From: Zijun Hu <quic_zijuhu@quicinc.com>

For devm_phy_put(), its comment says it needs to invoke phy_put() to
release the phy, but it will not actually invoke the function since
devres_destroy() does not call devm_phy_release(), and the missing
phy_put() call will cause:

- The phy fails to be released.
- devm_phy_put() can not fully undo what API devm_phy_get() does.
- Leak refcount of both the module and device for below typical usage:

  devm_phy_get(); // or its variant
  ...
  err = do_something();
  if (err)
      goto err_out;
  ...
  err_out:
  devm_phy_put(); // leak refcount here

  The file(s) affected by this issue are shown below since they have such
  typical usage.
  drivers/pci/controller/cadence/pcie-cadence.c
  drivers/net/ethernet/ti/am65-cpsw-nuss.c

Fixed by using devres_release() instead of devres_destroy() within the API.

Fixes: ff764963479a ("drivers: phy: add generic PHY framework")
Cc: stable@vger.kernel.org
Cc: Lorenzo Pieralisi <lpieralisi@kernel.org>
Cc: "Krzysztof Wilczyński" <kw@linux.com>
Cc: Bjorn Helgaas <bhelgaas@google.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Jakub Kicinski <kuba@kernel.org>
Cc: Paolo Abeni <pabeni@redhat.com>
Cc: Johan Hovold <johan@kernel.org>
Signed-off-by: Zijun Hu <quic_zijuhu@quicinc.com>
---
 drivers/phy/phy-core.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/phy/phy-core.c b/drivers/phy/phy-core.c
index f053b525ccff..f190d7126613 100644
--- a/drivers/phy/phy-core.c
+++ b/drivers/phy/phy-core.c
@@ -737,7 +737,7 @@ void devm_phy_put(struct device *dev, struct phy *phy)
 	if (!phy)
 		return;
 
-	r = devres_destroy(dev, devm_phy_release, devm_phy_match, phy);
+	r = devres_release(dev, devm_phy_release, devm_phy_match, phy);
 	dev_WARN_ONCE(dev, r, "couldn't find PHY resource\n");
 }
 EXPORT_SYMBOL_GPL(devm_phy_put);

-- 
2.34.1

On Sat, Nov 02, 2024 at 11:53:43AM +0800, Zijun Hu wrote:
> From: Zijun Hu <quic_zijuhu@quicinc.com>
> 
> For devm_phy_put(), its comment says it needs to invoke phy_put() to
> release the phy, but it will not actually invoke the function since
> devres_destroy() does not call devm_phy_release(), and the missing
> phy_put() call will cause:

I still think the above should have been split in two sentences, but
this also works.

> - The phy fails to be released.
> - devm_phy_put() can not fully undo what API devm_phy_get() does.
> - Leak refcount of both the module and device for below typical usage:
> 
>   devm_phy_get(); // or its variant
>   ...
>   err = do_something();
>   if (err)
>       goto err_out;
>   ...
>   err_out:
>   devm_phy_put(); // leak refcount here
> 
>   The file(s) affected by this issue are shown below since they have such
>   typical usage.
>   drivers/pci/controller/cadence/pcie-cadence.c
>   drivers/net/ethernet/ti/am65-cpsw-nuss.c
> 
> Fixed by using devres_release() instead of devres_destroy() within the API.

Nit: in the future, try to use imperative mood in your commit messages
(e.g. "fix" instead of "fixed") as the process documentation suggests.

> Fixes: ff764963479a ("drivers: phy: add generic PHY framework")
> Cc: stable@vger.kernel.org
> Cc: Lorenzo Pieralisi <lpieralisi@kernel.org>
> Cc: "Krzysztof Wilczyński" <kw@linux.com>
> Cc: Bjorn Helgaas <bhelgaas@google.com>
> Cc: "David S. Miller" <davem@davemloft.net>
> Cc: Eric Dumazet <edumazet@google.com>
> Cc: Jakub Kicinski <kuba@kernel.org>
> Cc: Paolo Abeni <pabeni@redhat.com>
> Cc: Johan Hovold <johan@kernel.org>
> Signed-off-by: Zijun Hu <quic_zijuhu@quicinc.com>

Reviewed-by: Johan Hovold <johan+linaro@kernel.org>