
the resistor or put another one in series, not easy, especially with SMD's.
You can do so much easier by "piggy-backing" a larger capacitor in parallel
with the one in the network to lengthen the time after a cold boot that the
reset pin is held low.  This shouldn't significantly affect the reset timing
for a warm boot, and it beats the hell out of a two step cold boot or a
bunch or nop's in the code to slow the penguin down on any boot.

Regarding the adaptation to smaller machines, how much work would be
involved in adapting a slightly "non-standard" kernel to run from a 486's
BIOS at boot up?  Has anyone done any work on trying Linuxbios on older 486
machines like the Compaq Deskpro series?

The reason this may be a worthy goal is these machines were nearly the
industry standard for many years and are available in large numbers, free
for the taking, usually with 16MB of RAM, a 3com NIC, and a small (80-240MB)
HD.  The Linuxbios project could breathe new life in these machines since
they make ideal routers and other embedded projects that do not need
bleeding edge speed, and they are built to last.

My interest in this is that during downtime, my company builds routers and
other internet  appliances out of old 486's for non-profit businesses to get
them online in cooperation with Telocity, a DSL ISP.

You'd be amazed to find out how many non-profits have no web presence and
use several PII's they were given with one on a dialup account (usually
someone's personal account at Earthlink or AOL) and sneakernet for file
transfer.  This severely limits their effectiveness and thier ability to use
contributions to achieve the best results.  By using my staff to build and
install these at little to no cost to them (depending on their legal
status), I aim to change that.  We are currently in negotiations with
Telocity to provide non-profits with free or low cost broadband access and
the interface, and we would provide the router, web server, and network for
them to connect all their PC's.

The router / webserver / firewall / cacheing DNS server package we usually
install fits on a single floppy, but we generally install it on a bootable
partition of the Hard Disk to increase the speed of boot up and simplify
administration.  It requires 16MB of RAM, and a 386 or higher CPU (Mathco is
not needed) and the default scripts support up to 2 protected nets, dail up
access, dial backup, and Intranet administation.  The our current
configuration uses DR-DOS and Loadlin to boot, and then moves everything to
RAM.  I'd like to move the kernel to BIOS for simplicity and security, and
reserve the HD for FTP and HTTP use.

Regards,

Dan Harper

Clark's Law: Any sufficiently advanced technology is indistinguishable from
magic
Corollary to Clarke's Law: Any technology distinguishable from magic is
insufficiently advanced

Reply via email to