Respinning my test-devicetree branch

[Jeremy Kerr]

Hi Grant,

> I've respun my test-devicetree branch and dropped a bunch of patches
> in preference for ones in your arm-dt branch.  Here are my notes as I
> was working on it.

OK, cool. The patches aren't 100% confirmed yet - Nicolas and I are
still chatting about the specifics of machine-id allocation and probing
for device tree boot. I'm proposing a single machine ID for device tee
boot, and looking up a mdesc later (probing via the device tree). His
preference is a per-SOC machine IDs (eg, "omap4 with device tree
support", and no later DT probe.

However, it won't be difficult to cater for either as we go along.

> - skipped 94b9eca arm/dt: Allow CONFIG_OF on ARM
>    The version in my tree reflects changes to the drivers/of code

OK.

> - cherry picked 5769648 arm/dt: postpone mdesc lookup until setup_arch
> when booting with a DT
>   - qemu would no longer boot on normal versatilepb with this patch.
> Had to change compare to use "beq" to get it to boot.

OK, sounds reasonable - I hadn't tested that change on non-DT. I should
figure out why testing the carry bit doesn't work, as it made sense at
the time :)

> - cherry picked 4ac36ea arm/dt: allow bootmem reservation for device tree blob
>   - Please add my s-o-b line to this patch.  A chunk of that patch
> seems to be based on one of the patches I wrote ("arm/devicetree:
> Reserve memory used by dtb blob".  Correct me if I'm not remembering
> correctly).

Yes; it's a patch with multiple contributions from both of us; I must
have dropped your signoff at some point, sorry about that.

> - cherry picked fbdf9cf arm/dt: hook device tree setup into boot code
> - cherry picked 47ae655 arm/dt: probe for platforms via the device tree
> - cherry picked f528a85 drivers/amba: separate periphid probe logic
> from amba_device_register
> - cherry picked bf5ef5c drivers/amba: probe via device tree
> 

> I cherry picked the generic clock stuff, but I should probably merge
> in your clock tree instead.  None of this is stuff that I will be
> carrying in my tree over the long term:
> - cherry picked b156e35 Add a common struct clk
> - cherry picked 0a2e3d8 clk: Generic support for fixed-rate clocks
> - cherry picked 122a467 arm/clkdev: Allow common struct clk usage
> - cherry picked 8a9c892 arm/plat-versatile: convert to common struct clk

OK. These clock changed are merged in my DT branch to make the versatile
DT port easier; but it's possible to merge that separately. I've posted
the base clk infrastructure changes to rmk's tree, I think we're making
progress to have that accepted. Once that is done, the versatile port to
the new clk code should be fairly unintrusive.

> For the of-clocks I've cherry picked the following and then made some
> changes on top:
> - cherry picked 0177ef5 of: add clock providers
>   - Fixed up Kconfig conflicts
> - cherry picked 6bdbbd7 of/clk: parse fixed clocks

This one could do with some sanity checking; the DT parsing may be a
little naïve.

> - cherry picked c0e03fa arm/clkdev: lookup clocks from OF clock providers
> - Added patch "of/clock: rework to new binding"
>   - I'm still working on this.  I'll post to the ML it I've tested it.

OK, cool.

> Finally I've cherry picked the modified versatile platform, and made a
> few changes for my own purpose.
> - cherry picked 91eb8b8 arm/versatile: add a device tree versatile platform
> - squashed into 91eb8b8:  b1e67d2 arm/versatile: probe devices from device tree
> 
> Also have a couple of patches on top to fix a phandle endian issue and
> to fix booting of versatilepb when the dt patches are applied:

All looks good, thanks for that!

Cheers,


Jeremy