Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
MIME-Version: 1.0
References: <20180831212007.247186-1-dbasehore@chromium.org>
In-Reply-To: <20180831212007.247186-1-dbasehore@chromium.org>
From:   "dbasehore ." <dbasehore@chromium.org>
Date:   Fri, 19 Oct 2018 16:07:21 -0700
Message-ID: <CAGAzgsqayRfva3om-3juAhE+VDuY0hV6q-DuyiJ-QU=szPSaew@mail.gmail.com>
Subject: Re: [PATCH] clk: fix clk_calc_subtree compute duplications
To:     sboyd@kernel.org
Cc:     Michael Turquette <mturquette@baylibre.com>,
        linux-clk@vger.kernel.org,
        linux-kernel <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Fri, Aug 31, 2018 at 2:20 PM Derek Basehore <dbasehore@chromium.org> wrote:
>
> clk_calc_subtree was called at every step up the clk tree in
> clk_calc_new_rates. Since it recursively calls itself for its
> children, this means it would be called once on each clk for each
> step above the top clk is.
>
> This is fixed by adding a non-recursive function called at every
> step in clk_calc_new_rates that fills in new_rate, new_parent, etc.
> Since the clks not called directly for clk_calc_new_rates can only
> change their rate, we only set new_rate in clk_calc_subtree.
> clk_calc_subtree is also only called on the top clk after it's found
> via clk_calc_new_rates to remove the duplicate recursive calls.
>
> Signed-off-by: Derek Basehore <dbasehore@chromium.org>
> ---
>  drivers/clk/clk.c | 20 ++++++++++++--------
>  1 file changed, 12 insertions(+), 8 deletions(-)
>
> diff --git a/drivers/clk/clk.c b/drivers/clk/clk.c
> index d31055ae6ec6..52032fb1a8a2 100644
> --- a/drivers/clk/clk.c
> +++ b/drivers/clk/clk.c
> @@ -1609,11 +1609,18 @@ static int __clk_speculate_rates(struct clk_core *core,
>         return ret;
>  }
>
> -static void clk_calc_subtree(struct clk_core *core, unsigned long new_rate,
> -                            struct clk_core *new_parent, u8 p_index)
> +static void clk_calc_subtree(struct clk_core *core, unsigned long new_rate)
>  {
>         struct clk_core *child;
>
> +       core->new_rate = new_rate;
> +       hlist_for_each_entry(child, &core->children, child_node)
> +               clk_calc_subtree(child, clk_recalc(child, new_rate));
> +}
> +
> +static void clk_set_change(struct clk_core *core, unsigned long new_rate,
> +                          struct clk_core *new_parent, u8 p_index)
> +{
>         core->new_rate = new_rate;
>         core->new_parent = new_parent;
>         core->new_parent_index = p_index;
> @@ -1621,11 +1628,6 @@ static void clk_calc_subtree(struct clk_core *core, unsigned long new_rate,
>         core->new_child = NULL;
>         if (new_parent && new_parent != core->parent)
>                 new_parent->new_child = core;
> -
> -       hlist_for_each_entry(child, &core->children, child_node) {
> -               child->new_rate = clk_recalc(child, new_rate);
> -               clk_calc_subtree(child, child->new_rate, NULL, 0);
> -       }
>  }
>
>  /*
> @@ -1709,7 +1711,7 @@ static struct clk_core *clk_calc_new_rates(struct clk_core *core,
>                 top = clk_calc_new_rates(parent, best_parent_rate);
>
>  out:
> -       clk_calc_subtree(core, new_rate, parent, p_index);
> +       clk_set_change(core, new_rate, parent, p_index);
>
>         return top;
>  }
> @@ -1910,6 +1912,8 @@ static int clk_core_set_rate_nolock(struct clk_core *core,
>         if (ret)
>                 return ret;
>
> +       clk_calc_subtree(top, top->new_rate);
> +

Oops. This is wrong.  Calling here will overwrite new rate settings
such as when determine_rate/round_rate return something different than
clk_recalc(core, parent_rate). I'm working on a larger patch series
where this will be fixed.

>         /* notify that we are about to change rates */
>         fail_clk = clk_propagate_rate_change(top, PRE_RATE_CHANGE);
>         if (fail_clk) {
> --
> 2.19.0.rc1.350.ge57e33dbd1-goog
>