Re: [PATCH v2 3/5] phy: mediatek: add helpers to update bits of registers

[AngeloGioacchino Del Regno]

Il 30/12/21 03:06, Chunfeng Yun ha scritto:
> On Fri, 2021-12-24 at 11:10 +0100, AngeloGioacchino Del Regno wrote:
> > Il 18/12/21 09:28, Chunfeng Yun ha scritto:
> > > Add three helpers mtk_phy_clear/set/update_bits() for registers
> > > operation
> > >
> > > Signed-off-by: Chunfeng Yun <chunfeng.yun@xxxxxxxxxxxx>
> > > ---
> > > v2: new patch, add register access helpers,
> > >       Add updatel() macro suggested by Vinod, here add more ones
> > > instead.
> > > ---
> > >    drivers/phy/mediatek/phy-mtk-io.h | 38
> > > +++++++++++++++++++++++++++++++
> > >    1 file changed, 38 insertions(+)
> > >    create mode 100644 drivers/phy/mediatek/phy-mtk-io.h
> > >
> > > diff --git a/drivers/phy/mediatek/phy-mtk-io.h
> > > b/drivers/phy/mediatek/phy-mtk-io.h
> > > new file mode 100644
> > > index 000000000000..500fcdab165d
> > > --- /dev/null
> > > +++ b/drivers/phy/mediatek/phy-mtk-io.h
> > > @@ -0,0 +1,38 @@
> > > +/* SPDX-License-Identifier: GPL-2.0 */
> > > +/*
> > > + * Copyright (C) 2021 MediaTek Inc.
> > > + *
> > > + * Author: Chunfeng Yun <chunfeng.yun@xxxxxxxxxxxx>
> > > + */
> > > +
> > > +#ifndef __PHY_MTK_H__
> > > +#define __PHY_MTK_H__
> > > +
> > > +#include <linux/io.h>
> > > +
> > > +static inline void mtk_phy_clear_bits(void __iomem *reg, u32 bits)
> > > +{
> > > +	u32 tmp = readl(reg);
> > > +
> > > +	tmp &= ~bits;
> > > +	writel(tmp, reg);
> > > +}
> > > +
> > > +static inline void mtk_phy_set_bits(void __iomem *reg, u32 bits)
> > > +{
> > > +	u32 tmp = readl(reg);
> > > +
> > > +	tmp |= bits;
> > > +	writel(tmp, reg);
> > > +}
> > > +
> > > +static inline void mtk_phy_update_bits(void __iomem *reg, u32
> > > mask, u32 val)
> > > +{
> > > +	u32 tmp = readl(reg);
> > > +
> > > +	tmp &= ~mask;
> > > +	tmp |= val & mask;
> > > +	writel(tmp, reg);
> > > +}
> > > +
> > > +#endif
> >
> > These helpers are almost exactly duplicating what
> > regmap_update_bits() is doing.
> > I appreciate the effort to stop open-coding the same sequences over
> > and over by
> > adding such helper functions,
> I agree with you.
> > but I think that the proper way of doing what you
> > are proposing is not to add custom functions but rather reuse what
> > the Linux APIs
> > give you.
> I also like to use common APIs ASAP, but not found suitable ones.
> This may be a problem, I found that some similar custom helps already
> added under phy fold.
>
> > What about doing a conversion to use regmap on this driver?
> No, we don't use regmap here, these registers are monopolized by t-phy,
> it's not syscon.

Hello,

The regmap API allows this kind of usage, registers don't necessarily have
to be part of a syscon.

Regards,
- Angelo