XO-1.75 no indication of EC reflash, risk of bricking

[Quozl]

After flashing q4b06jc on a unit with q4b06jb, there was a pause while the EC firmware was updated, which was visible on the serial port, but the display showed only an XO icon and single dot.

Perhaps some indication should be given on the display, so that the laptop is not interfered with by the user during the very brief interval.

I'm not sure what indication would be both easy enough and appropriate.

For XO-1.5, OpenFirmware and EC flash mediated by olpc.fth, we simply say it is happening.

[wmb@…]

Good point. Perhaps you could drive the design and implementation of this enhancement. For starters, just doing "text-on" before reflashing might be good.

[Quozl]

Neither visible nor text-on immediately before reflash-ec in do-ec-update seem to have any effect. Also means the failure cases aren't displayed, since they use visible.

[Quozl]

Fixed in svn 2488 which uses the storage LED to indicate write in progress, and adds screen icons on the bottom left; a laptop and an ellipsis ...

Test case:

• install q4b08jb from http://dev.laptop.org/~quozl/q408jb.rom, or later, which bundles EC firmware 0.2.03,
• download and flash-ec EC firmware 0.2.02 (cl2-4_0_2_02.img), or a prior general release version that is earlier than the version bundled with the version of OpenFirmware you are using.

Expected outcome: storage LED will glow during flash-ec, unit will reboot, then OpenFirmware will reflash to 0.2.03, during which the screen shall show icons on bottom left, and storage LED will glow, then unit will reboot again.

[Quozl]

( link fix, http://dev.laptop.org/~quozl/q4b08jb.rom )

[wad]

As we use other flashing LEDs to indicate error conditions, I'm uncomfortable with only having the storage LED flash as an indicator.

I suggest some simple icons indicating to wait for 30 sec, and that power shouldn't be removed.

[Quozl]

The LED does not flash, and it is not the only indicator. I've added simple icons indicating a wait, but not for 30 seconds.

The user would have to remove the power cord and battery to brick, as I understand the power button should be inhibited during EC SPI FLASH write. Paul, Richard, is that correct? With XO-1 and XO-1.5, we inhibited the power button.

(I don't know if power button is inhibited on OpenFirmware SPI FLASH write though, different issue.)

[rsmith]

Replying to Quozl:

The user would have to remove the power cord and battery to brick, as I understand the power button should be inhibited during EC SPI FLASH write. Paul, Richard, is that correct? With XO-1 and XO-1.5, we inhibited the power button.

Yes. The EC is in reset at that time and can't do any signal management.

(I don't know if power button is inhibited on OpenFirmware SPI FLASH write though, different issue.)

Here however, it can and will turn the power off if you hold down the power button. Perhaps we need some sort of inhibit signal to the EC.

[Quozl]

Fixed in q4b09. Opened #11215 for inhibit signal to EC for OpenFirmware flashing.

Please add q4b09 to build.

[Quozl]

In os42.

[Quozl]

Tested, works okay.