From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1752640AbbKYU3O (ORCPT ); Wed, 25 Nov 2015 15:29:14 -0500 Received: from proxima.lp0.eu ([81.2.80.65]:59591 "EHLO proxima.lp0.eu" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751847AbbKYU3L (ORCPT ); Wed, 25 Nov 2015 15:29:11 -0500 Subject: Re: [PATCH 4/10] watchdog: bcm63xx_wdt: Handle hardware interrupt and remove software timer To: Jonas Gorski , Guenter Roeck References: <5650BFD6.5030700@simon.arlott.org.uk> <5650C08C.6090300@simon.arlott.org.uk> <5650E2FA.6090408@roeck-us.net> <5650E5BB.6020404@simon.arlott.org.uk> <56512937.6030903@roeck-us.net> <5651CB13.4090704@simon.arlott.org.uk> <5651CB9C.4090005@simon.arlott.org.uk> <20151124182138.GB12289@roeck-us.net> Cc: Florian Fainelli , "devicetree@vger.kernel.org" , Ralf Baechle , Thomas Gleixner , Jason Cooper , Marc Zyngier , Kevin Cernekee , Wim Van Sebroeck , Maxime Bizon , Linux Kernel Mailing List , MIPS Mailing List , linux-watchdog@vger.kernel.org, Rob Herring , Pawel Moll , Mark Rutland , Ian Campbell , Kumar Gala From: Simon Arlott Message-ID: <56561A0B.9020700@simon.arlott.org.uk> Date: Wed, 25 Nov 2015 20:28:59 +0000 User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:38.0) Gecko/20100101 Thunderbird/38.3.0 MIME-Version: 1.0 In-Reply-To: Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On 25/11/15 20:14, Jonas Gorski wrote: > On Tue, Nov 24, 2015 at 7:21 PM, Guenter Roeck wrote: >> On Sun, Nov 22, 2015 at 02:05:16PM +0000, Simon Arlott wrote: >>> There is a level triggered interrupt for the watchdog timer as part of >>> the bcm63xx_timer device. The interrupt occurs when the hardware watchdog >>> timer reaches 50% of the remaining time. >>> >>> It is not possible to mask the interrupt within the bcm63xx_timer device. >>> To get around this limitation, handle the interrupt by restarting the >>> watchdog with the current remaining time (which will be half the previous >>> timeout) so that the interrupt occurs again at 1/4th, 1/8th, etc. of the >>> original timeout value until the watchdog forces a reboot. >>> >>> The software timer was restarting the hardware watchdog with a 85 second >>> timeout until the software timer expired, and then causing a panic() >>> about 42.5 seconds later when the hardware interrupt occurred. The >>> hardware watchdog would not reboot until a further 42.5 seconds had >>> passed. >>> >>> Remove the software timer and rely on the hardware timer directly, >>> reducing the maximum timeout from 256 seconds to 85 seconds >>> (2^32 / WDT_HZ). >>> >> >> Florian, >> >> can you have a look into this patch and confirm that there is no better >> way to clear the interrupt status ? > > While the watchdog interrupt can't be masked, it should be able to be > cleared by writing 1 to the appropriate bit in the timer block's > interrupt status register. At least the broadcom sources do so. Not according to the hardware itself: [ 6.674626] watchdog watchdog0: warning timer fired, reboot in 7499ms [ 6.681212] irq_bcm6345_l2_timer: bcm6345_timer_write_int_status: b0000083=08 [ 6.688583] watchdog watchdog0: warning timer fired, reboot in 7486ms [ 6.695181] irq_bcm6345_l2_timer: bcm6345_timer_write_int_status: b0000083=08 [ 6.702554] watchdog watchdog0: warning timer fired, reboot in 7472ms [ 6.709158] irq_bcm6345_l2_timer: bcm6345_timer_write_int_status: b0000083=08 [ 6.716529] watchdog watchdog0: warning timer fired, reboot in 7458ms [ 6.723135] irq_bcm6345_l2_timer: bcm6345_timer_write_int_status: b0000083=08 [ 6.730538] watchdog watchdog0: warning timer fired, reboot in 7444ms [ 6.737121] irq_bcm6345_l2_timer: bcm6345_timer_write_int_status: b0000083=08 [ 6.744482] watchdog watchdog0: warning timer fired, reboot in 7430ms [ 6.751090] irq_bcm6345_l2_timer: bcm6345_timer_write_int_status: b0000083=08 typedef struct Timer { uint16 unused0; byte TimerMask; #define TIMER0EN 0x01 #define TIMER1EN 0x02 #define TIMER2EN 0x04 byte TimerInts; #define TIMER0 0x01 #define TIMER1 0x02 #define TIMER2 0x04 #define WATCHDOG 0x08 ... -- Simon Arlott From mboxrd@z Thu Jan 1 00:00:00 1970 From: Simon Arlott Subject: Re: [PATCH 4/10] watchdog: bcm63xx_wdt: Handle hardware interrupt and remove software timer Date: Wed, 25 Nov 2015 20:28:59 +0000 Message-ID: <56561A0B.9020700@simon.arlott.org.uk> References: <5650BFD6.5030700@simon.arlott.org.uk> <5650C08C.6090300@simon.arlott.org.uk> <5650E2FA.6090408@roeck-us.net> <5650E5BB.6020404@simon.arlott.org.uk> <56512937.6030903@roeck-us.net> <5651CB13.4090704@simon.arlott.org.uk> <5651CB9C.4090005@simon.arlott.org.uk> <20151124182138.GB12289@roeck-us.net> Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: Sender: linux-watchdog-owner-u79uwXL29TY76Z2rM5mHXA@public.gmane.org To: Jonas Gorski , Guenter Roeck Cc: Florian Fainelli , "devicetree-u79uwXL29TY76Z2rM5mHXA@public.gmane.org" , Ralf Baechle , Thomas Gleixner , Jason Cooper , Marc Zyngier , Kevin Cernekee , Wim Van Sebroeck , Maxime Bizon , Linux Kernel Mailing List , MIPS Mailing List , linux-watchdog-u79uwXL29TY76Z2rM5mHXA@public.gmane.org, Rob Herring , Pawel Moll , Mark Rutland , Ian Campbell , Kumar Gala List-Id: devicetree@vger.kernel.org On 25/11/15 20:14, Jonas Gorski wrote: > On Tue, Nov 24, 2015 at 7:21 PM, Guenter Roeck wrote: >> On Sun, Nov 22, 2015 at 02:05:16PM +0000, Simon Arlott wrote: >>> There is a level triggered interrupt for the watchdog timer as part of >>> the bcm63xx_timer device. The interrupt occurs when the hardware watchdog >>> timer reaches 50% of the remaining time. >>> >>> It is not possible to mask the interrupt within the bcm63xx_timer device. >>> To get around this limitation, handle the interrupt by restarting the >>> watchdog with the current remaining time (which will be half the previous >>> timeout) so that the interrupt occurs again at 1/4th, 1/8th, etc. of the >>> original timeout value until the watchdog forces a reboot. >>> >>> The software timer was restarting the hardware watchdog with a 85 second >>> timeout until the software timer expired, and then causing a panic() >>> about 42.5 seconds later when the hardware interrupt occurred. The >>> hardware watchdog would not reboot until a further 42.5 seconds had >>> passed. >>> >>> Remove the software timer and rely on the hardware timer directly, >>> reducing the maximum timeout from 256 seconds to 85 seconds >>> (2^32 / WDT_HZ). >>> >> >> Florian, >> >> can you have a look into this patch and confirm that there is no better >> way to clear the interrupt status ? > > While the watchdog interrupt can't be masked, it should be able to be > cleared by writing 1 to the appropriate bit in the timer block's > interrupt status register. At least the broadcom sources do so. Not according to the hardware itself: [ 6.674626] watchdog watchdog0: warning timer fired, reboot in 7499ms [ 6.681212] irq_bcm6345_l2_timer: bcm6345_timer_write_int_status: b0000083=08 [ 6.688583] watchdog watchdog0: warning timer fired, reboot in 7486ms [ 6.695181] irq_bcm6345_l2_timer: bcm6345_timer_write_int_status: b0000083=08 [ 6.702554] watchdog watchdog0: warning timer fired, reboot in 7472ms [ 6.709158] irq_bcm6345_l2_timer: bcm6345_timer_write_int_status: b0000083=08 [ 6.716529] watchdog watchdog0: warning timer fired, reboot in 7458ms [ 6.723135] irq_bcm6345_l2_timer: bcm6345_timer_write_int_status: b0000083=08 [ 6.730538] watchdog watchdog0: warning timer fired, reboot in 7444ms [ 6.737121] irq_bcm6345_l2_timer: bcm6345_timer_write_int_status: b0000083=08 [ 6.744482] watchdog watchdog0: warning timer fired, reboot in 7430ms [ 6.751090] irq_bcm6345_l2_timer: bcm6345_timer_write_int_status: b0000083=08 typedef struct Timer { uint16 unused0; byte TimerMask; #define TIMER0EN 0x01 #define TIMER1EN 0x02 #define TIMER2EN 0x04 byte TimerInts; #define TIMER0 0x01 #define TIMER1 0x02 #define TIMER2 0x04 #define WATCHDOG 0x08 ... -- Simon Arlott -- To unsubscribe from this list: send the line "unsubscribe linux-watchdog" in the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org More majordomo info at http://vger.kernel.org/majordomo-info.html