[ECOS] Re: Fwd: RE: [ECOS] Still having problems getting networking up

[ECOS] Re: Fwd: RE: [ECOS] Still having problems getting networking up

[Lewin A.R.W. Edwards]

> > > Just FYI, it seems that template is currently broken, as eCos won't build
> > > successfully with "ecosconfig new edb7xxx net". I'll redo it and send a
> > > make.out if you want, but it should be easy to reproduce.
> > >
> >
> >I tried this "six ways from Sunday" against the current [master] sources and
> >I could not reproduce your error.

I successfully reproduced it; I've attached the ecos.ecc and make.out from the
session. The only manual change I made to ecos.ecc was to set the processor type to
EP7212. I am building under Red Hat 6.1.

> > > Anyway, I got everything built OK, but it seems that more configuration is
> > > required. When my app starts, it hangs repeating "CS8900: Tx interrupt
> > lost".
> > >
> >
> >If you have an EDB7212 board then you'll need to remove a resistor to get
> >interrupts to work.  Remove R168 and all jumpers from JP45.

Aha! R168 is not fitted on my board, but there was a jumper on one of the JP45
pairs.

Removing the jumper has not made any difference to RedBoot's behavior (still can't
find the bootp server). However, my application now works for DHCP, which is great
and I can get my main functionality up & running.

In my app I will eventually need to implement the "AutoIP" scheme (I'm implementing
UPnP) if I don't detect a DHCP/BOOTP server. Is there a recognized process for
using an internally-generated (at runtime) IP address if DHCP fails? I guess I just
poke stuff into the bootp structure for eth0?

> > > support for setting the MAC address in this release." Does the networking
> > > code in eCos [not redboot] read the MAC address set by fconfig and
> >
> >Currently, if you run RedBoot and enable the network address (turn on
> >networking support in RedBoot) then eCos will just use that.  Note: eCos does
> not get the
> >ESA from the FLASH, it assumes that RedBoot has set it up.

ESA? Is that a networking term I don't know or an eCos term? I'm not a TCP/IP
expert, having only ever programmed it at a high level in the past. Very much of a
learning curve.

Anyway, I understand by your reply that since the EEPROM doesn't work on the
EDB7212 and eCos doesn't support it anyway, since there is no documented API for me
to tell the networking layer the MAC address, I will need to modify the network
code myself? (I have to find out from Cirrus how they expect us to handle this. I
originally assumed that the CS8900 had the MAC factory-burned, I didn't think about
the EEPROM until I looked at the board).


-- Lewin A.R.W. Edwards (Embedded Engineer)              * Linux system
Work:     http://www.digi-frame.com/
Personal: http://www.larwe.com/ and http://www.zws.com/

Re: [ECOS] Re: Fwd: RE: [ECOS] Still having problems getting networking up

[Bart Veer]

>>>>> "Lewin" == Lewin A R W Edwards <larwe@larwe.com> writes:

    >> > > Just FYI, it seems that template is currently broken, as
    >> > > eCos won't build successfully with "ecosconfig new edb7xxx
    >> > > net". I'll redo it and send a make.out if you want, but it
    >> > > should be easy to reproduce.
    >> > >
    >> >
    >> >I tried this "six ways from Sunday" against the current
    >> >[master] sources and I could not reproduce your error.

    Lewin> I successfully reproduced it; I've attached the ecos.ecc
    Lewin> and make.out from the session. The only manual change I
    Lewin> made to ecos.ecc was to set the processor type to EP7212. I
    Lewin> am building under Red Hat 6.1.

Please do not send 16000 line attachments to public mailing lists like
ecos-discuss unless absolutely necessary! Some people on the mailing
list may have only limited network connectivity, including dialup
access.

Instead you can use tools like "ecosconfig export" to generate a
minimal ecos.ecc savefile. Similarly when sending make output most of
the information is irrelevant: usually it is just the last couple of
commands and the error messages that are of interest, not the ~3500
lines that work without problems.

If somebody is willing to help resolve the problem and needs more
information, private email should be used. Obviously 16000 line
messages should also not be sent unsolicited to individuals.

Bart

Re: [ECOS] Re: Fwd: RE: [ECOS] Still having problems getting networking up

[Lewin A.R.W. Edwards]

Hello Bart,

>    Lewin> I successfully reproduced it; I've attached the ecos.ecc
>     Lewin> and make.out from the session. The only manual change I
>
>Please do not send 16000 line attachments to public mailing lists like

Sorry about that. I've become accustomed to hitting "reply all" and editing 
the reply list, and I was either in a hurry or not thinking (or both).

On a related note, would it be possible for the list to be reconfigured so 
that the Reply-to: is the list address, rather than the poster's address?

=== Lewin A.R.W. Edwards (Embedded Engineer)
Work: http://www.digi-frame.com/
Personal: http://www.zws.com/ and http://www.larwe.com/

"Und setzet ihr nicht das Leben ein,
Nie wird euch das Leben gewonnen sein."