linux-kernel - Re: [PATCH v1 01/43] gpio: ep93xx: split device in multiple

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: <a8b44c9c5dbd938eb2d826aa0d25af38ad613a17.camel@maquefel.me>
Date:   Thu, 15 Jun 2023 19:56:10 +0300
From:   Nikita Shubin <nikita.shubin@...uefel.me>
To:     andy.shevchenko@...il.com
Cc:     Alexander Sverdlin <alexander.sverdlin@...il.com>,
        Arnd Bergmann <arnd@...db.de>,
        Linus Walleij <linus.walleij@...aro.org>,
        Hartley Sweeten <hsweeten@...ionengravers.com>,
        Russell King <linux@...linux.org.uk>,
        Lukasz Majewski <lukma@...x.de>,
        Bartosz Golaszewski <brgl@...ev.pl>,
        Michael Peters <mpeters@...eddedts.com>,
        Kris Bahnsen <kris@...eddedts.com>,
        linux-arm-kernel@...ts.infradead.org, linux-kernel@...r.kernel.org,
        linux-gpio@...r.kernel.org
Subject: Re: [PATCH v1 01/43] gpio: ep93xx: split device in multiple

Hello Andy !

On Fri, 2023-06-02 at 04:50 +0300, andy.shevchenko@...il.com wrote:
> Thu, Jun 01, 2023 at 08:33:52AM +0300, Nikita Shubin kirjoitti:
> > This prepares ep93xx SOC gpio to convert into device tree driver:
> > - dropped banks and legacy defines
> > - split AB IRQ and make it shared
> > 
> > We are relying on IRQ number information A, B ports have single
> > shared
> > IRQ, while F port have dedicated IRQ for each line.
> > 
> > Also we had to split single ep93xx platform_device into multiple,
> > one
> > for each port, without this we can't do a full working transition
> > from
> > legacy platform code into device tree capable. All GPIO_LOOKUP were
> > change to match new chip namings.
> 
> First of all, check if you added In-Reply-to email header to the
> previous
> thread, at least `b4` downloaded 188 messages in this one so far.
> Second,
> the previous was kinda v0, while we usually assume that non-versioned
> series
> is v1. This is a bit ambiguous.
> 
> ...
> 
> > +               GPIO_LOOKUP_IDX("gpio-ep93xx.4", 1,     NULL, 1,
> > GPIO_ACTIVE_HIGH),
> 
> TAB used instead of space.
> 
> ...
> 
> >  struct device __init *ep93xx_init_devices(void)
> >  {
> >         struct device *parent;
> > +       int i;
> 
> It's unsigned, right?
> 
> > +       for (i = 0; i < ARRAY_SIZE(ep93xx_gpio_device); i++)
> > +               platform_device_register(ep93xx_gpio_device[i]);
> 
> ...
> 
> >         writeb(eic->int_debounce,
> > -              epg->base + eic->irq_offset +
> > EP93XX_INT_DEBOUNCE_OFFSET);
> > +              eic->base + EP93XX_INT_DEBOUNCE_OFFSET);
> 
> Now this can be a single line. Also some other cases may be
> optimized.
> 
> ...
> 
> > +       void __iomem *intr =
> > devm_platform_ioremap_resource_byname(pdev, "intr");
> 
> It's less error prone if the assignment is split from definition and
> moved
> closer to its (first) user...
> 
> > +
> 
> ...here.
> 
> > +       if (IS_ERR(intr))
> > +               return PTR_ERR(intr);
> 
> ...
> 
> > +       egc->eic = devm_kcalloc(dev, 1,
> > +                               sizeof(*egc->eic),
> > +                               GFP_KERNEL);
> 
> Why kcalloc(1), is this a part that will be (slightly) modified in
> the next
> patches in the series?
> 
> > +       if (!egc->eic)
> > +               return -ENOMEM;
> >  
> ...
> 
> > +               irq = platform_get_irq(pdev, 0);
> 
> No return value check?
> 
> > +               ret = devm_request_irq(dev, irq,
> > +                               ep93xx_ab_irq_handler,
> > +                               IRQF_SHARED, gc->label, gc);
> > +               if (ret) {
> > +                       dev_err(dev, "error requesting IRQ : %d\n",
> > irq);
> > +                       return ret;
> 
> If it's soslely part of the ->probe() flow, you may use
> dev_err_probe().
> 
> > +               }
> >  
> > +               girq->parents[0] = irq;
> 
> ...
> 
> >                 for (i = 0; i < girq->num_parents; i++) {
> > +                       irq = platform_get_irq(pdev, i);
> > +                       if (irq <= 0)
> 
> == 0 is never happen case. Why?
> 
> > +                               continue;
> > +
> > +                       girq->parents[i] = irq;
> >                 }
> 
> > +       ret = bgpio_init(gc, &pdev->dev, 1, data, NULL, NULL, dir,
> > NULL, 0);
> > +       if (ret) {
> > +               dev_err(&pdev->dev, "unable to init generic
> > GPIO\n");
> > +               return ret;
> 
>                 return dev_err_probe(...);
> 
> >         }
> 
> ...
> 
> > +       if (platform_irq_count(pdev) > 0) {
> 
> Do you need this check?

Only A/B/F ports have irq's so we don't bother setting up for other
ports. 

> 
> > +               dev_dbg(&pdev->dev, "setting up irqs for %s\n",
> > dev_name(&pdev->dev));
> > +               ret = ep93xx_setup_irqs(pdev, egc);
> > +               if (ret)
> > +                       dev_err(&pdev->dev, "setup irqs failed for
> > %s\n", dev_name(&pdev->dev));
> 
> If it's an error, why continuing?

Well - it's not fatal, we can still use gpios even without irq's but we
should warn.

All other comments acknowledged and fixed, thank you.

> 
> > +       }
>