Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Thu, 15 Sep 2022 13:33:33 +0200
From:   Horatiu Vultur <horatiu.vultur@microchip.com>
To:     Andy Shevchenko <andy.shevchenko@gmail.com>
CC:     "open list:GPIO SUBSYSTEM" <linux-gpio@vger.kernel.org>,
        "Linux Kernel Mailing List" <linux-kernel@vger.kernel.org>,
        Linus Walleij <linus.walleij@linaro.org>,
        Microchip Linux Driver Support <UNGLinuxDriver@microchip.com>
Subject: Re: [PATCH v3] pinctrl: ocelot: Fix interrupt controller
Message-ID: <20220915113333.uwgazi6hoeskpeoi@soft-dev3-1.localhost>
References: <20220909145942.844102-1-horatiu.vultur@microchip.com>
 <CAHp75VfNjQTVjsjWNJ1oOGuovxt_=ZEZrEfTvj=-Ym9R_ZzPoQ@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset="utf-8"
Content-Disposition: inline
In-Reply-To: <CAHp75VfNjQTVjsjWNJ1oOGuovxt_=ZEZrEfTvj=-Ym9R_ZzPoQ@mail.gmail.com>
Precedence: bulk

The 09/09/2022 18:09, Andy Shevchenko wrote:
> 
> On Fri, Sep 9, 2022 at 5:55 PM Horatiu Vultur
> <horatiu.vultur@microchip.com> wrote:
> 
> Thanks for an update, my comments below.

Thanks for all the help and sorry for late reply.

> 
> ...
> 
> > -       dev_set_drvdata(dev, info->map);
> > +       dev_set_drvdata(dev, info);
> 
> I would also change it to platform_set_drvdata() to keep symmetry with
> ->remove().

Yes, I will change this.

> 
> ...
> 
> > +static int ocelot_pinctrl_remove(struct platform_device *pdev)
> > +{
> > +       struct ocelot_pinctrl *info = platform_get_drvdata(pdev);
> 
> > +       destroy_workqueue(info->wq);
> 
> Is it a synchronous operation? Anyway, what does guarantee that after
> this no other task can schedule a new work due to unmasking an
> interrupt? I think you need to be sure your device is quiescent before
> killing that workqueue. Something like synchronize_irq() +
> disable_irq() or equivalent? (I don't know for sure, you need to
> investigate it yourself and find the best suitable way).

I have look at descriptions of the functions (synchronize_irq(),
disable_irq()) and I think is enough to use only disable_irq().
I also tried something but it didn't have the expected result so I would
need to look more into this. I tried to use disable_irq on returned irq
inside ocelot_gpiochip_register but I was still getting interrupts after
that.
Also I was thinking actually to use gpiochip_remove() here in
ocelot_pinctrl_remove() before calling destroy_workqueue(). But then I
might have problems inside ocelot_irq_work(). I need to check more this.

> 
> > +       return 0;
> > +}
> 
> --
> With Best Regards,
> Andy Shevchenko

-- 
/Horatiu