Re: [PATCH v1] iio: hid-sensor-attributes: Fix divisions for 32-bit platforms

["Pandruvada, Srinivas" <srinivas.pandruvada@intel.com>]

[-- Attachment #1: Type: text/plain, Size: 4918 bytes --]

On Thu, 2019-09-05 at 14:27 +0300, Andy Shevchenko wrote:
> The commit 473d12f7638c
> 
>   ("iio: hid-sensor-attributes: Convert to use int_pow()")
> 
> converted to use generic int_pow() helper. Though, the generic one
> returns
> 64-bit value and, in cases when it is used as divisor, it compels 64-
> bit
> division from compiler.
> 
> In order to fix this, introduce a temporary 32-bit variable to hold
> the result
> of int_pow() and use it as divisor afterwards.
> 
> In couple of cases, replace int_pow() with a predefined unit factors
> for time
> and frequency.
> 
> Fixes: 473d12f7638c ("iio: hid-sensor-attributes: Convert to use
> int_pow()")
> Reported-by: kbuild test robot <lkp@intel.com>
> Reported-by: Nathan Chancellor <natechancellor@gmail.com>
> Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Acked-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

> ---
>  .../hid-sensors/hid-sensor-attributes.c       | 42 ++++++++++++-----
> --
>  1 file changed, 28 insertions(+), 14 deletions(-)
> 
> diff --git a/drivers/iio/common/hid-sensors/hid-sensor-attributes.c
> b/drivers/iio/common/hid-sensors/hid-sensor-attributes.c
> index b9dd19b34267..442ff787f7af 100644
> --- a/drivers/iio/common/hid-sensors/hid-sensor-attributes.c
> +++ b/drivers/iio/common/hid-sensors/hid-sensor-attributes.c
> @@ -10,10 +10,14 @@
>  #include <linux/irq.h>
>  #include <linux/kernel.h>
>  #include <linux/slab.h>
> +#include <linux/time.h>
> +
>  #include <linux/hid-sensor-hub.h>
>  #include <linux/iio/iio.h>
>  #include <linux/iio/sysfs.h>
>  
> +#define HZ_PER_MHZ	1000000L
> +
>  static struct {
>  	u32 usage_id;
>  	int unit; /* 0 for default others from HID sensor spec */
> @@ -93,8 +97,10 @@ static void simple_div(int dividend, int divisor,
> int *whole,
>  
>  static void split_micro_fraction(unsigned int no, int exp, int
> *val1, int *val2)
>  {
> -	*val1 = no / int_pow(10, exp);
> -	*val2 = no % int_pow(10, exp) * int_pow(10, 6 - exp);
> +	int divisor = int_pow(10, exp);
> +
> +	*val1 = no / divisor;
> +	*val2 = no % divisor * int_pow(10, 6 - exp);
>  }
>  
>  /*
> @@ -129,6 +135,7 @@ static void convert_from_vtf_format(u32 value,
> int size, int exp,
>  
>  static u32 convert_to_vtf_format(int size, int exp, int val1, int
> val2)
>  {
> +	int divisor;
>  	u32 value;
>  	int sign = 1;
>  
> @@ -136,10 +143,13 @@ static u32 convert_to_vtf_format(int size, int
> exp, int val1, int val2)
>  		sign = -1;
>  	exp = hid_sensor_convert_exponent(exp);
>  	if (exp < 0) {
> +		divisor = int_pow(10, 6 + exp);
>  		value = abs(val1) * int_pow(10, -exp);
> -		value += abs(val2) / int_pow(10, 6 + exp);
> -	} else
> -		value = abs(val1) / int_pow(10, exp);
> +		value += abs(val2) / divisor;
> +	} else {
> +		divisor = int_pow(10, exp);
> +		value = abs(val1) / divisor;
> +	}
>  	if (sign < 0)
>  		value =  ((1LL << (size * 8)) - value);
>  
> @@ -202,12 +212,12 @@ int hid_sensor_write_samp_freq_value(struct
> hid_sensor_common *st,
>  	if (val1 < 0 || val2 < 0)
>  		return -EINVAL;
>  
> -	value = val1 * int_pow(10, 6) + val2;
> +	value = val1 * HZ_PER_MHZ + val2;
>  	if (value) {
>  		if (st->poll.units ==
> HID_USAGE_SENSOR_UNITS_MILLISECOND)
> -			value = int_pow(10, 9) / value;
> +			value = NSEC_PER_SEC / value;
>  		else if (st->poll.units ==
> HID_USAGE_SENSOR_UNITS_SECOND)
> -			value = int_pow(10, 6) / value;
> +			value = USEC_PER_SEC / value;
>  		else
>  			value = 0;
>  	}
> @@ -296,6 +306,7 @@ EXPORT_SYMBOL(hid_sensor_write_raw_hyst_value);
>  static void adjust_exponent_nano(int *val0, int *val1, int scale0,
>  				  int scale1, int exp)
>  {
> +	int divisor;
>  	int i;
>  	int x;
>  	int res;
> @@ -309,9 +320,10 @@ static void adjust_exponent_nano(int *val0, int
> *val1, int scale0,
>  			return;
>  		}
>  		for (i = 0; i < exp; ++i) {
> -			x = scale1 / int_pow(10, 8 - i);
> +			divisor = int_pow(10, 8 - i);
> +			x = scale1 / divisor;
>  			res += int_pow(10, exp - 1 - i) * x;
> -			scale1 = scale1 % int_pow(10, 8 - i);
> +			scale1 = scale1 % divisor;
>  		}
>  		*val0 += res;
>  		*val1 = scale1 * int_pow(10, exp);
> @@ -321,13 +333,15 @@ static void adjust_exponent_nano(int *val0, int
> *val1, int scale0,
>  			*val0 = *val1 = 0;
>  			return;
>  		}
> -		*val0 = scale0 / int_pow(10, exp);
> -		rem = scale0 % int_pow(10, exp);
> +		divisor = int_pow(10, exp);
> +		*val0 = scale0 / divisor;
> +		rem = scale0 % divisor;
>  		res = 0;
>  		for (i = 0; i < (9 - exp); ++i) {
> -			x = scale1 / int_pow(10, 8 - i);
> +			divisor = int_pow(10, 8 - i);
> +			x = scale1 / divisor;
>  			res += int_pow(10, 8 - exp - i) * x;
> -			scale1 = scale1 % int_pow(10, 8 - i);
> +			scale1 = scale1 % divisor;
>  		}
>  		*val1 = rem * int_pow(10, 9 - exp) + res;
>  	} else {

[-- Attachment #2: smime.p7s --]
[-- Type: application/x-pkcs7-signature, Size: 3290 bytes --]