Re: [PATCH] firmware: ti_sci: Convert CPU latency constraint from us to ms

[Kendall Willis <k-willis@ti.com>]

On 4/25/25 14:08, Nishanth Menon wrote:
> On 10:37-20250425, Kendall Willis wrote:
>> Fix CPU resume latency constraint units sent to TI SCI firmware.
>> CPU latency constraints are set using the PM QoS framework. The PM QoS
>> framework uses usecs as the units for latency whereas the device manager
>> uses msecs, so a conversion is needed before passing to device manager.
>>
> 
> If this is a bug fix (sounds like it), follow the stable kernel rules.
> 
> Also please do not expect reviewers in the community know what this
> means, I think you intent to point folks to the url
> https://software-dl.ti.com/tisci/esd/latest/2_tisci_msgs/pm/lpm.html#tisci-msg-lpm-set-latency-constraint
> 
> If so, add the reference to your commit message.
>  >
>> Signed-off-by: Kendall Willis <k-willis@ti.com>
>> Reviewed-by: Dhruva Gole <d-gole@ti.com>
>> ---
>> Test log [1] shows entry to MCU Only low power mode by sending a CPU
>> resume latency constraint of 100000 us using PM QoS. MCU Only is shown
>> to be entered by 0x1 as the printed mode.
>>
>> [1] https://gist.github.com/kwillis01/059a2ca38232387b414bc6f4b87c7475
>> ---
>>   drivers/firmware/ti_sci.c | 9 +++++++--
>>   1 file changed, 7 insertions(+), 2 deletions(-)
>>
>> diff --git a/drivers/firmware/ti_sci.c b/drivers/firmware/ti_sci.c
>> index 806a975fff22..bc138a837430 100644
>> --- a/drivers/firmware/ti_sci.c
>> +++ b/drivers/firmware/ti_sci.c
>> @@ -3670,6 +3670,7 @@ static int __maybe_unused ti_sci_suspend(struct device *dev)
>>   	struct ti_sci_info *info = dev_get_drvdata(dev);
>>   	struct device *cpu_dev, *cpu_dev_max = NULL;
>>   	s32 val, cpu_lat = 0;
>> +	u16 cpu_lat_ms;
>>   	int i, ret;
>>   
>>   	if (info->fw_caps & MSG_FLAG_CAPS_LPM_DM_MANAGED) {
>> @@ -3682,9 +3683,13 @@ static int __maybe_unused ti_sci_suspend(struct device *dev)
>>   			}
>>   		}
>>   		if (cpu_dev_max) {
>> -			dev_dbg(cpu_dev_max, "%s: sending max CPU latency=%u\n", __func__, cpu_lat);
>> +			/* PM QoS latency unit is usecs, TI SCI uses msecs */
>> +			cpu_lat_ms = cpu_lat / USEC_PER_MSEC;
> 
> round_down or a round_up? I assume you intent round_down, please
> document that in the comments.
> 
>> +			dev_dbg(cpu_dev_max, "%s: sending max CPU latency=%u ms\n", __func__,
>> +				cpu_lat_ms);
>>   			ret = ti_sci_cmd_set_latency_constraint(&info->handle,
>> -								cpu_lat, TISCI_MSG_CONSTRAINT_SET);
>> +								cpu_lat_ms,
>> +								TISCI_MSG_CONSTRAINT_SET);
>>   			if (ret)
>>   				return ret;
>>   		}
>>
>> base-commit: 393d0c54cae31317deaa9043320c5fd9454deabc
>> -- 
>> 2.34.1
>>
> 

Thanks for looking this over, Nishanth. I will implement your comments 
in V2.

Best,
Kendall Willis