2023-04-02 11:30:34

by H. Nikolaus Schaller

[permalink] [raw]
Subject: [PATCH] leds: tca6507: fix error handling of using fwnode_property_read_string

Commit 96f524105b9c ("leds: tca6507: use fwnode API instead of OF")

changed to fwnode API but did not take into account that a missing property
"linux,default-trigger" now seems to return an error and as a side effect
sets value to -1. This seems to be different from of_get_property() which
always returned NULL in any case of error.

Neglecting this side-effect leads to

[ 11.201965] Unable to handle kernel paging request at virtual address ffffffff when read

in the strcmp() of led_trigger_set_default() if there is no led-trigger
defined in the DTS.

I don't know if this was recently introduced somewhere in the fwnode lib
or if the effect was missed in initial testing. Anyways it seems to be a
bug to ignore the error return value of an optional value here in the
driver.

Fixes: 96f524105b9c ("leds: tca6507: use fwnode API instead of OF")
Cc: Marek Behún <[email protected]>
Cc: NeilBrown <[email protected]>
Cc: Linus Walleij <[email protected]>
Signed-off-by: H. Nikolaus Schaller <[email protected]>
---
drivers/leds/leds-tca6507.c | 5 +++--
1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/drivers/leds/leds-tca6507.c b/drivers/leds/leds-tca6507.c
index 07dd12686a696..634cabd5bb796 100644
--- a/drivers/leds/leds-tca6507.c
+++ b/drivers/leds/leds-tca6507.c
@@ -691,8 +691,9 @@ tca6507_led_dt_init(struct device *dev)
if (fwnode_property_read_string(child, "label", &led.name))
led.name = fwnode_get_name(child);

- fwnode_property_read_string(child, "linux,default-trigger",
- &led.default_trigger);
+ if (fwnode_property_read_string(child, "linux,default-trigger",
+ &led.default_trigger))
+ led.default_trigger = NULL;

led.flags = 0;
if (fwnode_device_is_compatible(child, "gpio"))
--
2.38.1


2023-04-03 09:15:04

by Marek Behún

[permalink] [raw]
Subject: Re: [PATCH] leds: tca6507: fix error handling of using fwnode_property_read_string

On Sun, 2 Apr 2023 13:12:59 +0200
"H. Nikolaus Schaller" <[email protected]> wrote:

> Commit 96f524105b9c ("leds: tca6507: use fwnode API instead of OF")
>
> changed to fwnode API but did not take into account that a missing property
> "linux,default-trigger" now seems to return an error and as a side effect
> sets value to -1. This seems to be different from of_get_property() which
> always returned NULL in any case of error.
>
> Neglecting this side-effect leads to
>
> [ 11.201965] Unable to handle kernel paging request at virtual address ffffffff when read
>
> in the strcmp() of led_trigger_set_default() if there is no led-trigger
> defined in the DTS.
>
> I don't know if this was recently introduced somewhere in the fwnode lib
> or if the effect was missed in initial testing. Anyways it seems to be a
> bug to ignore the error return value of an optional value here in the
> driver.
>
> Fixes: 96f524105b9c ("leds: tca6507: use fwnode API instead of OF")
> Cc: Marek Behún <[email protected]>
> Cc: NeilBrown <[email protected]>
> Cc: Linus Walleij <[email protected]>
> Signed-off-by: H. Nikolaus Schaller <[email protected]>
> ---
> drivers/leds/leds-tca6507.c | 5 +++--
> 1 file changed, 3 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/leds/leds-tca6507.c b/drivers/leds/leds-tca6507.c
> index 07dd12686a696..634cabd5bb796 100644
> --- a/drivers/leds/leds-tca6507.c
> +++ b/drivers/leds/leds-tca6507.c
> @@ -691,8 +691,9 @@ tca6507_led_dt_init(struct device *dev)
> if (fwnode_property_read_string(child, "label", &led.name))
> led.name = fwnode_get_name(child);
>
> - fwnode_property_read_string(child, "linux,default-trigger",
> - &led.default_trigger);
> + if (fwnode_property_read_string(child, "linux,default-trigger",
> + &led.default_trigger))
> + led.default_trigger = NULL;
>
> led.flags = 0;
> if (fwnode_device_is_compatible(child, "gpio"))

Reviewed-by: Marek Behún <[email protected]>

2023-04-03 15:32:58

by Pavel Machek

[permalink] [raw]
Subject: Re: [PATCH] leds: tca6507: fix error handling of using fwnode_property_read_string

On Sun 2023-04-02 13:12:59, H. Nikolaus Schaller wrote:
> Commit 96f524105b9c ("leds: tca6507: use fwnode API instead of OF")
>
> changed to fwnode API but did not take into account that a missing property
> "linux,default-trigger" now seems to return an error and as a side effect
> sets value to -1. This seems to be different from of_get_property() which
> always returned NULL in any case of error.
>
> Neglecting this side-effect leads to
>
> [ 11.201965] Unable to handle kernel paging request at virtual address ffffffff when read
>
> in the strcmp() of led_trigger_set_default() if there is no led-trigger
> defined in the DTS.
>
> I don't know if this was recently introduced somewhere in the fwnode lib
> or if the effect was missed in initial testing. Anyways it seems to be a
> bug to ignore the error return value of an optional value here in the
> driver.
>
> Fixes: 96f524105b9c ("leds: tca6507: use fwnode API instead of OF")
> Cc: Marek Beh?n <[email protected]>
> Cc: NeilBrown <[email protected]>
> Cc: Linus Walleij <[email protected]>
> Signed-off-by: H. Nikolaus Schaller <[email protected]>

Acked-by: Pavel Machek <[email protected]>

BR, Pavel
--
People of Russia, stop Putin before his war on Ukraine escalates.


Attachments:
(No filename) (1.25 kB)
signature.asc (201.00 B)
Download all attachments

2023-04-05 15:31:52

by Lee Jones

[permalink] [raw]
Subject: Re: [PATCH] leds: tca6507: fix error handling of using fwnode_property_read_string

On Sun, 02 Apr 2023, H. Nikolaus Schaller wrote:

> Commit 96f524105b9c ("leds: tca6507: use fwnode API instead of OF")
>
> changed to fwnode API but did not take into account that a missing property
> "linux,default-trigger" now seems to return an error and as a side effect
> sets value to -1. This seems to be different from of_get_property() which
> always returned NULL in any case of error.
>
> Neglecting this side-effect leads to
>
> [ 11.201965] Unable to handle kernel paging request at virtual address ffffffff when read
>
> in the strcmp() of led_trigger_set_default() if there is no led-trigger
> defined in the DTS.
>
> I don't know if this was recently introduced somewhere in the fwnode lib
> or if the effect was missed in initial testing. Anyways it seems to be a
> bug to ignore the error return value of an optional value here in the
> driver.
>
> Fixes: 96f524105b9c ("leds: tca6507: use fwnode API instead of OF")
> Cc: Marek Behún <[email protected]>
> Cc: NeilBrown <[email protected]>
> Cc: Linus Walleij <[email protected]>
> Signed-off-by: H. Nikolaus Schaller <[email protected]>
> ---
> drivers/leds/leds-tca6507.c | 5 +++--
> 1 file changed, 3 insertions(+), 2 deletions(-)

Applied, thanks

--
Lee Jones [李琼斯]