Subject: Re: Indigo kernel downloads, boots w/ FreeBSD, not NetBSD
To: Gerald Heinig <Gerald.Heinig@post.rwth-aachen.de>
From: Rafal Boni <rafal.boni@eDial.com>
Date: 02/02/2001 22:26:41
In message <200102021749.MAA21421@ninigret.metatel.office>, I wrote:
-> Ahh!! I figured it must be on the client side, rather than server, as some
-> of the I2 code (the diag. environment, for one) managed to TFTP just fine.
-> Since we have the source for the FreeBSD & NetBSD TFTPd's, I think this is
-> fixable!! 8-)
Well, the answer is that it's a boot PROM bug, but tickled by how the server
sends back replies. See the following URL and search for `tftp':
-> That's really good to know, since if I can tfp kernels to the box, it makes
-> my life much easier (no more booting IRIX to write new kernel into the vol.
-> header, pulling SASH & IDE out of volume header to make room for the NetBSD
-> kernel, ...).
I think I found the answer, but as my wife is getting annoyed with my
hacking binge, didn't get a change to verify. SGI notes in a 5.X kernel
rollup patch that the boot PROM does not grok ports with the high bit set,
hence their tftpd was fixed to not send packets from ports > 32K.
Unsurprisingly, the NetBSD tftpd sends replies back from ports ~ 64K, and
the PROM spits out a 'tftprecv: pkt from server port 65528, not -8' message
(thanks to Jeff Smith for the 'setenv DEBUG 1' trick... I didn't know you
could make the boot PROM spit out more info that way!).
I imagine once this is fixed, I'll be able to tftp my kernels straight from
the boot PROM which will make life much happier!
BTW, I'm considering starting a FAQ/HOWTO for getting the sgimips port
booting for people interested in playing (with things like the SASH bug
Soren forwarded me, this TFTP bug, ...) with the sgimips port. Please
send me any info you think might be useful and I'll try and compile it
and post it to the list (and maybe get one of the web mavens to HTML'ize
it and post in on www.netbsd.org). Any other comments on the topic also
appreciated (even if you think it's a worthless exerise).
Thanks for all the hint, folks!
Rafal Boni rafal.boni@eDial.com
PGP key C7D3024C, print EA49 160D F5E4 C46A 9E91 524E 11E0 7133 C7D3 024C
Need to get a hold of me? http://800.edial.com/rafal.boni@eDial.com