experimental fastboot images

Chia-I Wu olv at openmoko.com
Sat Nov 1 11:00:56 CET 2008


On Fri, Oct 31, 2008 at 07:08:47PM +0100, Adrian-Ken Rueegsegger wrote:
> I noticed patch 0004-fastboot-udev-static-only.patch. Is there a
> specific reason to used udev-118 instead of udev-124? I have used
> udev-124 and didn't notice any issues so far. 124 uses the same method
> (untar'ring /dev) that you implemented in your patch and thus saves
> quite some boottime compared to udev-118.
I haven't tested it, but it looks good.  The device nodes created on
first boot are remembered.  Later on, subsystems usually do not change
are skipped for coldplug.  And it nevers waits udevd to settle before
continuting.

There is one race on NEO though.  If SD card is not inserted in the
first boot, it is possible that mountall.sh runs before /dev/mmcblk0p1
is created in later boots.  The race asides, it looks good and the
fastboot image might switch to it in a latter version.

But, I have been thinking the usefulness of udev recently.  udev helps
us configure the devices dynamically.  Which means, if you have SD card
inserted, it creates the device node automatically and mounts it.  If
you have a bluetooth or usb device connected to NEO, it tries to help
you configure it.  It might work, might partially work, or might not
work at all.  God knows someone would use NEO to control robot legs,
right?  http://www.youtube.com/watch?v=SG3fAHFNs2c

Same to the other services being removed in current image, there is some
rooms for discussion about the necessity of udev.  We can mount SD card
anyway without udev saying yes.  Bluetooth keyboards should simply work
without udev.  As for robot legs, geeks knows how to solve it :)

-- 
Regards,
olv



More information about the devel mailing list