回复: [PATCH v1] net:thunder_bgx:Fix resource leaks in device_for_each_child_node() loops

[王明-软件底层技术部]

Thank you.Ok, I will repost [PATCH net v2] net: thunderx:... in 24 hours. 
-----邮件原件-----
发件人: Simon Horman <simon.horman at corigine.com> 
发送时间: 2023年7月5日 4:30
收件人: 王明-软件底层技术部 <machel at vivo.com>
抄送: Sunil Goutham <sgoutham at marvell.com>; David S. Miller <davem at davemloft.net>; Eric Dumazet <edumazet at google.com>; Jakub Kicinski <kuba at kernel.org>; Paolo Abeni <pabeni at redhat.com>; linux-arm-kernel at lists.infradead.org; netdev at vger.kernel.org; linux-kernel at vger.kernel.org; opensource.kernel <opensource.kernel at vivo.com>
主题: Re: [PATCH v1] net:thunder_bgx:Fix resource leaks in device_for_each_child_node() loops

On Tue, Jul 04, 2023 at 10:14:47PM +0800, Wang Ming wrote:
> The device_for_each_child_node() loop in
> bgx_init_of_phy() function should have
> fwnode_handle_put() before break which could avoid resource leaks.
> This patch could fix this bug.

Hi Wang Ming,

If this fixes a bug then it probably warrants a fixes tag.
And it should also probably be targeted at the next tree:

	[PATCH net v2] ...

If so, please allow 24h to elapse before posting v2.

Else it should be targeted at the net-next tree, and posted after net-next reopens after July 10th:

	[PATCH net-next v2] ...

Also, looking at git history, I think a more appropriate subject prefix would be 'net: thunderx: '

	[PATCH net v2] net: thunderx: ...

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

> Signed-off-by: Wang Ming <machel at vivo.com>
> ---
>  drivers/net/ethernet/cavium/thunder/thunder_bgx.c | 5 ++++-
>  1 file changed, 4 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/net/ethernet/cavium/thunder/thunder_bgx.c 
> b/drivers/net/ethernet/cavium/thunder/thunder_bgx.c
> index a317feb8d..dad32d36a 100644
> --- a/drivers/net/ethernet/cavium/thunder/thunder_bgx.c
> +++ b/drivers/net/ethernet/cavium/thunder/thunder_bgx.c
> @@ -1478,8 +1478,10 @@ static int bgx_init_of_phy(struct bgx *bgx)
>  		 * cannot handle it, so exit the loop.
>  		 */
>  		node = to_of_node(fwn);
> -		if (!node)
> +		if (!node) {
> +			fwnode_handle_put(fwn);
>  			break;
> +		}
>  
>  		of_get_mac_address(node, bgx->lmac[lmac].mac);
>  
> @@ -1503,6 +1505,7 @@ static int bgx_init_of_phy(struct bgx *bgx)
>  		lmac++;
>  		if (lmac == bgx->max_lmac) {
>  			of_node_put(node);
> +			fwnode_handle_put(fwn);
>  			break;
>  		}
>  	}

Should this change also apply to the 'goto defer' case in the same loop?
If so, perhaps a more idiomatic approach of releasing resources in a ladder of goto labels is appropriate. Something like this (untested!):

diff --git a/drivers/net/ethernet/cavium/thunder/thunder_bgx.c b/drivers/net/ethernet/cavium/thunder/thunder_bgx.c
index a317feb8decb..3091c96134e4 100644
--- a/drivers/net/ethernet/cavium/thunder/thunder_bgx.c
+++ b/drivers/net/ethernet/cavium/thunder/thunder_bgx.c
@@ -1469,6 +1469,7 @@ static int bgx_init_of_phy(struct bgx *bgx)
 	struct fwnode_handle *fwn;
 	struct device_node *node = NULL;
 	u8 lmac = 0;
+	int err = 0;
 
 	device_for_each_child_node(&bgx->pdev->dev, fwn) {
 		struct phy_device *pd;
@@ -1479,7 +1480,7 @@ static int bgx_init_of_phy(struct bgx *bgx)
 		 */
 		node = to_of_node(fwn);
 		if (!node)
-			break;
+			goto out_handle_put;
 
 		of_get_mac_address(node, bgx->lmac[lmac].mac);
 
@@ -1501,10 +1502,8 @@ static int bgx_init_of_phy(struct bgx *bgx)
 		}
 
 		lmac++;
-		if (lmac == bgx->max_lmac) {
-			of_node_put(node);
-			break;
-		}
+		if (lmac == bgx->max_lmac)
+			goto out_node_put;
 	}
 	return 0;
 
@@ -1519,8 +1518,12 @@ static int bgx_init_of_phy(struct bgx *bgx)
 		}
 		lmac--;
 	}
+	err = -EPROBE_DEFER;
+out_node_put:
 	of_node_put(node);
-	return -EPROBE_DEFER;
+out_handle_put:
+	fwnode_handle_put(fwn);
+	return err;
 }
 
 #else

--
pw-bot: changes-requested