Re: [PATCH 1/5] soc: hpe: Add GXP SROM Control Register Driver

[Clay Chang]

Hi Krzysztof,

Thank you for the time to review my code.

On Tue, Jan 10, 2023 at 10:46:57AM +0100, Krzysztof Kozlowski wrote:
> On 10/01/2023 05:25, clayc@xxxxxxx wrote:
> > From: Clay Chang <clayc@xxxxxxx>
> > 
> > The GXP SROM control register can be used to configure LPC related
> > legacy I/O registers. Currently only the SROM RAM Offset Register
> > (vromoff) is exported.
> > 
> > The GXP SOCLIB is a common library used for creating the common
> > "soc" class in the kernel.
> > 
> > Signed-off-by: Clay Chang <clayc@xxxxxxx>
> > ---
> >  drivers/soc/Kconfig          |   1 +
> >  drivers/soc/Makefile         |   1 +
> >  drivers/soc/hpe/Kconfig      |  29 +++++++
> >  drivers/soc/hpe/Makefile     |   2 +
> >  drivers/soc/hpe/gxp-soclib.c |  17 +++++
> >  drivers/soc/hpe/gxp-soclib.h |   9 +++
> >  drivers/soc/hpe/gxp-srom.c   | 141 +++++++++++++++++++++++++++++++++++
> >  7 files changed, 200 insertions(+)
> >  create mode 100644 drivers/soc/hpe/Kconfig
> >  create mode 100644 drivers/soc/hpe/Makefile
> >  create mode 100644 drivers/soc/hpe/gxp-soclib.c
> >  create mode 100644 drivers/soc/hpe/gxp-soclib.h
> >  create mode 100644 drivers/soc/hpe/gxp-srom.c
> > 
> > diff --git a/drivers/soc/Kconfig b/drivers/soc/Kconfig
> > index 5dbb09f843f7..faff0f036b61 100644
> > --- a/drivers/soc/Kconfig
> > +++ b/drivers/soc/Kconfig
> > @@ -10,6 +10,7 @@ source "drivers/soc/bcm/Kconfig"
> >  source "drivers/soc/canaan/Kconfig"
> >  source "drivers/soc/fsl/Kconfig"
> >  source "drivers/soc/fujitsu/Kconfig"
> > +source "drivers/soc/hpe/Kconfig"
> >  source "drivers/soc/imx/Kconfig"
> >  source "drivers/soc/ixp4xx/Kconfig"
> >  source "drivers/soc/litex/Kconfig"
> > diff --git a/drivers/soc/Makefile b/drivers/soc/Makefile
> > index fff513bd522d..d257b9d654b3 100644
> > --- a/drivers/soc/Makefile
> > +++ b/drivers/soc/Makefile
> > @@ -14,6 +14,7 @@ obj-$(CONFIG_MACH_DOVE)		+= dove/
> >  obj-y				+= fsl/
> >  obj-y				+= fujitsu/
> >  obj-$(CONFIG_ARCH_GEMINI)	+= gemini/
> > +obj-$(CONFIG_ARCH_HPE)		+= hpe/
> >  obj-y				+= imx/
> >  obj-y				+= ixp4xx/
> >  obj-$(CONFIG_SOC_XWAY)		+= lantiq/
> > diff --git a/drivers/soc/hpe/Kconfig b/drivers/soc/hpe/Kconfig
> > new file mode 100644
> > index 000000000000..88f5d46b06b6
> > --- /dev/null
> > +++ b/drivers/soc/hpe/Kconfig
> > @@ -0,0 +1,29 @@
> > +#
> > +# HPE GXP SoC drivers
> > +#
> > +menu "HPE GXP SoC drivers"
> > +	depends on ARCH_HPE || COMPILE_TEST
> > +
> > +config HPE_GXP_SOCLIB
> > +	bool "GXP Common SoC Library"
> > +	default y
> > +	depends on ARCH_HPE_GXP || COMPILE_TEST
> > +	help
> > +	  This is for the common library for all HPE SoC drivers. It
> > +	  creates the root soc class (/sys/class/soc) for all GXP SoC
> > +	  drivers. It must be yes if any one of the GXP SoC drivers were
> > +	  added, so the config of all GXP SoC drivers must select this.
> 
> Don't open-code Kconfig dependencies in free-form text. IOW, drop last
> sentence.

Understood, I will correct this part in the next revision.

> 
> > +
> > +
> > +config HPE_GXP_SROM
> > +	bool "GXP SROM Configuration Driver"
> > +	default y
> > +	depends on ARCH_HPE_GXP || COMPILE_TEST
> > +	select HPE_GXP_SOCLIB
> > +	help
> > +	  Say yes here to add support for SROM Configuration. The GXP SROM
> > +	  control register can be used to configure LPC related legacy I/O
> > +	  registers. Currently only the SROM RAM Offset Register (vromoff)
> > +	  is exported.
> > +
> > +endmenu
> > diff --git a/drivers/soc/hpe/Makefile b/drivers/soc/hpe/Makefile
> > new file mode 100644
> > index 000000000000..78de24ecb606
> > --- /dev/null
> > +++ b/drivers/soc/hpe/Makefile
> > @@ -0,0 +1,2 @@
> > +obj-$(CONFIG_HPE_GXP_SOCLIB) += gxp-soclib.o
> > +obj-$(CONFIG_HPE_GXP_SROM) += gxp-srom.o
> > diff --git a/drivers/soc/hpe/gxp-soclib.c b/drivers/soc/hpe/gxp-soclib.c
> > new file mode 100644
> > index 000000000000..11b0afe09070
> > --- /dev/null
> > +++ b/drivers/soc/hpe/gxp-soclib.c
> > @@ -0,0 +1,17 @@
> > +// SPDX-License-Identifier: GPL-2.0-only
> > +/* Copyright (C) 2023 Hewlett Packard Enteprise Development Company, L.P. */
> > +
> > +#include <linux/device.h>
> > +#include <linux/module.h>
> > +
> > +struct class *soc_class;
> > +
> > +static int __init gxp_soclib_init(void)
> > +{
> > +	soc_class = class_create(THIS_MODULE, "soc");
> > +	if (IS_ERR(soc_class))
> > +		return PTR_ERR(soc_class);
> > +	return 0;
> > +}
> > +
> > +module_init(gxp_soclib_init);
> 
> I don't see a point of making it a shared object and a "kernel library".
> Module inits are not libraries. Drop entire file.

Sure, I will re-write this part in the next revision.

> 
> > diff --git a/drivers/soc/hpe/gxp-soclib.h b/drivers/soc/hpe/gxp-soclib.h
> > new file mode 100644
> > index 000000000000..eb0e72b67aee
> > --- /dev/null
> > +++ b/drivers/soc/hpe/gxp-soclib.h
> > @@ -0,0 +1,9 @@
> 
> 
> > +
> > +static int sysfs_register(struct device *parent,
> > +			  struct gxp_srom_drvdata *drvdata)
> > +{
> > +	struct device *dev;
> > +
> > +	dev = device_create_with_groups(soc_class, parent, 0,
> > +					drvdata, srom_groups, "srom");
> > +	if (IS_ERR(dev))
> > +		return PTR_ERR(dev);
> > +	drvdata->dev = dev;
> > +	return 0;
> > +}
> > +
> > +static int gxp_srom_probe(struct platform_device *pdev)
> > +{
> > +	struct gxp_srom_drvdata *drvdata;
> > +	struct resource *res;
> > +	int ret;
> > +
> > +	drvdata = devm_kzalloc(&pdev->dev, sizeof(struct gxp_srom_drvdata),
> 
> sizeof(*)

Ok, I will correct this in the next.

> 
> > +				GFP_KERNEL);
> > +	if (!drvdata)
> > +		return -ENOMEM;
> > +
> > +	drvdata->pdev = pdev;
> > +	platform_set_drvdata(pdev, drvdata);
> > +
> > +	res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
> > +	if (!res) {
> > +		dev_err(&pdev->dev, "no srom resource defined\n");
> > +		ret = -ENODEV;
> > +		goto out;
> > +	}
> > +
> > +	drvdata->base = devm_ioremap_resource(&pdev->dev, res);
> 
> Combine these two. There is a helper for this.
> 

Sure, I will combine these two with devm_platform_ioremap_resource.

> > +	if (IS_ERR(drvdata->base)) {
> > +		dev_err(&pdev->dev, "ioremap_resource error\n");
> > +		ret = PTR_ERR(drvdata->base);
> 
> return dev_err_probe().
> > +		goto ioremap_err;
> > +	}
> > +
> > +	mutex_init(&drvdata->mutex);
> > +
> > +	ret = sysfs_register(&pdev->dev, drvdata);
> 
> Missing ABI documentation.

Ok, I will provide with the proper ABI documentation in the next revision.

> 
> > +	if (ret != 0) {
> > +		dev_err(&pdev->dev, "sysfs error\n");
> 
> return dev_err_probe

Ok, will correct this in the next.

> 
> > +		goto sysfs_err;
> > +	}
> > +
> > +	dev_info(&pdev->dev, "initialized\n");
> 
> Drop silly probe successes.

Sure, will drop this in the next revision.

> 
> > +	return 0;
> > +
> > +sysfs_err:
> > +ioremap_err:
> > +	platform_set_drvdata(pdev, NULL);
> > +out:
> > +	if (drvdata)
> > +		devm_kfree(&pdev->dev, (void *)drvdata);
> 
> 1. Why? if it is here, it must be in remove() callback. If it is invalid
> (because it is really not correct) for remove(), it cannot be here, right?
> 2. Why cast?

Ok, sorry for my mis-understanding about the error handling path in
device probing. I will re-write this part in the next revision.

> 
> > +	return ret;
> > +}
> > +
> > +static const struct of_device_id gxp_srom_of_match[] = {
> > +	{ .compatible = "hpe,gxp-srom" },
> > +	{},
> > +};
> > +MODULE_DEVICE_TABLE(of, gxp_srom_of_match);
> > +
> > +static struct platform_driver gxp_srom_driver = {
> > +	.probe = gxp_srom_probe,
> > +	.driver = {
> > +		.name = "gxp-srom",
> > +		.of_match_table = of_match_ptr(gxp_srom_of_match),
> 
> That will cause a warning. Drop of_match_ptr.

Got it!

> 
> > +	},
> > +};
> > +module_platform_driver(gxp_srom_driver);
> > +
> > +MODULE_AUTHOR("Clay Chang <clayc@xxxxxxx>");
> > +MODULE_DESCRIPTION("HPE GXP SROM Configuration Driver");
> > +MODULE_LICENSE("GPL");
> 
> Best regards,
> Krzysztof
> 

Thanks,
Clay