[PATCH 25/38] rtc: m48t35: support the SGI IP35 timekeeper's 1968 year base

[PATCH 25/38] rtc: m48t35: support the SGI IP35 timekeeper's 1968 year base

[Imre Kaloz]

The M48T35 of SGI IP35-family machines counts its year from 1968. Take
the epoch from a platform id table, "rtc-m48t35" for IP27 and
"rtc-m48t35-ip35" for IP35, and bound the year to the part's century: on
IP27 a year past 2069 now fails to set.

Signed-off-by: Imre Kaloz <kaloz@kernel.org>
---
 drivers/rtc/rtc-m48t35.c | 29 +++++++++++++++--------------
 1 file changed, 15 insertions(+), 14 deletions(-)

diff --git a/drivers/rtc/rtc-m48t35.c b/drivers/rtc/rtc-m48t35.c
index 92f19bf997b2..69c313b4b6e8 100644
--- a/drivers/rtc/rtc-m48t35.c
+++ b/drivers/rtc/rtc-m48t35.c
@@ -49,6 +49,7 @@ struct m48t35_priv {
 	struct m48t35_rtc __iomem *reg;
 	size_t size;
 	unsigned long baseaddr;
+	unsigned int epoch;
 	spinlock_t lock;
 };
 
@@ -86,10 +87,7 @@ static int m48t35_read_time(struct device *dev, struct rtc_time *tm)
 	 * Account for differences between how the RTC uses the values
 	 * and how they are defined in a struct rtc_time;
 	 */
-	tm->tm_year += 70;
-	if (tm->tm_year <= 69)
-		tm->tm_year += 100;
-
+	tm->tm_year += priv->epoch - 1900;
 	tm->tm_mon--;
 	return 0;
 }
@@ -108,18 +106,10 @@ static int m48t35_set_time(struct device *dev, struct rtc_time *tm)
 	min = tm->tm_min;
 	sec = tm->tm_sec;
 
-	if (yrs < 1970)
-		return -EINVAL;
-
-	yrs -= 1970;
-	if (yrs > 255)    /* They are unsigned */
+	if (yrs < priv->epoch || yrs >= priv->epoch + 100)
 		return -EINVAL;
 
-	if (yrs > 169)
-		return -EINVAL;
-
-	if (yrs >= 100)
-		yrs -= 100;
+	yrs -= priv->epoch;
 
 	sec = bin2bcd(sec);
 	min = bin2bcd(min);
@@ -164,6 +154,7 @@ static int m48t35_probe(struct platform_device *pdev)
 				     pdev->name))
 		return -EBUSY;
 
+	priv->epoch = platform_get_device_id(pdev)->driver_data;
 	priv->baseaddr = res->start;
 	priv->reg = devm_ioremap(&pdev->dev, priv->baseaddr, priv->size);
 	if (!priv->reg)
@@ -178,11 +169,20 @@ static int m48t35_probe(struct platform_device *pdev)
 	return PTR_ERR_OR_ZERO(priv->rtc);
 }
 
+/* The year register on SGI IP35-family machines counts from 1968. */
+static const struct platform_device_id m48t35_id_table[] = {
+	{ "rtc-m48t35", 1970 },
+	{ "rtc-m48t35-ip35", 1968 },
+	{ }
+};
+MODULE_DEVICE_TABLE(platform, m48t35_id_table);
+
 static struct platform_driver m48t35_platform_driver = {
 	.driver		= {
 		.name	= "rtc-m48t35",
 	},
 	.probe		= m48t35_probe,
+	.id_table	= m48t35_id_table,
 };
 
 module_platform_driver(m48t35_platform_driver);
@@ -191,3 +191,4 @@ MODULE_AUTHOR("Thomas Bogendoerfer <tsbogend@alpha.franken.de>");
 MODULE_DESCRIPTION("M48T35 RTC driver");
 MODULE_LICENSE("GPL");
 MODULE_ALIAS("platform:rtc-m48t35");
+MODULE_ALIAS("platform:rtc-m48t35-ip35");
-- 
2.47.3

Re: [PATCH 25/38] rtc: m48t35: support the SGI IP35 timekeeper's 1968 year base

[sashiko-bot]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [High] rtc: m48t35: null pointer dereference in m48t35_probe() via driver_override
--

commit 697b757d9c2a3da93baaf70e4fd5d745beae964e
Author: Imre Kaloz <kaloz@kernel.org>

rtc: m48t35: support the SGI IP35 timekeeper's 1968 year base

This commit updates the M48T35 RTC driver to support the 1968 year base for 
SGI IP35-family machines. It assigns a base epoch using the platform device ID
table and bounds the year to the appropriate century.

> diff --git a/drivers/rtc/rtc-m48t35.c b/drivers/rtc/rtc-m48t35.c
> index 92f19bf997b2f..69c313b4b6e84 100644
> --- a/drivers/rtc/rtc-m48t35.c
> +++ b/drivers/rtc/rtc-m48t35.c

[ ... ]

> @@ -164,6 +154,7 @@ static int m48t35_probe(struct platform_device *pdev)
>  				     pdev->name))
>  		return -EBUSY;
>  
> +	priv->epoch = platform_get_device_id(pdev)->driver_data;

[Severity: High]
Does this code handle devices bound via sysfs driver_override?

If the driver is manually bound to a device using the sysfs driver_override
interface, the platform bus platform_match() function returns success without
calling platform_match_id(). This leaves pdev->id_entry as null.

Since platform_get_device_id(pdev) returns pdev->id_entry, wouldn't
dereferencing its result directly cause a null pointer dereference regression?

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/cover.1790857966.git.kaloz@kernel.org?part=25