[support-gang] is rolling back firmware dangerous? (on XO-1s especially)

[James Cameron]

On Tue, Feb 18, 2014 at 09:04:52AM -0500, Kevin Gordon wrote:
> Thanks for this detailed update.  If i may summarize,  if one is
> using a 12 or 13 O/S on any architecture, I'm assuming the
> recommended f/w is the most current,  whether on XO-1, 1.5, 1.75 or
> 4.  If running 11, well, one is on their own :-)

Your summary disagrees with my view, which is more general: when you
have a choice of firmware, you should always use the latest.

The latest supports any operating system version, and contains bug
fixes that reduce your cost of ownership.  For volunteers, cost of
ownership is critical, as it increases the number of children you can
serve.

> Any concern I had stems from the cases where some non-XO
> machines/appliances/ gadgets I have advise to not roll back
> firmware, ever, as changes to the NVRAM configuration can introduce
> changes that cause the older firmware to be loaded, but then
> unusable, thereby turning the device into the proverbial brick.  My
> understanding, from the above discussion, is that this risk is not
> present with the XO.  I may be getting old, but I seem to remember
> being expressly told to *NOT* downgrade my XO-1's from C to B f/w,
> but that was maybe back in the days when my laptop was made out of
> wood.

I'm happy to get back into discussion after your summary.

In the general case, we also recommend that you do not downgrade,
firstly because downgrades to version N from version N+1 are not
tested before version N is released (doh), but also because we knew of
certain versions that would brick (in the Q2B and Q2C series).

However, since then we have tested more downgrades to the point we
know they are safe, and the only bricks due to downgrades happened
before mass production.

If you are using a prototype XO then you've decided to take the risk.
A prototype may brick on upgrade.  An upgrade may be forced by an
operating system install.

> Anecdotally, I have successfully upgraded *many* XO-1s to Q2F19, and
> they are happily running a custom 11.3 build, so there is certainly
> no need to 'roll-back' in that environment.  In fact, compared to
> the f/w "as-shipped", I get some nice USB added
> functionality/reliability with some weird-ass hubs/ sticks using the
> newest f/w.

Yes, those are examples of the bugs fixed.  Others include reset of
the clock to a valid date if the clock battery is lost.

You have all regularly pointed out how important the clock battery is.

> I have not yet upgraded any 11.3 XO 1.5 f/w to current yet.  But,
> our 11.3 XO 1.5 deployments are now on Q3C07, and that seems to be
> just peachy.  It's on the list to test Q3C16.

This will bring:

- antenna testing utility,
- booting from USB drives with embedded hubs, or external USB hubs,
- allow booting if the internal microSD card is missing,
- primitive laptop cloning capability,
- detection of incomplete download of .zd file,
- allow booting from ISO 9660 images, and hybrid images,
- reset of the clock to a valid date if the clock battery is lost,
- lid test,
- serial terminal and remote diagnosis tools,
- startup sound customisation,
- compatibility with USB drives with corrupt directories,
- compatibility with USB drives with empty extended partitions.

Are you hitting any of these?

> Why do we continue to use 11.3? [... lack of time ...]

No worries.  Hopefully I've given you some tools that will increase
the time you have.

(It is irritating that Adam, who started this thread, did not explain
_why_ he wanted to downgrade firmware, leaving us all to speculate.)

-- 
James Cameron
http://quozl.linux.org.au/