From mboxrd@z Thu Jan 1 00:00:00 1970 Return-path: Received: from bh-25.webhostbox.net ([208.91.199.152]:49892 "EHLO bh-25.webhostbox.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751063AbbAKQQm (ORCPT ); Sun, 11 Jan 2015 11:16:42 -0500 Received: from mailnull by bh-25.webhostbox.net with sa-checked (Exim 4.82) (envelope-from ) id 1YALBw-0049hE-Q1 for linux-watchdog@vger.kernel.org; Sun, 11 Jan 2015 16:16:43 +0000 Message-ID: <54B2A1E5.9030500@roeck-us.net> Date: Sun, 11 Jan 2015 08:16:37 -0800 From: Guenter Roeck MIME-Version: 1.0 To: Lars-Peter Clausen , Maarten ter Huurne CC: Ralf Baechle , Wim Van Sebroeck , Paul Burton , Paul Cercueil , linux-mips@linux-mips.org, linux-watchdog@vger.kernel.org Subject: Re: [PATCH 3/3] MIPS: jz4740: Move reset code to the watchdog driver References: <1420914550-18335-1-git-send-email-lars@metafoo.de> <1420914550-18335-3-git-send-email-lars@metafoo.de> <54B1CF4B.3070503@roeck-us.net> <1766434.QjfqQROysC@hyperion> <54B24566.5010109@metafoo.de> In-Reply-To: <54B24566.5010109@metafoo.de> Content-Type: text/plain; charset=windows-1252; format=flowed Content-Transfer-Encoding: 7bit Sender: linux-watchdog-owner@vger.kernel.org List-Id: linux-watchdog@vger.kernel.org On 01/11/2015 01:41 AM, Lars-Peter Clausen wrote: > On 01/11/2015 02:43 AM, Maarten ter Huurne wrote: >> On Saturday 10 January 2015 17:18:03 Guenter Roeck wrote: >>> On 01/10/2015 10:29 AM, Lars-Peter Clausen wrote: >>>> @@ -186,9 +208,20 @@ static int jz4740_wdt_probe(struct platform_device >>>> *pdev)> >>>> if (ret < 0) >>>> goto err_disable_clk; >>>> >>>> + drvdata->restart_handler.notifier_call = jz4740_wdt_restart; >>>> + drvdata->restart_handler.priority = 128; >>>> + ret = register_restart_handler(&drvdata->restart_handler); >>>> + if (ret) { >>>> + dev_err(&pdev->dev, "cannot register restart handler, %d\n", >>>> + ret); >>>> + goto err_unregister_watchdog; >>> >>> Are you sure you want to abort in this case ? >>> After all, the watchdog would still work. >> >> That raises a similar question: what about the opposite case, where the >> watchdog registration fails? If the resource acquisition part of the probe >> fails, neither the watchdog nor the restart functionality is going to work, >> but if the call to watchdog_register_device() fails, the restart handler >> would still work. > > I think this is fine, if either the watchdog or the restart handler registration fail then the system is probably already in a rather unusable state. > > But that got me thinking, maybe instead of having each watchdog driver register and implement its own restart handler we should maybe add this as a functionality to the watchdog framework. Something along the lines off. > > watchdog_set_timeout(wdt, wdt->min_timeout); > watchdog_start(wdt); > mdelay(wdt->min_timeout * 2000); > Agreed, something like that might be useful as fallback mechanism. Guenter