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;
MIME-Version: 1.0
References: <20220415165505.30383-1-schspa@gmail.com>
In-Reply-To: <20220415165505.30383-1-schspa@gmail.com>
From:   Andy Shevchenko <andy.shevchenko@gmail.com>
Date:   Sat, 16 Apr 2022 11:30:19 +0300
Message-ID: <CAHp75Veo30c0BWb4Fykgvd-czSnEXsyA8wyMTeKQdS49=e5-nw@mail.gmail.com>
Subject: Re: [PATCH] gpio: use raw spinlock for gpio chip shadowed data
To:     Schspa Shi <schspa@gmail.com>
Cc:     Linus Walleij <linus.walleij@linaro.org>,
        Bartosz Golaszewski <brgl@bgdev.pl>, opendmb@gmail.com,
        Florian Fainelli <f.fainelli@gmail.com>,
        Hoan Tran <hoan@os.amperecomputing.com>,
        Serge Semin <fancer.lancer@gmail.com>,
        "open list:GPIO SUBSYSTEM" <linux-gpio@vger.kernel.org>,
        Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
        linux-arm Mailing List <linux-arm-kernel@lists.infradead.org>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Sat, Apr 16, 2022 at 3:45 AM Schspa Shi <schspa@gmail.com> wrote:
>
> In case of PREEMPT_RT, there is a raw_spinlock -> spinlock dependency
> as the lockdep report shows.
>
> __irq_set_handler
>   irq_get_desc_buslock
>     __irq_get_desc_lock
>       raw_spin_lock_irqsave(&desc->lock, *flags);  // raw spinlock get here
>   __irq_do_set_handler
>     mask_ack_irq
>       dwapb_irq_ack
>         spin_lock_irqsave(&gc->bgpio_lock, flags); // sleep able spinlock
>   irq_put_desc_busunlock
>
> Replace with a raw lock to avoid BUGs. This lock is only used to access
> registers, and It's safe to replace with the raw lock without bad
> influence.

> [   15.090359][    T1] =============================
> [   15.090365][    T1] [ BUG: Invalid wait context ]
> [   15.090373][    T1] 5.10.59-rt52-00983-g186a6841c682-dirty #3 Not tainted
> [   15.090386][    T1] -----------------------------

Please, reduce this very noise traceback to the point, usually it
takes ~3-5 lines only.

...

>  drivers/gpio/gpio-amdpt.c                 | 10 +++----
>  drivers/gpio/gpio-brcmstb.c               | 12 ++++----
>  drivers/gpio/gpio-cadence.c               | 12 ++++----
>  drivers/gpio/gpio-dwapb.c                 | 36 +++++++++++------------
>  drivers/gpio/gpio-grgpio.c                | 30 +++++++++----------
>  drivers/gpio/gpio-hlwd.c                  | 18 ++++++------
>  drivers/gpio/gpio-idt3243x.c              | 12 ++++----
>  drivers/gpio/gpio-ixp4xx.c                |  4 +--
>  drivers/gpio/gpio-loongson1.c             |  8 ++---
>  drivers/gpio/gpio-menz127.c               |  8 ++---
>  drivers/gpio/gpio-mlxbf2.c                | 18 ++++++------
>  drivers/gpio/gpio-mmio.c                  | 22 +++++++-------
>  drivers/gpio/gpio-sifive.c                | 12 ++++----
>  drivers/gpio/gpio-tb10x.c                 |  4 +--
>  drivers/pinctrl/nuvoton/pinctrl-npcm7xx.c |  8 ++---
>  include/linux/gpio/driver.h               |  2 +-

Looking into the code I think we can't easily (without too much
additional churn) to split this to the series of patches. But I leave
this up to the maintainers to decide.

-- 
With Best Regards,
Andy Shevchenko