Re: [ECOS] Redboot and edb7211

[Gary Thomas <gthomas@cambridge.redhat.com>]

On 20-Apr-2001 Hugo Tyson wrote:
> 
> Jonathan Larmour <jlarmour@redhat.com> writes:
>> Gary Thomas wrote:
>> > On 19-Apr-2001 Jonathan Larmour wrote:
>> > > Perhaps the message is the initial checksum error reported when no Flash
>> > > Image System has been created yet. If you do a "fis init" it may just go
>> > > away.
>> > 
>> > These two items are not related in any way.
>> > 
>> > *  'fis init' rebuilds only the FIS directory.
>> > 
>> > *  The warning about checksum failures is about the 'fconfig' database.  One
>> >    needs to run 'fconfig' to fix that.
>> 
>> But presumably you can't do an fconfig until you've done an "fis init", so
>> both steps are required in order. 
> 
> Not sure.  I think the fconfig stuff is placed where it's placed and that's
> that.  "fis init" happens to make an entry that decribes where fconfig is
> placed, to help the user know what flash is used for what, and to keep the
> fis from using that flash itself.  But the fconfig does not refer to the
> fis to determine what flash to use; it's not that way round.
>  

Totally correct.

> If RedBoot startup reports a bad checksum, and the system appears to hang
> without a RedBoot> prompt, it's probably trying to use BOOTP to get an IP
> address.  Either wait a while, or build a RedBoot with no networking
> included and try that, use it to initialize the flash then upgrade to one
> with net - no net => no BOOTP => no delay at startup.
> 

Most likely this would make things seem to get "stuck".  Wait for a little
while (maybe 30 seconds) and the RedBoot prompt should appear.

> Warning, if you have a valid fconfig block in flash, but a new
> configuration of RedBoot adds some new keys, you might have to explicitly
> erase (using "fis erase -f 0x503e0000 -l 0x10000" or whatever) the old
> fconfig data and reset it all to get the new keys to "take".  Doesn't
> happen often that a new configuration changes things like this - except
> when debugging a whole new port and messing with extra fco settings - but
> mentioned just in case...

This should no longer be the case, with the newest 'fconfig' layout and
32 bit checksums.  [It did use to be the case however]