Re: [PATCH 05/10] hwmon: generic-pwm-tachometer: Add generic PWM based tachometer
On Thursday 08 March 2018 08:01 PM, Guenter Roeck wrote: On 03/07/2018 10:06 PM, Laxman Dewangan wrote: The RPM is measured speed via PWM signal capture which is output from fan. So should we have the fan[1..n]_output_rpm? No. I hear you clearly that you for some reason dislike fan[1..n]_input. While ABIs are not always to our liking, that doesn't mean that we get to change them at our whim. If that is not acceptable for you, I can't help you. And you can't change inX_input to inX_voltage either, sorry. My opinion is only to not use "input" as this is not really the input to fan. -- To unsubscribe from this list: send the line "unsubscribe linux-hwmon" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH 05/10] hwmon: generic-pwm-tachometer: Add generic PWM based tachometer
On Wednesday 07 March 2018 07:50 PM, Guenter Roeck wrote: On 03/07/2018 01:47 AM, Rajkumar Rampelli wrote: While I am not opposed to ABI changes, the merits of those would need to be discussed on the mailing list. But replacing "fan1_input" with "rpm" is not an acceptable ABI change, even if it may measure something that turns but isn't a fan. If this _is_ in fact supposed to be used for something else but fans, we would have to discuss what that might be, and if hwmon is the appropriate subsystem to measure and report it. This does to some degree lead back to my concern of having the "fan" part of this patch series in the pwm core. I am still not sure if that makes sense. Thanks, Guenter I am planning to add tachometer support in pwm-fan.c driver (drivers/hwmon/) instead of adding new generic-pwm-tachometer.c driver. Measuring RPM value will be done in pwm-fan driver itself using pwm capture feature and will add new sysfs attributes under this driver to report rpm value of fan. There is an existing attribute to report the RPM of fans. It is called fan[1..n]_input. "replacing "fan1_input" with "rpm" is not an acceptable ABI change" Preemptive NACK. The RPM is measured speed via PWM signal capture which is output from fan. So should we have the fan[1..n]_output_rpm? -- To unsubscribe from this list: send the line "unsubscribe linux-hwmon" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH 2/3] iio: adc: ina3221: Add support for IIO ADC driver for TI INA3221
On Wednesday 08 June 2016 08:34 PM, Andrew F. Davis wrote: On 06/07/2016 05:30 PM, Guenter Roeck wrote: On Fri, Jun 03, 2016 at 10:17:55AM -0500, Andrew F. Davis wrote: On 06/03/2016 09:14 AM, Laxman Dewangan wrote: On Friday 03 June 2016 06:59 PM, Guenter Roeck wrote: On 06/03/2016 03:06 AM, Jonathan Cameron wrote: On 01/06/16 13:34, Laxman Dewangan wrote: The INA3221 is a three-channel, high-side current and bus voltage monitor with an I2C interface from Texas Instruments. The INA3221 monitors both shunt voltage drops and bus supply voltages in addition to having programmable conversion times and averaging modes for these signals. The INA3221 offers both critical and warning alerts to detect multiple programmable out-of-range conditions for each channel. Add support for INA3221 SW driver via IIO ADC interface. The device is register as iio-device and provides interface for voltage/current and power monitor. Also provide interface for setting oneshot/continuous mode and critical/warning threshold for the shunt voltage drop. Signed-off-by: Laxman Dewangan <ldewan...@nvidia.com> Hi Laxman, As ever with any driver lying on the border of IIO and hwmon, please include a short justification of why you need an IIO driver and also cc the hwmon list + maintainers. (cc'd on this reply). I simply won't take a driver where the hwmon maintainers aren't happy. As it stands I'm not seeing obvious reasons in the code for why this should be an IIO device. Me not either. I have a hwmon driver for the same chip pending from Andrew Davis (TI) which I am just about to accept. We had directed Andrew back in April to write a hwmon driver for the chip, which he did. Thanks Guenter, I found the series [PATCH v2 0/2] Add support for INA3221 Triple Current/Voltage Monitors Looks fine to me. I can use the hwmon. If you search even further back you can see I originally went with an IIO driver as well, comparing the IIO and hwmon version for the simple use cases I needed the hwmod version turned out much simpler, so it was probably the right framework for now. However, some of the stuff from my patch are not there which I will add later once original patch applied: - Dynamic mode changes for continuous and one-shot from sysfs. - In one shot, when try to read voltage data, do conversion and then read. My attempt is rather minimal, to be honest I like your stab at this driver better in some ways, especially relating to DT putting each channel in its own node with labels, I think this is a bit more clean and will be more extendable if/when new multi-channel monitor chips are made. Not sure whether exporting the following will help or not. Can you please confirm? - Oversampling time i.e. average sample - conversion time for bus voltage and shunt voltage if default is not suited for system. These can always be added as needed, but the DT changes are not as easy. I would like it if this got in this cycle but if you think something will be needed to help your improvements, that can not be added incrementally, it would probably be best to get them into the initial DT binding doc now. Andrew, with this, the hwmon driver is pretty much on hold. What do you want me to do ? Should I wait for DT updates ? If Laxman would like to commit to making these changes I do not problem waiting a bit to get this right the first time. If we don't hear back soon then I'd just take it as is and anything needing to be fixed can be later. I will be able to post the patch tomorrow. I think this series can be applied and then on top of it, I will post the dt changes and other my changes. As this is new driver and no one using now, it will be fine to have dt changes if everything complete in one cycle. I will work from the tree on which this applied to post my patches. -- To unsubscribe from this list: send the line "unsubscribe linux-hwmon" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH 2/3] iio: adc: ina3221: Add support for IIO ADC driver for TI INA3221
On Friday 03 June 2016 05:39 PM, Jonathan Cameron wrote: On 03/06/16 12:26, Laxman Dewangan wrote: On Friday 03 June 2016 03:36 PM, Jonathan Cameron wrote: I thought that all ADC or monitors are going to be part of IIO device framework. I saw the ina2xx which is same (single channel) which was my reference point. That had a rather specific use case IIRC - they needed the buffered support to get the data fast enough. I think in our particular requirements, we dont need the buffering support but HW keep monitor and check with warning/critical threshold to generate HW signal. Funily enough I know this datasheet a little as was evaluating it for use on some boards at the day job a week or so ago. Various comments inline. Major points are: * Don't use 'fake' channels to control events. If the events infrastructure doesn't handle your events, then fix that rather than working around it. * There is a lot of ABI in here concerned with oneshot vs continuous. This seems to me to be more than it should be. We wouldn't expect to see stuff changing as a result of switching between these modes other than wrt to when the data shows up. So I'd expect to not see this directly exposed at all - but rather sit in oneshot unless either: 1) Buffered mode is running (not currently supported) 2) Alerts are on - which I think requires it to be in continuous mode. Other question to my mind is whether we should be reporting vshunt or (using device tree to pass resistance) current. This is bus and shunt voltage device for power monitoring. In our platforms, we use this device for bus current and so power monitor. We have two usecases, one is one shot, read when it needs it. And other continuous when we have multiple core running then continuous mode to get the power consumption by rail. That's fine, but continuous should be using the buffered interfaces really as that's there explicitly to support groups of channels captured using a sequencer. Then the abi ends up much more standard which is nice. Also allows for high speed ish continuous monitoring which is what the was I think the point of the single channel driver. The requirement for continuous monitoring is to ADC generate alert when the current on bus cross the threshold of warning/critical level so that alert signal can be used for throttling. So in my this particular usecase, we may not need buffered data. Yaah, alert is used only on continuous mode and mainly used for throttling when rail power goes beyond some limit. Of interesting in Linux, or routed directly to hardware? Yaah, In some platform this is routed to the hardware for throttling. -- To unsubscribe from this list: send the line "unsubscribe linux-hwmon" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH 2/3] iio: adc: ina3221: Add support for IIO ADC driver for TI INA3221
On Friday 03 June 2016 05:34 PM, Jonathan Cameron wrote: On 03/06/16 12:31, Laxman Dewangan wrote: On Friday 03 June 2016 03:46 PM, Jonathan Cameron wrote: On 03/06/16 11:06, Jonathan Cameron wrote: Code looks good, bu these more fundamental bits need sorting. Another minor point - why do the power calculations in driver? no hardware support for it, so why not just leave it to userspace? Device supports the bus and shunt voltage monitoring. So even no current. Also the warning/critical limit is for the voltage across shunt. So should we only expose the shunt/bus voltage, no power/current? I am thinking that user space should not know the platform and hence shunt resistance and so exposing the current and power on bus is better option. I'd go for current and voltage rather than current and power, but otherwise agree. OK, I will have the voltage (bus voltage) and current (on bus). User space can get power out of this. Thanks for suggestion. The critical limits will be still in current. And it should be customized sysfs instead of the iio_chan_spec i.e. iio_device attribute interface. -- To unsubscribe from this list: send the line "unsubscribe linux-hwmon" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH 2/3] iio: adc: ina3221: Add support for IIO ADC driver for TI INA3221
On Friday 03 June 2016 03:46 PM, Jonathan Cameron wrote: On 03/06/16 11:06, Jonathan Cameron wrote: Code looks good, bu these more fundamental bits need sorting. Another minor point - why do the power calculations in driver? no hardware support for it, so why not just leave it to userspace? Device supports the bus and shunt voltage monitoring. So even no current. Also the warning/critical limit is for the voltage across shunt. So should we only expose the shunt/bus voltage, no power/current? I am thinking that user space should not know the platform and hence shunt resistance and so exposing the current and power on bus is better option. -- To unsubscribe from this list: send the line "unsubscribe linux-hwmon" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH 2/3] iio: adc: ina3221: Add support for IIO ADC driver for TI INA3221
On Friday 03 June 2016 03:36 PM, Jonathan Cameron wrote: On 01/06/16 13:34, Laxman Dewangan wrote: Add support for INA3221 SW driver via IIO ADC interface. The device is register as iio-device and provides interface for voltage/current and power monitor. Also provide interface for setting oneshot/continuous mode and critical/warning threshold for the shunt voltage drop. Signed-off-by: Laxman Dewangan <ldewan...@nvidia.com> Hi Laxman, As ever with any driver lying on the border of IIO and hwmon, please include a short justification of why you need an IIO driver and also cc the hwmon list + maintainers. (cc'd on this reply). I simply won't take a driver where the hwmon maintainers aren't happy. As it stands I'm not seeing obvious reasons in the code for why this should be an IIO device. I thought that all ADC or monitors are going to be part of IIO device framework. I saw the ina2xx which is same (single channel) which was my reference point. Funily enough I know this datasheet a little as was evaluating it for use on some boards at the day job a week or so ago. Various comments inline. Major points are: * Don't use 'fake' channels to control events. If the events infrastructure doesn't handle your events, then fix that rather than working around it. * There is a lot of ABI in here concerned with oneshot vs continuous. This seems to me to be more than it should be. We wouldn't expect to see stuff changing as a result of switching between these modes other than wrt to when the data shows up. So I'd expect to not see this directly exposed at all - but rather sit in oneshot unless either: 1) Buffered mode is running (not currently supported) 2) Alerts are on - which I think requires it to be in continuous mode. Other question to my mind is whether we should be reporting vshunt or (using device tree to pass resistance) current. This is bus and shunt voltage device for power monitoring. In our platforms, we use this device for bus current and so power monitor. We have two usecases, one is one shot, read when it needs it. And other continuous when we have multiple core running then continuous mode to get the power consumption by rail. Yaah, alert is used only on continuous mode and mainly used for throttling when rail power goes beyond some limit. -- To unsubscribe from this list: send the line "unsubscribe linux-hwmon" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
