From mboxrd@z Thu Jan 1 00:00:00 1970 From: Linus Walleij Subject: Re: [PATCH] gpiolib: handle probe deferrals better Date: Wed, 6 Apr 2016 15:39:13 +0200 Message-ID: References: <1459511048-24084-1-git-send-email-linus.walleij@linaro.org> <56FE7785.1010507@ti.com> <570294A4.2070801@ti.com> Mime-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Return-path: In-Reply-To: <570294A4.2070801@ti.com> Sender: linux-gpio-owner@vger.kernel.org To: Grygorii Strashko Cc: "linux-gpio@vger.kernel.org" , Alexandre Courbot , Alexander Stein , Linux Input , Tomeu Vizoso , Guenter Roeck , Bjorn Andersson List-Id: linux-input@vger.kernel.org On Mon, Apr 4, 2016 at 6:21 PM, Grygorii Strashko wrote: > Below is RFC patch to prove above consent. I've had offlist > debugging session with Alexander and He mentioned that this change > fixes boot issue for him. Thanks for looking into this. > Of course, there are some question to discuss: > 1) [+] It should sync initialization of GPIOchip and GPIOirqchip > 2) [+] This approach requires changes in gpiolib/gpio drivers only, from other side > It will require to add fixes all over the Kernel if gpiod_to_irq() will > start returning -EPROBE_DEFER. Yes, so it will need to be cross-coordinated with IRQ maintainers Marc and TGLX. > 3) [-] has_irq might need to be initialized by non-DT drivers Yes. Every driver in the kernel need to be audited. > 4) [-] irq_ready might need to be set manually by drivers which do not use GPIO irq > helpers (see change in gpio-mpc8xxx.c) Yes. That too. Every driver in the kernel need to be audited. > 4) irq_ready access synchronization on SMP? atomic? Uhhh.... I don't even understand the question. > job done with commit e6918cd 'gpiolib: handle probe deferrals better' > reverted. I have taken that out of my tree as well. My naive approach doesn't work. Yours, Linus Walleij