andy at openmoko.com
Thu Oct 16 17:52:10 CEST 2008
-----BEGIN PGP SIGNED MESSAGE-----
Somebody in the thread at some point said:
| (The first message I sent does not seem to have arrived)
| On 12/10/08 18:00:55, Andy Green wrote:
|> Yes always-on MPU can deliver consistent power behaviours we can't do
|> our current way of relying on PMU. You would basically make the PMU
|> slave of the MPU. Stuff like debricking scheme for a programmable
|> so brickable MPU that controls the PMU... needs careful thought.
| That shouldn't be a problem, because microcontrollers support in
| circuit serial programming, so just make sure we can get to those pins
| and have a doc that specifies the programming protocol for the brave.
The issue is that if we allow user-updateable MPU, it can always be
bricked. So for example we put out a new package with some MPU update
that is broken, suddenly many devices could be bricked before we pull
it. We definitely need some credible sequence of actions for the
end-user that can unbrick the devices. Just telling him where some pins
are doesn't really cut it.
If the MPU is master of the CPU, then when it is bricked a lot of assets
we might otherwise call on are unavailable. So it needs thinking
through being aware of specific capabilities of the MPU.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (GNU/Linux)
Comment: Using GnuPG with Fedora - http://enigmail.mozdev.org
-----END PGP SIGNATURE-----
More information about the community