On Mon, Aug 07, 2023 at 01:47:14PM +0000, Xiang Yang wrote:
> From: Xiang Yang <xiangyang3@xxxxxxxxxx>
>
> The reference of pdev->dev is taken by of_find_device_by_node, so
> it should be released when error out.
>
> Fixes: 7dc54d3b8d91 ("net: pcs: add Renesas MII converter driver")
> Signed-off-by: Xiang Yang <xiangyang3@xxxxxxxxxx>
> ---
Ok, but if of_find_device_by_node() requires a subsequent call to
put_device(), then doesn't this mean that in the non-error case, the
pdev refcount remains elevated, and will thus never be freed?
I would expect another put_device(&pdev->dev) after the device_link_add(),
which itself takes another set of references on the supplier and
consumer devs, which are sufficient for this driver's runtime operation.
Also, "if (!pdev)" and "if (!platform_get_drvdata(pdev)" here need
different treatment. One needs to call put_device(&pdev->dev) and the
other one doesn't.
pdev = of_find_device_by_node(pcs_np);
of_node_put(pcs_np);
if (!pdev || !platform_get_drvdata(pdev))
return ERR_PTR(-EPROBE_DEFER);
pw-bot: cr
> drivers/net/pcs/pcs-rzn1-miic.c | 4 +++-
> 1 file changed, 3 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/net/pcs/pcs-rzn1-miic.c b/drivers/net/pcs/pcs-rzn1-miic.c
> index e5d642c67a2c..d097f123635a 100644
> --- a/drivers/net/pcs/pcs-rzn1-miic.c
> +++ b/drivers/net/pcs/pcs-rzn1-miic.c
> @@ -318,8 +318,10 @@ struct phylink_pcs *miic_create(struct device *dev, struct device_node *np)
> return ERR_PTR(-EPROBE_DEFER);
>
> miic_port = kzalloc(sizeof(*miic_port), GFP_KERNEL);
> - if (!miic_port)
> + if (!miic_port) {
> + put_device(&pdev->dev);
> return ERR_PTR(-ENOMEM);
> + }
>
> miic = platform_get_drvdata(pdev);
> device_link_add(dev, miic->dev, DL_FLAG_AUTOREMOVE_CONSUMER);
> --
> 2.34.1
>
