Re: [patch] set prefix on PPC - briQ results

[Hollis Blanchard]

On Apr 19, 2005, at 12:36 PM, Marco Gerards wrote:

> Hollis Blanchard <address@hidden> writes:
>
> > I've tested the patch on briQ. As far as I can tell, CodeGen's
> > "SmartFirmware" is crap, frequently suffering mysterious failures that
> > require unplugging and waiting (unplugging and replugging rapidly
> > isn't enough). It also suffers frequent ATA failures, where the disk
> > or controller simply stop responding.
> >
> > 1. (major) Firmware does not properly set /chosen/bootpath to include
> > the partition number.
>
> What happens?  Is it something I could try to reproduce on the 
> PegasosII?

Boot GRUB, run suspend, "dev /chosen .properties". You'll notice the 
device arguments are missing from bootpath.

> > Problem 1 could be worked around by taking the yaboot approach:
> > iterate over all disk partitions looking for a file named
> > "/grub/grub.cfg" (i.e. /boot/grub/grub.cfg). This fallback can be used
> > if we do not find a grub.cfg in prefix.
>
> Right...
>
> > Problem 2 could be worked around by blacklisting firmwares based on
> > the /openprom properties. Here is the briQ's output:
>
> With blacklist you mean setting some additional flags that describe
> which bugs the firmware has?

Yes.

> >         ok dev /openprom
> >         ok .properties
> >         relative-addressing
> >         model                 "BRIQ,1.0.2.60"
> >         SmartFirmware-version "1.1"
> >         CodeGen-copyright     "SmartFirmware(tm) Copyright 1996-2000
> > by CodeGen, Inc.  All Rights Reserved."
> >         name                  "openprom"
> >
> > Does the Pegasos have a SmartFirmware-version property?
>
> It does.  It has both a CodeGen and bplan copyright property.  The
> SmartFirmware-version is 1.2 here.  The model is "Pegasos2,1.2".  It
> seems to me that this SmartFirmware-version is not interesting because
> it looks like this version is not changed for every build.
>
> The most interesting property is "built-on", which we can use to
> detect if a specific version has the bug or not.

Right now I'm inclined to set a "0_BASED_PARTITION" flag if the 
SmartFirmware-version property exists. Later on if bplan fixes it then 
they will need to provide some other property we can examine.

After I check in my patch I will work on these workarounds.

-Hollis