Re: stmmac: pci: Add dwmac support for Loongson

From: Andrew Lunn
Date: Mon Jun 21 2021 - 12:54:18 EST


On Mon, Jun 21, 2021 at 04:51:28PM +0100, Colin Ian King wrote:
> Hi,
>
> Static analysis by Coverity on today's linux-next has found an issue in
> function loongson_dwmac_probe with the following commit:
>
> commit 30bba69d7db40e732d6c0aa6d4890c60d717e314
> Author: Qing Zhang <zhangqing@xxxxxxxxxxx>
> Date: Fri Jun 18 10:53:34 2021 +0800
>
> stmmac: pci: Add dwmac support for Loongson
>
> The analysis is as follows:
>
> 110 plat->phy_interface = device_get_phy_mode(&pdev->dev);
>
> Enum compared against 0
> (NO_EFFECT)
> unsigned_compare: This less-than-zero comparison of an
> unsigned value is never true. plat->phy_interface < 0U.
>
> 111 if (plat->phy_interface < 0)
> 112 dev_err(&pdev->dev, "phy_mode not found\n");
>
> Enum plat->phy_interface is unsigned, so can't be negative and so the
> comparison will always be false.
>
> A possible fix is to use int variable ret for the assignment and check:
>
>
> ret = device_get_phy_mode(&pdev->dev);
> if (ret < 0)
> dev_err(&pdev->dev, "phy_mode not found\n");
> plat->phy_interface = ret;
>
> ..however, I think the dev_err may need handling too, e.g.
>
> ret = device_get_phy_mode(&pdev->dev);
> if (ret < 0) {
> dev_err(&pdev->dev, "phy_mode not found\n");
> ret = -ENODEV;
> goto cleanup; /* needs to be written */
> }
> plat->phy_interface = ret;

Potentially a cleaner fix is to use of_get_phy_mode(), which separates
the success/error value from the phy_interface_t.

Andrew