[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CAN4iqtTcdZ8tqwPx=WeG6j6k17OHF3rj_YcXUgJAwseoTUKDeQ@mail.gmail.com>
Date: Thu, 12 Dec 2024 20:11:37 -0500
From: Mingwei Zheng <zmw12306@...il.com>
To: Antonio Borneo <antonio.borneo@...s.st.com>
Cc: marex@...x.de, peng.fan@....com, make24@...as.ac.cn,
linus.walleij@...aro.org, linux-kernel@...r.kernel.org,
linux-gpio@...r.kernel.org, Jiasheng Jiang <jiashengjiangcool@...il.com>,
fabien.dessenne@...s.st.com, mcoquelin.stm32@...il.com,
linux-stm32@...md-mailman.stormreply.com,
linux-arm-kernel@...ts.infradead.org
Subject: Re: [Linux-stm32] [PATCH v3] pinctrl: stm32: Add check for clk_enable()
Hi Antonio,
On Sat, Dec 7, 2024 at 9:25 AM Antonio Borneo
<antonio.borneo@...s.st.com> wrote:
>
> On Fri, 2024-12-06 at 16:43 -0500, Mingwei Zheng wrote:
> > Convert the driver to clk_bulk*() API.
> > Add check for the return value of clk_bulk_enable() to catch
> > the potential error.
> >
> > Fixes: 05d8af449d93 ("pinctrl: stm32: Keep pinctrl block clock enabled when LEVEL IRQ requested")
> > Signed-off-by: Mingwei Zheng <zmw12306@...il.com>
> > Signed-off-by: Jiasheng Jiang <jiashengjiangcool@...il.com>
> > ---
> > Changelog:
> >
> > v2 -> v3:
> >
> > 1. Convert clk_disable_unprepare to clk_bulk_disable
> > and clk_bulk_unprepare.
> >
> > v1 -> v2:
> >
> > 1. Move int ret declaration into if block.
> > ---
> > drivers/pinctrl/stm32/pinctrl-stm32.c | 28 +++++++++++++--------------
> > 1 file changed, 13 insertions(+), 15 deletions(-)
> >
> > diff --git a/drivers/pinctrl/stm32/pinctrl-stm32.c b/drivers/pinctrl/stm32/pinctrl-stm32.c
> > index 5b7fa77c1184..0ef912e82736 100644
> > --- a/drivers/pinctrl/stm32/pinctrl-stm32.c
> > +++ b/drivers/pinctrl/stm32/pinctrl-stm32.c
> > @@ -86,7 +86,6 @@ struct stm32_pinctrl_group {
> >
> > struct stm32_gpio_bank {
> > void __iomem *base;
> > - struct clk *clk;
> > struct reset_control *rstc;
> > spinlock_t lock;
> > struct gpio_chip gpio_chip;
> > @@ -108,6 +107,7 @@ struct stm32_pinctrl {
> > unsigned ngroups;
> > const char **grp_names;
> > struct stm32_gpio_bank *banks;
>
> Thanks for your patch!
> While it's growing wider than your initial scope, it looks as a nice improvement.
>
> > + struct clk_bulk_data *clks;
>
> Maybe I've missed it but, where does this array get allocated?
> There exist clk_bulk_get_all() and devm_clk_bulk_get_all() that do the allocation for you, but they require all the clocks to be listed in the same "clocks" property, while this driver has a clock in
> each DT subnode!
> Some suggestion below.
>
> > unsigned nbanks;
> > const struct stm32_pinctrl_match_data *match_data;
> > struct irq_domain *domain;
> > @@ -1308,7 +1308,7 @@ static int stm32_gpiolib_register_bank(struct stm32_pinctrl *pctl, struct fwnode
> > if (IS_ERR(bank->base))
> > return PTR_ERR(bank->base);
> >
> > - err = clk_prepare_enable(bank->clk);
> > + err = clk_prepare_enable(pctl->clks[pctl->nbanks].clk);
> > if (err) {
> > dev_err(dev, "failed to prepare_enable clk (%d)\n", err);
> > return err;
> > @@ -1397,7 +1397,7 @@ static int stm32_gpiolib_register_bank(struct stm32_pinctrl *pctl, struct fwnode
> > return 0;
> >
> > err_clk:
> > - clk_disable_unprepare(bank->clk);
> > + clk_disable_unprepare(pctl->clks[pctl->nbanks].clk);
> > return err;
> > }
> >
> > @@ -1631,11 +1631,10 @@ int stm32_pctl_probe(struct platform_device *pdev)
> > fwnode_handle_put(child);
> > return -EPROBE_DEFER;
> > }
> > -
> > - bank->clk = of_clk_get_by_name(np, NULL);
> > - if (IS_ERR(bank->clk)) {
> > + pctl->clks[i].clk = of_clk_get_by_name(np, NULL);
>
> This is the first time the array is used, during stm32_pctl_probe(), so the array should be already allocated before this line.
> The number of banks pctl->nbanks is computed few lines below, so it should be anticipated, or being replaced by a simpler
> pctl->nbanks = gpiochip_node_count(dev);
>
> then allocation could be done with
> pctl->clks = devm_kmalloc_array(dev, pctl->nbanks, sizeof(*pctl->clks), GFP_KERNEL);
>
> Also, you never assign `pctl->clks[i].id` which is then used by the error messages in, e.g., clk_bulk_enable().
> We don't mandate the property "clock-names" in the DT subnodes, so no way to get a reasonable value from the DT.
> I presume it's fine to set it to NULL... to be tested.
>
> Regards,
> Antonio
>
>
Thanks for your suggestion. We used devm_kcalloc to allocate clk. We
set `pctl->clks[i].id` to "".
Best,
Mingwei
> > + if (IS_ERR(pctl->clks[i].clk)) {
> > fwnode_handle_put(child);
> > - return dev_err_probe(dev, PTR_ERR(bank->clk),
> > + return dev_err_probe(dev, PTR_ERR(pctl->clks[i].clk),
> > "failed to get clk\n");
> > }
> > i++;
> > @@ -1646,8 +1645,8 @@ int stm32_pctl_probe(struct platform_device *pdev)
> > if (ret) {
> > fwnode_handle_put(child);
> >
> > - for (i = 0; i < pctl->nbanks; i++)
> > - clk_disable_unprepare(pctl->banks[i].clk);
> > + clk_bulk_disable(pctl->nbanks, pctl->clks);
> > + clk_bulk_unprepare(pctl->nbanks, pctl->clks);
> >
> > return ret;
> > }
> > @@ -1726,10 +1725,8 @@ static int __maybe_unused stm32_pinctrl_restore_gpio_regs(
> > int __maybe_unused stm32_pinctrl_suspend(struct device *dev)
> > {
> > struct stm32_pinctrl *pctl = dev_get_drvdata(dev);
> > - int i;
> >
> > - for (i = 0; i < pctl->nbanks; i++)
> > - clk_disable(pctl->banks[i].clk);
> > + clk_bulk_disable(pctl->nbanks, pctl->clks);
> >
> > return 0;
> > }
> > @@ -1738,10 +1735,11 @@ int __maybe_unused stm32_pinctrl_resume(struct device *dev)
> > {
> > struct stm32_pinctrl *pctl = dev_get_drvdata(dev);
> > struct stm32_pinctrl_group *g = pctl->groups;
> > - int i;
> > + int i, ret;
> >
> > - for (i = 0; i < pctl->nbanks; i++)
> > - clk_enable(pctl->banks[i].clk);
> > + ret = clk_bulk_enable(pctl->nbanks, pctl->clks);
> > + if (ret)
> > + return ret;
> >
> > for (i = 0; i < pctl->ngroups; i++, g++)
> > stm32_pinctrl_restore_gpio_regs(pctl, g->pin);
Powered by blists - more mailing lists