Re: [PATCH 3/5] arm64: dts: exynos: gs101: use correct clocks for usi8
From: Tudor Ambarus
Date: Mon Jan 29 2024 - 11:39:04 EST
On 1/27/24 04:00, Tudor Ambarus wrote:
>>> Wrong pclk clocks have been used in this usi8 instance here. For USI
>>> and I2C, we need the ipclk and pclk, where pclk is the bus clock.
>>> Without it, nothing can work.
>> Empty line is missing here?
>>
>>> It is unclear what exactly is using USI8_USI_CLK, but it is not
>>> required for the IP to be operational at this stage, while pclk is.
>> From [1] it looks like DIV_CLK_PERIC0_USI8_USI is a common parrent for
>> these two leaf gate clocks:
>> 1. GOUT_BLK_PERIC0_UID_RSTNSYNC_CLK_PERIC0_USI8_USI_IPCLKPORT_CLK
>> 2. GOUT_BLK_PERIC0_UID_PERIC0_TOP0_IPCLKPORT_IPCLK_7
>>
>> So IIUC, you replace clock #1 with clock #2 in this patch? If so, I
> No, GOUT_BLK_PERIC0_UID_PERIC0_TOP0_IPCLKPORT_IPCLK_7 is already used by
> IPCLK, the one that controls the clock frequency.
>
> So I understand Andre' replaces a child of the USI8 with something else.
>
> I don't think this works. We shall at least test it. I tested my usi8
> patches with the eeprom that's populated on the battery connector. I'll
> sync with Andre' offline and redo the tests on Monday.
Andre' is right, I messed up the bus clocks for USI. I tested the IPCLK,
the one that feeds USI clients, but I failed to correctly test the bus
clock. I retested by removing the clk_ignore_unused bootargs param and
verified that the patch is correct.
Reviewed-by: Tudor Ambarus <tudor.ambarus@xxxxxxxxxx>
Tested-by: Tudor Ambarus <tudor.ambarus@xxxxxxxxxx>