Re: [PATCH v2 3/4] iio: common: ssp_sensors: drop conditional optimization for simplicity

[Jonathan Cameron <jic23@kernel.org>]

On Wed, 18 Dec 2024 22:13:55 +0100
Vasileios Amoiridis <vassilisamir@gmail.com> wrote:

> On Wed, Dec 18, 2024 at 09:17:44AM -0600, David Lechner wrote:
> > On 12/17/24 5:41 PM, Vasileios Amoiridis wrote:  
> > > On Mon, Dec 16, 2024 at 03:57:44PM -0600, David Lechner wrote:  
> > >> On 12/14/24 1:14 PM, Vasileios Amoiridis wrote:  
> > >>> Drop conditional in favor of always calculating the timestamp value.
> > >>> This simplifies the code and allows to drop usage of internal private
> > >>> variable "scan_timestamp" of the struct iio_dev.
> > >>>
> > >>> Signed-off-by: Vasileios Amoiridis <vassilisamir@gmail.com>
> > >>> ---
> > >>>  drivers/iio/common/ssp_sensors/ssp_iio.c | 9 +++------
> > >>>  1 file changed, 3 insertions(+), 6 deletions(-)
> > >>>
> > >>> diff --git a/drivers/iio/common/ssp_sensors/ssp_iio.c b/drivers/iio/common/ssp_sensors/ssp_iio.c
> > >>> index caa404edd9d0..6b86b5315694 100644
> > >>> --- a/drivers/iio/common/ssp_sensors/ssp_iio.c
> > >>> +++ b/drivers/iio/common/ssp_sensors/ssp_iio.c
> > >>> @@ -8,6 +8,8 @@
> > >>>  #include <linux/iio/kfifo_buf.h>
> > >>>  #include <linux/module.h>
> > >>>  #include <linux/slab.h>
> > >>> +#include <linux/unaligned.h>
> > >>> +#include <linux/units.h>
> > >>>  #include "ssp_iio_sensor.h"
> > >>>  
> > >>>  /**
> > >>> @@ -70,7 +72,6 @@ EXPORT_SYMBOL_NS(ssp_common_buffer_postdisable, "IIO_SSP_SENSORS");
> > >>>  int ssp_common_process_data(struct iio_dev *indio_dev, void *buf,
> > >>>  			    unsigned int len, int64_t timestamp)
> > >>>  {
> > >>> -	__le32 time;
> > >>>  	int64_t calculated_time = 0;
> > >>>  	struct ssp_sensor_data *spd = iio_priv(indio_dev);
> > >>>  
> > >>> @@ -82,11 +83,7 @@ int ssp_common_process_data(struct iio_dev *indio_dev, void *buf,
> > >>>  	 */
> > >>>  	memcpy(spd->buffer, buf, len);
> > >>>  
> > >>> -	if (indio_dev->scan_timestamp) {
> > >>> -		memcpy(&time, &((char *)buf)[len], SSP_TIME_SIZE);
> > >>> -		calculated_time =
> > >>> -			timestamp + (int64_t)le32_to_cpu(time) * 1000000;
> > >>> -	}
> > >>> +	calculated_time = timestamp + get_unaligned_le32(buf + len) * MEGA;  
> > >>
> > >> Don't we still need to cast to 64 bit to avoid multiplication overflow?
> > >>  
> > > 
> > > Hi David,
> > > 
> > > Thanks for your message!
> > > 
> > > Aren't we already covered by the fact that MEGA is defined as an
> > > unsigned long?  
> > 
> > That is only 64-bits on 64-bit architectures, so could still overflow on
> > 32-bit architectures where long is 32-bit.
> >   
> 
> Hi David,
> 
> Hmmm, I think you are right. I can fix it in next, iteration. I will
> wait also for Jonathan's comments on the rest of the series.
> 
I tweaked it and applied.  Also dropped the initial assignment of calculated_time given it is now
always set before use.

Thanks,

Jonathan


> Cheers,
> Vasilis
> 
> > > 
> > > 	include/linux/units.h:12:#define MEGA 1000000UL
> > > 
> > > Cheers,
> > > Vasilis
> > >   
> > >>>  
> > >>>  	return iio_push_to_buffers_with_timestamp(indio_dev, spd->buffer,
> > >>>  						  calculated_time);  
> > >>  
> >