[PATCH v1] meson saradc: fix clock divider mask length

George Stark gnstark at sberdevices.ru
Tue Jun 6 05:53:51 PDT 2023


On 6/5/23 23:18, Martin Blumenstingl wrote:
> Hi George,
>
> On Thu, Jun 1, 2023 at 10:56 PM George Stark <gnstark at sberdevices.ru> wrote:
> [...]
>> Here the test I promised:
>> Question: what's the real size of clock divder field in SAR_ADC_REG3 register in saradc in meson8 socs?
>> The current kernel code says 5 bits
>> The datasheet says 6 bit
>>
>> The parent clock of adc clock is 24Mhz
>> I can check it here by:
>>
>> # cat /sys/kernel/debug/clk/clk_summary
>>    xtal                                 4        4        1    24000000          0     0  50000         Y
>>       c1108680.adc#adc_div              1        1        0      1142858         0     0  50000         Y
>>          c1108680.adc#adc_en            1        1        0      1142858         0     0  50000         Y
>>
>> for divider width 5bit min adc clock is 24Mhz / 32 =  750KHZ
>> for divider width 6bit min adc clock is 24Mhz / 64 =  375KHz
>>
>> I suppose that the lower adc clock rate the higher measurement time
>> so I need to get measurement time at both clk freqs and the times differ so
>> 6bit divider is really applied
>>
>> I performed test at Odroid-C1, kernel 6.2-rc8
>> Two kernel patches must be applied:
>>
>> the topic starter patch and the helper patch at the end of the letter
>> In the helper patch I turn on CLOCK_ALLOW_WRITE_DEBUGFS to change clock rate from she shell
>> and use ktime_get_raw_ts64 to measure measurement time
>>
>> So the the test itself:
>> cat /sys/devices/platform/soc/c1100000.cbus/c1108680.adc/iio:device0/in_voltage3_raw
>> [ 1781.226309] ==== freq: 1142858 time 42408000
>>
>> # echo 750000 > /sys/kernel/debug/clk/c1108680.adc#adc_en/clk_rate
>> # cat /sys/devices/platform/soc/c1100000.cbus/c1108680.adc/iio:device0/in_voltage3_raw
>> [ 1790.728656] ==== freq: 750000 time 49173000
>>
>> # echo 375000 > /sys/kernel/debug/clk/c1108680.adc#adc_en/clk_rate
>> # cat /sys/devices/platform/soc/c1100000.cbus/c1108680.adc/iio:device0/in_voltage3_raw
>> [ 1816.955477] ==== freq: 375000 time 68245000
>>
>> # cat /sys/kernel/debug/clk/clk_summary
>>    xtal                                 4        4        1    24000000          0     0  50000         Y
>>       c1108680.adc#adc_div              1        1        0      375000          0     0  50000         Y
>>          c1108680.adc#adc_en            1        1        0      375000          0     0  50000         Y
> These results looks excellent - thanks for sharing the test results!
> Could you please check one last thing:
> $ grep -i adc /sys/kernel/debug/meson-clk-msr/measure_summary
> It should confirm that the clock rate is 375kHz (or close to it, SoC
> internal clock measurement is not 100% precise)
Hello Martin

Looks like it works as expected:
# grep -i adc /sys/kernel/debug/meson-clk-msr/measure_summary
   sar_adc                 1140625    +/-3125Hz
# echo 375000 > /sys/kernel/debug/clk/c1108680.adc#adc_en/clk_rate
# grep -i adc /sys/kernel/debug/meson-clk-msr/measure_summary
  sar_adc                  371875    +/-3125Hz

So I'm re-sending the patch with fixed commit message

-- 
Best regards
George

>
> Once we have that confirmation: can you please re-send the patch with
> the description updated so it's clear which SoC generations are
> affected and by stating that the fix was tested on a Meson8b Odroid-C1
> board.
>
>
> Thank you and best regards,
> Martin




More information about the linux-amlogic mailing list