Netboot Mailing List (by thread)

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: First experience with Netboot 0.6.0 remote booting



Hello,

George Shin <gshin@ix.netcom.com> wrote:
> I'm currently trying out the Netboot 0.6.0 by first building
> the bootprom utility, "mknbi-dos" using precompiled kernel
> binaries (i'm using standard kernel).  I was able to successfully
> do the "./configure" but when doing "make dep" (from ~/netboot),
> i ran into some errors (i haven't had the chance to analyze this).

You can safely skip the 'make dep'.

> I then tried to create the boot image file from the /dev/fd0
> but i kept getting the error message, "/dev/fd0: not a file or
> directory". I don't know what's going on here.

Thanks, this is a bug in mknbi-dos. I will fix it as soon as possible.
Simply use

	dd if=/dev/fd0 of=fdImage

before calling mknbi-dos with the file fdImage as the ramdisk image.

> Currently we are getting
> the error, "PKTDRV ERROR: Invalid PKTDRVR".

This is the result of a bug in your packet driver. The bootrom asks the
packet driver about it's MTU size using function 10h, which is what the
specs call a high performance function. If the packet driver is too old,
it doesn't have this call, and should therefore return with the carry
flag set. This condition is checked in the bootrom. However, since that
error message came out, your packet driver returned successfully to that
function call, but since it can't support the function it returned illegal
values. You are the first one who reported such a problem to me.

> I've contacted the vendor and they state that their packet driver
> is based on Packet Driver Spec. v1.06 and this seems to be the
> valid specification to date.

No, the current Packet Driver Spec. version is 1.09! Actually, I never
used an older packet driver, as I don't have one. Would it be (legally)
possible for you to send me your packet driver so that I can take a look
at it? (Please uuencode it before sending). Thanks for your help.

> I've looked at the module, "bootp.c" and it seems
> the timeout is done via count to 32.

No, the timeout is handled the way prescribed by the relevant RFC. A bit
mask is shifted left for every retry, and a random value is masked with this
bitmask and then used as the timeout in seconds (in the call to udp_read the
timeout value is multiplied by 18 to convert it into timer ticks). The number
of retries is in fact 32.
Ok, I just removed the network cable from one of my clients and in fact it
seems that the timeout is not increasing as it should. I will investigate
this further, and let you know.

> Is there a way to adjust
> the timeout in BOOTP and TFTP without having to rebuild the
> bootprom kernel (how about infinite wait)?

No, it isn't adjustable. What do the others think about an infinite number
of retries? There is still the option of pressing ESC to get out of the BOOTP
loop and let the client boot from the floppy, so this might not be a bad idea.

gero.

--
Pessimists have already begun to worry about what is going to replace
automation.		- John Tudor
--
Gero Kuhlmann, Hannover     0511/6497525 (Voice)        gero@gkminix.han.de



For requests or suggestions regarding this mailing list archive please write to netboot@gkminix.han.de.