[PATCH] net: sfp: Fix memory leak of hwmon_name on hwmon registration failure

Krishan Singh posted 1 patch 1 month, 4 weeks ago
drivers/net/phy/sfp.c | 4 ++++
1 file changed, 4 insertions(+)
[PATCH] net: sfp: Fix memory leak of hwmon_name on hwmon registration failure
Posted by Krishan Singh 1 month, 4 weeks ago
    hwmon_sanitize_name() allocates sfp->hwmon_name before
    hwmon_device_register_with_info() is called. If the registration
    fails, sfp->hwmon_dev is left as an error pointer while
    sfp->hwmon_name remains allocated.

    Later, when the SFP module is removed, sfp_hwmon_remove() is still
    called. However, it frees sfp->hwmon_name only when
    !IS_ERR_OR_NULL(sfp->hwmon_dev) is true. Since sfp->hwmon_dev is an
    error pointer in the failure case, the cleanup block is skipped and
    hwmon_name is leaked.

    Fix this by cleaning up hwmon_name independently of hwmon_dev.
    Continue to unregister the hwmon device only when hwmon_dev is valid,
    but free hwmon_name whenever it is a valid allocated pointer.

    Fixes: 3f118c449c8e ("net: sfp: use hwmon_sanitize_name()")
    Suggested-by: Andrew Lunn <andrew@lunn.ch>
    Signed-off-by: Krishan Singh <krishanmohan298@gmail.com>
---
 drivers/net/phy/sfp.c | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/drivers/net/phy/sfp.c b/drivers/net/phy/sfp.c
index f52020673..bfa2b821f 100644
--- a/drivers/net/phy/sfp.c
+++ b/drivers/net/phy/sfp.c
@@ -1916,7 +1916,11 @@ static void sfp_hwmon_remove(struct sfp *sfp)
 	if (!IS_ERR_OR_NULL(sfp->hwmon_dev)) {
 		hwmon_device_unregister(sfp->hwmon_dev);
 		sfp->hwmon_dev = NULL;
+	}
+
+	if (!IS_ERR_OR_NULL(sfp->hwmon_name)) {
 		kfree(sfp->hwmon_name);
+		sfp->hwmon_name = NULL;
 	}
 }
 
-- 
2.34.1
Re: [PATCH] net: sfp: Fix memory leak of hwmon_name on hwmon registration failure
Posted by Andrew Lunn 1 month, 3 weeks ago
On Wed, Aug 05, 2026 at 10:06:43AM +0530, Krishan Singh wrote:
>     hwmon_sanitize_name() allocates sfp->hwmon_name before
>     hwmon_device_register_with_info() is called. If the registration
>     fails, sfp->hwmon_dev is left as an error pointer while
>     sfp->hwmon_name remains allocated.
> 
>     Later, when the SFP module is removed, sfp_hwmon_remove() is still
>     called. However, it frees sfp->hwmon_name only when
>     !IS_ERR_OR_NULL(sfp->hwmon_dev) is true. Since sfp->hwmon_dev is an
>     error pointer in the failure case, the cleanup block is skipped and
>     hwmon_name is leaked.
> 
>     Fix this by cleaning up hwmon_name independently of hwmon_dev.
>     Continue to unregister the hwmon device only when hwmon_dev is valid,
>     but free hwmon_name whenever it is a valid allocated pointer.
> 
>     Fixes: 3f118c449c8e ("net: sfp: use hwmon_sanitize_name()")
>     Suggested-by: Andrew Lunn <andrew@lunn.ch>
>     Signed-off-by: Krishan Singh <krishanmohan298@gmail.com>
> ---

Please read

https://docs.kernel.org/process/submitting-patches.html

There should be a version number in the Subject: line, and under the
--- a version history.

https://www.kernel.org/doc/html/latest/process/maintainer-netdev.html

For netdev, we want the tree to be indicated in the Subject: line. For
this patch i would suggest net-next.

    Andrew

---
pw-bot: cr
Re: [PATCH] net: sfp: Fix memory leak of hwmon_name on hwmon registration failure
Posted by krishan mohan 1 month, 3 weeks ago
Hi All,

Please find v2 of this patch.
Changes since v1: - Move hwmon_name cleanup to sfp_hwmon_remove(). - Free
hwmon_name independently of hwmon_dev.

---
v2:
- Move hwmon_name cleanup to sfp_hwmon_remove().
- Free hwmon_name independently of hwmon_dev.
---
 drivers/net/phy/sfp.c | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/drivers/net/phy/sfp.c b/drivers/net/phy/sfp.c
index f52020673..bfa2b821f 100644
--- a/drivers/net/phy/sfp.c
+++ b/drivers/net/phy/sfp.c
@@ -1916,7 +1916,11 @@ static void sfp_hwmon_remove(struct sfp *sfp)
  if (!IS_ERR_OR_NULL(sfp->hwmon_dev)) {
  hwmon_device_unregister(sfp->hwmon_dev);
  sfp->hwmon_dev = NULL;
+ }
+
+ if (!IS_ERR_OR_NULL(sfp->hwmon_name)) {
  kfree(sfp->hwmon_name);
+ sfp->hwmon_name = NULL;
  }
 }

-- 


Thanks&Regards
Krishan Mohan Singh


On Wed, Aug 5, 2026 at 5:28 PM Andrew Lunn <andrew@lunn.ch> wrote:

> On Wed, Aug 05, 2026 at 10:06:43AM +0530, Krishan Singh wrote:
> >     hwmon_sanitize_name() allocates sfp->hwmon_name before
> >     hwmon_device_register_with_info() is called. If the registration
> >     fails, sfp->hwmon_dev is left as an error pointer while
> >     sfp->hwmon_name remains allocated.
> >
> >     Later, when the SFP module is removed, sfp_hwmon_remove() is still
> >     called. However, it frees sfp->hwmon_name only when
> >     !IS_ERR_OR_NULL(sfp->hwmon_dev) is true. Since sfp->hwmon_dev is an
> >     error pointer in the failure case, the cleanup block is skipped and
> >     hwmon_name is leaked.
> >
> >     Fix this by cleaning up hwmon_name independently of hwmon_dev.
> >     Continue to unregister the hwmon device only when hwmon_dev is valid,
> >     but free hwmon_name whenever it is a valid allocated pointer.
> >
> >     Fixes: 3f118c449c8e ("net: sfp: use hwmon_sanitize_name()")
> >     Suggested-by: Andrew Lunn <andrew@lunn.ch>
> >     Signed-off-by: Krishan Singh <krishanmohan298@gmail.com>
> > ---
>
> Please read
>
> https://docs.kernel.org/process/submitting-patches.html
>
> There should be a version number in the Subject: line, and under the
> --- a version history.
>
> https://www.kernel.org/doc/html/latest/process/maintainer-netdev.html
>
> For netdev, we want the tree to be indicated in the Subject: line. For
> this patch i would suggest net-next.
>
>     Andrew
>
> ---
> pw-bot: cr
>
Re: [PATCH] net: sfp: Fix memory leak of hwmon_name on hwmon registration failure
Posted by Andrew Lunn 1 month, 3 weeks ago
On Fri, Aug 07, 2026 at 12:52:58PM +0530, krishan mohan wrote:
> Hi All,
> 
> Please find v2 of this patch. 
> Changes since v1: - Move hwmon_name cleanup to sfp_hwmon_remove(). - Free
> hwmon_name independently of hwmon_dev.

Please read:

https://docs.kernel.org/process/submitting-patches.html

and then submit a proper patch, following the process defined in this
document.

It is also important you start a new thread, otherwise the CI probably
does not work.

    Andrew

---
pw-bot: cr