2023-12-12 09:21:11

by Dan Carpenter

Subject: [PATCH] cdx: Unlock on error path in rescan_store()

We added locking to this function but these two error paths were
accidentally overlooked.

Fixes: f0af81683466 ("cdx: Introduce lock to protect controller ops")
Signed-off-by: Dan Carpenter <[email protected]>
---
drivers/cdx/cdx.c | 14 +++++++++-----
1 file changed, 9 insertions(+), 5 deletions(-)

diff --git a/drivers/cdx/cdx.c b/drivers/cdx/cdx.c
index d84d153078d7..f4f9f0c88c09 100644
--- a/drivers/cdx/cdx.c
+++ b/drivers/cdx/cdx.c
@@ -572,12 +572,16 @@ static ssize_t rescan_store(const struct bus_type *bus,

/* Rescan all the devices */
for_each_compatible_node(np, NULL, compat_node_name) {
- if (!np)
- return -EINVAL;
+ if (!np) {
+ count = -EINVAL;
+ goto unlock;
+ }

pd = of_find_device_by_node(np);
- if (!pd)
- return -EINVAL;
+ if (!pd) {
+ count = -EINVAL;
+ goto unlock;
+ }

cdx = platform_get_drvdata(pd);
if (cdx && cdx->controller_registered && cdx->ops->scan)
@@ -585,7 +589,7 @@ static ssize_t rescan_store(const struct bus_type *bus,

put_device(&pd->dev);
}
-
+unlock:
mutex_unlock(&cdx_controller_lock);

return count;
--
2.42.0

2023-12-12 17:53:35

by Christophe JAILLET

Subject: Re: [PATCH] cdx: Unlock on error path in rescan_store()

Le 12/12/2023 à 10:20, Dan Carpenter a écrit :
> We added locking to this function but these two error paths were
> accidentally overlooked.
>
> Fixes: f0af81683466 ("cdx: Introduce lock to protect controller ops")
> Signed-off-by: Dan Carpenter <[email protected]>
> ---
> drivers/cdx/cdx.c | 14 +++++++++-----
> 1 file changed, 9 insertions(+), 5 deletions(-)
>
> diff --git a/drivers/cdx/cdx.c b/drivers/cdx/cdx.c
> index d84d153078d7..f4f9f0c88c09 100644
> --- a/drivers/cdx/cdx.c
> +++ b/drivers/cdx/cdx.c
> @@ -572,12 +572,16 @@ static ssize_t rescan_store(const struct bus_type *bus,
>
> /* Rescan all the devices */
> for_each_compatible_node(np, NULL, compat_node_name) {
> - if (!np)
> - return -EINVAL;
> + if (!np) {
> + count = -EINVAL;
> + goto unlock;
> + }
>
> pd = of_find_device_by_node(np);
> - if (!pd)
> - return -EINVAL;
> + if (!pd) {
> + count = -EINVAL;
> + goto unlock;

Unrelated to your patch, but should we have a of_node_put(np); here, on
early exit?

CJ

> + }
>
> cdx = platform_get_drvdata(pd);
> if (cdx && cdx->controller_registered && cdx->ops->scan)
> @@ -585,7 +589,7 @@ static ssize_t rescan_store(const struct bus_type *bus,
>
> put_device(&pd->dev);
> }
> -
> +unlock:
> mutex_unlock(&cdx_controller_lock);
>
> return count;

2023-12-13 04:22:54

by Agarwal, Nikhil

Subject: RE: [PATCH] cdx: Unlock on error path in rescan_store()

> -----Original Message-----
> From: Dan Carpenter <[email protected]>
> Sent: Tuesday, December 12, 2023 2:50 PM
> To: Gangurde, Abhijit <[email protected]>
> Cc: Gupta, Nipun <[email protected]>; Agarwal, Nikhil
> <[email protected]>; Greg Kroah-Hartman
> <[email protected]>; [email protected]; kernel-
> [email protected]
> Subject: [PATCH] cdx: Unlock on error path in rescan_store()
>
> We added locking to this function but these two error paths were accidentally
> overlooked.
>
> Fixes: f0af81683466 ("cdx: Introduce lock to protect controller ops")
> Signed-off-by: Dan Carpenter <[email protected]>

Acked-by: Nikhil Agarwal <[email protected]>

2023-12-13 06:14:29

by Gangurde, Abhijit

Subject: RE: [PATCH] cdx: Unlock on error path in rescan_store()

> Le 12/12/2023 à 10:20, Dan Carpenter a écrit :
> > We added locking to this function but these two error paths were
> > accidentally overlooked.
> >
> > Fixes: f0af81683466 ("cdx: Introduce lock to protect controller ops")
> > Signed-off-by: Dan Carpenter <[email protected]>
> > ---
> > drivers/cdx/cdx.c | 14 +++++++++-----
> > 1 file changed, 9 insertions(+), 5 deletions(-)
> >
> > diff --git a/drivers/cdx/cdx.c b/drivers/cdx/cdx.c
> > index d84d153078d7..f4f9f0c88c09 100644
> > --- a/drivers/cdx/cdx.c
> > +++ b/drivers/cdx/cdx.c
> > @@ -572,12 +572,16 @@ static ssize_t rescan_store(const struct bus_type
> *bus,
> >
> > /* Rescan all the devices */
> > for_each_compatible_node(np, NULL, compat_node_name) {
> > - if (!np)
> > - return -EINVAL;
> > + if (!np) {
> > + count = -EINVAL;
> > + goto unlock;
> > + }
> >
> > pd = of_find_device_by_node(np);
> > - if (!pd)
> > - return -EINVAL;
> > + if (!pd) {
> > + count = -EINVAL;
> > + goto unlock;
>
> Unrelated to your patch, but should we have a of_node_put(np); here, on
> early exit?

Yes. of_node_put(np) is needed here.

Thanks,
Abhijit

2023-12-13 09:47:43

by Dan Carpenter

Subject: Re: [PATCH] cdx: Unlock on error path in rescan_store()

On Tue, Dec 12, 2023 at 06:53:13PM +0100, Christophe JAILLET wrote:
> Le 12/12/2023 ? 10:20, Dan Carpenter a ?crit?:
> > We added locking to this function but these two error paths were
> > accidentally overlooked.
> >
> > Fixes: f0af81683466 ("cdx: Introduce lock to protect controller ops")
> > Signed-off-by: Dan Carpenter <[email protected]>
> > ---
> > drivers/cdx/cdx.c | 14 +++++++++-----
> > 1 file changed, 9 insertions(+), 5 deletions(-)
> >
> > diff --git a/drivers/cdx/cdx.c b/drivers/cdx/cdx.c
> > index d84d153078d7..f4f9f0c88c09 100644
> > --- a/drivers/cdx/cdx.c
> > +++ b/drivers/cdx/cdx.c
> > @@ -572,12 +572,16 @@ static ssize_t rescan_store(const struct bus_type *bus,
> > /* Rescan all the devices */
> > for_each_compatible_node(np, NULL, compat_node_name) {
> > - if (!np)
> > - return -EINVAL;
> > + if (!np) {
> > + count = -EINVAL;
> > + goto unlock;
> > + }
> > pd = of_find_device_by_node(np);
> > - if (!pd)
> > - return -EINVAL;
> > + if (!pd) {
> > + count = -EINVAL;
> > + goto unlock;
>
> Unrelated to your patch, but should we have a of_node_put(np); here, on
> early exit?
>

Let me resend this along with a patch 2/2 which adds the of_node_put().

regards,
dan carpenter