linux-kernel - [PATCH v8 09/14] iio: afe: rescale: fix accuracy for small fractional scales

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-Id: <20210820191714.69898-10-liambeguin@gmail.com>
Date:   Fri, 20 Aug 2021 15:17:09 -0400
From:   Liam Beguin <liambeguin@...il.com>
To:     liambeguin@...il.com, peda@...ntia.se, jic23@...nel.org,
        lars@...afoo.de
Cc:     linux-kernel@...r.kernel.org, linux-iio@...r.kernel.org,
        devicetree@...r.kernel.org, robh+dt@...nel.org
Subject: [PATCH v8 09/14] iio: afe: rescale: fix accuracy for small fractional scales

From: Liam Beguin <lvb@...hos.com>

The approximation caused by integer divisions can be costly on smaller
scale values since the decimal part is significant compared to the
integer part. Switch to an IIO_VAL_INT_PLUS_NANO scale type in such
cases to maintain accuracy.

Signed-off-by: Liam Beguin <lvb@...hos.com>
---
 drivers/iio/afe/iio-rescale.c | 27 +++++++++++++++++++++++++--
 1 file changed, 25 insertions(+), 2 deletions(-)

diff --git a/drivers/iio/afe/iio-rescale.c b/drivers/iio/afe/iio-rescale.c
index c408c4057c08..7304306c9806 100644
--- a/drivers/iio/afe/iio-rescale.c
+++ b/drivers/iio/afe/iio-rescale.c
@@ -22,7 +22,7 @@ int rescale_process_scale(struct rescale *rescale, int scale_type,
 			  int *val, int *val2)
 {
 	s64 tmp;
-	s32 rem;
+	s32 rem, rem2;
 	u32 mult;
 	u32 neg;
 
@@ -38,8 +38,31 @@ int rescale_process_scale(struct rescale *rescale, int scale_type,
 		tmp = (s64)*val * 1000000000LL;
 		tmp = div_s64(tmp, rescale->denominator);
 		tmp *= rescale->numerator;
-		tmp = div_s64(tmp, 1000000000LL);
+
+		tmp = div_s64_rem(tmp, 1000000000LL, &rem);
 		*val = tmp;
+
+		/*
+		 * For small values, the approximation can be costly,
+		 * change scale type to maintain accuracy.
+		 *
+		 * 100 vs. 10000000 NANO caps the error to about 100 ppm.
+		 */
+		if (scale_type == IIO_VAL_FRACTIONAL)
+			tmp = *val2;
+		else
+			tmp = 1 << *val2;
+
+		 if (abs(rem) > 10000000 && abs(*val / tmp) < 100) {
+			 *val = div_s64_rem(*val, tmp, &rem2);
+
+			 *val2 = div_s64(rem, tmp);
+			 if (rem2)
+				 *val2 += div_s64(rem2 * 1000000000LL, tmp);
+
+			 return IIO_VAL_INT_PLUS_NANO;
+		 }
+
 		return scale_type;
 	case IIO_VAL_INT_PLUS_NANO:
 	case IIO_VAL_INT_PLUS_MICRO:
-- 
2.32.0.452.g940fe202adcb