linux-kernel - [PATCH] media: tc358746: improve calculation of the D-PHY timing registers

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: <20250107160702.641703-2-matthias.fend@emfend.at>
Date: Tue,  7 Jan 2025 17:07:01 +0100
From: Matthias Fend <matthias.fend@...end.at>
To: Mauro Carvalho Chehab <mchehab@...nel.org>
Cc: m.felsch@...gutronix.de,
	linux-media@...r.kernel.org,
	linux-kernel@...r.kernel.org
Subject: [PATCH] media: tc358746: improve calculation of the D-PHY timing registers

When calculating D-PHY registers, using data rates that are not multiples
of 16 can lead to precision loss in division operations. This can result in
register values that produce timing violations against the MIPI standard.

An example:
cfg->hs_clk_rate = 294MHz
hf_clk = 18

If the desired value in cfg->init is 100us, which is the minimum allowed
value, then the LINEINITCNT register is calculated as 1799. But since the
actual clock is 18.375MHz instead of 18MHz, this setting results in a time
that is shorter than 100us and thus violates the standard. The correct
value for LINEINITCNT would be 1837.

Improve the precision of calculations by using Hz instead of MHz as unit.

Signed-off-by: Matthias Fend <matthias.fend@...end.at>
---
 drivers/media/i2c/tc358746.c | 19 +++++++------------
 1 file changed, 7 insertions(+), 12 deletions(-)

diff --git a/drivers/media/i2c/tc358746.c b/drivers/media/i2c/tc358746.c
index 5eccd4ffd84d..883ca40709a9 100644
--- a/drivers/media/i2c/tc358746.c
+++ b/drivers/media/i2c/tc358746.c
@@ -485,24 +485,20 @@ static int tc358746_apply_misc_config(struct tc358746 *tc358746)
 	return err;
 }
 
-/* Use MHz as base so the div needs no u64 */
-static u32 tc358746_cfg_to_cnt(unsigned int cfg_val,
-			       unsigned int clk_mhz,
-			       unsigned int time_base)
+static u32 tc358746_cfg_to_cnt(unsigned long cfg_val, unsigned long clk_hz,
+			       unsigned long long time_base)
 {
-	return DIV_ROUND_UP(cfg_val * clk_mhz, time_base);
+	return div64_u64((u64)cfg_val * clk_hz + time_base - 1, time_base);
 }
 
-static u32 tc358746_ps_to_cnt(unsigned int cfg_val,
-			      unsigned int clk_mhz)
+static u32 tc358746_ps_to_cnt(unsigned long cfg_val, unsigned long clk_hz)
 {
-	return tc358746_cfg_to_cnt(cfg_val, clk_mhz, USEC_PER_SEC);
+	return tc358746_cfg_to_cnt(cfg_val, clk_hz, PSEC_PER_SEC);
 }
 
-static u32 tc358746_us_to_cnt(unsigned int cfg_val,
-			      unsigned int clk_mhz)
+static u32 tc358746_us_to_cnt(unsigned long cfg_val, unsigned long clk_hz)
 {
-	return tc358746_cfg_to_cnt(cfg_val, clk_mhz, 1);
+	return tc358746_cfg_to_cnt(cfg_val, clk_hz, USEC_PER_SEC);
 }
 
 static int tc358746_apply_dphy_config(struct tc358746 *tc358746)
@@ -517,7 +513,6 @@ static int tc358746_apply_dphy_config(struct tc358746 *tc358746)
 
 	/* The hs_byte_clk is also called SYSCLK in the excel sheet */
 	hs_byte_clk = cfg->hs_clk_rate / 8;
-	hs_byte_clk /= HZ_PER_MHZ;
 	hf_clk = hs_byte_clk / 2;
 
 	val = tc358746_us_to_cnt(cfg->init, hf_clk) - 1;
-- 
2.34.1