From: Arnd Bergmann <arnd@arndb.de>

gcc runs into undefined behavior at the end of the three led_polarity_set()
callback functions if it were called with a zero 'modes' argument and it
just ends the function there without returning from it.

This gets flagged by 'objtool' as a function that continues on
to the next one:

drivers/net/phy/aquantia/aquantia_leds.o: warning: objtool: aqr_phy_led_polarity_set+0xf: can't find jump dest instruction at .text+0x5d9
drivers/net/phy/intel-xway.o: warning: objtool: xway_gphy_led_polarity_set() falls through to next function xway_gphy_config_init()
drivers/net/phy/mxl-gpy.o: warning: objtool: gpy_led_polarity_set() falls through to next function gpy_led_hw_control_get()

There is no point to micro-optimize the behavior here to save a single-digit
number of bytes in the kernel, so just change this to a "return -EINVAL"
as we do when any unexpected bits are set.

Fixes: 1758af47b98c ("net: phy: intel-xway: add support for PHY LEDs")
Fixes: 9d55e68b19f2 ("net: phy: aquantia: correctly describe LED polarity override")
Fixes: eb89c79c1b8f ("net: phy: mxl-gpy: correctly describe LED polarity")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
---
 drivers/net/phy/aquantia/aquantia_leds.c | 2 +-
 drivers/net/phy/intel-xway.c             | 2 +-
 drivers/net/phy/mxl-gpy.c                | 2 +-
 3 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/drivers/net/phy/aquantia/aquantia_leds.c b/drivers/net/phy/aquantia/aquantia_leds.c
index 00ad2313fed3..951f46104eff 100644
--- a/drivers/net/phy/aquantia/aquantia_leds.c
+++ b/drivers/net/phy/aquantia/aquantia_leds.c
@@ -156,5 +156,5 @@ int aqr_phy_led_polarity_set(struct phy_device *phydev, int index, unsigned long
 	if (force_active_high || force_active_low)
 		return aqr_phy_led_active_low_set(phydev, index, force_active_low);
 
-	unreachable();
+	return -EINVAL;
 }
diff --git a/drivers/net/phy/intel-xway.c b/drivers/net/phy/intel-xway.c
index b672c55a7a4e..e6ed2413e514 100644
--- a/drivers/net/phy/intel-xway.c
+++ b/drivers/net/phy/intel-xway.c
@@ -529,7 +529,7 @@ static int xway_gphy_led_polarity_set(struct phy_device *phydev, int index,
 	if (force_active_high)
 		return phy_clear_bits(phydev, XWAY_MDIO_LED, XWAY_GPHY_LED_INV(index));
 
-	unreachable();
+	return -EINVAL;
 }
 
 static struct phy_driver xway_gphy[] = {
diff --git a/drivers/net/phy/mxl-gpy.c b/drivers/net/phy/mxl-gpy.c
index db3c1f72b407..a8ccf257c109 100644
--- a/drivers/net/phy/mxl-gpy.c
+++ b/drivers/net/phy/mxl-gpy.c
@@ -1014,7 +1014,7 @@ static int gpy_led_polarity_set(struct phy_device *phydev, int index,
 	if (force_active_high)
 		return phy_clear_bits(phydev, PHY_LED, PHY_LED_POLARITY(index));
 
-	unreachable();
+	return -EINVAL;
 }
 
 static struct phy_driver gpy_drivers[] = {
-- 
2.39.5

On Tue, Dec 17, 2024 at 09:10:34AM +0100, Arnd Bergmann wrote:
> From: Arnd Bergmann <arnd@arndb.de>
> 
> gcc runs into undefined behavior at the end of the three led_polarity_set()
> callback functions if it were called with a zero 'modes' argument and it
> just ends the function there without returning from it.
> 
> This gets flagged by 'objtool' as a function that continues on
> to the next one:
> 
> drivers/net/phy/aquantia/aquantia_leds.o: warning: objtool: aqr_phy_led_polarity_set+0xf: can't find jump dest instruction at .text+0x5d9
> drivers/net/phy/intel-xway.o: warning: objtool: xway_gphy_led_polarity_set() falls through to next function xway_gphy_config_init()
> drivers/net/phy/mxl-gpy.o: warning: objtool: gpy_led_polarity_set() falls through to next function gpy_led_hw_control_get()
> 
> There is no point to micro-optimize the behavior here to save a single-digit
> number of bytes in the kernel, so just change this to a "return -EINVAL"
> as we do when any unexpected bits are set.
> 
> Fixes: 1758af47b98c ("net: phy: intel-xway: add support for PHY LEDs")
> Fixes: 9d55e68b19f2 ("net: phy: aquantia: correctly describe LED polarity override")
> Fixes: eb89c79c1b8f ("net: phy: mxl-gpy: correctly describe LED polarity")
> Signed-off-by: Arnd Bergmann <arnd@arndb.de>

Reviewed-by: Andrew Lunn <andrew@lunn.ch>

    Andrew