I thought I'd finally get around to trying this driver.
No luck.
The proximate cause is that Atmel's second stage "at91boot"
loader disables the watchdog. That's because the reset state
of this watchdog is "reset after 16 seconds" ... so it must
be tended carefully until the Linux "watchdog" daemon takes
over. Since there's no code in at91boot, U-Boot, or Linux to
do that, it gets disabled. (How was this driver tested??)
Needless to say, this isn't exactly my model of how to make
a useful watchdog. That "write once" rule is pure trouble.
I suggest merging the following patch, to reduce confusion.
- Dave
======== CUT HERE
The at91sam9 watchdog is a kind of annoying bit of hardware: since its
mode register is write-once, it can't be reconfigured. Moreover, Atmel's
standard second stage loader "at91boot" always this watchdog (at least
on Atmel's reference boards), ensuring that Linux normally can't use it.
This patch removes some confusion by not registering the watchdog device
on systems where it was disabled before Linux starts.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
--- a/arch/arm/mach-at91/at91sam9260_devices.c 2008-06-06 15:00:06.000000000 -0700
+++ b/arch/arm/mach-at91/at91sam9260_devices.c 2008-06-06 20:40:38.000000000 -0700
@@ -21,6 +21,7 @@
#include <asm/arch/at91sam9260.h>
#include <asm/arch/at91sam9260_matrix.h>
#include <asm/arch/at91sam9_smc.h>
+#include <asm/arch/at91_wdt.h>
#include "generic.h"
@@ -666,6 +667,9 @@ static struct platform_device at91sam926
static void __init at91_add_device_watchdog(void)
{
+ /* WDT_MR is write-once; if it was disabled, we're stuck */
+ if (at91_sys_read(AT91_WDT_MR) & AT91_WDT_WDDIS)
+ return;
platform_device_register(&at91sam9260_wdt_device);
}
#else
--- a/arch/arm/mach-at91/at91sam9261_devices.c 2008-06-06 15:00:06.000000000 -0700
+++ b/arch/arm/mach-at91/at91sam9261_devices.c 2008-06-06 20:40:45.000000000 -0700
@@ -25,6 +25,7 @@
#include <asm/arch/at91sam9261.h>
#include <asm/arch/at91sam9261_matrix.h>
#include <asm/arch/at91sam9_smc.h>
+#include <asm/arch/at91_wdt.h>
#include "generic.h"
@@ -654,6 +655,9 @@ static struct platform_device at91sam926
static void __init at91_add_device_watchdog(void)
{
+ /* WDT_MR is write-once; if it was disabled, we're stuck */
+ if (at91_sys_read(AT91_WDT_MR) & AT91_WDT_WDDIS)
+ return;
platform_device_register(&at91sam9261_wdt_device);
}
#else
--- a/arch/arm/mach-at91/at91sam9263_devices.c 2008-06-06 15:00:06.000000000 -0700
+++ b/arch/arm/mach-at91/at91sam9263_devices.c 2008-06-06 20:40:48.000000000 -0700
@@ -24,6 +24,7 @@
#include <asm/arch/at91sam9263.h>
#include <asm/arch/at91sam9263_matrix.h>
#include <asm/arch/at91sam9_smc.h>
+#include <asm/arch/at91_wdt.h>
#include "generic.h"
@@ -911,6 +912,9 @@ static struct platform_device at91sam926
static void __init at91_add_device_watchdog(void)
{
+ /* WDT_MR is write-once; if it was disabled, we're stuck */
+ if (at91_sys_read(AT91_WDT_MR) & AT91_WDT_WDDIS)
+ return;
platform_device_register(&at91sam9263_wdt_device);
}
#else
--- a/arch/arm/mach-at91/at91sam9rl_devices.c 2008-06-06 15:00:06.000000000 -0700
+++ b/arch/arm/mach-at91/at91sam9rl_devices.c 2008-06-06 20:40:52.000000000 -0700
@@ -21,6 +21,7 @@
#include <asm/arch/at91sam9rl.h>
#include <asm/arch/at91sam9rl_matrix.h>
#include <asm/arch/at91sam9_smc.h>
+#include <asm/arch/at91_wdt.h>
#include "generic.h"
@@ -501,6 +502,9 @@ static struct platform_device at91sam9rl
static void __init at91_add_device_watchdog(void)
{
+ /* WDT_MR is write-once; if it was disabled, we're stuck */
+ if (at91_sys_read(AT91_WDT_MR) & AT91_WDT_WDDIS)
+ return;
platform_device_register(&at91sam9rl_wdt_device);
}
#else
--
| Ingo Molnar | [patch 12/13] syslets: x86: optimized copy_uatom() |
| Greg Kroah-Hartman | [PATCH 017/196] aoechr: Convert from class_device to device |
| Yinghai Lu | Re: 2.6.26, PAT and AMD family 6 |
| Jan Engelhardt | intel iommu (Re: -mm merge plans for 2.6.23) |
git: | |
| Gerrit Renker | [PATCH 27/37] dccp: Integration of dynamic feature activation - part 2 (server side) |
| David Miller | [GIT]: Networking |
| David Miller | Re: [PATCH] pkt_sched: Destroy gen estimators under rtnl_lock(). |
| Natalie Protasevich | [BUG] New Kernel Bugs |
