From mboxrd@z Thu Jan 1 00:00:00 1970 Return-path: Date: Tue, 22 Dec 2015 10:33:37 -0500 From: Damien Riegel To: Guenter Roeck Cc: linux-watchdog@vger.kernel.org, Wim Van Sebroeck , Pratyush Anand , Hans de Goede , linux-kernel@vger.kernel.org Subject: Re: [PATCH 1/5] watchdog: Create watchdog device in watchdog_dev.c Message-ID: <20151222153337.GB6164@localhost> References: <1450645503-16661-1-git-send-email-linux@roeck-us.net> <1450645503-16661-2-git-send-email-linux@roeck-us.net> <20151221173123.GD12696@localhost> <56788B27.20706@roeck-us.net> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <56788B27.20706@roeck-us.net> List-ID: On Mon, Dec 21, 2015 at 03:28:39PM -0800, Guenter Roeck wrote: > On 12/21/2015 09:31 AM, Damien Riegel wrote: > >On Sun, Dec 20, 2015 at 01:04:59PM -0800, Guenter Roeck wrote: > >>The watchdog character device is currently created in watchdog_dev.c, > >>and the watchdog device in watchdog_core.c. This results in > >>cross-dependencies, since device creation needs to know the watchdog > >>character device number as well as the watchdog class, both of which > >>reside in watchdog_dev.c. > >> > >>Create the watchdog device in watchdog_dev.c to simplify the code. > >> > >>Inspired by earlier patch set from Damien Riegel. > > > >Hi Guenter, > > > >The main purpose of my patch was to inverse the device creation and the > >cdev registration to avoid a racy situation, bu you have dropped that in > >this version. Is there a reason for that? > > > Every other driver I looked at does it in the same order (cdev first, device > second). I don't really know if doing it differently has any undesired > side effect, so I wanted to play safe. > > It would help a lot if someone listening to this exchange can confirm > that it is ok to create the device first, followed by the character device. The issue is that some drivers use watchdog_device->dev in their watchdog_ops functions. With a quick grep, I could spot 3 examples: - bcm2835_wdt_stop in bcm2835_wdt.c - gpio_wdt_hwping in gpio_wdt.c - a21_wdt_set_timeout in mena21_wdt.c Maybe we should simply fix these drivers and keep watchdog_device->dev for core internal usage? Thanks, Damien