[PATCH v11]: clk: Add common clock support for Mediatek MT8135 and MT8173

[Matthias Brugger]

2015-04-15 17:11 GMT+02:00 Matthias Brugger <matthias.bgg at gmail.com>:
> 2015-04-14 13:57 GMT+02:00 Sascha Hauer <s.hauer at pengutronix.de>:
>> On Tue, Apr 14, 2015 at 01:01:30PM +0200, Matthias Brugger wrote:
>>> Hi Sascha,
>>>
>>> 2015-04-14 12:08 GMT+02:00 Sascha Hauer <s.hauer at pengutronix.de>:
>>> > Hi Matthias,
>>> >
>>> > On Tue, Apr 07, 2015 at 01:47:58PM +0200, Matthias Brugger wrote:
>>> >> Hi Sascha,
>>> >>
>>> >>
>>> >> 2015-03-31 20:16 GMT+02:00 Sascha Hauer <s.hauer at pengutronix.de>:
>>> >> >
>>> >> > The following changes since commit 9eccca0843205f87c00404b663188b88eb248051:
>>> >> >
>>> >> >   Linux 4.0-rc3 (2015-03-08 16:09:09 -0700)
>>> >> >
>>> >> > are available in the git repository at:
>>> >> >
>>> >> >   git://git.pengutronix.de/git/imx/linux-2.6.git tags/v4.0-clk-mediatek-v11
>>> >> >
>>> >> > for you to fetch changes up to ae9129219143cfdefe8b3a463deb8c5cb8955525:
>>> >> >
>>> >> >   dt-bindings: ARM: Mediatek: Document devicetree bindings for clock/reset controllers (2015-03-31 20:08:46 +0200)
>>> >> >
>>> >> > ----------------------------------------------------------------
>>> >> > This patchset contains the initial common clock support for Mediatek SoCs.
>>> >> > Mediatek SoC's clock architecture comprises of various PLLs, dividers, muxes
>>> >> > and clock gates.
>>> >>
>>> >> I tried the patch set on my mt8135 eval board. I used the dts bindings
>>> >> from a former version of this set [1], but it does not boot the board
>>> >> (based on v4.0-rc7).
>>> >> Do you have any hint, what is happening, or are the bindings wrong?
>>> >
>>> > I just tried on a v4.0 with
>>> > - this series applied
>>> > - the dts patch applied (which is still up-to-date)
>>> > - multi_v7_defconfig
>>> >
>>> > And it still works. What do you mean with "does not boot the board"? No
>>> > console output? Could you try with earlyprintk?
>>>
>>> The probelms I see is, that with the clock patches, I'm not able to
>>> boot into a initramfs [1].
>>> Whereas if I just comment topckgen and preicfg in the dts, I'm able to
>>> get the a serial console of my initramfs [2].
>>>
>>> I wonder if you are able to get serial console from the initramfs with
>>> the clock patches + dts patch?
>>
>> So you get kernel messages but no output from initramfs? In this case
>> the kernel disables the unused clocks in a late_initcall. The UART
>> driver still uses the dummy clock provided in the dtsi, so the real
>> UART clk gets disabled in the initcall. Try passing clk_ignore_unused to
>> the kernel
>
> OK, I can boot to the initramfs with this kernel parameter, but it
> fails to boot, when I change the uart clock. I suppose that PERI_UART3
> for debug port uart3 should be fine, but it looks like it doesn't
> work.
> I can see on the mt6589 datasheet that you have a register
> PERI_UART_CK_SOURCE_SEL where you can decide if you want a 26 MHz or a
> 52 MHz clock for the UART block.
> Does this register exists on mt8135? If so, why is it not implemented?

I found that PERI_UARTx_CK is connected to axi_sel but should be
connected to uart_sel. Like this I'm able to reach to initramfs
console, but the baud rate values are screwed up.
As uart_sel is at 52 MHz, I suppose we need to change the entries in
PERI_UART_CK_SOURCE_SEL.

Cheers,
Matthias

-- 
motzblog.wordpress.com