linux-kernel - Re: [PATCH RESEND v7 4/4] clk: meson: add one based divider support for sclk divider

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <4da764c237b8f752af1dc33a011e2a4b73068f02.camel@baylibre.com>
Date:   Tue, 11 Dec 2018 18:16:53 +0100
From:   Jerome Brunet <jbrunet@...libre.com>
To:     Jianxin Pan <jianxin.pan@...ogic.com>,
        Neil Armstrong <narmstrong@...libre.com>
Cc:     Kevin Hilman <khilman@...libre.com>,
        Carlo Caione <carlo@...one.org>,
        Michael Turquette <mturquette@...libre.com>,
        Stephen Boyd <sboyd@...nel.org>, Rob Herring <robh@...nel.org>,
        Miquel Raynal <miquel.raynal@...tlin.com>,
        Boris Brezillon <boris.brezillon@...tlin.com>,
        Martin Blumenstingl <martin.blumenstingl@...glemail.com>,
        Yixun Lan <yixun.lan@...ogic.com>,
        Liang Yang <liang.yang@...ogic.com>,
        Jian Hu <jian.hu@...ogic.com>,
        Qiufang Dai <qiufang.dai@...ogic.com>,
        Hanjie Lin <hanjie.lin@...ogic.com>,
        Victor Wan <victor.wan@...ogic.com>, linux-clk@...r.kernel.org,
        linux-amlogic@...ts.infradead.org,
        linux-arm-kernel@...ts.infradead.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH RESEND v7 4/4] clk: meson: add one based divider support
 for sclk divider

On Tue, 2018-12-11 at 00:04 +0800, Jianxin Pan wrote:
> When CLK_DIVIDER_ONE_BASED flag is set, the sclk divider will be:
> one based divider (div = val), and zero value gates the clock
> 
> Signed-off-by: Jianxin Pan <jianxin.pan@...ogic.com>
> ---
>  drivers/clk/meson/clkc-audio.h |  1 +
>  drivers/clk/meson/sclk-div.c   | 28 ++++++++++++++++++----------
>  2 files changed, 19 insertions(+), 10 deletions(-)

Such a patch should be done earlier in the series, at least before using sclk
in your controller, otherwise thing will be broken in between

In general, I would prefer if you had added two helper function to deal with
the translation between register value and divider value.

Only these function should care about CLK_DIVIDER_ONE_BASED, the rest should
just call them.

This, we will be able to deal the with HI (duty cycle) part as well, which you
completly skiped.

I know your device does not have this, but still the code has to make sense.

> 
> diff --git a/drivers/clk/meson/clkc-audio.h b/drivers/clk/meson/clkc-audio.h
> index 0a7c157..9bd6ced 100644
> --- a/drivers/clk/meson/clkc-audio.h
> +++ b/drivers/clk/meson/clkc-audio.h
> @@ -20,6 +20,7 @@ struct meson_sclk_div_data {
>  	struct parm hi;
>  	unsigned int cached_div;
>  	struct clk_duty cached_duty;
> +	u8	flags;
>  };
>  
>  extern const struct clk_ops meson_clk_triphase_ops;
> diff --git a/drivers/clk/meson/sclk-div.c b/drivers/clk/meson/sclk-div.c
> index bc64019..d98707b 100644
> --- a/drivers/clk/meson/sclk-div.c
> +++ b/drivers/clk/meson/sclk-div.c
> @@ -24,22 +24,23 @@
>  	return (struct meson_sclk_div_data *)clk->data;
>  }
>  
> -static int sclk_div_maxval(struct meson_sclk_div_data *sclk)
> -{
> -	return (1 << sclk->div.width) - 1;
> -}
> -
>  static int sclk_div_maxdiv(struct meson_sclk_div_data *sclk)
>  {
> -	return sclk_div_maxval(sclk) + 1;
> +	if (sclk->flags & CLK_DIVIDER_ONE_BASED)
> +		return clk_div_mask(sclk->div.width);
> +	else
> +		return clk_div_mask(sclk->div.width) + 1;

seems over complicated.
why no call clk_div_mask just once, and add 1 if necessary ?

>  }
>  
>  static int sclk_div_getdiv(struct clk_hw *hw, unsigned long rate,
>  			   unsigned long prate, int maxdiv)
>  {
>  	int div = DIV_ROUND_CLOSEST_ULL((u64)prate, rate);
> +	struct clk_regmap *clk = to_clk_regmap(hw);
> +	struct meson_sclk_div_data *sclk = meson_sclk_div_data(clk);
> +	int mindiv = (sclk->flags & CLK_DIVIDER_ONE_BASED) ? 1 : 2;

This is why I want helpers, don't like this above

>  
> -	return clamp(div, 2, maxdiv);
> +	return clamp(div, mindiv, maxdiv);
>  }
>  
>  static int sclk_div_bestdiv(struct clk_hw *hw, unsigned long rate,
> @@ -47,7 +48,7 @@ static int sclk_div_bestdiv(struct clk_hw *hw, unsigned
> long rate,
>  			    struct meson_sclk_div_data *sclk)
>  {
>  	struct clk_hw *parent = clk_hw_get_parent(hw);
> -	int bestdiv = 0, i;
> +	int bestdiv = 0, i, mindiv;
>  	unsigned long maxdiv, now, parent_now;
>  	unsigned long best = 0, best_parent = 0;
>  
> @@ -64,8 +65,9 @@ static int sclk_div_bestdiv(struct clk_hw *hw, unsigned
> long rate,
>  	 * unsigned long in rate * i below
>  	 */
>  	maxdiv = min(ULONG_MAX / rate, maxdiv);
> +	mindiv = (sclk->flags & CLK_DIVIDER_ONE_BASED) ? 1 : 2;
>  
> -	for (i = 2; i <= maxdiv; i++) {
> +	for (i = mindiv; i <= maxdiv; i++) {
>  		/*
>  		 * It's the most ideal case if the requested rate can be
>  		 * divided from parent clock without needing to change
> @@ -153,10 +155,14 @@ static int sclk_div_get_duty_cycle(struct clk_hw *hw,
>  static void sclk_apply_divider(struct clk_regmap *clk,
>  			       struct meson_sclk_div_data *sclk)
>  {
> +	unsigned int div;
> +
>  	if (MESON_PARM_APPLICABLE(&sclk->hi))
>  		sclk_apply_ratio(clk, sclk);
>  
> -	meson_parm_write(clk->map, &sclk->div, sclk->cached_div - 1);
> +	div = (sclk->flags & CLK_DIVIDER_ONE_BASED) ?
> +		sclk->cached_div : (sclk->cached_div - 1);

helpers again.

> +	meson_parm_write(clk->map, &sclk->div, div);
>  }
>  
>  static int sclk_div_set_rate(struct clk_hw *hw, unsigned long rate,
> @@ -223,6 +229,8 @@ static void sclk_div_init(struct clk_hw *hw)
>  	/* if the divider is initially disabled, assume max */
>  	if (!val)
>  		sclk->cached_div = sclk_div_maxdiv(sclk);
> +	else if (sclk->flags & CLK_DIVIDER_ONE_BASED)
> +		sclk->cached_div = val;
>  	else
>  		sclk->cached_div = val + 1;

same ...

>