Re: [PATCH v1 1/3] PCI: dwc: refactor common code

[Bjorn Helgaas <helgaas@kernel.org>]

On Thu, Feb 29, 2024 at 09:11:34PM -0800, Shashank Babu Chinta Venkata wrote:
> Refactor common code from RC(Root Complex) and EP(End Point)
> drivers and move them to a common repository. This acts as placeholder
> for common source code for both drivers avoiding duplication.
> 
> Signed-off-by: Shashank Babu Chinta Venkata <quic_schintav@quicinc.com>
> ---
>  drivers/pci/controller/dwc/Kconfig         |  5 ++
>  drivers/pci/controller/dwc/Makefile        |  1 +
>  drivers/pci/controller/dwc/pcie-qcom-cmn.c | 85 ++++++++++++++++++++++
>  drivers/pci/controller/dwc/pcie-qcom-cmn.h | 30 ++++++++
>  drivers/pci/controller/dwc/pcie-qcom-ep.c  | 39 +---------
>  drivers/pci/controller/dwc/pcie-qcom.c     | 67 ++---------------
>  6 files changed, 133 insertions(+), 94 deletions(-)
>  create mode 100644 drivers/pci/controller/dwc/pcie-qcom-cmn.c
>  create mode 100644 drivers/pci/controller/dwc/pcie-qcom-cmn.h

Hmm.  I'm a little ambivalent about adding two new files.  Overall I
think I prefer the drivers that include both RC and EP mode in a
single source file because one file is easier to browse than four and
more things can be static.

A single file would also reduce quite a bit more duplication between
pcie-qcom.c and pcie-qcom-ep.c, e.g., register names and fields with
needlessly different names:

  #define AUX_PWR_DET                     BIT(4)  # pcie-qcom.c
  #define PARF_SYS_CTRL_AUX_PWR_DET       BIT(4)  # pcie-qcom-ep.c

I do see PCIE_QCOM is bool and PCIE_QCOM_EP is tristate, so that and
other considerations might make a single source file impractical.

> +++ b/drivers/pci/controller/dwc/Makefile
> @@ -27,6 +27,7 @@ obj-$(CONFIG_PCIE_UNIPHIER) += pcie-uniphier.o
>  obj-$(CONFIG_PCIE_UNIPHIER_EP) += pcie-uniphier-ep.o
>  obj-$(CONFIG_PCIE_VISCONTI_HOST) += pcie-visconti.o
>  obj-$(CONFIG_PCIE_RCAR_GEN4) += pcie-rcar-gen4.o
> +obj-$(CONFIG_PCIE_QCOM_CMN) += pcie-qcom-cmn.o

If we have to have pcie-qcom-cmn.o, at least move this next to the
existing lines:

  obj-$(CONFIG_PCIE_QCOM) += pcie-qcom.o
  obj-$(CONFIG_PCIE_QCOM_EP) += pcie-qcom-ep.o

> +++ b/drivers/pci/controller/dwc/pcie-qcom-cmn.c
> @@ -0,0 +1,85 @@
> +// SPDX-License-Identifier: GPL-2.0
> +/*
> + * Copyright (c) 2014-2015, 2020 The Linux Foundation. All rights reserved.
> + * Copyright 2015, 2021 Linaro Limited.
> + * Copyright (c) 2024 Qualcomm Innovation Center, Inc. All rights reserved.
> + *

Spurious blank line.

> +int qcom_pcie_cmn_icc_get_resource(struct dw_pcie *pci, struct icc_path *icc_mem)

I don't see the value of adding "cmn" in the middle of the names.

> +{
> +	int ret = 0;
> +
> +	if (IS_ERR(pci))
> +		return PTR_ERR(pci);
> +
> +	icc_mem = devm_of_icc_get(pci->dev, "pcie-mem");
> +	if (IS_ERR(icc_mem))
> +		return PTR_ERR(icc_mem);
> +
> +	return ret;

No need for the "ret" variable since it's never assigned.  "return 0"
here would be easier to read.

> +int qcom_pcie_cmn_icc_init(struct dw_pcie *pci, struct icc_path *icc_mem)
> +{
> +	int ret = 0;

Unnecessary initialization.

> +++ b/drivers/pci/controller/dwc/pcie-qcom-cmn.h
> @@ -0,0 +1,30 @@
> +/* SPDX-License-Identifier: GPL-2.0 */
> +/*
> + * Copyright (c) 2014-2015, 2020 The Linux Foundation. All rights reserved.
> + * Copyright 2015, 2021 Linaro Limited.
> + * Copyright (c) 2024 Qualcomm Innovation Center, Inc. All rights reserved.
> + */
> +
> +#include <linux/pci.h>
> +#include "../../pci.h"
> +#include "pcie-designware.h"
> +
> +#ifdef CONFIG_PCIE_QCOM_CMN

Why the #ifdef wrapper?  And why do we need the stubs when
CONFIG_PCIE_QCOM_CMN isn't defined?

> +#else
> +static inline int qcom_pcie_cmn_icc_get_resource(struct dw_pcie *pci, struct icc_path *icc_mem)
> +{
> +	return 0;
> +}