RE: [PATCH V3 1/2] soc: imx: Add SCU SoC info driver support

[Anson Huang]

Hi, Leonard

> -----Original Message-----
> From: Leonard Crestez
> Sent: Thursday, May 16, 2019 6:07 PM
> To: Anson Huang <anson.huang@xxxxxxx>; shawnguo@xxxxxxxxxx
> Cc: catalin.marinas@xxxxxxx; will.deacon@xxxxxxx;
> s.hauer@xxxxxxxxxxxxxx; kernel@xxxxxxxxxxxxxx; festevam@xxxxxxxxx;
> agross@xxxxxxxxxx; maxime.ripard@xxxxxxxxxxx; olof@xxxxxxxxx;
> horms+renesas@xxxxxxxxxxxx; jagan@xxxxxxxxxxxxxxxxxxxx;
> bjorn.andersson@xxxxxxxxxx; marc.w.gonzalez@xxxxxxx;
> dinguyen@xxxxxxxxxx; enric.balletbo@xxxxxxxxxxxxx;
> l.stach@xxxxxxxxxxxxxx; Aisheng Dong <aisheng.dong@xxxxxxx>; Abel Vesa
> <abel.vesa@xxxxxxx>; robh@xxxxxxxxxx; linux-arm-
> kernel@xxxxxxxxxxxxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx; dl-linux-imx
> <linux-imx@xxxxxxx>
> Subject: Re: [PATCH V3 1/2] soc: imx: Add SCU SoC info driver support
> 
> On 16.05.2019 06:24, Anson Huang wrote:
> > Add i.MX SCU SoC info driver to support i.MX8QXP SoC, introduce driver
> > dependency into Kconfig as CONFIG_IMX_SCU must be selected to support
> > i.MX SCU SoC driver, also need to use platform driver model to make
> > sure IMX_SCU driver is probed before i.MX SCU SoC driver.
> 
> > +#define imx_scu_revision(soc_rev) \
> > +	soc_rev ? \
> > +	kasprintf(GFP_KERNEL, "%d.%d", (soc_rev >> 4) & 0xf,  soc_rev &
> 0xf) : \
> > +	"unknown"
> 
> > +	id = of_match_node(imx_scu_soc_match, pdev->dev.of_node);
> > +	data = id->data;
> > +	if (data) {
> > +		soc_dev_attr->soc_id = data->name;
> > +		if (data->soc_revision)
> > +			soc_rev = data->soc_revision();
> > +	}
> > +
> > +	soc_dev_attr->revision = imx_scu_revision(soc_rev);
> > +	if (!soc_dev_attr->revision)
> > +		return -ENODEV;
> 
> The imx_scu_revision macro returns either kasprintf or "unknown", never
> NULL. So it's not clear what this return -ENODEV does exactly.

The kasprintf could return NULL though.

> 
> It makes more sense to return -ENODEV if get_soc_revision fails, so maybe
> check "soc_rev != 0" instead?
> 
> If you really want to check the kasprintf result then you should return -
> ENOMEM for it. It would be clearer if you dropped the imx_scu_revision
> revision macro and open-coded instead.

This makes more sense, I think maybe we can remove the imx_scu_revision macro,
just use below code instead, and return -ENOMEM if kasprintf returns NULL.

113         soc_dev_attr->revision = soc_rev ? kasprintf(GFP_KERNEL,
114                                                      "%d.%d",
115                                                      (soc_rev >> 4) & 0xf,
116                                                      soc_rev & 0xf) : "unknown";
117         if (!soc_dev_attr->revision)
118                 return -ENOMEM;

BTW, the soc-imx8.c looks like also having same issue, do you think we should fix it
as well?

Anson

> 
> --
> Regards,
> Leonard