Re: [PATCH v2 2/3] phy: phy-rockchip-inno-usb2: Handle failed extcon allocation better

[Dragan Simic]

Hello Christophe,

On 2024-08-21 13:17, Christophe JAILLET wrote:
> Le 21/08/2024 à 09:37, Dragan Simic a écrit :
> > Return the actual error code upon failure to allocate extcon device, 
> > instead
> > of hardcoding -ENOMEM.  Use dev_err_probe() to also log appropriate 
> > messages,
> > which is fine because the containing function is used in the probe 
> > path.
> >
> > Helped-by: Heiko Stubner <heiko@xxxxxxxxx>
> > Signed-off-by: Dragan Simic <dsimic@xxxxxxxxxxx>
> > ---
> >   drivers/phy/rockchip/phy-rockchip-inno-usb2.c | 3 ++-
> >   1 file changed, 2 insertions(+), 1 deletion(-)
> >
> > diff --git a/drivers/phy/rockchip/phy-rockchip-inno-usb2.c 
> > b/drivers/phy/rockchip/phy-rockchip-inno-usb2.c
> > index 113bfc717ff0..05af46dda11d 100644
> > --- a/drivers/phy/rockchip/phy-rockchip-inno-usb2.c
> > +++ b/drivers/phy/rockchip/phy-rockchip-inno-usb2.c
> > @@ -435,7 +435,8 @@ static int rockchip_usb2phy_extcon_register(struct 
> > rockchip_usb2phy *rphy)
> >   						rockchip_usb2phy_extcon_cable);
> >     		if (IS_ERR(edev))
> > -			return -ENOMEM;
> > +			return dev_err_probe(rphy->dev, PTR_ERR(edev),
> > +					     "failed to allocate extcon device\n");
>
> Returning PTR_ERR(edev) may make sense, but I don't think that adding
> a dev_err_probe() really helps.
>
> devm_extcon_dev_allocate() can only return -EINVAL if it 2nd argument
> is NULL. It is trivial to see that it can't happen here,
> rockchip_usb2phy_extcon_cable is a global variable.
>
> in all other cases, it returns -ENOMEM because of a failed memory
> allocation. In this case, usually it is not needed to log anything
> because it is already loud enough.

True, using dev_err_probe() in this case is somewhat redundant,
but it falls under the "still fine to use" category, [1] because
the error code passed to dev_err_probe() is received from another
function that was invoked.

On the other hand, another patch in this series tries to be strict
in another direction, by not using dev_err_probe() where the error
code passed to it is basically hardcoded. [2]

I hope this makes sense.

[1] 
https://lore.kernel.org/linux-phy/cover.1724225528.git.dsimic@xxxxxxxxxxx/T/#mc3af7d24e31ed885732e6f26ca0d107b157d478b
[2] 
https://lore.kernel.org/linux-phy/cover.1724225528.git.dsimic@xxxxxxxxxxx/T/#ma26b614412787814dab7923987b8c814f7e86beb

> >     		ret = devm_extcon_dev_register(rphy->dev, edev);
> >   		if (ret) {