u-boot low-battery handling

[Werner Almesberger]

Andy Green wrote:
> What do you mean here as "over-optimized", the dynparts business?

Yup.

> Holger would like to see the NAND nuked so we can enable his hardware
> ECC patches by default, which is a nice win on speed to balance the
> hassle so people will be okay with it I think.

Good. One more :-)

> Qi can do GTA01 without too much hassle, but it's a really big step for
> that device since blowing the bootloader there is scary, I don't know it
> is a good idea to be proposing it.

Hmm, yes. Let's keep GTA01 for last. About 50% of the GTA01 users
should have a debug board, but that still leaves many who may not
have a convenient means of rescue.

> Reality is we have to engage with this at pcf50633 kernel driver for
> GTA03,

Or maybe even user space: bring up user space really quickly, fire
up the power policy daemon, then let it figure out what to do next.
Power-wise, it shouldn't really matter whether we're running kernel
code or user-space, and if there are no secret handshakes between
USB and the PMU, that's just a few race conditions or deadlocks who
will not come to haunt us later.

> For example we may need the spinning action in U-Boot in pcf50633 for
> GTA03 under 100mA / PC USB power circumstance, since at some point Linux
> would otherwise bring up GSM.

GSM, maybe some blinkenlights, ... sounds like a great task for user
space :-)

- Werner